Targeting the IRE1α/XBP1s pathway suppresses CARM1-expressing ovarian cancer

CARM1 is often overexpressed in human cancers including in ovarian cancer. However, therapeutic approaches based on CARM1 expression remain to be an unmet need. Cancer cells exploit adaptive responses such as the endoplasmic reticulum (ER) stress response for their survival through activating pathways such as the IRE1α/XBP1s pathway. Here, we report that CARM1-expressing ovarian cancer cells are selectively sensitive to inhibition of the IRE1α/XBP1s pathway. CARM1 regulates XBP1s target gene expression and directly interacts with XBP1s during ER stress response. Inhibition of the IRE1α/XBP1s pathway was effective against ovarian cancer in a CARM1-dependent manner both in vitro and in vivo in orthotopic and patient-derived xenograft models. In addition, IRE1α inhibitor B-I09 synergizes with immune checkpoint blockade anti-PD1 antibody in an immunocompetent CARM1-expressing ovarian cancer model. Our data show that pharmacological inhibition of the IRE1α/XBP1s pathway alone or in combination with immune checkpoint blockade represents a therapeutic strategy for CARM1-expressing cancers.

The development of therapeutic strategies for epithelial ovarian cancer (EOC) remains a major obstacle to overcome. High-grade serous ovarian cancer (HGSOC) is the most common and fatal subtype of EOC. EOC is genetically heterogeneous1. Thus, it is imperative that therapeutic strategies need to be personalized by targeting distinct molecular subsets of EOC2. CARM1 (also known as PRMT4) is a type I protein arginine methyltransferase (PRMT) that asymmetrically dimethylates arginine residues on protein substrates3,4. Overwhelming evidence suggests that CARM1 functions as an oncogene in human cancers5. High levels of CARM1 expression have been observed in several major cancer types, including breast, colon, and prostate6,7,8. Notably, CARM1 amplification/overexpression occurs in ~20% of HGSOCs9, the highest among all cancer types in The Cancer Genome Atlas (TCGA) databases. However, the role of CARM1 and the associated therapeutic vulnerabilities conferred by its expression in cancers including EOC remain to be explored.

The endoplasmic reticulum (ER) stress response or the unfolded protein response (UPR) orchestrates adaptive programs to promote cancer cell survival10,11,12. Thus, inhibition of the UPR represents a therapeutic approach for cancers with hyperactive ER stress response12. The mammalian UPR is governed by three stress transducers that sense ER stress. They include inositol-required enzyme 1 alpha (IRE1α), activating transcription factor 6 (ATF6), and protein kinase RNA-like ER kinase (PERK)12. When functions of the ER are severely impaired, cells undergo apoptosis through effector proteins such as C/EBP homologous protein (CHOP) downstream of PERK13. IRE1α signaling is the most conserved and well-studied UPR. In response to ER stress, the endoribonuclease activity of the IRE1α RNase domain is activated by conformational changes12. IRE1α RNase processes the mRNA encoding the transcription factor X-box binding protein 1 (XBP1), excising a 26-nucleotide intron in the XBP1 mRNA12. This splicing event shifts the coding reading frame, leading to the translation of a transcription factor termed spliced XBP1 (XBP1s). XBP1s translocates into the nucleus to promote the transcription of genes involved in protein folding among other targets to alleviate ER stress12. However, the mechanism that controls the IRE1α/XBP1s pathway remains poorly understood. For example, how XBP1s target genes were regulated at the molecular and genome-wide levels is not clear.

In addition to tumor-intrinsic function, the ER stress response such as the IRE1α/XBP1s pathway is implicated in intratumoral immune cells10,11. For example, activation of the IRE1α/XBP1s pathway is known to be immune suppressive in various populations of immune cells10. Targeting the ER stress response may reinvigorate endogenous antitumor immunity, which could synergize with immunotherapies, such as immune checkpoint blockade10. Thus, there are substantial ongoing efforts in exploring the pathway as a cancer therapeutic target. However, the molecular determinant of therapeutic response to the inhibition of the IRE1α/XBP1s pathway remains poorly understood.

