Essential pro-Bmp roles of crossveinless 2 in mouse organogenesis

Embryogenesis is a complex process involving a number of cell-cell interactions mediated by extracellular signals. The Bmp family is a class of extracellular signaling proteins that are well conserved across species,including vertebrates and invertebrates(Hogan, 1996). Bmp proteins were first isolated on the basis of their bone-inducing activity in mammalian tissues (therefore, named bone morphogenetic proteins) and play key regulatory roles in skeletal development (reviewed by Wan and Cao, 2005). Interestingly, the Bmp family signals have been shown to play a variety of roles in the control of embryogenesis, including in cell-type specification,maturation, cell growth, apoptosis and dorsoventral axis determination(Hogan, 1996; De Robertis and Sasai, 1996; Massague and Chen, 2000; Hammerschmidt and Mullins,2002).

A characteristic feature of Bmp signals is their function as a morphogen;they generate an activity gradient and evoke multiple-threshold responses in recipient cells (Gurdon and Bourillot,2001). In Xenopus, for example, Bmp4 induces the graded ventralization of mesodermal and ectodermal tissues in a dose-dependent manner(Dale et al., 1992; Jones et al., 1992; Fainsod et al., 1994). In Drosophila, a gradient of the Dpp (fly Bmp4) activity in the ectoderm determines the dorsoventral specification(Ferguson and Anderson, 1992a; Wharton et al., 1993). Therefore, the fine spatial control of Bmp signals is important for tissue formation to occur in the right place.

Over the past decade, several classes of factors that negatively regulate Bmp signals in the extracellular space have been identified. A typical example is a class of secreted antagonist proteins that bind to and inactivate Bmp proteins, such as noggin, chordin (Chd), follistatin, cerberus and gremlin(Smith and Harland, 1992; Lamb et al., 1993; Sasai et al., 1994; Sasai et al., 1995; Hemmati-Brivanlou et al.,1994; Glinka et al.,1997; Hsu et al.,1998).

Extracellular factors regulate Bmp signals not only negatively but also positively. For example, detailed genetic analyses in the fly have shown that Sog (fly Chd) functions both as an antagonist (anti-Bmp, hereafter) and as a potentiator (pro-Bmp) of Bmp (Dpp) signals in a context-dependent manner. The sog^-/- mutant has a strong ventral defect (e.g. reduction of the ventral neurogenic ectoderm) with expanded dorsal non-neural ectoderm(Zusman et al., 1988; Ferguson and Anderson, 1992b; François et al., 1994). In this dorsalization phenotype, Sog is shown to act antagonistically against Dpp (Ferguson and Anderson,1992b). Nevertheless, the sog mutant shows impaired formation of the dorsalmost tissue (amnioserosa), which requires the highest level of Bmp activity (Zusman et al.,1988; Ashe and Levine,1999; Decotto and Ferguson,2001). In this particular context, Sog is thought to act as a pro-Bmp factor.

Chd also seems to play some pro-Bmp roles in vertebrate development,although the mechanistic details are poorly understood. In zebrafish, for example, formation of the ventral tail fin, which is a derivative of the ventral-most tissue, requires the highest level of Bmp signaling(Wagner and Mullins, 2002; Rentzsch et al., 2006). The fish Chd mutant (chordino) has a reduced ventral tail fin(Fisher et al., 1997; Schulte-Merker et al., 1997; Hammerschmidt and Mullins,2002; Rentzsch et al.,2006), suggesting that Chd is necessary to enhance the Bmp signaling in the ventralmost tissue just as is indicated for fly Sog in the amnioserosa.

Structurally, the Chd/Sog protein contains four cysteine-rich domains. At least two of these domains interact physically with Bmp proteins and play an essential role in inhibiting Bmp activity(Larrain et al., 2000). We have previously reported the isolation of a secreted protein, named kielin,which contains 27 repeats of the Chd-type cysteine-rich domains(Matsui et al., 2000). Kielin is expressed specifically in the dorsal axis (the notochord and floor plate)of the Xenopus embryo. Microinjection of kielin mRNA weakly promotes paraxial mesoderm differentiation but does not induce strong dorsalization (such as ectopic neural differentiation, which is typical for the Chd-mediated Bmp inhibition in Xenopus)(Sasai et al., 1995). This raises the possibility that kielin does not simply block Bmp signals, even though it contains numerous cysteine-rich domains.

