TGFβ secreted by B16 melanoma antagonizes cancer gene immunotherapy bystander effect

Abstract

Tumor-targeted delivery of immune stimulatory genes, such as pro-inflammatory cytokines and suicide genes, has shown to cure mouse models of cancer. Total tumor eradication was also found to occur despite subtotal tumor engineering; a phenomenon coined the “bystander effect”. The bystander effect in immune competent animals arises mostly from recruitment of a cancer lytic cell-mediated immune response to local and distant tumor cells which escaped gene modification. We have previously described a Granulocyte–Macrophage Colony Stimulating Factor (GM-CSF) and Interleukin 2 (IL2) fusokine (aka GIFT2) which serves as a potent anticancer cytokine and it here served as a means to understand the mechanistic underpinnings to the immune bystander effect in an immune competent model of B16 melanoma. As expected, we observed that GIFT2 secreted by genetically engineered B16 tumor cells induces a bystander effect on non modified B16 cells, when admixed in a 1:1 ratio. However, despite keeping the 1:1 ratio constant, the immune bystander effect was completely lost as the total B16 cell number was increased from 10⁴ to 10⁶ which correlated with a sharp reduction in the number of tumor-infiltrating NK cells. We found that B16 secrete biologically active TGFβ which in turn inhibited GIFT2 dependent immune cell proliferation in vitro and downregulated IL-2Rβ expression and IFNγ secretion by NK cells. In vivo blockade of B16 originating TGFβ significantly improved the immune bystander effect arising from GIFT2. We propose that cancer gene immunotherapy of pre-established tumors will be enhanced by blockade of tumor-derived TGFβ.

Electronic supplementary material

The online version of this article (doi:10.1007/s00262-008-0453-1) contains supplementary material, which is available to authorized users.

Introduction

Many immunogene therapy strategies have been developed for the treatment of malignant melanoma. These approaches include the introduction of “suicide genes”, the expression of tumor suppressor genes by tumor cells or the inactivation of oncogene expression, as well as the introduction of genes encoding pro-inflammatory proteins such as co-stimulatory molecules and cytokines. In particular, the genetic modification of melanoma cells to secrete pro-inflammatory cytokines enhances the immunogenicity of these cells by providing signals required to trigger an effective cell mediate immune response [16]. Since it is not possible to modify all pre-existing tumor cells with suicide or proinflammatory genes in situ by any contemporary gene transfer technology, an important feature to consider for cancer gene immunotherapy is the bystander effect [20]. Whereas a small fraction of tumor is gene modified to initiate an immune response in vivo against a vastly bulkier pre-established native cancer. Indeed, a cytokine-secreting live cell cancer vaccine approach is also wholly dependent upon a robust immune “bystander” effect for clinical effectiveness [5]. A vast array of pro-inflammatory cytokines and derivatives has been studied as a means to initiate an anticancer immune response in animal models of melanoma and in clinical trials. However, despite tightly controlled conditions, we––as others [13]- have observed treatment failures in mouse models of melanoma where the immune “bystander” effect was lost as the experimental tumor burden was increased at onset of treatment. As a possible explanation for this observation, we may invoke the balance between inhibitory and stimulatory signals essential in the maintenance of homeostasis and in the regulation of the immune response as possibly antagonistic to the immune bystander effect. During cancer progression this balance is disrupted and inhibitory signals may prevail, leading to an immunosuppression in the tumor-microenvironment and resulting in tumor growth. Various mechanisms have been proposed for tumor cell evasion from physiological immunosurveillance and these include the dysregulation of MHC class I and tumor antigen expression as well as adhesion/accessory molecules expression, induction of anergy or clonal deletion of effector cells, and the secretion of suppressive soluble factors [16].

Previous studies from our laboratory have shown that the fusion protein between GM-CSF and IL-2, aka GIFT2––has novel immunological properties compared to both cytokines in combination, such as greater melanoma site recruitment of macrophages and functional NK cells, and circumvents the limitations of each individual cytokine [21]. With the use of GIFT2 fusokine as means to initiate an anticancer immune response, we here analyzed the immune bystander effect of GIFT2-secreting melanoma cells on wild type B16 present in the tumor site in vivo. We observed that the bystander effect is lost as tumor burden increases and that B16-derived TGFβ was responsible in good part of this acquired refractoriness by its direct effect on innate effector cells despite local production of a potent pro-inflammatory fusokine. These data strongly support the need to target tumor-derived suppressor cytokines––such as TGFβ, for an optimal immune bystander response to a cancer immunogene platform.

