Review of the Effects of Anti-Angiogenic Compounds on the Epiphyseal Growth Plate

Abstract

The formation of new blood vessels from a pre-existing vascular bed, termed “angiogenesis,” is of critical importance for the growth and development of the animal since it is required for the growth of the skeleton during endochondral ossification, development and cycling of the corpus luteum and uterus, and for the repair of tissues during wound healing. “Vasculogenesis,” the de novo formation of blood vessels is also important for the proper function and development of the vascular system in the embryo. New blood vessel formation is a prominent feature and permissive factor in the relentless progression of many human diseases, one of the most important examples of which is neoplasia. It is for this reason that angiogenesis is considered to be one of the hallmarks of cancer. The development of new classes of drugs that inhibit the growth and proper functioning of new blood vessels in vivo is likely to provide significant therapeutic benefit in the treatment of cancer, as well as other conditions where angiogenesis is a strong driver to the disease process. During the preclinical safety testing of these drugs, it is becoming increasingly clear that their in vivo efficacy is reflected in the profile of “expected toxicity” (resulting from pharmacology) observed in laboratory animals, so much so, that this profile of “desired” toxicity may act as a signature for their anti-angiogenic effect. In this article we review the major mechanisms controlling angiogenesis and its role during endochondral ossification. We also review the effects of perturbation of endochondral ossification through four mechanisms—inhibition of vascular endothelial growth factor (VEGF), pp60 c-Src kinase and matrix metalloproteinases as well as disruption of the blood supply with vascular targeting agents. Inhibition through each of these mechanisms appears to have broadly similar effects on the epiphyseal growth plate characterised by thickening due to the retention of hypertrophic chondrocytes resulting from the inhibition of angiogenesis. In contrast, in the metaphysis there are differing effects reflecting the specific role of these targets at this site.

Introduction

Angiogenesis, the growth of new capillaries from a pre-existing vascular bed, is a rare event in most adult tissues, with a turnover rate of approximately 0.01%. However it is required for tissue growth and repair, and is a prerequisite for endochondral ossification (Gerber and Ferrara, 2000), wound healing (Howdieshell et al., 2001), corpus luteum formation (Ferrara et al., 1998), uterine endometrial development (Ferrara et al., 1998), and neonatal growth (Gerber et al., 1999). Vasculogenesis, the de novo growth of new blood vessels, is required for the development of the embryo. Pathological angiogenesis, on the other hand represents a non-physiological form of angiogenesis that is associated with, or even drives the disease process. It is involved in a plethora of diseases (reviewed by Carmeliet, 2003) the most important of which is, arguably, neoplasia.

Human neoplasia and metastasis is absolutely dependent upon angiogenesis one of the six “hallmarks of cancer” (Hahn and Weinberg, 2002; Hanahan and Weinberg, 2000) since in its absence solid masses are limited to a maximum diameter of 2–3 mm—the diffusion limit for nutrients and oxygen (Folkman, 2002). This dependency is illustrated by the observations that tumour “hotspots” (areas of increased vascular density and mitotic activity) carry considerable clinical relevance, correlating with poorer survival in many types of human cancer (Jannink et al., 1995; Belien et al., 1999; Ahn et al., 2001; Mineo et al., 2004). The degree of vascularisation in turn correlates with the likelihood of progression and metastasis (Clayton, 1991; Weidner et al., 1991; Biesterfeld et al., 1995; Hasan et al, 2002).

The discoveries of the dependency of neoplasia (and other diseases) upon angiogenesis, and the elucidation of many of the important molecules controlling angiogenesis have prompted pharmaceutical and biotechnology companies to develop novel treatments that inhibit this process. The results of this effort have propelled the field of angiogenesis, and in particular, inhibition of VEGF into a key area of clinical importance (Cristofanilli et al., 2002; Ferrara and Alitalo et al., 1999; Kerbel and Folkman, 2002) resulting in the development of new compounds to treat human neoplasia. The generation of these novel treatments has of course resulted in significant toxicity to a number of target organs relying on angiogenesis for maintenance and development. This review describes the effects of some of these new compounds on endochondral ossification.

Control of Angiogenesis

Activation of the Angiogenic Switch

Cells readily upregulate pro-angiogenic molecules in response to metabolic stress (hypoxia, decreased pH, and hypocalcaemia), physical stress (pressure), inflammation, and oncogene activation (reviewed by Carmeliet and Jain, 2000). Hypoxia is a potent stimulator of angiogenesis by signalling through a family of hypoxia inducible transcription factors (HIFs) that can upregulate VEGF by up to 30-fold within a few minutes (Pugh and Ratcliffe, 2003). In tumours, this change to the pro-angiogenic phenotype in response to positive and negative regulators of angiogenesis (“the angiogenic switch”) is accompanied by the acquisition of activated oncogenes (Rak et al., 2000; Hanahan and Folkman, 1996). The dual roles that activated oncogenes, such as EGF (epidermal growth factor) (Hirata et al., 2002) and Ras (Thompson et al., 1989) play in the initiation and development of cancer, provides a mechanistic rational for the coupling of continued tumour growth with increased angiogenesis.