A mouse database search has identified several kielin/Chd-related factors(see Fig. S1 in the supplementary material). In mammals, Cv2 (Bmper - Mouse Genome Informatics) and kielin-chordin-related protein (Kcp) are the most closely related to kielin among these kielin/Chd-related proteins, and form a subfamily with it (O'Connor et al.,2006). Like kielin, Cv2 and Kcp contain multiple Chd-type cysteine-rich (CR) domains (five and 18 repeats, respectively; see Fig. S1A in the supplementary material). In addition, a von Willebrand factor type D-like(vWD) domain and a trypsin inhibitor-like cysteinerich (TIL) domain are located at the C terminus of each protein.

cv2 was first identified in the fly mutant study as a gene required for the formation of cross-veins in the fly wing(Garcia-Bellido and de Celis,1992). Genetic studies in flies showed that the formation of these veins requires high Bmp signaling activity (involving Dpp and Gbb), and that Cv2 is essential for enhancing the local Bmp signal near the receiving cells(O'Connor et al., 2006). This pro-Bmp role of Cv2 was also demonstrated by the reduction of phosphorylated Mad protein in the fly cv2 mutant(Conley et al., 2000; Ralston and Blair, 2005). In addition, forced expression of cv2 can antagonize the effect of sog overexpression in the wing crossvein formation(Ralston and Blair, 2005).

By contrast, the in vivo role of the vertebrate counterpart of Cv2 remains rather nebulous. Two opposing activities have been proposed for Cv2. One is to attenuate Bmp signals. Microinjection of Cv2 mRNA into the Xenopus embryo results in the formation of a secondary axis similar to the Chd-induced one (Moser et al.,2003; Coles et al.,2004). In vitro, purified recombinant human CV2 (at excessive doses) inhibits the Bmp-dependent osteoblast and chondrocyte differentiation in cultured cells (Binnerts et al.,2004). Moreover, the transfection of 293T cells with a Cv2 plasmid reduces cellular response to Bmp4 protein in a Bmp-responding luciferease reporter assay(Moser et al., 2003). The other theory is that Cv2 enhances Bmp signaling, as does its fly homologue. A report (Kamimura et al., 2004)has shown that transfection of a Cv2 plasmid enhances the cellular response to Bmp4, as measured by Smad phosphorylation. We have also observed that the addition of Cv2 protein (6 nM) enhances Bmp4 (25 nM)-induced cartilage differentiation in cultured E14.5 mouse embryonic fibroblasts (M.I. and Y.S., unpublished). In Xenopus, co-injection of Cv2 and Bmp4 mRNA into animal pole blastomeres synergistically induces the ectopic expression of Xbra (Coles et al., 2004). Thus, these gain-of-function analyses have so far failed to provide a consistent answer to the question `do vertebrate Cv2 acts as an anti-Bmp or a pro-Bmp factor in embryogenesis?'. Although a recent zebrafish Cv2 morphant study(Rentzsch et al., 2006)reported a pro-Bmp role in the dorsoventral axis patterning, the same study also indicated that Cv2 protein may acts as an anti-Bmp factor depending on its proteolytic processing, suggesting a possible bidirectional role for vertebrate Cv2.

In the present study, to elucidate the exact role of Cv2 in vivo, we performed a loss-of-function study by generating the `null' Cv2mutant. Based on loss-of-function evidence and genetic interaction data, we demonstrate that Cv2 plays essential roles as a local enhancer of Bmp signals in mouse organogenesis.

The Cv2 (tau-lacZ), Cv2 (nlacZ) and Kcp mutants (RIKEN Accession Numbers CDB0402K, CDB0451K and CDB0401K,respectively) were established as follows. Cv2 and Kcpgenomic clones were isolated from C57BL/6 genomic phage libraries or BAC libraries. We constructed a targeting vector by inserting lacZ-neocassettes into the Cv2 or Kcp locus, so that the first exon of each gene be replaced with lacZ at the first ATG. To generate the Cv2 (nlacZ) knockout line, we constructed a targeting vector using a combined in vivo/in vitro recombination technique (the EG construction method) described previously (Ikeya et al., 2005).

To generate the targeted ES cells, TT2 ES cells(Yagi et al., 1993a) were transfected with the resultant targeting vector made with pMCDTA(A+T/pau)(Yagi et al., 1993b) and selected with G418. Recombinant ES cells were injected into an eight-cell embryo (inside of the zona pelucida) of CD-1 mice. For the Cv2mutant, we obtained a germline chimera from one recombinant line each for tau-lacZ and for nlacZ. No differences in the null mutant phenotypes were observed between the two mouse lines. The nLacZ line gave better β-galactosidase staining (probably because lacZ was used instead of tau-lacZ), and was used for in vivo expression analysis. For the Kcp mutant, we obtained germline chimeras from two independent recombinants. To remove the pgk-neo cassette from the Kcp mutant genome, pCAG-cre plasmid was injected into the fertilized egg. The correct excision was confirmed by Southern blot analysis.

