K562 (ATCC) cells were cultured in IMDM (Gibco, Cat. No. 12440053) supplemented with 10% FCS (TICO Europe, Cat. No. FBSEU500). Capan-2 (ATCC) cells were cultured in RPMI-1640 (Gibco, Cat. No. 61870036) supplemented with 15% FCS (TICO Europe, Cat. No. FBSEU500). MiaPaca-2 (ATCC) cells were cultured in DMEM (Gibco, Cat. No. 31885023) supplemented with 15% FCS (TICO Europe, Cat. No. FBSEU500) and 2.5% horse serum (Gibco, Cat. No. 16050130). All cells were maintained at 37°C in a humidified incubator containing 21% O₂ and 5% CO₂ (Sanyo Electric).

Previously established (PANCO-11a and PANCO-22a) (27) and newly established (PANCO-45a and PANCO-51) organoid cultures were included in the current study. Tumor tissue from which the new organoids were established, was obtained from patients with pancreatic ductal adenocarcinoma who underwent resection at the Maastricht University Medical Center+ from 2017 to 2021. This study was approved by the local medical ethics committee (METC 2019-0977) and patients provided informed consent prior to inclusion. Organoids were established as previously described (27). Organoids were cultured in three droplets of 15 μL basement membrane extract (BME) (Geltrex^® LDEV‐Free Reduced Growth Factor Basement Membrane Matrix, Gibco, Cat. No. 1413202) with 500 μL of organoid Medium 1 or Medium 3 (Supplementary Table 1) per well of a 24‐wells non-tissue-treated culture plate (Corning, Cat. No. 351147) and kept in a humidified incubator at 37°C/21% O₂/5% CO₂. The medium was changed every 2–3 days and organoids were passaged approximately every 7 days (26–28). Briefly, the BME droplet domes were resuspended in ice‐cold Advanced DMEM/F12 (Gibco, Cat. No. 12634-010) supplemented with Penicillin/streptomycin (50 units/mL penicillin and 50 μg/mL streptomycin, Gibco, Cat. No. 15140-122), 10 mM HEPES (Gibco, Cat. No. 15630-080), and 1x Glutamax (Gibco, Cat. No. 35050-061) (AdvDF+++). The organoids suspensions were pelleted by 5 min centrifugation at 350 xg at 4°C and the organoids were mechanically sheared with narrowed glass Pasteur pipettes.

NK cells were isolated from buffy coats of healthy anonymous volunteers (Sanquin blood bank, Maastricht, The Netherlands). The use of buffy coats does not need ethical approval in the Netherlands under the Dutch Code for Proper Secondary Use of Human Tissue. First, peripheral blood mononuclear cells (PBMCs) were isolated by centrifugation on a Lymphoprep density gradient (Axis-Shield, Cat. No. 07851). From the PBMCs, NK cells were obtained by negative selection, using an NK cell isolation kit (Miltenyi Biotec, Cat. No. 130-092-657) according to the manufacturer’s protocol. The purified NK cells were activated with 1000 U/mL IL-2 (Proleukin, Novartis) and cultured overnight in RPMI-1640 medium (Gibco, Cat. No. 11875093), supplemented with 10% FCS, 100 U/mL penicillin, and 100 µg/mL streptomycin (Gibco, Cat. No. 15140122) in a humidified incubator at 37°C/21% O₂/5% CO₂. Effects of organoid growth medium and the culture medium supplement nicotinamide on NK cell cytotoxic capacity were assessed in a 4-hour standard cytotoxicity assay as described before (15). Briefly, the cytotoxic potential of NK cells against K562 was determined by a flow-cytometry based assay. K562 target cells, labeled with CellTracker™ CM-DiI Dye (Thermo Fisher Scientific, Cat. No. C7000), were resuspended in IMDM + 10% FCS or IMDM + 10% FCS + nicotinamide (10 mM). IL-2-activated (1000 U/mL) NK cells were resuspended in RPMI + 10% FCS and co-cultured with K562 cells at a 1:1 or a 5:1 effector-to-target cell (E:T) ratio. After 4-hour co-culture, the cells were stained with Live/Dead^® Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific, Cat. No. L34957).

