VEGF and its role in the early development of the long bone epiphysis


Michael J. F. Blumer, Department of Anatomy, Histology and Embryology, Division of Clinical and Functional Anatomy, Innsbruck Medical University, Müllerstrasse 59, A-6020 Innsbruck, Austria. T: + 43 512 900371120; F: + 43 512 900373112; E:


In long bones of murine species, undisturbed development of the epiphysis depends on the generation of vascularized cartilage canals shortly after birth. Despite its importance, it is still under discussion how this event is exactly regulated. It was suggested previously that, following increased hypoxia in the epiphyseal core, angiogenic factors are expressed and hence stimulate the ingrowth of the vascularized canals. In the present study, we tested this model and examined the spatio-temporal distribution of two angiogenic molecules during early development in mice. In addition, we investigated the onset of cartilage hypertrophy and mineralization. Our results provide evidence that the vascular endothelial growth factor is expressed in the epiphyseal resting cartilage prior to the moment of canal formation and is continuously expressed until the establishment of a large secondary ossification centre. Interestingly, we found no expression of secretoneurin before the establishment of the canals although this factor attracts blood vessels under hypoxic conditions. Epiphyseal development further involves maturation of the resting chondrocytes into hypertrophic ones, associated with the mineralization of the cartilage matrix and eventual death of the latter cells. Our results suggest that vascular endothelial growth factor is the critical molecule for the generation of the epiphyseal vascular network in mice long bones. Secretoneurin, however, does not appear to be a player in this event. Hypertrophic chondrocytes undergo cell death by a mechanism interpreted as chondroptosis.


In long bones of mammals, endochondral bone formation starts with the establishment of the primary ossification centre (POC) within the diaphysis, followed by the development of the secondary ossification centre (SOC) within the epiphysis. Between these two sites of ossification the metaphyseal growth plate is situated, a structure critical for the longitudinal growth of the bones (for review see Aszódi et al. 2000 and Olsen et al. 2000).

In murine species, an indispensable requirement for the development of the SOC is the formation of cartilage canals shortly after birth (Kugler et al. 1979; Holmbeck et al. 1999; Davoli et al. 2001; Lee et al. 2001, 2009; Álvarez et al. 2005a,b; Blumer et al. 2007, 2008b). The canals start off as invaginations of the perichondrium, invade the epiphyseal cartilage matrix, and contain vessels as well as perivascular cells. During progressing development these cells start to synthesize bone matrix (type I collagen) that is laid down onto non-resorbed calcified cartilage. Thus several small ossification nuclei are formed that finally coalesce into a large SOC (Álvarez et al. 2005a; Blumer et al. 2007; for review see Holmbeck & Szabova, 2006 and Blumer et al. 2008a).

Early formation of the canals is accompanied by the disintegration of the major components of cartilage matrix (type II collagen and aggrecan), and notably several matrix metalloproteinases (MMPs) are required to clear a path for both vessels and perivascular cells (Vu et al. 1998; Holmbeck et al. 1999; Zhou et al. 2000; Lee et al. 2001, 2009; Davoli et al. 2001; Stickens et al. 2004; Álvarez et al. 2005a; for review see Ortega et al. 2004; Krane & Inada, 2008). In particular, membrane-bound type-1 MMP (MT1-MMP = MMP 14) appears to be an important tool for the lysis of the cartilaginous components, as vascularized canals do not develop in the MT1-MMP knock-out mouse. Nevertheless, the epiphysis later ossifies by an alternative mechanism. However, the epiphyseal vascular defect leads to a malformation of the growth plate, eventually resulting in a reduced growth of the long bones (Holmbeck et al. 1999; Zhou et al. 2000). Recent studies have clearly demonstrated that MMP 13 located at the canal blind end, cleaves type II collagen, and MT1-MMP is needed to allow the enzyme to attain its mature, active form (Lee et al. 2009). Surprisingly, the MMP 13−/− mouse does not show a failure in the development of the skeleton, implying that the epiphyseal vascularization and formation of the SOC is hardly disturbed. However, mice lacking both MMP 9 and 13 have an impaired vascular recruitment in the epiphysis, and their bones are considerably shortened (Stickens et al. 2004; for review, see Ortega et al. 2004). Taken together, these studies clearly demonstrate that distinct MMPs are crucial in enabling a normal epiphyseal development.

Tartrate-resistant acid phosphatase (TRAP), normally expressed in multinucleated cells (chondroclasts and osteoclasts) that resorb the calcified extracellular matrix (Hollberg et al. 2002), is also encountered in the lumen of cartilage canals at the onset of their formation (Álvarez et al. 2005a; Blumer et al. 2008b). Remarkably, the TRAP−/− mouse has a phenotype equal to that of the MT1-MMP null and the MMP 9/MMP 13 double knock-out mouse (Hayman et al. 1996, 2000; Suter et al. 2001; Hollberg et al. 2002; Roberts et al. 2007). Consequently, it seems very likely that the TRAP enzyme is a further important player during epiphyseal bone development. However, its exact role has yet to be elucidated.

