GLI2 Facilitates Tumor Immune Evasion and Immunotherapeutic Resistance by Coordinating WNT and Prostaglandin Signaling

Abstract

Therapeutic resistance to immune checkpoint blockade has been commonly linked to the process of mesenchymal transformation (MT) and remains a prevalent obstacle across many cancer types. An improved mechanistic understanding for MT-mediated immune evasion promises to lead to more effective combination therapeutic regimens. Herein, we identified the hedgehog transcription factor, GLI2, as a key node of tumor-mediated immune evasion and immunotherapy resistance during MT. GLI2 generated an immunotolerant tumor microenvironment through the upregulation of WNT ligand production and increased prostaglandin synthesis. This pathway drove the recruitment, viability, and function of granulocytic myeloid-derived suppressor cells while also impairing type I conventional dendritic cell, CD8+ T-cell, and NK cell functionality. Pharmacologic inhibition of EP2/EP4 prostaglandin receptor signaling or WNT ligand secretion each reversed a subset of the immunomodulatory effects of GLI2 and prevented primary and adaptive resistance to anti-PD-1 immunotherapy, respectively. A transcriptional GLI2 signature correlated with resistance to anti-PD-1 immunotherapy in patients with stage IV melanoma. Together, these findings provide a translational roadmap to direct combination immunotherapies in the clinic. Significance: WNT and prostaglandin signaling generate an immunotolerant environment in GLI2-active tumors and can be targeted as a component of immunotherapeutic combination strategies to overcome resistance in tumors exhibiting mesenchymal plasticity.

Figure 1.. GLI2 activation is associated with anti-PD-1 resistance.

(A) Depiction of canonical and non-canonical activation of GLI2. (B) Spearman correlation of GLI2 expression with WNT5A, FOXL1, and FOXC2 in the human melanoma TCGA. (C) Left, Tumor FoxL1 and FoxC2 expression in an autochthonous BRAF^V600EPTEN^−/− melanoma model following anti-PD-1 antibody escape based on RNAseq (n = 3 tumors/group). Right, FOXL1 and FOXC2 expression in human melanoma tissues derived from anti-PD-1 responders and non-responders based on RNAseq. (D) Western blot of the active form of GLI2, GLI2ΔN, after anti-PD-1 escape in the murine BRAF^V600EPTEN^−/− melanoma model. Representative of 3 independent experiments. (E) Nanostring analysis of melanoma tissues derived from stage IV melanoma patients undergoing anti-PD-1 therapy. R, responder. LR, late relapser. NR, nonresponder. ab, antibody. Data presented as mean±SEM. Two group comparisons analyzed based on unpaired student’s t test. *p<0.05,***p<0.001,****p<0.0001.

Figure 2.. GLI2 drives tumor growth and immunosuppression in a murine melanoma model.

(A) Western blot analysis of WNT5a in BRAF^V600EPTEN^−/− GLI2^CA and control BRAF^V600EPTEN^−/−NTC control melanoma cell lines. (B) Western blot analysis of WNT5a in the BRAF^V600EPTEN^−/− GLI2^CA melanoma cell line treated with the direct GLI2 inhibitor, GANT61. This is the same experiment as shown in Figure 4D and includes the same control (β-Actin). (C) Gene Ontogeny (GO) Pathway Analysis of the BRAF^V600EPTEN^−/− GLI2^CA melanoma cell line compared to control. Number of significantly upregulated genes adjacent to GO term. (D) Western blot analysis of FoxC2 and FoxL1 TFs in the BRAF^V600EPTEN^−/−GLI2^CA and control melanoma cell lines. (E) BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/− GLI2^CA melanoma tumor measurements in syngeneic mice. Flow cytometric analysis of tumor-infiltrating (F) CD3⁺CD8⁺ T cells, (G) CD11b⁺Ly6G⁺F4/80⁻ PMN-MDSCs, and (H) ratio of cDC2s to cDC1s. (n = 6 tumors/group) (I) Whole tumor transcriptional analysis of BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA melanomas. NTC, non-target control. CA, constitutively active. Data presented as mean±SEM. Representative of 2–3 independent experiments. Data analyzed by E: two-way ANOVA; F,G,H: unpaired student’s t-test. *p<0.05, ***p<0.001.

