Tumour vasculature--a potential therapeutic target.

The tumour vasculature is vital for the establishment, growth and metastasis of solid tumours. Its physiological properties limit the effectiveness of conventional anti-cancer strategies. Therapeutic approaches directed at the tumour vasculature are reviewed, suggesting the potential of anti-angiogenesis and the targeting of vascular proliferation antigens as cancer treatments.

Chemotherapy, radiotherapy and biological therapy (the use of cellular or humoral components of the immune system in cancer treatment) have all been shown to have a direct cytotoxic effect on malignant cells. However, a solid tumour is composed not only of a parenchymal compartment containing malignant cells, but also of a supportive stromal compartment containing vascular elements derived from surrounding tissues. It has been suggested that anti-cancer strategies directed against the stromal component of tumours could effect an indirect tumour cell kill. This paper emphasises the potential of the tumour vasculature as a target in solid tumour therapy.
The vasculature of tumours has been shown to possess distinct anatomical, physiological and, recently, immunological characteristics that distinguish it from that of normal tissues. Many solid tumours are inadequately perfused, resulting in conditions of hypoxia and acidosis which, paradoxically, protect malignant cells from the standard treatment modalities of chemotherapy and radiotherapy (Thomlinson and Gray, 1965). The delivery of drugs and immunoconjugates by the tumour vasculature is inefficient, undermining therapeutic strategies which rely upon vascular access to the parenchymal compartment of tumours (Dvorak et al., 1991). Given the constraints upon therapy normally imposed by the tumour vasculature, treatment options which seek to exploit its properties have evoked considerable interest.
The process by which an avascular aggregate of tumour cells establishes a blood supply derived from the host stroma is known as tumour angiogenesis. The acquisition of new vascular elements by an established tumour is also dependent on this process. Angiogenesis is a complex process which is tightly regulated under normal physiological conditions with multiple levels of control. It is conceivable that antiangiogenesis strategies could provide the clinician with novel alternatives for cancer therapy (Folkman, 1972).
The established (as opposed to the developing) vasculature of solid tumours has also been singled out as a potential therapeutic target, largely as a result of research into the proliferation kinetics of tumour endothelium. The realisation that tumour endothelium is highly proliferative relative to normal endothelium, suggests this as a discriminating feature which might allow for anti-proliferating endothelial therapy (APET) (Denekamp, 1982). The anti-endothelial approach has been broadened to encompass any treatment acting primarily at the level of the tumour vasculature, under the heading 'anti-vascular therapy'.

Tumour angiogenesis
The term angiogenesis was first used to describe the formation of new blood vessels in the placenta (Hertig, 1935 Angiogenesis is also a feature of the developing embryo, healing wounds, psoriasis, arthritis and diabetic retinopathy (Folkman, 1985). The most extensive study of angiogenesis, however, has been in solid tumours. The first microscopic observations of the vascularisation of tumour implants were made in 1939 using the Sandison-Clarke rabbit ear chamber (Ide et al., 1939). The most striking feature of these early studies was the ability of the tumour to elicit a rapid and continuous ingrowth of new capillary endothelium from the host tissue (Algire et al., 1945), a process which has become known as tumour angiogenesis.
A possible mechanism of angiogenesis, originally proposed by Folkman, in solid tumours begins with the retraction of pericytes and the proteolytic degradation of the basement membrane from host post-capillary venules adjacent to the tumour. Endothelial cells begin to proliferate and migrate in the direction of the tumour, resulting in three distinct zones of angiogenesis: the migratory zone, a proliferative zone and a zone of maturation, where functional vessels can be identified. The endothelial cells become organised into tubular structures (capillary loops) and form anastomoses between themselves and elements of the host vasculature in the establishment of a primitive tumour circulation (Folkman, 1984). The formation of a basement membrane and the incorporation of pericytes into the vascular structures, which are features of vascular maturation, are commonly deficient in tumour angiogenesis (Bicknell and Harris, 1992). Recently, another model of tumour angiogenesis has been described in which capillary loops are derived principally from precapillary arterioles (rather than venules), and the tumour vasculature expands by remodelling the established host vasculature based on bifurcation of, and anastomosis with, existing host vessels (Hori et al., 1990). This system has a greater emphasis on a dynamic process involving the vascular unit as a whole. It differs significantly from Folkman's description of the angiogenic process, which is highly mechanistic and focuses on the endothelial cell component of the tumour vasculature. Several possible control points in tumour angiogenesis are identified by virtue of this emphasis on the role of endothelial cells, which demonstrate capabilities of migration, proliferation and differentiation at different phases of the angiogenic process. The investigation of each of these features has broadened the scope for intervention in tumour angiogenesis.
