Sfrp1 and Sfrp2 regulate anteroposterior axis elongation and somite segmentation during mouse embryogenesis

Cell-cell communication is essential for embryogenesis with respect to cell fate specification, cell polarity, cell behavior and embryonic patterning. Intercellular communication is critically controlled by signaling molecules that include ligand-receptor pairs and their antagonists. Although activation of signaling via interactions between ligands and receptors exerts profound effects, the regulation of activity by secreted antagonists is indispensable for proper cell-cell communication(Niehrs, 1999; Perea-Gomez et al., 2002).

Wnt, a family of secreted glycoproteins, stimulates signals through the seven transmembrane receptors of Frizzled(Kawano and Kypta, 2003). Wnt interactions with Frizzled and the co-receptor Lrp5/Lrp6 activate the canonical Wnt/β-catenin pathway, which leads to the stabilization ofβ-catenin as a transcriptional regulator in the nucleus(Kawano and Kypta, 2003; Kelly et al., 2004). A proportion of Wnt molecules activate the non-canonical planar cell polarity(PCP) pathways, which results in activation of Rho GTPases, or the Wnt/Ca²⁺ pathway, which leads to intercellular Ca²⁺release, and PKC and CamKII activation(Kühl, 2002). Nineteen Wnt genes have been identified in the human and murine genomes. Studies of loss-of function mutations in the mouse, and knockdown or overexpression in Xenopus, chicken and zebrafish, have revealed that Wnt proteins play diverse roles during embryogenesis in vertebrates. Wnt signaling is required for primitive streak formation (Liu et al., 1999), mesoderm cell movement(Heisenberg et al., 2000; Ulrich et al., 2003),generation of the tail organizer (Agathon et al., 2003), posterior patterning and somitogenesis(Takada et al., 1994; Aulehla et al., 2003) (see also http://www.stanford.edu/~rnusse/wntwindow.html).

Wnt signaling is inhibited by several secreted antagonists: Dkk, Wise, Wif and Sfrp (Kawano and Kypta,2003; Yamaguchi,2001). Dickkopf (Dkk) and Wnt modulator in surface ectoderm (Wise)interact with co-receptor Lrp6, leading to the inhibition of active ligand-receptor complex formation(Semënov et al., 2001; Itasaki et al., 2003). Secreted frizzled-related protein (Sfrp) and Wnt inhibiting factor (Wif)possess a region related to the cysteine-rich domain (CRD) of Frizzled that interacts with the Wnt ligand (Hsieh et al., 1999; Kawano and Kypta,2003). This inhibition of the ligand might regulate the degree of active Wnt in tissues (i.e. regulation of ligand activity and gradient generation). Regulation of Wnt signaling by secreted antagonists has been implicated in anterior embryonic patterning. During mouse embryogenesis,inhibition of Wnt activity by Dkk1 is necessary for anterior head specification and limb patterning(Mukhopadhyay et al., 2001). In Xenopus, overexpression experiments revealed that Dkk1 and FrzB (a member of the Sfrps) antagonize Wnt in the Spemann organizer, which is required for head organizer activity(Glinka et al., 1997; Leyns et al., 1997; Wang et al., 1997; Niehrs, 1999).

The Sfrp gene family consists of five members in both the human and mouse genomes (Kawano and Kypta,2003), which are divided into Sfrp1 and FrzB subfamilies based on amino acid sequence similarity. In addition, Crescent and Sizzled are unique members found in Xenopus and zebrafish. Sfrp1, Sfrp2 and Sfrp5 belong to the Sfrp1 subfamily(Jones and Jomary, 2002),whose members share homology in the CRD domain. Consequently, Sfrp1, Sfrp2 and Sfrp5 suppress the canonical Wnt/β-catenin signal, decreaseβ-catenin levels and downregulate target gene expression, such as Myc expression, in cultured cells(Suzuki et al., 2004). Tlc,the zebrafish ortholog of Sfrp1 and Sfrp5, antagonizes Wnt8b in order to establish the telencephalon (Houart et al., 2002). In vitro data and Xenopus experiments suggest that Sfrp1 (FrzA) interacts with Wnt1, Xwnt8 and Wnt2, but not with Wnt5a(Xu et al., 1998; Dennis et al., 1999). Furthermore, Sfrp1 also interacts with Wnt7b, and is capable of attenuating the non-canonical Wnt signaling pathway, leading to inhibition of axon guidance (Rosso et al., 2005). Thus, Sfrps exhibit a wide spectrum of functions with respect to regulation of the activity of Wnt proteins in both the canonical and non-canonical pathways. However, the functions of the Sfrps in embryogenesis have not been fully elucidated.

