Address reprint requests to James M. Pluda, Investigational Drug Brunch, CTEP DCTDC, NCI, 6130 Executive Blvd, Executive Plaza North, Room 715, Rockville, MD 20852.

This is a US government work. There no restrictions on its use

0093-7754/97/2402-0005$0.00/0

TUMOR-ASSOCIATED ANGIOGENESIS: MECHANISMS, CLINICAL IMPLICATIONS, AND THERAPEUTIC STRATEGIES

James M. Pluda

ANGIOGENESIS-DEPENDENT TUMOR GROWTH AND METASTASIS

RELATIONSHIP OF TUMOR ANGIOGENESIS TO CLINICAL OUTCOME

INITIATION OF TUMOR ANGIOGENESIS

KEY AREAS IN THE PATHOGENESIS OF TUMOR- ASSOCIATED ANGIOGENESIS

REGULATORS OF ANGIOGENESIS IN TUMOR SPECIMENS AND PATIENTS

THERAPEUTIC AGENTS

DEVELOPMENT STRATEGIES

CONCLUSION

ACKNOWLEDGMENT AND REFERENCES

Semin Oncol 24:203-218. Copyright © 1997 by W.B. Saunders Company.

IN 1971, THE NOVEL theory that tumors lay dormant yet viable, unable to grow beyond 2 to 3 mm3 in size in the absence of neovascularization was put forth by Judah Folkman (1). He postulated that the production by the tumor of a disable product he termed tumor angiogenic factor initiated the process of angiogenesis and resulted in the shift of nearby vascular endothelial cells from a resting state into one resulting in the formation of new vessels with subsequent vascularization of the tumor. The angiogenic process allowed for further growth, invasion, and ultimately metastasis of the tumor. This hypothesis resulted in the search for and discovery of naturally occurring simulators of angiogenesis (Table 1) as well as a number of endogenous inhibitors of angiogenesis (2). Substantial evidence has since accumulated supporting the hypothesis that tumor growth, invasion, and metastasis are directly related to the process of angiogenesis, and Folkman's model is as relevant today as it was novel in 1971. The indirect data supporting the role of neovascularization in tumor growth has been reviewed elsewhere by Folkman (3).

ANGIOGENESIS-DEPENDENT TUMOR GROWTH AND METASTASIS

The direct evidence implicating tumor-associated neovascularization playing a central role in the growth, and metastasis of tumors continues to accumulate. The in vivo administration of an immunoneutralizing monoclonal antibody (MoAb) against basic fibroblast growth factor (bFGF) to a nude mouse bearing a highly tumorigenic bFGP-secreting tumor resulted in significant inhibition of tumor growth (4). Significant suppression of tumor-associated neovascularization was observed histologically. In contrast the direct exposure of these tumor cells to the anti-bFGF MoAb had no effect on in vitro tumor cell growth. Thus, it appeared that the in vivo antitumor effect of the MoAb was due to the in vivo inhibition of bFGF-dependent angiogenesis. Subsequently, it was shown that the in vivo administration of bFGF to mice implanted with tumor cells lacking a receptor for bFGF resulted in increased size, invasiveness, and neovas-cularization of the implanted tumor (5). Similarly, administration of an MoAb specific for vascular endothelial growth factor (VEGF) inhibited the in vivo growth of human tumor cell lines implanted into nude mice (6). This anti-VEGF MoAb had no effect on the in vitro growth of these same cell lines. Induction of the high affinity tyrosine kinase receptors for VEGF, Flt-1 and Flk-l, on the vascular endothelial cells within tumors but not on normal brain endothelium, has been shown in both glioblastoma clinical specimens obtained from patients and in a rat brain tumor model (7,8). The in vivo transfection of endothelial cells using a retro-virus encoding a dominant-negative mutant of the Flk-1/VEGF receptor resulted in the inhibition of VEGF-mediated signal transduction and subsequent growth arrest of a wide variety of tumor types (9,10). Data suggest that the contribution of neovascularization to tumor growth lies not only in perfusion of the tumor but in the paracrine effects on tumor cells of factors secreted by stimulated endothelial cells (11,12). Microvascular endothelial cells release factors, such as bFGF and insulin-like growth factor types 1 and 2, that have been shown to promote tumor cell growth and migration (11,13,14). Finally, a variety of agents or factors that specifically inhibit the process of angiogenesis have been shown to arrest tumor growth in vivo (15-17).

Angiogenesis is also implicated as part of the process of tumor metastasis. The local shedding of tumor cells into the circulation has been shown to commence only after the tumor has become vascularized (18). The number of cells entering the circulation is related to the surface area of new tumor vessels represented by the degree of tumor neovascularization. In addition, the proliferating capillaries of tumor neovasculature are leaky and contain fragmented basement membranes, thereby increasing the entry of tumor cells into the circulation (18,19). Furthermore, the processes of angiogenesis and metastasis both require the presence and activity of matrix metalloproteinases or collagenases (20). Thus, the process of angiogenesis and tumor neovascularization appears to play a central role in the growth and spread of tumors.

