Essential role of B-Raf in ERK activation during extraembryonic development

Abstract

The kinases of the Raf family have been intensively studied as activators of the mitogen-activated protein kinase kinase/extra-cellular signal-regulated kinase (ERK) module in regulated and deregulated proliferation. Genetic evidence that Raf is required for ERK activation in vivo has been obtained in lower organisms, which express only one Raf kinase, but was hitherto lacking in mammals, which express more than one Raf kinase. Ablation of the two best studied Raf kinases, B-Raf and Raf-1, is lethal at midgestation in mice, hampering the detailed study of the essential functions of these proteins. Here, we have combined conventional and conditional gene ablation to show that B-Raf is essential for ERK activation and for vascular development in the placenta. B-Raf-deficient placentae show complete absence of phosphorylated ERK and strongly reduced HIF-1α and VEGF levels, whereas all these parameters are normal in Raf-1-deficient placentae. In addition, neither ERK phosphorylation nor development are affected in B-raf-deficient embryos that are born alive obtained by epiblast-restricted gene inactivation. The data demonstrate that B-Raf plays a nonredundant role in ERK activation during extraembyronic mammalian development in vivo.

The placenta is the first organ to develop during embryogenesis, and it supports the growth of the developing embryo by mediating the exchange of nutrients and wastes between the fetal and maternal circulatory systems. Placentation includes extensive angiogenesis, and reduced placental vascular development is associated with early embryonic mortality. Genetic studies have demonstrated a crucial role of VEGF, FGF, and their receptors in placental angiogenesis. In addition, the ablation of several signaling molecules operating downstream of receptor tyrosine kinases results in defects in placentation, often at the stage of labyrinth formation (1).

The Raf kinases (A-Raf, B-Raf, and Raf-1) relay signals from tyrosine kinase receptors to the mitogen-activated protein kinase kinase (MEK)/extracellular signal-regulated kinase (ERK) signaling module. Although most of the early work on the activation of the MEK/ERK module was focused on Raf-1, evidence has accumulated that B-Raf is the main MEK kinase. Raf kinases from lower organisms (Caenorhabditis elegans lin-45 and Drosophila D-Raf) are more similar to B-Raf than to the other two mammalian Raf kinases. Biochemical studies have indicated that B-Raf is the main MEK kinase found in fibroblast and brain lysates (2-5). Consistently, among the three Raf kinases, B-Raf binds best to MEK (6) and has the highest basal MEK kinase activity, both in vitro (7) and in fibroblasts, when expressed as a conditionally oncogenic form (8). Finally, B-Raf mutations resulting in increased MEK/ERK activation have been discovered in a broad range of human tumors (9). All these observations hint at B-Raf as the archetypal mammalian MEK kinase, whereas Raf-1 and A-Raf have probably diverged to perform other functions. Growth-factor-stimulated ERK activation is reduced in cells lacking B-Raf but not in A-Raf- or Raf-1-deficient cells (10-14). However, none of the kinases that activate MEK in vitro or in cultured cells has been shown to be essential for the activation of the ERK module in vivo.

Ablation of B-Raf, Raf-1, MEK-1, and ERK2 results in embryonic death between embryonic day (E)8.5 (ERK-2) and E12.5 (Raf-1) (13, 15-19). Defects at various stages of placental development have been observed in embryos lacking Raf-1 (13, 14), MEK-1 (15), and, depending on the targeting strategy used, ERK-2 (16, 19). In contrast, B-Raf ablation has been reported to compromise the survival of mature endothelial cells in the embryo proper (18). Although the availability of a conditional knockout has helped establish a MEK-independent role of Raf-1 in apoptosis and migration in vivo (20, 21), follow-up work on the effects of B-Raf and MEK-1 ablation has been difficult because of early embryonic lethality. Here, we use conditional mutagenesis to show that the essential role of B-Raf in intrauterine life is restricted to extraembryonic development and that the anomalies observed in the B-Raf knockout (KO) embryos are secondary to placental defects. In addition, we show that B-Raf ablation abrogates ERK phosphorylation in the trophoblast but not in the epiblast. Lack of phosphorylated ERK (pERK) is accompanied by a dramatic reduction in HIF-1α and VEGF levels. In contrast, ERK activation is unperturbed in Raf-1 KO placentae, and epiblast-restricted ablation fails to rescue embryonic lethality. These data show a nonredundant role of B-Raf as a MEK/ERK activator in the developing placenta in vivo and highlight the significance of the B-Raf/MEK/ERK pathway for angiogenesis in this organ.