The cytotoxic potential of NK cells against pancreatic tumor organoids was determined using live cell imaging (IncuCyte S3 live cell imaging system (Sartorius)). Three days before the cytotoxicity assay, organoids were dissociated into small organoids using TryplE (Gibco, Cat. No. 12605‐010) supplemented with 10 μM Rho kinase inhibitor and allowed to grow for three days into organoids of homogenous size.

On the day of the cytotoxicity assay, the small organoids were harvested by resuspending the droplet domes in AdvDF+++, and filtered using a 100 µm filter (Greiner Bio-One, Cat. No. 542000). The suspension containing organoids with >100 μm diameter size was discarded. The organoids with ≤100 µm diameter were labeled using 5 µM CellTracker Red CMTPX Dye (Thermo Fisher Scientific, Cat. No. C34552) in AdvDF+++ for 30 min at 37°C. In order to prevent 2D outgrowth of the organoids, 96-wells flat bottom culture plates were coated with ice-cold 100% BME diluted 1:1 in ice-cold PBS (50% BME-PBS scaffold), and incubated at 37°C for at least 30 minutes to solidify before seeding NK cells and organoids. Labeled organoids (≤100 µm diameter) were resuspended in culture medium without Rho Kinase Inhibitor (organoid medium-noROCKi) and counted using a Burker counting chamber. 200 organoids were seeded onto the 50% BME-PBS scaffold and NK cells were immediately added. BME was added to reach a final concentration of 10%. To quantify apoptosis, Caspase-3/7 dye (Sartorius, Cat. No. 4440) was added to each well at a final concentration of 2.5 µM. The NK cells and Caspase-3/7 dye were added immediately after the organoids were seeded and before the 10% BME in the medium was solidified, which would result in inconsistent NK cell migration towards the organoids. Moreover, the co-culture plates were briefly centrifuged (3 min at 300 xg) to ensure localization of both organoids and NK cells in one plane. We approximated that a single organoid with a diameter of 100 μm consists of 50 cells. Hence, co-cultures of 2x10⁴ or 5x10⁴ NK cells with 200 organoids represent approximately 2:1 and 5:1 effector-to-target-cell-ratios, respectively.

To assess the effects of ADCC-inducing antibodies, the organoids (≤100 µm diameter) were pre-incubated with ADCC-inducing antibodies or organoid medium-noROCKi in ultra-low attachment round bottom 96 well plates (Corning, Cat. No. 7007) for 30 min at 37°C. Trastuzumab and avelumab were diluted in organoid medium-noROCKi to the following concentrations: 1 µg/mL trastuzumab (Roche), 2 µg/mL avelumab (MedChemExpress, Cat. No. HY-108730).

Organoid-NK cell co-cultures were imaged every 30 min (using the Multi Spheroid scan settings at 10x magnification and one field/well) for 3 days using the IncuCyte S3 live cell imaging system (Sartorius). Organoid segmentation was performed using the IncuCyte S3 Software (Sartorius, version 2019B). A mask selecting the organoids was created based on brightfield images and the fluorescence signal of the CellTrackerRed. The level of apoptosis was quantified from the mean fluorescent intensity of the Caspase-3/7 dye within the organoid masks. As the brightness of the apoptotic signal at baseline varied per well, the level of fluorescence at each time point was divided by the level at baseline in each well. Moreover, the total level of apoptosis over the 24 h cytotoxicity assay was calculated from the area under the curve (AUC) of the caspase-3/7 signal [comparable to (29)]. Images and videos were generated after background subtraction using the IncuCyte S3 software (Sartorius, version 2019B).

