Influence of host microvascular environment on tumour vascular endothelium

Abstract

This review will focus on the tumour microvascular endothelium; how it is derived, modulated by angiogenic factors, and how the structure and function is influenced by the host tissue microenvironment.

Significant progress has been made in the detection and treatment of many cancers in the past 10 years, however, incidence and mortality of metastatic disease is still the most serious concern in the war against cancer (Wingo et al. 1998). There is a significant interest in attacking the tumour vasculature using antiangiogenesis or vascular targeting therapies, due to the ability of tumour cells to escape conventional radio/chemotherapies (Moore & West 1991; Strawn et al. 1996; Huang et al. 1997; O'Reilly et al. 1997; Pasqualini et al. 1997). While there are many reports and reviews describing tumour cell differences between the primary and metastatic tumours, relatively few studies have examined the differences in the microvasculature between metastatic and primary tumours. However, in order to fully capitalize on the antiangiogenic and vascular targetting therapies, it will be necessary to better understand how the host tissue microvascular environment influences the structure, function, and protein expression of the resulting tumour vessels. Reports describing host tissue influence on tumour cells are numerous (for review see Nicolson 1988; Alonsa-Varona et al. 1996; Umansky et al. 1996), therefore this topic will only be briefly discussed.

Vascular differences among normal tissues

The mammalian microvascular endothelium differs in structure and function throughout the body. There are characteristic differences which appear to be expressed in specific tissues, organs or parts of organs. Presently, there are three recognized types of differentiated microvascular endothelium. The first is found in the microvasculature of the lung, skeletal and visceral muscles and in the body wall, referred to as continuous endothelium. This vasculature is characterized, in most cases, by a large population of plasmalemmal vesicles, or caveolae, which are involved in transcytosis between blood plasma and interstitial fluid (Bruns & Palade 1968a). Moreover, within the continuous microvascular category, there are substantial differences in function and transport or permeability among organs, for example skeletal muscle, lung, and brain (Bruns & Palade 1968b; Simionescu & Palade 1971). The second type of differentiated microvascular endothelium is provided with fenestrae, regularly shaped ‘holes’, which appear in highly attenuated regions of endothelial cells and provide an opening from the lumen to the extravascular space. There are two types of these fenestrated endothelia, one has a diaphragm in practically every fenestrae while the other has similar openings (fenestrae) without diaphragms. Endothelia with fenestrae are spatially very restricted, occurring in capillaries of endocrine organs, visceral mucosae, and the kidney glomerulus (Farquhar & Palade 1961; Clementi & Palade 1969; Bearer & Orci 1985). These organ locations are typically characterized by their increased filtration or transendothelial transport. Fenestrated capillary endothelium is more permeable to water and small solutes than continuous endothelium, but equally impermeable to large proteins (Clementi & Palade 1969; Granger et al. 1979). The permselective nature of the fenestrae is due primarily to the anionic glycoproteins on the diaphragm rather than the diaphragm itself (Levick & Smaje 1987; Rostgaard & Qvortrup 1997). Finally, there is a type of microvascular endothelium which also has larger, irregular openings referred to as discontinuous endothelium. These endothelial cells are found lining the sinusoids of the liver. However, there are variants for each main type of microvascular endothelium and many endothelia have both caveolae and fenestrae.

There are recognized differences in the protein composition among various microvascular endothelia, specifically growth factor receptors, cell adhesion molecules, and transport channels (Dejana 1996; Gerber et al. 1997). Although not all the factors and mechanisms which differentiate these microvascular beds are understood, recent evidence strongly suggests growth factors, extracellular matrix, and haemodynamic forces specific to the microenvironment are primarily responsible (Risau 1995; Roberts & Palade 1995). In fact, it is now well-established that not all microvascular beds respond similarly in their expression of adhesion molecules when exposed to inflammatory cytokines (Belloni et al. 1992; Mantovani et al. 1992; Fukumura et al. 1995). These data suggest that microvasculature in different regions or organs of the body will necessarily vary in response to identical angiogenic stimuli.