RELATIONSHIP OF TUMOR ANGIOGENESIS TO CLINICAL OUTCOME

A growing body of clinical data links the degree of angiogenesis in the primary tumor to the risk or developing metastatic disease, and more importantly to duration of disease-free and overall survival. Weidner et al were among the first to report a correlation between the degree of primary tumor neovascularization as measured by the angiogenic index (number of vessels per microscopic field) in primary breast- carcinoma surgical specimens, and the subsequent development of metastatic disease (21). The angiogenic index alone was reported to be responsible for the association between tumor size and grade, the occurrence of lymph node metastasis, and ultimately early death in patients with breast cancer (22). Subsequently, it was shown that a significant correlation existed between the degree of tumor angiogenesis (microvessel density) and survival in patients presenting with lymph node-negative breast carcinoma (23,24). More recently, it has been shown using multiivariate analysis that while microvessel density, p53 protein expression, tumor size, and perilymphatic tumor invasion were all significant prognostic factors for disease-free survival, only, tumor microvessel density and tumor size were significant-independent predictors of overall survival (25). In node-positive breast cancer, intratumoral microvessel density was the strongest independent predictor of clinical outcome in patients who received either adjuvant hormones or chemotherapy (26,27). Similarly, the microvessel density of the primary tumor correlated with the pathologic stage and the presence of metastasis in patients with prostate carcinoma (28-30). There are now data correlating microvessel density with metastasis, recurrence, or mortality in other neoplastic diseases such as colorectal carcinoma (31-33), non-small cell lung cancer (34,35), gastric carcinoma (36,37), squamous cell carcinoma of the head and neck (38), melanoma (39-41), testicular germ cell tumors (42), bladder cancer (43), ovarian carcinoma (44), and pediatric brain tumors (45).

The processes or angiogenesis and tumor invasion appear to be related. Both involve the dissolution of basement membrane and the migration of proliferating cells into the interstitial stroma. Moreover, the shedding of tumor cells into the circulation occurs at the onset of angiogenesis and is quantitatively related to the surface area of tumor vessels. Recently, a significant correlation was shown between vascular density of the primary tumor and the intraoperative detection of circulating tumor cells in 16 women with breast cancer (46). The patient with the highest microvessel density had detectable circulating tumor cells before surgery and the highest recorded intraoperative circulating tumor cell count. There was also a suggestion that the microvessel density of the primary tumor was directly related to the absolute number of intraoperative circulating cells. Thus, it may be that the greater the number of angiogenic cells present within a tumor, the greater the degree of neovascularization that develops with increased possibility for shedding of angiogenic tumor cells and induction of neovascularization and tumor growth at metastatic sites. It is possible that the association between increased tumor angiogenic index, increased risk of metastasis, and decreased survival may, in part, be explained by shedding of a greater number of highly angiogenic cells into the circulation by highly angiogenic tumors, thereby increasing the likelihood of metastatic foci, which are also angiogenic and thus capable of rapid growth and further metastasis.

INITIATION OF TUMOR ANGIOGENESIS

Angiogenesis is important in the process or tumor growth, invasion, and metastasis; however, at what point does a tumor acquire the angiogenic phenotype? Most tumors initially arise without angiogenic activity, incapable of growing beyond 2 to 3 mm3 in size, enlarging beyond this size only after a "switch" to the angiogenic phenotype (14). The factors responsible for this switch are not completely understood. In transgenic mice containing an oncogene in the beta cells of the pancreatic islets, angiogenic activity was observed in a subset of hyperplastic islet cells Wore the onset of tumor formation (47). The timing and frequency of this angiogenic conversion within islet cells in vitro corresponded with the in vivo incidence of neovascularized tumors. Thus, hyperplasia per se may not have led to the development of tumors; but the induction of angiogenesis in the areas of hyperplasia did appear to result in tumor formation. In a tumor model of transgenic mice containing the genome of the bovine papilloma virus type 1, the switch to the angiogenic phenotype appeared to be associated with the ability to export bFGF from the cell (48). It is important to note that a tumor can vascularize, grow, invade, and metastasize without all cells within the tumor acquiring the angiogenic phenotype. An angiogenic growth factor-producing subpopulation of cells confers increased tumorogenicity to the whole tumor (49).

The genetic events associated with the acquisition of the angiogenic phenotype are not yet understood. Some data implicate tumor suppressor genes in this process. Bouck first postulated that angiogenesis was controlled by a tumor suppressor gene whose product mediated a high level of transcription for a second gene that encoded an inhibitor of angiogenesis (50,51). She subsequently showed that a tumor suppresser-dependent inhibitor of angiogenesis in hamster cells was immunologically and functionally indistinguishable from a fragment of thrombospondin (52). Ultimately, through an elegant series of experiments using cultured fibroblasts from patients with Li-Fraumeni syndrome, Bouck et al showed that control of angiogenesis in human fibroblasts was via the suppresser gene p53: The p53 gene product positively regulated the synthesis of thrombospondin-1 (TSP-1), a potent inhibitor of angiogenesis (53). Loss of p53 gene function was associated with reduced expression of TSP-1 and the subsequent switch to the angiogenic phenotype. Furthermore, it has been shown that in vitro restoration of p53 function in human glioblastoma cells-resulted in the secretion of a novel and potent inhibitor of angiogenesis termed glioma-derived angiogenesis inhibitory factor (54). Thus, one mechanism for acquiring the angiogenic phenotype appears to involve an alteration of a tumor suppressor gene with subsequent decreased production of an angiogenesis inhibitor.

