Adar1 loss decreases tumor burden and prolongs survival, which is further enhanced by DNMTi treatment

To test the hypothesis that Adar1 loss synergizes with DNMTi in vivo, we combined DNMTi and Adar1 knockdown in an aggressive model of OC. This murine OC model uses implantable Trp53^-/- ID8 ovarian carcinoma cells which are syngeneic to C57BL/6 mice and thus enable the study of an intact immune system.10 24 35 The CRISPR-engineered Trp53^-/- ID8 line is more representative of the human disease because TP53 mutations occur in over 90% of HGSOCs.35 46 Mice in this ID8 model develop hemorrhagic ascites (a buildup of malignant fluid in the peritoneum), and tumor nodules are disseminated throughout the peritoneal cavity, closely modeling advanced disease in humans.35 47 It has been previously shown that accumulation of ascites is a marker of advanced, disseminated OC in humans,47 and can be used as a surrogate measure of tumor burden.10

Mice were implanted with control tumors (shGFP in the Trp53^-/- ID8 cell line background) or Adar1 knockdown tumors (shAdar1 in the Trp53^-/- ID8 cell line background) and given either Mock or THU/DNMTi treatment (figure 1A). THU is a competitive inhibitor of cytidine deaminase, the enzyme which rapidly breaks down 5-Aza and decitabine/Dac in vivo.43 48 With this regimen, Dac and 5-Aza were alternated to exploit cross-priming mechanisms as shown in previous studies.48 49 This new THU/DNMTi treatment regimen offers significant survival advantage in the Trp53^-/- mouse model of OC (online supplemental figure 1A–C) compared with our previous treatment regimen (online supplemental figure 1D–F).10 24 Importantly, we began treating mice much later post-tumor implantation for the Adar1 studies (4 weeks compared with 1 week) to achieve a more clinically relevant model since patients with HGSOC are often diagnosed and treated at advanced stages of disease.46

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Adar1 loss decreases tumor burden and prolongs survival, which is further enhanced by DNMTi treatment and correlates with increases of CD8+ T cells and NK cells in the tumor microenvironment. (A) Schematic of mouse experiment. ID8 Trp53^-/- shGFP (control) or shAdar1 knockdown tumor cells were grown in tissue culture dishes and then 5e6 cells were i.p. injected into mice. Mock or THU/DNMTi treatment began around week 4 (advanced stage of disease), and around week 6, mice began developing ascites (a buildup of fluid in the peritoneum). (B) Survival curve. shGFP/Mock n=9 (one mouse was excluded: it did not develop ascites and was identified as an outlier using the ROUT method), shGFP/DNMTi n=10 (no mice excluded/censored), shAdar1/Mock n=9 (one mouse was excluded: it did not develop ascites and was identified as an outlier using the ROUT method), shAdar1/DNMTi n=7 (three mice were censored: one mouse was sacrificed due to a severe dermatitis condition, one mouse had an accidental/fatal fall, and one mouse was still alive and well at the end of the predefined, 20-week long study). One limitation of this model is that mice do not typically develop palpable tumors, thus making it difficult to know whether each mouse was effectively inoculated prior to start of treatment. (C) Ascites fluid drained at week 7 (attempted from all mice that were still alive from each group). Ascites is a sign of advanced stage of disease and ascites volume is an indicator of tumor burden. (D) Ascites fluid collection over time. (E–H) Immunophenotyping was performed on ascites drained from mice in figure 1A via flow cytometry. For flow analysis days, all mice that were still alive had an attempted ascites drain (despite body weight and circumference), and all samples obtained were analyzed. One limitation of this model is that ascites is required for flow analysis of the TME, and the mice in different groups do not develop ascites at the same time. To minimize potential batch effect, drain and flow stain days were chosen to maximize the number of samples obtained from two or more groups at one time. Due to difference in disease progression in each group, the shAdar1 Mock group (group 3) only resulted in two data points (two mice), which is too few data points to perform statistical analysis on. Therefore, all data are shown, although statistical comparisons have been excluded for group 3. Group 1: mice injected with shGFP (control) cells and given mock treatment (n=9). Group 2: mice injected with shGFP (control) cells and given THU/DNMTi treatment (n=8). Group 3: mice injected with shAdar1 (Adar1 knockdown) cells and given mock treatment (n=2). Group 4: mice injected with shAdar1 (Adar1 knockdown) cells and given THU/DNMTi treatment (n=5). (E) Absolute lymphocyte cell count (per mL of ascites). Absolute lymphocyte cell count (per mL of ascites) plotted against survival. (F) Absolute myeloid cell count per (per mL of ascites). Absolute myeloid cell count per (per mL of ascites) plotted against survival. (G) Absolute count of CD8+ T cells (per mL of ascites). Absolute count of CD8+ T cells (per mL of ascites) plotted against survival. (H) Absolute count of NK cells (per mL of ascites). Absolute count of NK cells (per mL of ascites) plotted against survival. Spearman’s correlation was performed for statistical significance on linear regression graphs. A one-way ANOVA was performed for statistical significance on box-and-whisker graphs. Gehan-Breslow-Wilcoxon method was performed for statistical significance on Kaplan-Meier survival curve. A one-way ANOVA was performed for statistical significance on column graph. *P<0.05. ANOVA, analysis of variance; DNMTi, DNA methyltransferase inhibitor; i.p., intraperitoneally; NK, natural killer; THU, tetrahydrouridine; TME, tumor microenvironment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

DNMTi treatment enhances secretion of pro-inflammatory chemokines from murine OC cells and increases migration of T cells. (A) ID8 Trp53^-/- Adar1 knockdown (shAdar1) and control cells (shGFP) were plated on day 0 and treated with DNMTi for three consecutive days without media change. Cell culture supernatant was collected on day 4 for chemokine/cytokine analysis with the BioLegend LEGENDplex Mouse Anti-Virus Response Panel. Chemokine/Cytokine levels were normalized by number of live cells. (C) Graphs show ratio of migrated CD4+ or CD8+ T cells to tumor cells. (D) Graphs show migration index, which is the fold change of DNMTi-treated migrated T cells to mock-treated migrated T cells (red and purple bars), or alternatively, the fold change of migrated T cells in the +CCL5 condition to the −CCL5 condition (gray bars). A one-way analysis of variance was performed for statistical significance. *P<0.05; **p<0.01; ***p<0.001. 5-AZA, 5-zaacytidine; DMSO, dimethyl sulfoxide; DNMTi, DNA methyltransferase inhibitor; IFN, interferon; OC, ovarian cancer.

