Twenty-eight patient ascites samples with advanced malignancies were collected at the Departments of Gynecology and Obstetrics at the University Hospital Cologne, Cologne and the University Hospital Essen, Essen, Germany. Among them, 14 patients with initial and 6 patients with recurrent diagnosis of high-grade serous ovarian carcinoma, and 2 patients with initial diagnosis of invasive mucinous ovarian carcinoma, one small cell carcinoma of the ovary, hypercalcemic type (SCCOHT) and one ovarian signet ring cell carcinoma, have been enrolled. In addition, one patient with initial advanced colon carcinoma and one patient with adenocarcinoma of unknown primary have been included. Besides these, one patient with diffuse large cell B-cell lymphoma and one patient with benign ascites due to congestive heart failure were part of the cohort. The ascites samples were collected during clinically indicated surgery or palliative paracentesis after written information and consent. The isolated ascites was transferred to Falcon 50 tubes and centrifuged at 2,000g for 10 min. After centrifugation, only acellular supernatant was collected and sterile filtered using bottle top filters (Sarstedt, 0.2 μm). Patient and healthy donor serum was obtained using clotting activator serum collection tubes (Sarstedt). Samples were centrifuged at 2,000g for 10 min. All samples were aliquoted and frozen at −80°C for long-term storage. The procedure is illustrated in Figure 1A . The study has been approved by the local Ethics Committee of the medical faculties of the Universities of Duisburg-Essen and Cologne.

Several EGFR-positive cell line models were used in coculture assays: The human ovarian cancer cell line IGROV1 was kindly provided by LIMES Institute, University of Bonn, Bonn, Germany. SKOV-3 ovarian cancer cell lines were kindly provided by the Department of Obstetrics and Gynecology, University of Bonn, Germany. The colorectal cancer cell line A431 was obtained from DSMZ, Braunschweig, Germany. All used cancer cell lines showed positive expression for EGFR (epidermal growth factor receptor). IGROV1, SKOV3, and A431 cell lines were cultured in Roswell Park Memorial Institute (RPMI Gibco) supplemented with 10% (v/v) heat-inactivated fetal calf serum (FCS Gibco), 100 U/mL penicillin, and 100 mg/mL streptomycin (PenStrep, Gibco by Life Technologies) (supplemented complete media). Cell cultivation was done in plastic flasks (Sarstedt) at 37°C and 5% CO₂ and deattached for passaging by Accutase (Gibco) treatment for 5 min at 37°C. Before use, the cell lines were tested to exclude mycoplasma contamination. Cell line identity was confirmed by STR Typing done by the Leibniz-Institute DSMZ Molecular Biology Group.

Healthy donor blood was collected in trisodium citrate blood collection tubes (Sarstedt). Blood was diluted (1:1 ratio) with Dulbecco’s Phosphate Buffered Saline (DPBS, Gibco, Life Technologies Limited) and overlaid on a separation medium (1.077 g/mL, Biocoll, Merck Millipore). Separation by density was done by centrifugation at room temperature (400g for 30 min) without acceleration or brake. Collected PBMCs were incubated in a T175 plastic flask (Sarstedt) at 37°C and 5% CO₂ for 1 h for depleting monocytes. NK isolation was performed using NK MACS Isolation Kit (Miltenyi Biotec) according to the manufacturer’s instructions. After isolation, 5 U/mL recombinant human IL-15 (50 µg, Immuno Tools) was added to NK cells that were used for functional experiments after incubation overnight.

Isolated NK cells were cocultured (1:1 ratio, 25,000 each) with ovarian cancer cells in a flat-bottom 96-well plate with or without different patient ascites or ascites-isolated antibody samples (25% addition), bovine serum albumin [fraction V (pH 7.0), PanReac], or Rituximab (MabThera, 10 mg/mL, Roche). The human anti-EGFR-antibody Cetuximab 1 μg/mL [Erbitux, 5 mg/mL, Merck (Serono)] was added to the coculture to induce NK cell-mediated ADCC. For induction of unspecific degranulation, PMA (Sigma-Aldrich, 50 ng/mL) and Ionomycin (Sigma-Aldrich, 1 µg/mL) were added. NK cells were labeled with anti-CD107a-FITC (25 µg/mL, clone H4A3, Mouse IgG1, k, BD Biosciences) or isotype control and incubated for 1 h at 37°C and 5% CO₂. Afterwards, Golgistop Monensin (BD Biosciences) was added (1:600). Following 5 h of incubation, NK cells were stained with CD56-BV421 (12 µg/mL, NCAM 16.2, IgG2b, k, BD Biosciences). CD107a and CD56 expressions were analyzed by flow cytometry.