Whereas the processes governing cartilage disintegration during the formation of the SOC are predominantly well-established, the exact role of the angiogenic factors stimulating vessels to invade the epiphysis is still under discussion (Maes et al. 2004; Álvarez et al. 2005a; Blumer et al. 2008a). Vascular endothelial growth factor (VEGF), the prototypical angiogenic cytokine, attracts endothelial cells and was shown to be pivotal for vascularization of the POC (Gerber et al. 1999; Harper & Klagsbrun, 1999; Carlevaro et al. 2000; Maes et al. 2002; Grellier et al. 2009). However, its impact as an angiogenic stimulator in the epiphysis remains uncertain due to conflicting findings (Maes et al. 2004; Álvarez et al. 2005a). During formation of the POC, a first event is the cartilage hypertrophy, followed by mineralization of the extracellular matrix. Subsequently, hypertrophic chondrocytes start to synthesize VEGF, triggering capillary invasion from the outer connective tissue into the cartilage template. This process is then associated with the death of the terminal chondrocytes, resorption of the mineralized cartilage by chondroclasts, and formation of new bone. Contrary to this sequence of events, the formation of the vascularized cartilage canals occurs prior to hypertrophy and mineralization of the epiphyseal cartilage (Álvarez et al. 2005a). In rats, the latter study showed that VEGF is not expressed during early formation of the canals but is essential in later developmental stages for the establishment of the SOC and its bone marrow cavity. However, Maes et al. (2004) presented a model in mice suggesting that a short-term hypoxic condition in the epiphyseal core shortly after birth is the principal stimulus for VEGF synthesis. VEGF, existing as several isoforms, is then sequestered into the extracellular matrix, where it diffuses to the perichondrium and makes it possible for pre-existing vessels to grow into the epiphysis of the long bones. This plausible model has yet to be verified by further in vivo studies.

Secretoneurin (SN), a neuropeptide expressed in many neuronal, neuroendocrine and endocrine tissues under normal conditions, is highly conserved during evolution (Leitner et al. 1998). The protein represents a 33-amino acid polypeptide which is generated by proteolytic processing of secretogranin II, formerly also called chromogranin C. Under pathophysiological conditions, especially in the case of cellular hypoxia, SN expression can be induced in non-endocrine tissues such as muscle cells, pneumocytes or tumour epithelial cells (Fischer-Colbrie et al. 2005; Egger et al. 2007). In vivo investigations showed that muscle cells of ischaemic mouse hind limbs expressed increased levels of SN, inducing an angiogenic response (Egger et al. 2007). Because of the chemotactic properties of SN on vessel endothelial cells and smooth muscle cells in vitro, the close interaction of SN-containing nerve fibres with blood vessels, and the increased production of SN in the setting of tissue ischaemia, it is assumed that SN is an inducer of angiogenesis and therefore comparable to VEGF (Kähler et al. 1997; Kirchmair et al. 2004a,b; Troger et al. 2005). Based on the fact that decreased oxygen levels occur in the epiphyseal centre during early development (Maes et al. 2004), we wanted to examine whether an up-regulation of SN follows hypoxia. Thus, SN could be an additional player triggering the formation of the epiphyseal vascular network.

The present study focused first on the chronology of the mineralization and hypertrophy of the epiphyseal cartilage in the distal femur of mice. Secondly, we investigated the fate of the epiphyseal hypertrophic chondrocytes. Thirdly, we checked whether the TRAP enzyme is expressed before the onset of the formation of cartilage canals. Finally, we tested the model of Maes et al. (2004) and attempted to add new data on the role of angiogenic factors during the early vascularization of the epiphysis.

Materials and methods

Mice (C57Bl6) were obtained from the central laboratory animal facilities of the Innsbruck Medical University. Postnatal stages day (D) 4, 5, 6, 7, 8, 9, 10, 18 and 20 were used, and three animals per age group were investigated. Mice were anaesthetized with CO2, and killed by cervical displacement. Subsequently, the legs were amputated, and the femur was isolated from the tibia. However, at D 4 the bones were too small to be separated from each other. The distal part of the femur was examined.

Light (LM) and transmission electron microscopy (TEM)

The bones from mice aged 5, 6 and 7 days were fixed in 2.5% glutaraldehyde, 2% paraformaldehyde (PFA) buffered in sodium cacodylate (0.1 m), pH 7.4 for 4 h at room temperature and rinsed in the same buffer. They were postfixed in 0.5% osmium tetroxide, 1% potassium hexacyanoferrat III in distilled water overnight at 4 °C, rinsed and decalcified in 3% ascorbic acid in sodium chloride (0.15 m) for 12 h at 4 °C. This was followed by dehydration in graded ethanol series and embedding in Spurr’s epoxy resin. Ribbons of consecutive semithin sections (2 μm) were cut on a Reichert Ultracut S microtome (Leica Microsystem, Wetzlar, Germany) with a histo-jumbo-diamond knife (Diatome, Biel, Switzerland) (Blumer et al. 2002) and stained with toluidine blue for 20 s at 60 °C. Complete series of ultrathin sections (80 nm) were cut on a Reichert Ultracut S microtome with an ultra-diamond knife, mounted on dioxan-formvar coated copper slot-grids and stained with an aqueous solution of uranyl acetate (1%) for 30 min at 20 °C, followed by lead citrate for 3 min at 20 °C. The ultrathin sections were examined with a Philips CM 120 transmission electron microscope (FEI, Eindhoven, the Netherlands) equipped with a MORADA digital camera (Olympus SIS, Münster, Germany) and the software olympus tem imaging platform was used.

Tissue preparation for enzyme histochemistry, immunohistochemistry and in situ hybridization

Bones were fixed with 4% PFA in phosphate-buffered saline (PBS, 0.1 m) for 4 h, rinsed in the same buffer and decalcified as described before. Subsequently, the bones were dehydrated in graded isopropanol and xylene series and embedded in paraffin. Serial sections (6–7 μm) were made on an HM 355S microtome (Microm, Walldorf, Germany) and collected on SuperFrost®Plus slides. For histological purposes only, sections were stained with Masson’s trichrome.

Enzyme histochemistry (HC)

Tartrate-resistant acid phosphatase (TRAP) is an accepted marker for the chondro-/osteoclast lineage. To show TRAP activity, sections were deparaffinized, rinsed in PBS and incubated in a solution containing 50 mm sodium acetate (pH 5.2), 0.15% naphthol-AS-TR-phosphate, 50 mm sodium tartrate, and 0.1% Fast Red T.R. (Sigma Aldrich Chemie Gmbh, Taufkirchen, Germany) for 30–40 min at room temperature. Subsequently, the sections were rinsed in PBS.

