An IL-4 signalling axis in bone marrow drives pro-tumorigenic myelopoiesis

Abstract

Myeloid cells are known to suppress antitumour immunity¹. However, the molecular drivers of immunosuppressive myeloid cell states are not well defined. Here we used single-cell RNA sequencing of human and mouse non-small cell lung cancer (NSCLC) lesions, and found that in both species the type 2 cytokine interleukin-4 (IL-4) was predicted to be the primary driver of the tumour-infiltrating monocyte-derived macrophage phenotype. Using a panel of conditional knockout mice, we found that only deletion of the IL-4 receptor IL-4Rα in early myeloid progenitors in bone marrow reduced tumour burden, whereas deletion of IL-4Rα in downstream mature myeloid cells had no effect. Mechanistically, IL-4 derived from bone marrow basophils and eosinophils acted on granulocyte-monocyte progenitors to transcriptionally programme the development of immunosuppressive tumour-promoting myeloid cells. Consequentially, depletion of basophils profoundly reduced tumour burden and normalized myelopoiesis. We subsequently initiated a clinical trial of the IL-4Rα blocking antibody dupilumab^2-5 given in conjunction with PD-1/PD-L1 checkpoint blockade in patients with relapsed or refractory NSCLC who had progressed on PD-1/PD-L1 blockade alone (ClinicalTrials.gov identifier NCT05013450 ). Dupilumab supplementation reduced circulating monocytes, expanded tumour-infiltrating CD8 T cells, and in one out of six patients, drove a near-complete clinical response two months after treatment. Our study defines a central role for IL-4 in controlling immunosuppressive myelopoiesis in cancer, identifies a novel combination therapy for immune checkpoint blockade in humans, and highlights cancer as a systemic malady that requires therapeutic strategies beyond the primary disease site.

Extended Data Figure 1.. A role for myeloid IL-4Rα expression in NSCLC development.

(a-b) Tumor burden in KP-lung tumor (a) and B16 melanoma metastasis (b) bearing mice treated with isotype or αIL-4 antibodies. Images at 20X. n=4 mice per group in panel a; n=5 mice per group in panel b. Representative of two independent experiments. (c) KP lung tumor burden in Il4ra^ΔDC, Il4ra^ΔRTM, and Il4ra^ΔT mice normalized to WT littermate controls (n=8-24 mice per group). n=24, 22, 16, 9, 8, 11 mice per group. Pooled from three independent experiments. (d) Number of viable KP cells at indicated seeding values after 5 days of culture in complete media with indicated concentrations of IL-4. Two technical replicates per timepoint. Representative of two independent experiments. (e) IL-4Rα expression in indicated myeloid populations of tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice, normalized to mean MFI of Il4ra^+/+ group. n=15 Il4ra^+/+ mice and 9 Il4ra^ΔMs4a3 mice for neutrophils; 15 Il4ra^+/+ mice and 10 Il4ra^ΔMs4a3 mice for monocytes, RTMs, and DCs; 10 Il4ra^+/+ mice and 7 Il4ra^ΔMs4a3 mice for mo-macs, and 10 Il4ra^+/+ mice and 8 Il4ra^ΔMs4a3 mice for GMPs. Pooled from three independent experiments. (f) KP tumor burden in Ms4a3-Cre heterozygous mice bearing no floxed allele (n=11) compared to age-matched WT controls (n=9). Scale bar=2mm. One experiment. (g) Number of circulating blood neutrophils in naïve and KP tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice. n=3, 3, 15, 17 mice per group. Pooled from 3 independent experiments. (h) Number of lung myeloid populations in naïve and KP tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice. n=3, 3, 7, 7 mice per group. Representative of two independent experiments. Unpaired two-tailed Student’s t-test for all statistical analyses shown. Data are mean ± s.d.

Extended Data Figure 2.. Transcriptional and histological analyses of lung tumors in Il4ra^+/+ and Il4ra^ΔMs4a3 mice.

(a) Heatmap showing myeloid scRNA-seq clusters (y axis) along with cluster-defining genes (x axis). (b) Gene expression in indicated lung immune clusters of tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice (n=3 mice per group). One experiment. (c) Average number of CD4 T cells per mm² of tumor in KP lesions of Il4ra^+/+ (n=13) and Il4ra^ΔMs4a3 (n=10) mice. Representative of three independent experiments. (d) Heatmap showing scRNA-seq clusters from sorted T cells (left) and fine-clustered CD4 (middle) and CD8 (right) T cells (y axis) along with cluster-defining genes (x axis). One experiment. (e) Proportion of the “Exhausted CD8” and “Effector CD8” clusters from panel d among all lung T cells in Il4ra^+/+ and Il4ra^ΔMs4a3 mice (n=3 mice per group). One experiment. Data are mean (b), median (c), or mean ± s.d (e). Mann-Whitney test (c) or unpaired two-tailed Student’s t-test (e).

