This study included 28 treatment-naïve patients with high-grade serous OC undergoing primary debulking surgery in 2019-2024 ( Table 1 ). Primary resected patients with grading G3 and FIGO stage III-IV were included in this comprehensive investigation. Prior to surgery, all patients were informed about the study and gave their consent for the examination of obtained samples at the University Hospital Carl Gustav Carus Dresden. The Ethics Committee of TUD Dresden University of Technology approved the retrospective analysis of clinical and biological data (EK44022018, EK74032013). PB (n = 23, median = 16 ml, range = 9–18 ml) was collected in EDTA-containing monovettes. Intraoperatively, fresh tumor tissues resected from peritoneal carcinomatosis herds and ovarian tissue (n = 24, median = 1.42 g, range = 0.03-18.1 g), as well as ascites (n = 21, median = 110 ml, range = 10–440 ml) were taken and processed as described in the following sections. Of 24 patients, at least two matched samples of different OC-associated compartments (PB, ascites, and tumor tissue) were obtained forming the main cohort of this study. Furthermore, PB samples were received from four additional patients with the same patient criteria and immune cell profiles as the patients in the main cohort. Therefore, they were included in the association analyses of immune cell frequencies with clinical parameters in order to expand the patient number.

Moreover, immune cell frequencies of PB samples from OC patients were compared to PB samples from gender- and age-matched healthy donors (HD; median: 66 years, range: 60–72 years). Buffy coats from female blood donors were received from the German Red Cross blood donation center (Dresden, Germany; EK138042014) with the consent of the donors.

Peripheral blood mononuclear cells (PBMCs) were isolated from PB samples by density gradient centrifugation (980 × g, 20 min) at room temperature (RT) using Pancoll^® (PAN-Biotech, Aidenbach, Germany). Hereinafter, PBMCs were harvested from the interphase, washed with cold phosphate buffered saline (PBS, Sigma-Aldrich, St. Louis, Missouri, United States), and passed over a 40 µm cell strainer (Falcon™, Thermo Fisher Scientific, Schwerte, Germany).

Further, ascites samples were centrifuged several times. First, the whole ascites samples were centrifuged at 260 × g for 10 min (RT) to eliminate larger debris fragments and aggregates. Resultant pellet was discarded and the supernatant was centrifuged again at 1800 × g for 10 min (RT). Hereinafter, the resultant supernatant was discarded and the pelleted cells contained the immune cells.

For immune cell detection in tumor tissues, the samples were manually minced with a scalpel and further enzymatically processed using Tumor Dissociation Kit and gentleMACS™ Dissociator (both from Miltenyi Biotec, Bergisch Gladbach, Germany). After digestion according to the standard protocol (1 h, 37 °C), the suspension was filtered through a 70 µm cell strainer (Falcon™, Thermo Fisher Scientific) and washed with Dulbecco’s Modified Eagle Medium (DMEM cell culture media, Gibco™, Thermo Fisher Scientific).

In case of remaining red blood cells in the pellets, patient samples were additionally incubated in ammonium-chloride-potassium lysing buffer (ACK buffer, Gibco™, Thermo Fisher Scientific) for 10 min at RT and washed with PBS. The pelleted immune cells were counted in acetic acid (5%, Merck KGaA, Darmstadt, Germany) or trypan blue (0.4%, Gibco™, Thermo Fisher Scientific) using the Neubauer counting chamber (Marienfeld GmbH & Co.KG, Lauda Königshofen, Germany). For subsequent staining procedure, at least 3 × 10⁵ cells (blood) or 1 × 10⁶ cells (ascites and tumor tissue) for each individual staining panel were used. Thus, depending on the count of isolated PBMCs from each patient sample, different quantities of immune cell panels were stained and analyzed.

