Frs2α and Shp2 signal independently of Gab to mediate FGF signaling in lens development

The fibroblast growth factor (FGF) signaling cascade is initiated by the binding of FGF and heparan sulfate proteoglycan (HSPG) to the FGF receptor (FGFR), which undergoes dimerization to activate its intrinsic tyrosine kinase activity (Schlessinger, 2004). This leads to phosphorylation of multiple tyrosine residues on FGFR, generating potential docking sites for signaling molecules containing SH2 (Src homology-2) or PTB (phosphotyrosine binding). However, only a handful of proteins, including Frs2 (Frs2α and Frs2β, encoded by the genes Frs2 and Frs3, respectively), Crk (Crk and Crkl) and PLCγ have been identified to interact directly with activated FGFR (Mohammadi et al., 1991; Wang et al., 1996; Kouhara et al., 1997; Larsson et al., 1999). Among these factors, Crk proteins appear to function in a context-dependent manner in FGF signaling (Moon et al., 2006). PLCγ mainly triggers activation of PKC and intercellular calcium release. Mutation of the PLCγ binding residue on FGFR1 elevates FGF signaling in vivo, suggesting that the interaction between FGFR and PLCγ is unlikely to explain most of the FGF signaling effects (Partanen et al., 1998). Instead, biochemical studies show that FGFR signaling through Frs2 proteins is the primary pathway to activate downstream Ras–MAPK and PI3K–Akt cascades (Ong et al., 2000; Hadari et al., 2001).

Frs2 proteins are composed of N-terminal PTB domains that bind to the juxtamembrane domain of FGFR and C-terminal domains that can be phosphorylated at multiple tyrosine residues by activated FGFR (Gotoh, 2008). In Frs2α, for example, four of these phosphorylated tyrosine residues bind to the adaptor protein Grb2, which recruits its constitutively bound partner Sos, a guanine nucleotide exchange factor (GEF) that directly activates Ras signaling (Hadari et al., 2001). It is thought that another Grb2-associated protein, Gab1, can further engage the p85 protein, a regulatory subunit of PI3-kinase (PI3K) (Ong et al., 2001). This provides a direct route for the Frs2–Grb2–Gab1 complex to stimulate PI3K–Akt signaling. However, recent genetic studies have begun to question this Frs2-centric view of FGF signaling mechanism. Unlike Frs2β, which is restricted to nervous system, Frs2α is ubiquitously expressed during development. However, Frs2α-knockout embryos survive until mouse embryonic day 8 (E8), significantly late compared with some of the FGF-signaling knockouts such as Fgf4, which dies at E4 (Feldman et al., 1995; Gotoh et al., 2005). Deletion of the Frs2 binding site in Fgfr1 (Fgfr^ΔFrs2) does not disrupt Fgfr1 signaling during gastrulation and somitogenesis, but results in neurulation, tail bud and pharyngeal arch defects at late embryogenesis (Hoch and Soriano, 2006). Furthermore, no major defects were reported in an Fgfr2 mutant (Fgfr2^LR) deficient in Frs2 binding (Eswarakumar et al., 2006; Sims-Lucas et al., 2011). Therefore, the functional significance of Frs2 proteins in FGF signaling remains a major question in embryonic development.

The development of vertebrate lens is an excellent model for studying FGF signaling. At E10.5, the invaginating lens ectoderm detaches from the anterior head epithelium to give rise to the lens vesicle. This is followed by differentiation of the posterior lens vesicle, which generates primary lens fibers to occupy the lumen of lens vesicle by E12.5. The subsequent growth of the lens requires proliferation and migration of the anterior lens epithelial cells, which transition at the lens equator region into orderly arrays of secondary lens fibers. This establishes structural integrity and clarity of mature lens. Previous explants and transgenic studies have demonstrated that FGF signaling is necessary and sufficient for inducing lens fiber development (McAvoy and Chamberlain, 1989; Robinson et al., 1995; Lovicu and Overbeek, 1998). In support of this, deletion of the FGF receptors Fgfr1, Fgfr2 and Fgfr3 or their co-receptor heparan sulfates at various stages of lens development disrupts both lens vesicle formation and lens fiber differentiation (Pan et al., 2006; Zhao et al., 2008; Garcia et al., 2011; Qu et al., 2011a). In addition, a systemic Frs2α mutant (Frs2α^2F) lacking two tyrosine residues required for binding of Shp2 protein displays lens and retina developmental failure, whereas genetic ablation of the Ras signaling molecules Shp2, Nf1 and ERK disrupts lens development (Gotoh et al., 2004; Pan et al., 2010; Carbe and Zhang, 2011; Upadhya et al., 2013). However, it was recently reported that a lens-specific knockout of Frs2α results in a well-differentiated lens with only moderately reduced size (Madakashira et al., 2012). This result raises the question whether Frs2α is indeed essential for lens development.

