Pharmacological Activation of Estrogen Receptor Beta Overcomes Tumor Resistance to Immune Checkpoint Blockade Therapy.

The emerging immune checkpoint blockade (ICB) therapy has ushered the cancer therapeutics field into an era of immunotherapy. Although ICB treatment provides remarkable clinical responses in a subset of patients with cancer, this regimen fails to extend survival in a large proportion of patients. Here, we found that a combined treatment of estrogen receptor beta (ERβ) agonist and PD-1 antibody treatment improved therapeutic efficacy in mouse tumor models, compared with monotherapies, by reducing infiltration of myeloid-derived suppressor cells (MDSCs) and increasing CD8+ T cells in tumors. Mechanistically, LY500307 treatment reduced tumor-derived CSF1 and decreased infiltration of CSF1R+ MDSCs in the tumor bed. CSF1 released by tumor cells induced CSF1R+ MDSC chemotaxis in vitro and blockade of CSF1R demonstrated similar therapeutic effects as ERβ activation in vivo. Collectively, our study proved combined treatment of ERβ agonist and PD-1 antibody reduced MDSC infiltration in the tumor and enhanced tumor response to ICB therapy.

Introduction

The human immune system is responsible for the clearance of pathogens and transformed cells. This usually requires functionally redundant counterbalance mechanisms to ensure safety and avoid overreaction (Wykes and Lewin, 2018). In the tumor microenvironment, cancer cells often hijack this counterbalance system to avoid self-destruction and mediate immune evasion. Immune checkpoint molecules are inhibitory receptors expressed on immune cells that elicit immunosuppressive signaling pathways, which constitute an important part of this system. These molecules play critical roles in sustaining self-tolerance and for modulating the length and magnitude of effector immune responses (Fritz and Lenardo, 2019; Pauken et al., 2019; Sanmamed and Chen, 2018). Recently, the checkpoints guarded by the programmed cell death-1 (PD-1) and cytotoxic T-lymphocyte-associated antigen-4 (CTLA-4) receptors have been intensively explored owing to the availability of antibodies that can inhibit their function (Gordon et al., 2017; Minn and Wherry, 2016; Patel and Minn, 2018). With the remarkable therapeutic effects of anti-CTLA-4, anti-PD-1, and anti-PD-L1 monoclonal antibodies in preclinical models and clinical trials, the US Food and Drug Administration has approved their clinical use for the treatment of a variety of cancers.

Although immune checkpoint blockade (ICB) therapy has greatly improved objective response rates, time to progression, and overall survival in some patients with cancer, the majority of patients still fail to respond to ICB therapy. The reported molecular mechanisms include a variety of factors within the tumor microenvironment, for instance, the infiltration of immunosuppressive immune cells including Treg cells, myeloid-derived suppressor cells (MDSCs), and indole 2,3-dioxygenase (IDO) activity. Furthermore, tumor-cell-autonomous factors including mutational load, oncogenic signaling pathways, expression of PD-L1, and down-regulation of major histocompatibility complex (MHC) class I likely contribute as well (Conway et al., 2018; Keenan et al., 2019; Pitt et al., 2016). Apart from these tumor-intrinsic influences, other host-related and environmental factors affecting immune system function could also be involved in the development of checkpoint blockade therapy resistance, like the heterogeneity of gut microbiota (Gopalakrishnan et al., 2018; Routy et al., 2018). Thus, it is important to find ways to improve the response rate of patients with cancer to ICB therapy.

Estrogenic actions are mediated mainly through two distinct estrogen receptor (ER) subtypes: ERα and ERβ. In contrast to the tumor-promoting role of ERα in hormone-responsive cancer, ERβ is reported to be tumor suppressive and has a major role in the immune system. In this study, we aim to evaluate whether pharmacological activation of ERβ could exert therapeutic effects for ICB therapy-resistant tumors and the possible mechanisms of effects.

