Noggin is a secreted Bone Morphogenetic Protein (BMP) antagonist, expressed in a number of tissues throughout embryonic development. Its protein sequence is strongly conserved across species, and its homologues can be found in all chordates investigated so far, from C. intestinalis to H. sapiens (http://www.ensembl.org). It was initially identified as a dorsalizing agent in X. laevis (Smith et al., 1993). Subsequent biochemical analysis of NOGGIN revealed that it could bind in picomolar range to BMP2 and 4, to a lesser extent to BMP7 (Zimmerman et al., 1996; Chang and Hemmati-Brivanlou, 1999) and to other BMP-related proteins such as GDFs (Merino et al., 1999). It did not, however, interact with other members of the Transforming Growth Factor family of ligands such as TGFβ or activin (Zimmerman et al., 1996).
BMPs were initially identified as proteins inducing ectopic, endochondral bone formation (Urist, 1965; Sampath and Reddi, 1981; Wozney et al., 1988), and their proposed function in this process points to their involvement in the regulation of different stages of chondrogenic differentiation (Ju et al., 2000). Additional work revealed their multiple functions in development (Hogan, 1996). They bind to Type II and Type I tetrameric receptor complexes and activate SMAD-dependent or -independent signaling cascades (reviewed by Nohe et al., 2004). Elucidation of the NOGGIN crystal structure in a complex with BMP7 revealed the molecular basis of the antagonistic action of NOGGIN. Specifically, the interaction is based on blocking of the epitopes on the ligand responsible for binding to both Type I and Type II receptors (Groppe et al., 2002).
The proposed biological function of Noggin during early developmental stages of X. laevis embryos is to antagonize the action of BMP ligands in Spemann organizer during mesoderm patterning and neural induction (De Robertis and Kuroda, 2004). Targeted inactivation of the Noggin gene in mice led to severe defects in somitogenesis as well as multiple skeletal defects (Brunet et al., 1998; McMahon et al., 1998). The defects in somitogenesis indicated that Noggin played a greater role in patterning of the neural tube and somite compartment than in neural induction (McMahon et al., 1998). During later stages of embryonic development, Noggin is required for hair follicle formation (Botchkarev et al., 1999).
While complete Noggin inactivation in mice led to multiple and severe phenotypes, no haploinsufficiency for Noggin was reported (Brunet et al., 1998; McMahon et al., 1998; Botchkarev et al., 1999; Groppe et al., 2002). In contrast, in humans, NOGGIN haploinsufficiency leads to a number of defects, such as proximal symphalangism, multiple synostosis syndrome, tarsal/carpal coalition syndrome, or autosomal dominant stapes ankylosis (Gong et al., 1999; Dixon et al., 2001; Brown et al., 2002; Debeer et al., 2005; Online Mendelian Inheritance in Men at http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM).
The targeted inactivation of Noggin was carried out in 129X1/SvJ (formerly known as 129/SvJ- http://www.informatics.jax.org/mgihome/nomen/strain_129.shtml), or mixed 129X1/SvJ/C57BL/6 mouse genetic backgrounds (McMahon et al., 1998). To study the function of Noggin in more detail during skeletal development, we decided to breed the Noggin heterozygous mice in three other genetic backgrounds: two inbred, Dba/1 and C57BL/6, and one outbred, CD1. We demonstrated that the consequences of Noggin inactivation were dependent on the genetic background and were sufficiently mild in the CD1 strain, thus allowing for the analysis of phenotypes at later developmental stages. We also discovered an effect on myogenic differentiation in developing limbs. Finally, taking into consideration the apparent discrepancy between the phenotypic manifestations of haploinsufficiency for Noggin in mice and men, we set out to reexamine mice heterozygous for Noggin. We were able to show that, contrary to previously published observations, mice display Noggin haploinsufficiency akin to the human phenotype.
Mice were backcrossed into the selected genetic backgrounds for at least 14 generations and analyzed throughout the different stages of development in all three backgrounds. We analyzed a minimum of six embryos of each genotype obtained from three or more litters for each time point tested. Since the CD1 and Dba/1 mice displayed identical skeletal and muscular developmental defects (unpublished results), the remaining experiments described were carried out in the CD1 background, unless indicated otherwise.