Materials and methods

Animals, cell lines, and reagents

All experimental mice were females 6–8 weeks old (Jackson Laboratory, Bar Harbor, ME). The C57Bl/6-derived B16F0 (B16) mouse melanoma cells (American Type Culture Collection [ATCC], Manassas, VA) as well as a polyclonal population of B16 derivative (B16GIFT2 [21]) were maintained in Dulbecco’s modified Eagle’s medium (Wisent Technologies, Rocklin, CA), supplemented with 10% fetal bovine serum (Wisent Technologies) and 50 U/ml Pen/Strep (Wisent Technologies). The cell lines CTLL-2 and mouse embryonic fibroblasts (MEF) (American Type Culture Collection [ATCC], Manassas, VA) were grown according to ATCC’s recommendations. Recombinant mouse TGFβ and IL-2, as well as TGFβ neutralizing antibody (anti-TGFβ1, 2, 3 isoforms) and soluble TGFβ receptor II (TβRII) were obtained from R&D Systems, Minneapolis, MN; antiphosphorylated SMAD2 and SMAD3 antibodies were obtained from Cell Signalling Technology, Danvers, MA; α-tubulin antibody was obtained from Santa Cruz Biotechnology, Santa Cruz, CA. Anti-mouse FcR III/II, CD3, CD8, CD4, CD25, NK1.1, CD80, CD86, CD105, MHC class I, MHC class II, CD122 (IL-2Rβ chain) and the isotype control antibodies for flow cytometry were obtained from BD Biosciences, San Diego, CA. The enzyme-linked immunosorbent assay (ELISA) kit for mouse IFN-γ was obtained from BD Biosciences.

Murine B16F0 tumor implantation in immunocompetent C57Bl/6 mice and immune infiltrate analysis

Wild type and genetically modified GIFT2 fusokine-secreting B16 cells (0.7 ± 0.2 pmol per 10⁶ cells per 24 h) were injected subcutaneously in C57Bl/6 mice, and tumor growth was monitored over time by performing external measurements in two dimensions and calculating using the equation volume = lengthxwidth² × 0.5. For immune infiltrate analysis, 10⁴ or 10⁶ genetically modified cytokine secreting and/or non modified B16 cells were mixed with 500 μl Matrigel (BD Biosciences) at 4°C and injected subcutaneously in C57Bl/6 mice. Implants were surgically removed 6 days after transplantation and enzymatically dissociated as reported previously [21]. After incubation with anti-FcR III/II mAb for 1 h, infiltrated cells were incubated for 1 h at 4°C with appropriate antibodies and analyzed by flow cytometry using a FACS Calibur cytometer (BD).

Flow cytometry analysis of non modified and genetically modified cytokine expressing B16 cells

Flow cytometry analysis was performed in phosphate-buffered saline (PBS) with 2% FBS with the following antibodies: R-phycoerythrin (PE)-conjugated anti-mouse H-2 Kb (MHC class I, clone AF6-88.5), I-Ab (MHC class II, clone AF6-120.1), CD80 (clone 16-10A1), CD86 (clone GL1) and CD105. Isotype control analysis was performed in parallel. Non modified and genetically modified B16 cells expressing mouse GIFT2 were incubated with the appropriate antibodies for 1 h at 4°C and the expression of these cell surface markers was determined by using FACS Calibur cytometer (BD) and analyzed using Cellquest software (BD).

Cytokine-dependent CTLL-2 proliferation assay

CTLL-2 cells were pre-incubated in medium conditioned by non-modified B16 cells and pretreated with TGFβ neutralizing antibody or isotype control, as well as in medium conditioned by MEF. The cells were plated at 10⁴ cells/well of a 96-well plate with increasing concentration of mouse recombinant IL-2. The cells were incubated for 48 h and 20 μl of 5 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) solution were incorporated for the last 4 h of incubation. The reaction was stopped by adding 200 μl of dimethyl sulfoxide and absorbance read at 570 nm. The expression of CD8 and CD122 (IL-2Rβ chain) on CTLL-2 cells pre-incubated as previously described were determined by flow cytometry analysis using FACS Calibur Cytometer (BD).