Angiogenesis and angiogenic factors

Angiogenesis, the generation of new blood vessels from preexisting (host) vessels, has become a well-accepted concept in tumour biology (Folkman & Shing 1992; Hanahan & Folkman 1996). Tumour growth and metastasis are angiogenesis dependent and by inhibiting angiogenesis, it is possible to block tumour growth and spread (Claffey et al. 1996). Due to this dependence of tumour growth on angiogenesis, the concept of targeting tumour neovasculature or preventing angiogenesis has been increasingly considered (Jain 1996; Pasqualini & Ruoslahti 1996; Folkman 1997). It has been repeatedly demonstrated that by targeting the vessels directly or by inhibiting angiogenesis, tumours can be reduced and even eradicated (Nelson et al. 1988; Moore & West 1991; Strawn et al. 1996; Wickham et al. 1997; Lin et al. 1998). There are reports describing the selective induction of proteins on tumour vessels, including integrins (αvβ3, αvβ5) (Gladson 1996; Brooks et al. 1994) and the vascular endothelial growth factor (VEGF) receptors (Fetal liver kinase (flk)-1, fms-like tyrosine (flt)-1) (Strawn et al. 1996; Lin et al. 1998) to name a few. The efficacy of antiangiogenesis or vascular targeting therapies is dependent upon similar reactions toward tumour vasculature regardless of tumour location.

There are a number of growth factors which induce endothelial or smooth muscle cell proliferation in vivo resulting in angiogenesis, such as the fibroblast growth factor family (bFGF, in particular), angiogenin, epidermal growth factor (EGF), platelet-derived endothelial cell growth factor (PD-ECGF), transforming growth factor-α (TGF-α), platelet derived growth factor (PDGF), and VEGF (Lindner et al. 1990; Rifkin & Klagsbrun 1987; Jackson et al. 1994; Bohling et al. 1996; Ferrara & Davis-Smyth 1997; Fu et al. 1998). Of these factors, research efforts have recently focused on VEGF, due in part, to its endothelial cell specificity, spatial and temporal expression at times of vasculogenesis and angiogenesis, and prominent association with numerous vascular proliferative disorders (Breier & Risau 1996).

Originally described as a tumour-secreted protein that potently increased vascular permeability (Senger et al. 1983), VEGF is produced by many cell types other than tumours, including keratinocytes, macrophages, and folliculostellate cells (Ferrara & Henzel 1989; Brown et al. 1992b). VEGF is believed to be primarily responsible for angiogenesis observed in wound healing, solid tumour growth, diabetic retinopathy, rheumatoid arthritis, psoriasis (Koch et al. 1994; Berkman et al. 1993; Aiello et al. 1994; Detmar et al. 1994), as well as vasculogenesis in the embryo (Peters et al. 1993; Breier et al. 1992). VEGF is upregulated in response to hypoglycaemia and hypoxia (Shweiki et al. 1992; Stein et al. 1995; Suri et al. 1996), conditions commonly observed in solid tumours and ischaemic tissues (Shweiki et al. 1995; Brogi et al. 1996; Brown & Giaccia 1998). VEGF is an N-glycosylated protein most commonly found as a homodimer with an apparent MW of 46 000. There are 3 common isoforms, generated by alternative mRNA splicing, resulting in peptides of 121, 165 and 189 amino acids. Although these isoforms have different biochemical properties (e.g. heparin binding differences), it is not clear whether they have significant biological differences (effect and potency) (Park et al. 1993). There are two high affinity receptors known for VEGF which appear to be exclusively expressed and functional on endothelium, fms-like tyrosine kinase (Flt-1) and fetal liver kinase (Flk-1) (De Vries et al. 1992; Quinn et al. 1993; Morishita et al. 1995). Although other forms of VEGF (VEGF-B, -C, -D) and receptors (Flt-4) have recently been identified, their role in tumour angiogenesis has not been clearly demonstrated (Jeltsch et al. 1997; Ferrara & Davis-Smyth 1997; Laitinen et al. 1997). Flk and Flt are expressed on tumour vascular endothelium and are up-regulated on neovascular endothelium in normal development and pathologies (Brown et al. 1995; Fong et al. 1995; Mustonen & Alitalo 1995; Shalaby et al. 1995). Similar to VEGF expression, Flt-1 and Flk-1 expression is closely regulated to times of vasculogenesis and angiogenesis (Yamaguchi et al. 1993; Fong et al. 1996)

VEGF modification of vessels

Early studies on the mechanism of endothelial permeability largely focused on the effects of inflammatory mediators, such as histamine, serotonin, and bradykinin. These mediators increase microvascular permeability by opening endothelial intercellular junctions solely at the level of the postcapillary venules (Majno & Palade 1961; Majno et al. 1961). None of the inflammatory mediators appeared to affect capillaries or arterioles. These observations, made repeatedly using a variety of procedures since the 1960s, have led to the generally accepted belief that increased permeability was the result of opened or loose post capillary venular intercellular junctions.

