Patients with stage IV cancer with ME and patients with cancer who planned to undergo surgical resection between April 2012 and December 2017 at the Severance Hospital were prospectively enrolled. Inclusion criteria were as follows: 1) over 20 years old; 2) stage IV cancer with malignant pleural effusion or ascites confirmed by cytology, or cancer with scheduled surgery; and 3) written informed consent. We collected 300 cc of effusions and simultaneously obtained 10 cc of PB from patients with stage IV cancer with ME, if available. In patients who underwent surgery, we collected TM-adjacent normal tissue, and 10 cc of PB to isolate peripheral blood lymphocytes (PBLs). The study was approved by the Institutional Review Board of Severance Hospital. We categorized the samples into three groups: 1) PBLs, 2) ME from patients with stage IV cancer, and 3) TM from patients with cancer who underwent surgery. To analyze the characteristics of T_reg cells in TME, we also collected paired peritumoral tissue lymphocytes (pTILs), tumor-infiltrating lymphocytes (TILs), and PBLs at the same day from 12 patients with NSCLC who underwent curative resection.

PB mononuclear cells were isolated from 10 cc PB collected into EDTA tubes by separation over a Percoll (Sigma-Aldrich) gradient. Lymphocytes were isolated from 500 cc of ME by discontinuous density gradient centrifugation on Percoll. To isolate TILs, lung TMs were chopped and then incubated with a solution containing 1 mg/mL collagenase type IV (Worthington Biochemical) and 0.01 mg/mL DNaseI (Sigma-Aldrich) at 37 °C for 25 min. TILs were isolated by Percoll gradient after washing dissociated tissues with ice-cold RPMI1640.

Flow cytometry was performed using a FACS CANTOII (BD Biosciences, Franklin Lakes, NJ, USA) and CytoFLEX (Beckman Coulter, IN, USA). Data were analyzed using FlowJo software (Tree Star, OR, USA).

For immunolabeling of human samples, fluorophore-conjugated monoclonal antibodies against the following proteins were used: CD4 (RPA-T4), CD3 (OKT3), PD-1 (EH12.2H7), and CTLA-4 (BNI3) (all from Biolegend, San Diego, CA, USA); TIM-3 (344823) and TIGIT (741182) (both from R & D Systems, Minneapolis, MN, USA); CD25 (M-A251) (BD Biosciences, Franklin Lakes, NJ, USA); and Foxp3 (PCH101) (eBioscience, San Diego, CA, USA). The LIVE/DEAD Fixable Red Dead Cell Stain kit was from Invitrogen (Carlsbad, CA, USA). T_reg cells labeled with various antibodies (except for the antibody against Foxp3) were fixed and permeabilized with Foxp3 fixation/permeabilization solution (eBioscience). Foxp3 antibody was then administered for intracellular labeling of T_reg cells. The proportion of CD4⁺ and CD8⁺ T cells among total lymphocytes was determined, and the fraction of Foxp3-positive CD4⁺ T cells was quantified.

For immunolabeling of mouse samples, fluorophore-conjugated monoclonal antibodies against the following proteins were used: CD4 (RM4–5), Ly5.1 (A20), PD-1 (29F.1A12), TIM-3 (RMT3–23), NK1.1 (PK136), and DX5 (DX5) (all from Biolegend); and CD8 (53-6.7), CD25 (PC61.5), CTLA-4 (UC10-4B9), TIGIT (G1GD7), and F4/80 (BM8) (all from eBioscience); and CD11b (M1/79) (BD Biosciences). The LIVE/DEAD Fixable Near-IR Dead Cell Stain kit was from Invitrogen. T_reg cells labeled with various antibodies (except for the antibodies against Foxp3 and CTLA-4) were fixed and permeabilized with Foxp3 fixation/permeabilization solution (eBioscience, San Diego, CA, USA). Foxp3 antibody was then administered for intracellular labeling of T_reg cells. The proportions of CD4⁺ and CD8⁺ T cells among lymphocytes were determined, and the fraction of Foxp3-positive CD4⁺ T cells was quantified. To prevent myeloid cells from non-specific staining, samples were preincubated with anti-CD16/32 (93) (eBioscience) before immunolabeling with fluorophore-conjugated antibodies.