Phenotypes of Noggin Null Mice in Selected Mouse Strains
At 10.5 dpc, the Noggin null mouse embryos were readily identifiable with craniofacial, somatic, and apparent vascular defects (Fig. 1A). The defects were similar to those already described (McMahon et al., 1998). An interesting exception was the limb bud phenotype, which was less pronounced in the CD1 background (Fig. 1A, inset) as compared to published observations (McMahon et al., 1998). Specifically, the observed size of the limb buds was similar between Noggin wild type and null mice in the CD1 mouse strain, whereas in 129X1/SvJ/C57BL/6, the hind limb buds were reduced in size (see Fig. 8 in McMahon et al., 1998).
At 12.5 dpc, the most striking consequence of Noggin inactivation was the degeneration of the tail structure, frequently accompanied by a large hematoma (Fig. 1B). No tail remnants were recovered at later developmental stages.
At 14.5 dpc, dorsal edemas as well as hematomas in other parts of the body were detected (Fig. 1C, dotted arrows). Two distinct head phenotypes were observed (also reported by McMahon et al., 1998), including exencephaly (Fig. 1C, white arrows). The frequency of the exencephalic phenotype was lower in the DBA/1 background (25%) than in CD1 background (60%).
Inactivation of Noggin in the C57BL/6 mouse strain led to midgestation embryonic lethality, and at 14.5 dpc, the embryos were completely necrotic (Fig. 1C′). Since we were interested in the midgestation and perinatal limb phenotypes, analysis of the Noggin null mutation in this background was not pursued further.
Both CD1 and Dba/1 mouse strains could carry the Noggin null embryos to term. In both cases, however, the embryos died perinatally (Fig. 1D and D′, respectively). The lungs of newborn embryos could, in some cases, float on water, indicating that the newborns were able to take at least one breath. The newborn Noggin null embryos were shorter along the rostral-caudal axis than the wild type littermates, had stubby limbs, and lacked tails.
Analysis of the consequences of Noggin inactivation on skeletal development revealed novel features. The effect on bone mineralization during development was different depending on bone type (Fig. 2A and A′). In some cases, mineralization was apparently accelerated, as in the cervical vertebrae (Fig. 2A and A′, yellow ellipse), while in other cases, the absence of Noggin had an opposite effect or no effect on ossification. For instance, the ossification of the mandible was comparable between wild type and knockout embryos, but zygomatic and parietal bones were not ossified in the knockout animals at 14.5 dpc (Fig. 2A and A′). In contrast, and as previously reported (Brunet et al., 1998), the bones of the developing forelimbs displayed extensive chondrogenesis followed by ossification spanning the stylopod and zeugopod, which resulted in elbow fusions at 14.5 dpc. The ossification of the hind limbs in Noggin null mice was, however, delayed (Fig. 2A and A′). Upon birth, the extent of the skeletal ossification was comparable between CD1 and DBA/1 mouse strains (Fig. 2B and B′ and data not shown).
Noggin null mice did not produce detectable Noggin mRNA (Fig. 2C), while Gdf5 mRNA level was reduced 5-fold as measured by quantitative PCR (Fig. 2D). Analysis of the Gdf5 mRNA expression pattern using DISH (Tylzanowski et al., 2003) revealed that at 12.5 dpc there was no detectable difference between wild type and knockout embryos (Fig. 2E), whereas at 13.5 dpc., this difference was readily visible (Fig. 2F). One of the striking features of Gdf5 mRNA expression at 13.5 dpc was its prominent expression in the carpal region (Fig. 2F, arrowhead). The expression of Indian hedgehog (Ihh) was maintained in the central part of the monophalangeal digit in the Noggin null mice (Fig. 2G).
Muscle and Skin Development
Next, we investigated the myogenic differentiation in the developing limbs of Noggin null mice. One of the striking features observed was the severe reduction of muscle tissue in newborn mouse embryos (Fig. 3C–E′). The onset of myogenesis in the developing limb was not affected as demonstrated by normal expression of a marker gene for the myogenic commitment, Pax3 (McMahon et al., 1998; and our unpublished data). Since the onset of myogenic differentiation was unaffected, but terminal muscle differentiation was impaired in the absence of NOGGIN protein, we have extended our investigation to determine at which point muscle differentiation was affected.
