Vegfa regulates perichondrial vascularity and osteoblast differentiation in bone development

During endochondral bone formation osteochondroprogenitors form cartilage templates of future bones. Following chondrocyte hypertrophy, perichondrial osteoblast precursors, endothelial and hematopoietic cells and osteoclasts migrate into primary ossification centers (POCs) and replace cartilage by bone marrow and trabecular bone (Karsenty, 2003; Kronenberg, 2003; Zelzer and Olsen, 2003).

The transcription factor osterix (Osx; Sp7 – Mouse Genome Informatics) is expressed in perichondrial osteoblast precursors and, at lower levels, in prehypertrophic chondrocytes (Nakashima et al., 2002). Osx expression first appears in the perichondrium of bone templates at embryonic day (E) 13.5 in mice. After E15.5, strong expression is associated with trabecular and cortical bone formation (Nakashima et al., 2002) as the precursors differentiate into collagen I (Col1)-expressing osteoblasts (Karsenty and Wagner, 2002; Maes et al., 2010b).

Several factors regulate endochondral bone formation, including Indian hedgehog (Ihh), Wnt/β-catenin, Notch, Bmps, Fgfs and Pthrp (Kronenberg, 2003). Ihh induces osteoblast differentiation during perichondrial maturation into periosteum (Chung et al., 2001; Hu et al., 2005; Long et al., 2004). Canonical Wnt/β-catenin signaling is essential for the differentiation of osteochondroprogenitors (Day et al., 2005; Hill et al., 2005) and promotes hypertrophic chondrocyte differentiation and osteoblast differentiation and maturation (Hu et al., 2005). Notch2 is a negative regulator of osteoblast differentiation (Hilton et al., 2008).

Vegfa is essential during key stages of endochondral bone formation. Mice with Vegfa conditionally deleted in collagen II (Col2)-expressing cells show a delay in blood vessel invasion into POCs, delayed hypertrophic cartilage removal and chondrocyte apoptosis (Haigh et al., 2000; Zelzer et al., 2002, 2004). Mice expressing only the non-heparin-binding Vegf120 isoform of Vegfa show decreased skeletal mineralization and reduced expression of osteoblastic markers (Maes et al., 2002; Zelzer et al., 2002). Overexpression of Vegfa in osteochondroprogenitor cells results in increased bone mass (Maes et al., 2010a).

Osx-positive osteoblast precursors not only migrate into POCs in response to Vegfa secreted by chondrocytes, but they also express high levels of Vegfa (Maes et al., 2010b). In this study, we show that Vegfa produced by these cells is required for blood vessel recruitment and early stages of osteoblast differentiation in perichondrial regions of bone templates. This function is mediated by Vegfr2 and is likely to involve Ihh, β-catenin and Notch2-dependent pathways.

To assess the function of Vegfa expressed by Osx-positive precursor cells in developing bone, we generated mice with conditional loss of Vegfa alleles in these cells. Newborn Vegfa^fl/fl; Osx-Cre:GFP mice had thinner bones and reduced skeletal mineralization compared with wild-type (WT) (Vegfa^+/+; Osx-Cre:GFP) mice (Fig. 1A,B). MicroCT showed reduced tibia length and that secondary ossification centers were largely missing in mutants at postnatal day (P) 9 (supplementary material Fig. S1A). Mutant femurs had increased hypertrophic zones, decreased mineralization and fewer tartrate-resistant acid phosphatase (Trap)-positive osteoclasts and GFP-labeled Osx-expressing (Osx/GFP⁺) cells (Fig. 1B). Anti-CD31 (Pecam1) staining indicated reduced density of endothelial and Osx/GFP⁺ cells within primary spongiosa in mutants (Fig. 1B).

Visualization of Vegfa expression by X-gal staining of P1 femurs from heterozygous Vegfa-lacZ^KI/WT (Vegfa-lacZ) mice revealed positive staining of prehypertrophic chondrocytes and cells located within primary spongiosa; by contrast, WT tissue showed only low-level false-positive staining in osteoclasts (supplementary material Fig. S1B). In situ hybridization indicated that loss of Vegfa expression in Osx⁺ precursors resulted in reduced Col1 (Col1a1)-expressing cells within primary spongiosa (Fig. 1C). Low Col1 expression in mutant bones confirmed that the majority of Osx/GFP⁺ cells are osteoblast progenitor cells, which are decreased in mutants (Fig. 1B,C).

