To direct the expression of Dll1 throughout somites including rostral somite compartments, a full length Dll1 cDNA under the control of the mesodermal specific cis-regulatory element (msd) was fused to the Dll1 minimal promoter 20 (Fig. 1A). Two independent and stable transgenic lines (Tg(msd/Dll1)1Ieg and Tg(msd/Dll1)2Ieg) were established by pronuclear injection. Transgenic mice showed tails with multiple kinks and a reduction in tail and body length with varying severity (Fig. 1B, C). Both transgenic lines exhibited comparable dysmorphologies of the axial skeleton with varying severity. 25% of the transgenic mice born had an externally visible alteration of the phenotype. 18 transgenic animals without externally visible transgenic phenotypes from both the transgenic lines were examined for morphological changes of the axial skeleton by X-ray imaging (Table 1). Abnormal bony fusions of vertebral bodies were found in the thoracic, lumbar and tail regions in 9 out of the 18 transgenic animals. Vertebrae with reduced rostro-caudal length in the thoracic or tail regions were evident in 7 out of the 18 transgenic mice without externally visible phenotype. In 6 out of the 18 animals the number of thoracic and lumbar vertebrae was altered as compared to wild-type littermates and in 1 out of the 18 animals the number of sacral vertebrae was changed. Neural arches were irregular (lacking the spinous process or without bony fusion in the midline) in 11 out of the 18 transgenic mice without externally visible phenotype changes. We did not find dysmorphologies of the vertebral column by X-ray imaging in one of the 18 transgenic animals. Bones of limbs and skull were not affected in any of the transgenic animals. The shortening of the vertebral axis in adult transgenic mice with severe transgenic phenotype was due to frequent fusions of vertebral bodies from thoracic to caudal regions and a reduction of the R/C length of vertebrae caudal to the cervical region (Fig. 1D, E). Thus, the Dll1 overexpression phenotype was characterized by various dysmorphologies of the axial skeleton (analysed in detail below) with high penetrance but varying severity.

Dll1 expression in wild-type and transgenic mouse embryos was examined from E7.5 to E10.5 by whole-mount RNA in situ hybridisation. At E7.5 no differences between wild-type and transgenic embryos were observed (Fig. 2A). Ectopic Dll1 expression in rostral halves of recently formed somites was first detected at E8.0 in transgenic embryos when somitogenesis is initiated and has lasted at least until E10.5 (Fig. 2B to 2E). Variations in the expression level of the transgene were observed but did neither correlate with one of the transgenic lines nor with a specific stage of development. Embryos from whole-mount in situ hybridisations were used for histological sections to determine Dll1 expressing cells. In wild-type embryos at E10.5, sections revealed significant staining in cells of the somitic epithelium of caudal somite halves (Fig. 2D, E). A weak staining was detected in the somitocoele adjacent to the caudal epithelium. In transgenic embryos, Dll1 expression was also detected in epithelial cells but without restriction to caudal somite compartments (Fig. 2D, E). Similar to wild-type embryos a weak staining was detected in the somitocoele.

Histological sections of paraffin embedded embryos at E10.5 revealed no differences between somites of wild-type and transgenic embryos (Fig. 2F). Nascent somites developed a normal epithelial layer surrounding the somitocoele. The size of somites was identical in five age-matched pairs of transgenic and wild-type embryos between E9.5 and E10.5 in raster electron microscopic images (data not shown). Thus, the over-expression of Dll1 in the paraxial mesoderm did not affect the early generation of somites with regards to epithelialisation, size and histological appearance at least until E10.5.

We analysed the expression of various rostral and caudal somite marker genes and Notch signalling targets by whole-mount in situ hybridisation of wild-type and transgenic embryos. From E9.5 to E10.5 we found no differences in the expression levels of Dll3, Jag1, Notch1, Notch2, Lfng, Uncx4.1, Hes5, Mesp1, Mesp2, Paraxis (Tcf15), Pax1, Pax9, Myf5, Epha4, Myog and Cer1 (Fig. 3A to 3J, and data not shown) 132122232425. The R/C marker and Notch pathway genes Notch1, Notch2, Mesp1, Mesp2, Myf5, Uncx4.1, Pax1, and Pax9 were analysed until E12.5 and did not reveal differential gene expression between transgenic and wild-type embryos (Additional file 1). Many of these genes also have dynamic patterns of expression in the PSM. In accordance with the normal and regular epithelial somites observed, we did not find evidence for a change in phased gene expression in the PSM of transgenic embryos in comparison to their wild-type littermates. In particular, the cyclic expression of Lfng in the PSM 2627 was not affected in Tg(msd/Dll1)Ieg embryos. Distinct phases of Lfng expression were evident in transgenic embryos at E10.5 (Additional file 2). Homozygous Tg(msd/Dll1)Ieg mice were generated. Dysmorphologies of the axial skeleton of these mice were not different from the dysmorphologies that were identified in heterozygous transgenic mice. We did, therefore, not undertake a gene expression analysis of homozygous transgenic embryos. We cannot exclude that even higher doses of ectopic Dll1 signal in the PSM may lead to a different phenotype with regards to cycling gene expression and the generation of R/C polarized epithelial somites.

