Mutation and Analysis of Dan, the Founding Member of the Dan Family of Transforming Growth Factor β Antagonists

Abstract

The Dan family of transforming growth factor β antagonists is a large, evolutionarily conserved family of proteins. Little is known about either the specificity of these antagonists or the biological roles of these proteins. We have characterized Dan, the founding member of this family, with regard to both its biochemical specificity and its biological roles. Although DAN is not an efficient antagonist of BMP-2/4 class signals, we found that DAN was able to interact with GDF-5 in a frog embryo assay, suggesting that DAN may regulate signaling by the GDF-5/6/7 class of BMPs in vivo. Intriguingly, in developing neurons, Dan mRNA was localized to axons, suggesting a potential role for the DAN protein in axonal outgrowth or guidance. Mice lacking Dan activity were generated by gene targeting and displayed subtle, background-dependent defects.

Bone morphogenetic proteins (BMPs) were first isolated biochemically as activities that could induce ectopic bone formation in rat soft tissues. When the active proteins were purified and sequenced and the genes encoding them were cloned, they were found to be homologous to the Drosophila gene decapentaplegic (dpp), a member of the transforming growth factor β (TGF-β) superfamily of secreted signals (25, 36, 39). dpp and the two other Drosophila BMP family members, screw and glass-bottomed boat/60A, play roles in a wide variety of developmental processes, including the establishment of initial dorsal-ventral polarity in the Drosophila ectoderm and development of various imaginal disc-derived structures (8). This family of proteins has undergone dramatic evolutionary expansion: vertebrate genomes contain 20 or more BMPs (23). Loss-of-function mutants in the vertebrate BMPs display a tremendous variety of defects, and BMP overexpression can cause a wide variety of developmental defects (11).

Vertebrate genomes also encode many proteins that block signaling by TGF-β family members; these are generally called TGF-β or BMP antagonists. The largest group of these antagonists is the DAN family of proteins (13, 26, 35). There are at least seven DAN family members in the mouse, including DAN (differential screening-selected gene aberrative in neuroblastoma) itself, mCer1, gremlin, PRDC, and several genes that have only been identified as expressed sequence tags (3, 13, 22). This family is marked by a shared cysteine-rich domain and by a resemblance to the mucins, a family of secreted factors found in mucus (13, 26). It is unclear whether the similarity to mucin is functionally meaningful. DAN family members are believed to act by binding BMPs within the extracellular space (13, 27).

Dan was originally cloned as a transcript downregulated in src-transformed rat fibroblasts (24). DAN has been shown to be able to bind BMP-2 in vitro; however, interaction between DAN and any specific TGF-β under physiological conditions remains to be demonstrated (13). Although DAN can bind BMP-2 at high concentrations, the early phenotype of DAN overexpression in the frog embryo is distinct from those of other BMP antagonists, and hence argues against the BMP-2/4 and BMP-7 classes of BMPs as physiological ligands for DAN (13, 32).

In order to understand better the in vivo roles of DAN, we designed bioassays for BMP blockade within the Xenopus embryo, examined Dan expression in the mouse, and generated mice that lacked DAN function. In this way, we have shown that DAN may bind GDF-5 in vivo, that Dan mRNA is localized in many developing axon tracts, and that Dan mutant mice have only subtle defects.

MATERIALS AND METHODS

Library screening and generation of targeting constructs.

IMAGE clone 441413, encoding mouse DAN, was obtained from Research Genetics; an 873-bp PstI-PvuII fragment from this clone was used to screen a mouse 129/Sv genomic library by standard methods (28). Multiple clones were isolated and restriction mapped to verify that rearrangements had not occurred in the library. One lambda clone, designated λ10, was used to generate the regions of homology used in both targeting vectors.

(i) pDANλ10F/R.

For construction of pDANλ10F/R, a 12-kb phage insert which contained the entire DAN cDNA was cloned as follows. λ10 was cut with NotI and cloned into NotI-cut, phosphatase-treated pBluescript SK− (Stratagene). The F orientation indicates that the DAN gene reads from the SacI end to the KpnI end of the polylinker, and the R orientation indicates the reverse.

