Map2k1 and Map2k2 genes contribute to the normal development of syncytiotrophoblasts during placentation

Mitogen-activated protein kinase (MAPK) signaling pathways consist in protein kinase cascades linking extracellular stimuli with various targets scattered in the different cell compartments(Rubinfeld and Seger, 2005; Seger and Krebs, 1995). The classical ERK/MAPK pathway, which appears to be the major one in growth factor signaling, involves the extracellular-signal-regulated kinases ERK1 and ERK2(renamed MAPK3 and MAPK1, respectively) and the ERK kinases (MAP2K1 and MAP2K2, also known as MEK1 and MEK2). MAP2K1 and MAP2K2 are dual-specificity kinases that activate MAPK1 and MAPK3 upon agonist binding to receptors(Crews et al., 1992). The ERK/MAPK pathway is implicated in cell-fate determination in Caenorhabditis elegans, Drosophila melanogaster and Xenopus laevis (Hsu and Perrimon,1994; Kornfeld et al.,1995; Umbhauer et al.,1995; Wu et al.,1995). In mammals, this cascade plays a crucial role in development, including fate determination, differentiation, proliferation,survival, migration, growth and apoptosis(Fischer et al., 2005; Johnson and Vaillancourt,1994; Ussar and Voss,2004).

Although two different MAP2K proteins are present in the ERK/MAPK cascade in mammals, a single Map2k gene fulfills this role in C. elegans,D. melanogaster and X. laevis. Sequence analysis revealed that the murine MAP2K1 is more related to the X. laevis MAP2K1 (X-MEK)than to the mouse MAP2K2 (see Fig. S1 in the supplementary material)(Brott et al., 1993; Crews et al., 1992; Russell et al., 1995). MAP2K2 protein is 80% identical and 90% similar to MAP2K1, whereas X-MEK presents 91%identity and 96% similarity to murine MAP2K1. Two short regions of MAP2K1 show reduced homology with MAP2K2: (1) the amino-terminus (25% identical/53%similar over 36 residues), and (2) the proline-rich insert (17% identical/33%similar over 18 residues). These two motifs are essential for MAP2K function,as they affect the efficiency of ERK phosphorylation, the phosphorylation of residue Ser386 of MAP2K1/2 by MAPK1/3, and the subcellular localization of MAP2K1/2 (Rubinfeld and Seger,2005). The amino-terminus of MAP2K1/2 proteins contains the MAPK1/3 docking site and a nuclear export sequence. The proline-rich domain is shared by MAP2K proteins from different species (see Fig. S1B in the supplementary material) (Dang et al.,1998). The proline-rich domain of MAP2K1 contains a PAK phosphorylation site (Ser298) important for its function(Coles and Shaw, 2002). This domain is also involved in the interaction with the Raf family members(Catling et al., 1995; Dang et al., 1998; Eblen et al., 2002; Nantel et al., 1998; Papin et al., 1996). In vitro,PAK phosphorylation of MAP2K1 on residue Ser298 stimulates MAP2K1 autophosphorylation on the activation loop, which leads to the phosphorylation of MAPK1/3 by MAP2K1. Although Ser298 is present in MAP2K2, MAP2K2 is a poor substrate of PAK1 (Frost et al.,1997; Park et al.,2007). In addition, the Thr residues found at positions 286 and 292 of MAP2K1, which are involved in feedback regulation of the ERK/MAPK cascade once phosphorylated, are not conserved in MAP2K2(Rubinfeld and Seger, 2005). Thus, protein sequence differences between MAP2K1 and MAP2K2 suggest that these two proteins have diverged to achieve unique functions in mammals.

A differential response of MAP2K1 and MAP2K2 to certain stimuli also exists. For instance, only MAP2K1 is activated in Swiss 3T3 and macrophage cells in response to bombesin and tumor necrosis factor α, respectively(Seufferlein et al., 1996; Winston et al., 1995). This is further supported by our observations that the ERK/MAPK cascade cannot be activated in response to bombesin in Map2k1^-/- mouse embryonic fibroblasts (M. Tremblay, S.R. and J.C. unpublished). By contrast,MAP2K2 is specifically activated by lactosylceramide in human aortic smooth muscle cells or by estradiol in mouse cerebral cortex(Bhunia et al., 1996; Setalo et al., 2002). Differential activation of MAP2K1 and MAP2K2 by Raf family members in EGF-stimulated HeLa cells has also been described(Wu et al., 1996). In addition, only MAP2K1 can form a signaling complex with Ras and c-Raf in serum-stimulated NIH 3T3 cells, suggesting that in these conditions the c-Raf signaling preferentially transits via MAP2K1(Jelinek et al., 1994). The Rac-PAK pathway has been shown to be involved in the activation of the ERK/MAPK cascade by regulating the formation of a specific MAP2K1-MAPK1/3 signaling complex, which is under the control of PAK and MAPK1/3 phosphorylation sites unique to MAP2K1 in the proline-rich sequence(Eblen et al., 2002; Eblen et al., 2004). Finally,MAP2K1 and MAP2K2 act at different steps during the cell cycle: MAP2K1 provides proliferative signals whereas MAP2K2 induces growth arrest at the G1/S boundary (Ussar and Voss,2004). The role of MAP2K1 in proliferation seems to depend on its ability to translocate MAPK1 to the nucleus following MAP2K1 Thr292 and Ser298 phosphorylation (Skarpen et al.,2007). Altogether these data support a model in which the transduction of specific signals transits via distinct protein kinase isoforms through the ERK/MAPK cascade (Acharya et al., 1998). The presence of various isoforms at the different levels of the pathway may reflect the complexity of the controls required for the fine regulation of the multiple processes in mammalian cells.

