The biology of the combretastatins as tumour vascular targeting agents

Abstract

The tumour vasculature is an attractive target for therapy. Combretastatin A-4 (CA-4) and A-1 (CA-1) are tubulin binding agents, structurally related to colchicine, which induce vascular-mediated tumour necrosis in animal models. CA-1 and CA-4 were isolated from the African bush willow, Combretum caffrum, and several synthetic analogues are also now available, such as the Aventis Pharma compound, AVE8062. More soluble, phosphated, forms of CA-4 (CA-4-P) and CA-1 (CA-1-P) are commonly used for in vitro and in vivo studies. These are cleaved to the natural forms by endogenous phosphatases and are taken up into cells. The lead compound, CA-4-P, is currently in clinical trial as a tumour vascular targeting agent. In animal models, CA-4-P causes a prolonged and extensive shut-down of blood flow in established tumour blood vessels, with much less effect in normal tissues. This paper reviews the current understanding of the mechanism of action of the combretastatins and their therapeutic potential.

Introduction

The tumour vasculature is an attractive target for therapy because it is easily accessible by blood-borne anticancer agents and most tumour cells rely on an intact vascular supply for their survival (Denekamp 1982; Folkman 1997; Chaplin & Dougherty 1999). Most of the research effort in this area has focused on understanding the process of angiogenesis (the out-growth of new blood vessels from pre-existing ones) and identifying pro- and antiangiogenic factors (Lewis et al. 1997). Measurement of proliferation indices for endothelial cells have confirmed active angiogenesis, not only in rapidly growing animal tumours (Denekamp & Hobson 1982) but also in human tumours (Eberhard et al. 2000). Physiological angiogenesis occurs in only a limited number of situations such as wound healing and during the menstrual cycle and therefore antiangiogenic strategies have potential to be highly specific to tumours with little toxicity. In addition, the target endothelial cells are nontransformed and this implies that treatment-resistant mutations are unlikely to emerge (Boehm et al. 1991; Kerbel 1997). The early stages of angiogenesis (breakdown of basement membrane, endothelial cell migration and proliferation) have been the main targets for the development of antiangiogenic strategies and several classes of antiangiogenic agents are currently in clinical trials for cancer (see Eatock et al. 2000 and Los & Voest 2001 for recent reviews). These include inhibitors of matrix-metalloproteinases (e.g. Marimistat), agents that inhibit vascular endothelial growth factor or its receptor (e.g. SU5416), inhibitors of endothelial cell growth (e.g. TNP-470) and the endogenous peptide inhibitors of endothelial cell proliferation, angiostatin (O'Reilly et al. 1996) and endostatin (O'Reilly et al. 1997). In addition, it has been recognized that many conventional chemotherapeutic agents that target the proliferative capacity of tumour cells may also be active against proliferating/angiogenic endothelial cells. Exploitation of this effect is likely to require modification of conventional acute dosing regimes to continuous dosing strategies, in order to have a substantial effect on the dynamic process of new blood vessel development (Klement et al. 2000).

In contrast to the antiangiogenesis approach, antivascular strategies aim to cause a rapid and extensive shut-down of the established tumour vasculature, leading to secondary tumour cell death. Cell death following blood flow shut-down, induced by clamping or ligation of the tumour-supplying blood vessels, is characterized by an early and extensive tumour cell necrosis (Denekamp et al. 1983; Kim et al. 1973). Therefore, this pattern of cell death following treatment is indicative of vascular-mediated cytotoxicity. There is potential for specific targeting of the tumour vasculature by development of drugs, antibodies or genetic vectors, based on selective expression of proteins on tumour endothelial cells. Recent development of techniques for the isolation of tumour-derived endothelial cells and gene expression have led to the identification of a number of gene transcripts, which are specifically elevated in tumour-associated endothelium. In endothelial cells isolated from human colorectal cancer, 46 of 170 transcripts, predominantly expressed in endothelial cells, were significantly elevated above levels in the corresponding normal cells (St Croix et al. 2000).

The function of the majority of these genes is unknown (St Croix et al. 2000). However, their recognition holds the promise of selective tumour vascular targeting. In addition to active angiogenesis, they are likely to be associated with a range of characteristics of the established tumour vasculature, which may eventually be possible to target at the molecular level. For example, tumour blood vessels are often described as ‘immature’, relatively lacking investment with vascular smooth muscle cells or pericytes (Eberhard et al. 2000) and highly permeable (Dvorak et al. 1988), which contributes to high interstitial fluid pressure (Boucher et al. 1990). Tumour vascular networks are complex, seemingly chaotic, leading to heterogeneous blood perfusion (Tozer et al. 1990) and oxygenation (Vaupel et al. 1992).

Current approaches to tumour vascular targeting

Selection of phage display libraries in experimental animals bearing human tumour xenografts have been used to identify specific markers on tumour vasculature. Using this technique, peptides that recognize integrin binding sites on endothelial cells were found to home to tumour vasculature (Pasqualini & Ruoslahti 1996). In order to test the therapeutic possibilities of this approach, integrin binding peptides containing the Arg-Gly-Asp (RGD) and Asn-Gly-Arg (NGR) motifs were conjugated to the anticancer drug doxorubicin. Results showed that the efficacy of doxorubicin was increased for the conjugated drug compared to doxorubicin alone (Arap et al. 1998).

Monoclonal antibodies designed to target tumour endothelial cell-specific epitopes have also been tested in animal studies. Antibodies that bind specifically to tumour endothelial cells have been identified from the isolation of these cells from solid tumours (Makimoto et al. 1999). Monoclonal antibodies targeted to endothelial-specific receptors such as endoglin, a component of the transforming growth factor beta receptor complex, which is up-regulated in angiogenic areas of tumours, have shown promise (Burrows et al. 1995). Cytotoxic strategies have involved conjugation of the antibodies with a range of agents including radio-isotopes, cytotoxic drugs and tissue factor for inducing coagulation specifically in tumour blood vessels (Huang et al. 1997).

