Tumor Cell–Independent Estrogen Signaling Drives Disease Progression through Mobilization of Myeloid-Derived Suppressor Cells

Estrogens are pleiotropic steroid hormones known to influence many biological processes that ultimately affect homeostasis, such as development and metabolism. Estrogens bind to two high-affinity receptors (ERα and ERβ) that activate similar but not identical response elements and are differentially expressed in multiple tissues. Due to their pathogenic role in accelerated malignant progression, ER⁺ breast cancers have been commonly treated with tamoxifen. Tamoxifen, however, has mixed antagonist/agonist effect on ERs, depending on cell type (1). Correspondingly, alternative interventions are currently evolving as results from clinical testing emerge (2). In contrast to breast cancer, antiestrogen therapies have proven to be effective in only some patients with ovarian cancer (3–7). However, these studies were exclusively focused on patients with ER⁺ cancer, who represent only 31% of patients with ovarian cancer for ERα and 60% of patients for ERβ, and did not provide any insight into the effects of estrogen activity on nontumor cells.

Besides tumor cells, the tumor microenvironment plays a critical role in determining malignant progression as well as response to various therapies. In particular, it is becoming evident that tumors elicit immune responses that ultimately affect survival. In ovarian cancer, for instance, the presence of tumor-infiltrating lymphocytes is a major positive prognostic indicator of tumor survival (8), and multiple T-cell inhibitory pathways have been identified (9–11).

In addition to tumor cells, both ERs are expressed by most immune cell types, including T cells, B cells, and natural killer cells, in which ERα46 is the predominant isoform (12). Correspondingly, estrogens influence helper CD4 T-cell differentiation favoring humoral Th2 over cell-mediated Th1 responses (13). Premenopausal women have higher levels of estrogen than men, which may contribute to differences in the incidence of certain autoimmune diseases. Notably, various cancers exhibit sex biases that are at least partly explained by hormonal differences. Obesity, which is associated with increased adipocyte production of estrogens, is also a risk factor for a number of cancers. Changes in estrogen levels in women caused by menstruation, menopause, and pregnancy are associated with changes in the immune system, which could ultimately affect disease susceptibility. Despite growing evidence implicating estrogen as a fundamental mediator of inflammation, currently little is known about its potential role in antitumor immune responses, and particularly in patients without direct estrogen signaling on tumor cells but with a strongly responsive immuno-environment.

Among suppressors of antitumor immune responses, factors driving tumor-associated inflammation universally induce aberrant myelopoiesis in solid tumors, which fuels malignant progression in part by generating immunosuppressive myeloid cell populations (14). In ovarian cancer, deregulated myelopoiesis results in the mobilization of myeloid-derived suppressor cells (MDSC) from the bone marrow (BM; ref. 15) and, eventually, the accumulation of tumor-promoting inflammatory dendritic cells (DC) with immunosuppressive activity in solid tumors (16, 17), while canonical macrophages build up in tumor ascites (16, 18). Although all these cell types express at least ERα and are influenced by estrogen signaling (19–21), how estrogens affect the orchestration and maintenance of protective antitumor immunity remains elusive. Here, we show that estrogens, independently of the sensitivity of tumor cells to estrogen signaling, are a crucial mechanism underlying pathologic myelopoiesis in ovarian cancer. We report that estrogens drive MDSC mobilization and augment their immunosuppressive activity, which directly facilitates malignant progression. Our data provide mechanistic insight into how augmented estrogenic activity could contribute to tumor initiation (e.g., in BRCA1-mutation carriers; ref. 22), and provide a rationale for blocking estrogen signals to boost the effectiveness of anticancer immunotherapies.

