The Role of the Tumor Microenvironment in Regulating Angiogenesis

Abstract

The tumor-associated stroma has been shown to play a significant role in cancer formation. Paracrine signaling interactions between epithelial tumor cells and stromal cells are a key component in the transformation and proliferation of tumors in several organs. Whereas the intracellular signaling pathways regulating the expression of several pro- and antiangiogenic proteins have been well characterized in human cancer cells, the intercellular signaling that takes place between tumor cells and the surrounding tumor-associated stroma has not been as extensively studied with regard to the regulation of angiogenesis. In this chapter we define the key players in the regulation of angiogenesis and examine how their expression is regulated in the tumor-associated stroma. The resulting analysis is often seemingly paradoxical, underscoring the complexity of intercellular signaling within tumors and the need to better understand the environmental context underlying these signaling mechanisms.

Carcinoma-associated fibroblasts deposit large amounts of extracellular matrix. This creates a hypoxic environment in tumors which may lead to the induction of proangiogenic factors (e.g., VEGF).

In the earliest stages of cancer, epithelial tumors, carcinomas, are physically confined within the region of the tissue from whence they arise. These early lesions (carcinomas in situ) are separated from the tissue parenchyma by the basement membrane (Hanahan and Weinberg 2000). Opposite the basement membrane are a myriad of cells consisting of fibroblasts, myofibroblasts, immune/inflammatory cells, and endothelial cells (Ronnov-Jessen et al. 1996). In addition to these cell types are the extracellular matrix proteins which they secrete and to which they, and tumor cells, attach (Ronnov-Jessen et al. 1996).

To progress to a clinically relevant and potentially lethal disease, tumor cells must also acquire the ability to escape the confines of the epithelial compartment and thus invade locally and disseminate systemically. To make this escape, tumor cells must degrade the basement membrane separating the epithelial compartment from the tissue parenchyma. The process of invading the tissue parenchyma, or being invaded by cells from the tissue parenchyma, initiates a new phase of tumor progression in which tumor growth becomes partially regulated by non-cell-autonomous processes regulated by paracrine and juxtacrine interactions with the tumor microenvironment (Chung and Davies 1996; Henshall et al. 2001; Tuxhorn et al. 2001). The tumor-associated stroma provides oxygen and nutrients via the vasculature as well as soluble and matrix-bound growth factors and enzymes that promote tumor proliferation and progression (Hanahan and Folkman 1996). The evidence for stromal cells in tumor progression suggests that they play a key role in matrix remodeling, tumor invasion, and metastatic spread (Picard et al. 1986; Grey et al. 1989; Camps et al. 1990). More importantly, epithelial cancer cells are able to alter their surrounding stromal fibroblasts to enhance tumor growth (Olumi et al. 1999). Specifically, paracrine signaling interactions between epithelial tumor cells and stromal cells have been shown to be a key component in the transformation and proliferation of tumors in several organs (Cunha et al. 1985; Donjacour and Cunha 1991; Hom et al. 1998). It has been shown, for instance, that stromal fibroblasts isolated from a prostate tumor induce tumor formation of immortal but nontransformed prostate epithelial cells when the mixture is injected orthotopically into nude mice (Olumi et al. 1999).

One process in tumor progression enhanced by tumor–stromal signaling is the induction of angiogenesis. As the endothelium does not cross the basement membrane in normal tissue architecture, in order for tumors to gain access to blood vessels they must first invade the surrounding stroma. Once the tumor cells have invaded the tissue parenchyma they must be able to transmit paracrine signals to the stromal cells to induce a proangiogenic environment. Although a myriad of pro- and antiangiogenic factors have been discovered and studied, the major effort in understanding their regulation has been in a cell-autonomous fashion. To date, the tumor cell-autonomous regulation of vascular endothelial growth factor (VEGF) (Rak et al. 1995; Damert et al. 1997; Wojta et al. 1999; Xiong et al. 2001; Akiyama et al. 2002) and Thrombospondin-1 (Tsp-1) (Rak et al. 2000; Watnick et al. 2003), two of the major positive and negative regulators of angiogenesis, have been described in detailed biochemical fashion.

The regulation of angiogenesis via signaling between epithelial tumor cells and stromal fibroblasts and endothelial cells may also be important in the establishment and proliferation of metastases. It has been well documented that tumors from various organs have distinct metastatic profiles (Chambers et al. 2002). For example, prostate cancer metastasizes preferentially to bone and liver, whereas breast cancer metastasizes to bone and lung, although in xenograft models it is possible to isolate variants of tumor cell lines that metastasize preferentially to lymph nodes (Pettaway et al. 1996). The ability of a tumor cell to survive and proliferate in a metastatic environment may ultimately rely on its ability to manipulate the angiogenicity of the stroma in this new environment.

The tumor-associated stroma, or tumor microenvironment, can grossly be categorized into two types of cells: (1) cells that are present in the normal tissue parenchyma before tumor development; and (2) cells that are recruited to the tumor-associated stroma from distal sites (i.e., the circulation or bone marrow). The first type is largely comprised of fibroblasts and endothelial cells, whereas the second type of cells is largely comprised of immune/inflammatory cells, including T- and B-cells, macrophages, neutrophils, mast cells, and other bone marrow–derived cells. In this work, the different cell types, as well as the extracellular matrix, and their contribution to tumor progression will be detailed and explained.

