Dissecting Wnt/β-catenin signaling during gastrulation using RNA interference in mouse embryos

Signaling molecules of the Wnt family regulate many cellular behaviors,including differentiation, proliferation and morphogenesis, and are involved in gastrulation and axial development (reviewed by Sokol, 1999; Yamaguchi, 2001).β-Catenin, besides acting as a central component of the cadherin cell-adhesion complex, plays an essential role in the canonical Wnt/β-catenin signaling pathway (reviewed by Huber et al., 1996). Upon Wnt stimulation, β-catenin enters the nucleus and acts in a complex with members of the Tcf/Lef (T-cell factor/Lymphoid enhancer factor) family of transcription factors to activate target genes(Hecht and Kemler, 2000).

Mutations in a number of Wnt genes and Wnt signaling components highlight the crucial role of the Wnt/β-catenin signaling pathway in the initiation of primitive streak formation, as well as in the patterning and morphogenesis of the gastrulation-stage embryo(Beddington and Robertson,1999; Lu et al.,2001). Mutant mouse embryos that lack functional Wnt3 orβ-catenin fail to establish an anterior-posterior (A-P) axis and do not form a primitive streak, thus they fail to generate endoderm and mesoderm,resulting in an arrest of development before gastrulation(Liu et al., 1999; Haegel et al., 1995; Huelsken et al., 2000). Both Wnt3a null mutants and Lef1;Tcf1 compound homozygous null mutants fail to differentiate paraxial mesoderm, do not form somites caudal to the forelimb buds and exhibit severe posterior truncations(Takada et al., 1994; Galceran et al., 1999). In addition, Wnt3a controls directly the expression of Axin2and Dll1 in the paraxial mesoderm, and thereby, links the Notch and Wnt signaling pathways in the processes of somitogenesis(Aulehla et al., 2003; Galceran et al., 2004; Hofmann et al., 2004). By the end of gastrulation and the beginning of neurulation, secreted Fgf8 and Wnt1 molecules from the isthmic organizer play an important role in patterning the mid/hindbrain region along the A-P axis (reviewed by Liu and Joyner, 2001; Wurst and Bally-Cuif,2001).

Recently, using the Cre/loxP system, we have conditionally inactivated β-catenin in the visceral endoderm (VE) and the anterior primitive streak (APS) by using a Cytokeration 19 (K19)-driven Cre(Lickert et al., 2002). Similar to in Wnt3 and β-catenin null mutants, A-P axis formation was affected; however, the conditional β-catenin mutants proceeded through gastrulation. This revealed a crucial function for β-catenin during later developmental processes, such as posterior axis elongation and somite formation, processes affected in other Wnt mutants. Additionally, the node, an embryonic structure functionally equivalent to the Spemann/Mangold organizer in frog, failed to form in these mutants. Taken together, these results are consistent with the hypothesis that Wnt/β-catenin signaling is important for the induction of the mouse embryonic organizing centers, the formation of somites, and the proper morphogenesis of the gastrulating embryo.

Here, we have used a functional genomic approach combining Affymetrix GeneChip analysis, whole-mount in situ screening and rapid functional assessment by RNAi in embryonic stem (ES) cell-derived embryos to dissect the Wnt/β-catenin signaling pathway during gastrulation. Intriguingly, the knock-down phenotypes of two potential target genes, Grsf1 and Fragilis2 (Ifitm1 – Mouse Genome Informatics),recapitulate specific but distinct aspects of Wnt pathway mutants, suggesting that these genes are components of the downstream Wnt response. In summary,this approach represents a highly efficient and rapid methodology with which to unravel developmental pathways in the mouse.

Cytokeratin 19 (K19) promoter-driven Cre mice(K19-Cre) were previously generated by a knock-in of the Cre recombinase gene into the ATG translation initiation codon of exon1 of K19 (Harada et al.,1999). The β-catenin floxed (flox) allele and theβ-catenin floxed deleted (floxdel) allele were previously described (Brault et al.,2001). K19-Cre mice were mated withβ-catenin floxdel mice and the offspring, which inherited both alleles, were crossed with homozygous β-catenin flox mice; a quarter of the offspring was positive for K19-Cre, together with one flox and one floxdel allele. Littermates, which inherited the floxed and floxdel β-catenin alleles but did not carry the K19-Cre allele served as heterozygous controls. Mutant animals were bred on a mixed 129Sv×C57Bl/6 background. PCR genotyping was performed as described previously(Lickert et al., 2002).

