grb2 heterozygosity rescues embryonic lethality but not tumorigenesis in pten^+/- mice

Abstract

PTEN is a tumor suppressor gene implicated in both sporadic cancers and inherited tumor-prone syndromes. Here we show that pten^+/- mice display a partially penetrant embryonic lethality. This lethality is associated with defects in both neural and placental development. Notably, this lethality is completely rescued by grb2 haploinsufficiency. In contrast, grb2 heterozygosity did not alter tumorigenesis in either pten^+/- or T cell-specific pten^-/- mice. grb2^-/hypomorph murine embryonic fibroblasts (MEFs) show decreased activation of both PKB and Erk upon stimulation with epidermal growth factor, whereas grb2^-/hypomorph;pten^+/- MEFs activate PKB but not Erk normally. Similarly, grb2^-/hypomorph fibroblasts die in low serum, and this phenotype is rescued by pten haploinsufficiency. Activation of both PKB and Erk as well as survival in low serum-containing media are all rescued by reexpression of Grb2 containing mutations within the N-terminal Src homology 3 (SH3) domain, but not by C-terminal SH3 domain mutants. The N-terminal SH3 domain mutants fail to bind to Sos, whereas the C-terminal SH3 domain mutants fail to bind to Gab1, suggesting that Erk and PKB activation in fibroblasts in response to epidermal growth factor depends on Gab1 or other C-terminal SH3 domain-interacting proteins, but not on Sos. Thus, PTEN/phosphatidylinositol 3′ kinase signaling requires Grb2 during both embryonic development and fibroblast survival, but Grb2 heterozygosity does not effect tumorigenesis in pten-deficient mice. In fibroblasts, survival signals emanating from the epidermal growth factor receptor appear to be PKB-dependent, and this activation depends on the C-terminal SH3 domain of Grb2, likely through the interaction of Grb2 with Gab1.

The phosphatidylinositol 3′ kinase (PI3′K) signaling pathway is a crucial mediator of biological responses from a variety of cellular receptors, including the receptors for mitogenic growth factors, insulin, and the T cell receptor (1). Upon the binding of mitogenic growth factors such as epidermal growth factor (EGF), transforming growth factor α, fibroblast growth factor, and insulin-like growth factors to their respective receptor tyrosine kinases (RTKs), these RTKs dimerize and transphosphorylate on tyrosine residues (2). The resulting phosphotyrosine residues serve as recognition sites for Src homology 2 (SH2) domain-containing molecules such as Grb2, Shc, and the p85 subunit of type I PI3′Ks. Binding of the p85 regulatory subunit to phospho-YXXM motifs results in the activation of the p110 catalytic subunit of PI3′K, which then phosphorylates phosphatidylinositols at the D3 position of the phosphatidylinositol ring to generate phosphoinositol-3-phosphate, phosphoinositol-3,4-diphosphate [PI(3,4)P₂], and phosphoinositol-3,4,5-triphosphate [PI(3,4,5)P₃]. PI(3,4,5)P₃ recruits proteins containing PH domains, including the serine/threonine kinase PDK1 and the PDK1 substrate, PKB/Akt. The net result of PI3′K activation is cell-context dependent, and includes increased glucose uptake and glycogen deposition (3), resistance to apoptosis (4), increased proliferation (5), and increased cell size (6).

Although PI3′K can be activated by direct binding of p85 to phosphorylated RTKs, additional PI3′K regulatory mechanisms exist (Fig. 1). Ras can bind to and activate the p110 catalytic subunit of PI3′K (7). Ras itself is activated through the canonical Grb2-Sos-Ras pathway downstream of receptor tyrosine kinase phosphorylation (2), so PI3′K can be activated by RTK phosphorylation through Grb2-Sos-Ras. In this pathway, the SH2 domain of Grb2 binds to phosphotyrosine motifs on RTKs resulting in membrane recruitment of Sos, a protein constitutively associated with the N-terminal SH3 domain of Grb2. At the membrane, Sos acts as a guanine nucleotide exchange factor for Ras, thereby leading to Ras activation. Ras has multiple downstream targets, including Raf, PI3′K, Ral, and Rac (2). Thus, both Ras-dependent and -independent mechanisms for PI3′K activation exist. In addition, PI3′K can be activated through an alternative Grb2-dependent mechanism, independently of Ras. In this third pathway, the C-terminal SH3 domain of Grb2 binds to Gab1 (8), resulting in phosphorylation of Gab1 on tyrosine residues. This phosphorylation generates binding sites for p85, thereby activating PI3′K (9). Gab1 also activates Ras, and this activation appears to be dependent on the phosphatase Shp2, although the mechanism responsible for this activation is still unclear (10). Grb2 has also been reported to interact directly with p85, although the physiological significance of this interaction is unclear (11). Because the SH2 domain of p85 binds to phosphotyrosine motifs where a methionine is present at the +3 position (i.e., YXXM), whereas the SH2 domain of Grb2 binds to YXN motifs, activation of PI3′K downstream of Grb2 may provide a mechanism through which PI3′K can be activated in response to phosphorylation at non-YXXM motifs (12). This may have a number of biological consequences, including PI3′K activation in response to the phosphorylation of RTKs without YXXM motifs, or a decreased threshold of activation for PI3′K when both YXXM and YXN motifs are phosphorylated on the same RTK. This second situation would result in two independent signals to PI3′K emanating from the same RTK.

