Dynamics of T-cell infiltration during the course of ovarian cancer: The gradual shift from a Th17 effector cell response to a predominant infiltration by regulatory T-cells

Authors


Abstract

The type of immune cells that are present within the tumor microenvironment can play a crucial role in the survival of patients. However, little is known about the dynamics of the tumor-infiltrating immune cells during disease progression. We studied the immune cells that infiltrated the tumor tissues of ovarian cancer patients at different stages of disease. The early stages of development of ovarian carcinomas were characterized by a strong Th17 immune response, whereas in stage II patients, recruitment of high numbers of Th1 cells was observed. In disseminated tumors (Stages III–IV), we detected a dominant population of Helios+ activated regulatory T cells (Tregs) along with high numbers of monocytes/macrophages and myeloid dendritic cells (mDCs). Tumor-infiltrating Tregs had markedly lower expression of CCR4 than circulating Tregs, and the numbers of tumor-infiltrating Tregs significantly correlated with the levels of CCL22 in ovarian tumor cell culture supernatants, suggesting their recruitment via a CCR4/CCL22 interaction. CCL22 was mainly produced by tumor cells, monocytes/macrophages and mDCs in the primary ovarian tumors, and its expression markedly increased in response to IFNγ. Taken together, the specific recruitment of Tregs, probably triggered by inflammatory stimuli, leads to a significant immune suppression in the advanced stages of ovarian cancer.

Introduction

Ovarian cancer is one of the 10 most common malignancies in females worldwide and is known to have the highest mortality rate among gynecologic cancers.1 Because of the lack of sensitive and specific biomarkers and due to the fact that the disease tends to develop and spread rapidly, almost 70% of patients are diagnosed at an advanced stage of tumor dissemination with poor prognosis. Although conventional treatment leads to a significant reduction in malignant cell numbers in more than 80% of ovarian carcinomas, most of the patients experience an eventually lethal relapse of the disease within 2–5 years because of the persistence of a small number of chemotherapy-resistant tumor cells.2 To improve the prognosis of ovarian cancer patients, there is a need for superior prognostic markers and novel therapeutic strategies that can identify high-risk patients and diminish the likelihood of relapse, respectively.

Immune surveillance has been proposed to play a crucial role in cancer development and progression. Experiments in immunodeficient mice have shown that the immune system is able to recognize and eradicate tumors.3–5 However, although cancer cells can prime tumor-specific immune responses, interactions between the tumor and the host immune system may not result in clinical regression of the disease but instead to the development of an immunosuppressive microenvironment within the tumor tissue, consequently promoting immune evasion.6

Despite being unable to eradicate established tumors, the presence of certain tumor-infiltrating immune cells can represent strong prognostic markers in cancer patients. Zhang et al.7 has reported that high numbers of tumor-infiltrating CD3+ T cells are associated with a favorable clinical outcome in advanced ovarian cancer patients. More recent studies have reported that improved survival is associated with enhanced numbers of CD8+ cytotoxic T cells.8, 9 By contrast, high numbers of CD4+CD25+FoxP3+ regulatory T cells (Tregs),10 plasmacytoid dendritic cells,11 and B7-H4+ macrophages12 within the tumor tissue can indicate a poor prognosis. The role of CD4+ T helper cells is less well documented; however, there is strong evidence that Th17 cells may be substantial players in cancer immunity. Proinflammatory Th17 cells have primarily been associated with autoimmune diseases and mucosal immunity.13 Further studies have demonstrated that Th17 cells are present in different types of tumors, including ovarian cancer, head and neck cancer, gastric cancer, breast cancer, colorectal cancer and prostate cancer.14–19 Although extensively studied, the exact role of Th17 cells in tumor immunity and the survival of patients remains controversial. On the one hand, IL-17-producing cells have been reported to promote antitumor immunity,20–22 but on the other hand, IL-17 is known to act as an angiogenic factor that can enhance tumor growth.23–25

Most published studies have focused on the pattern and prognostic significance of different types of tumor-infiltrating immune cells; however, little is known about the dynamics of the immune response within the tumor tissue during disease progression. In this study, we evaluated the distribution, phenotype and clinico-pathological significance of CD4+ and CD8+ T lymphocytes, dendritic cells (DCs) and monocytes/macrophages in different stages of epithelial ovarian cancer (EOC). Our findings demonstrate an alteration in the tumor-infiltrating immune cell pattern during cancer development and clearly show a dynamic shift from an effector Th17 cell infiltration, which prevail over Tregs in the early stages of EOC, to a significant Treg accumulation in the advanced stages of disease.

