IL-33-activated ILC2s induce tertiary lymphoid structures in pancreatic cancer

Abstract

Tertiary lymphoid structures (TLSs) are de novo ectopic lymphoid aggregates that regulate immunity in chronically inflamed tissues, including tumours. Although TLSs form due to inflammation-triggered activation of the lymphotoxin (LT)-LTβ receptor (LTβR) pathway¹, the inflammatory signals and cells that induce TLSs remain incompletely identified. Here we show that interleukin-33 (IL-33), the alarmin released by inflamed tissues², induces TLSs. In mice, Il33 deficiency severely attenuates inflammation- and LTβR-activation-induced TLSs in models of colitis and pancreatic ductal adenocarcinoma (PDAC). In PDAC, the alarmin domain of IL-33 activates group 2 innate lymphoid cells (ILC2s) expressing LT that engage putative LTβR⁺ myeloid organizer cells to initiate tertiary lymphoneogenesis. Notably, lymphoneogenic ILC2s migrate to PDACs from the gut, can be mobilized to PDACs in different tissues and are modulated by gut microbiota. Furthermore, we detect putative lymphoneogenic ILC2s and IL-33-expressing cells within TLSs in human PDAC that correlate with improved prognosis. To harness this lymphoneogenic pathway for immunotherapy, we engineer a recombinant human IL-33 protein that expands intratumoural lymphoneogenic ILC2s and TLSs and demonstrates enhanced anti-tumour activity in PDAC mice. In summary, we identify the molecules and cells of a druggable pathway that induces inflammation-triggered TLSs. More broadly, we reveal a lymphoneogenic function for alarmins and ILC2s.

Fig. 1. IL33 expression correlates with TLS transcriptional signatures in human tumours.

a,b, Unbiased correlation of bulk tumour mRNA gene expression to TLS transcriptional signatures (TLS signatures 1 (ref. ), 2 (ref. ) and 3 (ref. )) in human PDAC (a, top). Correlation of bulk tumour IL33 mRNA expression to TLS transcriptional signatures and LTB in human PDAC (a, bottom) and human tumours (TLS signature 1; b). c,d, Immunohistochemistry (c, left) and immunofluorescence (c, right), and the immunofluorescence-quantified frequency of bulk IL-33⁺ cells (c) and IL-33⁺CD45⁺ cells (d), and association with survival (c and d) in human PDAC. The dotted red line in c shows the putative TLS. Adj. panc., adjacent pancreatic tissue; HR, hazard ratio. n represents the number of tumours (a–c) or patients (survival; c and d). High and low (in the survival plots in c and d) represent ≥ (high) or < (low) the median frequency of the cohort. Horizontal bars in dot plots indicate the median (c, d). P values were calculated using two-sided Pearson correlation (a and b), two-tailed Mann–Whitney U-test (c (left) and d (left and middle)) and two-sided log-rank test (c (right) and d (right)). For c, scale bars, 20 μm (left) and 50 μm (right). Source Data

Fig. 2. The alarmin domain of IL-33 induces TLSs in PDAC and chemical colitis.

a, Haematoxylin and eosin staining (H&E) (top) and quantification of intratumoural TLSs (bottom left) and intracolonic lymphoid aggregates (LA; bottom right) in anti-LTβR-treated PDAC (left) and DSS-colitis (right) WT and Il33^−/− mice. b, H&E (top) and immunofluorescence (IF; middle) staining, and the TLS density (H&E quantified; bottom) in vehicle (veh.) and rIL-33-treated PDAC mice. PDAC 1–6, orthotopic PDAC mice established with six PDAC cell lines. c,d, Intratumoural TLS maturation states (immunofluorescence quantified) (c), and the B cell frequency and clonality and somatic hypermutation (d) in vehicle-treated and rIL-33-treated PDAC mice. e, H&E and colonic LA quantification in vehicle-treated and rIL-33-treated DSS-colitis mice. Data were collected 2 weeks (a, left), 3 weeks (b, PDAC 3 and 4; c and d) and 3–5 weeks (b, PDAC 1, 2, 5 and 6) after tumour implantation, and 2 weeks after DSS initiation (a (right) and e), pooled from ≥2 independent experiments with n ≥ 2 mice per group with consistent results. n represents the number of individual tumours or organs from individual mice analysed separately, or B cell clones (d, right). The dotted lines in a, b and e show putative TLS/LA; the horizontal bars in a–e show the median. P values were calculated using two-way analysis of variance (ANOVA) with Tukey’s multiple-comparison test (a) and two-tailed Mann–Whitney U-test (b–e). Scale bars, 20 μm (a, b (top) and e) and 10 μm (b (bottom)). Source Data