NK cells were coincubated in 1:1 ratio with A431 cells, Cetuximab (1 µg/mL) and 25% addition of isolated ascites antibodies for 18 h. The following antibodies for the flow cytometric analysis of the NK cell marker expression were used: CD56-BV421 (12 µg/mL, clone NCAM 16.2, mIgG2b,k, BD Biosciences), DNAM-1-PerCP-Cy5.5 (CD226, 200 µg/mL, clone 11A8, mIgG1, k, BioLegend), and CD16-PE-Cy7 (50 µg/mL, clone 3G8, mIgG1, k, BD Biosciences). Stained cells were analyzed with BD FACS Canto II using DIVA 8.01 software (BD Biosciences) or FlowJo10 (LLC, Ashland, Oregon, USA).

Ovarian cancer cells were prestained with Cell Tracker Green (Thermo Fisher Scientific) and NK cells were prestained with Cell Tracker Red (Thermo Fisher Scientific) according to the manufacturer’s instructions. NK cells (50,000 cells) and tumor cells (200,000 cells) were coincubated (1:4 ratio) in the presence of 1 μg/mL Cetuximab with different concentrations of albumin, Rituximab, or ascites isolated antibodies. After centrifugation [20g (570 rpm), 1 min], tubes were placed in a water bath at 37°C for 45 min. Retrieved samples were briefly vortexed and 300 µL of ice-cold 0.5% PFA in PBS was added. Using flow cytometry, events positive for both cell trackers were interpreted as conjugates.

For the assessment of intracellular calcium, NK cells were stained with Fluo-4 (Thermo Fisher Scientific) according to the manufacturer’s specifications. Prestained NKs (250,000 cells) from original suspensions were taken and incubated in either phenol-free media and Rituximab or albumin-supplemented media in 96-well plates. This has allowed for consistent dye loading and comparison between samples. The baseline fluorescence was recorded for 1 min and then PMA (Sigma-Aldrich, 50 ng/mL) and Ionomycin (Sigma-Aldrich, 1 µg/mL) were added to induce NK cell activation. Intracellular calcium flux was measured for 10–15 min in intervals of every 3 s at 37°C using spectrophotometer Synergy2 (BioTek). Absorbance was recorded at 480/25. Data were normalized according to the average unstimulated baseline and presented as fold change of starting point measurement. Comparison between calcium flux peak values and ROC curve analysis was used to confirm significance between different measured flux curves.

Potentiometry was done to determine the concentrations of sodium, potassium, and chloride by using an ion-selective electrode on the Atellica CH Analyzer (Siemens Healthineers). Total protein (pyrogallol red method) and albumin (bromocresol green method) were photometrically determined on the Atellica CH Analyzer (Siemens Healthineers). IgG subclasses were measured by immunonephelometry on the BN-II (Siemens Healthineers). Instrument controls were performed according to the manufacturer’s instructions. Parameters are accredited according to DIN EN ISO 15189:2014. The original obtained clinical chemistry data of ascites, serum samples, and their permeates has been previously published in (24).

Healthy donor serum was fractionalized using ultracentrifugation filters (Amicon, 0.5 mL, MCWO 3 kDa), which were centrifuged at 14,000–15,000g. The centrifugation was interrupted after 15 min and the remaining ascites was resuspended before continuing centrifugation for another 15 min. For dialysis of ascites samples, dialysis tubes were filled up with ascites and dialyzed against 1 L of cell culture medium RPMI 1640 (Pur-a-lyzer, Sigma-Aldrich, 1 kDa and 25 kDa) for 24 h. Heat inactivation of untreated or dialyzed ascites was performed at 56°C for 30 min.

Cytokine concentrations in patient ascites samples were assessed using ELISA kits, according to the manufacturer’s instructions. The following cytokines were assessed: CXCL1, IL-6, IL-8, IL-10, IL-12, IL-13, IL-17, IL-27, MIF, MIP-1b, TGF-β1, TNF-α, and VEGF. All used ELISA kits were obtained from R&D Systems, Wiesbaden, Germany.