Results and Discussion

To circumvent early embryonic lethality by B-raf inactivation, a conditionally targeted B-raf allele (B-raf ^f) was generated (22). B-raf^f (Fig. 1A) contains loxP sites cloned 5′ and 3′ of exon 12, which encodes the start of the kinase domain. B-raf^f/f animals were bred to Mox2^+/cre transgenic mice (23) to obtain Mox2^+/cre;B-raf^Δ/+ mice. Offspring were genotyped by PCR (Fig. 1A). Mox2^+/cre;B-raf^Δ/+ animals were bred to WT to obtain B-raf^-/+ mice with a WT Mox2 locus. B-raf^-/+ crosses did not yield any viable B-raf^-/- offspring after E11.5, at which stage 12% of B-raf^-/- embryos could be recovered (Fig. 1B). Analysis of mouse embryonic fibroblasts (MEFs) showed that exon-12 excision completely abrogated B-Raf expression (Fig. 1C) and B-Raf kinase activity (Fig. 1D). In addition, as described for the ablation of B-Raf exon 3 (10, 11), growth factor-induced ERK phosphorylation was impaired over a wide range of concentrations (data not shown) and time points (Fig. 1E) in B-Raf KO MEFs. Regardless of the defect in pERK, however, B-Raf ablation decreased proliferation only marginally in the presence of FCS (Fig. 1F).

Fig. 1.

Exon-12 excision abrogates B-Raf expression and kinase activity and causes lethality at midgestation. (A) Targeting strategy and PCR analysis. Primers 1 and 2 amplify a 357-bp fragment of the endogenous allele and a 413-bp fragment of the floxed allele, and primers 1 and 3 amplify a 282-bp fragment of the targeted allele. Tail biopsies were used for the PCR. (B) The percentage of B-raf^-/- conceptuses of 153 viable embryos recovered from B-raf^+/- intercrosses is shown as a function of gestational age. Sixty embryos were recovered at E10.5, 63 at E11.5, and 30 at E12.5. (C) Immunoblot of whole-cell lysates from MEFs. (D) Immunocomplex kinase assay of B-Raf i.p.s from +/+ or -/- MEFs stimulated with EGF (33 nM for 5 min). One representative experiment of three is shown. ADU, arbitrary densitometric units. (E) +/+ and -/- MEFs were stimulated with 30 ng/ml FGF for different time periods before cell lysis. The presence of pMEK, pERK, and MEK2 (loading control) was detected by immunoblotting. (F) Proliferation of primary +/+ and -/- MEFs monitored for 8 days in culture. The plot shows the mean of three independent WT and KO lines, respectively. Vertical bars, standard deviation of the mean.

B-Raf KO embryos at E11.5 were significantly smaller than WT or heterozygous littermates (Fig. 2A), and extensive apoptosis could be detected in liver, brain, and heart (data not shown). Mutant placentae were similar in diameter to those of WT or heterozygous littermates, but the fetal part was less vascularized and thinner. E11.5 and E10.5 WT or heterozygous placentae displayed an organized three-layered structure, in which the intensely vascularized innermost layer, the labyrinth, was lined by a continuous layer of spongiotrophoblasts. A third layer of giant trophoblast cells demarcated the boundary to the maternal deciduas (Fig. 2B). In the mutant placenta, the spongiotrophoblast and the giant trophoblast layers were discontinuous, and the labyrinth layer was severely underdeveloped. Whereas the WT labyrinth contained abundant blood vessels of fetal origin (characterized by the presence of nucleated fetal erythrocytes) in close contact with the maternal sinuses (filled with mature, enucleated RBCs), the KO labyrinth contained few patent blood vessels that did not make proper contact to the maternal blood sinuses and large hypocellular areas filled with stroma (Fig. 2B). Dilated blood vessels and hemorrhage were observed in E11.5 placenta (Fig. 2B), and the embryonic part was characterized by extensive apoptosis (Fig. 2C). Defects in placenta development were already evident at E10.5 (Fig. 2B Bottom), although the density of proliferating (Ki67+) cells at this time was comparable between WT and KO organs (Fig. 2D), and apoptosis was undetectable (data not shown).

