Abstract
Our previous studies in both mouse and human identified the Bapx1 homeobox gene, a member of the NK gene family, as one of the earliest markers for prechondrogenic cells that will subsequently undergo mesenchymal condensation, cartilage production and, finally, endochondral bone formation. In addition, Bapx1 is an early developmental marker for splanchnic mesoderm, consistent with a role in visceral mesoderm specification, a function performed by its homologue bagpipe, in Drosophila. The human homologue of Bapx1 has been identified and mapped to 4p16.1, a region containing loci for several skeletal diseases. Bapx1 null mice are affected by a perinatal lethal skeletal dysplasia and asplenia, with severe malformation or absence of specific bones of the vertebral column and cranial bones of mesodermal origin, with the most severely affected skeletal elements corresponding to ventral structures associated with the notochord. We provide evidence that the failure of the formation of skeletal elements in Bapx1 null embryos is a consequence of a failure of cartilage development, as demonstrated by downregulation of several molecular markers required for normal chondroblast differentiation (alpha 1(II) collagen, Fgfr3, Osf2, Indian hedgehog, Sox9), as well as a chondrocyte-specific alpha 1 (II) collagen-lacZ transgene. The cartilage defects are correlated with failed differentiation of the sclerotome at the time when these cells are normally initiating chondrogenesis. Loss of Bapx1 is accompanied by an increase in apoptotic cell death in affected tissues, although cell cycling rates are unaltered.
INTRODUCTION
During vertebrate development, the initial overt step in axial skeletogenesis is the generation of the paired paraxial mesodermal somites, which are spheres of epithelial cells located on both sides of the neural tube. Positioned immediately ventral to the neural tube is the notochord, which releases local signaling molecules critical for patterning of the ventral somite. Additional inductive signals derived from adjacent lateral plate mesoderm, the dorsal neural tube and surface ectoderm direct the dorsolateral patterning of the somite (reviewed in (Lassar and Munsterberg, 1996; Pourquie et al., 1996). Sonic hedgehog (Shh) is a signaling molecule produced by the notochord and its expression coincides with the activation of the transcription factors Bapx1 and Pax1 in the ventromedial presclerotome cells of the somite (Borycki et al., 1998; Johnson et al., 1994). Bapx1 and Pax1 are the earliest ventral markers for the onset of the somite dorsoventral polarization (Balling et al., 1996; Tribioli et al., 1997). Presclerotome cells subsequently undergo an epithelial-to-mesenchymal transition, migrate ventromedially to surround the notochord and lateral neural tube, and differentiate into chondroblasts. Cells of the chondrogenic lineage produce a cartilaginous matrix (primarily collagen type II) which, in association with proteoglycans, establishes the framework of the fetal skeleton. This framework is subsequently mineralized following chondrocyte hypertrophy and osteoblast differentiation, through the process of endochondral ossification (Cancedda et al., 1995).
The vertebrate skeleton is almost entirely mesodermal in origin, aside from several bones in the skull, which are derived from neural crest (Couly et al., 1993; Noden, 1992) reviewed in (Hanken and Thorogood, 1993). In Drosophila, one of the principal genes controlling mesoderm differentiation is the bagpipe homeobox gene (Azpiazu and Frasch, 1993; Azpiazu et al., 1996). We have previously isolated from mouse and human, homologues of bagpipe termed Bapx1. Examination of the expression of Bapx1 during embryogenesis revealed an expression almost exclusively restricted to paraxial and lateral plate mesoderm, with earliest expression detectable in the presclerotome cells of the somite and in splanchnic mesoderm surrounding the gut endoderm (Tribioli et al., 1997; Tribioli and Lufkin, 1997). During subsequent stages of embryogenesis, Bapx1 is expressed in essentially all cartilaginous condensations that will subsequently undergo endochondral ossification and in splanchnic mesoderm-derived tissues giving rise to intestinal smooth muscle, parts of the peritoneal body wall and spleen. To investigate the genetic role of Bapx1 in embryogenesis, we have undertaken a loss-of-function study. Mice lacking Bapx1 are asplenic, but have otherwise normal visceral development. Furthermore, Bapx1 is dispensable for sclerotome migration and early proliferation but is essential for appropriate prechondroblast-to-chondrocyte transition in mesenchymal cells most closely associated with the notochord.
MATERIALS AND METHODS
Construction of the Bapx1 targeting vector
The targeting vector was constructed from a mouse genomic clone that contains the Bapx1 gene, which was isolated from a 129/Sv genomic phage library (Tribioli et al., 1997). To construct the targeting vector, a 17 kb genomic SalI fragment was subcloned into pTZ18R (US Biochemicals). A 2.1 kb BssHII fragment encompassing the Bapx1 exon I-coding region, the single intron and part of exon II-coding sequences, including the homeobox, was deleted (Tribioli and Lufkin, 1997) and replaced by a 1.8 kb SpeI fragment that contained the bacterial neomycin resistance gene (neo) selectable marker under the transcriptional control of the phosphoglycerate kinase (PGK) promoter followed by the PGK poly(A) signal. This neo cassette is flanked by two loxP sites (floxed). The PGK neo cassette was cloned in the opposite transcriptional orientation relative to the endogenous Bapx1 gene and it introduced two new SphI restriction sites into the Bapx1 allele (Fig. 1A). The targeting vector was linearized at the unique XbaI site for electroporation into ES cells.