Activation of the angiogenic switch, either through mutation and oncogene activation or due to physiological angiogenesis in the growing animal causes vessel destabilisation, endothelial basement membrane/matrix dissolution, and finally sprouting/intussusception, migration and proliferation of endothelial cells. Eventually strings of endothelial cells organise themselves into hollow vascular tubes. Many of these effects can be achieved by vascular endothelial growth factor (VEGF) alone (Figure 1) since it not only promotes endothelial cell activation, migration, proliferation and survival, but also promotes MMP-2 (matrix metalloproteinase 2), MMP-9 and MT1-MMP (membrane type-1 MMP) secretion, and increased vascular permeability (reviewed in Rundhaug, 2003). Increased vascular permeability itself promotes the formation of a provisional matrix scaffold via leakage of fibrin, fibronectin and other plasma proteins (Dvorak, 2002). The provisional stroma is able to bind integrin receptors (e.g.,α5β1 and αvβ1) on activated endothelial cells further promoting migration and endothelial cell survival.

VEGF and the Angiopoietins

The recent discovery of organ specific activators (Ferrara et al., 2004; Pisani et al., 2004; LeCouter et al., 2002) and inhibitors (Albig and Schiemann, 2004; Kee et al., 2002) indicate that there are additional levels of control over angiogenesis. The functional significance that many of these factors contribute to the angiogenic process; however, is still not fully understood. Two key molecules that have received considerable interest and have fairly well defined roles are VEGF and the angiopoietins. The receptors for both these ligands are predominantly expressed upon endothelial cells and regulate the proliferation and survival of endothelial cells.

The discovery of VEGF in 1989 (Leung et al., 1989) (also known as vascular permeability factor (Senger et al., 1983)) catalysed research into angiogenesis. Deficiencies in either vascular endothelial growth factor receptor (VEGFR) or its ligand, VEGF, result in embryonic lethality. VEGFR-1/2 null mice die at embryonic day 8.5 (reviewed in Hanahan, 1997) whereas VEGF+/− hemizygous mouse embryos die at E11 to 12 due to ubiquitous defects in their vasculature. Histologically, defects in yolk sac and embryo vessel morphology are seen with accompanying tissue necrosis (Patan, 2004).

The VEGFs are dimeric endothelial cell mitogens encoded by 5 genes in mammals designated VEGF-A, VEGF-B, VEGF-C, VEGF-D, and placental growth factor (PlGF) (Huang and Bao, 2004). Variations in mRNA splicing generate isoforms of several of the VEGFs, adding to the complexity of the family (McColl et al., 2004). They bind to their cognate receptors, VEGFR-1 (also known as Flt-1) and VEGFR-2 (also known as Flk-1, KDR).

The Angiopoietin-Tie receptor system consists of angiopoietin (Ang) –1, which binds to its cognate receptor tyrosine kinase, Tie-2 (Huang and Bao, 2004). Other angiopoietins have been found (Ang2, Ang3, Ang4), which also bind to Tie-2 but do not necessarily induce phosphorylation (reviewed by Thurston, 2003). No known ligand has been found for Tie-1. Tie-1 and Tie-2 receptor null mice, like VEGF/VEGFR deficient mice die in utero, between E9.5-10.5 (Tie-2) and E13.5-(E18.5 depending on strain) (Tie-1) (Dumont et al., 1994; Sato et al., 1995; Puri et al., 1995). Both Tie-1 and Tie-2 receptor knockout mice develop vascular anomalies, with the Tie-2 mutant mice developing abnormally dilated vessels that lack mural cell support (pericytes and vascular smooth muscle cells (vSMCs)) (Patan, 2004). Recently, the Tie-2 receptor has been shown to be expressed on three distinct cell populations involved in tumour neovascularisation: endothelial cells, proangiogenic bone marrow derived monocyte/macrophages and pericyte precursors of mesenchymal origin (De Palma et al., 2005).

Angiopoietin-2 (Ang-2) binds to the Tie-2 receptor tyrosine kinase and induces a loosening of the attachment between endothelial cells and supporting stromal cells (pericytes and vSMCs) (Maisonpierre et al., 1997). In the absence of VEGF, Ang-2 induces regression of blood vessels (Sato et al., 1995; Dumont et al., 1994). In the presence of VEGF, Ang-2 facilitates vessel sprouting (the formation of blind ending capillary tubes) (Asahara et al., 1998) and intussusception (the folding of slender tissue pillars or posts into the vessel lumen (Patan et al., 1992)). Vessel sprouting is accompanied by basement membrane/matrix dissolution and the loss of endothelial/matrix attachments. VEGF promotes endothelial vascular permeability and vasodilatation as well as survival, migration and proliferation. Under its influence, a provisional scaffold of plasma proteins and fibrin forms.