The importance of angiogenesis to the establishment, growth and metastasis of solid tumours may be inferred from a variety of observations. Tumours grown in the rabbit corneal micropocket have demonstrated two phases of growth: a prevascular phase characterised by slow expansion of the implant as a thin plate and a vascular phase characterised by the formation of a rapidly growing exophytic mass (Gimbrone et al., 1974). Similarly, homologous tumour implants grown in the anterior chamber of the rabbit eye formed dormant spheroids which, when placed on the iris, where vascularisation could occur, grew rapidly and became locally invasive (Gimbrone et al., 1972). Studies of the vascularisation of hepatic metastases, by making silicone rubber casts of hepatic vasculature, demonstrated that tumour metastases were avascular up to 1 mm in diameter and were consistently vascularised beyond this (Lien and Ackerman, 1970), suggesting that the acquisition of a vascular supply from the host is essential for the local establishment of a tumour. Furthermore, tumours grown on the chick chorioallantoic membrane (CAM) at different times have resulted in growth rates directly proportional to the [3H]thymidine labelling index of the vascular endothelial cells, suggesting that the tumour growth rate may be related to the intensity of the host neovascular response (Knighton et al., 1977). The relationship between angiogenesis and the development of the malignant phenotype was examined in an experiment in which diploid mouse fibroblasts were passaged subcutaneously and separately assessed for their angiogenic potential using the corneal micropocket assay system. It was found that the fibroblasts could initiate angiogenesis by passage 5, but became malignant at passage 15, indicating that the capability of the cells to evoke an angiogenic response preceded development of the frankly malignant phenotype (Ziche and Gullino, 1982). It has been observed that some in situ human breast carcinomas exhibit a prominent angiogenic stromal response, suggesting that, before becoming agents of local invasion and metastasis, malignant cells require capabilities over and above the ability to initiate an angiogenic response (Weidner et al., 1991). Increased angiogenic activity has been suggested as a marker of neoplastic and in situ bladder carcinoma (Chodak et al., 1980), and the intensity of the angiogenic response evoked by tumours has been positively correlated with the probability of metastasis for cancer of the breast (Weidner et al., 1991;Bosari et al., 1992;Horak et al., 1992;Weidner et al., 1992), melanoma (Srivastava et al., 1988), non-small-cell lung cancer (Macchiarini et al., 1992), prostatic cancer (Fregene et al., 1993) and squamous cell carcinoma of the head and neck (Albo et al., 1994). It has been suggested that microvessel density (MVD) may represent a new prognostic indicator in solid tumours (Weidner, 1993).
Most of the evidence outlined above has provided indirect support for the hypothesis that tumour establishment, growth and metastasis are angiogenesis dependent. The impact of molecular biological approaches on the field of tumour angiogenesis is beginning to provide direct supportive evidence. It seems likely that inhibitory genetic mechanisms normally keep angiogenesis in check. Two candidate genes have been identified, one coding for a glycoprotein with anti-angiogenic properties (Rastinejad et al., 1989), subsequently identified as thrombospondin (Good et al., 1990), and the other, nm23, coding for a protein positively correlated with low tumour metastatic potential (Rosengard et al., 1989), which possibly operates by interfering with the signal transduction pathway of the angiogenic peptide transforming growth factor P (TGF-P) (Leone et al., 1991).
The concept that tumours are angiogenesis dependent has been summarised by Folkman (1972) in the following statement. 'Once tumour take has occurred, every increase in tumour cell population must be preceded by an increase in the new capillaries that converge upon that tumour'. It is this assumption that underpins the case for anti-angiogenesis strategies in cancer therapy.

Tumour angiogenesis factors
A landmark paper in the study of tumour angiogenesis described the isolation of a tumour angiogenesis factor (TAF) from rat Walker 256 carcinoma cells (Folkman et al., 1971).
A variety of human angiogenic peptides have subsequently been identified and have had their structures determined by protein sequencing and cDNA cloning ( Table I).
Comparison of the actions of these angiogenic proteins has demonstrated that several are mitogenic to endothelial cells in vitro. The mitogenic angiogenic peptides have trophic effects on diverse tissues with the exception of vascular endothelial growth factor (VEGF), which is specific for endothelial cells (Leung et al., 1989). It has been suggested that VEGF functions as a hypoxia-induced angiogenic factor. The production of VEGF in human glioblastoma multiforme has been specifically localised to tumour cells which are juxtaposed to regions of necrosis, by in situ hybridisation, using radiolabelled antisense riboprobes with specificity for VEGF mRNA. The same workers have confirmed that VEGF mRNA can be induced by hypoxia in vitro using cultured rat glioma cells, skeletal muscle myoblasts and mouse fibroblasts (Shweiki et al., 1992). Additional support for the role of VEGF in tumour angiogenesis has been provided by a similar study which confirmed VEGF production by palisading tumour cells in anaplastic gliomas, and demon-   t, Positive; +, inhibitory; +, no action; t4,, both mitogenic and inhibitory action reported; ?, not known; *, lacks signal peptide for secretion. strated strong expression of the high-affinity VEGF receptor, flt, on tumour endothelial cells but not on endothelial cells in normal brain tissue (Plate et al., 1992). VEGF is probably identical to vascular permeability factor (VPF) (Keck et al., 1989), which, by virtue of its ability to increase vessel permeability, causes the extravasation of plasma proteins, thus providing a suitable microenvironment for angiogenesis (Kondo et al., 1993). The most studied angiogenic peptide, basic fibroblast growth factor (bFGF), and also platelet-derived endothelial cell growth factor (PD-ECGF) lack secretory signal peptides necessary for extracellular secretion, casting doubt on their significance in tumour angiogenesis. However, it has been suggested that bFGF, which is associated with the extracellular matrix and basement membrane, may be released by enzymic action, thus permitting its role as a mediator of angiogenesis (Klagsbrun and D'Amore, 1991). PD-ECGF has been clearly shown to stimulate angiogenesis (Ishikawa et al., 1989), and, like VEGF, was originally thought to be a specific endothelial cell mitogen. However, the mechanism of action of PD-ECGF has recently been called into question when it was demonstrated that cDNA coding for a 120 amino acid sequence of human thymidine phosphorylase (TP) was identical to the sequence of PD-ECGF (Furukawa et al., 1992). Recombinant human PD-ECGF has been shown to have thymidine phosphorylase activity (Usuki et al., 1992), casting doubt on the validity of in vitro assays of endothelial mitogenicity based on [3H]thymidine uptake. rPD-ECGF was not shown to have any action on endothelial proliferation in vitro using a direct cell counting technique (Moghaddam and . However, a proliferation assay based on acid phosphatase production has confirmed endothelial mitogenicity (Finnis et al., 1993), possibly resulting from an indirect action mediated by local levels of thymidine (Finnis et al., 1993). The mechanism by which PD-ECGF/TP stimulates angiogenesis remains unclear, but it is not a classical growth factor since direct contact with a cell-surface receptor is not required for its mitogenic capability (Finnis et al., 1993).