Female C57BL/6, C57BL/6-Rag2^−/−, and C57BL/6-Ly5.1 congenic mice (5–6 weeks) were purchased from Charles River Laboratories (Wilmington, MA, USA) and Jackson Laboratories (Bar Harbor, ME, USA). To generate lung TM bearing mice, 5 × 10⁵ TC-1 cells were intravenously injected into C57BL/6 mice via the tail-vein. Mice were sacrificed on day 21 post-injection. Lymphocytes were isolated from the spleen, normal lung, and lung tumor as previously described [9]. The number of tumor nodules on left upper lobe of the lung was counted at day 12, 16, and 21 post-injection. All animal protocols were approved by the Institutional Animal Care and Use Committee of the Yonsei University Laboratory Animal Research Center (2013–0115).

For the T_reg cells suppression assay, CD4⁺CD25⁺ T_reg cells (10⁵/well) were co-cultured with CD8⁺ T cells (10⁵/well) with Dynabeads mouse T-activator CD3/CD28 (Thermo Fisher Scientific, Waltham, MA, USA) in a 96-well U-bottom plate at 37 °C for 72 h. For the CellTrace Violet dilution assay, CD8⁺ T cells were isolated from the spleen of naïve mice using a CD8⁺ T Cell Isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) and labeled with 5 μM CellTrace Violet (Thermo Fisher Scientific). CD4⁺CD25⁺ T_reg cells were separately isolated from the spleen and tumor of TM-bearing mice on day 21 post-injection using CD4⁺CD25⁺ Regulatory T Cell Isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany). To inhibit cell migration, Transwell membranes (0.4 mm pore; BD Biosciences) were inserted into 24-well plate. CD4⁺CD25⁺ T_reg cells (10⁶/well) were co-cultured with CD8⁺ T cells (10⁶/well) with Dynabeads mouse T-activator CD3/CD28 (Thermo Fisher Scientific) in a 24-well plate at 37 °C for 72 h.

For PD-1 blockade in tumor-infiltrating T_reg, CD4⁺CD25⁺ T_reg cells (2.5 × 10⁴/well) isolated from tumor lymphocytes of TM bearing mice on day 14 post-injection were preincubated with 10 μg/mL anti-PD-1 antibody (RMP1–14) or rat IgG2a isotype control (2A3) (Bio X Cell) at 4 °C for 1 h, washed twice, and then co-cultured with CD8⁺ T cells (10⁵/well) in the presence of mouse T-activator CD3/CD28 Dynabeads for 68 h.

To examine the functionality of TIL T_reg (PD-1^hi) and spleen T_reg (PD-1^lo) cells, CD4⁺CD25⁺ T_reg cells were isolated from the tumor and spleen of TM bearing mice at day 21 post-injection using CD4⁺CD25⁺ Regulatory T Cell Isolation kit (Miltenyi Biotec). Ly5.1⁺ CD8⁺ T cells were isolated from naïve C57BL/6-Ly5.1 congenic mice. Ly5.1⁺ CD8⁺ T cells (2 × 10⁶) were injected i.v. into recipient Rag2^−/− mice alone or with Ly5.2⁺ TIL T_reg or spleen T_reg (1 × 10⁶). At day 7 after cell transfer, splenocytes isolated from Rag2^−/− mice were analyzed for homeostatic expansion of the Ly5.1⁺ CD8⁺ T cell population using FACS.

CD4⁺CD25⁺ T_reg cells were isolated from the tumor tissue and peripheral blood of NSCLC patients, using human CD4⁺CD25⁺CD127^dim/− Regulatory T Cells Isolation Kit II (Miltenyi Biotec, Bergisch Gladbach, Germany). CD8⁺ T cells were isolated from the paired peripheral blood of NSCLC patients using human CD8⁺ T Cell Isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) and subsequently labeled with 5 μM CellTrace Violet. The CD8⁺ T cells (10⁵/well) were co-cultured with CD4⁺CD25⁺ T_reg cells (5 × 10⁴/well) isolated from either tumor tissue or peripheral blood in the presence of 2.5 μl/well of Dynabeads human T-activator CD3/28 (Thermo Fisher Scientific) at 37 °C for 72 h.

For multicolor immunofluorescence analysis, lungs were isolated, fixed with 2% paraformaldehyde/phosphate buffered saline overnight at 4 °C, and then embedded in OCT compound (Sakura). Tissue blocks were frozen in 2-methyl butane and cooled on dry ice. Frozen blocks were cut to a thickness of 8 μm and mounted on the silane-coated slide. Sections were stained with 4,6-diamidino-2-phenylindole (DAPI; Invitrogen) and with antibodies for anti-CD8α (Clone 53–6.7), anti-CD4 (clone RM4–5), anti-CD279 (clone RMP1–30), and anti-GFP (clone 1GFP63) for amplification of Foxp3-GFP signals (Biolegend). Streptavidin-conjugated horseradish peroxidase was used for staining of biotin-conjugated antibodies, and TSA Cyanine 3 Tyramidetetramethylrhodamine reagent (SAT704A001EA; PerkinElmer) was subsequently added for amplification. Images was acquired using a microscope (Carl Zeiss Co. Ltd) and analyzed with ImageJ 1.50b software.