Fig. 2.

B-raf ablation perturbs embryonic development. (A) Phenotype of an E11.5 -/- fetus and a +/+ littermate. -/- embryos are developmentally retarded and smaller than their +/+ littermates, and the placentae are less vascularized. (B) Placental defects in -/- embryos. Hematoxylin and eosin-stained radial sections of E11.5 and E10.5 placentae from +/+ and -/- littermates are shown. In the -/- placentae, the spongiotrophoblast (Sp) and labyrinth layer (L) are severely underdeveloped and disorganized compared with +/+. (Top and Bottom) The dotted lines mark the boundaries between the labyrinth and the spongiotrophoblast layer and between the giant cell layer and the maternal decidua (De). (Insets) Higher magnifications. Thick filled arrows, giant cells; open arrows, hemorrhages. (Middle) Labyrinth architecture in +/+ and -/- E11.5 placentae. t, trophoblasts; s, stroma; thin filled arrows, endothelial cells; filled arrowheads, nucleated embryonic RBCs; open arrowheads, maternal RBCs. (C) Massive apoptosis in -/- placenta at E11.5 revealed by TUNEL staining. (Lower) Adjacent sections stained with hematoxylin to show tissue structure. (D) proliferating cells (Ki67⁺, brown staining) are present in E10.5 +/+ and -/- placentae. (Scale bar, 500 μm.)

To determine whether the severe placental defects caused the death of the B-Raf KO embryos, we performed epiblast-restricted ablation by crossing B-raf^f/f animals to Mox2^+/cre;B-raf^Δ/+ animals, completely rescuing the placental phenotype (Fig. 3A Top) as well as the widespread apoptosis observed in the embryo proper (data not shown). Mox2^+/cre;B-raf^Δ/- pups were born at a Mendelian ratio (n = 135) and were indistinguishable from their B-raf^f/+, B-raf^f/-, or Mox2^+/cre;B-raf^Δ/+ littermates (data not shown), although complete conversion of flox to null alleles could be observed in E10.5 embryos and in all adult tissues examined (Fig. 3B). However, Mox2^+/cre;B-raf^Δ/- animals showed progressive growth retardation and died around postnatal day 21 of an aggressive neurodegenerative disease (G.G.-K., D.M., and M.B., unpublished work).

Fig. 3.

Epiblast-restricted ablation rescues embryonic lethality due to lack of B-raf. (A) Lack of placental defects in embryos with epiblast-restricted B-Raf or Raf-1 ablation. Radial sections of E11.5 Mox2^+/cre;B-raf ^f/-, Mox2^+/cre;c-raf-1^f/-, and the respective WT placentae. De, decidua; L, labyrinth. (B) Complete conversion of the B-raf ^f/- to the B-raf^Δ/- genotype in tissue samples of three E10.5 Mox2^+/cre;B-raf-1^Δ/- embryos (embryo codes 230, 233, and 234; Upper) and of a Mox2^+/cre;B-raf-1^Δ/- adult animal as determined by PCR analysis (Lower); B, brain; S, spleen; Li, liver; H, heart; L, lung; K, kidney; St, stomach; N, negative control. (C) Fetal liver apoptosis in Mox2^+/cre;c-raf ^Δ/- embryos. Parasagittal sections of E11.5 livers stained with TUNEL are shown. (Scale bar, 500 μm.)