The related mitogenic growth factors, transforming growth factor x (TGF-x) and epidermal growth factor (EGF) may be important mediators of tumour angiogenesis. Immunohistochemical staining using an antibody to TGF-a has shown perivascular staining in experimental neoplasms, and TGF-x mRNA has been demonstrated in several solid tumours, suggesting a possible role in tumour angiogenesis (Schreiber et al., 1986). The EGF receptors belong to a group of proto-oncogenes, including c-erbB-2, which are overexpressed in a number of human tumours. The production of TGF-a/ EGF, coupled with high-level EGF receptor expression by tumour cells, confers a selective growth advantage on tumour cells, which may be supported by the mitogenic action of both TGF-x and EGF on endothelial cells, and the promotion of an angiogenic response (Yates et al., 1991).
The mechanism of action of other angiogenic factors is less clear. Both angiogenin (Fett et al., 1985;Kurachi et al., 1985) and platelet-derived growth factor (PDGF) (Bicknell and Harris, 1992) have been shown to have no mitogenic activity for cultured endothelial cells, while TGF-P (Roberts et al., 1986;Frater-Schr6der et al., 1987), tumour necrosis factor-x (TNF-o) (Leibovich et al., 1987) and interleukin 6 (Motro et al., 1990) have been demonstrated to be inhibitory to endothelial cells in vitro. The explanation for these paradoxical observations has not been fully elucidated, although PDGF has been shown to promote endothelial migration . The other peptides presumably have an indirect mechanism of action, such as the stimulation of other cells (e.g. macrophages) to release angiogenic factors. Alternatively, they might promote angiogenesis by causing endothelial differentiation (e.g. tube formation or matrix production), rather than endothelial proliferation (Klagsbrun and D'Amore, 1991).
The relative importance of these and other angiogenic peptides, such as angiotensin II (Fernandez et al., 1985) and substance P (Ziche et al., 1990), as mediators of tumour Tumour vasculature -a potential therapeutic agent CT Baillie et al angiogenesis is uncertain. The majority of the angiogenic peptides have been demonstrated to be present in at least some tumours, and others may be released by white cells and macrophages associated with the stromal response provoked by malignant cells.
Low molecular weight angiogenic factors, defined as less than 2000 daltons in a recent review (Odedra and Weiss, 1991), although less well characterised than their peptide counterparts, may play a significant role in tumour angiogenesis. Both endothelial cell stimulating angiogenesis factor (ESAF) and endothelial stimulating factor (ESF) have been isolated from tumours, although their precise structure remains unknown (McAuslan and Hoffman, 1979;Weiss et al., 1979). ESAF has been shown to dissociate neutral matrix metalloproteinases from their specific inhibitor, tissue inhibitor of metalloproteinase (TIMP), thus permitting the matrix proteolysis which is an essential feature of angiogenesis. The original TAF was probably composed of ESAF/ESF in combination with bFGF, and subsequent observations have confirmed that tissues containing bFGF have high levels of ESAF. The demonstration that bFGF stimulates microvascular endothelial cells to synthesise procollagenase suggests a synergistic relationship between ESAF and bFGF, in which collagenase production under the influence of bFGF is supported by the action of ESAF in keeping the collagenase in its active form (Odedra and Weiss, 1991).
Many of the reported low molecular weight angiogenic factors are either metabolic co-factors or products of anaerobic metabolism. Whether any of these factors have any particular relevance to tumour angiogenesis is questionable.
The prostaglandins El and E2 (BenEzra, 1978;Form and Auerbach, 1983) and nicotinamide derivatives (Kull et al., 1987;Morris et al., 1989) have been implicated in tumour angiogenesis by virtue of their isolation from tumour extracts. Prostacyclin analogues (Ohtsu et al., 1988) and longchain fatty acids such as erucamide (Wakamatsu et al., 1990), although not confirmed as tumour-secreted angiogenic agents, may also have a role in tumour angiogenesis.

Anti-angiogenesis strategies
The original discovery of a tumour-derived diffusible mediator of angiogenesis (TAF) (Folkman et al., 1971) led to Folkman's (1972) suggestion of an anti-angiogenesis approach for cancer therapy. The ability of anti-TAF antisera to block in vivo angiogenesis caused by TAF emphasised the potential value of this approach .