Von Kossa staining was performed to visualize the mineralized cartilage matrix. Thus, several tissue samples were not decalcified after fixation. Sections were deparaffinized, rinsed in PBS, incubated in 5% Silver nitrate and exposed to light (60 Watt lamp) for 1 h. Subsequently, the sections were rinsed in distilled water, briefly incubated with 2% ascorbic acid (pH = 4.0) and fixed with 5% sodium thiosulfate. The sections were counterstained with haematoxylin.

Immunohistochemistry (IHC)

To evaluate programmed cell death, goat anti-human caspase 3 (catalogue no. sc-1226; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) was used. The sections were deparaffinized, rinsed in PBS, and endogenous peroxidase activity was blocked with 0.5% H2O2 in 30% methanol for 30 min in the dark. The sections were then washed in PBS, and heat-induced epitope/antigen retrieval with citric buffer (1 : 10, pH = 6.0) was performed for 5 min in a micro-wave oven at 400 Watt. After a cooling step of 20 min, the sections were washed in PBS, incubated with caspase 3 antibody (1 : 100, 0.1% Triton × 100 in antibody diluents) overnight at 4 °C, rinsed, and the unspecific sites were blocked with 10% bovine serum albumin (BSA) in PBS. The sections were eventually exposed to a secondary antibody (rabbit anti-goat IgG/HRP-conjugated 1 : 1000 in antibody diluents) (catalogue no. P0449; DakoCytomation, Glostrup, Denmark) for 4 h at room temperature. After washing in PBS, chromogenic detection of the antigen-antibody complex was performed with 0.05% 3,3-diaminobenzidine (DAB), 0.01% H2O2 in distilled water for 10–15 min in the dark.

For VEGF, a rabbit anti-mouse polyclonal antibody (1 : 250, 0.1% Triton × 100 in antibody diluents) (catalogue no. sc-152; Santa Cruz Biotechnology, Inc.) was used. The sections were deparaffinized, rinsed in PBS, and endogenous peroxidase activity was blocked as described before. The subsequent protocol comprised a proteolytic digestion step using protease 1 for 5 min (Ventana, Strasbourg, France), primary antibody incubation overnight at 4 °C and saturation of the unspecific sites with 10% normal goat serum (NGS) for approximately 20 min. This step was followed by the application of a secondary antibody (goat anti-rabbit IgG/HRP-conjugated 1 : 500 in antibody diluents) (catalogue no. P0448; DakoCytomation) for 4 h at room temperature, and antibody binding sites were detected as described before. Sections through samples (4% PFA fixed and paraffin embedded) of the human kidney and breast cancer served as positive controls.

For SN, a rabbit polyclonal anti-SN antibody was used. The generation of the antibody was described in detail by Kirchmair et al. (1993). In short, we used the same protocol as outlined for VEGF; however, the primary antibody dilution was 1 : 1000 in antibody diluents. Negative controls were obtained by substituting the primary antibodies with 0.1% Triton × 100 in antibody diluents. All negative controls yielded no labelling.

Double labelling (HC and IHC)

For double labelling, TRAP reaction was performed first, followed by IHC. The sections were counterstained with Gils’ haematoxylin, but several sections were not stained.

mRNA in situ hybridization (ISH)

For ISH we used a method similar to the procedure recently published by Illig et al. (2009). In doing so, the sections were deparaffinized in xylene overnight and rehydrated in a descending series of alcohol. Subsequently, they were incubated in HCl (2 m) for 30 min at 30 °C, rinsed in PBS, dehydrated in an ascending series of alcohol and incubated with chloroform for 5 min at room temperature. They were dried for 5 min, prehybridized with a HybriBuffer (Biognostik GmbH, Göttingen, Germany), 0.2% Triton × 100 in a humid chamber for 3 h at 50 °C and then hybridized with the 6-carboxyfluorescein (FAM)-labelled probes (180 pmol probe mL−1 HybriBuffer, 0.2% Triton × 100) at the same temperature for 15 min and at 30 °C overnight. Single-stranded antisera DNA oligomer probes, 30 bases in length, were purchased from Microsynth (Balgach, Switzerland) and had the following sequences: VEGF-A 1753 5′-AGG ACT GTT CTG TCA ACG GTG ACG ATG ATG-3, VEGF-A 1894 5′-ATA AGA AAA TGG CGA ATC CAG TCC CAC GAG-3′, VEGF-A 2114 5′-CTC CCA ACA CAA GTC CAC AGC AGT CAA ACA-3′. After hybridization a stringent wash treatment in 0.5× SSC was performed and unspecific binding was blocked with 10% BSA in PBS. Probes were made visible by incubation with a rabbit anti-FITC/alkaline phosphatase-conjugated antibody (1 : 100 in PBS) (catalogue no. A4843; Sigma Aldrich Chemie Gmbh) for 3 h at room temperature, and chromogenic reaction was obtained using a ready-to-use nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate substrate (NBT/BCIP) (DakoCytomation) for 30–45 min. The sections were not counterstained. Controls for staining specificity were performed by addition of 1000-fold excess of non-FAM-labelled oligonucleotide probes, which led to an entire loss of the signal.

All sections were examined with a Zeiss Axioplan 2 (Zeiss, Oberkochen, Germany) and photographed as colour images using Zeiss AxioCam HR and AxioVision 4.1. software running on a Pentium 4 (Intel Inc., Santa Cruz, CA, USA) with WindowsXP (Microsoft Inc., Redmond, WA, USA).


