Human ILC3 Exert TRAIL-Mediated Cytotoxicity Towards Cancer Cells

Group 3 helper Innate Lymphoid Cells (ILC3s) are cytokine-producing lymphocytes that respond to stress signals released during disturbed tissue homeostasis and infection. Upon activation, ILC3s secrete IL-22 and IL-17, and orchestrate immune responses against extracellular pathogens. Their role in cancer remains poorly explored. To determine their anti-cancer effector potential, we co-cultured cytokine-activated human ILC3s with cancer cells of different origins. ILC3s were able to directly respond to tumor cells, resulting in enhanced IFN-γ production. Upon tumor cell encounter, ILC3s maintained expression of the transcription factor RORγt, indicating that ILC3s preserved their identity. ILC3s were able to directly kill both hepatocellular carcinoma and melanoma tumor cells expressing cell-death receptor TRAILR2, through the activation of Caspase-8 in target cells. Moreover, liver-derived cytokine-activated ILC3s also expressed TRAIL and were able to eliminate hepatoblastoma cells. Together, our data reveal that ILC3s can participate in anti-tumor immune response through direct recognition of tumor cells resulting in IFN-γ release and TRAIL-dependent cytotoxicity. Thus, ILC3s might be ancillary players of anti-tumor immunity in tissues, acting as primary responders against transformed or metastasizing cells, which might be further exploited for therapies against cancer.


INTRODUCTION
The family of innate lymphoid cells (ILCs) is a heterogeneous group of immune cells belonging to the innate immune system. They lack the somatic rearrangements of the antigen receptor genes, as described for B or T cells, and comprise cytotoxic natural killer (NK) cells, lymphoid tissue inducer cells (LTis), and several subsets of helper innate lymphocytes (1). LTis and helper ILCs express CD127 (IL-7Ra) and, unlike NK cells, depend on the cytokine IL-7 for their development (1,2).
ILC3s support tissue repair and immunity against extracellular pathogens (13). However, their role in the tumor immunity is incompletely understood. Several studies reported ILC3 infiltration into the tumor tissue (14)(15)(16), but data so far indicate an ambiguous role of ILC3s in cancer (17). It has been shown that ILC3s can both promote and inhibit tumor growth (18,19). In particular, they were reported to promote the development of colon cancer (19) and the proliferation of pancreatic cancer cells (16) via the production of IL-22. In the context of liver cancer, it was shown that IL-23-stimulated mouse ILC3s promoted the development of hepatocellular carcinoma (HCC) via IL-17 production (20). In addition, ILC3s were described to facilitate the formation of lymph node metastasis of breast cancer via the interaction with stromal cells (21). On the other hand, NKp46+ ILC3s were crucial for IL-12dependent tumor suppression in a mouse model of melanoma (18). Recently, it was shown that tonsillar CD94+ ILCs could kill human leukemia cells, and IL-12 could induce this phenotype in noncytotoxic fraction by upregulating cytolytic machinery and several NK cell-expressed molecules, including CD94 and Eomes (22). Moreover, human NKp44+ ILC3s were shown to support the formation of protective tertiary lymphoid structures and thereby constrain tumor growth of lung (14) and colorectal cancer (23).
Here, we demonstrate a direct cytotoxic effector potential of human ILC3s towards tumor cells. Our data reveal that bloodderived human ILC3s can directly respond to tumor cells by increased production of IFN-g, and cytotoxicity mediated via the interaction of TRAIL and its receptor on cancer cells. We show that both intra-hepatic and blood-derived cytokine-stimulated ILC3s are able to kill hepatic tumor cells, which holds promise for the exploitation of the cytolytic and cytokineproducing potential of ILC3s as additional immune effectors against cancer.
Generation of CRISPR/Cas9-Mediated TRAILR2 Knockout Cell Lines TRAILR2-deficient cancer cell lines were generated as described (25). HepG2, HuH6 and SK-Mel-37 cells were first transduced with the lentiCas9-Blast (Addgene #52962) vector, encoding Cas9 protein and Blasticidin-resistance gene, used to select Cas9-expressing cells. For generating the TRAILR2 knockout cell lines, sgRNA were designed using the https://design. synthego.com/#/ website. Three sgRNA with the lowest offtarget ratio were selected and cloned individually into pLenti_HDCRISPR_NoC sgRNA-puromycin expression vector (kindly provided by Prof. Dr. Michael Boutros, DKFZ, Heidelberg). SgRNA-containing vectors were transfected separately into Cas9-expressing cell lines using jetOPTIMUS (Polyplus #117-15). After selection with 2.5 μg/ml of Puromycin (Sigma-Aldrich) for 2 days, the expression of TRAILR2 was measured using flow cytometry, and TRAILR2negative population was purified by flow cytometric sort from the cell lines with the highest knockout efficiency.

