Angiogenesis Is Required for Successful Bone Induction During Distraction Osteogenesis



The role of angiogenesis during mechanically induced bone formation is incompletely understood. The relationship between the mechanical environment, angiogenesis, and bone formation was determined in a rat distraction osteogenesis model. Disruption of either the mechanical environment or endothelial cell proliferation blocked angiogenesis and bone formation. This study further defines the role of the mechanical environment and angiogenesis during distraction osteogenesis.

Introduction: Whereas successful fracture repair requires a coordinated and complex transcriptional program that integrates mechanotransductive signaling, angiogenesis, and osteogenesis, the interdependence of these processes is not fully understood. In this study, we use a system of bony regeneration known as mandibular distraction osteogenesis (DO) in which a controlled mechanical stimulus promotes bone induction after an osteotomy and gradual separation of the osteotomy edges to examine the relationship between the mechanical environment, angiogenesis, and osteogenesis.

Materials and Methods: Adult Sprague-Dawley rats were treated with gradual distraction, gradual distraction plus the angiogenic inhibitor TNP-470, or acute distraction (a model of failed bony regeneration). Animals were killed at the end of distraction (day 13) or at the end of consolidation (day 41) and examined with μCT, histology, and immunohistochemistry for angiogenesis and bone formation (n = 4 per time-point per group). An additional group of animals (n = 6 per time-point per group) was processed for microarray analysis at days 5, 9, 13, 21, and 41.

Results and Conclusions: Either TNP-470 administration or disruption of the mechanical environment prevented normal osteogenesis and resulted in a fibrous nonunion. Subsequent analysis of the regenerate showed an absence of angiogenesis by gross histology and immunohistochemical localization of platelet endothelial cell adhesion molecule in the groups that failed to heal. Microarray analysis revealed distinct patterns of expression of genes associated with osteogenesis, angiogenesis, and hypoxia in each of the three groups. Our findings confirm the interdependence of the mechanical environment, angiogenesis, and osteogenesis during DO, and suggest that induction of proangiogenic genes and the proper mechanical environment are both necessary to support new vasculature for bone induction in DO.


ANGIOGENESIS PLAYS A critical role in bone development and postnatal fracture repair.(1–9) Despite abundant evidence suggesting that angiogenesis and osteogenesis are closely linked, the complex progression of transcriptional signals that couple osteogenic and angiogenic processes after injury remains to be elucidated. Given the enormous biomedical burden of skeletal injury, and the need to treat congenital, posttraumatic, and postsurgical conditions in which bone regeneration is unsuccessful, defining the role of angiogenesis in osteogenesis and bone repair will likely have important clinical implications.

Insights into postnatal osteogenesis and fracture healing in particular have been gained using gene-targeted studies to define the specific signaling molecules, matrix proteins, growth factors, transcription factors, and other participants in the repair program. In addition to long-bone fracture models, the robust nature of bone induction in distraction osteogenesis (DO) has made it an important system to study the role of angiogenesis in bone regeneration.(10,11) A unique and powerful form of endogenous bone tissue engineering, DO relies on the application of controlled mechanical forces to promote bone induction between two osteotomy fronts, and has become the treatment of choice for many craniofacial hypoplasias and other conditions requiring bone lengthening.(11) Whereas mechanical strain (a result of the controlled separation applied during the gradual distraction phase of DO) has been postulated to affect osteoblast proliferation and differentiation through mechanotransduction, the molecular signals that translate mechanical stimuli into osteogenic induction have not been completely determined in DO. Furthermore, despite evidence of angiogenic mediator upregulation and increased blood vessel formation during DO, it is unknown whether angiogenesis is a prerequisite for the profound bone induction seen in DO and whether disrupting the controlled mechanical forces that occur during DO affects the angiogenic program.(12)

With the development of anti-angiogenic agents for cancer therapy, it has become possible to directly determine if inhibiting angiogenesis affects postnatal healing. TNP-470, a synthetic analog of fumagillin, has been shown to block new capillary formation in vivo and inhibit endothelial cell proliferation in vitro.(13,14) Whereas the exact mechanism responsible for the specificity of TNP-470 remains under study, it has been recently identified that the molecular target of TNP-470 is methionine aminopeptidase-2 (MetAP2), which is involved in cell cycle progression.(15–17) Inhibition of this enzyme leads to endothelial cell cycle arrest in late G1, with subsequent inhibition of endothelial cell proliferation and migration.(15–17) TNP-470 may also have endothelial cell-independent antitumor effects.(18) More recently, it has been shown using a murine endochondral femoral fracture model that TNP-470 administration was associated with radiographic and histological evidence of failed fracture healing or nonunion.(19) Whereas this observation supported earlier evidence in a similar model that application of a vascular endothelial growth factor (VEGF) blocker (i.e., a soluble VEGF receptor) prevented normal fracture healing, the molecular basis for unsuccessful osteogenesis after angiogenic inhibition remains to be elucidated.(20) Furthermore, the effect of angiogenic blockade on the direct, or membranous, bone induction seen during DO is not known.