The differential role of MAP2K1 and MAP2K2 in signal transduction during mouse development was revealed by the characterization of mutant mouse lines in which Map2k1 and Map2k2 genes were disrupted(Bélanger et al., 2003; Bissonauth et al., 2006; Giroux et al., 1999). The null mutation of Map2k1 results in a recessive lethal phenotype, the mutant embryos dying at 10.5 days of gestation (E10.5) due to abnormal placenta development with marked undergrowth of the labyrinth region and reduction of its vascularization. However, Map2k1^-/- mice survive if the development of the extra-embryonic structures is rescued(Bissonauth et al., 2006). The Map2k1^-/- phenotype is observed in the presence of normal Map2k2 expression levels in the placenta providing genetic evidence that Map2k2 is unable to make up for the absence of Map2k1(Giroux et al., 1999). By contrast, Map2k2^-/- mice show no obvious phenotype,suggesting compensatory effects by Map2k1(Bélanger et al., 2003). These data support the idea of an essential role for Map2k1 during mammalian development.

To circumvent the potential functional redundancy between MAP2K1 and MAP2K2 that precludes the definition of a specific physiological role for MAP2K2 during development, Map2k alleles were successively mutated. Mice with combined Map2k1 and Map2k2 haploinsufficiency showed reduced survival rate at birth. The Map2k1^+/-Map2k2^+/- embryos dying during gestation presented an underdevelopment of the placenta labyrinth region and a reduction of its vascularization. The Map2k1^+/-Map2k2^+/- survival rate can be rescued by restricting the Map2k1 mutation to the embryo. The placenta phenotype is Map2k1/Map2k2 gene dosage-dependent with a predominant role of Map2k1. Map2k2 haploinsufficiency in the absence of one Map2k1 allele also leads to the abnormal differentiation of the syncytiotrophoblast (SynT) layer II and to the accumulation of multinucleated trophoblast giant (MTG) cells.

All experiments were performed according to the guidelines of the Canadian Council on Animal Care and approved by the institutional animal care committee. The genotyping by Southern blot and PCR analyses of the Map2k1 and Map2k2 mutant mouse lines, the Sox2Credeleter mouse line and the ROSA26 Cre reporter (R26R) mouse line have been previously described(Bélanger et al., 2003; Bissonauth et al., 2006; Giroux et al., 1999; Hayashi et al., 2003; Soriano, 1999). The mouse lines are in the 129/Sv genetic background except the R26R line,which is in a 129Sv-C57BL/6 hybrid background. The Gcm1Cre deleter mouse line was generated in the laboratory and will be described in detail in a subsequent manuscript.

Specimens were collected and processed for paraffin inclusion. Serial sections of 4 μm were stained according to standard histological procedures to identify specific cell types: Hematoxylin and Eosin, Periodic acid/Schiff(for glycogen trophoblasts) (Schmitz et al., 2007), DAPI/Phalloidin (for multinucleated cells), MSB(martius/scarlet/blue for fibrin deposits)(Lendrum et al., 1962) and alkaline phosphatase (ALP, for SynT)(Bissonauth et al., 2006).

Proliferation was assessed by immunostaining with a rabbit polyclonal antibody against the phosphorylated histone H3 (pH3; Upstate Biotechnology,Lake Placid, NY) (Aubin et al.,2002) and by BrdU labeling. Intraperitoneal injection of BrdU (100μg/g body weight) was performed 4 hours before sacrifice. BrdU incorporation was detected as described in Hayashi et al.(Hayashi et al., 1988) with some modifications. Paraffin sections were incubated in 1 N HCl for 1 hour at 40°C followed by a trypsin digestion (0.025% w/v) at 40°C for 20 minutes before detection with a mouse monoclonal antibody (1:1000; Chemicon International, Temecula, CA). Apoptotic cells were detected by terminal transferase (TdT) DNA end labeling (TUNEL) according to Giroux and Charron(Giroux and Charron, 1998). The placenta vascular network was revealed with a rat monoclonal antibody against CD31/PECAM (1:50; Pharmingen, Mississauga, ON), whereas MAP2K1 and MAP2K2 were detected using rabbit monoclonal antibodies (1:100 and 1:250,respectively; Epitomics, Burlingame, CA). Antigen retrieval was performed by microwaving samples in 0.01 M citrate buffer at pH 6.0. The Vectastain HRP ABC Reagent (Vector Laboratories, Burlingame, CA) was used for detection and the sections were counterstained with Hematoxylin. For these analyses, at least six specimens per genotype were tested and the most representative fields are presented.