To detect the tau-lacZ knock-in allele of Cv2, PCR primers were designed for the sequence upstream and downstream of the start codon of Cv2 (Cv2-5′ and Cv2-3′, respectively), and for the sequence of tau (Cv2-T). The lengths of the amplified cDNA fragments are 564 bp for the wild type and 412 bp for the mutant. The primers are as follows: Cv2-5′, 5′-AGTCGCCCGGGATTCCCTCCAGGT-3′;Cv2-3′, 5′-AGATGCTGCCTAGCACCGTGGATTT-3′; Cv2-T,5′-TGTCATCGGGTCCAGTCCCATCTTT-3′. To detect the nLacZknock-in allele of Cv2, a primer for β-galactosidase (mCv2-L,5′-TAACCGTGCATCTGCCAGTTTGAGG-3′) was used. The length of the mutant fragment is 375 bp. To detect the Kcp mutant allele, primers were designed for the sequence upstream and downstream of the start codon of Kcp (Kcp-5′ and Kcp-3′, respectively), and for the sequence of tau sequence (Kcp-T). The lengths of the amplified cDNA fragments are 553 bp for the wild type and 390 bp for the mutant. The primers are as follows: Kcp-5′, 5′-GAGCTTGGAAGACTGCTATGGGTCA-3′;Kcp-3′, 5′-AGCCTCTGCTTCACCGCTACTTAGG-3′; Kcp-T,5′-GGTTTCAGAGCCTGGTTCCTCAGAT-3′. All the lines were backcrossed with C57BL/6 genetic background for two or three generations prior to analysis.

Bmp4 mutant mice (Lawson et al., 1999) were crossed to the Cv2 mutant strain. To detect the Bmp4 mutant allele, primers were designed for the sequence upstream and downstream of the start codon of Bmp4 (Bmp4-5′ and Bmp4-3′, respectively), and for the sequence of sequence ofβ-galactosidase sequence (Bmp4-L). The lengths of the amplified cDNA fragments are 495 bp for the wild type and about 300 bp for the mutant. The primers are as follows: Bmp4-5′,5′-GCAGCTGGTGTGTGTGTGTGTAGGG-3′; Bmp4-3′,5′-GTTCCCTGGCTCTGCTCTTCCTCCT-3′; Bmp4-L,5′-TTCACCCACCGGTACCTTACGCTTC-3′.

For histological analyses, embryos were fixed in 7.6% formamide/distilled water or Bouin's fixative overnight. They were then embedded in paraffin,sectioned at 10 μm and stained with Hematoxylin-Eosin (HE) as described previously (Ikeya et al., 1998).

Skeletons were prepared as described previously(Parr and McMahon, 1995). Briefly, E14.5, E18.5 and P0 mice were eviscerated, skinned, fixed in 80%alcohol and stained with Alcian Blue and Alizarin Red.

For immunohistochemistry, samples were fixed in 4%paraformaldehyde/phosphate-buffered saline for 30 minutes and processed as described previously (Mizuseki et al.,2003). Primary antibody dilutions were as follows: anti-human Ki67 at 1:200 (BD Pharmingen, mouse monoclonal), anti-Zic1 at 1:3000(Su et al., 2006), anti-rat aquaporin 5 (AQP5) at 1:5 (CHEMICON, rabbit polyclonal), anti-Cc10 (T18) at 1:10 (Santa Cruz, goat polyclonal), anti-Hnf3b at 1:100 (DSHB), anti-Isl1 at 1:200 (DSHB), anti-Mash1 at 1:10 (Pharmingen, mouse monoclonal), anti-Msx1/2 at 1:100 (DSHB, 4G1), anti-Pax2 at 1:200 (Babco, rabbit polyclonal), anti-Pax7 at 1:200 (DSHB), anti-Pecam (CD31) at 1:25 (BD Pharmingen, rat monoclonal),anti-human prosurfactant protein C (proSP-C) at 1:200 (Chemicon, rabbit polyclonal), and anti-smooth muscle actin (SMA) at 1:4000 (Sigma, mouse monoclonal, clone 1A4).

Whole-mount immunostaining was performed with an anti-Wt1 antibody (at 1/100 dilution; Santa-Cruz) and anti-E-cadherin antibody (at 1/500 dilution;TAKARA, ECCD2) as described previously(Grieshammer et al., 2005). In situ hybridization was performed as described previously(Mizuseki et al., 2003).

For the statistical analysis of proliferating chondrocytes, we counted Ki67-positive cells among DAPI-positive cells in a 150 μm × 150 μm square area around the notochord in four to six randomly selected sections of each embryo (three samples for each genotype). The glomeruli in developing kidneys were counted in HE-stained sections that represented the maximum area along the longest axis, and the numbers from six to 16 sections for each genotype were scored.

RT-PCR was performed as described previously(Mizuseki et al., 2003). The primers used for RT-PCR were as follows:

aggrecan 1, 5′-CCAAGTTCCAGGGTCACTGT-3′ and 5′-CCAAGTTCCAGGGTCACTGT-3′;

Col2a1, 5′-GCCAAGACCTGAAACTCTGC-3′ and 5′-CTTGCCCCACTTACCAGTGT-3′;

Gapdh, 5′-GACCCCTCATTGACCTCAACTACA-3′ and 5′-GGTCTTACTCCTTGGAGGCCATGT-3′;

Cv2, 5′-ATTACCTGCTGCGTCTTGCT-3′ and 5′-TTCTCTCACGCACTGTGTCC-3′;

Pax1, 5′-CACATTCAGTCAGCAACATCCTG-3′ and 5′-TGTATACTCCCTGCTGGTTGGAA-3′.

