Introduction

Malignant gliomas are the most common intrinsic tumors of the central nervous system (CNS) and are refractory to most standard therapies, including surgical resection, radiation therapy, and chemotherapy.1 The biology of these tumors, including their highly infiltrative nature, aggressive growth characteristics, potent immunosuppression, and anatomic location have presented significant problems for development of effective therapy. Because of their highly infiltrative growth patterns, complete surgical resection without significant local damage to the brain is not attainable. Similarly, because of the diffuse infiltrative elements of these tumors, elimination by directed external beam radiotherapy, including stereotactic radiosurgery, is not entirely achievable.2,3

Many investigators have suggested immunotherapy as an attractive complement to conventional therapeutic modalities. There was early skepticism regarding whether elements of the peripheral, adaptive immune system could gain sufficient access to tumors in the CNS to have a significant impact on tumor growth. However, recent studies have clearly shown that CNS tumors can be infiltrated by immune cells from the periphery,4,5 and more importantly, that immunization in the periphery can promote a therapeutically meaningful attack against established CNS tumors.6,7,8,9

Although it is now clear that immune effectors can gain access to CNS tumor site, additional issues have been raised regarding the functional capabilities of these infiltrating cells. For instance, gliomas release factors such as transforming growth factor-beta (TGF-β), which inhibit a number of functions of cytolytic T cells and NK cells.10,11 Gliomas also express Fas ligand (CD95L), which induces apoptosis in immune effector cells.12 A more recent study has demonstrated that human glioma cells produce substances that also suppress the function of antigen presenting cells (APCs).13 This suggests that achieving optimal participation of the immune system in treatment of gliomas will require either immuno-enhancing treatments, reversal of the immunosuppressive effects of the tumor, or both.

In considering the development of immunotherapy for gliomas, a number of reports have indicated that local delivery of non-adaptive effectors, such as lymphokine-activated killer (LAK) cells,14,15 are not highly effective approaches. Attempts at active immunization with whole tumor cells or extracts have also proven ineffective.16,17,18 However, recent data demonstrate that effective anti-CNS tumor immune responses can be generated through immunization with dendritic cells (DC) pulsed with a tumor-specific antigen peptide or with bulk tumor antigens,7,19,20,21 or through immunization with cytokine-gene modified tumor cells as vaccines.6,8,9 Also, as an alternative strategy to circumvent the relatively cumbersome steps of ex vivo tumor-cell processing, culture and transfection for vaccine preparation,22 cytokine genes have been delivered directly into the brain tumor site with various vector systems and therapeutic benefit has been demonstrated.23,24,25 More recently, interleukin-4 (IL-4) transfected neuronal progenitor cells have been used as a vehicle to deliver IL-4 at the brain tumor site.26 Although this strategy seems to be promising, whether any other cytokine offers advantages over IL-4 is not yet clear.

To determine the most effective means of inducing a therapeutic response to intracranial (i.c.) tumors using a single cytokine gene, we directly compared i.c. delivery of cytokine genes and i.d. vaccination protocols using the rat 9L gliosarcoma line transduced to express IL-4 (9L-IL4), interferon-α (9L-IFNα), granulocyte–monocyte colony-stimulating factor (9L-GMCSF) or IL-12 (9L-IL12). These cytokines were chosen because potent induction of specific cellular immune responses has been demonstrated when they were used in gene therapy or as adjuvants in vaccines for other types of cancers.23,27,28,29,30,31,32,33,34,35,36,37 For comparison, therapeutic responses to both established subcutaneous (s.c.) and i.c. parental 9L tumors were analyzed, including evaluations of immune cell infiltration, degree of vascularization, tumor volume, and survival.

Results

Cytokine transfected tumors were rejected i.d., but not i.c.

In order to compare the rates of growth of cytokine-transfected tumors, animals were injected either i.c. or i.d. with cytokine-expressing 9L cells. Figure 1a illustrates the i.d. growth of tumors at various time intervals following injection of 2 × 106 cells of each tumor type. Parental 9L and 9L-neo tumors grew at comparable rates. In contrast, 9L-IL4, 9L-IL12, 9L-GM-CSF and 9L-IFNα tumors grew and formed small i.d. tumor nodules, and were rejected by day 22. There was no statistically significance difference in the rates of growth among 9L-IL4, 9L-IL12, 9L-GM-CSF or 9L-IFNα.

Figure 1
figure 1

Cytokine transfected 9L tumors are efficiently rejected when injected intradermally, but not when implanted intracranially. (a) Following i.d. injection of 2 × 106 control or cytokine transfected 9L tumor cells on day 0, serial measurements were made and expressed as mean tumor area (n = 4). Each data point represents means of four tumors/group. (b) Aliquots of 1 × 105 control or cytokine transfected 9L tumor cells were implanted stereotactically into the right frontal lobe of each rat on day 0. The data shown are cumulative survival results for animals in two independent experiments. Open diamonds, parental 9L; open squares, 9L-neo; closed triangles, 9L-IL4; crosses, 9L-IL12; closed squares, 9L-GM-CSF; closed circles, 9L-IFNα.

