Development of NR0B2 as a therapeutic target for the re-education of tumor associated myeloid cells

Abstract

Immune checkpoint blockade (ICB) has had limited utility in several solid tumors such as breast cancer, a major cause of cancer-related mortality in women. Therefore, there is considerable interest in alternate strategies to promote an anti-cancer immune response. A paper co-published in this issue describes how NR0B2, a protein involved in cholesterol homeostasis, functions within myeloid immune cells to modulate the inflammasome and reduce the expansion of immune-suppressive regulatory T cells (T_reg). Here, we develop NR0B2 as a potential therapeutic target. NR0B2 in tumors is associated with improved survival for several cancer types including breast. Importantly, NR0B2 expression is also prognostic of ICB success. Within breast tumors, NR0B2 expression is inversely associated with FOXP3, a marker of T_regs. While a described agonist (DSHN) had some efficacy, it required high doses and long treatment times. Therefore, we designed and screened several derivatives. A methyl ester derivative (DSHN-OMe) emerged as superior in terms of (1) cellular uptake, (2) ability to regulate expected expression of genes, (3) suppression of T_reg expansion using in vitro co-culture systems, and (4) efficacy against the growth of primary and metastatic tumors. This work identifies NR0B2 as a target to re-educate myeloid immune cells and a novel ligand with significant anti-tumor efficacy in preclinical models.

1. Introduction

In the last decade, immune therapy has revolutionized cancer treatment. Chimeric antigen receptor (CAR) T cells are remarkably effective but are limited to tumors with known and unique antigens, or at least antigens on normal cells that are dispensable for life (i.e. CD19 on B cells). In theory, immune checkpoint blockade (ICB) promised to apply to all solid tumors. However, in practice, it is only effective in certain tumor types or subtypes. Metastatic breast cancer is an example of this, where ICB has proven to be largely unsuccessful, and is currently only approved for those diagnosed with stage IV triple-negative breast cancer (TNBC) whose tumors also stain positive for programmed death-ligand 1 (PD-L1). Even amongst these patients receiving anti-PD-1 (αPD1) in combination with nab-paclitaxel, only 20–30 % show response with the majority being refractory [1,2]. Genentech recently announced that it will voluntarily withdraw atezolizumab (αPDL1) in combination with nab-paclitaxel for consideration of accelerated approval for metastatic PDL1-positive TNBC (August 2021) [3]. Unfortunately, since TNBC lacks endocrine targets such as the estrogen receptor (ERα) or HER2, no targeted therapies exist other than for BRCA1/2 mutant tumors, resulting in a high recurrence and mortality rate. Therefore, there is an urgent need to better understand alternate immune-suppressive networks in solid tumors such as TNBC and develop therapies that combine ICB with pro-immune drugs that work in a checkpoint-independent manner.

Although undoubtedly multifactorial, one major obstacle to ICB is the highly immune-suppressive microenvironment of certain tumors such as breast – a phenomenon that is strongly maintained by myeloid immune cells, including macrophages, and regulatory T cells (T_regs) [4,5]. Importantly, however, myeloid cells are also critical for antigen presentation and a robust anti-tumor response. Therefore, it is now appreciated that rather than just eliminating or inhibiting myeloid cells, strategies are required to ‘re-educate’ myeloid cells away from being pro-tumorigenic and towards being anti-tumorigenic [6–8].

T_regs are highly immune suppressive. Although protective against autoimmune diseases, they have been associated with poor prognosis and poor response to therapy [9–11]. They lack a unique surface marker, precluding the use of a monoclonal antibody to target them. Rather, they are identified based on their expression of FOXP3 (also known as Scurfin). FOXP3 is a forkhead box protein and attempts to specifically target this family of proteins with small molecules have proven challenging. Thus, the lack of ability to reduce the number or function of T_regs is seen as a major barrier to the treatment of solid tumors [12,13].