In this study, we confirmed the relatively modest lens defect in an Frs2α conditional knockout mouse, but we observed that knockouts of both Frs2α and Shp2 abrogated lens development by E14.5. Although genes involved in early differentiation of the lens were unaffected in the Frs2α, Shp2 double mutant at E12.5, there was an arrest of proliferation and differentiation accompanied by a loss of ERK signaling. We further showed that the Frs2α mutant lacking the Shp2-binding site mimicked the Frs2α-null mutant in genetic interaction with Shp2, whereas gain-of-function mutations in either Shp2 or Kras could ameliorate the Frs2α lens defect. Finally, we showed that Gab1 and Gab2, two potential substrates of Shp2 and binding partners of the Frs2–Grb2–Shp2 complex, were dispensable for lens development. Taken together, these results show that Frs2–Shp2–Ras signaling is essential in lens development.

To determine downstream mediators of FGF signaling in lens development, we first generated a conditional knockout of Frs2α using Le-Cre, which is known to be active during lens development as early as E9.5. Although RNA in situ hybridization indicated that Frs2α transcripts had indeed disappeared in E12.5 Le-Cre; Frs2α^flox/flox mutant lens (Fig. 1A,B, arrows), immunohistochemistry revealed persistent Frs2α protein as late as E14.5, especially in the transition zone of the lens (Fig. 1B,C,F,G, arrows). This is consistent with a previous study of Le-Cre; Frs2α^flox/flox mutant, which found that Frs2α protein was significantly reduced but not eliminated in the E15.5 lens (Madakashira et al., 2012). We did not detect any overt phenotype in Le-Cre; Frs2α^flox/flox mutants until E12.5, when the primary lens fibers failed to reach the anterior rim of mutant lens (Fig. 1I,J, arrows). However, this modest delay in elongation of lens fiber cells was no longer visible by E13.5, when the lens cavities were filled in both wild-type and mutant lenses (Fig. 1M,N, arrows). At E14.5, the Frs2α mutant lens was reduced by 50% in size compared with the wild type (n = 8 controls and 7 mutants; P<0.01, Student's t test) (Fig. 1Q,R, arrows). As we previously reported, ablation of Shp2 protein in Le-Cre; Shp2^flox/flox mutants also resulted in a similar lens phenotype (Fig. 1K,O,S, arrows) (Pan et al., 2010). Combined deletion of both Frs2α and Shp2, however, significantly disrupted lens development, leading to a complete loss of lens in E14.5 Le-Cre; Shp2^flox/flox; Frs2α^flox/flox embryos (Fig. 1L,P,T, arrows, n = 8). This genetic interaction suggests that Frs2α and Shp2 together play important roles in lens development.

