Lysophosphatidic acid (LPA) is a bioactive lipid enriched in the tumor microenvironment of immunosuppressive malignancies such as ovarian cancer. Although LPA enhances the tumorigenic attributes of cancer cells, the immunomodulatory activity of this phospholipid messenger remains largely unexplored. Here, we report that LPA operates as a negative regulator of type I interferon (IFN) responses in ovarian cancer. Ablation of the LPA-generating enzyme autotaxin (ATX) in ovarian cancer cells reprogrammed the tumor immune microenvironment, extended host survival, and improved the effects of therapies that elicit protective responses driven by type I IFN. Mechanistically, LPA sensing by dendritic cells triggered PGE2 biosynthesis that suppressed type I IFN signaling via autocrine EP4 engagement. Moreover, we identified an LPA-controlled, immune-derived gene signature associated with poor responses to combined PARP inhibition and PD-1 blockade in patients with ovarian cancer. Controlling LPA production or sensing in tumors may therefore be useful to improve cancer immunotherapies that rely on robust induction of type I IFN.

Significance:

This study uncovers that ATX–LPA is a central immunosuppressive pathway in the ovarian tumor microenvironment. Ablating this axis sensitizes ovarian cancer hosts to various immunotherapies by unleashing protective type I IFN responses. Understanding the immunoregulatory programs induced by LPA could lead to new biomarkers predicting resistance to immunotherapy in patients with cancer.

See related commentary by Conejo-Garcia and Curiel, p. 1841.

This article is highlighted in the In This Issue feature, p. 1825

Lysophosphatidic acid (LPA) is a bioactive lipid overproduced by aggressive tumor types such as ovarian, pancreatic, and breast cancers (1–3). This phospholipid messenger has been demonstrated to act as a potent mitogen that stimulates the proliferation, migration, invasiveness, and chemoresistance of malignant cells (4–8). LPA is highly enriched in the ascites of epithelial ovarian cancer patients compared with effusions from healthy individuals and women with benign ovarian neoplasms or nonmalignant ascitic transudate (9–12). Importantly, overexpression of cell adhesion–related genes induced by this phospholipid correlates with poor prognosis in patients with ovarian cancer (13). Autotaxin (ATX), encoded by ENPP2 and frequently overexpressed by ovarian cancer cells (14), is a secreted enzyme that generates LPA by cleaving the choline group from the abundant plasma phospholipid lysophosphatidylcholine (LPC; refs. 15, 16). LPA sensing occurs via six G protein–coupled receptors (LPA1–6) that are variably expressed in multiple cell types, and signaling through these receptors induces diverse intracellular responses that can alter cell metabolism, migration, and proliferation (17). Of note, although the direct protumoral role of ATX–LPA in cancer cells has been well established (18), whether this pathway facilitates malignant progression and resistance to therapy by suppressing antitumor immunity remains largely unexplored.

The vast majority of metastatic ovarian cancer patients are refractory to standard treatments and current forms of immunotherapy (19–21), indicating that potent, yet unidentified, immunosuppressive mechanisms are actively engaged in these aggressive tumors. Type I IFNs, mainly IFNα and IFNβ, have been demonstrated to be required for the development of robust anticancer immune responses and for the optimal efficacy of immunotherapy in several malignancies (22, 23). Nonetheless, the precise mechanisms by which ovarian cancer disables type I IFN responses to hinder antitumor immunity have not been determined. Here, we report that LPA operates as a negative regulator of type I IFN production by dendritic cells (DC). Genetic ablation of the LPA-generating enzyme ATX in malignant cells enhanced DC function, reprogrammed the immune contexture of metastatic ovarian cancer, prolonged overall host survival, and enhanced the immunotherapeutic effects of multiple anticancer interventions by boosting protective type I IFN responses.

Loss of ATX in Malignant Cells Delays Ovarian Cancer Progression and Extends Host Survival

We sought to define the immunoregulatory role of LPA in the setting of metastatic ovarian cancer. To this end, we used the orthotopic ID8-Defb29/Vegf-A ovarian cancer model, which engenders a highly chemoresistant and immunosuppressive peritoneal carcinomatosis that recapitulates the advanced stages of human ovarian cancer (24–26). ELISA and lipidomic analyses determined that the ascitic fluid generated by these mouse peritoneal tumors contained LPA levels and species similar to those present in the ascites of patients with ovarian cancer (Fig. 1A and B). Consistent with prior reports (9–12), we found high concentrations of diverse LPA species in the ascites of patients with ovarian cancer compared with plasma samples from patients with ovarian cancer or cancer-free women (Supplementary Fig. S1A). Similarly, the ascites generated by ID8-Defb29/Vegf-A ovarian tumors contained increased levels of multiple LPA species in comparison with peritoneal wash samples obtained from naïve mice (Supplementary Fig. S1B). Female mice bearing established ID8-Defb29/Vegf-A ovarian cancer also showed higher circulating levels of ATX than their tumor-free counterparts (Supplementary Fig. S1C). Of note, the Enpp2 mRNA (encoding ATX) was primarily expressed by cancer cells at tumor locations rather than diverse infiltrating leukocyte populations (Fig. 1C). Thus, we used CRISPR/Cas9 to abrogate Enpp2/ATX in malignant cells and further understand the function of this enzyme in the ovarian cancer microenvironment. Enpp2 expression and ATX production were markedly reduced in Enpp2-targeted ovarian cancer cells [Enpp2 single-guide RNA (sgRNA)] compared with their isogenic counterparts harboring a control sgRNA that does not target the mouse genome (Supplementary Fig. S2A and S2B). Ablating Enpp2 in malignant cells did not compromise their in vitro proliferation (Supplementary Fig. S2C) or early orthotopic establishment in vivo (Supplementary Fig. S2D), yet it drastically reduced the concentrations of ATX and LPA in the peritoneal cavity of mice bearing metastatic ovarian cancer for 5 to 6 weeks (Fig. 1D and E). Decreased tumor-induced splenomegaly and a ∼50% reduction in the amount of hemorrhagic ascites were concurrently observed in the same mice developing ATX/LPA-deficient ovarian cancer (Supplementary Fig. S2E–S2H). Notably, ablation of Enpp2/ATX in ovarian cancer cells diminished malignant ascites development and accumulation over time (Fig. 1F), and markedly extended overall host survival (Fig. 1G). Similar effects were observed when cancer cell–intrinsic Enpp2/ATX was ablated in the PPNM mouse model of high-grade serous tubo-ovarian carcinoma (HGSC) that carries the most common genetic abnormalities found in human HGSCs (ref. 27; Supplementary Fig. S2I; Fig. 1H and I). Hence, malignant cells are the main source of ATX–LPA at tumor locations, and this axis is crucial for the aggressive behavior and progression of metastatic ovarian cancer in two independent mouse models of disease.

Figure 1.

Genetic loss of ATX in malignant cells compromises metastatic ovarian cancer (OvCa) progression. A, LPA concentration was determined by ELISA in malignant ascites from patients with ovarian cancer (n = 17) and mice bearing ID8-Defb29/Vegf-A ovarian cancer for 5 to 6 weeks (n = 10). B, Lipidomic analyses were performed to evaluate specific LPA species in malignant ascites from patients with ovarian cancer (n = 15) and mice bearing advanced ovarian cancer (n = 4). C, Cancer cells, DCs, macrophages, and CD3+ T cells were sorted from peritoneal wash samples of mice bearing metastatic ovarian cancer for 3 to 4 weeks (n = 9). Expression of the Enpp2 transcript was determined by RT-qPCR, and data were normalized to endogenous levels of Actb. D, Expression of the Enpp2 transcript was determined in cancer cells sorted from the ascites of mice bearing control or Enpp2-null ovarian cancer for 40 days (n = 6/group). E, ATX and LPA concentrations were quantified in ascites from mice bearing control or Enpp2-null ovarian cancer for 40 days (n = 6–7/group). F and G, Ascites accumulation (F) and overall host survival (G) in mice developing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer (n = 15–16 mice/group). H, Quantification of peritoneal carcinomatosis in mice bearing luciferase-expressing control or Enpp2-null PPNM tumors for 13, 27, and 41 days (n = 24–25/group). I, Overall survival rates for the same mice described in H (n = 24–25 mice/group). Data in CF are shown as mean ± SEM. C, One-way ANOVA (Tukey multiple comparisons test). D and E, Two-tailed Student t test. F, Two-way ANOVA (Šídák's multiple comparisons test). G and I, Log-rank test for survival. H, Two-way ANOVA (Tukey multiple comparisons test). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Control sgRNA, scrambled single-guide RNA. Enpp2 sgRNA, ATX-targeting single-guide RNA.

Figure 1.

Genetic loss of ATX in malignant cells compromises metastatic ovarian cancer (OvCa) progression. A, LPA concentration was determined by ELISA in malignant ascites from patients with ovarian cancer (n = 17) and mice bearing ID8-Defb29/Vegf-A ovarian cancer for 5 to 6 weeks (n = 10). B, Lipidomic analyses were performed to evaluate specific LPA species in malignant ascites from patients with ovarian cancer (n = 15) and mice bearing advanced ovarian cancer (n = 4). C, Cancer cells, DCs, macrophages, and CD3+ T cells were sorted from peritoneal wash samples of mice bearing metastatic ovarian cancer for 3 to 4 weeks (n = 9). Expression of the Enpp2 transcript was determined by RT-qPCR, and data were normalized to endogenous levels of Actb. D, Expression of the Enpp2 transcript was determined in cancer cells sorted from the ascites of mice bearing control or Enpp2-null ovarian cancer for 40 days (n = 6/group). E, ATX and LPA concentrations were quantified in ascites from mice bearing control or Enpp2-null ovarian cancer for 40 days (n = 6–7/group). F and G, Ascites accumulation (F) and overall host survival (G) in mice developing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer (n = 15–16 mice/group). H, Quantification of peritoneal carcinomatosis in mice bearing luciferase-expressing control or Enpp2-null PPNM tumors for 13, 27, and 41 days (n = 24–25/group). I, Overall survival rates for the same mice described in H (n = 24–25 mice/group). Data in CF are shown as mean ± SEM. C, One-way ANOVA (Tukey multiple comparisons test). D and E, Two-tailed Student t test. F, Two-way ANOVA (Šídák's multiple comparisons test). G and I, Log-rank test for survival. H, Two-way ANOVA (Tukey multiple comparisons test). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Control sgRNA, scrambled single-guide RNA. Enpp2 sgRNA, ATX-targeting single-guide RNA.

Close modal

Tumor-Intrinsic ATX Controls the Ovarian Cancer Immunoenvironment

We next evaluated whether the loss of ATX in malignant cells and the ensuing reduction of LPA production in the tumor milieu altered the immune contexture of metastatic ovarian cancer. We focused our analyses on the middle stages of disease progression (days 35–40), when mice developing control or ATX-deficient ovarian cancer had similar numbers and proportions of SSChiCD45 cancer cells in the peritoneal cavity (Supplementary Fig. S3A; Fig. 2A and B). At this stage, mice developing ATX-null ovarian cancer demonstrated superior CD3+ T-cell infiltration with higher proportions of CD8+ cytotoxic T cells at tumor locations (Fig. 2A and C–E). Importantly, we found increased antigen-experienced (CD44+) CD4+ and CD8+ T cells expressing IFNγ and TNFα in the peritoneal cavity of mice developing ATX-deficient ovarian cancer compared with their ATX-sufficient counterparts (Fig. 2FJ; Supplementary Fig. S3B–S3H). These changes were accompanied by reduced proportions of both CD11bhiF4/80+ macrophages and total CD11c+MHC-II+ DCs (Supplementary Fig. S3I–S3K), whereas no alterations were observed in the proportion of CD11b+Gr1+ myeloid cells (Supplementary Fig. S3I and S3L). Importantly, loss of ATX in ovarian cancer cells further augmented natural killer (NK) cell infiltration into tumor sites (Supplementary Fig. S3I and S3M). These data indicate that tumor-derived ATX–LPA controls the immune composition of ovarian cancer and suggest that this phospholipid messenger might operate as a global immunomodulatory mediator.