Light microscopy and transmission electron microscopy

Histology of the epiphysis

At the fourth postnatal day (D 4), the distal femoral epiphysis of mice comprised exclusively cartilaginous tissue. Five days after birth, several short cartilage canals appeared, and these canals remained non-branched for 2–3 days. TEM observations clearly revealed that during the early development (D 4–6) the epiphysis consisted solely of resting cartilage. The resting chondrocytes were spherical-shaped and contained large amounts of rough endoplasmic reticulum, a distinct Golgi apparatus and numerous mitochondria (Figs 1A and 2A). However, in mice aged 7–9 days the histology of the epiphysis started to change and was composed of resting and mineralized hypertrophic cartilage (Fig. 1B–D). Cartilage canals were highly ramified within the hypertrophic area, whereas their segments within the resting cartilage remained non-branched. The was no clear zone of flattened proliferative chondrocytes arranged in columns and located between the resting and hypertrophic zone (Fig. 1E,F). Thus, resting chondrocytes spontaneously differentiated into hypertrophic chondrocytes, which finally appeared to undergo cell death. The transitional phase of this developmental process was accompanied by a disintegration of several cell organelles that eventually led to an electron-translucent cytoplasm with a distinct area of numerous lysosomes, few mitochondria and pieces of rough endoplasmic reticulum. However, other cell organelles were hardly discernible in the hypertrophic chondrocytes. Ultrastructural findings further revealed shrinkage of the cells, a patchy distribution of the nucleus chromatin, the presence of autophagic vacuoles, and blebbing of the cellular fragments into the extracellular space of hypertrophic cell lacuna (Fig. 2B–F). Light-microscopical investigations based on consecutive semithin sections (2 μm) did not demonstrate any signs of empty lacunae in mice aged 7 and 10 days, respectively. As an example, we demonstrated consecutive sections through the hypertrophic zone of a mouse aged 10 days (Fig. 3A–F). During advancing development the resting zone became thinner and the area of mineralized hypertrophic cartilage increased markedly. Progressing development of the epiphysis led to the formation of a massive SOC, and in mice aged 18 and 20 days histological differences were seen in the cartilage immediately surrounding the epiphysis. The distal border (facing the articular cartilage) of the SOC comprised hypertrophic chondrocytes, whereas the proximal border (facing the metaphysis) was exclusively composed of roundish, resting chondrocytes (Fig. 1G–I). Throughout development (D 4–20) the epiphyseal vessels consisted of endothelial cells only (Fig. 1E, inset); they did not form layers of smooth muscle cells. The metaphyseal growth plate was situated between the POC and the SOC. This well-organized structure consisted of resting as well as columns of proliferating and hypertrophic chondrocytes (Fig. 1G).

Figure 1.

 LM micrographs. (A–D) (von Kossa reaction on paraffin sections): Latero-medial sections through the distal part of the femur at different developmental stages (D 6, 7, and 9). Von Kossa reaction indicates the mineralized extracellular matrix (black staining). (A) At D 6 the epiphysis comprises solely resting cartilage (rz), and mineralized extracellular matrix is only visible within the primary ossification centre (poc). Inset (semithin section stained with toluidine blue) shows the roundish chondrocytes of the resting zone (rz); scale bar in the inset = 20 μm. (B) Within the epiphysis, mineralized cartilage is seen for the first time in mice aged 7 days. It is associated with the hypertrophy of the chondrocytes. (C–D) At D 9 the mineralized area has increased. (D) A cartilage canal (cc) with its ramified segment in the calcified hypertrophic zone (hz) is depicted. The extracellular matrix of the resting zone (rz) remains uncalcified; scale bar = 50 μm. (A–C) Same magnification; scale bar = 500 μm. (E,F) Semithin resin cross-section (2 μm) through the epiphysis stained with toluidine blue. (E) By D 7 cartilage canals (cc) are highly branched within the hypertrophic zone (hz), but the segment with the resting zone remains non-ramified; scale bar = 50 μm. Inset (higher magnification) shows that the canal vessels (red arrowheads) are solely composed of endothelium cells; scale bar = 20 μm. (F) No zone of flattened chondrocytes is encountered between the resting (rz) and the hypertrophic zone (hz); scale bar = 20 μm. (G–I) (Frontal paraffin sections through the distal part of the femur of mice aged 18 and 20 days, respectively. The sections are stained with Masson’s trichrome). (G) A large secondary ossification centre (soc) is seen within the epiphysis. The metaphyseal growth plate (gp) is located between the SOC (soc) and POC (poc); scale bar 500 μm. (H) The adjacent zone of the SOC distal border is formed by hypertrophic chondrocytes (hz), whereas the articular cartilage comprises resting chondrocytes (rz). (I) Proximally, the SOC is surrounded by resting chondrocytes (rz) followed by the proliferative zone (pz) of the metaphyseal growth plate. (H,I) Same magnification; scale bar = 50 μm.

Figure 2.

 TEM micrographs of mice aged 6 (A) and 7 days (B–F), respectively, show the transformation process from a resting (A) to a dying hypertrophic chondrocyte (F). (A) A typical resting chondrocyte (rCh) within its lacuna is shown. The cell is spherical in shape, contains a large amount of rough endoplasmic reticulum (rer) and has an equally roundish nucleus (n); scale bar = 5 μm. (B,C) During the transformation process the cytoplasm of the resting chondrocytes is altered, leading to a large homogeneous area (asterisks) and a less prominent rer as compared with (A). Mitochondria (m) and a distinct Golgi apparatus (g) are still noticeable; scale bars = 5 and 1 μm. (D,E) Hypertrophic chondrocytes (hCh) appear for the first time 7–8 days after birth, with an electron-translucent cytoplasm and numerous lysosomes (ly). They also contain several mitochondria (m) and fragments of rough endoplasmic reticulum (rer). The cells are surrounded by a mineralized cartilage matrix indicated by the asterisks; scale bars = 5 and 1 μm. (F) Dying hypertrophic chondrocytes appear shrunken and are detached from the lacunar wall. In addition, their cell fragments are sequestered into the lacunar cavity; scale bar = 5 μm.

