Dual G9A/EZH2 Inhibition Stimulates Antitumor Immune Response in Ovarian High-Grade Serous Carcinoma

Abstract Ovarian high-grade serous carcinoma (HGSC) prognosis correlates directly with presence of intratumoral lymphocytes. However, cancer immunotherapy has yet to achieve meaningful survival benefit in patients with HGSC. Epigenetic silencing of immunostimulatory genes is implicated in immune evasion in HGSC and re-expression of these genes could promote tumor immune clearance. We discovered that simultaneous inhibition of the histone methyltransferases G9A and EZH2 activates the CXCL10–CXCR3 axis and increases homing of intratumoral effector lymphocytes and natural killer cells while suppressing tumor-promoting FoxP3+ CD4 T cells. The dual G9A/EZH2 inhibitor HKMTI-1–005 induced chromatin changes that resulted in the transcriptional activation of immunostimulatory gene networks, including the re-expression of elements of the ERV-K endogenous retroviral family. Importantly, treatment with HKMTI-1–005 improved the survival of mice bearing Trp53−/− null ID8 ovarian tumors and resulted in tumor burden reduction. These results indicate that inhibiting G9A and EZH2 in ovarian cancer alters the immune microenvironment and reduces tumor growth and therefore positions dual inhibition of G9A/EZH2 as a strategy for clinical development.


Introduction
Despite ample evidence that the prognosis of patients with advanced ovarian high-grade serous carcinoma (HGSC) is strongly influenced by the immune microenvironment (1), current immunotherapies have failed to produce a meaningful survival benefit for patients (2). HGSC cells can evade immune responses by altering their epigenome, and targeting ovarian cancer epigenetics can reactivate cancer testis antigens (3), induce viral mimicry (4), and alter the tumor immune microenvironment and immune cell function (5). DNA methylation and histone deacetylation are two mechanisms that play a role in cancer immune evasion (6), and, although inhibitors of both DNA methylation and histone deacetylation are currently used in some hematologic malignancies, their use in solid malignancies has been limited due to toxicity and limited efficacy (7). More recently, histone methylation mediated by both G9A and EZH2 has been identified as an important pathway that influences the immune system in ovarian cancer and multiple other malignancies, including multiple myeloma and hepatocellular carcinoma (8)(9)(10)(11)(12).
Increased levels of the chemokines CXCL9, CXCL10, CXCL11, and CCL5 are all associated with an immune-reactive ovarian cancer microenvironment and improved patient prognosis (13). CXCL9, CXCL10, and CXCL11 are IFN-inducible and bind to CXCR3. The Cancer Genomic Atlas (TCGA) Research Network identified a subgroup of patients with HGSC with an activated CXCR3/CXCL9-11 pathway (14). Critically, when these chemokines are present at high concentrations within tumors, patients achieve a longer disease-free interval and overall survival (15). The primary role of these IFNginducible chemokines is trafficking of activated CD8 þ , CD4 þ T cells, and natural killer (NK) cells. In preclinical models of ovarian cancer, increased expression of CXCL10 can reduce tumor burden and ascites accumulation (16). CCL5 is also associated with T-cell infiltration and tumor control in other carcinomas (17). Coukos and colleagues recently showed that CCL5 hi CXCL9 hi ovarian tumors are immunoreactive and responsive to immune checkpoint blockade, with tumorderived CCL5 driving expression of CXCL9 from intratumoral immune cells, such as antigen-presenting cells, which in turn supports T-cell engraftment in the tumor (18). We reasoned that pharmacologic approaches to activate the CXCR3/CXCL9-11 pathway might be of therapeutic benefit in ovarian cancer.