Nuclear expression of ERs specifically in neoplastic cells has been identified in human ovarian carcinomas of all histologic subtypes, with positive signal in approximately 60% of high-grade serous tumors (23). ERα is the predominant estrogen receptor in mouse hematopoietic cells (12). To define the expression of ERα in human ovarian cancer–infiltrating leukocytes, we performed immunohistochemical analysis in 54 serous ovarian carcinomas. Supporting previous reports, we found specific nuclear staining in tumor cells in approximately 35% of tumors (Fig. 1A, left). In addition, we identified weaker signals in individual cells in the stroma with leukocyte morphology (different from tumor cell nuclei) in approximately 20% of ovarian tumors, independent of the ERα status of tumor cells (Fig. 1A, center). We finally identified two specimens that showed specific signals restricted to stromal fibroblasts (Fig. 1A, right). To confirm that hematopoietic cells at tumor beds express ERα, we sorted (CD45⁺) cells from 7 different dissociated human ovarian tumors. As shown in Fig. 1B and Supplementary Fig. S1A, both tumor-infiltrating (CD11b⁺) myeloid cells and (CD11b⁻) nonmyeloid leukocytes express variable levels of ERα. In addition, both myeloid and lymphoid cells sorted from either the BM of a patient with cancer or the peripheral blood of 5 patients with ovarian cancer were also ERα⁺ (Fig. 1B and C; and Supplementary Fig. S1A and S1B), suggesting that in addition to potentially having tumor cell–intrinsic effects, estrogens may also play a wider role in shaping the tumor immuno-environment. To determine the role of estrogen signaling in tumor-promoting inflammation or antitumor immunity, we used a preclinical model of aggressive ovarian cancer in which syngeneic epithelial ovarian tumor cells (ID8-Defb29/Vegfa) develop intraperitoneal tumors and ascites that recapitulate the inflammatory microenvironment of human ovarian tumors (9, 15, 17, 24). Importantly, no ERα was detected in these tumor cells, unlike tumor-associated myeloid cells (Fig. 1D). Of note, ID8-Defb29/Vegfa cells fail to respond to estradiol (E2) treatment or the ER antagonist fulvestrant in vitro, unlike established estrogen-responsive MCF-7 cells (Fig. 1E). Supporting a tumor cell–independent role of estrogen signaling in malignant progression, oöphorectomized (estrogen-depleted) wild-type (WT) mice survived significantly longer than non-oöphorectomized, aged-matched controls after orthotopic tumor challenge in multiple independent experiments (Fig. 1F), whereas estrogen supplementation further accelerated malignant progression and reversed the protective effects of oöphorectomy (Fig. 1F; Supplementary Fig. S1C). Strikingly, the survival benefit imparted by oöphorectomy disappeared in tumor-bearing immunodeficient RAG1-deficient KO mice (Fig. 1G), indicating that an adaptive immune response is required for the protective effects of estrogen depletion.

Interestingly, ad libitum E2 supplementation resulted in augmented T-cell inflammation at tumor (peritoneal) beds (Fig. 2A). However, the proportions of antigen experienced (CD44⁺), recently activated (CD69⁺) tumor-associated CD4 and CD8 T cells were significantly higher in oöphorectomized tumor-bearing hosts, with corresponding decreases in E2-supplemented animals (Fig. 2B). Accordingly, the frequencies of T cells isolated from the peritoneal cavity of oöphorectomized tumor-bearing mice producing IFNγ in response to cognate tumor antigens were significantly higher than those generated by control (non-oöphorectomized) mice in conventional ELISpot analysis (Fig. 2C), indicative of superior T-cell–dependent antitumor immunity in the former. Consistently, tumor-associated T cells from E2-treated mice responded significantly worse than either group (Fig. 2C). Taken together, these results demonstrate that human ovarian cancer microenvironmental hematopoietic cells express ERα, and that, independent of a direct effect on tumor cells, estrogens accelerate ovarian cancer progression through a mechanism that blunts protective antitumor immunity.

The benefits of estrogen depletion were not restricted to ID8-Defb29/Vegfa tumors, because the progression of estrogen-insensitive (Supplementary Fig. S1D), intraperitoneal Lewis lung carcinomas (LLC) was also significantly delayed in oöphorectomized mice, ultimately resulting in decreased survival (Fig. 3A, left), whereas E2 supplementation accelerated flank tumor growth (Fig. 3A, right). Further supporting the general applicability of this mechanism, E2 supplementation also accelerated the growth of estrogen-insensitive (Supplementary Fig. S1E) syngeneic A7C11 mammary tumor cells, derived from autochthonous p53/KRAS-dependent mammary tumors (ref. 15; Fig. 3B). Finally, E2 treatment also increased the number of lung metastases in a model of i.v. injected (estrogen-insensitive) B16 melanoma cells (Fig. 3C; Supplementary Fig. S1F).

To determine the mechanism by which estrogen signaling accelerates malignant progression, we next investigated differences in the mobilization of immunosuppressive cells. We identified strong estrogen-dependent differences only in the accumulation of MDSCs, both in the spleen (Fig. 3D and E) and at tumor beds (Fig. 3F and G), which persisted in tumors of similar size (Supplementary Fig. S1G). Hence, estrogen treatment increased the percentage and total numbers of both Ly6C^hiLy6G⁻ myelomonocytic (M-MDSC) and Ly6C⁺Ly6G⁺ granulocytic MDSCs (G-MDSC) in tumor-bearing mice, whereas estrogen depletion through oöphorectomy significantly decreased their percentage and total numbers both in the spleen and at tumor beds (Fig. 3D–G).