Results

Combination of ERβ Activation and ICB Therapy Improves Therapeutic Efficacy in TNBC and Colorectal Cancers

Previous evidence has proved expression of ERβ in triple-negative breast cancer (TNBC) (Reese et al., 2018; Zhao et al., 2018) and colorectal cancer tissues (Ibrahim et al., 2019; Williams et al., 2016) and its potential role as therapeutic targets for these tumors. In addition, TNBC and colorectal cancer have demonstrated resistance to ICB therapy (Kim et al., 2014). In an effort to explore whether ERβ activation could overcome ICB resistance in TNBC and colorectal cancers, we first evaluated the specificity of LY500307 for ERβ activation. Based upon the results demonstrated in a previous report (Reese et al., 2018), we examined whether LY500307 could induce increased expression of ERβ target genes. It was demonstrated that LY500307 could up-regulate the expression of CXCL14, KRT17, IGFBP4, KRT13, and Ankrd33, which are known ERβ target genes (Figure S1A). We further analyzed the bona fide targets of ERβ by analyzing the publicly available Chip sequencing dataset GSE108979. Principle component analysis (PCA) demonstrated that each replicate consistently showed altered DNA binding profiles (Figure S1B). The differential ERβ-binding regions on DNA were demonstrated (Figure S1C). After comparing genes that are both up-regulated after ERβ activation identified by RNA sequencing and are found to be regulated by ERβ identified by Chip sequencing, we found 111 overlapping genes, which are bona fide transcriptional targets of ERβ (Figure S1D). Among them, the peak view of IGFBP4 as a representative ERβ-target gene was illustrated (Figure S1E). These data suggested that LY500307 specifically activates ERβ.

Next, we established a BALB/c mouse model bearing subcutaneous 4T1 and CT26 tumors for the evaluation of therapeutic effects of combined PD-1 antibody and a selective ERβ agonist, LY500307, therapies as well as PD-1 antibody or LY500307 monotherapies (Figure 1A). We observed that, although either PD-1 antibody or LY500307 monotherapies had minimal impact on TNBC tumor growth, the combination of PD-1 antibody and LY500307 showed synergistic efficacy in targeting TNBC growth (Figure 1B). No obvious effects on the body weight of those mice were observed in each treatment group (Figure 1C). At indicated time point, the volume and weight of the tumors in the combined treatment group were significantly reduced compared with monotherapy groups and control (Figures 1D and 1E). Similarly, in a corroborating study with CT26 model, combined therapy of PD-1 antibody and LY500307 also showed synergistic efficacy in targeting colorectal cancer growth (Figure 1F), as revealed by tumor volume and tumor weight at indicated time point (Figures 1H and 1I). Mouse body weight was unaffected in each group in CT26 models as well (Figure 1G). In an orthotopic breast cancer model, we also observed that combined treatment with PD-1 antibody and LY500307 showed significantly better therapeutic efficacy compared with each monotherapy or control (Figure 1J), in terms of tumor volume (Figure 1L) and tumor weight (Figure 1M). No significant difference was observed in the body weight of mice in each group as well (Figure 1K). Moreover, we observed a significantly increased level of apoptosis and diminished proliferation rate of tumors in the combined treatment group compared with each monotherapy group and control, as revealed by immunohistochemistry analysis of cleaved caspase 3 and Ki-67 staining, respectively (Figures S2A and S2B).

Figure 1

Selective ERβ agonist LY500307 Overcomes ICB Therapy Resistance in Tumors

(A) Schematic model for the evaluation of ERβ activation in ICB-resistant tumors and the integrative analytical strategy for underlying molecular mechanisms.

(B) BALB/c mice were injected subcutaneously with 4T1 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 4–5 for each group).

(C) The body weight of 4T1 murine model in each group at indicated time points (n = 5 for each group).

(D) The tumor volume of 4T1 murine model in each group at 15th day after tumor injection (n = 4–5 for each group).

(E) The tumor weight of 4T1 murine model in each group at 15th day after tumor injection (n = 5 for each group).

(F) BALB/c mice were injected subcutaneously with CT26 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 4 for each group).