The development of bioassays for most of the component steps of angiogenesis has enabled the precise mechanism of action of some angiostatic compounds to be determined (Table II). Tumour angiogenesis could be subject to therapeutic intervention at several key points, which are illustrated in Figure 1. The strategic links identified in the chain of events that bring about tumour angiogenesis include the inhibition of tumour cell release of angiogenic factors, the antibody-mediated blocking of angiogenic factors or their receptors, the inhibition of microvascular endothelial proliferation or migration, the disruption of endothelial differentiation into organised capillary tubes and the prevention of anastomosis formation between newly formed vascular elements and the host vasculature (Bicknell and Harris, 1992).
Reports of the inhibition of angiogenic factor production by tumours are scanty, but partial (50%) inhibition of unidentified angiogenic factors produced by a hepatoma and a bladder carcinoma have been attributed to the interferons-x and -P (Sidky and Borden, 1987). [Interferon-o has been used successfully in the treatment of haemangioendotheliomas, prompting the suggestion of a possible anti-angiogenic mechanism of action (Orchard et al., 1989).] The most studied aspect of the anti-angiogenesis approach has been the attempted blockade of angiogenic factors by neutralising antibodies. In one such study human colonic adenocarcinoma xenografts were grown in nude mice. The resulting tumours showed both an increased rate of growth and increased vascularity in the presence of systemically administered bFGF (Gross et al., 1990). It was demonstrated that bFGF was not mitogenic to the tumour cells in culture. Neutralising monoclonal antibodies to bFGF were able to suppress tumour growth in vivo, and autoradiography of tumour sections demonstrated that the receptors for bFGF were located on the vascular endothelium. Several workers have reported varying degrees of vascular-mediated solid tumour control in experimental animals using antibodies directed against bFGF (Hori et al., 1991;Reilly et al., 1989).
However, the fact that others have noted that anti-bFGF monoclonal antibody-secreting hybridomas readily form vascularised solid tumours in athymic Inice (Matsuzaki et al., 1989) serves to demonstrate the limitations of this strategy.
The problem with strategies directed at tumour  basement membrane angiogenesis factors, whether the inhibition of their synthesis by tumour cells or their neutralisation once secreted, is that several angiogenic factors may be produced by a given tumour, so that the inhibition of a single factor is unlikely to prevent tumour establishment and growth. Given the lack of uniformity in the type and number of angiogenic factors secreted by solid tumours, the chance of generalised success with these approaches is dependent on the identification of an angiogenic factor which is common to most tumours. It has been suggested, for this reason, that VEGF would be a better candidate than bFGF for this type of strategy (Bicknell and Harris, 1992). Recently, tumour inhibition has been demonstrated using antibodies directed against VEGF (Kim et al., 1993;Kondo et al., 1993), and signal transduction from the Flk-I receptor for VEGF has been blocked using a retrovirus encoding a dominant-negative Flk-I mutant in a murine glioblastoma model, with resultant inhibition of tumour growth (Millauer et al., 1994). In a variation on antibody-mediated blockade of angiogenic factors, the angiogenic activity of angiogenin has been shown to be neutralised by the highly specific binding of placental ribonuclease inhibitor (Shapiro and Vallee, 1987). An alternative approach involves the blockade of angiogenic factors at receptor level, as is illustrated by the ability of suramin to inhibit binding of bFGF (Bicknell and Harris, 1991;Danesi et al., 1993), and the reduction in binding of aFGF at the endothelial cell surface caused by interferon-y (Klagsbrun and D'Amore, 1991).
The majority of tumour angiogenesis factors operate by causing either endothelial cell proliferation or migration (or by a combination of both mechanisms). Clearly, the ability to disrupt the process of endothelial proliferation and migration would be important features of any putative anti-angiogenic compound. Platelet factor-4 (PF-4) (Maione et al., 1990), fumagillin (Ingber et al., 1990) and a 16 kDa fragment of human prolactin (PRL) (Clapp et al., 1993) have been shown to inhibit growth factor-stimulated endothelial proliferation in vivo. Thrombospondin (Good et al., 1990), protamine (Taylor and Folkman, 1982) and the sulphated chitin derivative, SCM-chitin III (Murata et al., 1991), have all been demonstrated to inhibit endothelial migration in vitro, while linomide has been shown to be both cytostatic and inhibitory to endothelial chemotaxis, suggesting that its antiangiogenic properties may account for its in vivo anti-tumour effects in both rats and mice (Vukanovic et al., 1993).
The migration/proliferation phase of angiogenesis is associated with increased synthesis of proteolytic enzymes by endothelial cells. A variety of anti-angiogenic agents operate by inhibiting this process, thus preventing invasion of the endothelial basement membrane and migration through the extracellular matrix by endothelial cells. An anti-angiogenic, 28.5 kDa glycoprotein, tissue inhibitor of metalloproteases type 1 (TIMP-1), which complexes activated interstitial collagenase with 1:1 stoichiometry, has been isolated from fibroblasts (Carmichael et al., 1986) and cartilage (Moses et al., 1990). The resulting TIMP-l -collagenase complex has no proteolytic activity. A similar, naturally occurring 21 kDa metalloproteinase inhibitor, TIMP-2, with anti-angiogenic properties, has been isolated from human melanoma cells . TIMP-2 complexes 1:1 with type IV *procollagenase. These inhibitory actions of TIMP-1 and -2 illustrate the importance of matrix proteolysis to the angiogenic process.
