To investigate the biological consequences of endothelial cell activation by endogenous FGF2, immortalized Balb/c mouse aortic endothelial cells (MAE cells) and brain microvascular cells (MBE cells) were transfected with a retroviral expression vector harboring a human FGF2 cDNA (Gualandris et al., 1996a; Gualandris et al., 1996b). FGF2 transfectants express all FGF2 isoforms and are characterized by a transformed morphology and an increased saturation density. FGF2 transfectants show invasive and morphogenetic behavior in three-dimensional gels which is prevented by anti-FGF2 antibody, revealing the autocrine modality of the process (Gualandris et al., 1996a).

The biological consequences of this autocrine activation were investigated in vivo. FGF2-transfected MAE cells induce the growth of highly vascularized tumors.

Tumorigenic activity of FGF2-transfected and parental MAE cells in nude mice. Male Swiss nu/nu mice were inoculated s.c. with 0.510⁶ (squares), 1.010⁶ (triangles), or 2.0 10⁶ (circles) parental MAE cells (open symbols) or pZipbFGF2-MAE cells (closed symbols). The size of the lesion was measured with calipers.

Lesions were observed also when cells were injected in x-irradiated syngeneic mice but grew poorly in immunocompetent syngeneic animals, indicating that the growth of these lesions is dependent on the immunological status of the host. Histologically, the tumors have the appearance of hemangioendothelioma with spindled areas resembling KS (Gualandris et al., 1996a) and with numerous CD31-positive blood vessels and lacunae. Southern blot analysis revealed that less than 10% of the cells in the tumor mass were transplanted FGF2-transfected MAE cells. Accordingly, disaggregation of the lesion and in vitro cell culture demonstrate that less than 10% of total cell population retain FGF2 overexpression and neomycin-resistance. These data indicate that FGF2-overexpressing endothelial cells cause vascular lesions in the immunocompromised host that are sustained to a large extent by recruitment of host cells, including endothelial cells.

In agreement with these observations, FGF2-transfected MAE cells induce an angiogenic response when implanted in the avascular rabbit cornea (Gualandris et al., 1996a). Also, they cause an increase in vascular density (Gualandris et al., 1996a) and formation of hemangiomas in the chorioallantoic membrane when injected into the allantoic sac of the chick embryo.

The above data demonstrate that pZipbFGF2-MAE cells induce highly vascularized spindle-cell hemangioendotheliomas in immunodeficient mice that are sustained by recruitment of host elements, including endothelial cells. The sulfonated distamycin derivative PNU 153429 is a non cytotoxic compound endowed with antiangiogenic and antimetastatic activity and of FGF2-binding capacity (40, 45).

Since vascular tumors may represent an interesting model for the evaluation of angiostatic therapies, we evaluated the effect of PNU 153429 on tumor growth induced by s.c. injection of 0.510⁶ pZipbFGF2-MAE cells into nude mice. PNU 153429 administered i.p. once a week at the dose of 100 mg/kg causes a significant delay in tumor take, resulting in inhibition of tumor growth. To assess whether the inhibitory effect exerted by PNU 153429 on vascular lesions was due, at least in part, to its angiostatic activity, microvessel density was measured in tissue sections immunostained with anti-CD31 antibody to highlight newly formed blood vessels. For each tumor section the most vascularized area was selected and CD31-positive microvessels in a 160-field were counted. The results demonstrate that PNU 153429 causes 50% decrease (p < 0.01) in CD31-positive microvessel density of pZipbFGF2-MAE cell-induced lesions.

Inhibitory activity of PNU 153429. A) Swiss nu/nu mice (10 animals per group) were inoculated s.c. with 0.510⁶ pZipbFGF2-MAE cells. After 24 h, animals were treated i.p. with PNU 153429 at 100 mg/kg (closed circles) or with vehicle (open circles). Mice were monitored three times a week and the size of the lesion was measured with calipers. Treatment was repeated once a week for 35 days and animals were sacrificed 3 days after the last treatment. B) Density of CD31-positive microvessels was then evaluated on tumor biopsies obtained from control and treated animals as described in "Materials and Methods". Data are expressed as mean SEM.