Figure 3.. Activation of GLI2 modulates immunity throughout tumor development and promotes anti-PD-1 resistance.

(A) Time course of tumor growth measurements of BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA melanoma models in syngeneic mice. d, day of tissue harvest. n = 4 mice/group/time point. Flow cytometric analysis of tumor-infiltrating (B) CD3⁺CD8⁺ T cells, (C) PMN-MDSCs, (D) cDC1s (left) and cDC2s (right), (E) CD49b⁺NK1.1⁺CD3⁻ NK cells. (F) Transcriptional analysis of viable CD45⁺CD11c⁺MHCII⁺CD19⁻CD49b⁻CD3⁻F4/80⁻ DCs sorted from BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA melanomas. (G) Tumor growth curves from BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/− GLI2^CA melanomas treated with anti-PD1 antibody (aPD1) versus IgG control. Data are representative of 2 independent experiments. Flow cytometric analysis of tumor-infiltrating (H) CD3⁺CD8⁺ T cells, (I) PMN-MDSCs, (J) NK cells, (K) CD103⁺XCR1⁺ DCs. n = 5–6 mice/group. Data are representative of 3 independent experiments. Data presented as mean±SEM. Data are analyzed by A,G-K: two-way ANOVA; B-E: unpaired student’s t-test. *p<0.05,**p<0.01,***p<0.001.

Figure 4.. Prostaglandin signaling is driven by GLI2 in tumors.

(A) Volcano plot of differentially expressed genes in the BRAF^V600EPTEN^−/−GLI2^CA and BRAF^V600EPTEN^−/−NTC melanoma cell lines. (B) Western blot analysis of COX2 expression in the BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA melanoma cell lines. (C) ELISA of PGE₂ in conditioned media derived from the BRAF^V600EPTEN^−/−GLI2^CA and BRAF^V600EPTEN^−/−NTC melanoma cell lines. (D) GANT61 treatment of BRAF^V600EPTEN^−/− melanoma cells followed by Western blot analysis of COX2. This is the same experiment as shown in Figure 2B and includes the same control (β-Actin). (E) Western blot analysis of the mouse AKP-GLI2^CA colon cancer cell line. (F) Transcriptional analysis of multiple prostaglandin pathway members in BRAF^V600EPTEN^−/−GLI2^CA and BRAF^V600EPTEN^−/−GLI2^KO melanoma cell lines compared with their controls. (G) GLI2 binding motif in the Ptgs2/PTGS2 gene in mice and humans (UCSD genome browser). (H) GLI2 ChIP-qPCR analysis of the BRAF^V600EPTEN^−/−Cas9-NTC and BRAF^V600EPTEN^−/−GLI2^KO cell lines. (I) Correlation of an established prostaglandin signature and GLI2-associated genes in the melanoma (left), CRC (middle), and solid non-CNS tumors (right) in TCGA. KO, knockout. Data presented as mean±SEM. B-E, F, H: representative of 2–3 independent experiments. Data analyzed by C: unpaired student’s t-test; I: Spearman’s correlation. **p<0.01.

Figure 5.. GLI2-active tumors promote PMN-MDSC recruitment via WNT ligands while potentiating their suppressive function and survival through prostaglandins.