Our previous studies have implicated cholesterol metabolism and homeostasis in regulating myeloid immune cell function. One metabolite, 27-hydroxycholesterol (27HC) was found to work through the liver X receptors (LXRs) in myeloid cells to robustly impair T cell expansion [14,15]. In normal cholesterol homeostasis, the nuclear receptor NR0B2 (small heterodimer partner, SHP) feeds back to bind to and attenuate LXR activity [16,17]. Therefore, we postulated that NR0B2 may impair the negative effects of 27HC within myeloid cells. To our surprise though, we found that NR0B2 serves to shift macrophage function in a way that impairs the expansion of T_regs [13]. Specifically, overexpression of NR0B2 in bone marrow derived macrophages (BMDMs) led to decreased T_reg expansion when activated T cells were co-cultured with them. When T cells engineered to recognize the OVA peptide antigen (OTII) were grafted into mice where NR0B2 was specifically knocked out of myeloid immune cells (NR0B2^fl/fl;LysMcre⁺) and then challenged with OVA, an increased expansion of T_regs was observed [13]. Subsequent work has found that NR0B2 regulates several aspects of the inflammasome in myeloid immune cells, ultimately resulting in decreased active IL1β; IL1β driving the expansion of T_regs [18–20] [https://doi.org/10.1016/j.canlet.2024.217086]. Importantly, E0771 mammary tumors and B16–F0 tumors grew at a faster rate in NR0B2^fl/fl; LysMcre⁺ mice, demonstrating the pathophysiological relevance of myeloid cell NR0B2 [https://doi.org/10.1016/j.canlet.2024.217086].

The involvement of NR0B2 in the expansion of T_regs is exciting as NR0B2 is a nuclear receptor. Nuclear receptors typically have well-defined ligand binding domains, making them highly amenable to small molecule targeting [16]. Various ligands for NR0B2 have been reported. One such compound, DSHN, has previously been described as an NR0B2 agonist [21]. Given that NR0B2 expression in myeloid immune cells reduces T_reg expansion, in this manuscript, we explored whether this agonist would have beneficial activities in terms of tumor control. We focus on breast cancer as proof-of-principle given it is often refractory to ICB, but our findings likely apply to all solid tumors. While DSHN did exhibit efficacy, it required high doses and chronic treatment. Thus, we also screened a series of DSHN derivatives and identified one compound with superior efficacy both in vitro and in vivo. Therefore, we establish that NR0B2 is a viable target to modulate myeloid cell function and subsequent frequency of T_regs, removing an immune-suppressive arm within the tumor microenvironment.

2. Results

2.1. NR0B2 is inversely correlated with FoxP3, a marker of T_regs in human breast and ovarian tumors

Our preclinical work has found that NR0B2 within myeloid immune cells results in decreased expansion of T_regs [13]. In order to explore whether this biology was relevant in human breast tumors, we first explored the METABRIC dataset [22]. We found an inverse correlation between NR0B2 and FoxP3 expression, FoxP3 being a marker of T_regs (Fig. 1A). This was apparent when breast tumors were parsed into low and high NR0B2 based on median, where high NR0B2 expression had significantly lower FoxP3 expression (Fig. 1B). Since the interaction between myeloid immune cells and T cells would be expected in all solid tumors, we also explored this correlation in a different female malignancy. As expected, there was an inverse correlation between NR0B2 mRNA and FOXP3 mRNA in human ovarian cancer samples (Fig. 1C). Therefore, highlighting the relevance to human disease, NR0B2 was inversely correlated with the mRNA expression of a marker of T_regs.

Fig. 1. Tumoral expression of NR0B2 is associated with decreased expression of the T_reg marker FoxP3.

(A) Linear regression comparing NR0B2 to FoxP3 expression in human breast tumors. Data obtained from METABRIC. Slope of the line is significantly different than 0 (N = 1904, P = 0.0106). (B) FoxP3 expression is lower in human tumors with high NR0B2 expression (upper quartile compared to lower quartile depicted here, N = 477–478/group, t-test). Data obtained from METABRIC. (C) Linear regression comparing NR0B2 to FoxP3 expression in human ovarian tumors. Data obtained from the TCGA Firehose Legacy via cBioPortal. Slope of the line is significantly different than 0 (N = 558, P = 0.0035). (D) Representative micrographs indicating in situ hybridization (ISH) staining for NR0B2 and anti-FOXP3 staining (IHC) in serial sections from human breast tumors. Sections were counterstained with cytokeratin to differentiate between tumoral and stromal regions. Red arrows point out representative positive staining. ISH staining is black. FOXP3 staining is brown. The cytokeratin counterstain is red-maroon. Example micrographs at a higher digital magnification are shown on the far right. Scale bars for each miocrograph are in the bottom lower left. (E) There is an inverse correlation between NR0B2 and FOXP3 in a French cohort of breast cancer samples. Table indicating Spearman correlation and P value.

