Jagged2 targeting in lung cancer activates anti-tumor immunity via Notch-induced functional reprogramming of tumor-associated macrophages

SUMMARY

Signaling through Notch receptors intrinsically regulates tumor cell development and growth. Here, we studied the role of the Notch ligand Jagged2 on immune evasion in non-small cell lung cancer (NSCLC). Higher expression of JAG2 in NSCLC negatively correlated with survival. In NSCLC pre-clinical models, deletion of Jag2, but not Jag1, in cancer cells attenuated tumor growth and activated protective anti-tumor T cell responses. Jag2^−/− lung tumors exhibited higher frequencies of macrophages that expressed immunostimulatory mediators and triggered T cell dependent anti-tumor immunity. Mechanistically, Jag2 ablation promoted Nr4a-mediated induction of Notch ligands DLL1/4 on cancer cells. DLL1/4 initiated Notch1/2 signaling in macrophages induced expression of transcription factor IRF4 and macrophage immunostimulatory functionality. IRF4 expression was required for the anti-tumor effects of Jag2 deletion in lung tumors. Antibody targeting of Jagged2 inhibited tumor growth and activated IRF4-driven macrophage-mediated anti-tumor immunity. Thus, Jagged2 orchestrates immunosuppressive systems in NSCLC that can be overcome to incite macrophage-mediated anti-tumor immunity.

Graphical Abstract

eTOC/“In Brief’ Paragraph

Signaling via Notch receptors intrinsically regulates tumor cells, promoting tumor progression. Mandula et al. demonstrate that in lung cancer, deletion of Notch ligand, Jagged2, promotes expansion of immunostimulatory macrophages and anti-tumor T cell immunity. Mechanistically, in the absence of Jagged2, the Notch ligand DLL1/4 on tumor cells rewires macrophages via Notch-mediated induction of the transcription factor IRF4.

INTRODUCTION

The immunosuppressive tumor microenvironment (TME) impairs protective immunity, thereby limiting the therapeutic efficacy of cancer immunotherapies^1,2. In cancer hosts, emergency myelopoiesis expands myeloid-derived suppressor cells (MDSC), macrophages, and dendritic cells (DC) in tumor beds and represents a primary process for the evasion of anti-tumor immunity and the asymmetrical clinical responses to cancer immunotherapy³. Paradoxically, under specific therapeutic strategies, immunoinhibitory myeloid cells can be reprogrammed into subsets that activate protective anti-tumor immunity⁴. Despite the therapeutic potential of this scenario, the cancer cell-driven programs that activate anti-tumor myeloid cells remain poorly characterized.

The Notch family of receptors mediates a conserved pathway that controls the function and differentiation of many cell types, including immune cells⁵. Mammals have four Notch receptors (Notch1-4) that are bound by five ligands from the Jagged (Jagged1 and 2) and Delta-like (DLL1, 3, and 4) families⁶. Ligand binding of Notch receptors induces a two-step proteolytic activation, leading to the release and nuclear translocation of the Notch intracellular active domain (NICD)⁷. NICD-related signaling regulates immunity through modulating T cell and myeloid cell functions and differentiation^8-10. Notch activation by DLL ligands is linked to higher anti-tumor immunity, whereas Notch priming with Jagged ligands associates with cancer-related immune evasion^10-13. Our previous report demonstrates the therapeutic impact of an anti-Jagged1/2 blockade approach to overcome immunosuppressive myeloid cell actions in cancer-bearing hosts, which relates to intrinsic Jagged ligand effects on tumor-related MDSC¹³. However, the functional role of Jagged1 and 2 in cancer cells in restricting anti-tumor immunity remains elusive.

Here, we aimed to determine the mechanistic role of Jagged ligands in cancer cells in the modulation of anti-tumor immunity. Deletion of Jagged2, but not Jagged1, in non-small cell lung cancer (NSCLC) cells attenuated tumor growth through the induction of immunostimulatory macrophages and subsequent activation of anti-tumor T cell responses. Jagged2 ablation in lung tumors resulted in DLL1 and DLL4 upregulation, which consequently primed Notch signaling in macrophages and triggered immunostimulatory actions via interferon regulatory factor 4 (IRF4). Similar anti-tumor actions were found after treatment of NSCLC tumor-bearing mice with an anti-Jagged2 antibody (αJag2) or after transgenic promotion of Notch signaling in myeloid cells. Our results identified primary mechanisms whereby Jagged2 expression in lung cancer cells promotes immune evasion and highlights the potential of Jagged2-targeting therapies in reprogramming immunosuppressive myeloid cells for the potential benefit of cancer immunotherapy.

RESULTS

Jagged2 in NSCLC cells restricts protective anti-tumor T cell immunity

The impact of the Jagged ligands in NSCLC clinical outcome and anti-tumor immunity remains unknown. Analyses of different NSCLC datasets, including The Cancer Genome Atlas (TCGA, 474 lung adenocarcinoma patients), Schabath (398 lung adenocarcinoma patients¹⁴), and ORIEN (114 NSCLC patients after ICI-treatment), indicated that patients with lung tumors expressing increased JAG2 (Jagged2-encoding gene) mRNA had reduced survival compared to counterparts with low JAG2 mRNA (Figure 1A). Conversely, this stratification of lower patient survival was not observed in lung tumors with elevated JAG1 (Jagged1-encoding gene) mRNA expression (Figure S1A). Thus, Jagged2, but not Jagged1, limits clinical outcome in human NSCLC. Next, we determined whether Jagged2 expression in human NSCLC cells (Pan-Cytokeratin⁺, PCK⁺) alters the expansion of intra-tumor CD3⁺ T cells in two replicates of a human NSCLC tissue microarray (TMA; total 160 tumors) (Table S1 and S2). In NSCLC tumors, Jagged2 was augmented in PCK⁺ regions relative to PCK⁻ areas (Figure S1B), and when stratified by JAG2 expression, a higher proportion of CD3⁺ T cells was found in PCK⁺ Jagged2^Lo (JAG2^Lo) tumors compared to PCK⁺ Jagged2^Hi (JAG2^Hi) counterparts (Figure 1B and 1C). These results suggest the primary expression of Jagged2 in NSCLC cells and the negative impact of heightened Jagged2 in lung cancer cells on intra-tumor T cell infiltration.

Figure 1. Jagged2 ablation directs anti-tumor T cell immunity in lung cancer.

(A) Probability of survival ± 95% CI in NSCLC patients (TCGA PanCancer Atlas lung adenocarcinoma (LUAD); Schabath LUAD; ORIEN NSCLC) stratified by JAG2 mRNA expression (TCGA: JAG2^Lo n = 79, JAG2^Hi n = 395; Schabath JAG2^Lo n = 55, JAG2^Hi n = 343; ORIEN JAG2^Lo n = 15, JAG2^Hi n = 99). HR: log rank hazard ratio.

(B,C) From two NSCLC TMAs, (B) representative automated multispectral images showing Pan-cytokeratin (PCK; cyan), JAG2 (yellow), CD3 (green), and DAPI, and (C) the median ± IQR of intra-tumor CD3⁺ T cells in DAPI⁺ cells within all tumors (JAG2^Lo n = 91, JAG2^Hi n = 28) or primary tumors (JAG2^Lo n = 71; JAG2^Hi n = 37). See also Table S1 and Table S2.

(D, E) (D) H&E staining and (E) median ± quartiles in lung tumor areas from mice orthotopically inj ected with Scramble or Jag2^−/−LLC (n = 6 mice/group). Triangles note total tumor regions.

(F, G) (F) H&E staining and (G) median ± quartiles in lung tumor areas from mice orthotopically injected with Scramble or Jag2^−/− KPM^Kin (n = 5 mice/group). Triangles note representative tumor regions.

(H) Tumor volume ± SEM in C57BL/6 mice bearing subcutaneous wildtype (n = 6), Scramble (n = 18) or Jag2^−/− (n = 19) LLC tumors.

(I) Tumor volume ± SEM in mice bearing subcutaneous wildtype (n = 3), Scramble (n = 7), or Jag2^−/− (n = 15) KPM^Kin tumors.

(J, K) (J) H&E staining and (K) median ± quartiles in lung tumor areas from KP (n = 14) or KPJ (n = 12) mice intratracheally challenged with Cre-expressing adenovirus. Triangles note representative tumor regions.

(L) Tumor volume ± SEM of Scramble and Jag2^−/− tumors in C57BL/6 or Rag1^−/− mice (n = 5/group).

(M) Tumor volume ± SEM in mice bearing Scramble or Jag2^−/− tumors treated with isotype (ISO; n = 10), αCD4 (n = 9;), or αCD8 (n = 9).

(N, O) (N) Immunohistochemistry images and (O) percent per area ± SEM of CD8⁺ and CD4⁺ from Scramble (22 fields) and Jag2^−/− (20 fields) LLC tumors (n = 3/group).

(P, Q) (P) Immunohistochemistry images and (Q) percent per area ± SEM of CD8⁺ and CD4⁺ from Scramble (22 fields) and Jag2^−/− (5 fields) KPM^Kin tumors.

(R, S) Proportion ± SEM of intratumor (R) CD4⁺ or CD8⁺ T cells within CD45⁺ cells and (S) CD44⁺CD69⁺ or IFNγ⁺TNFα⁺ CD8⁺ T cells from Scramble (n = 4, 8) and Jag2^−/− (n = 4, 14) LLC tumors.

Statistics were applied using log-rank Mantel-Cox test, one-way ANOVA, Student’s xt-test, or Mann-Whitney t-test *, p<0.05; **, p<0.01, ***, p<0.001. See also Figure S1.

To test whether modulation of Jagged2 intrinsically regulated tumor growth in mice, we eliminated Jag2 in lung cancer cell lines or in Cre recombinase-driven autochthonous lung tumors. Deletion of Jag2 (Jag2^−/−) was assessed in: i) Lewis lung carcinoma (LLC) cells (Figure S1C), or ii) a tumor cell line generated from spontaneously developed lung tumors induced after intranasal challenge with adenoviral-encoded Cre recombinase into Kras^LSL–G12D/+Trp53^F/FMsh2^F/F mice, which were then reconstituted with DNA repair driver Msh2 to prevent further antigenic drifting (KPM^Kin) (Figure S1D and S1E). Jag2 ablation in LLC and KPM^Kin cells lowered tumor growth when administered orthotopically or subcutaneously, as compared to controls (Figure 1D-I). Next, autochthonous lung tumors were induced via intratracheal delivery of adenoviral-encoded Cre recombinase in Kras^LSL–G12D/+Trp53^F/FJag2^F/F (KPJ) and Kras^LSL–G12D/+Trp53^F/F (KP) mice. Adenoviral-Cre driven induction of lung tumors plus Jag2 elimination in mice reduced tumor burden relative to identically treated Jagged2-competent KP controls (Figure 1J and 1K). These results support the role of Jagged2 in models of transplantable and spontaneous lung tumors.

To explore the role of protective lymphocyte activity in the delayed progression of Jag2^−/− tumors, we implanted Scramble or Jag2^−/− LLC cells into immunodeficient Rag1^−/− mice or wildtype mice followed by CD4⁺ or CD8⁺ T cell depletion. The anti-tumor effect triggered by Jagged2 deletion in LLC cells in wildtype mice was partially prevented in Rag1^−/− mice, or in Jag2^−/− LLC-bearing wildtype mice treated with antibodies against CD4 or CD8 (Figure 1L and 1M; Figure S1F). Also, opposite to the data found in immunocompetent mice (Figure 1D-G), orthotopic injections of Jagged2-deficient LLC or KPM^Kin tumors into Rag1^−/− mice showed similar lung tumor burdens compared to Scramble controls (Figure S1G-J). Furthermore, in agreement with the protective role of T cells in Jag2^−/− tumors, higher intra-tumor proportions of CD4⁺ and CD8⁺ T cells were noted in Jag2^−/−-LLC and Jag2^−/−-KPM^Kin tumors compared to Scramble counterparts (Figure 1N-R). Also, heightened frequency of recently primed CD69⁺CD44⁺CD8⁺ T cells and polyfunctional IFNγ⁺TNFα⁺CD8⁺ T cells were detected in tumors from mice bearing Jag2^−/− LLC cells relative to Scramble controls (Figure 1S). Despite small anti-tumor effects, no change in T cell infiltration was found in mice bearing Jag1^−/− LLC tumors (Figure S1K-N). Thus, these results demonstrate the anti-tumor protective action of T cells in Jag2^−/− lung tumors.