Microarray experiments have been submitted to GEO in a MIAME compliant format (Minimum Information About a Microarray Experiment)(Brazma et al., 2001). The accession number is GSE2519.

Embryos from the above described intercrosses were dissected in PBS and separated into embryonic and extraembryonic portions. The embryonic portions were stored in PBS at –80°C until the PCR genotyping was carried out using the extraembryonic portions. The pooled embryos were homogenized in 250μl of TRIzol Reagent (Invitrogen), and total RNA was extracted according to the manufacturer's protocol.

For each sample, 5 μg total RNA was used for cDNA synthesis according to the Expression Analysis Technical Manual (Affymetrix). The in vitro transcription and labeling of cRNA was carried out using the BioArray High-Yield Transcription Labeling Kit (Enzo). Then 25 μg of labeled cRNA was used to hybridize all three GeneChips from the Affymetrix U74v2 according to the standardized Affymetrix protocol.

MAS 5.0 software was used to generate the expression data set for each GeneChip.dat file and scale normalized to a target value of 150. Comparisons were made to calculate signal log ratios of expression between mut:wt and mut:het using either wild type (wt) or heterozygotes (het) as a baseline,respectively. The resulting table was exported to Microsoft Excel to filter out probe sets with absent calls across all samples (P-detection>0.04) and probe sets with no change in expression (0.003 > P-signal log ratio <0.997) in both mut:het and mut:wt comparisons. Illogical combinations of absent and present calls with a significant change call were deleted, e.g. an absent call in baseline and present call in mutant with a decrease change. Genes that were consistently up- and downregulated in both comparisons (absolute signal log ratio of ≥0.5) were hierarchically clustered using Cluster 3.0 software[http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/software.htm,based on the method of Eisen et al. (Eisen et al., 1998)].

Using the Affymetrix Gene Ontology (GO) analysis software(www.affymetrix.com)the numbers of probe sets on U74v2A GeneChip corresponding to the GO terms`Transcription' (GOID:6350), `Cell Communication' (GOID:7154), `Pattern Specification' (GOID:7389) and `Morphogenesis' (GOID:9653) were calculated as of the November 2003 annotation build. The total numbers of all other probe sets with a GO term were denoted as `Others'. Additionally, the numbers for the same GO terms were calculated for the 49 downregulated genes from the U74Av2 GeneChip. Percentiles for each GO term category were calculated by dividing the numbers in each category by the total number of probe sets with any GO term.

For whole-mount in situ hybridization and histology, embryos were processed as described previously (Lickert et al.,2002). Sense and antisense in situ probes were in vitro transcribed using ESTs available from ATCC (Manassas, VA, USA) with the following I.M.A.G.E. Clone IDs: Fragilis2, 657273; Zic3,5120056; Scap2, 3599914; Punc, 3514346; EST10,3980327; EST16, 1328681; Sox2, 5707193. Additionally, ESTs from the NIA 15K Mouse cDNA Clone Set with the following H3 clone IDs were used for probe synthesis: Grsf1, H3046G05; EST6, H3074B02. Additional probes for known genes were as follows: Axin2(Aulehla et al., 2003), Wnt3a (Gavin et al.,1990), Wnt8 (Bouillet et al., 1996), Frzb1(Hoang et al., 1998), Notch1 (Conlon et al.,1995), Dll1 (Hrabe de Angelis et al., 1997), Gbx2(Wassarman et al., 1997), Hoxb1 (Marshall et al.,1992), T (Herrmann et al., 1990), Tbx6(Chapman et al., 1996), PAPC (Rhee et al.,2003), and Krox20(Swiatek and Gridley,1993).

To identify germ cells, embryos were fixed in 4% paraformaldehyde in PBS for 30 minutes and stained for tissue non-specific alkaline phosphatase for 5 minutes [25 mM Tris-maleic acid (pH 9.0), 0.4 mg/ml α-naphthyl phosphate(Sigma), 1 mg/ml Fast Red TR salt (Sigma), 8 mM MgCl₂, 0.01%Na-desoxycholate, 0.02% NP-40].