ILC3 Isolation and Culture
Buffy coats from healthy donors were obtained from DRK-Blutspendedienst Baden-Württemberg-Hessen (Mannheim, Germany). Written informed consent from the blood donors was obtained and ethical approval 87/04 was granted by the Ethik Kommission II of the Medical Faculty Mannheim (Mannheim, Germany). Adjacent non-cancerous tissue from human livers was acquired from patients undergoing surgical resection for hepatocellular carcinoma (HCC) or colorectal cancer liver metastasis. The collection and use of patient material was approved (2012-293N-MA) by the Ethik Kommission II of the Medical Faculty Mannheim. The samples were collected according to the principles of the Declaration of Helsinki. All patients gave their written consent.
Liver samples were cut into small pieces and were enzymatically digested using a gentleMACS ™ Octo Dissociator (Miltenyi) to obtain cell suspension. Mononuclear cells from liver and blood were enriched using a gradient centrifugation, and red blood cells lysis. To purify ILCs, non-ILCs (CD3+, CD19+, CD14+ and CD16+ cells) were depleted using magnetic beads (Miltenyi). Blood and liver ILC3s and NK cells were then isolated from ILCenriched and ILC-depleted fraction, respectively, using BD FACSAria ™ Fusion cell sorter.

Co-Culture of ILCs or NK Cells With Tumor Cell Lines and Functional Assays
ILC3s and NK cells were washed and seeded at the indicated ratios with target cells into 96-well v-bottomed plates (Sarstedt) for 24 h in media containing 10 U/ml IL-2 (NIH). The supernatants were collected for further analysis, while the cells were harvested and used for flow cytometry.
To detect cytokines, collected cell-free supernatants were stored at -20°C until further analysis. IFN-g cytokine concentrations were measured using ELISA (Biolegend) according to manufacturer's instructions. Measurements of cytokines IL-22, IL-8, GM-CSF, and TNF-a were performed using a bead-based assay (Legendplex, Biolegend) according to the manufacturer's instructions.
To determine target cell killing, LDH-Glo ™ Cytotoxicity Assay Kit (Promega) was used according to the manufacturer's instructions. ILC3s and NK cells were seeded either alone (control) or together with 1000 target cells in 96-well vbottomed plates, in a total volume of 100 μl of RPMI 1640 AQmedia ™ supplemented with 10% human serum, 1% penicillin/streptomycin and 10 U/ml IL-2. The conditions were run in duplicates. The purity of ILC3s and NK cells (95-98%) was confirmed by flow cytometry. Spontaneous release of the lactate dehydrogenase (LDH) was measured from cultures of target cells without effectors. In order to determine the maximum release of LDH, target cells were lysed in 0.2 % Triton X-100 (Sigma-Fluka) for 15 min at RT. All controls (including medium) were set up in quadruplicates. Luminescence was measured using a plate reader (Tecan). For every time point, LDH activity measured from medium was subtracted from values of all other conditions, and LDH activity in supernatant of effector cells cultured alone was subtracted from the co-culture-derived values. Finally, % specific cytotoxicity was calculated as 100 x [(experimental LDHminimum LDH)/(maximum LDH-minimum LDH)].

Time-Lapse Live-Cell Imaging
Time-lapse live-cell imaging analysis was performed as recently published (24). HeLa-CD48 cells, which express NES-ELQTD-GFP-T2A-NES-VGPD-mCherry and CD48, a ligand for 2B4, were seeded into a silicon-glass microchip (size of wells 350 μm). We have confirmed that effector cells (cytokine-stimulated ILC3s) did not express the 2B4 receptor (Supplementary Figure 1A). Tumor cells were left to adhere overnight and then washed once with medium. The microchip was placed into an incubation chamber (37°C, 5% CO 2 and humidity device S1) at the microscope. 20000 ILC3s resuspended in RPMI 1640 AQmedia ™ containing 10% human serum, 1% penicillin/streptomycin and 10 U/ml IL-2, were added into the microchip (2-10 ILC3s per micro-well). Imaging was performed using ApoTome System with Axio Observer 7 microscope (Zeiss) with a 20x/0.8 Plan-Apochromat objective, equipped with an incubator chamber and environmental control system (37°C), in a humidified atmosphere with 5% CO 2 . Fluorescence of GFP and mCherry was acquired in line sequential mode. To visualize mCherry and GFP, we used the Colibri7 LED-module 561 (filter set 64 HE LED) and the LED-module 488 (filter set 38 HE LED), respectively. Brightfield was acquired with the TL LED module. Images were captured at a rate of three z-stack every 3 min and z-slices were spaced by 1 μm. Images were acquired with the camera Axiocam 560 mono. All confocal images were analyzed using the ImageJ software (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https:// imagej.nih.gov/ij/, 1997-2018), as previously described (24,26). Briefly, reporter cleavage was quantified based on the redistribution of the fluorescence signal from cytosol to nucleus in image stacks from different time points. The fluorescence signal intensity (I) was measured in the cell nucleus and normalized to the cytosolic signal over time (t). The background was subtracted.