Generation of recombinant ES cell clones, transgenic and mutant null mice
ES cell transfection, chimera production and testing, and genotyping of offspring are essentially as previous described (Wang et al., 1998). The probe used for the identification of homologous recombinants by Southern blot analysis with XbaI+SphI digestion was a 0.5 kb HindIII-XbaI fragment (probe 1, Fig. 1A), containing sequences located distal to the 3′ arm of the homology region present in the targeting vector (Fig.1A). Chimeric male mice from two independent ES clones (two from clone 34 and two from clone 77) were used to transmit the mutant allele, designated Bapx1neo+, onto a congenic inbred 129/SvJ background (#000691, The Jackson Laboratory) and an outbred C57BL/6J background. Heterozygotes were intercrossed to obtain homozygous null embryos. Genotyping of embryos obtained from heterozygous crosses was performed by Southern blot analysis (with the probe 1, Fig. 1A) as previously described (Wang et al., 1998).
To generate the Bapx1neo− null allele from the Bapx1neo+ allele, Bapx1neo+ male heterozygotes were mated with CMV-Cre transgenic female mice, which express Cre recombinase in unfertilized oocytes and preimplantation embryos under the direction of the CMV promoter (Nagy et al., 1998). The resulting offspring were genotyped with XbaI+SphI digestion and the Bapx1 probe1 (Fig. 1B) and with BamHI digestion and a 1 kb SpeI fragment as probe from the plasmid pSL13, containing the Cre gene (Li et al., 1997). Heterozygotes with the neo− deleted allele, resulting from Cre-mediated recombination, were inbred to produce Bapx1neo− null homozygotes (Fig. 1B).
Transgenic mice carrying the lacZ reporter under the control of the mouse pro alpha 1(II) collagen promoter and enhancer were generated by pronuclear injection of 1-cell B6D2 embryos as previously described (Frasch et al., 1995). The mouse pro alpha 1(II) collagenlacZ (collagen-lacZ) transgene construct was kindly provided by Benoit de Crombrugghe and has been previously described (Metsaranta et al., 1995; Zhou et al., 1995). Three transgenic founders showing identical cartilage-specific expression were generated and used for this study. Embryo fixation, β-galactosidase staining and paraffin sectioning was performed as previously described (Frasch et al., 1995; Wang et al., 1998).
RT-PCR assays
RNA was isolated from entire E14.5 mouse embryos and yolk sac DNA was removed for genotyping. Total RNA preparation and firststrand cDNA synthesis were performed as previously described (Tribioli and Lufkin, 1997). The cDNA was then used as a substrate for PCR amplification assays using AMP Taq DNA polymerase (Perkin Elmer) and standard procedures with a final concentration of 10% DMSO in the reaction buffer (Tribioli and Lufkin, 1997). One pair of primers: (o1:5′GAAGAGAACGAGGGCAGGAG 3′ and o2: 5′GCAGTGGCAGAAGGGAAGGTG 3′) was used for the first round of amplification and the pair of primers (o3: 5′CCAAGGACCT-GGAGGAGGAA 3′ and o4: 5′GCAGAGG-CGAGCAGGTCGGC 3′) was used for the nested round. o1 and o3 primers are located in Bapx1 exon 1, which is deleted in the mutant allele, thus allowing us to distinguish the expression of the wild-type Bapx1 allele from the Bapx1 null allele. The specificity of the RT-PCR reactions was verified by Southern blot analysis using the Bapx1 p1205 plasmid insert as probe. The primers used in the control reactions for β-actin mRNA (antisense: 5′-TCTCCAGGGAGGAAGAGGAT03′; sense: 5′-ATG-TTTGAGACCTTCAACACC-3′) were employed as previously described (Liu et al., 1996).