Figure 2.

Ablation of ATX reprograms the immune microenvironment of metastatic ovarian cancer. A–E, Peritoneal wash samples were collected from mice developing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer for 35 to 40 days, and cells were analyzed by flow cytometry (n = 6–7/group). SSC, side scatter. F–J, The proportion of IFNγ- and TNFα-expressing cells within CD3+CD4+CD44+ and CD3+CD8α+CD44+ T cells was determined in the ascites of mice bearing control or Enpp2-null ovarian cancer (n = 5–6/group). Data in BE and GJ are shown as mean ± SEM. B–E and G–J, Two-tailed Student t test. ns, not significant; *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

Figure 2.

Ablation of ATX reprograms the immune microenvironment of metastatic ovarian cancer. A–E, Peritoneal wash samples were collected from mice developing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer for 35 to 40 days, and cells were analyzed by flow cytometry (n = 6–7/group). SSC, side scatter. F–J, The proportion of IFNγ- and TNFα-expressing cells within CD3+CD4+CD44+ and CD3+CD8α+CD44+ T cells was determined in the ascites of mice bearing control or Enpp2-null ovarian cancer (n = 5–6/group). Data in BE and GJ are shown as mean ± SEM. B–E and G–J, Two-tailed Student t test. ns, not significant; *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

Close modal

LPA Inhibits Type I IFN Production by Multiple DC Subsets

We hypothesized that LPA could directly alter immune cell functions in the tumor microenvironment. We conducted single-cell RNA sequencing (RNA-seq) analyses of total CD45+ leukocytes sorted from the ascites of mice developing metastatic ovarian cancer to evaluate the biodistribution and expression levels of Lpar1–Lpar6 genes encoding LPA receptors 1–6 (LPA1–6). These experiments revealed preferential expression of Lpar1, Lpar2, and Lpar6 in multiple ovarian cancer–associated immune cells (Supplementary Fig. S4A and S4B), with diverse intratumoral DC subsets, including type 1 conventional DCs (cDC1), type 2 conventional DCs (cDC2), and plasmacytoid DCs (pDC) demonstrating the highest expression of Lpar6 (Supplementary Fig. S4B). RT-qPCR analyses using total CD45+CD11c+MHC-II+ tumor-associated DCs (tDC) confirmed that these cells robustly expressed Lpar6, followed by Lpar2 and Lpar5 (Supplementary Fig. S4C). Of note, total splenic DCs (sDC) and bone marrow–derived DCs (BMDC) from naïve mice also showed high Lpar6 levels, accompanied by the moderate expression of Lpar3 and Lpar5 (Supplementary Fig. S4D and S4E). We therefore surmised that LPA could alter DC functions.

Mouse BMDCs exposed to LPA concentrations similar to those found in the ascites of patients with ovarian cancer exhibited drastic changes in global gene expression, with 1,264 genes demonstrating significant downregulation and 1,155 genes showing significant upregulation (fold change > 1.5; P < 0.05; FDR < 5%). These altered gene subsets were analyzed to identify potential upstream regulators mediating the observed transcriptional changes. Strikingly, LPA-exposed BMDCs showed a marked downregulation of gene networks induced by type I IFN signaling, with IFNβ, IFNAR1, IRF3, IRF7, and STAT1 emerging as the top predicted upstream regulators (Fig. 3A and B). Conversely, immunosuppressive gene programs controlled by PTGER4/EP4 signaling, SOCS1, and STAT3 were significantly activated in BMDCs exposed to LPA (Fig. 3A and C). Subsequent RT-qPCR analyses confirmed that LPA treatment downregulated multiple type I IFN–stimulated genes (ISG), such as Ddx58, Ifit1, Ifit2, and Isg15, in a time- and dose-dependent manner (Supplementary Fig. S5A and S5B). Hence, we hypothesized that LPA might suppress type I IFN expression by diverse DC types. Mouse BMDCs, sDCs, and pDCs stimulated with various Toll-like receptor (TLR) agonists demonstrated a dose-dependent decrease in IFNβ production upon LPA exposure (Fig. 3DG). Similar effects were observed when human monocyte-derived DCs (moDC) and pDCs obtained from the peripheral blood of cancer-free donors were stimulated with TLR agonists in the presence of LPA (Fig. 3H; Supplementary Fig. S5C). In LPS- or poly (I:C)–stimulated BMDCs, LPA treatment inhibited the phosphory­lation of TBK1 and IRF3 (Fig. 3I and J), which are signaling events required for optimal type I IFN expression (28). Hence, LPA acts as a negative regulator of type I IFN production by diverse human and mouse DC subsets. Consistent with these in vitro findings, tDCs sorted from mice developing Enpp2-null ovarian cancer with diminished LPA (Fig. 1E) demonstrated marked upregulation of Ifna and Ifnb1 transcripts (Fig. 3K), as well as overexpression of multiple type I ISGs (Fig. 3L), compared with their counterparts isolated from mice bearing ATX-sufficient tumors. Similar effects were observed in cancer cells and neutrophils sorted from the peritoneal cavity of mice growing Enpp2-null ovarian cancer (Supplementary Fig. S5D–S5G), indicating that tumor-derived LPA can further repress type I IFN production and sensing by additional cell types present in the same microenvironment.

Figure 3.

LPA blunts type I IFN production by DCs. BMDCs were left untreated or stimulated with LPA (100 μmol/L) for 2 or 6 hours, and global transcriptional profiles were analyzed by RNA-seq (n = 3 independent biological replicates per group). A, Ingenuity Pathway Analysis (IPA) for predicted upstream regulators of differentially expressed genes. B and C, Heat map representation of type I IFN (B) and PTGER4/EP4 (C) target genes. untr, untreated. D–G, BMDCs, sDCs, or pDCs were left untreated or pretreated for 2 hours with LPA at 10 μmol/L (+) or 100 μmol/L (++), and cells were then stimulated with LPS, poly (I:C), or CpG ODN1585, as described in the Methods section. Production of IFNβ in culture supernatants was determined by ELISA (n = 4). H, MoDCs or purified pDCs from cancer-free donors were treated with LPA (100 μmol/L) for 2 hours, and cells were then stimulated with LPS, poly (I:C), or CpG-C274 as described in the Methods section (n = 5–7). Production of IFNα or IFNβ in culture supernatants was determined by ELISA. I and J, Representative immunoblot analysis for phospho-TBK1 (pTBK1), total TBK1, phospho-IRF3 (pIRF3), and total IRF3 in LPA-exposed BMDCs stimulated with either LPS (100 ng/mL; I) or poly (I:C) (10 μg/mL; J) for the indicated times. K and L, RT-qPCR analysis of type I IFN transcripts (K) and type I ISGs (L) in tDCs sorted from the ascites of mice bearing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer. Data were normalized to Actb in all cases (n = 4 independent mice/group). Expr, log2 value of normalized reads; fold/average, fold change of expression relative to average normalized reads of all samples. Data in DG and K and L are shown as mean ± SEM. DG, One-way ANOVA (Tukey multiple comparisons test). H, Two-tailed paired Student t test. K and L, Two-tailed Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 3.

LPA blunts type I IFN production by DCs. BMDCs were left untreated or stimulated with LPA (100 μmol/L) for 2 or 6 hours, and global transcriptional profiles were analyzed by RNA-seq (n = 3 independent biological replicates per group). A, Ingenuity Pathway Analysis (IPA) for predicted upstream regulators of differentially expressed genes. B and C, Heat map representation of type I IFN (B) and PTGER4/EP4 (C) target genes. untr, untreated. D–G, BMDCs, sDCs, or pDCs were left untreated or pretreated for 2 hours with LPA at 10 μmol/L (+) or 100 μmol/L (++), and cells were then stimulated with LPS, poly (I:C), or CpG ODN1585, as described in the Methods section. Production of IFNβ in culture supernatants was determined by ELISA (n = 4). H, MoDCs or purified pDCs from cancer-free donors were treated with LPA (100 μmol/L) for 2 hours, and cells were then stimulated with LPS, poly (I:C), or CpG-C274 as described in the Methods section (n = 5–7). Production of IFNα or IFNβ in culture supernatants was determined by ELISA. I and J, Representative immunoblot analysis for phospho-TBK1 (pTBK1), total TBK1, phospho-IRF3 (pIRF3), and total IRF3 in LPA-exposed BMDCs stimulated with either LPS (100 ng/mL; I) or poly (I:C) (10 μg/mL; J) for the indicated times. K and L, RT-qPCR analysis of type I IFN transcripts (K) and type I ISGs (L) in tDCs sorted from the ascites of mice bearing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer. Data were normalized to Actb in all cases (n = 4 independent mice/group). Expr, log2 value of normalized reads; fold/average, fold change of expression relative to average normalized reads of all samples. Data in DG and K and L are shown as mean ± SEM. DG, One-way ANOVA (Tukey multiple comparisons test). H, Two-tailed paired Student t test. K and L, Two-tailed Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

LPA-Induced PGE2 Suppresses Type I IFN Responses in DCs via PTGER4/EP4

Next, we sought to determine how LPA blunts type I IFN expression in DCs. We examined the involvement of prostaglandin E2 (PGE2) in this process, as our transcriptomic analyses indicated that gene programs induced upon engagement of the PGE2 receptor PTGER4 (also known as EP4) were enriched in DCs exposed to LPA (Fig. 3A and C) and also because PGE2 overproduction in the tumor microenvironment can suppress anticancer immunity via multiple mechanisms (29, 30). LPA treatment rapidly induced the expression of Ptgs2/Cox2 (encoding prostaglandin-endoperoxide synthase 2) and triggered PGE2 production by BMDCs in a time- and dose-dependent manner (Fig. 4A; Supplementary Fig. S6A). This process was mediated by the activation of p38 MAPK (Fig. 4B), whereas other factors previously implicated in prostanoid generation, such as PPARδ, PPARα, CREB, JNK, NFκβ, or AP-1, did not play a role (Supplementary Fig. S6B; refs. 31–34). To ascertain whether LPA-induced PGE2 inhibited autocrine type I IFN responses, we treated BMDCs with the PTGER4/EP4 antagonist PGN 1531 (35) and then stimulated them with bacterial LPS in the presence or absence of LPA. Blocking EP4 engagement by PGE2 fully restored the expression of multiple type I ISGs in LPS-treated BMDCs exposed to LPA (Fig. 4C). These data reveal that the LPA–PGE2–EP4 axis disables autocrine type I IFN signaling in DCs.

Figure 4.