Fig. 1.

Schematic representation of the three possible mechanisms of PI3′K activation. See text for details.

The PI3′K pathway is subject to negative regulation by numerous mechanisms, including the tumor suppressor PTEN. PTEN dephosphorylates PIP3 at the 3′ position, thereby directly opposing PI3′K. PTEN mutations are found in sporadic tumors including tumors of the brain, prostate, kidney, and breast (13, 14). Germ-line mutations of PTEN result in Cowden's syndrome, a disease characterized by the appearance of multiple benign hamartomas and predisposition to malignancies of the breast and thyroid (15–18). Several groups have disrupted the pten locus in mice through gene targeting strategies (19–21). Mice with a null mutation of the tumor suppressor gene pten (pten^-/- mice) die by embryonic day 7.5 (E7.5) and show overgrowth of the cephalic and caudal regions of the embryo (19). Mice heterozygous for this mutation (pten^+/- mice) serve as models for tumorigenesis associated with pten mutations, because they develop tumors in multiple organ systems, and these lesions partially resemble the spectrum of tumors found in patients with germ-line pten mutations (20, 22). Furthermore, mice in which pten has been specifically disrupted in the thymus, breast, skin, germ cells, or prostate by using the Cre-loxP system rapidly develop tumors in those organ systems (23). These observations point to a role for PTEN as a critical regulator of tumorigenesis in both mice and humans.

To address the roles of Grb2-dependent and -independent PI3′K activation, we generated mice doubly heterozygous for pten and grb2. Although grb2 heterozygosity did not affect tumorigenesis, it completely rescued the lethality seen in pten^+/- embryos, suggesting that PI3′K signaling may depend on Grb2 in a context-specific fashion. In fibroblasts, PKB activation downstream of EGF stimulation also requires Grb2, and activation of PKB correlates with the ability of fibroblasts to survive in low serum-containing media. Activation of both Erk and PKB in fibroblasts depends on the interaction between Grb2 and Gab1, but not on the interaction between Grb2 and Sos1.

Materials and Methods

Mouse Strains. pten^+/-, tpten^-/-, and grb2^+/- mice have been described (19, 24, 25). The pten^+/- mice were backcrossed at least eight times into the C57BL/6 strain before embryonic lethality was investigated.

Embryo Histology and Immunohistochemistry. pten^+/+ and pten^+/- embryos were fixed in 4% paraformaldehyde in PBS. Fixed embryos were dehydrated through increasing concentrations of ethanol, then washed three times in xylene before being placed in a 50:50 xylene/wax mixture for 1 h. Embryos were then embedded in wax. Sections were cut at 4 μm. Immunohistochemistry was performed by using anti phospho-PKB (Cell Signaling Technology, Beverly, MA) and anti phospho-Erk (Cell Signaling Technology) antibodies as per manufacturer's recommendations.

Tumor Formation and Histological Analysis. Mice were killed at 1 year of age or upon signs of morbidity. Organs were fixed in neural-buffered formalin before being dehydrated and embedded in wax.

Fibroblast Lines. pten^+/-; grb2^+/- and pten^+/flox; grb2^-/E89K mice were crossed, and E9.5 embryos were used to establish fibroblasts. These fibroblasts were immortalized by retroviral infection with SV40 large T antigen. pten^-/- cell lines were established by isolating pten^-/flox fibroblasts, immortalizing these cells by retroviral infection with SV40 large T antigen, and then infecting these cells with a self-excising Cre retrovirus (26).