Histological and skeletal analyses
For histological studies, 7 μm paraffin-embedded sections were collected on glass slides, dewaxed and stained with either: HE, hematoxylin and eosin; or AR, alcian blue and nuclear fast red; or HGF, hematoxylin, fast green and basic fuschin, essentially as described (Sheehan and Hrapchak, 1987). Embryos were fixed in Bouin’s solution or in 4% paraformaldehyde overnight (with equivalent results) then dehydrated through graded ethanols, followed by Americlear (substituted for xylene) and paraffin embedding. HE staining was performed essentially as described (Lufkin et al., 1993). AR staining for cartilage was performed on dewaxed and rehydrated sections. Slides were treated for 30 minutes with 1% alcian blue 8GX (Mallinckrodt) in 3% glacial acetic acid, followed by 2 minutes washing in running water and then 5 minutes in nuclear fast red (Kernechtrot) counterstain (Vector Laboratories). Sections were then dehydrated in graded ethanols and coverslipped. HGF staining for collagen-associated proteoglycans was performed as follows. Rehydrated sections were stained in Weigert’s iron hematoxylin solution (Sheehan and Hrapchak, 1987) for 1 minute and rinsed with running water until the blue color fully developed. Sections were then transferred to fast green FCF stain (1:5000 aqueous solution) for 3 minutes, rinsed briefly in 1% acetic acid and then stained in 0.1% basic fuchsin (in 1:100 glacial acetic acid:water) for 4 minutes. Sections were then dehydrated in 95% and 100% ethanols and coverslipped. Unless indicated, all stains were obtained from Sigma. For staining and visualization of whole-mount cartilage and ossified skeletal elements, embryos or neonatal mice were dissected and stained with alizarin red and/or alcian blue, as previously described (Lufkin et al., 1992). For alcian blue/alizarin red combined staining, the skin and internal organs were removed and the samples were fixed overnight in 95% ethanol followed by staining with 0.02% alcian blue in 4:1 95% ethanol:glacial acetic acid for 2 days. The samples were washed in 95% ethanol rapidly and immersed in 2% KOH for several hours to overnight. The samples were then stained in 75 μg/ml alizarin red in 1% KOH overnight, then processed through a graded series of glycerols in ethanol and stored in 100% glycerol. For cartilage staining, E12.5-E14.5 embryos were fixed in Bouin’s solution overnight, rinsed rapidly with water several times, immersed in four changes of 1% ammonia in 70% EtOH for at least one hour each and stained overnight with 0.05% alcian blue in 5% acetic acid. Embryos were rinsed in 5% acetic acid three times for 1-2 hours each and then once overnight. Specimens were dehydrated through graded ethanols, cleared and stored in 2:1 benzyl alcohol:benzyl benzoate (Sigma) and photographed in glass dishes.
RNA in situ hybridization
In situ hybridization analysis was performed as previously described (Tribioli et al., 1997). The following cDNAs were used as templates for synthesizing antisense or sense strand [35S]UTP RNA probes: 0.7 kb Bapx1 cDNA (p1140) (Tribioli et al., 1997); 0.9 kb BMP4 (Lyons et al., 1989); 0.4 kb Fgfr3 (Goldfarb, 1990); 3.0 kb Mfh1 (Winnier et al., 1997); 1.5 kb Shh (Echelard et al., 1993); 0. 473 kb pro alpha1(II) collagen (Andrikopoulos et al., 1992); 0.313 kb Pax1 (Koseki et al., 1993); 0.270 kb Osf2 (pLA-Oa4) (Ducy et al., 1997); 1.8 kb Ihh (Yang et al., 1998); To distinguish the expression of the Bapx1 wild-type allele from the two null alleles, we employed the Bapx1 probe2 (pGIA115), which contains Bapx1 cDNA sequences from nt 268 to nt 687 (Tribioli et al., 1997), which are deleted in the Bapx1neo+ and Bapx1neo− mutant alleles.
Cell proliferation and apoptosis
To determine cell proliferation, DNA synthesis was examined by measuring 5-bromo-2′-deoxyuridine (BrdU) incorporation into cells of E10.5-E14.5 embryos. Pregnant mice were injected intraperitoneally with a mixture of BrdU (Sigma B-9285) and 5-fluorodeoxyuridine (FUdr, Sigma F-0503) at 50 μg and 10 μg per gram body weight, respectively. After 1 hour, the embryos were removed, fixed in 4% paraformaldehyde overnight at 4°C, dehydrated in a graded ethanol series and embedded in paraffin. 7 μm sections were cut and mounted on glass slides then dewaxed in Americlear and rehydrated through a graded series of ethanols with water as a final wash. BrdU-positive cells were identified using a mouse monoclonal antibody (clone BMC 9318, IgG1) followed by a sheep anti-mouse Ig-alkaline phosphatase and NBT/X-phosphate color reaction essentially as described by the supplier (Boehringer Mannheim). Antibody incubations were performed for 1 hour at 37°C in a humidified incubator and all washes were performed at room temperature. Following the AP color reaction at room temperature for 10-20 minutes, slides were washed three times in PBS and then coverslipped in DTG2; 2.5% DABCO, 17.5% 0.5 M Tris pH 8.6, 80% glycerol, the edges sealed with nail hardener and stored at 4°C. Following photography, cells of the prevertebrae were scored as labeled or unlabeled in ten or more sections from at least two independent embryo preparations and cells of the overlying neural tube or adjacent lateral plate mesoderm were scored for an internal control.
Apoptotic cell death was examined by analyzing the extent of oligonucleosomal DNA cleavage (TUNEL) in cells of E10.5-E14.5 embryos. Bapx1 wild-type and null embryos derived from Bapx1 heterozygote intermatings were genotyped, fixed, paraffin embedded and sectioned as described above. Dewaxed and rehydrated sections were incubated in permeabilisation solution (0.1% Triton X-100 in 0.1% sodium citrate) for 8 minutes, rinsed three times in PBS and incubated with terminal deoxynucleotide transferase (TdT) and fluorescein-conjugated dUTP for 1 hour at 37°C as described by the supplier (Boehringer Mannheim). Slides were washed three times in PBS at room temperature and the incorporated fluorescein-dUTP was detected using an alkaline phosphatase-labeled anti-fluorescein antibody followed by NBT/X-phosphate color reaction. Section washing, coverslipping, photography and cell counting are as described above.