FIBROBLASTS

The tissue parenchyma of most organs is largely comprised of fibroblasts that, along with the extracellular matrix they secrete, make up the structural scaffolding of organs (Fig. 1). There are two distinct types of fibroblasts in normal tissue: fibroblasts and myofibroblasts. These two types of fibroblasts are distinguished by, among other markers, the expression of smooth muscle actin, which is expressed by myofibroblasts but not normal fibroblasts. Fibroblasts, although morphologically distinct, are poorly defined molecularly and were originally characterized more for what they are not: vascular, inflammatory, or epithelial cells (Tarin and Croft 1969). Fibroblasts are responsible for the synthesis and deposition of the extracellular matrix (ECM), the regulation of epithelial cell differentiation, and the regulation of inflammatory response to tissue insults (Parsonage et al. 2005; Tomasek et al. 2005). Fibroblasts are also key mediators of wound repair, where they invade the wound, synthesize and secrete ECM to anchor other cells recruited to the site of the wound, and facilitate healing wound contractions via the contractile properties of collagen (Gabbiani 2003). Fibroblasts in wound repair attain what has been termed an activated phenotype (Castor et al. 1979).

Figure 1.

Schematic of tumor microenvironment.

Activated fibroblasts proliferate at a faster rate, produce greater amounts of ECM and, as stated above, express α-smooth-muscle actin (Gabbiani 2003). Fibroblasts can be activated by growth factors released from damaged epithelial cells, such as TGF-β and basic fibroblast growth factor (bFGF), or via direct cell–cell contacts with leukocytes at the sites of wounds (Clayton et al. 1998; Choi and Tseng 2001). To accommodate the production and secretion of large amounts of ECM proteins, activated fibroblasts contain an oval-shaped euchromatic nucleus, rough endoplasmic reticulum, and prominent Golgi apparatus (Castor et al. 1979). Activated fibroblasts also produce proteases, such as matrix metalloproteases (MMPs) that degrade the ECM. These proteases aid in the turnover and reorganization of the ECM and secrete growth factors like hepatocyte growth factor (HGF) and bFGF (Rodemann and Muller 1991). Interestingly, once the wound is repaired, the number of activated fibroblasts is greatly decreased, although the overall number of fibroblasts in the area is not significantly changed (Gabbiani 2003). Thus, it is not easily discernible whether the decrease in the number of activated fibroblasts is caused by cell death or apoptosis and subsequent repopulation of normal fibroblasts from neighboring tissue or if the activated fibroblasts revert back to normal fibroblasts. However, the general consensus is that the activation is transient and once wound repair is complete, the fibroblasts revert back to a quiescent phenotype.

CARCINOMA-ASSOCIATED FIBROBLASTS

Tumor progression from the in situ stage to metastatic disease has been shown to be promoted by fibroblasts present in the tumor microenvironment (Elenbaas and Weinberg 2001). These fibroblasts found in the tumor microenvironment have been termed carcinoma-associated fibroblasts (CAFs) (Olumi et al. 1999). The molecular and genetic characterization of CAFs indicates that they are similar, if not identical, to activated fibroblasts found in the stroma of tissues undergoing wound repair, described above (Durning et al. 1984; Tsukada et al. 1987; Schor et al. 1988a,b). Specifically, they express smooth muscle actin, EGF (Normanno et al. 1995; Panico et al. 1996), HGF (Montesano et al. 1991; Seslar et al. 1993; Jin et al. 1997; Vande Woude et al. 1997; To and Tsao 1998), IGF-1 and -2 (Yee et al. 1989; Cullen et al. 1992; Ellis et al. 1994), and matrix remodeling enzymes such as MMPs (Basset et al. 1990; Wolf et al. 1993; Basset et al. 1994; Engel et al. 1994; Newell et al. 1994; Heppner et al. 1996; Chambers and Matrisian 1997; Lochter et al. 1998; Masson et al. 1998; Noel et al. 1998; McCawley and Matrisian 2000).

Furthermore, in some cancers ∼80% of the fibroblasts within the tumor stroma are thought to become activated (Sappino et al. 1988). The signaling mechanism by which CAFs are generated has not been conclusively demonstrated; however, in vitro studies have shown that TGF-β can induce CAF-like properties in normal fibroblasts (Ronnov-Jessen and Petersen 1993). Moreover, studies suggest that human carcinoma cells can convert normal fibroblasts into CAFs in a mouse xenograft model (Orimo and Weinberg 2006). Once fibroblasts become CAFs, they can be cultured in the absence of carcinoma cells and retain their CAF phenotype in culture until they undergo senescence (Orimo et al. 2005). It is interesting to note that chickens infected with Rous sarcoma virus develop invasive carcinomas when wounded, showing the oncogenecity of tumor stroma (Dolberg et al. 1985). These studies show the importance of tumor stroma, and CAFs in particular, to the process of tumor progression.

When carcinomas progress to the invasive state the basement membrane is degraded and stromal cells, including CAFs, inflammatory response cells, and newly formed capillaries, come into contact with the tumor cells (Hanahan and Weinberg 2000). CAFs in the stroma of invasive carcinoma continue depositing large amounts of ECM, including tenascin C in some cases (Chiquet-Ehrismann et al. 1986; Inaguma et al. 1988). It has been shown that in breast and bladder carcinomas expression of tenascin C correlates with increased tumor invasiveness (Mackie et al. 1987; Brunner et al. 2004). The accumulation of ECM in tumors contributes to increased interstitial fluid pressure which hinders oxygen and nutrient diffusion (Netti et al. 2000; Brown et al. 2004). Thus, CAF-mediated hypoxia could lead to the expression of HIF-1α and the induction of VEGF, thus providing a mechanism by which CAFs can promote angiogenesis in tumors (Fig. 2).

Figure 2.

Tumor–stromal paracrine signaling. Tumors secrete factors into the microenvironment that act on the stromal cells to either promote or inhibit the growth of new blood vessels (angiogenesis).

