Cytotoxic innate lymphoid cells sense cancer cell-expressed interleukin-15 to suppress human and murine malignancies

Abstract

Malignancy can be suppressed by the immune system. However, the classes of immunosurveillance responses and their mode of tumor sensing remain incompletely understood. Here, we show that although clear cell renal cell carcinoma (ccRCC) was infiltrated by exhaustion-phenotype CD8⁺ T cells that negatively correlated with patient prognosis, chromophobe RCC (chRCC) had abundant infiltration of granzyme A-expressing intraepithelial type 1 innate lymphoid cells (ILC1s) that positively associated with patient survival. Interleukin-15 (IL-15) promoted ILC1 granzyme A expression and cytotoxicity, and IL-15 expression in chRCC tumor tissue positively tracked with the ILC1 response. An ILC1 gene signature also predicted survival of a subset of breast cancer patients in association with IL-15 expression. Notably, ILC1s directly interacted with cancer cells, and IL-15 produced by cancer cells supported the expansion and anti-tumor function of ILC1s in a murine breast cancer model. Thus, ILC1 sensing of cancer cell IL-15 defines an immunosurveillance mechanism of epithelial malignancies.

Extended Data Fig. 1. Cluster-defining marker plots for all clusters and heatmap of differential gene expression analysis among the three CD8⁺ T cell clusters

a, Marker plots showing normalized expression of selected common markers for lymphoid and myeloid populations (CD3D – T cells, CD4 – CD4⁺ T cells, CD8A – CD8⁺ T cells, KLRB1 – innate lymphocytes, CD14 – myeloid cells). b, Heatmap of expression of the top 30 differentially expressed genes by log fold change, [computed using Wilcox test in the FindMarkers() function of Seurat (FDR P <0.05 and LFC > 0)], across the three CD8⁺ T cell clusters within the chromophobe renal cell carcinoma (chRCC) and clear cell RCC (ccRCC) patients. Each column represents an individual cell. c, Plots showing the back gating strategy for plots shown in Fig. 1e.

Extended Data Fig. 2. Heatmap of differential gene expression analysis between clusters NK and ILC1 and comparison of CD56 expression between CD49a⁺CD103⁺ ILC1s and CD49a⁻CD103⁻ NK cells

a, Heatmap of expression of the top 30 differentially expressed genes by log fold change, [computed using Wilcox test in the FindMarkers() function of Seurat (FDR P <0.05 and LFC > 0)], across the two innate lymphocyte clusters in the chromophobe renal cell carcinoma (chRCC) and clear cell RCC (ccRCC) patients. Each column represents an individual cell. b, Plots showing the back gating strategy for plots shown in Fig. 2c. c, CD56 MFI in CD49a⁺CD103⁺ ILC1s (orange) compared to CD49a⁻CD103⁻ NK cells (green) in the indicated histology. Each pair of symbols connected by a line denotes an individual patient (chRCC n = 9, ccRCC n = 15). Paired ratio t test was used for statistical analysis, NS = non-significant, ****p < 0.0001.

Extended Data Fig. 3. Validation of the ILC1 signature

a, Table outlining the cell surface markers used to define each immune cell population sorted for downstream bulk RNA-sequencing. b, Area under the ROC curve and c, precision recall curve, when using the ILC1 signature to discriminate resident ILC1 populations (n = 12) vs all others (n = 57). The areas under the ROC and PR curves were calculated using the PRROC package in R.

Extended Data Fig. 4. CD49a⁺CD103⁺ ILC1s and CD49a⁻CD103⁻ NK cells are phenotypically distinct in terms of NKG2A and CD16 expression

a, Violin plot showing KLRC1 expression in the indicated clusters. b, (Left) Representative histograms of NKG2A expression in CD49a⁻CD103⁻ NK cells (green) and CD49a⁺CD103⁺ ILC1s (orange) from the same patient of the indicated histology. (Center) Quantification of NKG2A⁺ cells within the indicated cell type and histology. (Right) MFI of NKG2A in NKG2A⁺CD49a⁺CD103⁺ ILC1s compared to NKG2A⁺CD49a⁻CD103⁻ NK cells in tumor samples. chRCC n = 6, ccRCC n = 9. Each pair of symbols connected by a line denotes an individual patient. Two-tailed unpaired t test was used for statistical analysis of the percent NKG2A positive, and paired ratio t test was used for statistical analysis of MFI, ***p < 0.001, ****p < 0.0001. c, Violin plot showing log-normalized expression of the HLA-E gene in the TCGA ccRCC and chRCC cohorts. Two-sided Wilcoxon test was used for statistical analysis p < 2.2e⁻¹⁶. d, Correlation between level of HLA-E expression and ILC1 signature in chRCC cases from the TCGA database. Statistical analyses calculated using Spearman’s correlation. Error bands represent the 95% confidence interval. e, Association of HLA-E expression and overall survival across the TCGA chRCC cohort. High represents the top quartile and low represents the bottom 3 quartiles of IL-15 expression level. P value calculated using a Cox regression and log-rank test.

