Thrombospondin-1 and Thrombospondin-2

In normal tissues, the balance of pro- and antiangiogenic factors is in equilibrium. In glioblastoma tumours, the balance of pro- and antiangiogenic growth factors is skewed to favour proangiogenic factors, which favours neovascularisation. Several antiangiogenic factors have been described. Thrombospondin-1 (TSP1) is an extracellular antiangiogenic factor that is synthesised and secreted by multiple normal cells and tissues, such as platelets, endothelial cells, and smooth muscle cells (see Fig. 1 and Table 2) (reviewed by Refs Reference Adams and Lawler45, Reference Lawler46). The antiangiogenic effects of TSP1 treatment have been studied extensively in dermal microvessel endothelial cells propagated in vitro; TSP1 inhibits the migration and proliferation of these cells, and also induces apoptosis (Refs Reference Jimenez47, Reference Volpert48, Reference Nor49). The most potent antiangiogenic effects of TSP1 are induced through the binding of TSP1 to CD36, which is expressed on the surface of microvessel endothelial cells. This interaction is mediated by the type 1 repeat domain (TSR) of TSP1 and the CLESH1 domain of CD36 (Ref. Reference Simantov and Silverstein50). Other proteins can modify the TSP1 response by cooperating with or modulating CD36 function; for example, integrin αvβ3 modulates CD36 function in macrophages (Ref. Reference Bottcher51). In addition, TSP1 also can promote an antiangiogenic effect that is mediated through other domains of TSP1, or is caspase independent (Refs Reference Adams and Lawler45, Reference Elzie and Murphy-Ullrich52, Reference Armstrong53). Peptides derived from the TSP1 TSRs have been administered to mice bearing malignant glioma tumours and have been shown to inhibit tumour angiogenesis and tumour growth, as discussed in more detail later in this review (Ref. Reference Anderson, Grammer, Wang, Nabors, Henkin, Stewart and Gladson54). TSP1 has been shown to act downstream of the tumour suppressor p53 in cells propagated in vitro (Refs Reference Tenan55, Reference Harada56). Loss of p53 function is found in some glioblastoma tumours, yet TSP1 levels are increased in the tumour vasculature, probably because of a host stromal antitumour and antiangiogenesis response (See Table 2 and Fig. 1) (Refs Reference Pijuan-Thompson57, Reference Kawataki58).

Figure 1.

Endogenous inhibitors of angiogenesis in the normal brain and glioblastoma tumours. The cell of origin or cellular localisation. Angiostatin, endorepellin and endostatin are proteolytic cleavage products. Various cell types produce proteases but the relative contribution of each of these cellular sources of proteases to the generation of these cleavage products has not yet been defined. Abbreviations: BAI1, brain angiogenesis inhibitor-1; PF4, platelet factor 4; TIMP, tissue inhibitors of metalloproteinase; TSP, thrombospondin.

Table 2.

Antiangiogenic molecules expressed in glioblastoma tumours: binding partner(s) and localisation

Abbreviations: BAI1, brain angiogenesis inhibitor-1; MMP, matrix metalloproteinase; PF4, platelet factor 4; TIMP, tissue inhibitors of metalloproteinase; TSP, thrombospondin; TSR, thrombospondin type 1 repeat; VEGFR, vascular endothelial growth factor receptor.

TSP2 is highly homologous to TSP1 and also exhibits antiangiogenic properties that are due in part to the TSRs (Refs Reference Adams and Lawler45, Reference Simantov and Silverstein50, Reference Streit59). TSP2 is expressed by endothelial cells and other cells in the normal mouse brain (Ref. Reference Fears60), and its expression is downregulated in glioblastoma tumours compared with levels in the normal brain (see Fig. 1) (Ref. Reference Kazuno61). Furthermore, mouse malignant glioma tumours propagated orthotopically in the TSP2-null mouse showed an increased tumour size and an increased microvessel density compared with tumours propagated in the wild-type mouse (Ref. Reference Fears60), supporting a host antitumour and antiangiogenesis role for TSP2 in the brain.