Primarily as a result of screening studies, two main classes of drugs have been found to induce vascular damage and subsequent haemorrhagic necrosis in tumours. These are the tubulin-binding agents (Chaplin et al. 1996; Hill et al. 1993), which include the combretastatins, and drugs related to flavone acetic acid (FAA) (Bibby et al. 1989; Hill et al. 1989; Zwi et al. 1989; Hill et al. 1995). A Phase I clinical trial of dimethylxanthenone-acetic acid (DMXAA), a derivative of FAA, has just been completed in the UK and New Zealand. Its mechanism of action is not entirely clear, although there is evidence for involvement of both intratumoural production of tumour necrosis factor-α (TNF-α) (Mahadevan et al. 1990) and serotonin (5-HT) (Baguley et al. 1997) in its vascular effects. It is now well established that the haemorrhagic and coagulative necrosis induced by TNF-α in solid tumours is primarily due to selective damage to the tumour vasculature (Kallinowski et al. 1989). TNF-α itself has been considered as a tumour vascular targeting agent. However, its maximum tolerated dose in humans is approximately 10 times lower than the effective dose in animals and its systemic administration in the clinic, at tolerated doses, provided little evidence of antitumour effects (Lejeune et al. 1998). Nevertheless, TNF-α does have potential for regional treatments, such as in isolated limb perfusions for the treatment of advanced soft tissue sarcomas of the extremities (Eggermont et al. 1997).

It is worth noting that although we have drawn a distinction between antiangiogenic strategies and antivascular strategies for cancer therapy, there is likely to be an overlap between the two. For instance, VEGF has been described as a survival factor for endothelial cells (Alon et al. 1995). Its withdrawal can lead to endothelial cell apoptosis and vascular regression in an immature vascular bed such as found in a tumour (Benjamin et al. 1999; Yuan et al. 1996). Thus a strategy designed to prevent angiogenesis can also cause direct vascular damage in tumours, leading to the development of necrosis (Benjamin et al. 1999). Conversely, agents designed to cause acute vascular damage, such as the combretastatins, also have antiproliferative effects (see Discussion, below). A practical distinction between the two approaches is in the dosing strategies employed. In order to prevent angiogenesis, a chronic dosing schedule is appropriate, whereas single dose or split dose treatments are effective for inducing vascular shut-down.

Tubulin binding agents

The classic tubulin binding agent, colchicine, was recognized as having damaging effects on tumour vasculature as long ago as the 1930s and 1940s, causing haemorrhage and extensive necrosis in both animal and human tumours (Boyland & Boyland 1937; Seed et al.1940; Ludford 1948). Although its toxicity precluded further clinical development, sporadic reports of tumour vascular damage induced by related compounds, such as podophyllotoxin, continued to emerge (Algire et al. 1954). More recently, the tubulin-binding agents vincristine and vinblastine, which are potent, clinically active anticancer drugs, have also been recognized to have antivascular effects in animal tumours, at close to their maximum tolerated doses (Baguley et al. 1991; Hill et al. 1993; Hill et al. 1995). Arsenic trioxide which, amongst other effects, targets vicinal thiol groups on tubulin, has also been shown to elicit profound tumour vascular effects (Lew et al. 1999). TZT-1027 (auristatin PE), an antimicrotubule agent derived from the naturally occurring dolastatin 10, isolated from a marine mollusc, is currently in clinical trial in Japan and has similar properties (Otani et al. 2000). An analogue of colchicine, N-acetylcolchinol-O-phosphate or ZD6126, is also undergoing preclinical and clinical testing (Davis et al. 2000).

The combretastatins were originally derived from the African willow tree Combretum caffrum (Pettit et al. 1987a). They are structurally related to colchicine, containing two phenyl rings, which are tilted at 50–60° to each other and linked by a two-carbon bridge, with several methoxy substitutions on the ring system (Fig. 1). Combretastatin A-1 (CA-1) and combretastatin A-4 (CA-4) (Pettit et al. 1987b; Pettit et al. 1989) are the most active of 17 compounds isolated from the natural source, in terms of their effects on tumour cells in vitro and their interaction with tubulin, CA-4 being the most active (Lin et al. 1988; McGown & Fox 1990). CA-4 has been shown to interact with tubulin at or near the colchicine binding site, an action not shared with the vinca alkaloids (McGown & Fox 1989). The cis configuration along with other structural features, have been shown to be important for biological activity (Woods et al. 1995).

Figure 1.

Chemical structures of(a) combretastatin A-4-phosphate (CA-4-P); (b) combretastatin A-1-phosphate(CA-1-P); (c) AVE8062; (d) colchicine.

Further development of CA-1 and CA-4 was complicated by limited water solubility. In the mid-1990s, this was solved for CA-4 by the development of the sodium phosphate salt, disodium combretastatin A-4 3-O-phosphate (CA-4-P) (Pettit et al. 1995). This compound acts as a pro-drug, and is cleaved to the active CA-4 by endogenous nonspecific phosphatases, which are then taken up into cells. More recently, a sodium phosphate pro-drug of CA-1 (CA-1-P) has been developed (Pettit & Lippert 2000). In contrast to colchicine and other previously available tubulin binding agents, the antivascular effects of CA-4-P (Dark et al. 1997) and CA-1-P (unpublished data) in vivo are apparent at well below the maximum tolerated dose, offering a wide therapeutic window.

Numerous synthetic analogues of the combretastatins have been developed (Aleksandrzak et al. 1998; Hatanaka et al. 1998; Ohsumi et al. 1998; Pettit et al. 1998; Lawrence et al. 2001). Most information is available for the Aventis Pharma compound AVE8062 ((Z)-N-[2-methoxy-5-[2-(3,4,5-trimethoxyphenyl)vinyl]phenyl]-l-serinamide hydrochloride, formerly the Ajinomoto Co. Inc. compound, AC7700). This is a CA-4 derivative, the serine of which is cleaved by aminopeptidase, to form the active drug (Hori et al. 1999; Nihei et al. 1999a; Hori et al. 2001). The following discussion will be confined to the biological effects of CA-4/CA-4-P, CA-1/CA-1-P and AVE8062. Emphasis will be on CA-4-P, for which most information is available, the drug having entered Phase I clinical trial in the United Kingdom and United States in 1998 (Galbraith et al. 2001b; Rosen et al. 2001; Rustin et al. 2001).