We next examined the molecular defects in Frs2α and Shp2 double mutants. At E12.5, Le-Cre; Shp2^flox/flox; Frs2α^flox/flox embryos still contained a rudimentary lens vesicle marked by expressions of Pax6 and Foxe3, two essential regulators of early lens development (Fig. 2A–A′″ and B–B′″). However, whereas these transcriptional factors were normally excluded from the differentiating cells in the posterior lens vesicle at this stage, they were maintained in the entire lens vesicle in Frs2α and Shp2 double mutants. This suggests that combined deletion of Frs2α and Shp2 allows proper specification of lens progenitor cells in the lens vesicle, but disrupts their subsequent differentiation into lens fibers. Consistent with this, the lens fiber differentiation factor Prox1 was expressed in Frs2α and Shp2 double mutants at the low level typically found in wild-type lens epithelium, in contrast to the elevated level present in the posterior lens vesicle (Fig. 2C–C′″). By contrast, Frs2α and Shp2 double mutants ectopically expressed the anterior lens epithelial marker E-cadherin in the posterior lens vesicle (Fig. 2D–D′″). Lens differentiation is characterized by sequential induction of α-, β- and γ-crystallins during development. Although α- and β-crystallins were still expressed, the late onset expression of γ-crystallin was absent in the posterior lens vesicle (Fig. 2E–E′″; Fig. 2F–F′″ and Fig. 2G–G′″). Because none of these genetic abnormalities were observed in Frs2α or Shp2 single mutants, these results suggest that Frs2α and Shp2 are synergistically required for proper differentiation of lens fibers.

The Frs2α and Shp2 mutant phenotypes described above overlap with the lens defects previously observed when FGF receptors were ablated during lens vesicle development (Zhao et al., 2008). We thus investigated FGF downstream signaling in Frs2α and Shp2 mutants. In contrast to Frs2α or Shp2 single mutants, double deletion of Frs2α and Shp2 abolished ERK phosphorylation and expression of the FGF signaling downstream response genes Erm and Er81 (Fig. 3A–D; supplementary material Fig. S1A–D). D-type cyclins and their associated kinase CDK4 are required for G1–S transition, whereas p57 promotes cell cycle exit. In Frs2α and Shp2 double mutant lenses, both cyclin D1 and p57 were lost and CDK4 failed to be localized to nucleus (Fig. 3E–L; supplementary material Fig. S2A,B). Although the cell cycle arrest marker p21 appeared to be unchanged, the Frs2α and Shp2 mutant lens displayed significant reduction in cell proliferation marker phospho-histone H3 (pHH3) and increased cell death as shown by TUNEL staining (supplementary material Fig. S2D,E; Fig. 3M–Y). Therefore, Frs2α and Shp2 cooperate in regulating ERK signaling and in promoting cell cycle progression and survival in lens development.

Frs2α and Shp2 double mutants clearly exhibited more severe lens defects than those of single mutants, suggesting that Frs2α and Shp2 mediate FGF–ERK signaling together. To test this notion, we generated primary mouse embryonic fibroblast (MEF) cells carrying Frs2α ^flox or Shp2^flox alleles. This allows us to ablate Frs2α and Shp2 in vitro by infecting these MEF cells with a Cre-expressing adenovirus for 5 days (Fig. 4A). Seemingly consistent with our in vivo finding in the lens, FGF2-stimulated ERK phosphorylation was either unaffected or reduced modestly in Shp2 or Frs2α single mutant cells, but became further compromised in Frs2α and Shp2 double mutant cells. However, we noticed that there was still residual Frs2α and Shp2 protein in these MEF cells, which might obscure the essential functions of these factors in FGF signaling. We thus extended the virus infection to 7 days, further reducing the level of Frs2α and Shp2 protein (Fig. 4B). Interestingly, Frs2α and Shp2 single mutant cells now displayed downregulation of FGF2-induced ERK signaling similar to that of Frs2α and Shp2 double mutant cells. Therefore, a more complete depletion of Frs2α and Shp2 protein in MEF cells abolished their apparent genetic synergy in FGF signaling.

We have observed a similar delay in depletion of Frs2α protein in vivo (Fig. 1E–G), raising the possibility that residual Frs2α protein might be sufficient to sustain ERK signaling during the crucial period of lens development. In this scenario, because Frs2α and Shp2 operate in an epistatic cascade downstream to FGF, simultaneous depletion of both proteins in double mutants will synergistically destabilize the entire pathway, resulting in a drastic loss of ERK signaling and severe lens defects. To distinguish the epistatic and the parallel models of the function of Frs2α and Shp2 in FGF signaling, we took advantage of an existing Frs2α^2F allele, which was mutated at two amino acids critical for Shp2 binding to Frs2α (Gotoh et al., 2004). We reasoned that, if the parallel model is correct, addition of the Frs2α^2F allele to the Shp2 single mutant background should not aggravate its lens defects, because the Shp2-independent function of Frs2α remains intact in Frs2α^2F and Shp2 compound mutants. However, the epistatic model predicts that combining disruption of the Shp2–Frs2α interaction upon depletion of Shp2 interferes with two consecutive nodes of FGF signaling, thus resulting in severe lens defects similar to combined depletion of Frs2α and Shp2.