For construction of the Grsf1 and Fragilis2 shRNA transgenes, we have used the pcDNA3.1 RasGAP shRNA plasmid described recently (Kunath et al.,2003). RasGAP shRNA was released from the plasmid by Asp718, XbaI digestion, and annealed oligonucleotides corresponding to the target sequence were introduced into the same sites using the following sense- and antisense-strand oligonucleotides (target sequence in bold):

Grsf1 shRNA forward, 5′-GT ACC AAA GCA CAG GGA AGA AAT TGG TA C AAG AGA TA CCA ATT TCT TCC CTG TGC TTT TTT TTGG AAA T-3′ and Grsf1 shRNA reverse, 5′-CTA GAT TTC CAA AAA AAA GCA CAG GGA AGA AAT TGG TA T CTC TTG TA CCA ATT TCT TCC CTG TGC TTT G-3′ (corresponding to bases 982-1004 of the murine Grsf1 gene, NCBI accession no.: NM_178700); and

Fragilis2 shRNA forward, 5′-GT ACC GAA CAT CAG CTC CCT GTT CTT CA C AAG AGA TG AAG AAC AGG GAG CTG ATG TTC TTT TTT TGG AAA T-3′ and Fragilis2 shRNA reverse, 5′-CTA GAT TTC CAA AAA AA GAA CAT CAG CTC CCT GTT CTT CA T CTC TTG TG AAG AAC AGG GAG CTG ATG TTC G-3′ (corresponding to bases 295-317 of the murine Fragilis2 gene, NCBI accession no.: BK001123).

The resulting shRNA targeting constructs were confirmed by DNA sequencing. Transgenic ES-cell lines were established as described(Kunath et al., 2003). The mRNA expression level of the individual ES-cell lines for Grsf1 and Fragilis2 was determined by northern blotting using the NorthernMax-Gly™ Kit (Ambion), according to the manufacturer's protocol. Pre-selected ES-cell lines were used to generate totally ES cell-derived embryos using the tetraploid aggregation technique as described previously(Nagy et al., 1993; Kunath et al., 2003). Embryos with any contribution of tetraploid EGFP-positive cells were excluded from the analysis. Experimental animals were treated according to guidelines approved by the Canadian Council for Animal Care.

The co-cultivation of ES cells with NIH3T3 fibroblasts was carried out essentially as described previously(Lickert et al., 2000), with the exception that the ES cells were seeded in transwell filters(Transwell-COL, collagen-treated, 0.4 μm pore-size; Costar #3491). After 18 hours of co-cultivation, total RNA was isolated from the ES cells using the RNeasy Mini Kit (Qiagen). For each sample, 2 μg of RNA was treated with DNaseI and then reverse transcribed using oligo (dT)-primers and SuperSriptII reverse transcriptase (Invitrogen).

Quantitative PCR was performed using the LightCycler Fast Start DNA Master^Plus SYBR Green I Kit (Roche) according to the manufacturer's protocol. The following primers were used to amplify mRNAs for Gapdh,Fragilis2 and Grsf1: Gapdh-fwd,5′-ACCACAGTCCATGCCATCACT-3′; Gapdh-rev,5′-GTCCACCACCCTGTTGCTGTA-3′; Fragilis2-fwd,5′-GGGCTCCTCGACCACACCTCTT-3′; Fragilis2-rev,5′-CCCAGTCGTATCACCCACCATCT-3′; Grsf1-fwd,GATATTCGGCCTATGACGGCT-3′; Grsf1-rev,CAAAATCGACAGCCTCTGGAAG.