For specific staining of surface antigens, the cells were incubated with appropriate antibodies in the dark for 15 min at 4 °C ( Supplementary Table 1 , antibody dilution in brilliant stain buffer, BD Biosciences, Franklin Lakes, New Jersey, United States). Hereinafter, the samples were washed with PBS (360 × g, 3 min, 4 °C) or further treated to stain intracellular markers. In terms of FoxP3 staining, cells were fixed and permeabilized using FoxP3 staining buffer set according to manufacturer’s instructions (Miltenyi Biotec) followed by an incubation step (30 min, 4 °C, dark) with fluorophore-labeled anti-human FoxP3 antibody or isotype control ( Supplementary Table 1 ). On the other hand, cells were incubated with 4% paraformaldehyde solution (PFA, 10 min, on ice, dark, Sigma-Aldrich) and 0.1% saponin solution (3 min, 4 °C, dark, Carl Roth GmbH + Co. KG, Karlsruhe, Germany) for the intracytoplasmic staining of myeloperoxidase (MPO) and arginase 1 in the neutrophil staining panel (antibody dilution in saponin buffer). Prior to the flow cytometric measurement using flow cytometer BD LSR Fortessa™ and BD FACSDiva™ software (BD Biosciences), samples were filtered using nylon membranes (Sefar, Thal, Switzerland) and lastly 7-AAD (BD Biosciences) was added to stain for cell viability.

FlowLogic™ software (version 8.6; Inivai Technologies Pty. Ltd., Mentone Victoria, Australia) was used for data evaluation. For each sample, isotype control approaches served as controls for particular gating strategies ( Supplementary Figures 1 – 5 ). Hierarchical gating comprised duplicate exclusion (FSC-A vs. FSC-H) followed by dead cell exclusion (SSC-A vs. 7-AAD staining) and specific antigen stainings. Single patient samples, where fluorescence intensities of the marker proteins to be detected did not correspond to the generally valid gating strategy for all OC patients of the respective panel, failed quality control and were excluded from the respective analysis. Boolean functions were applied to the gating strategies in FlowLogic™ in terms of i) expression or absence of CD10 and CD16 by neutrophils, ii) expression or absence of CD4 and CD8 by γδT cells, and iii) phenotypic characterization of T cells with respect to the expression of several co-activating and co-inhibitory receptor combinations.

In order to further characterize CD127+ CD8+ T cells, frozen PBMC samples of 13 OC patients were used. After thawing the cells in PBS supplemented with 4% sodium citrate and 0.5% human serum, PBMCs were placed in RPMI 1640 medium (Gibco™, Thermo Fisher Scientific) with 10% human serum (C.C.Pro GmbH, Neustadt, Deutschland), 200 mM L-glutamine (Sigma-Aldrich), 100 mM sodium pyruvate (Sigma-Aldrich), 1% non-essential amino acids (Gibco™, Thermo Fisher Scientific), 100 µg/ml penicillin, and 100 µg/ml streptomycin (both Gibco™, Thermo Fisher Scientific) in 96 well U-bottom plates (3 × 10⁵ cells per well) and were stimulated by Phorbol-12-myristate-13-acetate (PMA; 10 ng/ml, Sigma-Aldrich) and ionomycin (1 µg/ml, Sigma-Aldrich). To block exocytosis of cytokines, brefeldin A (1 µg/ml, Sigma-Aldrich) was added simultaneously to the cell culture. After 4 h incubation (36 °C, humidified atmosphere containing 5% CO₂), the cells were harvested and stained with fixable viability stain 575V (FVS, BD Biosciences) according to manufacturer’s instructions to discriminate between viable and dead cells in the flow cytometric analysis. Further, cells were stained with antibodies directed against surface antigens (CD3, CD8, CD127, CCR7, CD45RA, Supplementary Table 1 , diluted in Brilliant Stain Buffer, BD Biosciences) for 15 min at 4 °C. After washing, cells were fixed with ice-cold 4% PFA (Merck) for 15 min and permeabilized with 0.1% saponin (Merck) for 3 min at 4 °C. Subsequently, staining of intracellular granzyme B (GrzB, Supplementary Table 1 , diluted in 0.1% saponin buffer) with an APC-labeled antibody followed for 15 min at 4 °C. Prior to flow cytometric analysis using LSR Fortessa™ and BD FACSDiva™ software (BD Biosciences), samples were filtered through nylon membranes (Sefar, Thal, Switzerland). FlowLogic™ software (version 8.6; Inivai Technologies Pty. Ltd., Mentone Victoria, Australia) was used for data evaluation. For each sample, isotype control approaches served as controls for particular gating strategies ( Supplementary Figure 6 ). Hierarchical gating comprised duplicate exclusion (FSC-A vs. FSC-H) followed by dead cell exclusion (SSC-A vs. FVS staining) and specific antigen stainings.