We then obtained serial sections from 93 human breast tumors and stained for NR0B2 and FOXP3 on serial sections. Extensive optimization failed to find an antibody specific or suitable for protein staining of NR0B2, so we instead used an in situ hybridization approach for this target. FOXP3 was detected with an antibody against this protein. Sections were counter-stained with cytokeratin allowing us to evaluate potential differences between tumor nests and stromal regions (Fig. 1D). In strong support of our preclinical findings [13] [https://doi.org/10.1016/j.canlet.2024.217086], NR0B2 expression was inversely correlated with FOXP3 (Fig. 1E). When assessing all tumor types together, this inverse correlation was observed in both the tumor nest itself and in the surrounding stromal tissue. The inverse correlation between NR0B2 and FOXP3 appeared to be driven by TNBC cases, since no significant correlations were found in the other subtypes. However, when considering the tumor nests themselves, all subtypes showed an inverse correlation (ERα⁺/PR⁺ with a P value of 0.09), although the analyses of individual subtypes were underpowered to make firm conclusions (Fig. 1E). In summary, within human breast tumors, NR0B2 is inversely correlated with both FoxP3 mRNA and FOXP3 protein.

2.2. NR0B2 mRNA expression within different solid tumors is associated with increased recurrence free survival time

We next wanted to further validate NR0B2 as a potential target for solid tumors. We assessed data from the Kaplan-Meier Plotter, which pulls data from GEO, EGA, and TCGA [23]. When breast tumors were parsed into “high” and “low” NR0B2 expression, we found that elevated NR0B2 was associated with a significantly improved recurrence-free survival time (Fig. 2A). Importantly, this was apparent when all breast cancer subtypes were considered together, or when Luminal A, B, or Basal were considered independently (Fig. 2B–D). The HER2 subtype was underpowered to draw firm conclusions (P < 0.2682, Fig. 2E). We also found that lung tumors with increased NR0B2 were associated with improved overall survival (Fig. 2F), an important finding given the prevalence of lung cancer, and that the lung is a clinically common metastatic site for breast cancer. Several other cancers that were assessed also showed an association between elevated NR0B2 expression and overall survival, including stomach cancer adenocarcinoma, ovarian cancer, bladder cancer, renal clear cell carcinoma, and lung adenocarcinoma (Supplemental Fig. 1A). Given the role of myeloid cell NR0B2 in reducing T_reg expansion, the association between NR0B2 and survival being consistent across different breast cancer subtypes and several different cancer types reinforced our premise that it is likely NR0B2 within myeloid cells that drives these correlations, since myeloid cells would be common across solid tumor types.

Fig. 2. NR0B2 mRNA expression within breast and lung tumors is associated with an increased recurrence free survival time.

The Kaplan-Meier plotter was used to probe associations between NR0B2 expression and survival. (A) NR0B2 expression in breast tumors is associated with increased recurrence free survival time when all breast cancer subtypes are considered. (B) Luminal A cases only. (C) Luminal B cases only. (D) Basal cases only. (E) HER2+ cases only. (F) NR0B2 expression in lung cancer is associated with increased overall survival time. (G) NR0B2 mRNA expression within tumors from patients treated with immune checkpoint blockers (ICB) is associated with an increased progression free survival time, and (H) overall survival. For G & H, the Kaplan-Meier plotter was used to probe associations between NR0B2 expression in both genders and all cancers within the database. Further data in Supplementary Fig. 1. Kaplan-Meier survival analysis and P values were determined using the Log-rank (Mantel-Cox) method.

Perhaps even more strikingly, we found a strong association between NR0B2 expression and both increased progression free and overall survival in patients treated with ICB (Fig. 2G and H, Supplemental Figs. 1B–C). These analyses were performed on tumors from both sexes and “all cancers”, which included 90 cases of bladder, 103 cases of esophageal adenocarcinoma, 28 cases of glioblastoma, 22 cases of hepatocellular carcinoma, 110 cases of head and neck squamous cell carcinoma, 570 cases of melanoma, 21 cases of non-small cell lung cancer and 348 cases of urothelial cancer. Although slightly underpowered, these trends are observed regardless of the ICB considered: αPD-1, αPD-L1 or αCTLA-4 (Supplemental Figs. 1B–C). In conclusion, NR0B2 expression in tumors is associated with improved prognosis, even for those cases treated with ICB, underscoring its potential utility as a therapeutic target.