Figure 3.

 LM micrographs. (A–F) Consecutive semithin cross-sections (2 μm, stained with toluidine blue) through the hypertrophic zone (hz) where numerous chondrocytes in different developmental stages are depicted. A photographic sequence of six successive sections that have the same magnification as well as orientation is shown. Some hypertrophic cells are detached from the lacunar wall and have already begun cell death, but others appear to be intact and fill their lacuna. Arrowheads in different colours point to the course of several chondrocytes. (C) The green arrowhead indicates an empty lacuna, but the following sections clearly show that it is occupied by shrunken cell, which presumably is assigned to die; scale bar = 20 μm.

Double labelling: immunohistochemistry and histochemistry

Caspase-3 and TRAP

To verify the ultrastructural observations implying that the epiphyseal hypertrophic chondrocytes underwent cell death starting at D 7, we examined whether these cells expressed caspase-3, an established marker for apoptosis (for review, see Cohen, 1997; Sabbagh et al. 2005). Caspase-3 immunoreactivity was noted in several hypertrophic chondrocytes adjacent to the cartilage canals at D 7 and 9. Furthermore, TRAP cells were observed within the canal lumens (Fig. 4A,B). In older animals (D 18, 20) caspase-3 staining was visible in numerous hypertrophic chondrocytes that were regularly distributed at the distal border of the SOC. However, on its proximal border considerably fewer resting chondrocytes displayed staining (Fig. 4C). Throughout development, many hypertrophic chondrocytes of the metaphyseal growth plate stained positively, whereas the chondrocytes of the proliferative zone showed no labelling (Fig. 4D,E).

Figure 4.

 IHC and HC: Caspase 3 (brown staining) and TRAP (red staining). (A–C) Latero-medial sections counterstained with haematoxylin through the distal femoral epiphysis. (A,B) In mice aged 7, 8, and 9 days, several caspase 3-positive chondrocytes are noted around the ramified endings of the cartilage canals (cc) within the hypertrophic zone (hz). Additionally, multinucleated TRAP-positive cells are present in the canal lumen; scale bars = 20 μm. (C) The epiphysis of a 20-day-old mouse is depicted. At the distal margin of the SOC (soc) numerous hypertrophic chondrocytes (hz) stained positively, whereas on the proximal chondro-osseous junction only a few resting chondrocytes (rz) are labelled. This staining pattern was also encountered in the 18-day-old mouse; scale bar = 100 μm. (D,E) Longitudinal section through the proliferative (pz) and hypertrophic zone (hz) of the metaphyseal growth plate of a 18-day-old mouse. Only the hypertrophic chondrocytes are immunoreactive for caspase 3. Numerous multinucleated chondroclasts (chc) and vessels (v) are found at the chondro-osseous junction of the POC. These labelling patterns were identical for all stages investigated (D 7–20). (E) A higher magnification of D; scale bars = 20 μm.

To investigate which molecules mediate the invasion of the blood vessels into the epiphysis we first focused on the distribution of VEGF and then on SN.


Examination of serial sections through the epiphysis of mice aged 4 days demonstrated that immunoreactivity for VEGF was encountered in the perichondrium and in certain areas of the resting cartilage immediately below it. In addition, a few small TRAP-positive cells could be detected in the perichondrium (Fig. 5A,B). However, chondrocytes in the centre of the epiphyseal cartilage displayed only a very faint staining for the growth factor (Fig. 5C). During progressing development (D 5), cartilage canals invaded the epiphysis, and VEGF was immunolocalized in several canal cells. Furthermore, a few mononuclear TRAP-positive cells were seen within the canal lumens. Serial sections revealed that some of these cells were present at the advancing tips of the canals, but that in general the TRAP cells were not distributed according to a distinct pattern; rather, they were scattered within the canal matrix. Expression of VEGF was noted in several resting chondrocytes in close proximity to the apical tips of the canals (Fig. 5D,E). There were no differences in the staining patterns of mice aged 5 and 6 days. As the staining intensity appeared to be different among chondrocytes, counterstaining with haematoxylin was omitted to detect weakly labelled cells (Fig. 5F–H). By D 7 the ends of the cartilage canals were intensively ramified within the region of the hypertrophic cartilage. Cells positive for either VEGF or TRAP were located within the canal lumens. Furthermore, numerous resting and the majority of hypertrophic chondrocytes expressed the growth factor. These staining patterns were also observed in 9-day-old mice (Fig. 5I,J). In older mice (D 18 and 20) multinucleated TRAP cells resorbed the remnants of the mineralized cartilage matrix within the SOC, and immunoreactivity for VEGF was seen in a small portion of the osteoblasts. The angiogenic factor was also detectable in the hypertrophic chondrocytes on the distal side as well as the resting chondrocytes on the proximal side of the SOC. However, a great number of the latter cells revealed only weak staining (Fig. 6A–D). Within the metaphyseal growth plate, a distinct expression of the VEGF was visible in all hypertrophic chondrocytes, whereas no staining was seen within the proliferative zone (Fig. 6C–E). Finally, numerous osteoblasts of the POC (Fig. 6E, inset), cells of the perichondrium as well as periosteum, were VEGF-positive. These staining patterns were discernible in all developmental stages (D 4–20). Throughout development, VEGF was noticeable in the skeletal muscle cells but the staining decreased with progressive age (data not shown). Human kidney tissue samples (PFA-fixed and paraffin-embedded) were used as a positive control, and VEGF was observed in the cells of the glomerulus and the tubules (Fig. 6F).