Brain angiogenesis inhibitor-1

Brain angiogenesis inhibitor-1 (BAI1) is a brain-specific antiangiogenic molecule (see Table 2 and Fig. 1) (Ref. Reference Nishimori62). It is a cell membrane protein that can be cleaved, releasing the extracellular domain, which contains five TSRs and is known as vasculostatin (Refs Reference Nishimori62, Reference Kaur63). In certain experimental conditions, BAI1 may be a p53 target gene (Refs Reference Nishimori62, Reference Kaur64). However, BAI1 expression in glioblastoma tumours is independent of p53 (Ref. Reference Kaur64). BAI1 is expressed differentially in normal brain and glioblastoma tumours. In the normal brain, there is a high expression of BAI1 in normal glia and neurons but no expression in blood vessels. In glioblastoma tumours, BAI1 expression is decreased, probably because of the loss of neurons and normal glia (Ref. Reference Kaur64). Despite the TSR domains, BAI1 functions independently of CD36 expressed on the endothelial cells and it appears to inhibit cell survival through an interaction of the TSRs with integrin αvβ5 (Ref. Reference Koh65). Expression of either full-length BAI1, or the extracellular domain of BAI1, in human glioma cells that were propagated in nude mice resulted in a significant inhibition of tumour growth and tumour angiogenesis compared with similarly propagated control glioma cells (Refs Reference Kaur63, Reference Kang66). This suggests that BAI1 has an antiangiogenic effect in vivo, and that the TSRs of BAI1 are probably necessary for this antiangiogenic effect.

Endostatin and tumstatin

Antiangiogenic molecules, including endostatin, can be produced during the proteolytic cleavage of the basement membrane in glioblastoma tumours (see Table 2 and Fig. 1) (Refs Reference Strik67, Reference Morimoto68). Endostatin is a C-terminal cleavage product of collagen type XVIII by elastase, cathepsin-L and an MMP (Refs Reference Folkman69, Reference Heljasvaara70). Endostatin appears to be capable of interacting with several different receptors or binding partners on the endothelial cells and to inhibit angiogenesis through multiple mechanisms, including: blockade of VEGFR2; binding to integrin α5β1; blockade of focal adhesion kinase (FAK)-mediated endothelial cell migration and survival; binding to tropomyosin; and altered regulation of multiple other targets (such as through the nucleolin receptor) (Refs Reference Folkman69, Reference Shi71, Reference Sudhakar72). Endostatin also can reduce expression of anti-apoptotic Bcl2 family members (Ref. Reference Folkman69). In addition, endostatin binds the proangiogenic protease MMP2 and blocks its activation and catalytic activity (Ref. Reference Folkman69). Endostatin is expressed in glioblastoma tumours (Refs Reference Strik67, Reference Morimoto68), and a higher level of endostatin expression was found in glioblastoma tumours than in lower grade glioma tumour biopsies upon examination by immunoblotting and immunohistochemistry (Ref. Reference Morimoto68). This suggests that endostatin expression is part of a host antitumour response. Several studies have shown that endostatin inhibits malignant glioma growth in rodent models, for example, when recombinant endostatin is delivered to tumours propagated in an ectopic or an orthotopic manner, or when the endostatin gene is delivered to orthotopic tumours (Refs Reference Barnett73, Reference Sorensen74).

Another antiangiogenic molecule derived from the proteolytic cleavage of the basement membrane is tumstatin. Tumstatin is generated by the cleavage of collagen type IV by MMPs, and is thought to inhibit neovessel formation through an interaction with integrin αvβ3 resulting in blockade of FAK, and the PI3K signalling pathway (Refs Reference Sudhakar72, Reference Maeshima75, Reference Maeshima, Colorado and Kalluri76, Reference Mundel and Kalluri77).

Angiostatin

One of the best known of the endogenous antiangiogenic proteins is angiostatin (see Table 2 and Fig. 1). Angiostatin is made up of the first four kringle domains (K1–K4) of plasminogen (Ref. Reference O'Reilly78). Plasminogen is primarily synthesised and secreted by the liver and is present at high concentrations in the plasma; it can be proteolytically digested by proteases, such as cathepsin D and MMPs, to generate angiostatin (Ref. Reference Wahl79). The antiangiogenic activity of angiostatin was first demonstrated in mice bearing subcutaneous Lewis lung carcinoma tumours (Ref. Reference O'Reilly78). A number of studies have now shown that angiostatin inhibits malignant glioma growth in rodent models, for example, when angiostatin is administered to ectopic or orthotopic tumours propagated in immunodeficient mice (Refs Reference Kirsch80, Reference Joe81). Angiostatin appears to mediate its antiangiogenic effects through binding to several different receptors or binding partners on the endothelial cell, including binding to integrin αvβ3, and ATP synthase (Refs Reference Wahl79, Reference Tarui, Miles and Takada82, Reference Chekenya83). Angiostatin has also been shown to upregulate proapoptotic proteins in activated or proliferating endothelial cells (Ref. Reference Wahl79).