LPA-induced PGE2 suppresses type I IFN responses in DCs via EP4. A, Left, BMDCs were left untreated or incubated with LPA (100 μmol/L) for 2 hours, and expression of Ptgs2 was determined by RT-qPCR (n = 6). Right, BMDCs were left untreated or incubated with LPA (100 μmol/L) for 6 hours, and PGE2 levels were determined in the culture supernatant by ELISA (n = 4). B, Left, BMDCs were pretreated with the p38 MAPK kinase inhibitor (MAPKi) SB203580 for 1 hour and then stimulated with LPA for 2 hours. Expression of Ptgs2 was determined by RT-qPCR (n = 8). Right, BMDCs were pretreated with the p38 MAPK kinase inhibitor SB203580 (10 µmol/L) for 1 hour and then stimulated with LPA for 6 hours. PGE2 levels were determined in the culture supernatant by ELISA (n = 4). C, BMDCs were pretreated with the EP4 antagonist [EP4 inhibitor (EP4i)] PGN 1531 (5 μmol/L) for 1 hour and then stimulated with LPA and LPS for 4 hours. Expression of type I ISGs was quantified by RT-qPCR. Data were normalized to Actb (n = 4). D, PGE2 was quantified in ascites fluid from mice bearing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer for 40 days (n = 6–7/group). Correlation of PGE2 levels with ATX (E) or LPA (F) concentration in the same samples. G–L, Proportion of infiltrating CD8α+ T cells or NK cells versus levels of ATX (G and H), LPA (I and J), or PGE2 (K and L) in the peritoneal cavity of mice bearing control (blue dots) or Enpp2-null (red dots) ovarian cancer (n = 13). M and N, PGE2 concentration versus levels of LPA (total), LPA (16:0), and LPA (18:2) in ascites from patients with ovarian cancer (n = 11). O and P, Overall survival curves for HGSOC patients in the TCGA cohorts classified by the expression ratios of ENPP2/IFNA1 (O) or PTGS2/IFNA1 (P). Numbers at the bottom of the graph denote the median overall survival (months) for each group. Data in AD are shown as mean ± SEM. A and D, Two-tailed Student t test. B and C, One-way ANOVA (Tukey multiple comparisons test). E–N, Spearman rank correlation coefficient (r). O and P, Log-rank test. A–D, *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. E–P, Exact P values are shown.

Figure 4.

LPA-induced PGE2 suppresses type I IFN responses in DCs via EP4. A, Left, BMDCs were left untreated or incubated with LPA (100 μmol/L) for 2 hours, and expression of Ptgs2 was determined by RT-qPCR (n = 6). Right, BMDCs were left untreated or incubated with LPA (100 μmol/L) for 6 hours, and PGE2 levels were determined in the culture supernatant by ELISA (n = 4). B, Left, BMDCs were pretreated with the p38 MAPK kinase inhibitor (MAPKi) SB203580 for 1 hour and then stimulated with LPA for 2 hours. Expression of Ptgs2 was determined by RT-qPCR (n = 8). Right, BMDCs were pretreated with the p38 MAPK kinase inhibitor SB203580 (10 µmol/L) for 1 hour and then stimulated with LPA for 6 hours. PGE2 levels were determined in the culture supernatant by ELISA (n = 4). C, BMDCs were pretreated with the EP4 antagonist [EP4 inhibitor (EP4i)] PGN 1531 (5 μmol/L) for 1 hour and then stimulated with LPA and LPS for 4 hours. Expression of type I ISGs was quantified by RT-qPCR. Data were normalized to Actb (n = 4). D, PGE2 was quantified in ascites fluid from mice bearing control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer for 40 days (n = 6–7/group). Correlation of PGE2 levels with ATX (E) or LPA (F) concentration in the same samples. G–L, Proportion of infiltrating CD8α+ T cells or NK cells versus levels of ATX (G and H), LPA (I and J), or PGE2 (K and L) in the peritoneal cavity of mice bearing control (blue dots) or Enpp2-null (red dots) ovarian cancer (n = 13). M and N, PGE2 concentration versus levels of LPA (total), LPA (16:0), and LPA (18:2) in ascites from patients with ovarian cancer (n = 11). O and P, Overall survival curves for HGSOC patients in the TCGA cohorts classified by the expression ratios of ENPP2/IFNA1 (O) or PTGS2/IFNA1 (P). Numbers at the bottom of the graph denote the median overall survival (months) for each group. Data in AD are shown as mean ± SEM. A and D, Two-tailed Student t test. B and C, One-way ANOVA (Tukey multiple comparisons test). E–N, Spearman rank correlation coefficient (r). O and P, Log-rank test. A–D, *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. E–P, Exact P values are shown.

Close modal

Mice bearing Enpp2-null ovarian cancer demonstrated lower PGE2 levels in the ascites than hosts bearing control ATX-sufficient ovarian cancer (Fig. 4D). In this setting, intrinsic ATX expression and LPA production correlated with the amount of PGE2 at tumor sites (Fig. 4E and F), and high levels of ATX, LPA, or PGE2 were associated with reduced CD8+ T-cell and NK-cell infiltration (Fig. 4GL). Importantly, the concentration of PGE2 in cell-free ascites samples from patients with high-grade serous ovarian cancer (HGSOC) was positively associated with the level of total LPA in the same milieu (Fig. 4M). Among the individual LPA species analyzed, only 16:0 and 18:2 demonstrated a significant positive association with PGE2 expression (Fig. 4N; Supplementary Fig. S6C). Furthermore, an analysis of ovarian cancer patient cohorts from The Cancer Genome Atlas (TCGA) determined that high expression ratios of ENPP2 or PTGS2 to IFNA1 (encoding IFNα1) in human tumor specimens correlated with decreased overall survival (Fig. 4O and P).

ATX Deficiency Enhances the Antitumor Effects of Interventions that Elicit Type I IFN

We hypothesized that the ATX–LPA arm could operate as a major mechanism of resistance to therapies that induce type I IFN responses against cancer. To test this new concept, female mice bearing ATX-deficient or sufficient ID8–based ovarian cancer were treated with the TLR3 agonist poly (I:C), a prototypical type I IFN inducer (36, 37), and malignant progression and host survival were monitored over time. Consistent with our initial results (Fig. 1F and G), vehicle-treated mice developing Enpp2-null ovarian cancer showed reduced ascites accumulation and prolonged survival compared with their counterparts bearing ATX-sufficient ovarian cancer (Fig. 5AC). Notably, although poly (I:C) increased the median survival of mice bearing control tumors by only 1 week (∼15%), treatment with this TLR3 agonist induced superior therapeutic effects in mice bearing Enpp2-null ovarian cancer devoid of ATX, leading to an extended delay in ascites accumulation (Fig. 5A and B) and a ∼60% increase in overall survival in comparison with vehicle-treated mice bearing control ovarian cancer (Fig. 5C). Validating these effects in an independent model of HGSC (27), mice bearing ATX-deficient PPNM tubo-ovarian tumors also demonstrated delayed metastatic progression and a drastic increase in overall survival upon poly (I:C) administration, whereas their ATX-sufficient counterparts did not respond to this treatment (Fig. 5DF). Of note, antibody-mediated blockade of the interferon α/β receptor 1 (IFNAR1) fully abrogated the therapeutic effects of poly (I:C) in mice bearing Enpp2-null ovarian cancer (Fig. 5G; Supplementary Fig. S7A and S7B), demonstrating that lack of tumor-derived LPA enhances protective type I IFN signaling elicited by this treatment.

Figure 5.

Therapeutic effects of poly (I:C) administration in mice bearing ATX-null ovarian cancer (OvCa). WT C57BL/6J mice were challenged intraperitoneally with control or Enpp2-null ID8-Defb29/Vegf-A (AC; n = 16–19 mice/group) or PPNM (DF; n = 7–8 mice/group) cancer cells. After 10 days, mice were treated with vehicle control or poly (I:C) as described in the Methods section. A, Ascites accumulation denoted as percentage weight gain over time. B, Changes in ascites development were analyzed by calculating the area above the curve in each experimental group starting on day 21 and a cutoff of 35% of weight gain. Data are represented as percentage change compared with the control sgRNA group treated with vehicle. C, Overall survival curves for the same mice described in A and B. Representative bioluminescence imaging at day 39 (D) and quantification of peritoneal carcinomatosis (E) in mice bearing luciferase-expressing control or Enpp2-null PPNM tumors for 25 and 39 days with or without poly (I:C) treatment. F, Overall survival curves for the same mice described in D and E. G, Experiments were repeated as in AC, but 3 days after tumor implantation, mice (n = 6–9/group) were treated with isotype control or anti-IFNAR1 blocking antibodies, as described in the Methods section. Host survival was monitored over time. H, Experiments were repeated as in A–C, but 9 days after tumor implantation, mice (n = 6–9/group) were orally treated with vehicle control or the EP4 agonist KAG-308 as described in Methods. Host survival was monitored over time. I, Overall survival in mice of the indicated genotypes implanted with control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer treated with poly (I:C) (n = 8–10/group). KO, knockout. J, Overall survival in Rag2/Il2rg double-KO mice implanted with control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer receiving the indicated treatments (n = 7–8/group). Data in A and B are shown as mean ± SEM. B, One-way ANOVA with Tukey multiple comparisons test. E, Two-way ANOVA (Tukey multiple comparisons test). C and FJ, Log-rank test for survival. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 5.

Therapeutic effects of poly (I:C) administration in mice bearing ATX-null ovarian cancer (OvCa). WT C57BL/6J mice were challenged intraperitoneally with control or Enpp2-null ID8-Defb29/Vegf-A (AC; n = 16–19 mice/group) or PPNM (DF; n = 7–8 mice/group) cancer cells. After 10 days, mice were treated with vehicle control or poly (I:C) as described in the Methods section. A, Ascites accumulation denoted as percentage weight gain over time. B, Changes in ascites development were analyzed by calculating the area above the curve in each experimental group starting on day 21 and a cutoff of 35% of weight gain. Data are represented as percentage change compared with the control sgRNA group treated with vehicle. C, Overall survival curves for the same mice described in A and B. Representative bioluminescence imaging at day 39 (D) and quantification of peritoneal carcinomatosis (E) in mice bearing luciferase-expressing control or Enpp2-null PPNM tumors for 25 and 39 days with or without poly (I:C) treatment. F, Overall survival curves for the same mice described in D and E. G, Experiments were repeated as in AC, but 3 days after tumor implantation, mice (n = 6–9/group) were treated with isotype control or anti-IFNAR1 blocking antibodies, as described in the Methods section. Host survival was monitored over time. H, Experiments were repeated as in A–C, but 9 days after tumor implantation, mice (n = 6–9/group) were orally treated with vehicle control or the EP4 agonist KAG-308 as described in Methods. Host survival was monitored over time. I, Overall survival in mice of the indicated genotypes implanted with control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer treated with poly (I:C) (n = 8–10/group). KO, knockout. J, Overall survival in Rag2/Il2rg double-KO mice implanted with control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer receiving the indicated treatments (n = 7–8/group). Data in A and B are shown as mean ± SEM. B, One-way ANOVA with Tukey multiple comparisons test. E, Two-way ANOVA (Tukey multiple comparisons test). C and FJ, Log-rank test for survival. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

To functionally determine whether impaired PGE2 production (Fig. 4D) improved the therapeutic effects of poly (I:C) in mice bearing ATX-deficient ovarian cancer, we carried out rescue experiments using the selective EP4 agonist KAG-308 that is suitable for in vivo dosing via oral administration (38–40). The direct engagement of EP4 using KAG-308 bypassed the PGE2 production defects observed in tumors devoid of ATX/LPA and significantly reduced the survival benefit induced by poly (I:C) treatment in this genetic context (Fig. 5H; Supplementary Fig. S7C and S7D). The anticancer effects elicited by TLR3 agonism in ATX-deficient tumors were mediated by DCs, as Batf3-deficient mice devoid of cDC1s (41) demonstrated impaired therapeutic responses to poly (I:C) administration (Fig. 5I). Furthermore, immunodeficient [Rag2/Il2rg double knockout (KO)] mice lacking NK, T, and B cells implanted with ATX-null ovarian cancer were totally refractory to this treatment (Fig. 5J). Hence, tumor-intrinsic ATX–LPA functions as a negative regulator of protective type I IFN in ovarian cancer, and ablating this pathway could be used to evoke therapeutic antitumor immunity via TLR3 stimulation.