To address this issue, we determined the expression patterns of selected proteins associated with muscle development. Markers of the intermediate stages of muscle development, such as MyoD (Fig. 3A,A′) or myogenin (data not shown), were detected at the same anatomical locations in wild type and knockout embryos. The expression of myosin heavy chain, the marker for terminal muscle differentiation associated with the formation of myocytes and myotubes (Andres and Walsh, 1996), was, however, different in Noggin null mice. The expression of this marker at 14.5 dpc was detected in both the wild type and Noggin null mice. In the wild type mice, the staining had the typical elongated appearance associated with formation of multinucleated myotubes. In contrast, in the Noggin null mice, the staining had a more rounded appearance suggestive of cells that failed to fuse into myotubes at this stage of development (compare Fig. 3B and B′).
At birth, the myofibers of the knockout embryos formed loose structures, unlike the wild type muscles that formed densely packed fibrils with visible striations. The nuclei of the cells in the muscle fibers of Noggin null mice also failed to migrate to the cell wall (Fig. 3D and D′).
One of the ways to investigate thickness of muscle fibers is to analyze their birefringence on H&E stained sections using polarized light. The thick fibers tend to shift the light towards the more reddish color, while thin fibers maintain the white coloring (Chow et al., 2001). The muscle fibers of Noggin null mice were apparently thinner and somewhat disorganized as indicated by their appearance in polarized light (Fig. 3E and E′).
Skin of the knockout animals, as reported for the 129X1/SvJ strain (Botchkarev et al., 1999), was affected in the same manner in CD1 (Fig. 3F–G′) and Dba/1 genetic backgrounds (not shown). At 14.5 dpc, the expression of β-catenin was prominent in the wild type mice, but in the Noggin null mice, it was not detected by immunofluoresence (Fig. 3F,F′). At 18.5 dpc, unlike in 129X1/SvJ mice, rudimentary hair follicles did form (Fig. 3G,G′).
Several previous publications reported a lack of Noggin haploinsuficiency phenotype in mice (Brunet et al., 1998; McMahon et al., 1998; Botchkarev et al., 1999; Groppe et al., 2002). Since we generated Noggin heterozygous mice in different genetic backgrounds, we reexamined the Noggin heterozygous animals.
Quantitative PCR analysis revealed that the heterozygous mice produced Noggin mRNA at levels slightly above 50% (Fig. 2C). Examination of the developing limbs of newborn Noggin heterozygous mice revealed carpal and tarsal fusions in 100% of animals in all three backgrounds investigated: CD1, Dba/1, and C57BL/6 (Fig. 4A–F”; and data not shown). These fusions formed two morphological classes, independent of the strain used (data not shown). In one, carpals 2, c, and 3 were fused, while in the second, only carpals c and 3 were fused (Fig. 4B–D).
Next, we investigated the developmental onset of these fusions taking into consideration the possibility that they could be primary or secondary in nature. To accomplish this, we analyzed the expression patterns of Gdf5 mRNA, a molecular marker associated with joint formation, and collagen type II mRNA, a molecular marker for developing cartilage. At 12.5 dpc, the expression patterns of Gdf5 and collagen type II mRNAs were the same in wild type and heterozygous animals (data not shown). At 14.5 dpc, the fusions were detected by the analysis of the expression pattern of Gdf5 mRNA. As shown in Figure 4B′–D′, the expression of Gdf5 mRNA was not detectable in parts of the developing carpal region (compare arrows in Fig. 4B′ and 4C′–D′).
The carpal/tarsal fusions remained in the limbs upon birth and throughout the life of the animal, and no additional joint fusions were detected at later stages of development in all three genetic backgrounds: CD1 (Fig. 4E–F″) as well as C57BL/6 and Dba/1 (data not shown).
In addition, the Noggin haploinsufficiency in the CD1 background exhibited kyphosis, detectable in 30% of the animals (Fig. 4G–I′). The ribcage of the embryos were reduced in size due to malformed ribs (Fig. 4H,H′ and I,I′). Also, the embryos had tails that were kinked at the lumbar bone (arrows in Fig. 4G and G′).
To better understand the consequences of changing balances in BMP signaling, we reinvestigated some aspects of the Noggin null phenotype in selected mouse strains, as well as reexamined the Noggin haploinsufficiency in those strains.