Jagged2 ablation in NSCLC cells impacts the infiltrating immune repertoire

Next, we evaluated how Jagged2 elimination in tumor cells impacted the distribution of tumor-infiltrating immune subsets using scRNA-seq. Analyses of CD45⁺ cells from Scramble and Jag2^−/− LLC tumors enabled the transcriptome-based identification of five main subsets: macrophages, neutrophils, DC, fibroblasts, and T cells (Figure 2A; Figure S2A-C). Also, higher proportion of T lymphocytes was found in Jag2^−/− tumors relative to Scramble controls, with an increased frequency of T cell-related cluster 3 (Figure 2B and 2C; Figure S2D). When compared to T cell cluster 5, which exhibited higher expression of IFN-associated transcripts, cluster 3 had increased expression of Ccl5, Cd52, and Pdcd1 (Figure S2E). Further analysis of T cell cluster 3 indicated augmented mRNA levels of Prf1, Ifng, Gzmb, and Pdcd1 (Figure 2D), aligning with an effector-like T cell phenotype. In agreement, a higher frequency of CD8⁺ CD62L⁻CD44⁺ effector memory T cells was detected in Jag2^−/− tumors relative to controls (Figure S2F). Moreover, comparing the immune infiltration clusters between Scramble and Jag2^−/− tumors, we observed a shift in the distribution of neutrophil and macrophage clusters (Figure S2G). Thus, we performed secondary clustering of the myeloid group (Table S3) and identified 11 clusters primarily represented by monocytes and macrophages in clusters 0, 1, and 2 (Figure 2E and 2F; Figure S2H). Jag2^−/− tumors showed increased expansion of the macrophage-linked cluster 2 and reduction in the macrophage clusters 0 and G1/S precursor cluster 5 relative to Scramble controls (Figure 2G). Also, in silico RNA velocity trajectory described the Jag2^−/− tumor-enriched macrophage cluster 2 as a polarized subset of the Scramble tumor-enriched macrophage cluster 0, G1/S precursor cluster 5, and monocyte cluster 1 (Figure 2H), indicating that macrophage cluster 2 in Jag2^−/− tumors potentially originated from clusters 0, 5, and 1. Additional transcriptional comparisons between the macrophage clusters 0 and 2 indicated the augmented mRNA expression of Mrc1, Maf, Ccl12, Spp1, and Fcrls in the Scramble tumor-enriched cluster 0 and elevated levels of Cxcr4, Klf4, Klf2, Nr4a1, Nr4a2, Cd83, Fosb, Nfkbia, and Cxcl2 in the Jag2^−/− tumor enriched cluster 2 (Figure 2I). Also, we validated the cluster identification of macrophage subsets by flow cytometry with a reduction in F4/80⁺CD206⁺MHC-II^Lo (cluster 0) and elevation in F4/80⁺CXCR4⁺MHC-II⁺ (cluster 2) proportions in Jag2^−/− tumors relative to Scramble counterparts (Figure S2I). To further characterize macrophage cluster 2, the top 50 differentially expressed transcripts in cluster 2 were compared to a mRNA signature associated with M1- or M2-like macrophages¹⁵. Macrophage cluster 2 from Jag2^−/− tumors showed increased transcripts linked to both M1- and M2-like macrophages (Figure S2J). Thus, Jagged2 ablation in NSCLC tumors drives the expansion of transcriptionally distinct macrophage subsets from specific precursors.

Figure 2. Jagged2 deletion in NSCLC cells reprograms intra-tumor immune repertoire.

(A) Uniform manifold approximation and projection (UMAP) embedding of CD45⁺ tumor-infiltrating leukocytes clusters from Scramble (n = 3) and Jag2^−/− (n = 2) LLC tumors.

(B, C) T lymphocyte cluster (B) UMAP and (C) proportions and cell counts.

(D) Heatmap displaying T Lymphocyte cluster genes, identified in (B, C), from Jag2^−/− relative to Scramble.

(E-H) The macrophage and neutrophil myeloid cell clusters (E) UMAP, (F) heatmap, (G) proportions and cell counts, and (H) psuedotime trajectory analysis. See also Table S3.

(I) Volcano plot comparing cluster 0 from Scramble-enriched macrophages to cluster 2 from Jag2^−/−-enriched macrophages.

See also Figure S2.

Macrophages exert anti-tumor functions in Jagged2 ablated lung tumors

To confirm the macrophage elevation revealed by scRNA-seq, we tested the infiltration of F4/80⁺ macrophages in Scramble and Jag2^−/− tumors by immunohistochemistry. Tumors from subcutaneously or orthotopically injected LLC and KPM^Kin Jag2^−/−, and autochthonous Jagged2-null KPJ tumors, exhibited a heightened frequency of macrophages compared to controls (Figure 3A; Figure S3A-D). Also, flow cytometry studies showed increased percentage of macrophages and reduction of PMN-MDSC and M-MDSC, with no changes cDC1, cDC2, or moDC, in Jag2^−/− tumors relative to controls (Figure 3B). The macrophage increase was tumor bed specific as macrophage proportions were similar in bone marrow (BM) or spleen of mice bearing control or Jag2^−/− tumors (Figure S3E). Further phenotyping of F4/80⁺ macrophages indicated an enhanced proportion of differentiated Ly6C^LoMHC-II⁺ and less differentiated Ly6C^HiMHC-II⁺ macrophages in Jag2^−/− LLC and KPM^Kin tumors, and in lungs of adenoviral-Cre exposed KPJ mice, compared to controls (Figure 3C; Figure S3F-H). Also, consistent with the expansion of macrophages, homogenates from Jag2^−/− tumors showed elevated levels of the macrophage polarizing cytokines, M-CSF and GM-CSF, and reduced levels of granulocyte driver G-CSF, relative to Scramble counterparts (Figure 3D). Concordantly, higher expression of CSF1R (M-CSF receptor) was detected in macrophages from Jag2^−/− tumors compared to controls (Figure 3E). To elucidate the persistence of these macrophages, tumor-macrophages were isolated from CD45.2⁺ mice carrying Scramble or Jag2^−/− LLC and intra-tumorally injected into CD45.1⁺ hosts bearing established wildtype LLC tumors. When compared to Scramble controls, CD45.2⁺ macrophages from Jag2^−/− tumors were found in higher proportion after transfer and showed lower apoptosis rates (Figure S3I), suggesting that Jag2^−/− tumors promote macrophages with higher expansion and survival capacity.

Figure 3. Jagged2 ablated tumors expand immunostimulatory macrophages.

(A) Immunohistochemistry images and quantification of F4/80 from mice bearing Scramble (18 fields) or Jag2^−/− (17 fields) LLC tumors (n = 3/group).

(B) Flow cytometry percentages ± SEM of conventional DC1 (cDC1), conventional DC2 (cDC2), monocytic DC (moDC), monocytic (M-MDSC) or polymorphonuclear (PMN-MDSC) MDSC subsets, and macrophages (Mac) within CD45⁺ cells in Scramble and Jag2^−/− LLC tumors (n = 7, 14/group).

(C) Flow cytometry percentages ± SEM of intratumor Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ subsets within CD45⁺ cells or macrophages from Scramble (n = 5) and Jag2^−/− (n = 6) LLC tumors.

(D) G-CSF, M-CSF, and GM-CSF from Scramble or Jag2^−/− LLC tumor homogenates (n = 5/group).

(E-H) Histogram and mean fluorescent intensity (MFI ± SEM) of (E) CSF1R, (F) IFNγ, (G) TNFα, and (H) MHC-II on macrophages from Scramble or Jag2^−/− LLC tumors (n = 6, 9/group).

(I) Proportion ± SEM of MHC-II⁺ cells in CTV-labeled Ly6C⁺ cells from Scramble tumors transferred into established Scramble or Jag2^−/− tumors (n = 4) after 24 h.

(J) eGFP MFI from tumoral total macrophages or Ly6C^HiMHC-II⁺ macrophages from eGFP⁺Scramble (n = 10) or eGFP⁺Jag2^−/− (n = 9) LLC tumors.

(K, L) (K) Histograms and (L) MFI ± SEM of tumor injected cleaved DQ-OVA in macrophages from Scramble (n = 5) or Jag2^−/− (n = 4) LLC tumors.

(M, N) F4/80⁺ cells from Scramble or Jag2^−/− LLC tumors (n = 4/group) were pulsed ex vivo with DQ-OVA and co-cultured with OT-II CD4⁺ T cells. (M) DQ-OVA MFI ± SEM and (N) OT-II CD4⁺ CD25 expression.

(O, P) (O) Tumor volume ± SEM in C57BL/6 (n = 5) or Ccr2^−/− (n = 4) mice bearing Scramble or Jag2^−/− LLC tumors and (P) intratumoral proportions ± SEM of Ly6C^HiMHC-II⁺ or Ly6C^LoMHC-II⁺ macrophages in CD45⁺ cells.

(Q, R) (Q) Tumor volume ± SEM in mice treated with isotype (ISO; n = 5) or aCSFIR (n = 5) bearing Scramble or Jag2^−/− tumors and I intratumoral proportions ± SEM of Ly6C^HiMHC-II⁺ or Ly6C^LoMHC-II⁺ macrophage subsets in total tumor-infiltrating CD45⁺ cells.

(S) Tumor volume ± SEM in C57BL/6 (n = 10, 12/group) or Rag1^−/− mice (n = 4/group) mice receiving 1:1 co-transfer of LLC tumor cells with F4/80⁺ cells isolated from LLC Scramble or LLC Jag2^−/− tumors.

Statistics were applied using one-way ANOVA or Student’s t-test, *, p<0.05; **, p<0.01, ***, p<0.001. See also Figure S3.

Next, we aimed to characterize the inflammatory phenotype of macrophages from Jag2^−/− tumors. Macrophages in Jag2^−/− tumors had higher levels of IFNγ, TNFaα, and MHC-II relative to controls (Figure 3F-H). Moreover, monocytes collected from Scramble tumors and transferred into Jag2^−/− tumors had elevated MHC-II levels compared to those injected into Scramble controls (Figure 3I). Furthermore, we implanted mice with Scramble-eGFP or Jag2^−/−-eGFP tumors and analyzed the eGFP signal in macrophages as a proxy system for engulfment of tumor materials. Bulk and Ly6C^HiMHC-II⁺ macrophages from Jag2^−/−-eGFP tumors showed higher eGFP relative to controls (Figure 3J). Additionally, we tested antigen processing via intra-tumoral administered DQ-OVA¹⁶ and noted that macrophages from Jag2^−/− tumors exhibited higher DQ-OVA cleavage relative to Scramble controls, specifically in the immature Ly6C^HiMHC-II⁺ subset (Figure 3K and 3L). In agreement, macrophages from Jag2^−/− tumors pulsed ex vivo with DQ-OVA showed augmented antigen processing relative to macrophages from Scramble tumors (Figure 3M), which correlated with a higher activation of OVA-specific CD4⁺ T cells (OT-II) as assessed by CD25 detection (Figure 3N). Overall, these results indicate the accumulation of macrophages with elevated proinflammatory potential and augmented capacity to engulf, process, and present tumor products in Jag2^−/− tumors. Also, macrophages from Jag2^−/− tumors showed similar direct tumoricidal abilities compared to those from controls (Figure S3J), ruling out elevated direct anti-tumor actions.