In this study, we established that in vitro administration of TNP-470 inhibits endothelial cell proliferation without affecting osteoblast growth or differentiation. To gain insight into the mechanisms relating osteogenesis to angiogenesis and mechanotransduction, we disrupt bone induction seen during gradual distraction (GD) using two distinct perturbations: direct angiogenic blockade with TNP-470 (GD + TNP) or mechanical disruption with acute distraction (AD), an established protocol of failed osteogenesis that alters pro-osteogenic mechanotransductive signaling.(21) Our data show that either TNP-470- or AD-mediated disruption of the mechanical environment results in a failure of osteogenesis and angiogenesis. Furthermore, we use microarray analyses to compare the transcriptional profiles of successful bony regeneration (GD) with each unsuccessful osteogenesis protocol (AD, GD + TNP). Our findings suggest that both angiogenesis and a controlled mechanical environment are necessary for bone induction to occur in the setting of DO. Furthermore, proper mechanical signaling may be a prerequisite for angiogenic and osteogenic programs to allow successful osteogenesis to occur.


Animal care and surgery

Neonatal and adult male Sprague-Dawley rats weighing between 350 and 400 g were purchased from Simonsen Laboratories (Gilroy, CA, USA). Animals were housed in a light- and temperature-controlled environment and given food and water ad libitum. All animal experiments were performed in accordance with Stanford University Animal Care and Use Committee guidelines.

Cell culture, proliferation, and osteoblast differentiation assays

Osteoblast-enriched cell cultures derived from neonatal rat calvaria were established as previously described.(22) Briefly, 2-day-old Sprague-Dawley rat pups (n = 20 per harvest) were killed with CO2. Calvaria were harvested, and the periosteum, dura mater, and all nonosseous tissue meticulously removed. After washing the calvaria with Betadine serially diluted in PBS, osteoblasts were subsequently released with five sequential 10-minute digestions with 0.1% collagenase and 0.2% hyaluronidase (Boehringer Mannheim, Indianapolis, IN, USA). Fractions 2-5 were pooled, centrifuged, resuspended, and plated in DMEM plus 10% FBS, 100 IU/ml penicillin, and 100 IU/ml streptomycin (Life Technologies) at 37°C with 5% CO2. Media were refreshed every 2 days, and second-passage osteoblast cultures were used for experiments.

Human umbilical vein endothelial cells (HUVECs; American Type Culture Collection, Rockville, MD, USA) were maintained in subconfluent cultures in Endothelial Cell Grow Media-2 (EGM-2; Cambrex Bioscience, Walkersville, MD, USA). Fourth passage cells were used in the studies. Osteoblasts or HUVECs were plated on 24-well plates (2000 cells/well), grown in their respective culture media for 48 h, and serum-starved for 24 h to promote cell cycle synchronization. We then replaced the media with DMEM/10% FBS or EGM-2/10% FBS containing various dilutions of TNP-470 (10−13 to 10−8 M) dissolved in 0.01% DMSO. This dose range has previously been used to study molecular targets of TNP-470 in vitro in several cell types including HUVECs and fibroblasts.(23) Control wells were cultured in media plus 0.01% DMSO, and all conditions were performed in quadruplicate. After 72 h, cell counts were performed by means of colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay as previously described.(24,25) Briefly, media was aspirated and replaced with 1 ml of MTT solution (0.25 mg/ml). After incubation for 2 h at 37°C, the MTT solution was discarded, and 1 ml of 20% SDS solution added to the wells. After gentle shaking for 30 minutes, the absorbance at 570 nm was determined using a spectrophotometer (Pharmacia Biotech, Cambridge, UK). Cell proliferation results for TNP-treated cells were normalized to vehicle-treated controls.

For osteogenic differentiation, early passage osteoblast-enriched cultures established as above were placed in DMEM/10% FBS supplemented with 1 μM dexamethasone, 5 mM β-glycerophosphate, 100 μg/ml ascorbic acid, and TNP-470 (10−8 M) dissolved in DMSO or vehicle. Medium (with TNP-470 or vehicle) was replaced every 2 days thereafter. Cells underwent staining for alkaline phosphatase activity at 14 days using the Alkaline Phosphatase Detection Kit (Sigma, St Louis, MO, USA). The percentage area stained by alkaline phosphatase was determined using Image J software (NIH, Bethesda, MD, USA). Extracellular mineralization was detected by staining osteoblasts cultured as above for 28 days with Alizarin red (Sigma; 100 mg/ml in 0.01% NaOH) or von Kossa staining. The percentage area stained by alkaline phosphatase and von Kossa was determined using Image J software (NIH). Alizarin staining was biochemically quantified by solubilizing retained Alizarin red dye in 20% methanol and 10% acetic acid and measuring absorbance at A450.(26) A Student's t-test was used to determine the statistical significance of all quantifications, with p < 0.05 considered significant.