All data sets were statistically analyzed and illustrated using GraphPad Prism 10.2.3 software (GraphPad Prism Inc., Boston, Massachusetts, United States). Dot plots show mean ± SEM and multivariate Cox proportional hazards regression model includes hazard ratios (HR) with 95% CI and corresponding p-values. Depending on the underlying data set, unpaired or paired t-test, one-way or two-way ANOVA with Bonferroni post-hoc test for multiple comparisons, and Log-rank test for Kaplan-Meier survival analysis were performed. In terms of correlation analysis, Pearson correlation coefficients and two-tailed p-vales were calculated. Values were considered significant according to significance levels of p ≤ 0.05 (*), p ≤ 0.01 (**), p < 0.001 (***). Statistical tests and significance levels are mentioned in the respective figure legends.

In this section, a general overview of innate immune cells is provided entailing DC, monocytes, neutrophils, MDSC, and NK cells. We investigated pDC (CD303+ CD123+ HLA-DR+) and cDC (CD11c+ HLA-DR+) subpopulations, with cDC being further subdivided in cDC1 (CD141+) and cDC2 (CD1c+). Among the investigated OC compartments, pDC frequency was highest in ascites followed by tumor samples, with average percentages of 5.18 ± 2.32% and 2.60 ± 1.90% of HLA-DR+ cells, respectively, but without reaching significant difference compared to PB (1.01 ± 0.16% of HLA-DR+ cells, Figure 2A ). Regarding the expression of additional surface proteins, especially tumor-infiltrated and to a lesser extent ascites-derived pDC expressed ICOS-L, CD86, and CD40 indicating higher pDC activity in OC tissue ( Figure 2B ). In contrast, the mentioned marker proteins were almost absent on PB-derived pDC. Furthermore, PD-L1+ pDC were increased in OC tissues without reaching significance compared to blood and ascites samples. The highest proportions of cDC1 and cDC2 were detected in ascites with 1.19 ± 0.37% and 3.55 ± 2.56% of HLA-DR+ cells, respectively ( Figure 2C ). Regarding their phenotypes, CD86+ cDC1 were more frequently detected than CD86+ cDC2 in tumor tissue, with 44.44 ± 10.62% and 7.72 ± 4.84%, respectively ( Figure 2D , Supplementary Figure 7 ). For both liquid samples, CD86+ cDC1 and CD86+ cDC2 frequencies varied between 18-30% on average. Whereas PD-L1 expression on cDC1 was almost absent (< 4% of cDC1), PD-L1+ cDC2 were detected in PB and tumor tissue in a few OC patients.

Further, PB and ascites samples were analyzed in terms of monocyte abundance resulting in similar proportions, with 14.83 ± 2.38% and 19.91 ± 4.07% of CD45+ cells, respectively ( Figure 2E ). Additionally, the frequencies of several monocyte subpopulations classified as CD14+ CD16- (classical), CD14+ CD16+ (intermediate), and CD14+ CD16++ (non-classical) monocytes were studied ( Figure 2F ). Classical monocytes occurred predominantly in blood samples (blood: 67.94 ± 3.25% of CD14+ cells, ascites: 24.27 ± 3.43% of CD14+ cells), whereas intermediate monocytes were significantly enriched in ascites (blood: 9.13 ± 1.64% of CD14+ cells, ascites: 27.69 ± 7.61% of CD14+ cells). In turn, non-classical monocytes represented the least common subtype in both samples. On average, less than one fifth of the monocyte subsets expressed PD-L1 on their surfaces, with no statistical differences between the investigated liquid samples ( Figure 2G ). In comparison to HD, OC patients showed a significantly increased frequency of CD14+ monocytes in PB as well as numerically higher amounts of PD-L1+ monocytes and PD-L1+ monocyte subpopulations ( Supplementary Figure 8 ).

Regarding neutrophil (CD15+ CD66b+ CD11b+) abundance, no significant differences among ascites- or tumor tissue-derived neutrophils were obvious, with 13.88 ± 5.84% and 10.01 ± 2.54% of CD45+ cells, respectively ( Figure 3A ). Of note, neutrophils were studied only in ascites and tumor tissue samples of OC patients, as the PB sample preparation by density gradient centrifugation excluded neutrophils from PBMCs. Looking at the neutrophil phenotypes in more detail, PD-L1 was expressed in < 12% of neutrophils on average ( Figure 3B ). In terms of neutrophil activation, proportions of MPO+ and arginase 1+ neutrophils were higher in ascites than in tumor, with a 1.4- to 2.6-fold change, respectively. According to the literature, CD16low/- CD10- neutrophils are referred to be immature neutrophils, whereas CD16high CD10low/+ neutrophils reflect the mature phenotype (45–48). In our study, CD10- CD16+ mature neutrophils were significantly more prominent in ascites (ascites: 67.82 ± 11.94% of neutrophils, tumor: 19.15 ± 7.06% of neutrophils), while CD10- CD16- immature neutrophils dominated in OC tissues (ascites: 24.74 ± 11.28% of neutrophils, tumor: 65.19 ± 8.99% of neutrophils). Similarly, mature neutrophils in ascites co-expressed both maturation marker CD10 and CD16 two times more frequently (5.63 ± 2.62% of neutrophils) than their counterparts in tumor samples (2.10 ± 0.76% of neutrophils).