As stated above, CAFs are associated with tumor cells at most stages of cancer progression. Many studies have shown the ability of fibroblasts to promote cancer. For example, patients genetically predisposed to breast cancer contained skin fibroblasts that proliferated more rapidly in vitro (Kopelovich 1982). It has also been shown that CAFs can promote tumor growth in a mouse xenograft model whereas normal fibroblasts cannot (Olumi et al. 1999). Olumi et al. (1999) demonstrated that massive tumors grew in the grafts containing CAFs, whereas no tumors grew in grafts containing normal fibroblasts. This shows how CAFs aid in the formation of tumors, probably through induction of tumorigenic changes in epithelial cells.

Tumor progression is also mediated by CAFs. It has recently been shown that mixing human breast carcinoma cells with CAFs in a mouse xenograft gave rise to tumors that were larger and more angiogenic than when mixed with normal fibroblasts (Orimo et al. 2005). Furthermore, it was shown that the increase in tumor cell proliferation was mediated by stromal cell–derived factor 1 (SDF1) secreted by the CAFs binding to the CXCR4 receptor on tumor cells. Additionally, in this model it was shown that secreted SDF1 stimulated angiogenesis by recruitment of endothelial progenitor cells (EPCs) into the tumor. It has been previously shown that EPCs are recruited during tumor angiogenesis and differentiate into vascular endothelial cells (Lyden et al. 2001).

Another study showed the ability of CAFs to induce invasiveness in vivo with rat colon carcinoma cells that were not invasive in vitro (Dimanche-Boitrel et al. 1994). CAFs also secrete MMPs that help degrade the basement membrane and promote tumor invasion. For example, MMP3 secreted by CAFs can promote tumor cell invasiveness (Lochter et al. 1997). This is accomplished by MMP3-mediated cleavage of the extracellular domain of the adhesive protein E-cadherin on the surface of mammary epithelial cells. Cleavage of E-cadherin causes mammary epithelial cells to disperse and undergo epithelial-to-mesenchymal transition, which promotes tumor cell invasiveness.

CAFs have also been implicated in tumor metastasis, by promoting the proliferation of tumor cells at the metastatic site. For example, a hepatic metastatic cell line was shown to secrete factors that activate fibroblasts in vitro (Olaso et al. 1997). These activated fibroblasts were shown to be within the tumor stroma of the metastasis, and quiescent fibroblasts taken from the liver of mice were activated when cultured with conditioned media (CM) from the melanoma metastasis. The tumor CM induced fibroblast migration, proliferation, and production of MMP2. This suggests that CAFs help to create a niche for tumor cells at metastatic sites (Olaso and Vidal-Vanaclocha 2003).

Yet another study showed that mice deficient for the Mts1 protein, which stimulates tumor metastasis, failed to grow metastases when highly metastatic mammary carcinoma cells were grafted onto these mice (Grum-Schwensen et al. 2005). Furthermore there was a significant delay in tumor uptake as well a decrease in tumor incidence as compared to wild-type mice injected with the carcinoma cells. When the tumor cells were mixed with Mts1 competent fibroblasts and injected into the mts1 knockout mice, the ability of these tumors to metastasize was partially restored. Additionally, it has been shown that conditioned media from metastatic human breast and prostate carcinoma cell lines are able to repress the expression of Tsp-1 in fibroblasts from tissues where the carcinoma is known to metastasize (Kang et al. 2009). This shows the ability of tumors to prime metastatic sites for angiogenesis by decreasing the levels of an endogenous angiogenesis inhibitor.

It is clear that angiogenesis is an essential step in the progression and metastasis of tumors. Fibroblasts play an important role in promoting not only tumor progression but angiogenesis as well. CAFs produce growth factors like VEGF which aid in the recruitment and activation of endothelial cells within the tumor stroma. During tumor invasion, CAFs produce not only angiogenic growth factors, but also produce proteases which break down not only the basement membrane of the tumor-associated tissue but can break down the basement membrane of stromal blood vessels, an essential step in angiogenesis. Finally, during tumor metastasis CAFs are able to create permissive environments for tumor growth and angiogenesis at metastatic sites. Studies from Kalas et al. (2005) and Kang et al. (2009) have shown that tumors secrete factors that are able to repress the expression of Tsp-1 in fibroblasts. These studies underscore the importance of Tsp-1 in the induction of angiogenesis. In order for tumor cells to induce angiogenesis, the balance of angiogenic stimulators and inhibitors must be shifted toward induction of angiogenic stimulators in the tumor stroma.

CELL SIGNALING MECHANISMS AND FACTORS INFLUENCING STROMAL ANGIOGENESIS

Vascular Endothelial Growth Factor

One of the most potent proangiogenic proteins, VEGF induces endothelial cell migration and proliferation as well as vascular permeability (Senger et al. 1983; Leung et al. 1989). The regulation of VEGF in carcinoma cells has been extensively studied. Signal transduction cascades emanating from Ras and PI3 kinase lead to the increased transcription of VEGF and consequently to its secretion into the extracellular space (Rak et al. 1995). It has also been discovered that carcinoma cells secrete proteins into the extracellular space that modulate VEGF production and secretion by stromal fibroblasts in the tumor-associated stroma, such as TGF-β, platelet-derived growth factor (PDGF), and bFGF (Brogi et al. 1994; Tsai et al. 1995).

Stromal VEGF expression was first shown to be regulated by carcinoma cells using a transgenic mouse model in which GFP expression was driven by the VEGF promoter (Fukumura et al. 1998). In this model, the activation of the VEGF promoter downstream from a signal transduction cascade would result in the expression of GFP. Examination of tumor xenografts in these VEGF-GFP mice by fluorescence microscopy revealed that the fibroblasts infiltrating and immediately surrounding the carcinoma cells were bright green, indicating that the VEGF promoter had been activated. In the absence of tumors, however, the fibroblasts in normal tissues in these mice did not express GFP. These results indicated that tumors secreted a protein into the extracellular space that stimulated VEGF expression in the stroma.

