Background

Immune checkpoint inhibitors (ICIs) mainly include antibodies that targeting programmed cell death protein-1 (PD-1), programmed cell death ligand 1 (PD-L1), and cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4).1 ICIs immunotherapies have revolutionized the treatment of patients with a wide range of advanced cancers, including malignant melanoma, renal cell carcinoma, non-small cell lung cancer, Hodgkin’s lymphoma, breast cancer, cervical cancer, gastric carcinoma, and head and neck squamous cell carcinoma.2 However, only a minority of cancer patients benefit from ICIs immunotherapy. Given the potential toxicities and the significant economic cost of these agents, it is of vital significance to identify patients who are more likely to benefit from ICIs.1 3

SETD2 is the sole histone methyltransferase responsible for the trimethylation of histone H3 at lysine 36 (H3K36me3) and plays a critical role through H3K36me3.4 H3K36me3 is recognized and functions in modulating genomic instability by regulating DNA mismatch repairing, homologous recombination repairing initiated by DNA double-strand breaks and participates in intragenic transcription initiation, transcription elongation and RNA processing in alternative splicing.5–12 Genomic mutations of SETD2 are prevalent in ~5% of human cancers and promote tumor progression and metastasis, including clear cell renal cell carcinoma,13 14 pancreatic cancer,15 colorectal cancer,16 glioma,17 gastrointestinal stromal tumor,18 prostate cancer and breast cancer.19 20 However, no small molecules or antibodies that specifically target SETD2-deficient tumors and show therapeutic potential have been developed. In addition, SETD2 inactivation promotes cancer metastasis through distinct mechanisms in various cancer types.15 19 21 For instance, SETD2 inactivation promotes prostate cancer metastasis by reducing EZH2 methylation at K735 and inhibiting EZH2 degradation.19 In Kras-driven murine pancreatic cancer, SETD2 inactivation enhances metastasis by promoting acinar-to-ductal metaplasia and epithelia-mesenchymal transition.15 Loss of SETD2-mediated H3K36me3 induces a genome-wide increase in chromatin accessibility, which increases MMP1 transcription and accelerates clear cell renal cell carcinoma metastases.21 Therefore, it is challenging to develop patient-oriented therapies based on the pathway downstream of SETD2 in a manner that is effective across multiple cancers.

Recently, bioinformatic analysis of TCGA datasets has shown that SETD2 mutation is potentially correlated with the efficacy of ICI immunotherapy22; however, the functional importance of the SETD2 inactivation and how to modulate immunotherapy response remains unclear. Although the tumor suppressor function of SETD2 has been well established, the role of tumor cell-intrinsic SETD2 on antitumor immunity has been largely unexplored. Here, using various murine and human cancer cell lines, syngeneic murine models, patient samples, publicly available multiomics and clinical datasets, we present a comprehensive evaluation of the functional implications of SETD2 inactivation, including its role in altering the immune microenvironment and shaping ICIs response. We further demonstrate that the SETD2-NR2F1-STAT1 axis modulates the sensitivity of various lethal cancers to ICIs.

ResultsSetd2-knockout sensitizes tumors to ICIs immunotherapy in syngeneic murine models

SETD2 encodes a histone H3K36 trimethyltransferase that is mutated with high prevalence in human clear cell renal cell carcinoma, pancreatic cancer, melanoma, non-small cell lung cancer, among others (online supplemental figure S1A). SETD2, with 85% amino acid similarity, is highly homologous between human and mouse. To test whether SETD2 inactivation sensitizes tumors to immune checkpoint blockade, we established a syngeneic mouse tumor cell line, namely, KC cells, which were derived from murine spontaneous pancreatic adenocarcinoma samples from Pdx1cre; LSL-KrasG12D C57BL/6 mice (online supplemental figure S1B).23 We selected the pancreatic adenocarcinoma because pancreatic cancer is still one of the most lethal diseases and novel therapeutic strategies are urgently needed in clinical practice.24 According to TCGA dataset, genomic mutations that inactivate SETD2 occur in ~1.4% of pancreatic cancers. Whole-exome sequencing (WES) and RNA-seq confirmed that Setd2 gene was not mutated and that it was normally expressed in KC cells (online supplemental table S1,S2). Setd2-knockout (KO) KC cells were established using CRISPR/Cas9 genome-editing technology (figure 1A). KC-sgSetd2 cells showed similar rates of proliferation in vitro (online supplemental figure S1C) and of tumor growth in vivo in BALB/C immunodeficient mice compared with the Setd2-expressing control cells (online supplemental figure S1D). As Cd274 was upregulated in KC-sgSetd2 cells (online supplemental table S2), anti-PD-L1 antibody was used as the ICI in our study. To determine the effect of Setd2 on the response to immunotherapy, equal numbers of KC-sgSetd2 and control cells were subcutaneously (s.c.) transplanted into C57BL/6 immunocompetent mice, followed by intraperitoneal (i.p.) injection with the anti-PD-L1 antibody. Anti-PD-L1 therapy administration markedly decreased the growth of KC-sgSetd2 tumors compared with the vehicle control but had no effect on KC-control tumors (figure 1B–E). These results demonstrate that Setd2 knockout sensitizes pancreatic adenocarcinoma to anti-PD-L1 immunotherapy.