In addition, MDSC were distinguished in three subpopulations (49): CD14+ CD15- CD11b+ HLA-DR- (M-MDSC), CD14- CD15+ CD11b+ HLA-DR- (PMN-MDSC), and CD14- CD15- HLA-DR- CD33+ (e-MDSC). A significantly higher proportion of PMN-MDSC was found in tumor tissues (1.80 ± 0.76% of CD45+ cells) compared to blood and ascites samples ( Figure 3C ). In contrast, percentages of M-MDSC and e-MDSC were similar for all investigated OC-associated compartments, with < 1% of CD45+ cells. Interestingly, on average every second tumor-infiltrated M-MDSC and e-MDSC displayed PD-L1 expression (PD-L1+ M-MDSC: 55.26 ± 10.70%, PD-L1+ e-MDSC: 49.17 ± 9.40%) being significantly elevated in comparison to both liquid samples (blood PD-L1+ M-MDSC: 14.96 ± 9.32%, ascites PD-L1+ M-MDSC: 8.38 ± 5.41%, blood PD-L1+ e-MDSC: 2.09 ± 1.27%, ascites PD-L1+ e-MDSC: 8.56 ± 2.16%, Figure 3D ).

In all investigated compartments, up to 10% of NK cells (CD3- CD56+) among CD45+ immune cells were detected ( Figure 3E ). A more detailed phenotypic characterization regarding the expression of NKG2D, NKp46, CD69, and CD57 showed that NKG2D and CD57 were mainly expressed by blood-derived NK cells, with 30.49 ± 4.60% and 40.53 ± 3.94%, respectively ( Figure 3F ), whereas NKp46 was found significantly more often on ascites-derived NK cells (59.40 ± 9.43% of NK cells). CD69 was the least expressed marker by NK cells in all three compartments. With regard to CD16 expression, defining two functional distinct NK cell subsets, significant differences between the individual compartments were evident as 75.91 ± 7.47% of blood-circulating NK cells carried this Fc receptor followed by NK cells in ascites (49.93 ± 9.54% of NK cells) and tumor-infiltrated NK cells (24.49 ± 6.18% of NK cells, Figure 3G ). Both CD16+ and CD16- NK cell subpopulations displayed similar distributions of NKG2D, NKp46, CD69, and CD57 expression in all OC-associated compartments ( Figure 3H ). However, in PB samples, significantly higher proportions of CD57+ CD16+ NK cells and NKp46+ CD16- NK cells were observed in comparison to their respective counterparts ( Supplementary Figure 9 ). Comparing NK cell subsets among HD and OC, a significantly reduced frequency of NKG2D+ NK cells and a numerically increased amount of NKp46+ NK cells in OC, especially in CD16+ NK cell subpopulations, was found ( Supplementary Figure 10 ).

Summarizing the phenotyping of innate immune cell populations in multiple compartments of high-grade serous OC patients, we found lower proportions of classical monocytes and mature NK cells with cytotoxic potential in ascites and tumor than PB. Ascites samples showed elevated percentages of cDC being key regulators of immune responses, mature neutrophils (CD10- CD16+), and NKp46+ NK cells potentially exerting antitumoral effects. On the contrary, immature neutrophils (CD10- CD16-), PMN-MDSC, PD-L1+ M-MDSC, and PD-L1+ e-MDSC, pDC expressing ICOS-L, CD86, or CD40 as well as CD16- NK cell proportions with limited antitumoral potential dominated in tumor tissues.

Since tumor-infiltrated T lymphocytes are supposed to decisively affect clinical outcomes of high-grade serous OC patients (39), we focused on their differentiation stages and (co-)expression of functional markers in our study. Of note, αβT cells were studied in all three OC-associated compartments, whereas γδT cells were investigated only in ascites and tumor tissue samples, as γδT cell frequency is reported to be less frequent in blood samples (50).