Abbreviations

DC, dendritic cell; EOC, epithelial ovarian cancer; FIGO, International Federation of Gynecology and Obstetrics; mDC: myeloid dendritic cell; PBMC, peripheral blood mononuclear cells; Treg: regulatory T cell

Material and Methods

Patients and tissue samples

Peripheral blood and primary epithelial ovarian cancer specimens were obtained from 44 patients undergoing initial cytoreductive surgery at the University Hospital Motol in Prague between March 2009 and October 2011. None of the patients enrolled in the study had received neoadjuvant chemotherapy prior to the surgery. All tissue specimens were collected with patient consent, and the study was approved by the Institutional Review Board of the University Hospital Motol. The clinico-pathological characteristics of the patients are summarized in Table 1.

Table 1. Clinicopathological characteristics of the EOC patients in the study
inline image

Tumor tissue was minced with scissors, digested in PBS containing 1 mg/ml of Collagenase D (Roche, Basel, Switzerland) at 37°C for 30 min, mechanically dissociated using the gentleMACS™ Dissociator (Miltenyi Biotec, Auburn, CA) and passed through a 100 μm nylon cell strainer (BD Biosciences, Franklin Lakes, NJ). Peripheral blood mononuclear cells (PBMCs) from blood samples were isolated using Ficoll-Paque density gradient solution (GE Healthcare, Waukesha, WI).

Cell lines

The ovarian cancer cell lines, OV-90 and SKOV3, were cultured in RPMI 1640 supplemented with 10% heat-inactivated FCS, L-glutamine and penicillin-streptomycin (all from Invitrogen, Carlsbad, CA) at 37°C and 5% CO2.

Flow cytometry analysis

To detect tumor infiltrating myeloid DCs (mDCs), plasmacytoid DCs (pDCs) and monocytes/macrophages, single cell suspensions were stained with specific antibodies against CD3, CD11c, CD14, CD16, CD19, CD20, CD45, CD56 (Exbio, Vestec, Czech Republic), CD123 (eBioscience, San Diego, CA) and HLA-DR (BD Biosciences). Regulatory T cells (Tregs) were identified by surface staining with anti-CD3 (Exbio), anti-CD4 (eBioscience), anti-CD8 (Exbio), anti-CD25, anti-CD127 and anti-CCR4 (BioLegend) antibodies followed by fixation and permeabilization with a Foxp3 Staining Buffer Set (eBioscience) and intracellular staining with anti-FoxP3 (eBioscience) and anti-Helios (BioLegend) antibodies. For detection of intracellular cytokines, cell suspensions were stimulated with 50 ng/ml of PMA and 1 μg/ml of ionomycin (Sigma-Aldrich, St. Louis, MO) for 4 hr in the presence of brefeldin A (BioLegend). After the 4-hr incubation, cells were stained with antibodies against CD3 (Exbio), CD4 (eBioscience), and CD8 (Exbio), fixed and permeabilized with the Foxp3 Staining Buffer Set (eBioscience) and stained with anti-IL-17 (BioLegend) and anti-IFNγ (BD Biosciences) antibodies. For CCL22 detection, cells were cultured in the presence of brefeldin for 4 hr and labeled with the antibodies used for DC/macrophage identification, EpCAM (BioLegend) and CCL22 (R&D, Minneapolis, MN) as described above. Samples were analyzed on a BD FACSAria (BD Biosciences) using FlowJo software (TreeStar, Ashland, OR).