Statistical Analysis
Data were tested for normal distribution using Shapiro-Wilk test. Statistical differences were then evaluated using the appropriate statistical test indicated in each Figure legend. Data are corrected for multiple comparison testing when necessary. Compared experimental groups were considered to be significantly different when p<0.05.

ILC3s Increase IFN-g Release in Response to Tumor Cells
To address the ILC3-cancer cell interaction, we first established a co-culture system using ILC3s derived from peripheral blood of healthy donors and hepatic cancer cell lines. In concordance with the literature (27,28), freshly-isolated circulating ILC3s, gated as viable, CD45+, CD127+, lineage negative (CD3-, CD19-, CD14-, CD34-, CD94-, CD123-, TCR a/b-, TCR g/d-, FCϵR1a-) cells, that express c-Kit, but not the ILC2 marker CRTH2, did not express RORgt ( Figure 1A). Neither expression of the transcription factor Eomes, nor the surface receptors NKp80, CD94, CD16 and NKp44 were detected, while cells expressed T-bet ( Figure 1A and Supplementary Figure 1B). These cells were previously reported to display a multipotent phenotype with the ability to differentiate into all ILC lineages (27). Here, we show that short exposure (16-24h) to type 3-stimulating cytokines IL-1b, IL-23, and IL-2 increased both T-bet and RORgt expression ( Figure 1A and Supplementary Figure 1B). These RORgt-expressing ILC3s produced GM-CSF, IL-8 and IFN-g, as well as lower amounts of IL-22 and TNF-a, while IL-17 and IL-6 were not detectable ( Figure 1B). Coculture with the hepatoblastoma or hepatocellular carcinoma cell lines, HuH6 and HepG2, respectively, increased IFN-g release by ILC3s, without affecting the production of other cytokines ( Figure 1B). Thus, ILC3s have the ability to directly recognize and respond to tumor cells, leading to increased IFN-g production.

ILC3s Exert Cytotoxicity Against Tumor Cells
While NK cells are defined as cytolytic ILCs, able to directly kill tumor cells either via release of cytotoxic granules or through engagement of death-receptor ligands, ILC3s are considered to be cytokine producers (1,29). So far, we observed that ILC3s can recognize and respond to tumor cells, resulting in increased IFN-g production. Next, we investigated whether ILC3s could induce lysis of cancer cells, by measuring enzyme activity of lactate dehydrogenase (LDH) released by tumor cells over time in supernatants of ILC3-tumor cell co-cultures. Indeed, the presence of ILC3s increased lysis of HuH6 cells in a timedependent manner (Figure 2A, left). Expression of RORgt by ILC3s remained stable during the co-culture (  Although NK cells are considered prototypical cytolytic antitumor effectors, when exposed to IL-1b, IL-23 and IL-2, ILC3s induced significantly higher cancer cell lysis compared to NK cells ( Figure 2C). In the absence of IL-1b and IL-23, and when concentration of IL-2 is increased for optimal NK cell activation (400 U/ml) (30,31), NK cell cytotoxic activity was restored ( Figure 2C). IL-2 (400 U/ml) might also favor ILC3 cytotoxic responses leading to increased tumor cell lysis ( Figure 2D). These data indicate that ILC3s have the capacity to directly kill tumor targets, and that the microenvironmental condition might dictate the extent of their responses and the relative contribution of different effector cells to target cell lysis.