MEF cells from each embryo were prepared as described previously(Hogan et al., 1994; Lengner et al., 2004) with some modification. Briefly, E14.5 embryos were dissected from heterozygous mouse intercrosses. The trunk tissues of each eviscerated embryo were separately sheared through an 18-G syringe once in 1 ml of 0.05% trypsin/1 mM EDTA/0.001% DNase I and incubated at 37°C for 10 minutes. The cells were plated on one 10 cm tissue culture dish per embryo and culture in DMEM/10%FCS. MEF cells from Cv2^+/- mice and those from Cv2^-/- mice contained similar percentages of lacZ-positive cells (∼25%). MEF cells (1×10⁵,passage two or three) were plated onto each well of a 24-well cell culture plate (BD Falcon). The recombinant human BMP4 (R&D) was added to the culture 24 hours after plating. RT-PCR was performed 72 hours after plating.

Homologous recombination using a Cv2-targeting vector was performed with mouse TT2 embryonic stem (ES) cells(Yagi et al., 1993a)(Fig. 1A,B). Germ-line transmission was obtained with one line of the revertant cells (line 32; Fig. 1C-E), which was used for further analysis. To rule out artifacts resulting from using a single line, we also generated another Cv2 mutant ES cell line using a different knockout vector with an nLacZ insert (see Fig. S2A in the supplementary material). The mutant mice of this second line showed perinatal and embryonic phenotypes (discussed below) that were indistinguishable from those of the first mutant line.

Heterozygous mice exhibited no particular phenotypes. By contrast,homozygous Cv2^-/- mutants died soon after birth (after breathing several times), and no neonates survived. At and before birth, Cv2^-/- mutants were found in a reasonable Mendelian ratio(see Table S1 in the supplementary material). At postnatal day 0 (P0), the Cv2^-/- mutants showed a phenotype of a short trunk and a short curved tail (Fig. 1F). There was no sign of milk intake in the stomach. At embryonic day 15.5(E15.5), the null mutant mice did not show drastic defects in their external appearance but were generally smaller than the control siblings(Fig. 1G). In addition, a large subcutaneous lucent space was found in the dorsal midline (arrows in Fig. 1G).

Previous studies have reported Cv2 expression in embryonic and adult mouse tissues as well as in ES cells(Coffinier et al., 2002; Moser et al., 2003). By using nLacZ-knocked-in Cv2^+/- mice, we analyzed in detail the Cv2 expression patterns during the organogenetic stages.

At E10.5 (Fig. 2A,B), Cv2 was expressed in the dorsal midline of the CNS (arrowhead),migrating neural crest (nc), head mesenchyme, trigeminal ganglion (t), otic vesicle (o), para-aortic region and mesonephros (mn). At E11.5(Fig. 2C,D), strong Cv2 expression was first detected in the presumptive vertebral body(arrowhead) and arch (arrow). At E14.5 and E18.5(Fig. 2E-H; data not shown), Cv2 expression was detected in a number of developing skeletal structures and internal organs, including the vertebral body (vb) and arch(va), ribs (r), skull and long bones, limb girdle bones, pharyngeal and tracheal cartilages, dorsal root ganglion (drg), lung (l; particularly in the developing alveoli), kidney (condensed nephrogenic mesenchyme, or metanephric mesenchyme), and relatively small areas of the CNS. Expression of Cv2in the developing nephrons of the kidney was also seen at P0(Fig. 2I-K; discussed later).

The short-trunk phenotype (Fig. 1F,G) promoted us to analyze the skeletal formation in Cv2^-/- embryos, and multiple defects were observed in bone and cartilage development at P0 (Fig. 3A′; also see Table S2 in the supplementary material). They included in the formation of the vertebrae(Fig. 3B′-F′;discussed below), ribs (lack of the 13th ribs, Fig. 3G′),pharyngeal/tracheal cartilages (small hyoid, thyroid and cricoid cartilages and a lack of tracheal cartilages, Fig. 3A′, J′), skull (a wider unossified area of the metopic suture and small interparietal and supraoccipital bones, Fig. 3K′ and see Table S2 in the supplementary material; a cavity in the basisphenoid bone, Fig. 3L′; loss of the retrotympanic process of the squamosal bone, Fig. 3M′), scapula (small or with a hole in the middle, Fig. 3N′) and humerus (lack of the deltoid tuberosity, Fig. 3N′) and the pubic bone (smaller body of the pubis and unclosed symphysis, Fig. 3O′). In particular,gross defects were found in the vertebral column. When compared with those of the control mice (Cv2^+/+and Cv2^+/-),the vertebral bodies of Cv2^-/- mice were smaller and showed reduced bone formation throughout the rostrocaudal axis(Fig. 3A′, G′). In addition, the vertebral arches (from the cervical to the sacral regions) were largely missing (Fig. 3B′-F′). Similar vertebral phenotypes were also seen in earlier embryos (E14.5; Fig. 3H′-I′). These vertebral arch defects were not accompanied by the neural tube defect such as an unclosed spinal cord(Fig. 1F,G; see Fig. S4 in the supplementary material).