Effects of CA-4-P on cells in vitro

CA-4-P targets microtubules of the cytoskeleton and interferes with their cyclic formation from the polymerization of tubulin dimers. Exposure to 0.1 µm−1 µm CA-4-P for 30 min causes complete depolymerization of microtubules in cultured human umbilical vein endothelial cells (HUVECs), as illustrated in Fig. 2 (a, b), with activity at concentrations as low as 1 nm (Kanthou & Tozer 2002). A severe destabilization of the tubulin cytoskeleton and mitotic spindle disrupts the cell's ability to successfully complete cell division, and a number of studies have demonstrated a dose- and time-dependent antiproliferative and cytotoxic effect of CA-4-P on both proliferating tumour and endothelial cells in culture (Böhle et al. 2000; Dark et al. 1997; Nabha et al. 2000; Boehle et al. 2001; Galbraith et al. 2001a). There is some evidence that endothelial cells are more sensitive than other cell types to these effects. Using the cellular uptake of neutral red, we showed that actively proliferating HUVECs were more sensitive to CA-4 and CA-4-P than a proliferating human breast cancer cell line (MDA-231) (Dark et al. 1997). Similarly, a greater antiproliferative effect of CA-4-P has been demonstrated on an immortalized murine endothelial cell line than on a murine fibroblast cell line (Böhle et al. 2000). We have shown that HUVECs are correspondingly more sensitive to CA-4-P-mediated tubulin depolymerization, as compared with either MDA-231 cells or vascular smooth muscle cells (Kanthou & Tozer 2000). Since both CA-4 and CA-4-P were used in our earlier study (Dark et al. 1997), endothelial cell sensitivity is presumably unrelated to their high alkaline phosphatase activity (Murray et al. 1989).

Figure 2.

The effect of CA-4-P on human umbilical vein endothelial cells (HUVECs) in culture. Treated cells (b and d) were analysed after a 30 minute exposure to 1.0 µm CA-4-P. Immunofluorescence of cells, using an antibody to β-tubulin, is shown in (a) and (b) and illustrates intact microtubules in untreated cells (a) compared with disruption in the treated cells (b). Actin staining of cells, using fluorescent phalloidin, is shown in (c) and (d) and illustrates the formation of stress fibers in the treated cells (d).

A 2-h exposure to 1 µm (2 µmol.h/L) CA-4-P was required to produce a significant cytotoxic/antiproliferative effect in rapidly dividing HUVECs (Dark et al. 1997), which is a high dose compared to that required for depolymerization of microtubules. 10 µm for 3 h (30 µmol.h/L) was required to produce a decline in proliferation of an immortalized endothelial cell line (Böhle et al. 2000). This compares with an area-under-the-curve (AUC) exposure of 5.8 µmol.h/L CA-4, following a single bolus dose of 25 mg/kg CA-4-P in the mouse (the lowest effective dose) (Rustin et al. 2001). This dose roughly equates to the estimated maximum tolerated dose of 68 mg/m² in humans, for which the AUC was 2.05 µmol.h/L CA-4, for five patients (Rustin et al. 2001). Bioavailability of CA-4-P/CA-4 to both tumour and endothelial cells is likely to be reduced in vivo due to considerable binding to plasma proteins (Dorr et al. 1996). These considerations suggest that direct cytotoxic/antiproliferative effects of CA-4-P, either on endothelial cells or tumour cells, are unlikely to be important following well-tolerated single doses of the drug in humans. However, they may contribute to drug effects observed following chronic or repeated dosing schedules. The very rapid decline in tumour blood flow, observed in animal models, following a single dose of CA-4-P (within the first few minutes) certainly cannot be explained by these delayed in vitro effects (see below).

Mechanisms leading to cytotoxicity have been investigated by a number of groups. Apoptotic effects have been reported in proliferating HUVECs, from around 6 h following continuous exposure to a high concentration (100 µm) of CA-4-P (Iyer et al. 1998). However, other investigators, using different cell lines, have reported that CA-4-P acts as a potent inhibitor of cell proliferation and leads to cell death via pathways different from apoptosis. Indeed Böhle et al. (Böhle et al. 2000) have demonstrated that, in transformed murine endothelial cells, 100 µm CA-4-P leads to a gradual build-up and arrest in the G2/M phase of the cell cycle and a corresponding decline in cell numbers in G1 and S phases, over a 48-h period. Cells subsequently died with no evidence of apoptosis. Similarly, in a panel of malignant human B-lymphoid cell lines, Nabha et al. (Nabha et al. 2000) reported that CA-4-P exerts a concentration-dependent growth inhibition, which leads to their arrest in G2/M. They described mitotic catastrophe, characterized by the formation of giant multinucleated cells, as being the predominant mechanism by which CA-4 induces cell death, rather than apoptosis. Cell death via mitotic catastrophe or mitotic failure probably occurs via inability of arrested cells to undergo protein synthesis, resulting in metabolic breakdown (Böhle et al. 2000).

Our studies using flow cytometry have shown that proliferating HUVECs, consistent with the other cell lines described above, become arrested in the G2/M phase of the cell cycle, after continuous exposure to a range of doses of CA-4-P (0.01–10 µm; unpublished data). In such cultures, a significant proportion of cells round and detach and exhibit chromosomal condensation, characteristic of mitotic cells. Such detached cells were further characterized as mitotic by flow cytometry. In addition and apparently in contrast to the transformed endothelial cells and lymphoid cell lines described above, they were found to express apoptotic markers of caspase 3 activation and PARP cleavage. This is consistent with the earlier study of HUVECs by Iyer et al. (Iyer et al. 1998) but at much lower CA-4-P concentrations. Apoptotic markers were absent from the remaining adherent cells, which were mainly in the G1/S phase of the cycle, indicating that CA-4-P initiates an apoptotic pathway as a result of failure to complete mitosis (unpublished data). Confluent cultures of HUVECs were found to be more resistant to CA-4-P-induced antiproliferative/cytotoxic effects than sparsely plated cells (Dark et al. 1997), consistent with such effects being strongly linked to proliferation status.

Summarizing the cytotoxic action of CA-4-P, the consensus is that CA-4-P leads to an arrest of cells of various types in the G2/M phase of the cell cycle, which is in accordance with effects of other microtubule damaging agents. However, the cell type, status of proliferation and amounts and duration of exposure to the compound probably determine the extent of cell damage and the pathway by which cell death is subsequently achieved, following the initial arrest. Similar factors may determine the sensitivity of cells and their capacity to recover from exposures to CA-4-P.