(ii) pLoxP2.1.

In order to simplify construction of the targeting constructs, two tandem loxP sites were inserted into the pBluescript polylinker. Two oligonucleotides were constructed: loxP1 (GAT CAT AAC TTC GTA TAA TGT ATG CTA TAC GAA GTT ATA AGC TTA GAT CT) and loxP2 (AGA TCT AAG CTT ATA ACT TCG TAT AGC ATA CAT TAT ACG AAG TTA T). These oligonucleotides were annealed by heating to 90°C and then allowing them to slowly return to room temperature. The resulting double-stranded fragment was phosphorylated with T4 polynucleotide kinase, resulting in a linker with the structure 5′ GATC end-loxP site-HindIII-Bg1II-blunt end. This linker was blunted with T4 DNA polymerase, cut with HindIII, and inserted between the EcoRV and HindIII sites in pBluescript SK−, resulting in pLoxP2. The (uncut) linker was then cloned between the BamHI and SmaI sites in pLoxP2, resulting in pLoxP2.1. The selected clone was sequenced to confirm that there were no mutations in the loxP sites; a base had been lost in the first ligation, resulting in the regeneration of the EcoRV site (and the loss of a planned BcII site).

(iii) pdko1.

Plasmid pdko.1 carried a targeting construct which removed the final exon of the Dan gene. A β-actin-neo-SV40pA cassette was cut from pBKNPA (W.C. Skarnes, unpublished observations) with BamHI and Bg1II; the resulting fragment was inserted into the Bg1II site in pLoxP2.1, resulting in the plasmid pLoxβneoF. pDANλ10F was cut with SpeI, blunted, and then cut with KpnI; the resulting fragment was inserted into pLoxβneoF between the KpnI and blunted ClaI sites, resulting in the plasmid pLoxβneo3′. Finally, pdko1 was generated by cutting pLoxβneo3′ with SpeI and inserting the 4-kb XbaI fragment from pDANλ10F.

(iv) pdko3.

Plasmid pdko3 carried a targeting construct which inserted an internal ribosome entry site (IRES)-placental alkaline phosphatase (PLAP) cassette into the Dan gene. pBKNPA was cut with XbaI, blunted, and then cut with EcoRI; this was inserted into PstI (blunt)-EcoRI-cut pLoxP2.1. The resulting plasmid was cut with EcoRI and EcoRV, and the PGK-neo-BPA cassette was inserted as an EcoRI-XhoI (blunt) fragment from pPGKneoTK (19). The IRES-PLAP cassette was cut from pGTO-IP (W.C. Skarnes, unpublished observations) with XbaI and inserted into the SpeI site, resulting in the plasmid pIP1. Two oligonucleotides were constructed: bglsph3 (GTG ATG ACA CCG GTG GAT GTC GTC GAC ATG) and bglsph4 (TCG ACG ACA TCC ACC GGT GTC ATC ACC TG). These oligonucleotides were annealed as described for pLoxP2.1, resulting in a linker with the structure 3′ CTG end-SgrAI-SalI-3′ CATG end. A 6.3-kb EcoRI-XhoI fragment was cut from pDANλ10F and inserted into pBluescript SK−. This plasmid was then cut with SphI and Bg1I, and the Bg1I-SphI linker above was inserted. The resulting plasmid was cut with ClaI and NotI, and the 3.2-kb NotI-ClaI fragment from pdko1 was inserted, resulting in pHRV9. pIP1 was cut with NotI and partly filled with G's, resulting in a 3′ GG overhang; this was then cut with SalI, liberating the IRES-PLAP-PGK-neo-BPA cassette. This fragment was ligated into pHRV9, which had been cut with SgrAI, partly filled with C's and then cut with SalI, producing pdko3.