Mice implanted with Adar1 knockdown cells and treated with DNMTi had significantly longer survival than all other groups (p<0.05) (figure 1B). Additionally, mice that were implanted with Adar1 knockdown tumors and received mock treatment lived significantly longer than both groups of mice that were implanted with the control (shGFP) tumor cells, with or without DNMTi. Average ascites volume was significantly decreased at week 7 for all groups compared with the group implanted with control (shGFP) cells and given mock treatment (figure 1C). Although the mice with Adar1 knockdown tumors did not present with ascites at week 7, most of these mice did develop ascites at later timepoints. Mice with Adar1 knockdown tumors had a reduced rate of ascites accumulation and a reduced total volume of ascites compared with the mice with control (shGFP) tumors (figure 1D). Using ascites as a surrogate marker of tumor burden,10 we thus conclude that overall tumor burden was decreased in mice implanted with Adar1 knockdown tumors compared with mice implanted with control (shGFP) tumors.

Combined DNMTi treatment and Adar1 loss increases lymphocytes in the tumor microenvironment

We hypothesized that the combined effect of Adar1 loss and DNMTi treatment on survival and tumor burden was due to changes in the immune microenvironment. To identify immune populations responsible for the reduction in tumor burden, we immunophenotyped cells in the ascites fluid from the four groups of mice (figure 1E–H, online supplemental table 1). When we analyzed absolute cell counts (per mL of ascites) of lymphocytes, we observed that lymphocytes significantly increased with Adar1 loss and DNMTi treatment (figure 1E). Additionally, the absolute lymphocyte count (per mL of ascites) positively correlated with survival (figure 1E). However, absolute cell counts of myeloid cells (figure 1F) did not significantly change, and there was no association between absolute myeloid cell count (per mL of ascites) and survival.

Combined DNMTi treatment and Adar1 loss increases CD8+ T cells and NK cells in the tumor microenvironment

When we performed a more comprehensive breakdown of the cellular composition of the ascites (figure 1G,H, online supplemental figure 1G–K), we found that absolute cell concentration (per mL of ascites) of CD8+ T cells (figure 1G) and NK cells (figure 1H) were significantly increased with the combination of Adar1 loss and DNMTi treatment (p<0.05). Additionally, survival positively correlated with the absolute number of CD8+ T cells (figure 1G) and the absolute number of NK cells (figure 1H). Except for IFN-γ+CD8+ T cells (online supplemental figure 1G), frequencies of lymphocyte subpopulations did not change significantly. Interestingly, although total numbers of CD8+ T cells were increased, the combination of Adar1 loss and DNMTi treatment significantly (p<0.05) decreased the percentage of CD8+IFN-γ+ T cells of total CD8+ T cells (online supplemental figure 1G). This reduction in CD8+IFN-γ+ T cell frequency was despite non-significant changes in the frequency of PD-1+CD8+ T cells (online supplemental figure 1G), suggesting that this reduction in functional CD8+ T cells was not due to exhaustion. We also did not observe any significant changes in frequency of NKT cells (online supplemental figure 1H), CD4+ T cells (online supplemental figure 1I) or memory/effector/naïve subsets of CD4+ T cells of CD8+ T cells (online supplemental figure 1J,K). Overall, the biggest impact of DNMT and Adar1 inhibition was a significant recruitment of lymphocytes to the TME. This change shifts the TME from immunosuppressive to more immune permissive, which is consistent with an increase in type I IFN signaling and release of downstream pro-inflammatory cytokines and chemokines.1 12

DNMTi enhances secretion of pro-inflammatory chemokines from murine OC cells and increases migration of T cells

We hypothesized that DNMTi-induced type I IFN signaling in tumor cells causes secretion of pro-inflammatory chemokines capable of inducing migration of lymphocytes to the tumor. To test this hypothesis, we treated ID8 Trp53^-/- control (shGFP) and Adar1 knockdown (shAdar1) cells in vitro with Mock (DMSO) or DNMTi (Dac, Aza) and assessed secreted chemokine levels (figure 2A,B).50 DNMTi-induced induced type I IFN signaling resulted in increased secretion of pro-inflammatory chemokines from tumor cells. 5-Aza treatment significantly (p<0.05) increased the levels of the chemokines CCL5 and CXCL10 in ID8 Trp53^-/- cells with Adar1 knockdown compared with control (shGFP) cells (figure 2B). Dac treatment significantly (p<0.05) increased the levels of these chemokines compared with mock-treated cells (online supplemental figure 2A). Overall, we conclude that the production of chemokines is largely driven by DNMTi rather than Adar1 loss alone or in combination with DNMTi.

