Dchs1–Fat4 regulation of polarized cell behaviours during skeletal morphogenesis

Skeletal shape varies widely across species as adaptation to specialized modes of feeding and locomotion, but how skeletal shape is established is unknown. An example of extreme diversity in the shape of a skeletal structure can be seen in the sternum, which varies considerably across species. Here we show that the Dchs1–Fat4 planar cell polarity pathway controls cell orientation in the early skeletal condensation to define the shape and relative dimensions of the mouse sternum. These changes fit a model of cell intercalation along differential Dchs1–Fat4 activity that drives a simultaneous narrowing, thickening and elongation of the sternum. Our results identify the regulation of cellular polarity within the early pre-chondrogenic mesenchyme, when skeletal shape is established, and provide the first demonstration that Fat4 and Dchs1 establish polarized cell behaviour intrinsically within the mesenchyme. Our data also reveal the first indication that cell intercalation processes occur during ventral body wall elongation and closure.

P lanar cell polarity (PCP), the co-ordinated behaviour and polarity of cells within the plane of a tissue, is essential for the appropriate morphogenesis, and ultimately function of organs and tissues 1 . In Drosophila, PCP regulates cell behaviour within an epithelium and is controlled by two distinct pathways: Fz-Vang PCP, which includes the asymmetrically distributed Fz and Vang proteins, and Ds-Fat PCP, which includes the asymmetrically distributed Fat and Dachsous (Ds) proteins 1 Fat and Ds are large cadherin family proteins that engage in heterophilic binding between neighbouring cells. In Drosophila, gradients of Ds and Four-jointed expression result in different levels of Ds-Fat activity on opposing sides of each cell within the gradient. This asymmetry, in turn, establishes polarized cell behaviour, and the gradients ensure that the cells are collectively polarized. Mice mutant for Fat4 or Dchs1 are characterized by a wide range of defects in organogenesis, but in general, the underlying mechanisms that cause these defects are unknown 3,5 . These abnormalities include changes in the development of the sternum and vertebrae that are endochondral bones forming via a cartilage intermediate.
Development of the endochondral skeleton involves stepwise formation of the pre-chondrogenic mesenchyme, and commitment towards the chondrogenic lineage through the expression of the transcription factor Sox9. The developing chondrocytes become round and secrete matrix, and the cartilage condensation becomes surrounded by a fibroblastic layer, the perichondrium 6 . Skeletal shape is determined during these early phases of morphogenesis, but the molecular networks and processes that determine skeletal shape are in general unknown. In chicks and mice, Fz-PCP signalling has been shown to regulate the orientation of cells within the developing digit condensations, and the integration of proliferating chondrocytes into columnar stacks within the growth plate of long bones [7][8][9] , but whether Dchs1-Fat4 PCP polarizes cells within the skeletal mesenchyme is unknown.
The sternum arises within the lateral plate mesoderm, and in mice and chicks, the sternal progenitors have been identified in histological sections as a 'stream of flattened cells' that condense at the tip of the ribs (Fig. 1a) [10][11][12] . Sternal progenitors are first molecularly detectable around E12.5 within the closing body wall at the tip of the ribs by the expression of chondrogenic/osteogenic markers, Sox9, Runx1 and Runx2 (refs 13,14). By E15.5 the two sternal bands have met, overt chondrocyte differentiation is apparent, as characterized by alcian blue staining, and the perichondrium has formed (Fig. 1a). Ossification has occured by E16.5 and by P0, the fully differentiated sternum has formed (Fig. 1a). How sternum shape is established is unknown.