(G) The body weight of CT26 murine model in each group at indicated time points (n = 5 for each group).

(H) The tumor volume of CT26 murine model in each group at 15th day after tumor injection (n = 4 for each group).

(I) The tumor weight of CT26 murine model in each group at 15th day after tumor injection (n = 4 for each group).

(J) BALB/c mice were orthotopically injected with 4T1 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 5 for each group).

(K) The body weight of orthotopic 4T1 murine model in each group at indicated time points (n = 5 for each group).

(L) The tumor volume of orthotopic 4T1 murine model in each group at 16th day after tumor injection (n = 5 for each group).

(M) The tumor weight of orthotopic 4T1 murine model in each group at 16th day after tumor injection (n = 5 for each group).

Data are shown as mean ± SEM. One-way ANOVA, ∗, p < 0.05; ∗∗, p < 0.01.

In addition, we further validated our findings by genetic approaches. We first constructed ERβ-overexpressing plasmid and obtained an ERβ-overexpressing 4T1 cell line (Figures S3A and S3B). We found that, in the orthotopic 4T1 model, the group of 4T1 cells overexpressing ERβ treated together with PD-1 antibody group, similar to that of combined treatment group of LY500307 and PD-1 antibody, grew much slower compared with other groups (Figure S3E). No obvious effects on the body weight of those mice were observed in each treatment group (Figure S3F). At indicated time point, the volume and weight of the tumors in the group of 4T1 cells overexpressing ERβ treated together with PD-1 antibody group, similar to those of combined treatment group of LY500307 and PD-1 antibody, were significantly reduced compared with other groups (Figures S3G and S3H). By contrast, we also constructed ERβ knockdown 4T1 cell line by transfecting shERβ plasmids (Figures S3C and S3D). It was demonstrated that, after ERβ knockdown, the therapeutic efficacy of combined treatment with LY500307 and PD-1 was abolished (Figures S3I, S3K, and S3L). No obvious effects on the body weight of those mice were observed in each treatment group (Figure S3J). Altogether, these data suggest that combination of ERβ activation and ICB therapy enhances therapeutic efficacy in TNBC and colorectal cancers.

Combined Therapy of ERβ Activation and ICB Therapy Reduces MDSC Infiltration and Increases Cytotoxic T Lymphocytes In Vivo

To investigate the underlying mechanisms for the enhanced therapeutic efficacy of combined therapy, we performed RNA sequencing (RNA-seq) analysis for the bulk tumors of the four groups of 4T1 mouse models. Similar transcriptional changes were observed within each replicate of the four models (Figure 2A). Unsupervised clustering by PCA analysis showed that each replicate consistently showed altered transcriptional profiles (Figure 2B). In addition, the statistically significantly altered genes between the combination treatment group with tumors treated with either drug as a monotherapy were also demonstrated (Figures S4A and S4B). Interestingly, gene ontology (GO) enrichment identified a list of pathways associated with immune alterations. Among those top enriched pathways, we found that a group of T cell-related pathways were up-regulated in the combined therapy group, including T cell proliferation, T cell receptor signaling pathway, interferon gamma response, and TNF-α signaling via NF-κB, whereas a group of enriched pathways involved in myeloid cell differentiation and activation were found to be dramatically down-regulated in the combined therapy group (Figure 2C). Therefore, we further investigated which subtype of myeloid cells in the tumor beds are significantly changed by combined treatment of ERβ activation and PD-1 antibody. After screening for a number of subtypes of myeloid cells (including neutrophils, macrophages, eosinophils, basophils, and MDSCs) using flow cytometry, we found that only the number of MDSCs (CD45+CD11b+Gr1+ cells) was dramatically reduced in the combined treatment group in three in vivo tumor models (Figures 2D–2F and S5A). Further flow cytometry analysis revealed that, in the combined treatment group, the number of CD8+ T cells was dramatically increased compared with monotherapy and control groups in three in vivo tumor models (Figures S5B and S6A–S6C). Thus, combined therapy with ERβ activation and PD-1 antibody suppresses MDSC infiltration and increases cytotoxic T lymphocytes in the tumor beds.