Full size image

As illustrated in Figure 1b, in experiments comparing the survival of animals bearing i.c. cytokine transfected tumors, all tumors were lethal following i.c. implantation, except for one of 10 animals injected with 9L-GM-CSF-transfected tumor cells. However, survival of animals was prolonged for all cytokine-transfected 9L tumors. The comparisons of median survival of animals bearing cytokine transduced tumors and of animals bearing 9L-neo (median survival 20 days) are as follows: 9L-IL4, 37.5 days, P = 0.0003; 9L-IL12, 37 days, P = 0.0028; 9L-IFNα, 38 days, P = 0.0002; and 9L-GMCSF, 39.5 days, P = 0.0003. Necropsies of animals showed that all died of tumor-burden, regardless of the cytokine used. 9L transfectants isolated from the brains of these animals were confirmed to express transgenes as demonstrated by ELISA on supernatants of tumor cells obtained at necropsy and cultured in vitro (data not shown).

Difference in numbers and phenotype of tumor infiltrating lymphocytes (TILs) isolated from i.d. and i.c. cytokine transduced 9L tumors

The striking disparity in tumor growth observed in i.c. and i.d. cytokine-transduced 9L tumors raised intriguing questions regarding the quality and magnitude of the respective immune responses in the CNS and periphery. In order to compare the immune responses in the i.c. and i.d. tumor microenvironment, TILs were isolated, quantitated and analyzed phenotypically. First, animals were injected either i.d. (2 × 106) or i.c. (1 × 105) with 9L transfectant lines (four animals/group) on day 0. On day 12, tumors were harvested, tumor size determined, and TILs were isolated and characterized phenotypically by flow cytometry. Data in Table 1 are representative of three separate experiments and illustrate the relative size of tumors and numbers of TILs in the various 9L transfectants. In comparing the relative size of 9L-neo versus cytokine-transduced tumors, there was ⩾92% reduction (range 92–98%) in size of cytokine-transduced i.d. tumors. Similarly, there was a ⩾85% reduction (range 85–89%) in size of all i.c. cytokine-transduced tumors compared with 9L-neo. In comparing the relative level of immune cell infiltration of the various cytokine-transduced 9L tumors versus 9L-neo, it was determined that in i.d. tumors there was a two-fold greater density of TILs in 9L-IL12, a 2.07-fold greater density of TILs in 9L-IFNα, a 2.54-fold greater density of TILs in 9L-IL4, and a 4.35-fold greater density of TILs in 9L-GM-CSF. For i.c. tumors, there was a 3.6-fold greater density of infiltrating TILs for 9L-IL4, a 5.61-fold greater density for 9L-IL12, a 6.04-fold greater density for 9L-GM-CSF, and a 6.73-fold greater density for 9L-INFα.

Table 1 Number and density of TILs in cytokine-transfected 9L tumors
Full size table

As an additional means of characterizing the relative differences in TILs in i.c. and i.d. tumors, isolated TILs were analyzed phenotypically by flow cytometry. In particular, we determined the relative percentages of cells of the dendritic cell (DC)/monocyte lineage using two parameter staining with anti-MHC class II and anti-CD11b/c; the relative percentages of T helper cells using anti-CD4 and anti-TCRαβ; and the relative percentages of cytolytic T lymphocytes (CTLs) using anti-CD8 and anti-TCRαβ. As illustrated in Table 2a, there was a marked increase in cells expressing DC/macrophage markers in i.d. 9L-IL12 (five-fold) and 9L-GM-CSF (three-fold) compared with 9L-neo. There were minimal increases in these cells in 9L-IL4 and 9L-IFNα. In contrast, i.c. tumors were found to have the greatest increase in DC/macrophage lineage cells in 9L-IL4 (four-fold), whereas 9L-IL12 and 9L-GM-CSF had two fold and 2.5-fold increases, respectively (Table 2b). There was no apparent difference in the relative percentage of DC/macrophage lineage cells in 9L-IFNα.

Table 2 Phenotypic analysis of TILs
Full size table

In comparing the relative percentages of T helper cells in the various 9L transfectants, it was determined that there were decreases in CD4+/TCRαβ+ cells in i.d. 9L-IL4 (14% reduction), 9L-IL12 (46% reduction), but increases in 9L-GM-CSF (1.5-fold) and 9L-IFNα (1.1-fold) (Table 2a). As regards the relative percentage of CD8+/TCRαβ+ cells in i.d. tumors, it was determined that there were decreases in 9L-IFNα (32% reduction) and 9L-IL12 (22% reduction), whereas there were increases in these cells in 9L-IL4 (1.6-fold) and 9L-GM-CSF (1.6-fold) (Table 2a).