Morphological and functional effects of combretastatins on endothelial cells

The time-course of antiproliferative/cytotoxic effects of CA-4-P on proliferating endothelial cells cannot explain the rapidity of blood flow shut-down observed in animal tumours (Galbraith et al. 2001a; Tozer et al. 2001). However, endothelial cells in culture contract within minutes of exposure to CA-4-P, losing their characteristic cobblestone morphology (Galbraith et al. 2001a; Grosios et al. 1999). It is these effects which are likely to relate to the early vascular effects in vivo. In our studies, a more profound shape change was observed in proliferating than in confluent HUVECs (Galbraith et al. 2001a), although the reasons for this and the significance for the situation in vivo are unclear. Although the fraction of proliferating endothelial cells in tumours is higher than in normal tissues, it is still a small fraction of the total cell number; 0.15–9.8% have been reported for different tumour types (Eberhard et al. 2000). This suggests that proliferation status may not be the most important factor for selectivity of CA-4-P for the tumour vasculature (Tozer et al. 1999). Different degrees of HUVEC shape change in sparse vs. confluent cell cultures may be related to differences in cell-cell or cell-substratum contact rather than proliferation status per se and the changes observed in confluent cell cultures are certainly large enough to be significant in vivo (see below). Interestingly, the shape change induced by CA-4-P recovered to normal by 24 h, whereas that induced by colchicine did not (Galbraith et al. 2001a). This is consistent with different binding kinetics for the two agents (Lin et al. 1989) that, in turn, probably relate to their different toxicity profiles (Galbraith et al. 2001a).

One of the most important functions of the endothelium is to provide a selective barrier to blood solutes. Using confluent HUVEC monolayers grown on microporous cell culture inserts, both CA-1 (Watts et al. 1997) and CA-4-P (Kanthou & Tozer 2002) have been shown to increase monolayer permeability to fluorescent dextrans, the time-course of which was consistent with observed contraction of the actin cytoskeleton. Interestingly, in the first study, the effect could be increased by exposure of the cells to tumour-conditioned medium. The rapid time-course of these effects (within 20/30 min) and activity at low concentrations (0.1–1 µm for CA-4-P) indicate that an increase in vascular permeability, mediated via changes in the actin cytoskeleton, is likely to be an important factor in the expression of early vascular damage, following treatment with combretastatins in vivo.

CA-4-P-induced activation of signalling pathwaysin endothelial cells

Microtubules present a large surface in the cell and act as scaffolds for binding signalling molecules and relaying or repressing their activities (Gundersen & Cook 1999). Therefore, tubulin binding agents can interfere with multiple important cellular processes. Contraction of endothelial cells following exposure to CA-4-P, is indicative of rapid alterations of the actin cytoskeleton (van Hinsbergh 1997). Actin reorganization, which is crucial for determining cell morphology and function, is regulated by complex signal pathways that can be activated with rapid dynamics (Bishop & Hall 2000). Using simultaneous fluorescent phalloidin staining for actin and immunostaining for vinculin, we have shown the appearance of actin stress fibres (Fig. 2c, d) and focal adhesions, within minutes of exposure of HUVECs to CA-4-P (Kanthou & Tozer 2002). Myosin light chain (MLC) was rapidly phosphorylated in response to CA-4-P, in a time and dose-dependent manner. MLC phosphorylation activates actinomyosin interactions and myosin ATPase activity, leading to increased contractility and stress fibre assembly (Kohama et al. 1996). Increased MLC phosphorylation and actin stress fibre formation with CA-4-P were found to be mediated via activation of the GTPase Rho and its associated Rho kinase (Kanthou & Tozer 2002), which are known regulators of actin dynamics (Bishop & Hall 2000; Ridley & Hall 1992). The extent of MLC phosphorylation and stress fibre formation correlated with the degree of microtubule disruption by CA-4-P (Kanthou & Tozer 2002). This link between microtubule disassembly and Rho actvation is in accordance with previous reports (Liu et al. 1998). Rho and Rho kinase inhibitors were also found to block the CA-4-P-induced increase in permeability of HUVEC monolayers, demonstrating the importance of this signalling pathway in functional, as well as morphological changes brought about by CA-4-P (Kanthou & Tozer 2002).

Another prominent feature of CA-4-P-treated HUVECs is rapid blebbing, which occurs immediately after exposure to CA-4-P (Galbraith et al. 2001a) and is also dependent on actin reorganization (Kanthou & Tozer 2002). Here, F-actin accumulates around surface blebs, stress fibres misassemble into a spherical network surrounding the cytoplasm and focal adhesions appear malformed. This occurs predominantly in postconfluent nondividing cells exposed to CA-4-P, with an incidence of approximately 16% (Kanthou & Tozer 2002). Blebbing was inhibited by Rho/Rho-kinase inhibitors, indicating that actin reorganization by the Rho/Rho kinase pathway is essential for its formation. Preliminary evidence suggests that bleb formation is linked to early cytotoxicity, distinct from the mitotic block and subsequent apoptosis described above (Kanthou & Tozer 2002). However, the occurrence and/or relevance of surface blebbing in vivo are unknown. If it occurs, it could cause narrowing of the vascular lumen, an increase in vascular resistance, vascular obstruction and an increase in permeability, all of which could contribute towards an impairment of the normal function of the tumour microvasculature.

Effects of combretastatins in vivo

Classic evaluation of CA-4-P, in tubulin binding assays and in measurements of cytotoxic activity towards a panel of human tumour cell lines, failed to recognize its potential for antivascular effects in vivo. Cytotoxic actions against tumour cell lines in vitro were encouraging, with IC₉₀ values equivalent to or better than those for doxorubicin or 5-fluorouracil (El–Zayat et al. 1993). However, activity was greatly reduced, when serum concentrations > 10% were used and only moderate regrowth delays were recorded for treatment of B-16 melanoma in vivo (Dorr et al. 1996). These results suggested that nonprotein-bound CA-4 was the active form of the drug both in vitro and in vivo (Dorr et al. 1996). Subsequent studies on a range of tumours in vivo have confirmed the limited effect of CA-4-P as a single agent (Chaplin et al. 1999; Grosios et al. 1999; Grosios et al. 2000; Horsman et al. 2000).