2.3. DSHN, a small molecule NR0B2 agonist regulates expected genes in mammary cancer cells, but has no effect on proliferation

DSHN has previously been described as an NR0B2 agonist [21]. In models of hepatocellular carcinoma, DSHN impaired cellular migration. Some evidence suggests that this was due to a decrease in CCL2 synthesis and secretion [21]. However, the effects of DSHN on cancers with lower NR0B2 expression than liver cancers has not yet been explored. We found that NR0B2 mRNA was detectable in mammary cancer cell lines, but its expression was quite low compared to liver cancer cell lines or primary liver tissue (Supplemental Fig. 2). Therefore, it was important to determine whether DSHN activation of NR0B2 in cancer cells themselves had functional implications.

It was first necessary to ascertain that DSHN was activating NR0B2 in mammary cancer cells. NR0B2 is involved in an important regulatory loop for cholesterol homeostasis. When cholesterol levels are high, oxysterol metabolites activate the liver × receptors (LXRs) which serve to upregulate genes involved in cholesterol catabolism and efflux such as ABCA1. When downstream cholesterol metabolites (bile acids) accumulate, they activate the farnesoid × receptor (FXR) which upregulates the transcription of NR0B2. NR0B2 then feeds-back to physically associate with and impair LXR activity [16]. As expected, when 4T1 mammary cancer cells were treated with the synthetic LXR agonist GW3965, an upregulation of the LXR target gene abca1 was observed (Supplemental Figs. 3A–B). When DSHN was co-treated with GW3965, the induced expression of abca1 was attenuated (Supplemental Figs. 3A–B). Importantly, when NR0B2 was knocked down with siRNA, the inhibitory properties of DSHN were lost (Fig. 3A and B), indicating that DSHN was working through its intended target, NR0B2. Thus, although the expression of NR0B2 was relatively low in 4T1cells, NR0B2 could still be activated by DSHN and was still functionally able to regulate cholesterol homeostasis.

Fig. 3. Treatment of BMDMs or tumor-isolated CD11C⁺cells with DSHN results in decreased T_reg expansion.

(A) Co-culture of DSHN pre-treated murine BMDMs with T cells results in decreased T_reg expansion in a dose dependent manner. BMDMs were treated for 24hrs with indicated doses of DSHN. They were then co-cultured with naïve CD4⁺ T cells that had been activated with CD3/CD28 Dynabeads. Co-culture was for 72hrs under suboptimal T_reg inducing conditions (0.5 μg/mL anti-CD3, 0.5 ng/mL TGFβ and 0.5 ng/mL IL-2). After 72hrs, resulting T_reg (CD4⁺;FoxP3⁺) was quantified by flow cytometry. (B) Dendritic-like cells (CD11B⁺; CD11C⁺) and CD4⁺ T cells were isolated from MMTV-PyMT tumors by FACS, co-cultured and treated with vehicle or DSHN (50 μM) for 3d. The resulting frequency of T_reg (CD4⁺;CD25⁺;FoxP3⁺) was quantified by flow cytometry, (N = 3–5/group, t-test). (C) T cells expanded in the presence of DSHN treated BMDMs were less suppressive than those expanded in the presence of vehicle treated BMDMs, in terms of a subsequent co-culture of naïve T cells. DSHN treated BMDMs were co-cultured with naïve CD4⁺ T cells for 72hrs under optimal T_reg inducing conditions (0.5 μg/mL anti-CD3, 1 ng/mL TGFβ and 1 ng/mL IL-2), resulting cells were co-cultured with CFSE/CTV labeled CD8 T cells in the presence of CD3/CD28 Dynabeads at 2 cells per bead ratio and 0.5 ng/mL IL-2. The percentage of divided CD8⁺ T cells in the second round was quantified. Statistically significant differences are denoted by a line and asterisk (*, P < 0.05, 1-way ANOVA followed by Tukey’s multiple comparison test). (D) Treatment CD11C⁺ enriched PBMCs from healthy human volunteers with DSHN (50uM) resulted in decreased T_reg expansion (N = 3, paired analyses using Fisher’s exact test, P < 0.05). (E) RNAseq of murine BMDMs treated with placebo (DMSO), DSHN, the LXR agonist GW3965, or a combination of GW3965 and DSHN. Multidimensional Scaling (MDS) plot shown here. (F) Heatmap illustrating log₂ of the counts per million with a false discovery rate of <0.01. (G) Weighted correlation network analysis (WGCNA) found 16 unique modules, including several indicating that DSHN attenuated the GW3965 response (see modules 0, 6, 7, 8, 9, 11 and 12). All modules had an adjusted P value of <0.05. (H) Signature of genes differentially upregulated by DSHN in BMDMs is associated with improved overall survival in breast cancer patients. (I) Signature of genes differentially downregulated by DSHN in BMDMs is associated with improved overall survival in breast cancer patients.