Extended Data Figure 3.. IL-4Rα expression in myeloid populations of tumor-bearing Il4raΔCx3cr1 and Il4raΔS100a8 mice compared to littermate controls.

Representative flow cytometry histograms showing IL-4Rα protein expression in indicated immune populations from tumor-bearing Il4ra^ΔCx3cr1 and Il4ra^ΔS100a8 mice along with WT littermate controls, normalized to mean MFI of WT group. n=14 Il4ra^+/+ and 10 Il4ra^ΔCx3cr1 mice; 10 Il4ra^+/+ and 10 Il4ra^ΔS100a8 mice for neutrophils; 11 Il4ra^+/+ and 10 Il4ra^ΔS100a8 mice for monocytes; mo-macs, and RTMs; 10 Il4ra^+/+ and 9 Il4ra^ΔS100a8 mice for GMPs. Pooled from three independent experiments. Grey histogram = Isotype control. Unpaired two-tailed Student’s t-test. Data are mean ± s.d.

Extended Data Figure 4.. No evidence for Th2-biased immune microenvironment in NSCLC.

(a) Expression of IL4 in matched NSCLC tumors and adjacent normal tissue from TCGA database. (b) Expression of indicated genes (y axis) across T cell clusters (x axis) from Leader et al human NSCLC scRNA-seq dataset. (c) IL-4 production after PMA/I stimulation (left) and IL4-eGFP expression in CD4 T cells (middle) and total live cells (right) from lungs of naïve and tumor-bearing mice. n=5 mice per group for IL-4^POS lung CD4 T cells, 10 naïve and 10 tumor-bearing mice for IL4-eGFP^POS CD4 T cells and total IL4-eGFP^POS cells. Representative of two independent experiments. (d) IL-4 production by lung-seeding KP cells in vivo after PMA/I stimulation. One experiment. Unpaired two-tailed Student’s t-test. Data are mean ± s.d.

Extended Data Figure 5.. IL-4 controls myelopoiesis.

(a) Top pathways enriched among DEGs from indicated lung monocyte populations in tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice. (b) Flow cytometry plots showing CD45^NEGGFP^POS cells in lung and BM of mice bearing GFP-expressing KP tumors. (c) Number of GMPs per femur of naïve mice or KP tumor-bearing mice treated with isotype or αIL-4 antibodies. n=5, 7, 7 mice per group. Representative of two independent experiments. (d) Number of Ly6c^NEG GMP, GP, and cMoP progenitors in BM of KP tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice. n=14 Il4ra^+/+ and 12 Il4ra^ΔMs4a3 mice for Ly6c^NEG GMPs and GPs and 6 Il4ra^+/+ and 7 Il4ra^ΔMs4a3 mice for cMoPs. Pooled from two independent experiments. (e) Methocult differentiation assay of Il4ra^+/+ and Il4ra^ΔMs4a3 BM. n=4 technical replicates per condition. Representative of two independent experiments. (f) Representative gating strategy for BM progenitor populations in vehicle and IL-4c treated mice. (g) Total BM cellularity from legs of WT mice treated with vehicle or IL-4c (n=10 mice per group). Pooled from two independent experiments. (h) Number of GMPs per leg of mice treated with vehicle or IgG1κ antibodies. n=5 mice per group. One experiment (i) Number of GMP per leg of Il4ra^+/+ (WT) or Il4ra^ΔMs4a3 (KO) mice treated with vehicle or IL-4c. n=3, 3, 4 mice per group. One experiment. (j) Expression of macrophage polarization markers in BMDMs differentiated under indicated conditions. Representative of two independent experiments. (k) scRNA-seq of BM myeloid cells and myeloid progenitors from naïve, IL-4c treated, and KP tumor-bearing mice. Heatmap shows indicated clusters (y axis) and cluster-defining genes (x axis). (l) Total number of genes up or downregulated in each condition relative to naïve in indicated BM scRNA-seq clusters. One experiment. Fisher’s Exact Test (a), Unpaired two-tailed Student’s t-test (c,d,g,h), or One-way ANOVA with post hoc Tukey’s test (i). Data are mean ± s.d. Lin = CD3e, B220, Ter-119, CD11b, Ly6g, NK1.1.

Extended Data Figure 6.. Dynamics of Type 2 granulocytes in BM.