Radioactive in situ hybridization on tissue sections was performed with Gcm1 and Tpbpa [³⁵S]UTP-labeled riboprobes as previously described (Bissonauth et al.,2006; Giroux et al.,1999). β-Galactosidase staining was performed on cryosections. Dissected placentas were fixed in 4% paraformaldehyde/0.2%glutaraldehyde in PBS at 4°C for 25 minutes and 60 minutes for E10.5 and E12.5 specimens, respectively. The placentas were then equilibrated in 30%sucrose in 0.1 M phosphate buffer pH 7.3 overnight at 4°C before embedding in OCT medium. Serial cryosections of 10 μm were postfixed for 5 minutes and X-Gal staining was performed as described(Bissonauth et al., 2006).

Protein extracts from E10.5 embryos and from their corresponding placenta were prepared (Bélanger et al.,2003; Bissonauth et al.,2006). Total protein lysates (20 μg) were resolved on denaturing 10% SDS-PAGE and detected with anti-phospho-specific MAPK1/MAPK3(Cell Signaling Technology Inc., Pickering, ON) and homemade rabbit polyclonal anti-MAP2K1, anti-MAP2K2 and anti-MAPK1 antibodies. At least four specimens per genotype were analyzed. The most representative western blots are presented.

For proliferation and apoptosis studies, the ratio of positive cells to total cell number was determined for a minimum of five random areas. Repeated measures for the linear mixed model were performed to assess the difference between genotypes at all stages studied when genotype is considered the fixed effect and area the random effect. The procedure PROC MIXED from the SAS System was used (Littell et al.,1998). For the various breedings, deviation to the Mendelian distribution was assessed using χ² test.

To establish if Map2k2 plays a specific function during mouse development, we investigated the consequences of the lack of Map2k2function in the context of a reduced dosage of Map2k1. First, Map2k1^+/- and Map2k2^-/- mice were intercrossed to generate double heterozygous mutants(Map2k1^+/-Map2k2^+/-; DH). Unexpectedly, among the 361 viable offspring produced, only 32 Map2k1^+/-Map2k2^+/- mice were obtained at weaning, corresponding to 9% of the progeny instead of the expected 50%Mendelian ratio (P<10^-15)(Table 1). Thus, the combined loss of one allele of each Map2k gene was more deleterious than the Map2k2 homozygous mutation(Bélanger et al.,2003).

To determine the time of death of the Map2k1^+/-Map2k2^+/- embryos, Mak2k1^+/- and Map2k2^-/- mice were bred and embryos were recovered from E10.5 to E18.5. Mendelian ratios of Map2k1^+/+Map2k2^+/- and Map2k1^+/-Map2k2^+/- live embryos were obtained at E10.5 (P>0.5) but not at E12.5 (P<0.08). From E10.5 onward, a substantial proportion of Map2k1^+/-Map2k2^+/- embryos was moribund or in the process of being resorbed(Table 1). At E14.5 and E18.5, Map2k1^+/-Map2k2^+/- embryos constituted only 19% and 11% of the live embryos, respectively(P<10^-10). Thus, the lethality of the Map2k1^+/-Map2k2^+/- embryos was detected as early as E10.5, and this phenotype increased with gestational age.

The surviving Map2k1^+/-Map2k2^+/-mice had a normal lifespan and were fertile, reminiscent of our previous observations regarding the Map2k1^-/- mice when the mutation was not targeted to the extra-embryonic tissues(Bissonauth et al., 2006). To investigate if Map2k1 and Map2k2 haploinsufficiency can perturb normal placenta development without directly interfering with embryo formation, we analyzed the placenta of E10.5 to E14.5 conceptuses. As Map2k1^+/- and Map2k2^+/- mutants were indistinguishable from their wild-type littermates, Map2k2^+/- specimens were used as controls and breedings between Map2k1^+/- and Map2k2^-/- mice were performed to generate DH mutants (Fig. 4; see Fig. S2 in the supplementary material)(Bélanger et al., 2003; Bissonauth et al., 2006). Histological examination of Map2k1^+/-Map2k2^+/- specimens at different stages revealed the presence of two categories of mutants(Fig. 1). From E10.5 to E14.5,25 out of 32 Map2k1^+/-Map2k2^+/-embryos, hereafter referred to as type I DH mutants, had a small placenta and corresponded to the most affected conceptuses (moribund or resorbed embryos)(Table 4). Histological analyses of type I DH mutant placentas showed a severe reduction of the labyrinth size (Fig. 1). The labyrinth region was compact and disorganized, with numerous empty spaces,large maternal sinuses and aberrant MTG cells. By contrast, the histology of type II DH specimens was similar to that of Map2k2 heterozygous placentas. Thus, these results implicate Map2k2 in placenta formation.