Here we show that pharmacological targeting of the IRE1α/XBP1s pathway selectively suppresses CARM1-expressing ovarian cancer, which further synergizes with immune checkpoint blockade. Mechanistically, CARM1 determines ER stress response through controlling the IRE1α/XBP1s pathway by forming a complex with XBP1s to regulate its target gene expression.

CARM1 determines XBP1s target gene expression

To systematically profile CARM1 distribution pattern genome-wide, we performed the cut-and-run analysis for CARM1 in CARM1-expressing A1847 HGSOC cells9. The analysis revealed 22,398 significant peaks (false discovery rate (FDR) < 5% with at least fourfold over input). Overall distribution of CARM1 peak signal coincided with active histone mark H3K27ac binding and proximity to transcription starting sites (TSS) (Fig. 1a), which is consistent with the reported role of CARM1 as a transcriptional activator4. Notably, 25% of all CARM1-binding sites overlapped with genes’ TSS with 7859 genes occupied by CARM1 at the promoter (Fig. 1b). Analysis of enrichment of known regulators for those genes revealed MYC as the top regulator, which is consistent with previous reports that CARM1 regulates the c-MYC pathway3. Notably, XBP1s was identified as a top predicted regulator of the identified putative CARM1 target genes (Fig. 1c). Indeed, de novo motif analysis of 100 bp regions around TSS of CARM1-bound genes by HOMER revealed a significant enrichment of ACGTCA motif (P = 10−101) matching the core of known XBP1s-binding motif (Fig. 1d). This raised the possibility that CARM1 may regulate XBP1s function during ER stress response. To test this possibility, we performed XBP1s and CARM1 cut-and-run analysis with or without ER stress-inducer tunicamycin treatment (Supplementary Fig. 1a). The analysis revealed that CARM1- and XBP1s-binding signal significantly correlated in tunicamycin treatment condition and both factors increased binding compared to vehicle control-treated cells (Fig. 1e–g). Thus, we conclude that CARM1 is associated with XBP1s target genes.

a Heatmap of cut-and-run profiles of CARM1, H3K27ac, and input signal. Binding sites are sorted by strength of CARM1 signal. 3 kb around CARM1 peak centers are shown with 100 60 bp bins. bGenomic-wide distribution of CARM1 peaks relative to gene. c Ingenuity Pathway analysis of known regulators enriched among 7859 genes bound by CARM1. Transcriptional regulators with activation prediction Z-score of at least 2 and P < 0.001 are shown. The P value in Ingenuity Pathway analysis measures whether there is a statistically significant overlap between dataset genes and the genes that are regulated by a regulator. It is calculated using two-tailed Fisher’s Exact Test. d De novo motif analysis of 100 bp TSS regions for 7859 genes bound by CARM1. e Heatmap of cut-and-run profiles of XBP1s, CARM1, and input signal for vehicle (V) or ER stress-inducer tunicamycin (Tu)-treated cells around center of XBP1s peaks. Binding sites are sorted by strength of XBP1s signal. 3 kb around XBP1 peak centers are shown with 100 60 bp bins. f Correlation of XBP1s and CARM1 binding signal in XBP1s peak regions in tunicamycin-treated cells. P value was calculated using two-tailed test of significance. g Average profiles of binding signals for the indicated samples as shown in f.

Full size image

We next sought to identify direct CARM1 and XBP1s target genes that are subjected to ER stress regulation. Toward this goal, we performed RNA sequencing (RNA-seq) in control, CARM1 knockout, and XBP1s knockdown A1847 cells treated with or without tunicamycin (Fig. 2a, b). The analysis revealed that 3313 genes were upregulated by tunicamycin in A1847 cells. XBP1s knockdown significantly downregulated response of 779 of the 3313 genes induced by tunicamycin. In addition, CARM1 knockout significantly downregulated response of 1722 of the 3313 genes induced by tunicamycin. Five hundred and forty-three of the 3313 genes were downregulated by both XBP1s knockdown and CARM1 knockout, which represents a statistically significant overlap (P < 10−10 by hypergeometric test; Fig. 2c, d). We next cross-referenced the 543 genes with XBP1s and CARM1 peaks identified in A1847 cells treated with ER stress-inducer tunicamycin to identify direct ER stress-induced XBP1s and CARM1 target genes. The analysis revealed that 363 of the 543 genes are directly associated with both XBP1s and CARM1 in their promoters (P < 10−10 by hypergeometric test; Fig. 2e, fand Supplementary Data 1). In addition, we observed a significant correlation between strength of XBP1s binding and the increase in CARM1 binding under ER stress condition induced by tunicamycin (Fig. 2g). These findings suggest that CARM1 regulates XBP1s’s association with its target genes during ER stress response.