Using a medium-throughput screening library of epigenetic compounds, we sought to discover epigenetic mechanisms that can augment immune responses in HGSC. We discovered that dual inhibition of G9A and EZH2 histone lysine methyltransferases induces potent release of lymphocyte chemotactic chemokines, including CXCL9, CXCL10, CXCL11, and CCL5, confirming these results in a panel of human cell lines and primary patient samples. We also showed that the dual G9A/EZH2 inhibitor HKMTI-1-005 (19) powerfully modified accessible chromatin in a syngeneic HGSC model, accompanied by transcriptional upregulation of immune pathways and, critically, substantial modulation of the tumor immune microenvironment. Importantly, we describe how G9A/EZH2 inhibition generated a significant influx of effector CD8 þ T cells, NK cells, activated conventional type 1 dendritic cells (cDC1) while depleting tumors of CD4 þ T regulatory cells. We observed a significantly extended survival of mice treated with HKMTI-1-005, indicating that G9A/EZH2 inhibition may provide a useful tool to overcome the poor immune reaction to ovarian cancer in patients.

Ethics statements
All

In vivo syngeneic mouse model of ovarian cancer and cell lines
We utilized the ID8 syngeneic murine model (20) with bi-allelic Trp53 deletions that we described previously (21). A total of 5 Â 10 6 Trp53 À/À ID8 cells/mouse were injected intraperitoneally in 6-weekold C57BL/6J female mice. At defined endpoint, ascites, intraabdominal tumors (formed in omentum and porta hepatis) and spleens were collected ( Supplementary Fig. S1). When no ascites was present, peritoneal cells were collected by lavage with 5 mL PBS. For survival experiments, humane endpoints included weight loss of 20% or more, ascites equivalent to full term pregnancy, reduced/slow activity, pale feet, and visible symptoms of distress such as hunching, piloerection, closed eyes, and isolation from cage mates.
HKMTI-1-005 (patent WO/2013/140148; ref. 19) was synthesized at Imperial College London and converted to HCl salt, which was used in the biological experiments. It was dissolved in DMSO for long term storage and reconstituted in 1% Tween/0.9% NaCl vehicle, just prior to injection, and given as twice daily intraperitoneal injections of 20 mg/kg. Mice were randomly assigned to a 2-week treatment with HKMTI-1-005 or vehicle alone (control) starting on day 21 following intraperitoneal cell inoculation. The investigators deciding on endpoint were blinded to the treatment administered.

Gene expression assays
RNA was extracted from cells with the Qiagen RNeasy protocol (Catalog No. 74004). Quality control and quantification were performed using the NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). RNA was aliquoted and stored at À80 C. cDNA synthesis was performed using high-capacity cDNA Reverse Transcription Kit from Thermo Fisher Scientific (4368814) and iTaq/universal probes mastermix (Bio-Rad, 1725131) was used for single-gene RT-qPCR reactions (Supplementary Table S2). Chemokine expression was quantified using RT 2 Profiler PCR array for mouse chemokines/cytokines (Qiagen PAMM-150ZA, 330231) with data analysis performed using the Qiagen online tool PCR Array data analysis Web portal (https://www.qiagen.com/gb/resources/ resourcedetail?id¼20762fd2-8d75-4dbe-9f90-0b1bf8a7746b&lang¼en). The chemokines tested, quality control, and normalization analysis are found in Supplementary Tables S3 to S5.

Human ascites
After sterile collection, spheroids were captured on a 40 mm membrane and placed into T75 ultra-low attachment flasks (ULA, Corning, 3814) and cultured in advanced DMEM/F12 medium (Life Technologies, 12634010), supplemented with 10% autologous ascites, 10 mmol/L HEPES, 1Â N-2 supplement (Thermo Fisher Scientific, 17502048), 1Â serum-free B-27 supplement (Thermo Fisher Scientific, 17504044), 100 U/mL penicillin plus 100 mg/mL streptomycin (penicillin/streptomycin, Thermo Fisher Scientific, 15140-122), and 2 mmol/L L-glutamine (Thermo Fisher Scientific, 25030 -081 ). Spheroids were allowed to grow for up to 72 to 96 hours after which they were dissociated and treated as a monolayer. Cells were stained for PAX8 and sequenced for TP53 (Illumina Ampliseq) as detailed in Supplementary Materials and Methods. Patient details and cell line data are given in Supplementary Table S6.