To test these predictions, we first generated Le-Cre; Frs2α^flox/2F mutants, which exhibited mild lens defects similar to Le-Cre; Frs2α^flox/flox and Le-Cre; Shp2^flox/flox single mutants (supplementary material Fig. S3A–N). Le-Cre; Shp2^flox/flox; Frs2α^flox/2F compound mutants, however, closely resembled Le-Cre; Shp2^flox/flox; Frs2α^flox/flox double mutants in their defective FGF signaling and lens differentiation, indicated by loss of ERK phosphorylation and reduced Prox1 expression at E12.5 (Fig. 4C–D″). Aberrant cell cycle controls were also evident in both compound mutants because cyclin D1 and p57 were abolished, CDK4 remained cytoplasmic and p21 was unaffected (Fig. 4E–F″; supplementary material Fig. S2C,F). This led to reduced cell proliferation and increased cell death as shown by Ki67 expression, BrdU incorporation and TUNEL staining in the hollow lens vesicles (Fig. 4E–J). At E14.5, neither Le-Cre; Shp2^flox/flox; Frs2α^flox/2F (n = 5) nor Le-Cre; Shp2^flox/flox; Frs2α^flox/flox (n = 8) mutants contained any obvious lens structure. Collectively, these results are consistent with the idea that the Shp2-interacting residues on Frs2α are essential for FGF–ERK signaling in lens development.

The binding of Shp2 to Frs2 can generate additional docking sites for the downstream molecule Grb2, but it could also serve to activate Shp2 phosphatase activity. To test the significance of the latter possibility, we crossed Le-Cre; Frs2α^flox/flox mutant with the LSL-Shp2^D61Y allele, which contains an oncogenic Shp2^D61Y knock-in mutation silenced by a floxed transcription stop cassette (Chan et al., 2009) (Fig. 5A). Expression of Cre recombinase from Le-Cre transgene will thus result in lens-specific induction of Shp2^D61Y, which is known to exhibit constitutive phosphatase activity. The experiments presented in Fig. 5B first show that genetic depletion of Frs2α in MEF cells infected with Cre adenovirus effectively reduced FGF-induced phosphorylation of ERK and Gab1(Tyr627). Addition of LSL-Shp2^D61Y caused a notable increase in phospho-Gab1 and phospho-ERK levels. These results were next confirmed in vivo, where western blots and immunostaining showed that ERK phosphorylation was reduced in Le-Cre; Frs2α^flox/flox mutant lens, but recovered in Le-Cre; Frs2α^flox/flox; LSL-Shp2^D61Y mutants (Fig. 5C–F). As a result, a statistically significant increase in lens size was observed in Le-Cre; Frs2α^flox/flox; LSL-Shp2^D61Y mutants compared with Le-Cre; Frs2α^flox/flox mutants (Fig. 5G–J). These results demonstrate that Shp2 gain-of-function mutation can compensate for the loss of Frs2α in lens development.

Our above results show that phosphorylation of Gab1 at Tyr627, a known Shp2 binding site, is strictly dependent on Frs2α during FGF signaling, suggesting that Gab1 is a partner of the Frs2α–Shp2 complex (Schaeper et al., 2000). Indeed, a previous study has suggested that recruitment of Gab1 by Frs2α leads to activation of PI3K–Akt signaling (Hadari et al., 2001). We thus investigated function of Gab1 and its close homologue Gab2 using genetically modified MEF cells. In this experiment, Gab1^flox/flox; Gab2^−/− MEF cells were infected with Cre-expressing adenovirus and then subject to FGF stimulation. The samples were separated by SDS-PAGE followed by immunoblotting with phospho-specific antibodies. The experiment presented in Fig. 6A shows that simultaneous depletion of Gab1 and Gab2 in MEF cells failed to disrupt FGF-induced phosphorylation of Shp2, ERK and Akt, indicating that Gab1 and Gab2 are dispensable for FGF signaling in vitro.

