Site-specific anti-tumor immunity: Differences in DC function, TGF-β production and numbers of intratumoral Foxp3+ Treg

Authors


Abstract

Gliomas localized within the CNS are generally not rejected by the immune system despite being immunogenic. This failure of the immune system has been associated both with glioma-derived immunosuppressive molecules and the immune-privileged state of the CNS. However, the relative contribution of tumor location to the glioma-mediated immunosuppression, as well as the immune mechanisms involved in the failure of glioma rejection are not fully defined. We report here that syngeneic GL261 gliomas growing either intracranially or subcutaneously in mice are infiltrated by DC and T cells. However, only subcutaneous gliomas elicit an effective anti-tumor immune response. In contrast to DC infiltrating subcutaneously grown GL261 gliomas, tumor-infiltrating DC from intracranial gliomas do not activate antigen-dependent T-cell proliferation in vitro. In addition, brain-localized GL261 gliomas are characterized by significantly higher numbers of Foxp3+ Treg and higher levels of TGF-β1 mRNA and protein expression when compared with GL261 gliomas in the skin. Our data show that gliomas in the CNS, but not in the skin, give rise to TGF-β production and accumulation of both Treg and functionally impaired DC. Thus, not the tumor itself, but its location dictates the efficiency of the anti-tumor immune response.

Introduction

Malignant gliomas, especially the most frequent and aggressive form known as glioblastoma multiforme, are among the most fatal types of tumors. The best standard of care for glioblastoma multiforme, consisting of surgery followed by radiotherapy and chemotherapy with temozolomide, is associated with a median overall survival of 14.6 months following diagnosis 1. Gliomas have been shown to result in reduced peripheral T-cell responses 2, down-modulated function of circulating APC 3 and particularly to suppressed maturation of peripheral DC 4. In addition, the localization of the glioma within the immune-privileged CNS, combined with the production of tumor-derived immunosuppressive molecules including TGF-β 5–7, IL-10 8, VEGF 9 and prostaglandins 10, is suggested to account for the absence of a successful anti-tumor immune response. As a consequence, glioma patients fail to generate an effective immune response against the intracranially growing tumors. Although it occurs rarely, gliomas are able to metastasize to locations outside of the brain 11. The low frequency of peripheral glioma metastases might be due to an efficient recognition of tumor cells outside of the CNS. Gliomas express tumor-associated antigens such as ER-2, gp100, and MAGE-1, which can be recognized by cytotoxic T lymphocyte clones 12. Furthermore, T-cell clonal expansion has been detected in glioma patients 13 and vaccination of patients with autologous tumor lysate-pulsed DC or autologous tumor peptide-pulsed DC 14 can elicit tumor-specific T-cell responses and infiltration of cytotoxic and memory T cells into the tumor. Similarly, in the murine glioma model GL261, vaccination with DC loaded with tumor RNA 15 or tumor extract 16 prolongs the survival of mice bearing intracranial tumors. However, the extent to which the location of the tumor modulates the glioma-mediated immunosuppression is not known. Likewise, the precise mechanisms accounting for the failure of the immune system to reject intracranial gliomas are not fully defined.

DC are highly effective APC that have the ability to initiate primary immune responses but can also induce T-cell tolerance 17. Their role in the immune response against tumors is not yet fully understood. Although DC infiltration has been reported in several tumor types, no absolute correlation between the presence of tumor-infiltrating DC (TIDC) and patient outcome has been found. Indeed, DC infiltrations in lung cancer, gastric cancer, melanoma or papillary thyroid carcinoma 18 are associated with positive outcomes. In contrast, DC infiltration does not influence survival in bronchio-alveolar carcinoma 19. Furthermore, despite the presence of TIDC, immune-mediated tumor rejection is seldom successful.

Tregs represent a subset of peripheral CD4+ cells that mitigate the immune response. Tregs play a key role in the prevention of T-cell responses to self-antigens. However, tumors can take advantage of Tregs by attracting them and manipulating Treg-mediated immunosuppression in order to escape recognition and elimination by the host 20. Their role in tumor progression is highlighted by the observation that systemic depletion of Tregs results in reduced tumor growth and increased survival in several models 21. Although Tregs were recently found within gliomas 22–24, to the best of our knowledge, there has been no report describing the presence of TIDC in gliomas or their function during tumor progression.

This publication thus represents the first report of the phenotypical and functional characterization of TIDC in experimental gliomas. We found that the function of TIDC depends on the site of tumor growth. GL261 gliomas grown in the skin express low levels of TGF-β, harbor DC that activate T cells, and induce a productive anti-tumor immune response. In contrast, intracranial GL261 tumor tissue expresses higher levels of TGF-β, accumulates Tregs, and contains DC, which are unable to stimulate T cells but instead promote the development of Tregs. These findings suggest that gliomas do not unequivocally suppress immune surveillance, but do so only when growing in their natural microenvironment.