To define whether estrogen-driven MDSC mobilization is sufficient to explain accelerated tumor growth, we depleted MDSCs with anti-Gr1 antibodies in A7C11 breast tumor–bearing mice. As expected, oöphorectomized mice injected with an irrelevant IgG again showed accelerated tumor progression when E2 was supplemented ad libitum throughout disease progression (Fig. 3H). However, differences in tumor growth were completely abrogated when MDSCs were depleted with anti-Gr1 antibodies (Fig. 3H).

Estrogens primarily signal through the nuclear receptors ERα and ERβ, the former being expressed in virtually all murine hematopoietic cells (20). Further supporting that differences in the ovarian cancer immuno-environment are independent of estrogen signaling on tumor cells, we identified ERα expression in MDSCs derived from tumor- bearing mice (Fig. 1D). Importantly, myeloid cells sorted from tumor-bearing mice were also highly effective at suppressing T-cell proliferative responses and therefore are true immunosuppressive MDSCs and not merely immature myeloid cells (Fig. 3I; Supplementary Fig. S2A), supporting their role in estrogen-dependent abrogation of antitumor immunity. Interestingly, G-MDSCs from E2-depleted (oöphorectomized) mice exhibit weaker immunosuppressive potential compared with vehicle- or E2-treated mice.

To confirm that ERα signaling is sufficient to mediate accelerated malignant progression, we then challenged ERα^−/− and WT control mice with orthotopic ID8-Defb29/Vegfa tumors. As shown in Fig. 4A, E2 supplementation failed to accelerate tumor progression in ERα KO hosts but again had significant effects in WT controls, indicative that estrogen's tumor-promoting responses are attributable to ERα signaling. In addition, accelerated tumor growth depends on ERα signaling specifically on hematopoietic cells because in response to E2 treatment, tumors progress significantly faster in lethally irradiated mice reconstituted with WT BM, compared with identically treated mice reconstituted with ERα-deficient BM (Fig. 4B). Together, these results indicate that ERα signaling on hematopoietic cells accelerates malignant progression independently of the stimulation of neoplastic cells, through a mechanism that results in the mobilization of (ERα⁺) immunosuppressive MDSCs.

To rule out that estrogen-dependent myeloid expansion in tumor-bearing mice was the result of subtle differences in tumor burden or inflammation, we reconstituted lethally irradiated mice with a 1:1 mixture of CD45.2⁺ERα^−/− and (congenic) CD45.1⁺ERα⁺ BM and challenged them with orthotopic ovarian tumors. As shown in Fig. 4C and D, a significantly higher percentage (3.6-fold) of total (CD11b⁺Gr1⁺) MDSCs arose from ERα⁺ hematopoietic progenitors, compared with ERα-deficient cells. Because reconstitution of total hematopoietic cells occurred at a similar ratio (Fig. 4C) and MDSC mobilization took place in the same host under an identical milieu, dissimilar ERα-dependent MDSC accumulation can only be attributed to cell-intrinsic ERα⁺ signaling on myeloid precursors (Fig. 4E).

To understand how estrogen signaling promotes MDSC expansion, we next differentiated MDSCs in vitro by treating naïve WT (ERα⁺) BM with GM-CSF and IL6. As reported (25), these inflammatory cytokines induced the generation of immature myeloid cells that express Ly6G and Ly6C similar to MDSCs seen in vivo (Fig. 5A, left).

Normal cell culture medium drives estrogen signaling due to the presence of various estrogens in FBS (26) in addition to the estrogenic properties of phenol red. Blocking the estrogen activity of cell culture medium with methylpiperidino pyrazole (MPP), a selective antagonist of ERα, severely inhibited the expansion of both M-MDSCs and G-MDSCs (Fig. 5A, left, and Fig. 5B), similar to in vivo in tumor-bearing mice (Fig. 4E) and, to an even greater extent, BM-MDSCs expanded with ID8-Defb29/Vegfa–tumor conditioned medium (Supplementary Fig. S2B). In addition, the presence ERα antagonists allowed spontaneous differentiation of more mature CD11c⁺MHC-II⁺ myelo-monocytic cells (Fig. 5A, right). Corresponding to in vivo observations (Fig. 3F), further addition of E2 resulted in G-MDSCs that were more potently immunosuppressive whereas abrogation of ERα signaling prevented the acquisition of stronger immunosuppressive activity by G-MDSCs (Fig. 5C, top). In contrast, E2 did not affect the inhibitory activity of M-MDSCs (Fig. 5C, bottom), suggesting that the role of estrogens in the accumulation of M-MDSCs is primarily to drive their expansion, although the low yields of BM-MDSCs obtained in the presence of estrogen antagonists preclude testing their suppressive activity.