Thereafter, we elaborated upon the contribution of macrophages in the delayed growth of Jag2^−/− tumors by injecting tumors into Ccr2⁻ mice, or treating tumor-bearing mice with either a CSF1R blocking antibody or a CCR2 antagonist^17-20. Ablation of CCR2 or CSF1R restored the progression of Jag2^−/− tumors in mice, while modestly impacting growth of Scramble controls, which correlated with a lower rate of intra-tumor Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages (Figure 3O-R; Figure S3K and S3L). Concordantly, a partial restoration of Jag2^−/− tumor growth was found after depletion of macrophages via clodronate-containing liposomes relative to empty liposome controls (Figure S3M). We next evaluated whether macrophages from Jag2^−/− tumors limit tumor growth via lymphocyte activation. Macrophages from Jag2^−/− tumors co-injected with LLC tumor cells delayed tumor growth in immunocompetent mice, but not in immunodeficient Rag1^−/− mice, whereas macrophages from Scramble controls failed to alter tumor growth (Figure 3S). Together, these results indicate the role of CSF1R- and CCR2-dependent macrophages in the anti-tumor responses found in Jag2^−/− lung tumors.

Jagged2 ablation in lung tumor cells promotes DLL1/4

To assess whether structural or soluble mediators in Jag2^−/− tumor cells promote the expansion of macrophages, BM-derived monocytes were co-cultured with irradiated Jag2^−/− or Scramble LLC cells with and without trans-wells; or in media supplemented with tissue culture supernatants (TCS) or tumor explant supernatants (TES) from Jag2^−/− or Scramble LLC tumors. Direct co-culture with irradiated Jag2^−/− LLC cells dramatically augmented the differentiation of monocytes into macrophages relative to counterparts co-cultured with irradiated Scramble LLC cells, an effect absent in trans-well co-cultures (Figure 4A and 4B). Also, irradiated Jag2^−/− LLC cells improved the DQ-OVA processing of co-cultured macrophages compared to Scramble counterparts (Figure 4C). Furthermore, a slight increase in the differentiation of monocytes into macrophages was noted after treatment with TCS or TES from Jag2^−/− tumors compared to Scramble controls (Figure 4D), suggesting a minor role of tumor soluble components. However, Jag2^−/− tumors drive macrophage differentiation and function primarily through cell: cell interactions.

Figure 4. Jagged2 null tumor cells provoke contact dependent expansion of macrophages via DLL1 and DLL4 induction.

(A,B) Fold change in the proportion ± SEM of BM-derived macrophages (BM-Mac) in CD45⁺ from BM-derived monocytes (BM-Mono) co-cultured with irradiated Scramble or Jag2^−/− LLC tumor cells at varying ratios (A) without or (B) with transwells (2 independent experiments each).

(C) Mean fluorescent intensity (MFI) ± SEM of cleaved DQ-OVA from BM-Mac post-coculture with irradiated Scramble or Jag2^−/− LLC cells or BM-Mac cultured alone (2 experiments).

(D) Fold change in the proportion ± SD of BM-Mac in CD45⁺ cells from BM-Mono cultures supplemented with TCS or TES from Scramble or Jag2^−/− LLC.

(E, F) I Volcano plot indicating bulk RNAseq transcripts differentially expressed in Scramble LLC-eGFP versus Jag2^−/− LLC-eGFP from mice (n = 3 replicates/group) and (F) GSEA and enrichment plots for differentially expressed genes. See alsoTable S4 and Table S5.

(G) MFI ± SEM of DLL1 and DLL4 on Scramble or Jag2^−/− LLC tumor cells (n = 4 mice/group).

(H-K) (H) Tumor volume ± SEM from Scramble (n = 9), Jag2^−/−x (n = 10), Dll1^−/−Dll4^−/− (n = 9), and Jag2^−/−Dll1^−/−Dll4^−/− (n = 9) LLC tumors and (I) the intratumoral proportions ± SEM of total macrophages in tumor-infiltrating CD45⁺ cells. After intratumoral DQ-OVA injection, macrophages were measured for (J) DQ-OVA MFI ± SEM (n = 3/group) then isolated and co-cultured for (K) proliferation of OT-II CD4⁺ T cells.

(L-N) (L) Representative automated multispectral images of a NSCLC TMA showing Pan-cytokeratin (PCK; cyan), JAG2 (red), CD3 (orange), CD8 (magenta), DLL1 (yellow), DLL4 (green), and DAPI and the median ± quartiles of intra-tumor CD8⁺ T cells in DAPI⁺ cells stratified by PCK⁺ region expression of (M) DLL1^LoJAG2^Hi (n = 16) or DLL1^HiJAG2^Lo (n = 48) or (N) DLL4^LoJAG2^Hi (n = 18) or DLL4^HiJAG2^Lo (n = 45).

(O) Immunoblot of Nor1 and Nurr7 in CD45 Scramble or Jag2^−/− LLC tumor cells.

(P) Relative fold change in DLL1 and DLL4 MFI ± SEM in Jag2^−/− tumors transduced with a non-targeting (control) siRNA, Nor1 siRNA and Nur77 siRNA alone or both, and treated with LPS and IFNγ or LLC TES (3 independent experiments).

(Q) Fold change in the mean ± SEM relative luciferase units (RLU) of the Nr4a-reporter in Scramble or Jag2^−/− LLC treated with vehicle, LPS and IFNγ, or LLC TES (3 experiments).

(R) ChIP of Dll1 and Dll4 promoter pulldown with IgG or Nor1 in Scramble or Jag2^−/− LLC cells. Mean of 2 biological replicates.

Statistics were applied using one-way ANOVA, Student’s t-test, or Welch’s t-test *, p<0.05; **, p<0.01, ***, p<0.001. See also Figure S4

To investigate the crosstalk between transcriptional changes in Jag2^−/− tumors with their ability to promote immunostimulatory macrophages, we compared RNA-seq profiles of Jag2^−/−-eGFP and Scramble-eGFP tumor cells collected from mice. Differential transcriptional profile expression was found in Jag2^−/− tumors relative to controls (Figure 4E; Table S4). Also, higher Notch target systems, TNFα, IFNγ, CXCR4, and IL-12 mRNA gene set enrichment analysis (GSEA) related pathways were found in Jag2^−/− tumor cells (Figure S4A and S4B; Table S5). Notably, increased Notch, DLL1, and DLL4 GSEA pathways, and higher DLL1 and DLL4 expression were noticed in Jag2^−/−-eGFP tumors compared to controls (Figure 4F and 4G; Figure S4C). In agreement, Jag2^−/− tumor cells exposed in vitro to inflammatory stimuli increased DLL1 and DLL4 expression relative to controls (Figure S4D and S4E). To interrogate whether the induced DLL1 and DLL4 on Jag2^−/− tumors contributed to the reduced tumor burden in mice, we treated Scramble or Jag2^−/− LLC-bearing mice with DLL1 and/or DLL4 blocking antibodies. DLL1 and/or DLL4 blockade restored Jag2^−/− tumor growth in mice without altering progression of Scramble tumors (Figure S4F-H). Next, we ablated DLL1 or DLL4 and found that Dll4 mRNA was elevated in Dll1^−/− LLC cells, while Dll1 mRNA increased in Dll4^−/− LLC cells (Figure S4I, J), indicating the compensatory crosstalk among DLL ligands after genetic deletion. Elimination of DLL1 and DLL4 in Jag2^−/− LLC cells partially restored tumor growth and blunted macrophage expansion relative to controls (Figure 4H and 4I). Additionally, DQ-OVA processing was reduced in macrophages from Jag2^−/−Dll1^−/−Dll4^−/− tumors compared to Jag2^−/− counterparts (Figure 4J), which correlated with a decreased intrinsic capacity to promote proliferation of OT-II CD4⁺ T cells (Figure 4K). Collectively, these findings indicate that Jagged2 ablation triggers DLL1/DLL4 on tumor cells to provoke macrophage expansion and anti-tumor effects.

To determine the impact of DLL1 and DLL4 elevation with Jagged2 reduction in a human context, we utilized a human NSCLC TMA and stratified the PCK⁺ areas into high DLL1 and DLL4 with low Jagged2 (Figure 4L, left) and low DLL1 and DLL4 with high Jagged2 (Figure 4L, right). PCK⁺ tumor regions with low Jagged2 and high DLL1 or DLL4 had an elevated frequency of CD8⁺ T cells (Figure 4M, N), indicating that in settings of low Jagged2 levels, DLL1 and DLL4 expression positively correlates with CD8⁺ T cell expansion in human NSCLC tumors.

Next, we investigated the mechanistic drivers regulating the induction of DLL1 and DLL4 in Jag2^−/− tumors. The Nr4a transcription factor family members, Nr4a3 (Nor1) and Nr4a1 (Nur77), known mediators of Jagged signaling²¹, were elevated in Jag2^−/− tumors compared to controls (Figure 4E, O; Figure S4K). Ablation of Nr4a1 and/or Nr4a3 blunted DLL1 and DLL4 induction in Jag2^−/− tumors cells compared to controls (Figure 4P). Also, Nur77 overexpression (Nur77^OE) induced DLL1 and DLL4 in Scramble and Jag2^−/− tumors (Figure S4L). Next, we tested the transcriptional activity of the Nr4a family using a Nr4a response element-luciferase reporter²². Upon treatment with inflammatory stimuli, Jag2^−/− tumor cells showed higher Nr4a-driven luciferase activity than identically treated Scramble controls (Figure 4Q). To directly assess the transcriptional regulation of DLL1 and DLL4 by Nor1 and Nur77, we conducted chromatin immunoprecipitation (ChIP) assays. In Jag2^−/− tumor cells, Nor1 bound more to the Dll1 and Dll4 promoters, while Nur77 bound at levels similar to Scramble controls (Figure 4R; Figure 4SM). Furthermore, elevated NR4A1 and/or NR4A3 mRNA expression correlated with augmented DLL1 and DLL4 mRNA levels in human lung tumors (Figure S4N-P). Collectively, these results suggest that Jagged2 ablation in NSCLC cells directs DLL1/DLL4 expression via Nr4a family members.

Notch signaling regulates macrophage expansion in Jagged2 ablated tumors

Given that DLL1 and DLL4 function as Notch ligands, we investigated Notch in macrophages from Jag2^−/− tumors. At the mRNA and protein levels, macrophages from Jag2^−/− tumors showed dramatic increased expression of full-length Notch2, slight Notch1 elevation, lower Notch3 levels, and no Notch 4 changes (Figure 5A-C; Figure S5A and S5B). Also, higher Notch2 was noted in macrophages from lungs of orthotopic Jag2^−/− LLC and autochthonous KPJ mice, relative to Scramble or flox controls (Figure S5C and S5D). Similarly, Notch2 increased in Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages from Jag2^−/− tumors, or when co-cultured with irradiated Jag2^−/− LLC cells (Figure 5D and 5E). However, Notch1 was only slightly augmented in Ly6C^HiMHC-II⁺ macrophages from Jag2^−/− tumors, but not in macrophages co-cultured with irradiated Jag2^−/− tumors (Figure 5D and E). Compared to controls, macrophages from Jag2^−/− tumors had higher expression of cleaved Notch2 with corresponding upregulation of the Notch induced transcript Hes1 (Figure 5F, G). In agreement, higher NICD2 levels were detected in bulk, Ly6C^HiMHC-II⁺, and Ly6C^LoMHC-II⁺ macrophages from Jag2^−/− tumors compared to controls (Figure 5H; Figure S5E). To study the preferential Notch activation in the macrophage subset expanding in Jag2^−/− tumors, we compared Scramble tumor-enriched cluster 0 to Jag2^−/− tumor-expanded cluster 2 and found heightened expression of multiple Notch-driven transcripts (Figure 5i) with an increased proportion of cells expressing Notch2, Hes1, Rbpj, and Rras (Figure 5J). Overall, these results indicate the elevated Notch signaling in macrophages from Jag2^−/− tumors.

Figure 5. Jagged2 deficient tumors direct Notch-mediated programming of macrophages.

(A) Relative fold change of Notch1 and Notch2 mRNA expression in tumor-infiltrating F4/80⁺ cells from Scramble (n = 9) and Jag2^−/− (n = 8 or 11) LLC tumors.