In this study, Sfrp1 and Sfrp2 double mutant mice were generated to reveal the functions of Sfrp1 subfamily members during embryogenesis. The findings demonstrate that Sfrp1 and Sfrp2exhibit highly redundant functions during embryogenesis. Furthermore, the results suggest that regulation of Wnt signaling by Sfrp1 and Sfrp2 is required for normal AP axis elongation and somitogenesis.

A 6.5-kb BamHI-NheI genomic fragment of Sfrp1,containing the first exon, was obtained from the C57BL/6 BAC clone (BACPAC resources) and subcloned into a plasmid to produce a Sfrp1 knock-in(KI) vector. The KI vector contained a 5′ 1.4-kb BamHI-SacII fragment and a 3′ 5.3-kb EagI-NheI fragment of the initially subcloned fragment (see Fig. S1 in the supplementary material). The coding sequence of the first exon was deleted and replaced with a nuclear-localized lacZ KI cassette(generated by Dr Maki Wakamiya, M. D. Anderson Cancer Center, The University of Texas). β-galactosidase (β-gal) staining revealed that reporter expression mimicked endogenous Sfrp1 expression (data not shown). The Sfrp2 targeting construct was generated from a 14-kb BamHI fragment containing the entire Sfrp2 coding sequence in the C57BL/6 BAC clone. The vector contained a 5′ 2.0-kb Asp718-EcoRI fragment and a 3′ 6.5-kb EcoRV-BamHI fragment of the subcloned fragment (see Fig. S1 in the supplementary material). Electroporation of the vectors into embryonic stem (ES) cells and identification of homologous recombination events were performed with a strategy similar to that of Shimono and Behringer(Shimono and Behringer,2003).

For whole embryo immunostaining against neurofilament, embryos were fixed in 3.5% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) at 4°C overnight. The embryos were hydrated in an ascending methanol series (25, 50,75, 100%) in PBS containing 0.1% Tween-20 (PBT) before being re-hydrated in a descending methanol series (75, 50, 25, 0%). Following treatment with PBT containing 6% H₂O₂ and a wash in Tris-buffered saline containing 0.1% Tween-20 (TBST), embryos were treated for blocking in TBST containing 10% sheep serum. The embryos were incubated with a 2H3 monoclonal neurofilament antibody (supernatant, 1/50 dilution) at 4°C overnight. Embryos were then incubated with anti-mouse IgG antibody conjugated with HRP(1/1000 dilution). Immunoreactivity was detected with diaminobenzidine. Antibody staining against diphosphorylated ERK was conducted according to a previous report (Corson et al.,2003).

To detect non-phospho β-catenin, embryos were fixed in 8% PFA in PBS. Cryosections were subsequently generated according to a general protocol(Wakamatsu et al., 1993). Primary antibody against non-phospho β-catenin (Upstate; mouse monoclonal 8E4) was used at a 1:400 dilution, with TBST containing 5% skim milk. Immunoreactivity was detected with Alexa488-conjugated anti-mouse IgG antibody(Molecular Probes) diluted 1:200 with TBST containing 3% BSA. Images were captured on a BioRad Radiance 2100 Laser Scanning Confocal Microscope System equipped with a Zeiss Axiovert, and processed with Adobe Photoshop CS.

In situ hybridization using digoxigenin (DIG)-labeled riboprobes was performed on whole-mount embryos according to Wilkinson(Wilkinson, 1992). Some embryos were processed to generate cryosections. Double whole-mount in situ hybridizations were conducted according to Wilkinson(Wilkinson, 1992), with modifications, using two differentially labeled probes: one labeled with DIG and the other with fluorescein (FITC). After hybridization using two different probes at same time, DIG-labeled RNA probe was detected with alkaline phosphatase-conjugated anti-DIG antibody in color solution containing NBT(Nitro-Blue Tetrazolium Chloride)/BCIP (5-Bromo-4-chloro-3-indolyl phosphate,toluidine salt). Following the primary detection procedure, the antibody was removed by washing with 0.1 M Glycine-HCl containing 0.1% Tween-20. A secondary detection procedure was then performed with an alkaline phosphatase-conjugated anti-FITC antibody in INT[2-(4-Iodophenyl)-3-(4-nitrophenyl)-5-phenyltetrazolium chloride]/BCIP solution.

Hoxa7 (BC036986) and Hoxd10 (BC048690) cDNA were obtained as IMAGE clones from Invitrogen. Axin2 (AK084644), Fgf17(AK077555) and Nkd1 (AK082367) cDNA were obtained as FANTOM clones(RIKEN). A 1.5-kb fragment containing the 3′ UTR of Hoxb2 cDNA was isolated from the E8.5 cDNA library (a gift from Dr H. Hamada, Osaka University). Images were captured with a Pixera Pro600ES digital camera, with a Zeiss Stemi2000-C stereomicroscope.