We next sought to correlate expression of CARM1 with identified CARM1/XBP1s target genes under normal condition. Toward this goal, we cross-referenced RNA-seq data of control A1847 cells and CARM1 knockout or XBP1 knockdown A1847 cells to identify genes commonly regulated by CARM1 and XBP1s without ER stress induction. In addition, we examined the binding of both CARM1 and XBP1s on these genes based on the cut-and-run datasets. The analysis revealed 430 CARM1/XBP1s direct target genes that are downregulated by CARM1 knockout or XBP1 knockdown at the basal level (Supplementary Fig. 2a and Supplementary Data 2). We explored the correlation between CARM1 and the identified CARM1/XBP1s target genes in the TCGA HGSOC dataset. Indeed, there is a significant correlation between expression of CARM1 and these genes in the TCGA HGSOC dataset (Fig. 2h). A similar positive correlation was also obtained in the Broad Institute Cancer Cell Line Encyclopedia database (Supplementary Fig. 2b).

We validated the downregulation of the identified CARM1/XBP1s genes such as HSPA5 and DNAJB9 by CARM1 knockout using quantitative reverse-transcriptase (qRT-PCR) analysis in A1847 cells (Supplementary Fig. 2c). Similar downregulation of CARM1/XBP1s target genes by CARM1 knockout was also observed in CARM1-expressing PEO4 and OVSAHO cells (Supplementary Fig. 2d–g). Conversely, we sought to determine whether upregulation of CARM1 in CARM1-low cells upregulates CARM1/XBP1s target genes. Toward this goal, we specifically upregulated endogenous CARM1 expression using a CRISPR (clustered regularly interspaced short palindromic repeats)-mediated activation system (Fig. 2i)14. Indeed, CARM1 activation promoted CARM1/XBP1s target gene expression in both EF027 and COV362 cells (Supplementary Fig. 2h). These findings suggest that the observed effects are not cell line specific. Notably, the upregulation of these genes triggered by ER stress-inducer tunicamycin was also suppressed by CARM1 knockout (Fig. 2j and Supplementary Fig. 2i). Consistently, XBP1s reporter activity was significantly lower in CARM1 knockout cells compared with controls in response to tunicamycin treatment (Fig. 2k). In contrast, XBP1s reporter activity was significantly increased by CARM1 activation in tunicamycin-treated cells (Fig. 2l). Together, these results show that CARM1 determines the expression of XBP1s target genes during ER stress response.

We next validated the binding of both CARM1 and XBP1s with CARM1/XBP1s target genes such as HSPA5 and DNAJB9 by chromatin immunoprecipitation (ChIP) analysis (Fig. 3a, b). Notably, CARM1 knockout significantly impaired the recruitment of XBP1s to these genes (Fig. 3b), which is consistent with the observed downregulation of CARM1/XBP1s target genes in CARM1 knockout cells (Fig. 2g). As a control, the level of H3R17me2a, a product of the CARM1’s enzymatic activity4, on the promoters of these genes was also decreased by CARM1 knockout (Fig. 3b). Conversely, we sought to determine whether XBP1s affects CARM1 recruitment to CARM1/XBP1s target genes during ER stress response. Indeed, XBP1 knockdown significantly impaired the recruitment of CARM1 to the promoters of the CARM1/XBP1s target genes, which correlated with a decrease in H3R17me2a on these genes’ promoters (Fig. 3c). Finally, sequential ChIP analysis using an anti-CARM1 antibody followed by an anti-XBP1 antibody showed a significant enrichment of CARM1/XBP1s on these genes’ promoters (Fig. 3d). Together, we conclude that XBP1s and CARM1 are mutually dependent upon each other for their optimal recruitment to the CARM1/XBP1s target genes.