The significance of the stromal expression of VEGF in tumor growth and progression was not immediately obvious from these results. It was clear that VEGF expression was being stimulated; however, it was not clear whether this stimulation was required for tumor growth, supportive of tumor growth, or merely a by-product of tumor growth. The evidence that stromally produced VEGF was important for tumor growth came from experiments testing the efficacy of the human specific anti-VEGF antibody bevacizumab (Avastin). In these experiments human tumor cells were implanted into immunocompromised mice and the resulting tumors were treated with the human-specific VEGF antibody (Kim et al. 1993). Whereas the antibody was effective at significantly slowing tumor growth, it was not 100% effective. It was postulated that perhaps a residual angiogenic stimulus was being provided by the production of murine VEGF, whose activity was not inhibited by the human-specific antibody. To test this hypothesis, human tumor xenografts were treated with human-specific VEGF antibodies as well as mFlt(1-3)-IgG, which inhibits both human and murine VEGF by acting as a decoy receptor (Gerber et al. 2000). Treatment with both VEGF therapies resulted in the complete blockade of tumor growth, thus demonstrating the requirements for tumor- and stromal-produced VEGF.

BASIC FIBROBLAST GROWTH FACTOR

Basic fibroblast growth factor (bFGF, FGF2), is another extremely potent proangiogenic growth factor (Shing et al. 1984; Klagsbrun et al. 1986; Folkman and Klagsbrun 1987). Interestingly, despite the presence of high-affinity cell surface receptors (Dionne et al. 1990) and the fact that bFGF stimulates endothelial cell proliferation and angiogenesis both in vivo and in vitro, it lacks a signal sequence that would otherwise direct its secretion (Abraham et al. 1986). Although the role of bFGF as a stimulator of angiogenesis is unequivocal, its paracrine regulation in stromal fibroblasts and subsequent effect on tumor angiogenesis is clouded by the fact that that it also potently stimulates tumor cell proliferation via FGFR signaling via both autocrine and paracrine signaling (Rogelj et al. 1988, 1989; Gleave et al. 1991). Nevertheless, it has been shown that bFGF expression in the stroma of lung adenocarcinoma inversely correlates with disease progression and patient survival (Guddo et al. 1999). In addition to its paracrine regulation in fibroblasts, it has been shown that stem cell factor (SCF) and TGF-β are potent stimulators of bFGF expression in inflammatory cells, including macrophages, mast cells, and neutrophils (Qu et al. 1998). The role of these cells in tumor angiogenesis will be detailed later.

Thrombospondin-1

Tsp-1 is an endogenous inhibitor of angiogenesis which functions via a bimodal approach: it binds to CD36 on the endothelial cell surface and renders the cell insensitive to both VEGF and bFGF, via an as yet undetermined mechanism. Tsp-1 also binds to and functionally inactivates MMP-9, a MMP shown to liberate VEGF from the ECM (Bergers et al. 2000; Rodriguez-Manzaneque et al. 2001). One signal transduction pathway that has been shown to induce the repression of Tsp-1 leads from PI3-kinase to the Rho GTPase to ROCK to Myc, which represses Tsp-1 in a phosphorylation-dependent manner (Watnick et al. 2003). This pathway has been shown to be active in several human breast cancer cell lines in which Tsp-1 expression was virtually silenced (Watnick et al. 2003). Furthermore, in a majority of the surveyed breast cancer cell lines the pathway previously described (Watnick et al. 2003) was shown to be responsible for the silencing. Thus, this pathway represents the first biochemical elucidation of a cell-autonomous “angiogenic switch.”

Although the expression of VEGF in the tumor-associated stroma is widely accepted to have a positive correlation with tumor progression (Fukumura et al. 1998; Brown et al. 1999; Mueller and Fusenig 2002), the role of Tsp-1 expression in the tumor-associated stroma is unclear. Tsp-1 expression by epithelial tumor cells is observed infrequently and ectopic expression of Tsp-1 is inhibitory to tumor growth (Streit et al. 1999; Watnick et al. 2003). Stromal Tsp-1, meanwhile, has been correlated with a desmoplastic response and increased invasiveness in a subset of breast cancers (Wong et al. 1992; Bertin et al. 1997; Brown et al. 1999), whereas it has been shown to be inhibitory to early-stage breast cancers (Clezardin et al. 1993). Expression of Tsp-1 by stromal fibroblasts has been shown to be inhibitory to tumor formation and growth (Filleur et al. 2001). Intriguingly, the same report showed that tumors that arose in an environment high in Tsp-1 eventually overcame the inhibitory effects of this protein by increasing their production of VEGF. Thus, the complex interrelationship between these two proteins and their relative expression levels in the tumor-associated stroma can play a key role in the induction and maintenance of the angiogenic phenotype in human tumors.

The work described above shows that VEGF expression in the stroma is a critical component in tumor-mediated angiogenesis. Conversely, Tsp-1 expression in the tumor-associated stroma can be a potent inhibitor of tumor angiogenesis and growth. The question that arises, then, is how do tumors stimulate the expression of VEGF in the stroma while concomitantly repressing the expression of Tsp-1?

TGF-β

One of the most interesting paradoxes involving tumor-derived growth factors involves the effects of TGF-β. It had been shown that TGF-β displays a potent proangiogenic activity in vivo (Roberts et al. 1986). In seemingly diametric opposition are the in vitro data that demonstrate the growth-inhibitory activity of TGF-β on cultured endothelial cells (Baird and Durkin 1986; Frater-Schroder et al. 1986). These diverse activities were resolved with the discovery that TGF-β stimulated the expression of VEGF in both fibroblasts and tumor cells, indicating that tumor-secreted TGF-β could elicit proangiogenic effects via the induction of VEGF in stromal fibroblasts (Brogi et al. 1994; Pertovaara et al. 1994). Moreover, TGF-β has been shown to have profound effects on the expression of bFGF in fibroblasts, stimulating its expression by more than sixfold (Goldsmith et al. 1991). These results suggest that at lower levels tumor-secreted TGF-β may act on tumor-associated fibroblasts to stimulate VEGF and bFGF production to stimulate angiogenesis. Conversely, when expressed at higher levels, TGF-β may act directly on endothelial cells to inhibit their proliferation and thus inhibit angiogenesis.