Results

Tumor localization influences the anti-glioma immune response

We first sought to determine whether the growth of a glioma outside of the immune-privileged CNS would elicit an immune response different from that induced against an intracranial glioma. To this end, WT (C57BL/6) mice and Rag1-deficient (Rag1−/−) mice, which are devoid of both T and B lymphocytes, were inoculated with GL261 cells either intracranially or subcutaneously, and were monitored daily for tumor appearance. Kaplan–Meier survival analysis of animals injected intracranially with GL261 glioma cells revealed similar tumor growth rates in WT and Rag1−/− mice (Fig. 1A). In contrast, peripheral inoculation with glioma cells resulted in significantly delayed tumor development in WT mice compared with that in Rag1−/− mice (Fig. 1B; p=0.0074). Thus, a glioma grown in the periphery of an immuno-competent mouse induced an immune response resulting in reduced tumor growth, whereas the growth of an intracranial glioma was not retarded in WT mice compared with Rag1−/− mice.

Figure 1.

Difference in growth of tumors from intracranial versus subcutaneous locations. WT type or Rag1−/− mice were injected with GL261 cells either intracranially (A, n=8 for each strain) or subcutaneously (B, n=10 for each strain) and monitored for survival or tumor appearance, respectively. Statistically significant difference was reached for subcutaneous tumors (p=0.0074, Wilcoxon rank test).

Different immune cell infiltration patterns in intracranial and subcutaneous gliomas

Because the anti-tumor response relies on the ability of the cells of the immune system to access the tumor, we investigated whether the difference in tumor immune surveillance was due to a difference in the number of DC and T cells present in intracranial gliomas and subcutaneous gliomas. Intracranial and subcutaneous tumors were isolated at various time points and sections were stained for T-cell and DC markers. CD4+ and CD8+ T cells, as well as CD11c+ cells with typical DC morphology, were found to infiltrate intracranial gliomas at day 9 post-inoculation (data not shown). The number of CD11c+ TIDC further increased until the peak clinical stage, which was reached after 4 wk of tumor growth (Fig. 2A). DC were only found within the tumor, not in the surrounding healthy brain parenchyma (data not shown). Only low numbers of CD4+ and CD8+ T cells were observed within intracranial gliomas 2 wk post-injection (Fig. 2A). In contrast to DC infiltration, T-cell infiltration of intracranial tumors decreased further over time (Fig. 2A and C). In the subcutaneously grown GL261 tumors, DC were present 2 wk after inoculation and were still present after 4 wk of tumor growth, which was the time point at which mice with intracranial tumors had to be sacrificed due to neurological symptoms and weight loss (Fig. 2B). Both CD4+ and CD8+ T cells were detected in subcutaneous tumors 2 wk after inoculation (Fig. 2B). In contrast to the decrease in tumor-infiltrating lymphocytes (TIL) observed in intracranial tumors, the numbers of TIL in peripheral subcutaneous tumors were not decreased after 4 wk of tumor growth (Fig. 2B and C). Interestingly, in both locations, tumor infiltration by DC did not require DC–T-cell interaction, because TIDC were also found in Rag1−/− mice (data not shown). Thus, T cells and DC were able to reach both tumor sites. However, the observed difference in T-cell persistence might explain the difference in tumor growth seen between the intracranial and subcutaneous microenvironments. Hence, we addressed whether differences in TIDC phenotype and function were responsible for the disappearance of T cells from the intracranial tumor.

Figure 2.

Infiltration of intracranial and subcutaneous tumors by immune cells. Cryosections of either intracranial (A) or subcutaneous (B) tumors were stained to detect infiltration by CD11c+ DC, CD4+ T cells and CD8+ T cells during tumor development. Magnification, ×200. (C) Quantitative representation of (A) and (B) depicting the number of stained DC and CD4+ or CD8+ T cells per field during tumor development (p<0.001, Student's t-test). Error bars represent SEM. Three mice per time point were analyzed and ten fields per section were counted.

TIDC from both intracranial and subcutaneous gliomas display an immature myeloid phenotype

Tumors have been observed to interfere with DC maturation and with the up-regulation of proteins involved in antigen presentation and T-cell stimulation 25. Thus, we decided to quantify the expression of MHC class II as well as B7 and CD40 co-stimulatory molecules and the co-inhibitory molecules PD-1, PD-L1 and PD-L2 on TIDC isolated from gliomas that grew in the skin or the brain. We found that TIDC isolated from intracranial tumors expressed intermediate levels of MHC class II molecules, but only low levels of B7.1. B7.2 and CD40 co-stimulatory molecules were barely detectable on these cells (Fig. 3A). This expression pattern suggests an immature phenotype similar to that of in vitro-generated immature BM-derived DC (BMDC) (Fig. 3A and Table 1). BMDC matured with LPS expressed much higher levels of MHC class II, B7.1, B7.2 and CD40 than did TDIC from intracranial gliomas (Fig. 3A). No differences between intracranial and subcutaneous TIDC were found in the expression of the co-inhibitory molecules PD-1, PD-L1 or PD-L2 (Supporting Information Fig. 1). TIDC isolated from gliomas in the skin showed an immature phenotype similar to that of both TIDC from intracranial gliomas and immature BMDC (Fig. 3A and Table 1). The weak expression of B7.1 and the absence of B7.2 was confirmed by immunohistochemistry performed on frozen brain sections (data not shown), indicating that the phenotype of the TIDC was not influenced by the isolation method.