Western blot analysis

HUVECs and osteoblasts grown in subconfluent cultures were serum-starved for 24 h and treated with TNP-470 as described above. After 72 h, the plates were rinsed with cold PBS and lysed in 200 μl of cold RIPA buffer (50 mM Tris pH 8.0, 1% NP-40, 150 mM NaCl, 0.5% deoxycholic acid, 0.1% SDS) with an EDTA-free protease inhibitor cocktail (Roche, Indianapolis, IN, USA). The lysates were centrifuged to remove debris and quantitated. Fifty micrograms of protein was electrophoresed on 12% SDS-polyacrylamide gels and transferred to polyvinylidine difluoride (PVDF) membranes. Membranes were probed using a proliferating cell nuclear antigen (PCNA) and βactin antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, USA) and the electrochemoluminescence (ECL) detection kit (Amersham Biosciences, Piscataway, NJ, USA).

Gradual distraction and acute distraction protocols

GD and AD surgical protocols have been previously described and extensively used to analyze the biology of bone formation during distraction osteogenesis (Fig. 1).(21,27–33) Rats were anesthetized, given a preoperative dose of antibiotics (10 mg/kg cefazolin), and prepped in a sterile fashion. After incisor clipping, an incision was made over the right hemimandible, and the masseter muscle was divided to expose the underlying mandible. An osteotomy was performed between the second and third molars using a diamond disc (Brassler, Savannah, GA, USA), and two Flexi-post pins were inserted 4 mm anterior and posterior to the osteotomy site. The muscle and skin were closed in layers, and a modified Lewa Jackscrew (Pearson Dental, Sylmar, CA, USA) was fixed to the Flexi-post pins.

Figure FIG. 1..

Schematic representation of experiment outline. At day 0, rats assigned to the GD group underwent placement of a mandibular distraction device. This was followed by a latency period (days 0-5, in which no distraction was performed), gradual distraction (days 5-13, at a rate of 0.25 mm every 12 h each day for a total of 4 mm), and consolidation (days 13-41, in which the mandibular osteotomies were allowed to heal further without additional distraction). Animals undergoing GD received vehicle injections every 48 h for 41 days, whereas animals designated GD + TNP underwent the aforementioned protocol but received TNP-470 treatment (30 mg/kg SC every 48 h for 41 days). Animals undergoing the AD protocol underwent distraction device placement followed by immediate (on-table) mandibular separation by 4 mm. Arrowheads indicate tissue harvest time-points for imaging, histological, and/or RNA analysis.

Briefly, the protocol consisted of a 5-day latency period followed by 8 days of gradual distraction and 28 days of maturation/consolidation (in which no additional distraction is applied). Mandibles were distracted by 0.25 mm every 12 h each day for a total of 4 mm. Animals in the control group (GD alone, n = 38) received vehicle injections every 48 h for 41 days, whereas animals designated GD + TNP (n = 38) received 30 mg/kg TNP-470 subcutaneously every 48 h for the entire 41 days (generously supplied by Takeda Chemical Industries, Osaka, Japan) as previously described for in vivo studies.(19,34,35) Animals undergoing the AD protocol (n = 38) underwent distraction device placement as described above, followed by immediate, on-table mandibular lengthening by 4 mm. Unlike GD, in which controlled mechanical stimuli are applied to the osteotomy over 8 days, the immediate separation of the two osteotomy fronts in AD precludes the creation of a mechanical environment favorable to bony union and results in a fibrous scar.(21) All animals tolerated an oral diet without difficulty or evidence of weight loss during the entire postoperative period, and animals in all groups manifested the expected cross bite produced by hemimandibular lengthening.

μCT and histologic analysis

A subset of animals (n = 4 per time-point per group) were killed at the end of distraction (day 13) and at the end of consolidation (day 41) for right hemimandible harvest, μCT, and histology. Specimens for μCT were fixed in 4% paraformaldehyde for 48 h and scanned at 40-μm resolution, 50 KV, and 80 μA (Scanco μCT40; Scanco, Bassendorf, Switzerland). Serial axial images of the regenerate were used to generate 3-D reconstructions, and calibration was performed with a hydroxyapatite phantom following manufacturer's instructions.