Proportions of CD3+ T cells (32.45 ± 6.66% of CD45+ cells) in OC tissues were significantly lower compared to blood (57.67 ± 2.80% of CD45+ cells) and ascites samples (64.54 ± 6.73% of CD45+ cells, Figure 4A ). Among CD3+ T cells, γδT cells were detected in low percentages in ascites (1.10 ± 0.71% of CD3+ T cells) and tumor tissue samples (2.85 ± 1.31% of CD3+ T cells, Figure 4B ). In addition, γδT cells were subdivided according to CD4 and CD8 expression pattern (11), whereby a CD4+ phenotype was more abundant in ascites and tumor samples, with > 50% of γδT cells each ( Figure 4C ). Likewise, CD8+ γδT cells and CD4- CD8- γδT cells displayed similar percentages in both investigated compartments. Further, γδT cells were distinguished into four subpopulations with respect to their differentiation stages based on CD45RA and CD27 expression (51, 52): T_N (naïve γδT cells: CD45RA+ CD27+), T_CM (central memory γδT cells: CD45RA- CD27+), T_TD/EX (terminally differentiated exhausted γδT cells: CD45RA+ CD27-), and T_EM (effector memory γδT cells: CD45RA- CD27-). Both memory γδT cell subsets (T_CM and T_EM) represented the most frequently occurring ones in both investigated compartments ( Figure 4D ), with γδT_CM being significantly enriched in tumor tissues (50.95 ± 7.04% of γδT cells) and γδT_EM being more abundant in ascites samples (56.19 ± 8.04% of γδT cells) than in tumor samples (42.99 ± 7.14% of γδT cells). Moreover, significantly high proportions of tumor-infiltrated CD69+ γδT cells and PD-L1+ γδT cells were detected being at least 3-times more abundant than ascites-derived γδT cell counterparts ( Figure 4E ).

Besides γδT cells, the frequencies of αβT cell populations among CD3+ T cells were analyzed. In all investigated OC-associated compartments, CD4+ T cells displayed higher percentages than CD8+ T cells ( Figure 4F ), with a significantly increased proportion of blood-circulating CD4+ T cells (76.77 ± 3.03% of CD3+ T cells) and least tumor-infiltrated CD4+ T cells (47.99 ± 5.20% of CD3+ cells). Within the CD4+ T cell population, T_reg (CD4+ FoxP3+ CD25+ CD127-) are a special subset eliciting immunosuppressive effects. On average, T_reg proportions tended to increase from blood to ascites to tumor samples, although T_reg frequency was < 4% of CD4+ T cells ( Figure 4G ). Regarding the expression of immune checkpoint molecules, T_reg in ascites showed increased expression of PD-1 (42.45 ± 8.02% of T_reg) and PD-L1 (30.75 ± 10.84% of T_reg), whereas the proportions of tumor-infiltrated PD-1+ T_reg and PD-L1+ T_reg were 18.30 ± 5.64% and 36.57 ± 15.48%, respectively ( Figure 4H ). The investigated immune checkpoint markers were almost absent on blood-circulating T_reg with < 10% marker expression on average.

Moreover, CD4+ and CD8+ αβT cells were classified according to their differentiation stages: T_N (naïve T cells: CD45RA+ CCR7+ CD28+ CD95-), T_SCM (stem cell-like memory T cells: CD45RA+ CCR7+ CD28+ CD95+), T_CM (central memory T cells: CD45RA- CCR7+ CD28+), T_TM (transitional memory T cells: CD45RA- CCR7- CD28+), T_EM (effector memory T cells: CD45RA- CCR7- CD28-), and T_TE (terminal effector T cells: CD45RA+ CCR7- CD28-). In general, the differentiation profiles of CD4+ and CD8+ T cells in the three investigated OC-associated compartments were very similar. For both CD4+ and CD8+ T cells, significantly elevated T_N proportions were found in PB samples, with 29.31 ± 2.97% of CD4+ T cells and 15.07 ± 2.57% of CD8+ T cells, compared to the other two compartments ( Figures 4I, J ). The predominance of less differentiated blood-circulating CD4+ and CD8+ T cells was also visible for T_SCM and T_CM, although not reaching a significant difference compared to the other two OC-associated compartments. Further, CD4+ T_TM dominated in ascites and tumor tissues with 53.64 ± 10.37% and 46.63 ± 8.79% of CD4+ T cells, respectively. Likewise, CD8+ T_TM and CD8+ T_EM proportions were the most abundant differentiation stages in ascites (T_TM: 27.42 ± 6.06% of CD8+ T cells, T_EM: 21.11 ± 6.46% of CD8+ T cells) and tumor samples (T_TM: 28.62 ± 5.97% of CD8+ T cells, T_EM: 38.84 ± 7.75% of CD8+ T cells). In comparison to PB of HD, OC patients showed significantly reduced frequencies of CD4+ and CD8+ T_N as well as higher amounts of CD4+ and CD8+ T_SCM ( Supplementary Figure 11 ).