To support the relevance of ERα signaling in boosting pathologic expansion of MDSCs, we finally procured BM from 5 different patients with lung cancer and expanded myeloid cells with GM-CSF and IL6 (25), in the presence of different concentrations of the ERα-selective antagonist MPP. As shown in Fig. 5D, this system results in reproducible expansion of CD11b⁺CD33⁺CD15⁺CD14⁻MHC-II⁻ granulocytes and CD11b⁺CD33⁺CD15^−/loCD14⁺MHC-II⁻ monocytic cells, corresponding to the human counterparts of granulocytic and monocytic MDSCs, along with more mature myeloid cells (Supplementary Fig. S2C). Notably, blockade of ERα signaling resulted in a dramatic dose-dependent reduction in the expansion of both MDSC lineages, at the level of both proportions (Fig. 5D) and, especially, absolute numbers (Fig. 5E). Equally important, analysis of 266 patients with serous ovarian cancer in The Cancer Genome Atlas (TCGA) datasets confirmed that patients with expression levels of the aromatase gene CYP19A1 (the enzyme responsible for a key step in the biosynthesis of estrogens) above the median also exhibit lower expression of CD3e and perforin, indicators of cytotoxic activity and total T-cell infiltration, respectively (Fig. 5F).

Together, these data show that estrogen signaling through ERα influences myelopoiesis in both mice and humans to boost the expansion of highly immunosuppressive MDSCs in response to inflammatory signals and block their differentiation into MHC-II⁺ myeloid cells. These combined functions of ERα signaling in myeloid cells promote malignant progression through MDSC-mediated immune-suppression.

To determine the mechanism by which estrogen signaling promotes MDSC mobilization, we focused on the effect of estrogen signaling on STAT3 signaling, which plays a major role in regulating myeloid lineage cells and MDSC expansion (14). As shown in Fig. 6A, levels of pSTAT3^Y705 were significantly increased in both monocytic and granulocytic MDSCs immunopurified from the peritoneal cavity of oöphorectomized ovarian cancer–bearing mice supplemented with E2, compared with control oöphorectomized mice receiving vehicle. Accordingly, antiestrogen treatment of in vitro BM-MDSC cultures also inhibited STAT3 signaling, resulting in lower phospho-STAT3 (pSTAT3) in both M-MDSCs and G-MDSCs (Fig. 6B), confirming that pSTAT3 signaling is enhanced by estrogen activity.

Because STAT3 activation is triggered by IL6, which was used for in vitro MDSC expansion, we next investigated the role of estrogen signaling on IL6R. Treating BM-MDSCs with E2 or antiestrogens did not elicit changes in surface expression of the IL6Rα chain (Fig. 6C, left), whereas gp130 was paradoxically upregulated by MPP (Fig. 6C, right). We therefore focused on downstream JAK and SRC kinases, both of which mediate STAT3 phosphorylation, subsequent dimerization, and nuclear translocation following cytokine receptor engagement (27, 28). As shown in Fig. 6D, E2 supplementation induced transcriptional upregulation of JAK2 in cytokine-induced BM MDSCs of both lineages, whereas no detectable expression or changes were identified for other JAK members (not shown). Accordingly, E2 also induced a reproducible JAK2 upregulation at the protein level, including higher levels of active (phosphorylated) JAK2 after a short pulse (Fig. 6E and F). Notably, estrogen antagonists also reduced the levels of (active) phospho-Src in both M-MDSCs and G-MDSCs, whereas E2 supplementation also increased Src activity in the former (Fig. 6G).