Data were analyzed using Prism 5.0 software (GraphPad Inc., CA, USA). The Student’s t-test, one-way analysis of variance, and the least significant difference test were used, where appropriate, to evaluate the significance of differences among groups. All statistical analyses were conducted with a significance level of α = 0.05 (P < 0.05).

We enrolled 103 patients: 72 were stage IV cancer patients with ME, and 31 were patients with operable disease (not stage IV) who underwent surgical resection. Detailed information of the patients from which PB, ME, or TM were obtained is described in Additional file 6: Table S1. The total number of tumor specimens was divided into three groups based on the specimen type: PB, 20.7% (28/135); ME, 56.2% (76/135); and TM, 23.1% (31/135). Detailed analyses of immune subsets as well as the levels of their immune checkpoints were performed in PBLs (PB group), effusion-infiltrating lymphocytes (EILs) (ME group), and TILs (TM group). Primary cancer types in the ME group were NSCLC, 43.1% (31/72); gastric cancer, 22.2% (16/72); colon cancer, 5.6% (4/72), and breast cancer, 5.6% (4/72). The types of ME were ascites, 59.7% (43/76) and pleural effusion, 45.8% (33/76), with four patients having both (Additional file 6: Table S1). The presence of malignant cancer cells and TILs in TM or ME was pathologically or cytologically confirmed (Fig. 1a).Fig1Fig. 1

T cell characteristics and PD-1 expression in T_conv of patients with cancer. Malignant effusions, such as ascites, and pleural effusions were extracted from patients with stage IV cancer. Tumor-infiltrating lymphocytes were obtained from the tumors of patients with non-small cell lung cancer (NSCLC) and colon cancer. a Left, computed tomography images showing malignant ascites (upper), pleural effusion (middle), and lung cancer lesions in left lower lobe (bottom) of a patient with NSCLC. Right, cytological analysis of malignant effusion and histological analysis of lung cancer tissue. Red arrows indicate malignant effusion and cancer (left column) and tumor cells (right column). b Proportions of CD4⁺ and CD8⁺ T cells among CD3⁺ T cells. Representative plots (upper) and statistics (bottom) are shown. c, d PD-1 and TIM-3 expression on CD4⁺ and CD8⁺ T cells. Representative plots of PD-1 and TIM-3 expression (upper) and percentages of total PD-1⁺ (bottom left) and TIM-3⁺ (bottom right) cells among CD4⁺ and CD8⁺ T cells are shown. Peripheral blood lymphocytes (PBLs), effusion-infiltrating lymphocytes (EILs), and tumor-infiltrating lymphocytes (TILs) were isolated from healthy control donors (HC) and patients with cancer (PBLs of HC, n = 16; PBLs, n = 28; EILs, n = 76; TILs, n = 31). Lines in the scatterplot represent the mean values. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001 (Mann-Whitney test)

To investigate T cell subsets in three different tumor specimens, we compared the ratio of CD4⁺ and CD8⁺ T cells in PBLs, EILs, and TILs isolated from PB, ME, and TM, respectively. The percentage of CD4⁺ T cells was higher among TILs than among PBLs or EILs. In contrast, the percentage of CD8⁺ T cells was markedly lower among TILs than among PBLs or EILs (Fig. 1b), suggesting that the migration of cytotoxic lymphocytes (CTLs) into the TM was inhibited.

The phenotypes of PBLs, EILs, and TILs were compared by quantifying CD4⁺ and CD8⁺ T cells expressing PD-1 and TIM-3. The percentage of PD-1- or TIM-3-expressing CD4⁺ or CD8⁺ T cells was the highest among TILs, with lower percentages in EILs and PBLs (Fig. 1c, d), suggesting that T cells derived from TM and ME show more pronounced T cell exhaustion than those derived from PBLs.

