Eradication of Rat Malignant Gliomas by Retroviral-mediated, in Vivo Delivery of the Interleukin 4 Gene¹

The growth of malignant gliomas takes advantage of an immunosuppressive environment (1). Glioma patients are often characterized by a broad depression of both humoral and cell-mediated immunities, and particularly of T-cell functions (2, 3). These observations encourage the testing of immunotherapy approaches that may up-regulate the immune system and create an immunostimulatory tumor microenvironment.

The identification of the brain as a site of immunological privilege could be somehow contradictory with such approaches. However, the definition of this privilege has been challenged recently by the evidence that tight connections exist between the CNS⁴ and the immune system through cervical lymphatics and the blood-brain barrier (4). It has been demonstrated that both the afferent arm (i.e., the presentation of CNS-derived antigens to the immune system and the activation of lymphocytes in the lymphoid organs) and the efferent arm (the migration of activated lymphocytes back to the CNS) of the immune system are functional and that a selective admission of lymphocytes into the CNS can take place through the expression of specific adhesion molecules (4, 5). A redefinition of the CNS immune privilege should now include selective immunoregulatory events that down-regulate cell-mediated immunity and enhance noninflammatory humoral responses (4, 6).

Although infiltrating mononuclear cells can be found in malignant gliomas, their function can be impaired by tumor-derived immunosuppressive factors like TGF-β (7), IL-10 (8), and Fas ligand (9). Strategies that may overcome immunosuppression should therefore be considered for improving the treatment of gliomas. In this context, we decided to test the therapeutic potential of IL-4 gene transfer.

IL-4 is a cytokine with a strong antitumor effect that involves, at the early stages, nonspecific effector cells, such as eosinophils and macrophages and, later on, T-helper lymphocytes that stimulate a humoral response (10, 11). In mixed tumor transplantation assays, the growth of U-87 glioblastomas was inhibited if J588L plasmocytoma cells engineered to express high levels of IL-4 were coinjected into nude mice (12). The intracranial and s.c. growth of the more malignant C6 glioblastoma could also be limited in nude mice by retroviral and adenoviral-mediated transfer of the IL-4 gene (13, 14). Nevertheless, all IL-4-treated mice injected intracranially were dead by day 55 after tumor injection, likely because of the lack of the late T-cell response. When glioma cells were transfected in vitro by the IL-4 gene and injected into the brain of immunocompetent rats or mice, ∼70% of the animals survived for at least 70 days (15, 16). In these experiments, however, no gene transfer was taking place in vivo, and all tumor cells were overexpressing IL-4 at the moment of the injection, two factors that decrease the significance of these observations in view of the definition of gene therapy strategies for malignant gliomas.

In initial experiments, we found that coinjection of glioblastoma and RPCs transducing IL-4 into rat brain is followed by survival after 3 months (long-term survival) of a relatively low percentage of rats (23%; Ref. 17). Furthermore, data based on IL-4 transduction in vivo by defective herpes simplex viruses showed 0% survival at the same time point (18). In this report, we show that injection of RPCs producing increasing amounts of IL-4 raises strikingly the fraction of glioma survivors. Our results also suggest that high levels of IL-4 during the initial phases of tumor development are critical for tumor rejection and that CD8+ T-lymphocytes are important mediators of the antitumor function of IL-4.

C6 cells were obtained from the American Type Culture Collection and were cultured as suggested by this supplier. 9L cells (a kind gift of Dr. K. Plate, Freiburg University, Department of Neuropathology, Freiburg, Germany) and the mouse T-cell line HT-2 were cultured in RPMI supplemented by 10% FCS.