In conclusion, the sulfonated distamycin A derivative PNU 153429 inhibits the growth and vascularization of pZipbFGF2-MAE cell-induced lesions. Previous observations had shown that the growth of vascular tumors induced by PmT-transformed endothelial cells (eEnd.1 cells) is susceptible to the inhibitory activity of the matrix metalloprotease inhibitor batimastat (Taraboletti et al., 1995) and that batimastat is able to inhibit in vitro endothelial cell chemoinvasion and in vivo neovascularization induced by eEnd.1 cell supernatant (Taraboletti et al., 1995). Interferon a-2a, that inhibits angiogenesis in mice, induces regression of hemangiomas of infancy in humans (Ezekowitz et al., 1992). Taken together, the data indicate that vascular tumors are suitable targets for angiostatic therapy.

The combination of angiogenesis inhibitors have been hypothesized to be useful in tumor treatment (Brem et al., 1993). Also, anti-angiogenic agents interact in a positive manner with cytotoxic therapies (Teicher et al., 1996). Relevant to this point, preliminary observations have shown that doxorubicin (four i.v. injections at 6 mg/kg given at 4-day interval starting at day 14) exerts an inhibitory activity on the growth of pZipbFGF2-MAE cell-induced lesions similar to that exerted by PNU 153429 causing however only a limited decrease in microvessel density (F. Sola, M.P. Molinari Tosatti, and M. Presta, unpublished observations).

This suggests that the two molecules may exert their inhibitory activity on pZipbFGF2-MAE cell-induced lesions via different mechanisms of action. The opportunistic vascular tumors caused by pZipbFGF2-MAE cells may therefore represent an useful model for the study of angiostatic compounds in single agent or combination therapy of vascular tumors, including KS.

The use of purine analog antimetabolites in combination chemotherapy of solid tumors has been proposed. To assess the possibility that selected purine analogs may affect tumor neovascularization, 6-methylmercaptopurine riboside (6-MMPR), 6-methylmercaptopurine, 2-aminopurine, and adenosine were evaluated for the capacity to inhibit angiogenesis in vitro and in vivo.

6-MMPR inhibited FGF2-induced proliferation and delayed the repair of mechanically wounded monolayer in endothelial GM 7373 cell cultures. 6-MMPR also inhibited the formation of solid sprouts within fibrin gel by FGF2-treated murine brain microvascular endothelial cells and the formation of capillary-like structures on Matrigel by murine aortic endothelial cells transfected with FGF2 cDNA. 6-MMPR affected FGF2-induced intracellular signaling in murine aortic endothelial cells by inhibiting the phosphorylation of extracellular signal-regulated kinase-2. The other molecules were ineffective in all the assays.

In vivo, 6-MMPR inhibited vascularization in the chick embryo chorioallantoic membrane (CAM) and prevented blood vessel formation induced by human endometrial adenocarcinoma specimens grafted onto the CAM. Also, topical administration of 6-MMPR caused the regression of newly formed blood vessels in the rabbit cornea. Thus, 6-MMPR specifically inhibits both the early and the late phases of the angiogenesis process in vitro and exerts a potent anti-angiogenic activity in vivo. These results suggest a new rationale for the use of selected purine analogs in combination therapy of solid cancer.

Effect of 6-MMPR on newly formed blood vessel maintenance in the rabbit cornea. Slow-release pellets containing 100 ng of FGF2 (, ) or vehicle () were implanted in the cornea of New Zealand white rabbits (4 animals per group). At day 14, animals underwent topic treatment with 6-MMPR (, ) or vehicle () given as ophthalmic ointment. Daily observations of the implants were made with a slit lamp stereomicroscope without anesthesia and data were expressed as angiogenesis score, calculated as vessel density x distance from limbus in mm. No effect on FGF2-induced neovascularization was observed when animals were given 6-MMP under the same experimental conditions.