To determine whether these levels of CCL5 and CXCL10 could induce T cell migration, we performed transwell chemotaxis/cell migration assays (online supplemental figure 2B). We treated Adar1 knockdown versus control ID8 Trp53^-/- cells with DNMTi for three consecutive days, and on day 4, performed a transwell chemotaxis/migration assay using activated, murine T cells from non-tumor bearing syngeneic mice (figure 2C,D, online supplemental figure 2C,D). The ratio of migrated CD4+ and CD8+ T cells to tumor cells was significantly higher (p<0.05) with DNMTi treatment irrespective of Adar1 status (figure 2C). Furthermore, ~2.5-fold more CD4+IFN-γ+ and CD8+IFN-γ+ T cells migrated in wells with DNMTi-treated ID8 Trp53^-/- cells (shGFP or shAdar1) compared with T cells in mock-treated wells (figure 2D). We thus conclude that DNMTi treatment increases secretion of inflammatory chemokines CCL5 and CXCL10, causing increased T cell migration relative to number of live tumor cells.

Combined Adar1 loss and DNMT inhibition significantly increases numbers of tumor-specific, cytotoxic CD8+ T cells

To understand the difference in cytotoxic lymphocytes in the TME and survival conferred by Adar1 knockdown, we assessed the cytotoxicity of CD8+ T cells from the different mouse groups in figure 1. We hypothesized that mice implanted with shAdar1 tumors receiving DNMTi treatment were generating more tumor-specific CD8+ T cells compared with the control mice (shGFP tumors with mock treatment). To test this hypothesis, we first independently assessed the effect of DNMTi treatment on CD4+ and CD8+ T cells isolated from spleens of healthy, non-tumor-bearing mice. After treating with DNMTi or mock for three consecutive days, we observed significantly higher concentrations of secreted IFN-γ in the cell culture supernatant (figure 3A). It has previously been shown that loss of methylation through genetic deletion of DNA methyltransferase enzymes in T cells,51 52 or DNMTi treatment of exhausted T cells in models of chronic viral infection,53 increases IFN-γ expression, which we confirm in our ex vivo model.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Combined Adar1 loss and DNMT inhibition significantly increases numbers of tumor-specific, cytotoxic CD8+ T cells. (A) CD4+ or CD8+ T cells were isolated from spleens of healthy, non-tumor-bearing mice and treated with mock or 5-Aza for 3 days. IFN-γ protein concentration was then assessed in the cell culture media supernatant. (B) Total CD8+ T cells isolated from spleens of shGFP mock-treated mice or shAdar1 THU/DNMTi-treated mice. (C,D) Co-culture cytotoxicity assay. Per cent (%) specific lysis measured at 4 hours from co-cultures of: (1) shGFP mock-treated tumor cells+CD8+ splenic T cells from the shGFP mock-treated mice or from the shAdar1 THU/DNMTi-treated mice; (2) shGFP DNMTi-treated tumor cells+CD8+ splenic T cells from the shGFP mock-treated mice or from the shAdar1 THU/DNMTi-treated mice; (3) shAdar1 mock-treated tumor cells+CD8+ splenic T cells from the shGFP mock-treated mice or from the shAdar1 THU/DNMTi-treated mice; and (4) shAdar1 DNMTi-treated tumor cells+CD8+ splenic T cells from the shGFP mock-treated mice or from the shAdar1 THU/DNMTi-treated mice. (D) Per cent (%) specific lysis measured at 48 hours of co-culture (same groups as in figure 3C). *P<0.05; **p<0.01; ***p<0.001. 5-AZA, 5-zaacytidine; DNMT, DNA methyltransferase; DNMTi, DNA methyltransferase inhibitor; IFN, interferon; ns, not significant; THU, tetrahydrouridine.

Next, we isolated CD8+ T cells from spleens of mice in the shGFP/Mock group and the shAdar1/DNMTi group and performed co-culture cytotoxicity assays with pretreated (mock or DNMTi), luciferase-expressing shGFP or shAdar1 Trp53^-/- ID8 cells. Not surprisingly, we isolated significantly more CD8+ T cells from the spleens of mice implanted with shAdar1 tumors and given DNMTi treatment compared with the mice implanted with shGFP tumors and given mock treatment (figure 3B). The shAdar1/DNMTi spleens were larger by eye, which suggested to us that this group had undergone more lymphoproliferation, and thus likely had more primed, tumor-specific CD8+ T cells.

Regardless of tumor Adar1/treatment status, CD8+ T cells from shAdar1/DNMTi mice exhibited higher baseline cytotoxicity compared with CD8+ T cells from shGFP/Mock mice (figure 3C,D, online supplemental figure 3). These data further suggest that the mice in the shAdar1/DNMTi group have more tumor-specific CD8+ T cells. At the 4-hour timepoint, we observed that shGFP DNMTi-treated tumor cells had significantly increased cytotoxicity (figure 3C, online supplemental figure 3A), likely due to increased MHC-I presentation.54 55 Activated CD8+ T cells can recognize peptides presented by tumor cell MHC-I and kill the tumor cell if it is an antigen that the T cell specifically recognizes.56 Furthermore, we have previously shown that DNMTi treatment upregulates MHC-I expression on the surface of OC cells.24 For quick killing to occur, a high tumor-specific T cell frequency would need to be present. For a high tumor-specific T cell frequency, the T cells in the spleen would have been primed and undergone proliferation weeks before T cell isolation. We did see evidence of lymphoproliferation in the shAdar1 DNMTi-treated mice (figure 3B), suggesting that T cells from that group were capable of specifically killing and lysing tumor cells. Alternatively (or in addition too), IFN-γ secreted from T cells can induce endoplasmic reticulum stress and trigger apoptosis in tumor cells.57

Interestingly, this cytotoxicity was not sustained by the 48-hour timepoint (figure 3D, online supplemental figure 3B). We observed greatest sustained cytotoxicity in shAdar1 mock or DNMTi-treated tumors (~35% specific lysis) when co-cultured with CD8+ T cells from mice in the shGFP Mock group or the shAdar1 DNMTi group. The cytotoxicity observed at the 48 timepoint likely involves both (1) direct tumor cell recognition and killing by CD8+ T cells through specific peptide presentation through MHC-I, but also (2) indirect killing through IFN-γ-mediated mechanisms. These data further show that Adar1 loss sensitizes tumor cells to IFN-γ-mediated cytotoxicity.

