Therapeutic application of human type 2 innate lymphoid cells via induction of granzyme B-mediated tumor cell death

Summary

The therapeutic potential for human type 2 innate lymphoid cells (ILC2s) has been underexplored. Although not observed in mouse ILC2s, we found that human ILC2s secrete granzyme B (GZMB) and directly lyse tumor cells by inducing pyroptosis and/or apoptosis, which is governed by a DNAM-1−CD112/CD155 interaction that inactivates the negative regulator FOXO1. Over time, the high surface density expression of CD155 in acute myeloid leukemia cells impairs expression of DNAM-1 and GZMB, thus allowing for immune evasion. We describe a reliable platform capable of up to 2,000-fold expansion of human ILC2s within 4 weeks, whose molecular and cellular ILC2 profiles were validated by single-cell RNA sequencing. In both leukemia and solid tumor models, exogenously administered expanded human ILC2s show significant antitumor effects in vivo. Collectively, we demonstrate previously unreported properties of human ILC2s and identify this innate immune cell subset as a member of the cytolytic immune effector cell family.

In brief

Human type 2 innate lymphoid cells, unlike their murine counterparts, show cytolytic activity in tumor cells through granzyme-mediated induction of multiple cell death pathways. Ex-vivo expansion of ILC2s opens up possibilities for use of these cells in cancer therapy.

Graphical Abstract

Introduction

Cancer immunotherapy is a promising and constantly evolving modality of cancer treatment. Cell-based immunotherapies such as chimeric antigen receptor T cells (CAR-T)¹, CAR-NK cells², and CAR-macrophages³ are revolutionizing the treatment of multiple cancers. However, although adoptive immunotherapy has successfully treated B-cell lymphoma, acute lymphoblastic leukemia^4–6, and multiple myeloma⁷, challenges exist. For example, adoptive immunotherapy can trigger high levels of inflammation and result in severe consequences such as cytokine release syndrome (CRS) that can be lethal¹. Furthermore, its efficacy in treating other hematological malignancies (e.g., acute myeloid leukemia, AML) and solid tumors has been modest to date, highlighting the need for further innovation in cell-based therapy.

Group 2 innate lymphoid cells, also known as ILC2s, are characterized by expression of GATA3 and the production of Th2 cell-associated cytokines, including IL-4, IL-5, IL-9, and IL-13 as well as amphiregulin (AREG), in response to stimulation with the cytokines IL-25, IL-33, and thymic stromal lymphopoietin (TSLP)⁸. In mice, two main subgroups of ILC2s have been identified: the IL-33-induced steady-state natural ILC2 and the IL-25-elicited inflammatory ILC2^9,10. In humans, ILC2s have been characterized as CD127⁺CRTH2⁺c-Kit^+/− cells that express the IL-33 receptor ST2 and the IL-17RB subunit of the IL-25 receptors^11–13. Functionally, ILC2s have been implicated in regulating inflammation and act as a crucial bridge between innate and adaptive type 2 immunity¹⁴. Controversially, ILC2s have been reported to both promote and suppress antitumor immunity in different mouse tumor models^15–18. On one hand, mouse ILC2s can suppress anti-tumor immunity by promoting the expansion and activity of myeloid-derived suppressor cells (MDSCs) via producing type-2 cytokines^19–21. They can also suppress the NK cell-mediated anti-metastatic immune response against B16.F10 melanoma cells in the lungs of mice via recruitment and activation of eosinophils in an IL-5-dependent manner²². By contrast, more recent publications suggest that mouse ILC2s can reduce metastatic dissemination in pancreatic²³, lung^17,24, and colorectal²⁵ cancer models by exerting antitumor effects. While these contrasting data on the role of ILC2s in different tumor types may reflect ILC2 heterogeneity and/or differences in various tumor microenvironments (TME), they underscore the need for a better understanding of this unique cell type. Moreover, the majority of ILC2 research has focused on mice, with limited studies on human ILC2s. These gaps led us to the present work further characterizing human ILC2s in both normal and disease settings.

We report a previously unknown strategy to study ILC2 function and to explore their potential use in adoptive immunotherapy. Our work demonstrates that human, but not mouse, ILC2s represent a member within the family of cytolytic immune effector cells with broad antitumor activity, and may add to the armamentarium of adoptive cellular therapies for cancer.

Results

ILC2s isolated from human peripheral blood can be reliably expanded ex vivo with authenticity confirmed by full-length single-cell RNA-sequencing

ILC2s have been mainly described as tissue-resident cells, but they can also be detected at low levels in human peripheral blood (PB). However, unlike mouse ILC2s, there is still no consistent methodology for purifying and expanding human ILC2s to enable in-depth analysis. To establish a system where human ILC2s could be reliably isolated and expanded with high purity from PB, we explored a variety of lineage depletion techniques to obtain total ILCs (detailed in STAR Methods), followed by the culture conditions depicted in Figure 1A. Regardless of co-culture on either DL1-expressing or DL4-expressing OP9 stromal cells, fluorescence-activated cell sorting (FACS)-sorted CD161⁺CRTH2⁺CD117⁺ ILC2s on day 14 remained over 91% pure on day 28 (Figures 1A and 1B, Data S1, page 1A). Cells harvested on day 28 were referred to as Ex ILC2s. The use of bulk RNA-sequencing (RNA-seq) analysis of CD161⁺CRTH2⁺CD117⁺ Ex ILC2s revealed that these cells expressed previously reported core transcriptional signatures of human ILC2s²⁶ (Figure S1A), including IL1RL1, IL17RB, and KLRG1 alongside several transcripts that encode products involved in signaling via receptors of the prostaglandin response (HPGDS, PPARG, and PTGER2), environmental sensing (SLAMF1, P2RY1, ITGAM, and FASLG), and some transcription factors (e.g., GATA3, RORA, MAF, BCL11B, GFI1, and NFIL3). However, Ex ILC2s did not have elevated expression of certain cytokine-encoding transcripts that were expressed in activated ILC2s or tissue-resident ILC2s (i.e., IL4, IL5, IL9, and AREG)²⁶. GATA3 protein expression was confirmed, while the NK and ILC1 signature genes T-BET and EOMES were not expressed (Figure 1C). These cells also did not produce either IFNγ or TNF (Figure S1B), two cytokines produced by NK cells and ILC1s¹⁴, nor did they express RORγt, a transcription factor used to define ILC3s¹⁴ (Figure 1C). Compared to ILC2s isolated from freshly peripheral blood mononuclear cells (PBMCs) of healthy donors (HD ILC2s), Ex ILC2s produced comparable ILC2-associated cytokines such as IL-4, IL-9, and IL-13, upon phorbol-12-myristate-13-acetate (PMA)/ionomycin stimulation, except for IL-5 (Figures 1D and S1C). However, when compared to HD ILC2s, upon IL-33 stimulation Ex ILC2s exhibited significantly higher levels of IL-9 and IL-13 but produced similar levels of IL-4 and IL-5 (Figures S1D and S1E); Ex ILC2s also expressed a higher level of the IL-33 receptor but not NKp30 (Figures S1F and S1G). Ultimately, our culture conditions resulted in up to a 2,000-fold ILC2 expansion from HD ILC2s by day 28 (Figure 1E). Initially, our Ex ILC2s originated from a mixed culture of total ILCs including ILC precursors. To determine how much of our final population was derived from these precursors, we isolated ILC precursor cells (Lin⁻ CD127⁺CD161⁺CRTH2⁻CD117⁺)²⁷ and cultured them as depicted in Figure 1A. We achieved a modest expansion of up to 8-fold (Figure S1H), suggesting that mature ILC2s, rather than ILC precursors, contributed substantially towards the 2,000-fold expansion from bulk ILCs.

Figure 1. ILC2s isolated from human peripheral blood can be reliably expanded ex vivo with authenticity confirmed by full-length single-cell RNA-sequencing.

(A) Schematic of culture conditions. (B) Representative flow cytometry plots of the percentage of CD161+ cells (top) and CRTH2+CD117+ cells (bottom) after 28-day expansion of total ILCs from PBMC described in A (n = 4). (C) Representative flow cytometry plots and bar graphs of GATA3, EOMES, T-BET, and RORγt expression in expanded ILC2s (Ex ILC2; n = 4). (D) The percentage of IL-4, IL-5, IL-9, and IL-13 produced by HD and Ex ILC2s (n =4). (E) The fold change of harvested Ex ILC2s vs. pre-seeded ILC2s isolated from PBMCs after 28 days (n = 8). (F and G) A UMAP analysis of human ILCs from 4 groups (Ex ILC2s, HD ILCs, Traced ILC2s, and AML ILCs) identified seven distinct clusters. Cells are color-coded according to the defined subsets. (H) Graphics showing the relative representation from each group within each annotated cluster in (G). Subsets are color-coded as in F and G. Data are from two independent experiments. (I and J) Dot plot analysis from scRNA-seq displaying expression of selected and previously described genes encoding specific cell surface markers, cytokines, cytokine receptors, and TFs used to annotate clusters (n = 3 in Ex ILC2s, HD ILCs, and Tranced ILC2s groups; n = 4 in AML ILCs group). (K) Representative flow cytometry plots of expression of GATA3, EOMES, T-BET, and RORγt in Ex and Traced ILC2s. (L) Representative histograms and bar charts showing expression of CRTH2, KIT, IL-33R, IL-17RB, KLRG1, and GATA3 in Ex and Traced ILC2s (n = 5). Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant. See also Figure S1.

To further validate the identity of Ex ILC2s, we performed 10× Genomics high-throughput droplet-based single-cell RNA sequencing (scRNA-seq). For this purpose, we enriched ILCs either from freshly isolated human PBMCs of healthy donors or from patients with AML, named “HD ILCs” and “AML ILCs”, respectively. Additionally, we performed lineage tracing experiments by injecting Ex ILC2s into NSG mice. Seven days later, human CD45⁺ cells were sorted from the bone marrow (BM) of mice and named “Traced ILC2s”. Unsupervised clustering of all sequenced cells of HD ILCs, AML ILCs, Traced ILC2s, and Ex ILC2s revealed seven distinct transcript signature clusters after projecting cells into two dimensions through uniform manifold approximation and projection (UMAP) analysis (Figures 1F, 1G, and S1I). The analysis of the transcriptomic profile of the sorted and enriched cells for core ILC cell signatures^27–29 identified that clusters 0, 1, 2, 4, and 6 were consistent with ILC2s and expressed core ILC2 signature genes (e.g., GATA3, MAF, PTGDR2, and HPGDS); Cluster 3 was ILC progenitors since it expressed both low levels of some ILC2 core signature genes (e.g., GATA3 and PTGDR2) and ILC3 characteristic genes (e.g., RORC and IL23R)²⁷. Cluster 5 was ILC1/NK cells that expressed their core signature genes (e.g., TBX21, IKZF3, EOMES, and KIR2DL3) (Figures S1J-S1M). Of note, with AML ILCs and HD ILCs as references, the UMAP analysis and ILC2 core signature genes revealed that 97.36% of Ex ILC2s were identified as true ILC2 subsets, represented by cluster 0, 1, 2, 4, and 6 (Figures 1G and 1H); additionally, Ex ILC2s all but excluded those known cytotoxic ILCs, as only 1 out of 4798 cells in Ex ILC2s was grouped into the ILC1/NK cells cluster (Cluster 5) (Figure 1G).

Lineage tracing experiments revealed that Ex ILC2s maintained their scRNA-seq phenotypes after being transferred into mice. This was demonstrated by (1) UMAP analysis, which revealed similar cluster distributions of both CD45⁺ ILC2 types (Ex ILC2s and Traced ILC2s) (Figure 1G), and (2) Traced ILC2s maintained ILC2 core signature genes (e.g., PTGDR2 encoding CRTH2, KIT encoding c-KIT, IL1RL1 encoding IL-33R, IL17RB encoding IL-25R, KLRG1, and GATA3) but did not express ILC1/NK core signature genes (e.g., TBX21 and EOMES) or ILC progenitor/ILC3 core signature genes (e.g., RORC encoding RORγt) (Figures 1I and 1J). Flow cytometry further validated the protein expression of CRTH2, KIT, IL-33R, IL-17RB, KLRG1, and GATA3, but not T-BET, EOMES, and RORγt, in Traced ILC2s (Figures 1K and 1L). Compared to Ex ILC2s, Traced ILC2s exhibited decreased expression of KIT, IL-33R, KLRG1, and GATA3 and increased expression of IL-17RB, while CRTH2 expression remained (Figure 1L). These data support that Ex ILC2s have not been converted to other cell types.

ScRNA-seq and flow cytometric analysis also showed very low expression levels of CD5 in both CRTH2⁺ HD and Ex ILC2s (Figures S1N and S1O). A high-level expression of CD5 was previously defined as a functionally immature marker of ILCs³⁰. Thus, Ex ILC2s remain in a mature state. HD ILC2s mainly expressed the resting marker CD45RA but less the activation marker CD45RO of ILC2s³¹, consistent with previous reports^29,31; interestingly, Ex ILC2s maintained expression of both CD45RA and CD45RO (Figure S1P), suggesting that Ex ILC2s may be in an intermediate state of activation compared to resting and activated ILC2s.

We observed that Ex ILC2s acquired expression of the IL-15 receptor CD122, and so we included IL-15 in our culture conditions during ex vivo expansion (Data S1, page 2A). Compared with those cells without IL-15, hallmark pathway analysis of ILC2s expanded with IL-15 revealed up-regulated gene-enriched pathways associated with cell division, hypoxia, glycolysis, mammalian target of rapamycin (mTOR), and c-Myc signaling (Data S1, pages 2B and 2C), suggesting a role for IL-15 in ILC2 proliferation and aerobic glycolysis. IL-15-mediated improvement of aerobic glycolysis was reported to contribute to lymphocyte cell proliferation³². Consistently, the expansion capacity of ILC2s was increased in the presence of IL-15 compared to expansion without IL-15 (Data S1, page 2D). RNA-seq results also showed that expression of GZMB was higher in ILC2 cultured with IL-15 than those without IL-15, while ILC2s maintained expression of their core cytokines and surface markers (Data S1, page 2E). The increase in GZMB expression was confirmed at the protein level (Data S1, page 2F). The cytotoxicity of ILC2s against tumor cells was also elevated in the presence of IL-15 compared to its absence (Data S1, pages 2G and 2H).

ILC2s induce AML cell death in vitro and delay tumor growth in vivo

Mouse ILC2s have been shown to exert both pro- and anti-tumor effects⁸, and their precise role in human tumor immunity remains poorly defined. To further clarify the function of human ILC2s within the context of cancer immunity, we explored their interaction with AML, a disease with an overall poor prognosis³³. Among eight AML patients whose blood was sampled at disease onset and compared to age- and sex-matched HDs, we noted a highly significant reduction in total PB ILC2s among both lineage-negative (Lin⁻) cells and ILC subsets (defined as Lin⁻CD127⁺) (Figures S2A and S2B, Data S1, pages 1B and 1C). The use of the Cancer Genome Atlas (TCGA) database identified a significant correlation between the signatures of ILC2s and leukemia cells (leukemia stem cells and blasts)^26,34 (Figure S2C). Ex ILC2s lysed AML cell lines at very low effector to target (E:T) ratios of 2:1, 1:1, or 0.5:1 in a dose-dependent manner, as evidenced by both a visually appreciable size reduction in the pellet of AML cells under microscopy (Figure 2A), and an increase in the fraction of dead AML cells using both luminescence (Figure 2B) and flow cytometry-based assays (Figures 2C and S2D). Likewise, Ex ILC2s were able to lyse primary AML blasts isolated from the PB of patients in a dose-dependent manner (Figure S2E). However, both Ex ILC2s and HD ILC2s demonstrated greater potency in lysing MOLM13 and U937 compared to THP1 (Figures 2C, S2D, and S2F), suggesting that ILC2s exhibit variable responses among distinct subtypes of AML cells. The use of a single-cell cytotoxicity assay further confirmed the intrinsic capacity of human ILC2s to lyse tumor targets (Figures S2G-S2I).