Murine NK cell isolation

Mouse NK cell population was obtained by resuspending 10⁷ splenocytes/ml in PBS containing 0.5% bovine serum albumin and treated with biotin-antibody cocktail containing anti-CD5 (Ly-1), CD8a (Ly-2), CD4 (L3T4), Gr-1 (Ly-6G/C), CD19 and Ter-119 MicroBeads (Miltenyi Biotec, Gladbach, Germany). After incubation for 15 min at 4°C, cells were washed and the cell populations were depleted by magnetic cell sorting (MACS) system with an autoMACS^TM column (Miltenyi Biotec) according to the manufacturer’s instructions. NK cell population purity assessed by flow cytometry was 90% (data not shown). NK cells were incubated in the conditioned media from non-modified or genetically modified B16 cells expressing GIFT2 in the presence or absence of TGFβ neutralizing antibody for 72 h and IL-2Rβ (CD122) expression was determined by flow cytometry. After 72 h, the supernatant was collected and IFNγ production was determined by ELISA. NK cell extracts were immunoblotted using anti-phosphorylated SMAD2, SMAD3, total SMAD2/3 antibodies and anti α-tubulin antibody as loading control.

In vivo blockade of TGFβ 10⁴ genetically modified cytokine secreting and/or non modified B16 cells were mixed with 500 μl Matrigel (BD Biosciences) at 4°C plus 40 μg TGFβ neutralizing antibody or isotype control, and injected subcutaneously in C57Bl/6 mice. Tumor volume was monitored over time and statistic analysis was performed. Similar experiments were carried out using 10 μg of soluble TGFβ receptor II as a TGFβ blocking agent.

Results

MHC class I and II expression in B16 melanoma cells

B16 melanoma derived from C57Bl/6 mice [6, 7] is known as “poorly immunogenic” [15] yet under certain circumstances, it is possible to upregulate MHC I and MHC II expression in these cells in vitro [4]. To verify the immune phenotype of B16 and B16GIFT2 cells here utilized as model systems, we performed flow cytometric analysis for cell surface expression of MHC I, MHC II and the co-stimulatory molecules CD80 and CD86. As shown in supplementary data 1, wild type B16 cells and their B16GIFT2 derivatives do not express detectable levels of these surface proteins, yet robustly express CD105 (endoglin) a co-receptor for TGFβ. This phenotype is consistent with low immunogenicity and is unaffected by expressing GIFT2 fusokine.

Immune bystander effect is lost with increased B16 melanoma tumor burden

To test the effect of tumor burden on immune bystander effect, we admixed B16 melanoma cells with a polyclonal population of GIFT2-secreting B16 cells (hereafter B16GIFT2) at a constant 1:1 ratio. We measured tumor growth in mice having received an initial tumor cell inoculum of either: 1 × 10⁴, 1 × 10⁵ or 1 × 10⁶ of each cell type. As previously reported, Fig. 1a shows that mice implanted with B16GIFT2 melanoma cells remain tumor free long term and mice implanted with B16 cells promptly develop palpable tumors within 20 days. The cohort of immunocompetent mice implanted with 1 × 10⁴ B16GIFT2 cells mixed with 1 × 10⁴ B16 cells displayed the highest percentage of survival and cure (40% of mice), indicating that the paracrine secretion of GIFT2 from B16GIFT2 cells induced a local bystander antitumor response against B16 cells present at the tumor site. However, this bystander effect is lost in a cell dose dependent manner as the number of total B16 cells increases (despite a constant 1:1 mix of B16 and B16GIFT2 cells), with a complete loss of the immune bystander effect with a 1 × 10⁶ cell dose. Figure 1b, c details a replicate in vivo experiment where B16:B16GIFT2 at 1:1 ratio either 1 × 10⁴ (B) or 1 × 10⁶ (C) per cell type were implanted at day 0 and tumor growth monitored over time. Whereas mice receiving solely B16GIFT2 remained tumor-free independently of the size of the day 0 inoculum, virtually all mice receiving 1 × 10⁶ of each B16:B16GIFT2 at 1:1 grew tumors as quickly as B16 controls. Only in the “low tumor burden” group receiving 1 × 10⁴ of each B16:B16GIFT2 at 1:1 had a long term 50% cure rate (P < 0.05 log rank).