A similar mechanism of action has been postulated to explain the properties of the angiostatic steroids. In developing the chick CAM as an angiogenic assay, it was discovered that the combination of heparin with steroids was inhibitory to angiogenesis. The most potent steroid was shown to be tetrahydrocortisol, which was previously thought to be without biological activity, and was thus considered to define a new class of angiostatic steroids (Folkman and Ingber, 1987). Non-anticoagulant heparin fragments were more effective than intact heparin, and the synthetic heparinoid, 1Bcyclodextrin tetradecasulphate, provided the most potent angiostatic steroid/heparin combination ). It has recently been shown that angiostatic steroids are able to increase the synthesis of plasminogen activator inhibitor by endothelial cells. The resulting reduced levels of fibrinolytic proteases might explain the inhibitory action of the steroid-heparin combination to the angiogenic process (Blei et al., 1993). Other anti-angiogenic agents which may operate by the inhibition of matrix proteolysis include PF-4 and the synthetic tetracycline minocycline (Tamargo et al., 1991).
The mechanisms that bring about the organisation of endothelial cells into tubular structures are beginning to be understood at the molecular level, leading to the identification of a new potential control point for angiogenesis (Ingber, 1991). It has become apparent that insoluble extracellular matrix (ECM) components promote capillary tube formation by mechanochemical interactions with endothelial cells (Ingber and . The attachment of endothelial cells to the ECM and the cytoskeletal events that result in lumen formation are both believed to be mediated by laminin. A site on the laminin A-chain named PA 21, containing the Arg-Gly-Asp (RGD) sequence, has been demonstrated to mediate the initial endothelial cell attachment to laminin. A separate B1-chain domain containing the Tyr-Ile-Gly-Ser-Arg (YIGSR) sequence is of importance in cell-cell interaction and tube formation (Grant et al., 1989). If these mechanisms could be inhibited, the establishment of a primitive tumour circulation could effectively be prevented. The use of synthetic PA 21 or YIGSR peptides has been shown to inhibit neovascularisation on the developing chick CAM and YIGSR peptides have been shown to prevent vascular invasion of the rabbit cornea (Grant et al., 1989).
Antibodies have been raised to the endothelial cell-surface integrin, which binds to the RGD-containing laminin Achain peptide, and are capable of inhibiting endothelium-ECM interactions. Similarly, antibodies with specificity for the 32 kDa endothelial YIGSR-binding protein are capable of inhibiting the morphological changes and cell-cell interactions responsible for the formation of capillary tubes. Other anti-angiogenic agents which may disrupt endothelium-ECM interactions include fumagillin and the 16 kDa PRL fragment.
The PECAM (CD 31) molecule has been localised predominantly to endothelial cell intercellular junctions (Muller et al., 1989), leading to the suggestion that it may play an important role in the adhesive reactions between endothelial cells which accompany tubular differentiation.
The possibility that anti-PECAM antibodies might be able to disrupt this interaction is being investigated (Bicknell and Harris, 1992).
Anti-angiogenesis strategies that have been suggested are numerous, reflecting the complexity of the process and the number of levels at which intervention might be possible. The list of agents with anti-angiogenic properties is rapidly expanding, and includes several for which a mechanism of action has not yet been established, nor has a definite role in the inhibition of tumour angiogenesis been proven. As yet, success with this approach has been limited and essentially confined to the experimental setting.
Properties of the tumour vasculature Studies of tumour vascular morphology have identified a variety of structural differences between tumour and normal vasculature. Tumour vasculature is composed of abnormal vascular elements including sinusoidal vessels, giant capillaries and blood channels with a discontinuous endothelial lining. Normal vessels parasitised from the host tissues, capillary sprouts and arteriovenous anastomoses also contribute (Warren, 1979). The vasculature is not arranged in an efficient network, as is seen in normal tissue, but forms disorganised network patterns.
The physiological properties of tumour vasculature are strikingly different from normal vasculature. Tumour vasculature lacks innervation and therefore has an impaired 261 go capacity for autoregulation (Mattsson et al., 1979). The absence of a collateral potential and the reduced effective capillary density (Gunduz, 1981) combine to promote the conditions of hypoxia and acidosis often seen in tumours (Vaupel et al., 1989). Tumour endothelial cells, in contrast to normal endothelial cells which might divide only twice in a lifetime, are highly proliferative  and lack some of the differentiated features of normal endothelium such as alkaline phosphatase and 5'-nucleotidase activity (Murray et al., 1989). The tumour vasculature contains vessels, particularly at the tumour margin, which are leaky to macromolecules (Dvorak et al., 1988). The combination of leaky blood vessels and poor lymphatic drainage result in the raised interstitial hydrostatic pressure which is a feature of many tumours and which further impairs tumour perfusion (Vaupel et al., 1989).