Figure 1

Setd2 knockout sensitizes murine tumors to anti-PD-L1 immunotherapy. (A) Western blotting confirmation of Setd2 knockout in KC cells. (B–E) Setd2 knockout enhances sensitivity to anti-PD-L1 therapy in KC cells. KC-ctrl or KC-sgSetd2 tumors treated with anti-PD-L1 or vehicle control are shown (n=5 for each group). A total of 4×106 KC-ctrl or KC-sgSetd2 cells were injected s.c. and anti-PD-L1 antibody or PBS was injected i.p. into C57BL/6 mice at the indicated time points. (B) Schematic of anti-PD-L1 administration; (C, E) growth curves (multiple t-test, n.s. p>0.05, *p<0.05), tumor weight at the end point (unpaired Student’s t-test, n.s. p>0.05, **p<0.01) and (D) representative tumor images are shown. Error bars represent the mean±SEM. (F) Western blotting confirmation of SETD2 expression in B16F10 cells. (G–J) Setd2 knockout improves sensitivity to anti-PD-L1 therapy in B16F10 cells. B16F10-ctrl or B16F10-sgSetd2 tumors treated with anti-PD-L1 or vehicle control are shown (n=5 for each group). A total of 1×106 cells were injected s.c. and anti-PD-L1 antibody or PBS was injected i.p. into C57BL/6 mice at the indicated time points. (G) Schematic of anti-PD-L1 administration; (H, J) growth curves (multiple t-test, n.s. p>0.05, *p<0.05, ****p<0.0001), tumor weight at the end point (unpaired Student’s t-test, n.s. p>0.05, ***p<0.001) and (I) representative tumor images are shown. Error bars represent the mean±SEM.

To provide additional evidence supporting this, we extended these findings to a second mouse tumor cell line (B16F10, murine melanoma) that expressed Setd2 (figure 1F), and B16F10 cells are completely resistant to immune checkpoint blockade with antibodies targeting the PD-1 and/or CTLA-4 receptors.25 Similar to KC cells, ablation of Setd2 in B16F10 cells did not appreciably decelerate cell proliferation in vitro (online supplemental figure S1E) or in immunodeficient mice (online supplemental figure S1F) but had a profound effect on the efficacy of anti-PD-L1 treatment in immunocompetent mice (figure 1G, H and I). In contrast, anti-PD-L1 treatment was ineffective against control B16F10 tumors (figure 1I, J). Hence, tumor-intrinsic Setd2 inactivation sensitizes both KC pancreatic adenocarcinoma and B16F10 melanoma to immune checkpoint blockade.

SETD2 inactivation activates the IFNγ pathway by upregulating Stat1

To determine the global gene expression patterns regulated by Setd2 loss, we performed RNA-seq on paired KC-KO/KC-WT C57BL/6 allograft tumors in vivo and KC-sgSetd2/KC-control cells in vitro (online supplemental table S2). GSEA revealed that gene sets that regulate the immune response, cytokine production, interferon-γ (IFNγ) response, antigen processing and presentation were significantly upregulated in both KC-KO tumors and KC-sgSetd2 cells (online supplemental figure S2A, figure 2A). The IFNγ pathway plays a critical role in the tumor response to immunotherapy.26 Notably, STAT1, a key transcription factor in the IFNγ pathway,27 was significantly upregulated and activated in KC-KO tumors and KC-sgSetd2 cells (figure 2B, C, online supplemental figure S2B,C). SETD2-STAT1 regulation was further demonstrated by western blotting and RT-qPCR in B16F10-sgSetd2 murine melanoma cells (figure 2D, E). Furthermore, SETD2 inactivation upregulated the IFNγ-induced chemokine genes Cxcl9, Cxcl10, and Cxcl11 in vitro (figure 2F, G, online supplemental figure S2D) and in vivo (figure 2H, online supplemental figure S2E,F). The STAT1-IFNγ regulating axis has been reported to promote antitumor immunity by recruiting CXCR3+ CD8+ T cells and CXCR3+ NK cells.28 Taken together, these results demonstrate that Setd2 inactivation increases STAT1 expression and activation, and then activates the IFNγ pathway, which promotes immune-related genes expression and immune-related pathway activation.