(B, C) Histogram and mean fluorescent intensity quantification (MFI) ± SEM of Notch1 (B) and Notch2 (C) in macrophages from Scramble (n = 4, 5) or Jag2^−/− (n = 5) LLC tumors.

(B) MFI ± SEM of Notch2 and Notch1 on Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages from Scramble ox Jag2^−/− LLC tumors (n = 5/group).

(C) Notch2 and Notch1 MFI ± SEM on Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages from co-cultures with irradiated Scramble or Jag2^−/− LLC cells (2 experiments).

(D) Immunoblot of cleaved Notch2 in tumor-infiltrating CD11b⁺ cells from Scramble or Jag2^−/− LLC tumors. Representative of 3 replicates from 2 distinct experiments.

(E) Relative fold change ± SEM of Hes1 mRNA in tumor-infiltrating F4/80⁺ cells from Scramble (n = 7) and Jag2^−/− (n = 6) LLC tumors.

(F) MFI ± SEM of cleaved Notch2 (NICD2) in macrophage subsets from mice bearing Scramble or Jag2^−/− LLC tumors (n = 9/group).

(I, J) The (I) relative expression and (J) distribution of genes related to Notch signaling between Scramble-enriched macrophage cluster 0 and Jag2^−/−-enriched macrophage cluster 2.

(K) Proportion of BM-Mac in CD45⁺ cells from NICD inhibitor pre-treated BM-Mono then co-cultured with irradiated Scramble or Jag2^−/− LLC cells. Mean ± SEM (2 experiments).

(L-O) (L) LLC tumor volume ± SEM in RosaSoichy^NotchIC (n = 5) or Rosa^NotchICLysM^Cre/+− (n = 14) mice and intratumoral proportions ± SEM of (M) Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages in CD45⁺ cells, (N) CD4⁺ and CD8⁺ T cells in CD45⁺ cells, and (O) CD44⁺CD69⁺ CD4⁺ or CD8⁺ T cells in CD45⁺ cells.

(P) Notch2 MFI on F4/80⁺ macrophages from Scramble (n = 4), Jag2^−/− (n = 3), Dll1^−/−Dll4^−/− (n = 4) or Jag2^−/−Dll1^−/−Dll4^−/− (n = 4) LLC tumors.

Statistics were applied using one-way ANOVA or Student’s t-test, *, p<0.05; **, p<0.01, ***, p<0.001. See also Figure S5.

To ascertain the requirement of Notch signaling in macrophage differentiation, monocytes were pre-treated with a NICD inhibitor and co-cultured with Jag2^−/− tumor cells. NICD inhibition blunted macrophage expansion in monocytes co-cultured with irradiated Jag2^−/− tumors (Figure 5K). To test the impact of Notch signaling in macrophage anti-tumor functionalities, we crossed floxed Rosa-driven NICD-GFP mice²³ with mice carrying Lysozyme 2-driven Cre recombinase (Rosa^NotchICLysM^Cre+/−) to activate Notch signaling (NICD-GFP) in the myeloid compartment. Rosa^NotchICLysM^Cre+/− mice showed delayed LLC and KPM^Kin tumor growth relative to controls (Figure 5L; Figure S5F). Furthermore, heightened intra-tumoral proportions of Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages, CD8⁺ and CD4⁺ T cells, and recently primed CD69⁺CD44⁺CD8⁺ or CD4⁺ T cells were detected in Rosa^NotchICLysM^Cre+/− mice compared to tumors from Rosa^NotchIC controls (Figure 5M-O). Moreover, macrophages from Jag2^−/−Dll1^−/−Dll4^−/− tumors had lower Notch2 compared to macrophages from Scramble or Jag2^−/− tumors, suggesting a role of DLL land DLL4 on Jag2^−/− tumors in macrophage Notch2 induction (Figure 5P). Altogether, these results indicate the primary role of Notch signaling in myeloid cells in the expansion of anti-tumor macrophages in Jag2^−/− tumors.

DLL1/4-directed Notch signaling induces IRF4-driven immunostimulatory macrophages

IRF4 promotes macrophage differentiation after inflammatory challenges^24-26. We detected higher mRNA levels for Irf4, and lower Irf8, in macrophages from Jag2^−/− tumors relative to controls (Figure S6A), which translated into increased IRF4 protein (Figure S6B and S6C). To assess the IRF4 expression in the macrophage subsets, and the potential upstream role of the induced DLL1 and DLL4 in Jag2^−/− tumors in the expression of IRF4 in macrophages, we injected mice with Scramble and Jag2^−/− tumors ablated or not for DLL1 and DLL4. Similar to the IRF4 increase noted in bulk macrophages (Figure 6A and 6B), higher IRF4 was found in Ly6C^HiMHC-II⁺ and Ly6C^LoMHC-II⁺ macrophages from Jag2^−/− tumors compared to Scramble controls, which was reduced in macrophages from DLL1 and DLL4 deficient tumors (Figure 6C). Similarly, the elevated IRF4 noted in monocytes co-cultured with irradiated Jag2^−/− LLC cells was blunted after DLL1 and DLL4 ablation on tumor cells (Figure 6D). Thus, these results suggest the role of DLL1 and DLL4 in Jag2^−/− tumors as drivers of IRF4 expression in macrophages.

Figure 6. Notch2 directs IRF4-dependent anti-tumor macrophage polarization.

(A, B) (A) Relative fold change of Irf4 mRNA expression and (B) immunoblot of IRF4 in F4/80⁺ cells from Scramble, Jag2^−/−, Dll1^−/−Dll4^−/−, and Jag2^−/−Dll1^−/−Dll4^−/− LLC tumors.

(C) Mean fluorescent intensity (MFI) ± SEM of IRF4 in total macrophages or Ly6C^HiMHC-II⁺ or Ly6C^LoMHC-II⁺ subsets from Scramble, Jag2^−/−, Dll1^−/−Dll4^−/−; or Jag2^−/−Dll1^−/−Dll4^−/− LLC tumors (n = 4/group).

(D) MFI ± SEM of IRF4 in total macrophages co-cultured with irradiated Scramble, Jag2^−/−, Dll1^−/−Dll4^−/−, or Jag2^−/−Dll1^−/−Dll4^−/− LLC tumors cells (2 experiments).

(E) ChIP of Irf4 and Hes1 promoter pulldown with IgG, Notch1, or Notch2 antibodies in F4/80⁺ cells from Scramble or Jag2^−/− LLC tumors. Mean of 3 independent experiments with 3 biological replicates.

(F) MFI ± SEM of IRF4 in BM-Mac from NICD inhibitor pre-treated BM-Mono then co-cultured with irradiated Scramble or Jag2^−/− LLC cells. Mean ± SEM (2 experiments).

(G) MFI ± SEM of IRF4 in GFP⁺ (NICD⁺) macrophages from Rosa^NotchICLysM^Cre+/− mice and total macrophages from Rosa^NotchIC bearing wildtype LLC tumors (n = 5/group).

(H, I) Spleen macrophages (sMac), (H) proportions and (I) cleaved DQ-OVA MFI ± SEM, derived from splenic monocytes (sMono) of Irf4^F/F or Irf4^F/FLysM^Cre+/− mice (n = 4/group) co-cultured with irradiated Scramble or Jag2^−/− LLC tumor cells.

(J, K) (J) Tumor volume ± SEM of Scramble or Jag2^−/− LLC tumors implanted in Irf4^F/F (n = 18, 21) or Irf4^F/FLysM^F/F (n = 14, 18) mice and (K) intratumor proportions of Ly6C^LoMHC-II⁺ macrophages in total macrophages or tumor-infiltrating CD45⁺ cells.

Statistics were applied using one-way ANOVA or Student’s t-test, *, p<0.05; **, p<0.01, ***, p<0.001. See also Figure S6.

To explore whether Notch signaling directly stimulates IRF4 in macrophages, consensus predicted NICD-binding sequences were identified in the Irf4 promoter. ChIP assays showed higher Notch2, but not Notch1, binding to the Irf4 promoter and the canonical NICD-binding region on the Hes1 promoter²⁷ in macrophages from Jag2^−/− tumors compared to controls (Figure 6E). In agreement with the upstream role of Notch signaling in the induction of IRF4 in macrophages by Jag2^−/− tumors, NICD inhibitor pre-treatment ablated IRF4 expression in monocytes co-cultured with irradiated Jag2^−/− LLC cells, compared to co-cultured vehicle-treated counterparts (Figure 6F). Moreover, NICD-expressing macrophages from LLC or KPM^Kin tumors previously injected into Rosa^NotchICLysM^Cre+/− mice showed heightened IRF4, as compared to macrophages from controls (Figure 6G; Figure S6D). Thus, these results support the upstream role of Notch2 signaling in the induction of IRF4 in macrophages from Jag2^−/− lung tumors.

Next, we tested the role of IRF4 in the expansion of macrophages occurring in Jag2^−/− tumors. Reduced differentiation into macrophages and lowered DQ-OVA processing were noted after co-culturing Irf4^F/FLysM^Cre+/− monocytes with irradiated Jag2^−/− tumors, compared to identically cultured Irf4^F/F counterparts (Figure 6H, I). Additionally, restored tumor growth and impaired expansion of Ly6C^LoMHC-II⁺ macrophages were observed in Jag2^−/− tumors implanted into Irf4^F/FLysM^Cre+/− mice compared to Irf4^F/F or LysM^Cre+/− controls (Figure 6J and 6K; Figure S6E). Together, these results suggest that in Jag2^−/− lung tumors, DLL1/4 binding to primarily Notch2 directs IRF4 expression to differentiate and promote anti-tumor activities of macrophages.

Therapeutic Jagged2 targeting delays tumor growth and drives protective immunity

We interrogated the therapeutic potential of an αJag2 blocking antibody^28,29. Treatment of LLC bearing mice with αJag2 induced significant anti-tumor effects in immunocompetent mice, but not in Rag1^−/− mice (Figure 7A and 7B), indicating a protective role of lymphocytes. Mice tolerated well αJag2 with no significant weight change compared to vehicle controls (Figure S7A), ruling out potential toxicity. Furthermore, consistent with Jagged2-competent tumor cells as a primary target of αJag2 therapy, Jagged2 ablation in tumors exhibited no additive reduction in tumor growth in mice upon treatment with αJag2 (Figure S7B). Also, in agreement with the protective role of immune cells in αJag2-treated mice, higher intra-tumor proportions of bulk macrophages, CD4⁺ T cells, CD44⁺CD69⁺CD4⁺ T cells, and Ly6C^HiMHC-II⁺ macrophages were observed in tumor-bearing mice treated with αJag2 compared to isotype controls (Figure 7C-G). Furthermore, we studied whether αJag2 therapy elicits anti-tumor actions through similar modulators as in Jag2^−/− tumors. Relative to vehicle-treated controls, αJag2 treatment induced Nr4a1 and Csf1 mRNA and DLL1 and DLL4 expression in tumor cells (Figure 7H and 7I), which correlated with higher IRF4, Tnfa, and Ifng levels in tumor-associated macrophages (Figure 7J and 7K; Figure S7C and S7D). Moreover, macrophages from LLC-bearing mice treated with αJag2 had heightened DQ-OVA processing and NICD2 expression compared to controls (Figure 7L and 7M). Conversely to the anti-tumor effects observed in control mice, treatment with αJag2 failed to delay tumor progression in Ccr2^−/− mice or Irf4^F/FLysM^Cre+/− mice (Figure 7N and 7O), revealing the primary protective role of macrophages, regulated via CCR2 and IRF4. Thus, similar to the results found in Jag2^−/− tumors, αJag2 induces anti-tumor effects, which correlate with the development of protective immunity.

Figure 7. Therapeutic targeting of Jagged2 recapitulates anti-tumor effects of Jagged2 genetic ablation.

(A) LLC tumor volume ± SEM in mice injected with isotype (ISO) or αJag2 (n = 15/treatment).

(B) LLC tumor volume ± SEM in Rag1^−/− mice injected with ISO or αJag2 (n = 5/treatment).

(C, D) (C) Immunohistochemistry images and (D) percent ± SEM of F4/80, CD8, or CD4 in tumors from (A); ISO: 15 fields; αJag2: 10 fields.