Fig. 3. IL-33 activates lymphoneogenic ILC2s.

a, Single-cell RNA-sequencing (scRNA-seq) analysis of purified tumour and DLN ILC2s from rIL-33-treated PDAC mice. The uniform manifold approximation and projection (UMAP) representations show cells by unsupervised clusters (top left), organs (bottom left) and select genes (right). The heat map (middle) shows the top 25 genes by cluster. The heat map scale bar represents the z score. The UMAP scale bar represents gene expression. b, Gating (left) and LT expression by flow cytometry (right) and confocal imaging (bottom) in intratumoural KLRG1⁺ ILC2s from rIL-33-treated WT and Ltb^−/− PDAC mice. FMO, fluorescence minus one; MFI, mean fluorescence intensity. c,d, Gating, and the frequency and number of intratumoural (c) and gut (d) ILC2s in vehicle-treated and rIL-33-treated PDAC (c) and DSS-colitis (d) mice. e, Gating (left), ILC2 frequency (middle) and intratumoural TLS density (right) in rIL-33 and DT-treated Nmur1^icre-eGFPRosa26^LSL-DTR and littermate control Rosa26^LSL-DTR PDAC mice. f, Intratumoural TLS density (left), gating, frequency and the number of ILC2s, and the number of DLN immune cells (right) in rIL-33-treated Nmur1^icre-eGFPLtb^fl/fl and littermate control PDAC mice. g, Tumour KLRG1⁺ ILC2s were sort-purified from rIL-33-treated Il7r^cre/+Ltb^fl/fl or Il7r^cre/+Ltb^+/+ littermate PDAC mice and transferred to Il7r^cre/+Rora^fl/fl PDAC mouse recipients. The TLS density, KLRG1⁺ ILC2 frequency and number in recipient PDACs. Data were collected at 10 days (a), 2 weeks (b and d–f) and 5 weeks (c) after tumour implantation, and 2 weeks (g) after transfer, pooled from ≥3 independent experiments with n ≥ 3 mice per group with consistent results. n represents the number of individual tumours or organs from individual mice analysed separately. The horizontal bars show the median. In f, littermate controls include Ltb^fl/+ and Nmur1^icre-eGFPLtb^fl/+ combined. P values were calculated using two-tailed Mann–Whitney U-test (b, e (right), f and g) and two-way ANOVA with Holm’s test (c, d and e (left)). Scale bars, 1 μm (b (left)), 5 μm (b (all others)), 20 μm (e and f (left)) and 100 μm (f (right)). Source Data

Fig. 4. Lymphoneogenic ILC2s migrate to tumours through a microbiota- and PDAC-modulated gut–blood circuit.

a, Parabiotic mice with PDACs in recipient pancreata. Donors were treated with vehicle or rIL-33. Quantification (top) and gating (bottom) of donor-derived ILC2s in recipient blood and PDAC. b, KikGR PDAC mice were treated daily with rIL-33 for 6 days. On day 6, the gut or peritoneum was photoconverted. Experimental schematic (left) and density of gut-derived (RFP⁺) ILC2s in pancreatic PDACs (right). c, Gut Il33 mRNA was analysed by digital droplet PCR in sham-treated and PDAC mice. d, Gut Il33 mRNA (left), blood KLRG1⁺ ILC2s (middle) and the intratumoural TLS density (right) in rIL-33-treated PDAC mice treated with or without antibiotics (Abx). e, The intratumoural KLRG1⁺ ILC2 frequency, number and TLS density in skin (s.c.) PDACs in vehicle-treated and rIL-33-treated mice with s.c. alone (single PDAC) or s.c. and pancreatic PDACs (dual PDAC). f, Parabiotic mice with s.c. PDACs in recipients with or without pancreatic PDACs in donors. Donors were treated with vehicle or rIL-33. The donor-derived ILC2 frequency in recipient s.c. PDACs (bottom) is shown. g, Parabiotic mice with s.c. PDACs in recipients and pancreatic PDACs in donors. Donors were treated with rIL-33; donor and recipients were treated with antibiotics. The donor-derived ILC2 number and frequency in recipient s.c. PDACs and gut (bottom) are shown. Data were collected at 2 weeks (a, e (left), f and g), as shown in the experimental schema (b), 4 weeks (c and d) and 5 weeks (e (right)) after tumour implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n values represent the number of individual tumours or organs from individual mice analysed separately. The horizontal bars show the median. P values were calculated using two-tailed Mann–Whitney U-test (a, c, d, e (right) and g), one-way ANOVA with Kruskal–Wallis multiple-comparison test (b), two-way ANOVA with Kruskal–Wallis test (d, middle) and two-way ANOVA with Sidak’s multiple-comparison test (e (left) and f). Scale bar, 20 μm (e). Source Data