We next examined how T_reg cells expressing forkhead box (Fox) p3 are distributed and differ phenotypically in PB, ME, and TMs. T_reg cells showed greater accumulation in TILs than in PBLs and EILs of patients or in PBLs of healthy controls (Fig. 2a). Interestingly, T_reg cells in TILs expressed a higher level of PD-1 than those in PBLs and EILs; moreover, the PD-1-expressing Foxp3⁺ population among CD4⁺ T cells was also larger in TILs than in EILs, which, in turn, had a larger population than PBLs (Fig. 2b). To further characterize CD4⁺ T cells in different tissues, we compared PD-1 in Foxp3⁺ and Foxp3⁻ CD4⁺ T cells (Fig. 2c). The proportion of PD-1-expressing cells in both CD4⁺ cells was larger in EILs and TILs than in PBLs. These results indicate that PD-1 expression by T_reg cells and T_conv cells clearly reflect the TME, as PD-1 expression increased in the following order: TILs > EILs > PBLs.Fig2Fig. 2

PD-1 expression in Foxp3⁺ T_reg in different tissue types of patients with cancer. a Representative plots of CD25 and Foxp3 expression (left) and proportion of Foxp3⁺ T cells (right) among CD4⁺ T cells. b Representative plots of PD-1 and Foxp3 expression (left) and proportion of PD-1 and Foxp3 co-expressing cells among total CD4⁺ T cells (right). c Summary of PD-1-positive fraction of Foxp3⁺ T_conv (left) and Foxp3⁻ T_reg (right) cell populations among CD4⁺ T cells. Peripheral blood lymphocytes (PBLs), effusion-infiltrating lymphocytes (EILs), and tumor-infiltrating lymphocytes (TILs) were isolated from healthy control donors (PBL, n = 16) and patients with cancer (PBL, n = 28; EIL, n = 76; TIL, n = 31). Lines in the scatterplot represent the mean values. ns, not significant; **P < 0.01, ***P < 0.001 (Mann-Whitney test)

To clarify the characteristics of T_reg cells in the TME, we compared the frequency of T_reg cells and the expression of IC-molecules such as PD-1, TIM-3, TIGIT, and CTLA-4 in paired sets of tissue-derived lymphocytes, such as PBLs, pTILs, and TILs collected from 12 patients with NSCLC. As expected, T_reg cells were more highly enriched in TILs than in pTILs and PBLs (Fig. 3a). Moreover, more T_reg-expressing ICs were found among TILs than among pTILs and PBLs (Fig. 3b).Fig3Fig. 3

PD-1-expressing tumor-infiltrating T_reg and their activated phenotype in patients with non-small cell lung cancer (NSCLC). a CD25 and Foxp3 expression in CD4⁺ T cells (upper) and proportion of Foxp3⁺ cells among total CD4⁺ T cells (lower) in peripheral blood lymphocytes (PBLs), peritumoral infiltrating lymphocytes (pTILs), and tumor-infiltrating lymphocytes (TILs) derived from patients with NSCLC. b Representative plots of PD-1, TIM-3, TIGIT, CTLA-4, and Foxp3 expression in CD4⁺ T cells (left) and percentage of CD4⁺ T cells co-expressing PD-1, TIM-3, TIGIT, CTLA-4, and Foxp3 (right). c PD-1, TIM-3, TIGIT, and CTLA-4 expression in Foxp3⁺ T_reg, Foxp3⁻ T_conv and CD8⁺ T_conv of these patients. d Enhanced suppression of CD8⁺ T cells by PD-1-expressing tumor-infiltrating T_reg from NSCLC patients. T_reg were isolated from the peripheral blood and tumor tissue from NSCLC patients. Peripheral blood T_reg and tumor-infiltrating T_reg expressed low and high levels of PD-1, respectively. CellTrace Violet (CTV)-labeled CD8⁺ T cells were stimulated in vitro with CD3/CD28 Dynabeads for 96 h in the absence or presence of each T_reg population. CTV dilution in proliferating CD8⁺ T cells is indicated in each histogram. Histograms represent the percentages of proliferating cells. Lines in the bar graph represent the mean and mean ± SEM, respectively. ns, not significant; **P < 0.01, ***P < 0.001 (Mann-Whitney test)

We previously showed that immune-exhaustion markers were highly expressed in tumor-infiltrating T_reg cells of patients with NSCLC. We therefore investigated the T_reg phenotype in greater detail in different tissues, using a mouse lung cancer model. We compared the expression levels of IC-molecules, such as PD-1, TIM-3, and TIGIT, on CD4⁺ and CD8⁺ T cells in different tissues from naïve and TM-bearing mice. As in patients with cancer tissue, the expression of IC-molecules in CD4⁺ and CD8⁺ T cells was much higher in lung TM than in PB or spleen (Fig. 4a, b). Among the populations expressing IC-molecules, PD-1-expressing CD4⁺ and CD8⁺ T cells were more abundant in the TM.Fig4Fig. 4