Immunohistochemical markers of normal endothelium, such as factor VIII-related antigen and angiotensinconverting enzyme, are often deficient in the tumour vasculature, possibly reflecting the lack of differentiation of tumour endothelium (Denekamp, 1990). The corollary to this observation is the existence of novel proliferation antigens on tumour endothelium which serve to discriminate it from normal endothelium and which might provide a basis for specifically targeting tumour vasculature (Clarke and West, 1991).

Endothelial proliferation kinetics
Most studies of tumour proliferation kinetics have concentrated on the malignant cells of the tumour parenchyma to the exclusion of cells in the stromal compartment of solid tumours. The first detailed study of the proliferation kinetics of tumour stromal elements obtained mean comparative [3H]thymidine labelling indices of 35%, 11.4% and 9.1% for the tumour cells, endothelial cells and fibroblasts, respectively, in C3H murine mammary tumours. It was observed that the mitotic index and the labelling index of malignant cells decreased with increasing distance from the capillary lumen. Based on these observations it was suggested that the endothelial proliferative rate might be the rate-limiting step in solid tumour growth (Tannock, 1970). In a study of proliferation kinetics in pulmonary metastases of spontaneous mammary tumours in C3H/He mice, it was observed that the labelling index and growth fraction of carcinoma cells decreased with increasing tumour volume, and that the mean labelling index of endothelial cells was both higher than that of the carcinoma cells in the larger metastases and independent of tumour volume. This contradicted the earlier view that the endothelial proliferative rate was rate limiting for tumour growth, and prompted the suggestion that with increasing tumour size the decreasing effective capillary density was the rate-limiting parameter (Gunduz, 1981).
The questions relating to tumour vascular proliferation kinetics have been well summarised by Hirst et al. (1982). Do tumour growth rates decline owing to inadequate endothelial proliferation? Is endothelial proliferation adequate but the three-dimensional organisation of the tumour vasculature inadequate and rate limiting? Alternatively, could it be that tumour growth rates are independent of the endothelial proliferative rate and reflect some intrinsic property of the tumour cells themselves? These questions have not been satisfactorily answered by the studies to date, although it would appear from a review of the proliferation kinetics of 131 experimental tumours that no correlation exists between the tumour cell and endothelial cell labelling indices, nor has one been demonstrated between either of these variables and the tumour volume doubling time . However, the emphasis that emerged from these pioneering studies of endothelial proliferation kinetics in experimental solid tumours was the approximately 50-fold increased proliferative rate of tumour endothelium relative to normal endothelium (Weiss et al., 1988). This observation prompted Denekamp (1982) to suggest the tumour endo-thelium as a target for cancer therapy, an approach termed anti-proliferating endothelium therapy.
Recent proliferation data for both human and animal tumours have suggested that the magnitude of the increased proliferative rate of tumour endothelium over normal endothelium may not be as dramatic as was originally thought. A study of endothelial proliferation in 20 specimens of human breast cancer demonstrated mean labelling indices of 7.3% and 2.2% for tumour and endothelial cells respectively, although no indices were quoted for normal breast tissue (Fox et al., 1993). The technique of double immunostaining, both for bromodeoxyuridine (BrdUrd), providing a visual marker of proliferation, and for the CD 31 endothelial surface antigen, was considered to be a more accurate technique than the previous reported methods based on autoradiography without specific endothelial staining. The authors suggested that the true endothelial labelling index may have been even lower, owing to the failure to identify endothelial cell nuclei in immunoreactive vessels deep within the substance of the tumour. However, their observation that endothelial proliferation was maximal at the periphery of tumours could mean that the labelling of peripheral endothelial cells was underestimated for the same reason. This heterogeneity of endothelial proliferation status within a tumour was confirmed in a similar study of ten human squamous cell carcinomas labelled in vitro with BrdURd (Schultz-Hector and Haghayegh, 1993). This showed a mean endothelial labelling index of 1.8% compared with 0.16% in adjacent normal mucosa, but showed increased endothelial labelling in discrete foci expressing bFGF within the same tumour. The association of bFGF expression and an increased endothelial labelling was confirmed in murine squamous cell carcinomas by the same authors.
These recent studies have expressed a degree of pessimism about the prospects of the APET approach in solid tumours. However, they have demonstrated considerable heterogeneity of endothelial proliferation even within the same tumour, possibly related to oxygen and nutrient availability and local expression of angiogenic factors. Even if a proportion of the endothelial component of a solid tumour could be targeted on the basis of its proliferation status, the effects on global tumour perfusion might be sufficient to cause widespread tumour cell death.

Anti-vascular therapy
Denekamp's original suggestion of targeting the vasculature of solid tumours was exclusively concerned with the endothelial component of the tumour vasculature. The antivascular approach has since been broadened to encompass a variety of strategies designed to exploit the properties of the tumour vasculature (Denekamp, 1990;Denekamp, 1993). It has been suggested that an occult anti-vascular effect may be operating in a number of conventional and experimental anti-cancer strategies, further emphasising the potential of anti-vascular strategies directed at solid tumours.
Many solid tumours contain areas in which the perfusion is precariously balanced between adequacy and insufficiency. Given the lack of collateral reserve, a relatively small insult could be enough to precipitate vascular failure accompanied by the ischaemic death of numerous tumour cells (Denekamp, 1990). The pathophysiology of failure of the tumour circulation may result from both local and systemic mechanisms. Local mechanisms operate as a result of direct endothelial damage. Endothelial cells respond to injury by shifting the local balance towards a procoagulant state. Platelet aggregation and white cell margination lead to sludging of nutrient vessels. Increased vascular permeability results in increased interstitial hydrostatic pressure, which tends to shut off the tumour microcirculation. Systemic mechanisms precipitating failure of the tumour vasculature include hypotension and alterations in both blood coagulability and viscosity.