Fig. 5. An engineered human IL-33 therapeutic stimulates TLSs and anti-tumour activity in PDAC.

a, ST2 activation in HEK blue human IL-33 reporter cells with H-rIL-33, H-e-rIL-33 and H-e-rIL-33–Fc. EC₅₀, half-maximal effective concentration. b–d, The intratumoural KLRG1⁺ ILC2 frequency (b), TLS density (c) and tumour volume (d) in PDAC mice treated with vehicle, mouse rIL-33, H-rIL-33 or escalating H-e-rIL-33–Fc. Curves in a were fit by nonlinear regression. Data were collected at 3–4 weeks (b and c) after tumour implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n values represent the number of individual tumours from individual mice. The horizontal bars show the median. P values were calculated using extra-sum-of-squares F test (a) and two-way ANOVA with Tukey’s multiple-comparison correction test (b–d). Source Data

Extended Data Fig. 1. Intratumoural IL33 expression in human tumours.

a, b, Correlation of intratumoural IL33 mRNA expression to TLS transcriptional signatures (TLS signature: 1, 2, 3), and LTB mRNA expression in human PDAC cohorts (top, Memorial Sloan Kettering Cancer Center (MSK) cohort; bottom, International Cancer Genome Consortium (ICGC) cohort) and human tumours from the Cancer Genome Atlas (b). c, IL-33 immunohistochemistry in human PDACs with TLSs. Dotted red lines = putative TLS. d, Immunofluorescence images of TLSs in human PDACs. e, f, Serum IL-33 in rIL-33-treated PDAC mice (e) and in human PDAC patients (f). g, Correlation of IL-33-activated human ILC2 transcriptional signature to TLS signatures (TLS signatures: 1, 2, 3) in human tumours from the Cancer Genome Atlas. n = number of tumours (a, b, g), mice (e) or patients (f). Data collected at 4 weeks (e) after tumour implantation, pooled from 2 independent experiments with n ≥ 3 mice per group with consistent results. P values by two-sided Pearson’s correlation (a, b, g). Source Data

Extended Data Fig. 2. Disease and immune activity with IL-33 modulation in PDAC and DSS-colitis.

a, Tumour weight in anti-LTβR-treated WT and Il33^−/− PDAC mice. b, Weight loss and survival in WT and Il33^−/− DSS-colitis mice. c, Frequency of CD4⁺ and CD8⁺ T cells in PDAC mouse models with low and moderate T cell infiltration. d, H&E of TLSs in T cell low and T cell moderate tumours. Dotted white lines = putative TLS. e, f, Immunofluorescence images of intratumoural lymphoid aggregates and primary follicles (e) and TLSs with BCL6⁺ B cells (f) in rIL-33-treated PDAC mice. g, Weight loss and survival in vehicle-treated and rIL-33-treated DSS-colitis mice. Data collected 2 (a), 3 (c, d PDAC 3, 4) and 3–5 (c, d PDAC 1, 2, 5, 6, e, f) weeks after tumour implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n = individual tumours from individual mice or individual mice analysed separately. Horizontal bars = median. P values by one-way ANOVA with Tukey’s multiple comparison test (a), two-tailed Mann-Whitney U-test (c), two-way ANOVA with Sidak’s multiple comparison test (b, g, weight loss), and log rank test (b, g, survival). Source Data

Extended Data Fig. 3. IL-33 expands LT-expressing KLRG1⁺ ILC2s in PDAC mice.