VEGF has long been known to increase vascular permeability, but the cellular and molecular mechanisms have only recently been elucidated. A.M. Dvorak and coworkers hypothesized that VEGF increases microvascular permeability in tumour vessels by acting on clusters of fused caveolae, referred to as vesiculo-vacuolar organelles (VVO) (Kohn et al. 1992). These structures are also seen in postcapillary venules but are not hyperpermeable. Although endothelial gaps and fenestrae were observed in occasional micrographs of tumour vascular endothelium, they believed that these structural modifications were either artifacts or too infrequent to account for the increased microvascular permeability characteristic of tumour vessels or VEGF application (Feng et al. 1997). However, there is some controversy regarding the structural modifications in endothelium which are responsible for increased tumour microvascular permeability. Tracer or physiological studies have demonstrated that sterically stabilized liposomes or latex beads as large as 600 nm in diameter are able to escape from tumour vasculature, suggesting the presence of large endothelial gaps (Yuan et al. 1994a; 1995). These tracers are too large to exit the vessels via any other proposed mechanism, including transcytosis or through VVO, which often have stomatal diaphragms between the fused caveolae restricting the passage of anything larger than 9 nm in diameter.

Recently our laboratory decided to look at the endothelial morphological modifications induced by VEGF using the original model described in the early 1960s, the rat cremaster muscle. The cremaster is a thin skeletal muscle overlying the testes which is ideally suited for studying mammalian microvascular permeability because it is noninvasive, easily accessible, and essentially provides a 2-dimensional microvascular bed ideal for microscopic observation (Joris et al. 1990). As expected, a topical application of rhuVEGF₁₆₅ resulted in increased microvascular permeability within 10 min by opening endothelial junctions. However, we were surprised to find that within this brief exposure time many of the vessels normally lined with continuous endothelium now had fenestrated endothelium (similar results have been obtained with VEGF₁₂₁ and VEGF₁₈₉) (Roberts & Palade 1995). Although others had hypothesized that VEGF might be involved in inducing and/or maintaining fenestrated endothelium in adult kidney and brain due to its persistent expression only in cells adjacent to fenestrated endothelium (Breier et al. 1992; Brown et al. 1992a), our results were the first demonstration that VEGF can directly induce fenestrations in continuous endothelium in vivo. Continued studies in our laboratory demonstrated that chronic exposure of vessels to VEGF (from implanted tumour cells, VEGF-transfected cells, or slow-release pellets) also resulted in fenestrated endothelium (Figure 1) (Roberts & Palade 1997). These morphometric studies determined that a significant proportion of tumour vessels have fenestrated endothelium and open endothelial gaps. Interestingly, the fenestrae found in VEGF-induced neovasculature had a significantly decreased permselective glycoprotein coat, which allows the fenestrae to become hyperpermeable to proteins. Recently, Risau and coworkers demonstrated that VEGF can also induce fenestrations in endothelium in vitro (Esser et al. 1998b). Other studies have demonstrated similar results by perfusing the brain with VEGF (Dobrogowska et al. 1998). Recent efforts have demonstrated that VEGF increases endothelial permeability in vitro by opening interendothelial junctions by phosphorylating junctional proteins, VE-cadherin and occludin, through the mitogen-activated protein kinase signal transduction pathway (Esser et al. 1998a; Kevil et al. 1998). Whether VEGF increases microvascular permeability in vivo by a similar or identical mechanism remains to be seen. However, there is substantial data indicating that VEGF can induce fenestrations and open gaps in endothelium, consistent with the physiological measurements of tumour hyperpermeability.

Figure 1.