Here we show that Dchs1-Fat4 signalling acts via PCP to control polarized cell behaviours within the mesenchyme during the early phases of sternum morphogenesis. In wild-type embryos, sternal progenitors intercalate across the medio-lateral axis simultaneously narrowing the width of the sternum, while increasing its length and depth. These cell intercalation events are typical of convergent-extension processes that are regulated by Fz-PCP in other regions of the embryo. We show that these polarized cell behaviours are established by differential Dchs1-Fat4 expression and that mosaic ablation of Dchs1 disrupts cell polarity. This is the first demonstration that Dchs1-Fat4 regulate  Dchs1  (a) Diagram illustrating mouse sternum development between E12.5 and P0; the darker areas at E12.5 and E13.5 show the pre-cartilaginous condensations and the black regions at E16.5 and P0 indicate ossification of the intercostal regions. (b-i) Alcian blue and alizarin staining of P0 (b-e), E15.5 (f,g), E16.0 (h,i) sterna and ribs in wild-type (b,f,h), Dchs1 À / À (c,g), Fat4 À / À (d,i) and Dchs1 À / À Fat4 À / À (e) mutants; cartilage is stained blue, bone is red. (j-m) Ventral views of E12.5 wild type (j,l), Dchs1 À / À (k) and Fat4 À / À (m) sterna showing in situ hybridizations of Sox9 (j,k) and Runx1 (l,m) expression; the sternal plates are arrowed. (n-q) Flatmounts of E12.5 (n,o) and E13.5 (p,q), wild type (n,p) and Dchs1 À / À (o,q) sterna that have been immunostained for Sox9 (green), which is expressed in the sternal plate and ribs: the width of the sternal plate is indicated by the white bar in (n,o) and the sternum is outlined in (p,q); the ribs are numbered. (r-w) Haematoxylin and eosin stained sections through the dorso-ventral axis of the developing sternum at E12.5 (r,u), E13.5 (s,v) and E14.5 (t,w) in wild-type (r-t) and Dchs1 À / À (u-w) embryos; the sternum is outlined by the white dashes. r, rib. (z) Bar chart showing % change in width, length and depth in E14.5 Dchs1 À / À sterna compared with average sterna measurement in wild-type embryos that was standardized to 100%. The lines indicate mean (thicker black bar), 25 and 75%. Student's t-test: *Po0.05, ***Po0.001. Scale bars, 1 mM (b-e,j-m); 500 mM (f-i); 200 mM (n-q); 100 mM (r-w).
PCP within the mesenchyme and is the first evidence that cell intercalation occurs during ventral body wall development.
Analysis of chondrocyte cell density and differentiation at E14.5 revealed no significant changes between wild-type and Dchs1 À / À mutant embryos (Fig. 1t,w; Supplementary Figs 1t and 2c-g; Dchs1 þ / þ , n ¼ 3 and Dchs1 À / À , n ¼ 4). The changes in sternum shape must, therefore, be due to an altered arrangement of chondrocytes and cannot be due to differences in chondrocyte differentiation, such as reduced hypertrophy. However, there are changes in the timing and pattern of ossification of the sternum at later stages of development in mutants, which can be attributed to the altered dimensions of the sternum. In the wild-type sternum, ossification is inhibited by signals from the ribs that results in the formation of the separate intercostal regions 15 . Reflecting the alteration in sternum morphology, the timing of ossification was slightly delayed at E16.5 presumably due to the proximity of the ribs to each other in Fat4 À / À and Dchs1 À / À mutants (Fig. 1h,i; Supplementary Fig. 1a,b). By P0, the sternum was almost fully ossified in Fat4 À / À and Dchs1 À / À mutants with only discrete regions adjacent to the ribs remaining cartilaginous, which is again consistent with an inhibitory ossification signal from the ribs (Fig. 1c,d).
Immunohistological analysis of sternal flatmounts and tissue sections together with histological analyses showed there is a subtle increase in width at E12.5, although no obvious histological changes were identifiable and the sternal progenitors were found loosely organized at the tip of the rib in both wild-type and Dchs1 À / À or Fat4 À / À embryos (Fig. 1n,o,r,u; Supplementary   Fig. 1g,h,k,l). The phenotype is, however, clearly apparent by E13.5: in Dchs1 À / À and Fat4 À / À embryos the sternal condensation is wider along the medio-lateral axis and shorter along the dorso-ventral axis when compared with wild-type or heterozygous littermates (Fig. 1p,q,s,v; Supplementary  Fig. 1i,j,m,n,s). Furthermore, unlike in wild-type embryos, the sternal bar does not significantly narrow between E13.5 and E14.5 (Fig. 1s,t,v,w; Supplementary Fig. 1s; wild type, Po0.05; Dchs1 À / À not significant). These defects are before the formation of the perichondrium and overt chondrocyte differentiation as characterized by alcian blue staining and therefore occur within the early-condensing mesenchyme. In contrast, early development of the limb skeletal elements, which are also derived from the lateral plate mesoderm, is unaffected ( Supplementary Fig. 1q,r; Fat4 À / À , n ¼ 4).