Right hemimandibles for histology were harvested and fixed with 4% PFA for 48 h and demineralized with Formical 2000 (Decal Chemical Corp., Congers, NY, USA) for 3 weeks. The demineralized mandibles underwent paraffin embedding and serial sectioning at 5-μm intervals. H&E and Safranin orange stains were performed on all specimens. Immunohistochemistry (IHC) for platelet endothelial cell adhesion molecule (PECAM) was performed using a PECAM-1 antibody (SC-8306; Santa Cruz Biotechnology), the DAKO LSAB2 System, and horseradish peroxidase (HRP; K0673; DAKO) per the manufacturer's recommendations. PECAM-stained slides were counterstained with 0.02% Fast green. Photomicrographs were obtained using a Zeiss Axioplan 2 Imaging system and AxioCam digital Camera (Carl Zeiss, Hallbergmoos, Germany).

RNA extraction and microarray analysis

A subset of animals (n = 6 for each time-point per group) were killed on postoperative days (POD) 5, 9, 13, 28, and 41 by CO2 asphyxiation. The right hemimandibles were exposed and dissected free from overlying muscle and soft tissue, allowing the distraction regenerates (plus 1 mm cortical bone on either side) to be sharply excised. RNA was isolated from the harvested regenerates using Trizol reagent (Invitrogen, Carlsbad, CA, USA) with a Polytron 1200 tissue homogenizer (Kinematica). Two groups of RNA pooled from three separate animals were generated for each time-point. RNA from three animals that had not undergone osteotomy was collected as a control.

Samples were hybridized to rat U34A GeneChip arrays (Affymetrix, Santa Clara, CA, USA) containing known and annotated rat genes (∼7000) and unknown expressed sequence tags (ESTs; ∼1800) per manufacturer's instructions. Briefly, 20 μg of total RNA from each pooled sample triplicate was reverse transcribed to double-stranded cDNA using the SuperScript Choice System (Invitrogen). We synthesized cRNA using the T7 MegaScript In Vitro Transcription Kit (Ambion, Austin, TX, USA), followed by biotin labeling using Bio-11-CTP and Bio-16-UTP (Enzo Diagnostics, Farmingdale, NY, USA). After 37°C incubation for 6 h, the labeled cRNA was purified using RNeasy spin columns (Qiagen, Valencia, CA, USA). Fifty micrograms of each cRNA sample was fragmented by mild alkaline treatment at 94°C for 35 minutes, combined with aliquots of a master mix containing 0.1 mg/ml of herring sperm DNA (Sigma), 1 M sodium chloride, 10 mM Tris (pH 7.6), and 0.005% Triton X-100, and prepared for hybridization. Arrays were washed, stained with streptavidin phycoerythrin (Molecular Probes, Eugene, OR, USA) in Affymetrix fluidics stations, and scanned using a Hewlett Packard Gene Array Scanner (Hewlett Packard Corp., Palo Alto, CA, USA).

Microarrays were performed in duplicate for each time-point/treatment, and the expression values for each gene averaged across replicates. Based on previously described methods, microarray images were analyzed for quality control, chip defects, and hybridization signal abnormalities.(36) We performed array normalization, hierarchical clustering, and analysis using DNA-Chip Analyzer (dChip) Version 1.3 (available on-line from the Wong Laboratory, Harvard School of Public Health and Dana-Farber Cancer Institute).


Effect of TNP-470 on HUVECs and osteoblasts in vitro

TNP-470 has been shown to inhibit angiogenesis and in vitro vascular tube formation by blocking endothelial cell proliferation.(14,37–39) Whereas TNP-470 may also have a direct (and endothelial-cell independent) effect on tumors, its effect on primary osteoblasts has not been elucidated.(35) To determine if TNP-470 inhibits osteoblast proliferation, we treated primary rat osteoblast-enriched cultures with escalating doses of TNP-470 for 72 h and assessed proliferation. The dose range used in this study is derived from previous in vitro studies of TNP-470 in HUVECs and fibroblasts.(23) Relative to vehicle-treated cells, osteoblasts treated with 10−13 to 10−8 M TNP-470 did not show a significant alteration in proliferation (Fig. 2A). In contrast, HUVECs treated with TNP-470 using the same protocol showed a significant decrease in proliferation at 10−9 and 10−8 M. Immunoblotting for PCNA of osteoblasts and HUVECs treated as above confirmed that TNP-470 specifically inhibited HUVEC proliferation at 10−9 and 10−8 M without affecting osteoblasts at any of the doses examined (Fig. 2B). We also verified that treatment of primary preosteoblasts in vitro with TNP-470 at 10−8 M had no significant effect on either early (alkaline phosphatase activity) or late (extracellular matrix mineralization) assays of osteoblastic differentiation (Figs. 2C and 2D). Taken together, these findings suggest that TNP-470 does not inhibit in vitro proliferation and differentiation of primary rat osteoblasts at the doses examined and are consistent with prior reports of a selective inhibitory effect of TNP-470 on endothelial cells.(14)

Figure FIG. 2..