The data above establish that the essential function of B-Raf during intrauterine life is the control of placental development and that the defects observed in the KO embryos from B-raf^+/- crosses are a consequence of placental failure. Placental anomalies, albeit less pronounced than those caused by B-raf ablation, are also a hallmark of embryos lacking Raf-1 (13). However, epiblast-restricted KO did not rescue embryonic lethality due to Raf-1 ablation, and, although Mox2^+/cre;c-raf-1^Δ/- embryos were present at a Mendelian ratio on E10.5, live offspring of this genotype could not be obtained (n = 50; data not shown). Consistently, the fetal liver apoptosis caused by Raf-1 ablation (13) was still evident in the Mox2^+/cre;c-raf-1^Δ/- embryos (Fig. 3C), regardless of the rescue of the placental defects (Fig. 3A Lower), indicating that this and possibly other alterations in the Raf-1 KO embryos are not secondary to placental insufficiency and are the cause of embryonic lethality.

To gain some insight into the molecular mechanisms downstream of B-Raf in the placenta, we examined pERK in WT and B-Raf KO organs. At E10.5, massive ERK phosphorylation was detectable in the WT labyrinth, in particular in the vicinity of developing blood vessels and in giant trophoblast cells; in contrast, pERK was virtually absent in B-raf^-/- placentae. In contrast, similar levels of pERK could be detected in c-raf-1^-/- and control placentae (Fig. 4A). Thus, B-Raf, but not Raf-1, is required for ERK activation during placental development in vivo and in MEFs in vitro (Figs. 1E and 4A; and see ref. 13).

Fig. 4.

B-Raf is essential for ERK activation in the developing placenta but not in the embryo proper. (A) pERK immunohistochemistry of E10.5 WT, B-raf^-/-, and c-raf-1^-/- placentae. pERK reactivity can be readily visualized as brown staining in WT and Raf-1 KO but not in B-Raf KO placenta. Dotted lines, boundaries between the labyrinth (L) and the spongiotrophoblast layer and between the giant cell layer and the maternal decidua (De). Thick filled arrows, giant cells. (Lower) Higher magnifications of the labyrinth. t, trophoblast; filled arrowheads, developing embryonic blood vessels with nucleated RBCs, open arrowheads, maternal sinuses; thin filled arrows, endothelial cells. (B) Whole-mount pERK staining of E10.5 WT and Mox2^+/cre;B-raf^Δ/- embryos. pERK staining in eye primordia (ey), branchial arches (ba), frontonasal processes (fnp), limb buds (lb), and liver primordia (l). One representative pair of three is shown. (Scale bar, 250 μm.) (C) Expression of B-Raf, Raf-1, and A-Raf in E11.5 WT placenta (P) and embryo (E). P_M, maternal, P_E, embryonic part of the placenta; H, head; B, body of the embryo. Fifty micrograms of organ lysates were loaded in each lane. A MEK-2 immunoblot is shown as a loading control.

To assess whether B-Raf was essential for ERK phosphorylation in the embryo proper, we performed whole-mount immunohistochemistry on E10.5 Mox2^+/cre;B-raf ^Δ/- embryos to avoid recording effects due to the placental insufficiency. As described in ref. 24, pERK was evident in eye primordia, branchial arches, frontonasal processes, limb buds, and liver primordia, and pERK was unimpaired in B-Raf KO embryos (Fig. 4B). Thus, B-Raf is not required for ERK activation in the embryo proper in vivo.

The data above suggested a differential expression of Raf kinases in the E11.5 embryo and placenta. Indeed, immunoblot analysis revealed that, at E11.5, B-Raf was expressed at higher levels in the placenta, more precisely in the embryonic part of this organ, than in the embryo proper. Raf-1 showed the opposite distribution, whereas A-Raf was expressed at very similar levels in both placenta and embryo (Fig. 4C). The distribution of B-Raf is in line with its prominent role in ERK activation and placental development, yet it is surprising that, although Raf-1 and A-Raf are expressed in the placenta, they cannot compensate the lack of B-Raf in ERK activation and tissue development.