Cell polarity is altered in Fat4 and Dchs1 mutants. The phenotype emerges within the condensing mesenchyme preferentially affecting the medio-lateral and dorso-ventral axes. In wild-type embryos, the sternum narrows between E13.5 and E14.5, while increasing in length along the dorso-ventral axis (Fig. 1s,t; Supplementary Fig. 1s), suggestive of polarized cell behaviours, such as those occuring during convergent extension 16,17 . Convergent extension describes the intercalation of cells towards each other, which narrows the width of the tissue while simultaneously increasing the length and/or thickness. Hallmarks of convergent extension within the gastrulating mesenchyme in Xenopus, where this process has been well characterized, include the flattening and elongation of cells across the medio-lateral axis and the formation of extended filopodia at the medial and lateral edges of the cells [17][18][19] . Orientated cell divisions (OCD) can also contribute to directed tissue growth by directing the cell divisions across one axis of the tissue. If convergent extension and OCD contribute to sternum narrowing and thickening in wild-type embryos, we would expect cells to be orientated along the medio-lateral axis to allow cell intercalation and cell divisions to be directed along the dorso-ventral axis, which would increase the depth of the sternum.
To analyse nuclear alignment in three-dimensions, confocal z-stacks were used to generate reconstructions of the nuclei in the wild-type and Dchs1 À / À embryos. The orientation and length of b c  Dchs1 -/-Dchs1 +/+ Fat4 E12.25 12 *** * the nuclei along the three axes, together with the shape (sphericity) and total nuclear volume, were then determined. This revealed that the mean sphericity of nuclei was comparable in Dchs1 þ / þ and Dchs1 À / À sterna (0.8775 ± 0.0032 for Dchs1 þ / þ n ¼ 3 and 0.8798 ± 0.0036 for Dchs1 À / À n ¼ 3, P ¼ 0.6395) and the average volumes of nuclei analysed also did not differ (277.9 ± 13.28 mm 3 in Dchs1 þ / þ and 274.7 ± 8.67 in Dchs1 À / À , P ¼ 0.847). However, as shown in the previous analysis, the orientation of the nuclei did differ. This was revealed by projecting nuclei onto the anterior-posterior, medio-lateral and dorso-ventral axes ( Supplementary Fig. 3a-f). In wild-type embryos, nuclear lengths are longer along the mediolateral axis (mean 9.37 ± 0.10 mm) than the anterior-posterior axis (mean 7.87±0.24 mm; Fig. 2i; Supplementary Fig. 3a,b). Conversely, in Dchs1 À / À embryos a preferential alignment along the anterior-posterior axis versus the medio-lateral axis was observed (mean 9.02 ± 0.11 versus a mean 8.34 ± 0.52 mm along the medio-lateral axis; Fig. 2i; Supplementary Fig. 3d,e). There were no significant differences in the length of the nuclei along the dorso-ventral axis ( Fig. 2i; Supplementary Fig. 3c,f).