(a) Flow cytometry strategy for BM populations analyzed in Fig. 3a. (b) Percentage of IL4-eGFP^POS cells among indicated BM populations in WT KP tumor-bearing mice. n=5 mice. Representative of two independent experiments. (c) Number of BM eosinophils and basophils in naïve and tumor-bearing WT mice. n=5 mice per group. Representative of two independent experiments. (d) Number of CD45-IV negative eosinophils and basophils in lungs of tumor-bearing WT mice. n=5 mice per group. Representative of two independent experiments. (e) Representative microscopy of BM of tumor-bearing mice showing colocalization of basophils (green) with hematopoietic progenitors (red) (representative of 3 mice). (f) Basophil depletion efficiency in indicated organs after using αFCER1A (left) or αCD200R3 (right) depletion strategies n=5 isotype and 5 αFCER1A-treated mice; 7 isotype and 5 αCD200R3-treated mice for BM; 6 isotype and 3 αCD200R3-treated mice for blood. Unpaired two-tailed Student’s t-test. Data are mean ± s.d.

Extended Data Figure 7.. Gating Strategies.

(a) Flow cytometry gating strategy for indicated mouse immune populations. Note “basophil alternative gating” using CD200R3 was used in experiments where mice were treated with αFCER1A antibodies. (b) CyTOF gating strategy for human whole blood.

Extended Data Figure 8.. Sorting Strategies.

(a-b) Sorting strategies for scRNA-seq experiments presented in Figure 1 (a) and Figure 2 (b).

Figure 1.. Targeted deletion of IL-4Rα in early myeloid progenitors restricts lung cancer progression.

(a) Top gene pathways enriched in NSCLC infiltrating mo-macs compared to nLung RTMs from human and mouse datasets. (b) Tumor burden in KP lung tumor-bearing Il4ra^ΔMs4a3 mice (n=27) and littermate controls (n=23). Scale bar=2mm. Pooled from four independent experiments. (c) Number of lung mo-macs and blood monocytes in naïve and tumor-bearing Il4ra^ΔMs4a3 mice and littermate controls (n=3, 3, 10, 9 mice per group for mo-macs and n=3, 3, 15, 12 mice per group for monocytes). Pooled from three independent experiments. (d) Gene expression in lung mo-mac clusters from tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice. n=3 mice per group. One experiment. (e) Quantification of T cells in lung tumors of Il4ra^+/+ (n=140 tumors from 11 mice) and Il4ra^ΔMs4a3 (n=95 tumors from 13 mice) mice. Pooled from three independent experiments. Scale bar=50μm. (f) Quantification of NK cells in lung tumors of Il4ra^+/+ (n=59 tumors from 4 mice) and Il4ra^ΔMs4a3 (n=22 tumors from 7 mice) mice. Pooled from 3 out of 4 independent experiments. Scale bar=50μm. (g) IHC co-staining of CD3e, CD8a, and GZMB in lung tumors of Il4ra^+/+ and Il4ra^ΔMs4a3 mice. (h) Quantification of CD8 T cells in lung tumors of Il4ra^+/+ (n=13) and Il4ra^ΔMs4a3 (n=10) mice. Pooled from three independent experiments. (i) Average number and percentage of GZMB^POS CD8 T cells in lung tumors of Il4ra^+/+ (n=7) and Il4ra^ΔMs4a3 (n=4) mice. Representative of three independent experiments. (j) Diagram of myelopoiesis showing activities of indicated Cre drivers. (k) Tumor burden in KP lung tumor-bearing Il4ra^ΔCx3cr1 mice and Il4ra^ΔS100a8 mice compared to littermate controls (n=13, 11, 9, 11 mice per group). Pooled from three independent experiments. Scale bar=2mm. Fisher’s exact test (a), Mann-Whitney test (b,e,f) or Unpaired two-tailed Student’s t-test (c,h,i,k). Median (b), median ± first and third quartiles (e,f), mean (d) or mean ± s.d. (c,h,i,k).

Figure 2.. Local IL-4 signaling in bone marrow fuels immunosuppressive myelopoiesis.