Angiogenesis is regulated by VEGF in both the placenta and the embryo proper. During placentation, VEGF is produced by the giant trophoblast cells (25) in a pattern consistent with paracrine stimulation of the developing vascular endothelium. ERK signaling has been implicated in regulating the VEGF promoter in response to a variety of stimuli under hypoxic or normoxic conditions (26, 27). In particular, ERK activation is required for A isoform of VEGF (VEGF-A) expression by human cytotrophoblast cells in vitro (28, 29), suggesting that lack of VEGF production may correlate with decreased ERK phosphorylation in B-Raf KO placentae. Indeed, immunohistochemical analysis of E10.5 B-Raf-deficient placentae revealed a dramatic decrease in VEGF-A expression compared with control littermates (Fig. 5A).

Fig. 5.

B-Raf is essential for VEGF production in the developing placenta and in MEFs in culture. (A) Immunohistochemistry of +/+ and -/- E10.5 placentae. VEGF-A and HIF-1α reactivity can be readily visualized as brown staining in +/+ but not in -/- placentae. Note the VEGF staining in the maternal part of the -/- placenta. (Insets) Higher magnifications. (B) Accumulation of nuclear HIF-1α is impaired in B-raf^-/- MEFs treated with CoCl₂ (100 μM for 3 h). Fifty micrograms of nuclear extracts were loaded in each lane. A CREB immunoblot is shown as a loading control. (C) Lack of B-Raf reduces FCS- and CoCl₂-induced VEGF transcription. +/+ and -/- MEFs were cotransfected with a VEGF reporter construct (-1176/+54) and with a plasmid-encoding β-galactosidase to normalize for transfection efficiency. Transfected cells were either left unstimulated or were treated with (FCS 10% for 2.5 h) or CoCl₂ (100 μM, for 24 h) before lysis and luciferase activity measurements. (Scale bar, 250 μm.)

ERK can regulate VEGF transcription through a number of elements contained in the proximal region of the VEGF-A promoter (27). Activated ERK helps recruit the transcription machinery to the AP-2/Sp1 site (26). In addition, ERK can modulate the activity of the transcription factor HIF-1, which binds to the hypoxia-responsive element in the VEGF promoter and is the main player in hypoxia-induced VEGF transcription (30). HIF-1 is composed of two subunits, the constitutively expressed HIF-1β and the unstable HIF-1α subunit. HIF-1α stabilization is the rate-limiting event in HIF-1 activation and is strongly induced by low oxygen tension but also, under normoxic conditions, by a variety of growth factors and cytokines (31). Functional HIF-1 is required for mouse placental development, and it has been implicated in murine (32, 33) and human trophoblast differentiation (34). In human cytotrophoblasts, inhibition of ERK activation impairs both HIF-1α accumulation and VEGF synthesis induced under normoxic conditions by IL-1β and TGF-β (28, 29). In good correlation with the lack of ERK phosphorylation, HIF-1α expression was clearly reduced in E10.5 B-Raf KO placentae (Fig. 5A). In addition, nuclear accumulation of HIF-1α was reduced in B-raf^-/- MEFs treated with CoCl₂ to mimic hypoxia (Fig. 5B). Consistent with these results, transcription of a luciferase construct driven by the VEGF-A promoter was severely impaired in B-Raf KO cells stimulated with either growth factors or CoCl₂ (Fig. 5C). Thus, B-Raf contributes to VEGF expression in vivo and in vitro. At this point, it is unclear whether the profound deficiency in VEGF expression observed in the KO placenta is due to reduced HIF-1α expression, or to the impaired activation of the AP-2/Sp-1 promoter element, or to a combination of both.