Cytokine and chemokine detection

Tumor tissue-derived cell suspensions (1 x 106 cells/ml) were cultured in RPMI 1640 supplemented with 10% heat-inactivated FCS, L-glutamine and penicillin-streptomycin (Invitrogen) in the presence or absence of PMA + ionomycin. For IFNγ-mediated induction of chemokines, ovarian cancer cell lines and primary tumor tissue-derived cell suspensions (1 x 106 cells/ml) were cultured in the presence of 10, 50 and 100 ng/ml of recombinant human IFNγ (Invitrogen). After 24 hr of incubation, culture supernatants were harvested and stored at −80°C until use. Concentrations of IL-1β, IL-2, IL-4, IL-6, IL-7, IL-10, IL-12(p70), IL-13, IL-17a, IL-21, IFNγ and TNFα released into the culture supernatants were determined using a MILLIPLEX™ Human Cytokine/Chemokine Kit (Millipore, Billerica, MA). Concentrations of chemokines (CCL2, CCL5, CCL17, CCL19, CCL20, CCL21, CCL22, CXCL9 and CXCL10) were analyzed using a Quantibody® Array Kit (Raybiotech, Norcross, GA).

Extraction of genomic DNA and quantitative real time PCR-based methylation assay

Genomic DNA (gDNA) was isolated from a lysate containing 2 x 106 tumor tissue-derived cells using the PureLink Genomic DNA Mini kit (Invitrogen, Carlsbad, USA). Concentrations of extracted gDNA were measured with a Nanodrop© 2000c UV-Vis spectrophotometer (Thermo Scientific, Waltham, USA). Four to five micrograms of gDNA was treated with sodium bisulfite using the MethylCode Bisulfite Conversion kit (Invitrogen). Quantitative real time PCR amplification of the methylated and demethylated FoxP3 Treg-specific demethylated region (TSDR) was performed using bisulfite-treated gDNA, 0.5 U Platinum Taq DNA polymerase (Invitrogen), PCR buffer without MgCl2, 3 mM MgCl2, 0.2 mM dNTPs (each), 1 μM primers and 0.2 μM TaqMan© probe. PCR reactions were performed using a CFX 96 cycler (BioRad, Hercules, USA). Methylation-specific and demethylation-specific primers and TaqMan© probes were commercially synthesized (TIB MOLBIOL, Berlin, Germany). Amounts of methylated and demethylated FoxP3 TSDR DNA were estimated from calibration curves by crossing points linear regression using the second derivative maximum method as described by Rasmussen et al.26 The proportion of cells with a demethylated FoxP3 TSDR was calculated as the ratio of demethylated FoxP3 TSDR DNA to the sum of methylated and demethylated FoxP3 TSDR DNA.27, 28

Plasmid constructs

DNA fragments of the methylated and demethylated FoxP3 TSDR (86 bp) were cloned into the pUC18 plasmid (Amersham Biosciences, Amersham, UK) using HindIII and EcoRI restriction sites. The identity of the plasmid constructs was verified by sequencing followed by plasmid DNA purification using a Wizard Plus Midipreps kit (Promega, Madison, USA) and dilution to seven final concentrations of 100, 10, and 1 pg/μl and 100, 10, 1 and 0.1 fg/μl, representing a range of 3.34 x 10-7 to 3.34 x 10-1 plasmid copies. Serial dilutions demonstrated the specificity of the assay and were used as standards for quantification of the methylated and demethylated FoxP3 TSDR.

RNA extraction and quantitative real time PCR

Total RNA was extracted from 2 x 106 tumor-tissue derived cells using an RNA Easy Mini Kit (Qiagen, Hilden, Germany). RNA concentrations were determined with a Nanodrop© 2000c UV-Vis spectrophotometer (Thermo Scientific), and RNA integrity was assessed using an Experion automated system (BioRad). Complementary DNA was synthesized from total RNA using the M-MLV reverse transcriptase (Invitrogen) and amplified by quantitative real time PCR in the presence of primers and TaqMan© probes specific for TGFβ and the β-actin housekeeping gene, which was used as an internal reference. All primers and probes were commercially synthesized (TIB MOLBIOL). The identity of qPCR products in each assay was verified by sequencing. The relative expression of the target genes was normalized to the expression of β-actin.