CARM1 and XBP1s interact with each other

Given the observed mutual dependence of CARM1 and XBP1s for their optimal recruitment to the CARM1/XBP1s target genes in response to ER stress, we next determined whether CARM1 and XBP1s interact with each other. Indeed, co-IP revealed an interaction between CARM1 and XBP1s in response to tunicamycin treatment (Fig. 4a). The observed interaction was validated by glutathione S-transferase (GST) pulldown using GST-tagged CARM1 in lysates of A1847 cells treated with tunicamycin (Fig. 4b). We next sought to examine the interaction between XBP1s and CARM1 in situ by proximity ligation assay (PLA) in cells treated with or without tunicamycin. While we observed a basal level of interaction between XBP1s and CARM1 in control cells compared with negative immunoglobulin G (IgG) control, the interaction was significantly enhanced by ER stress-inducer tunicamycin (Fig. 4c, d). Together, we conclude that ER stress enhances CARM1 and XBP1s interaction, which correlates with the induced expression of XBP1s.

We next sought to map the CARM1 domain that interacts with XBP1s. GST-pulldown analysis of GST-tagged control or mutant CARM1 revealed that the catalytic domain of CARM1 (141–480) interacts with XBP1s in response to ER stress (Supplementary Fig. 3a, b). Indeed, ectopic CARM1 141–480 expression suppressed the expression of CARM1/XBP1s target genes (Fig. 4e, f and Supplementary Fig. 3c). This result suggests that the catalytic domain of CARM1 may function as a dominant-negative mutant in regulating CARM1/XBP1s target gene expression. Consistently, inhibition of CARM1 enzymatic activity does not affect the expression of CARM1/XBP1s target genes (Supplementary Fig. 3d, e). Notably, inhibition of CARM1 enzymatic activity does not affect the inhibition of IRE1α activity by B-I09 as determined by the ratio between total and spliced XBP1 mRNAs (Supplementary Fig. 3f). This result indicates that the catalytic activity of CARM1 is not required for the observed phenotype and instead the catalytic domain serves as an interaction domain with XBP1s. Together, we conclude that the catalytic domain of CARM1 mediates its interaction with XBP1s in an enzymatic activity-independent manner to regulate CARM1/XBP1s target gene expression.

CARM1 determines response to inhibition of the IRE1α/XBP1s pathway

Since we show that CARM1 promotes the IRE1α/XBP1s pathway, we next determined whether inhibition of the IRE1α/XBP1s pathway is selective against CARM1 expression. Toward this goal, we treated control and CARM1 knockout A1847 cells with a selective IRE1α inhibitor B-I0915. We chose B-I09 for our study because of its ability to specifically target IRE1α RNase activity and safety profile in vivo in preclinical studies15,16,17,18,19. Indeed, compared with controls, CARM1 knockout increased the IC50 of B-I09 in A1847 cells (Fig. 5a, b). Likewise, CARM1 knockout reduced the growth rate inhibition by B-I09 (Supplementary Fig. 4a, b)20. Consistent with the notion that unresolved ER stress leads to apoptosis, B-I09 treatment upregulated apoptotic markers such as cleaved lamin A and cleaved poly ADP-ribose polymerase (PARP) p85 in A1847 cells (Fig. 5c). As a control, B-I09 failed to induce apoptotic markers in CARM1 knockout A1847 cells (Fig. 5c). Similar increase in IC50 was also observed for another IRE1α inhibitor 4μ8c (Supplementary Fig. 4c). Likewise, CARM1 knockout increased the IC50 of IRE1α inhibitor in PEO4 and OVSAHO cells (Supplementary Fig. 4d–f). Conversely, endogenous CARM1 upregulation decreased the IC50 of B-I09 in CARM1-low cell lines (Fig. 5d). Notably, B-I09 is equally effective in suppressing IRE1α enzymatic activity in these cells as determined by the ratio between XBP1s and XBP1 in these cells (Supplementary Fig. 4g). Finally, we determined the IC50 of B-I09 in a panel of HGSOC cell lines. The IC50 of B-I09 was significantly lower in CARM1-high cell lines compared with those cell lines with low CARM1 expression (Fig. 5e–g). Consistently, in the Project Achilles synthetic lethality database21, XBP1 short hairpin RNA (shRNA) was more effective in suppressing the growth of cell lines with high CARM1 expression compared with those with low CARM1 expression (Supplementary Fig. 4h, i). Notably, the observed differences in response to B-I09 in the panel of cell lines did not correlate with c-MYC expression (Fig. 5e). Consistently, CARM1 knockout did not affect c-MYC expression (Supplementary Fig. 4j). Notably, inhibition of the enzymatic activity of CARM1 did not affect the sensitivity to B-I09 (Supplementary Fig. 4k), which is consistent with the finding that inhibition of CARM1 enzymatic activity does not affect the expression of CARM1/XBP1s target genes (Supplementary Fig. 3d, e). In addition, the observed effects were not due to variation in the expression of XBP1, XBP1s, or the ratio between XBP1s and XBP1 among these cell lines (Supplementary Fig. 4l). Likewise, CARM1 knockout did not affect XBP1s induction in response to tunicamycin treatment (Supplementary Fig. 4m). Conversely, CARM1 activation did not affect either IRE1α expression or XBP1s induction in response to tunicamycin (Supplementary Fig. 4n). These findings support the notion that CARM1-regulated recruitment of XBP1s to its target genes instead of its expression dictates the response to IRE1α inhibitor B-I09. Together, we conclude that inhibition of the IRE1α/XBP1s pathway is selective against CARM1 expression in HGSOC cells.