To define which kinase (JAK2 vs. Src) plays a predominant role in estrogen-dependent MDSC expansion, we expanded BM-MDSCs in the presence of specific Src (dasatinib), JAK1/2 (ruxolitinib), or ERα (MPP) inhibitors. As shown in Fig. 6H and I, JAK1/2 inhibition had a dramatic negative effect in the differentiation of M-MDSCs. G-MDSC expansion was also heavily decreased upon JAK1/2 inhibition, but concurrent use of ERα antagonists significantly potentiated these suppressive effects (Fig. 6H and I). On the other hand, Src inhibition had no effect on preventing G-MDSC differentiation, but resulted in a significant decrease in M-MDSC mobilization, which was further enhanced by additional ERα inhibition (Fig. 6I and J). Therefore, concomitant inhibition of ERα signaling and STAT3-activating kinases has stronger negative effects on the expansion of both monocytic and granulocytic MDSC mobilization than inhibition of either pathway individually. Taken together, these data indicate that ERα signaling on myeloid precursors promotes MDSC expansion by driving STAT3 phosphorylation. In M-MDSCs, this occurs through both enhanced (phosphorylated) Src activity and the necessary function of JAK2, whereas only JAK2 activity is relevant in G-MDSCs.

Finally, to rule out that differences in malignant progression are due to the direct effect of estrogens on effector T cells, we performed mixed BM chimera experiments in which mice received a 1:1 mixture of ERα^−/− and congenic WT BM. Compared with ERα^−/− T cells, E2-responsive WT CD4 and CD8 T cells displayed a less activated phenotype characterized by lower expression of CD44 (Fig. 7A). Correspondingly, the frequencies of WT T cells responding to tumor antigens in IFNγ ELISpot rechallenge assays were lower than those of their counterpart ERα^−/− T cells, sorted from the same microenvironment (Fig. 7B).

To determine the relative importance of these differences in direct ERα signaling in T cells, independent of estrogen-dependent MDSC activity, WT and ERα^−/− T-cell splenocytes were identically enriched for tumor-reactive populations by ex vivo priming against tumor lysate-pulsed bone marrow dendritic cells (BMDC; refs. 29, 30), and then adoptively transferred into ovarian cancer–bearing mice. Confirming previous reports (29, 30), both WT and ERα^−/− T cells significantly extended survival. However, there was no difference between WT and ERα KO T cells regardless of whether mice were treated with E2 (Fig. 7C). Therefore, although E2 has measurable T-cell–intrinsic effects, these are not sufficient to drive differences in malignant progression, and, therefore, E2 effects on immunosuppressive cells, namely MDSCs, are the main driver underlying estrogen-driven tumor acceleration.

Here, we show that ERα signaling on myeloid precursors is a major contributor to pathologic myelopoiesis in cancer, resulting in MDSC expansion and augmented immunosuppressive activity. Accordingly, oöphorectomized mice exhibit delayed malignant progression upon challenge with different estrogen-insensitive tumor models, whereas E2 supplementation has the opposite effects. Supporting the crucial role of spontaneous antitumor immunity in this mechanism, differences in tumor growth disappear in T-cell–deficient mice.

Although the role of estrogen signaling in the progression of breast tumors and a subset of patients with ovarian cancer has been underscored by the clinical use of ER antagonists, our results demonstrate that estrogens have a profound effect on antitumor immunity and tumor-promoting inflammation, independent of their direct activity on tumor cells. Our data therefore provide mechanistic insight into how enhanced estrogenic activity contributes to malignant progression in established tumors. Furthermore, our data support that antiestrogen drugs that, unlike tamoxifen (1), have no agonistic effects on nonbreast cell types may have benefits in a wide range of cancers in premenopausal women, independently of the expression of ERs in tumor cells. Therefore, antiestrogens, especially when used as an adjuvant therapy, could synergize with immunotherapies such as checkpoint inhibitors to extend survival significantly. Thus, although bilateral oöphorectomy is standard in ovarian cancer treatment, our data suggest that ER⁻ breast tumors and other malignancies in at least premenopausal female patients with cancer could be delayed by specifically blocking ERα in a systemic manner, especially if complementary immunotherapies are implemented as adjuvant therapy.

Our results also have implications for understanding gender-dependent differences in tumor initiation and malignant progression in different malignancies. For example, we showed that responses to αPD-L1 immunotherapy were sex-dependent in hormone-independent melanoma (31). This could be particularly relevant in BRCA1-mutation carriers, where augmented estrogenic signals have been recently demonstrated (22, 32). Furthermore, ERα expression is regulated by BRCA1-dependent ubiquitination (33), so that cancer-predisposing heterozygous BRCA1 mutations could result in increased ER expression, and therefore increase estrogen activity. Whether mobilization of MDSCs in the context of additional inflammatory signals contributes to tumor initiation in BRCA1 mutation carriers demands further experimental proof, but our study suggests this as a likely pathogenic mechanism.