To determine HER2 and PD-L1 surface expression, organoids were dissociated into single cells using TryplE supplemented with 1:1000 Rho kinase inhibitor. 25,000 single tumor cells were stained with Live/Dead^® Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific, Cat. No. L34957) for 30 min on ice. Tumor cells were washed with PBS/EDTA buffer (PBS/2 mM EDTA/0.5% HSA) and stained with mouse anti-human HER2-APC (Neu24.7, BD Biosciences, Cat. No. 340554), anti-PD-L1-PE (29E.2A3, BD, Cat. No. 568079), or matched isotype controls IgG1-APC (MOPC21, BD Biosciences, Cat. No. 550854) and IgG2b,k-PE (27-35, BD, Cat. No. 555058) for 30 min at 4°C. All acquisitions were performed on a BD FACS Canto II. Data were analyzed with FlowJo v10.6.1 (TreeStar).

Genomic DNA was extracted from organoid cultures and NK cells using the QIAamp DNA blood mini kit. The HLA class I genotype of the included organoid cultures and NK cell donors, and the presence of individual KIR genes for the NK cell donors were determined by Luminex- sequence-specific oligonucleotides (SSO) (One Lambda). The NK cell donor genotypes and licensing statuses are presented in Supplementary Table 2. HLA-Bw4 is recognized by KIR3DL1, HLA-C1 by KIR2DL2/3, and HLA-C2 by KIR2DL1. The matched and mismatched NK cell populations were identified based on the genotypic expression of the respective HLA epitope in organoid cultures. Moreover, mismatches were only taken into account when NK cells could be licensed through the mismatched KIR, i.e. the corresponding HLA gene was also genotypically present in the NK cell donor. Matched and mismatched conditions between the NK cell and organoid cultures are presented in Supplementary Table 3.

Statistical analyses were performed with GraphPad Prism 9 software using Kruskal-Wallis test or paired ANOVA with Tukey’s multiple comparisons test when appropriate (described per analysis). Values are expressed as mean ± standard error or deviation, where applicable. P-values of <0.05 were considered statistically significant. Group size, population information, and error bar definitions are indicated in the figure legends.

To investigate NK cell cytotoxicity towards pancreatic cancer organoids, we established a three-dimensional co-culture method that can be monitored in real-time using brightfield and fluorescent imaging. The included organoid cultures demonstrated heterogeneous phenotypes: one grew in solid cell clusters (PANCO-45a), while others (PANCO-11a, PANCO-22a, PANCO-51) had the more common cystic structures (27). To test the capability of NK cells to migrate towards organoids within culture medium containing BME, we first optimized the seeding procedure of the organoids to be used in the cytotoxicity assays. Based on a previous publication (30), we seeded organoid suspensions on top of a solidified BME scaffold. Immediately after seeding (t=0 h), the organoids, showed a normal morphology (Supplementary Figure 1A). Moreover, we observed that this setup was sufficient to prevent organoid 2D outgrowth for the duration of the cytotoxicity assays (24 hours, Supplementary Figure 1A).

Next, we optimized the seeding procedure of the NK cells to be used in the cytotoxicity assays. Initially, the organoids were seeded and allowed to settle for 24 h, during which the 10% BME in the organoid medium solidified as well. After 24 h, the NK cells in suspension were added and apoptosis was quantified by live cell imaging (Supplementary Figure 1B). When the BME within the organoid medium was pre-solidified, NK cells migrated towards the organoids but localized at different depths within the culture well. This hampered focusing on the correct plane during live cell imaging. Subsequently, we seeded NK cells and organoids at the same time on top of the BME scaffold. This resulted in the localization of both NK cells and organoids in one plane (Supplementary Figure 1C), allowing accurate focusing during live cell imaging. When NK cells were tracked for the duration of the cytotoxicity assay, efficient trafficking of NK cells towards organoids could be observed. We therefore continued with this setup (Figure 1A).

Organoids grow in medium containing nicotinamide, which has been shown to inhibit the cytotoxic capacity of the NK92 cell line (30). We therefore tested whether NK cell viability and cytotoxicity were hampered by the nicotinamide in the organoid medium. NK cells were co-cultured with 2D grown K562 cells, a cell line that is highly sensitive to NK cells and that is commonly used as positive control cell line for NK cell killing. Flowcytometric analysis showed that the addition of nicotinamide to the K562 culture medium did not affect NK cell viability nor did it significantly hamper killing of K562 during a standard cytotoxicity assay (Supplementary Figures 1D, E).