Figure 2

Setd2 inactivation induces STAT1 expression and activation. (A) Leading-edge plots from the GSEA highlighting the enriched pathways associated with STAT1 activation in KC-sgSetd2 cells. (B–E) Loss of Setd2 promotes STAT1 expression and activation in (B) KC cells and (D) B16F10 cells. Setd2 inactivation increases the mRNA levels of Stat1 in (C) KC cells and (E) B16F10 cells. *p<0.05, ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (F, G) Setd2 inactivation increases the expression of STAT1-activated chemokine genes in (F) KC cells and (G) B16F10 cells. Control or Setd2-KO cells were treated with or without 50 ng/mL IFNγ for 20 hours and total RNA was extracted. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (H)Setd2 inactivation in allograft tumors increases Cxcl9, Cxcl10 and Stat1 expression. Total mRNA was extracted from KC-ctrl or KC-sgSetd2 tumors from C56BL/6 mice (n=5 for each group). *p<0.05 by unpaired Student’s t-test. Error bars represent the mean±SEM. GSEA, gene set enrichment analysis.

Loss of SETD2 promotes PD-L1 expression and antigen presentation

It is well known that the IFNγ-JAK-STAT1 axis is the prominent pathway that contributes to PD-L1 expression and MHC-I antigen processing and presentation.29 30 In addition to chemokine genes, we found that PD-L1 expression and antigen presentation were enhanced on IFNγ-STAT1 pathway activation in our Setd2-deficient models. The upregulation of Pdl1 was confirmed at the transcriptional level by RT-qPCR in Setd2-knockout cells and allograft tumors (figure 3A, B, online supplemental figure S3A,B). PD-L1 protein expression on the cell membrane was increased, as shown by flow cytometry analysis (figure 3C, D, online supplemental figure S3C). It has been reported that upregulation of PD-L1 results in potent immune suppression and tumor immune escape, which explains the potential cause of the immunosuppression in Setd2-deficient tumors.31

Figure 3

Setd2 inactivation promotes IFNγ-STAT1-induced PD-L1 expression and MHC-I antigen presentation. (A)Setd2 knockout promotes Pdl1 expression in KC cells and B16F10 cells. Setd2- knockout or control cells were treated with or without 50 ng/mL IFNγ for 20 hours and total RNA was extracted. **p<0.01 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (B) Setd2 inactivation increases Pdl1 and H2-K1 expression in allograft tumors. Total mRNA was extracted from KC-ctrl or KC-sgSetd2 tumors from C57BL/6 mice (n=5 for each group). *p<0.05 by unpaired Student’s t-test. Error bars represent the mean±SEM. (C, D)Setd2 knockout promotes PD-L1 expression on the cell surface of (C) KC cells and (D) B16F10 cells. Setd2-knockout or control cells were treated with or without 50 ng/mL IFNγ for 12 hours and PD-L1 expression was detected by flow cytometry. Iso, isotype. MFI, mean fluorescence intensity. **p<0.01, ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (E) Setd2 knockout promotes H2-K1 mRNA expression in KC cells. Cells were treated with or without 50 ng/mL IFNγ for 20 hours and total RNA was extracted. ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (F) Setd2 knockout promotes H-2Kb/Db expression on the cell surface of KC cells. Cells were treated with or without 50 ng/mL IFNγ for 12 hours and H-2Kb/Db expression was assessed by flow cytometry. ***p<0.001, ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (G)Setd2 knockout promotes SIINFEKL-H-2Kb expression on the cell surface in KC cells. Cells were transduced with OVA lentivirus and treated with or without 50 ng/mL IFNγ for 12 hours and SIINFEKL-H-2Kb expression was assessed by flow cytometry. ***p<0.001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (H) Setd2 deficiency sensitizes KC cells to effector CD8+ T cell-mediated killing. Apoptosis was analyzed in CD45− tumor cells at the indicated time points. Error bars represent±SEM of three replicates. ****p<0.0001 by unpaired Student’s t-test. T, tumor cell. E, effector CD8+ T cell. VNA, viable non-apoptotic cell. VA, viable apoptotic cell. NVA, non-viable apoptotic cell. NVNA, non-viable non-apoptotic cell.