Adding to the paradox of TGF-β’s role in tumor angiogenesis is the fact that TGF-β stimulates the expression of Tsp-1 and, in turn, is activated by Tsp-1 (Penttinen et al. 1988; Murphy-Ullrich et al. 1992; Schultz-Cherry and Murphy-Ullrich 1993; Schultz-Cherry et al. 1994a,b). TGF-β is produced and secreted by cells in an inactive, latent form that covalently attached to the latent associated peptide. TGF-β can be activated by proteases that cleave the latent form or by undergoing a conformational change, induced by Tsp-1 binding that exposes the receptor-binding region. TGF-β expression in fibroblasts has also been shown to be induced by hypoxia (Falanga et al. 1991). Thus, the Tsp-1-mediated inhibition of angiogenesis would result in a state of hypoxia, which would then trigger the expression of TGF-β by the tumor stroma. The latent TGF-β would subsequently be activated by Tsp-1 and induce the expression of VEGF, whose expression would also be stimulated by the Tsp-1-induced hypoxia (Brogi et al. 1994). In this way a tumor could overcome the effects of Tsp-1 and induce angiogenesis. However, if the levels of Tsp-1 were so high as to inhibit the excess VEGF produced, then the tumor would remain in a state of dormancy until the balance of angiogenic activity was tipped decisively toward the positive.

Intriguingly, although TGF-β induces the expression of both VEGF and Tsp-1, it accomplishes these diverse activities through the activation of different transcription factors. TGF-β binding to its receptors TGFβR1 and R2 activates a signal transduction cascade which culminates in the activation of the smad family of transcription factors (Hoodless et al. 1996; Liu et al. 1996). The TGF-β-mediated stimulation of Tsp-1 expression is achieved via activation of Smad2, whereas stimulation of VEGF is mediated via activation of Smad3 (Nakagawa et al. 2004). Thus, depending on the contextual environment within the tumor-associated stroma, TGF-β can stimulate the expression of VEGF, Tsp-1, or both proteins.

Platelet-Derived Growth Factor

Another growth factor that seemingly defies characterization as a pro- or antiangiogenic protein is PDGF. It was shown by Goldsmith et al. (1991) that PDGF was a potent stimulator of bFGF. Specifically, treatment of lung fibroblasts with recombinant PDGF resulted in a twofold increase in bFGF expression (Goldsmith et al. 1991). In response to the observations described above that the contribution of stromal VEGF needed to be inhibited to effectively block tumor growth, it was determined that stromal VEGF expression was stimulated by tumor-derived PDGF (Dong et al. 2004). Furthermore, inhibition of PDGF activity via administration of soluble PDGFR abrogated the stimulation of VEGF in the tumor-associated stroma and inhibited tumor angiogenesis. Moreover, a second member of the PDGF family, PDGF B, had previously been shown to up-regulate VEGF expression in vascular smooth muscle cells (Brogi et al. 1994). These data indicated that PDGF expression by tumor cells was a potent inducer of VEGF expression in the tumor-associated stroma.

One conclusion that could be drawn from the above study was that PDGF expression by tumor cells promotes angiogenesis via stromal VEGF induction. As with TGF-β, however, the activities of this growth factor are not as straight forward as these results would indicate. In addition to its ability to stimulate VEGF and bFGF expression to stimulate angiogenesis, PDGF has also been shown to stimulate Tsp-1 expression (Majack et al. 1987). The stimulation of Tsp-1 by PDGF was shown to be via the Raf-MAPK pathway in an analogous fashion to the stimulation of Tsp-1 by serum (Majack et al. 1985). Intriguingly, unlike the differential activation of transcription factors by TGF-β, the PDGF-mediated stimulation of VEGF is also via the Raf-MAPK pathway (Chang et al. 2006). Thus, its effect on angiogenesis mediated by tumor-associated fibroblasts is most likely mediated by the activation of different receptor/signal transduction pathways.

Hormones and Nuclear Receptors

The results presented above indicate that two of the most potent inducers of stromal VEGF and, in turn, angiogenesis also possess the seemingly counterproductive ability to stimulate the expression of Tsp-1. These seemingly diverse events downstream from TGF-β and PDGF ligation to its receptor indicate that tumor-derived TGF-β and PDGF expression should have no net angiogenic activity, because their activity stimulates the expression of both pro- and antiangiogenic proteins. However, inhibition of PDGF activity has been shown to inhibit tumor angiogenesis despite its effects on Tsp-1 (Dong et al. 2004). Similarly, TGF-β has been shown to stimulate angiogenesis, again, despite its stimulatory effect on Tsp-1. One explanation for the observed proangiogenic activities of these two proteins is that the expression of Tsp-1 in the tumor-associated stroma is somehow suppressed by an independent signaling mechanism. The suppression of Tsp-1 would result in the net stimulation of only the proangiogenic growth factors VEGF and bFGF by these two growth factors and would account for their observed proangiogenic activity.

Two candidates for such a Tsp-1-repressing factor are the hormones estrogen and androgen. Estrogen has, in fact, been shown to repress the expression of Tsp-1 (Sengupta et al. 2004). Similarly, androgen has been shown to repress Tsp-1 expression (Colombel et al. 2005). Although these two hormones have similar effects on Tsp-1 expression, the mechanisms by which they repress Tsp-1 are different. Estrogen inhibition of Tsp-1 is dependent on both ERK1/2 and JNK activity (Sengupta et al. 2004). Furthermore, the repression of Tsp-1 expression by estrogen appears to be mediated through a combination of transcriptional repression and inhibition of protein secretion. Conversely, androgen-mediated suppression of Tsp-1 expression appears to be solely via transcriptional repression, as an androgen-responsive element has been identified in the Tsp-1 promoter (Colombel et al. 2005).