Figure 5.

 IHC and HC: VEGF (brown staining) and TRAP (red staining). The micrographs exhibit longitudinal sections through the femoral epiphysis and demonstrate the spatio-temporal distribution of both VEGF and TRAP during advancing development. (A–C) (D 4). (A) VEGF is identified within the perichondrium (p) and in certain areas of the resting zone (rz) directly below it. The green arrowheads delineate an accumulation of the VEGF-positive resting chondrocytes. The section is not counterstained with haematoxylin; scale bar = 50 μm. Inset: the section is counterstained, and the yellow arrowhead points to a TRAP cell within the perichondrium; scale bar = 20 μm. (B) A higher magnification of the area depicted by the green arrowheads (panel A) is shown. The cells of the perichondrium (p) as well as the resting chondrocytes below it display a strong immunoreactivity for VEGF. (C) The resting chondrocytes in the centre of the epiphysis exhibit only a weak staining for the growth factor. (B,C) Same magnification; scale bar = 20 μm. (D,E) Consecutive sections (same magnification) through the apical tip of a cartilage canal at D 5. Within the canal lumen few VEGF- (black arrows) and TRAP-positive cells are seen. Furthermore, several resting chondrocytes ahead of the canal tip are immunostained with the growth factor (green arrowheads); scale bar = 20 μm. (F–H) Mice aged 6 days have the same staining pattern as described before. (F) An overview of the epiphysis with its growth plate is shown. VEGF is detected in the hypertrophic zone (hz) and in several resting chondrocytes around a cartilage canal (black arrowhead). Proliferating chondrocytes show no brown reaction product; scale bar = 100 μm. (G) A higher magnification of F. Several VEGF-positive resting chondrocytes are detectable in front of two cartilage canals (cc) that originate from the perichondrium (p); scale bar = 50 μm. (H) Section (counterstained) through the tip of a canal. The canal cells show a strong staining for VEGF. In addition, a TRAP-positive cell is seen. Green arrowhead denotes a VEGF-positive chondrocyte of the resting zone (rz); scale bar = 20 μm. (I,J) During progressing development (D 7 and 9) VEGF is encountered in the cells of the canals (cc) and the perichondrium (p). In addition, immunoreactivity is seen in the chondrocytes of the resting (rz) and hypertrophic zones (hz). The section in J is not counterstained; scale bar = 20 μm (I), and 50 μm (J).

Figure 6.

 IHC and HC: VEGF (brown staining) and TRAP (red staining). (A–D) Latero-medial sections through the distal femoral epiphysis of mice aged 18 and 20 days, respectively. (A) The SOC (soc), the proliferative (pz) and the hypertrophic zones (hz) of the metaphyseal growth plate are shown. Staining for VEGF is noted around the SOC and within the hypertrophic chondrocytes of the metaphyseal growth plate; scale bar = 100 μm. The following panels (B–D) reveal this staining pattern in detail. (B) Hypertrophic chondrocytes on the distal chondro-osseous junction of the SOC are VEGF-positive. Inset demonstrates a higher magnification of the labelled cells. Within the SOC, several TRAP-positive chondroclasts (chc) are attached to the mineralized cartilage; scale bars = 20 μm. (C,D) Consecutive sections (same magnification) are depicted. A signal for VEGF is discernible in the resting chondrocytes (rz) at the proximal border of the SOC but many of these cells reveal only a weak labelling. This is more obvious when sections are not counterstained with haematoxylin (D). In addition, VEGF is noticeable in the hypertrophic chondrocytes (hz) of the metaphyseal growth plate, whereas the proliferating chondrocytes (pz) are negative for the growth factor; scale bar = 50 μm. (E) The chondro-osseous junction of the primary ossification centre from a mouse aged 9 days is shown. Hypertrophic chondrocytes and numerous osteoblasts (black arrowheads and labelled as ob in the inset) are VEGF-positive. Multinucleated TRAP-positive chondroclasts (chc) are seen at the junction. These staining patterns are identical throughout development (D 4–20); scale bars = 20 μm. (F) A human kidney was used as a positive control, and VEGF is discernible in the cells of a glomerulus; scale bar = 20 μm.


We investigated SN expression in mice aged 4, 9, and 20 days. With respect to the study of Maes et al. (2004), which claimed a hypoxic condition in the core of the epiphysis, we checked for SN expression in the corresponding area. In addition, we speculated that the hypertrophic chondrocytes of the growth plate might up-regulate this angiogenic factor. In contrast to our hypothesis, SN expression was not detectable during epiphyseal growth, either in the innermost resting chondrocytes at D4 or in the hypertrophic chondrocytes throughout advancing development (Fig. 7A,B). However, we found SN in the skeletal muscle. In 4-day-old mice, a strong immunoreactivity was observed in most muscle fibres but it obviously decreased with advancing development, and by D 20, SN was only sporadically distributed within the muscle cells (Fig. 7A,B). The skeletal muscle served as a positive control because it was shown that muscle fibres expressed SN during normal development (Egger et al. 2007).

Figure 7.

 IHC and HC: SN (brown staining) and TRAP (red staining). (A,B) Frontal sections through the distal femoral epiphysis of mice aged 4 and 20 days, respectively. The sections are not counterstained. (A) An overview, demonstrating the resting (rz), proliferative (pz) and hypertrophic zone and the chondro-osseous junction of the POC (poc) is shown. No staining is found within the chondrocytes but numerous skeletal muscle fibres (m) are SN-positive; scale bar = 100 μm. (B) At D 20, a large SOC (soc) is seen within the epiphysis. The metaphyseal growth plate (gp) is located between the POC (poc) and the SOC (soc). Note that the staining for SN has decreased and fewer muscle fibres (m) are labelled; scale bar = 500 μm.