Generation of Dan mutant mice; genotyping and mouse husbandry.

pdko1 and pdko3 were grown as multiple 100-ml cultures in Escherichia coli. DNA was purified from bacterial cultures by using Qiagen Maxiprep kits. Each construct was linearized (pdko1 with SacI; pdko3 with XhoI) and electroporated into E14 ES cells as previously described (31), except that 75 μg of DNA and 5 × 10⁷ cells were used for each electroporation. Selection was carried out in 175 μg of G418 per ml (Gibco). ES cell lines were analyzed by Southern blotting with a 5′ flanking probe (the AseI-SacI fragment shown in Fig. 3A) and a probe specific for neomycin phosphotransferase, as previously described (12). With the dko1 vector, 310 lines were analyzed, of which 8 carried the desired mutant allele; with the PLAP vector, 300 lines were analyzed, of which 1 carried the desired mutant allele.

FIG. 3.

Targeting of the Dan locus. (A) Map of the wild-type Dan locus and the two knockout constructs. All genes in this figure are transcribed from left to right; all exons of the Dan gene are shown. The top construct is for Dan^dkol, which removed intron II and exon III; the bottom construct is for Dan^PLAP, which removed parts of exons I and II and intron I. Bracketed restriction sites are those used for genotyping by Southern blotting, with the probe shown (between the AseI and SacI sites at the left end of the DAN genomic fragment). Below is shown the amino acid sequence of the DAN protein. The underlined region is the conserved cysteine-rich domain shared by DAN family members; the boxed region was removed in Dan^PLAP, and the region in italic was removed in Dan^dkol. (B) RT-PCR analysis showing that the portions of the Dan protein retained in Dan^dkol mice had no detectable residual activity in a frog embryo dorsalization assay. Ventral marginal zones from embryos injected with 100 pg of wild-type Dan mRNA expressed high levels of muscle-specific cardiac actin (MA), a marker of paraxial mesoderm, while those from uninjected embryos or embryos injected with 1 ng of Dan^dkol mRNA displayed no muscle actin expression. (Wild-type DAN will robustly induce muscle actin at doses ranging from 50 pg to several nanograms.) EF-1α is a loading control; the lane labeled “−RT” is a control sample treated without reverse transcriptase. (C) Genotyping of Dan^dkol and Dan^PLAP mice by Southern blotting. The left-hand blot was made from tail DNA cut with EcoRV and AseI; the right-hand blot was made from tail DNA cut with BamHI and AseI. (D) Genotyping of Dan^dkol and Dan^PLAP mice by PCR. The left two lanes (aside from the ladder) were analyzed for Dan^PLAP; the right three lanes were analyzed for Dan^dkol. (E) Southern blot of genomic DNA cut with AseI and EcoRV and probed with a 500-bp BglII fragment from exon III of Dan. As predicted, hybridization was lost in DNA from Dan^dkol mice. This blot had been probed previously with the genotyping probe to confirm the genotypes of these mice and to verify that all lanes were loaded with similar amounts of DNA.

Two dko1 lines (101 and 237) and the PLAP line (K93) were injected into C57BL/6J blastocysts. All three lines produced mutant progeny. Chimeras were crossed onto C57BL/6J (Jackson) and 129S6SvEv (Taconic) females; lines were maintained by backcross onto wild-type females of the appropriate inbred strain. Animals were genotyped either by Southern blotting or by PCR as previously described (12). Genomic DNA for Southern blotting was cut with AseI and either BamHI (for Dan^PLAP) or EcoRV (for Dan^dko1). All PCR genotyping was done with Platinum Taq (Gibco). The cycling conditions were 2 min at 95°C and 35 cycles of 30 s at 95°C, 30 s at 60°C, and 1 min at 72°C. The following primers were used: for allele dko1, dan5 (CAC TCG GGT CCA GGG GAG ATG G), dan8 (CTG GAC CCT TTA GGA AAG TAG C), and neo5 (CAC ACC TCC CCC TGA ACC TG), resulting in a 361-bp wild-type product and a 280-bp mutant product, and for the PLAP allele, danap1-2 (ACA ATG CTT TGG GTC CTG GTG G), danap2-5 (TTC CTC TAG TTC TAG AGC GGC C), and danap 3-2 (CTG AAG CAT TGT CCT AGG CAC GC), resulting in a 300-bp wild-type product and a 170-bp mutant product. Noggin mice were genotyped by PCR as previously described (19).

Xenopus embryo assays.