To assess the cytotoxic potential of NK cells in the 3D co-culture system, we performed cytotoxicity assays with four independent pancreatic tumor organoid cultures and different doses of NK cells. After 24-hours of co-culture, organoid 2D outgrowth increased (Supplementary Figure 1A). Hence, the cytotoxicity assay was analyzed up to 24 hours as the later time points yielded less accurate data due to 2D growth.

Organoid death was quantified by assessing the level of apoptosis in organoids cultured alone (Video 1) or together with 2x10⁴ or 5x10⁴ NK cells (Videos 2, 3). The addition of 2x10⁴ NK cells to 200 organoids did not significantly increase the level of apoptosis in the organoids (1.6 ± 0.7 vs 1.4 ± 0.6, p=0.860 vs No NK cell control (Figures 1B, C). However, co-culturing 5x10⁴ NK cells with 200 organoids significantly increased cytotoxicity among the organoids compared to organoids not co-cultured with NK cells (2.1 ± 0.8 vs 1.4 ± 0.6, p=0.005) (Figures 1B, C). Overall, NK cells induced more organoid apoptosis at higher NK-to-organoid ratios, demonstrating that NK cells can effectively target patient-derived pancreatic cancer organoids.

NK cell induced apoptosis of organoids did not result in complete eradication of organoids and surviving organoids were able to continue expanding (Supplementary Figure 2). Therefore, we aimed to enhance NK cell-induced cytotoxicity by ADCC using the clinically available mAbs against PD-L1 (avelumab) and HER2 (trastuzumab). We first validated the surface expression of PD-L1 and HER2 protein on patient-derived pancreatic tumor organoid cultures by flow cytometry (gating strategy in Supplementary Figure 3A). All organoid cultures displayed PD-L1 surface expression (Figures 2A, B). The ratio between the median fluorescence intensity (MFI) from the α-PD-L1 antibody divided by the MFI of the isotype controls indicated that organoid cultures PANCO-11a and PANCO-22a had slightly higher PD-L1 expression compared to PANCO-45a and PANCO-51 (1.7 ± 0.1 and 1.6 ± 0.2 vs 1.4 ± 0.3 and 1.3 ± 0.1), respectively). Furthermore, all organoid cultures tested expressed HER2 on the cell surface (Figures 2C, D). PANCO-45a consistently had the highest HER2 expression (40.5 ± 11.8), followed by PANCO-11a (25.6 ± 8.4). PANCO-22a and PANCO-51 had the lowest expression (17.2 ± 2.0 and 15.6 ± 5.7) (Figures 2C, D). We compared the PD-L1 and HER2 expression levels of the organoids with two conventional pancreatic cancer cell lines: Capan-2 and MiaPaca-2. PD-L1 expression was comparable between the organoid cultures and MiaPaca-2, while Capan-2 had approximately 10-fold higher expression compared to the organoid cultures and the MiaPaca-2 cell line (Supplementary Figure 3B). Comparing HER2 expression on organoids with cell lines indicated that PANCO-45a had the highest expression level. Capan-2 cells had comparable expression levels with PANCO-11a, whereas MiaPaca-2 had the lowest expression levels, which was even lower compared to the organoid cultures (Supplementary Figure 3C). As all organoids expressed PD-L1 and HER2 on their surface, treatment with ADCC-inducing antibodies may increase NK cell cytotoxicity.

We next assessed whether the lytic activity of donor NK cells could be enhanced in vitro following pre-treatment of organoids with the ADCC-inducing antibodies avelumab (anti-PD-L1) or trastuzumab (anti-HER2). In the absence of NK cells, pre-incubation of organoids with trastuzumab or avelumab did not increase apoptosis, indicating the absence of direct toxicity mediated by the monoclonal antibodies (Supplementary Figures 4A, B).