To analyse the polarity and direction of the filopodia of sternal cells, another critical indicator of convergent extension, Dchs1/mT/mG mice were generated. In these mice, mosaic green fluorescent protein (GFP) expression can be induced by tamoxifen induction of Cre-recombinase allowing the visualization of isolated cells. The direction of the filopodia was scored and cells were assigned into the following three categories: cells that had filopodia that were predominantly directed along the mediolateral (M-L) axis, the anterior-posterior (A-P) axis or that had multiple randomly orientated protrusions. At E12.5 in both Dchs1 þ / þ and Dchs1 À / À mutants, the majority of the cells were oriented along the anterior-posterior axis and had multiple filopodia that projected randomly (Fig. 3a- The latter is characteristic of unpolarized mesenchymal cells. Nonetheless, differences in cell behaviour start to emerge at this stage as in Dchs1 þ / þ embryos significantly fewer cells had multiple randomly projected filopodia compared with Dchs1 À / À embryos ( Fig. 3q; Po0.05, n ¼ 3). Conversely, significantly more cells in Dchs1 þ / þ embryos had filopodia that were directed along the medio-lateral axis when compared with Dchs1 À / À mutants ( Fig. 3q; Po0.05, n ¼ 3). By E13.5, there were striking differences: in Dchs1 þ / þ sterna the majority of cells had filopodia orientated along the medio-lateral axis ( Fig. 3i-k,r; n ¼ 3), whereas in Dchs1 À / À embryos, the majority of cells still had multiple randomly projected filopodia ( Fig. 3m-o,r; Po0.0001). More cells in Dchs1 À / À sterna also had projections that were predominantly directed along the anterior-posterior axis when compared with wild-type embryos (Fig. 3r, Po0.0005).
In wild-type embryos at E13.5, sternal cells are also very closely packed and aligned across the medio-lateral axis (Fig. 3k,l): this 'alignment' of cells across the medio-lateral axis was also observed in the sections through the dorso-ventral axis at E13.5 and E14.5 ( Supplementary Figs 2c,e,f and 3h). Analysis of the actin cytoskeleton at E13.5 also revealed a preferential alignment of F-actin across the medio-lateral axis of the wild-type cells ( Supplementary Fig. 4a,c; Fat4 þ / þ , n ¼ 3). In contrast, cells and the F-actin cytoskeleton are less organized in Dchs1 and Fat4 mutants (Fig. 3o,p; Supplementary Figs 2d,g, 3i and 4b,d). In conclusion, in wild-type sternal cells, filopodia and the actin cytoskeleton become polarized, which in turn would be expected to alter cell shape, orientation and the relative nuclear dimensions.
Evidence of cell intercalation in the developing sternum. The above data are all consistent with cells undergoing convergent extension 17 and therefore, we also attempted to examine the   (Fig. 4). These cell movements would be predicted to decrease the width of the developing sternum along its medio-lateral axis.
Orientated cell divisions do not contribute to sternum shape. A combination of convergent extension and OCD are needed for the elongation of other embryonic structures, such as the kidney tubules and neural tube 17,19 . If OCD contribute to the morphogenesis of the sternum, we would predict that cell divisions would be directed along the dorso-ventral axis of the sternum, and in mutant sterna, this directed cell division would be lost or re-orientated along the medio-lateral axis. To determine the orientation of cell divisions, 4,6-diamidino-2-phenylindole staining was used to identify cells at terminal anaphase/early telophase ( Supplementary Fig. 5e). Five cells were identified (0.045%; 11,000 cells analysed), but there was no obvious bias in the orientation of cell division in wild-type embryos (Supplementary Table 1). Analysis of Dchs1 À / À and Fat4 À / À mutants also indicated that there is no predominant bias in the orientation of cell divisions (Supplementary Table 1). No OCD were identified at E13.5 either (Supplementary Table 1). This analysis indicates that OCD are very unlikely to contribute to the morphogenesis of the wild-type sternum and that an alteration in OCD cannot explain the widened sternum in Dchs1 À / À and Fat4 À / À mutants.