Treatment with PARP inhibitors has been shown to evoke beneficial type I IFN responses in the tumor microenvironment through the DNA-sensing pathway cGAS–STING (42–44). Indeed, previous reports indicated that ovarian cancer cells treated with PARP inhibitors accumulate cytosolic DNA that can subsequently induce immunostimulatory type I IFN by neighboring antigen-presenting cells via STING activation (42). We reasoned that the ATX–LPA axis could also limit the therapeutic effects of PARP inhibition in ovarian cancer. We prioritized talazoparib, as it is the most potent PARP inhibitor currently in the clinic (45) and because its immunotherapeutic effects have been proposed to be independent of the BRCA status of the cancer cell (43). Notably, LPA exposure blocked the induction of Ifna, Ifnb1, and type I ISGs in BMDCs cocultured with ovarian cancer cells pretreated with talazoparib (Fig. 6A and B). In vivo, oral talazoparib administration diminished ascites accumulation and extended survival by ∼30% in mice bearing ATX-sufficient ovarian cancer (Fig. 6CE). However, the therapeutic effects of talazoparib drastically improved in female mice bearing Enpp2-null ovarian cancer devoid of ATX/LPA, eliciting a further delay in ascites accumulation and a ∼63% increase in their median survival (Fig. 6CE). Next, to ascertain whether the enhanced therapeutic effects observed were mediated by the host STING pathway, we performed similar survival experiments in wild-type (WT) or Sting KO mice. STING ablation did not affect the survival of talazoparib-treated mice bearing control ATX-sufficient ovarian cancer, yet it significantly curtailed the therapeutic effects of this PARP inhibitor in host developing Enpp2-null ovarian tumors (Fig. 6F). Hence, ATX–LPA production by ovarian cancer cells can limit STING-driven type I IFN responses evoked by talazoparib administration.

Figure 6.

Therapeutic effects of the PARP inhibitor talazoparib in mice bearing ATX-null ovarian cancer (OvCa). A and B, RT-qPCR analysis of type I IFN transcripts (A) and type I ISGs (B) in BMDCs cocultured with talazoparib-treated ovarian cancer cells in the presence or absence of LPA. Data were normalized to Actb in all cases (n = 4). C–E, C57BL/6J mice (n = 16–19/group) were challenged via i.p. injection with 1.5 × 106 control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer cells. After 7 days, mice were treated once daily with vehicle or talazoparib (0.33 mg/kg) by oral gavage for up to 28 days. C, Ascites accumulation was denoted as percentage weight gain over time. D, Changes in ascites development were analyzed by calculating the area above the curve in each experimental group starting on day 35 and a cutoff of 35% of weight gain. Data are represented as percentage change compared with the control sgRNA group treated with vehicle. E, Overall survival curves for the same mice as described in C and D. F, Survival experiments were repeated as described in CE but including STING-deficient (KO) hosts (n = 6–10/group). Data in AD are shown as mean ± SEM. A and B, Two-way ANOVA (Tukey multiple comparisons test). D, One-way ANOVA with Tukey multiple comparisons test. E and F, Log-rank test analysis for survival. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 6.

Therapeutic effects of the PARP inhibitor talazoparib in mice bearing ATX-null ovarian cancer (OvCa). A and B, RT-qPCR analysis of type I IFN transcripts (A) and type I ISGs (B) in BMDCs cocultured with talazoparib-treated ovarian cancer cells in the presence or absence of LPA. Data were normalized to Actb in all cases (n = 4). C–E, C57BL/6J mice (n = 16–19/group) were challenged via i.p. injection with 1.5 × 106 control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer cells. After 7 days, mice were treated once daily with vehicle or talazoparib (0.33 mg/kg) by oral gavage for up to 28 days. C, Ascites accumulation was denoted as percentage weight gain over time. D, Changes in ascites development were analyzed by calculating the area above the curve in each experimental group starting on day 35 and a cutoff of 35% of weight gain. Data are represented as percentage change compared with the control sgRNA group treated with vehicle. E, Overall survival curves for the same mice as described in C and D. F, Survival experiments were repeated as described in CE but including STING-deficient (KO) hosts (n = 6–10/group). Data in AD are shown as mean ± SEM. A and B, Two-way ANOVA (Tukey multiple comparisons test). D, One-way ANOVA with Tukey multiple comparisons test. E and F, Log-rank test analysis for survival. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

An LPA-Controlled Gene Signature Predicts Resistance to Combined PARP and PD-1 Inhibition in the Clinic

Patients with ovarian cancer are resistant to multiple forms of cancer immunotherapy, especially to the blockade of typical immune checkpoints such as PD-1 (19–21). Nonetheless, recent reports indicate that the immunostimulatory effects of PARP inhibition, mainly mediated by type I IFN, can enhance responses to anti–PD-1 therapy in patients with ovarian cancer and mouse models of this disease (42, 46). We hypothesized that tumor-intrinsic ATX/LPA could represent a mechanism of resistance to this combination treatment. Mice bearing ATX-sufficient ovarian cancer failed to respond to PD-1 blockade (Fig. 7A). Although these mice demonstrated a modest increase in survival upon treatment with talazoparib, the addition of anti–PD-1 antibodies did not improve the observed therapeutic benefit (Fig. 7A). Of note, PD-1 blockade enhanced the effects of talazoparib only in mice bearing ATX-deficient ovarian tumors devoid of LPA (Fig. 7A). Hence, we next examined whether the expression status of LPA-regulated genes that we identified (Fig. 3AC) is associated with response to combined immunotherapy with a PARP inhibitor plus checkpoint blockade in patients with ovarian cancer. To this end, we analyzed NanoString mRNA expression data from 44 tumor specimens within the TOPACIO/KEYNOTE-162 trial (NCT02657889), in which patients with HGSOC were treated with a combination of the PARP inhibitor niraparib and the PD-1 blocker pembrolizumab (46). Importantly, we identified seven LPA-controlled genes whose expression was significantly decreased in tumors from the nonresponder versus the responder patients in this trial (Fig. 7B and C). These genes were not differentially expressed due to prior chemotherapy exposure (Supplementary Fig. S8A), and further correlation analyses indicated that they form a coregulated module selectively found in tumors from the nonresponders (Fig. 7B; Supplementary Fig. S8B and S8C). Indeed, downregulation of the LPA-controlled gene module was observed in 77% of the nonresponder group in unsupervised hierarchical clustering (Fig. 7D). We generated an LPA signature score denoting a summary of the gene module expression and found that, regardless of their homologous recombination deficiency (HRD) status, nonresponders demonstrated significantly higher scores than responder patients in this trial (Fig. 7E). Hence, the status of this LPA-controlled gene signature might be useful for predicting resistance to combined PARP and PD-1 inhibition in patients with HGSOC.

Figure 7.

Expression of LPA-controlled genes and clinical response to combined PARP and PD-1 inhibition. A, C57BL/6J female mice (n = 7–16/group) were challenged i.p. with control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer cells. After 7 days, mice were treated with talazoparib alone or in combination with isotype control or anti–PD-1 antibodies as described in Methods, and overall survival was monitored. B, Correlation analysis within the nonresponder group uncovering a distinct module of coregulated genes (highlighted with red dashed lines). C, Differential mRNA expression analysis between the groups showing that the seven coregulated genes are significantly decreased in nonresponder patients. D, Hierarchical clustering analysis of patient groups and LPA signature score, defined as the inverse of the median expression. E, LPA signature score distribution as a function of the response category of each sample (Kruskal–Wallis test). A, Log-rank test for survival. ***, P < 0.001; ****, P < 0.0001. NA, not associated.

Figure 7.

Expression of LPA-controlled genes and clinical response to combined PARP and PD-1 inhibition. A, C57BL/6J female mice (n = 7–16/group) were challenged i.p. with control or Enpp2-null ID8-Defb29/Vegf-A ovarian cancer cells. After 7 days, mice were treated with talazoparib alone or in combination with isotype control or anti–PD-1 antibodies as described in Methods, and overall survival was monitored. B, Correlation analysis within the nonresponder group uncovering a distinct module of coregulated genes (highlighted with red dashed lines). C, Differential mRNA expression analysis between the groups showing that the seven coregulated genes are significantly decreased in nonresponder patients. D, Hierarchical clustering analysis of patient groups and LPA signature score, defined as the inverse of the median expression. E, LPA signature score distribution as a function of the response category of each sample (Kruskal–Wallis test). A, Log-rank test for survival. ***, P < 0.001; ****, P < 0.0001. NA, not associated.

Close modal

Type I IFNs are required for optimal cancer immunosurveillance and for the effective immunologic control of malignant progression by diverse therapeutic modalities (23). However, the dominant mechanisms through which aggressive tumors inhibit protective type I IFN remain incompletely understood. Our study uncovers that ATX–LPA is a major immunoregulatory axis that curtails protective type I IFN responses in metastatic ovarian cancer.

We found that malignant cells in the ovarian tumor microenvironment are the main source of ATX and that LPA locally generated by this enzyme can readily control the immune contexture of metastatic ovarian cancer, modulating DC function and preventing the accumulation and activation of T and NK cells at tumor locations. Although recent reports suggest that LPA can directly block the activation and intratumoral infiltration of cytotoxic CD8+ T cells (47, 48), the role of this bioactive lipid as a major regulator of DC function and type I IFN responses in immunosuppressive tumors such as ovarian cancer had not been established.

We determined that multiple DC types present at tumor locations and lymphoid tissue exhibited high expression of genes encoding various LPA receptors, suggesting that these myeloid cells could readily sense and respond to this bioactive lipid. Indeed, LPA-exposed DCs demonstrated marked transcriptomic alterations and produced copious amounts of PGE2 via p38 MAPK activation, which inhibited autocrine type I IFN responses through the prostanoid receptor EP4. Defining the main LPA receptor(s) and the precise downstream signaling events mediating the observed modulation of type I IFN responses in DCs exposed to this phospholipid will be of significant interest. Whether tumor-derived LPA can simultaneously alter the function of other ovarian cancer–infiltrating leukocytes to evade immune control and promote malignant progression also warrants further investigation.

Loss of ATX in ovarian cancer cells decreased LPA and PGE2 production at tumor sites, enhanced type I IFN responses in tumor-associated DCs, and augmented the anti–ovarian cancer effects of treatment with canonical inducers of type I IFNs such as poly (I:C) or the PARP inhibitor talazoparib. The improved survival benefit conferred by these treatments in mice bearing ATX-deficient ovarian cancer was reduced upon IFNAR1 blockade or STING ablation, respectively. Nonetheless, untreated mice bearing ATX-deficient ovarian cancer still demonstrated a significant increase in overall survival that was not mediated by type I IFN signaling or STING activation. Although these results are consistent with the reported direct protumorigenic role of LPA in cancer cells, they also suggest that this bioactive lipid may coordinate additional immunoregulatory mechanisms beyond type I IFN suppression, which deserves further exploration.

Increasing evidence demonstrates the potent immunosuppressive role of tumor-derived PGE2 in multiple cancers (29, 49–51). This prostanoid has also been shown to inhibit type I IFN responses in macrophages stimulated with bacterial LPS and in the setting of viral infections (35, 52). Yet a role for tumor-derived LPA as a direct inducer of PGE2 that inhibits global type I IFN responses and limits the efficacy of ovarian cancer immunotherapies had not been established.

The recent phase II TOPACIO trial investigated the combination of PARP inhibition plus PD-1 blockade in relapsed, platinum-resistant ovarian cancer patients (53). The correlative analyses indicated that HRD status was associated with prolonged progression-free survival but not with the proportion of patients with a response (46), indicating that additional factors contribute to the response/resistance to this combination treatment. Indeed, our study identified a DC-derived, LPA-controlled gene module in which decreased expression in tumor specimens from patients with HGSOC was associated with resistance to combined PARP and PD-1 inhibition in this trial (46). Whether nonresponder patients demonstrate superior ATX–LPA production in the tumor or increased expression of LPA receptors on infiltrating immune cells deserves further investigation, as it could lead to novel actionable biomarkers predicting response or resistance to this combination therapy. Whether other mechanisms of immunoregulation predominant in ovarian tumors, such as persistent endoplasmic reticulum stress responses (54), cooperate with detrimental LPA signaling to exacerbate tumor growth, immune escape, and resistance to combination immunotherapy also deserves additional research.