Electron micrographs of neovasculature in murine rhabdomyosarcoma grown (a) subcutaneously or (b) intracranially and (c) Chinese hamster ovary cells stably transfected with huVEGF₁₈₉ grown s.c. Neovascular endothelium in all these tumour models was often found to have fenestrations (arrowheads), induced by secreted VEGF. The neovascular interendothelial junctions (j) were often open or discontinuous when tumours were grown s.c. (a), but when implanted into the brain (b) the junctions were largely unaffected (unless the tumours were glioblastoma in origin). (d) An Elvax 40P slow-release pellet incorporated with 10 ng rhuVEGF₁₆₅ was implanted on the rat cremaster for 15 days resulting in fenestrated vessels which were not proliferative, demonstrating that chronic application of low concentrations of VEGF can induce and maintain a fenestrated endothelium without inducing angiogenesis. a: ×20,500, b: ×32,200, c: ×14,750, d: ×17,850.

Host influence of tumour biology

Drug susceptibility differences between metastatic lesions and their primary tumours are well documented clinically (Baggetto 1997; O'Neill et al. 1998). This infers adaptation and differential gene expression between the metastasis and the primary tumour. These differences are believed to result from selective genetic changes between the metastatic and primary tumour cells (Kerbel 1990). In the clinical setting, some tumours are known to preferentially metastasize to certain organs which has led to the concept that some organs have unique molecular addresses, presumably on the endothelium and specifically recognized by circulating tumour cells (Nicolson 1988; Weinstat-Saslow & Steeg 1994). Molecular modifications in adhesion molecule, cytokine, or growth factor expression of metastatic tumour cells have been demonstrated to change depending on the tumour location in vivo, suggesting that host tissue environment can regulate the differential expression of tumour cell proteins (Singh et al. 1994; Gutman et al. 1995). Taken together, these findings suggest that certain tumours will preferentially metastasize to particular organs which will differentially modify the tumour physiology. Additionally, tumour-secreted cytokines/growth factors modify the local environment surrounding the tumour to modulate the immune response, inhibit vascular cell adhesion protein expression, and induce angiogenesis (Kerbel 1990; Aaronson 1991; McCormick & Zetter 1992; Zocchi & Poggi 1993). All of these factors affect and modulate the tumour physiology, especially tumour vascular physiology.

Host vessels determine the structure and function of tumour vessels

Although it is well accepted that tumours derive their vasculature from surrounding normal tissues, that the structure and function of the host microvasculature eventually regulates the morphology and physiology of the tumour neovasculature is not as well understood nor accepted. This concept can only become more important as therapies continue to target tumour vasculature or exploit tumour microvascular permeability to reach tumour cells. The microvasculature of an experimental or clinical tumour is notoriously heterogeneous in vascular architecture and haemodynamics which is rendered more complicated when the host tissues which supply the vessels are varied (Jain 1988; Skinner et al. 1990; Leunig et al. 1992; Wesseling et al. 1997).

When identical tumour cells or angiogenic stimuli are placed in the brain or the subcutous, the microvascular permeability is governed more by the inherent permeability characteristics of the host tissues than the stimulus (Dellian et al. 1996). The vasculature which supplies the majority of the central nervous system has characteristically low permeability (Peters et al. 1991). Vessels in the normal brain have such low permeability that the smallest solutes (e.g. glucose and ions) must be actively transported (Virgintino et al. 1997). Capillaries of the normal brain consist of a continuous endothelium whose cells are joined by well-developed and complex tight junctions and essentially have no caveolae (Simionescu et al. 1988). When tumours were either implanted in cranial window or dorsal skinfold preparations, there were characteristic vascular pore sizes and permeabilities associated with different tumours (Yuan et al. 1994b). However, all tumours grown in the dorsal skinfold had increased vascular pore sizes compared to identical tumours implanted in the cranial window (Hobbs et al. 1998). These findings are consistent with the normal vasculature of the skin and the brain. The injection of tracers had no effect on the percentage of vessels which had fenestrated endothelium or open gaps. These functional data are consistent with morphometric analyses of tumours (human glioblastomas, U87MG & U251MG, murine mammary carcinoma, EMT-6, and rhabdomyosarcoma, M1S) grown intracranially (i.c.) or subcutaneously (s.c.). These data demonstrated a general reduction of tumour vessels with open endothelial gaps and fenestrated endothelium when tumours were implanted i.c. compared to s.c., regardless of tumour type (Roberts et al. 1998). Interestingly, of the tumours implanted in the brain, only the glioblastomas had vessels with open gaps as seen by electron microscopy. Clinically, the neovasculature of glioblastoma multiforme is characteristically hyperpermeable compared to other tumour types. This may be an indication that, unlike other tumours, glioblastomas produce additional factors which allow them to modify the extremely tight blood vessels in the brain (Risau 1998). Moreover, U87 glioblastoma grown i.c. had 2 times greater permeability to BSA than a murine mammary carcinoma (i.c.) even though the pore cutoff size was 4–5 times less, suggesting a dramatic increase in the number of gaps (Hobbs et al. 1998).