Differential expression of immune checkpoint (IC) molecules on CD4⁺ and CD8⁺ T cells in mice with lung cancer. To induce lung adenocarcinoma, TC-1 cells were intravenously injected into syngeneic mice. a, b Tumor-bearing mice at 3 weeks after TC-1 cell injection and naïve control mice were sacrificed, and lymphocytes were isolated from peripheral blood (PB), spleen (SP), and lung (LG). (Left) Expression levels of PD-1, TIM-3, and TIGIT on CD4⁺ and CD8⁺ T cells were assessed. (Right) Summary of the proportions of IC molecules expressed on populations of CD4⁺ and CD8⁺ T cells in PB, SP, and LG at the tumor site. Numbers in the plot indicate percentages of the corresponding population. Data are representative of three independent experiments (n = 5 mice per group in each experiment). ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t-test). The symbol above each column is the P value obtained when SP samples were compared to the corresponding samples from naïve mice (control)

Among all IC-molecules examined, PD-1 was most highly upregulated in tumor-infiltrating T_reg cells. To determine the role of PD-1 on tumor-infiltrating T_reg cells, in the regulation of the CD8⁺ T cell response, we compared the suppressive activity of T_reg expressing high- and low-levels of PD-1 (PD-1^hi T_reg cells from lung TM 3 weeks after TC-1 injection vs. PD-1^lo T_reg cells from the spleen of the same TM-bearing mice). CD4⁺CD25⁺ T_reg cells, isolated using a microbead-based Treg isolation kit (CD4⁺CD25⁺ Regulatory T Cell Isolation kit), was confirmed to be ~ 90% purified Foxp3⁺ T_reg cells (Additional file 4: Figure S4). Each population was co-cultured with naïve CD8⁺ cells with or without stimulation by αCD3/CD28. CD8⁺ T cells proliferated at a high rate in the absence of T_reg cells and were more potently inhibited by PD-1^hi tumor-infiltrating T_reg cells than by PD-1^lospleen T_reg cells (Fig. 6a). Similarly, interferon (IFN)-γ production was also more strongly suppressed by PD-1^hi tumor-infiltrating T_reg than by PD-1^lo spleen T_reg cells.Fig6Fig. 6

Enhanced suppressive function of PD-1-expressing tumor-infiltrating T_reg. a Enhanced suppression of CD8⁺ T cells by PD-1-expressing tumor-infiltrating T_reg. At 3 weeks after intravenous injection of TC-1 cells, T_reg were isolated from the spleen (SP) and lung of mice with TC-1 cell-induced tumors. SP T_reg and tumor-infiltrating T_reg expressed low and high levels of PD-1, respectively. CellTrace Violet (CTV)-labeled CD8⁺ T cells were stimulated in vitro with CD3/CD28 Dynabeads for 72 h in the absence or presence of each T_reg population. CTV dilution in proliferating CD8⁺ T cells is indicated in each histogram. Histograms represent the percentages of proliferating (upper) and IFN-γ-producing (lower) cells. b Contact-dependent T_reg-mediated suppression of CD8⁺ T proliferation. CTV-labeled CD8⁺ T cells were stimulated in vitro with CD3/CD28 Dynabeads and cocultured with tumor-infiltrating T_reg for 72 h in the absence or presence of a transwell membrane. c Homeostatic proliferation of donor Ly5.1⁺CD8⁺ T cells in the spleen isolated from Rag2^−/− mice at 7 d after adoptive cell transfer. Representative plot (left) and absolute number (right) of donor Ly5.1⁺CD8⁺ T cells in the spleen. d PD-1-mediated suppressive activity of tumor-infiltrating T_reg isolated from the lungs of tumor-bearing mice 2 weeks after intravenous injection of TC-1 cells. At this time point, T_reg expressed intermediate levels of PD-1. CTV-labeled CD8⁺ T cells were stimulated as shown in (a). Before co-culture of CD8⁺ T cells with tumor-infiltrating T_reg, the latter were pre-incubated with an anti-PD-1 antibody or its isotype as control. CTV dilution in proliferating CD8⁺ T cells is shown in the histograms, which represent the percentages of proliferating (upper) and IFN-γ-producing (lower) cells. (e) Representative immunofluorescence images of mouse lung tumor samples. Antibodies against Foxp3, CD8, and PD-1 were used to label and examine the interaction between T_reg and CD8⁺ T cells expressing PD-1. Data are representative of two independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t-test)