Various ATX small-molecule inhibitors are currently being tested in human clinical trials, particularly in the setting of pulmonary fibrosis (55), raising the possibility that these compounds might be repurposed to improve the anticancer effects of interventions that rely on potent type I IFN responses. Beyond ovarian tumors, ATX–LPA is commonly overproduced in pancreatic, breast, and prostate cancers, implying that targeting this axis may also be useful to unleash type I IFN and enhance the efficacy of immune-activating treatments in these aggressive malignancies.

Patient-Derived Specimens

Plasma samples from cancer-free women were obtained from the New York Blood Center. Plasma and ascites samples from patients with stage III–IV HGSOC were obtained under written informed consent following appropriate institutional biospecimen collection and use protocols established at Weill Cornell Medicine and Memorial Sloan Kettering Cancer Center. All human specimens were deidentified for subsequent experimental analyses. The ascites was centrifuged at 4°C for 10 minutes at 1,300 rpm. Supernatants were then collected, depleted of cells by passing through 0.22-μm filters, and stored frozen at −80°C in small aliquots until use.

Mice and Experimental Ovarian Cancer Models

C57BL/6J, B6.129S(C)-Batf3tm1Kmm/J (Batf3 KO), and Tmem173gt/J (Sting KO) female mice were obtained from The Jackson Laboratory. C57BL/6NTac.Cg-Rag2tm1Fwa Il2rgtm1Wjl (Rag2/Il2rg double KO) female mice were obtained from Taconic Biosciences. Female mice were housed in pathogen-free microisolator cages at the animal facilities of Weill Cornell Medicine and used at 8 to 12 weeks of age. Mouse experiments were approved by the Institutional Animal Care and Use Committee of Weill Cornell Medicine under protocol number 2011-0098. Parental ID8 cells expressing luciferase (ID8-Luc) and the aggressive ID8-Defb29/Vegf-A derivate were cultured and used as previously described (24, 56). Both cell lines were obtained under MTA from Drs. K. Roby (University of Kansas Medical Center) and J. Conejo-Garcia (H. Lee Moffitt Cancer Center and Research Institute), respectively. The PPNM cell line (Trp53−/−R172HPten−/−Nf1−/−MycOE) was generously provided by Dr. R. Weinberg (The Whitehead Institute) under MTA (27). Briefly, 1.5 × 106 ID8-based ovarian cancer cells suspended in 200 μL of sterile PBS were i.p. injected into WT or transgenic mice. Alternatively, 1.5 × 106 PPNM cells were suspended in PBS containing Matrigel (Corning Matrigel matrix; cat. #47743-716) at a 1:1 ratio, and 200 μL of the mix were administered i.p. into WT mice, as reported (27). Metastatic progression, ascites accumulation, and host survival were monitored over time. Tumor burden in the peritoneal cavity was assessed by live bioluminescence imaging. Briefly, mice were given a single i.p. injection of VivoGlo luciferin (2 mg/mouse, Promega) and then imaged on a Xenogen IVIS Spectrum In Vivo Imaging System at the Weill Cornell Research Animal Resource Center.

Generation of ATX-Deficient Ovarian Cancer Cell Lines Using CRISPR/Cas9

The 20-nucleotide CRISPR RNA (crRNA) targeting murine Enpp2 (Mus musculus chromosome 15, 15; 15 D1, NC_000081.7) was directed at the genomic sequence TCTCCATGGACCAACACATCTGG (the three additional nucleotides highlighted in bold represent the protospacer adjacent motif, or PAM). This target sequence corresponds to exon 3 of the murine Enpp2 transcript and was manually chosen by identifying a 20-nucleotide fragment immediately upstream of the highlighted PAM (57). The on- and off-target effects of the manually selected CRISPR sequence were then analyzed using the Broad Institute's Genetic Perturbation Platform (https://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design). To validate the genomic editing capacity of the crRNA, RT-qPCR was performed on total RNA isolated from cells transfected with sgRNA–Cas9 complexes containing the Enpp2 crRNA described above. The primer for Enpp2 quantification via RT-qPCR anneals to the same nucleotides as the Enpp2 crRNA target site. The primers for evaluating deletion efficacy are listed in Supplementary Table S1. The scrambled crRNA contains a 20-nucleotide sequence that is computationally designed to be nontargeting within the human or murine genomes (http://sfvideo.blob.core.windows.net/sitefinity/docs/default-source/user-guide-manual/alt-r-crispr-cas9-user-guide-ribonucleoprotein-transfections-recommended.pdf?sfvrsn=1c43407_12). The sequence for this nontargeting sgRNA control was CGUUAAUCGCGUAUAAUACG. As ATX expression in cancer cells is commonly repressed in vitro via epigenetic mechanisms (58), ID8-Defb29/Vegf-A or PPNM cancer cells were treated for 4 hours with the histone deacetylase inhibitor TSA at 250 nmol/L (Sigma-Aldrich; cat. #T1952-200UL) and then transfected with ATTO-550–labeled sgRNA–Cas9 complexes using the Neon transfection system following the manufacturer's protocol (Thermo Fisher Scientific; https://assets.thermofisher.com/TFS-Assets/LSG/manuals/neon_device_man.pdf). All materials for sgRNA–Cas9 complex generation were purchased from Integrated DNA Technologies and prepared as instructed (http://sfvideo.blob.core.windows.net/sitefinity/docs/default-source/user-guide-manual/alt-r-crispr-cas9-user-guide-ribonucleoprotein-transfections-recommended.pdf?sfvrsn=1c43407_12). Twenty-four hours after transfection, ATTO-550+ cells were individually sorted, expanded as clones, and screened for ATX ablation for subsequent experiments.

Primary Cell Isolation and Analysis

Cancer cells (CD45SSChi), total DCs (CD45+CD11c+MHC-II+CD11b+), macrophages (CD45+CD11b+F4/80hi), T cells (CD45+CD3+), and neutrophils (CD45+CD11b+F4/80Ly6G+) were sorted from single-cell suspensions of malignant ascites or peritoneal lavage samples obtained from mice bearing metastatic ovarian cancer. Murine BMDCs were generated from bone marrow precursor cells isolated from the tibias and femurs of mice. Bone marrow cells were flushed and plated on bacteriologic plates at 3 × 106/plate in 10 mL of complete RPMI media (RPMI + L-glutamine + 10% FBS + HEPES + sodium pyruvate + nonessential amino acids + β-mercaptoethanol + penicillin/streptomycin) containing 20 ng/mL of recombinant granulocyte–macrophage colony-stimulating factor (GM-CSF; PeproTech; cat. #315-03). Three days later, an equal volume of the media described above was added to the culture, and nonadherent cells were harvested 3 to 4 days thereafter (days 6–7). BMDCs were further enriched with UltraPure CD11c MicroBeads (Miltenyi Biotech; cat. #130-125-835) and used directly for in vitro profiling and functional assays. Murine sDC and pDCs were magnetically purified from spleens (Miltenyi Biotec; cat. #130-125-835 and 130-107-093) and used directly for subsequent in vitro functional assays. Human monocyte–derived DCs were generated by isolating CD14+ cells (Miltenyi Biotec; cat. #130-050-201) from blood/buffy coats using Ficoll-gradient centrifugation and plated in complete RPMI media containing human recombinant GM-CSF at 1,000 IU/mL and IL4 at 500 IU/mL (both from PeproTech; cat. #300-03 for GM-CSF, 200-04 for IL4) for 7 days. Cells were then harvested and used for in vitro assays (59). Human pDCs from healthy donors were purified as previously described (60) using negative selection (Miltenyi Biotec; cat. #130-090-509).

RNA Extraction and Quantitative RT-PCR Analysis

Total RNA was isolated using the RNeasy Mini Kit or QIAzol lysis reagent (Qiagen) according to the manufacturer's instructions. RNA (0.1–1 μg) was reverse-transcribed to generate cDNA using the qScript cDNA synthesis kit (Quantabio). Quantitative RT-PCR was performed using PerfeCTa SYBR green fastmix (Quantabio) on a QuantStudio 6 Flex real-time PCR system (Applied Biosystems). Normalized gene expression was calculated by the comparative threshold cycle method using ACTB for human or Actb for mouse as endogenous controls.

Flow Cytometry

Analyses were conducted using fluorochrome-conjugated antibodies purchased from BioLegend, unless stated otherwise. Cells were washed with PBS, Fc-gamma receptor–blocked using TruStain fcX (anti-mouse CD16/32, clone 93; cat. #101319), and then stained for surface markers at 4°C in the dark for 30 minutes using the following antibodies: anti-CD45 (clone 30-F11; cat. #103115), anti-CD3 (clone 17A2; cat. #100216), anti-CD4 (clone RM4-5; cat. #100547), anti-CD8α (clone 53-6.7; cat. #100725), anti-CD44 (clone IM7; cat. #103032), anti-CD11c (clone N418; cat. #117309), anti–I-A/I-E (Tonbo Biosciences, clone M5/114.15.2; cat. #35-5321), anti-CD11b (clone M1/70; cat. #101228), anti-F4/80 (clone BM8; cat. #123110), anti-Gr1 (clone RB6-8C5; cat. #108416), anti-Ly6G (Tonbo Biosciences, clone 1A8; cat. #25-1276) and anti-NK1.1 (clone PK136; cat. #108748). Cells were then washed and stained with DAPI for live/dead discrimination. For intracellular cytokine staining, cells from malignant peritoneal wash samples were stimulated for 5 hours in a complete RPMI containing cell activation cocktail with brefeldin A (BioLegend; cat. #423304). Cells were collected and stained for surface markers and intracellular cytokines using anti-IFNγ (clone XMG1.2; cat. #505826) and anti-TNFα (clone MP6-XT22; cat. #506308), following the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific; cat. #00-5523-00). Flow cytometry was performed on LSRII or Fortessa-X20 instruments (BD Biosciences). Cell populations were sorted from peritoneal lavage or ascites samples from ovarian cancer–bearing mice using a FACSAria sorter (BD Biosciences), and flow cytometry data were analyzed using FlowJo (TreeStar).

Immunoblot Analysis

BMDCs were washed twice in 1× cold PBS, and cell pellets were lysed using RIPA lysis and extraction buffer (Thermo Fisher Scientific) supplemented with a protease and phosphatase inhibitor tablet (Roche). Homogenates were centrifuged at 14,000 rpm for 30 minutes at 4°C, and the supernatants were collected. Protein concentrations were determined using a BCA protein assay kit (Thermo Fisher Scientific). Equivalent amounts of protein were separated via SDS-PAGE and transferred onto PVDF membranes following standard protocols. The following antibodies were used: anti–beta actin (Cell Signaling Technology; cat. #4967), anti-pTBK1 (Cell Signaling Technology; cat. #5438), anti-TBK1 (Cell Signaling Technology; cat. #3504), anti-pIRE3 (Cell Signaling Technology; cat. #29047), anti-IRE3 (Cell Signaling Technology; cat. #4302), and goat anti-rabbit secondary antibodies conjugated with HRP (Thermo Fisher Scientific; cat. #G-21234). SuperSignal West Pico and Femto chemiluminescent substrates (Thermo Fisher Scientific) were used to image blots in a FluorChemE instrument (ProteinSimple).