Avascular tumors are not quiescent. Folkman et al showed that dormant, avascular microscopic tumor foci proliferate at a rate that is equivalent to that of rapidly growing tumors (55). However, in the dormant state, the rate of tumor cell apoptosis balances that of proliferation resulting in a nonex-panding tumor mass. It appears that the acquisition of the angiogenic phenotype results in a decrease in the apoptotic rate of tumor cells thus shirting the balance of proliferation versus cell death in favor of proliferation. Administration of an inhibitor of angiogenesis to mice bearing human tumor xenografts caused tumor regression to microscopic dormant foci (56). These now dormant tumors had the same proliferation rate as the untreated tumors, but they now had a significantly increased tumor cell apoptosis rate. Thus, the acquisition of the angiogenic phenotype was associated with a decrease in the tumor apoptosis rate resulting in the growth and subsequent invasion and metastasis of the tumor. These results suggest that specific inhibitors of angiogenesis might achieve not only cessation of further tumor neovascularization and growth, but also decrease tumor size and re-induce tumor dormancy.

KEY AREAS IN THE PATHOGENESIS OF TUMOR- ASSOCIATED ANGIOGENESIS

There are a number of steps involved in tumor associated angiogenesis, reviewed by Folkman (14), that are important clinically and suggest opportunities for specific therapeutic interventions. Factors such as bFGF, VEGF, and others, secreted by tumor, endothelial, and supporting cells are required for angiogenesis. Migration of tumor and endothelial cells through tissue extracellular matrix (ECM) is also required, facilitated by a group of enzymes known as matrix metalloproteinases (MMPs), reviewed by Ray and Stetler-Stevenson (57). These enzymes, which are secreted by the same cells that produce angiogenic factors, are responsible for the breakdown of tissue matrix surrounding the growing vessels and tumor. MMPs are secreted in an inactive proenzyme form and require activation via proteolytic cleavage. The exact mechanism of this activation is unclear; however, it appears that the cell surface proteolytic cascade is mediated at least in part via the urokinase plasminogen activator/ urokinase plasminogen activator receptor (uPA/ uPAR) system (58,59). In addition, a recently identified membrane-type MMP (MT-MMP-1) possesses cell surface proteolytic activity and is involved, in complex with tissue inhibitor of matrix metalloproteinase(TIMP)-2, in the activation of gelatinaseA/MMP-2 on the tumor cell surtace (60,62).

The process of angiogenesis also requires thedirect interaction of endothelial cells with their surrounding matrix. In addition to the activation of MMPs by the uPA/uPAR system mentioned above, there is new information that uPAR may also be involved in the regulation of integrin function. uPAR expressed on the cell surface forms a stable complex with BETA1-integrins (63). This interaction modifies the BETA1-integrin function, promoting adhesion to and migration toward the matrix glycoprotein vitronectin while suppressing the normal adhesion of the integrin to the matrix glycoprotein fibronectin, and also serving as a means by which uPAR may affect intracellular signaling in the absence of a transmembrane domain. Overexpression of uPAR might lead to destabilization of integrin-dependent adhesion with subsequent release of cells from established fibronectin-based matrix, in addition to increased cell surface activation of matrix degrading enzymes, thus enhancing a cell's ability to invade through the ECM. Indeed, data suggest that the increased expression of uPA or uPAR is associated with increased malignant tumor progression and invasiveness (64-67). Brooks et al recently showed that the adhesion receptor integrin alfa v beta 3, capable of interacting with a wide variety of ECM components, is selectively ex-pressed on growing neovasculature but not on quiescent blood vessels (68). In addition, they were able to show that binding of the alfa v beta 3 integrin provided a specific transmembrane signal that enhanced the in vivo survival of angiogenic vascular endothelial cells (69). This signal was associated with a decrease in intracellular p53 activity as well as a reduction in p21WAF/CIP1 and bax expression while increasing bcl2 expression resulting in a sharp increase in the bcl2/bax ratio (70). The consequence of these changes was direct promotion of endothelial cell survival through suppression of the bax cell death pathway. Thus, during angiogenesis, the binding of the endothelial cell alfa v beta 3 integrin to ECM components appears to be required for the suppression of endothelial cell apoptosis facilitating the proliferation and maturation of new blood vessels. More recently, another function of the alfa v beta 3 integrin has been identified. The alfa v beta 3 integrin binds activated. MMP-2 to the cell surface of endothelial cells and invasive melanoma cells in vivo facilitating cell-mediated collagen degradation (71). Thus, it appears that the alfa v beta 3 integrin may function in a cooperative manner with MMP-2 to promote cell adhesion, migration, and controled ECM degradation resulting in directed endothelial and tumor cell invasion.