Fluorescence-Activated Cell Sorting Analysis. Floating cells were collected along with the adherent cells. Adherent cells were trypsinized, collected by centrifugation, and washed in PBS before fixation in 70% ethanol for 30 min. These cells were then washed with PBS plus 1% BSA and permeabilized (250 μg/ml RNase/50 μg/ml propidium iodide).

Western Blotting. Cells were lysed in Chaps buffer [10 mM Tris, pH 7.5/1 mM MgCl₂/1 mM EGTA/0.5% Chaps/10% glycerol/100 μM NaVO₃/50 mM NaF and protease inhibitors (Roche Applied Science)]. Protein concentration was normalized by Bradford assay, and 30 μg of each sample was loaded. Antibodies directed against PKB (Santa Cruz Biotechnology), phospho-PKB, phospho-Erk, or total Erk (all from Cell Signaling Technology) were used at 1:1,000 dilutions.

Coimmunoprecipitation. Cells were lysed in Chaps (Sos1) or in Nonidet P-40 lysis (Gab1) (50 mM Tris, pH 8.0/150 mM NaCl/0.02% sodium azide/100 μM NaVO₃/50 mM NaF/protease inhibitors/1% Nonidet P-40) buffer. Sos1 (3.5 mg) or Gab1 (1.5 mg) were incubated overnight with an anti-myc antibody (Invitrogen). Gammabind protein G beads (50 μg, Amersham Pharmacia) were added and, 1 h later, the pellet was washed three times in the corresponding lysis buffer.

Results and Discussion

Partial Embryonic Lethality in pten^+/- Mice. pten^+/- mice described in ref. 19 were backcrossed into the C57BL/6 background. Upon backcrossing, a partially penetrant embryonic lethality was observed in pten^+/- animals with >30% penetrance by E10.5 (Table 1). This lethality was not previously observed in the mixed B6/129J background (19). To further investigate this lethality, embryos from pten^+/- × pten^+/- intercrosses were examined. Two distinct phenotypes of moribund pten^+/- embryos were identified. The first, more severely affected group of pten^+/- embryos was indistinguishable from pten^-/- embryos of the same age (Fig. 2A Inset). By E9.5, these pten^+/- embryos had asymmetrical head folds with little or no development of the major organ systems, including the heart and brachial arches (Fig. 2A). This group represents ≈5% of all pten^+/- embryos examined. The second and more common group of moribund pten^+/- E9.5 embryos had normal body symmetry and development of most organ systems but were smaller than their wild-type littermates and remained unturned (Fig. 2 B–E). The development of the heart and yolk sac appeared normal in these embryos, but the neural tube was kinked and open along the entire length of the embryo (Fig. 2 C–K). Interestingly, the open neural tube in these embryos lacked exencephalic features. This phenotype thus stands in contrast to the localized exencephaly commonly seen in animals mutated for genes involved in growth and apoptosis, such as the Tsc2^-/- and Caspase-3^-/- mice (27, 28). Our data show that haploinsufficiency for pten has effects of varying severity on embryonic development, especially in the neural tube.

Table 1. Pten^+/- partial embryonic lethality is rescued by Grb2^+/-.

		Genotype
Mating scheme	Age	Wild type	Grb2+/-	Pten+/-	Pten+/-;Grb2+/-
Pten+/- × Grb2+/-	P21	54 (54)	50 (54)	28* (54)	46 (54)
Pten+/- × Pten+/-;Grb2+/-	E10.5	13 (13)	13 (13)	17* (26)	27 (26)

Parentheses indicate expected numbers relative to wild type based on Medelian frequency. P21, postnatal day 21.

^*Statistically significant difference versus wild type (χ², P = 0.05)

Fig. 2.

Embryonic defects in pten^+/- embryos. (A and B) Whole mount analyses of pten^+/- and pten^+/+ littermate embryos at E9.5. Severely affected pten^+/- embryos are indistinguishable from pten^-/- embryos (A Inset), whereas the majority of pten^+/- embryos at E9.5 are smaller and unturned (B). Arrows point to the allantois (pten^+/- and pten^-/-) or the umbilical cord with blood vessels (pten^+/+). (C–E) Histological sections of a pten^+/+ and two moribund pten^+/- embryos. Both mutants show an enlarged allantois and kinked neuroepithelium. (F–K) Transverse sections through E9.5 embryos. Matched serial sections from wild-type (F–H) and pten^+/- (I–K) embryos are shown with the posterior of the embryo at the top of the figure. The neural tube in the pten^+/- embryo is open, whereas in wild-type littermates, the neural tube is in the process of closing. a, allantois; h, heart; hfr, head fold region; pl, placenta; nt, neural tube.