For intracellular assessment of T-cell activity, 20 Â 10 6 /mL tumor cells were plated in clear untreated U-bottom plates (SLS, 3879) after tumor digestion. After stimulation (PMA and ionomycin, eBioscience, 00-4970, 2 mL/mL, 1 hour), protein transport inhibitor cocktail (eBioscience, 00-4980, 2 mL/mL) was added. After 4 hours, cells were transferred to a V-bottom plate and stained for the membrane markers, viability dye Zombie Yellow, and fixed followed by intracellular staining using the intracellular staining permeabilization buffer (BioLegend, 421002). Samples were analyzed on a 3-LASER Cytek Aurora (Cytek Biosciences) cytometer and the software FlowJo 10.7.1. Only samples that reached a threshold of 200 events per sample were included in the quantitative analysis. The geometric mean fluorescence intensity (MFI) was calculated by subtracting by an average (minus FMO) fluorescence value from pooled samples from each individual test sample.

RNA and ATAC sequencing
Frozen mouse tumors (≤10 mg) were homogenized in a Precellys homogenizer using ceramic beads at 2,000 Â g for two pulses of 30 seconds. RNA was extracted from the lysate with the Qiagen RNeasy protocol (Catalog No. 74004). RNA with an RNA Integrity number (RIN) of >7 as measured in an Agilent 2200 TapeStation was used for downstream-sequencing analysis. Following ribosomal RNA depletion (NEBNext, E6350) from 250 ng total RNA, sequencing libraries were constructed using the NEBNext Ultra II Directional Library Prep Kit for Illumina (E7760S). Following QC (Agilent D5000 Screen Tape System) and quantification [Qubit dsDNA High-Sensitivity Assay Kit (Thermo Fisher Scientific, Q32854)], samples were sequenced [Nova6000 SP flow cell (Illumina) 50 bp PE, target 50 million read pairs per sample].
For the assay of transposase-accessible chromatin using sequencing (ATAC-seq), the published protocol Omni-ATAC (23) was optimized for Trp53 À/À ID8 tumors. 20 mg tumor deposits were homogenized with a glass Dounce homogenizer. The lysate was then mixed in an iodixanol concentration gradient (25%-29%-35% concentration gradient, iodixanol Sigma-Aldrich, D1556) and centrifuged in a swinging bucket at 4,000 Â g for 20 minutes. After centrifugation, 20,000 nuclei per sample were harvested from the nuclear band and treated with 100 nmol/L hyperactive transposase enzyme (Nextera Tagment DNA enzyme I #15027916) for 30 minutes at 37 C, shaking at 400 rpm, in an Eppendorf Thermomixer comfort incubator. The purified, transposed DNA was amplified using customized primers as published previously (24). After quality control (Agilent D5000 Screen Tape System), the amplified library was sequenced [Nova6000 S1 flow cell (Illumina), 50 bp PE].
Raw-sequencing reads were aligned to mouse genome version GRCm38.p4 (mm10) using the STAR aligner with default parameters (25). Raw counts were generated using the Rsubread package (26) and differentially expressed genes (DEG) were identified using the DESeq package (27). All analyses, statistical tests, and plots were generated in R version 3.3.3 unless specified otherwise. MultiQC was used to collate data across different programs (28). For functional annotation of DEGs, we used the Database for Annotation, Visualization and Integrated Discovery (DAVID) online Functional Annotation Tool (29) with access to Gene Ontology (GO; ref. 30), and KEGG (31) databases. For analysis of endogenous retroviruses (ERV), a mm10 annotation for mouse endogenous viral elements was obtained from the gEVE database (http://geve.med.u-tokai.ac.jp). During alignment, only primary alignments were taken into account, a method adapted by Haase and colleagues (32). For ATAC-seq methodology, the MACS2 tool was used to call peaks on all individual control and treatment samples (33). Immune cell composition was inferred using seq-ImmuCC (34)

Statistical analysis
Statistical analyses were performed in GraphPad Prism (v9.0.0). For mean comparisons between two groups, t test was used for populations with normal distribution and Mann-Whitney test for nonparametric distribution. One-way ANOVA was used for comparison of more than two groups. Matched-pair t test was used to compare mean values for patient samples. Log-rank test was used to compare differences in survival. When indicated, ROUT (Q ¼ 1%) method was used to identify outliers.