Tumour endothelial cells may be inherently more sensitive 26 I to injury as a result of the actions of tumour-derived cytokines which have been shown to alter the physiological properties of endothelial cells in vitro. Endothelial monocyteactivating polypeptides I and II (EMAP-I and -II), purified from meth A fibrosarcoma cells, have both been demonstrated to enhance tissue factor expression by cultured endothelial cells (Clauss et al., 1990;Kao et al., 1992). Both polypeptides are chemotactic for monocytes, and EMAP-II is additionally chemotactic for neutrophils. EMAP-I and the angiogenic peptide VEGF/VPF increase endothelial permeability in vitro, providing a possible biochemical explanation for this property of tumour vasculature.
It is becoming apparent that some cancer treatments are more effective in vivo than would be anticipated from in vitro testing. These observations suggest that a host effect, such as immune-mediated or vascular-mediated tumour cell destruction, might be a component of the treatment. The characteristic features of an anti-vascular effect are rapid-onset patchy areas of cell death which are more conspicuous in large rather than small tumours (Denekamp, 1990). The chemotherapeutic drugs bleomycin, cyclophosphamide and the nitrosoureas have been demonstrated to cause endothelial damage (Lazo, 1986). Radiotherapy has been shown to cause vascular occlusion by thrombosis in 10mm capillaries, with relative sparing of larger 20-30mm vessels in experimental tumour xenografts (Solesvik et al., 1984). A further indication of the anti-vascular action of radiotherapy is seen in the tumour bed effect, in which tumour cells implanted onto an irradiated site grow more slowly than cells implanted onto a normal site (Begg and Terry, 1983). Both radiotherapy and chemotherapy cause ultrastructural damage to endothelial cells, including autolytic vacuole formation, intracytoplasmic oedema, the formation of cytoplasmic extrusions on the luminal surface of endothelial cells and the detachment of degenerated endothelial cells (Freudenberg et al., 1983;Ward et al., 1983). The biological response modifiers endotoxin, interferon, TNFa and IL-2 all have anti-vascular actions which might contribute to their overall effects on solid tumours.
Endotoxin has been shown to cause haemorrhagic necrosis in experimental tumours, most probably by an indirect effect involving TNFa (Carswell et al., 1975;Bloksma et al., 1982). Despite having angiogenic properties, TNFa has several antivascular actions, including the promotion of neutrophil margination, direct endothelial toxicity and the induction of a procoagulant state at the endothelial cell surface (Kallinowski et al., 1989). The use of interferon oc/p in the treatment of experimental solid tumours has resulted in extensive rapidonset vascular endothelial damage causing ischaemic tumour cell death (Dvorak and Gresser, 1989). IL-2 has also been linked with endothelial damage, possibly by the stimulation of endogenous lymphokine-activated killer (LAK) cells (Kotasek et al., 1988).  (Reinhold et al., 1978;Endrich et al., 1979). The mechanism of the vascular injury is likely to be multifactorial, but a direct effect on the endothelium, and especially proliferative endothelium, has been demonstrated (Fajardo et al., 1985). However, the response of human tumour vasculature to hyperthermic conditions has been less dramatic.
A vascular effect has also been suggested as part of the mechanism of action of photodynamic therapy (Star et al., 1986). The selective accumulation of porphyrin derivatives in tumours as opposed to normal tissues has been attributed to the leaky tumour vasculature (Bugelski et al., 1981). However, evidence exists to suggest that haematoporphyrin is preferentially taken up by endothelial cells rather than tumour cells, and that proliferating endothelial cells exceed quiescent endothelial cells in this capacity (West et al., 1990). A number of agents have demonstrated unexpected vascular actions in solid tumours without having significant cytostatic or cytotoxic effects on cancer cells themselves. Flavone acetic acid (FAA) has a profound anti-vascular action in some experimental murine solid tumours (Hill et al., 1989;Zwi et al., 1990) and has been suggested as the prototype anti-vascular agent (Denekamp, 1990). However, phase I clinical testing of FAA failed to achieve significant tumour regression and was complicated by dose-limiting hypotension (Weiss et al., 1988). The electron-affinic radiosensitiser misonidazole has been demonstrated to have an anti-vascular action in addition to its redox capability (Murray et al., 1987). The accumulating evidence of a vascular effect in different forms of cancer therapy, and the realisation that some agents may operate by an exclusively anti-vascular action, has prompted the suggestion that novel chemicals be tested for anti-vascular effects over and above routine screening for potential tumoricidal properties (Denekamp, 1990).
The distinctive properties of the tumour vasculature make possible the use of bioreductive drugs in cancer therapy. Misonidazole is an example of a non-toxic prodrug which is reductively metabolised in hypoxic cells to a toxic form.