β-Catenin homozygous null mutant embryos arrest before gastrulation(Haegel et al., 1995; Huelsken et al., 2000). To study the effects of disrupting the Wnt/β-catenin signaling pathway during gastrulation, we generated embryos that were compound heterozygous for a floxed β-catenin allele and a β-catenin floxdelallele (where exon 2-6 were removed by Cre-mediated excision), carrying the K19-Cre allele (hereafter termed conditional β-catenin knock out(CKO) (Lickert et al., 2002). We pooled and isolated total RNA from 47 CKO (7.32 μg), 47 heterozygous(6.12 μg), and 41 wild-type (6.44 μg) embryos between late gastrulation and head-fold stage – the time when the first phenotypic alterations can be seen. The high number of embryos collected should normalize any variation in staging. The three individual RNA samples were used for labeled cRNA preparation and GeneChip hybridizations (Affymetrix U74v2 Series containing about 36,000 probe sets). No amplification of the cRNA was used to ensure accurate representation of the individual expression profiles; however, this precludes the use of biological replicates because of limiting amounts of embryonic material. To identify genes directly or indirectly deregulated in the CKO mutant embryos, we compared wild-type (wt) and heterozygous (het)expression profiles against the mutant expression profile. For further study,we focused on the genes with consistent alterations in gene expression, i.e. those differing in both comparisons with an absolute signal log ratio ≥0.5. This limited the set of candidates to 262 upregulated and 77 downregulated genes on the three Affymetrix GeneChips(Fig. 1A, and Table S1 in the supplementary material). Here, we concentrated only on the downregulated genes in β-catenin mutants. These genes included the previously described direct target genes Cdx1, brachyury (T), follistatin, Axin2, and Dll1, thus validating the expression profiling approach (Table 1). Downregulation of genes such as Wnt3a, Tbx6, T, Notch1, Dll1,Meox1, Lef1, follistatin, Fgfr1 and Axin2 is in agreement with defects in paraxial mesoderm specification and somite formation in the mutant embryos. Moreover, components of the retinoic acid signaling cascade, such as Raldh2, Crabp1 and Rbp1, as well as components of the FGF signaling cascade, like Fgfr1, and components of the Wnt signaling cascade, including Wnt3a, Wnt5a, Wnt8 and Frzb1, were also downregulated. These results indicate that all the major caudalizing activities were affected, offering a possible explanation for the lack of posterior development in CKO embryos. Additionally,downregulation of the general definitive endoderm marker Sox17, is consistent with our previous observation that endoderm formation is affected in CKO mutants. Interestingly, we also found that the downregulation of genes,like Notch1, Dll1 and Foxj1(Krebs et al., 2003; Raya et al., 2003; Chen et al., 1998), was implicated in left-right (L-R) asymmetry(Table 1), a process not previously identified as being affected in CKO embryos and which should now be analyzed further (Lickert et al.,2002). Classification of the downregulated genes by Gene Ontology(GO) terms revealed that transcription factors, as well as genes implicated in pattern specification and morphogenesis, were highly enriched among the downregulated genes when compared with all genes on the U74Av2 GeneChip(Fig. 1B). Comparing our data with the reported expression profiles of pre-gastrulation stage β-catenin homozygous null mutants at E6.0 and E6.5 gave almost no overlap in the datasets (Morkel et al., 2003)(see Table S2 in the supplementary material), suggesting that the transcriptional Wnt response changes dynamically over the course of development.

To confirm that downregulated genes are differentially expressed in domains of Wnt activity in wild-type and CKO embryos, in situ hybridization was performed at E7.5 (Fig. 2). Wnt ligands, such as Wnt3a and Wnt8, as well as the secreted Wnt inhibitor Frzb1, and the negative regulator Axin2, showed strong downregulation in the primitive streak (PS) of CKO embryos(Fig. 2A-D). These results suggest both positive- and negative-feedback regulation in the Wnt/β-catenin signaling pathway. We also found that the PS expression of Notch1 and its ligand Dll1 was downregulated inβ-catenin mutants (Table 1, Fig. 2E,F). In addition, several Hox genes were downregulated in β-catenin mutants(Table 1), and Hoxb1was absent in β-catenin mutants (Fig. 2H). Furthermore, Gbx2, which is normally expressed in the posterior epiblast at late gastrulation stage(Wassarman et al., 1997), was completely lost in CKO embryos (Fig. 2G). Concordant with the loss of Gbx2 expression, we found a posterior expansion of the expression of Sox2, a marker of the anterior epiblast (Fig. 2I)(Wood and Episkopou, 1999),which was upregulated in the microarray experiments(Table 1).

In summary, all nine genes tested by in situ hybridization accurately reflected the GeneChip results.

To discover new components of Wnt/β-catenin signaling in developing embryos, we investigated further the expression of 16 downregulated genes(EST1-16) for which there was little or no published evidence concerning their developmental roles at the time the screen was conducted. Upon in situ hybridization in wild-type embryos at E7.5 and E8.5, eight of the genes showed expression patterns that strongly overlapped with known regions of high Wnt reporter activity (Fig. 3)(Mohamed et al., 2004). All genes except EST6 showed expression in the PS region at gastrulation stages(Fig. 3). In addition, Grsf1, Punc and Zic3 showed expression in the mid/hindbrain region at E8.5, while Scap2, Punc, Zic3, Fragilis2 and EST16 were expressed in the paraxial mesoderm, somites or neurectoderm of the tailbud region (Fig. 3). Interestingly,EST6 showed strong expression in the extraembryonic ectoderm and in a row of cells anterior to the node, known regions of organizing activity(Fig. 3C). EST10 showed a specific expression pattern in the definitive endoderm around gastrulation by whole-mount in situ hybridization and histological sectioning of stained embryos (Fig. 3F and data not shown).