REGULATORS OF ANGIOGENESIS IN TUMOR SPECIMENS AND PATIENTS

A dramatic upregulation of VEGF mRNA has been found in human glioblastoma tissue but not in normal brain tissue (7). In addition, the VEGF high affinity receptor fit was highly expressed on glioblastoma vasculature and not on normal brain endothelium. There was also a significant upregulation of VEGF mRNA in the tissue of highly vascularized glioblastoma multiforme, capillary hemangioblastomas, and cerebral metastases (72). The in vivo transfection of endothelial cells using a retrovirus encoding a dominant-negative mutant of the Flk-1/VEGF receptor was shown to inhibit VEGF-mediated signal transduction (9). Using this transfection system, aggressive, angiogenic, and VEGF secreting rat C6 glioblastoma tumor formation could be prevented in nude mice by inhibiting VEGF-mediated signaling. More recently, the Flk-1/VEGF receptor was shown to be involved in the growth of a wide range of solid tumors (10). The same in vivo endothelial cell transfection with a dominant-negative mutant of the Flk-I/VEGF inhibited the growth of mammary, ovarian, lung, and glioblas-toma carcinoma cells inoculated subcutaneously or intracerebrally into nude mice (10), It has been postulated that VEGF may be induced by hypoxia (73). The highest levels of VEGF mRNA were expressed by glioblastoma cells closest to necrotic hypoxic areas (73). However, there was a significant upregulation of mRNA for the fibroblast growth factor (FGF) receptor in human glioblastoma tumor cells that was absent in normal brain cells and in endothelial cells from capillaries and large vessels within the tumor (74). Thus, FGF receptor signal transduction may be associated with increased autocrine tumor growth but is probably nor related to the increased endothelial cell proliferation and neovascularization associated with these tumors.

The von Hippel-Lindau (VHL) syndrome is a hereditary condition in which affected individuals are prone to the development of extremely vascular tumors including hemangioblastomas and renal clear cell carcinoma (75). The VHL gene has recently been identified (76,77). Several types of experiments suggest that the key tumor angiogenesis factor associated with VHL disease and VHL gene mutations is VEGF, and that the normal function of the VHL gene may be to suppress VEGF production. Upregulation of VEGF in stromal cells and of the VEGF receptor mRNA occurs in tumor endothelium of VHL-associated and sporadic hemangioblastomas (78). In vitro, renal cancer cells that lack the wild-type VHL gene or express a mutant gene demonstrate overexpression of VEGF mRNA, whereas introduction of the wild-type VHL gene produces a decrease to normal levels (79). In addition, the in vitro stimulation of endothelial cells by conditioned medium of tumor cells expressing the mutant VHL gene is reversed by the neutralization of VEGF.

Angiogenic factors have also been demonstrated in pancreatic cancer. Messenger RNA levels for acidic FGF (aFGF) and bFGF in human pancreatic carcinoma specimens were increased compared with specimens from normal pancreas (80). Sixty percent and 56% of 78 pancreatic carcinomas contained intracellular aFGF and bFGF, respectively. Moreover, there was a correlation between the presence of either factor and advanced tumor stage and between the presence of bFGF and decreased survival. The high affinity type 1 FGF receptor (FGFR-1) was also overexpressed in human pancreatic carcinoma cells (81). This finding suggests that aFGF and bFGF may act as paracrine or autocrine growth factors for pancreatic carci-noma cells in addition to producing angiogenesis.

The expression of bFGF, VEGF, or their recep-tors has been found in association with a variety of other tumors including head and neck squamous cell carcinoma (82), melanoma (83), colon and gastric carcinomas (84-86), breast carcinoma (87,88), lung carcinoma (89,90), hepatocellular carcinoma (91,92), bladder carcinoma (93), and chronic lymphocytic leukemia cells (CLL) (94). In vitro, CLL cells with high intracellular bFGF were more resistant to fludarabine (94). Furthermore, the addition of bFGF in vitro to fludarabine-treated cells resulted in a delay of apoptosis and prolonged leukemic cell survival. This suggests the hypothesis that bFGF could be related to resistance of CLL cells to an apoptotic timulus.

Numerous investigators have evaluated the presence of angiogenic peptides, in particular bFGF and VEGF, in clinical body fluids in an effort to explore the usefulness of these measurements as potential clinical parameters. Most of these studies are exploratory and, although not consistent, suggest there may be a correlation between detectable levels of these peptides and clinical outcome. However, confirmation in prospective adequately sized clinical trials is needed. There was a relatively good correlation between serum but not urinary bFGF levels and tumor stage or grade in a small number of patients with renal cell carcinoma (95). More recently, a significant correlation between urinary bFGF levels and the extent and status of disease in patients with bladder cancer was noted, and that urinary bFGF levels appeared to be more sensitive, but not more specific, than urine cytology for the diagnosis of bladder cancer (96). Looking at urinary bFGF levels in 950 cancer patients, increased levels of urinary bFGF were found in at least some patients with every tumor type examined except cervical carcinoma (97). Increased levels of bFGF correlated positively with extent of disease, the development of recurrent disease, and risk of death. It was not possible to determine whether the bFGF was derived from tumor or normal tissue; however, in a mouse model the increased levels of urinary bFGF originated from tumor (98). Thus, urinary bFGF might in part be tumor-derived. Basic FGF has also been detected in the cerebrospinal fluid (CSF) of 62% of 26 children with brain tumors (45). The presence of bFGF correlated with in vitro CSF-induced endothelial cell DNA synthesis and with the microvessel density of the primary tumor. Basic FGF levels were higher in men with prostate cancer compared to those without in the absence of correlation with clinical stage, Gleason score, or prostate volume.(99). An increased bFGF level (arbitrarily set as >1.0 pg/mL) in patients with a normal prostate-specific antigen (PSA) was associated with a significant risk of cancer (9 of 11 patients had cancer) with a sensitivity in this subset of 83%.