Adar1 loss increases sensitivity to type I interferon in murine cell lines

While we observed evidence of tumor extrinsic mechanisms of tumor control on shAdar1 loss/DNMTi treatment of OC cells (figures 1–3), we hypothesized that tumor intrinsic mechanisms could also be involved.58 DNMTi plus Adar1 loss in vivo resulted in longer survival and lower tumor burden (figure 1B–D), which could result from (1) direct tumor killing by immune cells (tumor extrinsic) or (2) type I IFN-driven cytotoxic or cytostatic effects (tumor intrinsic). To test the hypothesis that Adar1 loss increases sensitivity to type I IFN as previously shown in human lung cancer cell lines59 and breast cancer cell lines,60 we performed growth assays using shAdar1 and control cells (shGFP) in the ID8 Trp53^-/- cell line background (figure 4). Since previous studies have demonstrated that DNMTi treatment induces type I IFN (such as IFN-β),11 we used exogenous IFN-β as a control for the growth assays. We treated shGFP and shAdar1 cells with Mock (DMSO), DNMTi (Dac, 5-Aza) or IFN-β in growth inhibition assays (figure 4A). We first assessed secretion of IFN-β in the cell culture supernatant after 3 days of Mock or DNMTi treatment. We observed that 5-Aza treatment significantly increased IFN-β levels in ID8 Trp53^-/- cells with Adar1 loss compared with control (p<0.05) (figure 4B). We did, however, also see significantly increased IFN-β levels when comparing Dac-treated shGFP with mock-treated shGFP. Likely in response to this tumor intrinsic production of IFN-β, we observed robust induction of the IFN-inducible Adar1 p150 isoform by day 7 in the DNMTi-treated ID8 Trp53^-/- control (shGFP) cells (figure 4C). Although growth was not significantly different between mock-treated shGFP control cells (gray bar) and Adar1 knockdown cells (blue bar) (online supplemental figure 4A), we observed that Adar1 loss confers increased sensitivity to IFN-β, Dac, and 5-Aza treatment (figure 4D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Adar1 loss increases sensitivity to type I IFN in murine OC cell lines. (A) Schematic for growth inhibition assay. ID8 Trp53^-/- Adar1 knockdown (shAdar1) and control cells (shGFP) were plated on day 0 and then treated with DNMTi for three consecutive days without media change. Cell culture supernatant was collected on day 4 for chemokine/cytokine analysis, and cell pellets were harvested on day 4 for western blot analysis. Cells were re-plated on day 4 and harvested on day 7 for western blot analysis. (B) Cell culture supernatant from day 4 was analyzed using BioLegend LEGENDplex Mouse Anti-Virus Response Panel. IFN-β cytokine levels normalized by number of live cells. (C) Representative western blot analysis of Adar1 protein expression at day 7. (D) Growth inhibition assays showing IFN-β or DNMTi (Dac and 5-Aza) compared with Mock (each condition performed in triplicate). (E) Schematic for colony formation assay. Cells were plated on day 0, and then anti-IFNAR1 or isotype control (10 μg/mL) was given on days 1, 3, 6, and 9. Cells were treated on days 1, 2, and 3 with either DMSO, Dac (100 nM), or 5-Aza (1 μM). IFN-β condition was treated on day 3. Cells were incubated until day 10 when they were subsequently stained with crystal violet. (F) Colony formation assays showing Mock, 5-Aza, and IFN-β (each condition performed in triplicate). (G) Colony formation assays using 100 μg/mL of anti-IFNAR1 or isotype control and showing Mock, 5-Aza, and IFN-β (each condition performed in triplicate). A one-way analysis of variance was performed for statistical significance. *P<0.05; **p<0.01; ***p<0.001. 5-AZA, 5-zaacytidine; Dac, 2’-deoxy-5-azacytidine; DMSO, dimethyl sulfoxide; DNMTi, DNA methyltransferase inhibitor; IFN, interferon; IFNAR1, interferon alpha and beta receptor 1; OC, ovarian cancer.

We next assessed long-term growth using colony formation assays (10 days in culture) with and without blockade of the type I IFN receptor (figure 4E). For these experiments, we used an antibody against IFNAR1, to inhibit type I IFN signaling. Colony formation assays test the ability of a single cell to grow into a colony, and thus can be used to determine cell reproductive death after experimental perturbation.61 After treating with exogenous IFN-β, we observed that Adar1 knockdown cells were strikingly more sensitive to IFN-β compared with control shGFP cells (figure 4F). Additionally, we observed that IFNAR1 blockade did not rescue cell growth in Adar1 knockdown cells, although it rescued growth for control cells. We hypothesized that a higher concentration of the IFNAR1 blocking antibody was required for rescue of IFN-β-stimulated Adar1 knockdown cells. This would likely be due increased cell surface expression of IFNAR1, or compromised surface stability of IFNAR1 (resulting in increased receptor endocytosis and recycling).62 63 When we perform IFNAR1 blockade and subsequently treat with exogenous IFN-β (online supplemental figure 4B), we observe robust upregulation of the ISG Ifi27 in the Adar1 knockdown cells (striped red bar), but not in control cells (striped gray bar) (online supplemental figure 4C). IFN-β stimulation did not robustly induce expression of an ISG (Ifi27) at 24 hours poststimulation in the control (shGFP) cell line (solid gray bar), although this gene was robustly upregulated in Adar1 knockdown cell line at the same timepoint (solid red bar) (online supplemental figure 4C). This suggests that the Adar1 knockdown cell line is more sensitive to type I IFN. These combined data suggest that Adar1 knockdown cells have increased cell surface expression of the type I IFN receptor, potentially explaining their increased sensitivity to type I IFN in cells with Adar1 loss. For example, certain viruses have been shown to antagonize type I IFN signaling by inhibiting IFNAR1 surface expression,64 and other studies have reported that the absence of the intracellular Jak tyrosine kinase signaling domain Tyk2, mature IFNAR1 is weakly expressed on the cell surface.62 65