In similar experiments in 9L tumors implanted in the CNS, some differences were strikingly apparent (Table 2b). For instance, there were increases in the percentages of class II+/CD11b/c+ cells in 9L-IL12 (1.9-fold) and 9L-GM-CSF (2.4-fold) compared to 9L-neo. However, the most marked increase in class II+/CD11b/c+ cells was in 9L-IL4 (four-fold). It has been reported that OX62 represent a unique marker for DC.38 However, other investigators have reported OX62 is negative on matured rat DCs,39 and our unpublished data from rat bone marrow culture coincide with their results (data not shown). Therefore, we did not use this to evaluate DC/monocyte type TILs. In determining the relative percentages of CD4+/TCRαβ+ cells in i.c. tumors, it was observed that there was no change in the percentages of these cells in 9L-IL12 or 9L-IFNα, whereas there was a substantial increase in these cells in 9L-GM-CSF (6.2-fold) (Table 2b). It was not possible to determine the relative percentage of CD4+/TCRαβ+ cells in 9L-IL4 in these experiments due to minimal tumor recovered and to the lower number of TILs in these tumors (Table 1). A similar pattern was observed for CD8+/TCRαβ+ cells as there was a substantial increase in these cells in 9L-GM-CSF (five-fold), but minimal differences in 9L-IL12 and 9L-IFNα. Due to minimal tumor recovered and the lesser number of TILs in these tumors (Table 1), it was not possible to make a determination of the relative percentage of CD8+/TCRαβ+ cells in 9L-IL4 in these experiments. Overall in Table 2, it is noteworthy that i.d. tumors generally had higher relative percentages of CD4+/TCRαβ+ cells and CD8+/TCRαβ+ cells in comparison to i.c. tumors.

Cytokine expression at the tumor site inhibits angiogenesis

While the cytokines transduced into 9L tumor cells have all been reported to serve as immune enhancing agents, they also have the potential to exert other effects in the tumor microenvironment such as disruption of angiogenesis. To examine the effects of local production of cytokines in 9L tumors on the angiogenesis in the tumor microenvironment, animals were injected i.c. with 1 × 105 cytokine-transduced 9L tumor cells on day 0. Animals were then killed on day 14 and tumor tissues were analyzed by immunohistochemistry using anti-von Willebrand Factor (factor VIII-related antigen) that stains vascular endothelial cells, and that is commonly used as a means of quantitating vascularization. Figure 2 is a representative experiment comparing the degree of vascularization of 9L-neo and 9L-IFNα. For quantitation, numbers of vessels in different 9L-cytokine-transfectant tumors were counted in five randomly chosen visual fields under light microscopy (×200). Figure 3 illustrates that statistically significant suppression of angiogenesis was observed in all cytokine transduced 9L tumors relative to 9L-neo. The number of vessels quantitated in tumors represented findings of a mean reduction of 61% in 9L-IL4 compared with 9L-neo. Other transfectants including 9L-IL12, 9L-GM-CSF and 9L-IFNα represented 37%, 51%, and 88% reductions in vessels, respectively. Similar effects were observed in i.d. 9L-cytokine transfected tumors (data not shown).

Figure 2
figure 2

Immunohistochemical analyses of vascularization of cytokine transduced 9L tumors compared with sham-transduced 9L (9L-neo). Rats bearing i.c. 9L tumors were killed on day 14 following implantation of 1 × 105 tumor cells, and tumors were harvested and snap frozen. Sections of 4–5 μm were cut on a cryostat and stained with von Willebrand's factor specific antibody to visualize vascular endothelial cells as dark brown spots using diaminobenzidine. Sections were counterstained with crystal violet. (a) 9L-neo tumors; (b) 9L-IFNα. Similar results were obtained with 9L-IL4, 9L-IL12, or 9L-GM-CSF (not shown). The scale bars represent 340 μm.

Full size image
Figure 3
figure 3

Local expression of cytokines in intracranial 9L tumors resulted in decreased neovascularization. Rats were injected stereotactically with 1 × 105 9L tumor cells. On day 14 following implantation, tumors were harvested, snap frozen, sectioned (4–5 μm) on a cryostat and stained with anti-von Willebrand's factor antibody. Numbers of vessels were counted in at least five randomly chosen fields using a light microscope (×200). P values, based upon a Student's t test assuming unequal variances, were calculated for cytokine transduced 9L tumor compared with 9L-neo; and were 0.00139, 0.01814, 0.00323, and 0.00065 for 9L-IL4, 9L-IL12, 9L-GM-CSF and 9L-IFNα, respectively.