ILC3 Cytotoxicity Is Mediated by Caspase-8
To dissect the mechanisms of ILC3-mediated the cytotoxicity, we first investigated the expression of death-receptor ligands and cytolytic effector molecules by ILC3s. Purified ILC3s exposed to IL-1b, IL-23, and IL-2 expressed TRAIL, but not FasL or Perforin ( Figure 3A and Supplementary Figure 2). Granzyme B was detected only in a subset of cells ( Figure 3A). Similar to RORgt,  TRAIL expression was not detected in freshly-purified ILC3s ( Figure 3B). TRAIL induction required the presence of IL-1b, while the amount of T-bet and RORgt could be increased when ILC3s were stimulated with IL-1b or IL-23 or IL-2 cytokines individually ( Figure 3B). It was previously shown that the TNFSF10 gene associated with H3K4Me2 modification in c-Kit+ ILCs from peripheral blood, which marks active and poised regulatory elements (27). At the same time, the authors did not detect TNFSF10 among genes whose expression and poised epigenetic state overlaps, indicating that TRAIL-encoding transcript was also not detected in these cells. These data indicate transcriptional regulation of TRAIL expression by IL-1b.
To determine which pathway is used by ILC3s to lyse tumor cells, we applied a recently described method of time-lapse live-cell imaging (24). As tumor targets, we used the reporter HeLa cell line stably transduced with the NES-ELQTD-GFP-T2A-NES-VGPD-mCherry construct (26), leading to the equimolar expression of two fluorescent reporters: Granzyme B-sensitive nuclear export signal (NES) fused to mCherry and Caspase-8-sensitive NES fused to GFP. The NES retains both reporters in the cytosol, unless cleaved by Granzyme B or Caspase-8, which results in the diffusion of the fluorescent protein to the nucleus ( Figure 3C). As such, Granzyme B and Caspase-8 activities can be assessed by measuring the corresponding fluorescence in the nucleus (24).
Measurements of the fluorescence signal in the nucleus of HeLa-reporter cells co-cultured with ILC3s, revealed an increased GFP compared to mCherry signal in the target cell nuclei at the event of cell death ( Figures 3D-F), indicating mainly a Caspase-8-mediated apoptosis. Tumor cell apoptosis followed ILC3 and target cell interaction, indicating direct engagement of cell-cell contact between ILC3s and tumor cells ( Figures 3D, E and Supplementary Video 1). Moreover, ILC3s did not appear to behave as serial killers, as during the indicated observation time, one ILC3 killed only one HeLa cell ( Figure 3E and Supplementary Video 1). Taken together, these data demonstrate the ability of ILC3s to directly kill target cells via a Caspase-8-mediated mechanism.

ILC3s Kill Tumor Cells via TRAIL-TRAILR2 Pathway
Since stimulated ILC3s expressed TRAIL, but not FasL, we next investigated whether TRAILR1 or TRAILR2 were expressed on the tumor cell lines HuH6, HepG2 and SK-Mel-37. While TRAILR1 was not detected, all three cell lines expressed TRAILR2 ( Figure 4A). To investigate the role of TRAIL during the ILC3-mediated killing, we generated TRAILR2deficient tumor cell lines using a CRISPR/Cas9-mediated gene deletion. All cells lines showed comparable apoptosis when exposed to TNF-a and cyclohexamide, indicating unaltered apoptotic machinery (Supplementary Figure 3A). Absence of TRAILR2 abrogated the lysis of tumor cells by ILC3s ( Figure 4B). These results indicate that ILC3s use TRAIL-TRAILR2 receptor engagement to induce lysis of tumor cells. In support to these data, freshly-purified TRAIL neg ILC3s were unable to induce tumor cell lysis to comparable levels as cytokine-activated ILC3s (Supplementary Figure 3B).

Intra-Hepatic ILC3s Are Able to Kill Hepatic Tumor Cells
To determine if ILC3s can contribute to anti-tumor immune responses in tissues, we isolated ILC3s from liver tissue, and determined their lytic capacity. These tissue-resident ILC3s expressed RORgt ( Figure 5A), but required stimulation with cytokines to upregulate TRAIL ( Figure 5B). TRAIL-expressing ILC3s induced HuH6 target cell lysis, which increased over time ( Figure 5C), while maintaining stable RORgt expression. These data indicate that, in the presence of type 3-stimulating cytokines, liver-resident ILC3s can participate in tumor cell lysis.