a Representative images of colonies formed by control and CARM1 knockout A1847 cells treated with or without IRE1α inhibitor B-I09 at the indicated concentrations. b B-I09 dose-responsive curves based on the colony-formation assay for control and CARM1 knockout A1847 cells. cExpression of apoptotic markers cleaved PARP and cleaved Lamin A and a loading control β-actin in control and CARM1 knockout A1847 cells treated with or without B-I09 at the indicated concentrations determined by immunoblot. Immunoblots are representative of three biologically independent experiments with similar results. d B-I09 dose-responsive curves based on colony-formation assay of the indicated control and CRISPR-mediated activation of endogenous CARM1-expressing COV362 and EF027 cells. e Expression of CARM1, c-MYC, and a loading control β-actin in a panel of the indicated cell lines determined by immunoblot. Immunoblots are representative of two biologically independent experiments with similar results. f B-I09 dose-responsive curves based on colony-formation assay for the indicated cells. g B-I09 IC50 of CARM1-high (n = 3) and CARM1-low (n = 9) cell lines. h B-I09 dose-responsive curves based on colony-formation assay of the indicated control and shCHOP-expressing A1847 cells. i Expression of apoptotic markers cleaved caspase 3, cleaved PARP p85, and a loading control β-actin in control and shCHOP-expressing (#1) A1847 cells treated with or without B-I09 (5 μg/ml, 48 h). Immunoblots are representative of two biologically independent experiments with similar results. Data represent mean ± SEM, n = 4 biologically independent experiments unless otherwise stated. P values were calculated using two-tailed t test.

Full size image

Consistent with findings that B-I09 induces apoptosis in CARM1-expressing cells, the pro-apoptotic ER stress effector CHOP expression22 was induced by B-I09 in CARM1-expressing cells (Supplementary Fig. 4o, p). Indeed, CHOP knockdown significantly decreased the sensitivity of B-I09 in CARM-expressing cells (Fig. 5h and Supplementary Fig. 4q), which correlates with the suppression of markers of apoptosis (Fig. 5i). These results support that CARM1 coordinates pro-apoptotic vs. pro-survival ER stress response and that CHOP signaling contributes to apoptosis induced by B-I09 in CARM1-expressing cells.