Serum levels of VEGF in patients with a variety of cancers were higher than those in the serum of patients without cancer (100,101). Although measurable levels of VEGF were detectable in normal sera, the VEGF levels in sera from cancer patients were higher (101). Using a cutoff serum VEGF level of 180 pg/mL. increased VEGF levels were present in 9% of 137 primary and 30% of 38 recurrent breast cancer patients. Overall, a lower number of patients with a microvessel density < 100 counts/mm3 compared with >100 counts/mm3 had increased VFGF levels (3% versus 19%, respectively). Patients with increased intratumoral VEGF appeared to have increased serum level (101). Thus, monitoring the levels of bFGF, VEGF, and possibly other angiogenic peptides, in the urine, serum, or CSF of patients with cancer may prove to be a useful marker for tumor status and perhaps prognosis.

An increased presence of MMPs may be related to increased invasive, metastatic, and angiogenic potential of tumors. The upregulation of MMP-2, MMP-9, and stromolysin-3 mRNA has been detected to a greater degree in tissue specimens of breast carcinoma than in normal breast tissue (102-105). MMP-2 mRNA was more frequently expressed in stromal tissue than in tumor cells while TIMP-1 mRNA was more frequently detected in epithelial cells, well differentiated, and low grade noninvasive tumor cells (103). These data suggest that the control of matrix degradation may be related to increased expression of MMPs and the altered expression of endogenous TIMPs (103). Another role for the TIMPs was recently suggested. The activation of MMP-2 by MT-MMP-1 was shown to require the binding of TIMP-2 (62). Thus, TIMPs may play a dual role in the regulation of MMP activation. In malignant gliomas, an increase in both the mRNA expression and/or the levels of the activated enzymes, for MMP-2, MMP-9 and MT-MMP, as well as a decrease in the tissue levels of both TIMP-1 and TIMP-2 have been reported (106-109). Increased MMP-2 and MMP-9 mRNA in addition to an increase in the activity of these enzymes has also been found in colorectal carcinoma tissue. (110-112) Other tumors require evaluation to determine the incidence of altered expression of the MMPs and TIMPs and whether alterations are associated with patient prognosis.

Investigators are now beginning to report circulating plasma levels of MMPs. Increased plasma levels of MMP-2, MMP-2/TIMP-2 complexes, or MMP-9 have been detected in patients with gastrointestinal malignancies (colon, rectal, and gastric cancers), breast carcinoma, and gynecologic cancer (ovarian and uterine cancer) (110-113). Patients who had complete resection of bladder cancer accompanied by muscular invasion or lymph node metastasis with high MMP-2/TIMP-2 ratios had higher rates of recurrence and decreased disease-free survival than patients with low ratios (114). Similarly, patients with recurrent disease had a higher ratio than those without recurrence. There was also a positive correlation between an increased MMP 2/TIMP-2 ratio and the presence of lymph node metastasis.

There was no correlation between either MMP-2 or TIMP-2 levels alone. Again, prospective studies with adequate numbers are required to determine whether MMP levels or MMP/TIMP ratios might be prognostic.

THERAPEUTIC AGENTS

Therapeutic strategies aimed at inhibiting various steps in the process of angiogenesis have significant clinical potential. Most of the antiangiogenic compounds currently in clinical trials interfere with the response of endothelial cells to angiogenic peptides (Table 2). However, agents that inhibit the action of MMPs (MMPIs) are also in clinical trials. Additional strategies aimed at inhibiting tumor neovascularization and targeted therapies aimed at destroying the tumor neovasculature directly are under development. Therapies affecting an end target or pathway that cannot be circumvented by alternate mechanisms may significantly enhance efficacy and broaden applicability. Furthermore, preclinical studies have shown that the combination of an antiangiogenic agent with cytotoxic chemotherapy may significantly increase the activity of the cytocoxic agent (115-120). In addition to arresting the in vivo growth of primary tumors and metastases, these combinations reduced the size of established tumors and rendered some mice tumor-free. This combination approach has produced such benefit for several classes of agents including angiostatic steroids, MMPIs, and agents such as TNP-470 that inhibit, endothelial cell response to angiogenic factors. In addition, in antiangiogenesis agents may also potentiate the effects of radiation therapy (121). Thus, antiangiogenic agents may prove clinically useful when used alone as adjuvant therapy or in the setting of small volume disease, and when combined with cytotoxic therapies in patients with advanced or metastatic disease.