Extended Data Fig. 5. IL2RB expression in clusters NK and ILC1, IL15 expression in chRCC and ccRCC tumors from the TCGA, and IL-15 regulation of CD56 expression in CD49a⁺CD103⁺ ILC1s

a, Violin plot showing log-normalized expression of the IL2RB gene in the indicated innate lymphocyte clusters. b, Violin plot showing level of IL15 expression across chromophobe renal cell carcinoma (chRCC) and clear cell RCC (ccRCC) patients in the TCGA cohort. One-sided Wilcoxon test was used for statistical analysis, *p < 0.05. c, (Left) Representative histograms of CD56 expression in CD49a⁺CD103⁺ ILC1s treated with the indicated concentration of IL-15/IL-15Rα complex. (Right) MFI of CD56 in CD49a⁺CD103⁺ innate lymphocytes isolated from tumors treated with 100 ng/mL IL-15/IL-15Rα complex compared to 10 ng/mL IL-15/IL-15Rα complex. Each pair of symbols connected by a line denotes cells isolated from an individual patient (n = 5, 1 chRCC and 4 ccRCC), in 5 independent experiments. Paired ratio t test was used for statistical analysis, *p < 0.05. d, Representative images of chRCC tumor tissue that was stained with anti-CD103 (red), anti-E-Cadherin (white), anti-CD3 (green), and DAPI (blue). White arrows denote CD3⁻CD103⁺ innate lymphocytes. Scale bar = 20 μM. Quantification is representative of three independent chRCC patient tumor tissues, each dot represents one patient. Error bar represents mean ± SEM.

Extended Data Fig. 6. ILC1s function independently of DC- and macrophage-expressed IL-15

a, Schematic describing the IL-15^2A-eGFP reporter mouse strain. b, Types of cancer cells and stromal cells with the potential for IL-15 expression in PyMT tumors. c, Table listing the Cre recombinase lines used to delete IL-15 in the listed target cell populations. d, Schematic of an Il15 floxed allele. e, Gating strategy for determining eGFP expression in the indicated myeloid cell populations isolated from pooled tumors of a 20-week-old IL-15^2A-eGFPPyMT mouse. f, Flow cytometric analysis of eGFP expression in the indicated myeloid cell populations from pooled tumors of a 20-week-old IL-15^2A-eGFPPyMT (colored) or PyMT mouse (gray). g, qPCR analysis of Il15 mRNA expression in sorted DCs from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 3) or Itgax-CreIl15^fl/flPyMT (n = 3) mice. h, qPCR analysis of Il15 mRNA expression in sorted TAMs from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 3) or Itgax-CreIl15^fl/flPyMT (n = 3) mice. i, Representative plot and quantification of NK1.1⁺ cells out of total CD45⁺CD3⁻ cells isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 6) or Itgax-CreIl15^fl/flPyMT mice (n = 6). j, Representative plot and quantification of percentage of CD49a⁺CD103⁺ ILC1s out of total CD45⁺CD3⁻NK1.1⁺ cells isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 6) or Itgax-CreIl15^fl/flPyMT (n = 6) mice. k, Representative histogram and quantification of granzyme B (GzmB) expression in CD49a⁺CD103⁺ ILC1s from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 5) or Itgax-CreIl15^fl/flPyMT (n = 5) mice. l, (Left) Representative histogram and quantification of granzyme C (GzmC) expression in CD49a⁺CD103⁺ ILC1s from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 8) or Itgax-CreIl15^fl/flPyMT (n = 5) mice. m, Total tumor burden of Il15^fl/flPyMT (11 weeks n = 6, 20 weeks n = 11) and Itgax-CreIl15^fl/flPyMT (11 weeks n = 6, 20 weeks n = 11) mice monitored between 11 and 20 weeks of age. n, Representative histograms of CD49b (DX5) expression among total CD3⁻NK1.1⁺ cells in spleens of the indicated mouse genotype. (Right) Percentage of DX5⁺ NK cells quantified out of total splenic CD45⁺ immune cells (Il15^fl/flPyMT n = 7, Itgax-CreIl15^fl/flPyMT n = 5). o, (Left) Representative plots of CD27 and CD11b expression among total CD3⁻NK1.1⁺DX5⁺ NK cells in spleens of the indicated mouse genotype. (Right) Percentage of DX5⁺CD11b⁺CD27⁻ cells quantified out of total splenic CD45⁺ cells (Il15^fl/flPyMT n = 7, Itgax-CreIl15^fl/flPyMT n = 5). g-o, Each dot represents an individual mouse. Data are pooled from 3 or more independent experiments. All error bars represent the mean ± SEM. Two-tailed unpaired t test was used for statistical analysis, NS = non-significant, *p < 0.05, **p < 0.01, ***p < 0.001.