Similar findings have been observed when tumours are grown in the dorsal window compared to the liver. The liver vasculature is provided with sinusoidal endothelium which has nondiaphragmed fenestrae of 100 nm, making this vasculature hyperpermeable compared to the continuous endothelium of skin vessels. Tumour microvascular permeability was greater in tumour vessels derived from the liver compared to the skin, although the vascular density was increased in dorsal window implanted tumours (Fukumura et al. 1997).

Morphology is not the only characteristic determined by host microenvironment, neovascular endothelial gene expression can also be significantly affected by the type of host vessels from which they are derived. For example, the ability of endothelial cells to respond to cytokines, as measured by leucocyte adhesion and rolling is greater in tumour vessels derived from the skin compared to vessels derived from the liver (Fukumura et al. 1997). In this case, the tumour milieu (i.e. growth factors, extracellular matrix, cytokines) is identical, the only difference is whether the neovessels were originally from skin or liver. Additionally, using phage display peptide libraries, researchers have identified specific peptide sequences which are preferentially expressed by tumour vascular endothelium depending on the host tissue vasculature from which they were derived (Arap et al. 1998).

Previous reports have described a remarkable induction of the VEGF receptors, Flk-1 and Flt-1, on brain tumour vessels compared to surrounding normal brain vessels (Plate et al. 1993; 1994). Recently, using competitive PCR, we examined differences in VEGF receptor mRNA expression in tumours grown s.c. or in the brain. In all tumours, regardless of location, Flk-1 mRNA expression was equal or higher than Flt-1. More importantly, however, tumour vessels derived from the brain (i.c. tumours) had higher expression levels of both receptors compared to the identical tumour type grown s.c. There was a 2-fold increase in vascular density in tumours grown in the brain compared to s.c. tumours, but there was a 3- to 14-fold increase in Flt-1 and Flk-1 mRNA expression in i.c. tumours (Figure 2) (Roberts et al. 1998). This represents an increased inducibility of both Flk-1 and Flt-1 on brain-derived endothelium (i.e. higher receptor expression per endothelial cell) compared to skin-derived endothelium even though the angiogenic stimulus was identical in both locations.

Figure 2.

Competitive PCR was used to measure the amount of VEGF receptor mRNA (▪ Flk-1; □ Flt-1) in tumours grown subcutaneously (s.c.) or intracranially (i.c.). Tumours express more (or equal) levels of Flk compared to Flt regardless of tumour location. However, tumours grown in the brain express higher levels of the receptors compared to the same tumour grown s.c. This is due to an increased level of expression per endothelial cell and not an increase in vascular density. Although vascular density is 1.7–2x higher in i.c. tumours compared to s.c. tumours, the receptor amounts are more than 2x higher, suggesting a greater inducibility of receptor expression when tumour vessels are derived from the brain (from Roberts et al. 1998).

These data indicate that neovascular morphology, physiology, and endothelial gene expression are governed and affected not only by the factors and cytokines which induce angiogenesis, but by the normal surrounding host microvasculature from which the vessels are derived. The implications are powerful, regardless of whether new antitumour therapies target tumour cells or tumour vascular endothelium because the tumour vessel morphology affects tumour vascular permeability and therefore distribution of anticancer agents. Moreover, tumour vascular endothelium derived from one location may have unique vascular markers not present on neovessels supplying the metastases, implying that neovascular targets derived from the primary tumour may not be useful in treating metastases in other organs. However, questions remain as to whether the molecular, cellular, and morphological differences in tumour vascular endothelium must be addressed individually for each tumour site or whether common characteristics can be found and exploited.