Histological analysis showed that the cell-dense region corresponding to the dorsal vertebral arch was replaced with mesenchymal tissues in the null mutant at E14.5 (Fig. 4A-D). By contrast, no obvious histological difference in this region was found between the control and mutant mice at E12.5 (data not shown). Consistent with this finding, the expressions of Zic1 (an marker for dorsal sclerotome-derived mesenchyme) (Aruga et al., 1999) in the future vertebral arch region was unaffected in the null mutant at E12.5 (Fig. 4I,J). These findings suggest that the loss of Cv2prevents the further differentiation of the sclerotome-derived precursor cells between E12.5 and E14.5 in this region.

At E14.5, the sizes of the vertebral bodies (which express Cv2; Fig. 2E), but not those of the intervertebral discs (not expressing Cv2), were reduced in the Cv2^-/- mutant (Fig. 3I′, Fig. 4E-H). The early vertebral body marker Pax1 was expressed normally in E12.5 Cv2^-/- mice(Fig. 4K,L), suggesting that the defect is unlikely to be attributed to impaired initiation of the sclerotomal tissue formation in the vertebral body. RT-PCR analyses of trunk tissues also showed that there was no obvious change in the expression of the early sclerotomal markers Pax1 and Sox9 between the control and the null mutant (Fig. 4M). At E14.5, interestingly, an obvious reduction was seen in the amount of cartilage matrix present in the extracellular space (see reduced intercellular spaces; compare Fig. 4E,F). This phenotype could be interpreted as a defect in cartilage cell maturation,as the mutant vertebral body contained a higher percentage of Ki67-positive immature mitotic cells than the normal one(Fig. 4P-R; 38.8% in the mutant and 18.6% in the wild type, E12.5). No significant increase in apopotosis was observed by the TUNEL assay in the Cv2^-/- mutants at E14.5(data not shown).

We next sought to elucidate the mechanism of the Cv2 action as a regulatory signal of skeletal development, focusing on its relation to Bmp signaling. RT-PCR analyses using embryonic trunk tissues (axial and body wall tissues,excluding the internal organs) showed reduced levels of the early cartilage precursor markers aggrecan 1 (Glumoff et al., 1994) and collagen2 α1 (Col2a1)(Cheah et al., 1991) in Cv2^-/- embryos at E14.5 (see Fig. S3 in the supplementary material). By contrast, no significant differences in the transcription levels of Bmp genes (Bmp2, Bmp4 and Bmp7) were observed between the control and Cv2^-/- embryos at E12.5 and E14.5(Fig. 4M; see Fig. S3 in the supplementary material), suggesting that the Cv2^-/-phenotypes could not be explained simply by a reduction in the general expression levels of Bmp genes.

As Cv2^-/- mice showed a significant decrease in the postmitotic mature cartilage cells (Fig. 4P-R), we next studied the functional relationship between Bmp signals and Cv2 in primary culture experiments using MEF cells. We prepared MEF cells from the control and mutant mice, and examined the effects of exogenous Bmp4 on early cartilage markers. Although the expression of aggrecan 1 was induced in a dose-dependent manner by Bmp4 (after 48 hours treatment), Cv2^-/- MEF cells showed a reduced aggrecan 1 induction when compared with control MEF cells (Fig. 4S, rows 1,2). A similar reduction (albeit moderate) was observed for another early cartilage marker Col2a1(Fig. 4S, rows 3,4). In contrast to the induction by Bmp4 proteins, aggrecan 1 induction by a transfected plasmid bearing constitutively-active ALK2(CA-ALK2) was observed at similar levels in control and Cv2^-/- MEF cells (data not shown), suggesting that Cv2 function is needed for Bmp4-induced aggrecan 1 expression upstream of (or at)the Bmp-receptor level.

During skeletal development, Bmp ligands are expressed and play essential roles at multiple phases in cartilage and bone differentiation (reviewed by Canalis et al., 2003; Wan and Cao, 2005). For example, the retrovirus-mediated misexpression of noggin in the axial skeleton inhibits the cartilage formation in chick, whereas the expression of Pax1 is not affected (Murtaugh et al., 1999). Moreover, the Bmp7-null mutant, like the Cv2^-/- mutant, shows the defects of the vertebral arches(Jena et al., 1997). In this respect, the Cv2^-/- phenotype of cartilage and bone defects (Figs 3, 4) seems consistent with the idea that Cv2 acts in the same direction as Bmp signals.

To obtain direct genetic evidence for either a pro- or anti-Bmp role, we next examined the effect of Bmp attenuation on the Cv2^-/-mutant phenotype. In particular, we tested whether the vertebral defect, which is a most evident skeletal phenotype, was enhanced or rescued by reducing Bmp4 gene dose (Fig. 5). Unlike Cv2^-/- mutants, Cv2^+/-;Bmp4^+/- mice were externally healthy and fertile, and, consistently, did not show a vertebral phenotype[Fig. 5A,D; no dorsal vertebral defects were observed, unlike those reported for Bmp4^lacZ/+ recently in a different genetic background(Goldman et al., 2006)]. By contrast, the deletion of one copy of Bmp4 strongly and cooperatively enhanced the vertebral defect of Cv2^-/- mutants (compare Fig. 5C,F with 5B,E), causing(1) reduction of the vertebral body in size and of its ossification (arrow)and (2) suppression of vertebral arch development (arrowhead). These findings demonstrate that Bmp4 has strong genetic enhancement with Cv2, indicating that Cv2 and Bmp4 function cooperatively, rather than antagonistically.