The underdevelopment of the labyrinth in Map2k1^-/-placentas was attributed to the concerted reduction of proliferation and survival of labyrinth trophoblasts(Bissonauth et al., 2006). To verify if these processes were also affected in Map2k1^+/-Map2k2^+/- placentas, we performed proliferation assays on E10.5 to E14.5 placentas. In control placentas, the proliferation rate in the labyrinth region was elevated as previously observed (Fig. 2A)(Bissonauth et al., 2006). By contrast, the proliferation rate was significantly reduced in Map2k1^+/-Map2k2^+/- specimens at all ages analyzed (P<0.005). TUNEL assays were also performed to define if increased apoptosis contributed to the reduced Map2k1^+/-Map2k2^+/- labyrinth region. Very few apoptotic cells were detected in E10.5 to E14.5 control and Map2k1^+/-Map2k2^+/- placentas, but at E12.5 the augmented number of apoptotic cells in Map2k1^+/-Map2k2^+/- specimens was statistically significant (P=0.013)(Fig. 2B). Moreover, in all specimens examined apoptosis was more important in type I DH placentas when compared with controls or type II DH mutant placentas, this phenomena being more important at E14.5, suggesting that type I DH embryos might be in the process of being resorbed. Therefore, perturbed cell proliferation may underlie the reduced expansion of the labyrinth region of the placenta from Map2k1^+/-Map2K2^+/- mutants, whereas augmented apoptosis may be secondary to the resorption process.

The embryonic lethal phenotype of Map2k1^-/- mutants is due to the underdevelopment of the labyrinth associated with its hypovascularization (Bissonauth et al.,2006). To determine if the reduced survival of the Map2k1^+/-Map2k2^+/- embryos involved a placenta vascularization defect, we examined the vascular network by using the endothelial cell marker CD31 on sections of E10.5 to E14.5 control and Map2k1^+/-Map2k2^+/- placentas. In control and type II DH specimens, the fetal blood vessels invaded the whole labyrinth region and the density of blood vessels increased with age(Fig. 3A, not shown). By contrast, the number of embryonic blood vessels was reduced in type I DH placentas, a result clearly seen at E14.5(Fig. 3A). To visualize the maternal vascular network, we also performed alkaline phosphatase assays to label the differentiated SynT lining the maternal sinuses. A strong staining was detected around the maternal sinuses in both control and type I DH specimens (Fig. 3B). The number of sinuses increased in control placentas with embryonic age, whereas in type I DH specimens sinus formation was significantly reduced at all stages analyzed. These data indicate that the combined Map2k1Map2k2haploinsufficiency affects trophoblast proliferation and labyrinth vascularization, as both fetal and maternal vascular networks are perturbed in type I DH mutants. Thus, not only is Map2k1 required for normal placentation, but Map2k2 also appears involved in labyrinth growth and morphogenesis.

To directly address the importance of the placenta defects on survival, we have rescued the placenta phenotype using the Map2k1 conditional allele (Map2k1^flox allele)(Bissonauth et al., 2006). Map2k1^flox/ΔMap2k2^+/+ females were bred with Map2k1^+/+Map2k2^-/-Tg^+/Sox2Cremales to specifically ablate Map2k1 function in embryonic derivatives, retaining the undeleted Map2k1^flox allele in extra-embryonic tissues (Hayashi et al.,2002). This specific breeding should provide all the genotypes required for control and experimental specimens. The Map2k1^+/floxMap2k2^+/-Tg^+/+progeny was used to define the normal survival rate whereas the Map2k1^+/ΔMap2k2^+/-Tg^+/+and Map2k1^+/ΔMap2k2^+/-Tg^+/Sox2Crespecimens monitored DH defective survival. Finally, the Map2k1^+/floxMap2k2^+/-Tg^+/Sox2Creconceptuses addressed the contribution of extra-embryonic development to DH survival. In the latter case, genotype of the embryo differed from that of the extra-embryonic tissues: the embryo was Map2k1^+/ΔMap2k2^+/- whereas the extra-embryonic structures were Map2k1^+/floxMap2k2^+/-. As expected,the survival rate at weaning of Map2k1^+/ΔMap2k2^+/-Tg^+/+mutants was reduced compared with that of Map2k1^+/floxMap2k2^+/-Tg^+/+mice (8% instead of 25%, P<0.0001)(Table 2). However, Map2k1^+/ΔMap2k2^+/-Tg^+/Sox2Creand Map2k1^+/floxMap2k2^+/-Tg^+/Sox2Cremice could not be distinguished by genotyping, but together they were represented at the expected Mendelian ratio(Table 2). To distinguish the Map2k1^+/ΔMap2k2^+/-Tg^+/Sox2Creand Map2k1^+/floxMap2k2^+/-Tg^+/Sox2Crespecimens, litters were recovered at E18.5 to allow genotyping of extra-embryonic tissues (Table 2). The genotypes of the embryos and their yolk sacs were independently determined. The number of embryos obtained for each genotype diverged significantly from the expected ratio (P<0.018). The number of Map2k1^+/ΔMap2k2^+/-Tg^+/+and Map2k1^+/ΔMap2k2^+/-Tg^+/Sox2Creembryos was reduced when compared with that of Map2k1^+/floxMap2k2^+/-Tg^+/+controls. Most importantly, restriction of haploinsufficiency to the embryo in Map2k1^+/floxMap2k2^+/-Tg^+/Sox2Crespecimens had no impact on survival rate. Thus, the rescue of the placenta phenotype allows the normal development of DH embryos and establishes that Map2k1 and Map2k2 gene dosage is crucial for extra-embryonic structure formation.