The inactivation of Noggin in C57BL/6 mouse strains led to embryonic lethality prior to day 14.5 dpc. This was in contrast to inbred Dba/1 or outbred CD1 strains, in which the embryos died perinatally. The C57BL/6 strain of mice is known to be an average breeder, with a poor quality of skeletal tissues and nearly osteoporotic bones (Beamer et al., 1996, 2002; http://aretha.jax.org/pub-cgi/phenome/mpdcgi?rtn=docs/home). Previous reports have also demonstrated that C57BL/6 is more sensitive to gene inactivation than other strains (Kallapur et al., 1999; Fleming et al., 2001; Errijgers and Kooy, 2004; reviewed in Sanford et al., 2001). Our current work demonstrates that this strain is also more sensitive to the absence of Noggin. A recent report on the Noggin phenotype in C57BL/6 mouse strain (Wijgerde et al., 2005) showed that these embryos can be recovered as late as 16.5 dpc. We do not have an explanation for that discrepancy at this time, except for a different number of inbreeding steps (14 in our case, 7 in Wijgerde et al., 2005) possibly leading to a partial compensation of the Noggin deficiency by modifier genes still remaining from the 129X1/SvJ background. Interestingly, despite the fact that the CD1 mouse strain is an outbred one, the observed phenotype was remarkably stable, with a similar range of phenotypes as in the inbred Dba/1 mice. The significant phenotypic difference of the Noggin null mutation between C57BL/6 and CD1 or Dba/1 strains may be suitable for a genetic modifier screen aimed at identifying the genes responsible for these differences, and a recently described method would simplify the traditional screening protocol (Beier and Herron, 2004). Identification of such genes would certainly add to our understanding of Noggin biology during the midgestation development of mice.
Skeletal Development in the Absence of Noggin
The skeletal phenotype offers an interesting insight into BMP-regulated ossification during development. The absence of Noggin is expected to lead to oversignaling by BMPs, consequently allowing at least some BMPs to signal in the areas normally “protected” by NOGGIN. While other BMP antagonists, such as chordin, Xnr3, follistatin, or Cerberus, can compensate for the absence of Noggin, this compensation is only partial (Niehrs, 2005). Additionally, NOGGIN has different affinities to different BMPs, and, therefore, its absence is likely to have a complex effect on BMP-dependent tissue differentiation (Zimmerman et al., 1996; Chang and Hemmati-Brivanlou, 1999). This complexity is apparent when analyzing the skeletal development of Noggin null mice. For instance, in the developing skull, different bones responded very differently to derepression of BMP signaling. The ossification of parietal and occipital bones was delayed, that of the mandible was the same, while the ossification of zygomatic bone was accelerated. In other places in the skeleton, the situation was similar with the forelimbs and cervical vertebrae having accelerated ossification and the hind limbs and ribs delayed. Therefore, in spite of the expected excessive BMP signaling in Noggin null mice, the ossification of some of the bones was delayed or even inhibited, raising the possibility that BMPs can serve not only as positive but also as negative regulators of bone formation (endochondral or membranous), depending on the location and/or embryonic origin of the bones. Similar mechanisms may be responsible for the observed Noggin haploinsufficiency and are collectively discussed in the last section of the Discussion.
We have also reexamined joint development in Noggin null mice. Specifically, we were interested in distinguishing whether the absence of most joints in developing limbs of Noggin null mice was a primary or secondary defect. Gdf5 mRNA expression has been associated with the onset of joint development (Storm et al., 1994). Therefore, in the case of a primary defect, one would expect the absence of Gdf5 mRNA expression in the presumptive joint interzones. Should the defect be secondary, caused by initially formed joints subsequently fused by overgrowing cartilage anlagen, one would be able to detect the expression of Gdf5 mRNA in these joints. Indeed, the expression of Gdf5 mRNA was dramatically affected in the Noggin knockout mice. The overall transcription levels were suppressed 5-fold (our work) and the location of Gdf5 mRNA within the developing limb was changed (our work and Brunet et al., 1998). While the expression pattern of Gdf5 mRNA in the Noggin null mice was the same as reported (Brunet et al., 1998), we have additionally detected a strong carpal expression of Gdf5 mRNA at 13.5 dpc. Specifically, the newborn Noggin null mice, while missing most of the joints in the limb, did maintain a “T”-shaped carpal joint at the site where Gdf5 mRNA has been detected. This observation suggests that the areas of mesoderm able to sustain Gdf5 mRNA expression do form joints. Should the absence of joints be caused by overgrowth of properly spaced cartilaginous elements, the carpal joint would be missing as well. The absence of joints in the digits of Noggin null mice is thus likely to be a primary defect and not a consequence of overgrowing and fusing cartilage anlagens. This is also supported by the single site of Ihh expression in the central part of the phalanges and by the absence of Gdf5 mRNA anywhere in the phalanx. We cannot exclude, however, the possibility of a temporary Gdf5 expression that could be very quickly extinguished by the BMP oversignaling.