Figure 2

Integrative Analysis Identifies MDSCs as the Target Cell Types for ERβ Activation in ICB-Resistant Tumors

(A) Heatmap of the differentially expressed genes in each treatment group of 4T1 model.

(B) Principal component analysis (PCA) of RNA-seq data from each treatment group of 4T1 model.

(C) GO enrichment analysis comparing indicated treatment groups using the RNA-seq data.

(D) Flow cytometry analysis of Gr1+CD11b+ cells in the subcutaneous tumors in each treatment group of 4T1 model (n = 3 for each group).

(E) Flow cytometry analysis of Gr1+CD11b+ cells in the subcutaneous tumors in each treatment group of CT26 model (n = 3–5 for each group).

(F) Flow cytometry analysis of Gr1+CD11b+ cells in the orthotopic tumors in each treatment group of 4T1 model (n = 3–4 for each group).

Data are shown as mean ± SEM; one-way ANOVA, ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗∗, p < 0.0001.

CSF1R+ MDSCs Infiltrate in Tumor Beds and MDSC Depletion Enhances ICB Therapy

We next explored the mechanisms for MDSC infiltration in these tumors. As previous reports demonstrated that the CSF1/CSF1R axis was crucial for the chemotaxis of MDSCs into the tumor microenvironment (Kumar et al., 2017; Neubert et al., 2018; Soncin et al., 2018; Xu et al., 2013; Zhu et al., 2014) and CSF1 signatures could be predictive in breast cancer (Beck et al., 2009; DeNardo et al., 2011), we further examined whether CSF1R+ MDSCs played a role in ERβ activation-mediated suppression of ICB therapy-resistant tumors. Immunofluorescent analysis revealed a number of CSF1R+Gr1+ cells in the control group of both 4T1 and CT26 subcutaneous tumor models (Figures 3A and 3B). Although no significant changes were observed in the PD-1 antibody monotherapy group (Figures 3A and 3B), the number of infiltrated CSF1R+Gr1+ cells in the combined anti-PD-1/LY500307 treatment group was significantly reduced in both models (Figures 3A and 3B). These findings further indicated that ERβ activation reduced CSF1R+ MDSCs in the ICB-resistant tumors to regain sensitivity to ICB therapy.

Figure 3

ERβ Activation Mimics Depletion of MDSCs in ICB-Resistant Tumors

(A) Immunofluorescent analysis of abundance of CSF1R+Gr1+ in different treatment groups in 4T1 mouse model (n = 4 for each group).

(B) Immunofluorescent analysis of abundance of CSF1R+Gr1+ in different treatment groups in CT26 mouse model (n = 4 for each group).

(C) BALB/c mice were injected subcutaneously with 4T1 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 4 for each group).

(D) The body weight of 4T1 murine model in each group at indicated time points (n = 5 for each group).

(E) The tumor volume of 4T1 murine model in each group at 12th day after tumor injection (n = 4 for each group).

(F) The tumor weight of 4T1 murine model in each group at 12th day after tumor injection (n = 4 for each group).

(G) BALB/c mice were injected subcutaneously with CT26 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 5 for each group).

(H) The body weight of CT26 murine model in each group at indicated time points (n = 4 for each group).

(I) The tumor volume of CT26 murine model in each group at 15th day after tumor injection (n = 5 for each group).

(J) The tumor weight of CT26 murine model in each group at 15th day after tumor injection (n = 5 for each group).

Data are shown as mean ± SEM; one-way ANOVA, ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001; ∗∗∗∗, p < 0.0001.