The lack of phenotype in Map2k2 homozygous mutant mice contrasts with the lethality of the Map2k1^+/-Map2k2^+/- embryos. The Map2k2 mutation participates in the lethal phenotype only in the absence of one Map2k1 allele, suggesting a gene dosage effect in which Map2k1 and Map2k2 genes do not contribute equally to the development of the extra-embryonic structures. To define the share of each gene in placenta development, Map2k1^+/-Map2k2^+/- mutant mice were intercrossed and their progenies analyzed. As the number of Map2k1^+/-Map2k2^+/- mice obtained by Map2k2^-/- × Map2k1^+/- breeding was limiting, we also performed Map2k1^flox/flox × Map2k2^-/-Tg^+/Sox2Crebreedings to generate Map2k1^+/ΔMap2k2^+/- mice. The Map2k1 deletion restricted to the embryo allowed the complete survival of Map2k1^+/ΔMap2k2^+/-mutants. DH intercrosses, using either Map2k1^+/Map2k2^+/- or Map2k1^+/ΔMap2k2^+/- mice,generated similar results (Table 3A,B). Analysis of the genotype distribution at weaning revealed that the deletion of both Map2k1 alleles led to embryonic death whereas the lethality associated with the homozygous Map2k2 mutation required the absence of at least one Map2k1 allele. To establish at which embryonic age Map2k1Map2k2 compound mutants die, the analysis was also performed at E10.5 (Table 3). Most of the allelic combinations were obtained at the expected Mendelian ratio except for the Map2k1^-/-Map2k2^+/- and Map2k1^-/-Map2k2^-/- genotypes,indicating an earlier embryonic death. Histology of E10.5 placentas carrying the allelic combinations recovered showed a gradation in the severity of the phenotype (Fig. 4). Normal development of the labyrinth region was observed in Map2k1^+/+Map2k2^+/-, Map2k1^+/+Map2k2^-/- and Map2k1^+/ΔMap2k2^+/+ specimens when compared to wild-type controls (Fig. 4A-D). Some Map2k1^+/ΔMap2k2^+/- mutants displayed a normal phenotype (type II DH mutants)(Fig. 4E), whereas others presented placenta defects (type I DH mutants) as previously shown(Fig. 1). All Map2k1^Δ/+Map2k2^-/- specimens presented a more severe phenotype with an increased reduction of the labyrinth region and the presence of MTG cells (Fig. 4F). The most severe placenta phenotype was detected in Map2k1^Δ/ΔMap2k2^+/+ and Map2k1^Δ/ΔMap2k2^+/-specimens, with a very thin labyrinth region. In addition, Map2k1^Δ/ΔMap2k2^+/- embryos were moribund and in the process of being resorbed. Thus, the severity of the placenta phenotype increases with the number of Map2k mutant alleles,demonstrating that the correct development of extra-embryonic structures requires a proper Map2k gene dosage.

To determine whether the placenta defect was responsible for the reduced survival rate of the Map2k1^+/ΔMap2k2^-/- conceptuses,we performed rescue of the placenta phenotype to determine whether Map2k1^+/ΔMap2k2^-/- mice can survive to adulthood. When Map2k1^flox/floxMap2k2^-/- females were bred with Map2k1^+/+Map2k2^-/-Tg^+/Sox2Cremales, the expected Mendelian ratio of Map2k1^+/ΔMap2k2^-/- animals was recovered (12 over 22 pups born) indicating that only one Map2k1allele is sufficient for normal embryo development in the absence of Map2k2 gene function when the placenta can develop normally.

To determine if the gene dosage-dependent defects of placenta development were due to perturbed ERK/MAPK signaling, we performed western blot analyses with embryo and placenta whole protein extracts from E10.5 wild-type and Map2k1Map2k2 compound mutants using a phospho-specific MAPK1/MAPK3 antibody (Fig. 5A). To do so, Map2k1^+/ΔMap2k2^+/- mice were intercrossed and the conceptuses were used for protein extraction. The phosphorylation of MAPKs at specific residues of the activation loop is usually a good read-out of their state of activation(Gopalbhai et al., 2003). MAP2K1 and MAP2K2 protein levels were also evaluated. Signal for MAP2K1 and MAP2K2 was sometimes detected in placenta samples from homozygous null mutants, indicating cross-contamination with maternal tissues. The reduction in phosphorylation of MAPK1 and MAPK3 was the most severe in Map2k1^Δ/ΔMap2k2^+/+ and Map2k1^+/ΔMap2k2^-/- placenta and embryonic samples. A likely explanation for this decrease could be that Map2k1^Δ/ΔMap2k2^+/+ and Map2k1^+/ΔMap2k2^-/- embryos were dying and in the process of resorption. In this experiment, no Map2k1^-/-Map2k2^+/- specimen was obtained. Together, these results support the correlation between the gene dosage-dependent placenta defect, the reduction of the ERK/MAPK signaling and the predominant role of Map2k1.