ELISA

Total LPA, ATX, and PGE2 concentrations in malignant ascites from patients with ovarian cancer and mice bearing advanced ovarian cancer were measured by ELISA (Echelon Biosciences; cat. #K-2800S for LPA; Echelon Biosciences; cat. #K-5600 for ATX; Enzo Lifesciences; cat. #ADI-900-001 for PGE2). BMDCs were stimulated with the indicated concentration of LPA (Cayman Chemical Company; cat. #65528-98-5) for various time points, and PGE2 levels in the supernatant were measured using the PGE2 ELISA kit described above. Ovarian cancer cells were stimulated with TSA (Sigma-Aldrich; cat. #T1952) for 24 hours, and secreted ATX was quantified in the supernatant using the ATX ELISA kit. Murine BMDCs, sDCs, and pDCs were left untreated or pretreated for 2 hours with LPA (10 or 100 μmol/L), and cells were then stimulated for 4 hours with LPS (100 ng/mL; Invivogen; cat. #tlrl-eklps), 24 hours with Poly (I:C) (50 μg/mL for BMDCs or 10 μg/mL for sDCs; Invivogen; cat. #vac-plc), or 24 hours with CpG ODN1585 (1 μg/mL; Invivogen; cat. #tlrl-1585). Production of IFNβ in culture supernatants was determined by ELISA (PBL Assay Science; cat. #42400-2). Human moDCs or purified pDCs from healthy donors were pretreated with LPA (100 μmol/L) for 2 hours, and cells were then stimulated for 4 hours with LPS (100 ng/mL), 24 hours with poly (I:C) (10 μg/mL), or 24 hours with CpG-C274 (0.075 μmol/L). Production of IFNβ was quantified in culture supernatants by ELISA (PBL Assay Science; cat. #41435-1). IFNα production by human pDCs was determined by ELISA (Mabtech; cat. #3425-1H-6) according to the manufacturer's protocol. Plates were read using a Varioskan instrument (Thermo Fisher Scientific).

Treatment of BMDCs with Inhibitors and Antagonists In Vitro

BMDCs were independently pretreated for 1 hour with the p38 MAPK inhibitor SB203580 (10 μmol/L; Invivogen; cat. #tlrl-sb20), the PPAR alpha antagonist GW6471 (50 μmol/L; Selleck Chemicals; cat. #S2798), the PPAR delta antagonist GSK3787 (10 μmol/L; Selleck Chemicals; cat. #S8025), the CREB inhibitor 666-15 (10 μmol/L; MedChem Express; cat. #HY-128686), the JNK inhibitor JNK-IN-8 (10 μmol/L; Selleck Chemicals; cat. #S4901), the NFκB inhibitor BAY 11-7821 (50 μmol/L; MedChem Express; cat. #HY-13453), or the NFκB/AP1 dual inhibitor SP 100030 (5 μmol/L; Tocris; cat. #5309) and then stimulated with LPA (100 μmol/L) for 2 hours. Expression of Ptgs2 was determined by RT-qPCR. For the in vitro assessment of EP4 signaling, BMDCs were pretreated for 1 hour with the EP4 antagonist PGN 1531 (5 μmol/L; Tocris; cat. #5327) and then stimulated with LPA (100 μmol/L) and LPS (100 ng/mL) for 4 hours. Expression of type I ISGs was subsequently quantified by RT-qPCR.

BMDC RNA-seq and Bioinformatic Analyses

Purified CD11c+ BMDCs were cultured overnight in complete RPMI media (described above) lacking FBS to promote LPA sensing (61). Cells were then left untreated or exposed to LPA (100 μmol/L) for 2 or 6 hours, and total RNA was subsequently isolated using the RNeasy MinElute kit (Qiagen). All samples passed the RNA quality control examined by Agilent Bioanalyzer 2100, and mRNA libraries were generated and sequenced at the Weill Cornell Epigenomics Core Facility. RNA-seq data were aligned using bowtie2 (62) against the mm10 genome, and RSEM v1.2.12 software (63) was used to estimate gene-level read counts using Ensembl transcriptome information. DESeq2 (64) was used to estimate the significance of differential expression difference between any two experimental groups. Gene expression changes were considered significant if passed the FDR < 5% threshold at both 2 and 6 hours of LPA treatment against the untreated condition. Gene set enrichment analysis was done using Ingenuity Pathway Analysis (IPA, Qiagen) using the “Upstream regulators” option. Regulators at P < 10−6 with significant predicted activation/inhibition states (Z-score > 2) were reported. Data were deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE182062.

Single-Cell RNA-seq Analysis

Viable leukocytes (DAPICD45+) were isolated by FACS from the ascites of mice bearing ID8-Defb29/Vegf-A ovarian cancer for 25 days (n = 4 independent mice), and cells (∼8,000) were then processed for single-cell RNA-seq at the Genomics Resources Core Facility of Weill Cornell Medicine. Raw gene expression matrices were generated for each sample by the Cell Ranger (v.3.0.2) Pipeline coupled with mouse reference version GRCm38 (mm10). The output-filtered gene expression matrices were analyzed by R software (v.3.5.3) with the Seurat package (v.3.0.0). In brief, genes expressed at a proportion >0.1% of the data and cells with >200 genes detected were selected for further analyses. Low-quality cells were removed if they met the following criteria: <800 unique molecular identifiers (UMI), <500 genes, or >5% UMIs derived from the mitochondrial genome. Further, gene expression matrices were normalized by the Normalize­Data function, and 2,000 features with high cell-to-cell variation were calculated using the FindVariableFeatures() function. To reduce the dimensionality of the data sets, the RunPCA() function was conducted with default parameters on linear transformation–scaled data generated by the ScaleData() function. In the end, we clustered cells using the FindNeighbors() and FindClusters() functions and performed nonlinear dimensional reduction with the RunUMAP() function with default settings. All details regarding the Seurat analyses performed in this work can be found in the website tutorial (https://satijalab.org/seurat/v3.0/pbmc3k_tutorial.html). Data were deposited in the NCBI GEO under accession number GSE182047.

In Vitro Coculture of Cancer Cells and BMDCs

ID8-Defb29/Vegf-A cancer cells were treated with DMSO (vehicle) or talazoparib at 0.2, 2, or 10 μmol/L for 24 hours. Cells were then washed twice with PBS and cocultured with BMDCs at a 1:1 ratio in the presence or absence of LPA (100 μmol/L) for 24 hours. Nonadherent cells, corresponding to >90% BMDCs by FACS, were then collected for gene expression analyses.

NanoString mRNA Expression Analysis from Human Tissue

Forty-four formalin-fixed, paraffin-embedded HGSOC samples from the TOPACIO trial (n = 9 responders and 35 nonresponders) were analyzed with a NanoString assay of 780 genes, as previously described (46). The normalized log-transformed NanoString expression scores were filtered to keep only the genes (n = 36) that were differentially regulated by LPA exposure in BMDCs (Fig. 3B and C). Genes significantly associated with chemo-experience status of the samples were removed from the downstream analysis (VEGFA, CCL4, and HAVC2) using Kruskal–Wallis rank test (P < 0.1 and |fold change| > 2; Supplementary Fig. S8A). Spearman correlation within each response group was then used to identify coexpressed genes specific to the response category that were not shaped by an imbalance of categories. The LPA signature score for each sample was defined as the inverse of the median expression of the seven downregulated genes identified. Heat map visualization utilizes hierarchical clustering of Euclidean distances with complete linkage. The HRD status of the tumors analyzed has been previously described (46).

Lipidomics

LPA species in plasma, malignant ascites, or peritoneal wash samples were analyzed and quantified using LC-MS at the Lipidomics Core Facility of Wayne State University School of Medicine.

Measurement of Cancer Cell Proliferation and Viability by MTT Assay

Control or Enpp2-null ovarian cancer cells were seeded in 96-well plates at a density of 6,000/well and allowed to adhere overnight. Cells were then exposed to LPA for 24 and 48 hours. The proliferation and viability of cancer cells were measured by MTT (Sigma-Aldrich).

Analysis of Ovarian Cancer Patient Survival

Kaplan–Meier Plotter (65) was used to evaluate the correlation between ENPP2/IFNA1 and PTGS2/IFNA1 expression ratios in tumor specimens and the overall survival of patients with ovarian cancer in the TCGA data set. Analyses were performed using only JetSet validated probes and the autoselect cutoff feature in all tumors from the serous histology, irrespective of p53, debulking, or chemotherapy status (n = 557 patients).

In Vivo Treatments

WT C57BL/6J, C57BL/6NTac.Cg-Rag2tm1Fwa Il2rgtm1Wjl (Rag2/Il2rg double KO), or B6.129S(C)-Batf3tm1Kmm/J (Batf3 KO) mice were implanted via i.p. injection with 1.5 × 106 control or Enpp2-null ovarian cancer cells. After 10 days, mice were i.p. treated every 5 days a total of 5 times with vehicle control (distilled water) or poly (I:C) at 100 μg/mouse (Invivogen; cat. #vac-plc). For the in vivo assessment of type I IFN signaling, mice were i.p. treated every 3 days 15 times with isotype control antibodies (Bio X Cell; cat. #BE0083) or anti-IFNAR blocking antibodies (Bio X Cell; cat. #BE0241) at 200 μg/mouse, starting 3 days after tumor implantation. For the in vivo stimulation of EP4 signaling, mice were orally administered vehicle control or the EP4 agonist KAG-308 (3 mg/kg; MedChem Express; cat. #HY-128686) every day for a total of 25 days, starting 9 days after tumor implantation. WT C57BL/6J mice or Tmem173gt/J (Sting KO) mice were challenged i.p. with 1.5 × 106 control or Enpp2-null ovarian cancer cells. After 7 days, mice were treated daily via oral gavage with vehicle control [10% N,N-Dimethylacetamide (DMAc), Sigma, 1% Solutol, 89% PBS] or talazoparib at 0.33 mg/kg (MedChem Express; cat. #HY-16106) for the indicated times. For the in vivo assessment of checkpoint blockade using anti–PD-1 antibodies, mice were i.p. treated every 3 days a total of 7 times with isotype control (Bio X Cell; cat. #BE0089) or anti–PD-1 blocking antibodies (Bio X Cell; cat. #BE0146) at 200 μg/mouse, starting 10 days after ovarian cancer challenge.

Statistical Analyses

All statistical analyses were performed using GraphPad Prism software (version 9). Significance for pairwise correlation analyses was calculated using the Spearman correlation coefficient (r). Comparisons between two groups were assessed using unpaired or paired (for matched comparisons) two-tailed Student t test. Multiple comparisons were assessed by one-way ANOVA including Tukey multiple comparisons test. Grouped data were analyzed by two-way ANOVA including the Tukey or Šídák multiple comparisons test. Host survival rates were compared using the log-rank test. Data are presented as mean ± SEM or violin plots. Unless otherwise stated, P < 0.05 was considered statistically significant.

Data Availability

Data generated in this study are available in the GEO database under the accession numbers GSE182062 for bulk RNA-seq and GSE182047 for single-cell RNA-seq. Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Juan R. Cubillos-Ruiz (jur2016@med.cornell.edu).

C.-S. Chae reports grants from CRI Irvington during the conduct of the study, as well as a patent for modulation of dendritic cell function by the phospholipid messenger LPA (WO2021034414A2) pending. C. Salvagno reports grants from CRI Irvington and the Ovarian Cancer Research Alliance during the conduct of the study. A. Emmanuelli reports grants from the NIH during the conduct of the study. F.J. Barrat reports other support from Ipinovyx Bio and personal fees from AstraZeneca outside the submitted work. E.A. Romero-Sandoval reports grants from the NIH during the conduct of the study; personal fees from Dove Medical Press—Journal of Pain Research outside the submitted work; and is Deputy Editor-in-Chief of the Journal of Pain Research. D. Zamarin reports grants and personal fees from AstraZeneca and Genentech, personal fees from GSK, Synlogic Therapeutics, Trieza Therapeutics, Xencor, Memgen, Agenus, and Merck, grants from Plexxikon, other support from Immunos, and personal fees and other support from Calidi Biotherapeutics, Synthekine, Accurius, Mana Therapeutics, and Targovax outside the submitted work, as well as a patent for Newcastle disease virus for cancer therapy licensed to Merck. A.D. D'Andrea reports nonfinancial support from AstraZeneca and Constellation Pharma, personal fees from Bayer AG, Faze Medicines, GSK, LAV Global Management Company Limited, L.E.K. Consulting, Patheon Pharmaceuticals, and Pfizer, grants and other support from Bristol Myers Squibb and Lilly Oncology, personal fees and other support from Cedilla Therapeutics, Cyteir, Ideaya, Impact Therapeutics, Oncolinea, and Zentalis Phamaceuticals/Zeno Management, other support from Epizyme, grants and personal fees from Merck KGAa/EMD Serono and Tango Therapeutics, and grants from Moderna. A. Färkkilä reports grants from the Academy of Finland, the Cancer Society of Finland, the Sigrid Juselius Foundation, and the Finnish Medical Foundation during the conduct of the study. J.R. Cubillos-Ruiz reports grants from the U.S. Department of Defense, Stand Up To Cancer, the NIH, the Pershing Square Sohn Cancer Research Alliance, the Mark Foundation for Cancer Research, the Cancer Research Institute, and the Ovarian Cancer Research Alliance during the conduct of the study; personal fees from NextRNA Therapeutics and Quentis Therapeutics, and other support from Autoimmunity Biologic Solutions outside the submitted work; and a patent for modulation of dendritic cell function by the phospholipid messenger LPA (WO2021034414A2) pending. No disclosures were reported by the other authors.