Fig. 1.

In vivo immune bystander effect of GIFT2-secreting cells. Kaplan–Meier survival curve of: a cohort of 10 C57Bl/6 mice per each experimental group were injected subcutaneously with (filled circle) 1 × 10⁵ B16GIFT2 cells (positive control), (filled triangle) 1 × 10⁵ B16 cells (negative control), or a mixture of both at 1:1 ratio varying the cell number: (filled square) 1 × 10⁴, (open circle) 1 × 10⁵ and (open cross) 1 × 10⁶ of each cell type. b Cohorts of 10 mice per each experimental group were injected subcutaneously with (filled circle) 1 × 10⁴ B16GIFT2 cells, (filled triangle) 1 × 10⁴ B16 cells and (filled square) 1 × 10⁴ of each admixed cells at 1:1 ratio (2 × 10⁴ total cell number). c A similar experiment was performed using a cohort of 10 mice per group injected with (filled square) 1 × 10⁶ of each admixed cells at 1:1 ratio (2 × 10⁶ total cell number), (filled circle) B16GIFT2 cells and (filled triangle) B16 cells. These experiments were repeated three times with similar results and statistic analysis indicated significant differences between the test groups (P < 0.05 log rank)

Host derived cellular immune response to B16 melanoma

We observed that the bystander effect was operative in implants of 1 × 10⁴ of each B16:B16GIFT2 at 1:1 yet lost in similar implants inoculated at a dose of 1 × 10⁶ cells as shown in Fig. 1c. We speculate that the host-derived cell mediated immune response must be involved in this discrepancy. To analyze the differential response to low (1 × 10⁴) and high (1 × 10⁶) dose implants, we performed an in vivo matrigel cell infiltrate analysis as previously described [21]. In brief, melanoma cells were embedded in matrigel, injected in mice subcutaneously, surgically retrieved 6 days later and enzymatically dissociated to produce a cellular suspension amenable to flow cytometry analysis. We analyzed the recruitment of host-derived CD4⁺, CD8⁺, NK, NKT cells, and CD4⁺CD25⁺ cells to the tumor site. Immune infiltrate analysis of the matrigel plugs of 1 × 10⁴ of each B16:B16GIFT2 at 1:1 admixed cells per implant, exhibited a pattern of cell migration similar to that seen in implants where all B16 cells express GIFT2 (Fig. 2a). There was no significant difference in the proportion of CD4, CD8 or CD4/CD25 cells between B16GIFT2 and B16:B16GIFT2 admixed cells at low and high tumor cell dose. The cohort of mice injected with wild type B16 shows overall a poor recruitment of immune effector cells to the tumor site. However, the cohort of mice implanted with 1 × 10⁶ of each B16:B16GIFT2 at 1:1 admixed cells revealed a significantly reduced NK and NKT cell recruitment to the tumor site (Fig. 2b), suggesting a selective suppression of innate cellular effectors as B16 tumor burden increases and immune bystander effect is lost. We also performed a cell infiltrate analysis in implants containing 1 × 10⁵ of each B16:B16GIFT2 and we observed a pattern of cell migration similar to that seen in 1 × 10⁶ cell dose, although no statistically significant difference was observed (P = 0.08 log rank, supplementary data 2).

Fig. 2.

Immune infiltrated analysis of tumor implants: Immunocompetent C57Bl/6 mice were injected subcutaneously with (filled bars) B16 cells, (open bars) B16GIFT2 cells, (dotted bar) admixed cells and (chequered bar) mouse embryonic fibroblasts (MEF) embedded in matrigel. Implants were retrieved 6 days post implantation and digested with collagenase to collect immune infiltrated cells, which were analyzed by flow cytometry. a Cohorts of six mice were implanted with 1 × 10⁴ cells. b Cohorts of six mice were implanted with 1 × 10⁶ cells. Significant differences were observed of NK cells tumor recruitment between implants contained 1 × 10⁶ of each admixed cells and 1 × 10⁶ B16GIFT2 cells (P < 0.05). These experiments were performed in triplicate with similar results