In vitro effects of TNP-470 on HUVECs and primary rat osteoblasts (OBs). (A) Each cell type was treated with 10−13 to 10−8 M TNP-470 and cell number (relative to untreated controls) determined by MTT assay. HUVECs but not OBs are growth-inhibited with escalating TNP-470 administration at the dose range examined. (B) PCNA confirmed an inhibitory effect of TNP-470 on HUVECs but not OBs. (C) Primary rat calvarial osteoblasts were grown in osteogenic differentiation medium supplemented with vehicle (ODM) or TNP-470 (+TNP) at a concentration of 10−8 M. Neither alkaline phosphatase activity (AP; 31.2 ± 3.5% vs. 29.5 ± 3.1%, *p > 0.05) at 14 days nor extracellular matrix mineralization at 28 days (VK; 15.4 ± 0.3% vs. 14.6 ± 0.4%, *p > 0.05) was significantly inhibited by TNP-470 treatment. No staining was seen in control (undifferentiated) cells (data not shown). (D) Quantification of mineralization with Alizarin red also showed no significant difference between vehicle and TNP-470 treated osteoblasts (p > 0.05).

TNP-470 blocks osteogenesis during gradual distraction

We next investigated whether TNP-470 affects osteogenesis in vivo using a rat DO model. As discussed above, DO is a powerful mode of bone induction that, in both experimental and clinical settings, has been shown to result in successful bone. Although TNP-470 has previously been used to disrupt angiogenesis during healing of stabilized fractures in the appendicular skeleton, the effect of direct angiogenic blockade during DO has not been reported.(19) We treated rats with TNP-470 (GD + TNP) or vehicle (GD) throughout a 41-day gradual distraction protocol and assessed the degree of osteogenesis radiographically and histologically. The TNP-470 dose used in this study has been shown to inhibit rat femoral fracture healing and to have antitumor effects in mice and rats.(19,34)

μCT analysis showed that all animals undergoing GD had complete bony union of the distracted mandibular segments as expected by POD 41 (representing the end of the consolidation period; Fig. 3A). In contrast, all animals treated with TNP-470 for 41 days showed complete nonunion, with no visible bone between the two mandibular osteotomy fronts (Fig. 3B). Histologic analysis of GD- and GD + TNP-treated mandibles at end-consolidation confirmed the radiographic findings, with bony bridging seen in all GD specimens and fibrous union seen in all GD + TNP specimens (Figs. 3C and 3D). Fibrous union was also noted in all AD specimens as expected (data not shown).

Figure FIG. 3..

μCT and histologic analysis of hemimandibles harvested at day 41 (end of consolidation). (A) All animals that underwent GD plus vehicle injections showed complete bony bridging at day 41, (B) whereas all animals in the GD + TNP group showed no visible bone between the mandibular osteotomy fronts. (C) H&E-stained hemimandibles from vehicle-treated animals revealed bone that completely bridged the distraction gap, (D) whereas TNP-470-treated animals showed dense fibrous tissue (characteristic of failed healing or nonunion).

We next compared the histological characteristics of failed osteogenesis resulting from angiogenic blockade with TNP-470 to that of AD at the end of the distraction phase, the period of the DO protocol when the rate of bone deposition is maximal.(32) These data were also compared with the histology data from GD specimens. GD + TNP and AD specimens were both characterized by a lack of early bone induction compared with GD controls. In GD specimens, intramembranous bone formation was seen within the distraction gap at both middistraction and midconsolidation, along with a small amount of endochondral bone formation at the periphery of the distraction gap along the periosteum at middistraction (Fig. 4A). Although no bone formation was seen in GD + TNP specimens, a small amount of endochondral bone formation was seen in 50% of GD + TNP specimens, at locations analogous to the endochondral bone seen in GD specimens (Fig. 4D). No evidence of either direct bone formation or a cartilaginous intermediate was found in the AD specimens (Fig. 4G).

Figure FIG. 4..

Mechanical disruption (through AD) or direct angiogenic inhibition (GD + TNP) results in failed bone induction and a lack of endothelial cell staining. (A) At day 13 (end-distraction), Safranin O staining of GD specimens showed evidence of early bone formation within the distraction gap, including trabecular bone and hypertrophic chondrocyte staining (red). (D) Although some cartilage was seen adjacent to the perichondrium in GD + TNP specimens, no bone formation was seen in the distraction gap. (G) Neither bone nor cartilage was observed in AD specimens. (B and C) Gross and immunohistochemical localization of endothelial cells with PECAM staining (brown) shows numerous vessels in the newly formed trabecular bone and in the distraction gap of GD specimens. (E and F) Although some staining is seen adjacent to the periosteum in GD + TNP specimens, no evidence of vessels was detected in the mesenchyme of the distraction gap. (H and I) A few small vessels can be seen in the AD specimens adjacent to the osteotomy. Scale bars = 500 μm; C, F, and I are magnifications of the boxes in B, E, and H, respectively. *Cartilage.