Figure 3.

Phenotypic characterization of TIDC. (A) TIDC display a phenotype similar to immature BMDC. Tumor-infiltrating mononuclear cells were isolated at peak clinical stage as described in the Materials and methods and then stained for CD11c in combination with antibodies to MHC class II or co-stimulatory molecules. 7-AAD was used to exclude dead cells. Histograms are gated on live CD11c+ DC (for TIDC: black line, specific staining; gray line, isotype control. For BMDC: black line, mature BMDC; dark gray line, immature BMDC; light gray line, isotype control). Representative histograms of at least three experiments are shown. (B) Both intracranial and subcutaneous TIDC are of myeloid origin. Histograms are gated on live CD11c+ DC (black line, specific staining; gray line, isotype control). Data are representative of at least three experiments. Numbers indicate the mean fluorescence intensity of the marker analyzed. (C) TIDC isolated from intracranial gliomas as well as TIDC isolated from gliomas in the skin are blood-derived. Tumor cells were inoculated in CD45.2+ mice that were lethally irradiated and reconstituted with BM cells from CD45.1+-congenic C57BL6/J 6 wk previous to tumor inoculation. Tumor-infiltrating mononuclear cells were stained with anti-CD11c, anti-CD45.1 and anti-CD45.2 antibodies. Dot blots are gated on live cells. Numbers in quadrants indicate percentage of positive cells. The data are representative of eight mice with intracranial tumors and three mice with subcutaneous tumors.

Table 1. Summary of the activation status of the TIDC and the CD4+ or CD8+ TIL in intracranial and subcutaneous gliomasa)
  IntracranialSubcutaneous
  • a)

    a) Data represent the mean±SEM of flow cytometry analyses shown in Figs. 3 and 5 from at least three different mice. MFI for TIDC, percent positive for TIL.

TIDCMHCII69.4±17.282.7±7.2
 B7.113.8±16.529.7±15.0
 B7.24.9±3.14.5±1.9
 CD407.0±7.96.9±4.0
CD4+ TILCD6932.3±1021.0±9.1
 CD62L18.5±9.016.2±16.7
 CD2548.5±9.042.8±15.9
 CD4492.0±4.892.0±10.3
CD8+ TILCD6933.0±8.635.8±16.5
 CD62L35.5±8.520.2±17.7
 CD257.0±3.46.6±2.5
 CD4486.8±7.189.8±16.1

We also did not find phenotypic differences between intracranial and subcutaneous TIDC when analyzing other characteristic DC surface molecules (Fig. 3B). MHC class I was expressed to a similar extent by both populations of infiltrating DC. The finding that both DC types stained positively for CD11b and negatively for CD8 (Fig. 3B) strongly suggests that they are of myeloid origin. Notably, the fact that no CD4 or CD19 staining was detectable in the TIDC samples confirmed that no T cells or B cells contaminated the DC populations. Furthermore, if plasmacytoid DC or myeloid suppressor cells were present, they would have constituted only a marginal fraction of TIDC, because staining for Gr-1 and B220 was low-negative (Supporting Information Fig. 2).

It has been shown that DC recruited to the CNS during autoimmune or infectious diseases may originate from microglial cells 26, 27. In order to address whether intracranial TIDC originate from CNS-resident precursors and would thus differ from subcutaneous TIDC in their origin, intracranial and subcutaneous TIDC from CD45-congenic BM chimeric mice were studied. These experiments clearly showed that virtually all TIDC (>95%) in CNS gliomas were blood-derived and thus did not originate from microglia (Fig. 3C). Moreover, the data suggested that at both tumor sites, TIDC originated from the same blood-derived precursors.

The capacity of infiltrating DC to stimulate T cells depends on the site of tumor growth

Because we found no differences between intracranial and subcutaneous TIDC with respect to the expression of cell-surface molecules and activation markers, we aimed to characterize the functional properties of these cells. First, we examined the T-cell priming capacity of TIDC in an allogeneic proliferation assay. In vitro-generated, LPS-matured BMDC and DC isolated from subcutaneous tumors were able to induce a strong proliferative response by allogeneic CD4+ T cells (Fig. 4A). In contrast, intracranial TIDC failed to induce allogeneic T-cell proliferation, even at a high APC:T ratio (Fig. 4A). Because TIDC were previously reported to exert immunosuppressive functions on T cells 25, 28, we tested the suppressive potential of purified TIDC in a proliferation inhibition assay (Fig. 4B). Constant numbers of mature BMDC and allogeneic CD4+ T cells were cultured at a ratio of 1:10 (baseline stimulation index set to 100%). To evaluate the respective potential of each type of TIDC to enhance or inhibit the reference level of T-cell proliferation induced by the mature BMDC, increasing numbers of TIDC or mature BMDC (as control cells) were added. As expected from the proliferation assay, subcutaneous TIDC, as well as control mature BMDC, further increased T-cell proliferation compared with baseline (Fig. 4B). Intracranial TIDC, on the other hand, neither increased nor inhibited baseline proliferation (Fig. 4B). Thus, intracranial TIDC failed to act as stimulators in an allogeneic T-cell proliferation system and did not suppress mature BMDC-induced proliferation. We then confirmed this difference in TIDC function using an antigen-specific system. We found that OVA-specific TCR-transgenic OT-II responder T cells proliferated in response to OVA323–339 peptide (Fig. 4C) or whole OVA protein (Fig. 4D) in the presence of subcutaneous TIDC, but they did not exhibit an antigen-specific response in the presence of intracranial TIDC. In summary, in contrast to subcutaneous TIDC, intracranial TIDC were not able to induce allogeneic or syngeneic T-cell proliferation. Thus, despite similarities in surface marker expression, TIDC isolated from the same tumor grown at different anatomical sites exhibited drastically different stimulatory capacities.