B16 tumor cells secrete biologically active TGFβ

The loss of the immune bystander effect with tumor cell dose is associated with a decrease of host-derived NK and NKT cells infiltration at tumor site. This observation suggests that a secreted inhibitory factor is released by B16 cells (and their GIFT2 derivatives) which acts as a dominant negative modulator of the immune bystander effect driven by the GIFT2 fusokine. B16 have been previously shown to release latent TGFβ which would be a likely candidate suppressor of immunogene-driven bystander effect [18]. We here determined whether B16 cells produced functionally active TGFβ. An IL-2 dependent cell line (CTLL-2) was used to analyze this immunosuppressive property and was used as a bioassay for active TGFβ. CTLL-2 cells cultured in medium conditioned by B16 cells and increasing doses of recombinant IL-2 showed a significant reduced proliferation in MTT assay compared to the control (medium conditioned by mouse embryonic fibroblasts). The addition of TGFβ neutralizing antibody rescued the ability of these cells to proliferate in response to IL-2 (Fig. 3a). Active TGFβ induces de novo expression of CD8 on CTLL-2 cells and on normal immature thymocytes [8]. Based on this property, we observed the expression of CD8 on CTLL-2 cultured in medium conditioned by B16 cells. Similarly, recombinant TGFβ (1 ng/ml) induced de novo CD8 expression on CTLL-2 cells (Fig. 3b). The expression of CD8 was abrogated with TGFβ neutralizing antibody, indicating a specific property of active TGFβ secreted by B16 cells (Fig. 3c). We analyzed surface expression of components of the interleukin-2 (IL2) receptor complex by CTLL-2 in response to B16 conditioned media. We found that CTLL-2 cells cultured in medium conditioned by B16 cells downregulated the expression of IL-2Rβ after 72 h of incubation and such expression increased significantly after treatment with TGFβ neutralizing antibody (Fig. 3d). The expression of the IL-2Rα on CTLL-2 was not altered by B16 conditioned media (data not shown).

Fig. 3.

B16 cells secrete active TGFβ. a CTLL-2 cells were cultured in the absence of serum with medium conditioned by: (filled circle) B16 cells plus isotype control, (filled diamond) B16 cells plus TGFβ neutralizing antibody and (open square) mouse embryonic fibroblast (MEF). MTT assay was performed to assess the proliferation ability of CTLL-2 cells in the presence of increasing concentration of recombinant IL2. The results plotted represent the average of two independent experiments performed in triplicate. b The ability of active TGFβ of inducing the novo CD8 expression on CTLL-2 was determined by culturing these cells with recombinant TGFβ as control in the presence of recombinant IL2. TGFβ neutralizing antibody was used to assess the specificity of this property. CD8 expression on CTLL-2 was measured by flow cytometry. c CTLL-2 cultured in medium conditioned by B16 cells plus recombinant IL2 with or without TGFβ neutralizing antibody. d IL-2Rβ expression on CTLL-2 cultured with IL2 in medium conditioned by B16 cells with and without TGFβ neutralizing antibody also was assessed. Significant differences (P < 0.05) are indicated