As compared to the condition with only avelumab, co-culture with 2x10⁴ NK cells after pre-incubation with avelumab significantly increased organoid apoptosis (1.2 ± 0.2 vs 2.0 ± 0.5, p=0.001) (Figures 3A, B). Moreover, the level of apoptosis was dependent on the E:T ratio, increasing up to 5.2-times after co-culture with 5x10⁴ NK cells (3.4 ± 0.9, p<0.0001 vs avelumab only, p=0.003 vs 2x10⁴ NK cells + avelumab). Moreover, combination therapy with 5x10⁴ NK cells and avelumab resulted in complete disintegration of some organoids (Figure 3B white arrows).

Compared to the trastuzumab only control, organoid apoptosis was significantly increased upon pre-incubation with trastuzumab in the presence of 2x10⁴ NK cells (1.4 ± 0.4 vs 2.5 ± 0.9, p=0.005) (Figures 3C, D). The level of apoptosis was dependent on the NK cell number since addition of 5x10⁴ NK cells resulted in up to 7.3-times higher apoptosis compared to baseline (4.6 ± 1.6), which was significantly higher compared to the level of apoptosis observed with trastuzumab only or 2x10⁴ NK cells with trastuzumab (p=0.0001 vs trastuzumab only, p=0.001 vs 2x10⁴ NK cells + trastuzumab). Moreover, combination therapy with 5x10⁴ NK cells and trastuzumab resulted in complete disintegration of the majority of organoids (Figure 3D).

Comparing the ADCC conditions with the conditions without ADCC-inducing antibodies showed that the level of NK cell-mediated cytotoxicity was significantly enhanced upon pre-incubation with either avelumab or trastuzumab. On average, the level of organoid apoptosis was low when 2x10⁴ NK cells were used (Figures 4A, B). Moreover, pre-incubation of the organoids with avelumab did not significantly increase apoptosis compared to the condition without ADCC (2.0 ± 0.5 vs 1.6 ± 0.7, p=0.4). However, the inclusion of trastuzumab increased apoptosis compared to controls without ADCC (2.5 ± 0.9 vs 1.6 ± 0.7, p=0.005), but not compared to the avelumab condition (2.5 ± 0.9 vs 2.0 ± 0.5, p=0.16). Co-culture with 5x10⁴ NK cells in combination with either avelumab or trastuzumab resulted in higher levels of apoptosis compared to the conditions without ADCC-inducing antibodies (2.1 ± 0.8 for No ADCC vs 3.4 ± 0.9, p=0.004 and vs 4.6 ± 1.6, p=0.001 for avelumab and trastuzumab, respectively) (Figures 4C, D). Moreover, the level of apoptosis induced by trastuzumab was higher compared to that induced by avelumab (4.6 ± 1.6 vs 3.4 ± 0.9, p=0.01) (Figures 4C, D). In addition to increasing the caspase-3/7-derived fluorescent signal, combination therapy with NK cells and ADCC-inducing antibodies resulted in complete disintegration of organoids, which was especially clear for the conditions with trastuzumab (Figures 3B–D, 4D white arrows, Video 3-6). While variances in cytotoxic capacity amongst the NK cell donors were present, we did not observe NK cell donors with consistently higher cytotoxic capacity (Supplementary Figure 5A). Instead, color coding of organoid cultures suggested the presence of NK sensitive and resistant organoids (Supplementary Figure 5B).

We next studied the efficacy of NK cell therapy in combination with ADCC-inducing antibodies by tracking the level of organoid apoptosis during the 24 h live cell imaging assay. For this, the total level of apoptosis for each organoid culture that was induced by 5x10⁴ NK cells with or without ADCC was calculated as the area under the curve (AUC) in the 24 h cytotoxicity assay. In organoid cultures PANCO-11a, PANCO-22a, PANCO-45a, and PANCO-51, the combination of NK cells with trastuzumab resulted in significantly higher total apoptotic signal compared to the No ADCC control (p=0.01, p=0.02, p=0.01, and p<0.0001, respectively) (Figure 5). For organoid culture PANCO-11a, avelumab increased the total level of apoptosis, but this increase did not reach significance (p=0.06). The total level of apoptosis induced by NK cells with trastuzumab was higher compared to avelumab, but this again did not reach the level of significance (p=0.0524) (Figures 5A, B). For organoid cultures PANCO-22a, PANCO-45a, and PANCO-51, the total apoptotic signal from avelumab-induced ADCC was significantly higher compared to the conditions without ADCC (p=0.02, p=0.0003, and p=0.015, respectively) (Figures 5C–H). Moreover, for these organoid cultures, total apoptosis mediated by trastuzumab-induced ADCC was significantly higher compared to the avelumab-induced ADCC (p=0.04, p=0.0295, and p=0.001, respectively) (Figures 5C–H).