Combined G9A/EZH2 inhibition upregulates chemotactic chemokines in vitro
We initially screened 38 epigenetic drugs for CXCL10 production by IFNg stimulated Trp53 À/À ID8 (21) cells using ELISA. We wished to identify chromatin modifying drugs that could enhance IFNg-induced Cxcl10 transcription. Although no statistically significant increases were observed in an initial single concentration screen (Fig. 1A), a twodose re-screen was performed with 10 drugs that had caused a numerical increase in mean CXCL10 production and that covered a wide range of epigenetic targets (Fig. 1B). As G9A can cooperate closely with Enhancer of Zeste homolog 2 (EZH2; ref. 36), we combined UNC0642 with the EZH2 inhibitor UNC1999, the latter chosen based on prior published data (37) and drug availability. We also evaluated HKMTI-1-005 ( Fig. 2A), the first described dual G9A/EZH2 inhibitor (19), which, in contrast to other EZH2 inhibitors, has a peptide substrate competitive mechanism. The combination of UNC0642 and UNC1999 induced a greater increase of both Cxcl10 mRNA and CXCL10 protein than either drug alone  Fig. 2B). HKMTI-1-005 treatment also resulted in higher CXCL10 protein production than the individual inhibitors given alone (mean fold change 3.1 AE 0.03 vs. 1.9 AE 0.01 vs. 1.4 AE 0.01, P < 0.0001) and in combination (mean fold change 3.1 AE 0.03 vs. 2.2 AE 0.1, P < 0.0001, Fig. 2C).

Dual G9A/EZH2 inhibition alters transcription and chromatin conformation in vivo
We hypothesized that altered chromatin accessibility induced by G9A/EZH2 inhibition could explain the changes in gene expression. To investigate this, we used the ATAC-seq and RNA-seq on tumors harvested after 14 days of HKMTI-1-005 treatment. ATAC-seq showed more peaks representing areas of open chromatin in the HKMTI-1-005-treated samples compared with controls ( Fig. 3A), with most peaks in intergenic regions (58.4%). Approximately 33% of peaks were intronic and 6.5% were in promoter regions (Fig. 3B). Peaks were present in genes involved in the activation pathways for Cxcl9, Cxcl10, and Ccl5, including Stat1, Irf1, NF-kB p105 subunit (Nfkb1), and inhibitor of NF-kB kinase subunit b (Ikbkb). We found a statistically significant overlap of 1,106 genes in common between DEG identified by RNA-seq and those in an euchromatin state identified by ATAC-seq (Fig. 3C). Among these were the Toll-like receptor Tlr13 (Log 2 FC 2.9, FDR ¼ 7.42eÀ19), the IFN pathway mediator Stat1 (Log 2 FC 1.01, FDR
Purple color: Log 2 FC ≥ 1, gray color: À1 < Log 2 FC ≤ 1, and blue color: Log 2 FC ≤ À1. D, Biological processes (BP) subontology for 1,053 genes from C that overlapped with gene expression signatures from DAVID online Functional Annotation Tool. Gene count denotes the number of genes found to overlap with genes within the respective signature and the dot size represents the percentage of these genes within the signature. FE, fold enrichment. E and F, DEG following HKTMI-1-005 classified by BP and KEGG sub-ontologies, respectively. Gene count denotes the number of genes found to overlap with genes within the respective signature and the dot size represents the percentage of these genes within the signature. G, Volcano plot showing differentially expressed ERVs, following HKMTI treatment (n ¼ 7) versus control (n ¼ 7 cytokine-cytokine receptor interaction (mmu04660, FE 2.6, FDR ¼ 2.51eÀ06; Fig. 3F).