For further analysis, we used transgenic RNA interference (RNAi)(Kunath et al., 2003) to analyze the function of the newly identified potential Wnt/β-catenin target genes, Irx3, Scap2, Smarcd3, Fragilis2 and Grsf1(Table 1). Because the results for the Smarcd3 knock-down analysis was recently published(Lickert et al., 2004) and because in a first attempt we were not able to knock down Scap2 and Irx3, we focus here on the analysis of Grsf1 and Fragilis2 (Fig. 3A,G).

Grsf1 codes for the mouse ortholog of the human G-rich sequence specific binding factor1 (GRSF1), which was previously shown to bind to a specific consensus sequence in the 5′-UTR of influenza virus nucleocapsid mRNA and thereby act positively on translation(Park et al., 1999; Kash et al., 2002). Fragilis2 belongs to a family of interferon-inducible genes with five members (Fragilis1-5), clustered on 68 kb of mouse chromosome 7 and associated with germ-cell differentiation(Tanaka and Matsui, 2002; Lange et al., 2003).

Both Grsf1 and Fragilis2 contain several putative Wnt-responsive elements in inter- and intragenic regions (see Fig. S1 in supplementary material). To test the hypothesis that Grsf1 and Fragilis2 are components of the Wnt-signaling response, we first monitored mRNA expression in ES cells that had been co-cultured in transwell filters on top of Wnt1-expressing 3T3 fibroblasts(Fig. 4A)(Lickert et al., 2000). Quantitative RT-PCR revealed that Wnt1 induced endogenous Grsf1 mRNA expression in ES cells by twofold, but had no effect on Fragilis2expression, which is already highly expressed in ES cells per se(Fig. 4D). We then used in situ hybridization analysis to test the mRNA expression level of both genes in wild-type and CKO embryos at E7.5 (Fig. 4B,C). Consistent with the GeneChip experiments, the expression of both genes is absent in the PS of CKO embryos, but weak Fragilis2expression remained at the base of the allantois, where the primordial germ cells (PGCs) are located. Because PGCs and ES cells share many molecular markers and cellular properties, this might suggest that the regulation of Fragilis2 expression in PGCs and ES cells is independent of Wnt/β-catenin, but is dependent on this signaling system in the paraxial mesoderm emerging from the PS. This idea is further supported by transgenic enhancer studies of a PGC-specific enhancer located in intron1 of Fragilis2, which does not depend on two Wnt-responsive elements (data not shown). Taken together with the co-expression of Grsf1 and Fragilis2 in the regions of high Wnt activity, these results suggest that both genes represent good candidate Wnt/β-catenin target genes.

To test whether these genes have necessary functions in domains of Wnt activity in embryonic development, we used RNAi in ES cells(Kunath et al., 2003) to knock down Grsf1 and Fragilis2. Stably transfected R1 ES-cell clones were selected and tested for successful silencing at the mRNA level using northern blotting (Fig. 4D). We were able to obtain ES-cell clones for both genes with very low residual levels of specific mRNA expression(Fig. 4D; Grsf1,clones 5 and 6; Fragilis2, clones 1 and 2), indicating that neither Grsf1 nor Fragilis2 function is essential for ES-cell growth or maintenance. To examine the effects of the expected silencing of these genes in embryos, we then used the tetraploid aggregation technique(Nagy et al., 1993) to generate completely ES cell-derived embryos of the efficiently silenced ES-cell lines (Grsf1, clones 5 and 6; Fragilis2, clones 1 and 2). In situ expression analysis at E7.5 revealed substantial silencing of Grsf1 and Fragilis2 mRNA levels, indicating that the silencing effect seen in ES cells is stably maintained in ES cell-derived embryos (Fig. 4E,F).