Figure 2. ILC2s induce AML cell death in vitro and prevent tumor growth in vivo.

(A) Representative images of AML cell lysis by ILC2s (5 × magnification, scale bar, 200 μm, n = 4). (B) Luciferase activity in the wells with tumor cells expressing a luciferase gene and Ex ILC2s was measured (n = 7). (C) Summary data of cell percentages of dead cells of indicated three cell lines, identified by Annexin V and DAPI, in the presence of Ex ILC2s (n = 7). (D) Design and procedures for E and F. (E) Images represent bioluminescence (BLI) at the indicated time points (n = 4 in MOLM13 alone group; n = 5 in MOLM13 + ILC2 group). (F) Survival of mice injected with or without Ex ILC2s (n = 4 or 5). (G) Design and procedures for H and I. (H) Images represent BLI at the indicated time points (n = 6). (I) Survival of mice injected with or without Ex ILC2s (n = 6). (J) Design and procedures for K and L. (K) Images represent BLI at the indicated time points (n = 6). (L) Survival of mice injected with or without Ex ILC2s (n = 6). Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. See also Figure S2 and Table S1.

A single infusion of Ex ILC2s improved in vivo MOLM13, U937, and THP1 tumor control and significantly prolonged mouse survival compared to the control group (Figures 2D–2I, and S2J). The adoptively transferred Ex ILC2s mainly existed in mouse liver at the endpoint of the experiment (Figures S2K and S2L). To simulate the physiologic environment of the human body more closely, we administered an infusion of human PBMCs into NSG-SGM3 mice before the transplantation of MOLM13 cells and Ex ILC2s. This also led to suppression of tumor growth and significantly prolonged mouse survival (Figures 2J–2L). The levels of some pro-inflammatory cytokines such as ICAM1, IL-1β, IL-6R, IL-16, TIMP-1, MIP-1α, TNF RI, and TNF RII were also increased (Table S1). Despite these changes, histopathologic analyses after Ex ILC2 infusion did not reveal any evidence of graft-versus-host disease (GVHD), tissue destruction, or toxicity (Figure S2M). The tissue distribution of neutrophils, mast cells, eosinophils, Th2 cells, and dendritic cells showed no significant difference between the mice treated with and without Ex ILC2s (Figures S2N-S2U, Data S1, pages 1D-1F). Collectively, Ex ILC2s possess the potential to delay the progression of AML with relatively low toxicity.

ILC2-secreted granzyme B induces pyroptosis or apoptosis in AML cells

Ex ILC2 induced-cell death of the MOLM13, U937, and THP1 was not suppressed in the presence of a neutralizing antibody against the ILC2 cytokine IL-4, IL-5, IL-9, or IL-13, the combination of anti-IL-4 and anti-IL-13 antibodies, or anti-IL-4Rα antibody, excluding the involvement of these cytokines or their signaling (Figures S3A and S3B). Ex ILC2s also did not convert into cytotoxic NK cells (CD3⁻CD56⁺)/ILC1s (Lin⁻CD161⁺CRTH2⁻CD117⁻) or cytotoxic ILC3s (Lin⁻CD56⁺CD161⁺CRTH2⁻CD117⁺) under co-culture conditions with MOLM13, U937, or THP1 (Figure S3C), despite this conversion being previously reported in response to tissue inflammation³⁵. If a conversion of ILC2s to ILC3s were occurring, we would anticipate the disappearance of CRTH2; however, this was not the case for the co-culture of Ex ILC2s with MOLM13 or U937. Although we did observe downregulation of CRTH2 in the co-culture with THP1, these Ex ILC2s still exhibited high expression of GATA3, rather than RORγt (Figure S3D), suggesting that their ILC2 identity was maintained. Our results were consistent with previous reports, which demonstrated that IL-33³⁶ and prostaglandin D2 (PGD2; a ligand of CRTH2)³⁷, both expressed by THP1 but not MOLM13 and U937 (Figure S3E), suppress expression of CRTH2 on ILC2s. Collectively, these results suggest that the induction of AML cell death by ILC2s is not attributable to the secretion of IL-4, IL-5, IL-9, or IL-13, nor a conversion into cytotoxic NK cells/ILC1s, or ILC3s.

In the presence of ILC2s, MOLM13 and THP1, but not U937, showed progressive morphological changes indicative of cell death, including cellular swelling and bubble-like protrusions appearing on their membrane surfaces before their subsequent rupture (Figure 3A; Videos S1A-S1C). Activation of caspase 3 in these cell lines was also increased (Figures S3F and S3G). MOLM13 and THP1, but not U937, expressed gasdermin E (GSDME) (Figure S3H). Coculturing with Ex ILC2s resulted in cleavage of GSDME in MOLM13 and THP1 (Figures 3B and 3C). Caspase 3 cleavage was observed in all three AML cell lines in the presence but not absence of Ex ILC2s (Figures 3B–3D).

Figure 3. ILC2-secreted granzyme B induces pyroptosis or apoptosis in AML cells.

(A) Time-lapse microscopy images of co-cultures of Far red-labeled Ex ILC2s (magenta) with CFSE-labeled AML cells (green) in a medium containing DAPI (blue). Scale bar, 100 μm. (B) Immunoblotting showing cleavage of full-length (F)-GSDME into N terminal (N)-GSDME and caspase 3 into cleaved caspase 3 in MOLM13 incubated with Ex ILC2s. (C and D) Immunoblotting showing GSDME and caspase 3 cleavage in THP1 and U937 incubated with or without Ex ILC2s. (E) Representative flow cytometry plots and bar graphs of the percentage of GZMB and perforin in HD and Ex ILC2s (n = 4). (F) Confocal microscopy images of GZMB and perforin in Ex ILC2s (scale bar, 10 μm, n = 3). (G and H) Representative flow cytometry plots and statistics of the percentage of GZMB in Ex ILC2s co-cultured with or without MOLM13, U937, or THP1 (n = 4). (I) ELISA determined the production of GZMB by Ex ILC2s co-cultured with or without AML cells (n = 4). (J) Left: ELISA determined the production of GZMB by wild-type (WT) Ex ILC2s or GZMB knockdown ILC2s (GZMB KD Ex ILC2) co-cultured with or without AML cells. Right: Luciferase activity in the wells with tumor cells was measured (n = 3). (K and L) Immunoblotting showing GSDME and caspase 3 cleavages in MOLM13 and U937, incubated with or without WT Ex ILC2s or GZMB KD Ex ILC2s. Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant. See also Figure S3 and Videos S1A-S1C.

The above features are characteristic of pyroptosis that was previously reported to be mediated by GZMB³⁸. Therefore, we hypothesized that ILC2s share phenotypic and functional traits with cytotoxic lymphocytes (e.g., NK cells) capable of producing GZMB and perforin to induce pyroptosis in tumor cells³⁹. Indeed, both HD and Ex ILC2s produced GZMB and perforin upon stimulation with PMA/ionomycin (Figures 3E, 3F, and S3I). Supporting this, RNA-seq analysis of Ex ILC2s co-cultured with or without MOLM13 or U937 revealed that AML cells substantially upregulated GZMB expression in Ex ILC2s (Figure S3J). Flow cytometry and enzyme-linked immunosorbent assay (ELISA) validated that the production of GZMB but not perforin was significantly induced in Ex ILC2s when co-cultured with AML cells (Figures 3G–3I and S3K-S3M). We also investigated the activation state of cytotoxic ILC2s (GZMB-producing ILC2s) in both HD and Ex ILC2s. We assessed expression of CD45RA and CD45RO on HD and Ex GZMB⁺ ILC2s. While earlier we noted that ~70% of all HD ILC2s were in a resting state of CD45RA⁺CD45RO⁻ (Figure S1P), we found that over 92% of GZMB⁺ HD ILC2s were in a resting state of CD45RA⁺CD45RO⁻ (Figure S3N). However, like all Ex ILC2s (Figure S1P), almost all GZMB⁺ Ex ILC2s were in an intermediate activation state expressing both CD45RA and CD45RO (Figure S3N).

Knock down (KD) of GZMB by CRISPR-Cas9 in Ex ILC2s decreased these cells’ ability to lyse target AML (Figure 3J), suggesting that the mechanism of AML cell death mediated by Ex ILC2s is dependent on the release of cytotoxic granules. When we substituted wildtype (WT) Ex ILC2s with GZMB KD Ex ILC2s in co-cultures with MOLM13, we observed the absence of GSDME cleavage in MOLM13 (Figure 3K). Caspase 3 cleavage was not detected in both MOLM13 and U937 when co-culturing them with GZMB KD Ex ILC2s (Figures 3K and 3L, respectively). Altogether, these results suggest two key points: 1) GZMB produced by Ex ILC2s can induce pyroptosis in AML cells that express GSDME, such as in MOLM13, by specifically cleaving GSDME; and 2) GZMB produced by Ex ILC2s triggers caspase 3 cleavage in both GSDME expressing (e.g., MOLM13) and non-GSDME expressing (e.g., U937) AML cells.

The cleavage of caspase 3 in non-GSDME cells is a hallmark of apoptosis. This prompted us to determine whether non-pyroptosis forms of programmed cell death (e.g., apoptosis, necroptosis, or ferroptosis) were also involved in ILC2-mediated AML cell death. We then pretreated WT MOLM13 with various inhibitors before adding Ex ILC2s. Consistent with the apoptotic form of cell death, zVAD-fmk, a pan-caspase inhibitor, or zDEVD-fmk, a caspase-3 inhibitor, partially reduced Ex ILC2-induced MOLM13 cell death (Figure S3O), suggesting that caspase-dependent apoptosis may also be involved in cell death occurring in GSDME-expressing AML cells. Neither the necroptosis inhibitor necrostatin-1 (Nec-1) nor the ferroptosis inhibitor ferrostatin-1 (Fer-1) suppressed Ex ILC2-mediated MOLM13 cell death (Figure S3P), indicating that necroptotic and ferroptotic forms of cell death were not involved. We next used CRISPR-Cas9 to knock down GSDME in MOLM13 cells (named GSDME KD MOLM13). Knockdown of GSDME did not alter expression of caspase 3 in MOLM13 (Figure S3Q). Co-cultures of GSDME KD MOLM13 with Ex ILC2s did not show characteristics of pyroptosis such as cellular swelling and bubble-like protrusions on the surface of cellular membranes (Video S2). Knockdown of GSDME in MOLM13 did not block Ex ILC2-induced MOLM13 cell death compared with controls (Figure S3R). When zDEVD-fmk or zVAD-fmk was added to Ex ILC2s co-cultured with GSDME KD MOLM13, Ex ILC2-induced cell death was suppressed (Figure S3R). These findings demonstrate the transition from pyroptosis to caspase-dependent apoptosis and the predominance of caspase-dependent apoptosis in GSDME-deficient AML cells.

Collectively, these data suggest 3 conclusions: (1) Ex ILC2s can induce pyroptosis of GSDME⁺ AML cells (e.g., MOLM13) through the cleavage of GSDME mediated by GZMB; (2) Ex ILC2-mediated cleavage of caspase 3 occurs irrespective of GSDME expression, leading to apoptosis in AML cells such as U937; and (3) during Ex ILC2-mediated cell death of GSDME⁺ AML, both pyroptosis and apoptosis are involved. However, in GSDME⁻ AML cells, caspase-dependent apoptosis emerges as the primary mechanism of cell death.

ILC2s require cell-cell contact with AML cells to induce GZMB production through DNAM-1 interaction with its ligands, CD112 and CD155

Using a transwell system to separate Ex ILC2s and AML cells, we did not observe AML cell death (Figures S4A and S4B). GZMB production by Ex ILC2s was significantly reduced in transwell assay compared with control co-cultures (Figures 4A–4C); the cleavage of GSDME in MOLM13 and the cleavage of caspase 3 in both MOLM13 and U937 disappeared (Figures S4C and S4D). Thus, direct cell-to-cell contact is essential for triggering GZMB production by ILC2s, which in turn induces pyroptosis and/or apoptosis in AML cells. Next, we investigated the specific receptors and ligands involved in the recognition of target cells. Like NK cells, Ex ILC2s also expressed certain activating receptors, including natural killer group 2 member D (NKG2D), DNAM-1 (CD226), and NKp30 (Figure 4D). Among these, DNAM-1 exhibited the highest level of expression. Although the ligands for NKG2D (MICA, MICB, and ULBP1/2/5/6) were expressed at low levels in MOLM13, U937 exhibited high expression of ULBP1/2/5/6, and THP1 showed high expression of MICA and MICB (Figure S4E). The ligands for DNAM-1 (CD155 and CD112) and the ligand for NKp30 (B7H6) were both highly expressed in all three AML cell lines (Figure S4F). Blocking antibodies against NKG2D or DNAM-1 significantly reduced GZMB production and cytotoxicity of ILC2s against MOLM13, U937, or THP1 cells without significantly altering perforin production; in contrast, blocking NKp30 had no significant impact (Figures 4E–4G and S4G-S4I). Finally, the decrease in ILC2-mediated lysis of AML appeared to be more pronounced in the presence of an antibody blocking DNAM-1 compared to one blocking NKG2D (Figures S4H and S4I).

Figure 4. ILC2s require cell-cell contact with AML cells to induce GZMB production through DNAM-1 interaction with its ligands, CD112 and CD155.

(A and B) Representative flow cytometry plots (A) and Statistics (B) of the percentage of GZMB+ Ex ILC2s co-cultured with or without MOLM13, U937, or THP1 in the presence or absence of transwell (n = 8). (C) Supernatants were collected from (A) and subjected to ELISA to determine levels of GZMB (n = 8). (D) Representative histograms showing expression of NKG2D, DNAM-1, and NKp30 on Ex ILC2s (n = 4). (E and F) Representative flow cytometry plots (E) and Statistics (F) of the percentage of the percentage of GZMB+ Ex ILC2s co-cultured with or without MOLM13, U937, or THP1 in the presence or absence of an NKG2D, DNAM-1, or NKp30 blocking antibody (n = 8). (G) Supernatants were collected from (E) and subjected to ELISA to determine levels of GZMB (n = 8). (H and I) Ex ILC2s were co-cultured with or without WT, SKO-CD112, SKO-CD155, or DKO-CD112/CD155 MOLM13, U937, or THP1, separately. Representative flow cytometry plots of the percentage (H) and statistics (I) of GZMB+ Ex ILC2s (n = 8). (J) Supernatants were collected from (H) and subjected to ELISA to determine levels of GZMB (n = 8). Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant. See also Figure S4 and Video S2.

Compared with WT ILC2s, DNAM-1 knockout (DNAM-1 KO) ILC2s showed a significant decrease in their ability to lyse AML cells and to produce GZMB (Figures S4J-S4L). Like co-culture with AML cell lines, GZMB production was also elevated when Ex ILC2s were co-cultured with primary AML blasts, and this elevation was not observed in the presence of a DNAM-1 blocking antibody (Figures S4M and S4N).