The RPC line PA317.STK.SBA (here defined as SBA), expressing the HSV-tk gene, was obtained and characterized previously (19). The RPC line ψ2.L4SN was also described previously (11). This and other RPC cell lines were named according to the amount of IL-4 that they produced. ψ2.L4SN cells produced 20 ng of IL-4/5 × 10⁵ cells/48 h and were named ψ2.L4SN²⁰. To prepare E86.L4SN RPCs, the murine IL-4 cDNA was cloned in the EcoRI site of a pLXSN plasmid and transfected by lipofection (DOTAP; Boehringer Mannheim) in GP+E86 ecotropic packaging cells (obtained from Dr. A. Bank, Columbia University, New York, NY). After G418 selection (0.5 mg/ml of active drug), several clones were isolated and tested for retroviral titer as described previously (19) and for IL-4 production by an ELISA-based assay (Genzyme). Two clones were used in this study. Clone 6 has a titer of 3 × 10⁷ cfu/ml, produces 50 ng of IL-4/5 × 10⁵ cells/48 h, and was named E86.L4SN⁵⁰. Clone 31 had a titer of 9 × 10⁷ cfu/ml and produced 200 ng of IL-4/5 × 10⁵ cells/48 h and was therefore named E86.L4SN²⁰⁰.

For growth rate analysis, 10⁴ 9L or C6 cells were plated in 96-well plates in the presence of different dilutions of conditioned medium from 10⁶ wild-type or IL-4-transduced NIH-3T3 cells (secreting 150 ng of IL-4/48 h). Cell proliferation was measured after 24, 72, and 96 h by WST-1 colorimetric assay (Boehringer Mannheim), according to the manufacturer’s instructions. Statistical analysis was performed by paired t test.

IL-4 receptor expression was evaluated by reverse transcription-PCR. cDNA synthesized using random hexamers (Boehringer Mannheim) was amplified with primers IL-4R-1F 5′-TGCTTCTCTGACTACATCCGC-3′ and IL-4R-1R 5′-TGGAGTGTGAGGTTGTCTGG-3′ with 30 cycles of PCR consisting of 30 s at 94°, 30 s at 58°, and 30 s at 72°.

Western blot was performed loading 10 μg of proteins on a 8% acrylamide gel and labeling with polyclonal rabbit anti-mouse and rat interleukin-4 receptor α chain (Research Diagnostics, Inc.; 1:500). The ECL chemoluminescence kit (Amersham Life Science) was used for the detection.

Immunohistochemical analysis for the IL-4 receptor was performed on frozen and Carnoy-fixed sections of rat brains with C6 and 9L tumors, rat spleen, or mouse HT-2 cells using the same polyclonal antibody, 1:100 dilution.

Packaging cells were labeled with 10 μm BrdUrd (Sigma) for 48 h. BrdUrd staining was performed on Carnoy-fixed brains with anti-BrdUrd antibody (clone B44; Becton Dickinson; 1:100).

Protein sequences (for IL-4: rat Swiss Prot P20096, mouse P07750, and human P05112; for IL-4R: rat Gene Bank 56391, mouse Swiss Prot P16382, and human Swiss Prot P24394) were aligned using the GeneWorks 2.5.1 software.

Sprague Dawley or Fischer 344 rats (females, 225–250 g or 160–180 g, respectively; Charles River Italia) were anesthetized before intracerebral implantation using 4% chloralium hydrate (1 ml/100 g) after a brief exposure to 1% fluothane or ether. In coinjection experiments, animals were placed in a stereotactic frame, and 4 × 10⁴ C6 or 6 × 10⁴ 9L cells mixed or not with SBA or ψ2.L4SN²⁰ or E86.L4SN cells (1:1 ratio for most experiments, 10:1 ratio in one case) in 2–3 μl of PBS were inoculated into the left striatum with a Hamilton syringe for 4 min (coordinates for C6, with respect to bregma: 1 mm anterior, 3 mm left lateral, 5 mm depth; coordinates for 9L: 1 mm posterior, 2.5 mm left, 6 mm depth). Bone wax was used to seal the burr-hole and avoid the spreading of neoplastic cells outside the needle track. In experiments of intratumoral RPC injection, 4–6 × 10⁴ glioma cells were implanted as described above. Four or 7 days later, 1.6 × 10⁶ E86.L4SN cells in 8 μl of PBS were inoculated using the same stereotactic coordinates but at three different depths (4.5, 5.0, and 5.5 mm for Sprague Dawley and 5.5, 6.0, and 6.5 mm for Fischer rats). For tumor challenge experiments, rats treated with IL-4 and surviving over 3 months were injected into the right hemisphere with 4 × 10⁴ C6 or 6 × 10⁴ 9L cells.