Microarray analyses of overall gene profiles and of osteogenic genes specifically were consistent with our histological data. A comparison of the total temporal gene expression pattern of GD mandibles to intact, unperturbed rat mandibles showed that 755 genes had 2-fold or greater differential expression at day 5 (the earliest distraction time-point examined) relative to unperturbed, control bone (Fig. 5A). This difference gradually decreased over time, with only 59 genes differentially regulated at day 41. Thus, by the end of the consolidation period, the transcriptomes of the newly formed mandible and control mandible were very similar. This observation is supported by our finding that genes associated with successful osteogenesis (including BMPs 2, 4, and 6, collagens I and III, alkaline phosphatase, bone sialoprotein, prepro-bone-inducing protein, and osteocalcin; Fig. 5B) were indeed upregulated in GD and remained high throughout the consolidation period.

Figure FIG. 5..

Microarray analysis data. (A) Summary of overall gene expression of GD and GD + TNP specimens compared with control mandibles. At the beginning of distraction (day 5), GD (squares) and GD + TNP (triangles) groups were notable for a relatively large number of differentially expressed genes (>650 genes expressed by ≥2-fold relative to control mandibles). Whereas the number of differentially expressed genes remained high in the GD + TNP group throughout the 41 days, the transcriptional profile of GD-treated specimens approached that of control bone by day 41 (only 59 genes expressed at ≥2-fold relative to day 0). (B) Selected cluster analysis of osteogenesis-associated genes shows sustained proosteogenic gene expression in GD but not AD or GD + TNP. Cluster analysis of (C) angiogenic and (D) hypoxia-regulated genes in AD and GD + TNP at the indicated time-points relative to day 0 (unperturbed mandible). (C) Analysis of angiogenesis-related related genes reveals induction with GD and AD, but not with angiogenic blockade through GD + TNP (except for the hypoxia-inducible genes VEGF-A, VEGF-2, and NO synthase). (D) Additional hypoxia-sensitive genes were found to be upregulated markedly compared with GD + TNP or AD. Gene names listed more than once indicate replicate identical probes or distinct probes representing the same locus. Color scale indicates range of gene expression. bmp, bone morphogenic protein; prepro bone ind. prot., prepro bone inducing protein.

In contrast, although the number of genes differentially regulated at day 5 in GD + TNP was similar to that seen in GD, over 350 genes were still differentially regulated at day 41 compared with control mandibles. A similar number of genes remained upregulated by day 41 in AD mandibles (data not shown). The failure of successful bone formation in the AD and GD + TNP groups could account for the persistent transcriptional differences between control mandibles and the AD and GD + TNP mandibles. The osteogenic genes listed above (Fig. 5B) were only transiently increased in AD and displayed no (or very late and attenuated) induction in the GD + TNP group, consistent with an early divergence of gene expression relative to GD and, ultimately, failed osteogenic induction.

Failure of osteogenesis caused by TNP-470 administration or acute distraction is characterized by a lack of angiogenesis

To confirm that the lack of bone formation seen after treatment with TNP-470 was caused by a failure of appropriate angiogenesis and to determine the effect of disrupting the mechanical environment on angiogenesis, we evaluated the regenerate for the presence of blood vessels by gross histology and immunohistochemistry for PECAM. PECAM immunohistochemistry of GD-treated mandibles revealed positive endothelial cell staining and clear vascular channels throughout the distraction gap and at the leading fronts of newly formed trabecular bone. In contrast, the area surrounding the hypertrophic chondrocytes near the periosteal areas outside the distraction gap notably lacked PECAM staining (Figs. 4B and 4C). Analysis of specimens from the GD + TNP group revealed a complete absence of PECAM staining and no obvious blood vessels in either the fibrous tissue of the distraction gap or surrounding chondrocytes, suggesting that failure of angiogenesis was responsible for the lack of osteogenesis in this group (Figs. 4E and 4F). Interestingly, specimens from the AD-treated animals (who, like the GD + TNP group, formed a fibrous nonunion) only showed minimal PECAM staining or the presence of vasculature within the distraction gap adjacent to the osteotomized bone (Figs. 4H and 4I).

We next interrogated our microarray dataset for genes previously implicated in angiogenesis. As expected, we observed significant upregulation of the majority of proangiogenic genes in the GD group, including fibroblast growth factors (FGFs) and their receptors, angiotensinogen and angiotensin receptors, and VEGF-R1 (Fig. 5C). Consistent with our histological findings (and with TNP-470's underlying effect as a direct inhibitor of angiogenesis), the GD + TNP group showed no significant upregulation of the majority of these genes at any of the time-points examined. Interestingly, despite the minimal PECAM staining seen in AD mandibles, the AD microarray results showed that transcriptional activity for angiogenic genes was active.