Thus, comprehensive analysis of adaptive immune cells in multiple compartments of high-grade serous OC patients showed that the proportions of tumor-infiltrated CD3+ T cells, especially CD4+ T cell subsets, were diminished compared to both liquid samples. However, higher proportions of potentially immunosuppressive T_reg were found in OC tissues than in ascites and PB samples. Further, blood-circulating αβT cells predominantly displayed less differentiated phenotypes (T_N, T_SCM), whereas in the other two compartments γδT cells and αβT cells mainly possess memory differentiation stages (γδT_CM, γδT_EM, αβT_TM, αβT_EM) representing T cells with increased effector functions and cytotoxicity.

In addition, CD4+ and CD8+ T cells were comprehensively analyzed in terms of their expression of several activating or inhibitory marker proteins (53–55). In general, the proportions of all investigated receptors in all three studied OC-associated compartments were similar for CD4+ and CD8+ T cells ( Figures 5 , 6 ). Starting with T cell stimulating receptors, percentages of CD27 expression were highest on blood-circulating CD4+ T cells (87.52 ± 2.57% of CD4+ T cells) and CD8+ T cells (60.59 ± 4.98% of CD8+ T cells) followed by ascites samples and tumor tissues having least CD27+ T cell proportions ( Figures 5A , 6A ). This graduation in T cell proportions was also visible for CD127 (interleukin-7 receptor, IL-7R) expression, with PB having significantly elevated CD127+ T cell proportions (34.05 ± 5.40% of CD4+ T cells and 17.16 ± 3.05% of CD8+ T cells). Moreover, ascites-derived CD4+ T cells and CD8+ T cells exhibited significantly reduced proportions of CD28, with 63.60 ± 10.77% of CD4+ T cells and 16.05 ± 5.57% of CD8+ T cells, compared to PB and tumor samples. Considering the percentages of the other activating receptors ICOS, OX40, and 4-1BB by CD4+ T cells and CD8+ T cells, highest proportions were found in tumor-infiltrated T cells. In addition, proportions of ICOS+ 4-1BB+ T cells were significantly increased in tumor tissues being at least 3-times more abundant compared to both liquid samples ( Figures 5B , 6B ). In addition, CD8+ T cells demonstrated significantly higher proportions expressing the activation marker CD69+ in the tumor tissues (18.03 ± 4.17% of CD8+ T cells) compared to PB (3.14 ± 1.43% of CD8+ T cells). In terms of detailed inhibitory receptor expression analysis, significantly higher fractions of tumor-infiltrated CD4+ T cells expressed LAG-3 (13.06 ± 4.63% of CD4+ T cells) and PD-1 (14.01 ± 3.76% of CD4+ T cells, Figure 5A ). Similarly, significantly elevated percentages of CD8+ T cells expressing LAG-3 (31.50 ± 8.33% of CD8+ T cells), and PD-1 (13.01 ± 3.82% of CD8+ T cells) were detected in tumor tissues ( Figure 6A ), as well. In comparison, lower proportions of ascites-derived CD4+ T cells and CD8+ T cells expressed PD-1 (7.69 ± 2.84% of CD4+ T cells, 8.60 ± 2.96% of CD8+ T cells). Looking at T cells in the PB, inhibitory marker expression was almost absent with < 8% of CD4+ or CD8+ T cells. Furthermore, co-expression of PD-1 and LAG-3 or PD-1 and CD69 were the two most common marker combinations on T cells in OC tissues ( Figures 5B , 6B ).