To measure the levels of murine IL-4 secreted in vivo, brains from rats injected with 9L cells mixed with SBA or E86.L4SN²⁰⁰ cells and sacrificed after 1, 2, and 3 weeks were homogenized on ice in 3 ml of PBS in a glass potter. The suspension was then centrifuged 10 min at 2000 rpm, 4°C, and the supernatants were stored at −80°C. IL-4 levels in the supernatants were measured by an ELISA assay (Biosource), and the amounts of total proteins were determined by Lowry method.

MRI was performed using a 1.5 T equipment with a 10-cm surface coil. Coronal sequences (Vol 2 mm-T1-FFE, Vol 2 mm T2-TSE) were obtained using the following parameters: the number signal average, matrix, and field of view were, respectively, 4, 256 × 220, and 220 for T1 sequences and 6, 512 × 52, and 110 for T2 sequences. Contrast medium (0.3 ml of gadolinium diethylenetriaminepentaacetic acid) was injected i.p.

For histological analysis, rat brains were either frozen in isopentane or fixed in Carnoy and embedded in Paraplast. In both cases, sections were routinely stained with H&E. To identify subsets of inflammatory cells, immunohistological evaluations were performed on frozen or fixed sections using monoclonal antibodies directed against rat macrophages (antibody ED2; Biosource; 1:250 dilution), CD4 (antibody W3/25; Biosource; 1:100 dilution), CD8 (antibody MRC OX-8; Biosource; 1:100 dilution), B cells (antibody anti-CD45RA, clone OX-33; PharMingen, 1:100 dilution), and NK cells (antibody anti-NKR-P1, clone 3.2.3; Endogen; 1:100 dilution).

Statistical analysis was performed by the log-rank (Mantel-Cox) test in Kaplan-Meyer nonparametric analysis for survival, using the StatView 4.5 software.

To assess the therapeutic potential of retroviral-mediated transfer of the IL-4 gene, we inserted the murine IL-4 cDNA in plasmid pLXSN and transfected GP+E86 packaging cells with this construct. Several clones were obtained from this transfection, and for the initial experiments, we used clone E86.L4SN⁵⁰. This clone has a titer of 3 × 10⁷ cfu/ml and secretes 50 ng IL-4/5 × 10⁵ cells/48 h. The titer was evaluated by transduction of NIH-3T3 cells, and transduced cells produced 150 ng of IL-4/1 × 10⁶ cells/48 h.

We compared the efficacy of E86.L4SN⁵⁰ cells with that of a previously tested, less efficient IL-4 RPC, producing 5 × 10⁵ cfu/ml and secreting 20 ng of IL-4/5 × 10⁵ cells/48 h (ψ2.L4SN²⁰ cells). Coinjection of C6 and ψ2.L4SN²⁰ cells in the brain of Sprague Dawley rats (1:1 ratio; 4 × 10⁴ cells for each cell type) increased rat survival significantly (P = 0.0334 versus controls, log-rank test; Fig. 1), and long-term survivors (i.e., animals surviving >90 days) were 27%. The survival of rats inoculated with C6+E86.L4SN⁵⁰cells, however, was higher, and 56% of these rats remained alive more than 3 months (P = 0.0003 versus controls; Fig. 1). Raising to 10:1 the ratio of C6 to E86.L4SN⁵⁰ injected cells did not decrease the fraction of long-term survivors (data not shown). Furthermore, the injection of 1.6 × 10⁶ E86.L4SN⁵⁰ cells into established C6 glioblastomas (i.e., tumors formed by the inoculation of 4 × 10⁴ C6 cells 1 week before) was followed by tumor rejection in 40% of the animals (P = 0.0207 versus controls; Fig. 1).