The first antiangiogenic agent to enter the clinic was the sulfated polysaccharide xylanopolyhydro-gensulfate (pentosan polysulfate). In vitro, pentosan polysulfate inhibited bFGF-induced endothelial cell migration and proliferation (122,123); growth of an adrenal cancer cell line transfected with an FGF gene inducing constitutive bFGF production (124,125); and inhibited the paracrine effects of heparin-binding growth factors secreted by a variety of malignant tumor cell lines (126). In an initial phase I clinical trial of the administration of pentosan polysulfate via continuous intravenous infusion and subcutaneous injection to patients with acquired immunodeficiency syndrome (AIDS)-associated Kaposi's sarcoma, the agent proved toxic producing anticoagulation, thrombocytopenia, and increased transaminase levels without evident clinical activity (127). Similar results were obtained when pentosan polysulfate was administered subcutaneously to patients with solid tumors (128).

In 1990, Ingber et al reported the antiangiogenic properties of fumagillin, a secreted antibiotic product of the fungus Aspergillus fumigatus fresenius that was originally isolated in 1949 (15,129). Fumagillin produced excessive toxicity; therefore, analogues of fumagillin were synthesized. TNP-470 (AGM-1470) is a more potent, less-toxic fumagillin analog.15 TNP-470 inhibited in vitro proliferation and migration of normal, but not transformed, endothelial cells and also inhibited in vitro capillary tube formation at concentrations that were cytostatic but not cytotoxic (15,130-133). Antiangiogenic activity was shown in vivo using a variety of assays, including the chick chorioallantoic membrane (CAM) assay, the rat corneal micropocket assay, and the mouse sponge implantation assay (130). TNP-470 inhibited in vivo growth of a variety of human xenografts and murine tumors in the absence of direct in vitro growth inhibition of the same tumor cell lines(15,134-138). It also suppressed metastasis in both human xenograft and murine tumor models (135,136,139). TNP-470 is currently undergoing phase I testing in Kaposi's sarcoma and other tumors, and early phase II clinical trials in central nervous system (CNS) and solid tumors are beginning. Preliminary results indicate that the agent is well tolerated with CNS toxicity manifested as reversible cerebellar symptoms being dose-limiting (140-142).

Another naturally occurring agent with potent antiangiogenic activity is platelet factor-4 (PF4), a compound stored in the alpha-granules of platelets. PF4 inhibited both endothelial cell migration and proliferation at cytostatic but not cytotoxic concentrations (16,143,144). It also inhibited mitogenesis induced in vitro by the physical wounding of a Swiss 3T3 cell culture monolayer (145). PF4's antiangiogenic affect is thought to occur through its binding to glycosaminoglycans on the cell surface and subsequently blocking binding of bFGF to its receptor (144,145). PF4 has also been shown to inhibit the in vivo growth of human xenograft and murine tumors when administered intralesionally (143,146). Currently, phase I trials of both systemic and local I ad ministration of PF4 are ongoing in patients with solid tumors and AIDS-associated Kaposi's sarcoma, and a phase II study of intratumoral administration of PF4 in patients with primary brain tumors recently began. Preliminary results indicate the drug is well tolerated with mild toxicities manifested as a local reaction at sites of intralesional injection, mild phlebitis, fatigue, and ane-mia (14-150).

A sulfated polysaccharide peptidoglycan complex (SP-PG, DS-4152, tecogalan) derived from an Arthrobacter sp bacterial cell wall polysaccharide inhibited in vitro endothelial cell growth, particularly when combined with tamoxifen (151,152). In vivo antiangiogenic activity that was enhanced by cortisone acetate or tetrahydro S was shown in the CAM assay (153,154). In addition, tecogalan, combined with tetrahydro S, inhibited angiogenesis induced by the ovarian ascites tumor M5076 in vivo in the murine dorsal airsac assay (151). Tecogalan also has been reported to possess in vivo antitumor activity against human xenografts and murine tumors when co-administered with either cortisone acetate or tetrahydro S (151-153). This agent has recently under-gone phase I clinical testing in patients with solid rumors (155). The dose-limiting toxicity was anticoagulation manifested by an increased activated partial thromboplastin time. Other toxicities included fever and rigors.

Thalidomide, which was originally marketed as a sedative hypnotic in the 1950s, produced severe birth defects manifested by hypoplastic and aplastic malformations of the extremeties when ingested within the first 2 months of pregnancy during limb bud formation. In addition to its known immunomodulatory properties, it is now known that the embryopathy was probably due to its antiangiogenic activity (156). In vitro thalidomide alone did not affect bFGF-induced endothelial cell proliferation. In vivo it did not have antiangiogenic activity in the CAM assay (156). However, after oral administeration to rabbits, the agent did exhibit potent antiangiogenic activity in the corneal micropocket assay. After evaluating a number of thalidomide analogues, D'Amato et al concluded that a hepatically generated epoxide metabolite of thalidomide was responsible for its antiangiogenic activity but not for its immunomodulatory properties (156). Based on these findings, clinical trials administering thalidomide to patients with AIDS-associated Kaposi's sarcoma, breast cancer, prostate cancer, and primary brain tumors are currently underway.