The function of Gab proteins has never been directly investigated in lens development. We thus generated Le-Cre; Gab1^flox/flox mice and confirmed the loss of Gab1 protein in E12.5 and E14.5 lens (Fig. 6B–E, arrows). Le-Cre activity is known to also be present in the ocular surface and lacrimal gland (Pan et al., 2008; Qu et al., 2011b). Interestingly, ablation of Gab1 did not perturb development of lacrimal gland, an Fgf10 signaling-dependent process, but resulted in an open eyelid at birth, consistent with the previously reported role of Gab1 in EGF-signaling-dependent eyelid closure (supplementary material Fig. S4A–J, arrows) (Makarenkova et al., 2000; Schaeper et al., 2007; Pan et al., 2008; Qu et al., 2011b). Immunoblotting of E16 lenses showed that ERK and Akt phosphorylation in Gab1-depleted mutants remained unchanged (Fig. 6H). Consistent with this, there were no reductions in either lens size or immunostaining intensities for phospho-Akt and phospho-ERK (Fig. 6I,J,L,M,O,P). Because there is a detectable level of Gab2 in the lens upon immunostaining (Fig. 6F,G), we also generated Le-Cre; Gab1^flox/flox; Gab2^−/− mice, which again failed to display any reduction in lens size or Akt and ERK phosphorylation (Fig. 6K,N,Q). Finally, examination of Le-Cre; Gab1^flox/flox; Frs2α^flox/flox and Le-Cre; Gab1^flox/flox; Shp2^flox/flox mutants showed that Gab1 did not genetically interact with Frs2α or Shp2 (data not shown). These results show that Gab1 and Gab2 are not required for FGF signaling either in vitro or in vivo.

We next considered Ras signaling as the main target of the Frs2α–Shp2 complex in lens development. In a genetic rescue experiment, we crossed Le-Cre; Shp2^flox/flox; Frs2α^flox/flox mutants with the LSL-Kras^G12D allele, which can be induced to express a constitutively active Kras^G12D mutant by Cre-mediated recombination (Fig. 7A). The results presented in Fig. 7B–G show that, despite a loss of Shp2 expression in Le-Cre; Shp2^flox/flox; Frs2α^flox/flox; LSL-Kras^G12D mutant lens at E12.5, ERK phosphorylation was induced to 52±6% (mean ± s.e.m) of wild-type level (Student's t-test: n = 4; P = 0.0231). This was accompanied by a recovery of the lens development markers Prox1, cyclin D1 and p57, cell proliferation marker Ki67 and a substantial increase in lens size (Fig. 7H–Q; supplementary material Fig. S5A,B). Consistent with our previous report (Pan et al., 2010), abnormal cell apoptosis, as indicated by TUNEL staining, was not rescued by activated Kras signaling (supplementary material Fig. S5C,D). Nevertheless, the Le-Cre; Shp2^flox/flox; Frs2α^flox/flox; LSL-Kras^G12D mutant lens at E16.5 exhibited elongated lens fibers capped anteriorly by a single layer of epithelial cells, closely resembling wild-type controls (supplementary material Fig. S5E,F). Taken together, these results show that activated Ras signaling can significantly subvert the requirement of both Frs2α and Shp2 in lens development.

In this study, we investigated the function of Frs2α and its downstream targets in FGF signaling during lens development. Combining biochemical analysis and mouse genetics, we demonstrate that simultaneous depletion of Frs2α and Shp2 results in a more severe FGF signaling defect than either single mutant. This synergy between Frs2α and Shp2 is not a result of overlapping or compensatory functions of these two factors. Instead, we show that Frs2α acts directly upstream to Shp2 in a linear cascade to mediate FGF signaling. We provide further evidence that Gab, a commonly believed downstream target of Frs2α, is in fact dispensable for FGF signaling. However, both constitutively activated Shp2 and Kras signaling can ameliorate lens defects in Frs2α mutants. These results show that Frs2α is a crucial link in the FGFR–Shp2–Ras–ERK signaling cascade in lens development.