Since the contribution of distinct somitic regions to the elements of the axial skeleton has been rather well described 1172829, we examined vertebrae morphology with the aim to identify those sclerotome and arthrotome derived cell types that contribute to malformed skeletal elements in Dll1 transgenic mice. In alizarin red and alcian blue stained skeletons of transgenic mice we observed that structures derived from the somitocoele were frequently missing (Fig. 4). In particular, costal heads of ribs were absent or strongly reduced in their thickness (Fig. 4B, C; asterisks) and IVDs were often missing resulting in the fusion of vertebral bodies (Fig. 4B, C; arrows). The articular processes of the neural arches were either missing in transgenic mice with strong phenotype (Fig. 4F, black arrowheads) or reduced and malformed (Fig. 4F; black and white arrows). Despite the dysmorphologies of articular processes we did not observe fusions between adjacent neural arches even in transgenic animals with severe phenotype (Fig. 4F, H).

Additional dysmorphologies in the axial skeleton of transgenic mice included split vertebral bodies with ectopic pseudo-growth plates (Fig. 4B, C; arrowheads) and open neural arches with missing spinous processes (Fig. 4E, F; white arrowheads). Pedicles and laminae of neural arches were always present and did not fuse between adjacent vertebrae (Fig. 4H).

To further characterise the loss of intervertebral joints and midline clefts of the axial skeleton, the developmental progression of this transgenic phenotype was monitored from E12.5 to the newborn stage using alcian blue and alizarin red stainings (Fig. 5). Pairwise chondrocyte condensations of the presumptive axial skeleton adjacent to the notochord appeared normal in transgenic embryos at E12.5 compared to wild-type littermates (Fig. 5A, B). In wild-type embryos at E13.5 the pairs of cell condensations have come into contact at the axial midline. Here they now form single, rather cylindrical cell condensations that periodically encircle the notochord (Fig. 5C). In transgenic embryos at E13.5 pairs of cartilage condensations were not yet merged at the axial midline and regions between successive presumptive vertebrae stained with alcian blue (arrowhead in Fig. 5D). In wild-type embryos at E14.5 morphogenesis had proceeded and a regular pattern of strongly alcian blue stained presumptive vertebral bodies and weaker stained presumptive IVDs was evident (Fig. 5E). Contrary, in transgenic embryos of the same age, the developing vertebral column stained uniformly with alcian blue (Fig. 5F). The extrusion of notochordal cells in transgenic embryos was incomplete such that the rod of axial mesoderm in the regions of vertebral bodies was thicker than in wild-type embryos (Fig. 5E, F). In vertebral bodies of newborn wild-type mice, alizarin red staining showed a single, central core of mineralised extracellular matrix (Fig. 5G). Mineralisation had progressed also in the early neural arches and morphogenesis had resulted in the formation of superior and inferior articular processes forming the intervertebral joints. In contrast, transgenic newborn mice showed two mineralised cores lateral to the axial midline in each vertebral body (Fig. 5H). Intervertebral discs were missing and adjacent vertebral bodies fused. Superior and inferior articular processes were missing in the newborn transgenic mice with severe phenotype (Fig. 5H). Thus, the loss of IVDs and the failure of sclerotome cells to merge at the axial midline was traced back as early as E13.5 in transgenic embryos.

In addition to altered cell differentiation of presumptive IVD cells in transgenic mice, we considered the possibility that the loss of IVD cells might result either from apoptosis of presumptive IVD cells or, alternatively, from over-proliferation of cells of the prospective vertebrae displacing future IVD cells. To analyze these alternatives, we performed Bromodeoxyuridine (BrdU) and TUNEL assays (BrdU: E12.5 (n = 4), E13.5 (n = 4); TUNEL: E12.5 (n = 8), E13.5 (n = 3), E14.5 (n = 6), E15.5 (n = 4) and E17.5 (n = 3)). No significant differences between wild-type and transgenic animals were observed in BrdU (Additional file 3) and in TUNEL assays (Additional file 4).