The upregulation of the MHC-I complex core gene H2-K1 in Setd2-knockout cells was also shown at the transcriptional level by RT-qPCR (figure 3E, online supplemental figure S3D), and the H-2Kb/Db protein expression level on the cell surface was consistently increased (figure 3F, online supplemental figure S3E). The upregulation of H2-K1 was further demonstrated in vivo (figure 3B, online supplemental figure S3F). Taking the GSEA results together, we show that Setd2 loss promotes MHC-I antigen-processing and presentation.

Furthermore, we constitutively expressed neoantigen ovalbumin (OVA) in Setd2-KO KC cells, confirmed the upregulation of SIINFEKL-H-2Kb levels on the surface of sgSetd2-OVA cells (figure 3G, online supplemental figure S3G), and then cocultured with OT-1 effector CD8+ T cells or injected subcutaneously into OT-1 mice. Both in vitro and in vivo assays showed that Setd2 deficiency sensitized the KC cells to effector CD8+ OT-1 cell-mediated killing (figure 3H, online supplemental figure S3H,I). Setd2 knockout also sensitized B16F10 cells to effector CD8+ T cell cytotoxicity (online supplemental figureS3J). Activation of the PD-L1/PD1 axis inhibits T cell proliferation, activation and survival mainly by attenuating the T cell receptor (TCR) and CD28 signaling pathways.31 We found that IFNγ-stimulated PD-L1 expression was unable to inhibit the cytotoxicity of OT-1 effector CD8+ T cells activated by specific peptides in vitro, and this results might occur because the activation signaling due to the interaction of highly expressed SIINFEKL-MHC-I and high-affinity OT-1-TCR was stronger than the suppression signaling of the PD-L1/PD1 interaction.32

In addition, Setd2 deficiency elevated the tumor mutation burden (TMB), as determined by WES (online supplemental figure S3K), (online supplemental table S1), which was validated in multiple human cancers (online supplemental figure S3L).The result is consistent with a previous report that SETD2 functions in DNA double-strand break repair and the accumulation of genomic mutations in Setd2 deficient tumors.9 Taken together, our data indicate that Setd2 deficiency activates the IFNγ pathway by upregulating Stat1 and further increases Pdl1 expression, IFNγ-induced chemokine gene expression and MHC-I antigen presentation, which synergistically improving the effects of anti-PD-L1 immunotherapy.

SETD2 inactivation promotes Stat1 expression by downregulating Nr2f1

As the sole enzyme responsible for the trimethylation of lysine 36 on histone 3 (H3K36me3), a histone mark related to actively transcribed regions, most functions of SETD2 are mediated by H3K36me3.33 Chromatin immunoprecipitation sequencing (ChIP-seq) was performed to map the distribution of H3K36me3 in KC-KO and KC-WT cells. Consistent with the previous reports, H3K36me3 was widely distributed in whole genomic regions, including gene bodies, promoters, and intergenic zones (online supplemental figure S4A, online supplemental table S3).15 SETD2 loss induced alterations in chromatin accessibility and transcriptional output via H3K36me3 loss.21 Assay for transposase-accessible chromatin using sequencing (ATAT-seq) was also carried out in KC-ctrl and KC-sgSetd2 cells to assess the impact of chromatin accessibility due to Setd2 loss on Stat1 expression. Loss of SETD2 in KC cells led to genome-wide alterations in chromatin accessibility, including introns, promoters and intergenic regions (online supplemental figure S4B, online supplemental table S4).