Figure 4.

Unlike subcutaneous TIDC, TIDC isolated from intracranial tumors fail to induce allogeneic or antigen-specific T-cell proliferation. TIDC from intracranial or subcutaneous tumors were sorted by flow cytometry at peak clinical stage. As a control, mature BMDC were used. (A) For allogeneic T-cell proliferation, 105 CD4+ T cells from BALB/c mice were added to each well of titrated DC and the proliferation was measured by 3H-thymidine incorporation. (B) For proliferation inhibition assays, 105 CD4+ T cells from BALB/c mice, as well as 104 mature BMDC from C57BL/6 mice, were added to each well. Then TIDC from intracranial or subcutaneous gliomas, or alternatively BMDC as control cells, were added. The T-cell proliferation stimulated by the mature BMDC in the absence of TIDC is set to 100%. (C and D) For antigen-specific proliferation assays, 105 OT-II T cells were added with either OVA323–339 peptide (C), or OVA protein (D). (*p<0.05; **p<0.01; ***p<0.005 comparing intracranial with subcutaneous TIDCs, Student's t-test). Experiments were performed at least three times.

No differences in activation markers on intratumoral T cells

In order to determine whether the differences in T-cell infiltration observed at the peak clinical stage were paralleled by changes in the activation state of the TIL, we determined the phenotype of TIL from both tumor locations. CD4+ and CD8+ TIL were stained for L-selectin (CD62L), the very early activation antigen CD69, the adhesion receptor CD44 and the IL-2 receptor alpha chain CD25 (Fig. 5 and Table 1). Most of the TIL had down-regulated CD62L, except for a minority of the CD8+ T cells. CD69+ cells were detected to a similar extent in both the CD4+ and CD8+ TIL populations from both anatomical locations. All CD4+ and CD8+ T cells expressed CD44, indicating the presence of effector-memory (both activated and long-lived) T cells in both tumor locations. For all markers analyzed, no statistically significant differences were found between TIL isolated from subcutaneous versus intracranial gliomas. Strikingly, whereas only few CD8+ TIL expressed CD25 (<10%), more than 40% of the CD4+ TIL did, regardless of tumor location. This suggested that the CD4+ TIL contained a population of either activated effector or Treg. To find out whether tumor-infiltrating T cells, like DC, showed a functional difference when both tumor locations were compared, we monitored the capacity of isolated tumor-infiltrating T cells to produce IFN-γ upon re-stimulation with PMA and ionomycin. Intracellular cytokine staining showed that subcutaneous tumor-infiltrating T cells secreted higher amounts of IFN-γ compared with intracranial tumor-infiltrating T cells, albeit not reaching statistical significance (data not shown). Furthermore, we found no difference in the expression of FasL by CD8+ tumor-infiltrating T cells from subcutaneous versus intracranial tumors (data not shown).

Figure 5.

TIL from intracranial or subcutaneous GL261 gliomas show similar activation marker expression. (A) At peak clinical stage (days 25–28), tumor-infiltrating mononuclear cells were stained with either anti-CD4 antibody or (B) anti-CD8 antibody in combination with antibodies directed against T-cell activation markers. Dot plots are gated on live CD4+ T cells (A) or live CD8+ T cells (B) respectively, and numbers indicate the percentage of positive cells in the respective quadrant. One representative experiment out of four is shown.

Tregs and TGF-β expression are prominent in intracranial gliomas, but not in subcutaneous gliomas

Because our flow cytometric analysis of TIL showed a large population of CD4+ T cells expressing CD25 (Fig. 5A), we aimed to define the proportion of Tregs present in each tumor type. Staining for the transcription factor Foxp3, a specific murine Treg marker, revealed that intracranial TIL contained a significantly higher proportion of Tregs than did the TIL from subcutaneous tumors (Fig. 6A and B). This difference was only seen in the TIL population and not in the peripheral T-cell population, because similar percentages of Tregs were detected in the blood and spleens of animals bearing intracranial or subcutaneous tumors. Notably, intracranial Tregs proved to be functionally suppressive in vitro (Fig. 6D), to the same extent as splenic Tregs isolated from the same animals (Fig. 6C). Moreover, in vivo depletion of CD25+ Tregs prior to intracranial tumor inoculation significantly prolonged the mean survival (data not shown) as previously published 22, 23.