We performed Safranin O stainings on histological sections of paraffin embedded embryos to investigate cartilage development in prospective vertebral bodies (Fig. 6). These stainings revealed fewer hypertrophic chondrocytes in vertebral bodies of transgenic rather than of wild-type embryos (E16.5) (Fig. 6A, B; arrows). Ossification had progressed normally at E17.5 in wild-type animals but was delayed in transgenic embryos (Fig. 6C, D; arrows) and hypertrophic chondrocytes were dorsally displaced (Fig. 6D; arrow). RNA in situ hybridisations to detect Col10a1 gene expression, which is a marker for cells undergoing endochondral ossification 3031, also revealed fewer hypertrophic cells (data not shown).

Dll1 deficient embryos (Dll1^{tm1Gos/tm1Gos}) showed severe patterning defects in the paraxial mesoderm and died before E12.5 33. In these mutants, the R/C pattern of somites was not established and maintained as in wild-type embryos and nascent somites did not epithelialise normally. The requirement of Dll1 for the formation of the axial skeleton was also studied in a mouse model expressing a truncated and dominant-negative version of Dll1 (Dll1^dn) under the control of the msd cis-regulatory element (Tg(msd/Dll1^dn)) 35. In the axial skeleton such transgenic mice exhibited fusions of laminae of neural arches, reduction or loss of pedicles, split vertebral bodies as well as rostral homeotic transformations at the cervical-thoracic transition. Gene expression analyses revealed a reduction of Uncx4.1 expression in caudal compartments and an expansion of the expression domain of the rostral somite marker gene Tbx18. These data suggested a partial loss of caudal somite compartment characteristics in Tg(msd/Dll1^dn) animals. Despite the fact that Tg(msd/Dll1)Ieg animals exhibited dysmorphologies of the axial skeleton, Dll1 transgenic embryos did not display changes in R/C marker gene expression.

The midline defects in the axial skeleton, the vertebral fusions and dysmorphologies of ribs observed in Tg(msd/Dll1)Ieg animals were reminiscent of somitogenesis related phenotypes observed in other mutant mice. All of these mutant animals are, however, characterised by alterations of R/C marker gene expression. For example, the bilateral centres of ossification in the vertebral bodies of Paraxis 36373839, Dll3 404142 and Psen1 14254344 deficient mice as well as missing spinous processes in Uncx4.1^-/- 2324 and Dll3 deficient skeletons resemble the midline defects found in the Dll1 over-expressing mice. Proximal rib malformations and/or rib fusions occur consistently in Paraxis, Psen1, Dll3 and Uncx4.1 as well as in the Dll1 over-expressing transgenic mice. Fusions of vertebral bodies or precursors along the R/C axis and the loss of IVDs as described for Tg(msd/Dll1)Ieg animals were also found in Paraxis, Psen1 and occasionally in Uncx4.1 deficient mice. Our analysis of the Tg(msd/Dll1)Ieg mice clearly shows that these phenotypes can occur in mice with apparently normal R/C polarity of somites. The fact that rostral and caudal sclerotome derived structures were not missing in the Tg(msd/Dll1)Ieg mice additionally supports the hypothesis that R/C somite polarity is not affected and suggests an alternative origin of the axial dysmorphologies. A potential reason for the finding that transgenic Dll1 over-expression does not affect the polarity of somites maybe that the endogenous expression of Dll3 inhibts the ectopic activation of Notch signalling in anterior somite compartments 45.