Fibroblast Growth Factor and Angiogenesis

A third family of growth factors that have received considerable interest are the fibroblast growth factors (FGFs). Fibroblast growth factor receptor (FGFR) –1 mutant mouse embryos exhibit abnormal embryonic vascularisation with reduced endothelial cell migration and survival (Lee et al., 2000). Basic FGF (bFGF; also known as FGF-2) is a pleiotropic mitogen for endothelial and smooth muscle cells (Lindner et al., 1995), and can synergise with VEGF to stimulate angiogenesis in vivo (Asahara et al., 1995). It is secreted by fibroblasts and a variety of tumours such as colon, cervical, ovarian (Salgado et al., 2004) and prostate (Nicholson and Theodorescu, 2004) where it acts as a paracrine/autocrine growth factor to promote tumour growth and co-opt host vasculature. This is mediated through direct stimulation of endothelial cell migration and proliferation (Folkman and Shing, 1992) as well as potentiating the effects of VEGF (reviewed by Gerwins et al., 2000; Javerzat et al., 2002).

The Extracellular Matrix—Role in Angiogenesis

The extracellular matrix consists of a delicate basement membrane that is rich in type IV collagen and laminin. This membrane is supported by a coarser matrix rich in collagens, elastin, laminins, proteoglycans and fibronectin. The shared endothelial/mural cell basement membrane surrounds the pericytes and vSMCs. Pericytes and supporting mural cells sense the external environment through integrin receptors (e.g.,α5β1 (Kim et al., 2000), α1β1, α2β1 (Senger et al., 2002) αvβ3 and αvβ5 (Nisato et al., 2003)) that promote cell survival, growth, migration and/or tube formation. An example of how this operates is provided by fibronectin, which is widely distributed in the extracellular matrix and influences in vivo angiogenesis by binding and enhancing the biological activity of VEGF (Wijelath et al., 2002) as well as promoting cell adhesion, growth (Bourdoulous et al., 1998), migration, and survival (George et al., 1993; Ilic et al., 1998; Re et al., 1994).

Degradation of the extracellular matrix is a necessary prerequisite for angiogenic sprouting (Figure 2). A plethora of intricately balanced proteases and protease inhibitors are involved in this process. The matrix metalloproteinases are considered essential for matrix remodelling and consist of a family of zinc endopeptidases that cleave peptide bonds in extracellular matrix proteins (reviewed by Sottile, 2004). Mice lacking “RECK,” a membrane bound inhibitor of MMP-2, -9 and MT1-MMP (membrane type I-matrix metalloproteinase), die in utero at E10.5, due to defects in collagens, basement membrane and vascular development (Oh et al., 2001). Restoration of RECK function in cell lines suppresses invasion and metastasis (Takahashi et al., 1998). Their differential expression is regulated, at least in part, by growth factor stimulation (e.g., FGF and VEGF (Burbridge et al., 2002)).

Matrix degradation promotes vascular sprouting not only by removing a physical barrier, but by stimulating endothelial cell migration by revealing cryptic binding sites on collagens (Hangai et al., 2002) and virtually all other matrix molecules (Schenk and Quaranta, 2003), releasing latent matrix-bound pro-angiogenic factors e.g., VEGF, FGF (Saksela et al., 1990) transforming growth factor-β (TGF-β) (Rifkin et al., 1999; Egeblad and Werb, 2002), and activating angiogenic chemokines (egIL-1β). Matrix degradation appears to be balanced by protease inhibitors e.g., tissue inhibitors of metalloproteinases (TIMPs), and urokinase plasminogen activator inhibitor (PAI-1) that limit excess degradation, but also, by the action of the proteases themselves, liberate matrix fragments with anti-angiogenic activity e.g., endorepellin (Mongiat et al., 2003), endostatin (O’Reilly et al., 1997), a thrombospondin fragment (Good et al., 1990), angiostatin (O’Reilly et al., 1994) and tumstatin (Maeshima et al., 2000; Petitclerc et al., 2000). This is necessary since excess matrix degradation results in impaired angiogenesis (Bajou et al., 1998) possibly due to removal of critical support and guidance cues for migrating endothelial cells (Anderson et al., 2004).

Mural Cell Support

Pericyte/mural cell recruitment and differentiation accompanies endothelial proliferation. The origin of pericytes is unclear since they can arise from and differentiate into vSMCs (vascular smooth muscle cells) (Nehls and Dreckhahn, 1993; Nicosia and Villashi, 1995) and stromal (myo)fibroblast-like cells (Sundberg et al., 1996). Intravital video and electron microscopic analysis of sprouting capillaries in the rat mesentery indicates that stromal fibroblasts are recruited to sprouting vascular tips, attach and transform into umbrella-like pericytes (Rhodin and Fujita, 1989). The analysis of pericyte precursors in granulation tissue suggests that there is a morphological and phenotypic continuum between fibroblasts, myofibroblasts, pericytes, and vSMCs (Schurch et al., 1997). (Myo)fibroblasts may represent an intermediate phenotype with the capacity to differentiate into pericytes or other stromal cells (Rhodin and Fujita, 1989).