Full size image

Immunization with 9L-GMCSF, 9L-IL4 and 9L-IFNα induces therapeutic immunity to established i.d. parental 9L tumors

To address whether cytokine-transduced 9L tumor cells could be used as vaccines for systemic immunization to treat parental 9L tumors established in the periphery, rats were injected in the right flank i.d. with 8 × 105 parental 9L on day 0. On day 3, 2 × 106 cytokine-transfected 9L tumor cells were injected s.c. on the left flank. Anti-tumor responses were assessed by measuring the growth of parental 9L tumors over time. As shown in Figure 4, immunization with 9L-GM-CSF, 9L-IFNα, and 9L-IL4 induced comparable therapeutic responses to established 9L tumors. Statistical analyses of the tumor size in comparison to the 9L-neo treated group using a Student's t test (two-sample assuming unequal variances) on day 24 tumors revealed a significance of P = 0.021659, P = 0.087205, P = 0.01173 and P = 0.006437 for 9L-IL4, 9L-IL12, 9L-IFNα and 9L-GM-CSF, respectively. Only the 9L-IL12 treatment did not achieve a statistically significant reduction in tumor size. In fact, one of the 9L-IL12 treated animals had to be killed, as were all animals treated with 9L-neo or HBSS. In contrast, one of the 9L-IL4 animals and one of the 9L-GMCSF treated animals entirely rejected parental 9L tumors by day 30 (data not shown). These data support the conclusion that significant therapeutic responses to parental 9L tumors established in the periphery could be induced by 9L transfected with IL-4, GM-CSF or IFN-α, but not with IL-12.

Figure 4
figure 4

Systemic immunization induced therapeutic immunity to established, parental 9L tumors is comparable with 9L-IL4, 9L-IL12, 9L-GM-CSF or 9L-IFNα Rats bearing day 3, i.d. parental 9L tumors (8 × 105) were injected on the opposing flank with 2 × 106 cytokine transduced 9L tumor cells. Anti-tumor responses were assessed by serial measurements of parental 9L tumor area. Significantly reduced growth of parental 9L tumors was achieved by day 24 in rats given 9L-IL4, 9L-GM-CSF or 9L-IFNα. Rats treated with 9L-IL12 did not achieve a significant reduction in growth of parental 9L. Open diamonds, HBSS; open squares, 9L-neo; closed triangles, 9L-IL4; crosses, 9L-IL12; closed circles, 9L-IFNα; closed squares, 9L-GM-CSF.

Full size image

9L-IL4 vaccine confers the most potent protective immunity against i.c. challenge with parental 9L tumor cells

In order to compare the induction of immunity to CNS tumors by peripheral vaccination with cytokine transfected 9L cells, rats were first immunized i.d. with non-irradiated 9L-neo or cytokine transfected 9L cells (2 × 106) on day 0. These animals were then challenged i.c. with parental 9L cells (1 × 105) on day 28. As shown in Figure 5, 9L-IL4 induced the most effective protective immune response, with 90% long-term survivors (>100 days; P < 0.0001) compared with 9L-neo. 9L-GM-CSF and 9L-IFNα protected animals to a lesser extent, having 40% (P = 0.0084, compared with 9L-neo) and 30% (P = 0.0296, compared with 9L-neo) long-term survivors, respectively. None of the 9L-neo or 9L-IL12 immunized animals survived longer than 52 days after i.c. challenge with parental 9L tumor cells. These data support the conclusion that 9L-IL4 was the most effective immunogen for generating protective responses against i.c. tumor challenge with parental 9L cells.

Figure 5
figure 5

Systemic immunization to induce protective immunity against i.c. challenge with parental 9L tumor cells was most effective with 9L-IL4. Rats injected i.d. with HBSS or 2 × 106 9L-neo, 9L-IL4, 9L-IL12, 9L-GM-CSF or 9L-IFNα on day −28 were challenged by i.c. injection of 1 × 105 parental 9L tumor cells on day 0. The data shown are cumulative survival curves pooling results in two independent experiments. Open diamonds, HBSS; open squares, 9L-neo; closed triangles, 9L-IL4; crosses, 9L-IL12; closed squares, 9L-GM-CSF; closed circles, 9L-IFNα.