Figure 6.

More Tregs and increased levels of TGF-β are found in intracranial gliomas. (A) CD4+ T cells from spleen, blood and tumors were stained for Foxp3 expression. Numbers indicate the percentage of positive cells in the respective quadrant. (B) Quantification of (A). Increased Treg levels are found in intracranial (n=4) compared with subcutaneous (n=6) tumors (*p<0.005; Student's t-test), whereas no difference is seen in the blood or in the spleen of tumor-bearing animals (intracranial, n=4; subcutaneous, n=5). (C and D) Tregs isolated from intracranial tumors are functionally suppressive. Tn, CD4+CD25- T cells; Treg, CD4+CD25+ T cells. The data are normalized to the proliferation of 105 Tn. The data in (C) and (D) are representative of three experiments each. (E) Increased expression of TGF-β1 and TGF-β2 mRNA is found in intracranial tumors. Total RNA was isolated from intracranial or subcutaneous GL261 tumors and cytokine expression was quantified by real-time PCR. Results represent mean±SEM from three mice each (*p<0.05; **p<0.01, Student's t-test). (F) TGF-β1 protein was assayed in extracts of subcutaneous and intracerebral gliomas by ELISA. The data are shown as pg TGF-β1/mg of total protein. n=11 (i.c.) and 7 (s.c.). (G) TIDC from the brain induce more Tregs in vitro than subcutaneous TIDC. The bar graph shows the relative induction of Tregs by TIDC normalized to intracerebral TIDC (=2.8% Foxp3+ among CD4+ cells). The data show the mean of three experiments with SEM. p(no DC–i.c. DC)=0.035; p(i.c. DC–s.c. DC)=0.002, Student's t-test. All analyses were done at peak clinical stage (days 25–28).

Because CD25+Foxp3+ Tregs can be generated upon exposure of naive T cells to TGF-β and also secrete TGF-β themselves to suppress T-cell priming and T-cell effector mechanisms 29, we examined the expression of TGF-β1 and TGF-β2 in GL261 tumors resected from the two inoculation sites. Both TGF-β1 and TGF-β2 isoforms were expressed at higher levels by intracranial tumors compared with subcutaneous tumors (Fig. 6E). In contrast to TGF-β, no significant differences in the expression of IL-4, IL-6 or IL-10 were detected in gliomas from the brain or the skin (Fig. 6E). Furthermore, increased levels of TGF-β1 protein were detected in intracranial tumors (Fig. 6F). Collectively, these data showed increased TGF-β expression in GL261 tumors growing within the brain, its natural environment, as compared with tumors growing in peripheral sites.

DC from intracranial gliomas are more efficient inducers of Tregs than are subcutaneous TIDC

Because intracranial TIDC were neither able to induce T-cell priming nor inhibited T-cell proliferation triggered by stimulatory DC (Fig. 4), we assessed whether they had the potential to induce Treg cells from naïve CD25CD4+ T cells. CD25-depleted naive CD4+ T cells were cultured for 5 days in the presence or absence of either intracranial or subcutaneous TIDC. Intracellular Foxp3 staining revealed the induction of significantly more Tregs in the presence of intracranial TIDC compared with subcutaneous TIDC or no DC (Fig. 6G).

Discussion

The most aggressive malignant gliomas are generally associated with a median overall survival of around 1 year following diagnosis. This has been attributed to their growth within the immune-privileged CNS, but in addition, gliomas have been shown to release immunosuppressive molecules, including TGF-β. Here, we provide evidence that the lethality of gliomas is not only a function of the tumor itself, but also of the specific habitation of the glioma in its “natural” environment, the CNS. We show that the intracranial glioma GL261 is infiltrated by DC, which are not able to prime T cells, whereas the same tumor growing in the skin is infiltrated by immuno-competent DC (Figs. 2 and 4). DC infiltration has been reported in several tumor types, such as lung tumors, colon carcinoma, gastric cancer and papillary carcinoma 18, but as evidenced by the still devastating effects of cancer, their presence does not necessarily correlate with a positive outcome for patients. Generally, DC found in tumors display an immature phenotype 30, 31. This is in line with our findings in gliomas. Both intracranial and subcutaneous TIDC express only intermediate levels of MHC class II, low levels of B7.1, and almost no B7.2 or CD40 (Fig. 3A). Moreover, no difference was found in the production of IL-12p35, IL-12p40, IL-23p19, IL-10, IL-1β, IL-6 or TNF-α by TIDC isolated from subcutaneous or intracranial TIDC (data not shown). However, we have clearly shown that glioma TIDC isolated from the two sites differ functionally. TIDC from gliomas of the skin are able to induce T-cell proliferation, whereas TIDC harvested from gliomas of the brain fail to do so (Fig. 4). Furthermore, intracranial TIDC induce higher numbers of Treg cells than do subcutaneous TIDC (Fig. 6G).