Extended Data Fig. 7. ILC1s function independently of hematopoietic and stromal cell sources of IL-15

a, Gating strategy and YFP expression in the indicated populations from pooled tumors of a 20-week-old S100a4-CreRosa26^LSL-YFPPyMT mouse. b, Flow cytometric analysis of eGFP expression in the indicated populations from pooled tumors of 20-week-old IL-15^2A-eGFPPyMT (colored) or control PyMT (gray) mice. c, qPCR analysis of Il15 mRNA expression in sorted live CD45⁺ immune cells from 20–24-week-old Il15^fl/flPyMT (n = 3) or S100a4-CreIl15^fl/flPyMT (n = 2) mice. d, qPCR analysis of Il15 mRNA expression in sorted live CD45⁻CD31⁻Ter119⁻CD24⁻CD29⁺EpCAM⁻ stromal cells from 20–24-week-old Il15^fl/flPyMT (n = 3) or S100a4-CreIl15^fl/flPyMT mice (n = 3). e, Representative plot and quantification of percentage of NK1.1⁺ cells out of total CD3⁻ cells isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 7) or S100a4-CreIl15^fl/flPyMT mice (n = 9). f, Representative plot and quantification of percentage of CD49a⁺CD103⁺ ILC1s out of total CD45⁺CD3⁻NK1.1⁺ cells isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 7) or S100a4-CreIl15^fl/flPyMT (n = 9) mice. g, Representative histogram and quantification of granzyme B (GzmB) expression in CD49a⁺CD103⁺ ILC1s isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 6) or S100a4-CreIl15^fl/flPyMT (n = 7) mice. h, Representative histogram and quantification of granzyme C (GzmC) expression in CD49a⁺CD103⁺ ILC1s isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 5) or S100a4-CreIl15^fl/flPyMT (n = 5) mice. i, Total tumor burden of Il15^fl/flPyMT (11 weeks n = 5, 20 weeks n = 7) and S100a4-CreIl15^fl/flPyMT (11 weeks n = 7, 20 weeks n = 8) mice monitored between 11 and 20 weeks of age. j, Representative histograms of CD49b (DX5) expression among total CD3⁻NK1.1⁺ cells in spleens of the indicated mouse genotype. (Right) Percentage of DX5⁺ NK cells quantified out of total splenic CD45⁺ cells (Il15^fl/flPyMT n = 7, S100a4-CreIl15^fl/flPyMT n = 6). k, (Left) Representative plots of CD27 and CD11b expression among total CD3⁻NK1.1⁺DX5⁺ NK cells in spleens of the indicated mouse genotype. (Right) Percentage of DX5⁺CD11b⁺CD27⁻ cells quantified out of total splenic CD45⁺ cells (Il15^fl/flPyMT n = 7, S100a4-CreIl15^fl/flPyMT n = 6). c-k, Each dot represents an individual mouse. Data are pooled from 3 or more independent experiments. All error bars represent the mean ± SEM. Two-tailed unpaired t test was used for statistical analysis, NS = non-significant, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Extended Data Fig. 8. Mouse models utilized for characterization of tissue-resident ILC1 responses in PyMT tumors

a, Diagram denoting the Gzmc gene locus of Gzmc^tdT-T2A-iCre mice. b, Flow cytometric analysis of GzmC^tdT reporter expression and granzyme C protein expression among the indicated CD3⁻NK1.1⁺ innate lymphocyte populations in PyMT tumors. c, Diagram denoting the Cdh1 gene locus of Cdh1^mCFP mice. d, Diagram denoting the Polr2a gene locus harboring an expression cassette for a GCaMP5 calcium indicator and a tdT reporter. e, Expected fluorescent phenotype that results when calcium signaling is sensed in GCaMP5-expressing innate lymphocytes.