Statistical analysis

Statistical analyses were performed using Statistica® 10.0 software (StatSoft, Tulsa, OK). The parametric assumptions of the data were verified using the Kolmogorov-Smirnov test for normality. The homogeneity of variances was tested by the Levene test. Correlations between tumor-infiltrating immune cells were evaluated using the Pearson r coefficient. The differences between tumor tissue and peripheral blood samples were analyzed using a paired t-test. The remaining data were analyzed using an ANOVA followed by a Scheffé test. The results were considered statistically significant when p < 0.05.

Results

Th17 lymphocytes accumulate in the tumor tissue and their proportions negatively correlate with disease progression

Th17 lymphocytes were detected in 100% of the analyzed ovarian tumor samples (range, 0.06–22.8%). The proportion of Th17 cells within the CD4+ T cell subset was significantly higher in the tumor tissues obtained from patients with limited disease (FIGO Stage I) than in those from patients suffering from disseminated disease (FIGO Stages III–IV) (Figs. 1a and 1b). The frequency of Th17 cells within the whole tumor-derived single cell suspensions showed the same pattern (data not shown), indicating that the absolute numbers of Th17 cells significantly decreased in the advanced disease. In accord with these findings, the concentration of IL-17 and IL-21 in tumor tissue-derived cell culture supernatants of early stage patients was markedly higher in both unstimulated (data not shown) and PMA- and ionomycin-stimulated (Fig. 1c) cultures. Despite the significant decrease in Th17 cell proportions in the later stages of the disease, the percentage of tumor-infiltrating Th17 cells was still significantly higher when compared with the peripheral blood of patients (Fig. 1d). The frequency of Th17 cells in the peripheral blood was not predictive of the number or proportion of Th17 cells in the tumor tissue.

Figure 1.

Proportions of Th17 cells within the primary ovarian tumor tissue and peripheral blood according to the stage of the disease. (a) Data are expressed as the proportion of Th17 lymphocytes among CD4+ T cells, — represents the mean value in each stage. (b) Dot plots are gated on CD3+CD4+ cells and show IL-17 and IFNγ in two representative patients. (c) Columns represent mean levels of IL-17 and IL-21 produced by PMA + ionomycin stimulated tumor tissue-derived cells + S.E.M. (d) Columns represent mean proportions of Th17 cells in the tumor tissue and PBMC + S.E.M. *p < 0.05, **p < 0.01 (A, ANOVA followed by Scheffé test; D, paired t-test).

The proportion of tumor-infiltrating Tregs is enhanced in the tumor tissue and increases with disease progression

In contrast to Th17 cells, the proportion of CD4+CD25+CD127FoxP3+ Tregs was significantly higher in the tumor tissues obtained from patients with disseminated disease (FIGO Stages III–IV) when compared with the samples from patients with Stage I disease (Figs. 2a and 2b). The proportion of tumor-infiltrating Tregs (Ti-Tregs) was significantly higher than that found in the peripheral blood of patients in all stages of the disease (Fig. 2c). To better characterize the status of Ti-Tregs, we included an analysis of the transcription factor, Helios, which is a marker of Treg activation. The proportion of Helios+ Tregs was significantly higher in the tumor tissue (67.8 ± 2.5%) than in the peripheral blood (56.3 ± 4.2%) of EOC patients (Fig. 2d). There was no difference in the proportions of Helios+ Tregs in the different stages of disease. The increasing proportion of stable Ti-Tregs during disease progression was further confirmed using a quantitative real time PCR-based methylation assay, which assessed the percentage of stable Tregs that were demethylated at the FoxP3 TSDR in tumor tissue-derived single cell suspensions (Fig. 2e). The proportions of Tregs and Th17 lymphocytes were inversely correlated in the same tumors (r = −0.44, p < 0.01) (Fig. 2f), whereas the proportion of Tregs positively correlated with the number of monocytes/macrophages (r = 0.47, p < 0.01) (Fig. 2g).