We have previously shown that the ERK/MAPK cascade is highly activated in the SynT of the labyrinth and that this activation is lost in Map2k1^-/- specimens(Bissonauth et al., 2006). To correlate ERK/MAPK cascade activation with the specific expression of either MAP2K1 or MAP2K2 proteins, the expression profile of the MAP2K1 and MAP2K2 proteins was established by immunohistochemistry(Fig. 5B). The specificity of the antibodies was demonstrated on Map2k1^-/- and Map2k2^-/- placenta sections, and as expected no MAP2K1 and MAP2K2 proteins were detected on sections of Map2k1^-/- and Map2k2^-/- specimens, respectively. The expression of the MAP2K2 protein was ubiquitous in the placenta. By contrast, the MAP2K1 protein was strongly detected in cells lining the maternal sinuses(Fig. 5B, arrow), but not in mononuclear giant cells (Fig. 5B, arrowheads).

The formation of MTG cells in specimens lacking one or both Map2k2alleles in Map2k1^+/- conceptuses was an unusual Map2k2-dependent phenotype (Fig. 1; Fig. 4F). These MTG cells were observed as early as E10.5 in more than 50% of the Map2k1^+/-Map2k2^+/- specimens analyzed(three out of five). Even if they were not strictly associated with type I DH mutants, their number was more abundant in this type. MTG were observed throughout the labyrinth and restricted to this layer. The number and the size of MTG was greater in type I DH mutants, and they increased at E12.5 and decreased at E14.5, most likely because the most affected specimens died between these two ages (Table 4). The presence of the MTG cells suggested an aberrant SynT differentiation, which can contribute to the reduced vascularization of the placenta due to impaired branching morphogenesis(Fig. 3)(Bissonauth et al., 2006). A similar structure was previously reported in Lbp-1a mutants and referred to `amorphous material' (Parekh et al., 2004). It was suggested to represent fibrin deposits derived from extravasation of blood from a defective vascular bed. To reveal any fibrin deposit in our specimens, we performed MSB staining. Even though MSB staining was observed in wild-type and Map2k1^+/-Map2k2^+/- placentas, the MTG cells were negative (Fig. 6A). Hematoxylin and Eosin staining clearly showed the multinucleated nature of these structures, with the nuclei located at the periphery(Fig. 6B). This result was corroborated by staining the nuclei with DAPI and the cortical actin with phalloidin (Fig. 6C). The actin signal was restricted to the periphery of the MTG cells with the juxtaposed nuclei. No phalloidin staining was observed between the nuclei in the multinucleated cells, suggesting the absence of cytoplasmic membrane. Altogether, these results indicate that these structures are multinucleated cells devoid of fibrin deposits.

As the multinucleated cell type of the placenta is the SynT, we wanted to establish whether MTG cells were SynT-derived using specific markers such as alkaline phosphatase activity for the SynT and Gcm1 expression for the SynT layer II (SynT-II) (Simmons et al., 2008). Gcm1-positive cells were detected in Map2k1^+/-Map2k2^+/- labyrinth(Fig. 6D), but the MTG cells were negative for Gcm1 expression. Similarly, alkaline phosphatase activity was detected in cells lining the maternal sinuses but not in MTG cells (Fig. 6E). MTG cells were negative for other cell lineage markers such as the glycogen trophoblasts detected by PAS staining (Fig. 6F) and the spongiotrophoblasts (detected by in situ hybridization using Tpbpa; not shown). The MTG cells were also negative for CD31,indicating that they were not vascular endothelial cells(Fig. 6G). Finally, the MTG cells were negative for BrdU incorporation, indicating that they were post-mitotic cells like the SynT (Fig. 6H) (Cross et al.,2006). Few BrdU-positive nuclei were found adjacent to the nuclei of the MTG cells but it was difficult to assess if they were part of the MTG cells or not (Fig. 6H). Thus,the formation of the MTG did not seem to be due to defective cytokinesis.

To further investigate the origin of the MTG cells, we used a genetic-based approach to follow the Gcm1-expressing SynT-II cell lineage. A Gcm1Cre transgenic mouse line expressing the Cre recombinase in the SynT-II cells was produced and characterized by breeding to the R26RCre reporter mouse line for monitoring Cre expression(Gt(ROSA)26Sor^tm1Sor allele)(Soriano, 1999). At E8.5, Gt(ROSA)26Sor^+/tm1SorTg^+/Gcm1Creplacentas displayed X-Gal staining restricted to the chorionic cells, whereas at E10.5 staining was restricted to the cells lining the maternal sinuses(Fig. 7A,B). This reproduced the placenta lacZ expression profile previously described in Gcm1-lacZ knock-in mice as well as the expression pattern of the endogenous Gcm1 gene (Basyuk et al., 1999; Schreiber et al.,2000; Stecca et al.,2002). We thus used the Gcm1Cre mice to determine if the MTG cells were derived from the SynT-II cells. Map2k1^+/ΔTg^+/Gcm1Cre mice were mated to Gt(ROSA)26Sor^{tm1Sor/tm1Sor}Map2k2^-/-mice and E12.5 progenies were analyzed for the presence of MTG cells. The latter were detected in Map2k1^+/ΔMap2k2^+/-Gt(ROSA)26Sor^+/tm1SorTg^+/Gcm1Creand Map2k1^+/ΔMap2k2^+/-Gt(ROSA)26Sor^+/tm1SorTg^+/+specimens. However, the structures were positive for X-Gal staining only when positive for the Gcm1Cre transgene, indicating that the MTG cells derived from SynT-II (Fig. 7C).