A fraction of these animals was examined by MRI and had radiological evidence for complete tumor regression. One example is shown in Fig. 2,A. MRI allowed a close follow-up of tumor growth in the presence or absence of IL-4 RPC. In general, tumor growth was detected during the first 3–4 weeks after tumor graft and was followed by the death of the animals in untreated controls or by tumor regression in many IL-4-treated rats. To characterize the immune effector cells involved in tumor rejection, some animals were sacrificed during the regression phase, usually 40–50 days after the injection of C6 and E86.L4SN⁵⁰ RPC. Fig. 2,B shows an example of MRI sequences taken 22 and 50 days after injection of C6 and IL-4 RPCs. The large, bilobulated lesion enhanced by gadolinium in the MRI scan taken at day 22 is decreased at day 50, suggesting the presence of an antitumor reaction. This animal was sacrificed 2 days later, and the histological analysis showed large inflammatory infiltrates in the vicinity of residual neoplastic cells (Fig. 3,C). Similar observations were made in other rats sacrificed 6–7 weeks after tumor injection (Fig. 3, A, B, and D). The immunohistochemical analysis demonstrated that these infiltrates were mostly constituted by CD4+ and CD8+ T-lymphocytes, B-lymphocytes, and macrophages (not shown).

C6 cells are allogeneic to Sprague Dawley, thus favoring spontaneous tumor rejection (20). In fact, 6 of 33 controls injected with C6 tumors were tumor free at the moment of the sacrifice, a result consistent with those from other groups (21, 22). To evaluate the effects of IL-4 gene transfer in a syngeneic model, we performed an additional series of experiments using 9L gliosarcoma cells and Fischer 344 inbred rats. With this system, the amount of spontaneous rejections of brain tumors was considerably lower (1of 55 controls tested, 1.8%), although 9L cells are considerably immunogenic (23). We injected inbred 344 Fisher rats with 6 × 10⁴ 9L gliosarcoma mixed or not with equal amounts of E86.L4SN⁵⁰ cells or of PA317.STK.SBA (SBA; Ref. 19), a control RPC used to test nonspecific immunogenic effects of retroviral transduction. PA317.STK.SBA RPC transduce the “suicide” HSV-tk gene and were used in the absence of ganciclovir. Five of 12 rats inoculated with 9L+E86.L4SN⁵⁰ RPCs were long-term survivors, whereas 11 of 12 rats injected by 9L cells and 13 of 13 injected by 9L and PA317.STK.SBA cells died by day 46 (Fig. 4 A; 9L versus 9L+SBA was not significant, P = 0.4643; 9L+ E86.L4SN⁵⁰ versus 9L, P = 0.0007; 9L+ E86.L4SN⁵⁰ versus 9L+SBA, P = 0.0227).