The fifth kringle domain (K5) of plasminogen has also been reported to have antiangiogenic activity (Ref. Reference Cao84); however, generation of K5 in vivo through the proteolytic cleavage of plasminogen has not been reported. K5 has been shown to promote several antiangiogenic effects, such as cell-cycle arrest, inhibition of migration and induction of apoptosis, in non-brain endothelial cells propagated in vitro (Refs Reference Cao84, Reference Davidson85, Reference Lu86, Reference Zhang87). The antiangiogenic effects of K5 appear to require binding to cell-surface glucose-regulated protein 78 (GRP78), a heat shock protein family member (Ref. Reference Davidson85). When glioma cells stably expressing K5 are propagated in nude mice the tumours are smaller and demonstrate a reduced angiogenic response (Ref. Reference Perri88).

Angiopoietins

The angiopoietins can exhibit pro- or antiangiogenic activities and have diverse roles in the brain and in the neovasculature of glioblastoma tumours. In normal blood vessels angiopoietin-1 is expressed primarily by pericytes and binds the TIE2 receptor expressed on endothelial cells, leading to an increased association of pericytes with endothelial cells, and to a less leaky, normalised vasculature (Refs Reference Bergers and Song32, Reference Eklund and Olsen89). If this occurred in glioblastoma tumours it could be of importance therapeutically because normalisation of the vasculature may improve antitumour drug delivery and radiation effectiveness. However, in glioblastoma tumours, there is increased expression of both angiopoietin-1 and -2, and angiopoietin-2 is expressed by endothelial cells, whereas angiopoietin-1 is expressed by the tumour cells and is not expressed at the blood vessels (Ref. Reference Zagzag90). Angiopoietin-2 is thought to serve as a TIE2-receptor antagonist (Refs Reference Metheny-Barlow and Li33, Reference Zagzag90, Reference Stratmann, Risau and Plate91). These observations have led investigators to hypothesise that angiopoietin-2 may have a proangiogenic function in the presence of VEGF, such as in angiogenic vessels in which VEGF and angiopoietin-2 are both expressed.

Other antiangiogenic proteins expressed in the brain or glioblastoma tumours, which are not discussed here, include tissue inhibitors of matrix metalloproteinases (TIMPs), such as TIMP1 (Ref. Reference Groft92), platelet factor 4 (Refs Reference Tanaka93, Reference Bikfalvi94); endorepellin, a fragment of perlecan (Refs Reference Mongiat95, Reference Bix96); PEX, a fragment of MMP2 (Ref. Reference Brooks97); and pigment-epithelium-derived factor (Refs Reference Guan98, Reference Zhang99) (see Table 2 and Fig. 1).

Endothelial cells

VEGF/VEGFR

Endothelial cells offer the most logical target for antiangiogenic therapy owing to their relative genetic stability compared with tumour cells, and their direct contact with blood, which could facilitate drug delivery. VEGF-A and the VEGFR2 receptor are the targets of multiple new antiangiogenic drugs, including antibodies directed toward VEGF-A, the targeting of active VEGFR2 with blocking antibodies or small-molecule tyrosine kinase inhibitors, and chimaeric VEGFR2-like molecules or Traps (see Fig. 2 and Table 3) (Refs Reference Willett100, Reference Jain101, Reference Jain102). Antiangiogenic drugs targeting this pathway in a mouse model of glioblastoma tumour and in patients with recurrent glioblastoma tumour decrease the diameter of enlarged tumour vessels and the overall blood vessel size, decrease the thickness of the basement membrane and increase pericyte coverage, resulting in decreased tumour oedema (Refs Reference Jain101, Reference Jain102, Reference Batchelor103, Reference Winkler104). These data suggest that inhibition of the VEGF/VEGFR pathway has a more pronounced effect on the large or dilated vessels in glioblastoma tumours. As inhibition of the VEGF/VEGFR signalling axis in glioblastoma tumours has been reviewed recently (Ref. Reference Jain102), we will not discuss this approach further here. Therapeutic inhibition of VEGFR2 also may be achieved, in part, through the administration of endostatin, which interferes with VEGFR2 signalling (reviewed in Ref. Reference Folkman69). The FGFR1 receptor and IL-8 receptors (CXCR1 and CXCR2) offer similar opportunities for therapeutic targeting, although the specificity of such targeting to angiogenic endothelial cells in glioblastoma tumours has yet to be thoroughly explored (Refs Reference Morrison14, Reference Brat, Bellail and Van Meir25, Reference Stefanik105).