Extended Data Fig. 9. S100a8-Cre targets cancer cells, and splenic NK cells are unaffected in S100a8-CreIl15^fl/flPyMT mice

a, Gating strategy for determining eGFP expression in CD24⁺CD29⁺EpCAM⁺ epithelial cells from non-reporter mammary gland, IL-15^2A-eGFP mammary gland, PyMT tumors, and IL-15^2A-eGFPPyMT tumors. b, Gating strategy and YFP expression in CD24⁺CD29⁺ cancer cells from pooled tumors of a 20-week-old S100a8-CreRosa26^LSL-YFPPyMT mouse. c, qPCR analysis of Il15 mRNA expression in sorted live CD45⁻CD24⁺CD29⁺EpCAM⁺ cancer cells from 20–24-week-old Il15^fl/flPyMT (n = 3) or S100a8-CreIl15^fl/flPyMT mice (n = 3). d, (Left) Representative histograms of CD49b (DX5) expression among total CD3⁻NK1.1⁺ cells in spleens of the indicated mouse genotype. (Right) Percentage of DX5⁺ NK cells quantified out of total splenic CD45⁺ cells (n = 4 for each genotype). e, (Left) Representative plots of CD27 and CD11b expression among total CD3⁻NK1.1⁺DX5⁺ cells in spleens of the indicated mouse genotype. (Right) Percentage of DX5⁺CD11b⁺CD27⁻ cells quantified out of total splenic CD45⁺ cells (n = 4 for each genotype). c-e, Each dot represents an individual mouse. Data are pooled from 3 independent experiments. All error bars represent the mean ± SEM. Two-tailed unpaired t test was used for statistical analysis, NS = non-significant.

Figure 1. chRCC and ccRCC tumor exhibit differential immune cell infiltration and CD8 T cell phenotypes

a, t-distributed Stochastic Neighbor Embedding (tSNE) embedding of transcriptional profiles from leukocytes isolated from one chromophobe renal cell carcinoma (chRCC) tumor and one clear cell RCC (ccRCC) tumor. Each dot represents a single CD45⁺ cell, and colors represent clusters denoted by cell type inferred from lineage markers and differential gene expression. b, tSNE plot as in a, colored by histology (chromophobe or clear cell). c, For each CD8 cluster, the frequency out of all CD8α⁺ clusters at which it was found in chRCC (n = 1) and ccRCC (n = 1) tumors. d, Violin plots showing log-normalized expression of selected differentially expressed genes among three CD8 clusters. e, Representative plots of flow cytometric analysis of the percentage of CD3⁺CD8α⁺ T cells out of the lymphocyte gate (CD45⁺SSC^Low) in blood, adjacent normal kidney, and tumor samples from one patient of the indicated histology. Quantification is CD3⁺CD8α⁺ T cells out of total CD45⁺ cells. f, Representative histograms of PD-1 expression in CD3⁺CD8α⁺ T cells from blood, adjacent normal kidney, and tumor tissues from a single patient of the indicated histology. Quantification of flow cytometric analysis of percentage of PD-1⁺CD3⁺CD8α⁺ T cells in blood, adjacent normal kidney, and tumor samples from the indicated histology. e, f, Each pair of symbols connected by a line denotes an individual patient [chRCC blood n = 6 (e and f), kidney n = 9 (e) and 10 (f), tumor n = 9 (e) and 10 (f); ccRCC blood n = 14, kidney n = 15, tumor n = 16]. One-way ANOVA with Tukey’s multiple comparison test was used for statistical analysis, NS = non-significant, **p < 0.01, ****p < 0.0001.