Full size image

Immunization with 9L-IL4 induced the most effective therapeutic immunity against established i.c. parental 9L tumors

In a more clinically relevant model, the therapeutic efficacy of peripheral immunization with cytokine-transduced 9L tumor cells was tested against established, i.c., parental 9L tumors. Animals were first inoculated i.c. with parental 9L cells (1 × 104) on day 0. These animals were then treated i.d. on day 3 with non-irradiated 9L-neo or 9L cytokine-transfectants (2 × 106). As illustrated in Figure 6, long-term survival was observed only in the group of animals treated with 9L-IL4 (P = 0.0251 when compared with 9L-neo). In this group, six of 14 (43%) of animals receiving 9L-IL4 treatment survived for longer than 100 days. In contrast, none of other treatment groups (9L-IL12, 9L-IFNα or 9L-GM-CSF) or control treatment groups (9L-neo and HBSS) survived longer than 43 days, although the 9L-GM-CSF treated group did show a statistically significant prolongation of survival (P = 0.0111) compared with the 9L-neo treatment group. These data support the conclusion that 9L-IL4 was the most efficient in inducing a therapeutic response to parental 9L tumors established in the CNS.

Figure 6
figure 6

Systemic immunization with 9L-IL4 conferred long-term survival in rats bearing established parental 9L i.c. tumors. Rats bearing day 3, i.c. parental 9L tumors (1 × 104) were injected i.d. with HBSS or 2 × 106 9L-neo, 9L-IL4, 9L-IL12, 9L-GM-CSF or 9L-IFNα. The data shown are cumulative survival curves pooling results in two independent experiments. Open diamonds, HBSS; open squares, 9L-neo; closed triangles, 9L-IL4; crosses, 9L-IL12; closed squares, 9L-GM-CSF; closed circles, 9L-IFNα.

Full size image

Discussion

In this study, we have demonstrated: (1) that cytokine gene transfected tumors induce more potent anti-tumor immune responses in the periphery than in the CNS, as evaluated on the basis of both tumor growth/rejection and TIL infiltration; (2) that direct delivery of a single cytokine gene to a CNS tumor site prolongs the survival of animals by inhibiting tumor growth to some extent, but does not consistently lead to cure; (3) that local production of cytokines inhibits neovascularization, which may contribute to prolonged survival of i.c. tumor bearing animals; (4) that for treatment of i.d. tumors, immunization with 9L-GM-CSF or 9L-IFNα was as potent as 9L-IL4; and (5) that for parental 9L established as i.c. tumors, immunization with 9L-IL4 induces the most potent protective and therapeutic immune responses.

We used the 9L gliosarcoma model because of its clinically relevant characteristics such as invasive and aggressive growth in the brain of syngeneic animals and elaboration of TGF-β2 as an immunosuppressive factor.40 We used non-irradiated cell vaccines because we had found non-irradiated vaccines to be more effective than irradiated vaccines.6 Accordingly, we developed a vector system expressing a cytokine and herpes simplex virus-thymidine kinase (HSVtk) so that we could use nonirradiated tumor vaccines and eliminate vaccine site tumors by administration of ganciclovir if needed.6,22,41

In addition to peripheral vaccine strategies, direct cytokine gene transfer approaches to brain tumor sites have been also shown to be effective in some murine models.23,24,25,42 Although induction of inflammation following cytokine gene expression in the CNS may cause life-threatening brain edema or elevation of intracranial pressure,43,44 this approach would circumvent cumbersome procedures of ex vivo transfection and expansion of gene-transfected cells which are required for peripheral vaccine approaches. Therefore, we tested the efficacy and responses following direct injections of cytokine gene-transfected tumors directly into the brain. Although i.c. implantation of cytokine-transduced 9L tumor resulted in delayed growth and increased survival compared with 9L-neo, virtually all 9L transfectants in the CNS ultimately proved lethal (Figure 1). Quantitative analyses showed a much lower density of TILs, particularly in terms of fewer CD4+/TCRαβ+ and CD8+/TCRαβ+ cells, in i.c. tumors compared with i.d. tumors. As all i.d. tumors were rejected, this suggests a reduced capability for induction of immunoreactivity in the brain compared with the periphery (Tables 1 and 2). Our previous study, using adoptive transfer experiments, revealed important roles of CD4+ and CD8+ cells in 9L-IL4-induced rejection of parental 9L.45 The sparse density of CD4+/TCRαβ+ and CD8+/TCRαβ+ cells in most of the i.c. cytokine gene-transfected tumors may explain the lack of a tumor rejection. Interestingly, among i.c. cytokine gene transfected tumors, only 9L-GM-CSF was rejected in one of 10 animals associated with relatively high numbers of infiltrating CD4+/TCRαβ+ and CD8+/TCRαβ+ cells. These results were observed using 9L cells transfected and selected in vitro with cytokine gene vectors. In clinical settings, transfection of established i.c. glioma by direct delivery of cytokine gene vector would also lead to transfection of other cell types, such as endothelial cells and microglia, and may result in a different therapeutic outcome.23,25,46 Therefore, our results do not necessarily rule out the use of i.c. cytokine delivery for future application to human disease. Further studies comparing ex vivo transfection and in vivo transfection of established tumor are clearly warranted to aid in the assessment of the efficacy of the i.c. gene delivery paradigm.