To the best of our knowledge, this is the first report presenting functional data on glioma-infiltrating DC. The function of CNS-derived DC has been analyzed in murine models of autoimmune and infectious diseases, with a variety of functions observed for these DC, ranging from inhibition of T-cell activation 32 to promotion of disease spreading 33. Likewise, TIDC in non-glioma tumors have been described as playing several roles in the anti-tumor response. In hepatocarcinoma and colon carcinoma, functional impairment of TIDC can be reversed by IL-10 neutralization combined with CpG stimulation 34, 35. In contrast, melanoma-derived TIDC are potent APC and can mediate tumor rejection without prior re-stimulation 36. Our results add to the complexity by demonstrating site-specific differences in DC function.

The non-stimulatory, Treg-inducing DC in the GL261 intracranial gliomas may impair the anti-glioma T-cell response and thereby inhibit tumor immune surveillance. It is unlikely that other types of cells compensate for the lack of APC function of TIDC. Glioma cells work poorly as APC due to a low expression of MHC molecules and a lack of co-stimulatory molecules, which are needed to reactivate the primed T cells 37. Microglia, which have been described to function as APC in vitro38, are unable to present glioma cell antigens to cytotoxic T lymphocytes 39. Moreover, using congenic BM chimeras, we did not see any involvement of microglia in the anti-tumor response. The absence of priming functions by TIDC in brain gliomas strongly suggests that they are unable to reactivate glioma-infiltrating lymphocytes in situ, which could explain the reduction in T-cell numbers observed during the course of tumor development. Indeed, T-cell activation is critically important for the infiltration of T cells deep into the tumor tissue 40, and T cells which do not re-encounter their antigen disappear from the CNS, either via migration back to the periphery 41 or via apoptosis 42. Interestingly, in glioblastoma patients, invading T cells have been reported to undergo apoptosis 43. The re-activation of T cells within the tumor is not only required for survival and maintenance of cytotoxic activity, but also for the attraction of other effector cells such as neutrophils, macrophages and NK cells 44. Thus, the lack of stimulatory DC in intracranial gliomas is very likely to have a broad negative impact on the anti-tumor response.

DC that are unable to prime normal T cells have the potential to induce suppressive Treg 45. This is in line with both our detection of increased Foxp3-expressing Treg within intracranial tumors compared with the subcutaneous tumors (Fig. 6A and B) and with the fact that intracranial TIDC are superior to subcutaneous TIDC in inducing Treg cells in the absence of exogenous TGF-β (Fig. 6G). The presence of Tregs in human glioblastomas has recently been reported 24 and the function of TIDC in human glioblastomas is under investigation.

BM chimera experiments allowed us to determine the origin of the myeloid DC detected in gliomas. In experimental infectious and autoimmune diseases, subpopulations of CNS-DC were shown to be derived from microglia 26, 27. This is in contrast to our observations, in which we found that almost all the DC in intracranial GL261 tumors are derived from donor BM cells (Fig. 3C). Thus, the glioma-infiltrating DC migrate from the blood into the tumor parenchyma in the CNS and the functional differences between subcutaneous and intracranial TIDC cannot be attributed to different cellular origins.

The increased TGF-β present in intracranially grown gliomas compared with gliomas grown in the skin could contribute to the difference in immune responses observed between tumors from these two locations (Fig. 6E and F). TGF-β1 and TGF-β2 have pleiotropic functions (for review see 46) and have been described to be expressed very strongly by glioma cells 5–7. The importance of TGF-β in supporting glioma growth is attested by experiments blocking TGF-β production, which positively influence the survival of rats with established gliomas 47. TGF-β regulates T-cell proliferation and T helper cell differentiation and it interferes with the generation of cytotoxic T cells 48, an effect which is central in tumor surveillance because CD8+ T cells are necessary for efficient anti-tumor responses. Moreover, TGF-β converts CD4+CD25 T cells into functional Tregs 29 and blocks the maturation of DC, which then display diminished T-cell activation potential.

Based on the site-specific differences observed when analyzing experimental GL261 gliomas in the CNS and the skin, and taking into account the present knowledge of DC function, Tregs and TGF-β, we can conclude from our results that the high level of TGF-β in intracranial gliomas leads to accumulation of both Tregs and immature DC with the potential to induce Tregs. This milieu prevents T-cell priming and re-stimulation, and interferes with T-cell effector function, resulting in an impaired anti-tumor immune response. We initially suspected that the programmed death-signaling pathway involving the molecules PD-1, PD-L1 and PD-L2 could be involved in the pronounced Treg induction in intracranial gliomas. However, we found no differences in the expression of these molecules on the respective DC (Supporting Information Fig. 1). The reasons for this site-specific anti-tumor response are still unknown, but are likely to involve a role of the surrounding tissue and/or the tumor vasculature. Indeed, astrocytes, the type of glial cells which encircle intracranial gliomas, can produce significant amounts of VEGF as well as matrix metalloproteinases and may thus promote tumor growth and invasion 49. Moreover, neurons can promote Treg induction via TGF-β secretion, and thus may be at least in part responsible for the relative increase of Tregs observed in intracranial gliomas 50. We show here that a glioma residing in the CNS is not simply ignored by the immune system, and that the immuno-modulatory capacity of the GL261 glioma is not purely an inherent property of this tumor. However, the fatal nature of brain tumors might be due to the production of immuno-modulatory factors, including TGF-β, which are potentiated when the glioma grows within the CNS. Our results suggest that future research should target the increased Treg infiltration, impaired TIDC function and high TGF-β production associated with intracranial gliomas in the development of new therapies or in the improvement of existing treatments.