Figure 2.

Localised angiogenic targets in glioblastoma tumours. A schematic of the new targets of antiangiogenic therapy, as discussed in the text, in the neovasculature of glioblastoma tumours. Abbreviations: FAK, focal adhesion kinase; GRP78, glucose-regulated protein 78 kDa; NG2, nerve glial antigen 2; PDGFRβ, platelet-derived growth factor β; SDF-1α/CXCL12, cylokine stromal-derived factor-1α; VEGFR, vascular endothelial growth factor receptor.

Table 3.

Clinical trials with antiangiogenic agents targeting malignant gliomas^a

^aThe information in this table was extracted from the National Institutes of Health (NIH) website http://www.clinicaltrials.gov searching for ‘glioma, brain cancer, glioblastoma and angiogenesis’. Clinical trials that are open, closed, enrolling or completed are included in this table. Irrelevant drug listings are excluded. Many drugs have been utilised in multiple concurrent clinical trials. The primary investigator and clinical trial information can be found on this website.

^bVEGF/VEGFR inhibitors in clinical trials are not discussed in this review, as they were recently reviewed (Refs 101, 102).

Abbreviations: COX2, cyclo-oxygenase-2; NSAID, nonsteroidal anti-inflammatory drug; mAb, monoclonal antibody; PKC, protein kinase C; RTK, receptor tyrosine kinase; VEGF, vascular endothelial growth factor.

Pericytes

Circulating progenitor cells – CXCR4

The circulating progenitor cells that are recruited to angiogenic vessels and may promote angiogenesis are of great interest (Refs Reference Shaked157, Reference Dome158, Reference Jain and Duda159). Whether the recruited progenitor cells are solely endothelial progenitors or are also progenitors of supportive smooth muscle cells and pericytes is an area of active research. The CXCR4 receptor expressed on circulating EPCs is recognised by chemokine CXCL12/SDF-1α; thus, the CXCR4 on these EPCs could be targeted therapeutically with the goal of inhibiting EPC mobilisation from the bone marrow (see Fig. 2) (Ref. Reference Zernecke160). As CXCR4 is expressed on smooth muscle cell progenitors, this approach could also be used to target the recruitment of these cells to angiogenic vessels (Ref. Reference Zernecke160). Inhibition of CXCL12 signalling is likely to have multiple effects; for example, circulating EPCs also can express CXCL12 and can therefore recruit additional EPCs from the bone marrow. The importance of this signalling pathway in vasculogenesis is suggested by the severe vascular abnormalities and associated embryonic lethality exhibited by both CXCL12- and CXCR4-knockout mice (Ref. Reference Tachibana161). Interestingly, CXCL12 is expressed by the pleomorphic angiogenic vessels in the glioblastoma tumours (Refs Reference Salvucci39, Reference Rempel40), suggesting that this neovasculature may recruit circulating EPCs, which would support the continued neovascularisation described above (see Table 1). When mice bearing grafted Lewis lung carcinoma tumours were treated with the TSP1 TSR mimetic ABT-510 or with a monoclonal antibody to VEGFR2, there was a reduction in the number of circulating EPCs and circulating endothelial cells (CECs); in this study the antitumour therapeutic activity of the VEGFR2 antibody was verified by a reduced tumour volume (Ref. 157). These results support the concept that circulating EPCs contribute to angiogenesis in tumours. Thus, further investigation of the relative numbers of circulating EPCs, their biological role in angiogenesis, and the potential efficacy of agents that target these cells in glioblastoma tumours is warranted.