Muscle and Skin Development in Noggin Null Mice
In the developing mouse limb, the muscles are derived from hypaxial dermomyotome. Myogenic differentiation of mesenchymal cells progresses through well-characterized stages and is associated with expression of various molecular markers. These include Pax3 (early stages), MyoD or myogenin (later stages), or myosin heavy chain for the final stages of differentiation and myotube formation (reviewed by Francis-West et al., 2003).
It has been reported (McMahon et al., 1998) that expression of the early muscle marker Pax3 is not affected in the Noggin null mice. Interestingly, it appears that the entire myogenic program, as defined by the expression of established marker genes, is not affected in those mice. While early and mid-differentiation stages of myogenesis were not affected in Noggin null mice, the later stages were delayed and detected phenotypes were consistent with regenerating or underdeveloped muscle. The absence of Noggin and concomitant BMP oversignaling may be responsible for this phenotype in a similar fashion as it is responsible for myogenic defects in the in vitro model system based on interference with differentiation of C2C12 cells. In this model, the cells can be induced into myogenic differentiation by placing them under starvation conditions. Myogenic differentiation can then be inhibited by exposure of cells to members of the TGFβ family of ligands. While Tgfβ or Gdf5 will block muscle differentiations, BMP will reprogram the cells towards osteogenic lineage (Katagiri et al., 1994; Tylzanowski et al., 2001; Y. Li et al., 2004; G. Li et al., 2005; and our unpublished observations). It may thus be possible that over-activation of BMP signaling can convert some of the cells destined to enter myogenic differentiation into osteo/chondrogenic ones, as seen in vitro in C2C12 cells. Although we failed to detect differences in the phospho-SMAD levels (data not shown) in muscle tissues of wild type or Noggin null mice, other, SMAD-independent signaling pathways downstream of BMPs may be responsible for this phenotype (reviewed in Feng and Derynck, 2005).
Skin development is affected in the Noggin KO mice (McMahon et al., 1998; Botchkarev et al., 1999). There are, however, some differences between previous reports and our data. Unlike in 129X1/SvJ mouse strain, in the CD1 mouse genetic background the hair follicles begin to develop somewhat earlier, leading to a histologically better-defined follicle at 18.5 dpc in Noggin null mice with apparent downregulation of β-catenin protein levels as demonstrated by immunofluorescent staining at 14.5 dpc. This may not necessarily be in contradiction with published data (Botchkarev et al., 1999), since the authors investigated β-catenin expression only at 17.5 dpc. It is, therefore, quite possible that the differences seen at 14.5 dpc are equalized at 17.5 dpc. This delay may be directly responsible for the subsequent delay in follicular development.
As mentioned in the Introduction, NOGGIN haploinsufficiency in humans leads to a number of skeletal disorders. It is, therefore, surprising that Noggin haploinsufficiency in mice has not been reported previously (Brunet et al., 1998; McMahon et al., 1998; Botchkarev et al., 1999; Gong et al., 1999; Groppe et al., 2002).
Analysis of the expression levels of Noggin mRNA in Noggin heterozygous animals demonstrated that in the absence of one functional copy of the gene, the levels of Noggin mRNA were about 50% of the wild type in all embryos tested. This indicates that neither genetic imprinting nor allelic compensation is taking place in the absence of one functional Noggin allele.
Reexamination of the Noggin heterozygous mice revealed limb phenotype 100% penetrant in all strains tested. Specifically, the mice had carpal and tarsal fusions similar to the fusions described for Gdf5/Gdf6 double knockout mice (Settle et al., 2003), as well as human NOGGIN haploinsufficiency (Debeer et al., 2005), but with an interesting difference. The same carpal bones were fused in both the Gdf5/6 double null and Noggin heterozygous animals. In the hind legs, however, the fusions were different. In the Gdf5/6 double null mice, tarsals 2 and 3 were fused, but in the Noggin heterozygote animals, the central tarsal bone and talium did not fuse (Settle et al., 2003). Additionally, while Gdf5/6 double null mice showed scoliosis (lateral spine curvature), some Noggin heterozygote mice had kyphosis (dorso-ventral spine bending).