RESULTS
Targeted disruption of the murine Bapx1 gene
Because the Bapx1 deleted region comprises nearly all of the coding region and the neo substitution introduces multiple stop codons into Bapx1 exon II, the disruption would be predicted to result in a null allele. This allele is referred to as Bapx1neo+. The linearized targeting vector was electroporated into ES cells and clones were selected for resistance to G418. 80 resistant clones were analyzed by Southern blotting using a 3′ probe (probe1) external to the targeting vector, and 8 independent cell lines, including clone 34 and clone 77, yielded the 5 kb XbaI+SphI band expected from a homologous recombination event (Fig. 1A), giving an overall frequency of 10%. ES clones 34 and 77 were injected into C57BL/6J blastocysts to generate chimeras that transmitted the Bapx1 null allele to their progeny. The Bapx1 null allele was either outcrossed to the C57BL/6J strain or maintained on a congenic inbred 129/SvJ background. Heterozygous mice were identified by Southern blot analysis of tail genomic DNA, and were viable, fertile, healthy and born in appropriate Mendelian ratios. However, careful examination of skeletal preparations of E18.5 Bapx?1−/+ embryos revealed, in 78% of the individuals examined, mild abnormalities of the vertebrae, such as split or reduced ossification centers, which did not affect the overall length of the vertebral column nor the morphology of the neural arches. These defects were observed primarily in the lumbar vertebrae. This indicates that Bapx1 is haploinsufficient in some axial skeletal elements.
Using CMV-Cre transgenic mice (Nagy et al., 1998), Cre-mediated recombination of the Bapx1neo+ allele in vivo excised the inserted “floxed” PGKneo gene. Thus, heterozygous mice carrying a null mutant Bapx1neo− allele were established as an independent line and were interbred to produce mutant homozygotes. Analysis of homozygous mutant embryos from Bapx1 null heterozygous crosses at various developmental stages (E10.5-18.5) revealed that they were present in appropriate Mendelian ratios indicating that there was no significant embryonic lethality (Table 1). We observed an identical phenotype in Bapx1neo− homozygous and in Bapx1neo+ homozygous mice, suggesting that the mutant phenotype results from the loss of Bapx1 and not from the influence of PGKneo on other sequences. To ensure that Bapx1 was not expressed as a functional protein, we carried out RT-PCR assays with RNA extracted from homozygous Bapx1neo+ mutant and from wild-type mouse embryos. Homozygous mutant embryos lacked an amplified Bapx1 cDNA from the region spanning the Bapx1-coding sequences deleted in the mutant allele (Fig. 1C). In further confirmation of the functional disruption of the Bapx1 gene, RNA in situ hybridization on mouse embryo sections with the GIA115 riboprobe (probe 2, Fig. 1A) which is deleted in the Bapx1 null alleles, showed that no Bapx1 transcript was present in the homozygous Bapx1neo+ mutant embryos in two primary sites of Bapx1 expression: the sclerotome (prevertebrae) and the forelimb (Fig. 1D).
Disruption of Bapx1 gene results in perinatal recessive lethality
Analysis of the genotype of 10-day-old pups from numerous Bapx1 heterozygous mutant intercrosses did not reveal any Bapx1 null offspring. To characterize the time point of lethality in Bapx1 null mice, yolk sac or tail DNA from mice at various developmental stages (E10.5-18.5) and from newborn mice (either naturally born or from Cesarean delivery) was analyzed (Fig. 1B). The results (Table 1) suggested that Bapx1 null fetuses were not subject to prenatal lethality; however, we noticed that newborn Bapx1 null mice died within minutes of birth, probably as a result of failing to initiate normal respiration. Superficial examination of the Bapx1 null pups, revealed overall similarity in body weight relative to wild-type and heterozygous littermates but the Bapx1 null pups had severely truncated tails and appeared slightly shorter in stature and displayed a clearly distended circumference of the thoracic and abdominal areas (barrel chest appearance). Since no Bapx1 null pups survived birth, we concluded that loss of Bapx1 resulted in a recessive perinatal lethal phenotype. Identical phenotypes were obtained with mouse lines generated from each of the two independent ES clones 34 and 77 and in both C57BL/6J mixed and 129/SvJ inbred backgrounds and with neo present or following removal with Cre recombinase.
Cartilage and bone defects in Bapx1 null embryos
Anatomical examination and alcian blue/alizarin red staining of cartilage and bone at different stages of chondrogenesis and mineralization (E12.5-E18.5) revealed a dramatically hypoplastic axial skeleton in the Bapx1 null embryos (Figs 2A,F, 3A,F). The reduction in length of the vertebral column appears to be primarily the result of a reduction in the overall rostrocaudal length of the individual vertebrae. The total number of vertebrae appears approximately normal in the Bapx1 null embryos but, in all vertebrae, the vertebral bodies (vb) were hypoplastic, with only a small region of the dorsal vertebral body still present and with the complete absence of the ventral vertebral body. In all of the vertebral bodies at E18.5, we observed a total loss of ossification centers (oc, Fig. 2B,G), which was preceded, beginning at E12.5, with similar defects in chondrogenic condensations surrounding the notochord (Fig. 3). In addition, the neural arches (na) and intervertebral discs (id) were approximately 0.2-0.5 times the normal thickness and the notochordal remnant (n) was left exposed (Figs 2, 3). The ribs appeared normal but the thoracic cavity was distended radially. In the occipital and cervical region, the neural arches were reduced in size and displayed fissures (long arrow, Fig. 2). This defect was also observed for the exoccipital bone (eo, Fig. 2), which is the occipital equivalent of a neural arch. The anterior arch of atlas (aaa, Fig. 2) was completely absent. The floor of the cranial vault was highly affected in the Bapx1 null embryos, with either a reduction in size or a severe dysmorphology of the basisphenoid (bs) and basioccipital bones (bo, Fig. 2D,I). Note also the complete loss of the supraoccipital bone (so in Fig. 2C,H) as well as the total absence of the ossification center of the cartilage primordium of the body of the hyoid bone (arrows Fig. 2E,J, missing in Bapx1−⁄−). Because Bapx1 null embryos at E12.5 and E14.5 displayed identical defects in chondrogenesis of the axial skeleton (Fig. 3E,J) as were observed at later stages of fetal development, we concluded that the function of Bapx1 in chondrogenesis is required prior to E12.5. Interestingly, no skeletal defect was observed in the limbs at any stage (e.g. Fig. 3) despite that significant Bapx1 expression has been detected there (Tribioli et al., 1997; Tribioli and Lufkin, 1997).