Placental Defects in pten^+/- Embryos Are Associated with Elevated Phospho-PKB. Moribund pten^+/- embryos also showed defects in placental development (Fig. 3 A–C). In wild-type littermates, chorioallantoic fusion had occurred by E9. In contrast, the allantois failed to fuse to the chorion at E9 in 80% of the dying pten^+/- embryos (Figs. 2 A and B and 3B). The remaining 20% of moribund pten^+/- embryos displayed other placental abnormalities, including disorganization of the spongiotrophoblast layer and limited invasion of the trophoblast layer by the chorion (Fig. 3C).

Fig. 3.

Placental defects in pten^+/- embryos. (A–C) Histological sections of placental attachment in a wild-type and two pten^+/- embryos. The mutants show defects in chorioallantoic fusion (B) or functional placental development (C) relative to the pten^+/+ embryo (A). (D and E) Histological sections of a wild-type (D) and a moribund pten^+/- (E) embryo stained for phospho-PKB. Phospho-PKB is elevated in the trophoblast layers of the developing pten^+/- placenta. (F and G) Histological sections of a wild-type (F) and a moribund pten^+/- (G) embryo stained for phospho-Erk. a, allantois; ch, chorion; g, giant cells; h, heart; ma, maternal portion of the placenta; sp, spongiotrophoblast cells.

The placental defects in developing pten^+/- embryos are reminiscent of those seen in mutants deficient for other molecules invloved in RTK signaling. Mice lacking the fibroblast growth factor receptor fgfr2 display defects in both chorioallantoic fusion and subsequent placental development (29). Hypomorphic mutation of the adaptor protein grb2 also displays defective labyrinth development (30), and mutations in gab1, sos1, and some raf family members also result in placental defects (31). These data suggest that normal Grb2 and Ras signaling are critical for the proper formation of the placenta.

Although Ras can activate PI3′K, PI3′K can also activate Ras in a cell type- and stimulus-specific manner (32, 33). To determine whether the placental defects in pten^+/- mice were caused by improper Ras activation, activation of the Ras and PI3′K signaling pathways were examined in moribund pten^+/- embryos by immunohistochemical staining for phosphorylated Erk and PKB markers for Ras and PI3′K activation, respectively. No differences in phospho-Erk levels could be detected in tissues of pten^+/+ and moribund pten^+/- embryos (Fig. 3 F and G), indicating that the pten^+/- placental defects cannot be explained by elevated Erk signaling. In contrast, phospho-PKB levels were markedly increased in the spongiotrophoblast cells of the abnormally developing pten^+/- chorion relative to both the wild-type and normally developing pten^+/- chorion (Fig. 3 D and E). This substantial increase in phosphorylated PKB suggests that pten is haploinsufficient in chorionic tissue, because one copy of the pten gene is unable to maintain normal phospho-PKB levels. In all moribund pten^+/- embryos, elevated PI3′K but not Erk signaling in the placenta is associated with lethality.

Defects in both embryonic and extraembryonic tissues are associated with the pten^+/- partial lethality. It is difficult to determine whether defects in embryonic development are the cause or consequence of improper placental formation. Given the aberrant phospho-PKB staining in the pten^+/- chorion, it is unlikely that the placental abnormalities are secondary to the observed embryonic defects, as PKB is downstream of PI3′K and thus serves as a marker for tissues likely to be directly affected by decreased PTEN levels. However, it also seems unlikely that the defects seen in the neural tube can be explained solely by inadequate nutritional supply due to defective placental formation, because the embryonic defects occur concurrent with, not subsequent to, the observed placental defects. Therefore, it appears that there are multiple, simultaneous abnormalities in the pten^+/- embryo at E9, including placental and neuronal defects.