Analysis of the desmoplastic stromal response induced by invasive carcinomas suggests that myofibroblasts express a variety of markers, most commonly vimentin (V) and alpha smooth muscle actin (αSMA), and variably, desmin (D) and smooth muscle myosin heavy chain (myosin H) (Schurch et al., 1997). Pericytes (V + and variably D+, αSMA+) have an intermediate phenotype between myofibroblasts and vascular smooth muscle (V+, myosin H+, αSMA + and variably D+) (Frank and Warren, 1981; Gabbiani et al., 1981; Schmid et al., 1982; Travo et al., 1982; Osborn et al., 1987) leading to the theory that pericytes are specific precursors of vSMCs (Meyrick and Reid, 1979). Recent evidence suggests that the PDGFRβ receptor is a marker for perivascular progenitor cells. These cells subsequently differentiate into PDGFRβ ⁻ perivascular support cells that express the mature pericyte phenotype (aSMA, desmin and NG2) (Song et al., 2005). The heterogeneity of mature pericytes has been further demonstrated since the expression of NG2 is restricted to arteriolar and capillary perivascular cells (smooth muscle and pericytes) but not on venular pericytes (Murfee et al., 2005).

Recruitment of PDGFRβ+ (platelet derived growth factor receptor β) mesenchymal pericyte precursors, appears to be critically dependent upon PDGF-B (platelet derived growth factor-B) which is elaborated by the sprouting vascular tip. PDGF-B induced migration is not straight forward, but relies upon receptor “crosstalk,” since it requires PDGFRβ regulated activation of the EDG-1/ sphingosine-1-phosphate (S1P1) receptor system ( and Hla, 2002; Rosenfeldt et al., 2003). In return, pericytes synthesise and release VEGF, which is a potent endothelial mitogen, as well as a possible pericyte mitogen in areas of hypoxia, (Yamagishi et al., 1999) creating a local circuit promoting endothelial/pericyte co-migration. PDGF-B ligand stimulates pericyte expansion from mesenchymal precursors and subsequent population of nascent blood vessels (Hellstrom et al., 1999). Genetic ablation of PDGF-B or PDGFRβ results in perinatal lethality due to microaneurysm formation and widespread vascular leakage (Leveen et al., 1994; Soriano, 1994; Lindahl et al., 1997) as a result of increased vascular permeability and lack of structural support from pericyte loss.

Attachment of pericytes is promoted by Ang 1, which appears to enhance vessel stabilisation and reduce vascular permeability (Hawighorst et al., 2002; Asahara et al., 1998). TGF-β signalling, occurring mainly during the later stages of angiogenesis, further modifies the angiogenic response. At low concentrations, TGF-β upregulates the angiogenic switch whereas at high concentrations, it stimulates matrix synthesis, induces mesenchymal cell differentiation into (myo)fibroblasts (Pepper, 1997; Chambers et al., 2003) and pericytes ( and D’Amore, 2001) and promotes smooth muscle differentiation (Hirschi et al., 1998).

This is mediated, at least in part, through TGF-Rβ type I receptor/Smad signal transduction pathways that can either induce endothelial proliferation and migration via the ALK1 (activin receptor-like kinase) pathway or inhibit it via the ALK5 pathway (Goumans et al., 2002). Coincident with mural cell recruitment, endothelial cells re-establish cell-cell and cell-matrix junctional communication via, for example, vascular endothelial (VE) cadherins (Breviario et al., 1995), occludins (Wu et al., 2000), JAM-1 in tight junctions ( and Dejana, 2001), CD31 (PECAM-1) (Feng et al., 2004), connexins in gap junctions (Yeh et al., 1998) and integrins. Maturation of the nascent capillary network usually returns the tissue to homeostasis. Pathological angiogenesis however is marked by an absence of resolution.

Xenograft and transgenic mouse tumours demonstrate abnormalities in angiogenesis due to dysregulated synthesis and release of angiogenic factors, which in turn promote continuous capillary proliferation and an immature vascular phenotype. Nascent microvessels show aberrations in mural/pericyte cell recruitment and attachment (Morikawa et al., 2002) and abnormally dilated and tortuous microvessels with increased permeability (Carmeliet and Jain, 2002). It is hoped and believed that these differences will help distinguish tumour vessels from quiescent vessels, and so make them amenable to therapeutic intervention.

Endochondral Bone Ossification

At 6 weeks old, the Sprague–Dawley rat tibial epiphyseal growth plate is growing at approximately 250–300 μm per day (Hansson et al., 1972). At the mineralising front, this mass of cartilage is replaced by woven bone in a process known as endochondral ossification. The regulated coupling of cartilage growth and ossification serves to maintain the growth plate at a relatively constant thickness. During this process, chondrocytes within the epiphyseal growth plate undergo a carefully coordinated programme of chondrocyte proliferation with matrix (cartilage) synthesis, followed by maturation, hypertrophy, matrix calcification and ultimately chondrocyte apoptosis (summarised in Figure 3). This process is critically dependent upon vascular invasion that occurs at the cartilage-bone interface.