In the external phenotype, the reduction of the Bmp4 gene dose synergistically increased the frequency of the microphthalamic phenotype. A strong defect was found in 100% of Cv2^-/-; Bmp4^+/- eyes (n=6), while micropthalmia was seen only in 18% (n=22) and 1.7% (n=172) of Cv2^+/-; Bmp4^+/- and Cv2^-/-; Bmp4^+/+ eyes, respectively,suggesting that Cv2 and Bmp4 work together in the same direction with respect to eye development.

Taken together, these findings show that Cv2 plays a pro-Bmp role at least in vertebral and eye organogenesis.

To further elucidate the role of Cv2 in Bmp-related organogenesis,we next examined the phenotype of Cv2^-/- in kidney development, where Bmp signals also play essential roles(Godin et al., 1999; Simic and Vukicevic, 2005). During nephrogenesis, Bmp4 is expressed in the stromal mesenchyme,S-shaped bodies and paraureteric mesenchyme, whereas Bmp7 is expressed in the nephrogenic mesenchyme and ureteric bud/collecting duct(Godin et al., 1998; Miyazaki et al., 2000). Small kidneys and decreased numbers of nephrons have been reported for the Bmp4^+/- and Bmp7^-/- mutants(Dudley et al., 1995; Luo et al., 1995; Miyazaki et al., 2000). In Cv2^-/- mice, the size of the kidney was significantly reduced at birth (Fig. 6B). In the section cut along the longest axis, the kidney of the null mutant was 27.0±12.4% smaller (n=6; Fig. 6D) than that of the wild-type mouse (n=6; Fig. 6C). The number of glomeruli was also reduced in the null mutant(by 56.8±9.4%, n=6; also see Fig. 7A, lanes 1, 2), whereas no obvious morphological defect was found in the renal corpuscle(Fig. 6F).

A similar hypomorphic phenotype was also observed for the Cv2^-/- kidney at E14.5(Fig. 6G,H) and E18.5 (not shown). During nephrogenesis, strong Cv2 expression was found in the condensed nephrogenic mesenchyme (Fig. 2H-K) and in the comma- and S-shaped bodies (derivatives of the nephrogenic mesenchyme; Fig. 2H,K). The nephrogenic mesenchyme is known to have a major contribution to the generation of the nephron. At E14.5, the number of Pax2-positive masses of condensed nephrogenic mesenchymes was significantly reduced in the Cv2^-/- kidney(Fig. 6I-K). Immunohistochemical analysis revealed that the general histological arrangement of the Wt1⁺ nephrogenic mesenchyme (and S-shaped bodies) and E-cadherin⁺ collecting ducts was largely unaffected(Fig. 6L,M), although both the number of nephrons and the size of the kidney were reduced.

Taken together, these observations suggest that Cv2 has an essential function for the generation of the proper number of nephrons, presumably by acting on the nephrogenic mesenchymes, which themselves strongly expresses Cv2. As the relative reduction in nephron numbers is more evident at P0 than at E14.5 (Fig. 6K, Fig. 7A), the continuous expression of Cv2 (Fig. 2H-K) may be also necessary during a substantial period of kidney organogenesis after E14.5.

The kidney defect phenotype of the Cv2 mutant appears consistent with the idea that Cv2 plays a positive regulatory role in Bmp signaling. However, previous gene targeting studies have demonstrated that the blockade of Bmp signals (e.g., the lack of Bmp7) can cause an even more drastic reduction in kidney development(Dudley et al., 1995; Luo et al., 1995). To test the possibility that some other related molecules act in a partially redundant manner, we next examined the compound phenotypes of Cv2 and Kcp mutants. Like Cv2, Kcp is a Kielin-related molecule that contains multiple Chd-type cysteine-rich repeats, a vWD domain and a TIL domain (see Fig. S1 in the supplementary material), and physically binds to Bmp proteins(Lin et al., 2005). During embryogenesis, Kcp is expressed in the nephric duct and the mesonephric tubule (E9-10) and in the presumptive metanephric tubules (E16)(data not shown) (Lin et al.,2005). Luciferase assays in 3T3 cells have suggested a pro-Bmp activity for Kcp (Lin et al.,2005).

We generated Kcp^-/- mice by gene disruption, and found that the Kcp^-/- mice were viable and fertile. They exhibited no gross structural defects, including in kidney formation(Fig. 7B and data not shown),consistent with a recent report (Lin et al., 2005).