DISCUSSION
The function of tumor-infiltrating ILC3s, especially in primary human cancers is poorly explored. Although some studies reported the presence of ILC3s in tumor tissue (14)(15)(16), their role remains ambiguous. While described as tissue-protective, the ILC3 signature cytokine, IL-22, was shown to promote carcinogenesis, which depended on the balance between tissue regeneration and tumor development in the intestine (32). Conversely, the formation of tertiary lymphoid structures, supported by ILC3s, presented benefits for lung (14) and colorectal (23) cancer patients.
ILCs are considered to be tissue-resident and the first responders to tissue perturbation. While local alarmins can provoke early responses of resident ILCs, the inflammatory reaction can lead to an increase in ILC numbers in inflamed tissue. Currently, the relative contribution of local proliferation versus differentiation from tissue precursors to the increased ILC numbers during an immune reaction remains unclear. Moreover, it was also shown that certain ILC subsets could circulate towards sites of inflammation to support ongoing immune responses (33). Recently, two publications suggested that human blood ILCs provided a pool of precursor cells, which can be recruited to the tissues and develop into mature ILCs under local environmental cues (27,28). Lim et al. showed that c-Kit+ CRTH2-ILCs from the blood did not express RORgt, unless exposed to IL-23 and IL-1b (27,28), which is in concordance with our observations. Furthermore, ILC2s and ILC3s could be effectively recruited to the type 1 responses, by acquiring the ability to produce IFN-g through cellular plasticity, including upregulation of transcription factor T-bet (3,34). Cytokines, such as IL-12 and/or IL-15 could re-shape ILC functions, broadening the spectrum of their functional responses (3,35,36). Indeed, in the presence of IL-12, NKp46+ ILC3s suppressed tumor growth of mouse melanoma (18). Recently, it was shown that IL-12 could support the acquisition of ILC3 cytotoxicity by tonsillar ILC3s by inducing the expression of cytolytic molecules, along with several NK-expressed molecules, including CD94 and Eomes (22). Of note, these ILC3s preserved expression of RORgt. Similarly, CD127+CD94+ ILCs identified in human lamina propria were reported to secrete granulysin in response to IL-15 and to accumulate in Crohn's disease patients (37). The ability to eliminate K562 cells or to degranulate upon their encounter was also demonstrated earlier for human tonsillar and intestinal intraepithelial ILCs (38), NKp44+ ILC3s exposed to IL-12 and IL-15 in vitro (36), and for tonsillar CD94+ ILCs, generated from non-cytotoxic fraction when exposed to IL-12 (22). In these studies, ILCs were derived from tissues, such as pediatric tonsils or intestine, where their phenotype was pre-shaped by local microenvironments, and frequently further expanded using feeder cells and cytokines. In contrast, peripheral blood Lin-CD127+c-Kit+ cells, that by phenotype correspond to tissue ILC3s, were shown to rather display "naive" progenitor-like features and were able to generate other ILC lineages in vitro (27). Accordingly, these cells neither expressed ILC lineagedetermining transcription factors, such as RORgt, GATA-3 or Eomes, receptors NKp44, NKp80, CD16 or CD94, nor were able to secrete IL-17 and IL-22, in contrast to their gut-derived counterparts (27). In addition, NKp44-(CD62L-) tonsillar ILC3 subset was also suggested to represent a naive ILC3 state and did not produce IL-17 and IL-22 in response to IL-23 and IL-1b (39). Here, we describe that peripheral blood-derived ILC3-like cells can acquire RORgt and TRAIL expression upon short 16-24h exposure to type 3 cytokines. In contrast to previously described CD94+ cytotoxic ILCs, these cells rather exerted perforinindependent, TRAIL-mediated killing of tumor targets. IL-1b was necessary to induce expression of TRAIL, while NK cell-like features (Eomes, NKp80, CD16, and CD94) were not detected. In addition, ILC3s can produce TNF-a, which might also contribute to Caspase-8 activation in susceptible target cells (40). These data together indicate that ILC3 functional specialization might be even broader than previously anticipated, allowing these cell population to acquire different modes of cytolytic responses (perforin vs TRAIL) driven by different environmental conditions (type 1 vs type 3 cytokines). Type 1 immune responses are considered beneficial for effective anti-tumor immunity, and NK cells and T cells (CD8+ cytotoxic T lymphocytes and CD4+ Th1 cells) are considered prototypical producers of IFN-g. Recently it was shown that blood-and tonsil-derived ILC3s were also able to produce IFN-g in response to IL-23 and IL-1b (10). Here, we show that ILC3 IFN-g production can be further increased upon interaction with tumor cells. Therefore, IFN-g+ ILC3s might contribute to a proinflammatory environment that promotes efficient anti-tumor responses. It remains to be determined which recognition mechanisms and signaling machineries contribute to tumorinduced IFN-g release by ILC3s. It was previously shown that subset(s) of ILC3s can express certain NK cell receptors, including NKp44 and NKp30 (11), whose ligands are frequently upregulated by malignant cells. Moreover, ILC3s could also be regulated by soluble factors, including both cytokines and metabolites that can be released by tumor cells. In the tumor microenvironment, these factors could additively contribute to ILC3 functions.