Materials and methods

Cell lines and mice

The GL261 glioma cell line, originally induced by implantation of 3-methylcholanthrene pellets in the brain of a C57BL/6 mouse, was kindly provided by Dr. Paul R. Walker (Geneva University Hospital, Switzerland) and grown in DMEM containing 4500 mg/L D-glucose (Invitrogen, Basel, Switzerland) supplemented with 10% FCS, 2 mM N-acetyl-L-alanyl-L-glutamine (Biochrom AG, Berlin, Germany) and 20 μg/mL gentamycin (Invitrogen). C57BL/6 (H-2b) and BALB/c (H-2d) mice were purchased from Harlan (Horst, The Netherlands) or RCC (Füllinsdorf, Switzerland). B6.SJL-Ptprca Pep3b/BoyJ and Rag1−/− (H-2b) mice were kindly provided by Prof. Burkhard Becher (University of Zurich, Switzerland). OT-II mice (H-2b) were kindly provided by Prof. Hans Acha-Orbea (University of Lausanne, Switzerland). BM chimeras were generated by irradiation with 1100 rad (split dose) and i.v. reconstitution with 5 Mio. BM cells.

Reagents

Antibodies to CD4 (RM4-5), CD8α (53-6.7), CD11b (M1/70), CD25 (PC61), CD40 (3/23), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD62L (MEL-14), CD69 (H1.2F3), CD80 (16-10A1), CD86 (GL1), MHC I (AF6-88.5) and MHC II (2G9) were purchased from BD Biosciences (Allschwil, Switzerland); anti-Foxp3 (FJK-16s) and anti-PD-1, PDL1 and PD-L2 were from e-Bioscience (San Diego, CA); anti-CD11c (N418) was from Caltag (Burlingame, CA). 7-AAD was from Sigma (Buchs, Switzerland). rmGM-CSF was purchased from Immunotools (Friesoythe, Germany), OVA protein was purchased from Sigma, and OVA323–339 peptide from NeoMPS (Strasbourg, France).

GL261 glioma tumor model

A total of 5×104 GL261 cells were inoculated either in the frontal cortex (2 mm lateral to bregma and 2 mm deep) in 10 μL PBS using a 27G syringe or in 2 μL using a Hamilton syringe, or subcutaneously in the right flank of anesthetized C57BL/6 or Rag1−/−- males. Tumor appearance was checked daily by weight measurement and assessment of neurological symptoms associated with an intracranial tumor. Endpoint for intracranial tumor was determined by a weight loss of more than 15% of maximal weight (=peak clinical stage reached 25–28 days after tumor inoculation). Mice injected subcutaneously were considered tumor free as long as no tumor could be detected by palpation, and these mice had to be sacrificed when the tumor volume was >1 cm3. All animal experiments were performed in compliance with the law and approved by the Veterinary Office of the Canton of Zurich.

Histology

Brains and subcutaneous tumors were embedded in Jung tissue freezing medium (Leica Instruments GmbH, Nussloch, Germany) and frozen on a metal plate chilled by liquid nitrogen. Tissue sections of 6 μm thickness were cut in a cryostat and fixed in 2% paraformaldehyde/PBS for 5 min. Thereafter, sections were incubated with antibodies against CD4, CD8, CD11c or the respective isotype control for 1 h. Binding of primary antibodies was revealed by the following means: primary rat anti-murine CD4 and CD8 antibodies were detected by the alkaline phosphatase anti-alkaline phosphatase system (DAKO, Zug, Switzerland) and hamster anti-murine CD11c antibodies by goat anti-hamster followed by donkey anti-goat alkaline phosphatase antibodies (Jackson ImmunoResearch, West Growe, PA). Color development was performed with Fast Red (DAKO) and counterstaining with Mayer's Hemalaun.

Preparation of tumor-infiltrating mononuclear cells

Intracranial and subcutaneous tumors from animals perfused with Ringer solution (Braun Medical, Emmenbrücke, Switzerland) were minced with a scalpel blade and digested for 30 min at 37°C in HBSS containing 50 μg/mL DNase I and 0.1 mg/mL Collagenase/Dispase (both from Roche, Rotkreuz, Switzerland). The digestion was quenched on ice after addition of 10% FCS. The digested tissue was passed through a 100 μm nylon mesh (Falcon, BD Biosciences). The suspension was pelleted, resuspended in 30% Percoll (GE Healthcare, Uppsala, Sweden) and underlayed with 70% Percoll in HBSS. The gradient was centrifuged at 500 g for 20 min at RT without brakes. Thereafter, if applicable, the myelin was removed by aspiration, and the interphase and about 80% of the 30% Percoll was collected, and cells were washed with ice cold HBSS and stained for FACS analysis or sorting.