Histological and immunohistochemical analysis of skeletal and spleen defects in Bapx1 null mutants
Histological examination of Bapx1 wild-type and null embryos at E12.5 and earlier stages, however, showed no alterations in the number or density of cells surrounding the notochord (Fig. 4A,B) suggesting that the defects in the Bapx1 null embryos are less related to cell migration than to cell differentiation. Cartilage production and chondroblast differentiation was examined in Bapx1 wild-type and null embryos using AR (alcian blue and nuclear fast red) or HGF (hematoxylin, fast green, and basic fuschin) staining, which are indicators of cartilage and collagen-associated proteoglycan (primarily aggrecan) production, respectively. Examination of the vertebral bodies (vb) in Bapx1 null embryos revealed complete agenesis of the ventral portion of the vertebral body (Fig. 4C-H) and faulty differentiation of the cells of the dorsal vertebral body, with the most affected cells positioned in closest association with the notochord (arrows, Fig. 4E-H). The mechanism for the agenesis and failed chondrogenesis of the vertebral bodies was investigate by analyzing Bapx1 wild-type and null embryos for alterations in cell proliferation and cellular apoptosis (Fig. 4I-L). To determine cell proliferation, DNA synthesis was examined by measuring 5-bromo-2′-deoxy-uridine (BrdU) incorporation into cells of Bapx1 wild-type and null embryos between the ages of E10.5 and E14.5. The results from this assay (Fig. 4K,L) showed no significant differences in the cell proliferation rates in the cells of the sclerotome/vertebral body or in adjacent tissues at any stage examined. In contrast, when apoptotic cell death was examined by analyzing the extent of oligonucleosomal DNA cleavage (TUNEL) in cells of Bapx1 wild-type and null embryos, a 2.7-fold increase in the number of cells undergoing apoptosis was observed in the developing vertebral bodies of the Bapx1 null embryos, whereas no alterations in apoptosis rates were observed in adjacent tissues of these same embryos (Fig. 4I,J). Another significant domain of expression of Bapx1 in the developing embryo is the lateral plate mesoderm surrounding the midgut (Tribioli et al., 1997). Bapx1 wild-type and null embryos were examined for alterations in midgut development and morphology. Surprisingly, no obvious defects in smooth muscle development were observed in the Bapx1 null embryos (Fig. 4M,N). In contrast, the Bapx1 null newborns displayed fully penetrant asplenia (Fig. 4O,P). Examination of earlier stage embryos (E11.5-E16.5) showed that the earliest absence of spleen precursor cells in the Bapx1 null embryos coincided with the timing of the normal appearance of the spleen anlage in wild-type embryos (E11.5), which is derived from a condensation of coelomic epithelium and underlying mesenchyme of the dorsal mesogastrium (Green, 1967), both cells types that normally express Bapx1 (Tribioli et al., 1997).
Molecular analysis of Bapx1 null mutants
Histological analysis of the Bapx1 null embryos suggested a defect in both growth and differentiation of the chondrogenic regions of the axial skeleton, particularly those regions that would subsequently undergo endochondral ossification. To extend this analysis and to investigate with molecular markers the Bapx1 null phenotype, we generated transgenic lines that drive expression of β-galactosidase in cartilaginous condensations and mature cartilage using the regulatory elements from the mouse pro alpha 1(II) collagen gene which has been shown to confer restricted expression to nearly all cartilaginous condensations in the developing embryo (Metsaranta et al., 1995; Zhou et al., 1995). β-galactosidase staining of Bapx1 wild-type and null embryos carrying the mouse pro alpha 1(II) collagen-lacZ transgene revealed an altered lacZ expression pattern in the axial skeleton, which is most evident in whole-mount-stained embryos in the tail region (Fig. 5A,B,E,F). In wild-type embryos, the mouse pro alpha 1(II) collagen-lacZ transgene shows clearly defined segmental expression in the developing vertebrae, which is absent in the Bapx1 null embryos (Fig. 5C,H). Transverse sections of the vertebral column in Bapx1 null embryos with the mouse pro alpha 1(II) collagen-lacZ transgene revealed a complete absence of β-galactosidase activity in the cells surrounding the notochord (n, Fig. 5I) which in wild-type embryos show strong β-galactosidase expression and which will subsequently form the vertebral body (vb, Fig. 5D).