Figure 2.

Proportions of Tregs within the primary ovarian tumor tissue and peripheral blood according to the stage of the disease. (a) Data are expressed as the proportion of CD4+CD25+CD127-FoxP3+ Tregs among CD4+ T cells, — represents the mean value in each stage. (b) Dot plots are gated on CD3+CD4+ cells and show CD25+FoxP3+ Tregs in two representative patients. (c) Columns represent mean proportions of Tregs in the tumor tissue and PBMC + S.E.M. (d) Dot plots are gated on CD3+CD4+ CD25+CD127 cells and show FoxP3+Helios+ Tregs in the tumor tissue of a representative patient. Box plots represent the proportion of Helios+ Tregs among CD4+CD25+CD127FoxP3+ Tregs. The boundaries of the box indicate the 25th and the 75th percentiles, the line within the box represents the median. Whiskers indicate the 90th and 10th percentiles. (e) Columns represent proportions of Tregs demethylated in FoxP3 TSDR among cells isolated from the primary ovarian tumor tissue in different stages of the disease, error bars indicate S.E.M. (f) The line shows negative correlation between % of tumor-infiltrating Tregs and Th17 cells. (g) The line shows positive correlation between % of tumor-infiltrating Tregs and the number of tumor-infiltrating monocytes/macrophages. *p < 0.05, **p < 0.01 (a, ANOVA followed by Scheffé test; c, d, paired t-test).

Proportions of Th1 and CD8+ T lymphocytes, DCs and monocytes/macrophages do not show significant changes during disease progression

In addition to Th17 cells and Tregs, we also analyzed the presence of other immune cell subsets, including Th1 cells, CD8+ T cells, pDCs, mDCs and monocytes/macrophages, in the tumor tissue. Although there was a trend toward an increase in mDC and monocyte/macrophage proportions in the advanced tumors (Fig. 3), none of these cell types showed any statistically significant dynamics during ovarian cancer progression.

Figure 3.

Proportions of Th1 lymphocytes and monocytes/macrophages within the primary ovarian tumor tissue according to the stage of the disease. (a) Data are expressed as the mean proportion of CD3+CD4+IFNγ+ Th1 cells among CD4+ T cells + S.E.M. (b) Columns represent the mean number of CD45+LineageHLA-DR+CD14+ monocytes/macrophages in 1 × 106 of isolated tumor-derived cells + S.E.M.

Cytokine and chemokine profiles of tumor-derived single cell suspensions support the recruitment of polarized Th lymphocytes into the tumor tissue rather than their in situ priming and polarization

To better characterize the tumor microenvironment and to evaluate, whether the cytokine profile supports in situ priming and polarization of T lymphocytes, we assessed the cytokine and chemokine profile in the tumor-derived cell culture supernatants. In the unstimulated culture supernatants, only three cytokines were detected in significant amounts, IL-6, IL-10 and TNFα (Fig. 4a). IL-2, IL-4, IL-12 and IL-23 levels were below the limit of detection of the MILLIPLEX™ assay in most of the patient samples tested, and IL-17 was only detected in cell supernatants of patients in Stage I of disease. Importantly, only minimal concentrations of IL-1β (4.2 ± 2.5 pg/ml), which is crucial for Th17 polarization, were detected in culture supernatants from Stage I patients, who had the highest proportions of tumor-infiltrating Th17 cells (Fig. 4c). Except for IL-17, no other significant correlations with disease progression were observed in the unstimulated cultures. Upon PMA and ionomycin stimulation, most of the tested cytokines were produced, except IL-12 and IL-23 (Fig. 4a). The production of IL-21 was significantly decreased in culture supernatants from patients in the advanced stages of the disease when compared with culture supernatants from Stage I patients (Fig. 1c). The expression of TGFβ mRNA was observed in all of the patient samples and did not significantly correlate with disease stage (Fig. 4d).

Figure 4.