Dual G9A/EZH2 inhibition prolongs survival in vivo
We next wanted to understand if modulating chromatin accessibility and stimulating gene expression, most importantly of chemokines associated with T-and NK-cell infiltration, could alter tumor growth and the immune response in vivo (Fig. 4A and B). HKMTI-1-005 treatment resulted in a prolongation of median survival (48 days vs. 54.5 days, P < 0.0001; HR, 0.33; 95% CI, 0.17-0.64; Fig. 4C), a reduction of tumor weight at the end of treatment (135 mg AE 5.2 vs. 108 mg AE 5.6, P ¼ 0.001, Fig. 4D) and completely abrogated the development of ascites in this model (602 mL AE 297 mL vs. 0 mL, P ¼ 0.0012, Fig. 4E). There were no toxicity signals observed throughout treatment and no significant weight difference between treatment groups (Supplementary Fig. S5).

Discussion
The drivers of the immune microenvironment in ovarian cancer remain unclear, although the extent of immune cell infiltration is strongly prognostic (13). Here we investigated whether modulation of epigenetic pathways could augment immune cell infiltration in HGSC, the commonest subtype of ovarian cancer in light of previous data suggesting that epigenetic mechanisms could underpin immune evasion in ovarian cancer (4,43).
Using a mouse model that faithfully reproduces HGSC peritoneal dissemination, established cell lines, and ascites-derived primary cell cultures, we show that dual blockade of the histone methyltransferases G9A and EZH2 reprogrammed the immune TME and activated the  Dual G9A/EZH2 inhibition inhibits tumor growth and prolongs survival in a mouse ovarian cancer model. A, Experimental design for mechanism experiments. Mice bearing intraperitoneal Trp53 À/À ID8 cells were treated with either HKMTI-1-005 (20 mg/kg i.p. twice a day) or vehicle (1% Tween/3.4% DMSO in 0.9% NaCl i.p. twice a day) for 14 days starting on day 21, followed by omental and porta hepatic deposits harvest/weighting, measuring ascites and immunophenotyping by flow cytometry immediately after the end of treatment. Image created with BioRender.com. B, Experimental design for efficacy experiments. Mice were treated as per A, but treatment was followed by observation until mice reached humane survival endpoint. C, Kaplan-Meier survival curves for mice treated with vehicle (n ¼ 24) or 20 mg/kg HKMTI-1-005 (n ¼ 24) as per schedule on B. Median survival was 48 days for vehicle versus 54.5 days for HKMTI-1-005, P < 0.0001). Curves were compared using the Log-rank (Mantel-Cox) test, ÃÃÃÃ , P < 0.0001. Experiment was performed twice with n ¼ 12 per cohort for each experiment. D, Whole tumor weight (including both porta hepatis and omental tumor deposits) and E ascites volume for mice treated with either vehicle (n ¼ 20) or HKMTI-1-005 (n ¼ 20) as per schedule in A; comparisons were made using unpaired t test for whole tumor burden and Mann-Whitney test for ascites volume ( ÃÃ , P < 0.01). Experiment was performed twice with n ¼ 10 per cohort for each experiment.   Figure 5. Dual G9A/EZH2 inhibition changes intratumoral immune cell composition in a mouse ovarian cancer model. A, Quantitative result for NK cells (CD3 À DX5 þ ) cells in porta hepatis deposits with vehicle (n ¼ 7) versus HKMTI-1-005 (n ¼ 8) treatment and in the omental deposits in mice treated with vehicle (n ¼ 9) versus HKMTI-1-005 (n ¼ 8) treatment from mice bearing intraperitoneal Trp53 À/À ID8 cells were treated with either HKMTI-1-005 or vehicle as per Fig. 4A. Significance was tested using an unpaired t test. B, Percentage of effector CD8 þ cells (CD44 þ CD62L À ) within the total CD8 þ population in porta hepatis deposits and omental deposits in mice treated with vehicle (n ¼ 6 and 8, respectively) versus HKMTI-1-005 (n ¼ 7 and 8, respectively) treatment. Significance was tested using an unpaired t test. C, Percentage of na€ ve CD8 þ (CD44 À CD62L þ ) within the total CD8 þ population in porta hepatis deposits and omental deposits in mice treated with vehicle (n ¼ 6 and 8, respectively) versus HKMTI-1-005 (n ¼ 7 and 8, respectively) treatment. Significance