Immature and proliferating chondrocytes secrete and express angiogenic inhibitors (Moses et al., 1999) and an extracellular matrix (ECM) rich in aggrecan and type II collagen (reviewed in Orth, 1999). Enlargement and terminal differentiation into hypertrophic chondrocytes now allows these cells to act as regulators of the process and engage a genetic programme characterised by expression of type X collagen (Alini et al., 1994), VEGF (Gerber et al., 1999; Horner et al., 1999; Carlevaro et al., 2000; Maes et al., 2002), connective tissue growth factor (Ivkovic et al., 2003), acidic and basic FGF (Baron et al., 1994; Nagai and Aoki, 2002), MMP-13 (Nagai and Aoki, 2002) and the enzyme alkaline phosphatase (Vaananen, 1980). Hypertrophic chondrocytes also deposit mineral and partially degrade the ECM before undergoing programme cell death.

Cells migrating inwards from the marrow space also contribute to ECM degradation by secreting MMPs—the most important of which being MMP-2, MMP-9 and MMP-13 that degrade non mineralised matrix (reviewed in Blair et al., 2002) resulting in loss of the “last transverse septum.” This septum marks the boundary between apoptotic hypertrophic chondrocytes and the migrating vascular invasion front.

Hypertrophic chondrocytes express a number of VEGF isoforms (Maes et al., 2002; Petersen et al., 2002). Long splice forms, such as murine VEGF₁₆₄ and VEGF₁₈₈ show increased binding to heparin sulphate containing proteoglycans present on the cell surface and extracellular matrix (Ferrara and Davis-Smyth, 1997), whereas the short murine VEGF₁₂₀ splice form is completely soluble and unable to bind heparin (Ferrara and Davis-Smyth, 1997). The long VEGF splice forms (Maes et al., 2002) appear to be the most important isoforms mediating endochondral bone ossification (Maes et al., 2002). During ECM degradation, MMPs release the matrix bound long VEGF isoforms, which together with other processed angiogenic growth factors, further encourage the inwards migration of vascular endothelium (Petersen et al., 2002).

Capillary invasion into the empty lacunae of terminal hypertrophic chondrocytes now allows access and functional differentiation of other VEGFR expressing nonvascular cells—notably osteoblasts (Midy and Plouet, 1994; Deckers et al., 2000) and osteoclasts (Engsig et al., 2000; Maes et al., 2002). Apoptotic chondrocytes leave behind a mineralised cartilage matrix that becomes the scaffold for invading osteoblasts and osteoclasts. Osteoblasts differentiate on the plates of mineralised cartilage between columns of hypertrophic chondrocytes and lay down delicate spicules of woven bone (primary trabeculae), whereas osteoclasts remove mineralised matrix.

Analysis of MMP-9^{− / −} and MMP-13^{− / −} deficient mice indicate that these are the two most active MMPs involved in epiphyseal growth plate endochondral ossification and operate in concert to degrade the unmineralised hypertrophic chondrocyte septae. They are expressed in two different compartments at the chondro-osseous interface. MMP-9 appears to be the most important, being expressed at the invasion front by endothelial cells, monocytes, preosteoclasts, osteoclasts and chondroclasts (reviewed in Ortega et al., 2004). MMP-9 is not only necessary for ECM degradation, a prerequisite for angiogenesis, but also appears to play a role in the processing of kit ligand (stem cell factor) for the recruitment and mobilisation of bone marrow derived stem, haemopoietic and endothelial precursors (Hessig et al., 2002). MMP-13, derived from hypertrophic chondrocytes and osteoblasts at the chondro-osseous border, synergises with MMP-9 during ECM degradation.

Primary trabeculae are modelled into fewer and thicker secondary and tertiary trabeculae located more distally in the diaphysis (Porter et al., 1997). The modelling of bone consists of a “coupled” process of osteoclastic resorption and osteoblastic formation. Osteoblasts are thought to provide osteoclastic help by removing the organic phase to expose bone mineral (Chambers and Fuller, 1985). Osteoclasts remove mineralised matrix. They are polarised cells with a ruffled border that enables them to create an acidic microenvironment into which they release lysosomal enzymes, resulting in the decalcification and degradation of bone (Baron, 1989). The formation of the ruffled border, necessary for bone resorption, is dependent upon pp60 c-Src (Boyce et al., 1992; Soriano et al., 1991) and the micropthalmia transcription factor, Mitf (Hodgkinson et al., 1993). The uncoupling of these two processes—osteoblastic bone formation and osteoclastic resorption—is the basis of skeletal imbalances characterised by increased (osteopetrosis) or decreased (osteoporosis) bone density.

Vascular Endothelial Growth Factor Receptor-2 Tyrosine Kinase Inhibitors

A number of approaches have been taken to inhibit the VEGF/VEGFR-1/2 signalling complex through, for example, inhibition of VEGF ligand (Asano et al., 1998; Rowe et al., 2000; Ferrara et al., 2004), inhibition of the VEGF receptor by soluble antibodies (Roeckl et al., 1998; Brekken et al., 2000), or through inhibition of the tyrosine kinase domain by small molecules (Wedge et al., 2000; Wood et al., 2000; Beebe et al., 2003; Mendel et al., 2003; Sepp-Lorenzino et al., 2004). This approach has proved promising since pre-clinically, inhibition of VEGF has been shown to suppress xenograft tumour growth in nude mouse models (Kim et al., 1993). In addition, an anti-VEGFR-2 neutralising antibody was shown to synergise with low-dose vinblastine to produce full and sustained inhibition in a neuroblastoma xenograft model (Klement et al., 2000).