Our initial in vivo evaluation of CA-4-P arose from a wider investigation of the vascular effects of a range of conventional anticancer agents (Hill et al. 1993; Hill et al. 1995; Chaplin et al. 1996) and revealed that it possessed significant antivascular activity in tumours at approximately one-tenth the maximum tolerated dose (Dark et al. 1997). 100 mg/kg CA-4-P reduced the perfused vascular volume of a murine tumour to less than 10%, by 6 h after treatment, and this was accompanied by massive necrosis of all but a narrow rim of tumour tissue by 24 h (Dark et al. 1997). It is this vascular effect which holds promise as a new adjuvant treatment to conventional cancer therapy. Indeed, no antitumour activity was observed in avascular tumour nodules (Grosios et al. 1999). These, and subsequent studies (Beauregard et al. 1998; Horsman et al. 1998; Li et al. 1998; Chaplin et al. 1999; Tozer et al. 1999), led to initiation of Phase I clinical trials of CA-4-P in late 1998 (Galbraith et al. 2001b; Rosen et al. 2001; Rustin et al. 2001). Similar effects have been reported for AVE8062 (AC7700) (Nihei et al. 1999b). The peripheral sparing, common to these agents, accounts for the rapid regrowth of tumours after treatment (Chaplin et al. 1999). A major challenge is to understand why the periphery is spared and to overcome this problem (see below).

Effects of combretastatins on tissueblood flow rates

The vascular activity of CA-4-P has been reported in ectopically and orthotopically transplanted murine tumours, xenografted human tumours, spontaneous tumours and vascularized metastases (Beauregard et al. 1998; Dark et al. 1997; Horsman et al. 1998; Maxwell et al. 1998; Chaplin et al. 1999; Grosios et al. 1999; Tozer et al. 1999; Pedley et al. 2001). Vascular effects have also been reported in nontumour pathologies, characterized by rapid angiogenesis (Griggs et al. 2001). At 24 h following treatment, with CA-4-P well below the maximum-tolerated dose, most murine tumours present with the vast majority of cells necrotic. There is little evidence for induction of apoptosis (Böhle et al. 2000 and our own unpublished observations). In the first few hours after treatment, histology of tumours is characterized by large, distended blood vessels, packed with red cells, with some showing signs of coagulation. Haemorrhage is evident in some tumours, primarily at the tumour periphery.

There is a dose and time-dependent decrease in blood flow rate to animal tumours following administration of CA-4-P (Chaplin et al. 1999; Tozer et al. 1999; Murata et al. 2001a). We investigated the selectivity of CA-4-P for tumour vs. normal tissue vasculature, in a rat tumour system, using the uptake of radiolabelled iodo-antipyrine (¹²⁵I-IAP or ¹⁴C-IAP) for measurement of absolute tissue blood flow rates. At a dose of 100 mg/kg, CA-4-P reduced blood flow rate to the P22 rat tumour by 100-fold (almost complete shut-down) at 6 h after treatment and no recovery was observed within 24 h (Tozer et al. 1999). At lower doses (30 and 10 mg/kg), similar reductions in tumour blood flow were obtained at early times after treatment but blood flow partially or completely recovered by 24 h (unpublished data). A similar recovery pattern was observed in a mouse mammary carcinoma model, using a radiotracer method for measuring tissue distribution of the cardiac output (Murata et al. 2001a). Pharmacokinetic data in rats and humans show that 10 mg/kg CA-4-P in the rat is roughly equivalent to the maximum tolerated dose in humans (Rustin et al. 2001 and unpublished data), suggesting that this pattern of decrease and subsequent recovery of tumour blood flow is the most likely scenario for single dose treatment with CA-4-P in the clinic. Of course, this will depend on many other factors, such as tumour type. Using dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), significant decreases in tumour uptake of a contrast agent have been detected following CA-4-P administration to cancer patients (Galbraith et al. 2001b), which tend to be greater at 4 than at 24 h after treatment. This is consistent with the animal data.

In rats, there was an increase in mean arterial blood pressure (MABP) at 1 h after CA-4-P treatment but this tended to normalize by 6 h, sometimes with a slight reduction thereafter (Tozer et al. 1999 and unpublished data). This appeared to be mainly due to an increase in flow resistance (decrease in blood flow rate in the face of an increased or unchanged perfusion pressure) in normal tissues, as well as tumour, with skeletal muscle, skin and spleen being most affected (Tozer et al. 1999). Hori et al. (Hori et al. 1999) also reported a decrease in blood flow rate to rat liver, bone marrow and brain with AVE8062, the effects in liver and brain reversing within 8 and 24 h, respectively. No effect was observed in kidney. Despite these effects, the blood flow reduction for tumours, following treatment with either CA-4-P or AVE8062, were larger and generally longer-lasting than in any normal tissues studied (Hori et al. 1999; Tozer et al. 1999; Murata et al. 2001a).

Using quantitative autoradiography to determine the spatial distribution of tumour blood flow rate using ¹⁴C-IAP, we showed that the decrease in blood flow induced by CA-4-P was less in the periphery than in the centre of P22 tumours (Tozer et al. 1999). This is consistent with the sparing of tumour cells in the tumour periphery, reported in a whole range of tumour types (Beauregard et al.1998;; Dark et al.1997,Li et al.;Chaplin et al. 1999). Recovery of flow, at 24 h following 10 mg/kg in the rat, was throughout the tumour centre as well as the periphery (unpublished data), suggesting that it was due to renewed flow in pre-existing vessels.

Mechanisms leading to blood flow shut-down

We have investigated microvascular effects of CA-4-P in the rat P22 tumour growing in a dorsal skin flap window chamber implanted into BD9 rats (Tozer et al. 2001). Tumour vascular effects were visualized and monitored on-line, using intravital microscopy. Tumour blood flow reduction was extremely rapid following CA-4-P treatment, with red cell velocity decreasing significantly by 5 min and dropping to less than 5% of the starting value by 1 h. There was a visible loss of a large proportion of the smallest blood vessels, with some return of visible vasculature at 1 h after treatment but the blood in these vessels was static or near so and many of the vessels were distended, confirming histological studies.

Blood flow effects, even in these very small tumours, were most profound in the tumour centre. Direct vasoconstrictive effects could not account for the full effect of the drug. The haematocrit within larger draining tumour venules tended to increase at early times after CA-4-P (approximately 20 min), suggesting fluid loss from the blood. Abnormal rheology, involving the stacking of red cells to form rouleaux (Lominadze & Mchedlishvili 1999), was very apparent as blood flow slowed, and this would have made a significant contribution to the vascular shut-down observed. Haemorrhage from peripheral vessels was also apparent. Some of these effects, in two different tumour lines, are illustrated in Fig. 3.

Figure 3.