Many transcription factors, signaling molecules and extracellular matrix molecules have been shown to play a critical role in normal development of the fetal skeleton. alpha 1(II) collagen (Col2a1) is one of the earliest markers for cells entering the chondroblast lineage (Aubin et al., 1995; Cancedda et al., 1995). The expression of alpha 1(II) collagen was downregulated in the ventral vertebral bodies of Bapx1 null embryos (Fig. 6A,B,E,F) and the expression of alpha 1(II) collagen in other regions of the axial skeleton showed a reduced expression pattern similar to the AR and HGF histological staining results for chondrogenic regions (described above, Fig. 4). The runt domain-containing transcription factor Osf2/Cbfa1 which is expressed in a common progenitor for chondroblasts and osteoblasts and is required for differentiation of the later (Ducy et al., 1997) was also downregulated in the Bapx1 null embryos (Fig. 6M,N,Q,R). This result was not surprising given the complete absence of vertebral body ossification in the Bapx1 null embryos. In a similar manner, expression of the cell-cell signaling molecule Indian hedgehog (Ihh) which is expressed in prehypertrophic chondrocytes and is a regulator of the passage of cells from the proliferative to prehypertrophic chondrocyte stage, was dramatically reduced in the Bapx1 null embryos (Fig. 6C,D). The transcription factors Pax1 and Mfh1 are absolutely required for proper differentiation of the sclerotome and, in particular, for the formation of the vertebral bodies (Iida et al., 1997; Wilm et al., 1998; Winnier et al., 1997). In addition, both genes have been shown to be induced by Shh signals from the notochord (Furumoto et al., 1999). Both Pax1 and Mfh1 are expressed at normal levels in Bapx1 null embryos; however, both genes showed slightly altered patterns of transcript distribution in the vertebral bodies (Fig. 6G-J), and this alteration appears related to the histological disruption of the vertebral body tissue, rather than to a regulatory effect. The zinc-finger transcription factor Gli2 and the cell-cell signaling molecule sonic hedgehog (Shh), both of which play a positive regulatory role in the induction of the sclerotome and normal skeletal development, are unaffected in the Bapx1 null background (Fig. 6K,L,O,P). The third fibroblast growth factor receptor, Fgfr3, which negatively regulates osteogenesis by inhibiting chondrocyte proliferation and differentiation shows widespread expression throughout the prevertebrae of E12.5 wild-type embryos; however, expression of Fgfr3 is absent from the prevertebrae of Bapx1 null embryos (Fig. 6U,V). The expression of bone morphogenetic protein 4, Bmp4, which is a regulator of somite differentiation, and is under the negative feedback control of Ihh (Aubin et al., 1995; Vortkamp et al., 1998) is normally restricted to the perichondrial region of E12.5 wild-type embryos. In the Bapx1 null embryos, Bmp4 expression is observed throughout all of the cells of the prevertebrae and is no longer restricted to the perichondrial region (Fig. 6W,X). Sox9 is normally expressed throughout all cartilaginous condensations of the developing preskeleton (Ng et al., 1997; Wright et al., 1995; Zhao et al., 1997) and has been shown to be absolutely required for cartilage formation in loss-of-function studies (Bi et al., 1999) and to be a direct regulator of alpha 1(II) collagen expression (Bell et al., 1997; Healy et al., 1999; Lefebvre et al., 1997; Ng et al., 1997). A decrease in Sox9 expression is observed in Bapx1 null embryos, in mesenchymal cells migrating to form the sclerotome surrounding the notochord, the precursor of the vertebral body (Fig. 6Y,Z). During subsequent development, Sox9 is downregulated in the ventral prevertebral body and in prechondrogenic cells surrounding the notochord (n, Fig. 6AA,BB). A probe specific to the 3′ end of the Bapx1 gene, which is still present in the Bapx1 null mutation, showed normal cell distributions and cell densities for the population of prevertebral cells expressing the Bapx1 null allele at E12.5 (Fig. 6S,T). Taken together, our results suggest that the developmental function of Bapx1 is dispensable for sclerotome migration and proliferation but is required for directing ventral sclerotomal cells towards a chondroblast pathway, and this effect is most pronounced in cells in closest apposition to the notochord. An increase in the rate of programmed cell death may account in part for the loss of tissue in the Bapx1 null embryos.