(A) Expression of PMN-MDSC recruiting chemokines in BRAF^V600EPTEN^−/−GLI2^CA and BRAF^V600EPTEN^−/−NTC whole tumors based on RNAseq. (B) ELISA analysis of CXCL5 in BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA cell lines ± Yap inhibitor, verteporfin (Yapi), or WNT5a silencing (WNT5a^KD, ± recombinant WNT5a rescue). (C) Flow cytometric analysis of tumor-infiltrating PMN-MDSCs (left) and CD3⁺CD8⁺ T cells (right) in BRAF^V600EPTEN^−/−GLI2^CAWNT5a^KD and BRAF^V600EPTEN^−/−GLI2^CAScr control tumors. n = 5–6 mice/group. (D) UMAP (left) and scRNAseq quantification of neutrophil subsets (right). (E) Flow cytometric analysis of tumor-infiltrating PMN-MDSCs from BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA tumor-bearing mice ± ETC-159. (F) Expression of functional PMN-MDSC genes based on scRNAseq. (G) Quantification of PMN-MDSCs in TPST-1495 treated tumors by flow cytometry (left) and relevant apoptosis-related genes by scRNAseq (right). n = 5–6 mice per group. (H) Tumor volume ratios of BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA tumors following Ly6G depletion. Data normalized to corresponding tumor volumes in IgG control-treated mice. (I) CD3⁺CD8⁺ T cells, (J) NK cells, (K) CD103⁺XCR1⁺ DCs. n = 10 mice/group. Scr, scramble. KD, knockdown. aLy6g, anti-Ly6G ab. Data presented as mean ± SEM. A-C, G, H-K: representative of 2 independent experiments. Data analyzed by C: unpaired student’s t-test; E, G, I-K: one-way ANOVA; B, H: two-way ANOVA. *p<0.05,**p<0.01,***p<0.001,****p<0.0001.

Figure 6.. GLI2-active tumors suppress effector T and NK cells while altering DC function.

(A) UMAP (left) and scRNAseq quantification of relevant NK Cell subsets (right). (B) Expression of select genes in effector NK cells based on scRNAseq. (C) Annexin-V apoptosis flow cytometry analysis of NK cells. n = 6 mice/group. Right, representative flow plots in BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA tumors. (D) UMAP (left) and scRNAseq quantification of relevant CD8⁺ T cell subsets (right). (E) Gene expression profiles from CD8⁺ T cell subsets. (F) Intra-tumoral TCF1⁺CD44⁺CD8⁺ T cell flow cytometry analysis of BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA tumors ± ETC-159. n = 6 mice/group. (G) scRNAseq UMAP of DC subsets (above) quantification of scRNAseq DC analysis (below). (H) Flow cytometric verification of relevant cDC2:cDC1 ratios. Expression of functional (I) cDC1 and (J) cDC2 genes from scRNAseq. Data presented as mean±SEM. Apoptosis data representative of 2 independent experiments. Data analyzed by one way ANOVA (C, F). *p<0.05.

Figure 7.. GLI2-active tumors are sensitive to selective prostaglandin and WNT inhibition.

Tumor growth kinetics of (A) BRAF^V600EPTEN^−/−NTC and (B) BRAF^V600EPTEN^−/−GLI2^CA tumor-bearing mice treated with TPST-1495 ± anti-PD-1. n = 6 mice/group. Flow cytometric analysis of intra-tumoral immune populations in BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA tumors ±TPST-1495± anti-PD-1 including PMN-MDSCs (C), NK cells (D), CD8⁺ T cells (E), and the ratio of cDC2s:cDC1s (F). (G) Waterfall plot of BRAF^V600EPTEN^−/−NTC and BRAF^V600EPTEN^−/−GLI2^CA tumors treated with the PORCN inhibitor ETC-159 ±anti-PD-1±TPST-1495. n = 5–7 mice/group. (H) Quantification of TRP2-specific CD8⁺ T cells by IFN-γ ELISPOT assay. (I) Schematic illustrating the development of anti-PD-1 resistance (top) with pre- and post-anti-PD-1 RNAseq of relevant WNT ligand and Prostaglandin synthesis genes (below). (J) Treatment of the autochthonous BRAF^V600EPTEN^−/− melanoma model prior to (left) or following (right) anti-PD-1 escape with ETC-159 or TPST-1495 ± anti-PD-1. (K) Receiver-operator characteristic (ROC) curves predicting resistance to anti-PD-1 monotherapy in stage IV melanoma patients utilizing a GLI2 signature based on a Duke Nanostring dataset (left) and an external dataset (right). (L) Distribution of responding stage IV melanoma patients undergoing anti-PD-1 monotherapy according to Tcell:GLI2 signatures in an external data set. Data presented as mean±SEM. A-J: representative of 2–3 independent experiments. Data analyzed by two-way ANOVA (A-F, H, J). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.