The functional significance of Frs2α remains a controversial issue in the field of FGF signaling research. Although earlier biochemical studies have indicated that Frs2α mediates most, if not all, FGF signaling effects, more recent genetic studies have begun to cast doubt on this central role of Frs2α. Not only do knockouts of several Fgf and Fgfr family members cause embryonic lethality earlier than Frs2α null animals, mutation of the Frs2-binding sites on either Fgfr1 or Fgfr2 generally results in far less severe phenotypes than their null counterparts (Feldman et al., 1995; Gotoh et al., 2005; Eswarakumar et al., 2006; Hoch and Soriano, 2006). However, one should exercise caution when interpreting these genetic results because conformational changes in Fgfr mutants might cause fortuitous heterodimerization with other Fgfr family members, which could act in trans to rescue Frs2α-binding deficiency (Bellot et al., 1991; Ueno et al., 1992). In the case of systemic knockouts, deposition of long-lived maternal proteins in zygotes could distort the timing and severity of early embryonic defects. The complication of long-lasting proteins especially confounds conditional knockouts, where Cre-mediated genetic ablation will inevitably leave behind residual protein that persists with variable half-life dictated by the individual protein degradation rate. Indeed, by modeling protein degradation kinetics using MEF cells infected with Cre virus, we showed that trace amounts of Frs2α and Shp2 proteins existed several days after genetic deletion to sustain a significant FGF signaling response, which agrees with the extremely slow turnover rate of Shp2 previously reported (Siewert et al., 1999). Further protein depletion after prolonged exposure to Cre virus eventually revealed that both Frs2α and Shp2 are essential for FGF signaling. Therefore, the gradual depletion of Frs2α and Shp2 proteins expected in conditional knockouts might have significantly masked the true function of these two proteins in embryonic development.

The conditional knockouts of Frs2a and Shp2 described in this study should thus be construed as hypomorphic alleles, which can instead be valuable tools to probe protein–protein interaction. This is because hypomorphic mutations in two interacting factors are known to cause synergistic rather than additive effects, manifesting as genetic interactions in vivo. Indeed, we showed that conditional knockouts of Frs2α and Shp2 displayed mild phenotypes individually, but abolished lens development when combined. Lens induction failure has been previously reported in Frs2α^2F mutants that lack a Shp2-binding site, but these observations are confounded by the crucial role of FGF signaling in retinal development (Gotoh et al., 2004; Cai et al., 2010; Cai et al., 2013). We thus combined Frs2α^2F and Frs2α^flox alleles to generate a lens-specific disruption of Shp2 binding to Frs2α, showing that Frs2α^2F can substitute for Frs2α^flox to reproduce the genetic interaction between Frs2α and Shp2 in lens development. This result not only confirms the physical recruitment of Shp2 by Frs2α in vivo, but also suggests that the Frs2α–Shp2 complex is the signaling center of the FGF pathway during lens development. In this model, Frs2α acts directly upstream of Shp2 in a signaling cascade and cannot be compensated by other factors that mediate parallel signaling.

Although it has been generally assumed that Gab proteins play important roles in FGF signaling, this notion has never been tested in vivo (reviewed by Eswarakumar et al., 2006; Goetz and Mohammadi, 2013). Biochemical studies have previously suggested that FGF signaling induces an indirect coupling of Frs2 and Gab1 mediated by the Grb2 adaptor protein, leading to phosphorylation of Gab1 on multiple tyrosine residues (Hadari et al., 2001). These phosphorylation sites can attract binding of PI3K for activation of Akt signaling, RasGAP for suppression of Ras–MAPK signaling and protein tyrosine phosphatase Shp2 (Maroun et al., 2000; Rodrigues et al., 2000; Montagner et al., 2005). It has been proposed that Shp2 dephosphorylates Gab1 to eliminate the specific binding site for the SH2 domain of RasGAP, resulting in ultimate activation of Ras–MAPK signaling (Montagner et al., 2005). Therefore, Gab1 is not only an attractive mediator of PI3K–Akt signaling downstream of Frs2, but also serves as a phosphatase substrate enabling Shp2 to play a positive role in Ras–MAPK signaling. In primary MEF cells, we indeed observed FGF-induced phosphorylation of Gab1 on a tyrosine residue that engages Shp2. Genetic ablation of Gab1 and its homologue Gab2, however, did not affect either Akt or ERK signaling in vitro or in vivo. Consistent with this, we failed to observe any defects in Gab1- and Gab2-deficient lens and lacrimal gland, both of which require FGF signaling during embryonic development. To our knowledge, these studies provide the first genetic tests of Gab function in FGF signaling in vivo and they argue against a functional role of Gab in FGF–Frs2–Shp2 signaling in general.