DISCUSSION
Bapx1 function during embryonic development of the skeleton and spleen
The Bapx1 null embryos die perinatally and show a dramatically reduced and unossified axial skeleton and asplenia. The absence of spleen precursor cells is observed from the earliest stages of initiation of the splenic anlage from lateral plate mesoderm. The absence of ossification centers in the ventral axial skeletal elements in Bapx1 null newborns is preceded at earlier stages by failed chondrogenesis of the same tissues. Since endochondral ossification proceeds upon a cartilaginous outline, the absence of ossification in the Bapx1 null offspring is likely related to earlier defects in chondrogenesis, although a direct effect upon a chondro/osteoblast precursor cannot be ruled out. The analysis of prechondrogenic cells (E10.5-12.5) of the embryonic axial skeleton in Bapx1 wild-type and null embryos showed no significant differences in cell migration patterns or cell density and the overall morphology of the prechondrogenic prevertebrae appeared identical. In addition, identifying cells that normally express Bapx1 with a RNA in situ probe against mRNA sequences that are still present in the Bapx1 null mutation, showed normal cell distributions and densities for the population of Bapx1−/−-expressing cells, suggesting that Bapx function is dispensable for sclerotome migration and early proliferation, but is required for some later step in skeletal development. Cartilage is the primary product of chondroblasts and is composed of two principal types of molecules, alpha 1 (II) collagen fibrils and proteoglycans. During subsequent stages of development, the cells of the vertebral body fail to differentiate into chondroblasts as assayed by alcian blue (AR) staining (specific for cartilage) or fuschin (HGF) staining (specific for proteoglycans), or when assayed for alpha 1 (II) collagen (Col2a1) mRNA expression. Furthermore, using a cartilage-specific mouse pro alpha 1 (II) collagen-lacZ transgene as a marker, we observed a loss of β-galactosidase expression in the vertebral bodies of Bapx1 null embryos, but not in wild-type embryos. Hence the cells of the prevertebrae in the Bapx1 mutants appear to fail to make the prechondroblast-to-chondrocyte transition. No significant differences were observed in the cell proliferation rates in the axial skeleton between Bapx1 wild-type and null embryos; however, an increase in the number of cells undergoing apoptosis was observed in the Bapx1 null prevertebrae. An increase in the rate of programmed cell death may account in part, for the loss of tissue in the Bapx1 null skeleton. An increase in apoptosis in cells that are unable to complete their normal developmental pathway has been observed in other systems (D’Mello, 1998; Milligan and Schwartz, 1997; Sanders and Wride, 1995) and, in particular, for certain developmental pathways involving homeobox-containing genes (Tiret et al., 1998).
The role of Bapx1 in the developmental program of skeletogenic genes
Osf2/Cbfa1 is a runt domain-containing transcription factor strongly expressed in all mesenchymal condensations of the E12.5-E14.5 skeleton in a common progenitor for chondroblasts and osteoblasts. Later in development Osf2 is expressed primarily in cells of the osteoblast lineage and not in differentiated chondrocytes (Ducy et al., 1997; Komori et al., 1997; Otto et al., 1997). Analysis of Osf2 in Bapx1 null embryos showed that its expression was downregulated in the axial skeleton. No alteration in Osf2 expression was observed in other regions, such as the limbs, which co-express Bapx1, but interestingly show no defects in the Bapx1 null mice. It remains to be determined whether Bapx1 is a regulator of Osf2 expression in a common chondro/osteoblast progenitor, or whether loss of Osf2 expression in these affected Bapx1 null cells results as a secondary effect of their inability to complete their normal developmental program. Ihh is expressed in prehypertrophic chondrocytes and has been shown to regulate the rate of chondrocyte differentiation by stimulating parathyroid hormone-related protein (PTHrP) expression in the adjacent perichondrium. This in turn inhibits the transition from proliferating to prehypertrophic chondrocytes in the internal chondrogenic region (Lanske et al., 1996; Vortkamp et al., 1996). While Ihh showed strong expression in the caudal vertebrae of wild-type embryos, no expression of Ihh was detected in the caudal vertebrae of Bapx1 null embryos. This effect was restricted to the axial skeleton, as there was no effect of Bapx1 on Ihh expression in other tissues such at the gut or limb. These results indicate that the population of Ihh-expressing prehypertrophic chondrocytes of the axial skeleton is affected in the Bapx1 mutant mice.
Fgfr3 is the third member of the Fgf polypeptide growth factor receptor family. Point mutations in Fgfr3 in mouse or human lead to retardation in bone growth, achondroplasia, reduced proliferation of cartilage and overall bone shortening (Cohen, 1998; Wilkie, 1997). The expression of Fgfr3 was decreased in the vertebral bodies of the Bapx1 null embryos, but Fgfr3 expression was not affected in other tissues, like the neural tube, that do not express Bapx1. Fgfr3 is believed to play a role in chondrocyte proliferation and the transition from proliferative to prehypertrophic chondrocytes (Colvin et al., 1996; Wang et al., 1999). Bapx1 may be controlling the entry of certain prevertebral mesenchymal stem cells into the chondrogenic pathway and the loss of Fgfr3 expression in the Bapx1 null embryos is consistent with this. Bone morphogenetic protein 4, Bmp4 is a member of the TGFβ super family of signaling proteins. The Bmps are homologues of dpp in Drosophila, and function via diffusion within the extracellular matrix and interact either with their cognate receptors or with antagonists such as noggin and chordin (reviewed in Graff, 1997; Hirsinger et al., 1998). In the mouse, Bmp4 is expressed within the perichondrial region of both the axial and appendicular skeleton and its expression in the perichondrium appears to be under the negative control of the Ihh signaling pathway (Naski et al., 1998). In the Bapx1 null embryos, the expression on Bmp4 was altered such that it was no longer strictly confined to the perichondrial region, but instead had assumed a much more widespread and homogeneous domain of expression, with transcripts detectable throughout all the remaining cells of the vertebral body. Since Ihh was downregulated in the Bapx1 null embryos, the upregulation of Bmp4 in these same tissues is consistent with the proposed negative feedback control of Ihh (Naski et al., 1998), although a more direct effect of Bapx1 on Bmp4 regulation may also exist.