(E-G) Intratumoral proportions ± SEM of (E) CD4⁺ and CD8⁺ T cells (TIL) in CD45⁺ cells, (F) CD44⁺CD69⁺ CD4⁺ or CD8⁺ in TILs, and (G) Ly6C^HiMHC-II⁺ or Ly6C^LoMHC-II⁺ macrophages from (A).

(H) Relative fold change of Nr4a1 and Csf1 mRNA in CD45⁻ tumor cells from mice treated (n = 3/treatment) as in (A).

(I) MFI ± SEM of DLL 1 or DLL4 expression on LLC tumor cells from mice treated (n = 7, 9/treatment) as in (A).

(J, K) (J) Relative fold change ± SEM of Irf4 mRNA and (K) immunoblot of IRF4 in intra-tumor F4/80⁺ cells from mice treated (n = 3/treatment) as in (A).

(L) Representative histogram of tumor-injected cleaved DQ-OVA in total tumor-infiltrating macrophages from mice treated (n = 2/treatment) as in (A).

(M) MFI ± SEM of cleaved Notch2 (NICD2) in macrophage subsets from mice treated (n = 6/treatment) as in (A).

(N) LLC tumor volume ± SEM in C57BL/6 or Ccr2^−/− mice treated with ISO (n = 4, 5) or αJag2 (n = 5).

(O) LLC tumor volume ± SEM in Irf4^F/F (n = 5, 6/treatment), LysM^Cre+/− (n = 5/treatment), or Irf4^F/FLysM^Cre+/− (n = 5, 6/treatment) mice treated with ISO or αJag2.

Statistics were applied using one-way ANOVA or Student’s t-test, *, p<0.05; **, p<0.01, ***, p<0.001. See also Figure S7.

DISCUSSION

Here, we elucidated the mechanistic role of Jagged2 in NSCLC tumors in the evasion of anti-tumor immunity. Ablation of Jagged2 in NSCLC cells or treatment of lung tumor-bearing mice with αJag2 antibodies promoted protective T cell immunity activated by the expansion of macrophages directed by Notch-driven IRF4. Thus, Jagged2 blockade therapies represent a promising strategy to overcome immunosuppressive myeloid cells and restore protective immunity in tumors.

The role of Notch receptors in the regulation of anti-tumor immunity is emerging^13,30. However, the impact of Notch ligands, particularly Jagged, in immunology remains less understood. Recent reports show that Jagged1 and Jagged2 expression is elevated in lung tumors compared to normal tissue, but does not significantly differ across clinical stages¹². In our study, we could not formalize analyses on the role of tumor cell Jagged2 on T cell infiltration by NSCLC clinicopathological subtypes due to our limited sample number. Our previous report shows the intrinsic effect of an anti-Jagged1/2 antibody on limiting the immunoregulatory activity of tumor-related MDSC¹³. Also, anti-Jagged2 therapies can prevent expansion of IL-8-driven Jagged2⁺ neutrophils in tumors³¹. Here, we found a reduction in PMN-MDSC in mice bearing Jag2^−/− tumors, indicating that Jagged2⁺ tumor cells promote intra-tumor neutrophil accumulation as a potential mechanism of immune evasion. Upon Jagged2 ablation on NSCLC cells or αJag2 therapy, immunosuppressive myelopoiesis was rewired into the expansion of IRF4 and CCR2-dependent immunostimulatory macrophages. Different to directly targeting of Jagged2 in myeloid cells, we found that Jagged2⁺ NSCLC cells serve as a primary target for αJag2 therapy. Indeed, Jag2^−/− tumor cells regulated macrophages primarily through cell: cell interactions. A recent study demonstrates the impact of Jagged1 from breast tumor cells in driving tumor immune evasion by recruiting macrophages that suppress T cells³². In our study, we did not find major modulation of anti-tumor immunity upon Jagged1 ablation in NSCLC cells; although slight anti-tumor effects were noted. Therefore, further research needs to elucidate the role of Jagged forms in the modulation of myeloid cell activity in different tumors.

An additional layer of complexity arises from the divergent effects of Notch ligands. Although Jagged and DLL ligands can engage Notch proteins, their expression and interaction with Notch can trigger opposite immune effects^33-35. Our results showed that Jagged2 ablation in tumor cells directed DLL1 and DLL4 expression via the Nr4a family of transcription factors, suggesting a coordinated regulation of Notch ligand expression in lung tumor cells ^21,36. Notably, both DLL1 and DLL4 are necessary to fully activate tumor control upon ablation of Jagged2 in lung tumors. Therein, strategies directed at intercepting the immunoinhibitory role of Jagged ligands may exert dual function through blocking suppressive Jagged-directed signaling, but also by augmenting immunostimulatory actions of DLL1/DLL4-Notch signaling.

Our findings that Jagged2 ablation directed accumulation and immunostimulatory programming of intra-tumor macrophages underlines an intrinsic role for Notch signaling in the myeloid compartment. Aberrant Notch activation reportedly promotes accumulation of MDSC³⁷, whereas Notch stimulation in DC primes their maturation and function^9,38,39. Indeed, CD11c lineage-specific deletion of DLL1 accelerates growth of lung tumors concomitant with MDSC expansion and reduced T cell functions⁴⁰. In macrophages, Notch signaling ablation associates with their development into a “M2-like” phenotype characterized by pro-tumor functions⁴¹, while Notch upregulation coincides with their activation via toll like receptors^8,11. Our results showed that Jagged2 deletion in lung tumors triggered the expansion of macrophage subsets with a hybrid M1- and M2-like phenotype. Notch signaling induction by DLL4 ligand binding regulates macrophage function^42,43; however, within these contexts DLL4-Notch signaling coordinates pathogenic processes including atherosclerosis, diabetes-associated metabolic dysregulation, and fatty liver. Inversely, in the context of our study, we found that Notch signaling in macrophages provoked by DLL1/4 on Jagged2-null tumors primed anti-tumor actions. However, our results do not rule out additional detrimental impacts of Notch activation in myeloid cells or the potential direct crosstalk between tumor cell-DLL1/4 and Notch signaling in T cells.

Mechanistically, we noted that Notch2 activation on macrophages directs IRF4 expression, which mediated anti-tumor immune effects. Interestingly, prior studies describe the intracellular domains of Notch1 and Notch2 as functionally interchangeable during tumorigenesis⁴⁴. Whether the redundant functionality of Notch1 and Notch2 is applicable in macrophages remains to be studied. Increased expression of Notch2 relative to Notch1, divergent cytokine responsiveness, differential ligand affinities, and potential post-translational modifications of Notch1 relative to Notch2 may underlie the more pronounced role for Notch2 in inducing IRF4 in macrophages^45-47. Also, since IRF4 ablation limited macrophage expansion in mice bearing Scramble tumors, IRF4 could possibly serve as a key regulator for the accumulation of macrophage subsets in tumors. There is evidence of the regulation IRF transcription factors IRF8 and IRF5 by Notch activation and subsequent NICD signaling in macrophages^48,49. However, the relationship between Notch2 and IRF4 in macrophages has not been comprehensively interrogated. Although canonically reported as a driver of immunoregulatory macrophage development, IRF4 also exerts anti-tumor polarization of tumor-infiltrating immune subsets^25,50,51, suggesting a nuanced role of IRF4 in the development of anti-tumor immunity.

Taken together, our results elucidated the primary role for Jagged2 in NSCLC cells in regulating protective immunity and underline the anti-tumor therapeutic potential of targeting Jagged2 in lung cancer. We demonstrated that Jagged2 elimination in lung tumors triggers DLL1 and DLL4, which promotes expansion of immunostimulatory macrophages via Notch-driven induction of IRF4.

Limitations of the Study

Herein, we reported differences in T cell infiltration in human NSCLC tumors based on the degree of Jagged2 expression. However, the limited numbers of tumor samples in this dataset precluded any definitive conclusions on the contribution of Jagged2 on immune infiltration in NSCLC clinicopathological subtypes. Moreover, the identified macrophage subsets and phenotypes from our Jagged2-deficient murine models require validation in patients with cancer, including subsets carrying higher Notch and IRF4 signaling. Importantly, the impact of αJag2 therapies to reprogram the immunosuppressive milieu in human lung tumors remains to be studied, as well as the combinatory effects of αJag2 with other forms of immunotherapy.

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, Paulo C. Rodriguez, (Paulo.Rodriguez@Moffitt.org).

Materials availability

The authors declare that all the results supporting the findings of this study are available within the paper and its Supplemental Figures.

Data and Code Availability

RNAseq and survival datasets that are publicly available at the GEO repository with accessions GSE227671 (bulk RNAseq from CD45⁻ LLC), GSE228105 (scRNAseq from CD45⁺ in LLC), GSE72094 (Schabath_2016¹⁴). Additional survival analyses were conducted on datasets available from ORIEN Oncology Research Information Exchange Network Avatar database (ORIEN, https://www.oriencancer.org/) and The Cancer Genome Atlas (https://www.cancer.gov/tcga). Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

METHOD DETAILS

Mice

Experiments using mice complied with all the relevant ethical regulations for animal testing and research under an approved Institutional Animal Care and Use Committee (IACUC) protocol (IS00008833) and an active Institutional Biosafety Committee (IBC) study (#PROTO2020-043), reviewed by the Integrity and Compliance board at the University of South Florida and Moffitt Cancer Center. Mice were maintained under specific pathogen-free conditions, utilized between 6 – 10 weeks old, and randomly assigned to experimental groups. Wild type C57BL/6J mice were purchased from Envigo. Rag1^−/− (NOD.129S7 (B6)-Rag1^tm1Mom/J JAX:002216), (Ccr2^−/− (B6.129S4-Ccr2^tm1lfc/J; JAX:004999), LysM^Cre+/− (B6.129P2-Lyz2^tm1(cre)Ifo/J; JAX:004781), Irf4^F/F (B6.129S1-Irf4^tm1Rdf/J; JAX:009380), OT-II (B6.Cg-Tg(TcraTcrb)425Cbn/J; JAX:004194), KP (B6,129-Kras^tm4Tyj;Trp53^tm1Brn/J; JAX:032435), Msh2^F/F (B6.Cg-Msh2^tm2.1Rak/J; JAX: 016231) and CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ;JAX002014) mice were from The Jackson Laboratory. Floxed transgenic Rosa26-driven Notch1 intracellular domain-GFP (Rosa^NotchIC)²³ were backcrossed and maintained on a C57BL/6 background³⁰; herein Rosa^NotchIC were additionally crossed with LysM^Cre+/− to create a myeloid-specific GFP reporter for Cre-driven Notch signaling (Rosa^NotchIC;LysM^Cre+/−). Jagged^F/F were obtained from Dr. Mikhail M Dikov from the Ohio State University and crossed with KP mice to generate KPJ mice.

Human materials

All human studies are covered through the Institutional Review Board (IRB) protocol #19223, reviewed by the Regulatory Affairs Committee Board at Moffitt Cancer Center. We used two repeats of a NSCLC tissue microarrays (TMA) containing duplicated cores and within the 160 total tumor cores/patients across both TMAs, 120 are primary samples, 40 are metastatic samples (Biomax/Tissue Array LC953). Clinicopathological variables are available in Table S1. De-identified patients signed approved consent forms.