ChIP-seq and ATAC-seq analyses did not show significant differences in H3K36me3 deposition and chromatin accessibility for Stat1, suggesting that Stat1 expression was indirectly regulated by Setd2 loss. The reduction in Stat1 at the transcriptional level led us to determine that Setd2 deficiency may regulate the expression of potential transcription factors that increase the transcription of Stat1. Nr2f1 was the most significantly downregulated transcription factor among the differentially expressed genes and showed the consistent decreasing trends in H3K36me3 enrichment of ChIP-seq and chromatin accessibility of ATAC-seq in Setd2-inactivated cells (figure 4A, B, online supplemental figure S4C,D), (online supplemental table S5). NR2F1 is an orphan nuclear receptor belonging to the steroid/thyroid hormone receptor superfamily and functions as a transcriptional regulator that can both activate and repress target gene expression.34 We identified the NR2F1 binding motif ‘AGGTCA’in the Stat1 promoter region through a bioinformatics approach (online supplemental figure S4E).34 To experimentally investigate whether Stat1 transcription was regulated by NR2F1, Nr2f1 lentivirus was transduced into Setd2-knockout and control cells (figure 4C, online supplemental figure S4F) and the reduction in Stat1 expression was confirmed at both the mRNA level by RT-qPCR and the protein level by western blotting (figure 4D, online supplemental figure S4G), followed by the downregulation of a series of IFNγ-induced genes (figure 4E, online supplemental figure S4H). Furthermore, to assess whether NR2F1 directly binds to the Stat1 promoter, we constructed a Flag-Nr2f1 lentiviral vector and determined its ability to inhibit STAT1 expression and activation (online supplemental figure S4I). Direct binding was confirmed by ChIP‒qPCR assay (figure 4F). In addition, the GFP reporter assay showed that NR2F1 significantly inhibited GFP expression driven by the Stat1 promoter (figure 4G, H). These results indicated that Setd2 loss impaired the transcription of Nr2f1 by decreasing H3K36me3 deposition on the Nr2f1 gene body and reducing chromatin accessibility proximal to the Nr2f1 promoter region, which further increased the expression and activation of STAT1 in a negative feedback.

Figure 4

Setd2 inactivation induces STAT1 expression and activation by downregulating Nr2f1 expression. (A) Setd2 loss suppresses Nr2f1 expression in KC cells. ***p<0.001 by unpaired Student’s t-test. Error bars represent±SEM of three replicates. (B) ChIP-seq, ATAC-seq and RNA-seq tracks of the Nr2f1 gene locus in the indicated cells. (C, D) NR2F1 overexpression inhibits STAT1 expression and activation. KC-ctrl or KC-sgSetd2 cells were transduced with Nr2f1 or control lentivirus. EV, empty vector. (C) Nr2f1 mRNA levels and (D) STAT1 expression and activation were shown. **p<0.01, ***p<0.001, ****p<0.0001, n.s. p>0.05 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (E) NR2F1 overexpression impairs the expression of Stat1 downstream genes in KC-sgSetd2 cells. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by unpaired Student’s t-test. Error bars represent±SEM of three replicates. (F) NR2F1 directly binds to the Stat1 promoter region in KC cells, as determined by ChIP‒qPCR assay. KC-sgSetd2 cells were transduced with Flag-Nr2f1 and control lentivirus. ChIP assays were performed with anti-Flag and isotype IgG. **p<0.01, n.s. p>0.05 by unpaired Student’s t-test. Error bars represent±SEM of three replicates. (G, H) NR2F1 suppresses GFP expression driven by the Stat1 promoter in HEK293T cells. MFI of GFP and mCherry was assessed by flow cytometry. SP, Stat1 promoter. ****p<0.0001, n.s. p>0.05 by unpaired Student’s t-test. Error bars represent±SEM of three replicates.

SETD2 deficiency reprograms the tumor immune microenvironment

To systemically elucidate the regulation of SETD2-NR2F1-STAT1 axis to the tumor immune microenvironment, we established Setd2-knockout (KC-KO) and Setd2-wildtype (KC-WT) KC monoclonal cell lines (online supplemental figure S5A) and performed RNA-seq assays in C57BL/6 allograft tumors (online supplemental table S2). Immune cell infiltration was significantly higher in KC-KO tumors than that in KC-WT tumors by single sample gene set enrichment analysis (ssGSEA) and ESTIMATE analysis (figure 5A, online supplemental figure S5B).35 Many immune-related genes, including the immune inhibitory genes Cd274, Ctla4, Lag3 and Ido1, were upregulated in KO tumors. In addition, the expression of some immune stimulatory genes and chemokine genes, such as Ccl5 and Cxcl10, was also upregulated in KO tumors (figure 5B).36 37