Cytokine and chemokine profiles of tumor-derived single cell suspensions. (a) White columns represent the mean spontaneous cytokine production; black columns represent cytokine production upon PMA and ionomycin stimulation. (b) Columns represent mean spontaneous chemokine production. All error bars indicate S.E.M. (c) The plot expresses the dynamics of the mean spontaneous production of Th17 polarizing cytokines and IL-17 during the disease progression. (d) The plot indicates the dynamics of TGFβ mRNA expression during the disease progression presented as the mean proportion of β-actin mRNA expression.

Unstimulated tumor-derived cells produced high amounts of CCL20, CCL22 and inflammatory CXCL9 and CXCL10 (Fig. 4b) but low amounts of CCL2, CCL5, CCL17, CCL19 and CCL21. The levels of CCL22 (Fig. 5a), CXCL9 and CXCL10 (not shown) increased during disease progression; however, these data were not statistically significant. The levels of CXCL9 significantly correlated with the numbers of tumor-infiltrating CD4+ (r = 0.65, p < 0.01) and CD8+ (r = 0.64, p < 0.01) T lymphocytes. There was no correlation with a particular Th lymphocyte subset (not shown). Most important, the levels of CCL22 significantly correlated with the number of Ti-Tregs (r = 0.66, p < 0.01) (Fig. 5b). No other correlations between the chemokines and tumor-infiltrating immune cells were observed.

Figure 5.

Recruitment of Tregs into the ovarian tumor tissue is probably CCL22-mediated. (a) Columns represent CCL22 production in tumor tissue derived cell culture supernatants + S.E.M. (b) The scatter plot expresses the positive correlation between CCL22 production and number of Ti-Tregs. (c) Histograms show the CCR4 expression in CD4+CD25+CD127-FoxP3+ Tregs in the tumor tissue (black line) and peripheral blood (gray line, tinted) in a representative patient. Columns represent the mean fluorescence intensity of CCR4 on CD4+CD25+CD127FoxP3+ Tregs in the tumor tissue and peripheral blood + S.E.M. (n = 6). (d) Columns represent the mean proportion of proliferating Ki67+CD4+CD25+CD127FoxP3+ Tregs and CD4+FoxP3 conventional T cells (Tconv) in the tumor tissue and peripheral blood + S.E.M. (n = 6). (e) Dot plots are gated on CD3+CD4+ cells and show the proportion of Ki67+ FoxP3+ Tregs and FoxP3 Tconv in a representative Stage III patient. *p < 0.05 (paired t-test).

Tregs are probably recruited to the tumor tissue via a CCL22/CCR4 interaction and proliferate in situ

As described above, the number of Ti-Tregs significantly correlated with the level of CCL22 in the tumor cell culture supernatants. In accord with these observations, Ti-Tregs had significantly lower expression of CCR4 than Tregs from the peripheral blood (mean fluorescence intensity 6,280 ± 35 and 3,487 ± 333, respectively; n = 6) (Fig. 5c). To evaluate whether Ti- Tregs proliferated in situ, Ki67 staining was performed in tumor tissue-derived cell suspensions and PBMCs. A markedly higher percentage of Ti-Tregs expressed Ki67 when compared with Tregs in the peripheral blood (21.8 ± 5.1% and 12.9 ± 1.8%, respectively; n = 6). Ki67 was expressed in a significantly lower proportion of conventional CD4+FoxP3 T cells in comparison with Tregs (7.6 ± 2.7% in the tumor tissue and 2.3 ± 0.4% in the peripheral blood) (Figs. 5d and 5e).