Endothelial cell matrix and basement membrane of the glioblastoma tumour vasculature

Quantitative measures of angiogenesis

The monitoring of antiangiogenic therapeutic agents and delineation of their mechanism of action requires assays that can be used to accurately quantify changes in the tumour neovasculature. Angiogenesis can be quantified in biopsies of tumour tissue using direct counts of the vessel surface area, including small and large vessels, or counts of microvessel density after immunostaining the tissue section with an antibody directed toward von Willebrand factor, CD31 or factor VIII (Refs Reference Folkman4, Reference Anderson, Grammer, Wang, Nabors, Henkin, Stewart and Gladson54, Reference Perri88, Reference Reiher106, Reference Haviv107). Angiogenesis can also be assessed in biopsies of glioblastoma tumour tissue by visualising the decreased pericyte coverage or decreased pericyte density after immunostaining the tissue section with an antibody directed toward NG2, desmin or smooth muscle actin, all of which are markers of pericytes (Ref. Reference Bergers and Song32). Methods for the assessment of angiogenesis in the tumour when biopsy tissue is not available are still under development and this complicates the assessment of the clinical efficacy of antiangiogenic therapy. A radiolabelled anti-fibronectin ED-B isoform antibody (Refs Reference Mariani163, Reference Tarli164) and a radiolabelled RGD peptide (Ref. Reference Janssen124) have been tested as in vivo markers of the neovasculature. A promising approach is the quantification of indirect markers of tumour angiogenesis that can be assessed noninvasively, such as the quantification of CECs and circulating EPCs (Refs Reference Shaked157, Reference Duda172). However, the criteria for the identification of CECs and circulating EPCs are still being standardised. CECs are thought to reflect endothelial cell damage and increased numbers are seen with tumour angiogenesis. The number of circulating EPCs is thought to correlate with ongoing angiogenesis and recruitment from the bone marrow. Currently, multichannel flow cytometry analysing a combination of markers (CD45^–, CD34⁺, CD133⁺ and VEGFR2⁺) is used to identify circulating EPCs (Refs Reference Shaked157, Reference Dome158, Reference Duda172, Reference Mancuso173, Reference Mancuso174). Some investigators include the expression of c-kit (CD117⁺) and CD13 to identify circulating EPCs (Ref. Reference Shaked157). Alteration of the number of circulating EPCs has been used with some success to measure the vascular response to treatment with the TSP1 TSR mimetic ABT-510, a monoclonal antibody to VEGFR2 and a monoclonal antibody to VEGF-A (Refs Reference Willett100, Reference Markovic118, Reference Shaked157). Quantification of circulating EPCs and CECs also may offer a measure of the direct effect of antiangiogenic agents on the mobilisation of EPCs from the bone marrow and the sloughing of endothelial cells from angiogenic tumour vessels, respectively.

Serum or plasma levels of certain proangiogenic factors, such as bFGF and CXCL12, could also potentially be useful markers of angiogenesis. Antiangiogenic therapy with AZD2171, a pan-VEGFR tyrosine kinase inhibitor, results in a decrease in the levels of bFGF and CXCL12 (Ref. Reference Batchelor103). The expression of these two proangiogenic growth factors is upregulated on the vessels or on the pleomorphic angiogenic vessels in glioblastoma tumours (Refs Reference Karcher13, Reference Salvucci39, Reference Rempel40).

Perfusion MRI, also known as dynamic contrast-enhanced MRI, is a relatively new imaging technique that can be used for the assessment of tumour angiogenesis (reviewed in Ref. Reference Barrett175). Using this technique, images can be acquired in a relatively short period of time when a patient presents for a standard contrast-enhanced MRI. Recent reports suggest this technique may be useful in monitoring neovascularisation, because altered perfusion parameters have been demonstrated in brain tumours after antiangiogenic therapy with the RGD-peptide Cilengitide (Refs Reference Nabors120, Reference Akella176).

The blood–brain barrier and antiangiogenic therapy

The BBB is a unique functional unit that limits passive, active and facilitated transport of molecules between the blood and the brain (Refs Reference Hawkins and Egleton177, Reference Ballabh, Braun and Nedergaard178). Key components of the BBB include: tight junctions between endothelial cells, pericytes in the basement membrane and astrocyte endfoot processes (Refs Reference Hawkins and Egleton177, Reference Ballabh, Braun and Nedergaard178). There is a reduced number of fenestrations in brain capillaries compared with capillaries in other organs that are responsible for uptake of molecules, and this limits transport across the BBB (Ref. Reference Hawkins and Egleton177). In addition, brain endothelial cells express relatively high levels of the P-glycoprotein (Pgp/MDR1) drug efflux pump that controls the transport of many chemotherapeutic agents, and this limits active transport across the BBB (Ref. Reference Pardridge179).