The effect of CA-4-P on tumours growing in dorsal skin flap window chambers in rats and mice. The human colorectal tumour, HT29, xenografted into SCID mice, showed considerable haemorrhage around the periphery, 3 min after administration of 100 mg/kg CA-4-P (b compared to a), even though this is a relatively resistant tumour cell line in respect of necrosis induction (see text). Pairs of images of the P22 rat carcinosarcoma, at different magnifications and at different times after 30 mg/kg CA-4-P are shown in (c to h). The rectangles in the low power images(c, e and g, × 4 objective) represent the approximate areas covered by thehigh power images (d, f and h, × 20 objective). There is a loss of the visible vasculature, especially the smallest vessels, at early times after treatment(e and f vs. c and d). Some vessels return at later times (g and h) and these appear dilated and congested. Arrows in (h) indicate stacking of red cells.

Fluid loss from the blood and haemorrhage suggested that an early effect of CA-4-P was an increase in tumour vascular permeability to macromolecules, consistent with our in vitro results described above. Using a radiotracer technique, we have measured an increase in vascular permeability to bovine serum albumin of the P22 rat carcinosarcoma growing in the inguinal fat pad, within minutes of systemic administration of CA-4-P (Tozer et al. 2001). This increase is a likely trigger for vascular shut-down via development of tissue oedema, followed by an increase in blood viscosity and/or an increase in interstitial fluid pressure. Interestingly, the peripheral haemorrhage, which occurred in the window chamber tumours, suggests that the primary vascular damage induced by CA-4-P may occur in this area, with secondary vascular shut-down in the centre. A similar conclusion was reached for another vascular-damaging agent, TZT-1027 (Otani et al. 2000). If, in vivo, proliferating endothelial cells are more sensitive than quiescent ones, as suggested by the in vitro results (Galbraith et al. 2001a), this would be consistent with reports of more active angiogenesis at the tumour periphery (Vermeulen et al. 1997).

Evidence to date suggests that the rapid decrease in tumour blood flow following CA-4-P treatment in vivo is due to a combination of morphological and functional changes in endothelial cells, brought about by rapid cell signalling between the tubulin and actin cytoskeletons. Effects of the measured increase in tumour vascular permeability are probably enhanced by direct effects on tumour vascular resistance, from changes in endothelial cell shape, endothelial cell blebbing and even endothelial cell detachment. However, these characteristics of in vitro exposure of endothelial cells to CA-4-P have not been unequivocally demonstrated in the in vivo situation. In addition, there is some evidence for arteriolar constriction in tumours (Tozer et al. 2001), which may be a direct effect of the drug or an effect on the production of endothelial-derived vasoconstrictors/vasodilators. A summary of the proposed vascular events in tumours, leading to early vascular shut-down following exposure to combretastatins, is shown in Fig. 4.

Figure 4.

Potential mechanisms for initial vascular shut-down following combretastatin treatment. Endothelial cells, shown in purple lining the vascular wall in brown, are affected within minutes of drug exposure (in vitro results). Increased tumour vascular permeability to macromolecules within minutes of exposure to CA-4-P has been measured (see text). This protein leakage would lead to oedema and may cause an early increase in interstitial fluid pressure leading to vascular collapse. There is also likely to be an increase in blood viscosity. Rounding up and blebbing of endothelial cells would increase vascular resistance, exacerbated by active vasoconstriction (or interference with vasodilation) in supplying arterioles. Endothelial cells may even slough off. As blood flow slows, red cells (shown in red) will stack to form ‘rouleaux’, further increasing viscous resistance and slowing flow by a positive feedback mechanism. Exposure of the basement membrane will initiate later effects (see text).

Sensitivity of the tumour vasculature to combretastatins

Studies in animal models have clearly shown that combretastatins modify blood flow to various normal tissues but that blood flow to tumours is uniquely compromised (Hori et al. 1999; Tozer et al. 1999; Murata et al. 2001a). Initial results from clinical trials of CA-4-P are also promising (Galbraith et al. 2001b). However, the factors determining tissue sensitivity are unclear. In addition, the sensitivity of different tumour lines to CA-4-P is variable, in terms of both blood flow reduction and necrosis induction (Beauregard et al. 2001).

Differences between tumour and normal tissue endothelial cells may influence the primary damage inflicted on them. Differences in proliferation rate (Dark et al. 1997), post-translational modifications of tubulin (Galbraith et al. 2001a; Parkins et al. 2001), interaction between the tubulin and actin cytoskeletons and microenvironmental factors may all be important. On the other hand, or in addition to these factors, tumours may be uniquely sensitive to secondary effects of the drugs. It is possible that the ‘mature’ vasculature of normal tissues can sustain more endothelial cell injury than the ‘immature’ vasculature of tumours. Interestingly, blood vessels in the HT29 tumour, a relatively insensitive tumour (Beauregard et al. 2001), stain well for α-smooth muscle actin, indicating relatively mature vessel walls compared with several other more sensitive tumours (Fig. 5). Vascular permeability to macromolecules is another index of vascular maturity. For a panel of 5 different tumour types, tumour vascular damage following CA-4-P treatment positively correlated with pretreatment vascular permeability to albumin, as measured by a magnetic resonance imaging technique (Beauregard et al. 2001).

Figure 5.

Immunohistochemical staining for α-smooth muscle actin in tumour sections, showing positive brown staining in the HT29 human colorectal carcinoma (a) and very little staining in the MDA-231 human mammary carcinoma (b), both tumours xenografted into SCID mice.

Many tumour vessels have sluggish flow, which is variable over time, and these vessels are likely to be particularly susceptible to CA-4-P, because of the rheological behaviour of blood at low flow rates (Lominadze & Mchedlishvili 1999). An increase in vascular permeability in tumours, where the interstitial fluid pressure is already high (Boucher et al. 1990), could be catastrophic in tumours, whereas it may be less harmful in normal tissues. Pharmacokinetics also certainly influence the final damage inflicted on tumour vs. normal tissue. Exposure to drug is undoubtedly higher in tumours than it is in normal tissues because of self-trapping as blood flow decreases and potential differences in phosphatase activity (Tozer et al. 1999 and M Stratford, personal communication).

In summary, it is currently unclear as to whether the tumour vasculature is particularly sensitive to the initial damaging effects of combretastatins on endothelial cells or whether blood flow reduction in tumours is more profound than in normal tissues, for the same degree of initial damage. The characteristics of the tumour microcirculation would certainly indicate that damaging events at the endothelial cell level might have much more profound effects on blood flow maintenance in tumours than in normal tissues. These issues remain to be resolved.