We next characterized the functional role of MDSCs in the development of resistance to ICB therapy. MDSC depletion using Gr1 antibody in vivo significantly decreased infiltration of MDSCs in PD-1 antibody-treated 4T1 and CT26 subcutaneous tumor models, similar to that of the combined anti-PD-1/LY500307 treatment group (Figures 3A and 3B). Functionally, MDSC depletion enhanced the sensitivity of 4T1 cells to ICB therapy, as revealed in the growth curve in each treatment group (Figure 3C), tumor volume (Figure 3E), and tumor weight (Figure 3F), without affecting mouse weight in vivo (Figure 3D). This effect is comparable with combined treatment with LY500307 and ICB therapy but significantly better than any of the monotherapy regimens (Figure 3C). Similar results were also observed in the CT26 tumor model in vivo (Figures 3G–3J). These results demonstrated that CSF1R+ MDSCs infiltrate in tumor beds and MDSC depletion enhances ICB therapy.

Tumor Cells Secrete CSF1 to Attract CSF1R+ MDSC Infiltration

We assessed how CSF1R+ MDSCs infiltrated into tumor beds. We performed a Raybio cytokine array analysis to characterize the alteration of cytokines and chemokines in 4T1 cells treated with LY500307 (Figure S7A). It was demonstrated that, among all the down-regulated chemokines and cytokines, the expressions of M-CSF (CSF-1), CXCL9, CCL19, CXCL4, CCL1, TNFα, and VEGFA were reduced for over 30% percent (Figure S7B). By contrast, the expressions of TNFSF8, IL5, CXCL1, LIX, and CD62L were up-regulated after treatment with LY500307 in 4T1 cells (Figure S7C). This result led us to further characterize the functional role of tumor-derived CSF1 as a chemoattractant for CSF1R+ in tumor beds (Figure 4A). Tumors derived from both 4T1 and CT26 cells overexpress CSF1 compared with normal mouse breast and colorectal tissues, respectively (Figures 4B and 4C). We next examined whether ERβ activation could have any effects on the expression levels of CSF1 in the two cell lines. Treatment with LY500307 caused remarkable reduction of CSF1 in both models, as measured by qPCR (Figures 4D and 4E) and ELISA analysis (Figures 4F and 4G). These observations indicate that ERβ activation could lead to decreased release of CSF1 in tumor cells.

Figure 4

ERβ Activation Reduces Tumor Cell-Derived CSF1 and Blocks MDSC Chemotaxis via CSF1/CSF1R Axis

(A) Schematic model describing the procedure to analyze the mRNA levels or supernatant protein concentration of CSF1 in indicated samples.

(B) The CSF1 mRNA levels between normal mouse breast tissues and 4T1 tumor tissues.

(C) The CSF1 mRNA levels between normal colon tissues and CT26 tumor tissues.

(D) The CSF1 mRNA levels between control 4T1 cells and LY500307-treated 4T1 cells.

(E) The CSF1 mRNA levels between control CT26 cells and LY500307-treated CT26 cells.

(F) The CSF1 concentration in the supernatant of 4T1 cells and LY500307-treated 4T1 cells.

(G) The CSF1 concentration in the supernatant of CT26 cells and LY500307-treated CT26 cells.

(H) Schematic model illustrating the MDSC chemotaxis assay.

(I) The number of migrated MDSCs in each treatment group of 4T1 ex vivo model using Transwell chamber assay.

(J) The number of migrated MDSCs in each treatment group of CT26 ex vivo model using Transwell chamber assay.

(K) Schematic model illustrating how we examined the expression of functional molecules in MDSCs.

(L) The expression of up-regulated genes of MDSCs in combined treatment group (n = 3 for each group).

(M) The expression of down-regulated genes of MDSCs in combined treatment group (n = 3 for each group).

Data are shown as mean ± SEM; one-way ANOVA, ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001, ∗∗∗∗, p < 0.0001.