Figure 5

Bulk RNA-seq and single cell RNA-seq analysis for intratumoral immune infiltration. (A) Heatmap showing the profiles of intratumoral immune cells in KC-WT or KC-KO allograft tumors through bulk RNA-seq, n=3. The relative infiltration of each cell type is normalized into a z-score. (B) Heatmap showing the relative expression of immune-related genes in KC-WT or KC-KO allograft tumors, n=3. (C, D) Setd2 loss alters the intratumoral infiltration of myeloid cells and lymphocytes. (C) Clusters of intratumoral immune cell populations and (D) statistics of individual cell populations identified from single-cell transcriptomic data. Individual clusters are indicated by colors. (E, F)Setd2 loss reprograms macrophage compartments. (E) t-SNE plots showing secondary clusters of macrophages and (F) statistics of M1- and M2-like macrophages. (G, H) Setd2 loss reprograms T lymphocyte subgroups. (G) t-SNE plots showing secondary clusters of T lymphocytes and (H) population statistics of individual cell types. (I, J) Cytotoxic CD8+ T cells are significantly increased through anti-PD-L1 treatment in KC-sgSetd2 tumors. (I) Trajectory analysis showing CD8+ T cell status in different groups and (J) statistics of the indicated cell states. (K, L) Intratumoral T cells infiltrations was increased in (K) KC-sgSetd2 and (L) B16F10-sgSetd2 tumors treated with anti-PD-L1 by flow cytometry analysis. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM.

Single-cell RNA-seq (scRNA-seq) analysis of sorted CD45+ immune cells of Setd2-deficient and control KC tumors that were isolated from PBS-treated C57BL/6 immunocompetent mice showed that the proportions of infiltrating T cells, neutrophils, natural killer (NK) cells, and dendritic cells were significantly increased, while the proportion of infiltrating macrophages was significantly reduced in Setd2-deficient tumors compared with control KC tumors (figure 5C, D, online supplemental figure S5C,D). Given the dual function of tumor-associated macrophages (TAMs),38 we reclustered macrophages into M1-like and M2-like macrophages and found a decrease in both types of macrophage proportions, while a dramatic decrease in M2-like macrophages in Setd2-deficient tumors was shown (figure 5E, F, online supplemental figure S5E,F). In view of the crucial role of T cells in antitumor immunity,39 we annotated T cell subsets according to their function and marker genes (figure 5G, online supplemental figure S5G,J). Infiltration of naïve and effector T cell subsets, including naïve CD4+ T cells, regulatory CD4+ T cells (Tregs), CD4+ helper T cells, naïve CD8+ T cells and effector CD8+ T cells, was significantly enhanced in Setd2-knockout allografts, mainly manifesting as an increase in effector CD8+ T cells and Tregs (figure 5H). These results indicate that Setd2 loss inflames the tumor microenvironment (TME) by reducing M2 macrophage infiltration and increasing effector T cell infiltration.

The relative ratio of effector CD8+ T cells and Tregs indicated that effector CD8+ T cells exerted a major antitumor effect with anti-PD-L1 treatment (online supplemental figure S5K). To further explore the effect of CD8+ T cells in the context of anti-PD-L1 administration, we assessed the status of CD8+ T cells. Trajectory analysis demonstrated that there was a significant increase in the proportion of cytotoxic CD8+ T cells in KC-sgSetd2 tumors treated with anti-PD-L1 immunotherapy (figure 5I, J, online supplemental figure S5L), and the anti-PD-L1 immunotherapy regulated the differentiation of CD8+ T cells, turning them from the exhausted to the cytotoxic state in both Setd2-expressing and Setd2-deficient tumors (online supplemental figure S5M). Flow cytometry analysis further validated that anti-PD-L1 treatment significantly increased the infiltration of CD3+ T cells, CD4+ T cells and CD8+ T cells in Setd2-knockout subcutaneous allografts (figure 5K, L, online supplemental figure S5N). Compared with KC-ctrl group, Setd2 knockout significantly increased the infiltration of CD45+ immune cells and anti-PD-L1 therapy further increased the infiltration of CD3+ T cells in orthotopic allografts (online supplemental figure S5O). The findings were validated in human TCGA clear cell renal cell carcinoma (ccRCC) cohort. The immune cell infiltration was significantly higher in SETD2-deleted tumors than that in high-expressed tumors (online supplemental figure S5P,Q), which was consistent with the findings in mouse pancreatic orthotopic tumors (online supplemental figure S5O). These results demonstrated that the inflamed TME due to Setd2 loss contributed to anti-PD-L1 treatment and the main role played by cytotoxic CD8+ T cells in immunotherapy administration. All the results show that Setd2 deficiency has a complicated interaction with the TME. On the one hand, Setd2 deficiency promotes chemokines production to inflame immune microenvironment with increased infiltration of T cells and enhances antigen presentation to activate CD8+ T cell-mediated killing. On the other hand, activation of STAT1 promotes the expression of immune inhibitory checkpoint molecules such as PD-L1, which rivalries with antitumor immunity. All the above mechanisms facilitate the benefit of ICIs immunotherapy.