Using the CRISPR-Cas9 system, we performed single-knockout of CD112 (SKO-CD112), CD155 (SKO-CD155), and double knockout of both CD112 and CD155 (DKO-CD112/CD155) in MOLM13, U937, and THP1 cells (Figure S5A). Compared with co-culturing with WT AML cells, Ex ILC2s’ production of GZMB did not show consistent, significant changes when co-cultured with either SKO-CD112 or SKO-CD155 AML cells. However, there was a significant and consistent reduction in GZMB production when EX ILC2s were co-cultured with each of the three DKO-CD112/CD155 AML cell lines (Figures 4H–4J). A similar reduction in Ex ILC2 cytotoxicity was also observed (Figures S5B-S5D).

We further examined the potential role of NKG2D ligands MICA/MICB in GZMB production. This was particularly relevant in the context of experiments co-culturing Ex ILC2s with THP1, which exhibited high levels of MICA/MICB (Figure S4E). Blockade of MICA/MICB on THP1 significantly inhibited the production of GZMB by Ex ILC2s (Figure S5E). This inhibitory effect was comparable to that achieved by blocking NKG2D alone, or in combination with blocking MICA/MICB. Notably, the suppression of GZMB production in Ex ILC2s was more pronounced when both NKG2D and DNAM-1 signaling pathways were blocked, as opposed to blocking either one individually (Figure S5E). Our data indicate that the MICA/MICB-NKG2D pathway also plays a role in the production of GZMB by ILC2s when encountering AML cells with high surface densities of NKG2D ligands.

Together, these data suggest that: (1) the production of GZMB by ILC2s is mediated through interactions between DNAM-1⁺ ILC2s and CD112 and CD155 expressed on AML cells, as well as through the engagement of NKG2D on ILC2s with its ligands when expressed on AML cells; and (2) expression of DNAM-1 and NKG2D ligands on AML cells has a role in inducing the subsequent lysis of these AML cells by ILC2s.

DNAM-1-mediated inactivation of FOXO1 is required to produce GZMB by ILC2s

Next, we performed gene set enrichment analysis (GSEA) using our RNA-seq data from Ex ILC2s co-cultured with or without AML cells, and either in the presence or absence of a DNAM-1 blocking antibody (part of data presented in Figure S3J). Results revealed that Ex ILC2s co-cultured with U937 exhibited changes in signaling pathways, with the top 10 pathways displayed in Figure 5A. Four of the top 10 activated pathways were related to either protein kinase B (AKT) or forkhead box protein O1 (FOXO1) signaling. Gene sets associated with FOXO1 were downregulated in Ex ILC2s co-cultured with U937 when a DNAM-1 blocking antibody was present, compared with its absence (Figures 5B and S5F). This suggests that downstream signaling from DNAM-1 is linked to the modulation of AKT or FOXO1 pathways.

Figure 5. DNAM-1-mediated inactivation of FOXO1 is required to produce GZMB in ILC2s.

(A and B) Hallmark pathway analysis in Ex ILC2 RNA pools (A: ILC2s co-cultured U937 vs. ILC2s alone; B: ILC2s co-cultured with U937 in the presence of anti-DNAM-1 vs. ILC2s co-cultured with U937 in the absence of anti-DNAM-1; n = 3). (C and D) Immunoblotting measured the phosphorylation of FOXO1 and AKT in ILC2s (C: ILC2s co-cultured with or without WT U937 in the absence or presence of anti-DNAM-1 (The samples were run on the same gradient gel, showing an apparent similarity of distinct FOXO and p-AKT bands.); D: ILC2s co-cultured with or without WT, SKO-CD112, SKO-CD155, or DKO-CD112/CD155 U937. The results were from the same membrane). (E and F) Representative flow cytometry plots (E) and statistics (F) of the percentage of GZMB production by ILC2s when co-cultured with or without AML cells in the absence or presence of FOXO1 or AKT inhibitor (n = 6). (G) Supernatants were collected from (E and F) and subjected to ELISA to determine levels of GZMB (n = 6). Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. See also Figure S5.

The engagement of DNAM-1 with its ligands in NK cells results in the inactivation of its negative regulatory control over mouse NK cell effector function through the phosphorylation of FOXO1⁴⁰. We hypothesized that DNAM-1-mediated cytotoxicity in Ex ILC2s was also dependent on signaling through FOXO1. Indeed, incubation of Ex ILC2s with WT U937 increased the phosphorylation of FOXO1 (p-FOXO1) in Ex ILC2s. This increase was suppressed when a DNAM-1 blocking antibody was present during co-culture with WT U937 (Figure 5C). Further, compared with co-culture with WT U937, the increase of p-FOXO1 was reduced in Ex ILC2s co-cultured with SKO-CD112 and SKO-CD155 U937, with the greatest reduction seen in Ex ILC2s co-cultured with DKO-CD112/155 U937 (Figure 5D). Similar results were found when examining the phosphorylation of AKT (Figures 5C and 5D), suggesting that DNAM-1 also triggers AKT phosphorylation in Ex ILC2s and that FOXO1 could be a direct substrate of phosphorylated AKT, consistent with previous reports^40,41. Our results suggest that: (1) FOXO1 phosphorylation may be involved in inactivating the negative regulatory control over human ILC2 effector function; and (2) interaction between CD112 and/or CD155 and DNAM-1 leads to FOXO1 phosphorylation in human ILC2s. The lysis of AML cells (Figures S5G-S5I) and the production of GZMB (Figures 5E–5G) by Ex ILC2s were increased in the presence of a FOXO1 inhibitor (AS1842856), which can reduce the phosphorylation and expression of FOXO1⁴², and reduced in the presence of an AKT inhibitor (afuresertib), validating that DNAM-1-mediated inactivation of FOXO1 is required for the enhancement of ILC2 effector activity. Collectively, these data demonstrate that DNAM-1 engagement in Ex ILC2s triggers downstream inactivation of FOXO1, which then enhances the function of Ex ILC2s.

ILC2s exhibit antitumor efficacy at levels similar to those of NK cells in vitro and in vivo

Like NK cells, ILC2s express inhibitory and activating receptors, and the loss of their balance can change the effector function of these immune cells^17,23. We compared expression of inhibitory receptors in Ex ILC2s to expanded human NK cells (Ex NK cells) in the absence or presence of tumor cells in vitro. Ex ILC2s and Ex NK cells have comparable PD-1 expression, but the former have lower TIM3 expression. Although Ex NK cells highly express TIGIT, Ex ILC2s have negligible expression regardless of co-culture with or without tumor cells (Data S1; page 3A). The analysis of activating receptors demonstrated that HD ILC2s and HD NK cells have similar expressions of NKG2D and DNAM-1; however, Ex ILC2s had a much lower level of NKG2D expression compared to Ex NK cells yet both exhibited a comparable level of DNAM-1 expression (Data S1; pages 3B and 3C). NK cells maintained a higher NKp30 expression compared to ILC2s regardless of whether the NK cells and the ILC2s were freshly isolated or expanded (Data S1; pages 3B and 3C). Ex ILC2s and Ex NK cells exhibited a similar level of cytotoxicity against MOLM13 and U937 in vitro (Data S1; page 3D). Finally, in the in vivo MOLM13 AML model, infusion of either Ex ILC2s or Ex NK cells similarly improved both tumor control and survival compared to the control group without Ex ILC2 or Ex NK cell treatment (Data S1; pages 3E and 3F). Together, these results indicate that human ILC2s exhibit antitumor efficacy in vitro and in vivo at levels comparable to those of NK cells.

Immune evasion of AML by inhibiting DNAM-1 surface expression on ILC2s

We observed that DNAM-1 expression on ILC2s in the PB of AML patients (AML ILC2s) was profoundly reduced compared with HD ILC2s (Figure 6A). Likewise, the production of GZMB by AML ILC2s was significantly decreased relative to HD ILC2s (Figure 6B). Consistent with this, co-culturing Ex ILC2s with primary AML blasts resulted in a gradual reduction in DNAM-1 expression over time. However, GZMB production spiked on day 1 of co-culture, then followed a decreasing trend by day 2 (Figure 6C). This pattern suggests that a reduction of DNAM-1 and GZMB in AML ILC2s may be a result of inhibitory signals produced by AML blasts in vivo. Further, transwell separation abolished the downregulation of DNAM-1 expression in Ex ILC2s by AML blasts (Figure 6D), suggesting the reduction of DNAM-1 on AML ILC2s likely results from the interaction between ILC2s and AML blasts. Indeed, our analysis of DNAM-1 ligands showed high expression of CD112 and CD155 on AML blasts (Figure 6E). It was reported that CD155 suppressed DNAM-1 expression on both NK cells and T cells^43,44. We thus hypothesized that CD155 expressed on AML blasts resulted in decreased expression of DNAM-1 on ILC2s. Consistent with this, compared with ILC2 alone, DNAM-1 expression did not diminish when Ex ILC2s were co-cultured with either SKO-CD155 or DKO-CD112/CD155 AML cells. However, a reduction in DNAM-1 expression was noted when Ex ILC2s were co-cultured with SKO-CD112 AML cells (Figure 6F). These results suggest that CD155, rather than CD112, is instrumental in suppressing DNAM-1 expression on ILC2s in this setting.

Figure 6. Immune evasion of AML by inhibiting DNAM-1 surface expression on ILC2s.

(A and B) Representative histograms and statistics of DNAM-1 (A) and GZMB (B) expression in the indicated ILC2s (n = 4). (C) Ex ILC2s were co-cultured with or without primary AML blasts for 5 days. Representative histograms and statistics of DNAM-1 and GZMB expression on the Ex ILC2s (n = 4). (D) Representative histograms and statistics of DNAM-1 expression on the ILC2s co-cultured with or without WT AML blasts in the presence or absence of transwell (n = 4) for 3 days. (E) Representative histograms showing CD112 and CD155 expressions on AML blasts (n = 2). (F) Representative histograms and statistics of DNAM-1 expression on the ILC2s co-cultured with or without WT, SKO-CD112, SKO-CD155, or DKO-CD112/CD155 AML blasts for 3 days (n = 4). (G) U937 were treated with trametinib (100 nM) for 3 days. Histograms showing the MFI of CD155 and CD112. (H) ILC2s were co-cultured with or without U937 pre-treated with or without trametinib for 3 days. Histograms and graphics showing the MFI of DNAM-1 on ILC2s (n = 8). (I) NSG mice transplanted with U937 were treated with or without trametinib (0.8 mg/kg) for 7 days. Images represent BLI on day 7 (n = 5). (J)Graphics showing the MFI of CD155 in the U937 from the blood of NSG mice on day 7 (n = 5). (K) Histograms and graphics showing the MFI of DNAM-1 on ILC2s from the BM of NSG mice transplanted with or without U937; the mice were treated with or without trametinib. (L) Survival analyses based on ILC2 signatures in TCGA-AML individual cohort (n = 106). Log-rank Mantel-Cox test was used. Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant.

The mitogen-activated protein kinase kinase (MEK) inhibitor trametinib can suppress FLT3⁺ AML cell growth^45,46. It has also been observed that trametinib downregulates CD155 expression on several FLT3⁺ AML cell lines (e.g., MV4–11)⁴⁶. Extending this, we also found that trametinib had similar results within the FLT3⁻ U937 cell line, i.e., trametinib treatment reduced expression of CD155 but not CD112 (Figure 6G). By this, when co-cultured with U937 pre-treated by trametinib, Ex ILC2s exhibited a higher expression of DNAM-1 in vitro compared to those in the same culture system without trametinib (Figure 6H). Next, we then used trametinib to treat NSG mice transplanted with U937. FLT3⁻ U937 was chosen to exclude the interference from FLT3 targeting and/or tumor burden. Our imaging data show that treatment without or with trametinib yielded the same amount of tumor growth on day 7 (Figure 6I). We did observe that expression of CD155 was decreased on U937 isolated from the blood of NSG mice treated with trametinib compared to controls (Figure 6J). Next, the NSG mice were injected with Ex ILC2s. Seven days later, we observed that there was no loss of DNAM-1 on Ex ILC2s isolated from the BM of mice pre-treated with trametinib, as compared to loss of DNAM-1 on Ex ILC2s isolated from the BM of control animals (Figure 6K). Considering the relevance of DNAM-1 on ILC2 recognition and lysis of leukemic cells, the reduced expression of DNAM-1 on AML ILC2s appears to represent a mechanism of tumor escape or immune evasion. Analysis of 106 AML cases from TCGA showed that AML patients with high ILC2 gene signatures²⁶ had a significantly prolonged overall survival compared to AML patients with low ILC2 gene signatures (Figure 6L). Altogether, these data suggest that the functional roles of human ILC2s become dysregulated in the context of AML, and a high ILC2 gene signature in a retrospective analysis correlates with a more favorable clinical outcome in AML.

Mouse ILC2s do not have cytotoxicity against AML in vitro or in vivo

We analyzed scRNA-seq data from unbiased databases of mouse ILC2s isolated from BM, liver, and lung (BM: GSM3303970; liver: GSM5697024; lung: GSM5697025). These ILC2 populations were characterized by high levels of Gata3, Il1rl1, Il-25rα, and Klrg1; however, they did not express Gzmb (Data S1; Page 4A). Flow cytometry validated the absence of GzmB expression in ILC2s from both WT and spontaneous AML mice (Mll^PTD-Flt3^ITD) (Data S1; Page 4B). We found mouse ILC2s cannot be expanded under conditions similar to those used for human ILC2s. Thus, we sorted and expanded mouse ILC2s from BM, spleen, liver, and lung of WT mice using previously published methods^23,47 (Data S1; Page 4C). The expanded cells were verified as ILC2s based on their expression of surface markers (CD25 and IL-33R), the canonical transcription factor GATA3, and the production of canonical signature cytokines (IL-4, IL-5, IL-9, and IL-13) (Data S1; Pages 4D-4F). The expanded mouse ILC2s produced only small quantities of GzmB, with all stimulation conditions resulting in less than 10% GzmB⁺ ILC2s on average (Data S1; Pages 5A and 5B). This production level was dramatically lower compared to that seen in human HD and Ex ILC2s, where the majority of ILC2s were GZMB⁺ upon stimulation with PMA/ionomycin (Figure 3E). Following stimulation with tumor cells, the proportion of GzmB⁺ cells in expanded mouse ILC2s (~3%) was also substantially lower compared to human Ex ILC2s (~11–30%) stimulated under similar conditions (Data S1; Page 5C and Figure 3H, respectively). Consistently, the expanded mouse ILC2s had only modest in vitro cytotoxicity against AML targets compared to their human counterpart Ex ILC2s (~5% cell death vs. ~40–70% cell death, respectively) (Data S1; Pages 5D and 5E vs. Figures 2C and S4K). Our in vivo model also showed no significant change in survival of mice bearing AML between injected with and without expanded mouse ILC2s (Data S1; Pages 5F and 5G). There were no differences in tissue distribution of neutrophils, mast cells, eosinophils, Th2, and DCs within the BM, spleen, liver, or lung of the mice injected with expanded mouse ILC2s compared to those without (Data S1; Pages 5H-5M). These findings are consistent with the data from human Ex ILC2s (Figures S2N-2U). Our data suggest that expanded mouse ILC2s produce very low, almost negligible levels of GzmB and lack direct cytotoxicity against tumor cells.