a, KLRG1⁺ ILC2 gating strategy. b, Lta and Ltb mRNA expression in purified KLRG1⁺ ILC2s (digital droplet PCR – left; scRNA-seq – right). c, d, f, LT expression on KLRG1⁺ ILC2s by flow cytometry (c, rIL-33-treated PDAC mice; f, rIL-33-treated DSS-colitis mice), and confocal imaging of purified intratumoural KLRG1⁺ ILC2s (d). In d, LT expression by flow cytometry (left) and confocal imaging (right) on αCD3 and αCD28-stimulated WT and Ltb^−/− T cells (flow cytometry histograms) shown as positive and negative controls. e, Frequency of intratumoural KLRG1⁺ ILC2s in rIL-33-treated PDAC mice. PDAC 1–6 = orthotopic PDAC mice established with PDAC cell lines 1–6. g-i, scRNA-seq of 794 purified tumour ILC2s (g) and 1,668 purified tumour and draining lymph node (DLN) ILC2s (h, i) from PDAC mice. Heat map = top 25 genes for each cluster. UMAP, violin plots = ILC transcription factors (h), and markers (i) in unsupervised clusters in Fig. 3a, and ILC2 cytokines (h). Tbx21 was undetectable. j, ILC transcription factors on intratumoural KLRG1⁺ ILC2s from rIL-33-treated PDAC mice by flow cytometry. Data collected 10 days (g-j) and 2 weeks (b-f) after tumour implantation, pooled from ≥2 independent experiments with n ≥ 2 mice per group with consistent results. n = individual tumours from individual mice analysed separately. Horizontal bars = median; violin plots show the distribution with minima, maxima, and median (lines). P values by two-way ANOVA with Tukey’s multiple comparisons test (b left, e, f), Wilcoxon signed rank test (b, i violin plots), two-way ANOVA with Holm’s post-test (c), and two-way ANOVA with Sidak’s multiple comparisons test (d). Source Data

Extended Data Fig. 4. Acute and conditional ILC2 deletion does not impact intratumoural ILC1s, ILC3s or non-ILC lymphocytes.

a, NMUR1 expression on intratumoural immune cells in rIL-33-treated in Nmur1^iCre-eGFP PDAC mice. b, c, Gating and frequency of all ILCs (b) and immune cells (c) in organs of rIL-33 and diphtheria toxin (DT)-treated Nmur1^iCre-eGFPROSA26^LSL-DTR and littermate control ROSA26^LSL-DTR PDAC mice. d, Quantification of ILCs and immune cells in organs of rIL-33-treated Nmur1^iCre-eGFP Ltb^fl/fl and littermate control PDAC mice. e, Intratumoural KLRG1⁺ ILC2 gating, frequency and number (left), TLS density (middle) and immune cell frequency (right) in rIL-33-treated Il7r^CreLtb^fl/fl and littermate control IL7r^CreLtb^+/+ PDAC mice. Dotted line = putative TLS. Data collected 2 (b-e) and 4 (a) weeks after tumour implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n = individual tumours or organs from individual mice analysed separately. Horizontal bars = median. In d, littermate control = Ltb^fl/+ and Nmur1 ^iCre-eGFPLtb^fl/+ combined. P values by two-way ANOVA with Sidak’s multiple comparison test (b, e right) and two-tailed Mann Whitney test (e left, middle). Source Data

Extended Data Fig. 5. Intratumoural LTβR⁺ myeloid cells function as TLS organizers.

a, LTβR expression in intratumoural immune cells in rIL-33-treated WT PDAC mice. b, Gating and Ltbr mRNA expression by digital droplet PCR in purified intratumoural LTβR⁺ and LTβR^- myeloid cells. c-e, scRNA-seq of 8,409 LTβR⁺ and 6,220 LTβR^- intratumoural myeloid cells purified from rIL-33-treated PDAC mice. In c, UMAP = cells by unsupervised clusters. Heat map = top 50 differentially expressed genes in LTβR⁺ and LTβR^- cells. d, Intratumoural LTβR⁺ and LTβR^- myeloid cell phenotypes by scRNA-seq (left) and TLS-inducing chemokine expression by digital droplet PCR (right). UMAP = cells by LTβR expression and TLS signature genes. e, Violin plots of chemokines differentially (left) and globally (right) expressed by LTβR⁺ cells. f, Il33 mRNA expression (left) by digital droplet PCR and IL-33 protein (right) by flow cytometry in purified intratumoural myeloid cells. g, Intratumoural KLRG1⁺ ILC2 density in agonist anti-LTβR-treated WT and Il33^−/− PDAC mice. h, Experimental schema (top) and LT MFI of intratumoural KLRG1⁺ ILC2s and intratumoural TLS density (bottom) in rIL-33-treated PDAC mice co-implanted with LTβR⁺ cells purified from WT or Il33^−/− PDAC mice. Data collected 4 (a) and 2 (b-h) weeks after tumour implantation, pooled from ≥2 independent experiments with n ≥ 2 mice per group with consistent results. n = individual tumours from individual mice analysed separately. Horizontal bars = median; violin plots show the distribution with minima, maxima, and median (lines). MFI = mean fluorescence intensity. P values by one-way ANOVA test (a), two-tailed Mann-Whitney test (b, d right, f, h), Wilcoxon signed rank test (d, e violin plots), and two-way ANOVA with Sidak’s multiple comparison test (g, h LT MFI). Source Data