In glioblastoma tumours, the BBB is either intact or disrupted depending on whether the vessels are part of the tumour neovasculature or are co-opted vessels (Refs Reference Ballabh, Braun and Nedergaard178, Reference Claes180). The status of the BBB may also be dependent on the spatial localisation of the vessels in the tumour, in areas lacking a neovasculature (for example, some areas of the invasive margin) the BBB is intact; whereas in areas of the tumour where there is a neovasculature, the vessels are leaky (disrupted BBB) (Ref. Reference Muldoon181). A disrupted BBB probably promotes access of large molecules to the tumour. Disruption of the BBB should also result in an initial increase in passive transport of small molecules to the tumour; however, passive transport is probably diminished, because the disrupted BBB becomes chronic owing to an equalisation of the osmotic gradient. These observations result in difficulty in targeting the entire tumour with one drug at any given time, because different properties are required for the drug to gain access to the tumour depending on whether there is a disrupted or intact BBB.

Certain antiangiogenic agents, such as inhibitors of the VEGF/VEGFR signalling pathway, cause a ‘normalisation’ of the BBB; normalisation is defined as a reduction of vessel leakage and a concomitant reduction of interstitial fluid pressure in the brain parenchyma as assessed by perfusion or contrast-enhanced MRI (Ref. Reference Batchelor103). In a recent study, normalisation of the neovasculature occurred post therapy with a VEGFR inhibitor and correlated with an increased pericyte recruitment to neovessels in the tumour, suggesting that pericyte coverage could be an anatomical correlate of vascular normalisation in glioblastoma tumours (Ref. Reference Batchelor103). Other antiangiogenic agents have been shown to reduce vascular permeability associated with disruption of the BBB. For example, endostatin therapy reduced vessel density, vessel diameter and vessel permeability in malignant glioma tumours, as well as inducing a collapse of tumour vessels in a syngeneic intracerebral rat model of malignant glioma (Ref. Reference Barnett73). This reduction in vessel permeability may have been due to an increase in pericyte coverage of tumour vessels because the authors reported an increase in pericyte number with endostatin therapy; however, pericyte coverage was not shown and the quantification of pericytes was not described. Also, the antiangiogenic compound thalidomide administered at teratogenic doses that inhibit angiogenesis (Ref. Reference D'Amato182) may decrease brain vessel permeability induced by stimulation with a β-amyloid peptide (Ref. Reference Ryu and McLarnon183). However, thalidomide used at a similar teratogenic dose was reported to induce increased endothelial cell fenestrations and thinned endothelial cell processes, which correlated with inhibition of bFGF-induced neovascularisation in the rabbit corneal micropocket model of angiogenesis (Ref. Reference D'Amato182).

Normalisation of the tumour neovasculature with VEGF or VEGFR inhibitors may potentially alter the effectiveness of certain chemotherapeutic agents. Recently, a decreased tumour response to temozolomide therapy was observed when temozolomide was administered simultaneously with a VEGFR2 inhibitor in an intracerebral xenograft model of malignant glioma (Ref. Reference Claes180). This was interpreted to be due to a decreased localisation of temozolomide to the tumour, although temozolomide levels in the tumour were not measured. Also, the antiangiogenic compound O-(N-chloroacetyl-carbamoyl)-funagillol (TNP-470) may affect the BBB, because a decreased concentration of temozolomide in the tumour was reported when TNP-470 was administered to an orthotopic mouse model of malignant glioma followed by temozolomide therapy (Ref. Reference Ma184). However, other reports indicate that simultaneous administration of an antiangiogenic agent and a chemotherapeutic agent has beneficial results in orthotopic rodent models of malignant glioma. For example, thalidomide and cisplatin combination therapy resulted in a greater inhibition of tumour growth than that seen with either agent alone (Ref. Reference Murphy185), a synthetic fragment of endostatin and carmustine combination therapy increased survival in a synergistic manner (Ref. Reference Pradilla186) and a novel small-molecule tyrosine kinase inhibitor (AEE788), which blocks the activity of VEGFR2 and EGFR, combined with a mTOR inhibitor (RAD001) yielded greater increases in mouse survival than either agent alone (Ref. Reference Goudar187). Also, some clinical trials of combination therapy with an antiangiogenic agent and a chemotherapeutic agent in patients with malignant glioma or glioblastoma tumours yield promising results, including for example, a study utilising a monoclonal antibody to VEGF-A (bevacizumab) and irinotecan (Ref. Reference Vredenburgh188); a study utilising thalidomide and temozolomide (Ref. Reference Baumann189); and a study utilising thalidomide and carmustine (Ref. Reference Fine190). Therefore, the potential of certain antiangiogenic agents to restore the BBB and reduce the tumour localisation of certain chemotherapeutic agents suggests that the target of specific antiangiogenic therapy, the effects of different antiangiogenic therapies on the BBB and the time course of these effects must be determined to optimally combine antiangiogenic therapy with chemotherapy. Agents that normalise the tumour neovasculature could be used following cytotoxic therapy, if the cytotoxic agent of choice cannot cross the BBB after antiangiogenic therapy. Also, the BBB could potentially be disrupted transiently for chemotherapeutic delivery purposes using ultrasound, convection-enhanced delivery or hypertonic solutions after antiangiogenic therapy (Ref. Reference Kemper191). One caveat of the latter approach is that there is the potential for neurotoxicity or an inflammatory response with disruption of the BBB, because it allows a broad spectrum of plasma proteins to cross into the brain.