Gradients of Fat4 and Dchs1 determine cell polarity. We have previously shown that Fat4 and Dchs1 are expressed in opposing gradients across the hindbrain and that cells become polarized along these gradients 4 . This is reminiscent of the situation in Drosophila, in which gradients of Ds and a modulator of Ds-Fat binding, Four-jointed, direct Ds-Fat PCP 2 . To examine whether differential expression of Fat4 and Dchs1 could contribute to PCP in the sternum, the expression of Fat4 and Dchs1 was determined by immunolocalisation at E12.5 and E13.5, and their expression levels were quantified against either N-K-ATPase or b-catenin (Fig. 5a,c,e,g,i). Background antibody staining was determined by immunostaining Dchs1 À / À or Fat4 À / À sterna as appropriate (Fig. 5b,d,f,h). This analysis revealed that Dchs1 and Fat4 are differentially expressed along the medio-lateral axis of the E12.5 and E13.5 developing sternum, with Dchs1 expression highest laterally in the sternal cells near the ribs, whereas Fat4 expression is highest in sternal cells closest to the ventral midline (Fig. 5a,c,e,g,i). Quantification of the expression levels revealed evidence of opposing gradients of Fat4 and Dchs1 expression (Fig. 5i). These gradients would be expected to establish intracellular polarity within each sternal cell. Cells would have higher levels of Dchs1 on their medial sides where there are higher levels of Fat4 in the adjacent cells and conversely, higher levels of Fat4 on their lateral sides where the Dchs1 expression is highest (Fig. 5j). To confirm that Fat4 functions within the mesenchyme, Fat4 f/f Dermo1 cre were generated to specifically inactivate Fat4 expression within the mesoderm and not the overlying epithelium. Micro-computerized tomography scanning of Fat4 f/f Dermo1 cre P0 mice revealed they have a broadened sternum resembling Fat4 and Dchs1 null mice ( Supplementary Fig. 6a,b; n ¼ 3). We considered two potential models for how Dchs1-Fat4 polarize cells and the filopodia. In one model, Fat4 and Dchs1 may regulate cell adhesion, and cells may extend filopodia preferentially to regions of highest Fat4 and Dchs1 expression. In the second model, Dchs1-Fat4 gradients may establish an intracellular polarity (with different levels of Dchs1 and/or Fat4 on the medial versus lateral axis of the cell), which in turn determines cell polarity. The latter model is based on studies in Drosophila, where the establishment of Ds-Fat PCP involves the polarization of Ds and Fat proteins within each cell. Importantly, a cell with polarized Ds and Fat can serve as a template for a similar polarization of Ds and Fat in its neighbours [29][30][31] . This enables Ds-Fat PCP to spread throughout a tissue without requiring each individual cell to 'read' gradients of differential Ds expression. A characteristic feature of this type of polarization is that it is easily disrupted by small patches of mutant tissue, because this prevents the propagation of polarization. Indeed, in the hindbrain, we found that having even a small percentage of Dchs1 mutant cells was sufficient to severely disrupt PCP 4 . In contrast if filopodia extend to regions of highest Fat4 and Dchs1 expression, or Fat4/Dchs1 regulate cell adhesion, this would not be disrupted by the presence of a few mutant cells. To determine whether a gradient of Dchs1 establishes cell polarity during sternum morphogenesis, we generated mosaics in which Dchs1 was deleted in a Dchs1 þ / À background. Specifically, we generated Dchs1 f/ À /mT/mG embryos in which tamoxifen treatment would produce a mosaic of Dchs1 null GFPexpressing cells in a heterozygous background. These null cells would be predicted to disrupt cellular behaviours both cell and non-cell autonomously (Fig. 5j). The width and length of the sternum were analysed following Sox9 immunostaining at E14.5.
We found that mosaic disruption of Dchs1 following 2 or 4 mg of tamoxifen treatment resulted in an increase in the width of the sternum (Fig. 5k-m; Supplementary Fig. 6c,d). Treatment with 2 mg tamoxifen resulted in an average increase of 56% in width (n ¼ 6, Po0.01), while there was an average increase of 70% following treatment with 4 mg of tamoxifen ( Fig. 4m; n ¼ 9, Po0.005). The length of the sternum was also significantly reduced at the higher dose of tamoxifen ( Fig. 4m; Po0.05).
Significantly, a mosaic with as few as 3% GFP cells was 30% wider. These data clearly demonstrate that the presence of a few Dchs1 null cells causes the loss of polarized cell behaviour autonomously and non-cell autonomously.