Human ILC2s show antitumor activity against solid tumors

Finally, we investigated the role of human ILC2s in several models of solid tumors. We co-cultured three different solid tumor cell lines with Ex ILC2s, each line representing either pancreatic cancer, brain cancer, or lung cancer. These tumor lines were selected because ILC2s have been detected in the corresponding organs of healthy donors and/or patients with cancer^23,48–50, but there are limited data regarding the function of ILC2s in these cancer patients. We observed that Ex ILC2s lysed solid tumor cells, as measured by Annexin V/DAPI staining and real-time cell analysis (RTCA) assay (Figures 7A, 7B, and S6A-S6C). These solid tumor cells expressed various levels of GSDME (Figure S6D). Among the cell lines that we tested, the Capan-1 pancreatic cancer cell line, the GBM30 glioblastoma cell line, and the A549 non-small-cell lung cancer cell line showed a morphology consistent with pyroptosis in co-culture with Ex ILC2s (Figure S6E). The use of time-lapse microscopy further confirmed that these three cell lines underwent increasing pyroptotic membrane ballooning (Figure S6F; Videos S3A-S3C), suggesting that the pyroptosis of solid tumor cells was induced by ILC2s. All in vivo animal models of Capan-1, GBM30, and A549 also showed that infusion of Ex ILC2s significantly suppressed tumor growth in these highly aggressive mouse models compared with controls (Figures 7C, 7D, and S7A). Ex ILC2s also prolonged the survival of tumor-bearing mice in the Capan-1 and GBM30 models (Figures 7E and 7F). However, survival data were not available for the A549 model at the time the experiment was terminated. To mimic physiologic conditions, as we did for liquid tumors, we infused PBMCs into NSG-SGM3 mice before implantation of Capan-1 and Ex ILC2s. In this model, we validated that infusion of Ex ILC2s also inhibited tumor growth (Figure S7B). We again did not observe significant differences in the tissue distribution of neutrophils, mast cells, eosinophils, Th2 cells, and dendritic cells. The numbers of monocyte-derived MDSC (M-MDSCs) and granulocytic-derived MDSCs (G-MDSCs) were comparable between the mice injected with Capan-1 and Ex ILC2s for ten days and those injected with Capan-1 alone (Figures S7C-S7J). Taken together, our data suggest that as shown for the liquid tumor AML, human ILC2s also have therapeutic potential against solid tumors.

Figure 7. Human ILC2s protect against solid tumor progression.

(A) Ex ILC2s were cultured at the indicated ratios with various solid tumor cell lines. Statistics of the percentages of dead cells identified by Annexin V and DAPI in the cells (n = 4). (B) Ex ILC2s were cultured with various tumor cell lines. Real-time cell analysis of the cytotoxicity of ILC2s against solid tumor cells and presented as the growth index of the residual cancer cells. Tumor-alone control group: no Ex ILC2s (n = 5). (C and D) Images represent BLI at the indicated time points and quantification of the BLI images (n = 6). (E and F) Survival of mice injected with or without Ex ILC2s in the indicated groups from C and D, respectively (n = 6). Data are shown as mean ± S.D. ∗p < 0.05, ∗∗p < 0.01. See also Figures S6, S7, and Videos S3A-S3B.

Discussion

Our study has identified human ILC2s as innate immune cytolytic effector cells, demonstrating a potency comparable to NK cells. This finding suggests that human ILC2s could serve as an adoptive cell strategy for cancer immunotherapy. In contrast to NK cells, this cytolytic function is not seen in mouse ILC2s, emphasizing the importance of studying this and other cell types in both species. We show that GZMB released from human ILC2s can cleave GSDME in GSDME⁺ tumor cells and cleave caspase 3 in GSDME⁺ and GSDME⁻ tumor cells, providing mechanistic insight as to how human ILC2s can exert antitumor immunity through pyroptosis and/or apoptosis.

Inhibitory receptors expressed as checkpoint molecules on the surface of immune cells have been targeted as a means for cancer immunotherapy. The success of this strategy also depends on activating receptors on effector CD8⁺ T cells and NK cells recognizing their cognate ligands on tumor cells⁵¹. One such activating receptor, DNAM-1, plays a critical role in anti-tumor immunity. Its ligands, CD112 and CD155, are highly expressed on tumor cells and their engagement with DNAM-1 promotes DNAM-1⁺ CD8⁺ T cell and NK cell-mediated recognition and target lysis⁵². DNAM-1-deficient mice show impaired clearance of CD155-expressing sarcoma cells, increased tumor growth, and significantly worse mortality than their WT counterparts⁵³. Like CD8⁺ T cells and NK cells, human ILC2s have high surface density expression of DNAM-1. Herein we elucidate the mechanism by which this critical activation receptor regulates ILC2 effector function. We show that both AML cell lines and primary AML patient blasts interact with healthy Ex ILC2s through DNAM-1 to elevate GZMB production, which in turn induces AML pyroptosis and/or apoptosis, thereby enhancing antitumor immunity. We further demonstrate that the DNAM-1-CD112/CD155 receptor-ligand interaction results in downstream phosphorylation and inactivation of the inhibitor FOXO1, thereby enhancing the production of GZMB in ILC2s. These findings have relevance in the clinic because checkpoint therapy targeting the CD112 axis is now moving into patients (NCT03667716). The binding site on CD112 is shared by both DNAM-1 and CD112R⁵⁴. Thus, when DNAM-1 successfully competes with the inhibitory immune checkpoint CD112R and binds to CD112, T cell and NK cell activation ensues^54,55. Therefore, checkpoint therapy targeting the CD112-CD112R interaction with an anti-CD112R monoclonal antibody (mAb) can improve the CD112–DNAM-1 interaction and enhance T cell and NK cell, as well as ILC2 cytotoxicity. Importantly, our data show that, unlike prolonged DNAM-1 binding to CD155, prolonged binding of DNAM-1 to CD112 does not decrease the level of DNAM-1 and thus should not promote tumor escape.

During tumor progression cancer cells adopt a variety of changes to escape from immune detection and clearance. Our report demonstrates that human ILC2s become dysregulated in AML patients with a profound decrease in DNAM-1 expression as the result of prolonged exposure to CD155 expressed on AML blasts. This may lead to a reduced recognition of AML by ILC2s, NK cells, and T cells, allowing AML to evade detection and immune clearance. Indeed, Ex ILC2s from HDs placed in short contact with primary AML blasts become activated to kill AML, while prolonged contact leads to loss of DNAM-1 surface expression. This loss can be due to CD155 expressed on AML cells interacting with DNAM-1 expressed on ILC2s thereby resulting in the internalization and proteasome degradation of DNAM-1, as has recently been reported for T cells⁴³. Hence, blocking CD155 and forcing DNAM-1 to bind to its other cognate ligand, CD112 as described above, might prevent tumor escape from a variety of DNAM-1⁺ immune cells including ILC2s, ILC1s, and NK cells, as suggested in the current study and a previous study⁵⁶. It’s worth noting that the anti-CD155 antibody, NTX-1088, is currently being tested in the clinic (NCT05378425).

Although autologous CAR T cell therapy has led to remarkable improvements in patients with aggressive B cell malignancies, challenges remain. Toxicities such as CAR T cell-related CRS, neurotoxicity, and allogeneic T cell receptor (TCR)-induced GVHD can develop, sometimes requiring intensive care unit support⁵⁷. Therapy of AML also remains challenging due to the heterogeneity and high relapse rate of this disease. The latter often precludes the use of autologous CAR T cells, as it takes several weeks to prepare the product once extracted from AML patients in remission. In this context, it may be possible to deliver expanded innate immune effector ILC2s as unmatched, allogeneic, off-the-shelf therapies with or without CARs as we have demonstrated using NK cells (NCT05334329). In this setting, such off-the-shelf, cryopreserved, and affordable innate immune therapies can be delivered immediately once the AML patient achieves remission, thus avoiding the possibility of relapse that can preclude the use of cell therapy. A similar platform can also be applied to treat solid tumors. Like NK cell infusions in preclinical models^58,59 and in the clinic², ILC2 infusions did not show any signs of overactive inflammation or autoimmunity while halting cancer progression in our tumor models. Therefore, reprogramming autologous ILC2s in vivo, administration of ex vivo allogeneic Ex ILC2s (without or with CARs), and combining Ex ILC2s with an FDA-approved anti-tumor agent represent future opportunities worth exploring.

In summary, we present a system for the expansion of authentic human ILC2s and identify previously unknown antitumor properties of human Ex ILC2s, not seen in mouse ILC2s. We demonstrate that human Ex ILC2s can play an important role in controlling the progression of both hematological and solid tumors in our animal models. These studies provide us with an additional strategy to improve outcomes in a variety of cancers by defining the role of ILC2s and exploring their therapeutic potential as a member of the cytolytic immune effector cell family.