Interestingly, a substantial enhancement in the kidney defect was seen when the Cv2 and Kcp mutants were crossed. On the Cv2^+/+ or Cv2^+/- background, the loss of both Kcp alleles exhibited no obvious effects on kidney formation(Fig. 7A, lane 3 and Fig. 7B). On the Cv2^-/- background, however, the additional deletion of one or both alleles of Kcp caused a further reduction in kidney size(Fig. 7C,D,F,G) and the glomerulus number (Fig. 7A,lanes 4, 5). Furthermore, the kidney of the Cv2^-/-;Kcp^+/- or Cv2^-/-;Kcp^-/- mutant exhibited strong disorganization of the cortex-medullar arrangement with no clear border in between (Fig. 7F,G; a minor disorganization was also seen in the Cv2^-/- mutant alone; Fig. 6D). These findings of genetic interactions indicate that both Cv2 and Kcp function in the same direction to promote kidney development, and that these Cv2-related proteins play a major regulatory role for the morphogenetic signals of kidney development.

By contrast, the skeletal phenotypes of the Cv2^-/-mutant were not enhanced by the additional deletion of both Kcpalleles (data not shown). It is consistent with the fact that strong Kcp expression is not detected in the skeletal tissues (M.I. and Y.S., unpublished).

This study has demonstrated that Cv2 is essential for multiple organogenetic processes that involve Bmp signals. The null mutant phenotypes of Cv2 are in accordance with the idea of pro-Bmp functions of Cv2 in mammalian organogenesis such as vertebral and eye development.

A number of previous reports have shown that Bmp signals play positive regulatory roles in cartilage and bone differentiation. The Cv2^-/- mice exhibit major trunk defects and minor head deformities in their cartilages and bones(Fig. 3; see Table S2 in the supplementary material), which express Cv2(Fig. 2 and data not shown). In the trunk, severe defects are seen in both the axial structures (e.g. the vertebral body and arch) and the non-axial structures (e.g. the 13th rib,pharyngeal and tracheal cartilages, scapula, deltoid tuberosity of the humerus and unclosed symphysis) (Fig. 3G′,J′,N′,O′; see Table S2 in the supplementary material). In addition, minor malformations are found in the head, including the presence of a cavity in the basisphenoid bone, small interparietal and supraoccipital bones, and an enlargement of the metopic suture (Fig. 3K′,L′; see Table S2 in the supplementary material).

Importantly, many aspects of these Cv2^-/- phenotypes(including minor defects) coincide with those found in other mutant mice with attenuated Bmp signaling. For example, like the Cv2^-/-mouse, the Bmp7 mutant has a small basisphenoid with a cavity(Luo et al., 1995; Jena et al., 1997) and a partial loss of the vertebral arches (Jena et al., 1997). The Bmpr2 mutant mouse (hypomorphic allele) lacks the 13th ribs and has malformed interparietal bones(Delot et al., 2003), as is seen with the Cv2^-/- mouse. The deltoid tuberosity is also lost in Bmp7^-/-;Bmpr1b^-/- mutants(Yi et al., 2000). The enlarged metopic suture is found in Bmp1^-/- mice(Suzuki et al., 1996).

These phenotypes are thus consistent with a pro-Bmp role of Cv2 in skeletal development. Moreover, our findings of genetic enhancement of Cv2 and Bmp4 mutants strongly support this idea. To further exclude the possibility of having any anti-Bmp function (particularly as a minor contribution), it may be useful in future studies to examine the Cv2mutant phenotypes in detail on backgrounds mutant for other Bmp pathway genes. In this respect, one complex problem is how to explain the vertebral defect in the null Chd mutant (Bachiller et al., 2003), which is similar to the defective formation found in Cv2^-/- and Bmp7^-/- mutants. In addition, the null mutant of Tsg, which encodes a Chd co-factor, also exhibits a similar vertebral defect(Nosaka et al., 2003; Petryk et al., 2004; Zakin and De Robertis, 2004). As Tsg has been shown to act as a pro-Bmp factor at least in some developmental contexts (Zakin and De Robertis, 2004; Zakin et al.,2005), a possible pro-Bmp role of Chd in mouse vertebral development may be an intriguing topic that should be tested genetically in the future.

Parallels between the Cv2 mutant and the Bmp-related mutants are also found in the kidney phenotypes. Renal hypoplasia with decreased nephrons is observed in Bmp4^+/- and Bmp7^-/-mice, respectively (Dudley et al.,1995; Luo et al.,1995; Miyazaki et al.,2000), as is seen in the Cv2 mutant (Figs 6, 7). The hypoplastic phenotype of the kidney in the Cv2 mutant is further enhanced by combining it with a mutation of Kcp (Fig. 7A), which encodes a secreted protein belonging to the Cv2/kielin subfamily. As a recent report has indicated that Kcp acts as a pro-Bmp factor at least in vitro (Lin et al.,2005), these findings further support the idea that Cv2 acts as a positive regulator of Bmp signals. However, the conclusion on the pro-Bmp role of Cv2 in nephrogenesis should await further genetic crossing studies with mutants of Bmp-related genes in the future, as we have so far failed to observe a more drastic renal phenotype in Cv2^-/-;Bmp4^+/- mice beyond the simple additive effect (our preliminary observations).