Whole-embryo culture and DiI(1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate) labeling were performed according to the method of Shimono and Behringer (Shimono and Behringer,2003). DiI was injected at the primitive streak region in the tail bud in order to label mesoderm cells. Labeled cells within the endoderm layer and the neural plate did not spread during the shorter culture period. Mesoderm cell labeling was also confirmed in cryosections of the cultured embryos. Labeled embryos, which were selected based on a lateral view of the labeling of three germ layers, were cultured for two hours. The images were captured with a Zeiss AxioCam HRc digital camera with a Leica MZFLIII fluorescence stereomicroscope. The culture was extended up to 22 hours to confirm a contribution of the labeled cells to the pre-somitic mesoderm (PSM). Embryos displaying labeled PSM cells were adopted (n). Tail-bud culture was performed based on the method of Correia and Conlon(Correia and Conlon,2000).

Sfrp1 and Sfrp2 mutations were introduced via a deletion of the first exon encoding the first methionine and the CRD domain to generate mutant mice. For the Sfrp1 KI mutant, the first exon was replaced with a nuclear-localized lacZ knock-in cassette in ES cells (see Fig. S1A,B in the supplementary material). Similarly, the first exon of Sfrp2 was replaced with a PGKneobpAloxB cassette(Shimono and Behringer, 2003)to generate conventional knockout mice (see Fig. S1D,E in the supplementary material). Following germline transmission, mutant pups were obtained from the intercross of heterozygous mutant mice with 129 and C57BL/6 mixed backgrounds. Genotypes of the pups were determined by Southern hybridization or PCR amplification at weaning, suggesting that mice carrying the homozygous mutation in Sfrp1 (Sfrp1^-/-) appeared normal during embryogenesis (see Fig. S1C, Table S1 in the supplementary material)(Bodine et al., 2004). Similarly, most of the Sfrp2 homozygous mutant(Sfrp2^-/-) mice appeared normal and healthy (Fig. S1F,Table S1 in the supplementary material), although hindlimb syndactyly occurred at a lower frequency (3%; two out of 62 Sfrp2^-/- pups). The deleted exons encode the first methionine and the functional domain of the protein. As a result, the mutant mice were regarded as functionally null. In addition, expression of the transcript from the mutated alleles was undetectable (data not shown).

Functional redundancy between Sfrp1 and Sfrp2 has been suggested on the basis of the similarity of their expression patterns during embryogenesis. Overlapping expression of Sfrp1 and Sfrp2 is evident in the forebrain, midbrain and hindbrain region and in the posterior neural plate/tube at embryonic day (E) 8.5 and E9.5(Leimeister et al., 1998). A different study reported overlapping expression in the PSM(Lee et al., 2000). Therefore, Sfrp1^-/-;Sfrp2^+/- mice, which appeared normal and healthy, were intercrossed to generate Sfrp1 and Sfrp2double homozygous mutant mice(Sfrp1^-/-;Sfrp2^-/-). No Sfrp1^-/-;Sfrp2^-/- pups were recovered,suggesting pre-natal lethality of this mutation (see Tables S1, S2 in the supplementary material). Thus, Sfrp1 and Sfrp2 are functionally redundant.

Sfrp1^-/-;Sfrp2^-/- embryos, which died at around E16.5, exhibited edematous, craniofacial defects, limb outgrowth defects and extra digits (Fig. 1A-E; see also Table S2 in the supplementary material). The limb outgrowth defect is a typical phenotype induced by activated Wnt/β-catenin signaling, such as constitutively active β-catenin expression, Wnt overexpression and Wnt5a inactivation(Topol et al., 2003; Guo et al., 2004). Extra digits were frequently observed in the anterior portion of the right hindlimb(50% of embryos, n=16), but rarely in the left hindlimb(Fig. 1F-I), a finding that was correlated with extra stripes of Fgf8 expression domains on the ventral surface of the hindlimb bud at E10.5(Fig. 1J,K). Ectopic Fgf8 expression has been associated with an upregulation of Wnt/β-catenin signaling in the apical ectodermal ridge during limb morphogenesis (Mukhopaghyay at al.,2001; Barrow et al.,2003; Soshnikova et al.,2003). Sfrp1 is expressed in the ventral body wall,including the proximal region of the hindlimb bud, and Sfrp2 is expressed in the limb mesenchyme at E10.5(Leimeister et al., 1998).