We thus performed the assay again and treated with a higher concentration of the IFNAR1 blocking antibody and observed a nearly complete rescue of growth in the Adar1 knockdown cells (figure 4G). We conclude that the growth inhibitory effects of Adar1 knockdown in OC cells are through canonical type I IFN signaling. Furthermore, we treated Adar1 knockdown cells (shAdar1) or control cells (shGFP) with Dac or 5-Aza and observed that Adar1 loss confers increased sensitivity to DNMTi treatment independent of IFN-β sensing (figure 4G). A major mechanism of DNMTis in tumor cells is demethylation of tumor suppressor genes, including p16, whose expression can promote tumor cell apoptosis independent of type I IFN signaling.10

Besides the ID8 Trp53^-/- model, two additional murine OC cell lines show similar colony formation trends on Adar1 loss (online supplemental figure 4D–F). In these studies, we compared ovarian surface epithelial-derived CRISPR-modified ID8 cells: Trp53^-/- and Trp53^+/+ (the CRISPR control cell line).35 Although Trp53^-/- knockout more closely resembles the mutational background seen in most human OC, the ID8 cell line is not derived from the same tissue of origin as HGSOC. Since human HGSOC is most commonly derived from the fallopian tube, we also performed these in vitro comparisons using the fallopian tube-derived HGS2 cell line.36 Colony formation assays demonstrate that Adar1 loss dramatically increased IFN-β sensitivity of both Trp53^-/- cell lines (online supplemental figure 4F). We additionally compared IFN-β cytokine expression in the cell culture supernatant in the two Trp53^-/- cell lines (ID8 Trp53^-/- and HGS2) and the Trp53^+/+ cell lines (ID8 CRISPR control) following DNMTi treatment (online supplemental figure 4G–I). DNMTis (5-Aza, Dac) induce a more robust IFN response in p53 wild-type murine cells, although an increased response is still elicited in Trp53^-/- ID8 cells. Additionally, we observed similar trends across all three cell lines for secretion of CCL2, CCL5, and CXCL10 (online supplemental figure 4J–K). Overall, these data demonstrate that Adar1 inhibition increases type I IFN production and inhibits OC tumor cell growth in murine models when combined with DNMTi treatment. This is also consistent with studies in human cancer cells demonstrating that ADAR1 establishes a negative-feedback loop to restrict the DNMTi-induced IFN response.26

RNA editing of immunogenic RNA increases with DNMTi in human OC cells

As Adar1 knockdown plus DNMTi had significant benefit in a murine model of OC, we next assessed the effects of ADAR1 knockdown plus DNMTi treatment in a panel of human OC cell lines (figure 5). In our previous studies in human OC cell lines, we observed that TP53 mutant cells exhibited higher baseline TE expression, although they upregulated fewer TEs in response to DNMTi treatment compared with wild-type.13 We hypothesized that higher TE expression in human cell lines leads to the desensitization of IFN signaling, and that ADAR1 loss sensitizes these cell lines to IFN.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

RNA editing of immunogenic RNA increases with DNMTi in human OC cells. (A–C) TykNu and isogenic CRISPR-edited Hey cell lines (TP53 R175H mutant/TP53 wild-type) were treated for five consecutive days with Mock (PBS) or 5-Aza (500 nM) and then harvested on day 9. Another set of flasks were stimulated with IFN-β for 24 hours prior to harvesting. Protein was isolated and immunoblotted for ADAR1 and β-actin (loading control). (D–F) TykNu and isogenic CRISPR-edited Hey cell lines (TP53 R175H mutant/TP53 wild-type) were treated for five consecutive days with Mock (PBS) or 5-Aza (500 nM) and then harvested on day 9. RNA was isolated from three treatment replicates each, and Illumina sequencing libraries were prepared and sequenced, resulting in paired-end, stranded reads. (D) Alu editing index and LTR editing for TykNu TP53 R175H mutant. (E) Alu editing index and LTR editing for Hey TP53 R175H mutant. (F) Alu editing index and LTR editing for Hey TP53 wild-type. (G) Colony formation assays showing Mock, IFN-β Dac±IFN-β, 5-Aza±IFN-β, and across TykNu shSCR or shADAR1 (each condition performed in triplicate). (H) Colony formation assays showing Mock, IFN-β Dac±IFN-β, 5-Aza±IFN-β, and across Hey TP53 mutant shSCR or shADAR1 (each condition performed in triplicate). Unpaired t-tests were performed for statistical significance for figure 1D,F. *P<0.05; **p<0.01. ADAR1, adenosine deaminase 1; 5-AZA, 5-zaacytidine; Dac, 2’-deoxy-5-azacytidine; DNMTi, DNA methyltransferase inhibitor; IFN, interferon; LTR, long-terminal repeat; OC, ovarian cancer; PBS, phosphate-buffered saline.