Discussion
Here we have identified a new role for Dchs1-Fat4 PCP in vertebrates by establishing that Dchs1-Fat4 control cell orientation within the sternal mesenchyme to determine its final shape. In the developing sternum, cells polarize along the axis of differential Fat4 and Dchs1 expression, and mosaic disruption of Dchs1 perturbs sternal development, implying that differential expression of Dchs1 and Fat4 polarizes cells. In Drosophila, the graded expression of Ds and Fj results in a polarized sub-cellular localization of Ds and Fat, and this molecular polarization is responsible for directing cellular polarization; molecular polarization of Dchs1 protein has also been observed in the developing kidney 26 . We suggest that the polarization of sternal progenitors established by Dchs1 and Fat4 biases cell intercalations as sternal progenitors condense, to achieve the correct shape of the sternum.
Our results establish that Dchs1-Fat4 PCP can control cell orientation and behaviours within the mesenchyme, and that both vertebrate PCP pathways influence the shape and morphogenesis of skeletal structures. In Drosophila, Ds-Fat PCP has only been identified within epithelial cells. Dchs1-Fat4 are expressed in both epithelial and mesenchymal cell populations 5,32 , and our discovery establishes Dchs1-Fat4 as an important regulator of PCP within both epithelial and mesenchymal cell populations. Thus, like the Fz-Vang PCP pathway, in vertebrates the Fat-PCP pathway has also been co-opted to determine polarized cell behaviours within the mesenchyme in addition to the epithelium. In addition, as previously shown in the hindbrain, we have also shown that differential or gradients of Dchs1-Fat4 within the sternal mesenchyme determine cell polarity. Thus, we provide further support that mechanisms of Ds-Fat PCP are conserved between Drosophila and vertebrates. Dchs1 and Fat4 are widely expressed in mesenchymal tissues and such effects on cell behaviour are likely responsible for the diverse symptoms in human patients with mutations in DCHS1 or FAT4 (refs 24,33).
The Fz-Vang PCP pathway controls elongation of the long bones in chicks and mice, and pharyngeal arch cartilages in fish 7-9,34 and our studies here identify Ds-Fat PCP as a critical determinant of sternal shape in mice. Differential or combinational Fz-PCP and Fat-PCP could, therefore, be utilized to generate the different shapes of skeletal structures and may also be modulated to bring about evolutionary change. For example, our data raise the possibility that modulation of Dchs1-Fat4 signalling results in evolutionary adaptations of sternum. These evolutionary differences are particularly notable in avians, where the sternal complex has been modified to accommodate variations in locomotion and predatory behaviours. For example, the sternum can be 'square', as in ostriches and albatrosses, resembling the murine Dchs1 À / À and Fat4 À / À mutant sterna, or narrower and elongated, as in parakeet and penguins, more closely resembling sternal shape in wild-type mice. Thus, we speculate that modulation of the Dchs1-Fat4 PCP pathway could play a key role in morphological changes during evolution. Genotyping. For genotyping, DNA was extracted from tail snips or yolk sacs using reagent DirectPCR (Viagen) and Takara Ex-Taq PCR enzyme. Alternatively, DNA was isolated using Direct PCR-Ear Lysis reagent (PeqLab). The genotype was determined by PCR using allele-specific primers. Image and sternum analysis. The lengths of the medio-lateral, dorso-ventral and anterior-posterior axes were measured in E14.5 whole-mount (medio-lateral and anterior-posterior) or E13.5-E14.5 tissue sections (medio-lateral and dorso-ventral) between ribs 2 and 6. In whole mounts, the width was measured at every rib (2)(3)(4)(5)(6), and the average width was analysed. The length of the sternum was determined from r2 to r6. In tissue sections, the lengths were measured every 100 mm and averaged for each embryo. Significance was determined using the Students t-test. For the analysis of nuclei orientation/cell polarity, whole-mounted sternums were analysed between ribs 3 and 5. Circular analysis was performed using Oriana 4.0; sterna width was measured using AxioVision software by outlining the region of interest. The orientation of nuclei was analysed with CellProfiler2.0. Three-dimensional analysis of confocal z-stacks was performed using Imaris and nuclei were reconstructed based on SOX9 immunostaining. Surfaces were built following built-in protocols and the parameters adjusted accordingly. At least three embryos were analysed for each study with the exception of the cell orientation analysis along the dorso-ventral axis where n ¼ 2 ( Supplementary Fig. 3g). The numbers of embryos and cells analysed are stated within the text. Raos spacing test was used to determine the significance of the collective orientation of the nuclei, that is, whether nuclei orientation is random or is polarized. The Mardia-Watson-Wheeler test was used to compare the collective distribution/orientation of nuclei between wild-type and mutant embryos. Cell proliferation and cell death analysis was carried out in whole-mount sterna or on tissue sections and was analysed using Volocity (PerkinElmer) and ImageJ software.