Whereas hormone-mediated effects on tumor growth have been largely studied through their actions on hormone-responsive tumor cells, it has recently been shown that estrogen can have a systemic proangiogenic effect (Gupta et al. 2007). This study showed that estrogen receptor (ER)-positive stromal cells stimulate angiogenesis and promote tumor growth in response to estrogen even for ER-negative tumor cells. Although the regulation of VEGF and Tsp-1 was not the focus of that study, it is not unreasonable to assume that the stimulation of estrogen-mediated stimulation of angiogenesis observed could be partially mediated by a proangiogenic activity.

It has also been recently shown that another nuclear receptor family, the peroxisome proliferator-activator receptors (PPAR), can regulate the expression of VEGF and Tsp-1. Specifically, it has been shown that tumor cells injected into PPARα^−/− mice failed to grow beyond a microscopic size (Kaipainen et al. 2007). It was shown that the dormant state of these tumors was the result of increased Tsp-1 expression in the host stroma. Surprisingly, it was later determined that two ligands of PPARα, fenofibrate and WY14643, also stimulated the expression of Tsp-1 (Panigrahy et al. 2008). These seemingly discordant observations with respect to Tsp-1 expression could indicate that, in the absence of PPARα, another member of the PPAR family—perhaps PPARγ, which has been shown to stimulate expression of CD36 (Han et al. 2000), a receptor for Tsp-1—may compensate and stimulate the expression of Tsp-1. In keeping with these observations, it was shown that the PPARγ agonists rosiglitazone and pioglitazone inhibit bFGF and VEGF-mediated angiogenesis in the chick chorioallantoic membrane assay (CAM) (Aljada et al. 2008).

Nonprotein Mediators of Angiogenesis

The vast majority of paracrine signaling studies focus on the roles of cytokines and growth factors. However, one largely understudied signaling mechanism, mediated by lipids and phospholipids, has recently been investigated with respect to the regulation of Tsp-1 expression. Two independent studies revealed that generation of phospholipids and the resultant downstream signal transduction cascades could potently repress the expression of Tsp-1 in stromal fibroblasts. The first showed that generation of phospholipids, specifically sphingosine 1 phosphate (S1P), mediated by platelets could repress the expression of Tsp-1 in dermal fibroblasts via activation of G_i-protein-coupled S1P receptors (Kalas et al. 2005). It has been widely noted that platelets are commonly trapped in tumors, most likely caused by the presence of high levels of heparin-sulfates (Sack et al. 1977). Thus, platelet-mediated repression of Tsp-1, and increased angiogenesis and tumor growth, may be a by-product of this seemingly random event.

The second study investigating the role of S1P in repression of Tsp-1 showed that Ras-transformed cells secrete a low molecular weight (<3 kD) molecule that represses Tsp-1 in dermal fibroblasts via an S1P-dependent mechanism (Kalas et al. 2005). These results suggest that angiogenic tumor cells secrete factors that actively repress Tsp-1 in the surrounding tumor-associated stroma. As with the activity of estrogen and androgen noted above, the secretion of factor(s) by tumor cells that suppress the expression of Tsp-1 in the tumor-associated stroma may be a critical process in the escape from dormancy.

Matrix Metalloproteases

Local invasion across the basement membrane and within the tissue microenvironment of tumors is critical for tumor growth and ultimately progression to metastasis. One critical step in tumor migration and invasion is the degradation of the ECM and the resultant remodeling. A key driver of matrix degradation and remodeling is the MMPs. For example, it has been shown that co-injection of MCF7 breast cancer cells with fibroblasts significantly accelerated tumor growth (Noel et al. 1993). When the fibroblasts were engineered to express TIMP-2 (tissue inhibitor of metalloprotease 2), an inhibitor of MMP activity, the tumor-stimulating activity was lost (Noel et al. 1998). Analogously, when a general inhibitor of MMP activity, batimastat, was administered to mice that were co-injected with MCF7 cells and fibroblasts the tumors grew at the same rate as MCF7 cells alone (Noel et al. 1998).

In addition to allowing tumor cells to migrate and invade, MMP-mediated matrix remodeling also allows endothelial cells to migrate and form the leading edge of new blood vessels. Additionally, MMPs may function to liberate angiogenic growth factors like VEGF and bFGF that would otherwise be sequestered by the ECM. Evidence for the role of MMPs in angiogenesis was generated by crossing tumor-prone RIP-TAG2 mice with different matrix protease knockout mice (Bergers et al. 2000). In this model mice develop pancreatic islet cell tumors driven by the expression of the SV40 Large T antigen from the rat insulin promoter (Hanahan 1985). Crossing the RIP-TAG mice with MMP2 knockout mice impaired tumor growth but had no effect on angiogenesis (Bergers et al. 2000). Crossing the RIP-TAG mice with urokinase knockout mice had no effect on tumor growth. However, crossing RIP-TAG mice with MMP9 knockout mice inhibited tumor growth and angiogenesis (Bergers et al. 2000). In addition to cleaving matrix proteins, it has also been shown that MMP9 can cleave TGF-β, which is normally secreted as a pro-protein covalently bound to the latent associated peptide and is thus inactive. The ability of MMP9 to convert TGF-β from the latent to active form has been shown to stimulate growth of mammary tumor model (Yu and Stamenkovic 2000).

BONE MARROW–DERIVED CELLS

In addition to fibroblasts, the tumor microenvironment is made up several other types of cells that were present before tumorigenesis or migrated as a result. Most prominent among these cells are bone marrow–derived cells: mesenchymal stem cells (MSCs), macrophages, neutrophils, mast cells, and T cells. These cells migrate in response to the growing tumor mass, often interpreted as inflammation, and by the secretion of discrete growth factors and chemokines produced by the tumor cells.