In addition to its effects on VEGF production (Fig. 5), B-Raf may play a role in endothelial cell survival (18). Hence, defects in endothelial cell differentiation/survival might contribute to the failure to form a proper labyrinth in the absence of B-Raf. To test this idea, we generated Tie2-Cre;B-raf^f/f mice, which express Cre in all endothelial cells and in the majority of hematopoetic cells (35). Tie2-Cre;B-raf^f/f mice were born at a Mendelian ratio, were of normal size, and were healthy and fertile. The architecture of the placenta was normal at E11.5 (see Fig. 6, which is published as supporting information on the PNAS web site). Thus, B-Raf is not required for endothelial cell proliferation, differentiation, or survival in the embryo and/or in adult mice. Together with the data mentioned above, the results indicate that the defects in placentation of B-Raf KO embryos are due to the reduced amount of VEGF produced by the mutant placenta, rather than to a cell-autonomous defect of endothelial cells.

The data above identify B-Raf as the nonredundant ERK activator in mouse placenta in vivo. This result, corroborated by the concomitant investigation of Raf-1 KO mice, was unexpected, given the number of kinases that are able to activate the MEK/ERK module and the selective pressure imposed by the technique. Up to now, in fact, the genetic reconstruction of the Raf/MEK/ERK pathway could be achieved only in simpler organisms like Drosophila, which expresses only one Raf form. Although A-Raf and Raf-1 activate MEK less efficiently than B-Raf, they are expressed in the placenta (Fig. 4C), and one might have expected them to take over in the absence of B-Raf. Indeed, this argument has been regarded as a likely explanation for the unimpaired ERK activation in Raf-1 KO cells and tissues and has suggested that the presence of enzymes with similar or identical substrate specificity may prevent the identification of essential kinase-dependent functions in geneablation experiments. The lack of ERK activation in the B-Raf-deficient, but not in the Raf-1 KO, placenta demonstrates that this is not always the case. The reason why B-Raf is necessary for ERK activation in the context of the placenta and of MEFs in vitro, but not of the embryo proper in vivo, is unknown. It is conceivable that the stimulation of MEFs in vitro may not be representative of the situation in vivo, where the cells receive a mixture of signals in the context of a tissue. In vivo, the differences observed in the behavior of the placenta and the embryo proper may be cell-autonomous, i.e., some cells may regularly use B-Raf-independent mechanisms for ERK activation production or may up-regulate such mechanisms in the absence of B-Raf, whereas others may not be able to do so. The specular distribution of B-Raf and Raf-1 in placenta and embryo (B-Raf higher in the placenta than in the embryo and Raf-1 vice versa) lends some support to this hypothesis. Alternatively, the difference may lie in the nature of the activators (growth factors, adhesion molecules, or hypoxia) that impinge on the pathway in embryo and placenta in vivo.

Stimuli that activate the ERK cascade, notably FGFs and their receptors, have been implicated in extraembryonic ectoderm development and survival (1). FGF4 produced by the epiblast regulates the proliferation and maintenance of trophoblast stem cells (36). FGFR2 (37), the main FGFR in these cells, and its downstream adaptor molecule FRS2α (38) are essential for trophoblast stem cell renewal. FRS2α, in particular, is required for full-fledged ERK activation in the extraembryonic ectoderm. Embryos bearing the FGFR2 mutation or lacking FRS2α die much earlier (<E8.5) than B-Raf KO embryos (E11.5). These data imply that either B-Raf is not the molecule mediating ERK activation through FSR2α at E8.5, or another upstream activator can carry out this function in B-Raf's absence. Alternatively, impaired ERK activation may not be the essential reason for the reduced survival of FRS2α^-/- trophoblast stem cells. Consistent with the latter hypothesis, activation of the PI3-Kinase/Akt survival pathway is also reduced in FRS2α KO fibroblasts treated with FGF (39). At present, we cannot distinguish between these possibilities.

Our data show that B-Raf sustains labyrinth development by regulating the production of VEGF rather than by a cell-autonomous role in survival. This finding may be of significance in the context of tumor development. VEGF is the main inducer of tumor angiogenesis. VEGF is expressed, together with its receptors, in highly vascularized tumors, and expression has been used as an indicator of increased metastatic risk (40). Hence, although the molecular mechanism regulating the “angiogenic switch” by which quiescent endothelium becomes activated by the tumor is unknown, VEGF is regarded as an important therapeutic target. Our data predict that the B-Raf/ERK module may participate in the induction of this angiogenic switch. If so, a drug targeting B-Raf would affect proliferation both directly, by decreasing ERK activation, and indirectly, by reducing VEGF production, thus interrupting the paracrine loop securing nutrient supply to the tumor. Conditional ablation of B-Raf in mouse tumor models whose progression involves an angiogenic switch will help clarify this issue.