Grsf1 shRNA-silenced ES cell-derived embryos did not show any obvious phenotype at E7.5 (clone 5, n=7; clone 6, n=5). At E9.5 (clone 5, n=10; clone 6, n=25), we consistently found two phenotypes: (1) truncation of the posterior axis with formation of a large allantois; and (2) abnormal mid/hindbrain development(Fig. 5A). The phenotypes observed were restricted to the expression domains of Grsf1 in the posterior epiblast and in the mid/hindbrain region, suggesting that the effects seen are specific for Grsf1 gene knock down(Fig. 3A). Shortening of the tailbud region was clearly seen at E8.5 (clone 5, n=6; clone 6, n=8), when the mid/hindbrain region still looked morphologically normal (Fig. 5B). The abnormally large allantois is most likely a secondary effect due to the truncation of the posterior axis and failure of chorio-allantoic fusion. A more detailed histological analysis of the mutants at E9.5 revealed an overgrowth of neurepithelium in the mid/hindbrain region of Grsf1-silenced embryos (Fig. 5C). Loss of epithelial integrity was seen in the neural tube from the level of the septum transversum in the midtrunk region of Grsf1-silenced embryos and extending posteriorly(Fig. 5D), which might be due to secondary effects, because the neural tube rapidly degenerates following a failure of allantoic placental development. To avoid studying a secondary degeneration phenotype at this late stage, we analyzed the expression of marker genes in knock-down embryos prior to the 13-somite stage. At the 7- to 8-somite stage, Grsf1-silenced embryos showed slightly reduced levels of Fgf8 expression in the isthmus region, whereas expression in the tailbud was strongly reduced (Fig. 5E) (Crossley and Martin,1995). By the 10-somite stage, Grsf1-silenced embryos showed barely detectable levels of Fgf8 in the midline of the mid/hindbrain region and almost no expression in the lateral regions of the neural tube (Fig. 5F). Expression of Gbx2 in rhombomeres (r) 1 and 2 of knock-down embryos was strongly reduced at the 7- to 8-somite stage(Fig. 5G), whereas the hindbrain marker Krox20 appeared to be normally expressed in r3 and r5 in these mutants (Fig. 5H). Interestingly, Gbx2 expression in the tailbud appeared to be normal(Fig. 5G), in contrast to the reduced level of Fgf8 in this region(Fig. 5E), which indicates that the knock down of Grsf1 selectively alters the expression of marker genes. Onset of expression of the potential β-catenin target Gbx2 (Fig. 2G) in the epiblast seemed to be unaffected at head-fold stage(Fig. 5I), indicating that Grsf1 is important for maintaining Gbx2 expression at later developmental stages in the mid/hindbrain region.

We next tested the mRNA expression of known Wnt/β-catenin target genes, Cdx1 and T, in Grsf1 knock-down mutants. At gastrulation stages, Cdx1 and T were normally expressed in the PS of mutant embryos (Fig. 5J-K), whereas the expression domain of T in the axial mesoderm did not extend as far anteriorly in the mutants as it did in wild-type embryos (Fig. 5K,arrows). Interestingly, T expression in the axial mesoderm of Grsf1 knock-down embryos was greatly reduced at head-fold stage(Fig. 5L, arrows). Taken together, these results indicate that Grsf1 function is essential for posterior axis elongation, midbrain development and axial mesoderm specification.

Fragilis2-silenced embryos also did not show any obvious phenotype at E7.5 (clone 1, n=6; clone 2, n=4). Embryos analyzed between E9.0 and E9.5 (clone 1, n=8; clone 2, n=13) revealed problems in somite formation and a truncation of the posterior axis(Fig. 6A). Similar to Grsf1-silenced embryos, Fragilis2-silenced embryos also developed a large allantois, presumably due to the posterior truncation. In Fragilis2-silenced embryos at E8.5 (clone 1, n=12; clone 2, n=8), the somites appeared hollow and were irregular in shape and smaller in size (Fig. 6B, see b′,b″). Additionally, the neural tube appeared kinked, a phenotype frequently seen in mutants affecting somite formation(Conlon et al., 1995). Histological analysis of the Fragilis2-silenced embryos at E8.5 revealed abnormalities in epithelialization and/or maintenance of epithelial integrity of the somites (Fig. 6C). As the Fragilis gene family is implicated in PGC development and Fragilis2 is expressed in this cell population, we stained Fragilis2-silenced embryos for tissue non-specific alkaline phosphatase (AP). No difference in the AP staining between wild-type and knock-down embryos was observed, suggesting that the formation of PGCs was normal at head-fold stage (Fig. 6D).