Disruption of neovascularization is an important strategy being evaluated as a means to control tumor growth in vivo. As cytokines are highly pleiotropic and can disrupt neovascularization in glioma models, we assessed the effects of expression of various cytokine genes in 9L tumors. As demonstrated in Figure 3, there were marked decreases in the degree of vascularization in all cytokine-transfected 9L tumor lines tested, with 9L-IFNα causing the greatest decrease, followed by IL-4, GM-CSF and IL-12. These data are consistent with previous reports indicating that IL-4 disrupted neovascularization in gliomas,47 and while both IL-1248 and GM-CSF49 have been reported to have anti-angiogenic effects in several tumor types, these represent novel findings regarding disruption of neovascularization by IFN-α, IL-12 and GM-CSF in gliomas. What is intriguing in these observations is that there are increased numbers of TILs despite decreased numbers of tumor vessels, which are the route for TILs to gain access into the tumor site. Therefore, cytokines may possess a highly desirable characteristic of both inhibiting tumor neovascularization and concurrently enhancing TIL infiltration.

To date, there has not been a direct evaluation of IL-4, IL-12, GM-CSF and IFN-α for cytokine gene therapy of gliomas in a single model. Therefore, we compared therapy of established i.d. parental 9L tumors (day 3) using vaccination on the opposing flank with 9L-IL4, -IL12, -GM-CSF, or -IFNα (Figure 4). In these experiments, it was determined that significant reductions in the size of parental 9L tumors could be observed following immunization with 9L-IL4, 9L-GM-CSF or 9L-IFNα. In the case of vaccination with 9L-IL12, there was not a statistically significant decrease in the size of parental 9L tumors. Murine IL-12 is biologically active in the rat immune system, as we have observed a potent stimulatory activity of IL-12 secreted from 9L-IL12 on cultures of anti-9L syngeneic T cell lines (manuscript in preparation). Anti-rat glioma specific T cells have been successfully grown only in the presence of irradiated 9L-IL12 as the stimulator cells, but not with 9L-neo or other cytokine gene-transfected 9L cells. Other investigators have demonstrated biological activity of mouse IL-12 in the 9L rat gliosarcoma model.37,50 Jean et al37 demonstrated the continuous infusion of a very low dose of mouse IL-12, totaling 1 ng per day, showed the greatest antitumor effect. Therefore, it appears that the dose of IL-12 may be a critical factor in achieving optimum therapeutic results. We have assessed the efficacy of immunization with only 1 × 105 irradiated 9L-IL12, which is estimated to produce 2–3 ng/day at the vaccine site, but no therapeutic benefit was observed (not shown). The lack of any therapeutic response in our experiments may suggest that the use of IL-12 would require further optimization of the administration dose.

In the model most relevant to a clinical situation for patients with gliomas, it was determined that vaccination of rats, bearing established i.c. parental 9L tumors, with 9L-GM-CSF and 9L-IL4 resulted in therapeutic immunity (Figure 6). Therapeutic immunity achieved with IL-4-transfected glioma was, however, superior in this particular model. Again, one may raise a concern that the use of mouse cytokines in rats may complicate the assessment of relative efficacy. Sampson et al8 have reported efficacy of GM-CSF rather than IL-4 in a mouse B16 model, and the difference in the results on these studies could reflect the use of cytokines across the species rather than different tumor models. However, mouse GM-CSF has been demonstrated to induce potent anti-9L glioma immunity using subcutaneous pumps that release recombinant mouse GM-CSF that are co-localized with irradiated gliosarcoma cells.51 In addition, local expression of mouse GM-CSF at the i.c. rat C6 glioma induced potent immunity.52,53 Furthermore, we have found that recombinant mouse GM-CSF is capable of inducing potent DC in rat bone marrow culture (Brissette-Storkus SC et al, submitted). These data suggest the functional activity of mouse GM-CSF in rats. However, we recognize that the further studies in different glioma models will provide us with more generalizable information for the development of effective strategies in humans.