C.-S. Chae: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. T.A. Sandoval: Formal analysis, investigation, visualization, methodology. S.-M. Hwang: Data curation, formal analysis, visualization, methodology. E.S. Park: Investigation, visualization. P. Giovanelli: Formal analysis, validation, investigation, methodology. D. Awasthi: Validation, investigation. C. Salvagno: Validation, investigation, methodology. A. Emmanuelli: Investigation. C. Tan: Investigation. V. Chaudhary: Resources, investigation. J. Casado: Data curation, formal analysis, visualization, methodology. A.V. Kossenkov: Data curation, formal analysis, visualization. M. Song: Investigation, methodology. F.J. Barrat: Resources, investigation, methodology. K. Holcomb: Resources, funding acquisition, investigation, visualization. E.A. Romero-Sandoval: Formal analysis, funding acquisition, investigation, visualization. D. Zamarin: Resources, investigation. D. Pépin: Resources, investigation, methodology. A.D. D'Andrea: Resources, funding acquisition, investigation, writing–review and editing. A. Färkkilä: Resources, data curation, formal analysis, supervision, investigation, visualization, methodology, writing–review and editing. J.R. Cubillos-Ruiz: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank the members of the Flow Cytometry Core Facility, the Epigenomics Core Facility, and the Genomics Resources Core Facility at Weill Cornell Medicine for their excellent assistance with cell sorting, bulk RNA-seq, and single-cell RNA-seq experiments, respectively. This research was supported by the U.S. Department of Defense Ovarian Cancer Research Program grants W81XWH-16-1-0438, OC190443, OC200166, and OC200224 (J.R. Cubillos-Ruiz); a Stand Up To Cancer Innovative Research Grant, Grant Number SU2C-AACR-IRG-03-16 (J.R. Cubillos-Ruiz); a Stand Up To Cancer Phillip A. Sharp Innovation in Collaboration Award, Grant Number SU2C-AACR-PS-24 (J.R. Cubillos-Ruiz and A.D. D'Andrea); NIH grants R01NS114653 (J.R. Cubillos-Ruiz and E.A. Romero-Sandoval), R21CA248106 (J.R. Cubillos-Ruiz and E.A. Romero-Sandoval), 1R01AI132446 (F.J. Barrat), and F31CA257631 (A. Emmanuelli); a Pershing Square Sohn Cancer Research Alliance grant (J.R. Cubillos-Ruiz); the Mark Foundation for Cancer Research ASPIRE Award (J.R. Cubillos-Ruiz); a Wade F. B. Thompson/Cancer Research Institute CLIP grant (J.R. Cubillos-Ruiz); the Ann Schreiber Mentored Investigator Award of the Ovarian Cancer Research Alliance (C. Salvagno); a CRI Irvington Postdoctoral Fellowship Award (C. Salvagno and C.-S. Chae); a National Research Foundation of Korea (NRF) grant 2020R1C1C1010303 (M. Song) and National Cancer Center of Korea grant NCC-203175 (M. Song); the Cancer Foundation Finland, Instrumentarium, and Maud Kuistila Memorial Foundation (J.R. Cubillos-Ruiz); and the Academy of Finland, Sigrid Juselius Foundation, and Cancer Foundation Finland (A. Färkkilä). Stand Up To Cancer is a division of the Entertainment Industry Foundation. The indicated Stand Up To Cancer Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C.

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.