Wnt/β-catenin signaling is implicated in both the formation of paraxial mesoderm (Takada et al.,1994; Galceran et al.,1999) and the subsequent segmentation of presomitic mesoderm into somites (Auhlela et al., 2003; Galceran et al., 2004; Hofmann et al.,2004). To discriminate between possible defects in these two processes in the Fragilis2 knock-down embryos, we analyzed Tand Tbx6, genes implicated in paraxial mesoderm formation, and PAPC, a gene important for epithelialization of the somites(Rhee et al., 2003). Expression of T and Tbx6 was unaffected in mutant embryos at head-fold stage, but decreased in the tailbud region at E8.5(Fig. 5E). Expression of PAPC in wild-type embryos is restricted to two stripes at the anterior end of the presomitic mesoderm corresponding to the next two presumptive somites(Fig. 5E; somite 0 and–1). Analysis of Fragilis2 knock-down embryos revealed that PAPC expression is decreased at head-fold stage and was barely detectable in the two presomitic stripes at 10-somite stage(Fig. 5E). Taken together,these results demonstrate that knock down of Fragilis2 predominantly affects the epithelialization of somites, and to a lesser extent the formation of paraxial mesoderm, suggesting that Fragilis2 acts downstream of Wnt/β-catenin to regulate these processes.

Previous studies have highlighted the important role of Wnt/β-catenin signaling in the control of various developmental processes at the time of gastrulation and axial patterning in the early mouse embryo(Sokol, 1999; Beddington and Robertson, 1999; Yamaguchi, 2001; Lu et al., 2001). To understand mechanistically how the Wnt/β-catenin signaling cascade acts on these developmental processes, it is necessary to identify the target genes of Wnt signaling in different developmental domains. Using GeneChip analysis of conditional β-catenin mutants and wild-type embryos at late-gastrulation stages, we identified many potential Wnt/β-catenin target genes enriched for developmental components involved in pattern specification and morphogenesis, including 26 well-characterized signaling molecules and transcription factors (Table 1 and Fig. 1B). Further clustering of these genes into different functional groups, e.g. genes involved in paraxial mesoderm and somite formation, genes important for endoderm formation, factors involved in patterning and morphogenesis, and genes implicated in L-R axis formation(Table 1), reveals the different molecular programs potentially controlled by β-catenin during gastrulation.

As well as identifying sets of genes possibly downregulated by Wnt/β-catenin signaling at gastrulation, we also found a large number of genes that were upregulated in the conditional β-catenin mutants. One possible explanation for this is offered by the morphogenetic defects observed in β-catenin mutant embryos, such as retention of the visceral endoderm(VE) because of a lack of definitive endoderm formation, which normally displaces VE into the extraembryonic region. Consistent with this interpretation, Dab2 and Pem, two marker genes for VE, were upregulated in β-catenin mutants(Table 1). However, we also found genes specifically expressed in the embryo to be upregulated, e.g. Sox2, an anterior epiblast marker, which later becomes restricted to the neurectoderm. In situ expression analysis showed that Sox2becomes expressed in the posterior region of β-catenin mutants,suggesting that the posterior epiblast has acquired an anterior fate. Further mining of the upregulated set of genes in the conditional β-catenin mutants might reveal novel components of visceral endoderm formation or anterior specification.

The objective of target gene screens is not only to identify characterized and functionally annotated genes, but also to add new players and their respective function to the gene regulatory network. All the genes we tested,whether well characterized or less well annotated, showed the expected expression differences between wild type and β-catenin mutants by in situ hybridization. Thus, we expect that our dataset will provide a rich resource for future data mining to characterize Wnt/β-catenin pathways in gastrulation. Ideally, the relative importance of target genes needs to be tested by assessing their function during development. Using shRNA-mediated gene silencing in ES cells and then in ES-derived embryos(Kunath et al., 2003), we have identified and characterized two novel putative Wnt/β-catenin target genes, Grsf1 and Fragilis2, whose expression is required for normal development. Both genes, when knocked down, recapitulate specific but distinct aspects of the conditional β-catenin mutant phenotype,implicating them as crucial downstream mediators of the Wnt/β-catenin signaling pathway.