was tested using an unpaired t test (porta hepatis) and Mann-Whitney test (omental deposit). D, Representative contour plot from one of the omental deposits for effector and na€ ve CD8 þ cells, in mice treated with vehicle versus HKMTI-1-005 treatment. E, Percentage of granzyme-B (GZMB þ ) CD8 þ cells, following stimulation, within the total CD8 þ population in porta hepatis deposits and omental deposits from mice treated with vehicle (n ¼ 9 and 8, respectively) versus HKMTI-1-005 (n ¼ 9 and 8, respectively) treatment. Statistical significance was tested by unpaired t test. F, Representative contour plot showing (GZMB þ ) CD8 þ cells from one of the omental deposits from E. G, Quantitative result for T regulatory CD4 þ cells (FoxP3 þ CD4 þ ) cells in porta hepatis deposits and omental deposits with vehicle (n ¼ 9 and 8, respectively) versus HKMTI-1-005 (n ¼ 9 and 8, respectively) treatment. Unpaired t test (porta hepatis deposits) and Mann-Whitney test (omental deposits). H, CXCR3 MFI on CD3 þ (n ¼ 9 vehicle and n ¼ 8 HKMTI-1-005), CD4 þ (n ¼ 6 vehicle and n ¼ 6 HKMTI-1-005), CD8 þ (n ¼ 6 vehicle and n ¼ 7 HKMTI-1-005), and NK cells (n ¼ 7 vehicle and n ¼ 8 HKMTI-1-005) in the porta hepatis deposits. Statistical significance was tested by unpaired t-test. I, CXCR3 MFI on CD3 þ (n ¼ 9 vehicle and n ¼ 8 HKMTI-1-005), CD4 þ (n ¼ 9 vehicle and n ¼ 8 HKMTI-1-005), CD8 þ (n ¼ 8 vehicle and n ¼ 8 HKMTI-1-005), and NK cells (n ¼ 9 vehicle and n ¼ 8 HKMTI-1-005) in omental deposits. Unpaired t test. J, CD206 MFI on cDC1 dendritic cells (CD11b À MHCII þ CD11c þ ) in porta hepatis deposits (n ¼ 4 vehicle and n ¼ 1 HKMTI-1-005, statistics not performed as n < 3) and omental deposits (n ¼ 6 vehicle and n ¼ 5 HKMTI-1-005). Statistical significance was tested by unpaired t test. K, Ly6C þ macrophages (CD11b þ MHCII þ F4/80 þ ) in porta hepatis and omentum deposits with vehicle (both n ¼ 7) versus HKMTI-1-005 (n ¼ 7 and 8, respectively) treatment. Statistical significance was tested by the Mann-Whitney test. L, Representative flow cytometry plot with pseudocolour heatmap showing Ly6C þ macrophages from a representative omental deposit from K. cDC2 þ cells were subsequently gated on a CD11cþ ( ÃÃÃÃ , P < 0.0001; ÃÃÃ , P < 0.001; ÃÃ , P <0.01; Ã , P < 0.05; ns, nonsignificant). Error bars represent SEM. transcription of immune networks both in vitro and in vivo. Specifically, we identified accumulation of effector cytotoxic lymphocytes and NK cells, and reductions in immunosuppressive Treg CD4 þ cells. These changes were accompanied by a small but significant prolongation of survival in vivo. Furthermore, treatment also reduced the expression of the suppressive receptor CD206 on dendritic cells and macrophages, and blocked monocyte-to-macrophage differentiation in both the TME and peritoneal cavity. TAMs derive from the large population of CCR2 high Ly6C þ inflammatory monocytes that constantly contributes to the pool, and Ly6C expression gradually reduces as TAMs differentiate within tumors (44). HKMT-1-005 treatment increased the abundance of Ly6C þ macrophages, suggesting that this epigenetic modifier may impede the differentiation of the monocyte precursor pool into fully differentiated TAMs. The preclinical results presented here provide evidence that dual inhibition of G9A and EZH2 induces more robust chemokine induction than blockade of either methyltransferase alone. Recently, the co-dependence of EZH2 and G9A was established by Mozzetta and colleagues (36) and this has led to efforts to discover pharmacologic inhibitors that target both enzymes simultaneously, with HKMTI-1-005 being the described first (19). Curry and colleagues showed that treatment of the breast cancer cell line MDA-MB-231 with HKMTI-1-005 induced transcription of SPINK1, which did not occur when EZH2 or G9A were individually knocked down (19). The interplay between EZH2 and G9A in regulating CXCL10 transcription has also recently been observed in idiopathic pulmonary fibrosis by Coward and colleagues (45), further supporting our findings.