Next, we addressed potential heterogeneity in response to NK cell treatment between the pancreatic cancer organoids. Treatment of organoids with only NK cells resulted in significant differences in total apoptotic signal between PANCO-11a and PANCO-22a (34.8 ± 0.6 vs 28.6 ± 0.4, p=0.001); PANCO-11a and PANCO-45a (34.8 ± 0.6 vs 41.4 ± 0.6, p=0.003); PANCO-22a and PANCO-45a (28.6 ± 0.4 vs 41.4 ± 0.6, p=0.001); and PANCO-22a vs PANCO-51 (28.6 ± 0.4 vs 41.4 ± 2.1, p=0.02) (Figures 6A, B).

The addition of avelumab to NK cell treatment resulted in higher total apoptotic signal for PANCO-45a and PANCO-51 compared to PANCO-11a and PANCO-22a (56.1 ± 0.8 and 51.5 ± 1.1 vs 40.1 ± 0.8 and 42.7 ± 1.4, p=0.003 and p=0.01; and p=0.01 and p=0.03, respectively). Therefore, PANCO-45a and PANCO-51 were the organoid cultures with the highest response to NK cell with avelumab treatment from the four organoid cultures tested (Figures 6C, D). Interestingly, the level of PD-L1 expression was not in line with the level of sensitivity of the organoids towards NK cell treatment with avelumab. PANCO-45a and PANCO-51 had the lowest PD-L1 expression levels, while these organoid cultures were most responsive to NK cell therapy with avelumab.

The heterogeneity in response between the pancreatic cancer organoid cultures was further emphasized in the conditions treated with NK cells and trastuzumab. The addition of trastuzumab to NK cell treatment resulted in higher total apoptotic signal for PANCO-45a and PANCO-51 compared to PANCO-11a and PANCO-22a (78.4 ± 2.6 and 72.6 ± 1.6 vs 48.1 ± 1.7 and 56.1 ± 2.4, p= 0.001 and p=0.003; and p=0.01 and p=0.04, respectively). Similar to the observations with avelumab, PANCO-45a and PANCO-51 were the organoid cultures with the highest response to NK cell with trastuzumab treatment from the four organoid cultures tested (Figures 6E, F). However, the level of HER2 surface expression was not in line with the level of response after addition of trastuzumab, as the level of HER2 expression was higher for PANCO-11a compared to PANCO-51, whereas the apoptotic signal upon NK cells with trastuzumab treatment was consistently higher for PANCO-51 compared to PANCO-11a.

NK cell activation is determined by the balance between inhibitory and activating signaling it receives from a potential target cell. The interaction between the human killer cell immunoglobulin-like receptors (KIR) and human leukocyte antigen (HLA)-Class I is an important inhibitory signal for NK cells and thus an important regulator of NK cell activity. Introducing a mismatched situation between the NK donor-KIR and patient-HLA would create a situation where inhibitory signaling is removed, thereby reducing the amount of necessary activating signaling. We hypothesized that an increased number of KIR-HLA ligand mismatches between NK donors and organoids could result in stronger NK cell effector function. Therefore, we performed HLA and KIR typing as an initial approach to decipher a potential mechanism that could affect organoid sensitivity to NK cell treatment. The combinations of organoid cultures and NK cell donors were ranked based on the number of mismatches (ranging from no mismatches to two mismatches). We assessed the ability of 2x10⁴ and 5x10⁴ NK cells to induce apoptosis, calculated as area under the curve (AUC), and compared it to the AUC of the No NK cell control. We did not observe a clear relationship between the number of matched HLA-KIR pairs versus mismatched pairs at the genomic level and the level of NK cell cytotoxicity (Figure 7).