To address mechanisms underlying reduced numbers of Osx⁺ cells in P1 femurs, we analyzed Osx/GFP⁺ cells during formation of POCs. E15.5 tibia of mutant mice showed markedly decreased numbers of Osx/GFP⁺ cells (43.2±4.9 per defined region of perichondrium and POC) compared with WT (56.4±6.5) mice (Fig. 2A). The area of anti-CD31 staining in POCs as a percentage of stained areas in perichondrium and POC combined was 0.6±0.5% for mutant compared with 23.4±5.5% for controls. Likewise, the area of Trap staining for osteoclasts in POCs as a percentage of total staining in perichondrium and POC combined was 27.7±8.5% for mutant compared with 53.6±17.2% in controls. This is consistent with published findings that Vegfa derived from Osx⁺ prehypertrophic chondrocytes attracts these cells into POCs (Maes et al., 2010b; Zelzer et al., 2002, 2004).

Loss of Vegfa expression in Osx⁺ cells resulted in decreased numbers of Osx/GFP⁺ cells in the perichondrial region of E14.5 tibia and reduced GFP levels in the whole hindlimb as assessed by western blotting. Area of anti-CD31 staining indicated reduced, but not significant, differences in the perichondrium of mutant tibia (Fig. 2B), whereas the area of anti-CD34 staining, a marker of hematopoietic/endothelial cells, was reduced (supplementary material Fig. S2A). This suggests that reduced perichondrial vascularity might contribute to the decreased numbers of Osx/GFP⁺ cells. BrdU and TUNEL stainings confirmed that proliferating and apoptotic cell numbers were unaffected in E14.5 Vegfa^fl/fl; Osx-Cre:GFP mice (supplementary material Fig. S2B,C).

X-gal staining of hindlimbs of E14.5 Vegfa-lacZ mice indicated that perichondrial cells of tibia and femur express Vegfa (Fig. 2C; supplementary material Fig. S2D). No false-positive cells were observed in the control, as osteoclasts are hardly present (supplementary material Fig. S2E). At E14.5, the perichondrium contains specific pools of osteoblast lineage cells, including osteochondroprogenitors (Col2⁺), osteoblast precursors (Osx⁺) and mature osteoblasts (Col1⁺) (Maes et al., 2010b). In situ hybridization confirmed co-expression of Osx and Col2 (Col2a1) in perichondrial cells (Fig. 2D), and perichondrial osteochondroprogenitors and their progeny were present in mice carrying Tomato and Col2-Cre transgenes (supplementary material Fig. S2F).

Next, we compared mice carrying floxed Vegfa alleles and Osx-Cre:GFP or Col2-Cre transgenes. Loss of Vegfa expression in Osx/GFP⁺ cells in Vegfa^fl/fl; Osx-Cre:GFP mice was confirmed by in situ hybridization (supplementary material Fig. S2G). Osx expression was reduced and Col2 expression appeared increased in tibia sections of E14.5 mutant mice (Fig. 2D), suggesting reduced differentiation of Col2⁺ osteochondroprogenitors into Osx⁺ cells. Probing sections for Col1 mRNA (supplementary material Fig. S2H) or Col1 protein revealed that loss of Vegfa in Osx/GFP⁺ cells or Col2⁺ cells resulted in slightly reduced Col1 expression in perichondrial areas (Fig. 2E; supplementary material Fig. S2I). Thus, Vegfa produced by Col2⁺ and Osx⁺ osteoprogenitors appears to primarily function as a stimulator of osteolineage differentiation.

Most effects of Vegfa are mediated by Vegfr1 (Flt1 – Mouse Genome Informatics) and Vegfr2 (Kdr – Mouse Genome Informatics; also known as Flk1) (Ferrara et al., 2003). Vegfr2 regulates chemotaxis, mitogenesis and cytoskeletal reorganization (Carmeliet and Collen, 1999); Vegfr1 primarily acts as a decoy receptor, although its function depends on the developmental stage (Ferrara et al., 2003). Overexpression of Vegfa in Col2⁺ osteochondroprogenitors enhances bone mass by mechanisms involving Vegfr2 and β-catenin pathways (Maes et al., 2010a). Staining tibia sections of E14.5 Vegfa^+/+; Osx-Cre:GFP mice with anti-Vegfr2 indicated abundant Vegfr2 expression in perichondrial Osx/GFP⁺ and likely Col2⁺ cells (Fig. 3A).