Patterning in the forebrain, midbrain and hindbrain of Sfrp1^-/-;Sfrp2^-/- embryos appeared normal up to E9.5-E10.5. Although Sfrp1 and Sfrp2 are highly expressed in the hindbrain (Leimeister et al.,1998), rhombomere patterning was normal in Sfrp1^-/-;Sfrp2^-/- embryos, as evidenced by staining for Krox20 transcripts in rhombomeres 3 and 5, and immunoreactivity staining with the 2H3 anti-neurofilament monoclonal antibody(data not shown).

Gross morphology suggested that the entire AP body axis was shortened in Sfrp1^-/-;Sfrp2^-/- embryos at E14.5-E16.5. Moreover, cartilage staining revealed that the thoracic region was severely shortened, and that the number of thoracic vertebrae was reduced from thirteen to five (Fig. 1B-E). This observation suggested that AP axis formation, especially in the thoracic region, might be affected in Sfrp1^-/-;Sfrp2^-/-embryos in a manner that correlates axis elongation with somite segmentation. To elucidate the body axis at the marker level, expression of the following Hox genes was examined at E9.25-E10.5: Hoxb2, whose anterior expression boundary is located at rhombomere 3(Rossel and Capecchi, 1999); Hoxa7, which has an anterior boundary in the mesoderm at the thirteenth somite (Li and Shiota,1999); and Hoxd10, which is expressed up to the twenty-seventh somite (Hérault et al., 1998). Hoxb2 was expressed in the posterior region from rhombomere 3 in Sfrp1^-/-;Sfrp2^-/- embryos,as well as in the control (wild type, Sfrp1^-/-;Sfrp2^+/+ and Sfrp1^-/-;Sfrp2^+/-) embryos(Fig. 2A,B). The anterior boundary of Hoxa7 expression in the mesoderm was located around the thirteenth somite in the Sfrp1^-/-;Sfrp2^-/-embryos, a pattern similar to that of the control embryos at E9.25(Fig. 2C,D). However, using myogenin expression as the landmark in the double whole-mount in situ hybridizations (Fig. 2E-H)(Edmondson and Olson, 1989),the expression boundary of Hoxd10 was shifted to the twenty-first somite in Sfrp1^-/-;Sfrp2^-/- embryos at E10.5(Fig. 2E,F). These observations suggest that somites 11-23, which give rise to vertebrae in the thoracic region, are fused, and/or reduced in number, during somite segmentation(Fig. 2I).

Histological analysis disclosed randomized, incomplete segmentation of somites in Sfrp1^-/-;Sfrp2^-/- embryos at E9.5(Fig. 3A-D). A segmentation defect in the middle of the trunk at E10.5 was also revealed by immunostaining with an anti-neurofilament antibody (Fig. 3E,F) (Dodd et al.,1988). Histological analysis showed randomized segmentation after the eleventh somite in the region between the forelimb and the hindlimb(Fig. 3C,D). This finding was confirmed by marker analysis of Mox1(Fig. 4A-D), which is expressed in somites from the onset of segmentation and is gradually restricted to the posterior half-somite of sclerotome lineage(Candia et al., 1992), and of Pax3 (Fig. 4E-H), a marker for the dermatome lineage (Goulding et al., 1991). Regular, albeit smaller, somites were generated in a posterior region near the hindlimb, as well as around the forelimb level,suggesting restoration of somite segmentation around the hindlimb level (Figs 3, 4). The somites maintained AP identity within the region between the forelimb and the hindlimb, and the abnormal somites expressed Uncx4.1 and Pax1, markers for the sclerotome lineage (Mansouri et al.,1997; Neubuser et al.,1995) (Fig. 4I-K;data not shown). Although myotome differentiation was delayed in the embryos,as indicated by myogenin expression at E9.5(Edmondson and Olson, 1989)(Fig. 4L-N), these observations reveal that cell differentiation occurs in somites with impaired segmentation. Significantly, myogenin expression at E10.5 clearly demonstrated a reduced number of somites between the forelimb and hindlimb in Sfrp1^-/-;Sfrp2^-/- embryos (17 somites in Sfrp1^-/-;Sfrp2^+/+ or Sfrp1^-/-;Sfrp2^+/- embryos, n=3; 13 somites in Sfrp1^-/-;Sfrp2^-/- embryos, n=3; Fig. 4R-T). Moreover, the pre-somitic mesoderm (PSM) region was dramatically reduced in the Sfrp1^-/-;Sfrp2^-/- embryos, as evidenced by the expression pattern of the delta-like 1 gene (Dll1)(Bettenhausen et al., 1995) at E9.5 (Fig. 4O-Q). This observation, in addition to the small somite around the forelimb level, is suggestive of a defect in posterior embryonic patterning at an early stage. The defect of posterior patterning was apparent as a reduction in the posterior axis at E8.25, and appeared prior to small somite generation and irregular and incomplete somite segmentation.