Are there differential roles for Cv2 and Kcp in renal development? In fact,the kidney develops normally in the Kcp-null mutant, even in the Cv2^+/- background (Fig. 7B,E; M.I. and Y.S., unpublished)(Lin et al., 2005). Interestingly, the recent report showed that Kcp^-/- mice are more susceptible to developing renal fibrosis and pathological tubular lesions after injury (Lin et al.,2005). This raises the interesting possibility that Cv2 plays a predominant role in the organogenesis of the embryonic kidney while Kcp is an essential regulator of renal tissue maintenance/regeneration after birth. Detailed analyses will be required in future studies to elucidate the exact relationship between these two Cv2/kielin-related factors.

In addition to renal hypoplasia, the Cv2-null mutant has the impaired maturation and expansion of the alveoli (see Fig. S4 in the supplementary material; rather than differentiation of each cellular components), which are reminiscent of a phenotype in the transgenic mice with a dominant-negative Bmpr1b overexpressed in the lung(Weaver et al., 1999; Weaver et al., 2003). Therefore, a possible pro-Bmp role of Cv2 in lung development should be also an intriguing topic for future investigation.

This null mutant study has also indicated that Cv2 is not required for several other aspects of Bmp-dependent developmental processes. For example,the lack of Cv2 causes strong defects of cartilage and bone development, but not in all skeletal parts. Even in vertebral development, which is clearly defective, Cv2 is not needed for the early Bmp-dependent events of cartilage formation, such as the induction of early precursor markers (e.g. Pax1, Zic1, Sox9).

Moreover, Cv2 seems dispensable for developmental events during the early gestation period such as gastrulation and early neural patterning. At E12.5, the Cv2^-/- mice exhibit no remarkable gross defects and are found in a Mendelian ratio (see Table S1 in the supplementary material; M.I. and Y.S., unpublished), even though Cv2 is expressed in the rostral mesoderm and the caudal primitive streak during gastrulation(Coffinier et al., 2002).

The dorsoventral patterning of the neural tube is another example that is dependent on Bmp signals [emanating from the roof plate and overlying ectoderm(Lee and Jessell, 1999)] and is not affected in Cv2^-/- mutants. The dorsoventral markers of the neural tube (e.g. Msx1/2, Pax7, Isl1, Mash1 and Hnf3b) are normally expressed in Cv2^-/- mice at E10.5 (see Fig. S5 in the supplementary material). In addition, the interdigital shaping of the hand appears normal in Cv2^-/- mice(Fig. 1E and data not shown),although this region expresses both Cv2 and Bmp2/4(Kamimura et al., 2004). These findings indicate that the requirement for Cv2 is stage- and tissue-dependent in mouse embryogenesis.

The mechanism by which Cv2 enhances Bmp signals remains to be clarified in the future. Several possibilities can be considered for the mode of the Cv2 action at the molecular level. These include (1) enhancement of the binding between Bmp proteins and their receptors by Cv2 proteins, (2) competition with Bmp-binding inhibitors such as chordin and noggin for Bmp proteins, (3)control of the matrix-association/sequestration of Bmp, and (4) protection of the Bmp proteins from degradation (proteases).

In favor of the first model, a recent study suggested that the Cv2/kielin-related protein Kcp enhances the binding of the Bmp protein and its receptor by forming a tertiary complex(Lin et al., 2005). However,an extensive biochemical study using Biacore analysis did not seem to support the mechanism of the enhanced receptor binding of Bmps by Cv2 proteins(Rentzsch et al., 2006). Instead, the study demonstrated that Cv2 and Chd proteins compete each other for Bmp proteins as their binding partners. In fact, as epistatic analysis in zebrafish dorsoventral patterning has indicated that the Cv2 morphant phenotypes consist of both Chd-dependent and Chd-independent components, the Bmp signaling enhancement by Cv2 may involve more than one mechanism.

Previous in vitro studies have demonstrated that vertebrate Cv2 proteins bind to Bmp2, Bmp4, Bmp6 and Bmp7 with a high affinity, but not to Gdf5(Moser et al., 2003; Coles et al., 2004; Rentzsch et al., 2006). Although the present study implies strong functional interactions between Cv2 and Bmp4, the Cv2-binding specificity among the Bmp/Tgf β superfamily ligands should be an important topic to be analyzed in the in vivo context.

We are grateful to M. Hibi, H. Enomoto and H. Inomata for invaluable comments, to K. Miyazono, H. Beppu, T. Kitamura, and T. Nosaka for discussion on Bmp-related mutants, to F. Osakada for help in the statistical analysis,and to K. Furushima and T. Uesaka for technical advice. MI is thankful to Ayumi Ikeya for constant encouragement and support during this study. This work was supported by grants-in-aid (YS) from the Ministry of Education,Culture, Sports, Science and Technology, the Kobe Cluster Project and the Leading Project.