Agents that inhibit the activity of MMPs have also entered into clinical trials recently. Batimastat (BB-94) has broad inhibitory activity in vitro against a variety of MMPs including MMP-2, MMP-3, and MMP-9, but shows no direct effect on the in vitro growth of various tumor and fibroblast cell lines (157,158). However, it inhibits the growth and metastatic spread of human xenografts and mouse tumors (158-163). Phase I trials administering batimastat intraperitoneally and intrapleurally are being performed, and phase II trials are in development.

Intraperitoneal, intrapleural, or intravenous administration is impractical for agents that require frequent dosing and/or long-term administration and attempts are underway to develop orally available agents. An orally bioavailable MMPI, marimastat (BB2516), is now undergoing clinical evaluation. Marimastat possess in vitro MMP inhibitory activity similar to that of batimastat. In phase I trials of marimastat in pancreatic, ovarian, and prostate carcinoma patients, the drug was well tolerated, the main toxicity noted being joint and muscle pain and stiffness (164). Of note in preclinical studies, marmosets administered high doses of marimastat developed fibrosis and inflammation around the knee and ankle joints and muscle necrosis (164). Approximately 50% of patients had a decrease in the rate of increase of a tumor associated serum tumor marker (CA 19-9, PSA, and CA 125) (165-167). Phase III trials of marimastat in patients with various cancers have been initiated.

A group B streptococcus polysaccharide toxin (CM101) that is responsible for pulmonary disease in infected human neonates binds preferentially to the capillary endothelium of a variety of carcinomas but not to normal, mature endothelium (168). CM101 treatment of mice implanted with human or murine tumors resulted in an intense intratumoral inflammatory reaction associated with necrosis hemorrhage, thrombosis, and the release of large amounts of cytokines including tumor necrosis factor-alfa (TNF-alfa), interleukin-l alfa (IL-l alfa), IL-6, and macrophage inhibitory protein-1 (MIP-1) (168,169). Furthermore, CM1Q1 reduced tumor volume and prolonged survival, with mice observed for more than 5 months tumor free (168,170). A phase I trial of CM101 is being performed by DeVore et al (171). Preliminary results suggest the drug is well tolerated. Objective responses were noted in three of 15 patients with classical Kaposi's sarcoma, hepatocellular carcinoma, and metastatic small bowel adenocarcinoma (171). CM101 also increased systemic TNF-alfa, MIP-1alfa, IL-6, IL-8, and IL-10 levels (172). Soluble E-selectin levels suggesting endothelial engagement were also increased after CM101 administration. These findings suggest that CM101 acts by binding to tumor-associated neovasculature with the subsequent induction of vascular and cellular inflammatory reactions with the tumor (172). The National Cancer Institute in collaboration with CarboMed. Inc, plans phase II studies of CM101 in patients with Kaposi's sarcoma and other tumor types.

Finally, IL-12 is a cytokine that enhances proliferation of activated T and natural killer (NK) cells; enhances cytotoxic T- and NK-cell activity; and induces interferon-gamma (IFN-gamma) production (173). IL-12 enhances cell-mediated immunity by inducing differentiation of T-helper type 1 cells (Th1 cells) from uncommitted T cells while inhibiting T-helper type 2 cell (Th2) differentiation (174). This effect on T helper differentiation by IL-12 restores human immunodeficiency virus (HIV)-specific cell-mediated immune responses in HIV-infected T-cells (175). IL-12 also has antitumor and amimetastatic activity in preclinical models (176). Voest et al reported that IL-12 likewise possesses potent and angiogenic activity mediated by induction of IFN-gamma (177). A chemokine induced by IFN-gamma, human interferon-inducible protein 10 (IP-10), is a potent inhibitor of in vivo angiogenesis (178,179). Selective expression of IP-10 and IFN-gamma mRNA in inflammatory tumor tissue is found in vivo after administration of IL-12 to tumored mice (180). Blocking the effects of either IFN-gamma or IP-10 in vivo resulted in inhibition of the antiangiogenic effect of IL-12.181 Therefore, it appears that the antiangiogenic activity of IL-12 may be through its induction of local ifn- gamma production with subsequent upregulation of IP-10. Thus, IL-12 warrants clinical evaluation, not only for its direct antitumor effects, but also for its profound abilities to inhibit angiogenesis. Phase I and II clinical trials in patients with cancer and AIDS-associated Kaposi's sarcoma are now under way.

A common property possessed by most of the compounds discussed is their specificity for one or another portion of the angiogenesis pathway, or for established tumor neovasculature (CM101). They are not cytotoxic in vitro for either tumor or normal cells at concentrations that are biologically effective in vivo. Recently, a variety of compounds referred to as antiangiogenesis agents were found to produce antiangiogenic activity and direct in vitro cytotoxicity to tumor as well as a wide variety of normal cells. These compounds may have nonspecific cytotoxic or inhibitory properties for endothelial cells and are not true selective antiangiogenic agents. As newer agents are discovered and enter the clinic, the distinction, between true andangiogenesis inhibitors or antivascular agents and agents possessing nonspecific antiangiogenic activity needs to be maintained.