(a) Relative expression of statistically significant maturation-associated genes in lung monocyte populations from Il4ra^+/+ and Il4ra^ΔMs4a3 tumor-bearing mice (n=3 mice per group). (b) Number of bulk GMPs per femur in naïve and tumor-bearing Il4ra^+/+ and Il4ra^ΔMs4a3 mice (n=8, 8, 23, 23 mice per group) Pooled from three independent experiments. (c) pSTAT6 staining in indicated BM myeloid progenitors in naïve (n=5) and KP tumor-bearing (n=6) mice. Representative of two independent experiments. (d) Proportion of WT (CD45.1) and Il4ra^ΔMs4a3 (CD45.2) cells in indicated cell populations in mice reconstituted with 1:1 (WT : Il4ra^ΔMs4a3 ) BM cells 12 weeks post-transplant (n=5 mice per group). One experiment. (e) Number of indicated hematopoietic progenitor populations in BM of vehicle and IL-4c injected mice (n=10 mice per group). Pooled from two independent experiments. (f) Number of monocytes per ml of blood in vehicle and IL-4c injected mice (n=5 mice per group). Representative of two independent experiments. (g) Curated gene lists showing indicate genes in GMP, pre-monocyte, and monocyte clusters of IL-4c and KP-treated mice relative to naïve controls. Unpaired two-tailed Student’s t-test. Data are mean ± s.d. HSC-LT, Long term hematopoietic stem cell. HSC-ST, Short term hematopoietic stem cell.

Figure 3.. Type 2 granulocytes in BM upregulate IL-4 and control myeloid output in response to distal tumor cues.

(a) Proportion of indicated cell type among IL4-eGFP^POS cells in BM of tumor-bearing mice (n=4 mice). Representative of two independent experiments. (b) IL4-eGFP expression in BM eosinophils and basophils from naïve (n=10) and tumor-bearing (n=7) mice. Pooled from two independent experiments. (c) Percentage of lung parenchymal eosinophils and basophils in naïve and tumor-bearing mice, (n=5 mice per group). Representative of two independent experiments. (d) IHC for MCP-8 in indicated tissues. Representative of 12 mice. (e) Tumor burden in KP-bearing mice treated with αFCER1A or isotype antibodies (n=9 mice per group). Pooled from two independent experiments. Scale bar=2mm. (f) Tumor burden in KP-bearing mice treated with αCD200R3 (n=11) or isotype (n=12) antibodies. Pooled from two independent experiments. Scale bar=2mm. (g) Immune populations in tumor-bearing mice treated with αFCER1A or isotype antibodies. n=8 isotype and n=7 αFCER1A-treated mice for mo-macs. n=8 isotype and n=7 αFCER1A-treated mice for monocytes and Ly6c^NEG GMPs. Representative of two independent experiments. (h) Immune populations in tumor-bearing mice treated with αCD200R3 or isotype antibodies. n=11 mice per group for mo-macs. n=12 isotype and 11 αCD200R3-treated mice for monocytes. n=11 isotype and 9 αCD200R3-treated mice for Ly6c^NEG GMPs. Pooled from two independent experiments. (i) IL4-eGFP expression in BM basophils cultured in control or tumor-conditioned media. Representative of two independent experiments. (j) Expression of proteins in lung homogenate of naïve (n=5) and tumor-bearing (n=8) mice. Pooled from two independent experiments. (k) Inflammatory proteins upregulated in plasma of treatment-naïve NSCLC patients (n=29) compared to healthy controls (n=21). Analytes also upregulated in j are indicated in red. One experiment. (l) Expression of IL4-eGFP in BM basophils cultured with indicated cytokines normalized to control media. Two technical replicates per condition, representative of two independent experiments. Unpaired two-tailed Student’s t-test (b,c,g,h,j,k) or Mann-Whitney Test (e,f). Data are mean ± s.d. (a,b,c,g,h,j,k) or median (e,f).

Figure 4.. IL-4Rα blockade enhances response to immunotherapy in human NSCLC.

(a) Lung tumor burden in mice transplanted with HKP1 cells and treated with αPD-L1 antibodies, αIL-4 antibodies, or a combination of both (n=8, 8, 10, 10 mice per group). Pooled from two independent experiments. (b) Clinical trial design. (c) Averaged heatmap of Olink Inflammation Panel analytes in patients’ plasma at indicated timepoints post dupilumab treatment. (d) Levels of indicated immune cells in patient whole blood at indicated timepoints post dupilumab treatment as assessed by CyTOF, normalized to Day 1 (pre-dupilumab treatment). Dotted lines represent individual patients; solid line represents mean of all patients. (e) Number of CD8, DC-LAMP, and CD20-positve cells in tumor biopsies of relapsed/refractory NSCLC patients prior to and 36 days after initiation of dupilumab treatment, as measured by IHC (n=3 patients). Scale bar=50μm. (f) Chest CT stans of dupilumab responder patient before treatment and 56, 168, and 273 days after treatment. One way ANOVA with post hoc Tukey’s multiple comparison test (a). Panels b-f representative of one clinical cohort.