Limitations of the study

Although our mechanistic study identifies a cytotoxic function of human ILC2s against tumor cells and offers a potential innovative anti-cancer immunotherapy, there are still questions that warrant further investigation. First, although we demonstrate the quantity and the function of ILC2s are decreased in patients with AML, additional mechanisms underlying this phenomenon besides those reported here likely exist. Second, while we demonstrate that the DNAM-1-CD112/CD155 receptor-ligand interaction is involved in the production of GZMB and cytolytic function in ILC2s, it’s important to acknowledge that other activating and inhibitory ligand-receptor axes involving tumor cells and ILC2s may also contribute to ILC2 effector function. Finally, the task of reprogramming Ex ILC2s to maximize their immune responsiveness while minimizing toxicity will require additional investigation.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact: Jianhua Yu (Jiayu@coh.org).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All data and code to understand and assess the conclusions of this research are available in the main text and supplementary materials. Single-cell RNA-seq and bulk RNA-seq data have been uploaded to the GEO. The accession numbers have been listed in the Key resources table.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-mouse CD3ε-FITC	Biolegend	Cat# 152304, RRID: AB_2632667
Anti-mouse NK1.1-FITC	Biolegend	Cat# 108706, RRID: AB_313393
Anti-mouse CD11c-FITC	Biolegend	Cat# 117306, RRID: AB_313775
Anti-mouse CD19-FITC	BD Biosciences	Cat#553785, RRID: AB_395049
Anti-mouse CD5-FITC	Biolegend	Cat# 100606, RRID: AB_312735
Anti-mouse FceR1 alpha-FITC	Thermo Fisher Scientific	Cat# 11–5898-82, RRID: AB_465308
Anti-mouse IL33Rα-APC	Biolegend	Cat# 146606, RRID: AB_2728175
Anti-mouse CD127-PE/Cy7	Biolegend	Cat# 135014, RRID: AB_1937265
Anti-mouse CD127-BV421	Biolegend	Cat# 135024, RRID: AB_11218800
Anti-mouse CD25-BV421	BD Biosciences	Cat# 562606, RRID: AB_11153485
Anti-mouse CD45-AF700	BD Biosciences	Cat# 560510, RRID: AB_1645208
Anti-mouse CD25-BV605	Biolegend	Cat# 102036, RRID: AB_ 2563059
Anti-mouse NK1.1-FITC	BD Biosciences	Cat# 561082, RRID: AB_10563221
Anti-mouse IL-9-PE	Biolegend	Cat# 514103, RRID: AB_2126639
Anti-mouse IL-4-BV786	Biolegend	Cat# 564006, RRID: AB_2738538
Anti-mouse IL-13-PerCP-eFluor710	Thermo Fisher Scientific	Cat# 46–7133-80, RRID: AB_11218893
Anti-mouse Siglec-F-BV786	BD Biosciences	Cat# 740956, RRID: AB_2740581
Anti-mouse GATA3-BV711	BD Biosciences	Cat# 565449, RRID: AB_2739242
Anti-mouse Ly6G-BUV395	Thermo Fisher Scientific	Cat# 363–9668-82, RRID: AB_2925309
Anti-mouse CD4-BV785	Biolegend	Cat# 100453, RRID: AB_2565843
Anti-mouse CD3-PE-Cy7	Biolegend	Cat# 100220, RRID: AB_1732057
Anti-human CD3-FITC	BD Biosciences	Cat# 561802, RRID: AB_10893003
Anti-human CD4-FITC	BD Biosciences	Cat# 555346, RRID: AB_395751
Anti-human CD8-FITC	BD Biosciences	Cat# 555634, RRID: AB_395996
Anti-human CD14-FITC	BD Biosciences	Cat# 555397, RRID: AB_395798
Anti-human CD15-FITC	BD Biosciences	Cat# 555401, RRID: AB_395801
Anti-human CD16-FITC	BD Biosciences	Cat# 555406, RRID: AB_395806
Anti-human CD19-FITC	BD Biosciences	Cat# 555412, RRID: AB_395812
Anti-human CD20-FITC	BD Biosciences	Cat# 555622, RRID: AB_395988
Anti-human CD33-FITC	BD Biosciences	Cat# 555626, RRID: AB_395992
Anti-human CD34-FITC	BD Biosciences	Cat# 555821, RRID: AB_396150
Anti-human CD203C-FITC	Thermo Fisher Scientific	Cat# MA5–28586, RRID: AB_2745545
Anti-human FcepsilonRIalpha-FITC	Biolegend	Cat# 334608, RRID: AB_1227653
Anti-human CD56-FITC	Biolegend	Cat# 362546, RRID: AB_2565964
Anti-human CD56-AF700	BD Biosciences	Cat# 557919, RRID: AB_396940
Anti-human CD56-BUV395	BD Biosciences	Cat# 563554, RRID: AB_2687886
Anti-human CD56-APC	Biolegend	Cat# 981204, RRID: AB_2715759
Anti-human CD127-APC	BD Biosciences	Cat# 558598, RRID: AB_647113
Anti-human CD127-BV421	Biolegend	Cat# 351310, RRID: AB_10960140
Anti-human CD127-PE	Biolegend	Cat# 351340, RRID: AB_2564136
Anti-human CD117-BV711	Biolegend	Cat# 313230, RRID: AB_2566217
Anti-human CD117-PE	BD Biosciences	Cat# 555714, RRID: AB_396058
Anti-human CRTH2-PE-Cy7	Biolegend	Cat# 350118, RRID: AB_2562470
Anti-human CD5-BV510	Biolegend	Cat# 364018, RRID: AB_2565728
Anti-human CD161-BV785	Biolegend	Cat# 339930, RRID: AB_2563968
Anti-human CD161-PE	Biolegend	Cat# 307504, RRID: AB_2876602
Anti-human DNAM-1-AF647	Biolegend	Cat# 338328, RRID: AB_2728298
Anti-human NKG2D-PE	BD Biosciences	Cat# 557940, RRID: AB_396951
Anti-human NKp30-BV421	BD Biosciences	Cat# 563385, RRID: AB_2738171
Anti-human CD122-APC	Biolegend	Cat# 339007, RRID: AB_2248891
Anti-human CD112-PE	Biolegend	Cat# 337410, RRID: AB_2269088
Anti-human CD155-APC	Biolegend	Cat# 337618, RRID: AB_2565815
Anti-human MICA-APC	R&D	Cat# FAB1300A, RRID: AB_416836
Anti-human MICA-PE	R&D	Cat# FAB1300P, RRID: AB_416837
Anti-human MICB-APC	R&D	Cat# FAB1599A, RRID: AB_2297703
Anti-human TIGIT-BV605	BD Biosciences	Cat# 747841, RRID: AB_2872304
Anti-human PD-1-PE-Cy7	Thermo Fisher Scientific	Cat# 25–9969-42, RRID: AB_2688257
Anti-human TIM3-APC-Cy7	Biolegend	Cat# 345026, RRID: AB_2565717
Anti-human IL-33R-APC	Thermo Fisher Scientific	Cat# 17–9338-42, RRID: AB_2762446
Anti-human IL17RB-PE	R&D	Cat# FAB1207P, RRID: AB_2125555
Anti-human KLRG1-APC	Miltenyi Biotec	Cat# 130–117-703, RRID: AB_2733432
Anti-human CD45-BUV395	BD Biosciences	Cat# 563792, RRID: AB_2869519
Anti-human CD45-BV510	BD Biosciences	Cat# 563204, RRID: AB_2738067
Anti-human CD45RO-APC/Cy7	Biolegend	Cat# 304228, RRID: AB_10895897
Anti-human CD45RA-BV650	Biolegend	Cat# 304136, RRID: AB_2563653
Anti-human HLA-DR-BV510	Biolegend	Cat# 307645, RRID: AB_2561396
Anti-human CD14-BV421	Biolegend	Cat# 367143, RRID: AB_2810579
Anti-human CD16-PE	Biolegend	Cat# 360704, RRID: AB_2562749
Anti-human CD206-PE/Cy7	Biolegend	Cat# 321123, RRID: AB_10900995
Anti-human CD24-APC/Cy7	Biolegend	Cat# 311131, RRID: AB_2566346
Anti-human CD11c-BV785	Biolegend	Cat# 301643, RRID: AB_2565778
Anti-human CD66b-PE	Biolegend	Cat# 392903, RRID: AB_2750201
Anti-CD11b-PE/Cy7	BD Biosciences	Cat# 552850, RRID: AB_394491
Anti-human CD123-BV650	Biolegend	Cat# 306019, RRID: AB_11218792
Anti-human CD33-BV785	Biolegend	Cat# 303428, RRID: AB_2650888
Anti-human IL-4-BV510	Biolegend	Cat# 500835, RRID: AB_2650992
Anti-human IL-5-eFluor450	Thermo Fisher Scientific	Cat# 48–7052-82, RRID: AB_2802295
Anti-human IL-9-PE	Biolegend	Cat# 507605, RRID: AB_315487
Anti-human IL-13-APC	Biolegend	Cat# 501907, RRID: AB_315202
Anti-human Perforin-BV421	BD Biosciences	Cat# 563393, RRID: AB_2738178
Anti-human IFN-γ-PE-Cy7	Biolegend	Cat# 502528, RRID: AB_2123323
Anti-humanTNF-α-APC	Biolegend	Cat# 502912, RRID: AB_315264
Anti-GATA3-PE-CF594	BD Biosciences	Cat# 563510, RRID: AB_2738248
Anti-GATA3-PE	Biolegend	Cat# 653804, RRID: AB_2562723
Anti-T-bet Antibody-APC	Biolegend	Cat# 644814, RRID: AB_10901173
Anti-T-bet Antibody-BV421	Biolegend	Cat# 644816, RRID: AB_10959653
Anti-EOMES-BUV395	BD Biosciences	Cat# 567171, RRID: AB_2916488
Anti-human RORγt-AF647	BD Biosciences	Cat# 563620, RRID: AB_2738324
Anti-human Granzyme B-PE	BD Biosciences	Cat# 561142, RRID: AB_10561690
Anti-human/mouse Granzyme B-PE	Biolegend	Cat# 396406, RRID: AB_2801075
Granzyme B Monoclonal Antibody (GB11)	Thermo Fisher Scientific	Cat# MA1–80734, RRID: AB_931084
Perforin Polyclonal Antibody	Thermo Fisher Scientific	Cat# PA5–109315, RRID: AB_2854726
IL-13 Monoclonal Antibody (JES10–5A2)	Thermo Fisher Scientific	Cat# AHC0132, RRID: AB_2536246
IL-4 Monoclonal Antibody (MP4–25D2)	Thermo Fisher Scientific	Cat# 16–7048-81, RRID: AB_469210
IL-5 Monoclonal Antibody (TRFK5)	Thermo Fisher Scientific	Cat# 14–7052-85, RRID: AB_468421
Purified anti-human CD226 (DNAM-1)	Biolegend	Cat# 338302, RRID: AB_1279155
Purified anti-human CD314 (NKG2D)	Biolegend	Cat# 320802, RRID: AB_492956
Ultra-LEAF(TM) Purified anti-human IL-9	Biolegend	Cat# 512005, RRID: AB_2888812
Purified anti-human CD124 (IL-4Ralpha)	Biolegend	Cat# 355002, RRID: AB_11219599
Ultra-LEAF™ Purified anti-human CD337 (NKp30) Antibody	Biolegend	Cat# 325224, RRID: AB_2814183
Anti-phospho-Akt (Ser473)	Cell Signaling Technology	Cat# 4060S, RRID: AB_2315049
Anti-pan Akt	Cell Signaling Technology	Cat# 2920S, RRID:AB_1147620
Alexa Fluor 647 Annexin V	Biolegend	Cat# 640943, RRID: AB_2616658
7-AAD Staining Solution 2mL antibody	BD Biosciences	Cat# 559925, RRID: AB_2869266
DAPI Solution	BD Biosciences	Cat# 564907, RRID: AB_2869624
Violet Live Cell Caspase Probe	BD Biosciences	Cat# 565521, RRID: AB_2869682
Chemicals, Peptides, and Recombinant Proteins
Recombinant Human IL-2	NCI	N/A
Recombinant Human IL-7	NCI	N/A
Recombinant Human IL-15	NCI	N/A
Recombinant Human IL-33	PeproTech	Cat# 200–33
Recombinant Murine IL-33	PeproTech	Cat# 210–33
Recombinant Murine IL-7	PeproTech	Cat# 217–17
Z-VAD-FMK	MCE	Cat# HY-16658B
Z-DEVD-FMK	MCE	Cat# HY-12466
Necrostatin-1	MCE	Cat# HY-15760
Ferrostatin-1	MCE	Cat# HY-100579
AS1842856	MCE	Cat# HY-100596
Afuresertib	MCE	Cat# HY-15727
Trametinib	MCE	Cat# HY-10999
Cas9 Nuclease V3	IDT	Cat# 1081059
AmaxaTM P3 Primary Cell 4D-NucleofectorTM X Kit	Lonza	Cat# V4XP-3032
RIPA Lysis and Extraction Buffer	Thermo Fisher Scientific	Cat# 89900
Fixation/Permeabilization Solution Kit	BD Biosciences	Cat# 554714
Leukocyte Activation Cocktail, with BD GolgiPlug™	BD Biosciences	Cat# 550583
Biological Samples
Peripheral Blood Samples of Patients with AML	City of Hope National Medical Center	N/A
Peripheral Blood Samples of Healthy Donors	City of Hope National Medical Center	N/A
Critical Commercial Assays
RosetteSep™ Human NK Cell Enrichment Kit	STEMCELL	Cat# 15065
EasySep™ Human Pan-ILC Enrichment Kit	STEMCELL	Cat# 17975
Anti-Cy7 MicroBeads	Miltenyi Biotec	Cat# 130–091-652
Anti-FITC MicroBeads	Miltenyi Biotec	Cat# 130–048-701
CellTrace™ Violet Cell Proliferation Kit	Thermo Fisher Scientific	Cat# C34557
CellTrace™ Far Red Cell Proliferation Kit	Thermo Fisher Scientific	Cat# C34572
CellTrace™ CFSE Cell Proliferation Kit	Thermo Fisher Scientific	Cat# C34554
High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor	Thermo Fisher Scientific	Cat# 4374967
RNeasy Mini Kit (250)	QIAGEN	Cat# 74106
Decemotinib (VX-509)	Selleck	Cat# S7541
D-Luciferin, Potassium Salt	GOLDBIO	Cat# LUCK-10G
Caspase-Glo® 3/7 Assay System	Promega	Cat# G8090
Steady-Glo® Luciferase Assay System	Promega	Cat# E2520
Perforin Human ELISA Kit	Thermo Fisher Scientific	Cat# BMS2306
Human Granzyme B DuoSet ELISA Kit	R&D	Cat# DY2906–05
Deposited Data
ScRNAseq data	This manuscript	GSE247205
Bulk RNA Seq data	This manuscript	GSE247206 and GSE247207
Experimental Models: Cell Lines
MOLM13	DSMZ	Cat# ACC-554; RRID: CVCL_2119
U937	ATCC	Cat# CRL-1593.2, RRID: CRL-1593.2
THP1	ATCC	Cat# TIB-202, RRID: CVCL_0006
A549	ATCC	Cat# CRM-CCL-185, RRID: CVCL_0023
Capan-1	ATCC	Cat# HTB-79, RRID: CVCL_0237
MIA PaCa-2	ATCC	Cat# CRM-CRL-1420, RRID: CVCL_0428
Gli36	E. Antonio Chiocca	Department of Neurosurgery, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
U251	E. Antonio Chiocca	Department of Neurosurgery, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
GBM30	E. Antonio Chiocca	Department of Neurosurgery, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
LN229	ATCC	Cat# CRL-2611, RRID: CVCL_0393
OP9	ATCC	Cat# CRL-2749, RRID: CVCL_4398
C1498	ATCC	Cat# TIB-49, RRID: CVCL_3494
HEK293T	ATCC	Cat# CRL-3216, RRID: CVCL_0063
Experimental Models: Organisms/Strains
C57BL/6 mice	The Jackson Laboratory	Stock No: 000664; RRID: IMSR_JAX:000664
Mouse: MllPTD/WT: Flt3ITD/ITD mouse	Michael A. Caligiuri (Zorko et al.,2012)60	N/A
Mouse: C;129S4-Rag2tm1.1Flv Il2rgtm1.1Flv/J (Rag2− /−γc−/−)	The Jackson Laboratory	Stock No: 014593; RRID: IMSR_JAX:014593
NOD.Cg-Prkdcscid Il2rgtm1Wjl Tg(CMV-IL3,CSF2,KITLG)1Eav/MloySz J	The Jackson Laboratory	Stock No: 013062; RRID: IMSR_JAX:013062
Software and Algorithms
GraphPad Prism 9	GraphPad Software	https://www.graphpad.com/
FlowJo software (version 10)	FlowJo	https://www.flowjo.com/
Cell Ranger (Version 5. 0.0)	10 × Genomics	https://support.10xgenomics.com/single-cell-geneexpression/software/pipelines/latest/installation
Trimmomatic	(Bolger et al., 2014)76	http://www.usadellab.org/cms/?page=trimmomat ic
FASTP	(Chen et al., 2018)77	https://github.com/OpenGene/fastp
STAR (v. 020201)	(Anders and Huber, 2010)78	https://github.com/alexdobin/STAR
HTSeq v.0.6.0	(Anders and Huber, 2010)78	https://github.com/simon-anders/htseq
GSEA-2.2.3	(Subramanian et al., 2005)80	https://www.gsea-msigdb.org/gsea/index.jsp

The paper does not report the original code.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Human sample collection

AML specimens were collected from patients with AML registered at City of Hope National Medical Center (COHNMC) who consented to an Institutional Review Board-approved protocol (IRB# 18067); healthy donor specimens were collected from patients who consented to IRB# 06229. Patient sample acquisition was approved by the IRB at the COHNMC, in accordance with an assurance filed with and approved by the Department of Health and Human Services and met all requirements of the Declaration of Helsinki. Mononuclear cells were isolated using Ficoll separation. ILCs were isolated using EasySep^™ Human Pan-ILC Enrichment Kit (STEMCELL) or were sorted using a BD FACSAria™ Fusion.

Animal model

NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG), NOD.Cg-Prkdc^scid Il2rg^tm1Wjl Tg (CMV-IL3, CSF2, KITLG)1Eav/MloySzJ (NSG-SGM3) mice, and wild-type C57BL/6J mice were purchased from the Jackson Laboratory. Mll^PTD/WT: Flt3^ITD/ITD mice⁶⁰ on the B6 background were generated by our group. Both males and females were used. All mice were maintained by the Animal Resource Center of City of Hope. For animal studies, mice of the same age and sex were divided randomly into experimental groups. Experimenters were blinded to observe mouse survival. Mice were fed with PicoLab Rodent Diet 20 (5053) and housed in the City of Hope Animal Facility with a 12-h light/12-h dark cycle and temperatures of ~18–23 °C with 40–60% air humidity. Mouse care and experimental procedures were performed in accordance with federal guidelines and protocols approved by the Institutional Animal Care and Use Committee (IACUC) at City of Hope. Tumor-bearing mice were monitored twice per week and at more frequent intervals depending on the status of the mice. Mice exhibiting evidence of distress, discomfort, pain, lethargy, inability to properly groom, or inability to obtain food and/or water were killed immediately via CO₂ inhalation. Tumor-bearing mice with 20% weight loss from the age-matched controls without receiving tumor cell inoculation were killed.

Cell lines

OP9-mDL1 and OP9-mDL4 were maintained in MEM GlutaMAX media (Gibco) supplemented with 20% FBS (Gibco) and 50 μM β-mercaptoethanol. The human AML cell lines MOLM13, U937, and THP1 were cultured in RPMI 1640 (Gibco) with 10% FBS. All AML cell lines expressed luciferase. These cell lines were purchased from the American Type Culture Collection (ATCC) or Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (DSMZ). The human lung cancer cell line A549, pancreatic cancer cell lines Capan-1 and MIA PaCa-2, and brain cancer cell lines Gli36, LN229, and U251 were cultured in DMEM GlutaMAX media with 10% FBS. GBM30 spheroid cells derived from a patient with GBM and modified to express luciferase were maintained with neurobasal media (DMEM/F12) supplemented with 2% B27 (Gibco), human epidermal growth factor (StemCell), basic fibroblast growth factor (StemCell), heparin (StemCell), and Glutamax (Gibco) in low-attachment cell culture flasks. These cell lines were either purchased from the ATCC or obtained from Dr. E. Antonio Chiocca’s laboratory at Harvard University (Cambridge, MA). The mouse AML cell line (C1498) was cultured in RPMI 1640 with 10% FBS. The cell line was purchased from ATCC. All cell lines were routinely tested for the absence of Mycoplasma using the MycoAlert Plus Mycoplasma Detection Kit (Lonza). All cell culture media are supplemented with penicillin (100 U/mL) and streptomycin (100 mg/mL). Cultures were incubated at 37°C in a humidified atmosphere of 5% CO₂. Penicillin and streptomycin were from ThermoFisher Scientific.