Mesenchymal Stem Cells

MSCs are bone marrow–derived cells that have been characterized by the ability to differentiate into a myriad of mesenchymal cells: fibroblasts, osteoblasts, chondrocytes, adipocytes, pericytes, and muscle cells. MSCs are an exceedingly rare cell type within the bone marrow, comprising between 0.01% and 0.001% of the mononuclear cells (Civin et al. 1996; Pittenger et al. 1999). Human MSCs are defined by the expression of cell surface markers: CD44 adhesion molecule (HCAM), CD73, CD90, CD105 (endoglin), CD106 (VCAM-1), and STRO-1 (Dennis and Charbord 2002).

MSCs have been shown to be recruited to sites of wounding or inflammation, as well as to tumors (Hall et al. 2007). MSCs are recruited to tumors by multiple different growth factors and cytokines, including VEGF, bFGF, IL-8, EGF, HGF, and PDGF as well as CCL2, CCL7, and CXCL12 (SDF-1) (Schichor et al. 2006; Birnbaum et al. 2007; Dwyer et al. 2007; Kidd et al. 2008; Spaeth et al. 2008). In melanoma, a correlation has been demonstrated between MSCs and angiogenesis (Sun et al. 2005). Furthermore, following recruitment to the tumor, MSCs have been shown to secrete VEGF to stimulate angiogenesis (Coffelt et al. 2009).

In addition to correlation and expression studies examining the role of MSCs in angiogenesis, they have also been shown to be recruited and stimulate angiogenesis in vitro as well as in murine pancreatic xenografts (Beckermann et al. 2008). Tumors in mice injected with wild-type, vector control MSCs had twice as many blood vessels as normal tumors. Conversely, tumors in mice injected with MSCs in which VEGF had been silenced by lentiviral VEGF shRNA had comparable numbers of blood vessels to normal tumors (Beckermann et al. 2008). Thus, the ability of MSCs to home to, produce, and secrete VEGF can contribute to tumor growth via enhanced angiogenesis.

Macrophages

By far the most prevalent immune/inflammatory cell type present in tumors is the tumor associate macrophage (TAM; Balkwill and Mantovani 2001). Activated macrophages (i.e., those that have been recruited to sites of inflammation) are generally categorized into two types, M1 and M2, depending on the type of inflammation (Sher et al. 2003; Mantovani et al. 2004; Balkwill et al. 2005). M1 macrophages are effector cells that are able to potently kill microorganisms as well as tumor cells and secrete high levels of proinflammatory cytokines (Balkwill et al. 2005). M2 macrophages can have different response phenotypes based on the type of signals present in the inflammation. Thus, they are able to scavenge debris and stimulate angiogenesis, as well as tissue remodeling and repair (Goerdt and Orfanos 1999; Mantovani et al. 2002; Gordon 2003; Mosser 2003; Balkwill et al. 2005). TAMs, to the extent they have been studied, are most similar to, and share many properties with, M2 macrophages.

TAMs have been shown to display a growth-promoting activity for both human and experimental tumor models (Crowther et al. 2001). TAMs are preferentially recruited to sites of hypoxia which, in nontumorous tissue generally, is symptomatic of wounded or infected tissue (Crowther et al. 2001). Hypoxia stimulates the activity of the transcription factor HIF-1 which, in turn, stimulates the expression of the proangiogenic growth factors VEGF, bFGF, TNFα, and CXCL8 (Crowther et al. 2001). VEGF and bFGF directly stimulate endothelial migration and proliferation leading to new blood vessel growth into the hypoxic region of the tumor. Additionally, hypoxia stimulates the secretion of CXCL12 (SDF1), which potentiates the activity of VEGF and bFGF on endothelial cells (Salcedo et al. 1999; Schioppa et al. 2003). In addition, SDF1 has been shown to induce angiogenesis by recruiting bone marrow–derived endothelial precursor cells to tumors (Orimo et al. 2005).

Although for most of the time the net result of macrophage activity is growth promoting on tumors, macrophages can also inhibit the growth of tumors. Specifically, it has been shown that CSF stimulates macrophage production and secretion of metalloelastase (Dong et al. 1997, 1998). Metalloelastase is an extracellular protease which, among other substrates, cleaves plasminogen into multiple fragments, one of which is the antiangiogenic protein angiostatin (O’Reilly et al. 1994). Thus, macrophage activity with respect to tumor growth is highly context dependent and can, in certain circumstances, inhibit angiogenesis and tumor growth—in keeping with the macrophage’s general role of protecting the body from disease.

Mast Cells

Mast cells are multifunctional secretory cells, characterized by numerous large electron-dense granules composed of proteoglycans, of which heparin is the major component (Norrby 2002). Mast cells descend from pluripotent bone marrow progenitor cells that express the cell surface markers CD34, c-kit, and CD13 (Kirshenbaum et al. 1999). Mast cells in circulation are progenitor-like cells that differentiate/mature after being recruited to a given tissue. One consistent characteristic of precursor mast cells is the ability to produce MMP9, which is essential for migration into different tissue types (Tanaka et al. 1999). Mast cells express a variety of proteases including chymases, tryptases, and MMPs, which are stored in secretory granules (Norrby 2002). These proteases, especially the MMPs, specifically MMP2 and 9, are vital to mast cells’ ability to promote tissue remodeling and repair (Matrisian 1990). In addition to proteases, the secretory granules of mast cells are also depots for cytokines and growth factors, including TNF-α, GM-CSF, SCF, bFGF, EGF, PDGF, VEGF, and IFN-γ, IL-3, -4, -5, -6, -8, -10, -13, -14, and chemokines, such as (MIP)-1-α, I-309, (MCP)-1, and lymphotactin (Norrby 2002). The release of proteases, cytokines, and growth factors stored in the secretory granules of macrophages can be triggered by multiple cytokines, including IL-1, IL-3, and GM-CSF, Platelet factor 4, IL-8, SCF, (MCP)-1, and MIP-1-α (Tazzyman et al. 2009). Moreover, mast cells also produce and secrete MMPs 2 and 9, which have been shown to promote angiogenesis by liberating VEGF and bFGF from the ECM (Coussens et al. 1999, 2000). Interestingly, mast cells have been shown to be recruited to tumors by the proangiogenic proteins VEGF, bFGF, and TGF-β (Gruber et al. 1994, 1995). Thus, conditions within a tumor that necessitate the growth of new blood vessels recruit mast cells, which in turn further stimulate angiogenesis.