SETD2-NR2F1-STAT1 axis in human cancers

To confirm the findings in murine tumors that SETD2 inactivation inhibits STAT1 expression and activation in a NR2F1-dependent manner in human cancers, we knocked out SETD2 with individual SETD2 sgRNA in the human pancreatic ductal adenocarcinoma cell line YAPC, which normally expresses SETD2 and is widely used in pancreatic cancer research (figure 6A).40 The upregulation of STAT1 was confirmed at both the mRNA and the protein levels, and the upregulation of CXCL10 and PDL1 mediated by STAT1 activation was also validated in YAPC-sgSETD2 cells (figure 6A, B). Next, we performed GSEA analysis to obtain a global view of transcriptome differences in SETD2-knockout and control YAPC cells. Several pathways were enriched in SETD2-knockout YAPC cells, similar to the gene signatures in KC-sgSetd2 cells (figure 6C), and the expression of immune-related genes also increased as that in KC-KO cells (figure 6D).

Figure 6

SETD2-NR2F1-STAT1 axis in human ccRCC and pancreatic cancer. (A) SETD2 knockout promotes STAT1 expression and activation in YAPC cells. (B) SETD2 knockout increases the mRNA levels of STAT1, PDL1 and CXCL10 in YAPC cells. ***p<0.001, ****p<0.0001 by unpaired Student’s t-test. Error bars represent the mean±SEM of three replicates. (C) Leading-edge plots from the GSEA highlight the enriched pathways associated with STAT1 activation in YAPC-sgSETD2 cells. (D) Column charts showing the mRNA levels of the indicated genes in YAPC-ctrl or YAPC-sgSETD2 cells according to RNA-seq. (E) Spearman correlation of expression levels between SETD2 and STAT1 in human pancreatic cancer41 cohort, tumors with unknown-significance of SETD2 mutations were excluded, n=89). (F) Spearman correlation of SETD2 and NR2F1 mRNA levels in human pancreatic cancer (ICGC PACA-CA cohort, n=234). (G) Correlations between NR2F1 and STAT1, SETD2 and STAT1 expression in human pancreatic cancer41 cohort, tumors with unknown-significance of SETD2 mutations were excluded, n=89). SETD2-high represents the tumors of the top one-third and SETD2-low represents the tumors of the bottom one-third based on the SETD2 expression level. *p<0.05 by unpaired Student’s t-test. Error bars represent±SEM. (H, I) Correlation between SETD2 and NR2F1 expression levels in the Renji pancreatic cancer TMA cohort assessed by (H) immunohistochemistry staining and (I) representative pictures of expression levels (n=75). IHC, immunohistochemistry. High represents expression higher than or equal to tumor adjacent normal tissues. Low represents expression lower than tumor adjacent normal tissues. ***p<0.001 by Fisher’s exact test. (J–L) SETD2-NR2F1-STAT1 axis in human ccRCC (TCGA renal cell carcinoma cohort). Spearman correlations of (J) NR2F1 and STAT1 and (K) SETD2 and NR2F1 mRNA levels in human ccRCC cohort, n=569. (L) Correlation between SETD2 and NR2F1 expression. SETD2-high represents the top 50 tumors, and SETD2-low represents the bottom 50 tumors based on SETD2 mRNA levels in the SETD2 WT cohort. SETD2-loss represents 29 tumors with SETD2 inactivation mutations. ****p<0.0001, n.s. p>0.05 by unpaired Student’s t-test. Error bars represent±SEM.

Analysis of public datasets on the Nature, 2016 and PACA-CA (ICGC) pancreatic cancer cohorts revealed the inverse correlations between SETD2 and STAT1 expression, NR2F1 and STAT1 expression, and SETD2 and PDL1 expression (figure 6E, F