Xenopus embryos were staged, cultured and injected as previously described (18). Ventral marginal zone explants were cut at stage 10.25 and cultured in Sater's modified blastocoel buffer until sibling embryos had reached stage 20 for analysis of muscle actin gene expression (30). Reverse transcription-PCR (RT-PCR) analysis was carried out as previously described (38).

Skeletal preparations.

Skeletal preparations were performed as previously described (17).

Whole-mount in situ hybridization.

In situ hybridization was carried out essentially as described previously (37) with the following changes. All stages previous to probe hybridization were carried out on ice, and 0.1 M NaP (1 M NaP is 0.5 M Na₂HPO₄, with pH adjusted to 7.5 with H₃PO₄) was substituted for phosphate-buffered saline (PBS); mouse powder was not used; and antibody incubation and postantibody washes were carried out in a mixture of 1× MAB (1× MAB is 100 mM maleic acid plus 150 mM NaCl adjusted to pH 7.5 with NaOH), 0.1% bovine serum albumin (BSA) and 2% BM blocking reagent (Boehringer-Mannheim). Embryos at e12.5 or later were usually cut into pieces after initial fixation to aid reagent penetration.

PLAP staining.

Heads from embryos to be stained for alkaline phosphatase activity were fixed overnight in 4% paraformaldehyde–1× HeBS (10× HeBS is 82 g of NaCl per liter 59.5 g of HEPES acid per liter 1.05 g of Na₂HPO₄ pH per liter, with pH adjusted to 7 with NaOH) at 4°C. Roughly 18 h later, embryos were rinsed multiple times and then stored at 4°C in 1× HeBS for as long as several weeks.

For sectioning, embryos or heads were dried thoroughly with a Kimwipe and embedded in 5% low-melt agarose (Gibco)–1× HeBS. One hundred-micrometer sections were cut on a vibratome, allowed to dry at room temperature, and then immersed in 1× HeBS. Endogenous phosphatases were inactivated by heating to 65°C for 45 min. The slides were cooled to room temperature, and the HeBS was replaced with AP buffer (100 mM Tris [pH 9.5], 100 mM NaCl, 50 mM MgCl₂). After 3 min, the AP buffer was replaced with AP buffer supplemented with 175 μg of BCIP (5-bromo-4chloro-3-indolylphosphate) and 337.5 μg of nitroblue tetrazolium per ml. Sections were then allowed to react in the dark for 2 h at 37°C.

Once sections were stained, slides were rinsed in 1× PBS or 1× HeBS, allowed to sit for a few minutes, and then taken through 25, 50, and 75% ethanol (EtOH) in water to 100% EtOH). The 100% EtOH was changed once, and then slides were transferred to Hemo-De (Fisher). Hemo-De was changed twice, and then slides were mounted with Permount (Fisher).

Fluorescent antibody staining of sections.

Embryos were dissected on ice and fixed for 2 to 3 h in 4% paraformaldehyde–0.1 M sodium phosphate buffer (pH 7.5) at 4°C, rinsed twice in 0.1 M NaP buffer, and then transferred to 0.1 M NaP buffer–30% sucrose at 4°C and allowed to equilibrate for 24 h. At the end of this period, they were embedded in TBS tissue freezing medium (Fisher) and sectioned on a cryotome at a 20-μm thickness.

Slides were then washed three times for 5 min each in 1× PBS, which was then changed for PHT (PBS plus 1% heat-inactivated goat serum, 0.1% Triton X-100). After 30 min in PHT, slides were removed from the bath and placed upside-down on 200-μl drops of primary antibody mix (1:200 dilution of affinity-purified rabbit polyclonal anti-MATH1 antibody in anti-MASH1 antibody supernatant) that had been placed on Parafilm (American National Can Company). These were then incubated overnight at 4°C, washed three times in 1× PBS for 10 min each, and then placed upside-down as before on drops containing 1:150 Texas red-conjugated goat-anti-rabbit immunoglobulin G(IgG) and 1:200 Fluorescein isothiocyanate-conjugated goat-anti-mouse IgG in 1× PBS (secondary antibodies were from Jackson Immunoresearch); secondary antibodies were allowed to bind for at least 2 h at room temperature. Slides were then once again washed three times for 10 min each in 1× PBS and mounted with Vectashield (Vector Research).