To test this hypothesis, we used the following panel of human OC cell lines: TykNu (TP53 R175H hotspot mutant), Hey (TP53 wild-type), and A2780 (TP53 wild-type). We CRISPR-modified the Hey cell line to have the same TP53 hotspot mutation in the TykNu cell line (R175H), and we have used this cell line for these studies along with our TP53 wild-type matching isogenic CRISPR control (clones HH23 and HC2, respectively). We first assessed IFN induction of the ADAR1 p150 isoform after treating with DNMTi or IFN-β (figure 5A–C). Interestingly, IFN-β stimulation did not upregulate ADAR1 p150 protein expression as robustly in the TP53 mutant cell lines in comparison to the TP53 wild-type cell line. This suggests that TP53 mutant cell lines exhibit decreased sensitivity to IFN. To determine whether DNMTi treatment impacts RNA editing, we performed RNA editing analysis in DNMTi-treated human OC cell lines. Alu elements are targets of ADAR1 and, on induction by DNMTis, trigger a type I IFN response.26 We used the RNA Editing Indexer tool to quantify A-to-I RNA editing from RNA-sequencing data38 from DNMTi-treated OC cell lines. We observed a significant increase in the Alu editing index38 in the TykNu OC cell line (TP53 R175H hotspot mutant) with DNMTi treatment compared with mock (figure 5D),66–68 and in the HH23 cell line (CRISPER-edited Hey with TP53 R175H hotspot mutation) (figure 5E), but not in the HC2 OC cell line (CRISPR control Hey cell line with TP53 wild-type) (p<0.05) (figure 5F).

We previously reported that the TP53-mutant cell lines had a significantly higher baseline expression of TEs compared with TP53 wild-type cell lines.13 We have also previously shown that inhibition of DNA methylation increases levels of immunogenic dsRNAs in OC cells and induces type I IFN signaling in tumor cells via viral mimicry.11 This increased RNA editing is likely due to increased baseline TE expression in TP53 mutant cell lines, as we previously published.13 After performing validation of RNA editing for other repetitive elements using a publicly available dataset (online supplemental figure 5A), we also assessed RNA editing of LTRs, ERVs and LINEs in these OC cell lines (online supplemental figure 5B–D). Interestingly, RNA editing increases with DNMTi, but not with IFN-β treatment in the TykNu cell line (online supplemental figure 5C). Strikingly, the LINE and L1 (LINE1) RNA editing index are significantly increased with DNMTi treatment in the TykNu cell line (online supplemental figure 5B,C), although those increases are not evident in either of the Hey cell lines (online supplemental figure 5D).

To investigate the potential effect of RNA editing activity levels on sensitivity to DNMTi treatment and/or ADAR1 loss, we knocked down ADAR1 in the human OC cell lines (online supplemental figure 5E), performed growth assays (figure 5G–H, online supplemental figure 5F,G), and assessed expression of ISGs (online supplemental figure 5H–K). We observed that human OC cell lines have differential responses to ADAR1 loss and DNMTi treatment based on their TP53 status. The TykNu cell line (TP53 mutant) is strikingly sensitive to ADAR1 loss (figure 5G), likely due to high baseline levels of dsRNA expression.15 60 In the TykNu cell line, ADAR1 loss confers sensitivity to exogenous IFN-β stimulation or 5-Aza treatment, and increased levels of ISG expression are observed accordingly with combined ADAR1 loss and DNMTi treatment (online supplemental figure 5H). We observed a similar pattern with the Hey TP53 mutant isogenic clone (figure 5H). Strikingly, we did not observe sensitivity to DNMTi or IFN-β on ADAR1 knockdown in the Hey TP53 wild-type cell line (online supplemental figure 5F), which did not have as much effect on ISG response (online supplemental figure 5J). Interestingly, ADAR1 loss in A2780 (TP53 wild-type) confers increased sensitivity to exogenous IFN-β stimulation or 5-Aza treatment (online supplemental figure 5G), and we observe increases in ISG expression with DNMTi treatment (online supplemental figure 5K).

In conclusion, with combined ADAR1 loss and DNMT inhibition in human OC cell lines, we observed increased sensitivity to IFN, and increased expression of ISGs across a panel of human OC lines with different mutational backgrounds. We additionally show that RNA editing activity may be linked with DNMTi or IFN sensitivity in human OC cells. Overall, ADAR1 knockdown seemed to have the biggest effect on TP53 mutant human OC cells, which is translationally relevant as nearly all HGSOC tumors are TP53 mutant.69

Type I interferon signaling drives prolonged survival in mice with combined Adar1 loss and DNMT inhibition

