Background
Gastric cancer is a common disease in industrial countries and is associated with a poor prognosis. Over 50 percent of potentially curatively operated gastric cancer patients relapse within 5 years. Subsequent chemo- or radiation therapy is mostly insufficient [
1]. Therefore, the development of new adjuvant treatments with a favorable "therapeutic index", (i.e., good tolerability and demonstrated anti-tumor activity), are desperately needed. Active-specific immunotherapy (i.e., therapeutic vaccination) may represent such an option.
Active-specific immunotherapy aims to improve the patient's ability to mount a therapeutic immune response against cancer. Nevertheless, inducing an immune response against the tumor is by itself not sufficient, and clinical results with cancer vaccines have been sobering [
2], even though the first therapeutic vaccine based on autologous dendritic cells (DCs) called Provenge (sipuleucel-T, Dendreon Corp., Seattle, WA, USA) was recently approved for the treatment of hormone refractory prostate cancer [
3]. Few vaccination studies in patients with gastric cancer have been published, which demonstrated antibody responses or peptide-specific IFN-γ responses and cytotoxicity by isolated cytotoxic T cells, but did not show strong clinical responses [
4‐
6].
To increase the frequency of circulating tumor-specific T cells is likely to be one important minimal requirement for a successful therapy [
7]. To obtain sufficient expansion of such lymphocytes, several therapeutic strategies have been adopted, including prior lymphodepleting, non-myeloablative chemotherapy with cyclophosphamide followed by reconstitution of the lymphocyte pool by infusion of autologous immune cells [
8‐
10]. Lymphopenia naturally induces a proliferative response to maintain homeostasis [
11,
12]. This stimulates antigen-specific T cells directed towards antigens contained in the tumor vaccine. In preclinical models of melanoma, this strategy increased the frequency of tumor-specific T cells in tumor vaccine-draining lymph nodes (TVDLN) extensively and enhanced the therapeutic efficacy of active-specific and adoptive immunotherapy strategies [
13‐
15]. In addition to lymphopenia-induced proliferation, the elimination of regulatory T cells (Treg) and the creation of a beneficial host microenvironment by affecting components of the innate immune system are alternatively proposed as immunomodulatory effects of preparative chemotherapy with e.g. cyclophosphamide [
16‐
18].
A recently introduced strategy to increase the therapeutic efficacy of tumor vaccination is to combine different immunological approaches, i) applying multifaceted antigen vaccines to target a broad spectrum of tumor antigens, ii) providing co-stimulation, iii) reducing or eliminating suppressive immune cells, e.g. Tregs [
7], and iv) blocking tumor-induced immune suppression mediated by e.g. TGF-β [
19]. Such a multifactorial vaccination approach may be especially suitable for tumor entities that exhibit a low immunogenicity, as has been described for gastric cancer [
20]. Only a few tumor-associated antigens, mostly so-called cancer testis antigens, have been identified to be expressed in gastric tumors [
21‐
23], but this has not yet resulted in successful therapeutic approaches targeting these antigens [
24].
In order to explore novel therapeutic vaccination strategies for gastric cancer, we have established cell lines from the spontaneously growing gastric tumors of CEA424-SV40 TAg transgenic mice [
25,
26]. In the current study, we aimed to enhance the therapeutic anti-tumor immunity in a subcutaneous mouse model of gastric cancer by (i) combining a low immunogenic whole tumor cell vaccine (prepared from the established gastric cell lines) with granulocyte macrophage colony-stimulating factor (GM-CSF) to stimulate local antigen presentation and by (ii) pretreatment with cyclophosphamide to enhance proliferation of tumor-specific T cells and to reduce the frequency of Tregs. Here, we show that lymphodepletion by preparative treatment with cyclophosphamide followed by reconstitution with naïve spleen cells enhances the anti-tumor immunity induced by a whole cell vaccine. This treatment strategy, LRAST, induced a long-term anti-tumor immune response against subsequent tumor challenge and tended to slow down growth of established tumors. GM-CSF significantly reinforced the tumor-specific immune response induced by the tumor vaccine. Furthermore, we observed a transient reduction of Tregs, supporting the priming of a tumor-specific immune response.
Discussion
Several reports have shown that active-specific tumor vaccination administered to a lymphopenic host may result in significantly enhanced anti-tumor immune responses [
8,
13]. Meanwhile, this study design has been translated into early phase clinical trials for several tumor entities [
7,
9]. However, there are neither preclinical nor clinical studies that address this therapeutic strategy in gastric cancer. The goal of active-specific tumor vaccination is to induce a systemic tumor-specific immune response especially against low- or non-immunogenic tumors. The aim of this study was to increase the therapeutic efficacy of a vaccination with the low immunogenic gastric tumor cell line mGC8. Consistent with previous reports on other tumor entities [
8,
15,
39], we demonstrate here for the first time that the treatment with cyclophosphamide prior to tumor vaccination in the presence of GM-CSF can efficiently induce long-term protection against subcutaneous tumor growth in a gastric cancer model.