The data presented herein provide strong support for the continued evaluation of IL-4-transduced gliomas as a means of inducing systemic, therapeutic immunity for CNS tumors. At present, the underlying reasons for differences in the capacity of tumors transfected with various cytokines to induce therapeutic immunity have not been determined. In fact, this could be based upon several of the pleiotropic effects of these cytokines; and what might prove important in one tumor model might not be relevant for others. That is, in the setting of local gene delivery to gliomas, the effects of these cytokines on angiogenesis might be of fundamental importance to anti-tumor effects, but it might be of limited importance for a transfected tumor used for therapeutic vaccination in the periphery. What seems most likely to be important in the use of cytokine-transfected tumors as peripheral vaccines is the systemic immuno-enhancing effects of the various cytokines. For modulating immunity, it is unquestionable that there are fundamental differences in the outcome of the response generated using different cytokines. Presumably, these are dictated by factors produced by the tumor per se and by the microenvironment in which the immunity must manifest. That is, a combination of the cytokine used for transfection of a given tumor type and the cytokines produced by the tumor cells may cooperate in regulating the type of immune response induced. For instance, IL-4 is most often cited as being an important factor in driving Th2-type reactivity, which is not generally perceived as the most important for anti-tumor immunity.54 However, there are recent data indicating that IL-4 is essential for induction of anti-tumor immunity.55 A plausible explanation for this might be found in experiments indicating that specific immune responses induced with combinations of IL-4 and TGF-β resulted in Th1 reactivity,56 which is generally thought to be most important for anti-tumor reactivity. Given that the 9L gliosarcoma produces TGF-β2,40 it is possible that potent Th1 anti-tumor responses result from the use of 9L-IL4 as a vaccine. In fact, it has been reported that Th1-type responses result from local IL-4 expression at the tumor site.55 Furthermore, previous studies by us45 and others57 demonstrated robust tumor-reacting IFN-γ expression from both TILs or lymph node cells obtained from IL-4-expressing tumor immunized animals. It appears that the cytokine used and the tumor type will act in concert to influence the response, and that this must be taken into account in establishing effective therapy for a given tumor type. It has also been reported that the combination of GM-CSF and IL-4 enhances the efficacy of peripheral vaccine against i.c. B16 tumors.9 We created a 9L cell line that expresses both GM-CSF and IL-4 at equally high levels using retroviruses encoding each cytokine and with different selection markers. However, we were unable to observe any enhancement of therapeutic efficacy by this combination in our i.c. 9L model (data not shown).

Overall, it is apparent that peripheral immunization with cytokine-transfected gliomas cells can induce a degree of therapeutic immunity to gliomas established in the CNS. Further, it is likely that IL-4 is the most effective of the cytokines tested to date, and that vaccine strategies for therapy must be further refined to attain curative therapy. It is also clear that the success of this approach is influenced not only by the cytokines utilized, but also the tumor type used, and the microenvironment in which the tumor arises.

Materials and methods

Animals and animal care

Male, Fischer 344 rats (200 g) were obtained from Harlan Sprague–Dawley (Indianapolis, IN, USA) and were housed in the Central Animal Facility of the University of Pittsburgh Cancer Institute. All protocols were approved by the University of Pittsburgh Institutional Animal Care and Use Committee (IACUC).

Recombinant retroviral vectors and infection of 9L gliosarcoma cell line

Rat 9L gliosarcoma cells derived from Fischer 344 rats were maintained in complete medium (CM) composed of minimal essential medium (MEM) supplemented with 10% fetal bovine serum, 100 units/ml penicillin G and 100 units/ml of streptomycin sulfate (Gibco BRL, Gaithersburg, MD, USA). Modified MFG-based retroviral vectors,58 termed DFG-mIL4-Neo, DFG-mIL12-Neo, DFG-mIFNα-Neo and DFG-mGMCSF-Neo were constructed and used as previously described6,30,33,35 to establish 9L-IL4, 9L-IL12, 9L-IFNα and 9L-GMCSF lines, respectively. The cDNA for murine GM-CSF was a generous gift from Dr G Dranoff (Dana-Farber Cancer Institute, Boston, MA, USA).8 The MFG-Neo retroviral vector carrying only neoR was used to create a control transfectant line (9L-neo). Retroviral supernatant was generated by transfecting vector plasmids into the CRIP packaging line58 or BOSC23 packaging cell line59 (generously provided by Dr J Pear (Stanford University, Palo Alto, CA, USA) and Dr D Baltimore (The California Institute of Technology, Pasadena, CA, USA). 9L cells were infected by exposure to supernatant for 4 h in the presence of 8 μg/ml polybrene, and were then subjected to continuous selection in CM containing 1 mg/ml of G418. Selected colonies of transfected cells were expanded, maintained in culture, and regularly determined to be free of Mycoplasma sp. contamination using a Mycoplasma Detection Kit (Boehringer Mannheim, Indianapolis, IN, USA).

Cytokine production assay

Cytokine production by the transfected tumor cells was measured by ELISA using antibodies obtained from PharMingen (San Diego, CA, USA) and expressed as nanograms/106 cells/48 h. The production of cytokine was 118 ± 12, 90 ± 10, 105 ± 15 and 85 ± 7 nanograms/106 cells/48 h for 9L-IL4, -IL12, -GMCSF and -IFNα, respectively.

Intracranial (i.c.) and intradermal (i.d.) tumor models

Intracranial (i.c.) tumors were established by stereotactic implantation with a 10 μl Hamilton syringe at 2 mm anterior to bregma, 2.2 mm to the right of the sagittal suture and 4 mm below the surface of the skull of anesthetized rats using a Kopf stereotactic frame (Kopf Instruments, Tujunga, CA, USA).