B16 derived TGFβ blocks NK cell recruitment and function

IL-2/IL-2R interaction on NK cells has relevant implications in the expansion and development of NK cells, as well as in the cytotoxicity and cytokine production by these cells. In particular, the common IL-2/IL-15Rβ signaling activates transcription factors involved in the control of perforin expression and secretion of proinflammatory cytokines such as GM-CSF and IFNγ [24]. Since we observed a reduced recruitment of NK cells to the tumor site as B16 tumor burden increases, we evaluated the responsiveness of mouse NK cells to IL-2 by culturing NK cells in medium conditioned by B16 and B16GIFT2 cells. As shown in Fig. 4a, the expression of IL-2Rβ chain was significantly reduced on NK cells cultured in medium conditioned by B16 cells supplemented with recombinant IL-2 (23.5 pmol/ml). Pre-treating medium conditioned by B16 cells with anti-TGFβ neutralizing antibody restored the expression of IL-2Rβ on NK cells to similar levels as the positive control. On the other hand, NK cells cultured in medium conditioned by B16GIFT2 cells containing equimolar concentration of GIFT2, display a significantly greater expression of IL-2Rβ despite the presence of a similar concentration of active TGFβ secreted also by B16GIFT2 cells. This expression was not altered by TGFβ neutralizing antibody pre-treatment. This increase in the expression of IL-2Rβ demonstrated that GIFT2 fusokine released by B16GIFT2 cells overrides the suppressive effects of TGFβ on NK cells. We also evaluated the expression of IL-2Rα and IL-2Rγ on NK cells, and in both cases we detected low expression that was not altered with anti-TGFβ neutralizing antibody (data not shown). Although these cells secreted IFNγ in response to the stimulatory effect of the GIFT2 fusokine, a substantially greater amount of IFNγ was obtained with TGFβ neutralizing antibody pre-treatment (Fig. 4b). We investigated the direct effect of TGFβ on NK cells by analyzing the phosphorylation levels of Smad3 and Smad2 signalling molecules activated by TGFβ. Despite the fact that B16 and B16GIFT2 cells secrete equal amounts of active TGFβ (200 pg/ml), NK cells pre-cultured in medium conditioned by B16 cells show significant higher levels of Smad3 phosphorylation than those observed in the cell extracts of NK cells cultured in medium conditioned by B16GIFT2 cells (Fig. 4c). These results suggest that GIFT2 antagonizes the suppressive effect of TGFβ on NK cells. Regarding Smad2, we observed similar phosphorylation levels in cell extracts of NK cells cultured in medium conditioned by B16 and B16GIFT2 cells (Fig. 4c). Smad3 and Smad2 phosphorylation was completely abolished by anti-TGFβ neutralizing antibody.

Fig. 4.

Blockage of immunosuppressive effects of TGFβ enhances the activation state of NK cells in response to GIFT2. a Percentage of IL-2Rβ expressing NK cells cultured in medium conditioned by: B16 cells supplemented with 23.5 pmol/ml of recombinant IL-2 in the presence or absence of TGFβ neutralizing antibody, RPMI supplemented with 23.5 pmol/ml of recombinant IL-2 in the presence or absence of 1 ng/ml of recombinant active TGFβ, and B16GIFT2 cells containing equimolar concentration of GIFT2. Significant differences are indicated (P < 0.05 log rank). b IFNγ production by NK cells pretreated as described in panel a. c Western blot analysis of the activation status of Smad2 and Smad3 in NK cells pretreated with medium conditioned as described in the figure. Whole NK cell extract were subjected to gel electrophoresis and immunostained with an anti-phospho-Smad3-specific or an anti-phospho-Smad2-specific antibody. Nitrocellulose membrane were stripped and reprobed with anti-αtubulin antibody

In vivo blockade of TGFβ and effect on immune bystander effect

We assessed the contribution of TGFβ’s immune suppressive features, as well as its property to reduce GIFT2’s bystander effect in vivo, by embedding 1 × 10⁴ of each B16:B16GIFT2 at 1:1 ratio admixed in matrigel along with soluble TGFβ receptor II (TβRII) or anti-TGFβ neutralizing antibody. We observed a significant delay of tumor growth in mice injected with admixed cells and sTβRII (Fig. 5a) or TGFβ neutralizing antibody (Fig. 5b), indicating that active TGFβ secreted by B16 cells is an important variable which contributes to the reduction of the immune bystander effect of a potent pro-inflammatory fusokine such as GIFT2.

Fig. 5.

Blockage of TGFβ suppressive effects improved the bystander effect of GIFT2. a Cohorts of 10 C57Bl/6 mice per group were implanted subcutaneously with matrigel plugs containing (filled circle) 1 × 10⁴ B16GIFT2 cells, (filled triangle) 1 × 10⁴ B16 cells, and (open diamond) 1 × 10⁴ of each admixed cells at 1:1 ratio embedded in matrigel alone or (filled square) with soluble TGFβ receptor (TβRII) protein. b Similar experiments were performed using (filled square) TGFβ neutralizing antibody as TGFβ blocking agent. In both cases tumor volume was measure over time and statistic analysis was performed. These experiments were repeated three times and significant differences are indicated (P < 0.05 log rank)