Figure 2. chRCC tumors are highly infiltrated by CD56⁺CD49a⁺CD103⁺ ILC1s

a, Violin plots showing log-normalized expression of differentially expressed genes between NK and ILC1 clusters. b, For each innate lymphocyte (KLRB1⁺) cluster, the frequency out of all KLRB1⁺ clusters at which it was found in chRCC (n = 1) and ccRCC (n = 1) tumors. c, Representative plots of flow cytometric analysis of the percentage of CD3⁻CD56⁺ innate lymphocytes out of the lymphocyte gate (CD45⁺SSC^Low) in blood, adjacent normal kidney, and tumor samples from one patient of the indicated histology. Quantification is CD3⁻CD56⁺ innate lymphocytes out of total CD45⁺ cells. d, Representative plots of flow cytometric analysis of the percentage of CD49a⁺CD103⁺ ILC1s out of the innate lymphocyte gate (CD3⁻CD56⁺) in blood, adjacent normal kidney, and tumor samples from one patient of the indicated histology. Quantification is CD49a⁺CD103⁺ ILC1s out of CD3⁻CD56⁺ innate lymphocytes. c, d, Each pair of symbols connected by a line denotes an individual patient [chRCC blood n = 6 (c and d), kidney n = 10 (c) and 9 (d), tumor n = 10 (c) and 9 (d)), ccRCC blood n = 14, kidney n = 15, tumor n = 16]. One-way ANOVA with Tukey’s multiple comparison test was used for statistical analysis. Two-tailed unpaired t test was used to analyze significance of ILC1 abundance in chRCC versus ccRCC tumors. NS = non-significant, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 3. High expression of the ILC1 signature predicts better survival of chRCC patients

a, Heatmap showing enrichment of the CD8_2 signature genes in the indicated clusters. b, Survival analysis demonstrating association of the CD8_2 signature across the TCGA ccRCC (top, 535 patients) and chRCC (bottom, 66 patients) cohorts. c, Heatmap showing enrichment of the ILC1 signature genes in the indicated clusters. d, Survival analysis demonstrating association of the ILC1 signature across the TCGA ccRCC (top, 535 patients) and chRCC (bottom, 66 patients) cohorts. b, d, High represents the top quartile of the distribution of signature scores, low represents the bottom 3 quartiles. Statistical p values calculated using a Cox regression and log-rank test.

Figure 4. IL-15 governs the cytolytic effector function of ILC1s in RCC

a, Representative plots and quantification of granzyme A (GzmA) mean fluorescence intensity (MFI) in CD49a⁻CD103⁻ NK cells and CD49a⁺CD103⁺ ILC1s from chRCC (n = 5) and ccRCC (n = 14) tumors. Each pair of symbols connected by a line denotes an individual patient. FMO = fluorescence minus one control. b, Representative plots and quantification of GzmA expression in ILC1s in adjacent normal kidney and tumor from chRCC (n = 6) and ccRCC (n = 14) patients. c, Representative plots and quantification of GzmA expression in RCC tumor ILC1s (n = 4) cultured with the indicated concentrations of IL-15/IL-15Rα complex. d, Representative still images and quantification of cytolytic activities of ILC1s cultured with the indicated concentrations of IL-15/IL-15Rα complex. Times are in format of hour:minute. Blue cells are cell trace violet (CTV)-stained K562 target cells. Red arrows indicate ILC1s and white arrows indicate dead K562 target cells. Plot is representative of 3 independent experiments using cells isolated from 3 different RCC patients. e, Representative plots and quantification of Ki67 expression in RCC tumor ILC1s (n = 4) cultured with the indicated concentrations of IL-15/IL-15Rα complex. f, Correlation between level of IL-15 expression and ILC1 signature in chRCC cases from the TCGA database. Statistical analyses calculated using Spearman’s correlation. Error bands represent the 95% confidence interval. g, Association of IL-15 expression and overall survival across the TCGA chRCC cohort. High represents the top quartile and low represents the bottom 3 quartiles of IL-15 expression level. P value calculated using a Cox regression and log-rank test. a, b, c, Paired ratio t test was used for statistical analysis. *p < 0.05. e, Two-tailed unpaired t test was used for statistical analysis, *p < 0.05.

Figure 5. ILC1s are induced in human and murine breast cancers in association with IL-15 expression in tumor