Flow cytometry

Tumor-infiltrating mononuclear cells, resuspended in PBS containing 2% FCS, 10 mM EDTA and 0.01% NaN3, were first incubated with Fc-block (BD Biosciences) and subsequently with specific antibodies. For Foxp3 intracellular staining, mononuclear cells were enriched after the Percoll gradient by subsequent purification over a Lympholyte M (Cedarlane Labs, Burlington, Canada) gradient according to manufacturer's protocol. Then the cells were stained for CD4 and subsequently intracellularly stained for Foxp3 according to the manufacturer's instructions. All analyses were done on a CyFlow Space (Partec, Münster, Germany). Post-acquisition analysis was done with WinMDI 2.8 software (Scripps-Research Institute, La Jolla, CA).

For flow cytometric sorting, tumor-infiltrating mononuclear cells were stained in sterile PBS containing 2% FCS and 10 mM EDTA. TIDCs, CD4+ CD25 T cells and CD4+ CD25+ Tregs were sorted simultaneously on a FACSAria (BD Biosciences). Dead cells were excluded by 7-AAD staining.

Proliferation assays

Proliferation and inhibition assays were performed as described previously 32. Briefly, CD4+ T cells were isolated from BALB/c or OT-II spleens using anti-mouse CD4-biotin (BD Bioscience) and CELLection biotin binder beads (Dynal/Invitrogen, Basel, Switzerland). For proliferation assays, 105 T cells per well were cultured in a 96-well U-bottom plate (Falcon, BD) in the presence of titrated numbers of various DC preparations in 200 μL X-VIVO 20 medium (BioWhittaker, Walkersville, MD), supplemented with 2 mM acetyl-alanyl-glutamine, 1× MEM nonessential amino acids (Invitrogen), 1 mM sodium pyruvate (MP Biomedicals, Solon, OH), 50 mM β-mercaptoethanol (Sigma) and 25 μg/mL gentamycin. Mature BMDC were generated by cultivation of BM cells in GM-CSF and induction of maturation by LPS as previously described 32. For the inhibition assays, 105 CD4+ T cells and 104 mature BMDC were cultured in the presence of titrating numbers of various DC preparations. Trypan blue staining before co-cultivation of the cells for proliferation experiments ensured that only viable sorted cells were used. For all assays, DC were irradiated (2000 rad) before incubation with T cells. After 2 days, proliferation was assessed by measurement of 3H-thymidine incorporation. For Treg assays, 105 CD4+CD25- T cells (Tn) from either spleens or tumors were cultured on plates coated with anti-CD3 (3 μg/ml) in the presence or absence of 5 x 104 Tn or CD4+CD25+ Treg cells isolated from the same site. After three days, the T cell proliferation was assessed by 3H-thymidine incorporation.

In vitro Treg induction

Naive splenic CD4+ T cells were CD25-depleted by MACS (Miltenyi Biotech, Bergisch Gladbach, Germany). CD4+CD25 T cells were then cultured for 5 days in the presence or absence of flow cytometry-sorted CD11c+ DC from either intracranial or subcutaneous tumors at a ratio of 10:1 (T cell:TIDC). The development of Treg cells was determined by intracellular staining for Foxp3 and flow cytometry analysis.

Real-time PCR

Total RNA was prepared from resected intracranial or subcutaneous tumor tissue using Trizol (Invitrogen) according to manufacturer's instructions. cDNA was synthesized from total RNA using random hexamers and M-MuLV reverse transcriptase (Roche). Quantitative PCR was performed using an ABI PRISM 7700 detection system (Applied Biosystems, Foster City, CA) and the TaqMan Universal Master Mix (Applied Biosystems). Primer–probe sets were purchased from Applied Biosystems. The relative levels of each RNA were calculated by the 2−ΔΔCT method, with 18s rRNA used as housekeeping gene control. Results were normalized with respect to the internal control, and expressed relative to the levels found in intracranial GL261.

TGF-β1 ELISA

Resected tumors were homogenized in isotonic buffer. TGF-β1 in the homogenates was measured by a commercial ELISA (R&D Systems, Minneapolis, MN). TGF-β1 levels were normalized to total protein content that was determined by BCA (Pierce/Thermo Scientific, Rockford, IL).

Statistical analysis

Survival analysis was computed by the Kaplan–Meier method. Significances, by Wilcoxon rank test (for survival) and t-tests, were calculated using StatView (SAS Institute, Cary, NC).

Acknowledgements

The authors would like to thank Eva Niederer and Oralea Büchi (Institute for Biomedical Techniques, ETH Zürich), Hans-Rudolf Müller and Seppoei Muljana for technical assistance. The authors further thank Jan A. Ehses for critical reading of the manuscript. This study was supported by the Swiss National Science Foundation (grant 31-61136.00), the National Competence Centre (NCCR) Neuronal Plasticity and Repair, the Roche Research Foundation (to C.C.) and the Gianni Rubatto Stiftung.

Conflict of interest: The authors declare no financial or commercial conflict of interest.

Ancillary