IRE1α inhibition suppresses the growth of CARM-high EOCs in vivo

To determine the effects of IRE1α inhibition on CARM1-high EOCs in vivo, we utilized three different models. In the orthotopic model, the tumors established by CARM1-expressing A1847 cells in mouse bursa that covers mouse ovary were treated with vehicle control or B-I09 for 2 weeks (Supplementary Fig. 5a). At the end of treatment, we used tumor weight as a surrogate for tumor burden. Compared with vehicle control, B-I09 significantly reduced the tumor burden in the orthotopic A1847 xenograft model (Fig. 6a, b). This correlated with a significant improvement of survival of mice bearing CARM1-expressing A1847 tumors (Fig. 6c). As a control, B-I09 did not significantly reduce the tumor burden in the orthotopic model established by CARM1 knockout A1847 cells (Supplementary Fig. 5b). Likewise, B-I09 significantly reduced the tumor burden in CARM1-high, but not in CARM1-low, HGSOC patient-derived xenograft (PDX) models (Fig. 6d, e and Supplementary Fig. 5c, d). Notably, B-I09 was well tolerated. For example, B-I09 treatment did not significantly decrease the body weight of the treated mice, indicating that the treatment was not toxic (Supplementary Fig. 5e).

a, b A1847 cells were unilaterally injected into the ovarian bursa sac of immunocompromised mice. Tumor-bearing mice were randomized and treated with vehicle control or B-I09 (50 mg/kg, 5 weekdays/week by i.p.) for 2 weeks. After treatment, reproductive tracts with tumors from the indicated treatment groups were dissected (a). And tumor weights were measured as a surrogate for tumor burden (b). n = 6 mice per group. c After stopping the treatment, the mice were followed for survival. Kaplan–Meier survival curves for the indicated groups (n = 5 mice per group). P value was calculated by log-rank test. d Expression of CARM1 and a loading control β-actin in the indicated CARM1-high and CARM1-low PDXs determined by immunoblot. Immunoblots are representative of two biologically independent experiments with similar results. A1847 and OVCAR3 cells were used as controls for CARM1-high and CARM1-low expression, respectively. e, fMice bearing the indicated orthotopic CARM1-high and CARM1-low HGSOC PDXs were randomized and treated with vehicle control or B-I09 (50 mg/kg, 5 weekdays/week by i.p.) for 2 weeks. Reproductive tracts with tumors from the indicated treatment groups were dissected (e). And tumor weights were measured (f). g Tumors from CARM1-high group were subjected to immunological staining for an apoptotic marker cleaved caspase 3 and a cell proliferation marker Ki67. Scale bar = 100 μm. h Histological score (H score) of the indicated markers was quantified from three separate fields of each tumor, six tumors from vehicle group, and five tumors from B-I09 group were analyzed. Data represent mean ± SEM. P values were calculated using two-tailed t test except for c by log-rank test.

Full size image

We next sought to correlate the observed tumor-suppressive effects in vivo with the mechanism we have characterized in vitro. Toward this goal, we examined the expression of a cell proliferation marker Ki67 and an apoptosis marker cleaved caspase 3. Indeed, B-I09 treatment significantly decreased Ki67 expression, while it increased cleaved caspase 3 expression in CARM1-high tumors (Fig. 6g, h). Together, these data support that IRE1α inhibitor B-I09 suppresses CARM1-high HGSOC in vivo, which correlates with the induction of apoptosis and suppression of cell proliferation.