Conditional deletion of Vegfr2 in Vegfr2^fl/fl; Osx-Cre:GFP mice resulted in reduced numbers of Osx/GFP⁺ cells (Fig. 3A). Col1 levels appeared reduced in the perichondrium of Vegfr2^fl/fl; Osx-Cre:GFP (Fig. 3B) and Vegfr2^fl/fl; Col2-Cre:GFP mice (supplementary material Fig. S3A). Perichondrial regions of E14.5 femurs from Vegfr2^fl/fl; Tomato; Col2-Cre and Vegfr2^+/+;Tomato; Col2-Cre mice had similar numbers of Tomato-positive cells (supplementary material Fig. S3B). Therefore, Vegfa-dependent control of differentiation early in the osteoblast lineage appears to be mediated by Vegfr2. Vegfr1 is also abundantly expressed in perichondrial cells; however, conditional deletion of this receptor in osteoblast lineage cells had no clear effect on the numbers of Osx/GFP⁺ cells and Col1 levels in the perichondrium of E14.5 tibia (supplementary material Fig. S3C,D). Newborn Vegfr2^fl/fl; Osx-Cre:GFP and Vegfr1^fl/fl;Osx-Cre:GFP mice had no apparent defects compared with their littermate controls (supplementary material Fig. S3E).

Phosphorylation of Vegfr2, Akt and Gsk3β in hindlimb lysates of E14.5 Vegfa^fl/fl; Osx-Cre:GFP was decreased, whereas Mapk phosphorylation was unaffected (Fig. 3C). Furthermore, loss of Vegfa or Vegfr2 in Osx⁺ cells resulted in reduced β-catenin expression. Anti-β-catenin staining of tibia sections confirmed reduced β-catenin expression in perichondrial cells (Fig. 3D). Thus, Vegfa-induced Vegfr2 signaling and possibly also the β-catenin pathway regulate osteoblast differentiation in the perichondrium of developing bones.

Apart from Wnt/β-catenin, several other factors regulate endochondral osteoblast differentiation, including Bmps, Fgfs, Pthrp, Ihh (Colnot et al., 2005; Hu et al., 2005; Razzaque et al., 2005) and Notch2 (Hilton et al., 2008). To assess possible roles in Vegfa-dependent control of differentiation, we analyzed Ihh and Notch2 expression in hindlimb lysates from E14.5 Vegfa^fl/fl; Osx-Cre:GFP and Vegfr2^fl/fl; Osx-Cre:GFP mice. Loss of Vegfa or Vegfr2 in Osx⁺ cells resulted in decreased Ihh expression but increased Notch2 (Fig. 4A). Ihh is expressed in prehypertrophic chondrocytes and cells located in perichondrium during endochondral ossification (Hu et al., 2005; Mak et al., 2006). In situ hybridization of tibia sections from Vegfa^fl/fl; Osx-Cre:GFP and Vegfa^+/+;Osx-Cre:GFP mice indicated that Ihh expression was reduced in prehypertrophic mutant chondrocytes (Fig. 4B), whereas no Ihh expression was detected in mutant and control perichondrial cells (supplementary material Fig. S4A). Thus, reduced Ihh expression in prehypertrophic chondrocytes might directly affect osteoblast differentiation during perichondrial maturation. Anti-Notch2 staining of tibia sections showed increased Notch2 expression in perichondrial cells of mutants (Fig. 4C). These data suggest functional links between Vegfa/Vegfr2, Ihh and Notch2 in perichondrial osteoblast differentiation.

To test whether Vegfa directly regulates the expression of Ihh and Notch2, we used hindlimb cells from E13.5 Vegfa^fl/fl mice for in vitro differentiation assays. The majority of Vegfa-expressing cells showed nuclear Osx expression (Fig. 4Da). Adenoviral Cre-mediated Vegfa knockdown was confirmed by measuring cell-associated and secreted Vegfa levels (supplementary material Fig. S4Ba). Expression of the early and late osteoblast marker genes Runx2, Osx (Sp7) and Col1a1, and the mRNA levels of Ihh and the Ihh target genes patched 1 (Ptch1), Gli1 and hedgehog interacting protein (Hhip) were reduced upon loss of Vegfa expression (Fig. 4Db,c). Expression of Notch2 and target gene hairy/enhancer-of-split related with YRPW motif 1 (Hey1) was unaffected (supplementary material Fig. S4Bb), suggesting the possibility that cell-cell contact in vivo is important for regulation of Notch2 expression by Vegfa.