Materials and Methods

Mice. B-raf^f/f (22) and c-raf-1^f/f (41) mice were maintained on a 129/Sv background and crossed to mice expressing the Cre recombinase from the Mox-2 locus (23) for epiblast-restricted ablation. For endothelial-cell-restricted ablation, B-raf^f/f animals were bred to transgenic mice expressing Cre under the control of the Tie2 promoter (35) (kind gift of Bernd Arnold, Heidelberg).

PCR Analysis of Offspring and Conceptuses. Tail and embryonic tissue DNA was prepared as described in ref. 13. The following primers were used for genotyping B-raf alleles by PCR: primer 1, 5′-GCATAGCGCATATGCTCACA-3′; primer 2, 5′-CCATGCTCTAACTAGTGCTG-3′; and primer 3, 5′-GTTGACCTTGAACTTTCTCC-3′. Primers 1 and 2 amplify a 357-bp fragment of the endogenous B-raf allele and a 413-bp fragment of the floxed allele, whereas primers 1 and 3 amplify a 282-bp fragment of the targeted B-raf allele. Allele-specific PCR genotyping for c-raf-1 was described in ref. 41.

Histology and Immunohistochemistry. Hematoxylin and eosin staining and immunohistochemistry were performed on 3-μm-thick sections of 4% paraformaldehyde-fixed and paraffin-embedded tissues. Staining with the following antibodies was performed: αHIF-1α (Chemicon International, Temecula, CA), αpERK (Cell Signaling Technology, Beverly, MA), αVEGF-A (Santa Cruz Biotechnology), and αKi-67 (Novo-Castra, Newcastle-upon-Tyne, U.K.). For detection, we used avidin-biotinylated enzyme complex (ABC, Vector Laboratories) or the DAKO EnVision peroxidase system, followed by incubation with 0.01% diaminobenzidine (Sigma) and counterstaining with hematoxylin. TUNEL labeling was performed according to the manufacturer's protocol (In Situ Cell Death Detection kit, Roche) and counterstained with propidium iodide. Whole-mount staining of E10.5 embryos was performed as described in ref. 24.

Cell Culture, Transient Transfection, and Luciferase Assay. Primary MEFs were isolated, cultured, and immortalized as described in ref. 13. MEFs were starved in medium containing 0.5% FCS for 18 h before treatment with different stimuli (EGF, Biomedical Technologies; and FGF and CoCl₂, Sigma). Cells were transfected by using ExGen 500 (Fermentas, St. Leon-Rot, Germany) according to the manufacturer's instructions. MEFs in 60-mm dishes were transfected with 4 μg of reporter plasmid (the luciferase gene driven by the human VEGF-A promoter -1176/+54) together with 1 μg of pSV-β-galactosidase plasmid to control for transfection efficiency. Luciferase/β-galactosidase activity was assayed as described in ref. 26. The luciferase activity of each sample was normalized to the β-galactosidase value.

Immunoprecipitation, Assay of Raf Kinase Activity, and Immunoblot Analysis. Immunoprecipitation and i.p. kinase assays were performed as described in ref. 13, except that phosphorylation of GST-ERK was used as a readout. The primary antibodies were: αB-Raf IS11 (42), kind gift of J. V. Barnier (Institut de Neurobiologie Alfred Fessard, Gif sur Yvette, France); αCREB, αpMEK, and αpERK (Cell Signaling Technology); αRaf-1 and αMek-2 (Transduction Laboratories); αB-Raf (C-19) and αA-Raf (C-20) (Santa Cruz Biotechnology); and anti HIF-1α (Novus Biologicals).