To test further the effects of IL-4 concentration on tumor growth, Fischer 344 rats were injected with 9L cells mixed with E86.L4SN²⁰⁰, a new RPC clone with titer of the same order of magnitude as E86.L4SN⁵⁰ but producing four times the amount of IL-4. Three months after coinjection of 9L and E86.L4SN²⁰⁰ cells, 75% of the rats survived (Fig. 4,A; 9L+ E86.L4SN²⁰⁰ versus 9L, P = 0.0001; E86.L4SN²⁰⁰ versus 9L+SBA, P < 0.0001; E86.L4SN⁵⁰ versus E86.L4SN²⁰⁰ was not significant, P = 0.1126). This confirms, together with data obtained in C6 tumors (see Fig. 1), the correlation between the fraction of rats cured by the treatment, the amount of IL-4 due to the production by packaging cells, and the efficiency of transduction of the IL-4 gene. Although a direct evaluation of the concentration of IL-4 in the tumor microenvironment was not performed, the amount of IL-4 levels in brain homogenates of rats injected with 9L admixed with SBA cells or E86.L4SN²⁰⁰ cells was measured 1, 2, and 3 weeks after cell grafting. These preliminary data showed that IL-4 levels in brains of rats injected with 9L and E86.L4SN²⁰⁰ cells were 20, 25, and 7 times higher than in controls (average values of two rats for each time point). On the contrary, no relevant variation of IL-4 was found in 9L/SBA controls at any time point (5.8 ± 3.4 pg of IL-4/mg of protein, n = 5).

The effects of the inoculation of RPCs transducing IL-4 into established gliomas was also investigated in the 9L system. Using the same stereotactic coordinates, Fischer 344 rats were injected with E86.L4SN²⁰⁰ or SBA cells (as a control) 4 days after the brain injection of 9L cells. This experiment confirmed that the animals injected by IL-4 RPCs survived significantly longer than the controls (Fig. 4 B) and that a relevant fraction of them (50% in this experiment) can survive tumor injection.

In several animals, the packaging cells injected into the established gliomas had been incubated with BrdUrd 2 days before surgery. Immunohistochemistry with anti-BrdUrd antibodies was performed at different time points (1, 5, 11, 14, 18, and 45 days after surgery), and the results indicated that packaging cells were present, usually at the boundary of the tumor, until 18 days after surgery. This observation is in agreement with results reported by Ram et al. (24).

To evaluate whether IL-4 long-term survivors had developed an immunological memory that could reject a challenge with wild-type tumors, 6 × 10⁴ 9L cells were injected into the contralateral hemisphere of seven rats 3 months after the first injection of tumor cells. All of these animals survived the challenge. In a similar experiment on 10 long-term survivors to C6 glioblastomas, 8 animals survived the C6 challenge in the right hemisphere. Surprisingly, one of the two succumbing animals died because of a relapse of the first tumor in the left hemisphere.

Brains injected with 9L tumors were also analyzed by MRI and histological techniques. Of controls and treated rats dying spontaneously, all showed a very large malignant glioma compressing the remaining part of the brain. In treated animals sacrificed at different time points, but not in controls, inflammatory infiltrates, mostly constituted by lymphocytes, were detected. In general, these infiltrates appeared earlier (i.e., during the first 4 weeks after injection) and were larger in rat brains injected with E86.L4SN²⁰⁰ cells than in brains injected by ψ2.L4SN²⁰ and E86.L4SN⁵⁰ cells. Furthermore, these, but not the infiltrates associated with ψ2.L4SN²⁰ and E86.L4SN⁵⁰ cells, showed that CD8+T-lymphocytes were more numerous than CD4+ and B lymphocytes. NK cells and eosinophils were almost undetectable. This is an intriguing finding, because the presence of eosinophilic granulocytes was strongly associated with antitumor action of IL-4 (25). In IL-4-treated gliomas, eosinophils were detected in some cases (12, 13, 15) but not in others (Ref. 18 and this work). This may reflect variations in the amount of IL-4 (high levels may be initially necessary to recruit a large number of eosinophils) and/or variations in the presence of key factors in the brain, such as IL-5 or eotaxin, necessary to attract these cells. Recent data, however, suggest that the growth suppression of IL-4-secreting tumors is not dependent on the presence of the eosinophils (26).

An example is shown in Fig. 5, where the inflammatory cells were detected 23 days after injection of 9L and E86.L4SN²⁰⁰ cells. The MRI picture of this tumor, 3 days before the histological analysis, is shown in Fig. 6,B. The MRI of a control tumor at day 19 is shown for comparison in Fig. 6 A. The control tumor is clearly larger, suggesting that inflammatory cells give a major contribution to tumor shrinkage.