We have previously shown that DNMTi treatment prolongs survival and reduces tumor burden (as measured by ascites accumulation), and that this response is ablated with IFNAR1 blockade.10 Thus, we next sought to delineate the effects of increased type I IFN sensitivity on tumor burden and survival in vivo for mice implanted with Adar1 knockdown tumors and treated with DNMTi. We implanted mice with Adar1 knockdown Trp53^-/- tumors and treated with THU/DNMTi while blocking IFNAR1 (figure 6, online supplemental figure 6). Compared with Adar1 knockdown mice given DNMTi treatment and isotype control, IFNAR1 blockade significantly (p<0.05) decreased survival in mice implanted with Adar1 knockdown tumors and treated with THU/DNMTi (figure 6A). Furthermore, accumulation of ascites occurred sooner in combination Adar1 loss plus DNMTi-treated mice with blocked IFN response compared with isotype control (figure 6B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Prolonged survival in mice with combined Adar1 loss and DNMT inhibition is dependent on type I IFN signaling. (A) ID8 Trp53^-/- shAdar1 knockdown tumor cells were grown in tissue culture dishes and then 5e6 cells were i.p. injected into C57Bl6 wild-type mice 1 day after the first injections of antibodies. THU/DNMTi treatment began around week 2. Survival curve for the following groups: shGFP Mock/isotype control group (n=5), shGFP Mock/anti-IFNAR1 group (n=5), shAdar1 THU/DNMTi/isotype control (n=7), shAdar1 THU/DNMTi/anti-IFNAR1 (n=5). Two mice from the shGFP Mock isotype control group were censored from the study because they had not developed tumors by the study end point (week 20). (B) Ascites (a buildup of fluid in the peritoneum) collected over time for mice in figure 6A. Ascites is a sign of advanced stage of disease and ascites volume is an indicator of tumor burden in this model. (C) Survival curve (n=5 per group). ID8 Trp5^-/- shGFP (control) or shAdar1 knockdown tumor cells were grown in tissue culture dishes and then 5e6 cells were i.p. injected into mice. Mock or THU/DNMTi treatment began around week 2, and around week 6, mice began developing ascites (a buildup of fluid in the peritoneum). (D) Ascites (a buildup of fluid in the peritoneum) collected over time for mice in figure 6C. Ascites is a sign of advanced stage of disease and ascites volume is an indicator of tumor burden. (E) Ascites (a buildup of fluid in the peritoneum) collected over time for C57Bl6 wild-type mice that received shAdar1 tumors with THU/DNMTi treatment and either: (1) isotype control (n=10), (2) anti-IFNAR1 (n=10), (3) anti-CD8a (n=10), or (4) anti-NK1.1 (n=10) antibodies. ID8 Trp53^-/- shAdar1 knockdown tumor cells were grown in tissue culture dishes and then 5e6 cells were i.p. injected into mice 1 day after the first injections of antibodies. THU/DNMTi treatment began around week 2. (F) Average volume of ascites (a buildup of fluid in the peritoneum) for C57Bl6 wild-type mice that received shAdar1 tumors with THU/DNMTi treatment and either: (1) isotype control (n=10), (2) anti-IFNAR1 (n=10), (3) anti-CD8a (n=10), or (4) anti-NK1.1 (n=10) antibodies. ID8 Trp53^-/- shAdar1 knockdown tumor cells were grown in tissue culture dishes and then 5e6 cells were i.p. injected into mice 1 day after the first injections of antibodies. THU/DNMTi treatment began around week 2. *P<0.05; **p<0.01. DNMTi, DNA methyltransferase inhibitor; IFN, interferon; IFNAR1, interferon alpha and beta receptor 1; i.p., intraperitoneally; ns, not significant; THU, tetrahydrouridine.

Additionally, when we superimpose survival curves from the experiment in figure 1B onto the survival curves from the experiment in online supplemental figure 6A (online supplemental figure 6B), we observe that blockade of the type I IFN response in the shAdar1+DNMTi group (dotted purple line) shows similar survival as DNMTi treatment of the control (shGFP) group (solid red line). These survival trends, combined with data from our in vitro IFNAR1 blockade experiments (figure 4, online supplemental figure 4), show that the survival benefit of Adar1 loss is due to enhanced type I IFN signaling from the tumor cells.

Using an antibody to block IFNAR1, we have previously shown that blockade of type I IFN signaling ablates the DNMTi-induced effects on survival and ascites accumulation in mice implanted with ID8 murine ovarian tumors.10 We have also previously shown that inhibition of DNA methylation induces a type I IFN response in tumor cells through viral mimicry due to transcriptional induction of endogenous dsRNA.11 We next hypothesized that type I IFN produced by tumor cells signals to myeloid cells in the TME in a paracrine manner to amplify this immune response. To test this hypothesis, we injected control (shGFP) and Adar1 knockdown (shAdar1) tumor cells into mice lacking type I IFN receptor function (Ifnar1 null C57BL/6 mice). In contrast to total IFNAR1 blockade (figure 6A,B), survival and ascites accumulation were not dramatically affected in Ifnar1 mice (figure 6C,D) compared with wild-type mice (figure 1A–D). We thus conclude that potential paracrine signaling to myeloid cells in the TME has an inconsequential effect on survival and tumor burden in this mouse model of OC. Instead, chemokine and cytokine production from tumor cells, downstream of type I IFN signaling, is what attracts lymphocytes to control tumor burden.

CD8+ T cells control tumor outgrowth and prolong survival in mice with combined Adar1 loss and DNMT inhibition

We hypothesized that CD8+ lymphocytes, recruited by the above chemokines, were responsible for the control of tumor burden. In ex vivo co-culture assays, we demonstrated that CD8+ T cells from the shAdar1/DNMTi group displayed significantly higher levels of cytotoxicity at the 4-hour timepoint compared with CD8+ T cells from the shGFP/Mock group (figure 3C). We also showed that Adar1 knockdown tumors were more susceptible to CD8+ T cell killing by the 48-hour timepoint (figure 3D). To test this hypothesis in vivo, we depleted CD8+ T cells in mice with Adar1 knockdown tumors receiving DNMTi treatment (online supplemental figure 6A,C,D). Although we observed only a modest difference in survival in comparison to the mice that received isotype control (online supplemental figure 6A), we did observe notably earlier and more rapid rate of tumor burden, as measured by ascites accumulation (figure 6E). Strikingly, the CD8+ depletion group had a significantly increased average volume of ascites per mouse (figure 6F). Depleting NK cells in this model did not affect the average ascites volume or survival (figure 6E,F; online supplemental figure 6A,D). Altogether, these data indicate that CD8+ lymphocytes are critical in controlling tumor outgrowth, while also aiding in prolonging survival in mice with combined Adar1 loss and DNMT inhibition.