In earlier publications, tumor cell lines genetically modified to secrete GM-CSF or other immunostimulatory cytokines were compared with regard to their effectiveness as a cancer vaccine [
37,
40]. GM-CSF-secreting tumor vaccines appeared to be most potent to induce long-lasting tumor-specific immunity and have been used in clinical studies [
41,
42]. Due to the presence of GM-CSF at the vaccine site, antigen-presenting cells (APC) are recruited, activated and capable of activating tumor-specific T cells in the vaccine-draining lymph nodes [
33,
37]. A future aim of our immunotherapeutic approaches is to use autologous tumor samples for vaccination instead of cell lines. Since gene transfer into freshly derived tumor cells is laborious and may not be very efficient [
43], we aimed to apply GM-CSF separately to the tumor cells. The easiest way to do this would be the co-administration of recombinant GM-CSF to the irradiated tumor cells. However, this would require frequent applications of the cytokine due to its short half-life
in vivo [
44], and would probably yield less potent anti-tumor responses compared to GM-CSF secreting cells [
33,
45]. Approaches that encapsulate or modify GM-CSF to provide sustained release locally at the vaccine site have been shown to result in anti-tumor immune responses comparable to that of GM-CSF-secreting tumor cells [
44,
46]. In addition, emulsions with IFA have been described to induce a strong and long-term immune response and were suggested to be stable for a few weeks [
47,
48]. Therefore, we emulsified GM-CSF in IFA and we applied the emulsion subcutaneously at the vaccine site in order to enhance the immune response. Indeed, we found that application of emulsified GM-CSF, but not IFA alone, during vaccination increased the induction of tumor-specific T cells as measured by tumor-specific IFN-γ release from TVDLN cells. In addition, mice vaccinated with irradiated tumor cells in the presence of GM-CSF/IFA showed a significant enhancement of tumor-free survival as compared to lymphodepleted mice treated with GM-CSF/IFA without the tumor vaccine. This indicates the necessity of the presence of tumor antigens for successful LRAST treatment.
While low doses of GM-CSF as an adjuvant have been described to increase vaccine-induced immune responses (reviewed in [
49]), in our model the induction of a long-term therapeutic immune response
in vivo resulted only from the combination of cyclophosphamide treatment with GM-CSF application and not from GM-CSF alone. This emphasizes the expected potency of lymphodepletion applied prior to vaccination to enhance the therapeutic efficacy of a vaccination.
Unexpectedly, application of cyclophosphamide and reconstitution with naïve syngeneic splenocytes prior to the tumor vaccination with GM-CSF (LRAST) did not further increase but rather tended to decrease the tumor-specific immune response in vitro as determined by tumor-specific IFN-γ secretion and specific lysis of mGC8 tumor cells by TVDLN cells. This discrepancy between in vitro and in vivo observations may in part be explained by the fact that significantly less T cells could be recovered from TVDLN following LRAST as compared to TVDLN from other treatment groups. It is conceivable that the remaining LN cells may be more sensitive towards further handling than LN cells that were not affected by cyclophosphamide and that therefore the results do not reflect in vivo CTL activity in our setting. On the other hand, the in vivo CTL response may be influenced by other mechanisms, e.g. Treg, which do not necessarily have an inhibitory effect when studying CTL activity in vitro. Since the mGC8 GM-CSF/IFA-treated group shows a higher number of Treg than the LRAST group, it is conceivable the in vivo anti-tumor response is suppressed in the former group.
At least two mechanisms have been proposed for the positive effect of cyclophosphamide pre-treatment on tumor vaccination: (i) increased homeostatic expansion of antigen-specific T cells in a lymphopenic environment and (ii) depletion of regulatory T cells. We addressed the first mechanism by analyzing the tumor-specific cytokine release in T cells isolated from TVDLN 9 days after vaccination. TVDLN cells from LRAST-treated and LP mGC8 IFA-treated mice tended to secrete increased levels of tumor-specific IFN-γ compared with TVDLN cells from control mice. Considering the enhancement of anti-tumor immunity after the LRAST treatment, one may anticipate that an augmented secretion of IFN-γ reflects an increase in the number of tumor-specific T cells in the LRAST-treated mice. However, alternatively an increase in cytokine expression per cell may have occurred as well. A preliminary ELISpot analysis suggested that TVDLN from LRAST-treated mice had both a larger number of IFN-γ producing cells and released more tumor specific IFN-γ per cell as compared to control mice (not shown).