a, b, Survival analysis of ILC1 signature across the TCGA BRCA (1102 patients) cohort (a) and a subset of BRCA (231 patients) with hotspot activating mutations (p.H1047R, p.E545K, p.E542K) in PIK3CA (b). High represents the top quartile of the distribution of signature scores, low represents bottom 3 quartiles. Statistical p values calculated using Cox regression and log-rank test. c, Correlation between IL-15 expression and ILC1 signature in a cohort of BRCA patients with hotspot activating mutations in PIK3CA (231 patients). Statistical analyses calculated using Spearman’s correlation. Error bands represent the 95% confidence interval. d, tSNE embedding of transcriptional profiles from pooled tumors of a PyMT mouse. Each dot represents a single CD3⁻NK1.1⁺ cell, colors represent murine ILC1 (mILC1) and murine NK (mNK) clusters. e, GSEA of the list of DEGs between clusters mNK and mILC1 against the list of DEGs between human clusters NK and ILC1. P-value calculated using GSEA pre-ranked analysis with log fold changes as input. f, g, Violin plots showing log-normalized expression of selected differentially expressed genes between mNK and mILC1 clusters. h, A representative plot of CD49a and CD103 expression in CD3⁻NK1.1⁺ innate lymphocytes with representative plots and quantification of granzyme B (GzmB) and granzyme C (GzmC) expression within the CD49a⁻CD103⁻, CD49a⁺CD103⁻, and CD49a⁺CD103⁺ subsets. Data are pooled from 3 or more independent experiments. i, Representative plots and quantification of CD3⁻NK1.1⁺ innate lymphocyte subsets in healthy mammary glands (MG) and PyMT mammary tumors. Data are pooled from 3 or more independent experiments (MG n = 4, tumor n = 5–6). g, IL-15 ELISA on pooled MG and PyMT tumors. IL-15 quantity in ng was calculated for 1 mg of tissue. h, i, All error bars represent the mean ± SEM. One-way ANOVA with Tukey’s multiple comparison test was used for statistical analysis. j, Data representative of 3 separate experiments. Error bars represent the mean ± SEM. Two-tailed unpaired t test was used for statistical analysis. NS = non-significant, *p < 0.05, **p < 0.01.

Figure 6. ILC1s expand in transformed tissue where they are stationary but active

a, Representative immunofluorescence images of E-cadherin^CFP (white), PyMT (green), GzmC^tdT (red), and DAPI (blue) from a tumor section of a 13-week-old Gzmc^tdT-T2A-iCreCdh1^mCFPPyMT mouse. Scale bar, 150 μM. Green outline denotes transformed (PyMT⁺) area and white outline denotes non-transformed (PyMT⁻) area. Quantification is of total number of individual GzmC^tdT-positive cells in non-transformed or transformed areas, taken from tumor sections of two individual mice, each with 4–5 fields of view and a range of 1–8 distinct regions of non-transformed or transformed areas. Each dot represents discrete, individual E-Cadherin⁺ areas. Error bars represent the mean ± SEM. Two-tailed unpaired t test was used for statistical analysis, ***p < 0.001. b, Still images of live imaging time lapse of a mammary gland from a 13-week-old PyMT mouse with E-cadherin^CFP (teal), Ca²⁺ (green), and GzmC^tdT (red). Data shown is representative of two independent experiments. Scale bar, 10 μM. Circles denote cells in which we observe a calcium flux, indicated by a yellow arrow.

Figure 7. Cancer cell-expressed IL-15 promotes ILC1 responses in tumor

a, Flow cytometric analysis of eGFP expression in CD24⁺CD29⁺EpCAM⁺ epithelial cells from mammary tissues of 20-week-old IL-15^2a-eGFPPyMT (green solid), IL-15^2a-eGFP (green empty), PyMT (gray filled), or control mouse (gray empty) (n = 2 mice for each group). Numbers represent mean fluorescence intensity (MFI). Plot is representative of 2 independent experiments. Paired ratio t tests were used for statistical analysis, NS = non-significant, *p < 0.05. b, Representative plots and quantification of percentage of NK1.1⁺ cells out of total CD45⁺CD3⁻ cells isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 6) or S100a8-CreIl15^fl/flPyMT (n = 7) mice. c, Representative plots and quantification of percentage of CD49a⁺CD103⁺ ILC1s out of total CD45⁺CD3⁻NK1.1⁺ cells isolated from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 6) or S100a8-Crell15^fl/flPyMT mice (n = 7). d, A representative histogram and quantification of granzyme B (GzmB) expression in CD49a⁺CD103⁺ ILC1s from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 8) or S100a8-CreIl15^fl/flPyMT (n = 6) mice. e, A representative histogram and quantification of granzyme C (GzmC) expression in CD49a⁺CD103⁺ ILC1s from pooled tumors of 20–24-week-old Il15^fl/flPyMT (n = 6) or S100a8-CreIl15^fl/flPyMT (n = 6) mice. f, Total tumor burden of Il15^fl/flPyMT (11 weeks n = 11, 20 weeks n = 14) and S100a8-CreIl15^fl/flPyMT (11 weeks n = 13, 20 weeks n = 14) mice monitored between 11 and 20 weeks of age. Error bars represent the mean ± SEM. Two-tailed unpaired t test was used for statistical analyses, *p < 0.05. b-e, Data are pooled from 3 or more independent experiments. Each dot represents an individual mouse. All error bars represent the mean ± SEM. Two-tailed unpaired t test was used for all statistical analyses, **p < 0.01, ****p < 0.0001.