This study contributes to the understanding of the complexity of factors deregulating myelopoiesis (and therefore antigen presentation) in virtually all solid tumor–bearing hosts. Our data indicate that ERα signaling has a triple effect on myeloid BM progenitors by altering pSTAT3 signaling, which drives both expansion and increased survival in these cells (14): On the one hand, estrogens upregulate JAK2, which mediates STAT3 phosphorylation, as well as total STAT3 itself. Therefore, estrogenic activity prepares the BM for acute expansion of myeloid precursors, but estrogen-dependent mobilization of MDSCs occurs only in the presence of direct inflammatory signals that activate JAK kinases, which may not necessarily be present during the high estrogen phase of the menstrual cycle. These mechanisms appear to be important during pregnancy though, where E2 also drives the expansion and activation of MDSCs (21), but our study demonstrates their relevance in the pathogenesis of women's malignancies.

In summary, our study unveils the role of estrogen signaling in pathologic myelopoiesis and supports that more specific antiestrogen drugs could complement emerging immunotherapies to significantly extend the survival of patients with cancer, independently of the expression of ERs in tumor cells.

Female 5- to 8-week-old WT C57BL/6 and congenic Ly5.1 mice were purchased from the Charles River Frederick facility. ESR1 knockout (ERα KO) mice were purchased from The Jackson Laboratory. Oöphorectomies were performed by Charles River staff at 5 weeks of age. Mice were treated with vehicle (0.1% ethanol) or 10 μmol/L E2 (USP grade; Sigma) in drinking water refreshed every 3 to 4 days. All mice were maintained in specific pathogen-free barrier facilities. All experiments were conducted according to the approval of the Wistar Institute Institutional Animal Care and Use Committee.

ID8 cells were provided by K. Roby (Department of Anatomy and Cell Biology, University of Kansas, in 2011; ref. 34) and retrovirally transduced to express Defb29 and Vegfa (24). Clones were passaged fewer than 4 times. MCF-7, B16.F10, and LLC1 cells were obtained from the American Type Culture Collection in 2009, 2013, and 2009, respectively, and passaged fewer than 4 times. Cell lines used in this article were not authenticated by us.

Peritoneal tumors were initiated in mice by injecting 3 × 10⁶ ID8-Defb29/Vegfa cells intraperitoneally. Intraperitoneal cells were harvested from tumor-bearing mice by flushing the peritoneal cavity with PBS. Cells were maintained in vitro at 37°C, 5% CO₂ by culturing in RPMI+10% FBS or steroid-free medium (SFR10), which was comprised of phenol red–free RPMI+10% charcoal-stripped FBS. Cells were treated in vitro with vehicle (0.1% DMSO) or varying concentrations of E2, fulvestrant, or MPP purchased from Cayman Chemical. Cell proliferation was determined by MTS assay (Promega), and increased/decreased proliferation relative to vehicle was calculated.

The A7C11 mammary tumor cell line was generated by passaging sorted tumor cells from a mechanically disassociated p53/KRAS mammary tumor (35). Flank tumors were initiated by injecting 2 × 10⁴ cells into the axillary flank. Tumor volume was calculated as: 0.5 × (L × W²), where L is the length and W is the width.

For generating mixed BM chimeras, mononuclear BM cells were collected from adult age-matched CD45.1 (congenic) WT or CD45.2 Esr1^−/− donor mice, and 1–2 × 10⁶ cells were mixed in a 1:1 ratio and retro-orbitally injected into lethally irradiated (∼950 rad) adult recipients. Mixed chimeras were analyzed after 7 to 8 weeks as indicated.

Human ovarian carcinoma tissues were procured under a protocol approved by the Committee for the Protection of Human Subjects at Dartmouth-Hitchcock Medical Center (#17702) and under a protocol approved by the Institutional Review Board at Christiana Care Health System (#32214) and the Institutional Review Board of The Wistar Institute (#21212263). BM was obtained from patients with stage I–II lung cancer scheduled for surgical resection at the Hospital of the University of Pennsylvania and The Philadelphia Veterans Affairs Medical Center with approval from respective Institutional Review Boards. All patients selected for entry into the study met the following criteria: (i) histologically confirmed pulmonary squamous cell carcinoma or adenocarcinoma, (ii) no prior chemotherapy or radiotherapy within 2 years, and (iii) no other active malignancy. BM cell suspension was obtained from rib fragments that were removed from patients as part of their lung cancer surgery. Informed consent was obtained from all subjects.