We next wanted to determine whether DSHN regulated cancer cell proliferation; rapid, uncontrolled proliferation being a hallmark of cancer [24]. We found that DSHN treated mammary cancer cells were not significantly affected in terms of proliferation (Supplemental Fig. 3C). This was true for human and murine cells, and across different subtypes: MCF7 being ERα+, Sk-br-3 being HER2+, and MDA MB 231, 4T1 and Met1 cells being models of TNBC (Supplemental Fig. 3C). DSHN’s lack of effect on cellular proliferation was perhaps somewhat expected, since our data to date suggest that it is the role of NR0B2 in myeloid immune cells that is most important for reducing tumor growth [13] [https://doi.org/10.1016/j.canlet.2024.217086].

2.4. DSHN treated myeloid immune cells reduce subsequent expansion of T_regs

We have previously shown that overexpression of NR0B2 in different myeloid immune cell types results in decreased T_reg expansion when co-cultured with activated T cells. Similarly, increased T_reg expansion was observed when NR0B2 was knocked out of myeloid immune cells [13] [https://doi.org/10.1016/j.canlet.2024.217086]. Therefore, it was of importance to evaluate whether the small molecule NR0B2 agonist DSHN modulated T_reg expansion.

Murine BMDMs were cultured and treated with DSHN. T cells were activated by exposure to antibodies against CD3 and CD28, stained with CFSE (carboxyfluorescein succinimidyl ester) and then co-cultured with the BMDMs for 3 days. Importantly, DSHN reduced subsequent T_reg expansion in a dose-related manner (Fig. 3A). Since myeloid cells are thought to be uniquely ‘polarized’ within the tumor microenvironment, and in order to determine whether DSHN could overcome this, we isolated CD11B⁺;CD11C⁺ myeloid cells from mammary tumors of MMTV-PyMT mice. These mice express the Polyoma Virus middle T viral antigen under the control of the mouse mammary tumor virus promoter/enhancer, and thus spontaneously develop mammary tumors [25,26]. CD11B⁺;CD11C⁺ myeloid cells from these tumors were treated with vehicle or 100 μM DSHN in culture. They were then co-cultured with activated CD4⁺ T cells. Importantly, DSHN treatment of these myeloid cells resulted in decreased T_reg expansion, indicating that DSHN is likely effective across multiple myeloid cell types, even those extracted from an immune-suppressive tumor microenvironment (Fig. 3B).

In order to demonstrate that the T cells expanded in co-culture with BMDMs treated with DSHN were indeed less suppressive (as in, contained fewer T_regs) than those expanded in vehicle treated BMDMs, we assessed subsequent CD8⁺ cytotoxic T cell expansion. Specifically, BMDMs were treated with either vehicle, 50 or 100 μM DSHN. CD4⁺ T cells were activated with αCD3/28 antibodies and then were co-cultured with these BMDMs. Resulting expanded T cells were subsequently co-cultured with freshly activated and stained CD8⁺ T cells, for 72h. CD8⁺ cell expansion was assessed by flow cytometry for dilution of stain. As expected, we observed greater CD8⁺ expansion when cocultured with CD4⁺ T cells expanded in the presence of BMDMs treated with DSHN (Fig. 3C). However, this was only observed at the highest dose of DSHN tested (100 μM).