Subsequently, we explored whether the supernatants from 4T1 cells and CT26 cells could exert chemotactic effects for CSF1R+ MDSCs and whether blockade of CSF1/CSF1R axis could abolish these chemotactic effects in vitro (Figure 4H). The supernatant from control 4T1 cells and CT26 cells could attract CSF1R+ MDSCs to migrate to the lower layer of the Transwell chamber (Figures 4I and 4J). The number of CSF1R+ MDSCs attracted by the supernatants added with LY500307 was comparable with those in the control group in both models. However, the number of CSF1R+ MDSCs attracted by the supernatants from either LY500307-treated 4T1 or CT26 cells was significantly decreased, mimicked by the blockade of CSF1/CSF1R axis using CSF1R neutralization antibody ex vivo (Figures 4I and 4J). We further characterized the functional phenotype of MDSCs isolated from murine tumor tissues in each treatment group. It was revealed that the expression of genes encoding for TLR4, CD80, and CD86 molecules related to a pro-inflammatory phenotype was significantly increased in MDSCs of the combined treatment group with PD-1 antibody and LY500307, whereas IDO and NOS gene expressions, which are known to inhibit anti-tumor T cell responses, were reduced in MDSCs of the combined treatment group (Figures 4K–4M). Therefore, we concluded that LY500307 treatment could block the CSF1/CSF1R axis that mediated CSF1R+ MDSC infiltration in the ICB therapy-resistant tumors.

Blockade of CSF1R Mimics the Therapeutic Effects of ERβ Activation for PD-1-Resistant Tumors

We explored whether ERβ activation in ICB-resistant tumors could be mimicked by CSF1R blockade in vivo. We found that combined therapy of PD-1 antibody and CSF1R antibody in vivo remarkably attenuated the growth of 4T1 cells, with similar efficacy as combined therapy of PD-1 antibody and LY500307 (Figure 5A), without affecting mouse body weight (Figure 5B). These therapeutic effects were also revealed in the tumor volume (Figure 5C) and tumor weight at indicated time point (Figure 5D). Likewise, combined therapy of PD-1 antibody and CSF1R antibody also potently impaired CT26 tumor growth in vivo, comparable with that of combined therapy group of PD-1 antibody and LY500307, as shown by growth curve (Figure 5E), tumor volume (Figure 5G), and tumor weight at indicated time (Figure 5H). No impact of these treatments on mouse body weight was observed in CT26 models as well (Figure 5F). These data suggested that CSF1R blockade combined with PD-1 antibody therapy showed similar therapeutic effects with the combined therapy of ERβ activation and PD-1 antibody.

Figure 5

ERβ Activation Overcomes ICB Resistance via CSF1/CSF1R Axis In Vivo

(A) BALB/c mice were injected subcutaneously with 4T1 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 5 for each group).

(B) The body weight of 4T1 murine model in each group at indicated time points (n = 5 for each group).

(C) The tumor volume of 4T1 murine model in each group at 24th day after tumor injection (n = 3 for each group).

(D) The tumor weight of 4T1 murine model in each group at 24th day after tumor injection (n = 4 for each group).

(E) BALB/c mice were injected subcutaneously with CT26 cells and each treatment was given at the indicated time. Growth kinetics was recorded at indicated time (n = 5 for each group).

(F) The body weight of CT26 murine model in each group at indicated time points (n = 5 for each group).

(G) The tumor volume of CT26 murine model in each group at 15th day after tumor injection (n = 3–5 for each group).

(H) The tumor weight of CT26 murine model in each group at 15th day after tumor injection (n = 3–4 for each group).

Data are shown as mean ± SEM; one-way ANOVA, ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001;.

CSF1/CSF1R Axis Is Activated in Patients with TNBC and Colorectal Cancer and Informs Clinical Prognosis

Clinically, we examined the expression patterns of CSF1 and CSF1R in TNBC and colorectal cancer patient samples. We found that in TNBC clinical samples, the expression of CSF1 was primarily located in the epithelial cytoplasm (Figure 6A) and was up-regulated compared with normal breast tissue (Figures 6C and 6E). The expression of CSF1R, however, was primarily restricted to the stromal compartment (Figure 6A) and was overexpressed in TNBC clinical samples compared with normal breast tissue as well (Figures 6D and 6F). Similarly, in colorectal cancer clinical samples, CSF1 expression was also restricted to the epithelial cytoplasm (Figure 6B) and overexpressed compared with normal colorectal tissue (Figures 6G and 6I). CSF1R expression was located to the stromal cells (Figure 6B) and was overexpressed in patients with colorectal cancer compared with normal colorectal tissue (Figures 6H and 6J).