Our study thus reveals that the Frs2α–Shp2 complex is crucial for mediating FGF signaling in lens development. It is notable that our Frs2α Shp2 mutant at E12.5 closely resembles a previously reported Fgfr1, Fgfr2 and Fgfr3 triple mutant in forming a hollow lens vesicle with aberrant expressions of Pax6, Prox1, E-cadherin and γ-crystallin (Zhao et al., 2008). However, our mutant also displayed a more severe phenotype in its abrogation of Cyclin D1 and P57 expression, resulting in a loss of lens structure after E14.5. It has previously been shown that Fgfr4 is present at low levels as early as E9.5 in lens placode, and one of its ligands Fgf15 is also abundantly expressed in the eye (Garcia et al., 2011; Cai et al., 2013). It would be interesting to test whether Fgfr4 signaling partially compensates for the loss of Fgfr1, Fgfr2 and Fgfr3, allowing these mutant lens cells to retain a limited proliferative capacity. The FGF receptor is the only known receptor tyrosine kinase that can induce mammalian lens fiber differentiation, which has been proposed to depend on the unique capacity of Frs2α in orchestrating downstream signaling (Madakashira et al., 2012). By showing that constitutive Ras signaling is sufficient to obviate the requirement of the Frs2α–Shp2 complex in lens development, we demonstrate that the ultimate target of Frs2α is Ras signaling. Ras signaling is known to have multiple downstream effector pathways, including RAF–MEK–ERK, PI3K and RalGDS. Importantly, Upadhya and colleagues (Upadhya et al., 2013) recently showed that genetic deletion of Erk1 and Erk2 in the lens also caused significant reduction in cell proliferation and aberrant apoptosis, but unlike the eventual loss of lens structure in our Frs2α and Shp2 mutants, embryos lacking Erk1 and Erk2 appear to retain a rudimentary lens at E14.5 (Upadhya et al., 2013). Whether additional Ras downstream effector pathways complement ERK signaling in lens development will be an important topic for future study.

Mice carrying Frs2α^flox, Frsα^2F, Gab1^flox, Gab2^+/− and Shp2^flox alleles were bred and genotyped as described (Gotoh et al., 2004; Zhang et al., 2004; Bard-Chapeau et al., 2005; Teal et al., 2006; Lin et al., 2007). LSL-Shp2^D61Y mice were gifts from Rebecca Chan, Indiana University School of Medicine, Indianapolis, IN and Benjamin G. Neel, Ontario Cancer Institute, Ontario, Canada (Chan et al., 2009). Le-Cre mice were kindly provided by Ruth Ashery-Padan, Tel Aviv University, Tel Aviv, Israel and Richard Lang, Children's Hospital Research Foundation, Cincinnati, OH (Ashery-Padan et al., 2000). LSL-Kras^G12D mice were obtained from the Mouse Models of Human Cancers Consortium (MMHCC) Repository at National Cancer Institute (Tuveson et al., 2004). In all conditional knockout experiments, mice were maintained in mixed genetic background and Le-Cre only mice in the same litters were used as wild type controls. All experiments were performed in accordance with institutional guidelines.