Multiplex imaging

The formalin-fixed paraffin-embedded lung cancer tissue microarray (TMA) was stained using an automated OPAL-IHC system (PerkinElmer) in a BONDRX (Leica Biosystems). Briefly, slides were treated with the PerkinElmer blocking buffer for 10 min and incubated with the specific primary antibodies (JAG2, PCK, and CD3), followed by OPAL-HRP polymer and one OPAL fluorophore. Individual antibody complexes were stripped after each round of detection and DAPI applied as the last staining. Auto-fluorescence slides (negative control) included primary and secondary antibodies, omitting the OPAL fluorophores. Slides were imaged with a Vectra Automated Quantitative Pathology Imaging System. Multi-layer TIFF images were exported from In-Form (PerkinElmer) into HALO (Indica Labs) for quantitative image analysis. Each fluorophore was assigned to a dye color and positivity thresholds determined visually per marker based on nuclear or cytoplasmic staining patterns, and by intensity thresholds normalized for exposure (counts/2bit depth × exposure time × gain × binning area). Core were removed from analysis upon significant damage during staining or no marker staining in PCK⁺ regions. For Figure 1B, C in 138 cores from two TMAs the JAG2 Lo/Hi was based off the JAG2 expression in PCK⁺ neoplastic areas, split into top/bottom 50% expression therein, and the number of CD3⁺ cells in the cores was averaged. For analysis of Primary Tumor samples only, 108 cores were included. From one TMA of 64 cores for Figure 4L, the JAG2 Lo/Hi was stratified into the top/bottom 50% followed by DLL1 and DLL4 expression split into high and low expression and the number of CD8⁺ cells averaged.

Cell lines

Lewis lung carcinoma (LLC; #CRL-1642) and EL4 (#TIB-39) were from the ATCC. Cells isolated from an autochthonous model of lung carcinoma induced after intranasal injection with adenoviral-encoded Cre recombinase in Kras^LSL–G12D/+;Trp53^F/F;Msh2^F/F mice that were reconstituted with wild-type Msh2 (KPM^Kin) were provided by Brad Perez, MD PhD (Moffitt Cancer Center) and cultured in RPMI-1640, media supplemented with 2 mM L-glutamine, 25 mM HEPES, 100 U/ml Penicillin/Streptomycin, 5 μM β-mercaptoethanol and 10% heat-inactivated Fetal bovine serum (FBS) and maintained at 37°C in a humidified incubator with 5% CO₂. Cell lines were routinely screened and validated to be mycoplasma-free using the Universal Mycoplasma Detection Kit (#30-1012K, ATCC) and maintained in culture for fewer than 10 passages. For genetic editing of Jag1, Jag2 (deletion or overexpression), Dll1, Dll4, or Nur77 (overexpression) in LLC or KPM^Kin cells were transduced with lentivirus containing three targeting guide RNAs, spCas9, and a puromycin resistance cassette or eGFP fluorescent marker. After puromycin or FACS selection, singe cell clones were screened for efficient deletion of target genes via western blot and qRT-PCR. Scramble cell lines transduced with lentivirus containing non-targeted gRNA. Scramble or Jag2^−/− LLC were stably transduced with a luciferase reporter for Nr4a transcriptional activity the NBRE motif reporter published in Munoz-Tello et al.²² then treated with 1 μg/ml LPS and 100 ng/ml IFNγ or 50% LLC TES for 24 h prior to flow cytometry reading of the reporter activity.

In vivo tumor models

Mice were subcutaneously injected with 1×10⁶ LLC cells/mouse or 2.5×10⁵ KPM^Kin cells/mouse; mice were intravenously injected with 1×10⁶ LLC cells/mouse or 1×10⁶ KPM^Kin cells/mouse; mice were intratracheally challenged with 2×10⁸ PFU adenovirus/mouse⁵². Tumors were measured with digital calipers and tumor volume calculated as: ((smallest tumor diameter)² × (largest tumor diameter) × 0.5). Tumor-bearing mice were humanely terminated when i) subcutaneous tumors reached on average 2,000 mm³, or ii) 15 – 21 days after intravenous administration, or iii) 28 – 42 days after intratracheal infection. LLC bearing-mice were injected intraperitoneally (i.p.) with BioXCell antibodies to CD4 (BE0003), CD8 (BE0004), DLL1 (BE0155), DLL4 (BE0127), CSF1R (BE0204), and Jagged2 (BE0125). To deplete macrophages, 500 μg liposomes containing clodronate or vehicle/empty liposomes (Encapsula Nanosciences CLD-8901) were injected i.p. every 3 days. To deplete CCR2-responsive macrophages, CCR2 antagonist BMS-CCR2-22 (0.5 mg/kg, Tocris) was injected daily two days after LLC inoculation.

Tumor digestions

Resected and minced tumors were digested with 0.1 mg/ml each of DNase I and Liberase TM (Roche USA) for 1 hour in a 37°C water bath prior to lysis of red blood cells with ammonium-chloride-potassium buffer. Samples were strained with sterile 100 μm mesh filters to generate single cell suspensions for downstream analyses. Alternatively, to tumor digestions were plated at 10×10⁶ cells/ml in complete RPMI overnight then the media collected, pelleted cells discarded, and the supernatant sterilized with a 0.22 μm filter. The resulting tumor explant supernatant (TES) was stored at −80°C until utilized in experiments. Similarly, tissue culture supernatant (TCS) was obtained from Scramble or Jag2^−/− cell lines after achieving a high confluency in culture.

Lung digestions

Mice bearing orthotopic or autochthonous lung tumors were perfused with PBS via cardiac puncture. Lungs were minced and digested for 45 min at 37°C with collagenase A (0.6 mg/ml; Sigma-Aldrich) and deoxyribonuclease I (30 μg/ml; Sigma-Aldrich) in RPMI 1640 medium (Gibco). Digested lungs were further mechanically disrupted to obtained single-cell suspensions, red blood cells were lysed, and cell suspensions were then filtered through a 70-μm nylon strainer and used for flow cytometry phenotyping.

Flow cytometry phenotyping and sorting

Conjugated antibodies and probes used for flow cytometry are listed in the Key Resources Table. For surface staining, cells were labelled with antibodies in the presence of Fc blocker. For intracellular staining, surface-labeled cells were fixed with Cytofix/Cytoperm Solution (BD Biosciences), washed in Perm/Wash Buffer, and labelled with intracellular antibodies. For IFNγ, TNFα, and NICD2 detection in tumor-infiltrating T cells, tumor suspensions were treated with phorbol myristate acetate (PMA, 750 ng/mL, Sigma Aldrich) and ionomycin (50 mg/mL, Sigma-Aldrich) for 5 hours in the presence of Golgi stop 0.8 ml/ml, BD Biosciences). Cells were then washed in Perm/Wash Buffer then PBS. Cells were acquired on a CytoFlexII (Beckman Coulter) and analysis performed with FlowJo version 11 software. Flow cytometric cell sorting (FACS) was performed on a FACSAria-II SORP (BD Biosciences).