Generally, tumor-infiltrated CD4+ and CD8+ T cells displayed higher proportions expressing one or multiple activating and inhibitory markers than T cells in the liquid compartments ( Supplementary Figure 12 ). Comparing PB of OC with HD, we found significant higher proportions of CD4+ T cells expressing CD25, CD127, or 4-1BB as well as CD8+ T cells positive for CD25, CD127, ICOS, OX40, or 4-1BB in the blood of OC patients compared to HD ( Supplementary Figures 13 , 14 ).

In order to associate the immune cell profiles with clinicopathological OC patient characteristics, like FIGO stage, serum CA125 level at surgery, or residual tumor after surgery, the previous main cohort of 24 patients was expanded by four additional OC patients, where only PB samples were received for immune cell monitoring. Regarding immune cell abundancies and phenotypic marker expression, the results of the four additional patients did not differ significantly from the previous PB samples ( Supplementary Figure 15 ).

Immune cell phenotypes within the three investigated OC-associated compartments were compared between patients classified as FIGO stage III and FIGO stage IV ( Supplementary Figure 16 ). In our cohort, one third of the patients were classified as FIGO stage IV ( Table 1 ). Thereby, patients with FIGO stage IV exhibited significantly reduced CD8+ T cell proportions in tumor tissues (FIGO III: 30.98 ± 5.16% of CD3+ T cells, FIGO IV: 16.33 ± 3.70% of CD3+ T cells) being visible as trend also for ascites and PB samples ( Supplementary Figure 16B ). In PB samples of patients with FIGO stage IV, percentages of cDC2, CD57+ CD16- NK cells, and CD4+ T_TM were increased, whereby these differences were not seen in ascites or tumor samples. In addition, CA125 was measured in patient serum prior to surgery, as CA125 is the best-characterized biomarker in OC so far (40) and elevated serum CA125 level were linked to unfavorable clinical outcomes in OC patients (10, 11). We investigated, whether immune cell frequencies and phenotypes can be associated with serum CA125 level ( Supplementary Figure 17 ). The average serum CA125 value at surgery for our cohort was 487.5 IU/ml. In PB samples of OC patients with high CA125 values (> 487.5 IU/ml), the proportions of CD40+ pDC and TIGIT+ CD8+ T cells were significantly raised, with ascites and tumor tissue samples showing the same trend for both immune cell phenotypes ( Supplementary Figure 17B ). Patients with high CA125 level had also increased percentages of ICOS+ CD4+ T cell in ascites, with the other two compartments indicating same trends for both immune cell subtypes. In addition, patients were stratified according to their surgical outcome and immune cell frequencies were compared between patients with complete resection and residual tumor, whereby almost 80% of our patients underwent macroscopic complete resection ( Table 1 , Supplementary Figure 18 ). OC patients with residual tumor after surgery demonstrated elevated levels of PB-circulating PD-L1+ non-classical monocytes as well as tumor-infiltrated neutrophils, e-MDSC, γδT_CM, and PD-L1+ γδT cells, with the corresponding other OC-associated compartments showing the same trends for these immune cell phenotypes ( Supplementary Figure 18B ). On the contrary, γδT_EM cells were more pronounced in ascites and tumor samples of patients with complete resection.

In order to find a potential prognostic biomarker reflecting the local TME in one of the liquid biopsies, the immune cell characteristics in the OC-associated compartments were correlated with each other resulting in three comparisons (blood - ascites, blood - tumor, ascites - tumor). Therefore, we included only matched pairs of all three regions and excluded immune cell subsets with very low abundancies. For each immune cell phenotype, calculated Pearson correlation coefficients and corresponding p-values for the three comparisons are shown in separate heatmap-like graphs ( Figure 7A ). Generally, the correlation analysis showed that the three comparisons resulted in similar numbers of significant correlations. Looking at individual correlations of immune cell phenotypes among the OC-associated sites in more detail, CD127+ and PD-1+ T cell subpopulations correlated significantly positive between both liquid samples. In particular, CD127+ CD8+ T cells demonstrated significant or trending positive correlations between all three investigated compartments. Further, CD27+ CD4+ T cells and LAG-3+ CD8+ T cells displayed significant positive correlations between PB and tumor tissues, whereas a positive association of 4-1BB+ CD4+ or OX40+ CD4+ T cells derived from ascites and tumor samples was detected. Regarding NK cells, significantly positive correlations were obvious for NKp46+ NK cells and CD69+ NK cells in PB or ascites samples in association with corresponding tumor tissues, irrespective of CD16 expression.