Mouse embryos were fixed with 4% paraformaldehyde (PFA) in PBS overnight and paraffin embedded. The sections (10 µm) were rehydrated and stained with hematoxylin and eosin (H&E) for histological analysis. Lens sizes were measured as previously described (Pan et al., 2010). TUNEL staining and immunostaining were performed on the cryosections (8 µm) as previously described (Pan et al., 2006; Pan et al., 2008). For phospho-ERK, phospho-Akt, p21, Frs2α and Shp2 staining, the signal was amplified using a Tyramide Signal Amplification kit (TSA™ Plus System, PerkinElmer Life Sciences, Waltham, MA). Antibodies used were: anti-BrdU (G3G4, Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), anti-p21 (sc-397), anti-Shp2 (Sc-280), anti-γ-crystallin (SC-22415) (from Santa Cruz biotechnology, Santa Cruz, CA), anti-phospho-ERK1/2 (#4370), anti-phospho-AKT (#4060), anti-CDK4 (#2906), anti-cyclin-D1 (#2926), anti-Gab1 (#3232) and anti-Gab2 (#3239) (all from Cell Signaling Technology, Beverly, MA), anti-P57 (ab75974) and anti-Frs2α (ab10425) (both from Abcam, Cambridge, MA), anti-E-cadherin (U3254, Sigma, St Louis, Missouri), anti-Ki67 (#550609, from BD Pharmingen San Diego, CA), anti-Prox1 (PRB-238C) and anti-Pax6 (PRB-278P) (both from Covance, Berkeley, CA). Antibodies against α- and β-crystallins were kindly provided by Sam Zigler (National Eye Institute, Bethesda, MD). Cell proliferation and apoptosis were measured as the ratio of BrdU, pHH3, Ki67 or TUNEL-positive cells versus DAPI-positive cells, and analyzed by one-way ANOVA analysis. For the quantification of phospho-ERK staining, the fluorescent pixel intensity in the lens was measured by the ImageJ program. This value was further divided by the averaged phospho-ERK pixel intensity at the nearby retinal region to obtain the normalized lens phospho-ERK level.

RNA in situ hybridizations were performed on cryosections (8 µm) as previously described (Pan et al., 2006; Cai et al., 2010). Foxe3 probe was generously provided by Milan Jamrich (Baylor College of Medicine, Houston, TX). Erm and Er81 were both from Bridget Hogan (Duke University Medical Center, Durham, NC). The Frs2α probe was generated from a genomic fragment containing the last exon flanked by loxP sites in the Frs2α^flox allele (Cai et al., 2013). For all immunohistochemistry and RNA in situ hybridization procedures, at least three embryos of each genotype were analyzed.

Primary MEF cells were isolated from embryos at E12.5–E14.5 stages and cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS as described (Qu et al., 2011b). MEFs from the second passage were infected with Ad5CMVCre or Ad5CMVCre-eGFP (Gene Transfer Vector Core, University of Iowa, IA) in DMEM containing 2% FBS overnight at multiplicity of infection 500 plaque-forming units/cell and cultured for 5 or 7 days.

Lens tissues were collected at E15–16 stages and homogenized in RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS, 1 µg/ml aprotinin, 1 µg/ml pepstatin, 10 mM sodium pyrophosphate, 1 mM PMSF, 0.2 mM Na₃VO₄, 50 mM NaF). 4.24–6.36×10⁵ MEF cells were seeded in 60 mm dishes and serum starved (0.5% FBS in DMEM) for 36–48 hours before stimulated by 50 ng/ml FGF2 (R&D Systems, Minneapolis, MN) for 5 minutes at 37°C. After washed twice in cold PBS, MEF cells were lysed in 160 µl RIPA buffer and processed for infrared-based western blot analysis using an Odyssey SA scanner (LICOR Biosciences, Lincoln, NE) (Qu et al., 2011b). The signal intensity was quantified using the Odyssey software. The antibodies used were mouse anti-phospho-ERK1/2 (sc-7383), rabbit anti-Frs2α (sc-8318) (both from Santa Cruz Biotechnology, Santa Cruz, CA) and mouse anti-AKT (#4060), rabbit anti-phospho-AKT (#4060), anti-ERK1/2 (#4695), anti-Gab1 (#3232), anti-phospho-Gab1(#3233), anti-Shp2 (#3752) and anti-phospho-Shp2 (#3751) (all from Cell Signaling Technology).

The authors thank Drs Ruth Ashery-Padan, Rebecca Chan, Bridget Hogan, Milan Jamrich, Richard Lang, Benjamin G. Neel, Sam Zigler for reagents and mice.