RESULTS AND DISCUSSION

DAN bioassays.

We wished to understand DAN's physiological role better by identifying TGF-βs that could interact with DAN in a biological context. The Xenopus embryo is an excellent system for analyzing interactions between TGF-βs and their antagonists, since the embryo provides distinct readouts for different kinds of TGF-β activities. Several BMPs are expressed in the Xenopus embryo at this stage of development (4, 9, 10, 14). Overexpression of BMP-2/4-class or BMP-7-class signals in the early mesoderm induces ventral fates, while inhibitors of these signals (such as Noggin, Xnr3, Chordin, or Follistatin) induce dorsal fates (6). Dan can mimic the dorsalization caused by these inhibitors in mesodermal explants, although inefficiently (13, 32). This suggested to us that, while DAN was apparently able to antagonize BMP-4 or BMP-7 when provided in sufficient excess, the BMP-2/4 and BMP-7 classes of signals were unlikely to be physiological targets for DAN. We examined the GDF-5/6/7 class as potential DAN targets, because this group of BMPs is the next most closely related to the other two classes at a sequence level (16, 34).

The interaction between DAN and GDF-5 can be assayed indirectly, by testing whether GDF-5 can block the dorsalizing effects of DAN: in essence, by asking if GDF-5 is able to compete efficiently with the other BMPs in the early embryo to bind DAN. We used this assay because, in our hands, GDF-5 caused no obvious phenotype in the Xenopus embryo even when expressed at very high levels (data not shown). If GDF-5 is a physiological DAN ligand, it should compete with endogenous BMPs to bind Dan and hence efficiently prevent Dan-induced mesoderm dorsalization. We found that even small doses of GDF-5 were sufficient to reproducibly antagonize DAN-mediated dorsalization (Fig. 1A). Moreover, GDF-5 was unable to significantly reverse Noggin-induced dorsalization, even when Gdf5 mRNA was present in 100-fold excess over Noggin mRNA (Fig. 1B). This suggests that the reversal of Dan-induced dorsalization was a result of GDF-5 binding directly to DAN rather than a result of it signaling through BMP receptors. Although these experiments do not show direct binding of GDF-5 by DAN, they suggest that in the Xenopus embryo, DAN may interact with GDF-5 with a higher affinity than with the endogenous members of the BMP-2/4 and BMP-7 classes. Noggin has previously been shown to interact directly with GDF-5 in vitro (21); our results suggest that Noggin's affinity for GDF-5 in vivo may be significantly lower than that for BMP-4.

FIG. 1.

GDF-5 blocks DAN activity. Xenopus embryos were injected on the ventral side at the four-cell stage with the doses of Dan and Gdf5 mRNA shown. Embryos were cultured until the beginning of gastrulation, and then the ventral marginal zones were explanted and cultured until sibling embryos had reached stage 20, when they were lysed and assayed by RT-PCR for expression of muscle-specific cardiac actin, a marker of paraxial mesoderm. (A) In lanes 1 to 3, increasing doses of Dan induced increasing expression of muscle-specific cardiac actin (MA); in lanes 4 and 5, increasing doses of Gdf5 blocked this induction with increasing efficiency. Intriguingly, extremely high doses of Gdf5 blocked DAN less efficiently (compare lanes 5 and 6); this result was seen in three separate experiments and remains unexplained. As expected, Gdf5 alone did not induce MA above background levels, even at high doses (lanes 7 and 8), and when the RT was omitted, no background was visible. (B) In addition to Dan-Gdf5 coinjections, Gdf5 was also coexpressed with Noggin. In addition to repeating the result shown above, this experiment also showed that Gdf5 was unable to efficiently counter Noggin-induced dorsalization, as seen in lanes 9 to 11.

Dan mRNA localization in axons.