grb2 Heterozygosity Can Completely Rescue the pten^+/- Lethality. There are at least three mechanisms driving PI3′K activation, and the relative contributions of these mechanisms to the developing embryo were investigated. Grb2 plays a major role in receptor-proximal signaling in two of the three pathways leading to PI3′K activation. The Grb2-Sos-Ras and the Grb2-Gab1 pathways are both dependent on Grb2, whereas direct binding of PI3′K to phosphorylated RTK is independent of Grb2. Consequently, grb2 heterozygosity should distinguish between direct and indirect mechanisms of PI3′K activation. Mice doubly heterozygous for pten and grb2 were generated, and embryonic lethality was investigated. Notably, the pten^+/- embryonic lethality was completely rescued in a grb2^+/- background (Table 1). This result suggests that, in the absence of regulation by PTEN, PI3′K is abnormally activated by Grb2-mediated pathways, and that this excessive activation disrupts placental and embryonic development. Conversely, normal placental and embryonic development should rely on indirect PI3′K activation through Grb2.

grb2 Heterozygosity Does Not Affect Tumorigenesis in pten^+/- Mice. Mutations of both Ras and PTEN play major roles in human tumorigenesis (1, 34). Similarly, pten^+/- and tissue-specific pten^-/- mice develop tumors in multiple organ systems (22, 25, 35–38), as do mice expressing activated Ras (39, 40). Ras-induced tumorigenesis is often associated with increased PI3′K activity. For example, 3D cultures of mammary epithelial cells expressing Ras are hyperproliferative only when able to signal through PI3′K (41). Ex vivo, tumorigenic keratinocytes from the farnesylated Sos transgenic mouse show increased phosphorylation of PKB but not Erk (42).

grb2 heterozygosity is sufficient to delay tumor formation in the polyoma middle T antigen transgenic mouse (24), animals in which tumorigenesis is Grb2 dependent. grb2 heterozygosity should therefore be sufficient to reduce tumor formation in pten^+/- mice if these tumors are exquisitely sensitive to Grb2 levels. However, pten^+/-; grb2^+/- mice developed the same tumor types at the same frequency as pten^+/-; grb2^+/+ controls (Table 2), indicating that tumorigenesis in pten^+/- mice is not rescued by grb2 heterozygosity. The types of malignancies observed were consistent with those described in pten^+/- mice (20, 22), including neoplastic lesions in the breast, endometrium, prostate, adrenal gland, and lymphoid organs. Histological examination of all tumors from pten^+/-; grb2^+/- and pten^+/-; grb2^+/+ mice did not reveal any differences in tumor stage, grade, or metastatic potential.

Table 2. Percent tumorigenesis in aged Pten^+/- and Pten^+/-; Grb2^+/- mice.

	Hyperplasia
	Endometrium			Prostate			Breast	Lung	Adrenal	Lymphoid organs
Genotype	Low	High	Carc.	Low	High	Carc.				Lympho-prolif.	Lymphoma
Wild type	0	0	0	0	0	0	0	10	0	5	10
Grb2+/-	0	0	0	0	0	0	0	3	3	10	9
Pten+/-	12	41	47	50	10	10	8	7	43 (20)	43	36
Grb2+/-;Pten+/-	10	50	40	36	27	0	18	5	28 (10)	26	43

Numbers in parentheses indicate the percentage of adrenal tumors (pheochromocytomas) that metastasised to the lung. Carc, carcinoma; Lymphoprolif., lymphoproliferative disease. Low and high refer to hyperplastic grade. All numbers are as percentages. Females, n = 9 (wt), 14 (Grb2^+/-), 17 (Pten^+/-), 20 (Pten^+/-; Grb2^+/-). Males, n = 12 (wt), 17 (Grb2^+/-), 11 (Pten^+/-), 18 (Pten^+/-; Grb2^+/-). No statistically significant differences between Pten^+/- and Grb2^+/-;Pten^+/- were found in any tumor type or organ system by using the χ² test (P = 0.05) or the Mann–Whitney U test (P = 0.05).

Grb2 Heterozygosity Does Not Affect Tumorigenesis in T Cell-Specific pten^-/- Mice. Tumors in pten^+/- mice develop within 1 year (22). In the T cell-specific conditional mutant of pten (tpten^-/-), thymic lymphomas develop more rapidly, with a mean age of onset of 60 days (25). The grb2 null allele was introduced into the tpten^-/- strain and tumorigenesis was examined in these animals. Again, no differences in tumor penetrance, age of onset, or tumor grade were observed between tpten^-/-; grb2^+/+ and tpten^-/-; grb2^+/- mice (Fig. 4). Thus, although heterozygosity for grb2 rescues pten^+/- embryonic lethality, it cannot prevent tumor development in either the pten^+/- or the tpten^-/- mouse models.