This work has been continued in human patients, where the recombinant humanised anti-VEGF-A antibody (Avastin; Genetech) during phase II and phase III clinical trials showed synergy with standard chemotherapy in a number of malignancies, including colorectal (Midgley and Kerr, 2005) and advanced metastatic non-small cell lung carcinoma (Kabbinavar et al., 2003; Ferrara et al., 2004; Sandler et al., 2004). Avastin has now been approved in the United States as a first-line therapy for metastatic colorectal cancer (reviewed by Ferrara et al., 2004). In addition, several small molecule inhibitors of the VEGFR-1 and/or 2 tyrosine kinase, have now entered clinical trials and are expected to show clinical benefit.

The effect of inhibition of VEGF signalling on bone growth has been well described (Gerber et al., 1999; Wedge et al., 2000; Beebe et al., 2003) and results in thickening of the epiphyseal growth plate due to expansion of the hypertrophic zone (Figure 4b). These effects have been attributed to delayed vascular invasion of the epiphyseal growth plate resulting in a reduced rate of hypertrophic chondrocyte apoptosis. Similar effects have also been noted with a small molecule inhibitor of the FGF receptor (Brown et al., 2005). The changes in growth plate thickness can be morphmetrically quantified revealing a dose-dependent increase in the epiphyseal area of up to 481% (Wedge et al., 2005). These changes appear to be restricted to the hypertrophic zone characterised by the accumulation of relatively parallel columns of hypertrophic chondrocytes (Wedge et al., 2000). The reserve, proliferative and maturing zones of the growth plate appear normal. Minimal retention of hypertrophic chondrocytes also occurs in the articular-epiphyseal complex (epiphyseal growth cartilage) of the femoral and tibial condyles (Figure 4f). Since the normal rat femur would be expected to grow approximately 5 mm (female) to 5.5 mm (male) in length between 6 and 10 weeks of age (Hannson et al., 1972) it is likely that despite VEGFR inhibition, some endochondral ossification occurs at a basal level.

Inhibition of VEGF not only interferes with vascular invasion, resulting in growth plate thickening, but also impairs trabecular bone formation. Mice treated with a soluble VEGF receptor chimeric protein, mFlt(1-3)-IgG showed reduced length and number of primary trabeculae with thicker secondary trabeculae (Gerber et al., 1999) (Figure 5b). This observation can be accounted for by both reduced ossification of apoptotic hypertrophic chondrocytes (Gerber et al., 1999) and impaired recruitment and migration of monocyte/macrophage-derived chondroclasts and osteoblasts. These specialised cells are necessary for the formation of bone trabeculae, express VEGF-R1, and migrate in response to VEGF ligand (Midy and Plouet, 1994; Hiratsuka et al., 1998; Mayr-Wohlfart et al., 2002). Restoration of normal angiogenesis by discontinuation of mFlt(1-3)-IgG treatment allows rapid reversal of all growth plate changes within 2 weeks (Gerber et al., 1999). Cessation of treatment reverses femoral shortening and restores the growth plate to a normal histological appearance. This is attributed to renewed vascular invasion inducing resorption of hypertrophic chondrocytes, and modelling of newly formed primary trabeculae into secondary trabeculae.

In addition to changes in the growth plate induced by VEGFR antagonists, we have also observed dental dysplasia in the growing rodent incisor (unpublished observations), and ovarian atrophy (reduced numbers of corpora lutea) (Wedge et al., 2005). Ovarian atrophy is an expected consequence of VEGF inhibition since VEGF expression is essential for the capillary in-growth into the normally avascular Graafian follicle and conversion to the highly angiogenic corpus luteum (Phillips et al., 1990; Ferrara et al., 1998).

Vascular Disrupting/Targeting Agents

Vascular targeting agents destroy nascent endothelium in a selective way by exploiting unique markers on tumour endothelium. The concept of vascular targeting was demonstrated in 1993 when a ricin-A chain immunotoxin was directed to vascular endothelium by coupling ricin to anti-major histocompatibility complex class II (anti-MHC II) homing antibodies in a neuroblastoma xenograft transfected with IFN-γ (which upregulates MHC II expression) (Burrows and Thorpe, 1993). The uniqueness of tumour vasculature has since been demonstrated: Olson et al. (1997) linked diphtheria toxin to a recombinant VEGF and showed anti-tumour activity because tumour endothelium effectively over-expresses VEGFR compared to other vascular beds. Heinis et al (2004) have developed an Antibody-Directed Enzyme Prodrug Therapy strategy (ADEPT) in which an antibody that recognises the fibronectin extra domain B (EDB) (an oncofetal fibronectin and marker of angiogenesis) was coupled to an engineered prolyl endopeptidase mutant. This enzyme can efficiently activate a prodrug, glycylprolylmelphalan, and thus deliver melphalan to the tumour vasculature. Other authors have also reported successful targeting of tumour vasculature with homing peptides (Akerman et al., 2002; Ruoslahti, 2000), integrin targeted ligand (Hood et al., 2002; Yokoyama and Ramakrishnan, 2004) and a VEGFR-2 oral DNA vaccine (Niethammer et al., 2002).