Figure 6

The Expression of CSF1/CSF1R in Human TNBC and CRC Samples

(A) Immunohistochemistry analysis of CSF1 and CSF1R in human TNBC and normal breast tissue samples.

(B) Immunohistochemistry analysis of CSF1 and CSF1R in human colorectal cancer and normal colon tissue samples.

(C) The distribution of CSF1 expression in breast cancer epithelial cells.

(D) The distribution of CSF1R expression in breast cancer stromal cells.

(E) The distribution of CSF1 expression in normal breast epithelial cells.

(F) The distribution of CSF1R expression in normal breast stromal cells.

(G) The distribution of CSF1 expression in CRC epithelial cells.

(H) The distribution of CSF1R expression in CRC stromal cells.

(I) The distribution of CSF1 expression in normal colon epithelial cells.

(J) The distribution of CSF1R expression in normal colon stromal cells.

The red arrows point to the immunopositive cells.

We further investigated whether CSF1/CSF1R axis has any impact on patient prognosis. Colorectal cancer was used as an example to examine the correlation of expression of CSF1 and CSF1R with prognosis in public datasets. We found that, in GSE39582 dataset, both CSF1 (hazard ratio [HR] = 1.83(1.08–3.11), p = 2.21 × 10−2) and CSF1R (HR = 2.55(1.47–4.4), p = 4.99 × 10−4) overexpression was correlated with poorer relapse-free survival (RFS) in stage II colorectal patients (Figures S8A and S8B). This indicated that CSF1/CSF1R axis informs clinical prognosis in patients with cancer. Altogether, these data give support to our finding that combined therapy of ERβ activation and ICB could possibly suppress CSF1/CSF1R axis to impair MDSC infiltration and increase CD8+ T cells recruitment to tumor beds, which overcame the resistance to ICB therapy in tumors.

Discussion

Traditional cancer therapies include surgery, chemotherapy, radiotherapy, and molecularly targeted regimens. Recently, the discovery of immune checkpoint molecules has not only led to paradigm shift of our understanding of immune system but also offered a novel therapeutic option for patients with cancer: ICB therapy (Kalbasi and Ribas, 2019). The therapeutic antibody ipilimumab, targeting CTLA-4 as the first checkpoint inhibitor to be approved for patients with cancer in the clinical setting (Lo and Abdel-Motal, 2017; Rowshanravan et al., 2018), whereas the second immune checkpoint receptor, PD-1, which is expressed by activated T cells, is also considered important for driving T cells into an “exhausted” state (Blank et al., 2019; Wherry and Kurachi, 2015). Blocking either CTLA-4 or PD-1 has led to unprecedented durable responses with a generally favorable toxicity profile (Spallarossa et al., 2018). However, it is reported in large clinical trials that only a fraction of patients respond and many will relapse (Nishino et al., 2017). This has led to continuous investigation of mechanisms that lead to ICB therapy resistance and strategies to overcome the resistance (Minn and Wherry, 2016; Patel and Minn, 2018). For instance, Ishizuka et al. (2019) recently found that loss of function of the RNA-editing enzyme ADAR1 in tumor cells remarkably sensitizes tumors to immunotherapy and overcomes resistance to ICB. Mechanistically, in the absence of ADAR1, A-to-I editing of interferon-inducible RNA species is reduced, resulting in double-stranded RNA ligand sensing by PKR and MDA5, resulting in growth inhibition and tumor inflammation, respectively. Loss of ADAR1 overcomes resistance to PD-1 checkpoint blockade caused by inactivation of antigen presentation by tumor cells. More efforts are ongoing to find novel ways to battle against ICB resistance in patients with cancer. Our study demonstrated that combined therapy with an ERβ agonist and PD-1 antibody showed synergistic effects for tumor treatment compared with monotherapies.