Intradermal (i.d.) tumor growth and tumor immunizations were performed by injecting 8 × 105 or 2 × 106 exponentially growing, gene-modified or parental 9L cells in the right flank of syngeneic Fischer 344 rats. Tumor area was calculated using calipers by multiplying the long axis of the tumor by the greatest perpendicular axis. In the instances where irradiated tumor cells were used as vaccines, cells were dosed with 5000 rads using a Gammacell 1000 Elite cell irradiator (MDS Nordion, Kanata, Ontario, Canada) immediately before injection. In the protection model, 2 × 106 irradiated or non-irradiated cells were injected i.d. 28 days before i.c. tumor challenge with 1 × 105 parental 9L cells. In the treatment models, animals with day 3 established i.c. or i.d. tumors were vaccinated with 2 × 106 cytokine transduced or control irradiated 9L-neo cells. Animals bearing i.c. tumors were humanely killed at the onset of hemiparesis, or at ⩾25% of weight loss.

Isolation of tumor infiltrating lymphocytes (TILs)

Rats were injected i.c. or i.d. with 1 × 105 or 2 × 106 9L-transfectants, respectively. On day 12, tumors were resected and minced to yield 1–2 mm pieces. To release tumor cells and tumor infiltrating lymphocytes (TILs), tumor fragments were incubated in a mixture of 30 U/ml hyaluronidase (Sigma, St Louis, MO, USA), 500 U/ml DNAse (Sigma) and 0.01% w/v collagenase (Sigma) in Hank's balanced salt solution (HBSS) (Gibco BRL) at ambient temperatures for 45 min with constant stirring. The cell suspension was strained through a sterile grid and washed three times with HBSS. Lymphocytes were separated from tumor cells by centrifugation on a two-step gradient (75%/100% Ficoll (Uppsala, Sweden)) at 1000 r.p.m. for 20 min. The lymphocytes, which localized at the 75%/100% Ficoll interface, were harvested, and washed twice in CM.

Antibodies, immunostaining and flow cytometry

Fluorescein isothiocyanate (FITC)- or phycoerythrin (PE)-conjugated monoclonal antibodies (mAbs) R73.1 (anti-rat T cell receptor (TCR)-αβ), 341 (anti-rat CD8β), W3/25 (anti-rat CD4), OX6 (anti-rat major histocompatibility complex (MHC) class II), OX42 (anti-rat CD11b/c), were purchased from PharMingen. FITC- or PE-labeled, and unlabeled control mouse Ig, were purchased from Serotec (Raleigh, NC, USA).

Cells (2 × 105) were incubated with 0.5 μg FITC- or PE- labeled mAbs diluted in 25 μl PBS supplemented with 1% FBS (PBS/FCS) for 30 min at 4°C. The cells were washed twice with PBS/FBS, resuspended in PBS/FBS and fixed in 0.5% paraformaldehyde (Sigma) in PBS and analyzed using a FACScan or FACScan Plus cytometer (Becton Dickinson, Mountain View, CA, USA). A total of 5000 viable lymphocytes were analyzed using electronic gating of cells based upon orthogonal and forward light scatter. FITC- or PE-labeled isotype control mAbs were used for assessment of background signals and setting markers. FACS data analyses were performed using REPROMAN software (True Facts Software, Seattle, WA, USA).

To calculate the density of TILs, tumor volume was calculated according to the following formula: Tumor volume (mm3) = (long axis, mm)2 × the greatest perpendicular axis (mm)/2.

Immunohistochemistry for analyses of tumor vascularization

Vascularization in tumors was assessed by immunohistochemical staining with biotinylated anti-von Willebrand Factor antibody (Dako, Carpinteria, CA, USA), which specifically stains endothelial cells. Animals were injected with 1 × 105 cytokine-transduced 9L cells in the brain on day 0. Animals were killed on day 14, and brains containing tumors were frozen in OCT compound at −80°C. Thin sections (4–5 μm) were made on a microtome and were fixed with acetone for 10 min. After air-drying, sections were examined by immunohistochemistry with biotinylated anti-von Willebrand factor antibody (1:100), and the immunoreactivity was visualized using a streptavidin biotin kit (Dako) followed by diaminobenzidine (DAB; Sigma) staining. Cells were counter-stained with crystal violet. Vessels were quantitated by counting five randomly chosen fields under light microscopy (×200).

Statistical analyses

Survival estimates and median survival times were determined using the method of Kaplan and Meier. Survival data were compared using a log-rank test. Growth of i.d. tumors was compared by Student's t test for two samples with unequal variances. Statistical significance was determined at the <0.05 level.