Sfrp1^-/-;Sfrp2^-/- embryos were indistinguishable from wild-type and control embryos at the late head-fold stage. However, the Sfrp1^-/-;Sfrp2^-/- embryos,which had begun to generate somites, were distinct from control embryos in the posterior region in the increased thickness of the mesoderm layer (a bar in the tail region, Fig. 5A-C). The expression patterns of brachyury (T), Tbx6 and Dll1 were examined in order to gain insight into the defect in posterior axis extension. T expression in Sfrp1^-/-;Sfrp2^-/- embryos suggested that the primitive streak and the axial mesoderm (node and notochord) were generated normally in the early somite stage (∼three somites)(Fig. 5A-C)(Wilkinson et al., 1990). Tbx6 and Dll1 were expressed in the PSM region (between the arrow and arrowhead in Fig. 5D,E,G,H; data not shown) and in the paraxial mesoderm of the primitive streak region (between the bar and arrow in Fig. 5D,E,G,H; data not shown)in wild-type and control embryos at E8.5(Bettenhausen et al., 1995; Chapman et al., 1996). In the Sfrp1^-/-;Sfrp2^-/- embryos, Tbx6 and Dll1 expression were observed with high intensity in the paraxial mesoderm on both sides of the primitive streak (between the bar and arrow in Fig. 5F,I; data not shown);moreover, anterior extension of the expression domain was greatly reduced(between the arrow and arrowhead in Fig. 5F,I; data not shown). Sections of embryos at somite stages 6/7 and 11 revealed that unusual Dll1 staining was predominantly due to an increased number of paraxial mesoderm cells in the posterior embryonic portion (Fig. 5K,L,N,O; data not shown).

In order to identify the abnormality in posterior axis extension around E8.5, cell proliferation ratio was examined in Sfrp1^-/-;Sfrp2^-/- embryos. The cell proliferation rate, indicated by anti-phospho-Histone H3 antibody staining (a mitotic marker) (Chadee et al.,1995), was not elevated in the posterior region of Sfrp1^-/-;Sfrp2^-/- embryos in comparison with control embryos (7.36±1.30% of anti-phospho-Histone H3-positive cells in Sfrp1^-/-;Sfrp2^+/+ embryos; 6.48±1.63%in Sfrp1^-/-;Sfrp2^-/- embryos; n=3) at the 6 somite stage (data not shown). By contrast, the rate appeared to be reduced in Sfrp1^-/-;Sfrp2^-/- embryos at the 11 somite stage (7.83±0.73% of anti-phospho-Histone H3-positive cells in Sfrp1^-/-;Sfrp2^+/+ embryos; 3.76±0.88% in Sfrp1^-/-;Sfrp2^-/- embryos; n=3; data not shown). Thus, a reduction in cell proliferation may contribute to a diminished posterior axis at a later stage. However, a reduction of the posterior axis in conjunction with the increased thickness in the mesoderm cell layer is evident in Sfrp1^-/-;Sfrp2^-/-embryos at the end of the head-fold stage, when cell proliferation rate is unaffected. Therefore, accumulation of Dll1-expressing cells appears to be a consequence of cell migration defects in mesoderm cells along the AP axis.

Cell migration was directly evaluated via cell labeling with the lipophilic dye DiI. DiI was injected into the mesoderm cell population around the primitive streak region of control and Sfrp1^-/-;Sfrp2^-/- embryos. The DiI-labeled mesoderm cells migrated into the lateral region in the control embryos(Fig. 5P,Q; n=14). By contrast, migration of DiI-labeled mesoderm cells was greatly reduced in Sfrp1^-/-;Sfrp2^-/- embryos(Fig. 5R,S; n=5). Thus, Sfrp1 and Sfrp2 mediate posterior axis extension via regulation of cell migration in the paraxial mesoderm.

Wnt3a is expressed in the tail bud of developing embryos(Yoshikawa et al., 1997). Because Sfrp1 and Sfrp2 possess inhibitory activity against the Wnt pathway,activation of Wnt signaling was examined in Sfrp1^-/-;Sfrp2^-/- embryos at E8.5. Activation of the Wnt/β-catenin pathway was evaluated by antibody staining against stabilized non-phospho β-catenin. A higher staining intensity in the cellular membrane and nucleus was observed in the hindgut and in the mesoderm beneath the primitive streak in control embryos(Fig. 6A,B,G-I). In Sfrp1^-/-;Sfrp2^-/- embryos, a similar staining intensity was ectopically detected in the tail bud region in the mesoderm and neural ectoderm, as well as in the hindgut endoderm and the mesoderm beneath the primitive streak (n=2; Fig. 6D-E,J-O). Thus, inactivation of Sfrp1 and Sfrp2leads to activation of the Wnt pathway in the embryos.