In addition to midline defects, we observe the loss or reduction of IVDs, articular (zygapophyseal) joints and proximal ribs in the axial skeletons of Tg(msd/Dll1)Ieg animals. These phenotypic characteristics are present despite normal epithelialisation and R/C polarity in transgenic mice. One possible explanation for the transgenic phenotype affecting the intervertebral joints may be that the loss of restricted Dll1 expression results in a late defect in resegmentation of sclerotomal compartments. Alternatively, we noted that all the vertebral structures (IVDs, articular joints, and proximal ribs) that are affected in Dll1 over-expressing mice are located in a central position of the vertebral motion segment. Injection of single somite cells with fluorescent dye 46 and homotopical grafting experiments of quail and chicken somitocoele cells 1617 had previously suggested that somitocoele cells form a distinct somitic compartment, the arthrotome. During later development these cells are located at a central position of the vertebral motion segment forming vertebral joints, IVDs and the proximal ribs 18. Accordingly, the microsurgical removal of somitocoele cells from chick epithelial somites and preventing the epithelial cells from contributing to the somitocoele cell population, resulted in the loss of IVDs, fusion of vertebral bodies and the absence of intervertebral joints 19. These analyses on the immediate fate of somitocoele cells have consistently been performed in the avian system 17464748. A related study in mice showed Pax1 expression in somitocoele cells and the ventromedial sclerotome. Mice deficient for the Pax1 gene lack derivatives of the ventromedial sclerotome or have reduced IVDs, proximal ribs and articular processes 49. Considering the experimental evidence on the fate of somitocoele cells in the avian embryo and of Pax1 expressing cells in mice, we propose the hypothesis that the observed reduction or loss of central structures of the vertebral motion segment may be due to a failure in the development or specification of the arthrotome in Tg(msd/Dll1)Ieg transgenic embryos. However, we did not observe any changes in the expression of Pax1 in transgenic embryos. This observation together with the normal histological appearance of somites suggests that the Dll1 over-expression throughout somites and the PSM does not affect the initial formation of somitocoele cells. Instead we consider the possibility that the over-expression of Dll1 in epithelia of somites might affect the contribution of epithelial cells to the somitocoele 1946.

Tg(msd/Dll1)Ieg animals exhibited a delay in chondrocyte hypertrophy and endochondral bone formation in presumptive vertebral bodies. Mid-sagittal as well as lateral-sagittal sections through the vertebral column of transgenic embryos at E16.5 and E17.5 displayed less hypertophic chondrocytes that undergo endochondral ossification than in wild-type embryos (Fig. 6B, D). Previous lacZ reporter gene analyses in Dll1^+/tm1Gos animals 33 revealed no Dll1 expression in presumptive vertebral bodies, IVDs or the notochord 34. We therefore assume that chondrocyte maturation was inhibited in transgenic embryos due to the earlier sclerotomal over-expression of Dll1. These observations essentially confirm previous studies that have revealed that Delta/Notch signalling acts as a negative regulator for the transition from pre-hypertrophic to hypertrophic chondrocytes 505152.

The 1.4 kb msd – element was introduced upstream the Dll1 minimal promoter as previously described 2035, followed by a full length Dll1 cDNA cloned in frame with the start codon and the SV40 and PGK polyadenylation signals 32. The construct was isolated as fragment and microinjected into pronuclei of hybrid CByB6 fertilised eggs according to standard procedures. Two independent transgenic lines (Tg(msd/Dll1)1Ieg and Tg(msd/Dll1)2Ieg) were established and maintained on a C3H background.

Genomic DNA from yolk sacs or tail biopsies was used for PCR genotyping. With the primers 5'-CGATACCCAGGTTGTCTCC-3' (exon 6) and 5'-AGCACACTCATCTACTTCCAG-3' (exon 7) a 347 bp and a 225 bp PCR product for the wild-type and the transgene were amplified, respectively.

Skeletal preparations of embryos, newborn and adult mice were performed by alizarin red and alcian blue staining according to standard procedures 53. Safranin O staining on paraffin embedded sections was performed as described previously 54. H/E staining was performed according to standard procedures 55.

Whole-mount in situ hybridisation was performed according to standard procedures 5657. Embryos were fixed, dehydrated by ascending methanol series, bleached in 14% H₂O₂ for one hour and stored in 100% methanol at -20°C. Antisense riboprobes were generated using the DIG-RNA labelling system (Roche, Germany) and hybridisation was performed at 68°C over night. BM Purple AP substrate (Roche, Germany) was used for colour development at 4°C for 24 – 48 hours. After staining was complete embryos were post fixed in 4% PFA. For sections embryos were cryo-protected in 30% sucrose in PBS, OCT embedded, cryo-sectioned at 35 μm and mounted. The following probes were used: Dll1 32; Dll3 58; Jag1 59; Notch1 60; Notch2 61; Lfng 62; Uncx4.1 63; Hes5 64; paraxis 37; Mesp2 25; Pax1 65; Pax9 66; Myf5 ; Mesp1 67; Epha4; Myog; mCer; Col10a1.

Cell proliferation was analysed by detection of incorporated BrdU (Sigma, Germany) in mouse embryos of day 12.5 and 13.5 (E12.5 and E13.5) according to published protocols 2468. TUNEL assay was performed as described by Kingsley-Kallesen 69.