In Lbp-1a null mutants, the presence of multinucleated cells in the placenta cannot be rescued in tetraploid complementation assays,indicating that the defect is non-cell-autonomous(Parekh et al., 2004). The placenta phenotype in Lbp-1a^-/- embryos is secondary to a defective allantoic mesoderm. LBP-1 protein isoforms are phosphorylated by the ERK/MAPK cascade and their phosphorylation might affect their cellular localization as well as their DNA-binding activity(Pagon et al., 2003; Sato et al., 2005; Volker et al., 1997). We tested if the accumulation of MTG cells in Map2k1^+/-Map2k2^+/- was a cell-autonomous defect or not. Map2k1^flox/floxMap2k2^-/- females were first bred with Tg^+/Gcm1Cre males to restrict the Map2k1 deletion to the SynT-II cells(Fig. 7D-F). The specific deletion of one Map2k1 allele in SynT-II from the Map2k1^+/floxMap2k2^+/-Tg^+/Gcm1Creconceptuses led to formation of MTG cells, as detected in placentas from three E12.5 specimens analyzed (Fig. 7F). By contrast, the placenta from Map2k1^+/floxMap2k2^+/-Tg^+/Sox2Crespecimens did not include any MTG structure (seven specimens analyzed)(Fig. 7G). The complete deletion of the Map2k1^flox allele in embryonic tissues by Sox2Cre was confirmed by genotyping analyses (Fig. S3 in the supplementary material). Thus, the formation of MTG cells required the deletion of one allele of each Map2k1 and Map2k2 genes in the SynT-II, indicating that the phenotype is cell-autonomous. However, as the Map2k2 mutation is not restricted to the SynT-II, we cannot rule out a potential contribution of MAP2K2 in endothelial cells to the MTG phenotype.

The MTG are derived from SynT-II cells and the ERK/MAPK cascade is highly activated in SynT-II, leading us to investigate the ERK/MAPK activation in MTG. Both MAP2K1 and MAP2K2 were expressed in SynT-II-derived MTG, as assessed by immunohistochemistry (Fig. 8B,D). They were also activated in MTG, as seen with an anti-phospho-MAP2K1/2 (Fig. 8F). However, no phospho-MAPK1/3 staining was detected in MTG,revealing the absence of ERK/MAPK activation(Fig. 8H). These data indicate that the sustained activation of the ERK/MAPK cascade is required for the formation of the uniform SynT layer II lining the maternal sinuses.

The disruption of the murine Map2k1 gene leads to an embryonic lethal phenotype caused by the underdevelopment of the placenta labyrinth associated with its hypovascularization, revealing the essential role of Map2k1 in extra-embryonic development(Bissonauth et al., 2006; Giroux et al., 1999). By contrast, the absence of phenotype in Map2k2 mutants suggests that Map2k2 is dispensable as Map2k1 can compensate for the lack of Map2k2 gene function(Bélanger et al., 2003). We have previously shown that both Map2k1 and Map2k2 genes are widely expressed in the embryo and the placenta(Bissonauth et al., 2006; Giroux et al., 1999). In this study, we asked whether haploinsufficiency of the Map2k1 locus could reveal Map2k2 gene function during mouse development. Our data clearly show that the lack of one or both Map2k2 alleles in Map2k1^+/- mice causes embryonic death consistent with a role of Map2k2 in embryonic development. Map2k1^+/-Map2k2^+/- mice have a reduced survival rate whereas the Map2k1^+/-Map2k2^-/- embryos die at early gestation if Map2k1 is deleted in both embryonic and extra-embryonic tissues(Tables 1 and 3). However, when the deletion of Map2k1 is restricted to embryonic tissues, Map2k1^+/ΔMap2k2^+/- and Map2k1^+/ΔMap2k2^-/- mice are viable and represented at weaning according to the Mendelian distribution. These results indicate that both Map2k1 and Map2k2 genes contribute to the development of the extra-embryonic ectoderm whereas only one allele of Map2k1 is sufficient for embryo development and signaling in adult mice. It would be interesting to determine if a similar result is obtained when only one Map2k2 functional allele is retained in adult.

Gene duplication is a common event and has led to the production of numerous gene families in multiple species, including mammals. During evolution, two highly conserved Map2k genes, Map2k1 and Map2k2, have arisen from a single prototypic gene as a result of genome duplication. Following gene duplication, decreased selection pressure can sometimes allow production of proteins with novel function or expression profile resulting in partial redundancy among gene family members. The biochemical properties of MAP2K1 and MAP2K2 proteins suggest that these two enzymes have evolved to be differentially utilized(Skarpen et al., 2007; Ussar and Voss, 2004). If the shared function of duplicated genes becomes unequally distributed between gene copies, unequal redundancy among the gene family members may result, as observed for the Erk/Map2k gene family, where the Map2k2gene deletion does not cause any phenotype but enhances the Map2k1placenta defects in a unique manner. Our study uncovers unequal contribution and gene dosage effect of the Erk/Map2k genes that exhibit overlapping expression profile (Fig. 5) (Bissonauth et al.,2006; Giroux et al.,1999). Map2k2^-/- mutants show no obvious defect in placenta and embryo development, indicating that Map2k1expression is sufficient and above the MAP2K threshold levels essential for correct placenta development. As successive Map2k alleles are mutated, MAP2K1 and MAP2K2 levels became inadequate to maintain signaling and placenta defects occur.