Flow cytometry was performed by staining cells with Zombie Yellow viability dye, blocking with anti-CD16/32 (2.4G2), and staining for 30 minutes at 4°C with the following anti-mouse antibodies at the manufacturer's recommended dilution: CD45 (30-F11), CD45.1 (A20), CD45.2 (104), CD11c (N418), I-A/I-E (M5/114.15.2), CD3 (145-2C11), Ly6G (1A8), Ly6C (HK1.4), gp130 (4H1B35), IL6R (D7715A7), Gr1 (RB6-4C5) CD4 (RM4-5), CD8b (YTS156.7.7), CD44 (IM7), CD69 (H1.2F3), CD11b (M1/70); or anti-human antibodies: CD45 (HI30), CD11c (Bu15), HLA-DR-APC/Cy7 (L243), CD15 (HI98), CD14 (HCD14), CD11b (ICRF44), CD33 (WM53), CD19 (HIB19). Samples were subsequently run using an LSRII and analyzed using FlowJo.

Dendritic cells (BMDCs) were differentiated by culturing WT mouse BM for 7 days with 20 ng/mL GM-CSF (Peprotech 315-03), added on days 0 and 3, and 10 ng/mL GM-CSF added on day 6. BMDCs were subsequently primed with tumor antigen by pulsing for 24 hours with irradiated (100 Gy + 30 minutes UV) ID8-Defb29/Vegfa cells at a ratio of 10:1. ELISpot assay was performed by stimulating 1 × 10⁵ cells obtained from peritoneal wash with 1 × 10⁴ antigen-primed BMDCs in a 96-well filter plate (Millipore; MSIPS4510) coated with IFNγ capture antibody according to the manufacturer's guidelines (eBioscience; 88-3784-88). Following incubation at 37° C, 5% CO₂ for 48 hours, positive spots were developed using Avidin-AP and BCIP-NBT substrate (R&D Systems SEL002).

Naïve T cells were harvested from spleens of WT or ERα KO mice via red blood cell lysis followed by magnetic bead negative selection to remove non–T-cell B220⁺, CD16/32^+,CD11b⁺, and MHC-II⁺ cells and primed for 5 days with BMDCs pulsed with tumor antigen as reported (30). A total of 1 × 10⁶ T cells were injected i.p. 7 and 14 days after tumor injection.

Mouse MDSCs were expanded from mouse BM harvested by flushing tibias and femurs with media. Following red blood cell lysis, 2.5 × 10⁶ cells were cultured in 10 mL of RPMI+10% FBS augmented with recombinant mouse 40 ng/mL GM-CSF+40 ng/mL IL6 (Peprotech) and incubated at 37°C, 5% CO₂ for 3 or 6 days. Vehicle, estradiol, or MPP treatments were added as described above. For 6-day cultures, cytokines and estrogen treatments were refreshed on day 3. Following red blood cell lysis, 2.5 × 10⁶ cells were cultured in 10 mL of RPMI+10% FBS augmented with recombinant mouse 40 ng/mL GM-CSF + 40 ng/mL IL6 (Peprotech) and incubated at 37°C, 5% CO₂ for 3 or 6 days. Vehicle, E2, or MPP treatments were added as described above. For 6-day cultures, cytokines and E2 were refreshed on day 3. Following incubation, floating and adherent cells were collected, and M-MDSCs and G-MDSCs were isolated via a Miltenyi MDSC purification kit according to the manufacturer's protocol. Human MDSCs were expanded from human lung cancer patient BM acquired as single-cell suspensions (see above). Briefly, 2 × 10⁶ cells were cultured in 3 mL of IMDM+15% FBS supplemented with recombinant human 40 ng/mL GM-CSF + 40 ng/mL IL6 (Peprotech) and treated with vehicle, 2 μmol/L, or 10 μmol/L MPP (see above) for 4 days. Cells were subsequently harvested and analyzed by flow cytometry.

Naïve WT T cells were purified from spleens as described and labeled with the proliferation tracker Cell Trace Violet according to the manufacturer's protocol. T-cell proliferation was stimulated by adding anti-CD3/CD28 mouse T-activator beads (Thermo) at a 1:1 T-cell to bead ratio according to the manufacturer's protocol. T cells (2 × 10⁵) were subsequently cocultured with MDSC at 1:4, 1:8, or 1:16 MDSC to T-cell ratios and incubated for 3 days prior to flow cytometric analysis.