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies (in vivo)
Anti-Jagged2	BioXcell	BE0125
Anti-CSF1R	BioXcell	BE0204
Anti-CD4	BioXcell	BE0003
Anti-CD8	BioXcell	BE0004
Anti-DLL1	BioXcell	BE0155
Anti-DLL4	BioXcell	BE0127
Isotype control	BioXcell	BE0091
Antibodies (Flow cytometry)
TruStain anti-mouse CD16/32 Antibody	Biolegend	101320
CD45 Alexa Fluor 700	Biolegend	103128
CD45 APC	Biolegend	103111
CD4 Brilliant Violet 421	Biolegend	100443
CD8 Brilliant Violet 785	Biolegend	100749
CD44 APC	Biolegend	103011
CD69 PE/Dazzle 594	Biolegend	104535
CD62L BV785	Biolegend	104440
IFNγ Brilliant Violet 421	Biolegend	505829
TNFα APC	Biolegend	506308
CXCR4	Biolegend	146511
MHC-II Brilliant Violet 421	Biolegend	107632
CD206 Alexa Fluor 700	Biolegend	141733
CD11b PE/Dazzle 594	Biolegend	101256
CD11c PE	BD Biosciences	12-0114-82
F4/80 Brilliant Violet 785	Biolegend	123141
CD103 APC	BD Biosciences	17-1031-82
Ly6C FITC	Tonbo Biosciences	35-5932-U025
Ly6G APC	Tonbo Biosciences	20-1276-U100
Gr1 FITC	Biolegend	108406
IRF4 PE	Biolegend	646403
CSF1R PE	Biolegend	135505
DLL1 APC	Biolegend	128314
DLL4 PE	Biolegend	130807
DLL4 PE (human)	Biolegend	346505
Notch1 Brilliant Violet 421	Biolegend	130615
Notch2 APC	Biolegend	130714
Notch3 Alexa Fluor 647	Biolegend	130512
Notch4 PE	Biolegend	128407
NICD2	LS Bio	C744599
NICD2	Developmental Studies Hybridoma Bank	C651.6DbHN
Rat IgG2a	Invitrogen	14-4321-82
Goat anti-Rat IgG, Texas Red	Invitrogen	T-6392
Antibodies/reagents (Magnetic enrichment)
Biotinylated F4/80 (BM8)	Biolegend	123106
Biotinylated Ly6C (HK1.4)	Biolegend	128003
Biotinylated CD45 (30-F11)	Biolegend	103103
TruStain anti-mouse CD16/32 Antibody	Biolegend	101320
Streptavidin Nanobeads	Biolegend	480016
MojoSort magnet	Biolegend	480019
Antibodies (Vectra multispectral imaging)
Jagged2 Human	Biolegend	131001
DLL1 Human	Novus biologicals	MAB18181
DLL4 Human	Novus biologicals	MAB1389-SP
Antibodies (Western blot)
Jagged2	Santa Cruz Biotechnology	sc-5604
IRF4	Cell Signaling Technology	62834
Notch2/cleaved Notch2	Cell Signaling Technology	5732
Nor1	Novus Biological	NB100-56745
Nur77	Invitrogen	14-5965-82
β-Actin	Sigma	A2228
Vinculin	Sigma	V9131
GAPDH	Fitzgerald	10R-G109a
p84	Abcam	ab487
Antibodies (chromatin precipitation)
Notch1	Cell Signaling Technology	3608
Notch2	Cell Signaling Technology	5732
Nur77	Novus Biological	NB100-56745
Nor1	Novus Biological	NBP2-46246
Rabbit IgG isotype	Cell Signaling Technology	2729
Plasmids, bacterial and virus strains
Scrambled sgRNA CRISPR/Cas9 All-in-One Lentivector	Applied Biologicals	K010
Jag2 sgRNA CRISPR/Cas9 All-in-One Lentivector (Mouse) (Target 3)	Applied Biologicals	K4385608
Jag1 sgRNA CRISPR/Cas9 All-in-One Lentivector (Mouse)	Applied Biologicals	K4373928
eGFP (for Scramble, JAG2KO)	Addgene	17448
GenCrispr DLL4 espCas9-LentiCrispr v2-eGFP	GenScript (custom)/This paper	Dll4 −/−
GenCrispr DLL1 espCas9-LentiCrispr v2-eGFP	GenScript (custom)/This paper	Dll1 −/−
PMD2.G plasmid	Addgene	12259
PsPAX2 plasmid	Addgene	12260
Nur77 overexpression (Nur77OE)	Addgene	203859
Nor1 overexpression (Nor1OE)	Addgene	203863
Nr4a transcription activity luciferase reporter vector	Munoz-Tello et al. (2020)	NBRE-based luciferase reporter assay
Biological samples
Human lung carcinoma tissue microarray	Biomax/Tissue Array	LC953
Ovarian carcinoma ascites	This paper	OVCAR ascites
Chemicals, peptides, and recombinant proteins
nt-siRNA	Invitrogen	4390843
Nr4a1-siRNA (Nur77)	Invitrogen	4390771
Lipofectatime 3000	ThermoFisher	L3000001
Opti-MEM	ThermoFisher	31985062
Recombinant murine IFNγ	R&D Systems	485-MI
Recombinant murine GM-CSF	GeminiBiosciences	300-846P
Recombinant murine M-CSF	GeminiBiosciences	300-332P
Lipopolysaccharide (LPS)	Sigma	14011S
Puromycin	Millipore Sigma	54-022-225
CCR2 antagonist	Tocris	BMS-CCR2-22
Clodronate liposomes	Encapsula Nanosciences	CLD-8901
Empty liposomes	Encapsula Nanosciences	CLD-8901
NICD inhibitor (FLI-06)	MedChemExpress	HY-15860
CellTrace Violet	Invitrogen	C34557
ApoTracker Green	Biolegend	427403
DQ-Ovalbumin (DQ-OVA)	Invitrogen	D12053
RPMI 1640	Gibco	21870076
L-glutamine	Cytiva	SH30034.01
HEPES	Corning	MT25060CI
Penicillin/Streptomycin	Cytiva	SV30010
β-mercaptoethanol	Fisher	BP2664100
Heat-inactivated fetal bovine serum	GeminiBiosciences	100-106
DNase I	Roche USA	4716728001
Liberase TM	Roche USA	5401119001
Cytofix/Cytoperm	BD Biosciences	554722
Perm/Wash Buffer	BD Biosciences	554723
SYBR Green	Bio Rad	1708880
SYBR Green	Applied Biosciences	4309155
Verso cDNA synthesis kit	ThermoFisher	AB14553B
ACK lysis buffer	Gibco	A1049201
Aprotinin	Roche	10236614001
Leupeptin	Roche	91058027
Trypsin/chymotrypsin	Sigma	TP777-50MG
Phosphatase inhibitor cocktail 2	Sigma	P0044-1ML
Phosphatase inhibitor cocktail 3	Sigma	P576-1ML
Pierce BCA protein assay kit	ThermoFisher	23227
TRIzol	Invitrogen	15596026
UltraPure BSA	Invitrogen	AM2616
DAPI	Invitrogen	D3571
Propidium iodide	Biolegend	421301
eFluor450 viability dye	Invitrogen	65086314
Phorbol-12-myristate-13-acetate (PMA)	Millipore Sigma	524400
ionomycin	Millipore Sigma	I3909
GolgiStop	BD Biosciences	554724
Critical commercial assays
Opal 7-color automation IHC kit	Perkin Elmer	NEL801001KT
Spectral DAPI	Perkin Elmer	FP1490
Universal mycoplasma detection kit	ATCC	30-1012K
RNeasy Micro Kit	Qiagen	74004
SimpleChip kit	Cell Signaling Technology	9002
Mouse Cytokine/Chemokine 44-Plex	Eve Technologies	MD44
Deposited data
Bulk RNA-seq (CD45− LLC)	This paper	GSE227671
Single-cell RNA-seq (CD45+ in LLC)	This paper	GSE228105
Schabath_2016	doi.org/10.1038/onc.2015.375	GSE72094
Oncology Research Information Exchange Network Avatar database	https://www.oriencancer.org/	ORIEN
The Cancer Genome Atlas – lung adenocarcinoma	https://www.cancer.gov/tcga	TCGA – LUAD
Experimental models: Cell lines
EL4	ATCC	TIB-39
LLC/3LL (wildtype/parental)	ATCC	CRL-1642
LLC Scramble Crispr	This paper	Scramble
LLC Jagged1 Knockout Crispr	This paper	Jag1 −/−
LLC Jagged2 Knockout Crispr	This paper	Jag2 −/−
LLC eGFP Scramble Crispr	This paper	LLC-eGFP Scramble
LLC eGFP Jagged2 Knockout Crispr	This paper	LLC-eGFP Jag2−/−
KP Msh2-knock-in (parental)	Brad Perez, MD	KPMKin
KP Msh2-knock-in Scramble Crispr	This paper	KPMKinScramble
KP Msh2-knock-in Jagged2-KO Crispr	This paper	KPMKinJag2−/−
Experimental models: Organisms/strains
WT (C57BL/6)	Envigo/Inotiv	044
Rag1−/− (B6.129S7 (B6)-Rag1tm1Mom/J)	The Jackson Laboratory	002216
Ccr2−/− (B6.129S4-Ccr2tm1lfc/J)	The Jackson Laboratory	004999
RosaNotchIC	Murtaugh et al. (2003)	Sierra et al. (2014)
LysMCre+/− (B6.129P2-Lyz2tm1(cre)Ifo/J)	The Jackson Laboratory	004781
OT-II (B6.Cg-Tg(TcraTcrb)425Cbn/J)	The Jackson Laboratory	004194
Irf4F/F (B6.129S1-Irf4tm1Rdf/J)	The Jackson Laboratory	009380
Msh2F/F (B6.Cg-Msh2tm2.1Rak/J)	The Jackson Laboratory	016231
KP (B6.129-Krastm4Tyj;Trp53tm1Brn/J)	The Jackson Laboratory	032435
Jagged2 F/F	Tchekneva et al. (2019)	Jag2 F/F
CD45.1	The Jackson Laboratory	002014
Adenovirus expressing Cre	University of Iowa; Viral Vector Core	Ad5CMV-Cre
Adenovirus control (Cre negative)	University of Iowa; Viral Vector Core	Ad5CMV-empty
Software and algorithms
FlowJo	BD Biosciences
FCS Express	DeNovo Software
Halo	Indica Labs
GraphPad Prism	GraphPad by Dotmatics
ImageScope	Leica Biosystems
InForm	Perkin Elmer
ImageLab	Bio Rad
BD FACSDiva	BD Biosciences
StepOnePlus Real-Time PCR Software	Life Technologies
JASPAR	Computational Regulatory Genomics
10X Genomics CellRanger	10x Genomics
Seurat	Stuart et al. (2019)
DoubletFinder	McGinnis et al. (2019)
SingleR	BioConductor
scVelo	Stuart et al. (2019)
Python	Python
STAR	Dobin et al. (2013)
HTSeq	Anders et al. (2015)
Primers
Forward	Reverse	Source	Identifier
TGTCCTGGCGGGTAGAAGA	ACCTGACCACTAACCACACAC	This paper	ChIP Hes1
TTATCCGAAACAAGGTCTCCGT	GGGTGTACCGTGGCAGTAGA	This paper	ChIP Irf41
CAGCAACAGCAAGGCGAAAAAGG	TTTCCGCTTCCTGAGGCTGGAT	Origene MP206683	Ifng
GGTGCCTATGTCTCAGCCTCTT	GCCATAGAACTGATGAGAGGGAG	Origene MP217748	Tnfa
GTGCAGTCTGTGGTGACAATGC	CAGGCAGATGTACTTGGCGCTT	Origene MP209119	Nr4a1
AGGTGACGGTTGTGTAAAGGT	TGTCCGGCACATGTTTGAAAG	This paper	ChIP Dll1
CGAGGCAGGAGCTACCTAAAG	GGAAAGCAAAAAGCAGGACCA	This paper	ChIP Dll4
TGTGATGGTGGGAATGGGTCAGAA	TGTGGTGCCAGATCTTCTCCATGT	This paper	Actb
TATCGCCCAGACATTCTCGC	ACACACTGACGGGGATCAAC	This paper	Notch2
AGGACCTGTTGGAGTTCCCTC	TTTCGCCCTCACACTTGATGA	This paper	Csf1
			Qiagen QT00115703	Jag2
			Qiagen QT00115703	Jag1
			Qiagen QT00113239	Dll1
			Qiagen QT01053598	Dll4
			Qiagen QT00156982	Notch1
			Qiagen QT00313537	Hes1
			Qiagen QT00109984	Irf4
			Qiagen QT00174195	Irf8
			Qiagen QT01149547	B2m

Magnetic bead cell enrichment

Following tumor digestion, F4/80⁺ macrophages were magnetically enriched with biotinylated anti-F4/80⁺ (Biolegend 123106) with Fc Block and positively selected with Biolegend MojoSort Streptavidin Nanobeads (480015) according to the manufacturer’s protocol. F4/80⁺ macrophages were counted and used in assays, including protein lysis for western blot, TRIzol extraction for mRNA assessment by qRT-PCR, DQ-OVA pulsing for uptake and antigen presentation. Similarly, Ly6C⁺ cells were isolated from Scramble or Jag2^−/− LLC tumor digestions using the Biolegend MojoSort and biotinylated anti-Ly6C antibody (Biolegend 128003). Ly6C⁺ cells were counted and labeled with 5 μM CTV for 15 min in 37°C, washed and suspended in PBS. Then, 5×10⁵ CTV⁺Ly6C⁺ were injected into Scramble or Jag2^−/− LLC tumors and after 24 hours were collected and analyzed by flow cytometry. CD45⁻ cells were isolated from vehicle or αJag2-treated LLC tumor digestions using the Biolegend MojoSort by utilizing biotinylated anti-CD45 antibody (Biolegend 103103). The pooled tumor cells were used for qRT-PCR and western blot analyses. For co-cultures with DQ-OVA loaded F4/80⁺ macrophages, spleen OT-II CD4⁺ T cells from tumor-naïve mice were negatively enriched following the manufacturer’s protocol.

Macrophage intra-tumoral transfer

Scramble and Jag2^−/− LLC were injected subcutaneously into CD45.2⁺ wildtype mice and tumors harvested 16 days post-injection; 5 days prior to harvest recipient CD45.1⁺ mice were injected with wildtype LLC tumors. CD45.2⁺F4/80⁺ macrophages from Scramble or Jag2^−/− tumors were magnetically positively enriched and 4.4×10⁶ macrophages were injected directly into each tumor of recipient CD45.2⁺ tumor-bearing mice. 24 and 48 hours after transfer, tumors were harvested and assessed for macrophage persistence by flow cytometry.

DQ-OVA antigen processing assay

DQ-OVA (Invitrogen D12053) was reconstituted to 1 mg/ml in PBS. For intra-tumoral injections, 100 μg DQ-OVA in 100 μl was injected and tumors collected after 1 hour. For in vitro assessment, F4/80⁺ cells were magnetically enriched, and equal numbers of F4/80⁺ macrophages plated and loaded with 20 μg/ml in 200 μl DQ-OVA for the indicated time points, washed, and analyzed by flow cytometry. Also, DQ-OVA exposed tumor F4/80⁺ macrophages were plated at varying ratios of macrophage to splenic CD4⁺ OT-II T cells for 72 hours followed by flow cytometry assessment of T cell proliferation (CellTrace Violet dilution) or activation (CD25 expression).

Bone marrow-derived macrophage and splenic macrophage co-culture assays

BM cells from mice femurs were filtered into a single cell suspension, red blood cells lysed, washed, and plated at 2×10⁶/well overnight in 2 ml RPMI supplemented with 20 ng/ml recombinant murine M-CSF and GM-CSF (GeminiBiosciences #300-332P, #300-308P) and replaced again after 24 hours over the adherent cells. On day 3, immature adherent BM-Mono were harvested and co-cultured with irradiated tumor cells. To their prevent overgrowth in cocultures, Scramble and Jag2^−/− cell lines were pre-treated with 200 Gy and debris removed with three PBS washes. As indicated, BM-Mono cultures were also pretreated with 10 μM NICD inhibitor, FLI-06. Irradiated tumor cell:BM-Mono were cultured for 96 hours at 1:16-1 with direct cell: cell contact or for transwell co-cultures irradiated tumor cells were seeded into the 0.4 μm PET membrane transwell (Corning Falcon #353095) with the BM-Mono cultured underneath on the plate. Similarly, splenic monocytes (sMono) were co-cultured with irradiated tumor cells. BM macrophage (BM-Mac) or splenic macrophage (sMac) development was analyzed by flow cytometry.

Western blot

Cells were lysed in 50 mM HEPES, pH 7.5, 250 mM NaCl, 5 mM EDTA, 0.5 mM DTT, 0.1% NP-40 alternative, 10 μg/ml aprotinin, 10 μg/ml leupeptin, and 100 μg/ml trypsin/chymotrypsin inhibitor. Lysates were quantitated with the Pierce BCA protein assay kit (ThermoFisher) and equal protein concentrations were heated and denatured then electrophoresed in 8, 10, or 4-20% Tris-glycine mini gels (Novex-Invitrogen), transferred onto nitrocellulose membranes with a Bio-Rad Trans-Blot SD Semi-Dry Transfer Cell, and blotted with indicated primary and secondary antibodies (Key Resources Table). Imaging of membrane-bound immune complexes was performed with a ChemiDoc Imaging System (Bio-Rad #17001401) and exported through ImageLab (Bio-Rad #12012931).

Quantitative RT-PCR

Total RNA was isolated via TRIzol (Invitrogen 15596026). Reverse transcription was performed with the Verso cDNA synthesis kit (Thermo Scientific AB14553B). Quantitative PCR reactions were catalyzed with SYBR green (Bio-Rad 1708880 or Applied Biosystems 4309155) and performed on an Applied Biosystems Thermocycler (7900 HT) using primers detailed in Key Resources Table.