Recently, low molecular weight vascular disrupting agents have also been developed that target tumour vasculature, such as combretastatin A4 phosphate (Griggs et al., 2001) and ZD6126 (Blakey et al., 2002a). These compounds disrupt the tubulin cytoskeleton of immature tumour endothelium (Davis et al., 2002) and destabilise microtubules resulting in mitotic cell arrest/death (Kanthou et al., 2004) and rapid endothelial cell retraction/exfoliation. This culminates in exposure of vascular basement membrane, thrombus formation and, usually within 24 hours, large areas of ischaemic necrosis of the tumour core (Blakey et al., 2002b). These compounds have now entered clinical trials (Thorpe et al., 2003). Consistent with an anti-angiogenic effect, we have observed mild, often focal thickening of the epiphyseal growth plate together with necrosis of sub-physeal osteocytes (Figure 4c and 5f) (unpublished observations).

The local nature of both these lesions, metaphyseal osteocyte necrosis and growth plate thickening, may indicate relative insensitivity of the peripheral areas of the growth plate vasculature to vascular disruption by these agents. This would not be unexpected since treatment with combretastatin A4 phosphate and other tubulin-binding agents produces a characteristic pattern of necrosis in xenograft models, limited to the central core (West and Price, 2004). The rim is usually viable, reflecting relative vascular resistance.

pp60 c-Src Kinase Inhibition

pp60 c-Src kinase is a pleiotropic signalling molecule involved in several aspects of cancer biology such as angiogenesis (Mukhopadhyay et al., 1995; Kumar el al., 2003), metastasis (Sawyer, 2004) and cell migration (Kilarski et al., 2003). It is therefore thought to play an important role in the pathogenesis of cancer, rheumatoid arthritis and other diseases (Kilarski et al., 2003).

pp60 c-Src kinase is thought to form part of the VEGFR-2 (KDR) downstream signal transduction pathway necessary for the hypoxic induction of VEGF (Mukhopadhyay et al., 1995), and coupling of focal adhesion kinase to αvβ5 integrin for VEGF signalling (Eliceiri et al., 2002). Mice deficient in pp60 c-Src Kinase show defective VEGF induced angiogenesis and reduced VEGF induced vascular permeability (Eliceiri et al., 1999). Furthermore, expression of pp60 c-Src Kinase is associated with hypertrophic chondrocyte synthesis of type X collagen and subsequent mineralisation (Coe et al., 1992). However, despite the apparent dependence of VEGF signalling on pp60-Src Kinase and its expression within the growth plate, pp60-Src Kinase^{− / −} deficient mice develop metaphyseal rather than growth plate changes (Soriano et al., 1991).

pp60 c-Src forms part of a multimeric signalling protein involving gelsolin, pp60 c-Src and phosphatidylinositol 3′-kinase (Chellaiah et al., 1998, 2001). Gelsolin null mice show abnormalities in osteoclastic resorption causing thickening of metaphyseal trabeculae with abnormal trabecular architecture (Chellaiah et al., 2000). Homozygous c-Src^{− / −}-deficient mice develop osteopetrosis (Soriano et al., 1991). Histologically, the bone marrow space becomes occupied with increased bone volume that architecturally appears to be an extension of abnormally modelled primary trabeculae into the diaphysis (Figure 5c). This phenotype is thought to be mediated through an osteoclast defect that renders them inactive and unable to adhere to bone and/or form a bone resorbing ruffled border (Lowe et al., 1993). c-Src^{− / −}-deficient mice also show enhanced osteoblast differentiation and bone formation that contributes to the trabecular changes (Marzi et al., 2000).

The precise role that pp60 c-Src Kinase plays in osteoclastic bone resorption however is still unclear. Some authors suggest that only the adaptor function of pp60 c-Src is necessary for bone resorption (Schwartberg et al., 1997; Felsenfeld et al., 1999). Others (Miyazaki et al., 2004) have suggested that both Src kinase activity and Src SH2-dependent formation of the Pyk2/Src complex are necessary for bone resorption.