Fig. 4.

Kaplan–Meyer survival plot for tpten^-/-; grb2^+/+ and tpten^-/-; grb2^+/- mice. Mice were killed when tumors measured 1 cm across or at the first signs of morbidity. There is no significant difference between the survival or tumorigenesis in tpten^-/-; grb2^+/+ and tpten^-/-; grb2^+/- mice. tpten^-/-; grb2^+/+, n = 11; tpten^-/-; grb2^+/-, n = 10.

Grb2-Deficient Murine Embryonic Fibroblasts (MEFs). Early embryonic lethality prevents the generation of grb2^-/- fibroblasts (24). The grb2 E89K hypomorphic allele (30) was therefore used to establish fibroblast lines deficient in grb2. Although immortalized grb2^-/E89K fibroblasts grew normally in 10% serum (Fig. 5B and data not shown), these fibroblasts did not grow in 0.5% serum relative to wild-type fibroblasts (Fig. 5A). This growth defect appears to be due to increased cell death without any change in proliferation, as grb2^-/E89K cells grown in low serum have increased staining of a sub-G₁ cell population with a normal S-phase population (Fig. 5B). Interestingly, this defect is also rescued by heterozygosity for pten (Fig. 5A). These data suggest that growth factors present in serum signal through Grb2 to PI3′K to prevent cell death, and that this signaling is regulated by PTEN.

Fig. 5.

Grb2 hypomorphic fibroblasts are defective in Gab1-dependent signaling to PKB. (A) Fibroblasts (1 × 10⁴) of the indicated genotypes were plated per well of a six-well plate in media containing 10% serum. Six to 8 h later, the media was replaced with low serum media (0.5% serum) and the plates were incubated for 72 h. Viable cells were counted by trypan blue exclusion assay. Error bars indicate the SEM from three independent experiments with two cell lines from each genotype. (B) Propidium iodide staining of Grb2 hypomorphic cells grown in either 10% or 0.5% serum. M4 represents the population of cells undergoing cell death. (C) Phospho-PKB and phospho-Erk levels were determined in fibroblast lysates by Western blotting. Cells were starved in 0.5% serum media for 18 h and then stimulated with PBS (lanes 1–3) or 2 ng/ml EGF (Sigma; lanes 4–6) for 2 min. These blots were stripped and reprobed with anti-PKB and anti-Erk antibodies. Identical results were obtained when two independent cell lines from each genotype were used. (D) Grb2 hypomorphic cells were retrovirally infected with the indicated myc-tagged Grb2 constructs, and stable integrants were selected with puromycin. Survival in low serum was assayed as in C. (E) The stable integrants from D were starved for 18 h and then stimulated with EGF (10 ng/ml, 2 min). (F) Stable integrants of the indicated construct were immunoprecipitated with anti-myc, and the levels of Sos1 and Gab1 coimmunoprecipitating with each Grb2 construct was examined by Western blotting. Each of these cell lines contained an equal amount of Sos1 or Gab1 as indicated by Western blotting of whole cell lysates (WCL).

EGF signaling has been characterized in mammals, Caenorhabditis elegans, and Drosophila. In these systems, signaling from EGF family molecules depends on Grb2 homologues (43, 44). Therefore, EGF was used to stimulate immortalized grb2^-/E89K MEFs, and activation of Erk and PKB was examined. EGF stimulation of wild-type cells increased both phospho-Erk and phospho-PKB levels, whereas EGF stimulation of grb2^-/E89K cells did not result in either PKB or Erk phosphorylation (Fig. 5C). Therefore, efficient phosphorylation of PKB in response to EGF stimulation in fibroblasts depends on Grb2. Consistent with the ability to survive in low serum, grb2^-/E89K; pten^+/- MEFs show elevated phospho-PKB but not phospho-Erk in response to EGF (Fig. 5C). Thus, EGF stimulation of both the PKB and Erk pathways depends on Grb2, and PKB but not Erk phosphorylation correlates with growth in low serum.