IRE1α inhibition synergizes with immune checkpoint blockade

In addition to promoting survival of cancer cells, ER stress response such as the IRE1α/XBP1s pathway plays a critical role in eliciting an immune-suppressive tumor microenvironment11. This raises the possibility that inhibition of the IRE1α/XBP1s pathway may synergize with immune checkpoint blockade therapy10. To test whether IRE1α inhibition is synergistic with immune checkpoint blockade, we utilized a syngeneic immune-competent mouse model using UPK10 mouse ovarian cancer cell line23 that expresses CARM1 at a comparable level as those observed in A1847 cells (Fig. 7a). The orthotopically transplanted tumors were randomized into the following treatment groups: Vehicle/IgG, B-I09/IgG, Vehicle/anti-PD1, and B-I09/anti-PD1 (Fig. 7b). Consistent with our findings from orthotopic xenograft and PDX models, B-I09 treatment significantly reduced the tumor burden in the syngeneic model as a single agent (Fig. 7c, d). Notably, a combination of B-I09 and anti-PD1 was significantly more effective in reducing the tumor burden compared with either one of the individual treatments (Fig. 7c, d). Consistently, the combination significantly improved the survival of tumor-bearing mice compared with each of the individual treatments (Fig. 7e). As a control, B-I09 treatment did not further reduce the burden of tumors formed by CARM1 knockout UPK10 cells treated with anti-PD1 (Supplementary Fig. 6a, b). B-I09 treatment significantly decreased Xbp1s levels in both tumors and sorted T cells, indicating that B-I09 affects both tumors and the immune cells in the tumor microenvironment (Supplementary Fig. 6c–f). Indeed, immune cell profiling revealed that B-I09 significantly increased the intra-tumoral infiltration of B cells and CD4 T cells (Fig. 7f and Supplementary Fig. 6g). Although CD8 T cell infiltration was not affected by B-I09 treatment, activation of both CD4 and CD8 T cells was increased by B-I09 treatment (Fig. 7f). Consistent with previous reports19,24, B-I09 treatment decreased both intra-tumoral and spleen monocytic myeloid-derived suppressor cell (MDSC) and polymorphonuclear MDSC (Fig. 7f and Supplementary Fig. 6h). The combination treatment did not significantly decrease the body weight of the treated mice, and histological analyses of the liver, kidney, and spleen harvested at the end of the experiment from combination-treated mice revealed no overt abnormalities (Supplementary Fig. 6i, j), suggesting that the combination treatment was well tolerated. Together, we conclude that inhibition of the IRE1α/XBP1s pathway is synergistic with immune checkpoint blockade in CARM1-expressing EOCs in vivo.

Hyperactivation of the ER stress response in cancer cells is a therapeutic vulnerability that is actively explored25. However, the mechanisms underlying hyperactivation of the ER stress response are not fully understood. We show that CARM1 functions a co-activator of XBP1s in determining the expression of the IRE1α/XBP1s pathway target genes. Thus, our data support a model whereby CARM1 determines the ER stress response by controlling XBP1s’s association with its target genes. Notably, it has been reported that XBP1s controls the hypoxia-inducible factor-1α (HIF1α) transcriptional program in triple-negative breast cancer where XBP1s functions upstream of HIF1α26. Our results identify CARM1 as a critically important co-activator that functions upstream of XBP1s to drive the IRE1α/XBP1s pathway.

CARM1 is a well-studied transcriptional co-activator through its enzymatic substrates, most notably epigenetic regulators3,4. We show that CARM1 interacts with XBP1s through its catalytic domain. However, the enzymatic activity of CARM1 is not required for the expression of CARM1/XBP1s target genes. Thus, CARM1 may function as a scaffold factor to facilitate XBP1s recruitment/association with its target genes. In addition to functioning as a co-activator of gene transcription, CARM1 indirectly promotes the silencing of the EZH2/polycomb repressive complex 2 (PRC2) target genes by regulating the antagonism between PRC2 and the SWI/SNF chromatin remodeling complex3,9,27. Thus, CARM1 promotes cancer by both directly enhancing the activation of the IRE1α/XBP1s pathway and indirectly mediating the silencing of EZH2 target genes. One limitation of our studies is that, in addition to the IRE1α/XBP1s pathway, CARM1 may contribute to the activation of additional pathways including the c-MYC pathway3. Regardless, future studies will examine the relative contribution and cooperation among these CARM1-regulated pathways.

In summary, our results show that CARM1-expressing HGSOCs are hypersensitive to inhibition of the IRE1α/XBP1s pathway alone or in combination with immune checkpoint blockade. These findings suggest that CARM1 status could serve as a biomarker to predict response to inhibitors of the IRE1α/XBP1s pathway. CARM1 amplification/overexpression occurs in ~20% of HGSCOs and is mutually exclusive with deficiency in homologous recombination caused by genetic alterations such as BRCA1/2 mutations9. Thus, therapeutic approaches for CARM1-expressing HGSOCs are an unmet clinical need because these patients are unlikely to benefit from either standard-of-care platinum-based chemotherapy or emerging PARP inhibitors. Our findings indicate that inhibition of the IRE1α/XBP1s pathway alone or in combination with immune checkpoint blockade represents a therapeutic strategy for a number of cancer types with frequent CARM1 overexpression including HGSOCs6,7,8. Therefore, our findings will have broad implications in developing therapeutic approaches with precision based on CARM1 expression status.