That tumor rejection is mostly due to the action of specific and nonspecific inflammatory cells is also confirmed by in vitro experiments performed to test whether IL-4 may affect per se the proliferation of glioma cells. IL-4 receptors are expressed by astrocytes and by astrocytic tumors and may mediate negative growth signaling by IL-4 (27, 28). This negative effect on proliferation, however, seems only present in a subgroup of glioblastomas (28, 29). Immunohistochemistry and Western blotting with polyclonal antibodies failed to identify the α chain of the IL-4R in 9L and C6 cells (rat spleen and mouse HT-2 cells were the positive controls). Reverse transcription-PCR, however, gave a positive result (not shown), suggesting that the expression level of this receptor is rather low in both of these glioma cell lines. We subsequently evaluated whether the interaction of IL-4 with these receptors had some effect on proliferation. C6 and 9L cells were overlayed with different dilutions of the medium of NIH-3T3 cells transduced by the IL-4 gene and producing 150 ng of IL-4/1 × 10⁶/cells/48 h. As a control, we used the medium of untransduced NIH-3T3 cells. Under these conditions, IL-4 did not affect the proliferation of the tumor cells, and similar results were obtained by adding recombinant murine IL-4 to the culture medium in concentrations ranging from 10 pg to 100 ng/ml (data not shown).

Glioblastomas are very aggressive tumors characterized by the infiltrative capacity, the high proliferation rate, and the strong stimulation of neoangiogenesis. Progress in neurosurgery, radiotherapy, and chemotherapy did not modify significantly their prognosis (30). Gene therapy approaches based on transfer of “suicide” genes have been encouraging at the preclinical level, and clinical trials have been started (31, 32). Other gene therapy approaches have been based on the transfer of wild-type copies of tumor suppressor genes inactivated during the malignant progression of glioblastomas (33).

All of these strategies are hampered by the difficulty of targeting the majority of neoplastic cells and, particularly, the infiltrating cells, because in vivo transduction with defective viral vectors relies mostly on contiguity between viral producer cells or viral particles and the tumor cells. The search for new, nondefective vectors that may transfer genetic information more efficiently, maintaining the selectivity for neoplastic cells, is very active and includes the creation of new retroviral vectors with engineered envelope proteins or of replication-competent attenuated derivatives of herpes simplex virus (34, 35).

Gene therapy protocols based on the possibility of raising a strong immune response against the tumor have the potential to generate a therapeutic response affecting all glioblastoma cells without the need for a generalized transduction of neoplastic cells. In this perspective, we have developed a series of experiments to test the effects of IL-4 gene transfer on rat brain malignant gliomas.

We have first demonstrated that the coinjection of glioma cells with IL-4 RPCs causes a significant prolongation of survival with respect to controls and the virtual cure of a sizeable fraction of treated animals. Our previous results indicated that 23% of rats with C6 tumors coinjected with ψ2.L4SN²⁰ cells, whose viral titer and IL-4 production are, respectively, 5 × 10⁵ cfu/ml and 20 ng/5 × 10⁵ cells/48 h, survived >3 months (17), a finding now confirmed on a larger number of animals (27% on 18 rats; Fig. 1). Coinjection of 9L or C6 cells with E86.L4SN⁵⁰ cells, characterized by a viral titer of 3 × 10⁷ cfu/ml and by a release of 50 ng of IL-4/5 × 10⁵ cells/48 h, brought to 42–56% the percentage of long-term survivors. The injection of 9L and E86.L4SN²⁰⁰ cells (similar titer as E86.L4SN⁵⁰ but 200 ng of IL-4/5 × 10⁵ cells/48 h) raised to 75% such percentage.