Methods
Histology and immunolocalisation analyses. Alizarin red and alcian blue skeletal staining to whole mounts and haematoxylin and eosin staining to tissue sections were performed by standard procedures 37 . To orientate the sternum in wax (to cut perpendicular sections throughout the developing sternum), the ventral body wall was marked with carbon particles. For whole-mount antibody staining, E12.5 or older embryos were fixed in 4% paraformaldehyde/PBS for two to four hours at 4°C. Sternums were dissected, removing nearby skin and muscle, and washed with PBST (PBS, 1% bovine serum albumin (BSA), 0.1% Triton Â -100). Sternums were permeabilized in PBST overnight at 4°C, and then blocked in 5% donkey serum/ PBST (PBSTS solution) for 6 h, and then incubated with primary antibodies in PBSTS solution overnight at 4°C followed by overnight incubation at room temperature. Sternums were washed in PBST (five times, 1 h each), and then incubated with secondary antibodies in PBSTS solution following the same incubation procedure as for primary antibodies. Sternums were washed in PBST (five times, 1 h each), and then washed extensively (two times, 5 h each and then overnight) at room temperature before mounting. After antibody staining, whole sternums were mounted on slides with glass coverslips as spacers to prevent compression of the sternums. Alternatively immunolocalisation was carried out to paraffin wax sections. In brief, sections were dewaxed in xylene, processed through decreasing ethanol solutions into PBS. This was followed by antigen retrieval in citrate buffer (20 mM sodium citrate pH 6) by boiling for 10 min, sections were then washed in PBS, blocked for at least 1 h in 5% serum/PSBT and incubated overnight in the primary antibody at 4°C. Sections were then washed three times in PBS and incubated in the secondary antibody for at least 1 h at room temperature. This was followed by three more washes in PBS. Sections were mounted in Vectashield hardset. Primary antibodies used include: rabbit anti-Sox9 (Millipore, 1:400 or Santa Cruz 1:50), goat anti-beta-catenin (R&D Systems, 1:100), rat anti-Fat4 and Dchs1 (gift of T. Tanoue 38 , 1:200), mouse anti-phospho-histone H3 (cell signalling, 1:150) and active caspase 3 (cell signalling; 1:150). Secondary antibodies were from Jackson ImmunoResearch, and nuclei were stained with Hoechst 33,342 or 4,6-diamidino-2-phenylindole. Images were obtained using a Leica SP5 confocal.
Method for sternum culture and live imaging. B6.129(Cg)-Gt(ROSA) 26Sortm4(ACTB-tdTomato,-EGFP)Luo/J mice were crossed to FVB.Cg-Tg (CAG-cre/Esr1)5Amc/J mice (Jackson Lab). E10.5 pregnant mice were intraperitonially injected with tamoxifen at dose of 1.2 mg per mouse. The sterna were collected from GFP-positive E12.5 embryos, the skin and most of the rib was removed. Sternums were cultured on Costar Transwell inserts (#3,450) ventral side down in DMEM/F12 (Gibco, Cat. #11,320) with 10% fetal bovine serum plus antibiotics and antimycotics (Gibco, Cat. # 15,240). Live imaging was performed using PerkinElmer spinning disc confocal microscopy at 37°C and in 5% CO 2 . Images were taken at time intervals of 5 min for up to 10 h. z-Stacks (1.5 or 2 mm) were generated using a Â 20 objective lens with long working distance (3.5 mm) and water immersion. Both transmitted light and 488 nm laser were used to visualize sternums contour and individual GFP-positive sternum cells, respectively. Live images were analysed with Volocity software.