Developed resistance to antiangiogenic therapy

Glioblastoma tumours are able to escape the effectiveness of antiangiogenic monotherapy, even though the endothelial cells are relatively genetically stable. For example, one study showed that when dermal fibroblasts overexpressing intact TSP1 were injected subcutaneously along with rat C6 glioma cells into nude mice, there was an initial inhibition of tumour growth and angiogenesis (Ref. Reference Filleur192). However, prolonged exposure of the tumours to TSP1 resulted in an increased expression of proangiogenic growth factors by the tumour that counterbalanced the inhibitory effect of TSP1 on neovascularisation (Ref. Reference Filleur192). Also, in an orthotopic model of gliosarcoma propagated in the rat, downregulation of VEGF-A resulted in an increase in the expression of VEGF-D (Ref. Reference Moffat193), and a study in patients with brain tumours treated with a VEGF inhibitor showed only a temporary normalisation of the tumour vasculature (reviewed in Ref. Reference Jain102). In addition, in a study of 16 patients with recurrent glioblastoma tumours treated with AZD2171 monotherapy (directed toward VEGFR), the 11 patients who experienced disease progression had significantly elevated levels of serum bFGF and CXCL12 (Ref. Reference Batchelor103). These reports highlight the need to move from single antiangiogenic agent therapy to combined antiangiogenic therapy. Moreover, antiangiogenic therapies should be a component of a multi-modality strategy (cytotoxic and antiangiogenic), because progressive tumour growth can occur after antiangiogenic therapy alone, despite reduced angiogenesis (Ref. Reference Leenders194). The latter observation may be due to a co-option of the existing blood vessels by the tumour. In support of this hypothesis, treatment with a monoclonal antibody toward VEGFR2 inhibited glioma angiogenesis in an orthotopic mouse model of malignant glioma; however, numerous satellite tumours arose around the primary tumour and these satellite tumours had a central vessel core (Ref. Reference Kunkel195). These data support the need to combine antiangiogenic therapy with other treatment modalities.

Possible strategies for combining antiangiogenic therapy with other treatment modalities would be to use agents that target more than one pathway (e.g. targeting the fibronectin ED-B isoform in the endothelial cell basement membrane or VEGFR2 on the endothelial cell, with an antiangiogenic or cytotoxic conjugate coupled to an antibody). Alternatively, targeting small and dilated vessels in the tumour with the concurrent use of two antiangiogenic drugs – one that targets microvessels, such as the TSP1 TSR mimetic ABT-510, and a VEGF/VEGFR inhibitor, which has a more pronounced effect on enlarged (dilated) vessels in glioblastoma tumours – could be highly effective. Targeting of both endothelial cells and circulating EPCs (e.g. an antibody directed toward CXCR4 together with a TSP1 TSR mimetic such as ABT-510) warrants consideration. Finally, two recent studies – an animal study targeting VEGF signalling and a clinical trial with AZD2171 in patients with recurrent glioblastoma tumours – suggest that antiangiogenic agents inhibit angiogenesis temporarily, but on cessation of the therapy, renewed tumour neovascularisation occurs (Refs Reference Batchelor103, Reference Casanovas196). This would indicate that cycling of administration of antiangiogenic agents may be required. In conjunction with conventional chemotherapies and surgery, combinatorial antiangiogenic therapies could prove effective, especially if used simultaneously, thereby preventing tumour adaptation and thus drug resistance.