VEGF mRNA in situ hybridization 

We verified our data obtained from IHC by performing ISH for VEGF mRNA. In mice aged 4 days, a signal was detected in the perichondrium and certain resting chondrocytes below it. However, chondrocytes located in the centre of the epiphysis showed only a faint labelling (Fig. 8A,B). By 6 days, VEGF expression was visible in several resting chondrocytes surrounding the canal tips. In addition, some cells within the canals were labelled. In mice aged 20 days, VEGF-positive cells were observed in numerous hypertrophic chondrocytes at the distal border of the SOC (Fig. 8F,G). However, at the proximal border, VEGF expression was restricted to the outer edge (Fig. 8I), and in the midst the majority of the resting chondrocytes were not labelled (Fig. 8H). VEGF signal was also seen in several osteoblasts of the SOC and the POC, in cells of the periosteum, in the growth plate hypertrophic chondrocytes and skeletal muscle cells (Fig. 8J). These labelling patterns were found in all developmental stages, although the number of positive cells decreased during advancing development in the skeletal muscle (data not shown). In summary, our data obtained from ISH were consistent with the immunohistochemical findings. Controls for staining specificity, performed by the addition of a 1000-fold excess of non-FAM-labelled oligonucleotide probes, showed no labelling (Fig. 8C).

Figure 8.

 ISH: VEGF mRNA (blue staining) (A–E) Latero-medial sections through the distal femoral epiphysis at D 4 and 6 display the spatio-temporal localization of the VEGF mRNA. The sections are not counterstained. (A,B) In the 4-day-old mouse, VEGF is expressed in the perichondrium (p) and resting chondrocytes (rz) beneath it. Resting chondrocytes in the centre of the epiphysis show only a faint labelling. (C) Specificity of the labelling is demonstrated by comparing normally incubated sections with sections additionally hybridized with 1000-fold excess of non-FAM-labelled probes. These sections show no staining. (A–C) Same magnification; scale bar = 20 μm. (D,E) Consecutive sections through the tip of a cartilage canal (cc) of a 6-day-old mouse. VEGF expression is displayed in some perivascular canal cells and in the chondrocytes of the resting zone (rz). (D,E) Same magnification; scale bar = 20 μm. (F–I) Distribution of VEGF mRNA in a 20-day-old mouse. (F) Frontal section through the distal femoral epiphysis depicting the SOC (soc), the hypertrophic zone (hz) at its distal border and the resting zone (rz) at its proximal border. In addition, the proliferative zone (pz) of the metaphyseal growth plate is displayed; scale bar = 100 μm. The following micrographs demonstrate in detail the localization of the VEGF expression. (G) VEGF mRNA is encountered in the hypertrophic chondrocytes at the distal chondro-osseous margin of the SOC; scale bar = 20 μm. (H,I) On the proximal junction, VEGF mRNA in resting chondrocytes is mainly restricted to the outer edge (I), whereas in the innermost area (H) only a few chondrocytes (rz) are labelled. (H,I) Same magnification; scale bar = 20 μm. (J) The chondro-osseous junction of the POC (poc) is demonstrated. VEGF expression is discernible in the hypertrophic chondrocytes (hz) and in several osteoblasts (inset, black arrowheads); scale bar = 20 μm.


The fate of the epiphyseal hypertrophic chondrocytes

Studies on murine species have indicated that during the formation of the cartilage canals, several resting chondrocytes around the canal tips transform into a terminal stage and finally undergo cell death (Lee et al. 2001, 2009). In the distal femoral epiphysis of the mouse, the majority of the resting chondrocytes develop still further into hypertrophic chondrocytes. Our results, based on a histochemical approach (von Kossa reaction) and ultrastructural observations, provide clear evidence that hypertrophy of the chondrocytes correlates with the mineralization of the cartilage matrix (Figs 1D and 2D). In the chondrocyte maturation, the most noticeable feature is the alteration of their cytoplasm. In resting chondrocytes, abundant rough endoplasmic reticulum and numerous mitochondria are encountered, but a further differentiation of these cells leads to a significant degradation of the cell organelles, resulting in an almost electron-translucent cytoplasm in hypertrophic chondrocytes (Fig. 2A–E). These cells finally come off the wall of the lacuna, their nuclei show a patchy organization of the chromatin, and cellular fragments are sequestered into extracellular space (Fig. 2F). Our ultrastructural findings largely resemble those hallmarks typical for chondroptosis (Roach et al. 2004). This variant of apoptotic death leads to a complete self-destruction of the chondrocytes, eventually resulting in an empty lacuna in the growth plate, as demonstrated by the latter study. Nevertheless, we did not observe empty lacunae (Fig. 3A–F), bringing into question the absolute self-destruction of the epiphyseal hypertrophic chondrocytes. During the development of the SOC and erosion of the cartilage matrix, macrophages migrate into the epiphysis (Blumer et al. 2008b), and we assume that these cells might phagocytize the remnants of the hypertrophic cells after their lacunae are opened. This could be achieved by the TRAP-positive chondroclasts after disintegration of the mineralized cartilage (Fig. 4A). It should be noted that we did not perform consecutive semithin sections through the growth plate and therefore can not wholly exclude the complete self-destruction of the hypertrophic chondrocytes in this area. However, consistent with the notion of Roach et al. (2004) our immunohistochemical data indicate that caspase 3 is one of the executioners in chondroptosis (Fig. 4A–E).