The reduction of trophoblast proliferation and the hypovascularization of the labyrinth region, which both characterize the Map2k1^-/- placenta phenotype, were also observed in Map2k1^+/-Map2k2^+/- specimens. The contribution of Map2k1 to the phenotype appears predominant over that of Map2k2. Consistent with this unequal gene dosage-dependent model, the severity of the placenta phenotype increases with the number of Map2knull alleles: Map2k1^+/-Map2k2^+/-<Map2k1^+/-Map2k2^-/-<Map2k1^-/-Map2k2^+/+<Map2k1^-/-Map2k2^+/-, the Map2k1^-/-Map2k2^+/- placentas presenting the most severe phenotype. Moreover, this is paralleled by a gradation in ERK/MAPK activation (Fig. 5A).

In many cases, unequal redundancy between paralogs may be produced by differences in expression levels, distinct expression profiles or acquisition of new protein functions. Previous studies have shown that both MAP2K1 and MAP2K2 proteins are expressed at similar levels in several cell lines. However, in some cell lines MAP2K1 is present in slight excess(Xu et al., 1997). Furthermore, MAP2K2 has been shown to be the most active ERK activator(Zheng and Guan, 1993). Recombinant MAP2K1 has an activity approximately seven times lower than that of MAP2K2. However, this difference in activity does not take into account the presence in cells of various mechanisms that can increase the specificity and the enzymatic activity of MAP2K proteins. For instance, the MAP2K1 interacting protein 1 (MAP2K1ip1), a scaffolding protein, has been shown to enhance the enzymatic activity of MAP2K1 towards MAPK1 and MAPK3, to interfere with feedback inhibition of MAP2K1 by MAPK1, and to inhibit the ability of activated MAPK1 to phosphorylate the transcription factors ETS-1 and ETS-2(Brahma and Dalby, 2007; Foulds et al., 2004; Schaeffer et al., 1998). MAP2K2 does not form a stable complex with MAP2K1ip1. In the placenta, both MAP2K1 and MAP2K2 proteins are co-expressed, but MAP2K1 is expressed at higher levels in the SynT (Fig. 5B). This raises the possibility that MAP2K1 enzymatic activity is more important for this cell lineage than MAP2K2 one. In absence of both Map2k1functional alleles, ERK activation levels may be under threshold whereas the deletion of one allele of each Map2k gene may not have such an impact since some DH mutants survive at birth. Thus, the different phenotypes observed might be due to variations in Map2k expression levels or domains. However, we cannot exclude that each MAP2K protein may possess unique properties.

The loss of Map2k2 function when combined with Map2k1haploinsufficiency results in the accumulation of MTG cells in the placenta. A cell-lineage approach using the Gcm1Cre deleter mouse line has revealed the SynT-II origin of the MTG cells as well as the cell-autonomous nature of the defect. Even though the MAP2K1 protein is detected at high levels in SynT, no MTG cells were observed in Map2k1^+/-and Map2k1^-/- placentas(Fig. 4; see Fig. S2 in the supplementary material) (Bissonauth et al.,2006). Moreover, SynT-II are determined at the right time based on Gcm1 expression in Map2k1^-/- specimens, despite their restricted localization at the chorioallantoic interface that underlies the defective placenta vascularization(Bissonauth et al., 2006). This indicates the importance of Map2k2 mutation in the formation of MTG cells.

The presence of MTG cells has been reported for the Sos1 and Lbp-1a mouse mutants (Parekh et al., 2004; Qian et al.,2000). Sos1 is upstream of MAP2K1 and MAP2K2 in the ERK/MAPK cascade. Moreover, in E10.5 embryo, most ERK activity was shown to be Sos1-dependent (Qian et al.,2000). Therefore, it is not surprising that Sos1 and Map2k1Map2k2 compound mutants present a common placenta phenotype. By contrast, the link between Lbp-1a and the ERK/MAPK cascade is more indirect as the Lbp-1a placenta phenotype has a non-cell-autonomous origin. Normal development of the placenta includes non-cell-autonomous mechanisms involving placenta as well as embryonic paracrine factors necessary for maintaining the basic developmental program(Cross, 2005). Therefore, it is possible that LBP-1A and the ERK/MAPK cascade may complement each other,the first one by participating to the production of such paracrine signals,and the second by allowing its transduction into the cell. However, we cannot rule out the possibility that Map2k2 haploinsufficiency impacts on LBP-1A function in the endothelial cell whereas Map2k1 and Map2k2 haploinsufficiency is required in the SynT-II cells for the expression of the MTG phenotype.

The present study indicates that both Map2k1 and Map2k2genes are involved in extra-embryonic ectoderm formation and morphogenesis of the labyrinth, leading to its vascularization by the syncytiotrophoblasts and the embryonic blood vessels. The placenta phenotype is Map2k1/Map2k2gene dosage-dependent with a predominant role of Map2k1. Our results also reveal a unique role for Map2k2 in SynT II cell determination and differentiation.