Discussion

B16 melanoma cells implanted in C57Bl/6 mice serves as a robust and popular animal model system to test therapeutic platforms which recruit an immune anti-melanoma response. In previous published work by our group, we utilized this in vivo system to test the utility of a GMCSF and IL2 fusokine (aka GIFT2) and showed that gene transfer of GIFT2 encoding cDNA to B16 melanoma cells could serve as part of a cancer immunogene therapy strategy [21]. However, we had made the observation that as B16 tumor burden was increased the immune stimulation effects of GIFT2 were attenuated. Close reading of reports by other groups exploiting the B16-C57Bl/6 animal model in immunotherapy strategies reveals that an increased B16 tumor burden is apparently antagonistic to the use of immune rejection treatment strategies [13], including graft versus tumor arising from allogeneic marrow transplantation. Cures are readily achieved with small tumor implants, treatment failures abound when even modest increases in tumor cell dose are tested.

We have observed that tumor-associated immunosuppression is more prominent as B16 burden increases. Specifically, the significant reduction in the number of NK and NKT cells recruited at the tumor site observed in the implants of 1 × 10⁶ of each admixed cells (B16:B16GIFT2) compared to the implants comprising equal cell number of B16GIFT2, indicates that tumor derived suppressive factors are antagonistic to the GIFT2 dependent immune bystander effect. Several factors secreted by tumor cells have been described to have deleterious effects on the anti-tumor immune response. One of the most potent immunosuppressive cytokine secreted by tumor cells is transforming growth factor β (TGFβ), which stimulates tumor growth while abolishing effector functions of macrophages, NK cells, CTL and dendritic cells as well as cytokine secretion. Melanoma tumor cells release high levels of TGFβ1, 2 and 3 to the tumor microenvironment. However, these cells are resistant to TGFβ-induced growth inhibition by overexpressing SMAD inhibitors such as Ski and Sno [12]. In concordance with previous studies, we observed that B16 tumor-derived active TGFβ antagonizes the immune stimulatory effects of IL-2 on a murine IL-2 dependent CD8⁺ T cell line (CTLL-2). TGFβ preferentially antagonize IL-2 induced CTLL-2 cell proliferation and not IL-2 induced cell survival in a Smad3 dependent manner. Through the inhibitory activity of Smad3, TGFβ inhibits IL-2 induced expression of promitogenic genes such as c-myc, cyclin D, and cyclin E without affecting the activation status of upstream IL-2 receptor associated tyrosine kinase (JAK1) and signaling molecules Shc and STAT5 [14]. TGFβ1 also inhibits IL-2 induced proliferation in splenocytes and thymocytes through a SMAD3 independent mechanism [11] which may involve suppression Cdc25 expression [3] and inactivation of phosphatase 2A (PP2A) [17].

We also observed a downregulation in the expression of IL-2Rβ on CTLL-2 cultured in medium conditioned by B16 cells, which correlated with a dramatically decreased ability of these cells to proliferate in the presence of IL-2. These results confirm that B16 cells secrete active TGFβ, which antagonizes IL-2- induced proliferation and gene expression in lymphoid cell lines.

B16GIFT2 cells are always totally immune rejected independently of tumor cell dose, and this rejection is driven by GIFT2 mediated recruitment and activation of NK/NKT cells as well as CD8 T-cells [21]. Since we observed suppression of NK cell recruitment at tumor site in vivo, we tested whether B16-derived TGFβ would be a direct cause of this phenomenon. In order to evaluate the responsiveness of NK cells to IL-2 and GIFT2 in the presence of tumor derived TGFβ, we analyzed the expression levels of IL-2 receptor complex subunits. IL-2Rβ plays an important role in NK cell function since IL-2Rβ deficient mice display a reduction of NK1.1⁺CD3⁻ cells in the circulation and a complete absence of NK cytotoxicity activity in vitro [22]. We found that B16-derived TGFβ will markedly suppress IL-2Rβ expression on NK cells despite the presence of high concentrations of recombinant IL-2. In contrast, NK cells cultured in medium conditioned by B16GIFT2 cells, which also secrete similar concentrations of tumor derived TGFβ, display a striking expression of IL-2Rβ. These results suggest that GIFT2 exerts potent immunostimulatory effect on NK cells. Despite the greater immunostimulatory effect of GIFT2 on NK cells, tumor derived TGFβ down modulates IFNγ production by NK cells cultured in medium conditioned by B16GIFT2 cells. Notwithstanding TGFβ antagonism, we observed a substantial production of IFNγ (2 ng/ml) by NK cells cultured in medium conditioned by B16GIFT2 cells, which increased more than two fold by blocking TGFβ with specific antibody. In addition, the phosphorylation status of SMAD3 in this cell extract was significantly reduced despite B16GIFT2 cells secreting equal amounts of active TGFβ as non-modified B16 cells. These results suggest that GIFT2 antagonizes the suppressive effect of TGFβ on NK cells. GIFT2, in addition to directly affecting SMAD3 phophorylation, and therefore the responsive of NK cells to tumor derived TGFβ, also acts as potent chemoattracting factor for NK cells in vivo and promotes the secretion of substantial amount of IFNγ by NK cells. IFNγ inhibits TGFβ-induced phosphorylation of SMAD3 through JAK/STAT pathway, which induces the expression of SMAD7, an antagonist of SMAD. SMAD7 acts by preventing the interaction of SMAD3 with the TGFβ receptor [23]. The suppressive effects of tumor derived TGFβ on NK cells have been reported in several tumor models other than melanoma. For example, breast tumor cell line (MDA-231) derived TGFβ exerts deleterious effects on NK mediated cytotoxicity. Monoclonal antibody inducing blockade of all three mammalian TGFβ isoforms inhibited MDA-231 primary tumor or metastases formation in immunocompetent mice, but this effect was completely abolished in beige NK cell deficient nude mice [1, 2].