Further, immune cell populations showing significant correlations among the investigated compartments were associated with progression-free survival (PFS) and OS of OC patients. In this regard, only patients with a follow up period of at least two years were included ( Table 1 , Supplementary Figure 19 ). Due to the limited number of ascites samples from patients with suitable clinical data, only immune cells in PB and tumor tissue were appropriate for survival analysis. OC patients were stratified according to the median abundance of the respective immune cell subset and calculated p-values with respect to PFS and OS are shown in a heatmap-like manner ( Figure 7B ). OC patients with high frequencies of either tumor-infiltrating CD127+ CD4+ T cells or blood-circulating 4-1BB+ CD4+ T cells showed significantly reduced PFS and OS, respectively ( Supplementary Figure 20 ). However, CD127+ CD8+ T cells were the only immune cell subset in both, PB and tumor tissues, clearly associated with PFS and OS, with high CD127+ CD8+ T cell proportions being linked to unfavorable OC patients’ outcomes ( Figure 7C ). Further, multivariate Cox regression was performed to investigate whether proportions of CD127+ CD8+ T cells in PB might be an independent prognostic marker. Therefore, patient age at diagnosis, FIGO stage, serum CA125 at surgery, and residual tumor after surgery were included in the Cox proportional hazards regression models ( Figure 7D ). Of note, surgical outcome was only integrated for PFS and not for OS regression model due to missing OS events within the two-year follow-up period of patients with residual tumor. For both OS and PFS of our OC cohort, higher age at diagnosis (> 69 years), higher FIGO stage (IV), and higher CA125 values (> 487.5 IU/ml) were linked to increased hazard ratios (HR). The association with poor prognosis was also evident for CD127+ CD8+ T cell abundance, with a high proportion (> 15% of blood-circulating CD8+ T cells) being an independent prognostic marker for unfavorable PFS (HR = 8.61, p = 0.04).

In addition to the comprehensive investigation of immune cells freshly isolated from PB samples, remaining frozen PBMC samples from 13 OC patients were used to assess the expression of GrzB and maturation status of CD127+ CD8+ T cells compared to their CD127- counterparts by flow cytometry after stimulation. A significantly reduced GrzB expression by CD127+ CD8+ T cells was detected being six-times lower than in CD127- CD8+ T cells ( Figure 8A ). Comparing the frequency of GrzB+ CD127+ CD8+ T cells in OC patient samples with PB of HD, GrzB expression by CD127+ CD8+ T cells was significantly lower in OC ( Supplementary Figure 21A ).

Beyond a reduced cytotoxic potential, the proportion of terminally differentiated effector CD127+ CD8+ T cells (T_TDE: 4.86 ± 3.80% of CD8+ T cells) in OC, defined as CD45RA+ CCR7- T cell population, was significantly lower than corresponding CD127- CD8+ T cells (T_TDE: 20.10 ± 4.84% of CD8+ T cells, Figure 8B ). CD127+ CD8+ T cells showed predominantly memory phenotypes (T_CM: CD45RA- CCR7+, T_EM: CD45RA- CCR7-), whereby a significantly higher proportion of central memory CD127+ CD8+ T cells was measured (CD127+ CD8+ T_CM: 31.14 ± 5.01% of CD8+ T cells, CD127- CD8+ T_CM: 16.20 ± 4.78% of CD8+ T cells). In comparison to HD, OC patients showed significantly reduced proportions of CD127+ CD8+ T_N and CD127- CD8+ T_N ( Supplementary Figure 21B ), but increased frequencies of CD127+ CD8+ T_EM and CD127- CD8+ T_EM. Thus, the characterization of stimulated CD127+ CD8+ T cells revealed a memory T cell phenotype with reduced potential to produce GrzB in OC, explaining why an increased CD127+ CD8+ T cell population might contribute to an unfavorable prognosis of OC patient.

Taken together, the proportion of cytotoxic CD8+ T cells decreased in advanced stage OC (FIGO stage IV), especially in tumor tissues. Further, frequencies of CD127+ CD8+ T cells correlated positive between liquid samples and tumor tissues, with high proportions in PB and tumor tissue being associated with shorter PFS and OS of high-grade serous OC patients. Thereby, high proportions of blood-circulating CD127+ CD8+ T cells, as effector memory T cells with low cytotoxic activity, turned out to be an independent prognostic marker for a higher risk of tumor progression.