The role of neutrophils and nitric oxide

Experiments using the P22 rat tumour or normal skeletal muscle perfused with a saline-based buffer demonstrated that an increase in tumour vascular resistance could be achieved selectively with CA-4-P, in the absence of blood cells (Dark et al. 1997; Tozer et al. 1999). This is consistent with the hypothesis that direct damage to endothelial cells and/or vasoconstriction in tumours is important in the mechanism of action of the drug. However, a much larger effect was observed under in vivo conditions, implicating an additional role for blood cells. In normal tissue inflammation, neutrophil adhesion to the endothelium, with subsequent generation of damaging oxidizing species via the action of neutrophilic myeloperoxidase (MPO), causes vascular damage. Tumour invasion by inflammatory leucocytes probably contributes to the vascular damage and tumour cell kill following CA-4-P treatment, similar to that described for photodynamic therapy (Korbelik & Cecic 1999). We have measured an increase in MPO activity in murine tumours, 18 h after CA-4-P treatment (Parkins et al. 2000), which is consistent with immunohistochemical staining for neutrophils in tumour sections (Fig. 6). Recent studies, using an in vitro system for investigating HUVECs under flow conditions, have demonstrated that neutrophil adhesion to, and migration through, an endothelial cell monolayer could be induced by CA-4-P treatment and inhibited by a monoclonal antibody to E-selectin (Brooks et al. submitted for publication). Although the time-course of these effects (several hours) cannot account for the initial vascular events induced by CA-4-P, up-regulation of adhesion molecules and subsequent neutrophil infiltration are likely to be important events, leading both to additional vascular damage and direct neutrophil-mediated tumour cytotoxicity. These factors will certainly contribute to the final outcome of combretastatin treatment.

Figure 6.

Immunohistochemical staining for neutrophils in sections from the mouse CaNT tumour, showing very little staining in untreated tumours (a) and positive brown staining 18 h following treatment with 50 mg/kg CA-4-P (b). There may be some nonspecific staining. Results are consistent with measurements of the neutrophil-specific enzyme, myeloperoxidse (MPO), in these tumours (see text).

We found that nitric oxide (NO) production by tumours offers some protection from the vascular damaging effects of CA-4-P (Parkins et al. 2000; Tozer et al. 1999). Similar results have been found for ischaemia/reperfusion injury (Parkins et al. 1995; Parkins et al. 1998) and photodynamic therapy (Korbelik et al. 2000). Systemic NOS inhibition enhanced vascular damage in the P22 tumour whilst having little effect on normal tissue perfusion and may therefore have therapeutic potential (Tozer et al. 1999). MPO activity, in murine tumours following CA-4-P treatment, was enhanced by coadministration of the NOS inhibitor L-NNA (Parkins et al. 2000). Thus, part of the protective role of NO against CA-4-P-induced vascular injury may be mediated via an inhibitory effect on neutrophils. Alternatively, if the damage induced by CA-4-P involves the production of reactive oxygen species, the presence of NO may re-direct these reactions along less damaging pathways. The decrease in tumour blood flow by NOS inhibitors alone (Tozer et al. 1997) may also make the tumour vasculature uniquely susceptible to CA-4-P. Other pathways, which are currently under investigation, include the interaction of NO with tubulin (Parkins et al. 2001) and a putative role for NO in antagonizing CA-4-P-induced contraction of the actin cytoskeleton.

The involvement of neutrophils and, potentially, other immune-effector cells (Korbelik & Cecic 1999), suggests that the ultimate tumour cell cytotoxicity induced by CA-4-P and similar compounds may not directly correlate to the extent and time-course of blood flow shut-down. Indeed, we have found, in the CaNT mouse mammary tumour, that the enhancement of CA-4-P-induced tumour cell cytotoxicity by addition of L-NNA was more closely related to the extent of enhancement of neutrophil infiltration into the tumours than to the extent of enhancement of tumour blood flow reduction (unpublished data). This illustrates that ‘starving a tumour to death’ by shutting off its blood supply is too simplistic a concept to describe the complex series of events, which lead from binding of CA-4-P to the tubulin cytoskeleton and tumour cell death.

Therapeutic potential

Pre-clinical data indicate that CA-4-P has limited potential as a single agent (see above). However, it is possible to achieve a significant tumour growth delay with CA-4-P alone, by giving repeated low doses of drug. Indeed, splitting the dose into two equal fractions, separated by 1–6 h, increased tumour cell killing by at least a factor of two, in the CaNT mouse tumour (unpublished data). Repeated daily or twice daily doses of CA-4-P given over 2 or 3 weeks had a significantly greater antitumour effect than could be achieved with any single dose of drug. The reason for this increased effect is not clear. Repeated dosing may enhance self-trapping of the drug. If concentration of the drug is high enough, a direct cytostatic/cytotoxic effect on tumour endothelial cells and/or tumour cells may come into play.

Repeated administration of AVE8062, at doses typically half the MTD, produced significant growth retardation in a range of mouse and rat tumours, even producing some cures in the murine colon 38 (c38) tumour (Nihei et al. 1999a). The drug was active against orthotopically transplanted colon tumours (Nihei et al. 1999a) and chemically induced tumours (Hori et al. 2001). Similar antitumour activity was reported for tumours growing in immunocompetent hosts as for those growing in nude mice, demonstrating that lymphocytes are not required for drug activity (Nihei et al. 1999a). This characteristic discriminates the tubulin binding agents from other vascular-damaging agents, such as flavone acetic acid and its derivative, DMXAA, where lymphocytes play a larger part in drug activity (Bibby et al. 1991; Ching et al. 1992).

Newer agents such as AVE8062 may prove more effective than CA-4-P. Nihei et al. (Nihei et al. 1999b) compared AVE8062 with CA-4-P in a murine colon carcinoma model and found that AVE8062 produced a more profound suppression of tumour growth than CA-4-P, at approximately half the MTD for both drugs. Interestingly, the effective dose for AVE8062 was approximately 100-fold lower than for CA-4-P. Despite the fact that initial in vitro testing showed CA-4 to be the most promising of a whole range of natural and synthetic combretastatins, including CA-1 (Lin et al. 1988; McGown & Fox 1990), recent data suggests that CA-1-P may be more effective than CA-4-P in vivo (Holwell & Bibby 2001). We found that five daily doses of 5 mg/kg CA-1-P produced an equivalent growth retardation of the murine CaNT tumour to that produced by daily doses of 50 mg/kg CA-4-P (our unpublished data). Whilst the toxicity of CA-1-P is slightly increased relative to CA-4-P (MTD dose reduced by less than a factor of 2), this represents a significant improvement in the therapeutic window. The reason for this is not clear but may be linked to different rates of phosphate cleavage for the diphosphate, CA-1-P, compared to the monophosphate, CA-4-P. This requires further investigation.