In addition to IFN-g release upon tumor cell encounter, here, we show that both blood-and tissue-derived ILC3s were able to directly lyse tumor cells. This function was exerted under type 3 conditions, where ILC3s maintained RORgt expression, therefore excluding their differentiation towards an ILC1-like phenotype (1). The transcription factor T-bet that is detected in freshly-isolated ILC3s and upregulated upon their exposure to cytokines, might contribute to the ability of these cells to secrete IFN-g. In this microenvironment, NK cells were unable to induce comparable target cell lysis. This might be of high relevance for the transformed tissues with negligible type 1 immunity. Accordingly, it was recently shown that IL-23 was overexpressed in human HCCs (41) and melanoma (42), as well as in other tumor tissues, such as colorectal, ovarian and bladder cancer (43). It could be speculated that ILC3s might provide an alternative source of not only IFN-g, but also cytotoxicity, in the tumors with a type 3 microenvironment, and their functions might even increase with higher availability of IL-2. This can also be relevant for early tumor development in tissues enriched with local resident ILC3s, such as gut, where they can provide early elimination of transformed cells, before type 1 immunity emerges.
We showed that ILC3s preferentially used TRAIL and TRAILR2 engagement to kill their targets. In certain tissues, both developing tumors and arriving metastasizing disseminated cells were shown to be TRAIL-sensitive, and TRAIL represents important pathway for their elimination and control. These includes liver cancer (44), melanoma (45) and breast cancer (46). In these tissues, NK cells were considered main anti-tumor cytotoxic effectors. In addition, some tissues, such as liver, are enriched in ILC1s, which also can exert TRAIL-mediated cytotoxicity, although the proportion of TRAILexpressing ILC1s in human liver might be lower compared to mouse (47). Due to their residency, ILC1s and ILC3s might be able to provide protection even before NK cells are recruited. Which cells are preferentially engaged might be dictated by the microenvironment, and might dynamically change in a spatiotemporal manner as the response progresses. Although we also detected Granzyme B expression in a subset of ILC3s, these ILC3s did not express perforin. These data suggest that Granzyme B might have other functions, alternative to inducing target cell apoptosis. It was recently shown that ILC3s from the gut were able to express Granzyme B upon bacterial stimulation (48). Furthermore, Granzyme B has the ability to remodel the extracellular matrix by inducing anoikis and detachment of tumor cells, which prevents tumor cell migration and invasion (49). Therefore, Granzyme B produced by ILC3s might play a perforin-independent anti-metastatic function.
Taken together, our results revealed novel insights into the role of ILC3s in anti-tumor immunity. Our findings demonstrate that ILC3s can display anti-tumor functions, including the release of pro-inflammatory cytokines, and direct lysis of tumor targets via TRAIL-mediated cytotoxicity. Thus, we suggest that ILC3s could act as first-line local responders in tissues against transformed or metastasizing cells, which might be further exploited for therapies against cancer.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Ethik Kommission II of the Medical Faculty Mannheim. The patients/participants provided their written informed consent to participate in this study.    CD34 -, CD123 -, TCR a/b -, TCR g/d -, FCϵR1a -), CD127 -, CD56 + and NKp80 + cells], and ILC3s upon 24h stimulation with 100 ng/ml IL-1b, 100 ng/ml IL-23 and 10 U/ml IL-2. Data for n=2 blood donors are shown. (B) Histograms (left) depict representative staining of T-bet on NK cells and ILC3s treated as in A (FMO, control staining without antibody). Quantification of T-bet expression by ILC3s normalized to isotype control (n=4 blood donors, *** p<0.001 by unpaired Student's t-test).
Supplementary Video 1 | 16h time-lapse video of ILC3 and HeLa cell co-culture. 350-qm-wide micro well contains HeLa-CD48 target cells (large and adherent) stably expressing the NES-ELQTD-GFP-T2A-NES-VGPD-mCherry construct and blood-derived ILC3s (small round shaped). Example of reporter cleavage and cell death upon ILC3-target cell contact is depicted at 0:22 min. The green fluorescence in the nucleus indicates Caspase-8 activity.