1.
Benesch
MG
,
Tang
X
,
Maeda
T
,
Ohhata
A
,
Zhao
YY
,
Kok
BP
, et al
.
Inhibition of autotaxin delays breast tumor growth and lung metastasis in mice
.
FASEB J
2014
;
28
:
2655
66
.
2.
Chen
J
,
Li
H
,
Xu
W
,
Guo
X
.
Evaluation of serum ATX and LPA as potential diagnostic biomarkers in patients with pancreatic cancer
.
BMC Gastroenterol
2021
;
21
:
58
.
3.
Li
YY
,
Zhang
WC
,
Zhang
JL
,
Zheng
CJ
,
Zhu
H
,
Yu
HM
, et al
.
Plasma levels of lysophosphatidic acid in ovarian cancer versus controls: a meta-analysis
.
Lipids Health Dis
2015
;
14
:
72
.
4.
Liu
S
,
Umezu-Goto
M
,
Murph
M
,
Lu
Y
,
Liu
W
,
Zhang
F
, et al
.
Expression of autotaxin and lysophosphatidic acid receptors increases mammary tumorigenesis, invasion, and metastases
.
Cancer Cell
2009
;
15
:
539
50
.
5.
Yamada
T
,
Sato
K
,
Komachi
M
,
Malchinkhuu
E
,
Tobo
M
,
Kimura
T
, et al
.
Lysophosphatidic acid (LPA) in malignant ascites stimulates motility of human pancreatic cancer cells through LPA1
.
J Biol Chem
2004
;
279
:
6595
605
.
6.
Hayashi
M
,
Okabe
K
,
Kato
K
,
Okumura
M
,
Fukui
R
,
Fukushima
N
, et al
.
Differential function of lysophosphatidic acid receptors in cell proliferation and migration of neuroblastoma cells
.
Cancer Lett
2012
;
316
:
91
6
.
7.
Auciello
FR
,
Bulusu
V
,
Oon
C
,
Tait-Mulder
J
,
Berry
M
,
Bhattacharyya
S
, et al
.
A stromal lysolipid-autotaxin signaling axis promotes pancreatic tumor progression
.
Cancer Discov
2019
;
9
:
617
27
.
8.
Minami
K
,
Ueda
N
,
Ishimoto
K
,
Tsujiuchi
T
.
Lysophosphatidic acid receptor-2 (LPA2)-mediated signaling enhances chem-resistance in melanoma cells treated with anticancer drugs
.
Mol Cell Biochem
2020
;
469
:
89
95
.
9.
Fang
X
,
Gaudette
D
,
Furui
T
,
Mao
M
,
Estrella
V
,
Eder
A
, et al
.
Lysophospholipid growth factors in the initiation, progression, metastases, and management of ovarian cancer
.
Ann N Y Acad Sci
2000
;
905
:
188
208
.
10.
Fang
X
,
Schummer
M
,
Mao
M
,
Yu
S
,
Tabassam
FH
,
Swaby
R
, et al
.
Lysophosphatidic acid is a bioactive mediator in ovarian cancer
.
Biochim Biophys Acta
2002
;
1582
:
257
64
.
11.
Xiao
YJ
,
Schwartz
B
,
Washington
M
,
Kennedy
A
,
Webster
K
,
Belinson
J
, et al
.
Electrospray ionization mass spectrometry analysis of lysophospholipids in human ascitic fluids: comparison of the lysophospholipid contents in malignant vs nonmalignant ascitic fluids
.
Anal Biochem
2001
;
290
:
302
13
.
12.
Klymenko
Y
,
Bos
B
,
Campbell
L
,
Loughran
E
,
Liu
Y
,
Yang
J
, et al
.
Lysophosphatidic acid modulates ovarian cancer multicellular aggregate assembly and metastatic dissemination
.
Sci Rep
2020
;
10
:
10877
.
13.
Murph
MM
,
Liu
W
,
Yu
S
,
Lu
Y
,
Hall
H
,
Hennessy
BT
, et al
.
Lysophosphatidic acid-induced transcriptional profile represents serous epithelial ovarian carcinoma and worsened prognosis
.
PLoS One
2009
;
4
:
e5583
.
14.
Seo
EJ
,
Kwon
YW
,
Jang
IH
,
Kim
DK
,
Lee
SI
,
Choi
EJ
, et al
.
Autotaxin regulates maintenance of ovarian cancer stem cells through lysophosphatidic acid-mediated autocrine mechanism
.
Stem Cells
2016
;
34
:
551
64
.
15.
Perrakis
A
,
Moolenaar
WH
.
Autotaxin: structure-function and signaling
.
J Lipid Res
2014
;
55
:
1010
8
.
16.
Zhang
X
,
Li
M
,
Yin
N
,
Zhang
J
.
The expression regulation and biological function of autotaxin
.
Cells
2021
;
10
:
939
.
17.
Geraldo
LHM
,
Spohr
TCLS
,
Amaral
RFD
,
Fonseca
ACCD
,
Garcia
C
,
Mendes
FA
, et al
.
Role of lysophosphatidic acid and its receptors in health and disease: novel therapeutic strategies
.
Signal Transduct Target Ther
2021
;
6
:
45
.
18.
Brindley
DN
.
Lysophosphatidic acid signaling in cancer
.
Cancers
2020
;
12
:
3791
.
19.
Chae
CS
,
Teran-Cabanillas
E
,
Cubillos-Ruiz
JR
.
Dendritic cell rehab: new strategies to unleash therapeutic immunity in ovarian cancer
.
Cancer Immunol Immunother
2017
;
66
:
969
77
.
20.
Matulonis
UA
,
Sood
AK
,
Fallowfield
L
,
Howitt
BE
,
Sehouli
J
,
Karlan
BY
.
Ovarian cancer
.
Nat Rev Dis Primers
2016
;
2
:
16061
.
21.
Matulonis
UA
,
Shapira-Frommer
R
,
Santin
AD
,
Lisyanskaya
AS
,
Pignata
S
,
Vergote
I
, et al
.
Antitumor activity and safety of pembrolizumab in patients with advanced recurrent ovarian cancer: results from the phase II KEYNOTE-100 study
.
Ann Oncol
2019
;
30
:
1080
7
.
22.
Borden
EC
.
Interferons α and β in cancer: therapeutic opportunities from new insights
.
Nat Rev Drug Discov
2019
;
18
:
219
34
.
23.
Zitvogel
L
,
Galluzzi
L
,
Kepp
O
,
Smyth
MJ
,
Kroemer
G
.
Type I interferons in anticancer immunity
.
Nat Rev Immunol
2015
;
15
:
405
14
.
24.
Conejo-Garcia
JR
,
Benencia
F
,
Courreges
MC
,
Kang
E
,
Mohamed-Hadley
A
,
Buckanovich
RJ
, et al
.
Tumor-infiltrating dendritic cell precursors recruited by a beta-defensin contribute to vasculogenesis under the influence of Vegf-A
.
Nat Med
2004
;
10
:
950
8
.
25.
Cubillos-Ruiz
JR
,
Silberman
PC
,
Rutkowski
MR
,
Chopra
S
,
Perales-Puchalt
A
,
Song
M
, et al
.
ER stress sensor XBP1 controls anti-tumor immunity by disrupting dendritic cell homeostasis
.
Cell
2015
;
161
:
1527
38
.
26.
Song
M
,
Sandoval
TA
,
Chae
CS
,
Chopra
S
,
Tan
C
,
Rutkowski
MR
, et al
.
IRE1alpha-XBP1 controls T cell function in ovarian cancer by regulating mitochondrial activity
.
Nature
2018
;
562
:
423
8
.
27.
Iyer
S
,
Zhang
S
,
Yucel
S
,
Horn
H
,
Smith
SG
,
Reinhardt
F
, et al
.
Genetically defined syngeneic mouse models of ovarian cancer as tools for the discovery of combination immunotherapy
.
Cancer Discov
2021
;
11
:
384
407
.
28.
Fuertes
MB
,
Woo
SR
,
Burnett
B
,
Fu
YX
,
Gajewski
TF
.
Type I interferon response and innate immune sensing of cancer
.
Trends Immunol
2013
;
34
:
67
73
.
29.
Zelenay
S
,
van der Veen
AG
,
Bottcher
JP
,
Snelgrove
KJ
,
Rogers
N
,
Acton
SE
, et al
.
Cyclooxygenase-dependent tumor growth through evasion of immunity
.
Cell
2015
;
162
:
1257
70
.
30.
Wang
D
,
DuBois
RN
.
The role of prostaglandin E(2) in tumor-associated immunosuppression
.
Trends Mol Med
2016
;
22
:
1
3
.
31.
Singer
CA
,
Baker
KJ
,
McCaffrey
A
,
AuCoin
DP
,
Dechert
MA
,
Gerthoffer
WT
.
p38 MAPK and NF-kappaB mediate COX-2 expression in human airway myocytes
.
Am J Physiol Lung Cell Mol Physiol
2003
;
285
:
L1087
98
.
32.
Xu
L
,
Han
C
,
Lim
K
,
Wu
T
.
Cross-talk between peroxisome proliferator-activated receptor delta and cytosolic phospholipase A(2)alpha/cyclooxygenase-2/prostaglandin E(2) signaling pathways in human hepatocellular carcinoma cells
.
Cancer Res
2006
;
66
:
11859
68
.
33.
Li
H
,
Zhou
J
,
Wei
X
,
Chen
R
,
Geng
J
,
Zheng
R
, et al
.
miR-144 and targets, c-fos and cyclooxygenase-2 (COX2), modulate synthesis of PGE2 in the amnion during pregnancy and labor
.
Sci Rep
2016
;
6
:
27914
.
34.
Lai
ZZ
,
Yang
HL
,
Ha
SY
,
Chang
KK
,
Mei
J
,
Zhou
WJ
, et al
.
Cyclooxygenase-2 in endometriosis
.
Int J Biol Sci
2019
;
15
:
2783
97
.
35.
Perkins
DJ
,
Richard
K
,
Hansen
AM
,
Lai
W
,
Nallar
S
,
Koller
B
, et al
.
Autocrine-paracrine prostaglandin E2 signaling restricts TLR4 internalization and TRIF signaling
.
Nat Immunol
2018
;
19
:
1309
18
.
36.
Longhi
MP
,
Trumpfheller
C
,
Idoyaga
J
,
Caskey
M
,
Matos
I
,
Kluger
C
, et al
.
Dendritic cells require a systemic type I interferon response to mature and induce CD4+ Th1 immunity with poly IC as adjuvant
.
J Exp Med
2009
;
206
:
1589
602
.
37.
Scarlett
UK
,
Cubillos-Ruiz
JR
,
Nesbeth
YC
,
Martinez
DG
,
Engle
X
,
Gewirtz
AT
, et al
.
In situ stimulation of CD40 and Toll-like receptor 3 transforms ovarian cancer-infiltrating dendritic cells from immunosuppressive to immunostimulatory cells
.
Cancer Res
2009
;
69
:
7329
37
.
38.
Murahashi
Y
,
Yano
F
,
Chijimatsu
R
,
Nakamoto
H
,
Maenohara
Y
,
Amakawa
M
, et al
.
Oral administration of EP4-selective agonist KAG-308 suppresses mouse knee osteoarthritis development through reduction of chondrocyte hypertrophy and TNF secretion
.
Sci Rep
2019
;
9
:
20329
.
39.
Nishimura
R
,
Shirasaki
T
,
Tsuchiya
K
,
Miyake
Y
,
Watanabe
Y
,
Hibiya
S
, et al
.
Establishment of a system to evaluate the therapeutic effect and the dynamics of an investigational drug on ulcerative colitis using human colonic organoids
.
J Gastroenterol
2019
;
54
:
608
20
.
40.
Watanabe
Y
,
Murata
T
,
Amakawa
M
,
Miyake
Y
,
Handa
T
,
Konishi
K
, et al
.
KAG-308, a newly-identified EP4-selective agonist shows efficacy for treating ulcerative colitis and can bring about lower risk of colorectal carcinogenesis by oral administration
.
Eur J Pharmacol
2015
;
754
:
179
89
.
41.
Hildner
K
,
Edelson
BT
,
Purtha
WE
,
Diamond
M
,
Matsushita
H
,
Kohyama
M
, et al
.
Batf3 deficiency reveals a critical role for CD8alpha+ dendritic cells in cytotoxic T cell immunity
.
Science
2008
;
322
:
1097
100
.
42.
Ding
L
,
Kim
HJ
,
Wang
Q
,
Kearns
M
,
Jiang
T
,
Ohlson
CE
, et al
.
PARP inhibition elicits STING-dependent antitumor immunity in Brca1-deficient ovarian cancer
.
Cell Rep
2018
;
25
:
2972
80
.
43.
Shen
J
,
Zhao
W
,
Ju
Z
,
Wang
L
,
Peng
Y
,
Labrie
M
, et al
.
PARPi triggers the STING-dependent immune response and enhances the therapeutic efficacy of immune checkpoint blockade independent of BRCAness
.
Cancer Res
2019
;
79
:
311
9
.
44.
Pantelidou
C
,
Sonzogni
O
,
De Oliveria Taveira
M
,
Mehta
AK
,
Kothari
A
,
Wang
D
, et al
.
PARP inhibitor efficacy depends on CD8(+) T-cell recruitment via intratumoral STING pathway activation in BRCA-deficient models of triple-negative breast cancer
.
Cancer Discov
2019
;
9
:
722
37
.
45.
Boussios
S
,
Abson
C
,
Moschetta
M
,
Rassy
E
,
Karathanasi
A
,
Bhat
T
, et al
.
Poly (ADP-Ribose) polymerase inhibitors: talazoparib in ovarian cancer and beyond
.
Drugs R D
2020
;
20
:
55
73
.
46.
Färkkilä
A
,
Gulhan
DC
,
Casado
J
,
Jacobson
CA
,
Nguyen
H
,
Kochupurakkal
B
, et al
.
Immunogenomic profiling determines responses to combined PARP and PD-1 inhibition in ovarian cancer
.
Nat Commun
2020
;
11
:
1459
.
47.
Lee
SC
,
Dacheux
MA
,
Norman
DD
,
Balazs
L
,
Torres
RM
,
Augelli-Szafran
CE
, et al
.
Regulation of tumor immunity by lysophosphatidic acid
.
Cancers
2020
;
12
:
1202
.
48.
Matas-Rico
E
,
Frijlink
E
,
van der Haar Avila
I
,
Menegakis
A
,
van Zon
M
,
Morris
AJ
, et al
.
Autotaxin impedes anti-tumor immunity by suppressing chemotaxis and tumor infiltration of CD8(+) T cells
.
Cell Rep
2021
;
37
:
110013
.
49.
Wang
D
,
DuBois
RN
.
Role of prostanoids in gastrointestinal cancer
.
J Clin Invest
2018
;
128
:
2732
42
.
50.
Bottcher
JP
,
Bonavita
E
,
Chakravarty
P
,
Blees
H
,
Cabeza-Cabrerizo
M
,
Sammicheli
S
, et al
.
NK cells stimulate recruitment of cDC1 into the tumor microenvironment promoting cancer immune control
.
Cell
2018
;
172
:
1022
37
.
51.
Pelly
VS
,
Moeini
A
,
Roelofsen
LM
,
Bonavita
E
,
Bell
CR
,
Hutton
C
, et al
.
Anti-inflammatory drugs remodel the tumor immune environment to enhance immune checkpoint blockade efficacy
.
Cancer Discov
2021
;
11
:
2602
19
.
52.
Coulombe
F
,
Jaworska
J
,
Verway
M
,
Tzelepis
F
,
Massoud
A
,
Gillard
J
, et al
.
Targeted prostaglandin E2 inhibition enhances antiviral immunity through induction of type I interferon and apoptosis in macrophages
.
Immunity
2014
;
40
:
554
68
.
53.
Konstantinopoulos
PA
,
Waggoner
S
,
Vidal
GA
,
Mita
M
,
Moroney
JW
,
Holloway
R
, et al
.
Single-arm phases 1 and 2 trial of niraparib in combination with pembrolizumab in patients with recurrent platinum-resistant ovarian carcinoma
.
JAMA Oncol
2019
;
5
:
1141
9
.
54.
Chen
X
,
Cubillos-Ruiz
JR
.
Endoplasmic reticulum stress signals in the tumour and its microenvironment
.
Nat Rev Cancer
2021
;
21
:
71
88
.
55.
Zulfikar
S
,
Mulholland
S
,
Adamali
H
,
Barratt
SL
.
Inhibitors of the autotaxin-lysophosphatidic acid axis and their potential in the treatment of interstitial lung disease: current perspectives
.
Clin Pharmacol
2020
;
12
:
97
108
.
56.
Roby
KF
,
Taylor
CC
,
Sweetwood
JP
,
Cheng
Y
,
Pace
JL
,
Tawfik
O
, et al
.
Development of a syngeneic mouse model for events related to ovarian cancer
.
Carcinogenesis
2000
;
21
:
585
91
.
57.
Ran
FA
,
Hsu
PD
,
Wright
J
,
Agarwala
V
,
Scott
DA
,
Zhang
F
.
Genome engineering using the CRISPR-Cas9 system
.
Nat Protoc
2013
;
8
:
2281
308
.
58.
Li
S
,
Wang
B
,
Xu
Y
,
Zhang
J
.
Autotaxin is induced by TSA through HDAC3 and HDAC7 inhibition and antagonizes the TSA-induced cell apoptosis
.
Mol Cancer
2011
;
10
:
18
.
59.
Nair
S
,
Archer
GE
,
Tedder
TF
.
Isolation and generation of human dendritic cells
.
Curr Protoc Immunol
2012
;
Chapter 7:Unit7.32
.
60.
Ah Kioon
MD
,
Tripodo
C
,
Fernandez
D
,
Kirou
KA
,
Spiera
RF
,
Crow
MK
, et al
.
Plasmacytoid dendritic cells promote systemic sclerosis with a key role for TLR8
.
Sci Transl Med
2018
;
10
:
eaam8458
.
61.
Cao
P
,
Aoki
Y
,
Badri
L
,
Walker
NM
,
Manning
CM
,
Lagstein
A
, et al
.
Autocrine lysophosphatidic acid signaling activates beta-catenin and promotes lung allograft fibrosis
.
J Clin Invest
2017
;
127
:
1517
30
.
62.
Langmead
B
,
Salzberg
SL
.
Fast gapped-read alignment with Bowtie 2
.
Nat Methods
2012
;
9
:
357
9
.
63.
Li
B
,
Dewey
CN
.
RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome
.
BMC Bioinf
2011
;
12
:
323
.
64.
Love
MI
,
Huber
W
,
Anders
S
.
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol
2014
;
15
:
550
.
65.
Gyorffy
B
,
Lanczky
A
,
Szallasi
Z
.
Implementing an online tool for genome-wide validation of survival-associated biomarkers in ovarian-cancer using microarray data from 1287 patients
.
Endocr Relat Cancer
2012
;
19
:
197
208
.

Supplementary data