Ascites samples were diluted (1:1 ratio) with phosphate buffer saline (pH 7.2, Thermo Scientific), and protein A/G chromatography columns (Pierce, Thermo Scientific, 1 ml volume) were equilibrated with the BupH PBS. The sample was added and pushed through the column. The column was rinsed out with binding buffer and washed with elution buffer (Thermo Scientific), which released IgG1 antibodies from the column. Neutralization buffer (100 µL; 1 M Tris, pH 9) was added per 1 mL of collected fraction. After neutralization, collected fractions were gathered and concentrated until original ascites volume was reached (500 μL). The concentration of the eluted antibodies was determined by UV spectrophotometry (OD 1.4 ≈ 1 mg/mL IgG) before in vitro application.

Plotted data depict single values, mean as center value, and error bars for the standard deviation (SD). Heatmap data points represent mean values. Performed statistical tests are described in the corresponding figure legend. All graphs and statistical tests were prepared using GraphPad Prism 8 software.

For the initial assessment of antitumoral NK cell activity in ovarian cancer, we implemented an in vitro coculture system consisting of healthy donor NK cells, EGFR-positive tumor cells, and the anti-EGFR antibody Cetuximab that mediates ADCC ( Figure 1B ). In a further development, we studied different NK cell effector functions in the microenvironment of malignant ascites. For this, we supplemented our coculture with patient-derived malignant ascites after depletion of cellular components and additionally included one benign ascites derived from a patient with congestive heart failure in our study. Using this approach, we could demonstrate that crucial cytotoxic and secretory NK effector functions were substantially impaired by most of the malignant samples but not by the benign ascites sample. According to their inhibitory potential on NK ADCC, we were able to categorize all 23 ascites samples in very strong, strong, intermediate, and weak inhibitory samples and stimulatory sample ( Figure 1C ). To further study the underlying immunosuppressive mechanisms, we selected four malignant ascites samples with very strong and strong inhibitory potential (nos. 16, 21, 24, and 25) and the benign sample (no. 29) (24).

When we analyzed the content of ascites, we identified increased sodium concentration as the primary cause of immunosuppression. Using sodium channel blockers such as amiloride or lidocaine, we could confirm this inhibitory mechanism. However, their addition resulted in partial restoration of effector function (24), which indicated another potential inhibitory factor.

Considering the heterogeneous composition and the variety of proteins contained in ascites, we first assessed whether any correlations between overall protein content and suppressive activity existed. However, Pearson correlation ( Figure 2A ) and ROC analysis ( Figure 2B ) did not reveal any significant connection to NK cell ADCC inhibition. Furthermore, we observed that healthy donor serum contained higher albumin concentrations compared to patient serum and ascites ( Figure 2C ). In conjunction with our published data, these findings suggest that contribution of the protein component to the overall inhibition is lesser than electrolyte imbalance.

In order to further assess the inhibitory potential of the ascites proteinaceous component, we normalized electrolyte content in ascites samples with high and low protein content ( Figure 2D ). Using nephelometry and potentiometry, we confirmed that dialysis restored electrolyte levels to the level of culture medium, while ascites protein and antibody content remained unchanged ( Figure 2E ) (24). In selected ascites samples with high sodium (nos. 16, 21, and 24) and low sodium content (nos. 25 and 29), NK ADCC was determined before and after dialysis ( Figure 2F ). In line with our hypothesis, the normalization of imbalanced electrolytes leads to a significant restoration of NK ADCC in samples with previously high sodium content (nos. 16, 21, and 24). In contrast, NK ADCC in the presence of ascites with previously lower sodium content (nos. 25 and 29, but also no. 24) was only slightly restored by dialysis (24). These data indicate the existence of another, additional suppressive component and mechanism that becomes more relevant under conditions of no or moderate electrolyte imbalance.

To further substantiate this hypothesis, we assessed NK ADCC in the presence of heat-inactivated ascites samples. Indeed, heat inactivation restored NK ADCC only in the presence of samples with low sodium content ( Supplementary Figure 1A , nos. 9 and 20) while inhibitory property of sample no. 21 with high sodium content remained unaffected from heat inactivation. Furthermore, when already dialyzed ascites samples were heat inactivated, we could observe a further restoration of NK ADCC in ascites no. 24 and no. 29 ( Figure 2G ), supporting the hypothesis that the potential additional inhibitory factors are heat sensitive.