DEVELOPMENT STRATEGIES

For the most part, current antiangiogenic agents are cytostatic in nature, preventing tumor growth rather that inducing reduction in tumor size. Consequently, the usual paradigms for anticancer drug development where tumor responses in phase II trials prompt further development are not appropriate. In addition, antiangiogenic therapy may be appropriately administered in several different clinical situations: the adjuvant setting; to patients with high risk for relapse or recurrence after control of primary disease; as maintenance therapy to those with advanced, metastatic, or recurrent disease in whom tumor reduction is achieved by standard chemotherapy, radiation therapy, or surgery; in combination with chemothrapy or radiation therapy; perhaps even as a chemopreventive agent to patients at high risk for developing malignancies. To be suitable for long-term chronic uses, these agents should possess little acute or chronic toxicity. They should also be easily administered with oral administration preferred to intravenous. Oral or depot agents with long half-lives would require less frequent administration.

Establishing the maximally tolerated dose in phase 1 may not be appropriate. Rather, the determination of a biologically active dose which may possess less toxicity may be more relevant as the administration of treatment at a dose close to the maximally tolerated dose may not be as well tolerated on a chronic basis, especially if biological effects occur at lower doses. Unfortunately, the determination of clinical in vivo biological activity is complex. Validated surrogate markers for the biological activity of most of these agents are not available. The choice of phase I trial designs and appropriate end points may need, at least for now, to be guided by the mechanism of action of the agent under investigation. For example, if an agent inhibits MMP activity, perhaps measurement of circulating active MMP levels might be appropriate. Another potential approach would be to collect serum or plasma from patients who receive the investigational agent and perform in vitro tests assessing antiangiogenic activity in the systems where such activity has already been shown. In addition, in some early studies performed with the biologically active dose, serial tumor biopsies should be evaluated for determination of angiogenic index, the presence of active MMPs or other factors, or other pertinent biologic effects of the agent under evaluation. Other novel approaches for early evaluation of these agents may prove helpful. These studies may require more extensive laboratory and biologic evaluation in the phase I setting to ensure that the appropriate dose and schedule of an agent is chosen for further study.

New designs for trials to demonstrate the "activity" of such drugs are also required. It is still unclear whether some type of activity (but not objective response) should be demonstrated in phase-II-sized trials. The design of such trials would require careful consideration with respect to end points. Although controlled studies might be preferred, smaller studies, if properly designed with appropriate endpoints, could help prioritize the many agents in development. For example, because most of these agents are expected to be cytostatic, it is inappropriate to require objective response. However, the use of stable disease as an end point is also problematic. In the absence of studies designed to confirm tumor growth rare before initiation of treatment, observation of stable disease could be misleading. Even with study designs showing indisputable reductions in tumor growth, follow-up would be required to determine duration of tumor control. Because of the heterogeneity inherent in cancer patients even with the same tumor type, the use of historical controls is Inappropriate. Smaller, randomized trials in specific disease settings may prove useful because a standard treatment group would be included. If the characteristics of patients in future trials in the same disease settings are comparable, the data from this standard group may then be used as a control. The availability of markers proven to be appropriate surrogates for clinical activity would provide an important tool that could simplify and accelerate early development.

Randomized trials will ultimately be necessary to conclusively show the activity of these agents. The tumor types and clinical end points must be carefully chosen. Some studies should be performed in disease settings where the clinical end point would be reached in a relatively short period. Such studies could easily be performed in diseases where, despite chemotherapy-induced responses, survival with current therapy is of short duration such as extensive small cell lung cancer or metastatic breast cancer. Ultimately, randomized studies administering these agents to patients with high risk of recurrence in the adjuvant setting will be required. These trials will require large patient numbers, several years to accrue, and adequate follow-up.

Table 1. Endogenous Angiogenic Polypeptides

Fibroblast growth factor /basic (bFGF) and acidic (aFGF)/

Angiogenin

Transforming growth factor-alfa (TGF-alfa)

Transforming growth factor-beta (TGF-beta)

Tumor necrosis factor-alfa (TNF-alfa)

Vascular endothelial growth factor/

Vascular permeability factor (VEGF/VPF)

Platelet-derived endothelial cell growth factor (PD-ECGF)

Granulocyte-colony-stimulating factor (G-CSF)

Placental growth factor lnterleukin-8 (IL-8)

Hepatocyte growth factor/ scatter factor (HGF/SF)

Pleiotrophin (PTN)

Table2. Ahtiangiogenic/Antivascular Agents

Agents in clinical trials

Pentosan polysulfate

TNP- 470(AGM-1470)

Tecogalan (DS41S2.SP-PG)

Platelet factor 4 (PF4)

Thalidomide

Batimastat(BB-94)

Marirmastat(BB-2516)

CM101

lnterleukin l2(IL-12)

CT2584 Agents in preclinical development or about to enter clinical trials

Vitaxin

Meastat (Col 3)

Angiostatin

AG3340