The Mfh1 gene encodes a winged helix/forkhead domain transcription factor, which is normally expressed in paraxial and presomitic mesoderm, in the developing somites and in condensing sclerotome (Iida et al., 1997; Winnier et al., 1997). Mfh1 null mice have axial skeletal malformations in structures derived from cephalic and somitic mesoderm that are reminiscent of those observed in Bapx1 null mice (Iida et al., 1997; Winnier et al., 1997). Interestingly, no significant reduction in the expression of Mfh1 was observed in the Bapx1 null embryos; however, in both types of mutant animals a reduction in alpha 1 (II) collagen (Col2a1) expression was observed, suggesting that both of these transcription factors likely lay upstream of Col2a1. Sox9 is a high-mobility-group (HMG) domain transcription factor that is expressed in chondroblasts and other embryonic tissues and the onset of its expression in the embryo parallels the onset of Col2a1 expression (Wright et al., 1995; Zhao et al., 1997). Sox9 has also been shown to bind to critical elements in the Col2a1 enhancer and can trans-activate the Col2a1 enhancer in cell culture and transgenic animals and produce ectopic cartilage in vivo (Bell et al., 1997; Healy et al., 1999; Lefebvre et al., 1997; Ng et al., 1997). Embryonic stem cells that lack both copies of Sox9 are incapable of colonizing the chondroblast lineage and are excluded from all cartilages (Bi et al., 1999) and humans with only one copy of Sox9 suffer from campomelic dysplasia, a severe skeletal dysmorphology syndrome (OMIM 114290). In the Bapx1 null embryos, Sox9 expression is reduced in prechondroblasts of the sclerotome in early stage embryos and in cells of the ventral vertebral body in older embryos. The decrease in Sox9 expression is consistent with a decrease in Col2a1 expression in these same cells in the Bapx1 null embryos, suggesting that the effect of Bapx1 on Col2a1 expression may be mediated via changes in Sox9 expression levels.
Another transcription factor involved in sclerotome patterning is Pax1, which belongs to the paired box-containing gene family of transcription factors (Balling et al., 1996). The expression pattern of Pax1 in the developing vertebral column is very similar to Bapx1 at early stages where it is expressed in the presclerotome cells of the somite from E8.25 onward and then shows stronger expression in the caudal half of the somite by E10.5. The absence of vertebral bodies in the lumbar regions of the Pax1 null mice is strikingly similar to that of Bapx1 null animals. However, the morphological alterations in extravertebral components, like the sternum and the scapula, that are present in Pax1 mutants are not present in Bapx1 mutant mice. Overall the Bapx1 axial skeletal defects are more severe than those observed in the Pax1 null mice, as all of the vertebrae are dramatically affected in Bapx1 mutants, as well as the mesoderm-derived bones of the skull.
Human Bapx1 and redundant developmental function
The human homologue of Bapx1 has been identified and mapped to 4p16.1 (Tribioli and Lufkin, 1997) although pathogenic mutations have not yet been identified in this gene in humans. In murine Bapx1 null embryos, the skeletal defects are detectable at E12.5, this embryonic stage corresponds approximately to day 40 in humans. Based on the vertebral defects only, potentially the Bapx1 null embryo phenotype could have a human counterpart in an autosomal recessive form of neonatal lethal spondylodysplasia. However, the absence of limb defects in the murine null mutants makes it difficult to recognize the corresponding human disease in the group of the skeletal disorders that have been mapped by linkage to 4p16.1. Only Bapx1-expressing cells are affected in Bapx1 null mice, however, the morphological abnormalities are restricted to specific components of the axial skeleton and to the spleen. The concentration of skeletal defects in the Bapx1 null animals primarily to cells surrounding the notochord suggests a possible interaction between these cells and potential factors secreted from the notochord, such as Shh. Whether Bapx1 is regulated by Shh or whether other diffusible factors participate in the restriction of the Bapx1 null phenotype remains to be determined. Another significant domain of expression of Bapx1 in the developing embryo is the limbs (Tribioli et al., 1997; Tribioli and Lufkin, 1997), yet surprisingly, no obvious defect in limb morphogenesis or cellular differentiation was observed in the Bapx1 null embryos. The lack of an effect of Bapx1 may be the result of the expression of related Bapx family members that have yet to be identified (Sidow, 1996) or to a parallel regulatory network, which may functionally compensate for loss of Bapx1 in certain tissues.
Acknowlegment
We would like to thank Rudi Balling, Brigid Hogan, Mitch Goldfarb, Peter Gruss, Alex Joyner, Gerard Karsenty, Peter Koopman, Andy McMahon and Francesco Ramirez for providing RNA in situ probes. Andras Nagy for providing the CMV-Cre transgenic mice, Benoit de Crombrugghe for the mouse pro alpha 1(II) collagen-lacZ plasmid, and David Neustaedter and Maria Nikolova for technical assistance. The financial support of Telethon-Italy (Grant n.D.75) to C. T. is gratefully acknowledged.