Chromatin immunoprecipitation assay

ChIP assays utilized a SimpleChip kit (Cell Signaling Technology #9002) following the vendor's recommendations. Briefly, digested and cross-linked chromatin was prepared from 4 × 10⁶ magnetically enriched F4/80⁺ macrophages from Scramble or Jag2^−/− tumors or 4 × 10⁶ magnetically enriched CD45⁻ Scramble or Jag2^−/− tumor cells, followed by immunoprecipitation with the following antibodies for macrophages: rabbit IgG (Cell Signaling Technologies [CST] #2729), Notch1 (CST #3608), or Notch2 (CST #5732); for tumor cells: rabbit IgG (CST #2729), Nor1 (Novus Biologicals #NBP2-46246) Nur77 (Novus Biologicals #NB100-56745). Eluted and purified DNA was analyzed by qPCR with ChIP primers targeting the promoters for Irf4, Hes1, Dll1, or Dll4 based on the JASPAR prediction^53,54. Primers against Rpl30 promoter were used as a housekeeping gene-promoter control (provided in SimpleChip kit). The input DNA used to adjust for the chromatin concentration used in the ChIP. For each genotype, the adjusted mean input was used to normalize the percent of input DNA amplified for each primer set (Irf4, Hes1, Dll1, Dll4) for each immunoprecipitation antibody (IgG, Notch1, Notch2, Nor1, Nur77) and reported as % of input. To validate the Irf4, Hes1, Dll1, Dll4, and Rpl30 primers the efficiency of amplification was validated from a four-point standard curve of the 2% input diluted 5-fold (log-transformed input DNA ratios were used in a linear regression) and calculated as the primer set efficiency = ((10⁻¹/standard curve slope)-1)*100.

Histology and Immunohistochemistry

Tumors or PBS-perfused lungs bearing tumors were resected, cut transversely, and placed into tissue cassettes in 10% buffered formalin for 24–48 hours, transferred into 70% ethanol until the Moffitt Tissue Core Research Histology Services embedded the tumors into paraffin, sliced, and stained for indicated antigens or hematoxylin and eosin (H&E). The stained slides were digitized using the Aperio ScanScope XT (Leica Biosystems) at a 20X resolution. The Aperio ImageScope software (v12.4.3.5008, Leica Biosystems) was utilized to digitally draw the same size field or annotate lung-tumor regions for tumor burden quantitation. For IHC analysis, the Positive Pixel Count v9 algorithm with standard parameters selected strong positive cells for quantitation. To calculate cell infiltration into the tumor or lungs, the following formula was used: $({\mathrm{N}}_{\text{strong positive}}∕{\mathrm{N}}_{\text{total cells}})∕{\mathrm{A}}_{\text{total}}\times 100=\%$ strong positive cells out of total cells per mm² where N = cell count, A = area.

Multiplex ELISA

Measurement of cytokines and chemokines in tumor suspension protein extracts were quantitated using the Mouse Cytokine/Chemokine 44-Plex Discovery Assay (Cat #MD44, Eve-Technologies). Values were normalized based on protein concentration. Tumor homogenates were prepared from equal weights of resected tumor tissue subjected to three rounds of 60 second homogenizations with 1 mm glad beads (Sigma #1002619844) in Benchmark Bead Blaster homogenizer (3,000 RPM) in homogenate buffer: 20 mM Tris, 150 mM NaCl, 1% NP40, 100U/ml Leupeptin (Roche #91058027), aprotinin (Roche #10236614001), trypsin/chymotrypsin inhibitor (Sigma #TP777-50MG), phosphatase inhibitor cocktail 2 (Sigma #P0044-1ML) and phosphatase inhibitor cocktail 3 (Sigma #P576-1ML).

siRNA experiments

Tumor cells were plated one day prior to transfection with small interfering RNA (siRNA) targeting Nur77 (silencer select siRNA Nr4a1; #4390771, ID: s67611; Invitrogen), Nr4a3 (Nor1, #sc-38843; Santa Cruz Biotechnology) or nontargeting siRNA control (silencer select siRNA negative control #1; 4390843). Cells were transfected with 20 nM siRNA with the Lipofectamine 3000 system (ThermoFisher L3000001) in Opti-MEM media then incubated for 48 hours prior to experiments. The knockdown efficiency was validated via qRT-PCR. Jag−/− LLC were treated with non-targeting siRNA, Nor1 siRNA, Nur77 siRNA, or Nor1 and Nur77 siRNA and exposed to 1 μg/ml LPS and 100 ng/ml IFNγ or 50% LLC TES for 24 h prior to flow cytometry analysis.

Single cell RNA sequencing

Single cell suspensions of Scramble- or Jag2^−/−-LLC tumors were subjected to positive magnetic isolation for CD45⁺ cells (Biolegend, MojoSort), stained for CD45 (APC-CD45 clone F30-11) with mouse FcBlock for 30 min at 4°C followed by staining with 2 μg/ml propidium iodide (PI). Then, 600,000 CD45⁺PI⁻ cells were FACS sorted into 100% FBS. Post-sorting, the sample purity and viability was tested on a MACSQuant YVB as DAPI⁻CD45⁺ for scramble (94.27 ± 1.03%, n = 3) and Jag2^−/− (90.0 ± 3.5%, n = 3). The sorted CD45⁺ cells were suspended in BSA + 0.04% non-acetylated BSA and the viable cells were counted on a Nexcelom Cellometer K2 following staining with an acridine orange and PI cell viability kit for CD45⁺ samples from Scramble (6.62×10⁵ ± 6.61×10⁴, 77.1 ± 0.05% viable, n = 3) and Jag2^−/− (9.89×10⁵ ± 7.05×10⁴, 87.6 ± 0.04% viable, n = 3). Next, the cell suspensions were loaded onto the 10X Genomics Chromium Single Cell Controller at 1,000 cells/μl to encapsulate ~10,000 cells/sample. The single cells, reagents, and 10X Genomics gel beads were encapsulated into individual μl-sized Gelbeads in Emulsion (GEM) droplets wherein the poly-adenylated mRNA was reverse transcribed. The cDNA libraries were then completed in a single bulk reaction using the 10X Genomics Chromium NextGEN Single Cell 3’/5’ Reagent Kit v3.1, and 40,000 sequencing reads per cell were generated on the Illumina NovaSeq 6000 instrument. Demultiplexing, barcode processing, alignment, and gene counting was performed using the 10X Genomics CellRanger v6.0.0 software.

The scRNA-seq data was processed with Seurat (v.4.2.1)⁵⁵. Filters for low quality cells included high mitochondrial content (≥ 10%) or low feature counts (≤ 100 detected genes). Genes expressed in less than three cells were also excluded. DoubletFinder was additionally used to remove doublets⁵⁶. The remaining cells were integrated using the SCTransform function and normalized to their cell cycle phase using genes previously described⁵⁷. The resulting integrated SeuratObject utilized the top 2000 features in the FindVariableFeatures function, then clustered using the top 50 principal components and visualized by UMAP. Cell identity score was initially determined based on SingleR against the murine Immgen reference⁵⁸. Sub-clustering of the myeloid population was performed without T lymphocytes, fibroblasts, and DC to delineate myeloid subsets not currently noted in scRNAseq annotations. Analysis of the top differentially expressed genes (DEG) was done using the FindMarkers function. For myeloid cell identification, top DEGs in each cluster were utilized to correlate each cluster with known cell identity markers (Table S3). Trajectory analysis on the myeloid subcluster was done using scVelo on Python (v3.9.7)⁵⁹. Analysis of the top 50 DEG in cluster 2 Jag2^−/−-enriched macrophages were compared to the M1/M2 polarization dataset reported in Gerrick et al.¹⁵. To determine the upregulation of Notch signaling, the GSEA PID_NOTCH_PATHWAY list of Notch-related genes⁶⁰ was used to compare relative and absolute expression between cluster 0 Scramble-enriched and cluster 2 Jag2^−/−-enriched macrophages.

RNA sequencing

Single cell suspensions of eGFP Scramble and eGFP Jag2^−/− LLC were stained with APC-CD45 (Biolegend 103112) with mouse FcBlock for 30 min at 4°C followed by staining with 1 ng/ml DAPI. Then, 400,000 eGFP⁺CD45⁻DAPI⁻ cells were sorted into lysis Buffer RLT at a 1:1 ratio with sheath fluid and RNA was purified with the QIAGEN RNeasy Micro Kit (#74004) and quantitated by NanoDrop. RNA was sequenced and raw data files exported by Azenta/GENEWIZ (South Plainfield, NJ). Raw FASTQ files were aligned to the GRCm38 (mm10) genome using STAR (v2.7.7a)⁶¹ and raw counts were generated with HTSeq (v0.11.2)⁶². The top DEGs of Scramble and Jag2^−/− LLC were generated based on the standard Deseq2 pipeline⁶³.

Lung cancer patient RNAseq datasets

We used different lung cancer datasets, including TCGA (n = 474 lung adenocarcinoma patients), Schabath (n = 398 lung adenocarcinoma patients¹⁴), and ORIEN (n = 114 non-small cell lung cancer patients after ICI-treatment). JAG1 and JAG2 expression normalization methods conducted as described⁶⁴. Briefly, FPKM values or microarray abundance were log2 transformed and quantile normalized across each patient cohort. A cohort of patients treated with immune checkpoint inhibition (ICI) with RNA sequencing profiling was accessed through the Oncology Research Information Exchange Network (ORIEN) Avatar database. For overall survival probability analyses, samples were stratified by JAG2 mRNA by z-scores calculated relative to normal samples (log RNAseq V2 RSEM) where JAG2^Lo patients were designated with Z scores < −2 and JAG2^Hi designated with Z scores > 2. Samples were processed through the Aster Insights molecular data analysis protocol (BBDuk adapter trimming, STAR v2.7.3a mapping) and normalized to FPKM with RSEM⁶⁵. Patients with survival data were stratified by z-score cutoffs of JAG1 and JAG2 expression levels (low < −1; high ≥ −1). Survival curves were calculated using the Kaplan-Meier method, and the two-sided log rank test was used to calculate p values and log rank hazard ratios.

Quantification and statistical analysis

Statistical analysis was performed using GraphPad Prism 7 – 10 (San Diego, CA). For comparison of two groups in datasets with normally distributed with equal standard deviations, group means were compared by two-tailed unpaired Student’s t-test; for samples with significantly different standard deviations, unpaired t-tests with Welch’s correction were performed; for human data with non-parametric distributions unpaired Mann-Whitney t-tests were performed. For data sets with multiple groups and equal standard deviations, one-way ANOVA with Tukey’s multiple comparisons test was performed; for samples with unequal standard deviations, one-way ANOVA with Dunnett’s multiple comparisons test was performed. Statistically significant p values of < 0.05 are reported as *, p < 0.05; **, p < 0.01; and ***, p < 0.001.

Supplementary Material

3

Table S1. Clinicopathological variables of patient samples in TMA LC953, related to Figure 1.

4

Table S2. Frequency of JAG2 expression in PCK positive regions in TMA LC953, related to Figure 1.

5

Table S3. Differentially expressed genes in myeloid subcluster, related to Figure 2.

6

Table S4. Genes downregulated from RNAseq dataset of Jag2^−/− LLC compared to Scramble LLC, related to Figure 4.

7

Table S5. Pathways enriched in Jag2^−/− LLC analyzed by GSEA, related to Figure 4.

Highlights.

• Lung tumors expressing Jagged2 associate with poor outcome and immune evasion.

• Jagged2 deletion in lung tumors primes expansion of immunostimulatory macrophages.

• DLL1/4 in Jagged2-null tumors rewire macrophage function via Notch1/2-induced IRF4.

• Anti-Jagged2 therapy promotes anti-tumor responses through IRF4-driven macrophages.

Inclusion and Diversity Statement

We support inclusive, diverse, and equitable conduct of research.