In the next series of experiments, we aimed to elucidate the nature of these potential inhibitory components. For this purpose, proteins in the original ascites samples were depleted using 3 kDa ultracentrifugation obtaining protein depleted permeate fraction, which contained only electrolytes and small metabolites. We could show that permeates of ascites samples with high sodium content remained inhibitory to the same extent as the original untreated sample ( Figure 2H , red) (24). Following this, we assessed NK cell ADCC in the presence of healthy donor blood serum and its protein less permeate ( Figure 2H , green). While untreated healthy donor serum exerted inhibition, no inhibition could be observed after depleting proteins. In summary, these results support the hypothesis that heat-sensitive proteins are another immunosuppressive mediator in ascites, particularly mediating immunosuppression in an electrolyte balanced environment with physiological sodium content.

Since malignant ascites is a complex tumor microenvironment containing various proinflammatory and immunosuppressive cytokines, in the following part of our study, we assessed their impact on NK effector function. First, we quantified various cytokines in 21 ascites samples using ELISA ( Supplementary Figure 1B ) and correlated their content to NK effector function ( Supplementary Figure 1C ). Among these cytokines, IL-8 exhibited significant positive correlation to NK cell ADCC ( Supplementary Figure 1D ), while IL-10 showed an inverse correlation in trend ( Supplementary Figure 1E ). However, supplementing cell culture medium with corresponding concentrations of IL-8 or IL-10 in our experimental settings had no impact on NK ADCC ( Supplementary Figure 1F ). As in our samples ( Supplementary Figures 1G, H ), malignant ascites usually contains high IL-6 and VEGF levels, which are also of prognostic relevance (12, 25). However, their addition to our experimental settings also did not have any impact on NK ADCC (data not shown).

Since ascites-related cytokines did not seem to be main regulators of cytotoxic NK effector function in our present study, we re-analyzed our original nephelometric data of ascites content. Here, we noticed that all ascites samples considerably contained IgG immunoglobulins. However, IgG content of all original ascites samples did not correlate to NK ADCC ( Figures 2I, J ). Interestingly, after analyzing the IgG content of dialyzed ascites samples with balanced electrolytes, we could reveal that IgG content was significantly inversely correlated to NK ADCC ( Figure 2K ). Importantly, not only in dialyzed samples, but also in all samples with physiological or lower sodium content (<145 mM) was IgG concentration significantly negatively correlated to NK ADCC ( Figure 2L ).

The presented data of this part of the study indicate that ascites-related proteins and IgG immunoglobulins substantially contribute to ascites-mediated immunosuppression, which is masked in a high-sodium environment but becomes evident in an environment with physiological sodium content.

In the next part of our study, we aimed to investigate whether the inhibitory mechanism of immunoglobulins is distinct from immunosuppression mediated by overall high protein content. For this purpose, we added albumin to our in vitro coculture, which is the most abundant protein in serum and ascites. Furthermore, we used Rituximab as an irrelevant antibody without specific target and ADCC function in our experimental setting, because CD20, the target antigen of Rituximab, is not expressed on the cell lines used in our model. Firstly, we added different concentrations of albumin ( Figure 3A ) or Rituximab ( Figure 3B ) to the coculture and assessed NK cell degranulation under target-independent conditions in the absence of tumor cells (PMA/Ionomycin activation, left panels, Figures 3A, B ) and under ADCC conditions in the presence of Cetuximab and EGFR-positive tumor cells (right panels, Figures 3A, B ), respectively. Our results show that albumin had the capacity to inhibit NK cell degranulation in a dose-dependent manner in both ADCC- and target-independent conditions ( Figure 3A ). In contrast, Rituximab only impaired NK cell degranulation in the presence of Cetuximab in high dosage (1 mg/mL) ( Figure 3B ).

In further experiments, we investigated the impact of albumin and Rituximab on NK–tumor cell conjugation either in the presence ( Figure 3C ) or in the absence ( Figure 3D ) of Cetuximab. According to our data, the addition of albumin caused significant dose-dependent inhibition of NK–tumor cell conjugation in both conditions (left panels, Figures 3C, D ). According to the inhibition of NK ADCC, Rituximab only inhibited conjugation in the presence of Cetuximab in high dosage (right panel, Figures 3C, D .