MDSCs represent a heterogeneous subset of myeloid cells with major regulatory functions, which play important roles in diseases, including cancer, autoimmune disease, cardiovascular diseases, and metabolic disorders (Gabrilovich, 2017; Kumar et al., 2016; Pawelec et al., 2019). Specifically, the immuno-regulatory functions of MDSCs are critical for hallmarks of cancer (Kumar et al., 2016). For instance, Calcinotto et al. (2018) reported that MDSCs could secrete IL-23 to drive castration-resistant prostate cancer (CRPC) progression in mice and patients with CRPC. IL-23 secreted by MDSCs can activate the androgen receptor pathway in prostate tumor cells, promoting cell survival and proliferation in androgen-deprived conditions. MDSCs also impact on the therapeutic efficacy of ICB therapy. Sun et al. (2019) demonstrated that tumor-infiltrating CXCR2+ neutrophilic MDSCs (PMN-MDSCs) may prevent optimal responses following both PD-axis ICB and adoptive T cell transfer therapy. Abolishment of PMN-MDSC trafficking with SX-682 enhances T cell-based immunotherapeutic efficacy and may be of benefit to patients with MDSC-infiltrated cancers. Moreover, Zhu et al. demonstrated that CSF1R can functionally reprogram myeloid responses that enhance antigen presentation and productive antitumor T cell responses and synergize with ICB treatment to elicit tumor regressions in pancreatic ductal adenocarcinoma (Zhu et al., 2014). These are consistent with our results that inhibition of MDSC infiltration by selective ERβ agonist, possibly through suppression of CSF1/CSF1R axis, in the tumor microenvironment could potently overcome ICB therapy resistance.

ERβ, which is different from ERα, is primarily involved in the control of epithelial proliferation, neurodegeneration, and immune functions. Its tumor-suppressive functions have made ERβ agonists potential therapeutic options for patients with cancer (Nikolos et al., 2018; Zhao et al., 2019). Our group recently demonstrated that selective ERβ agonist LY500307 could suppress lung metastasis of TNBC and melanoma (Zhao et al., 2018). Mechanistically, although we observed that LY500307 potently induced cell death of cancer cells metastasized to lung in vivo, it does not mediate apoptosis of cancer cells in vitro, indicating that the cell death-inducing effects of LY500307 might be mediated by the tumor microenvironment. Further functional analysis indicated that LY500307 treatment induced significant infiltration of neutrophils in the metastatic niche. LY500307-treated cancer cells increased neutrophil chemotaxis and in vivo neutrophil depletion by administration of anti-Ly6G antibody could reverse the effects of LY500307-mediated metastasis suppression. LY500307 could induce up-regulation of IL-1β in TNBC and melanoma cells, which further triggered antitumor neutrophil chemotaxis. LY500307-mediated suppression of lung metastasis was attenuated in Il1b−/- murine models. The present study has provided another example of immune-modulatory function of ERβ activation in the treatment of cancer. Another interesting finding by the present study is that, from the RNA sequencing studies, a cluster of genes were uniquely up-regulated in tumors treated with combination therapy compared with each drug individually as a monotherapy. This set of genes might account for improved therapeutic effects for tumors compared with other monotherapy regimens.

Collectively, our study has identified a CSF1/CSF1R axis between cancer cells and MDSCs in the tumor microenvironment that ERβ activation could potentially target for the treatment of ICB-resistant tumors. This provides the rationale for the combined use of selective ERβ agonists and immune checkpoint inhibitors in patients with cancer.

Limitations of the Study

The limitations of the study include that the number of clinical samples included in this study is relatively small, which we might enlarge in the future investigations. Another drawback is that two cancer mouse models that are resistant to ICB therapy are used in this study.

Resource Availability

Lead Contact

Further information could be obtained by contacting the Lead Contact, Shengtao Zhou (shengtaozhou@scu.edu.cn).

Materials Availability

Materials are available upon request from Dr. Shengtao Zhou.

Data and Code Availability

RNA sequencing data have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE132529.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.