The human ortholog GRSF1 is a sequence-specific RNA-binding protein, and has been shown to act positively on translation in vitro and in a cell-culture system (Park et al., 1999; Kash et al., 2002). This raises the interesting possibility that Wnt induction of Grsf1 selectively activates the translation of other mRNA transcripts in the primitive streak and/or the mid/hindbrain region. Using a computational approach for predicting possible target genes of Grsf1, we have screened the genes expressed at late gastrulation stage according to our U74A Affymetrix wild-type data set. From 4694 annotated 5′ UTRs in the ENSEMBL database, we found 386 non-redundant genes with at least one high-affinity Grsf1 consensus binding site (5′-AGGGU-3′; see Table S3 in the supplementary material). Interestingly, among these genes we found developmental regulatory factor genes, such as T, Hoxb1, Hoxb8 and Frzb1, which are co-expressed with Grsf1 at gastrulation stage. We also found genes regulating cell proliferation, such as p53, cyclin B1, cyclin A2 and Cdk2, and genes regulating apoptosis, Bcl2 and Bax,as candidate Grsf1 target genes. In-depth analysis of these potential target genes will be required to dissect the mechanisms by which Grsf1 regulates mid/hindbrain development, posterior elongation and axial mesoderm specification.

Importantly, the observed Grsf1 knock-down phenotypes remarkably recapitulate distinct aspects of the CKO mutant phenotype and other Wnt pathway mutants (Lickert et al.,2002; McMahon and Bradley,1990; Thomas and Capecchi,1990; Brault et al.,2001), suggesting that Grsf1 is a crucial mediator of the Wnt/β-catenin signaling cascade. Interestingly, the lack of Texpression in the anterior primitive streak of Grsf1 knock-down embryos is comparable to lack of T expression in Wnt3a mutants(Yamaguchi et al., 1999),offering an explanation for the axis truncation in both mutants. The normal expression of the Wnt/β-catenin target genes, Cdx1 and Grsf1, in Grsf1 knock-down embryos suggests that Grsf1 acts downstream of the Wnt/β-catenin signaling pathway selectively on target mRNAs and is not involved in signal transduction, e.g. by stabilizing components of the pathway. This might also be the case for mid/hindbrain development, where Grsf1 is necessary for maintaining Fgf8and Gbx2 expression, two factors important for the establishment of the mid/hindbrain boundary. The comparison of putative mRNA targets of the RNA-binding factor Grsf1 (see Table S3 in the supplementary material) with all the deregulated genes from the β-catenin target gene screen (Table S1 in supplementary material) revealed several potentially coregulated transcripts(see Table S4 supplementary material), which might explain similarities in the Grsf1 and CKO mutant phenotypes.

Fragilis2 is expressed in the primitive streak, including the base of the allantois, where the PGCs are localized at late gastrulation stage, and in the paraxial and lateral mesoderm, as well as in the first forming somites at E8.5. Studies in the immune system suggest a role for Fragilis2 (human orthologs Leu13/9-27/IFITM1) as part of a transmembrane multiprotein signaling complex implicated in inhibition of cell proliferation and homotypic cell adhesion (Knight et al., 1985; Deblandre et al., 1995; Sato et al., 1997). Histological analysis of Fragilis2-silenced embryos revealed a defect in epithelialization of the somites, consistent with a function in homotypic cell adhesion. Additionally, marker gene analysis revealed that Fragilis2knock-down embryos show reduced expression of PAPC, a gene implicated in somite epithelialization, and reduced expression of the paraxial mesoderm markers T and Tbx6 at tailbud stage. These phenotypes are very similar to the paraxial mesoderm and somite segmentation defects seen in several different Wnt mutants (Lickert et al., 2002; Takada et al.,1994; Galceran et al.,1999; Aulehla et al.,2003; Galceran et al.,2004; Hofmann et al.,2004), thus it seems likely that Fragilis2 is a crucial downstream mediator of the Wnt/β-catenin signaling cascade in these processes, mediating homotypic cell adhesion.

By using RNAi-mediated gene functional studies in ES cell-derived embryos,we have shown that it is possible to rapidly evaluate the relative importance of putative target genes of developmental pathways identified from expression profiling of mutant versus wild-type embryos. The potential for parallel functional analyses of several candidate genes in a relatively high throughput manner is an important component of genome-wide approaches to developmental genomics in the mouse.

We are grateful to Sue MacMaster for excellent help in the generation of chimeras. We thank Ken Harpal for help with sectioning. We thank Randy Cassada and Tilo Kunath for suggestions and critical reading of the manuscript. We are also grateful to Jerry Gish and Tony Pawson for providing us the shRNA plasmid. This work was supported by the Canadian Institutes of Health Research. H.L. was supported by a CIHR post-doctoral research fellowship and is currently supported by the Emmy-Noether fellowship from the Deutsche Forschungsgemeinschaft. J.R. is a CIHR Distinguished Scientist.