There are two putative mechanisms through which Grb2 might signal to the PI3′K pathway. One of these mechanisms signals through Sos, whereas the second mechanism signals through Gab1. These pathways are not mutually exclusive, and may cooperate to elicit the appropriate biological response. To distinguish between these two mechanisms, truncations and point mutants were generated in the N- and C-terminal SH3 domains of Grb2. The W36K and W139K mutations abrogated binding of Grb2 to Sos1 and Gab1, respectively (Fig. 5F). Truncation mutants of either the N- or C-terminal SH3 domain prevented interaction with either Sos1 or Gab1, although both point mutants showed reduced affinity for both molecules relative to wild-type Grb2, likely because of a disruption of tertiary protein structure (Fig. 5F). These myc-tagged constructs were retrovirally infected into grb2^-/E89K MEFs, and the ability of these mutants to prevent apoptosis in low serum was examined. Whereas the N-terminal mutants prevented apoptosis in the grb2^-/E89K MEFs in low serum to the same extent as wild-type Grb2, both of the C-terminal mutants were unable to rescue cell death (Fig. 5D). This phenomenon also depends on the ability of Grb2 to bind to phosphotyrosine motifs, because a point mutant within the SH2 domain of Grb2 (R86K) was also unable to rescue this phenotype (Fig. 5D). Phosphorylation of both PKB and Erk in response to EGF was rescued by retroviral infection of grb2^-/E89K MEFs with either the N-terminal truncation mutant or by the W36K point mutant, but not by the C-terminal truncation mutant or the W139K point mutants (Fig. 5E). Signaling to both the PI3′K and Erk pathways in fibroblasts depends on the C-terminal SH3 domain of Grb2, possibly through its ability to interact with Gab1.

These data are somewhat in disagreement with the classical Grb2-Sos-Ras pathway for activation of Ras. However, in fibroblasts, it does not appear that this pathway is dominant, because Gab1-deficient fibroblasts display stark defects in Erk signaling (45), whereas Erk signaling defects in Sos1-deficient fibroblasts are less striking (46) and Sos2-deficient fibroblasts do not have Erk signaling defects (47). This may in part be explained by functional redundancy between the two mammalian Sos members, although this redundancy does not exist in all tissues (46). It seems likely that fibroblasts predominantly use the Gab1 pathway for Ras activation, whereas other systems activate Ras primarily through Sos (43).

Our data support a model in which Grb2 is able to decrease the threshold of activation for PI3′K. Although PI3′K can still be activated through direct binding of p85 to specific phosphotyrosine motifs, activation of Grb2 is able to decrease the amount of signal required for PI3′K-dependent biochemical changes. Thus, at higher levels of EGF stimulation, phosphorylated PKB is normal in grb2^-/E89K MEFs (data not shown), but at low levels of EGF stimulation, PKB activation depends on the presence of Grb2 (Fig. 5C). PTEN, as a negative regulator of PI3′K signaling, is also able to alter the threshold of activation. Thus, even in grb2^-/E89K MEFs, where there is little stimulation above background in response to low levels of EGF stimulation, pten heterozygosity is able to increase phospho-PKB and allow growth in low serum conditions (Fig. 5C). In accordance with this model, aberrant embryonic development in the pten^+/- embryos is rescued by heterozygosity at the grb2 locus, potentially because of decreased PI3′K signaling through Grb2 (Table 1).

The complete loss of PTEN renders the PI3′K pathway hyper-sensitive to stimulation by growth factors, and the effects of Grb2 become masked. For example, in pten^-/- MEFs, the presence or absence of Grb2 has no effect on either cell-based readouts such as survival in low serum or on levels of phosphorylated PKB (Fig. 5A and data not shown). Furthermore, in vivo phenotypes resulting from the complete loss of pten are unaffected by Grb2 heterozygosity. Because tumorigenesis in the pten^+/- mice involves loss of heterozygosity at the pten locus (22), the tumors arising in pten^+/- mice represent a pten-null situation. Thus, Grb2 heterozygosity is unable to alter pten^-/- embryonic lethality and tumorigenesis in tpten^-/- or pten^+/- mice (Table 2, Fig. 4, and data not shown), all of which are pten-null situations. Therefore, Grb2 is either dispensable for in vivo pten^-/- phenotypes, or the remaining grb2 wild-type allele is sufficient to transduce the required signal.

Our data demonstrate that PI3′K signaling is, in certain contexts, critically dependent on a Grb2-dependent pathway. This study highlights the complexity of interactions between signaling pathways regulated by PTEN and Grb2 during both embryonic development and tumorigenesis by using an in vivo loss of function model. Indeed, these differences in Grb2-PI3′K crosstalk underscore the importance of a thorough understanding of signaling pathways for the development of antitumor drugs by rational design.