TGFβ also promotes tumor invasion and metastases, acting as a pro-angiogenic factor, and induces the breakdown of extracellular matrix. This pleiotropic cytokine is secreted by a variety of cell types as a latent complex unable to bind to TGFβ signaling receptor (TGFR) and initiate signaling transduction. Different mechanisms have been proposed to induce the activation of latent TGFβ such as proteolytic cleavage, enzymatic deglycosylation and/or conformational changes of latent TGFβ complex, which expose the TGFβ receptor binding site [10]. Indeed, TGFβ may serve as a powerful means for B16 to escape immune surveillance but to also actively antagonize pro-inflammatory cancer gene therapy as we show here. A phenomenon which parallels human clinical experience in melanoma therapy. Indeed, recent observations indicate the presence of increased production of suppressive cytokines such as TGFβ and IL-10 in patients bearing more than one invaded lymph node [9].

These observations conciliate an apparent paradox: GIFT2 drives the activation of NK cells––as demonstrated by the upregulation of IL-2Rβ on effector NK cells––whilst contemporaneous production of TGFβ specifically blocks this effect. Reciprocally, GIFT2 can inhibit TGFβ-mediated signaling in NK cells. However, this reciprocal antagonism is skewed in favor of the TGFβ effect, since as tumor burden increases, suppression of NK cells wins out. The effect of GIFT2 can be rescued in part by specific blockade of TGFβ in vivo. We propose the following model: GIFT2 secreted by genetically modified B16 tumor cells, induces a potent chemotactic stimulus on NK cells, which respond to GIFT2 immunostimulatory effect by upregulating the expression of IL-2Rβ on their cell surfaces and producing substantial amounts of IFNγ. This creates a cytokine circuit that recruits other IFNγ-secreting immune cells, as well as GIFT2 responder cells such as macrophages and cytotoxic T cells. All these cytokines cooperate to antagonize the immunosuppressive effects of tumor-derived TGFβ. In contrast, at elevated tumor burden, the amount of active TGFβ in the tumor microenvironment overcomes the immunostimulatory effects of GIFT2 and prevents the recruitment of tumor infiltrating NK cells by blunting the expression of IL-2Rβ on their cell surface and therefore impairing their responsiveness to GIFT2 chemotactic effects, resulting in insufficient IFNγ secretion and recruitment of other immune effector cells.

In summary, eliciting an immune response to melanoma––and other types of cancers as well––will be antagonized by tumor-secreted TGFβ. The resistance of tumor to immune clearance will increase proportionally to cancer burden [19]. Blockade of TGFβ inhibitory effect on cell-mediated immune response will likely be an important component in cancer gene immunotherapy of bulky malignant disease.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

TGFβ

Transforming growth factor beta

IL-2

Interleukin 2

GM-CSF

Granulocyte–macrophage colony-stimulating factor

GIFT2

GM-CSF and IL-2 fusion transgene