The importance of the TRAP enzyme in the formation of cartilage canals

Whereas several MMPs are definitely needed for the early formation of the cartilage canals, the exact role of the TRAP enzyme during this event is yet not clear. The present results show that the enzyme is expressed in the perichondrium at D 4 (inset in Fig. 5A) and within the lumen of the canals the following day (Fig. 5D,E). Thus, an essential role for the TRAP molecule, similar to MT1-MMP, could be expected. However, our current studies on the femoral epiphysis of the TRAP knock-out mouse have revealed that cartilage canals are present 5 days after birth. Therefore, we assume that the enzyme does not initiate their formation. Nevertheless, TRAP appears to be critical in later developmental stages where it regulates the development of the SOC (unpublished observations).

The role of angiogenic factors during epiphyseal vascularization

To analyse which factors control the vascularization of the epiphysis, we examined the spatio-temporal distribution of two molecules (VEGF and SN) that are well-known to attract blood vessels. With respect to VEGF the localization of the protein was supported by an analysis of the corresponding mRNA.

Invasion of blood vessels is an important step for epiphyseal development and, if disrupted, leads to a delayed formation of the SOC, dysfunction of the growth plate and dwarfism (Holmbeck et al. 1999; Stickens et al. 2004). Mice lacking the isoforms VEGF120 and VEGF164, leaving expression only to VEGF188, exhibit a similar phenotype due to an altered vascular network in the epiphysis, which eventually results in impaired function of the growth plate. Thus, the data of Maes et al. (2004) indicate that activation of VEGF188 alone is not enough to allow for an undisturbed longitudinal growth of the bones.

In the present investigation we first examined the spatio-temporal expression pattern of VEGF during epiphyseal development. By day 4, VEGF expression is visible in the perichondrium and in several chondrocytes located immediately beneath it. Furthermore, the resting chondrocytes in the centre of the epiphysis synthesize the angiogenic molecule, but to a much smaller extent. Following the model presented by Maes et al. (2004), the avascular epiphyseal cartilage exceeds its critical size in early postnatal stages, leading to an insufficient supply of oxygen in its central area. This prompts the innermost resting chondrocytes to produce VEGF, and its soluble isoforms are released into the extracellular matrix from where they diffuse to the perichondrium. Our findings are partially in line with this model, but we want to emphasize that we also found the angiogenic factor in certain areas of the resting cartilage directly below the vascularized perichondrium, where a sufficient supply of oxygen can be expected. We therefore conclude that hypoxia is not necessarily required to induce VEGF synthesis, and that the chondrocytes of the latter areas have the major impact on the invasion of blood vessels. The cartilage canals emerge when the cells around the vessels migrate into the epiphysis. With respect to the three-dimensional reconstruction demonstrated by Blumer et al. (2007), only a few short canals are formed 5 days after birth, and we assume that those VEGF-positive areas below the perichondrium, noticed the previous day, give rise to their development. However, our results partially differ from those found in the rat tibia, where the onset of VEGF expression first occurred after the ingrowth of the canals (Álvarez et al. 2005a). Following the present results, VEGF is then discernible in several perivascular cells of the canals as well as in the resting chondrocytes surrounding the blind end of the canals, thus allowing them to advance deeper into the epiphysis (D 5 and 6) (Figs 5D–H and 8D,E). From D 7 onward, the latter cells mature into hypertrophic chondrocytes and VEGF obviously triggers an intense capillary sprouting, reflected by the highly branched canal endings and their vessels (Figs. 1D, E, and 5I,J). The expression of the angiogenic molecule further leads to the establishment of the SOC and its large marrow cavity. In mice aged 18 and 20 days, respectively, VEGF synthesis diminishes on the proximal chondro-osseous junction of the SOC, an observation also encountered in the rat tibia (Álvarez et al. 2005a). In summary, in the mouse femur VEGF is expressed before the onset of cartilage canal formation and is further continuously expressed throughout epiphyseal development.

In the present investigation we further tested whether SN might be an additional potent factor suitable for the establishment of the epiphyseal vascular network. Although this cytokine is up-regulated under hypoxic conditions (Egger et al. 2007), SN is apparently not expressed in the resting chondrocytes that are deprived of an adequate supply of oxygen. Furthermore, SN is not present within the cartilage anlage up to D 20. Accordingly, we conclude that SN has no impact on the vascularization of the epiphysis. In a cornea neovascularization assay it was shown that SN induces the formation of a capillary network, and the additional recruitment of smooth muscle cells implies that these vessels are mature. However, compared to VEGF, SN contributes to a somewhat smaller extent to the development of this vascular plexus (Kirchmair et al. 2004a; Fischer-Colbrie et al. 2005; Kirchmair, unpublished). We therefore speculate that SN has a less potent effect on endothelial cells and might preferentially attract vascular muscle cells, thereby regulating their proliferation and migration (Kähler et al. 1997). In mice until D 20, the epiphyseal vessels consist of endothelial cells only, without a muscular layer. Endothelial cells are the unambiguous target of VEGF (Gerber et al. 1999; Harper & Klagsbrun, 1999; Grellier et al. 2009), and hence we conclude that this molecule is the critical player during the early establishment of the epiphyseal capillary network.

In conclusion, our data provide evidence that hypertrophy of the epiphyseal chondrocytes correlates with mineralization of the extracellular matrix. During development of the epiphysis the latter cells undergo cell death by a mechanism interpreted as chondroptosis. Furthermore, the TRAP enzyme obviously has no role in the early formation of the cartilage canals. Finally, the findings reported here suggest that VEGF mediates the invasion of the epiphysis of the long bones by the blood vessels. SN does not appear to have any effect on this event.


This study was supported by the Austrian Science Fund (FWF) (grant P20758-B05). The authors thank Prof. VMD. H. Dietrich for helping to dissect the animals. We specially thank Mag. H. Hübl, and Dr R. Illig for her valuable assistance in the laboratory, Prof. MD. H. Fritsch and Prof. PhD. K. Pfaller for their helpful discussion, and C. Siemon and Travieso for carefully reading and correcting the manuscript.