The extensive ischaemic insult to tumours following CA-4-P treatment raises the possibility of hypoxia-induced angiogenesis. Indeed, there is some evidence for an increase in expression of both VEGF and basic fibroblast growth factor (bFGF) proteins in xenografted tumours following CA-4-P, although this was not associated with any change in microvessel density (Boehle et al. 2001). We have not detected any increased growth rate of re-growing tumours following CA-4-P treatment, compared with untreated controls and BUdR labelling of CaNT tumour cells in vivo showed no evidence of any change in proliferation rate following CA4P treatment (our unpublished data).

Most of the potential for CA-4-P resides in its combination with conventional therapies or with other emerging treatments based on high molecular weight entities. It is clear that CA-4-P can effectively ablate the tumour centre, resulting in up to 2 logs of cell kill. However, at least 2–3 logs of cell kill are required to induce a measurable growth response (and much more for a cure). The viable tumour rim remaining after CA-4-P treatment was seen as a potential target for conventional treatment approaches. This tumour region is often the best vascularized and oxygenated, factors working in favour of successful treatment by conventional chemotherapeutic drugs and radiation.

Several studies have demonstrated an enhancement of radiation damage in animal tumours by CA-4-P (Chaplin et al. 1999;Li et al. 1998; Horsman et al. 2000; Murata et al. 2001b). In a rat rhabdomyosarcoma model, this was only true for large tumours (7.1–14 cm³), which are much more sensitive to CA-4-P than smaller tumours of the same type (Landuyt et al. 2001). Although not the case for all tumours, most success was achieved with CA-4-P given after radiation treatment, consistent with the possibility that CA-4-P given first would increase tumour hypoxia and reduce the efficacy of the subsequent radiation treatment. There was evidence in one study that CA-4-P had a major impact on the radiation-resistant hypoxic cell population (Li et al. 1998). We have demonstrated efficacy of this combination in a fractionated regime. Weekly doses of CA-4-P added to a two-week radiation + carbogen schedule (8 doses of 5Gy) converted a noncurative treatment into one where 50% of the animals were cured (Chaplin et al. 1999). Hyperthermia is also enhanced by the addition of CA-4-P (Eikesdal et al. 2000; Horsman et al. 2000).

CA-4-P has also been tested in combination with conventional chemotherapeutic agents. The most effective regimes that we have tested, against a range of tumour types, involved cisplatin or Taxol (Chaplin et al. 1999), although a benefit was also seen for a variety of other agents, in at least one tumour type. The effects of some of these on the CaNT murine tumour are shown in Fig. 7. Others have also shown some benefit of CA-4-P in combination with doxorubicin (Nelkin & Ball 2001) and 5-fluorouracil (Grosios et al. 2000). Scheduling of these combinations is a complicated issue, especially because of the blood-flow modifying effects of CA-4-P. ‘Trapping’ of drugs in the tumour tissue, as blood flow is compromised, should be possible with judicious scheduling, as was noted above for CA-4-P itself. However, even in the absence of this effect, it is hoped that additive or synergistic effects may occur. This appeared to be the case for the combination with 5-fluorouracil, where CA-4-P in certain scheduling regimes, could enhance the delay in tumour growth induced by 5-FU alone, with no corresponding increase in the levels of 5-FU in tumour tissue (Grosios et al. 2000).

Figure 7.

Survival fraction per g tumour was measured in the CaNT mouse tumour by clonogenic assay, 18–24 h following treatment in vivo with a range of conventional chemotherapeutic agents, CA-4-P alone or the combination ofCA-4-P given 15 min following the other drugs. CA-4-P was administered at 50 mg/kg. The other drugs were administered at roughly half their maximum tolerated single dose, except for Taxol, which was administered at close to its maximum tolerated dose.

High molecular weight anticancer agents, with potent binding characteristics, are often found to localize in the tumour periphery, with very little penetration into the tumour centre. This limits their efficacy, as is the case for the anticarcinoembryonic antigen antibody, used for radio-immunotherapy (RIT) and antibody-directed-enzyme-prodrug-therapy (ADEPT) of colorectal tumours (Pedley et al. 1996, 1999). The combination of these approaches with combretastatins or other vascular targeting drugs, which preferentially cause tumour cytotoxicity in the tumour centre, was therefore attractive. Indeed, conversion of RIT into a curative treatment by the addition of CA-4-P has recently been demonstrated in a human xenografted colorectal tumour model (Pedley et al. 2001).

The addition of either tirapazamine or AQ4N, bioreductive agents which are inactive until metabolized under hypoxic conditions, significantly enhanced the growth delay induced in the CaNT tumour by CA-4-P alone, although they had no measurable effect in their own right (our unpublished data). These data suggest that, rather than dying as a result of ischaemic damage, an important subpopulation of cells become transiently hypoxic following the blood flow reduction induced by CA-4-P and eventually contribute to regrowth of the tumour. Thus, oxygen modifying treatments or hypoxia-selective cytotoxins may have a role in the development of antivascular therapies.

Finally, it may be possible to take advantage of the gross tumour necrosis and severe hypoxia induced by CA-4-P and similar agents, for therapeutic targeting of tumours using anaerobic bacteria. In order to test this, the apathogenic Clostridium acetobutylicum was genetically engineered to express E. coli cytosine deaminase (Theys et al. 2001). The levels of cytosine deaminase in a solid tumour, following administration of the Clostridium was significantly higher in the CA-4-P-treated animals than in controls (Theys et al. 2001).

Conclusion

New approaches for the treatment of solid tumours are required. Targeting of the tumour vasculature with combretastatins is one such approach. The lead compound, CA-4-P, is already being tested in clinical trials and initial results are promising, with reductions in tumour blood flow reported in patients with advanced disease. Pre-clinical studies suggest that more effective compounds than CA-4-P are now available. In order to progress with this approach, the priorities for biologists are to fully understand the mechanisms of action of these agents and the factors dictating susceptibility of different tumour types, tumour vs. normal tissue and tumour periphery vs. tumour centre.