Expression of Dan in the early mouse embryo has been described previously (26, 32). However, in examining Dan expression by in situ hybridization, we made the surprising observation that Dan mRNA was localized to many axon tracts in the developing mouse fetus (Fig. 2). This staining was visible as early as e11.5; embryos processed in parallel with a variety of other probes did not show similar staining. This staining appeared to be axonal and not specific to glia or other support cells, since the only glia which are present at this time and place are white-matter astrocytes, which display little resemblance to the cells stained (15, 29). Not all axon tracts were stained, in keeping with the restricted Dan expression seen in the developing brain. In some cases (including the animal shown), it was possible to trace the staining back along the axons all the way to the cell bodies; in these cases, the cell bodies appeared no more strongly stained than the axon. Such localization has previously been suggested to be characteristic of mRNAs encoding proteins which are required in distal portions of the developing axons (33).

FIG. 2.

Dan mRNA was found in axon tracts. Embryos were split in half sagittally and subjected to whole-mount in situ hybridization. (A) An e12.5 embryo. (B) An e14.5 embryo. In both cases, DAN mRNA was strongly localized in many of the descending projections from the forebrain and midbrain (arrowheads). The arrowhead in panel B points out a specific tract in which the cell bodies and the projections could be distinguished; based on its location and the developmental stage of the embryo, this may be the medullar projection from the arcuate nucleus of the hypothalamus.

Generation of Dan mutant mice.

In order to address biological roles for DAN, we generated mice carrying two mutant alleles of the Dan gene. Dan consists of three exons; the Dan^dkol allele replaces nearly all of the third exon of Dan with a β-actin::Neo^r cassette, while the Dan^PLAP allele replaces much of the first two exons and the entire first intron with an IRES—PLAP; PGK::Neo^r cassette (Fig. 3A). We chose to make mice carrying the PLAP marker because our observation that Dan mRNA was localized to axons led us to hypothesize that the mutant mice might exhibit some defect in axon guidance that would be most easily analyzed by being able to specifically stain DAN-expressing axons.

Both alleles delete significant portions of the DAN family conserved region (Fig. 3A). Moreover, Dan^PLAP separates the remaining C-terminal portion of the protein from the N-terminal secretory signal. In order to verify that Dan^dkol had little remaining activity, we injected mRNA encoding the predicted mutant protein into Xenopus embryos. No morphological defects resulted from this overexpression in the whole embryo (data not shown). Moreover, this mutant form had lost the ability to induce muscle actin expression in explanted ventral marginal zones, even when expressed at levels 10-fold higher than those required for robust induction by the wild-type protein (Fig. 3B), suggesting that the mutant allele was either a very strong hypomorph or a null allele.

Mutant mice were genotyped by Southern blotting with a 5′ probe (Fig. 3C), using a Neo^r probe (data not shown), or by PCR (Fig. 3D). Dan^dkol DNA was also blotted with a probe made to the deleted portion of the gene; no hybridization was visible with this probe in homozygous mutant mice (Fig. 3E).

Mice homozygous for either Dan mutant allele were viable, fertile, and displayed no obvious morphological or behavioral abnormalities, whether they were crossed onto a C57BL/6J background or onto a 129S6SvEv background. Dan^dkol mutant animals were born in the expected numbers from crosses of heterozygotes (for the C57BL/6J background, of 371 animals, 96 were wild type, 179 were heterozygous, and 96 were homozygous; for the 129S6SvEv background, of 183 animals, 46 were wild type, 91 were heterozygous, and 46 were homozygous).

Characterization of Dan mutants.

Early Dan expression (from e8 to e11.5) has previously been described; at these stages, Dan is expressed in head mesoderm, somites, facial structures, and limbs (26, 32). No defects were apparent in any of these structures in mutant mice (data not shown).

BMP family members have been found to be able to regulate dorsal-ventral patterning of the spinal cord and hindbrain, and we had observed diffuse Dan expression in the developing neural tube at e11.5 and e12.5 (Fig. 4A and B) (1, 16). Accordingly, we examined dorsal-ventral patterning of these structures in the Dan mutants. No alterations were observed in expression of Math-1 or Mash-1 in the spinal cord (Fig. 4C and D) or at the rhombic lip (data not shown) of Dan mutant mice.

FIG. 4.