Extended Data Fig. 6. ILC2 gating strategy and frequency in parabiotic PDAC mice.

a, Frequency of KLRG1⁺ ILC2s in blood of rIL-33-treated PDAC mice. b-d, Experimental schema and gating strategy (b) of donor and recipient KLRG1⁺ ILC2s in recipient blood. Gating (c) and frequency of donor-derived non-ILCs (c bottom), KLRG1⁻ ILC2s (d top) and CD45⁺ cells (d bottom) in recipient blood and tumours. e, Gating in sham and photoconverted KikGR PDAC mice as in Fig. 4b; gated on KLRG1⁺ ILC2s. Data collected 7 (a-d, blood) and 14 (c, d, tumour) days after tumour implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n = individual tumours or organs from individual mice analysed separately. Horizontal bars = median. P values by two-tailed Mann-Whitney test (a, c, d). Source Data

Extended Data Fig. 7. PDACs alter diversity and composition of gut microbiota.

a, Pancreatic KLRG1⁺ ILC2 frequency in sham-treated and PDAC mice. b, Experimental schema and faecal microbiome analyses in PDAC mice by alpha diversity (Shannon index, left) and principal coordinate analysis (right). c, Experimental schema, serum IL-33, and gut KLRG1⁺ ILC2s in rIL-33-treated PDAC mice with and without antibiotic (Abx) treatment. d, Gut KLRG1⁺ ILC2s in germ-free PDAC mice with and without faecal microbial transplantation (FMT) from WT mice. Data collected 3 (a, b) and 4 (c, d) weeks after tumour implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n = individual measurements from individual mice analysed separately. Horizontal bars = median. P values by two-tailed Mann-Whitney test (a, c, d) and Welch’s t-test (b). Source Data

Extended Data Fig. 8. Molecular determinants of rIL-33-mediated dual PDAC control.

a, Tumour volume in skin (subcutaneous (s.c.)) PDACs in vehicle-treated and rIL-33-treated mice with s.c. PDAC alone (single PDAC) or s.c. and pancreatic PDACs (dual PDAC). b, Gating of ILC2s from parabiotic mice with s.c. PDACs in recipients with/without pancreatic PDACs in donors as in Fig. 4f. c, Experimental schema (left), gating (top) and quantification of donor-derived ILC2s in donor pancreatic (bottom left), recipient s.c. PDAC (bottom middle) and blood (bottom right) in parabiotic PDAC mice treated with vehicle, rIL-33 or rIL-25. d, e, Tumour size in rIL-33-treated WT, Il1rl1^−/− and Ltbr^−/− s.c. PDAC (single) and s.c. + pancreatic (dual) PDAC mice. f, g, Tumour volume in s.c. PDACs in rIL-33-treated littermate Cre-control, Il7r^Cre/+Ltb^fl/fl (f) and Il7r^Cre/+Rora^fl/fl (g) single and dual PDAC mice. h, Gating of ILC2s from parabiotic mice with s.c. PDACs in recipients and pancreatic PDACs in donors treated with or without antibiotics (Abx) as in Fig. 4g. Data collected at 5–9 days (c, blood), 2 (b, c, tumour) and 3–5 (d) weeks post-implantation, pooled from ≥2 independent experiments with n ≥ 3 mice per group with consistent results. n = individual tumours or organs from individual mice analysed separately. Horizontal bars = median. Cre-control littermates = Il7r^Cre/+Ltb^+/+ (f) and Il7r^Cre/+Rora^+/+ (g). P values by two-way ANOVA with Tukey’s multiple comparison test (a), two-tailed Mann-Whitney test (d tumour weight) and two-way ANOVA with Sidak’s multiple comparison test (all else). Source Data

Extended Data Fig. 9. Human intratumoural KLRG1⁺ ILC2 gating and phenotype.

a-c, Gating strategy (a) and marker expression (b, c) on intratumoural KLRG1⁺ ILC2s in human PDACs. n = individual patients. MFI = mean fluorescence intensity. P values by paired two-tailed Mann Whitney test (b, c). Source Data

Extended Data Fig. 10. The IL-33–ILC2–TLS pathway.

Proposed schematic of IL-33–ILC2–TLS pathway. TLS: tertiary lymphoid structure. © 2024 Memorial Sloan-Kettering Cancer Center, Memorial Hospital for Cancer and Allied Diseases, and Sloan-Kettering Institute for Cancer Research, each in New York, NY. All rights reserved. Republished with permission.