This work has generated interesting questions with regards to mechanism of action of G9A/EZH2 blockade that will need further investigation. First, our transcriptional and chromatin accessibility analyses were based on whole-tumor sequencing and therefore do not identify the cell type subjected to transcriptional modifications by HKMTI-1-005 treatment. Single-cell sequencing may help to delineate whether HKMI-1-005 acts primarily on tumor or immune cells in vivo. Second, the contribution of ERV-K retroelements to the immune responses following HKMTI-1-005 treatment warrants further exploration. Recent evidence suggests that ERVs can potentiate antitumor immunity when they are transcriptionally active (4,46,47) and that activation of evolutionary young elements is associated with innate immune responses (48). ERVK, an evolutionary young element, was activated following treatment with HKMTI-1-005 and, interestingly, antibodies against ERVK have been detected in the serum of patients with ovarian cancer (49). Similarly, our team has previously shown that expression of ERVK elements correlate with a transcriptome indicative of strong immune cell infiltration in TCGA ovarian carcinoma datasets (50) and that epigenetic manipulation of ERV expression by DNA methyltransferase inhibition can result in augmented immune cell killing of tumor cells in vitro. Recently, Steiner and colleagues mapped human ERVs at a locus-specific resolution, creating the platform whereby ERVs and their relation to immune response can be further explored in greater granularity (51). The downregulation of the macrophage receptor MARCO following HKMTI-1-005 treatment is also an intriguing finding; inhibiting MARCO reprogrammes macrophages to acquire an antitumor phenotype, inhibiting tumor cell growth (52). In the work presented here, we used a single ovarian cancer mouse model ID8, engineered with Trp53 À/À deletion, which represents the only universal genomic alteration in HGSC. This model faithfully recreates the intraabdominal dissemination of HGSC with widespread peritoneal and omental deposits and formation of ascites, as commonly observed in human disease. Moreover, the models that we generated are now in widespread use and have supported critical studies on the nature of immune cell composition in the TME (53,54). However, it will be important in future work to assess the influence of tumor genotype in the response to HKMTI-1-005 in light of recent data showing that BRCA1 deficiency drives inflammation that supports both immunoreactivity and immune resistance (55). However, we used a series of established human ovarian cancer cell lines that are representative of HGSC as well as primary ascites-derived cultures to reinforce our findings with the Trp53 À/À ID8 model.
Although the primary aim of our study was to investigate epigenetic regulation of the TME, one critical outstanding question is whether the increase in survival seen following HKMTI-1-005 treatment is driven by the changes in immune cell composition. Certainly, the doses utilized in vitro were noncytotoxic, but detailed evaluation would require depletion of multiple immune lineages as well as complex combination experiments that lie beyond the scope of this study. However, the results that we present support the hypothesis that dual blockade of G9A/EZH2 histone methyltransferases modulates the tumor immune microenvironment within the peritoneal cavity, confers a survival benefit in an aggressive murine model of HGSC and warrants further investigation towards clinical development.
Cancer Research UK/NIHR Imperial Experimental Cancer Medicine Centre, the Imperial College Healthcare Tissue Bank, and the CRUK Imperial and Glasgow Centres. The authors would also like to thank the LMS/NIHR Imperial Biomedical Research Centre Flow Cytometry Facility and the Biological Services and Histology Service at the CRUK Beatson Institute for their support. We also thank Ms Hiromi Kudo for assistance with immunocytochemistry. The authors would also like to acknowledge the support of Drs Seth Coffelt, Josephine Walton, and Catherine Winchester.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.