Experimental evidence for the functional role of mast cells in angiogenesis and tumor growth was provided by an elegant murine genetic model in which Myc expression in β cells was driven via fusion to a mutant form of the ER (Soucek et al. 2007). In this model it was shown that Myc activation by systemic administration of 4-hydroxy tamoxifen, induced β-cell tumors characterized by blood vessel infiltration accompanied by mast cell recruitment. These findings indicated that mast cells are required for angiogenesis at the onset of tumorigenesis and for maintenance of angiogenesis during tumor growth and progression.

Neutrophils

Whereas macrophages, TAMS, are the most prevalent and common leukocyte present in the tumor microenvironment, neutrophils are the most abundant leukocyte in the circulation (Tazzyman et al. 2009). Leukocytes originate in the bone marrow from hematopoietic pluripotent stem cells and differentiate via a process termed myelopoiesis. Neutrophil recruitment from the bone marrow is, in part, mediated by CXCL12 (SDF-1) and its cognate receptor, CXCR4, is expressed at high levels on the cell surface of neutrophils (Suratt et al. 2004). There are two types of neutrophils present in the circulation: circulating neutrophils which, as their name suggests, are freely circulating; and marginated neutrophils, which are bound to the endothelium of small blood vessels (Tazzyman et al. 2009). The marginated pool can be mobilized into the circulating pool by the cytokines such as IL-6 (Steele et al. 1987; Suwa et al. 2000). Neutrophils from the circulating pool are those recruited to sites of inflammation and tumors (Kanwar and Cairo 1993; Friedman 2002).

Elevated levels of neutrophils have been observed in multiple human tumors, including colon, lung, myxofibrosarcoma, gastric carcinoma, and melanoma (Nielsen et al. 1996; Bellocq et al. 1998; Mentzel et al. 2001; Mhawech-Fauceglia et al. 2006). In addition to CXCL12, one of the most potent chemo-attractants of neutrophils is CXCL8, which is expressed by both tumor and stromal cells of many types of human tumors (Bellocq et al. 1998; Xie 2001). Once recruited to tumors, neutrophils are able to stimulate angiogenesis by directly secreting VEGF and by secreting MMPs, which can release angiogenic growth factors such as VEGF and bFGF from their sequestration by the ECM (Coussens and Werb 1996; Gaudry et al. 1997).

Experimentally, it has been shown that, in a genetic murine model of squamous cell skin carcinoma, the MMP9 produced and secreted by neutrophils is required for the angiogenic switch (Coussens et al. 2000). In this model, it was observed that the source of MMP9 in the skin tumors was not from the tumors themselves but from neutrophils. These results have since been recapitulated by anti-GR1-mediated neutrophil ablation in the RIP-TAG2 islet cell tumor model and in a human ovarian cancer xenograft model in MMP9 deficient mice (Huang et al. 2002; Nozawa et al. 2006). Thus, it is now firmly established that neutrophils are important and, in some cases, required for tumor angiogenesis.

Like macrophages and mast cells, neutrophils also possess antitumor activity. For example, as early as 1975 it was observed that neutrophils could kill tumor cells (Clark and Klebanoff 1975). It was originally thought that the killing was mediated exclusively by myeloperoxidase. However, it has since been shown that neutrophils can kill tumor cells by multiple different mediators, including the release of proteases, membrane perforating agents, reactive oxygen species, and cytokines such as TNFα and IL-1β (Di Carlo et al. 2001). Moreover, neutrophils can inhibit angiogenesis via two distinct mechanisms, both of which are mediated by the protease neutrophil elastase. First, neutrophil elastase can inhibit angiogenesis by degrading VEGF and bFGF (Ai et al. 2007). Second, neutrophil elastase can cleave plasminogen into angiostatin, which inhibits VEGF- and bFGF-mediated angiogenesis (Scapini et al. 2002). Thus, this is another example of a cell type which can influence angiogenesis in the opposite way depending on the contextual signals within the tumor microenvironment.

CONCLUSIONS

Angiogenesis is a complex process driven by many different growth factors and cytokines and inhibited by a diverse range of proteins. As such, the regulation of angiogenesis by the tumor microenvironment is equally, if not more, complex. The signaling molecules secreted by tumors that act on stromal cells in a paracrine fashion can often have different activities with respect to the production and secretion of pro- and antiangiogenic proteins. As such, the composition of the tumor microenvironment as well as the stage of the tumor have profound effects on determining whether the tumor microenvironment is proangiogenic and growth promoting or antiangiogenic and thus growth inhibitory. The complex signaling mechanisms provide a myriad of potential, and as yet largely untapped, targets for therapeutic intervention to inhibit tumor growth in patients. Ultimately, the strategy of targeting tumor–stromal signaling molecules may prove to be hugely successful as the accounts of genomic instability and mutation in the stroma are exceedingly rare. As such, stromal-based antiangiogenic therapy may encounter less acquired resistance than traditional therapeutic strategies.