METHOD DETAILS

Isolation and expansion of human ILC2s

To isolate ILC2s from human peripheral blood, we diluted blood cone samples 1:1 with phosphate-buffered saline (PBS). We layered the blood on the top of Ficoll-Paque (GE Healthcare) and centrifuged it according to the manufacturer’s instructions. The mononuclear cell fraction was aspirated and washed with PBS, and then the red blood cells were lysed. Total ILCs were initially isolated from PB to separate them from other lineage-negative cells using the RosetteSep^™ Human NK cell enrichment kit (StemCell). Subsequently, non-NK ILCs (ILC1s, ILC2s, and ILC3s) were enriched by excluding NK cells from the total ILC population using the EasySep^™ Human Pan-ILC enrichment kit (StemCell). Isolated ILCs were cultured on either DL1-expressing OP9 (hereafter referred to as DL1) or DL4-expressing OP9 (hereafter referred to as DL4) stromal cells in the presence of IL-2, IL-7, and IL-15. After 14 days, we sorted ILC2s using the lineage markers (Lin: CD3, CD4, CD8, CD14, CD15, CD16, CD19, CD20, CD33, CD34, CD123, CD203c, and FceRI), the surface marker CD161 (a human ILC marker), and a combination of CD56, chemoattractant receptor–homologous molecule expressed on Th2 lymphocytes (CRTH2), and CD117. As surface staining of CD127 (IL-7 receptor-α), a human pan-ILC marker, is affected by high IL-2 and IL-7 concentrations^61–63, we instead used CD161. The gating strategy of freshly isolated ILC2s and expanded ILC2s is described in Data S1, pages 1A and B, respectively. The purity of CD161⁺ ILC2s was approximately 97% after flow cytometry sorting on day 14 (Data S1, page 1B). Next, we further cultured and expanded FACS-sorted HD ILC2s on OP9 stromal cells in the presence of IL-2, IL-7, and IL-15. Human ILCs were cultured in the MEM GlutaMAX media (Gibco). The medium was supplemented with 10% human AB serum (Sigma-Aldrich), IL-2 (500 IU/ml), IL-7 (20 ng/ml), and IL-15 (20 ng/ml). Medium and cytokines were refreshed every two days by replacing half of the media containing 1 × concentration of cytokines. Every four days, ILC2s were re-cultured onto new plates with fresh OP9. Two weeks later, the ILC2s stained with lineage (anti-CD3, anti-CD4, anti-CD8, anti-CD14, anti-CD15, anti-CD16, anti-CD19, anti-CD20, anti-CD33, anti-CD34, anti-CD123, anti-CD203c, anti-FcεRIα, and anti-CD56), anti-CD161, anti-CRTH2, and anti-c-Kit (anti-CD117) antibodies were sorted using a BD FACSAria™ Fusion (BD Biosciences). FACS-sorted ILC2s were plated and cultured onto fresh OP9 cells in the fresh MEM GlutaMAX media with cytokines. All cytokines were provided by the National Institutes of Health.

Flow cytometry

Cell surface staining was done at 4 °C for 30 min in the dark. Briefly, single cells were stained with conjugated fluorescent or isotype controls and incubated for 30 min. Then cells were washed with ice-cold PBS (1% FBS in PBS, 5 mM EDTA) two times by centrifugation at 2000 rpm for 5 min. The cells are then ready for flow analysis.

ILC2s from human peripheral blood were identified using surface staining with a live/dead cell viability cell staining kit (Invitrogen) and the following monoclonal antibodies: anti-lineage cocktail (Lin; FITC-conjugated anti-CD3, anti-CD4, anti-CD8, anti-CD14, anti-CD15, anti-CD16, anti-CD19, anti-CD20, anti-CD33, anti-CD34, anti-CD123, anti-CD203c, and anti-FcεRIα), anti-CD56 (FITC, BV711 or BV421-conjugated), anti-CD127 (BV421 or APC-conjugated), anti-CRTH2 (PE-Cy7-conjugated), and anti-c-Kit or -CD117 (PE, PE-CF594, or BV711-conjugated). Expression of CD155 and CD112 on MOLM13, U937, and THP1 was identified by APC-conjugated anti-CD155 and PE-conjugated anti-CD112, respectively. Expression of IL-33R on human ILC2s was identified by APC-conjugated anti-IL-33R. Expression of DNAM-1, NKG2D, and NKp30 on human ILC2s was identified by FITC or AF647-conjugated anti-DNAM-1, PE-conjugated anti-NKG2D, and BV421 conjugated anti-NKp30, respectively. Expression of CD5, CD45RA, and CD45RO was identified by BV510-conjugated anti-CD5, BV650-conjugated anti-CD45RA, and APC-Cy7-conjugated anti-CD45RO, respectively. Human primary AML blasts were gated by Lin⁻CD45^dim. Human neutrophils, mast cells, eosinophils, Th2 cells, and dendritic cells were identified below using surface staining with a live/dead cell viability cell staining kit and the following monoclonal antibodies^64,65: neutrophils: mCD45⁻hCD45⁺CD14⁻CD206⁻CD16⁺CD24^−/+; mast cells: mCD45⁻hCD45⁺CD14⁻ FcɛRIα⁺CD117⁺; eosinophils: mCD45⁻hCD45⁺CD14⁻CD206⁻CD16⁻CD24⁺; Th2 cells: mCD45⁻hCD45⁺CD4⁺CRTH2⁺GATA3⁺; dendritic cells: mCD45⁻hCD45⁺CD14⁻CD206⁻CD123⁻CD11c⁺HLA-DR⁺; M-MDSCs: mCD45⁻hCD45⁺CD3/CD19⁻HLA-DA⁻CD11b⁺CD33⁺CD14⁺CD66b⁻; G-MDSCs: mCD45⁻hCD45⁺CD3/CD19⁻HLA-DA⁻CD11b⁺CD33⁺CD14⁻CD66b⁺.

Mouse ILC2s were identified using surface staining with a live/dead cell viability cell staining kit and the following monoclonal antibodies²³: lineage (Lin; FITC-conjugated anti-CD3, anti-CD11c, anti-CD11b, anti-CD5, anti-CD19, anti-NK1.1, anti-FcεRIα), PE-Cy7-conjugated CD127, BV421-conjugated CD25, and APC-conjugated IL-33R/ST2. The gating was Lin⁻CD127⁺CD25⁺IL-33R⁺. Mouse neutrophils, mast cells, eosinophils, Th2 cells, and dendritic cells were identified using surface staining, and dead cells were excluded by staining cells with a live/dead cell viability cell staining kit. The cell populations were identified as previously described ^66–69: neutrophils: Ly6G⁺Siglec⁻; mast cells: FcεRIα⁺c-Kit⁺; eosinophils: Ly6G⁻Siglec⁺; dendritic cells: Ly6G⁻Siglec⁻CD11c⁺MHC II⁺; and Th2 cells: CD3⁺CD4⁺GATA3⁺.

To examine intracellular cytokine production, we stimulated freshly isolated ILC2s from health donors or mice, expanded human ILC2s (Ex ILC2s), or expanded mouse ILC2s with or without Leukocyte Activation Cocktail in the presence of BD GolgiPlug^™ for 4 h. In some experiments, Ex ILC2s or expanded mouse ILC2s were co-cultured with or without tumor cells for 48 h (in the absence of stimulating the Leukocyte Activation Cocktail). Intracellular staining for granzyme B or perforin was performed using a Fix/Perm kit (BD Biosciences), followed by staining with granzyme B or perforin, respectively. Intracellular staining for caspase 3, IL-4, IL-5, IL-9, IL-13, IFNγ, or TNF was performed using a Fix/Perm kit, followed by staining with an anti-caspase 3, anti-IL-4, anti-IL-5, anti-IL-9, anti-IL-13, anti-IFNγ or anti-TNF antibody, respectively. Intracellular staining for human transcription factors (GATA3, RORγt, EOMES, or T-BET) or mouse transcription factors (GATA3) was performed using a Fix/Perm kit (ThermoFisher Scientific), followed by staining with anti-GATA3, anti-RORγt, anti-EOMES, or anti-T-BET for human cells and anti-GATA3 for mouse cells.

To examine tumor cell death, tumor cells were co-cultured with or without ILC2s for 48 h. Cells were then stained with PE- or APC-conjugated Annexin V and with 7-amino-actinomycin D (7-AAD, BD Biosciences) or 4′,6-diamidino-2-phenylindole (DAPI), followed by flow cytometric analysis. All human antibodies were used at 1:50, except for CD34, CRTH2, and CD117, which were used at 1:100. All mouse antibodies were used at 1:200. All analyses were performed on a Fortessa X-20 flow cytometer (BD Biosciences), and sorting was performed using a BD FACSAria™ Fusion (BD Biosciences). Flow Cytometry data were analyzed by FlowJo V10 (Treestar).

Single-cell RNA sequencing library preparation

Ex ILC2s, ILCs from the blood of healthy donors (HD ILCs), ILCs from the blood of patients with AML (AML ILCs), and human CD45⁺ cells from the BM of NSG mice (Traced ILC2s) were sorted using BD FACSAria. Cell samples from three donors in each of the Ex ILC2s, HD ILCs, and Traced ILC2s groups were pooled into single samples for single-cell sequencing (ScRNA-seq). In contrast, separate samples were obtained from individual patients with AML for their corresponding scRNA-seq analyses. Cell samples containing 3,000–6,000 cells were captured using a 10 × Genomics Chromium controller and the 10 × V3.1 Single Cell 3’ Solution kit (Chromium Next GEM Single Cell 3’ Regent kits V3.1, Cat# PN-1000268). All protocols were followed by the manufacturer’s instructions. The final sequencing libraries were assessed for library size using a High Sensitivity DNA Chip (Agilent, Cat# 5067–4626). Library concentration was determined using a Qubit High Sensitivity DNA assay Kit (ThermoFisher Scientific, Cat# Q32854). Sequencing was conducted using the Illumina NovaSeq 6000 platform with the paired-end setting, involving 28 cycles of read1, 101 cycles of R2, 10 cycles of index i7, and 10 cycles of index i5 read. The sequencing kit used was S4 Reagent kit v1.5 (Illumina, Cat# 20028313) at TGen. Demultiplexing, genome alignment, and feature-barcode matrix generation were performed using the 10 × Genetics Cell Ranger software pipeline.

Single-cell RNA-seq data analysis

Raw sequencing data from each sample were aligned back to the human genome (GRCh38) using the Cell ranger count command to produce expression data at a single-cell resolution according to 10 × Genomics (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/using/count). The Seurat R package⁷⁰ was employed to facilitate various critical data processing and analysis tasks, including gene and cell filtration, data normalization, principal component analysis (PCA), identification of variable genes, clustering analysis, and Uniform Manifold Approximation and Projection (UMAP) dimension reduction. To provide a succinct overview of the methodology, individual matrices containing gene-by-cell expression data were imported to create distinct Seurat objects for each sample. Subsequently, cells with fewer than 200 detectable genes and a mitochondrial gene content exceeding 15% were systematically excluded. Following this, the data from different samples were harmoniously integrated using the scTransform package for subsequent analytical procedures. Dimensionality reduction was carried out through PCA, retaining the first 23 principal components for subsequent clustering analysis. The resulting clusters were visually represented using UMAP embedding, and a comprehensive examination of gene expression patterns was conducted through the utilization of various visualization techniques, including VlnPlot, FeaturePlot, and DotPlot.

Lineage tracing assay

Six- to eight-week-old NSG mice were injected with 5 × 10⁶ Ex ILC2s. Recombinant human IL-2 (500 IU/mouse), IL-7 (2 μg/mouse), and IL-15 (2 μg/mouse) were intraperitoneally injected into recipient mice daily for 7 days. Seven days later, mice were sacrificed and human CD45⁺ cells from the BM of mice were sorted to perform scRNA-seq.

In vitro co-culture assay of ILC2s and tumor cells

For assays using human ILC2s co-cultured with AML cells in a cell-cell contact manner, AML cells were co-cultured with various numbers of HD ILC2s or Ex ILC2s for 48 h. For the co-culture of AML cells and ILC2s in the transwell co-culture system, AML cells were seeded in the lower chamber of a 24-well transwell plate, while ILC2s were seeded in the upper chamber for 48 h. The ratio of ILC2s: AML cells was 2:1. For co-culture assays with antibodies, ILC2s were co-cultured with anti-DNAM-1 (10 μg/ml) antibody, anti-NKG2D (10 μg/ml) antibody, or anti-NKp30 antibody (10 μg/ml) for 2 h, followed by adding AML cells at the 2:1 ratio of ILC2s: AML cells.

For co-culture assays of human ILC2s and solid tumor cells, solid tumor cells were co-cultured with various numbers of Ex ILC2s for 48 h.

For co-culture assays of mouse ILC2s and mouse AML cells, C1498 cells were co-cultured for 48 h with various numbers of expanded mouse ILC2s that were isolated from either BM, spleen, liver, or lung of wild-type mice.

For all co-culture assays, cells were harvested two days later and analyzed using flow cytometry. Annexin V and 7-AAD or DAPI were used to identify dead cells following the manufacturer’s instructions. Cell images or videos were taken by microscope (Zeiss AxioCam 702).

In vivo human and mouse tumor cell transplantation assay

For the human AML cell engraftment experiment, 6–8-week-old NSG were transplanted with 5 × 10⁴ MOLM13, 5 × 10⁴ U937, or 0.1 × 10⁶ THP1 cells via tail vein injection. Subsequently, mice implanted with MOLM13 or U937 cells received an intravenous (i.v.) injection of 2 × 10⁶ Ex ILC2s, while those implanted with THP1 cells received an i.v. injection of 4 × 10⁶ Ex ILC2s. Recombinant human IL-2 (500 IU/mouse) was intraperitoneally (i.p.) injected into recipient mice daily for 7 days. For the human AML cell engraftment experiment with human PBMCs, 5 ×10⁴ MOLM13 and 10 × 10⁶ PBMCs were co-transplanted via tail vein injection into 6–8-week-old NSG-SGM3 mice. This was followed by i.v. injection of 2 × 10⁶ Ex ILC2s or expanded human NK cells into the mice that were implanted with MOLM13.

For the trametinib treatment experiment, NSG mice transplanted with 1 × 10⁵ U937 cells were treated with or without trametinib (0.8 mg/kg) for 7 days. Next, Ex ILC2s were i.v. injected into these mice. Seven days later, the expression of DNAM-1 on Ex ILC2s was measured.

For the human lung cancer cell implantation experiment, 5 × 10⁴ A549 cells were transplanted via i.v. injection into 6–8-week-old NSG mice and followed by weekly i.v. injection of 5 × 10⁶ Ex ILC2s into the mice for three weeks. Recombinant human IL-2 (500 IU/mouse) was i.p. injected into recipient mice every two days for three weeks.

For the human pancreatic cancer cell implantation experiment, 5 × 10⁴ Capan-1 cells were injected via i.p. injection into 6–8-week-old NSG mice, followed by both i.v. and i.p. injection of 5 × 10⁶ Ex ILC2s. Recombinant human IL-2 (500 IU/mouse) was i.p. injected into recipient mice daily for 7 days. For the human Capan-1 engraftment with human PBMCs experiment, 10 × 10⁶ PBMCs were i.v. injected into 6–8-week-old NSG-SGM3 mice. One day later, 5 × 10⁴ Capan-1 cells were i.p. injected into the mice, followed by a separate i.v. injection and an i.p. injection of 5 × 10⁶ Ex ILC2s. Three days after the first treatment, the mice implanted with Capan-1 cells received a second i.p. injection of 5×10⁶ Ex ILC2s.

For the human brain cancer cell implantation experiment (orthotopic GBM models), we followed our previous procedures^71–74. Briefly, 1 × 10⁴ luciferase-expressing GBM30 cells were injected into the right frontal lobe of the brain (2-mm lateral and 1-mm anterior to bregma at a depth of 3-mm), followed by weekly intratumoral injection of 3 × 10⁶ Ex ILC2s into the same location as above for three weeks. Recombinant human IL-2 (500 IU/mouse) was i.p. injected into recipient mice every two days for three weeks.

For the mouse AML cell engraftment experiment, 1 × 10⁶ C1498 cells were transplanted via tail vein injection into 6–8-week-old C57BL/6J mice, followed by i.v. injection of 2 × 10⁶ expanded mouse ILC2s.

All mice were subsequently monitored frequently for AML, pancreatic, brain, or lung cancer disease progression and imaged to check tumor growth at different time points. Tumor burden was assessed via in vivo bioluminescence measurements using the IVIS Imaging System at the City of Hope Small Animal Imaging Core. For luciferase detection imaging, 200 μl of 15 mg/ml D-luciferin (Caliper Life Sciences) in PBS was injected i.p. before imaging. The observers were blinded to the group allocation. Recombinant human IL-2 (500 IU/mouse) was i.p. injected into recipient mice daily for three weeks.