One notable exception to the overall downregulation of angiogenic genes seen in GD + TNP was the expression of the hypoxia-inducible targets VEGF-A, VEGF-R2, and NO synthase-2. Induction of these genes in GD + TNP mandibles starting at end-distraction (day 13) led us to more closely examine additional hypoxia regulated genes. Consistent with the lack of angiogenic gene induction in GD + TNP, we noted sustained upregulation of hypoxia-sensitive targets (including anaerobic glycolytic enzymes, hypoxia inducible factor-1 α, hypoxia inducible gene-1, and several heat shock proteins; Fig. 5D and data not shown). No sustained upregulation of these genes was seen in either the GD or AD groups.


Despite clinical advances in DO and the development of numerous animal models aimed at studying the underlying molecular mechanisms, several important questions remain. One of these concerns the role of angiogenesis in bone induction in the setting of DO. It has long been known that DO-dependent bone regeneration is accompanied by a profound vascular response.(40,41) Gene-targeted studies and, more recently, macroarrays (in which ∼100 genes are examined simultaneously) have identified stage-specific activation of several vascular markers and proangiogenic factors within the bony regenerate.(42) Furthermore, large-scale microarrays have been applied in long-bone fracture models (but not in DO) to identify the expression pattern of thousands of genes at different stages of fracture healing, including factors implicated in angiogenic sprouting and maintenance.(42,43) Anti-angiogenic agents have also recently been used to block long-bone fracture healing to further show the significance of angiogenesis in osteogenic regeneration.(12) Nevertheless, it has not previously been determined whether angiogenesis is an absolute prerequisite for the bone induction seen in DO, a primarily intramembranous form of bony regeneration capable of overcoming radiation damage and other factors detrimental to bone formation.(44–46)

In this study, we used the anti-angiogenic agent TNP-470 to determine the effects of disrupting angiogenesis during unilateral mandibular distraction osteogenesis in the rat. We first determined whether TNP-470 affects osteoblast biology to ensure that any inhibition of osteogenesis was caused by a lack of angiogenesis. Our data show that TNP-470 is not directly toxic to osteoblasts, does not inhibit their proliferation, and has no significant effect on their ability to differentiate in vitro at doses used to inhibit endothelial cell proliferation. Consistent with data from long-bone fracture models, TNP-470 treatment in the setting of a 41-day distraction protocol resulted in a gross nonunion, with no histologic or radiographic evidence of bone formation.(19) Standard histological examination and immunohistochemical localization of endothelial cells using PECAM staining confirmed an absence of nascent and established vasculature in the regenerate of the TNP-470-treated mandibles, whereas abundant vessels were seen in GD mandibles. These findings suggest that angiogenic blockade alone is sufficient to prevent de novo osteogenesis during gradual distraction and is therefore a required component of the distraction osteogenesis process. At least in part, this may be caused by a lack of bioavailable VEGF at the fracture site, thus hindering osteoblast differentiation by limiting a growth factor with known pro-osteogenic effects.(47,48) Furthermore, the more global microenvironmental alterations stemming from a lack of a proper blood supply and ensuing tissue hypoxia may also directly hinder osteoblast differentiation.(49)

Despite the apparent absence of angiogenesis in the GD + TNP group (as determined by PECAM staining and microarray analysis), the proangiogenic factor VEGF-A was noted to be upregulated starting at end-distraction, as were numerous other hypoxia-inducible genes. Our histological studies showed the presence of chondrocytes at the periphery of the osteotomized bone in several of the GD + TNP specimens. Because hypoxia is known to induce VEGF expression in chondrocytes (through hypoxia inducible factor-1, also upregulated in GD + TNP specimens), production of VEGF by hypoxic chondrocytes may account for the increased transcript levels seen.(50,51) Alternatively, the hypoxic environment of GD + TNP (as suggested by the microarray results) may induce VEGF-A expression in the mesenchymal tissue of the distraction gap. Because the primary functional difference between the mesenchymal tissue in the distraction gap of GD versus GD + TNP is the ability to recruit a blood supply, it is possible that the mechanical environment created by gradual distraction induces new tissue formation, which is subsequently rendered hypoxic in our GD + TNP group, resulting in failed osteogenesis. It is interesting to note that no significant upregulation of hypoxic genes was seen in the AD group, where no tissue induction (and thus no need for vascularization of the distraction mesenchyme) is seen.