Wnt and Fgf8 gradients play a role in the establishment of somite segmentation boundaries (Aulehla et al.,2003). A steeper gradient in a shorter PSM region could lead to the generation of small somites. However, the shorter PSM is insufficient to account for the randomized and incomplete segmentation in Sfrp1^-/-;Sfrp2^-/- embryos. Coordinated somite segmentation is regulated by the cyclic expression of Notch related genes,such as Lfng and Hes7, in the PSM(Saga and Takeda, 2001; Bessho et al., 2001). Because Wnt3a is required for oscillating Notch signaling activity in the PSM(Aulehla et al., 2003), the expression of Wnt3a in Sfrp1^-/-;Sfrp2^-/- embryos was examined during incomplete somite segmentation and following the restoration of regular somite segmentation. The level of Wnt3a expression was normal in the tail bud region of Sfrp1^-/-;Sfrp2^-/- embryos displaying defective somite segmentation at E8.5 (see Fig. S2A-C in the supplementary material). In addition, Fgf8 was normally expressed in the tail bud of these embryos (see Fig. S2J-L′ in the supplementary material). Activation of Fgf signaling, visualized with an anti-diphospho-ERK antibody (Corson et al.,2003), was observed in posterior portion of the control embryos,and showed reduced activation in anterior region of the PSM (see Fig. S2M,N in the supplementary material). Similar staining patterns were observed in the Sfrp1^-/-;Sfrp2^-/- embryos (Fig. S2O in the supplementary material). However, Wnt3a expression decreased in the tail bud at E9.5, when somite segmentation was restored (see Fig. S3J-L in the supplementary material). In addition, Fgf8 expression was diminished in the tail bud, which was consistent with the findings of a previous report(Aulehla et al., 2003) (Fig. S3G-I in the supplementary material). By contrast, other marker genes of the tail bud region were normally expressed in Sfrp1^-/-;Sfrp2^-/- embryos(Maruoka et al., 1998;Yamguchi et al., 1999a) (Fig. S3A-F,M-O in the supplementary material). Hence,the defect is correlated with expression levels of Wnt3a. The reduction of Wnt3a expression in the tail bud may reduce the signaling activity elevated by Sfrp1 and Sfrp2 inactivation to levels that have no effect on somite segmentation.

This observation suggests the possibility that aberrant somitogenesis in Sfrp1^-/-;Sfrp2^-/- embryos occurs during the portion of the segmentation process related to Wnt3a expression(Aulehla et al., 2003). Therefore, expression of the Notch-related oscillator genes lunatic fringe(Lfng) and Hes7 was examined in Sfrp1^-/-;Sfrp2^-/- embryos at somite stages 10 to 13. Dynamic expression of Lfng, which is regulated by Notch signaling, is required for the somite segmentation process(Forsberg et al., 1998; Evrard et al., 1998; Cole et al., 2002; Morales et al., 2002),which is altered in Wnt3a mutants (Aulehla et al., 2003). Oscillating expression of Lfng was visualized via comparison between the half tail regions with and without culture incubation; in addition, the expression pattern was precisely compared between the explants according to Forsberg et al.(Forsberg et al., 1998)(Fig. 7G). All pairs of explants derived from control embryos at E8.5 exhibited the oscillating expression pattern (n=18; Fig. 7A,D,G). By contrast, the oscillations were perturbed in all pairs of Sfrp1^-/-;Sfrp2^-/- tail explants(n=7, Fig. 7B,C,E-G). An extra stripe was observed in the anterior-most region of the PSM,suggesting delayed somite segmentation (indicated by a dagger in Fig. 7C,F; two out of seven pairs of explants). Interestingly, cyclic expression of Lfng in the Sfrp1^-/-;Sfrp2^-/- explants at E9.5 was indistinguishable from that of control explants(Fig. 7H,I; n=11 for control, n=3 for Sfrp1^-/-;Sfrp2^-/-explants). Consequently, the abnormal expression pattern of the oscillator gene appears to be related to the defect in somite segmentation.

Hes7 expression is controlled by Notch signaling, and Hes7 protein represses Lfng expression (Bessho et al., 2001; Bessho et al.,2003). Although normal oscillating cycles of Hes7 were consistently observed in control explants(Fig. 7J,M; n=6),cyclic expression was affected in the Sfrp1^-/-;Sfrp2^-/- embryos (three out of five pairs of explants, Fig. 7L-O). In a manner similar to that of Lfng expression, a strong extra stripe frequently occurred in the anterior-most region of the PSM (indicated by a dagger in Fig. 7L,N; two out of five pairs of explants). Therefore, Sfrp1 and Sfrp2 affect Notch oscillator cycles.