In rescue experiments, Osx expression was eliminated in the presence of the hedgehog (Hh) antagonist GANT-58, whereas recombinant Vegfa had only minor effects and Runx2 and Col1a1 levels were not affected (Fig. 4Dd). This raises the possibility that Vegfa might require one or more co-factors to exert its effect on osteoblast differentiation; alternatively, Vegfa functions via intracrine rather than paracrine mechanisms, as described in adult mice (Liu et al., 2012). We also tested the possible involvement of Gαs (Gnas – Mouse Genome Informatics), which is a modulator of Wnt/β-catenin and Hh signaling activities in mesenchymal progenitors (Regard et al., 2011, 2013). In hindlimb lysates, Gαs expression was not affected in Vegfa^fl/fl; Osx-Cre:GFP mice (supplementary material Fig. S4Ca). Furthermore, the activity of protein kinase A (Pka), a major regulator of Hh signaling, was unaffected in these mice (supplementary material Fig. S4Cb).

In summary, our data demonstrate that Vegfa produced by Osx⁺ precursors regulates blood vessel recruitment and early stages of osteoblast differentiation during perichondrial maturation. Vegfa-dependent effects are mediated by Vegfr2 and mechanisms that include stimulated expression and activity of Ihh, increased expression of β-catenin and inhibition of Notch2 (Fig. 4De). Since many factors are crucial for osteoblast differentiation during endochondral bone formation, Vegfa/Vegfr2-dependent regulation of osteoblast differentiation might involve additional mechanisms as well. Furthermore, in addition to an autocrine/paracrine role of Vegfa, changes in perichondrial vascularity indicative of a general developmental delay might also contribute to the effects on osteoblast differentiation.

Floxed Vegfa, Vegfr2 (Flk1) and Vegfr1 (Flt1) mice were generated at Genentech. 129-Vegfatm1.1Nagy (Miquerol et al., 1999), Osx-Cre:GFP (Rodda and McMahon, 2006) and Col2-Cre (Schipani et al., 2001) mice have been described previously. B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J mice were purchased from Jackson Laboratory. All animal experiments were performed with protocols approved by Harvard Medical Area Standing Committee on Animals and in accordance with US Public Health Service Policy on Humane Care and Use of Laboratory Animals. Further details of mouse strains and genotyping are provided in the supplementary Materials and Methods and Table S1.

Alizarin Red and Alcian Blue staining of skeletal structures in P1 mice and microCT analyses on hindlimbs are described in the supplementary Materials and Methods. Frozen sections (7.5 µm) were prepared for histology, including Hematoxylin and Eosin (H&E), von Kossa and Trap staining, and analyzed with a Nikon 80i upright microscope or a Nikon Ti w/spinning disk confocal microscope as described in the supplementary Materials and Methods. X-Gal staining to detect Vegfa-lacZ expression, and BrdU and TUNEL staining of proliferating and apoptotic cells are detailed in the supplementary Materials and Methods.

IHC on frozen limb sections, ICH on fixed cells in culture chambers and western blotting of protein lysates of E14.5 hindlimbs were carried out with the antibodies and imaging techniques described in the supplementary Materials and Methods.

Vegfa protein levels in cell lysates (cell associated) or culture medium (secreted) were assessed and normalized as described in the supplementary Materials and Methods.

Gene expression analyses by RNA in situ hybridization and qRT-PCR were carried out using the probes and primers detailed in the supplementary Materials and Methods and Table S2.

Hindlimbs of E13.5 embryos were enzymatically digested and cells infected with either Ad-Cre or Ad-GFP followed by osteogenic induction as described in the supplementary Materials and Methods.

Quantifications of femur diameter, mineralization and cell counts (Osx, Trap, CD31, CD34 and Col1 staining) were performed as described in the supplementary Materials and Methods. Results are presented as mean±s.d. and unpaired Student's t-tests were used. P<0.05 was considered significant.

We thank Yulia Pittel for secretarial assistance; Sofiya Plotkina and Karen Cox for technical support; Andras Nagy (Lunenfeld-Tanenbaum Research Institute, Canada) for providing Vegfa-lacZ mice; Napoleone Ferrara (UC San Diego) and Genentech for providing mice with floxed Vegfa, Flk1 and Flt1 alleles; Beate Lanske (HSDM) for providing Col2-Cre mice; and Valerie Salazar (HSDM) for advice on X-gal staining. We acknowledge services of the MicroCT Core at HSDM and Nikon Imaging Center at HMS.