Since the influx of calcium is an important step in early activation of immune cells, we studied the impact of different concentrations of albumin and Rituximab on the influx of calcium in NK cells during their activation by PMA and Ionomycin. We could demonstrate that calcium influx was significantly inhibited by the addition of albumin, but not in the presence of Rituximab ( Figures 3E, F ).

In summary, these mechanistic studies indicate that proteins like albumin are able to reduce cytotoxic NK cell activity and impair NK–tumor cell conjugation by inhibiting influx of calcium in NK cells. Importantly, the inhibition was independent of the presence of ADCC-inducing antibody. In contrast, irrelevant antibodies inhibited NK cytotoxicity and conjugation only in the presence of Cetuximab, while the influx of calcium in NK cells remained unaffected. The irrelevant antibodies in our system are supposed to act as competitive inhibitors in the presence of specific targeted antibody, which is distinct from the mechanism of inhibition meditated by proteins and imbalanced electrolytes.

Since we have demonstrated that serum and ascites proteins and irrelevant antibodies can mediate inhibition, in the last part of our study, we further explored the inhibitory potential of IgG immunoglobulins isolated from ascites samples. Because of our nephelometric analysis, we correlated all ascites samples according to their sodium and IgG content ( Figure 4A ). Considering the results from our functional experiments with dialyzed and heat-inactivated samples in addition, we were able to divide our ascites cohort into four groups: Ascites sample nos. 16 and 21 displayed the highest sodium levels but a low amount of IgG immunoglobulins (upper left quadrant, red). Accordingly, the inhibitory effect of these samples was strongly heat resistant and completely restored after dialysis. In contrast, ascites samples no. 24 and 25 also showed increased sodium content, but also higher amount of IgG immunoglobulins (upper right quadrant, orange), which explained that NK suppression was restored after both dialysis and heat inactivation. The low, if any, NK suppression of benign ascites no. 29 (lower left quadrant, green) with physiological sodium and low IgG content did not change after sample dialysis but improved subsequent to heat inactivation. In contrast, impaired NK cytotoxicity in the presence of ascites sample no. 9 (lower right quadrant, blue) with increased IgG content was completely restored after heat inactivation, exhibiting similar properties to serum samples (purple dots, lower right quadrant).

To further explore the inhibitory potential of immunoglobulins on NK effector function, in the next series of experiments, we studied the impact of IgG in our coculture system. We selected ascites samples containing different combinations of sodium and IgG content ( Figure 4A , marked in yellow squares) and performed affinity chromatography to isolate the immunoglobulins. We confirmed that the isolation procedure did not lead to any significant loss of immunoglobulins using spectrometry and IgG ELISA ( Figure 4B ). After isolation, the original IgG concentration was restored using ultracentrifugation filters and ascites-derived IgGs were tested in our coculture. Initially, we observed that the isolated immunoglobulins of originally IgG-rich samples (nos. 11 and 13) significantly inhibited NK cell ADCC after 6 h of coculture ( Figure 4C ). The level of ADCC in the presence of IgGs isolated from ascites samples significantly correlated with their concentration ( Figure 4D ). The negative correlation further confirmed the suppressive effects of ascites IgGs.

Following this, we studied the effects of ascites immunoglobulins on NK cell activation and regulatory surface markers CD16 ( Figure 4E ) and DNAM-1 ( Figure 4F ). In both cases, the addition of ascites IgGs significantly prevented activation-induced downregulation of CD16 and DNAM-1 in the NK–tumor cell coculture. Considering that DNAM-1 is downregulated upon contact with its ligands CD112 and CD155, in the following experiment, we studied the impact of ascites IgG on the NK–tumor cell conjugation. Here, we noticed that IgG addition significantly impaired the conjugation of NK cells to target cells ( Figure 4G ).

In summary, the data of this study indicate the existence of two parallel inhibitory mechanisms of antitumoral NK effector functions in malignant ascites. While originally the correlation between ascites IgG and NK cell ADCC in unprocessed ascites was not significant, a negative trend could be noticed ( Figure 4H ). The trend became a significant correlation when NK ADCC was assessed in isolated IgG fraction ( Figure 4I ). This suggested that the IgG immunosupressive effect is masked by much stronger electrolyte-based inhibition and only becomes visible after dialysis or isolation.