Broad Spectrum Matrix Metalloproteinase Inhibitors

MMPs degrade most components of the extracellular matrix, including the endothelial/mural basement membrane to promote endothelial cell sprouting and migration (reviewed by Sottile, 2004 and Noel et al., 2004), a prerequisite for both physiological angiogenesis and tumour angiogenesis. It also facilitates tumour growth, metastasis and invasion (Seiki and Yana, 2003) through the processing of growth factors, cytokines, receptors and adhesion molecules (reviewed in Seiki et al., 2003). The role that a number of metalloproteinases play in angiogenesis and tumour progression has been well demonstrated (Kim and Kim, 1999; London et al., 2003; Jadhav et al., 2004; Zijlstra et al., 2004) and correlated with histologic grade in spindle soft tissue neoplasms (Roebuck et al., 2005), and growth, invasion and metastasis in gastric cancer (Huachuan et al., 2003). Small molecule MMP inhibitors such as FYK-1388, SI-27, and BMS-275291 have been shown to reduce angiogenesis in vitro (Shinoda et al., 2003), in animal models (Yoshida et al., 2004), and in patients (Lockhart et al., 2003). The role of specific MMPs, such as MMP-9 and MMP-13 has been demonstrated in growth plate angiogenesis. MMP-9 expressing cells, together with CD31 vascular endothelial cells, form part of the invasion front during vascularisation of the cartilage that precedes endochondral ossification (Takahara et al., 2004). MMP-13 expression has similarly been demonstrated in perivascular cells in the growth plate erosion zone (Nakamura et al., 2004) as well as localised to the terminal layer of the hypertrophic chondrocyte columns (Uusitalo et al., 2000; Nagai and Aoki, 2002). MMP-13 expression is also associated with the synthesis of collagen type X, which precedes chondrocyte lysis and mineralisation (Wu et al., 2002). The use of genetically modified mouse models, deficient in one or more MMPs has further helped to elucidate the role played by these proteases during endochondral ossification. Mice with MMP-9/gelatinase B null mutation and mice deficient in MMP-13 develop profound defects in growth plate cartilage (Vu et al., 1998; Inada et al., 2004) characterised by epiphyseal growth plate thickening and massive accumulation of hypertrophic chondrocytes (Figure 4d).

Genetic disruption of the FGFR-3 (Deng et al., 1996), treatment with a FGF receptor small molecule tyrosine kinase inhibitor (Brown et al., 2005) or supplementation with FGF-2 (Nagai and Aoki, 2002) all produce a similar effect on the growth plate. Growth plate thickening is thought to be mediated through one or more mechanisms-directly through inhibition of angiogenesis, or indirectly through inhibition of cartilage degradation, chondrocyte differentiation and vascular invasion (MMP-9 and FGF-2/MMP-13) preventing the exit of hypertrophic chondrocytes from the growth plate (Stickens et al., 2004) or through relief of chondrocyte growth suppression (FGFR-3). Recent studies with MMP-13 deficient mice indicate that MMP-13 expression is upstream of growth plate angiogenesis and that degradation of cartilage extracellular matrix is a prerequisite for vascular invasion (Stickens et al., 2004). This view is further supported by the observation that treatment of wild-type mice with galactin-3 (a matricellular protein) phenocopies the growth plate defect seen in MMP-9-deficient mice, indicating that this protein may be a physiologically relevant substrate for MMP-9 during endochondral bone ossification (Ortega et al., 2005).

The role of a number of MMPs has also been documented in the turnover and remodelling of bone in the metaphysis (Kusano et al., 1998; Pelletier et al., 2004). MMP-1, MMP-9, MMP-12 and MMP-13 expression has been immunolocalised under the ruffled borders of osteoclasts and on bone surfaces in rodents (Delaisse et al., 1993; Nakamura et al., 2004) and in osteoclasts in humans (Okada et al., 1995; Hou et al., 2004). MMP-13-deficient mice show metaphyseal changes characterised by increased trabecular bone (Figure 5d). This effect was thought to be due to MMP-13 deficiency in the osteoblasts and therefore independent of the effect in the growth plate cartilage (Stickens et al., 2004).

Systemic treatment of growing rats with FGF-2 causes metaphyseal changes which appeared morphologically as a disruption between the cartilage columns and trabecular bone, possibly mediated through diminished expression of MMP-13 in the terminal hypertrophic chondrocytes (Nagai and Aoki, 2002). These changes were characterised by a discontinuity between the metaphyseal trabecular bone and the hypertrophic chondrocyte columns (Figure 5e) resulting in separation and the appearance of horizontally laid vascular channels (Nagai and Aoki, 2002). Disruption of the chondro-osseous junction also been shown in a human case of achondroplasia (Briner et al., 1991) and the skeletal phenotype of FGFR-3 (G380) transgenic mice (Segev et al., 2000).

Concluding Remarks

The requirement for angiogenesis during the normal development of vascular tissues such as the growth plate and the Graafian follicle is well established. The key molecules that regulate this process during normal physiology also play crucial roles in the control of the aberrant angiogenesis found in tumours. It is not surprising therefore that inhibitors of tumour angiogenesis produce a recognisable anti-angiogenic signature in growing animals that can be assessed during preclinical safety studies. The growth plate (and other organs such as the ovary) appear to be a very sensitive biomarker of this effect and thus can contribute to the overall data set supporting the mechanism of action of these compounds in vivo.

It is likely as more experience is gathered in this area that other toxicities will be reported, specific for the intended target. These target organs will provide other opportunities to determine the efficacy and mechanism of action of novel angiogenic inhibitors, in the context of essentially normally regulated angiogenesis. It will be interesting therefore to speculate how predictive these biomarkers will be for human toxicity, and for the efficacy of anti-angiogenic compounds in human clinical trials when compared to efficacy data collected in vitro and in the standard xenograft model.