Caspase 3/7 activity assay

ILC2s were co-cultured with AML cells at an effector/target ratio of 2:1 for 12 h. Next, 100 μl of Caspase-Glo 3/7 reagent was added to each well. Plates were then shaken at 300 rpm for 1 min, incubated for 60 min at room temperature, and then read on a luminometer (Promega, Glomax). Background luminescence was determined with 100 μl of culture medium without cells and subtracted before fold changes were calculated.

Single-cell cytotoxicity assay

CFSE-labeled MOLM13, U937, or THP1 cells were mixed with DAPI to detect dead cells and then loaded into SIEVEWELL^Ⓡslides with nanowells of 50 μm (https://www.sievewell.com/wp-content/uploads/2019/09/2211_SIEVEWELL_Brochure_EN.pdf). Next, freshly FACS-sorted human ILC2s, which had been recovered overnight, were labeled with Far-red CellTrace Dye and were also loaded SIEVEWELL^Ⓡ slides. Both AML cells and ILC2s were then sedimented. Time-lapse images or videos were captured every 10 seconds (MOLM13 and U937) or 30 seconds (THP1) by microscope (Zeiss AxioCam 702).

Luminescence-based in vitro cell death assay

HD or Ex ILC2s were cultured at indicated ratios with AML cells. After 48 h, 100 μl of the mixture was transferred to a 96-well white luminometer plate. Next, 10 μl of the substrate (Promega) was added, and luminescence (RLU) was immediately determined. The results are reported as percent killing based on luciferase activity in the wells with tumor cells but not ILC2s [% killing = 100 – [(RLU) from well with effector and target cell co-culture) / (RLU from well with target cells) × 100)].

Gene expression analyses

For regular PCR analyses, RNA was isolated from 50,000 cells using a miRNeasy mini kit (QIAGEN) and reverse transcribed using a PrimeScript RT reagent kit with gDNA Eraser (TAKARA). PCR reactions were run on a ProFlex PCR System (Applied Biosystems) using 2×MyTaq Red Mix (Meridian Bioscience). RT–PCR analysis was conducted to assess expression of human IL-33 (Forward: 5’-GTGACGGTGTTGATGGTAAGAT-3’; reverse: 5’-AGCTCCACAGAGTGTTCCTTG-3’), human IL-25 (forward:5’-CAGGTGGTTGCATTCTTGGC; reverse: 5’-GAGCCGGTTCAAGTCTCTGT-3’), human TSLP (forward: 5’-ATGTTCGCCATGAAAACTAAGGC-3’; reverse: 5’-GCGACGCCACAATCCTTGTA-3’), PGD2 (forward: 5’-GGCGTTGTCCATGTGCAAG-3’; reverse: 5’-GGACTCCGGTAGCTGTAGGA-3’), and 18S rRNA (forward: 5′-GTAACCCGTTGAACCCCATT-3′; reverse: 5′-CCATCCAATCGGTAGTAGCG-3′). The primers were purchased from Integrated DNA Technologies.

For human ILC2 RNA-seq, the Ex ILC2s were co-cultured for 48 h with or without MOLM13 or U937 cells at an effector/target ratio of 2:1. The Ex ILC2s were sorted from ILC2-AML cell co-culture using a BD FACSAria Fusion. Total RNA was isolated from Ex ILC2s using a miRNeasy mini kit (QIAGEN). Poly(A) RNA-seq was performed in the Integrative Genomics Core of the City of Hope National Medical Center. A SMART-Seq ultra-low input RNA kit for sequencing v4 (Takara Bio) was used to generate double-stranded cDNA from each sample with 2 ng of input total RNA. The resulting cDNA was sheared using a Covaris LE220 sonicator. The sheared DNA was used to prepare a sequencing library using a KAPA HyperPrep kit. The final libraries were quantified using a Qubit assay kit (Thermo Fisher Scientific) and Bioanalyzer (Agilent). Sequencing was performed using the single-read mode of 51 cycles of read 1 and 7 cycles of index read with V4 reagents on a HiSeq 2500 system (Illumina). Real-time analysis 2.2.38 software was used to process the image analysis and base calling.

Confocal microscopy

The FACS-sorted Ex ILC2s were stimulated with Leukocyte Activation Cocktail in the presence of BD GolgiPlug^™ for 4 h. 1 × 10⁵ ILC2s were cytospun onto glass microscope slides. Cells were fixed in 4% formaldehyde for 10 min, permeabilized using Triton X-100 for 20 min, and washed three times using PBS, followed by a blocking step with PBS supplemented with 5% BSA for 60 min. The cells were then stained with GZMB recombinant rabbit monoclonal antibody (ThermoFisher Scientific) or perforin polyclonal antibody (ThermoFisher Scientific) diluted at 1:100, followed by counter-staining with DAPI overnight at 4°C. The cells were stained with Alexa Fluor 647-conjugated AffiniPure Goat Anti-Rabbit IgG (H+L) secondary antibody (Jackson ImmunoResearch Laboratories INC). Cells were washed, and images were acquired using a 63×/1.40 Plan-Apochromat oil immersion objective on an LSM 880 confocal microscope (Carl Zeiss AG).

ELISA

ILC2s were co-cultured with AML cells at an effector/target ratio of 2:1 for 48 h in MEM GlutaMAX media supplemented with IL-2, IL-7, and IL-15. Cell supernatants were collected and analyzed for GZMB and perforin content by ELISA according to the manufacturer’s protocols. Levels of GZMB and perforin production in culture supernatants were measured using the Human Granzyme B DuoSet ELISA Kit (Cat# DY2906–05, R&D) and Perforin Human ELISA Kit (Cat# BMS2306, ThermoFisher Scientific).

Immunoblot

ILC2s labeled with 5 mM CellTrace Violet (CTV; Thermo Fisher Scientific) were co-cultured with AML cells at a ratio of 2:1 for indicated times. AML cells or ILC2s were sorted and lysed in RIPA lysis and extraction buffer (ThermoFisher Scientific). Cell lysate in SDS loading buffer (ThermoFisher Scientific) was boiled and analyzed using 4–20% Mini-PROTEAN^® TGX^™ Precast Protein Gels (Bio-Rad). The gels were transferred to polyvinylidene fluoride (PVDF) or nitrocellulose membranes (Sigma-Aldrich) and were then blocked with 2% milk in TBST buffer (Bio-Rad) for 60 min before incubation with primary antibodies at 4 °C overnight. The membranes were washed 3 times and incubated with the appropriate fluorescent secondary antibody, Rabbit IgG HRP-conjugated Antibody or Mouse IgG HRP-conjugated Antibody for 60 min at room temperature. The immunoreactive proteins were detected using the Odyssey DLX Imaging System (LI-COR) or FluorChem E System (ProteinSimple). Antibodies used were anti-GSDME (ab215191, Abcam), anti-caspase 3 (9662S, Cell Signaling Technology), anti-GAPDH (60004–1-Ig, Proteintech), anti-β actin (66009–1-Ig, Proteintech), anti-phospho-FoxO1 (Thr24)/FoxO3a (Thr32)/FoxO4 (Thr28) (2599S, Cell Signaling Technology), anti-FoxO1 (14952S, Cell Signaling Technology), anti-phospho-Akt (Ser473) (4060S, Cell Signaling Technology), and anti-Akt (pan; 2920S, Cell Signaling Technology). For the experiments of GSDME and caspase 3 cleavage, tumor cells were co-cultured with or without Ex ILC2s at the ratio of 2:1. Six hours later, tumor cells were sorted using a BD FACSAria™ Fusion (BD Biosciences) for immunoblotting assay. For the experiments of the phosphorylation of FOXO1 and AKT, Ex ILC2s were co-cultured with or without tumor cells at the ratio of 2:1 for the indicated time. Ex ILC2s were sorted using a BD FACSAria™ Fusion for immunoblotting assay.

CRISPR–Cas9 knockout ILC2s using electroporation

The GZMB sgRNA (5’-GGCCCACAATATCAAAGAAC-3’ and 5’-GCTACCTAGCAACAAGGCCC-3’) as well as the DNAM-1 sgRNA (5’-GTTAAGAGGTCGATCTGACGGGG-3’; and 5’-CGATGACGCTCCACCTTCCGTGG-3’) were used to knock out GZMB and DNAM-1 in Ex ILC2s, respectively. Electroporation was performed after ILC2 expansion for two weeks. The Cas9-gRNA RNP complex was made by mixing 2.1 μl PBS, 1.2 μl GZMB gRNA (two sgRNA mix, 100 μM ) or DNAM-1 gRNA (two sgRNA mix, 100 μM), and 1.7 μl Cas9 Nuclease V3 (10 mg/ml, IDT, Cat# 1081059), followed by incubating the complex at room temperature for 15 min. 1 × 10⁶ ILC2s were washed with PBS two times and resuspended in 20 μl of P3 buffer according to the instruction of AmaxaTM P3 Primary Cell 4D-NucleofectorTM X Kit (Lonza, Cat# V4XP-3032), followed by mixing with the Cas9-gRNA RNP complex (5 μl). Cells in the electroporation buffer were then added and moved into electroporation cuvettes. Programme EO115 in a 4D-Nucleofector device was chosen for electroporation. After electroporation, ILC2s were immediately supplemented with a prewarmed medium, transferred out of the electroporation cuvettes, and then cultured at 37 °C with 5% CO₂.

CRISPR–Cas9 knockout AML cell lines or primary AML blasts by lentivirus

CD112 sgRNA (5’-CGAGTTTGCCACCTTCCCCA-3’ and 5’-ACCTGCGAACCACCAGAATG-3’), CD155 sgRNA (5’-CCAGCTATTCGGAGTCCAAA-3’ and 5’-CACGGAGTCGCCCAAGAAGC-3’), and GSDME sgRNA (5’-GAGTACATCGCCAAGGGTGA-3’ and 5’-AAGTTTGCAAACCACGTGAG-3’) were cloned into LentiCRISPRv2GFP (Cat# 82416, Addgene) as previously described⁷⁵. The resulting plasmids were transfected into HEK293T cells with pSPAX2 and pCMV-VSVG at a 1/1/2 ratio. Supernatants from HEK293T cells transfected with plasmids that express human CD112, CD155, and GSDME guide RNAs were collected 48 h later. The supernatants were transfected with plasmids expressing human CD112 and CD155 guide RNAs to transduce human MOLM13, U937, and THP1 cell lines, or primary AML blasts isolated from patients with AML. The supernatants were transfected with plasmids expressing GSDME guide RNAs to transduce human MOLM13. Five days later, cells were stained with CD112 and CD155 antibodies. CD112⁻CD155⁺GFP⁺, CD112⁺CD155⁻GFP⁺, and CD112⁻CD155⁻GFP⁺ cells were sorted using a BD FACSAria™ Fusion and were then cultured into 24-well plates. For GSDME knockdown MOLM13 cell screening, GFP⁺ cells were sorted for detecting GSDME expression by immunoblot.

Representative real-time cell analysis (RTCA)-based in vitro killing assay

Cell culture medium (50 μl) was added to each well of an E-plate (Cat# 300601010, Agilent). The E-plate is a standard 96-well plate with a glass-bottom coated with gold microelectrodes covering approximately 75% of the well area. The E-plate was then connected to the system to check for proper electrical contacts and to obtain background impedance readings in the absence of cells. Capan-1, MIAPaCa-2, Gli36, LN229, U251, GBM30, and A549 were used as target cells. 5,000 target cells in 50 μl of media were plated into the E-plate and cultured overnight in the RTCA system installed in the CO₂ incubator. Ex ILC2s in 100 μl media were added into the E-plate and co-cultured for at least an additional 40 h in the RTCA system. The growth and cell index of target cells were measured following the manufacturer’s instructions.

QUANTIFICATION AND STATISTICAL ANALYSIS

For continuous endpoints, Student’s t test or one-way ANOVA was used to compare two or more independent conditions. Paired t test or one-way ANOVA models with repeated measures were used to compare two or more donor-matched groups. Two-way ANOVA models with repeated measures were used to compare multiple groups on over-time repeated measures (e.g., tumor growth). For survival data, Kaplan–Meier method and log-rank tests were used to estimate survival functions and for group comparisons. All tests were two-sided. P values were adjusted for multiple comparisons by Holm-sidak’s procedure. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications⁵⁶. Data distribution was assumed to be normal, but this was not formally tested. For RNA-seq analysis, sequencing reads were trimmed from sequencing adapters using Trimmomatic⁷⁶ and poly(A) tails using FASTP⁷⁷ and mapped back to the human genome (mm10) using STAR (v. 020201)⁷⁸. The gene-level count table was created by a high-throughput sequence (HTSeq v.0.6.0)⁷⁸ and normalized by the trimmed mean of M values⁷⁹ method. General linear models based on negative binomial distributions (R package ‘EdgeR’) were used to compare gene expression levels between different treatment or response groups within each cell type. Genes with a false discovery rate-adjusted P value less than 0.05 and a fold change greater than 1.5 (upregulated) or less than 0.7 (downregulated) were considered as differentially expressed genes. Pathway analysis and gene set enrichment analysis (GSEA) were performed using the GSEA program, which runs the GSEAPreranked algorithm on a ranked list of genes⁸⁰. Data are presented as mean ± S.D. Prism software v.9 (GraphPad) was used to perform statistical analyses.