Cells were lysed in RIPA buffer supplemented with protease inhibitors, phosphatase inhibitors, and Na3VO4 (Thermo) according to the manufacturer's protocol. Protein quantification was determined via BCA assay, and protein was run on TGX 4-15% gradient gels. Following transfer, PVDF membranes were blocked with 5% BSA in TBS + 0.05% Tween-20. The following primary antibodies were added to membranes, as indicated, and incubated overnight: JAK2 (rabbit clone # D2E12), STAT3 (clone # 124H6), and pSTAT3 (Tyr705, rabbit clone # D3A7), from Cell Signaling Technology; and ERα (Thermo, clone # TE111.5D11) and beta-actin (Sigma, clone # AC-15). Following secondary staining with horseradish peroxidase–conjugated anti-mouse or rabbit IgG, membranes were developed using ECL prime (GE Healthcare).

ERα staining was initially performed in frozen sections from 54 ovarian cancer specimens with clone #TE111.5D11 as the primary antibody (Thermo), followed by a biotinylated goat anti-mouse and completion of immunohistochemical procedure according to the manufacturer's instructions (Vector Labs). Positive staining was confirmed in 19 paraffin-embedded ovarian cancer sections using FDA-approved ER tests (DAKO) in the Laboratory of Diagnostic Immunohistochemistry at the Hospital of the University of Pennsylvania, following the FDA-approved manufacturer guidelines.

Cells were lysed in TRIzol buffer and RNA was subsequently purified using an RNEasy kit (Qiagen). Reverse transcription was carried out using a High-Capacity Reverse Transcription Kit (Applied Biosystems). SYBR Green PCR Master Mix (Applied Biosystems) was used with an ABI 7500 Fast Sequence Detection Software (Applied Biosystems). The following primer sequences were used (5′→3′): Stat3 <F: GACTGATGAAGAGCTGGCTGACT, R: GGGTCTGAAGTTGAGATTCTGCT>; Jak2 <F: GTGTCGCCGGCCAATGTTC, R: CACAGGCGTAATACCACAAGC>; and Tbp (mRNA normalization) <F: CACCCCCTTGTACCCTTCAC, R: CAGTTGTCCGTGGCTCTCTT>.

Expression of human ESR1 was quantified with primers: ESR1 <F: CCACTCAACAGCGTGTCTC and R: GGCAGATTCCATAGCCATAC>, and normalized with primers: GAPDH <F: CCTGCACCACCAACTGCTTA R: AGTGATGGCATGGACTGTGGT>.

Expression of mouse ERα was determined with primers: ERα <F:GTGCAGCACCTTGAAGTCTCT and R: TGTTGTAGAGATGCTCCATGCC>.

Aligned Sequence files related to solid ovarian cancer samples including whole-exome sequencing and outcome data were downloaded from the TCGA data portal (2015). Scores (number of tags in each transcript) were obtained from each sample, normalized with respect to total tags in the sample as well as total tags in the chromosome, and expressed as FPKM (fragments/Kb of transcript/million mapped reads).

No potential conflicts of interest were disclosed.

Conception and design: N. Svoronos, M.R. Rutkowski, P. Zhang, R. Zhang, J.R. Conejo-Garcia

Development of methodology: N. Svoronos, A. Perales-Puchalt, M.R. Rutkowski, P. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Svoronos, A. Perales-Puchalt, M.J. Allegrezza, A.J. Tesone, M.G. Cadungog, S. Singhal, E.B. Eruslanov, J. Tchou

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Svoronos, A. Perales-Puchalt, M.J. Allegrezza, K.K. Payne, T.J. Curiel, S. Singhal, P. Zhang, J.R. Conejo-Garcia

Writing, review, and/or revision of the manuscript: N. Svoronos, A. Perales-Puchalt, K.K. Payne, A.J. Tesone, T.J. Curiel, S. Singhal, E.B. Eruslanov, J. Tchou, J.R. Conejo-Garcia

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.M. Nguyen

Study supervision: J.R. Conejo-Garcia

We thank the Gabrilovich lab at The Wistar Institute for outstanding technical insight.

This study was supported by grants to J.R. Conejo-Garcia [R01CA157664, R01CA124515, R01CA178687, The Jayne Koskinas & Ted Giovanis Breast Cancer Research Consortium at Wistar and Ovarian Cancer Research Fund (OCRF) Program Project Development awards]; T.J. Curiel (R01CA164122, CDMRP CA140355, P3054174, The Owens Foundation, The Skinner Endowment, and the OCRF Program Project Development award); E.B. Eruslanov (RO1CA187392); M.J. Allegrezza and N. Svoronos (T32CA009171); K.K. Payne (5T32CA009140-39); and A. Perales-Puchalt [Ann Schreiber Mentored Investigator Award (OCRF)]. Support for Shared Resources was provided by Cancer Center Support Grant (CCSG) CA010815 (D. Altieri).

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.