It is interesting to see that the absence of either two members of the ligand family (Gdf5 and 6) or a member of an antagonist family (Noggin) leads to partially overlapping developmental defects. This suggests that perhaps in some cases GDFs can act as antagonists rather than agonists of BMP signaling (see fig. 5 in Thomas et al., 1997), and the absence of either one may deregulate BMP signaling leading to absence of some of the joints. A similar case has been recently described for BMP3 that can, through the binding to the Type II receptor, prevent the subsequent binding and signaling by other members of the BMP family (Gamer et al., 2005). Targeted inactivation of BMP3 led to an increase in bone density and BMP3-conditioned medium could efficiently block BMP2-induced alkaline phosphatase in vitro, indicating that BMP3 can be an in vivo candidate antagonist for fine-tuning BMP signaling (Daluiski et al., 2001). While there are no publicly available data on BMP3 binding to NOGGIN, it is tempting to speculate that excessive BMP3 signaling in the absence of Noggin could be responsible for the phenotypes described above and further studies will be required to resolve this issue. It is also possible that the skeletal defects we discovered are absent in 129X1/SvJ strain but we did not investigate this particular background.
Thirty percent of mice heterozygous for Noggin allele showed additional, more severe defects in skeletal development than heterozygous littermates. This 30% phenotypic penetrance may be attributed to a number of factors, such as to the timing of correct levels of Noggin at specifically defined times and places during mouse development.
The haploinsufficiency of Noggin gene in at least three genetic backgrounds, as well as the 30% penetrance of a more severe phenotype, points to an increasingly appreciated concept of levels and ranges of signaling and the developmental consequences if either of those factors is perturbed. While we are not yet conducting quantitative developmental biology, a number of observations point to the need to address developmental and biological issues at the quantitative level.
Noggin heterozygous mice were kindly provided by Dr. Richard Harland (University of California, Berkeley). Mouse colonies were maintained according to Animal Welfare Guidelines. The timed matings were set up and noon of the day when the vaginal plug was detected was counted as day 0.5 dpc. The pregnant mice were sacrificed by cervical dislocation. All subsequent mouse work, including skeletal stainings, was carried out as described in the Mouse Laboratory Manual (Hogan et al., 1994). Histological stainings on paraffin-embedded sections were carried out according to standard protocols.
Mouse genotyping was carried out by two different methods. Noggin inactivation was achieved by inserting replacing coding region of the Noggin gene with LacZ expression cassette (Brunet et al., 1998). Therefore, when skeletal integrity of the embryo did not have to be preserved, the tail biopsies were stained for β-galactosidase activity (Hogan et al., 1994). In other cases, genotyping was carried out by PCR on DNA isolated form yolk sacs (Hogan et al., 1994). The PCR primers used to identify Noggin wt allele were Noggin-S: GCATGGAGCGCTGCCCCAGC and Noggin-AS: GAGCAGCGAGCGCAGCAGCG generating a fragment of 200 bp. Primers to identify Noggin null allele were Noggin-S as above and LacZ-AS: AAGGGCGATCGGTGCGGGCC generating a fragment of 170 bp. The PCR was carried out in Applied Systems 9600 PCR machine using the following conditions: 95°C for 1 min followed by 98°C for 10 sec and 72°C for 20 sec for 35 cycles. The cycling was followed by 72°C for 3 min. The samples were resolved on 1.5% agarose gel.
Immunofluorescence was carried out using standard protocols with the following primary antibodies used: anti-myosin heavy chain (Zymed, San Francisco, CA), anti- β-catenin (Santa Cruz Biotechnology, Santa Cruz, CA), anti-MyoD (Santa Cruz), myogenin (BD Pharmigen). The secondary antibodies were donkey-anti-rabbit (Alexa 488 Molecular Probes, Eugene, OR) and/or donkey anti-mouse (Alexa 555 Molecular Probes).
Dual in situ hybridization (DISH) was carried out as described previously (Tylzanowski et al., 2003).
Photography was carried out using SPOT2 digital camera (Diagnostic Instruments) placed on Leica DMR microscope. Digital photography was processed using the software provided by the camera's manufacturer.
The RNA for quantitative PCR was isolated from limbs of 14.5-dpc CD1 mouse embryos. The Quantitative PCR was carried out using Applied Biosystems (Foster City, CA) 7700 instrument and Noggin (Cat. No. Mm01297833_s1), Gdf5 (Cat. No. Mm00433564_m1), and GAPDH (Cat. No. Mm99999915_g1) Assay on Demand kits (Applied Biosystems) following the manufacturer's instructions.
We thank Drs. K. Verschueren and V. Jackiw for help with the manuscript.