Supplementary Material

1

Data S1 Extended supporting data, related to figures 1–5 and supplementary information

3

Figure S2 ILC2s induce AML cell death in vitro and prevent tumor growth in vivo, related to Figure 2. (A) Representative flow cytometry plots (left) and statistics (right) of the percentage of ILC2s among total ILCs in the blood of patients with AML (n = 8). (B) Statistics of the percentage of ILC2s among lineage-negative cells in the blood of patients with AML (n = 8). (C) Correlation analyses on ILC2 signatures (CD117, PTGDR2, GATA3, IL9, IL13, HPGDS, S1PR1, TLE4, IL1RL1, IL17RB, and ICOS) and primary leukemia cell signatures (GFP56, AKR1C3, CD34, NGFRAP1, EMP1, SMIM24, SOCS2, CPXM1, CDK6, KIAA0125, DPYSL3, MMRN1, LAPTM4B, ARHGAP2, NYNRIN, ZBTB46, and DNMT3B) in TCGA-LAML cohort. Two-tailed Pearson correlation test was used. (D) Representative flow cytometry plots of the percentage of AML cell death when co-cultured with Ex ILC2s at various ratios. (E) Statistics of the primary AML blast cell death percentage when co-cultured with Ex ILC2s (n = 3). (F) Graphics showing the percentage of AML cell death when co-cultured with HD ILC2s (n = 4). (G-I) MOLM13 (G), U937 (H), and THP1 (I) were co-cultured with or without HD ILC2s at a single-cell level. Time-lapse microscopy images of co-cultures of Far red-labeled Ex ILC2s (magenta) with CFSE-labeled AML cells (green) cells in a medium containing DAPI (blue). Scale bar, 100 μm. (J) Survival of mice injected with or without Ex ILC2s (n = 6). (K and L) Representative flow cytometry plots and statistics of the percentages of ILC2s in the BM, spleen, and liver of mice transplanted with MOLM13 (K) or U937 (L) and Ex ILC2s, representing organ distribution of ILC2s after being transplanted; mice were sacrificed 22–24 days after transplantation (n = 3). </P/> (M) Hematoxylin and eosin-stained tissues from mice treated with or without Ex ILC2s (scale bar: 100 μm. n = 4). (N-P) Representative flow cytometry dot plots showing the percentages of human neutrophils, eosinophils, Th2 cells, and dendritic cells in the NSG mice with human AML cells (MOLM13) that were treated with or without human Ex ILC2s in the presence of human PBMCs (n = 6). (Q-U) Percentages and statistical comparison of human neutrophils, mast cells, eosinophils, Th2 cells, and dendritic cells in the mice bearing MOLM13 tumor cells after being treated with or without human Ex ILC2s in the presence of human PBMCs (n = 6). Data represent one (M-U), two (E, F, J, K, and L), or three (A, B, D, G, H, and I) independent experiments. Data are shown as mean ± S.D. and were assessed by Student’s t test (A, B, Q-U) and one-way ANOVA with repeated measures (E). Survival data are representative of one experiment and were analyzed by Kaplan–Meier survival analysis and log-rank test (L). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

5

Figure S4 ILC2s require cell-cell contact with AML cells to induce GZMB production through DNAM-1 interaction with its ligands, CD112 and CD155, related to Figure 4. (A) Representative flow cytometry plots and statistics of the percentage of AML cell death (n = 4). (B) The percentage of AML cell death was measured by luminescence-based assay (n = 4). (C) Immunoblotting measured the cleavage of GSDME and caspase 3 in MOLM13 cells (n = 2). (D) Immunoblotting measured the cleavage of caspase 3 in U937 cells (n = 2). (E) Representative histograms showing expression of NKG2D ligands (MICA, MICB, and ULBP1/2/5/6) in AML cells. (F) Representative histograms showing expression of DNAM-1 ligands (CD112 and CD155) and NKp30 ligand (B7H6) in AML cells. (G) Representative flow cytometry plots (left) and statistics (right) of the percentage of perforin produced by Ex ILC2s co-cultured with or without AML cells (n = 4). (H) Representative images (5 × magnification, scale bar, 200 μm) of AML cells (n = 8). (I) MOLM13, U937, or THP1 were co-cultured with or without Ex ILC2s in the absence or presence of NKG2D, DNAM-1, or NKp30 blockade antibodies for 48 h. Luciferase activity in the wells with tumor cells was measured with a luminescence microplate reader. Statistics of the percentage of AML cell death (n = 8). (J) Knockout of DNAM-1 on Ex ILC2s using CRISPR-Cas9 system. Representative flow cytometry plots and histograms showing the surface expression of DNAM-1 on WT ILC2s and DNAM-1 knockout ILC2s. (K) Representative flow cytometry plots (left) and statistics (right) of the percentage of AML cell death when co-cultured with or without WT ILC2s or DNAM-1 KO ILC2s (n = 4). (L) Supernatants from AML cells co-cultured with or without WT ILC2s or DNAM-1 KO ILC2s were collected and subjected to ELISA to determine levels of GZMB (n = 4). (M) Representative flow cytometry plots and statistics of the percentage of GZMB production in Ex ILC2s when co-cultured with or without primary AML blasts in the absence or presence of DNAM-1 blockade antibody (n = 4). (N) Supernatants were collected from (M) and subjected to ELISA to determine levels of GZMB (n = 4). Data represent two (E, F, G, M and N) and three (A, B, C, D, H, I, J, K, and L) independent experiments. Data are shown as mean ± S.D. and were assessed by one-way ANOVA (A, B, I, K, L, M, and N). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant.

6

Figure S5 DNAM-1-mediated inactivation of FOXO1 is required to produce GZMB in ILC2s, related to Figures 4 and 5. (A) Knockout of CD112 and CD155 on AML cells using CRISPR-Cas9 system. Representative flow cytometry plots showing expression of CD112 and CD155 on AML cells. (B-D) Ex ILC2s were co-cultured with or without WT, SKO-CD112, SKO-CD155, or DKO-CD112/CD155 MOLM13, U937, or THP1, separately, for 48 h. Representative flow cytometry plots (left) and statistics (right) of the percentage of MOLM13 (B), U937 (C), and THP1 (D) cell death (n = 8). (E) Representative flow cytometry plots (up) and statistics (down) of the percentage of GZMB⁺ ILC2s (n = 6). (F) Gene set enrichment analysis (GSEA) plots showing enrichment of selected target genes in Ex ILC2s cocultured with U937 in the presence or absence of anti-DNAM-1. The rank orders (ILC2s co-cultured with U937 in the presence of anti-DNAM-1 vs. ILC2s co-cultured with U937 in the absence of anti-DNAM-1) of all genes are shown on the x-axis (n = 3). (G) Representative flow cytometry plots of the percentage of ILC2 cell death in the absence or presence of a FOXO1 inhibitor (AS1842856, 50 nM) or an AKT inhibitor (afuresertib, 1 μM) (n = 3). (H and I) Representative flow cytometry plots (H) and statistics (I) of the percentage of MOLM13 and U937 cell death when co-cultured with or without ILC2s in the absence or presence of the FOXO1 inhibitor or the AKT inhibitor (n = 6). Data represent one (F), two (E), and three (B, C, D, G, H, and I) independent experiments. Data are shown as mean ± S.D. and were assessed by one-way ANOVA with repeated measures (B, C, D, E, and I). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant.

7

Figure S6 Human ILC2s protect against solid tumor progression, related to Figure 7. (A-C) Ex ILC2s were cultured at the indicated ratios with Capan-1, MIAPaCa-2, Gli36, LN229, U251, GBM30, or A549 for 48 h. Representative flow cytometry of the percentage of dead cells identified by Annexin V and DAPI in the cells. (D) Immunoblotting analysis of expression of GSDME in A549, Gli36, LN229, U251, GBM30, Capan-1, and MIAPaCa-2 tumor cells. (E) Ex ILC2s were cultured at an effector/target ratio of 10:1 with Capan-1, MIAPaCa-2, Gli36, LN229, U251, GBM 30, or A549 for 2h. Representative microscopy images of these tumor cells. (F) Ex ILC2s were cultured with A549, Capan-1, or GBM30 for 2 h at an E/T ratio of 10:1. Time-lapse microscopy images of co-cultures of Far red-labeled Ex ILC2s (magenta) with CFSE-labeled these solid tumor cells (green) cells in a medium containing DAPI (blue). Scale bar, 100 μm. Data are representative of three independent experiments. The red (E) and black (F) arrows indicate bulbs formed by pyroptosis from tumor cells.

8

Figure S7 Human ILC2s protect against solid tumor progression, related to Figure 7. (A) 5×10⁴ A549 cells were injected via i.v. into NSG mice, followed by weekly i.v. injection of 5×10⁶ Ex ILC2s for three weeks. Images represent BLI at the indicated time points (top) and quantification of the BLI images (bottom) up to day 17 (n = 9). (B) 10×10⁶ PBMCs were i.v. injected into NSG-SGM3 mice. One day later, 5×10⁴ Capan-1 cells were i.p. injected into the mice, followed by a separate i.v. injection and an i.p. injection of 5×10⁶ Ex ILC2s. Three days after the first injection, the mice implanted with Capan-1 cells received a second i.p. injection of 5×10⁶ Ex ILC2s. Images represent BLI at the indicated time points (top) and quantification of the BLI images (bottom) up to day 17 (n = 6 in the Capan-1 alone group; n = 5 in the Capan-1 + ILC2 group). (C) The gating strategy of M-MDSCs and G-MDSCs. (D and E) Statistics of the absolute number of M-MDSCs and G-MDSCs in the BM, spleen, liver, lung, and pancreas of NSG-SGM3 mice that were transplanted with Capan-1 cells in the presence of PBMCs and were treated with or without Ex ILC2s (n = 4). (F-J) Statistics of the percentage of human neutrophils, mast cells, eosinophils, Th2 cells, and dendritic cells in the Capan-1 bearing mice that were treated with or without human Ex ILC2s in the presence of human PBMCs (n = 4). Data are shown as mean ± S.D. and were assessed by Student’s t test (D, E, F, G, H, I, and J) or two-way ANOVA with repeated measures (A and B). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant.

9

Table S1 The levels of cytokines in the sera of mice, related to Figure 2. NSG-SGM3 mice were i.v. injected with PBMCs prior to being transplanted with human AML cells (MOLM13). The levels of cytokines in the sera of mice that were treated with or without human Ex ILC2s were measured by an antibody-based array and shown in the table.

10

Video S1, related to Figure 3. Far red-labeled 4,000 Ex ILC2s (magenta) co-cultured with CFSE-labeled 2,000 WT tumor cell lines (Green; WT MOLM13 (A), WT U937 (B), or WT THP1 (C)) in a medium containing DAPI (blue) were incubated at 37°C in a humidified atmosphere of 5% CO₂ and imaged by live cell microscopy.

11

Video S2, related to Figure 4.Far red-labeled 4,000 Ex ILC2s (magenta) co-cultured with CFSE-labeled 2,000 GSDME knockdown MOLM13 (green) in a medium containing DAPI (blue) were incubated at 37°C in a humidified atmosphere of 5% CO₂ and imaged by live cell microscopy.

12

Video S3, related to Figure 7. Far red-labeled 10,000 Ex ILC2s (magenta) co-cultured with CFSE-labeled 1,000 tumor cell lines (Green; Capan-1 (A), GBM30 (B), or A549 (C)) in a medium containing DAPI (blue) were incubated at 37°C in a humidified atmosphere of 5% CO₂ and imaged by live cell microscopy.

Figure S1. ILC2s isolated from human peripheral blood can be reliably expanded ex vivo with authenticity confirmed by full-length single-cell RNA sequencing, related to Figure 1.

Figure S1 ILC2s isolated from human peripheral blood can be reliably expanded ex vivo with authenticity confirmed by full-length single-cell RNA-sequencing, related to Figure 1. (A) Heatmap showing the transcriptional signatures of human ILC2s. The black color means 0 (n = 3). (B) Representative flow cytometry plots of the percentage of IFNγ⁺ and TNF⁺ Ex ILC2s (n = 4). (C) Representative flow cytometry plots of the percentage of IL-4, IL-5, IL-9, and IL-13 produced by freshly isolated from healthy donors (HD ILC2s) or Ex ILC2s stimulated by PMA/ionomycin (n = 4). (D and E) Representative flow cytometry plots of the percentage (D) and statistics (E) of IL-4, IL-5, IL-9, and IL-13 produced by HD ILC2s or Ex ILC2s stimulated by human IL-33 (n = 4). (F) Representative histograms (left) and statistics (right) of expression of IL-33R on the Ex ILC2s (n = 7). (G) Representative histograms and statistical analysis on expression of NKp30 in the Ex ILC2s (n = 7). (H) The precursor cells (Lin⁻ CD127⁺CD161⁺CRTH2⁻CD117⁺) were sorted from the PBMCs of healthy donors using flow cytometry and cultured on OP9 stromal cells in the presence of IL-2, IL-7, and IL-15. Graphics showing the fold change of harvested ILC2s vs. pre-seeded precursor cells (n = 9). (I) Heatmap showing expression of specific genes in seven clusters (ILC2s: Cluster 0, 1, 2, 4, and 6; ILC progenitors: Cluster 3; ILC1/NK cells: Cluster 5). Genes are shown in rows and ranked by adjusted p values < 0.05. Gene expression is color-coded on a scale based on the z-score distribution. (J-M) Dot plot displaying expression of selected and previously described genes encoding specific cell surface markers, cytokines, cytokine receptors, and transcription factors (TFs), all of which were used to annotate clusters. (N) UMAP visualization of CD5 expression. Colored cells with shades of green (top panel) and brown (bottom panel) indicate expression of CD5 transcripts in HD ILC2 and Ex ILC2s, respectively. (O) Representative flow dot plots (left) and statistics (right) of the percentage of CD5⁺ HD and Ex ILC2s (n = 7). (P) Representative flow dot plots showing the CD45RA and CD45RO surface expression on HD and Ex ILC2s. Data are representative of one (A), two (B, C, D, E, F, G, H, I, J, K, L, M, and N), and three (O and P) independent experiments. Data are shown as mean ± S.D. and were assessed by Student’s t test (E, F, G, and O) and one-way ANOVA with repeated measures (H). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. ns, not significant.

Figure S3. ILC2-secreted granzyme B induces pyroptosis or apoptosis in AML cells, related to Figure 3.

Figure S3 ILC2-secreted granzyme B induces pyroptosis or apoptosis in AML cells, related to Figure 3. (A) Representative images (5 × magnification, scale bar, 200 μm) of AML cells (n = 4). (B) Graphics showing the percentage of AML cell death (n = 4). (C) Representative flow cytometry plots of the percentage of ILC2s after co-culturing with AML cells for 2 days (n = 4). (D) Representative flow cytometry plots of the percentage of GATA3 and RORγt expressed by Ex ILC2s. (E) RT-PCR analysis of human IL25, IL33, TSLP, PGD2, and 18S mRNA expression in MOLM13, U937, and THP1 cells. (F) Caspase 3 activity in the wells with MOLM13, U937, and THP1 cells was measured using a luminescence microplate reader (n = 4). (G) Statistics of the percentages of dead cells identified by Annexin V and DAPI in MOLM13, U937, and THP1 cells (n = 4). (H) Immunoblotting measured expression of GSDME in MOLM13, U937, and THP1 cells. (I) Representative of microscopy images of GZMB and perforin in Ex ILC2s (scale bar, 100 μm, n = 3). (J) Heatmap showing the differential gene expression in ILC2s co-cultured with or without MOLM13 or U937 (n = 3). (K and L) Representative flow cytometry plots (K) and statistics (L) of the percentages of perforin in Ex ILC2s co-cultured with or without MOLM13, U937, or THP1 cells (n = 4). (M) Supernatants from AML cells co-cultured with or without Ex ILC2s were collected and subjected to ELISA to determine levels of perforin (n = 4). (N) Representative flow cytometry plots (left) and statistics (right) of the percentage of CD45RA⁺CD45RO⁻, CD45RA⁻CD45RO⁺, and CD45RA⁺CD45RO⁺ cells (n = 5). (O) Representative flow cytometry plots (left) and statistics (right) of the percentage of WT MOLM13 cell death co-cultured with Ex ILC2s in the presence or absence of zVAD-fmk or zDEVD-fmk for 24 h (n = 4). (P) Representative flow cytometry plots (left) and statistics (right) of the percentage of WT MOLM13 cell death co-cultured with Ex ILC2s in the presence or absence of necrostatin-1 (Nec-1, 20 μM) or ferrostatin-1 (Fer-1, 2 μM) inhibitors for 24 h (n = 6). (Q) Immunoblotting analysis of expression of GSDME in MOLM13 knocked out for GSDME using CRISPR-Cas9. GAPDH serves as a loading control. (R) Representative flow cytometry plots (left) and statistics (right) of the percentage of WT and GSDME KD MOLM13 cell death co-cultured with Ex ILC2s in the presence or absence of zVAD-fmk (30 μM) or zDEVD-fmk (30 μM) for 24 h (n = 8). Data represent one (E, I, J, and Q), two (A, B, C, D, H, K, L, M, and N), and three (F, G, O, P, and R) independent experiments. Data are shown as mean ± S.D. and were assessed by one-way ANOVA with repeated measures (F, G, L, M, N, O, P, and R). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. ns, not significant.

Highlights.

• Human (h)ILC2s secrete GZMB and directly lyse tumor cells via pyroptosis/apoptosis

• ILC2s fail to produce GZMB in acute myeloid leukemia (AML) patients, a mechanism of tumor immune evasion

• hILC2s can be reliably expanded ex vivo, validated as authentic hILC2s by scRNAseq

• Infusion of expanded hILC2s protects against progression of liquid and solid tumors