Ovarian cancer cell lines produce chemokines, including CCL22, upon IFNγ stimulation

To assess whether ovarian cancer cell lines were able to produce CCL22, OV-90 and SKOV3 cells were cultured either without stimulation or upon stimulation with recombinant human IFNγ, as IFNγ has been shown to induce CCL22 production in breast cancer cell lines.29 In contrast to primary tumor-derived cell suspensions, spontaneous CCL22 secretion was undetectable in the ovarian cancer cell lines, a result that was similar to most of the other chemokines measured in this study. The SKOV3 cell line spontaneously produced small amounts of CXCL9 (Fig. 6a), whereas unstimulated OV-90 cells produced high amounts of CXCL9, CXCL10 and CCL20 (Fig. 6b). The addition of IFNγ induced robust secretion of CCL19, CCL21 and CCL22 in both of the cell lines tested. Surprisingly, production of CXCL9, CXCL10 and CCL20 increased in SKOV3 but decreased in OV-90 cells upon IFNγ stimulation (Figs. 6a and 6b).

Figure 6.

IFNγ induces chemokine production in both ovarian cancer cell lines and patients. (a, b) Columns represent the mean chemokine production in ovarian cancer cell lines with (black columns) or without IFNγ (white columns) stimulation + S.E.M. (c) Data are expressed as the mean fold-changes in chemokine levels in IFNγ-stimulated primary tumor-derived cell cultures in comparison to unstimulated cultures + S.E.M. (D) Histograms show the spontaneous expression of CCL22 in primary tumor-derived cells. Cells are gated on CD45-EpCAM+ tumor cells; CD45+LineageHLA-DR+CD14+ monocytes/macrophages and CD45+Lineage-HLA-DR+CD14CD11c+ mDCs.

Primary ovarian tumor-derived cells have markedly increased CCL22 production upon IFNγ stimulation

To evaluate the effect of IFNγ on primary tumor-derived single cell suspensions, cells isolated from patient tumor tissues (n = 6) were cultured in the presence of IFNγ. As shown in Figure 6c, only CCL21, CCL22 and CXCL10 concentrations increased in the tumor cell culture supernatants upon IFNγ stimulation, with the most significant increase observed for CCL22. As the source of the CCL22 secretion, EpCAM+ tumor cells, monocytes/macrophages and mDCs were identified in unstimulated as well as stimulated cultures (Fig. 6d).

Discussion

The presence and prognostic value of different types of tumor-infiltrating immune cells within tumor tissues have been extensively studied; however, little is known about the alterations in immune cell patterns within the tumor tissue during disease progression. To investigate the dynamics of immune cell distribution within cancer tissues, adequate selection of patients with both early and advanced disease is essential. However, with less than 15% of EOC patients diagnosed at Stage I,30, 31 it is very challenging to study the microenvironment of the early ovarian cancer lesions in prospective studies. Despite the extremely low availability of fresh tumor tissue samples from early disease patients, we were able to analyze six tumor samples from Stage I patients together with 38 tumor samples from patients with more advanced disease (Stages II–IV).

We found that different subsets of CD4+ T cells, but not other immune cell types, were present and exhibited a significantly different pattern during disease progression. Indeed, the early stages of EOC were characterized by a strong Th17 immune response (Fig. 1a), whereas in advanced stages of disease, a dominant population of regulatory T cells was detected (Fig. 2a). Surprisingly, CD8+ T lymphocytes and pDCs did not show any substantial alterations, whereas the proportions of mDCs and monocytes/macrophages appeared to increase during disease progression. The proportion of Th1 cells slightly increased in the Stage II tumors and then decreased again in the advanced stages of disease (data not shown). However, neither the alterations in mDCs and monocytes/macrophages nor those in Th1 cells were statistically significant. A similar kinetics of CD4+ effector T cells, i.e., an early burst of Th17 cells that was replaced by high numbers of IFNγ-producing cells, has been described in mouse models of EAE32 and infection.33 Consistent with these observations, Lohr et al.34 transferred antigen-specific T cells into transgenic mice expressing the antigen recognized by the T cells and found the sequential development of Th17 and then Th1 effector cells followed by Treg development in the late phase of disease. As far as we know, this dynamic has not yet been described in the microenvironment of human tumors.