Using this approach the hypoxic cancer cells are targeted rather than the genetic or proliferative status of malignant cells. The opposite strategy can be employed using the radioprotector WR2721 (Ethiophos) to protect normal cells from increased radiation doses. Ethiophos is poorly taken up by tumours, possibly as a result of reduced tumour endothelial alkaline phosphatase activity. Normal endothelium is able to phosphorylate Ethiophos, allowing transport of the radioprotector into normal cells (Denekamp, 1993).
The original concept of APET envisaged the linkage of an antibody with endothelial specificity to an S-phase cytotoxic drug, relying on the increased proliferative rate of tumour endothelium over normal endothelium for selectivity. Concerns about possible toxicity to normal tissues (Hart et al., 1981) and the demonstration of novel proliferation proteins on tumour endothelium (Clarke and West, 1991) have led to the modified aim of targeting tumour endothelium-specific antigens. A number of monoclonal antibodies with varying degrees of specificity for human tumour endothelium have been reported. The first such antibody to be described EN7/ 44, was derived by immunising mice with capillary-rich suspensions from human breast carcinoma, and was specific for endothelial cells in the tips of budding capillaries in proliferating tissues (Hagemeier et al., 1986). An alternative immunisation strategy employing capillary endothelial cells which had been cultured in tumour-conditioned medium generated antibodies HB6 and HU21, which stained the vasculature in a proportion of tumours without apparent vascular specificity in normal tissues (Clarke and West, 1991). The antibody, E9, has demonstrated a high degree of tumour endothelial specificity, despite being raised by immunising mice with unstimulated human umbilical vein endothelial cells. An interesting feature of this antibody was a considerable degree of heterogeneity of intensity of endothelial staining with individual tumours, prompting the authors to suggest that this might reflect unequal distribution of endothelial activating factors within the tumour substance (Wang et al., 1993). Incidental evidence for tumour endothelium-specific antigens has been demonstrated by workers investigating the immunobiology of osteosarcoma (Bruland et al., 1988) and renal carcinoma (Oosterwijk et al., 1986).
A 165 kDa cell-surface glycoprotein known as endosialin is probably the best characterised of the tumour endotheliumspecific antigens identified. Antibody FB5, which has specificity for endosialin, was derived by immunising mice with human fetal fibroblasts. Endosialin was shown to be expressed by endothelial cells in 67% of a total of 128 malignant tumours tested, and was not present on endothelial cells in normal tissues (Rettig et al., 1992). Recently, tumour endothelial specificity has been claimed for endoglin, the antigen recognised by monoclonal antibodies TEC4 and TECI1. Immunotoxins constructed from TEC11 have been shown to be selectively cytotoxic for proliferating rather than non-dividing endothelial cells in vitro (Thorpe et al., 1994). However, endoglin has previously been identified as a major endothelial TGF-P-binding glycoprotein (Cheifetz et al., 1992), with a pan-endothelial distribution in most normal tissues (Gougos and Letarte, 1990;Lastres et al., 1992). The explanation for the apparent tumour vascular specificity of antibodies TEC4 and TEC 1 remains unclear, given the staining characteristics of other antibodies with specificity for endoglin.
Antibody-mediated tumour vascular targeting has been tested in an animal system in which a neuroblastoma cell line was grown in nude mice. The tumour vasculature was induced to express MHC class II determinants by transfection of the tumour cell line with interferon-Ty. Frozen section immunohistochemistry was used to demonstrate that the tumour vasculature could be stained specifically using an anti-class II MHC antibody, and that the tumour cells themselves could be stained using an antibody directed at an MHC class I antigen present on the tumour allograft but absent from the host (Burrows et al., 1992). Classical tumour cell targeting and tumour vascular targeting were compared using ricin A immunoconjugates constructed from the same MHC class I and II antibodies. These experiments not only confirmed the theoretical superiority of vascular targeting, but also confirmed that combination therapy using both types of targeting resulted in the best tumour responses (Burrows and Thorpe, 1993). If antibodies of sufficient specificity could be raised against human tumour endothelial proliferation proteins, a number of strategies could be employed for therapeutic tumour vascular targeting such as conjugation with toxins, radioisotopes or enzymes, using the antibody-directed enzyme prodrug therapy (ADEPT) approach.

Conclusions
Antibody-mediated tumour vascular targeting could constitute an attractive alternative to similar humoral approaches directed at the parenchymal component of solid tumours. Unlike tumour-associated antigens expressed by cancer cells, tumour endothelial proliferation antigens are highly accessible to circulating antibodies. The failure of a single tumour capillary as a result of this and other anti-vascular strategies could lead to the ischaemic death of many malignant cells with nutritional dependence on the targeted vessel. The development of clinical applications for the related antiangiogenesis approach looks an increasingly realistic prospect given the plethora of agents with anti-angiogenic properties.
The targeting of tumour endothelial proliferation antigens and other anti-vascular approaches, in common with antiangiogenesis strategies, suffers from the theoretical problem that peripheral tumour cells could survive on a diffusiondependent basis. However, these surviving cells should be readily susceptible to conventional cancer treatments. An additional cause for concern is the effect of these approaches on wound healing, endometrial proliferation, placental development and other physiological processes involving proliferating endothelium. The theoretical benefits of vascular targeting and anti-angiogenesis strategies for cancer therapy still have to be realised in the clinical setting, but these approaches represent an attractive means of overcoming constraints imposed on conventional cancer therapy by the tumour vasculature.