Axin2 and Nkd1, negative regulators of the Wnt pathway,have been shown to exhibit an oscillatory expression pattern in the PSM(Aulehla et al., 2003; Ishikawa et al., 2004). On the one hand, Dynamic expression of Axin2, a Wnt-driven oscillator(Aulehla et al., 2003), was observed in Sfrp1^-/-;Sfrp2^-/- embryo explants(n=5), as well as in controls (n=19). On the other hand, the majority of control explants (89%; total n=9) exhibited changes in expression levels of Nkd1, an oscillator gene activated downstream of Wnt signaling (Yan et al.,2001) (Fig. 7S). By contrast, most of the explants (75%; total n=8) derived from Sfrp1^-/-;Sfrp2^-/- embryos retained higher expression levels of Nkd1 (Fig. 7T,U). This observation indicates that the downstream target of Wnt signaling is activated in the absence of Sfrp1 and Sfrp2. In concert, the results derived from the explant cultures suggest that inhibition of Wnt signaling by Sfrp1 and Sfrp2 is an essential component of the somite segmentation process.

In the present paper, we provide evidence indicating that Sfrp1and Sfrp2 are functionally redundant. Additionally, the results demonstrate the requirement of Sfrp1 and Sfrp2 with respect to embryonic AP patterning and somitogenesis. Reduction of the AP axis was associated with mesoderm cell migration, and aberrant somite segmentation was correlated with perturbed Notch oscillator cycles.

It has been reported that Sfrp1 and Sfrp2 exert opposing effects onβ-catenin stability in MCF-7 breast cancer cells: Sfrp1 decreasedβ-catenin stability, whereas Sfrp2 increased β-catenin cellular concentration (Melkonyan et al.,1997). Moreover, Sfrp2 acts as an antagonist of Sfrp1 with respect to Wnt inhibition (Yoshino et al.,2001). The generation of mice that are null for both Sfrp1 and Sfrp2 clearly reveals that these two proteins compensate for one another during embryonic development. Overlapping expression of the genes is observed from the early somite stage to the organogenesis period. However, the expression pattern is not completely coincident at later stages (Leimeister et al., 1998). The fact that Sfrps are secreted antagonists, rather than the differences between expression sites of the genes, could account for the functional redundancy of Sfrp1 and Sfrp2.

Inhibition of Wnt signaling is required for activity of the head organizer(Niehrs, 1999). Sfrp1is expressed in the anterior visceral endoderm (AVE) and the anterior mesendoderm (AME), essential tissues for head formation in the mouse(Shawlot et al., 1999),whereas wider expression of Sfrp2 is detected in the embryonic ectoderm at the primitive streak stage(Mukhopadhyay et al., 2003). Sfrp1 cDNA was isolated in a screen of differentially expressed genes in wild-type and Lim1 (also referred as to Lhx1) mutant AVE/AME at the late streak stage (Shimono and Behringer, 1999; Shimono and Behringer, 2000). Lim1 is necessary for head organizer activity (Shawlot and Behringer,1995; Shawlot et al.,1999). In addition, Sfrp1 expression is reduced in the AVE and AME of Lim1 homozygous mutant embryos (A.S., unpublished). However, no defect was evident in anterior patterning in the Sfrp1^-/-;Sfrp2^-/- embryos. Sfrp5 and Dkk1 are expressed in the AVE, which may compensate for the loss of Sfrp1 and Sfrp2 function in anterior patterning(Mukhopadhyay et al., 2001; Finley et al., 2003).

The results of this study provide genetic evidence for the in vivo function of Sfrp genes in embryonic development. The limb defect is consistent with the phenotype induced by Wnt/β-catenin signal activation(Mukhopaghyay at al., 2001; Barrow et al., 2003; Soshnikova et al., 2003). The defect in posterior axis elongation in Sfrp1^-/-;Sfrp2^-/- embryos at E8.25 appeared to be associated with an abnormality in migration of the paraxial mesoderm cells. In addition, we observed higher activation of the Wnt/β-catenin pathway in the posterior region of Sfrp1^-/-;Sfrp2^-/-embryos. Mesoderm migration is known to be dependent on the function of T, a downstream target of the Wnt/β-catenin pathway(Wilson et al., 1993; Yamaguchi et al., 1999b). Although hyperactivation of T expression was not detected in the double homozygous mutant embryos, it is possible that moderate levels of activation of expression affect mesoderm cell migration.