Notably, the origin of the tumor-infiltrating Th17 cells and their exact role in tumor immunity remain controversial. Kryczek et al.15 has reported that macrophages isolated from EOC tissues are capable of inducing Th17 cells in vitro. However, because IL-1β, IL-6 and IL-23 are supposed to be essential for the differentiation of human Th17 cells,35, 36 the cytokine profile of the tumor microenvironment that was observed in our study, especially the very low levels of IL-1β in stage I patients together with the complete lack of IL-23 (Fig. 4c), does not support the hypothesis of Th17 polarization in situ. It has been shown that Th17 trafficking could be mediated via the CCR6/CCL20 axis.37 Surprisingly, in our study we have not observed any correlation between the numbers of Th17 cells and CCL20 levels in the tumor tissue.

In comparison with Th17 cells, many more studies have focused on the trafficking properties and prognostic significance of Tregs in tumors. Curiel et al. has shown that Tregs infiltrating ovarian cancer tissues effectively inhibit TAA-specific immunity in vitro and in vivo and contribute to tumor growth. These authors also suggested that Treg tumor trafficking could be mediated by CCL22.10 Treg recruitment into the tumor tissue via CCL22/CCR4 has recently been confirmed in breast cancer patients.29, 38

In this study, we showed that Ti-Tregs were mainly Helios+ activated Tregs. A significant number of these Ti-Tregs was demethylated in the FoxP3 TSDR. Naturally occurring, but not in vitro TGFβ-induced, Foxp3+ Tregs have been shown to display stable Foxp3 expression that is associated with selective demethylation of the TSDR.39 Similarly, the Ikaros family member, Helios, has been reported to be a marker useful for discriminating between naturally occurring thymic-derived Tregs (nTregs) and those peripherally induced from naïve CD4+ T cells (iTregs),40 but these findings have very recently been questioned by Akimova et al.41 Thus, rather than discriminating between nTregs and iTregs, Helios appears to be useful for indicating activated Tregs regardless of their origin.41 Taken together, the high expression of Helios together with the demethylation of the TSDR found in the Ti-Tregs in the EOC tissues in our study indicate the stability and highly activated status of these cells.

In accord with the studies concerning Treg tumor trafficking, we demonstrated that the number of Ti-Tregs significantly correlated with the concentration of CCL22 in the ovarian tumor cell culture supernatants. Consistently, Ti-Tregs had significantly lower expression of CCR4 than circulating Tregs. As shown by Gobert et al.,38 low levels of CCR4 reflect its internalization due to an active engagement with CCL22. We also observed that a large proportion of Ti-Tregs (21.8 ± 5.1%) expressed the proliferation marker, Ki67, indicating their extensive local expansion.

Faget et al.29 has demonstrated that IFNγ triggers production of CCL22 by breast tumor cells. Indeed, the cooperation between tumor cells and the immune cell infiltrate (especially macrophages and NK cells) has been shown to be essential for inducing high quantities of CCL22. Similarly, we observed that ovarian cancer cell lines produced chemokines, including significant levels of CCL22, upon IFNγ stimulation. Most important, in primary ovarian cancer tissue-derived cells, recombinant IFNγ selectively augmented CCL22 production with only a small effect on the other chemokines tested (Fig. 6c). Tumor cells, monocytes/macrophages and mDCs were identified as the key source of CCL22 in the cancer tissue samples in our study.

In summary, our data suggest that the development of the antitumor immune response in ovarian cancer patients is highly dynamic. In the early stages of the disease, a strong recruitment of Th17 cells is observed, which is followed by a recruitment of Th1 cells, monocytes/macrophages and mDCs. In the later stages of EOC, enhanced levels of IFNγ, probably produced by the immigrating cells, together with increased proportions of monocytes/macrophages and mDCs may favor the production of CCL22 and the recruitment of Tregs via CCL22/CCR4, thus promoting the development of a suppressive microenvironment. Thus, as our findings clearly show a dynamic shift from an active antitumor immune response to a significant immune suppression in the advanced EOC, the development of novel cancer immunotherapy protocols should not only involve immune-boosting strategies but should also target the immunosuppressive tumor microenvironment in advanced disease.

Acknowledgements

The authors would like to thank Hana Ulčová for technical assistance.

Ancillary