These findings indicate a direct correlation between amounts of IL-4 at the tumor site and rat survival. In particular, the different survival rates obtained with RPCs of similar titer but producing different amounts of IL-4 suggest that if more IL-4 is available at the tumor site, the threshold needed to trigger the immune response is reached more rapidly. To be clinically efficacious, the treatment needs an early activation of the immune response before the formation of a large tumor. Thus, IL-4 that is present early during tumor development (measurements on brain homogenates demonstrated high levels during the first 2 weeks after injection of E86.L4SN²⁰⁰) could stimulate the presentation of tumor antigens by macrophages or B lymphocytes (these cell types were found in inflammatory infiltrates of IL-4-treated tumors) and elicit an antitumor response, mostly based on CD8+ T-lymphocytes. Furthermore, transduced tumor cells, which in vitro proliferate like wild-type cells (see “Results”), may contribute to maintain high levels of IL-4.

These two features of retroviral-mediated gene transfer, the temporary survival of RPC producing high levels of IL-4 before the formation of a large glioma and the stable integration of the transduced gene into actively dividing tumor cells, may underlie the high survival rates obtained in some of our experiments. Their lack, on the contrary, may explain why IL-4 transduction of GL261 gliomas by herpes simplex-based defective constructs could only increase slightly the survival of C57BL/6 mice (18).

It should be acknowledged that in our experiments we have matched mouse IL-4 with rat IL-4Rs. Human/mouse IL-4 species specificity has been demonstrated previously (36, 37). Rat/mouse IL-4 and IL-4R have higher similarity than human/mouse IL-4 and IL-4R (58% versus 39%, respectively, for IL-4; 79% versus 52%, respectively, for IL-4R). Furthermore, previous reports have demonstrated that mouse IL-4 can counteract graft rejection in the rat (38) and have an antitumor effect on rat C6 gliomas (15). These observations, in combination with our results, suggest that the species specificity of rat and mouse IL-4 is not as tight as for human IL-4. It is possible, however, that the use of species-specific IL-4 may contribute further to the antitumor efficacy of this cytokine.

In view of future clinical applications of this gene therapy approach, three other points need to be mentioned: (a) signs of toxicity that could be linked to IL-4 were never detected, and rats’ deaths were always due to a large brain tumor and not to inflammatory edema of the brain. Preliminary experiments based on injection into the brain of Sprague Dawley rats of ST14A neural progenitor cells derived from the same rat strain (39) and that we transduced to overexpress IL-4 confirmed the absence of brain or systemic signs of toxicity.⁵ This seems rather relevant in light of recent findings indicating that intracranial injection of cells overexpressing IL-2, IFN-γ or granulocyte/macrophage-colony-stimulating factor does not increase tumor survival and can cause severe CNS toxicity (40, 41, 42); (b) established 9L gliomas injected with IL-4 RPCs were cured in 50% of the rats. This experimental setting mimicks quite closely a clinical situation in which, after tumor debulking, RPCs could be injected at the boundaries of the tumor cavity; and (c) IL-4 has a relevant inhibitory effect on angiogenesis (43). This can be an important component of its antitumor effect (44) that may have contributed significantly to our results. The investigation of the mechanisms of the antiangiogenic effect of IL-4 on malignant gliomas and the precise identification of immune cells that in different districts of the immune system are responsible for tumor rejection (tumor-infiltrating lymphocytes and tumor-specific T lymphocytes in the spleen and in cervical lymph nodes) will be the focus of future research aimed at the characterization of the cellular and molecular mechanisms underlying IL-4 -driven rejection of experimental malignant gliomas.

We thank Santa Florio, Luciana Caposio, Marinella Cova, Silvia Benvenga, and Maurizio Gelati for their collaboration during MRI exams, Sonia Spinello for help in histological analysis, Daniele Bonsignorio for help in Western Blot analysis, Leopoldo Fertilanno for animal care, Sergio Ottolenghi for his collaboration during the preparation of retroviral producer cell lines, and Massimo Musicco for helpful suggestions during statistical analysis of rat survival.