Although abundant evidence exists showing that generation of the proper mechanical microenvironment is critical for successful bone induction during DO, whether alteration of the mechanical forces present during distraction would affect the angiogenic program has not been previously reported.(29–31,52) To this end, we evaluated the distraction gap of AD mandibles for evidence of nascent vasculature and/or endothelial cells. As with the TNP-470-treated mandibles, no gross histologic or immunohistochemical evidence of vessel formation was seen. This is in marked contrast to the numerous vascular channels and vessels seen in the GD specimens. Although it may be argued that PECAM staining does not conclusively show the presence of angiogenesis and an intact vasculature, we have previously observed the presence of greater blood flow and mature vasculature in rat mandibles treated with GD compared with AD using intravenously administered, radiolabeled nanoparticles conjugated to anti-αvβ3 antibodies (unpublished observations, 2002). Thus, the histological data presented in this study strongly suggest that a failure of angiogenesis in the distraction gap accompanies the failed osteogenesis seen in AD, resulting in a fibrous nonunion.

Despite the lack of histologic and immunohistochemical evidence of angiogenesis in AD, our microarray data showed a relatively unperturbed pattern of angiogenic gene expression in AD relative to GD. In other words, it seems that angiogenic transcriptional activity is not being “turned off” in AD. This upregulation of angiogenesis-related genes may reflect the sensitivity of microarray in picking up transcriptional activity resulting from angiogenesis and/or postinjury inflammation surrounding the mandible. Alternatively, it may imply that proangiogenic gene cascades are present at the injury site and would be capable of providing new vasculature to regenerating bone were the correct mechanical environment (i.e., GD) present to induce osteogenesis. We hypothesize that angiogenic signaling alone cannot fully support bone induction and that other factors acting downstream of the primary angiogenic cascade may prevent the formation of an intact vascular network. One candidate factor that may potentially disrupt or inhibit vasculogenesis is the mechanical environment itself. Several studies have shown that the presence of an appropriate mechanical environment supports the development of vasculature.(5,53) Furthermore, we have previously shown, using mechanical testing and finite element analyses, that the mechanical environments produced by gradual distraction and acute distraction are quite different.(32,33) In this study, targeted analysis of our microarray data showed that genes involved in mechanotransduction (including integrins and their downstream mediators, focal adhesion kinase and mitogen activated kinases) were upregulated in GD + TNP compared with AD (data not shown). Coupled with the similar initial angiogenic response to GD, these data support the notion that the primary perturbation in AD is a mechanical one.

Angiogenic induction and remodeling may, in fact, rely on many of the same mediators that are implicated in mechanotransductive signaling. The migration and invasion of endothelial cells is in large part mediated by integrin-dependent communication with the extracellular matrix.(54) In addition to affecting osteoblast gene expression and matrix deposition, mechanical factors such as shear stress play a major role in vascular homeostasis and pathophysiology and, by regulating the production of vasoactive mediators and expression of adhesion molecules, affect endothelial cell migration, proliferation, and survival.(32,54–58) Mechanical force has been shown to influence capillary growth and VEGF and matrix metalloproteinase expression.(59) Furthermore, we cannot exclude the prospect that adequate osteogenesis is a prerequisite for further development and maturation of a nascent vasculature. The possibility exists that bone formation itself creates an environment that subsequently induces or augments angiogenesis in the surrounding tissues and existing vasculature. VEGF is known to be released from the extracellular matrix during fracture repair, allowing it to bind to osteoblasts, osteoclasts, and endothelial cells, potentiating osteogenesis and angiogenesis.(47,48,60–62) Thus, both processes may be required for successful bone formation during GD, perhaps because they are closely linked or even dependent events (Fig. 6).

Figure FIG. 6..

Schematic showing that successful bone regeneration in the setting of distraction osteogenesis requires both a pro-osteogenic mechanical environment and intact angiogenic signaling.

In summary, we have attempted to examine the mechanisms relating angiogenesis and mechanotransductive signaling with osteogenesis by disrupting these processes separately in vivo. By showing that the anti-angiogenic agent TNP-470 does not alter osteoblast biology in vitro, our data show that angiogenesis is required for successful de novo bone formation during gradual distraction. Similarly, we were able to show that a maladaptive mechanical environment prevents both angiogenesis and osteogenesis. Whereas both TNP-470 treatment and alteration of the mechanical environment resulted in a lack of bone formation and angiogenesis, transcriptional analysis revealed important distinctions between these two perturbations in terms of mechanosensitive and angiogenic gene expression. Our data suggest that the osteoinductive mechanical environment created by gradual distraction sets up an environment favorable for blood vessel formation through induction of specific mechanosensitive genes and/or by providing mechanical cues required for vascular ingrowth. The information provided by studying bone formation during GD, GD + TNP, and AD will provide a framework for further delineating the relationship between angiogenesis and osteogenesis during successful distraction osteogenesis and may help design strategies for augmenting bone formation in a clinical setting.


This study was supported by RO1 DE 13028 and The Oak Foundation.

Drs Filvaroff and Carano own stock and are employees of Genentech Inc. All other authors have no conflict of interest.