Using ascites supernatant to analyze levels of chemokines and cytokines in vivo, we observed increased concentrations of CCL5 and CXCL10 in mice with Adar1 knockdown tumors and DNMTi treatment compared with mice with control tumors (shGFP) receiving Mock treatment (figure 7A), which is concordant with our in vitro data (figure 2). Furthermore, IFNAR1 blockade demonstrated that this increase in chemokines is dependent on IFN signaling (online supplemental figure 7A). Interestingly, significantly higher levels of IFN-γ and IFN-α were measured in the ascites supernatant of mice with Adar1 loss, DNMT inhibition, and blocked type I IFN signaling (online supplemental figure 5B,C). IFN-β was detected at very low levels in some groups, and below the limit of detection for other groups (online supplemental figure 5C). We hypothesize that mice are eventually succumbing to disease due to tumor extrinsic immunosuppressive factors or through another tumor intrinsic negative-feedback loop, which could be dampening the initially enhanced IFN response.70 71 When we assessed the ascites supernatant for immunosuppressive cytokines, we observed a significant decrease in TGF-β in the shAdar1 DNMTi group compared with the shGFP Mock group (figure 7B). Interestingly, we observed significant increase in interleukin (IL)-10 in mice with Adar1 knockdown tumors and DNMTi treatment compared with mice with control tumors (shGFP) and mock treatment (figure 7B). IL-10 could be acting to suppress the T cells that have been recruited to the TME.72

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

(A–D) Ascites was drained from mice in the experiment from figure 6A–D and analyzed for chemokines and cytokines present in the TME. (A) CCL5 and CXCL10 protein concentration in ascites supernatant. (B) TGF-β and IL-10 protein concentration in ascites supernatant. (C–E) Immunophenotyping of the ascites TME using flow cytometry. (C) Regulatory B cells-B_regs (% CD1d hi CD5+IL-10+of CD19+). (D) % MHC-II+ of CD5+ B cells. (E) Activated B cells (% IgM+IgD+ of CD19+). (F) IgM antibody concentration in ascites supernatant. (G) CITE-seq UMAP of cells in ascites of shGFP Mock-treated mice and shADdr1 THU/DNMTi-treated mice. (H) Differential expression of B cell populations of DC subpopulations in the ascites of shGFP Mock-treated mice and shADdr1 THU/DNMTi-treated mice. *P<0.05; **p<0.01. DC, dendritic cell; DNMTi, DNA methyltransferase inhibitor; IL, interleukin; MHC, major histocompatibility complex; ns, not significant; TGF-β, transforming growth factor beta; TME, tumor microenvironment.

Immunophenotyping reveals a recruitment of B cells to the tumor microenvironment on combined Adar1 knockdown and DNMTi treatment

Clinical evidence indicates that B cells may play an important role in OC,33 34 73 74 and recent studies have begun to investigate mechanisms in which B cells may function in OC.75 Specifically, B cells and tertiary lymphoid structures in the TME may predict a response to immune checkpoint blockade.34 Additionally, B cells can also play immunosuppressive roles in autoimmune diseases76 and cancer.77 These immunosuppressive B cells are called B regulatory cells (B_regs), and they are known to produce IL-10, which is a cytokine that suppresses T cells and the T helper 1 immune response.78 The peritoneal cavity, which is the TME for OC, is also known to be home to a prominent population of tissue resident B cells (B1 and B2 cells) that are known to express high levels of IgM.79 Knowing that B cells can play important roles that could influence either an antitumor or an immunosuppressive TME, we performed flow immunophenotyping on the B cell population. Since B cells have been largely understudied relative to T cells and NK cells in tumor immunology, they have not yet been comprehensively investigated in the ID8 mouse model of OC.

Indeed, we observe a significantly increased frequency of B_reg cells in the combination Adar1 and DNMT inhibition group (figure 7C). Although we also observed a significant increase in the frequency of MHC-II+CD5+ B cells in this group (figure 7D), there was additionally a significant reduction in the frequency of activated B cells (figure 7E). We additionally observed a significant decrease in IgM concentration in the ascites fluid (figure 7F). Natural antibodies are secreted by CD5+ B cells. These IgM or IgG natural antibodies can play a role in detecting neoantigens expressed by cancer cells, and they have recently been shown to play a role in tumor control.80–82 In this case, IL-10 may be acting as an immunosuppressant to CD8+ T cells and activated B cells, and B_reg cells may be a potential source of the IL-10.

To understand the transcriptional changes induced by Adar1 knockdown and DNMTi treatment at single-cell resolution, we comprehensively investigated the ascites TME using CITE-seq, a method that combines cell surface protein marker expression with RNA expression (figure 7G,H). Confirming our flow cytometry immunophenotyping results, the shAdar1 DNMTi group exhibited increases in the CD8+ T cell population as well as the B cell population (figure 7G). Beyond the lymphocyte response, we observed increase in the Ly-6G+population in the shAdar1 DNMTi group compared with the shGFP Mock group (figure 7G). Based on expression data, these cells are likely neutrophils and other myeloid-derived suppressor cells. Importantly, B cells and dendritic cells significantly upregulated ISGs in the shAdar1 DNMTi group, confirming a robust type I IFN response (figure 7H). Interestingly, we also found upregulation of immunoglobulin genes in the B cells in the shAdar1 DNMTi group, indicating that B cells are maturing to plasma cells. However, when we measured the concentrations of various immunoglobulin isotypes present in the ascites fluid, we did not observe significant increases (online supplemental figure 7J). The reduced concentration of antibodies in the TME could also be playing specific roles in tumor control in OC.83 Overall, Adar1 knockdown plus DNMTi treatment recruited B cells to the TME, but populations of these B cells may have immunosuppressive effects. Importantly, we observe evidence of lymphocyte dysfunction in the TME of the shAdar1/DNMTi mice, despite presence of a type I IFN response in immune cells, and despite significant recruitment of lymphocytes to the TME.