Interestingly, the AP axis reduction associated with small somites, and the defect in the craniofacial region and limbs of the Sfrp1^-/-;Sfrp2^-/- embryos are similar to the phenotype induced by Wnt5a inactivation(Yamaguchi et al., 1999a). Wnt5a is not capable of association with Sfrp1 in vitro(Xu et al., 1998; Dennis et al., 1999), which suggests that Wnt5a signaling is not directly regulated by Sfrp1 and Sfrp2. Previous reports indicated that Wnt5a, which leads to activation of the Wnt/Ca²⁺ pathway, plays a role in canonical Wnt/β-catenin pathway inhibition (Torres et al.,1996; Topol et al.,2003). Furthermore, the Wnt/Ca²⁺ pathway interacts with the non-canonical PCP pathway(Carreira-Barbosa et al.,2003; Veeman et al.,2003; Westfall et al.,2003). Further analysis is required in order to elucidate the relationship between the roles of Sfrp and Wnt5a. However, these observations suggest that inactivation of Sfrp1 and Sfrp2 causes misregulation of a wide spectrum of Wnt activities in the canonical and non-canonical pathways.

The segmentation clock in the PSM, which is generated by the expression of oscillator genes, is needed for somite segmentation (Aulehla et al., 2004). A number of oscillator genes, such as Lfng, Her1, chick hairy1, chick hairy2, Hes1/Hes2 and Hes7, have been identified as components of Notch signaling. In fact, mutations of Notch,Dll1 and Dll3 (Notch ligands), and Lfng and Hes7 (Notch effectors), all result in aberrant segmentation during somitogenesis(Saga and Takeda, 2001; Bessho et al., 2001). The disruption of somite segmentation exhibited by Sfrp1^-/-;Sfrp2^-/- embryos is associated with altered Notch oscillations, which is supported by abnormal Lfng and Hes7 expression cycles in Sfrp1^-/-;Sfrp2^-/- embryos.

A Wnt3a gradient in the PSM is thought to function in somitogenesis regulation. Expression of Notch-related oscillator genes is disrupted in Wnt3a mutant embryos, which suggests that Wnt acts upstream of Notch signaling (Aulehla et al.,2003). Upregulation of the Wnt/β-catenin pathway was suggested by non-phospho β-catenin staining in the tail bud of Sfrp1^-/-;Sfrp2^-/- embryos at E8.5. In addition,inactivation of Sfrp1 and Sfrp2 perturbed the oscillating expression of Nkd1, an oscillator gene activated downstream of Wnt signaling in the PSM (Ishikawa et al.,2004). Moreover, evidence that LEF1/TCF, a downstream mediator of Wnt signaling, activates expression of the Notch ligand Dll1 revealed the connection between Wnt and Notch signaling(Hofmann et al., 2004; Galceran et al., 2004). These data, and our observations, support the idea that inactivation of Sfrp1 and Sfrp2 affects Wnt signaling activity in the PSM in conjunction with the misregulation of oscillating Notch signaling, which results in abnormal of Lfng and Hes7 expression. Interestingly, dynamic expression of Axin2 in Sfrp1^-/-;Sfrp2^-/- embryos was observed,suggesting the presence of Axin2 oscillation. Rapid degradation of the protein may exert an effect on the maintenance of the oscillatory expression cycles even in the PSM of Sfrp1^-/-;Sfrp2^-/- embryos(Seidensticker and Behrens,2000; Aulehla et al.,2003). Thus, the intracellular oscillation cycle (Axin2expression) does not correlate with the intercellular cycle (Notchoscillators) in Sfrp1^-/-;Sfrp2^-/- embryos,suggesting that Notch signaling is a non-cell autonomous signal that is elevated in Wnt-receiving cells.

In summary, our findings clearly demonstrate the function of Sfrp1and Sfrp2 in AP axis elongation and somitogenesisis. Because Sfrps regulate a wide spectrum of Wnt activities and pathways, these results suggest that coordinated regulation in the canonical and non-canonical Wnt pathways plays a significant role in embryonic patterning.

We are grateful to Dr A. Bradley for the AB1 ES and NSL cell lines; to Dr R. Behringer for PGKneobpAloxB and pNTR-T-LacZ-PGKneobpAlox cassettes; to Drs R. Johnson, P. Gruss, C. Wright, E. Olson, A. Gossler, B. Herrmann, Y. Saga and R. Kageyama for probes; and to Dr K. Nakao, Animal Support Facility at CDB, for chimera production. We also thank Drs R. Behringer and Y. Mishina for helpful comments regarding the manuscript. The 2H3 monoclonal antibody developed by Drs Tom M. Jessell and Jane Dodd was obtained from the Developmental Studies Hybridoma Bank that was developed under the auspices of the NICHD and is maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242, USA. This study was supported by Grants-in-Aid for Scientific Research on Priority Areas (B) (No. 13138203)from the Ministry of Education, Science, Sports and Culture of Japan(A.S.).