Several studies have reported on a depletion of Tregs as another mechanism to explain the beneficial effect of cyclophosphamide treatment [
8,
16‐
18]. Tregs are known to efficiently down-modulate immune responses and depletion of these cells has been shown to enhance the anti-tumor immune response in various tumor models [
50,
51]. Consistent with other reports, we observed a rapid decline in white blood cells one day after a single i.p. application of cyclophosphamide and a gradual recovery of the cell numbers during the following week [
32]. Although the absolute numbers of lymphocytes in the peripheral blood normalized after 9 days (Additional file
1, Figure S1), the frequency and the absolute number of FoxP3
+ CD25
+ CD4
+ Treg cells were decreased in the spleen of LRAST-treated mice as compared to vaccinated mice without lymphodepletion (Figure
5A and
5B). This is consistent with previous findings that describe a transient reduction of Tregs in the spleens of mice in the first 10 days after cyclophosphamide (100 mg/kg) treatment [
16]. In that study, in addition to a reduction of CD4
+CD25
+ cells after cyclophosphamide treatment, a loss of
FoxP3 and
GITR gene expression as well as a reduction of Treg function was reported. In our experiments, the decline in the number of Tregs, the increase in the ratio of CD8
+ T cells to FoxP3
+ CD25
+ CD4
+ Tregs and the lymphopenic environment after cyclophosphamide treatment favor enhanced priming of tumor-specific immune responses during vaccination. This is consistent with the efficacy of the LRAST treatment against s.c. tumor growth
in vivo (Figure
2). The precise role of Treg in the induction of anti-tumor immunity is subject of planned investigations in our laboratory and will be analyzed by depletion of Treg from the cell population used for reconstitution as well as by adoptive transfer of Treg after cyclophosphamide treatment.
In the experiments using a therapeutic setting we aimed to boost the tumor-specific immune response by giving repeated vaccinations. Although some mice in the LRAST group showed benefit by displaying a delayed tumor growth, the mean growth was not significantly different from the group without cyclophosphamide treatment. We observed that approximately two months after LRAST treatment, the proportion of FoxP3
+CD25
+CD4
+ T cells had increased again to the frequencies of the other treatment groups without lymphodepletion. Thus, it seems that an initial decrease in Tregs after vaccination was followed by a secondary "induction" of Tregs. Interestingly, we also observed higher numbers of FoxP3
+CD25
+CD4
+ T cells in mice that showed a long-term protective response after LRAST (data not shown). Therefore, we assume that a later increase of Tregs does not necessarily affect the anticancer effect of the treatment. It remains to be determined whether late appearance of Tregs actually has an impact on the therapeutic efficacy of the overall anti-tumor response. A recent study reported that the use of multiple vaccinations had a negative effect on the generation of therapeutic effector T cells [
52]. The authors showed that multiple vaccinations increased the absolute number of CD4
+Foxp3
+ Tregs in the peripheral blood and in the spleens, which decreased the therapeutic efficacy of splenocytes when adoptively transferred into tumor-bearing mice. In support of these results, we have recently observed that repeated vaccination with irradiated autologous tumor vaccines did not maintain a long-term reduction of Foxp3
+ Tregs in the peripheral blood of non-small cell lung cancer patients after lymphodepleting chemotherapy (Van den Engel et al., manuscript in preparation).
Consistent with a previous report [
18], we detected high numbers of CD11b
+Gr1
+ cells in the spleen 9 days after pretreatment with cyclophosphamide. This increase in Gr1
+CD11b
+ cells in cyclophosphamide-treated mice suggests the presence of myeloid-derived suppressor cells that could limit the immune response, as has been suggested in several reports [
53,
54]. In contrast, other reports suggest a beneficial effect through inhibition of tumor growth by the MDSC [
18,
55]. It remains to be determined whether these cells have inhibitory influence on the immune response that is elicited by LRAST.
Recently, a related s.c. gastric cancer mouse model was used to test the therapeutic efficacy of a dendritic cell vaccine loaded with irradiated gastric tumor cells in combination with CpG oligonucleotides [
56]. In that study, tumor cells from the cell line mGC3 were used as the antigen source in the DC vaccine. The cell lines mGC3 and mGC8 were established from CEA424-SV40 TAg tumors and both cell lines display similar expression levels of epithelial cell surface markers, MHC class I molecules and the large-T antigen [
25], which suggests that they may exhibit comparable therapeutic potential. Indeed, prophylactic vaccination with the DC vaccine improved survival in wild type mice injected with mGC3 tumor cells and caused long-term protection, similarly to our results with LRAST using the cell line mGC8. However, neither active immunization using the DC tumor cell vaccine nor adoptive transfer of tumor-reactive splenocytes did change survival of transgenic CEA424-SV40 TAg mice developing spontaneous gastric tumors, suggesting immunological tolerance toward multiple tumor-associated epitopes in these mice [
56]. Correspondingly, we did not see a survival benefit in CEA424-SV40 TAg mice treated with LRAST in a pilot experiment (not shown). Therefore, we support the view that developing an immunotherapy, which is clinically effective in these transgenic mice will be challenging and will require additional immune-activating approaches, for example by inactivating cells that suppress immune responses.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
NKE and HW designed the animal experiments. DR provided support, discussed the data and reviewed the manuscript. NKE and MR planned and conducted the experiments and discussed the data. NKE coordinated the study and drafted the manuscript. HW discussed the data and reviewed the manuscript. RK established the cell lines and participated in coordination and design of the initial experiments. WZ participated in design of initial experiments and reviewed the manuscript. RH directed the laboratory where the studies were performed, participated in experimental design and obtained support for the project. All authors read and approved the final manuscript.