Dan^dkol/dkol animals showed no defect in patterning of the dorsal spinal cord. (A and B) Dan expression in the spinal cord at e11.5 (A) and e12.5 (B). (C and D) Transverse sections through the spinal cord at the level of the forelimb of a wild-type mouse (C) and a Dan mutant at e12.5 (D). Red indicates MATH1 expression, which marks dorsal interneuron precursors; green indicates MASH1 expression, which marks a variety of neuronal precursor types in the dorsomedial part of the ventricular zone.

Because of the localization we had previously observed of Dan mRNA in axons and because BMPs have previously been characterized as axon-guidance cues both in mammals and in Caenorhabditis elegans (2, 5), we examined mice carrying the Dan^PLAP allele for defects in axon tracts that were positive for PLAP activity at e14.5. No abnormalities were seen in these tracts, suggesting that Dan is not required for proper axonal pathfinding in the cells in which it is expressed.

Recent data suggest that Noggin is required for correct generation of limb tendons (C. Tabin, personal communication). Dan mRNA is expressed in the ventral limb tendons at e12.5 (data not shown). We assayed Dan mutant embryos for perturbations in expression of the tendon marker scleraxis at e12.5. No defects were apparent (data not shown). Additionally, no defects were apparent in visual examination of the ventral limb tendons of adult Dan mutant mice (data not shown).

Because DAN can antagonize BMP-class signals, we examined Dan mutant mice for skeletal defects (13, 32). No defects were apparent in Dan mutant pups on either genetic background (Fig. 5). We also examined mice that lacked Dan and were heterozygous for Noggin. On the 129S6SvEv background, 28% (n = 7) of these animals displayed an apparent transformation of the right-hand side of the last lumbar vertebra to a sacral fate (Fig. 6). No Dan-heterozygote or Dan-Noggin transheterozygote animals (n = 15) displayed this defect. This defect is particularly interesting in the context of recent work showing that Gdf11 acts as a global regulator of anterior-posterior pattern in the skeleton (20). Gdf11 mutant mice display posterior-to-anterior transformations throughout the axial skeleton, including extra ribs and posteriorward shifts of specific vertebral landmarks and the hindlimbs and posteriorward shifts in expression of a number of Hox genes (20). DAN and Noggin may be involved in region-specific regulation of the BMP or GDF signals in order to properly regulate antero-posterior identity in the posterior lumbar region. However, specific blocking of GDF-11 by DAN or Noggin seems unlikely: DAN was unable to block the activity of GDF-8, a close relative of GDF-11 (data not shown), and others have found that Noggin is unable to block GDF-11 signaling (7). A more likely possibility is that DAN and Noggin may be responsible for antagonism of other TGF-β family members in the region which would otherwise mis-specify the local regional identity. Another possibility is that the apparent lumbar-to-sacral conversion is not a true homeotic transformation, but rather an alteration in the morphology of the last lumbar vertebra so that it resembles a sacral vertebra.

FIG. 5.

No gross skeletal defects were apparent in newborn Dan^dkol/dkol animals. Blue staining indicates cartilage, while red is bone. (A) A Dan heterozygote. (B) A homozygous mutant, both on the 129S6SvEv inbred background.

FIG. 6.

The skeletal defect in Dan^dkol/dkol Nog^−/+ animals. The left panel shows the sacral vertebrae of an unaffected animal, while the right panel shows an animal in which the right-hand side of the last lumbar vertebra was affected. In each panel, the right-hand side of the last lumbar vertebra is indicated with an arrowhead.

We have presented data which suggest that DAN may be an antagonist of GDF-5/6/7 signals in vivo. Dan mRNA was localized within developing axons. This suggested a potential role for DAN in rendering the extending axons resistant to environmental GDF-5/6/7 signals; however, when we generated Dan mutant mice, we found no apparent defects in proper extension or pathfinding by the expressing axons. The Dan mutant mice were viable and fertile, with no obvious abnormalities. These mice displayed no defects in proper dorsal-ventral patterning of the spinal cord or in skeletal development. However, mice lacking Dan and heterozygous for Noggin displayed, at low penetrance, an apparent homeotic transformation of the last lumbar vertebra to a sacral fate.