To demonstrate that DSHN regulates this axis similarly in mice and humans, we made use of peripheral blood mononuclear cells (PBMCs) from healthy volunteers. CD11C + cells were enriched and treated with DSHN for 24h prior to co-culture with activated T cells. Importantly, DSHN-treated PBMC-CD11C⁺ cells resulted in significantly decreased T_reg expansion in all three volunteer samples assessed (Fig. 3D). Therefore, the putative NR0B2 agonist DSHN is able to work through myeloid immune cells to impair T_reg expansion, an activity which is conserved between mice and humans.

One major described function of NR0B2 is to inhibit the activities of LXR and LRH-1, thereby decreasing cholesterol catabolism and efflux. Other genes regulated by NR0B2 within myeloid cells are less clear. Therefore, we next wanted to explore how treatment with DSHN changed mRNA expression of genes as a single compound or when LXR was activated. Murine BMDMs were treated with vehicle or GW3965, a synthetic LXR agonist, plus or minus DSHN. RNAseq analysis was then performed. As illustrated on the multidimensional scaling (MDS) plot, GW3965 significantly altered the transcriptome in two major dimensions (Fig. 3E). DSHN did induce a unique transcriptional profile. Combined DSHN and GW3965 treatment led to a decrease dimension 2, back to values similar to vehicle. Weighted correlation network analysis (WGCNA) found 16 distinct modules of treatment-related differentially regulated clusters (Fig. 3F and G). Of those, six different modules found that the combined GW3965 plus DSHN treatment led to an attenuation or reversal of eigenvalues regulated by GW3965 alone (Fig. 3G). Other modules of were not attenuated but rather exaggerated by DSHN, suggesting that these were regulated by NR0B2 in an LXR-independent fashion. Interestingly, and further highlighting the clinical relevance of myeloid expressed NR0B2, are our findings that a signature of genes upregulated by DSHN was associated with increased overall survival time (Fig. 3H). A signature of genes downregulated by was also associated with improved overall survival (Fig. 3I), indicating that this nuclear receptor likely has multiple, pathologically relevant regulatory roles in the tumor microenvironment. Collectively, these data indicate that NR0B2 is a valid target in myeloid immune cells with an effect on T_reg expansion, and that DSHN is a ligand that could be used for proof-of-principle studies.

2.5. Tool compound, DSHN, an agonist of NR0B2 reduces breast tumor progression in preclinical models

T_regs have long been known to promote tumor progression and hinder the efficacy of various treatments. However, the main differentiating protein, FOXP3, has proven challenging to target therapeutically. Therefore, our finding that NR0B2 within myeloid cells results in skewed T_reg expansion [13] presents an opportunity to alter T_reg abundance indirectly. Our work thus far suggested that DSHN would be a good proof-of-principle NR0B2 agonist to test in murine models of mammary cancer.

Orthotopic E0771 mammary tumors were grafted and allowed to grow to 100 mm³, at which point daily treatment with placebo or DSHN commenced. For these studies, we also included groups treated with a control antibody or one against PDL1 (αPDL1, simulating ICB). As expected, αPDL1 alone did not have a significant effect on tumor growth. However, co-treatment with DSHN did result in modest but significantly reduced tumor growth after 19 days (Fig. 4A). We then evaluated DSHN in a model where a low suboptimal number of E0771-luc cells were grafted orthotopically. This model has a long latency, allowing us to investigate the longer-term actions of DSHN. Treatment was initiated one day post-graft and continued for 18d. Mice were then monitored through time. DSHN combined with αPDL1 significantly increased survival time compared to placebo treated mice (Fig. 4B). One DSHN-only treated mouse lived for more than 400 days post graft with no palpable tumor forming, although this may have been due to a non-DSHN related or stochastic cause. Therefore, DSHN had some efficacy against primary E0771 tumors.

Fig. 4. DSHN treatment reduces mammary tumor and metastasis growth in murine models.

(A) DSHN increases efficacy of αPDL1 therapy on orthotopic E0771 tumors. An optimal number of E0771 cells were grafted into mice and allowed to grow until tumors reached 100 mm³. Mice were euthanized on day 19 (waterfall plot shown with each bar representative of one mouse, with the dashed line indicating the threshold used to calculate the ratio used in the Fisher’s exact test). (B) The effects of DSHN on subsequent tumor outgrowth of a sub-optimal E0771 orthotopic graft. C57BL/6 Mice were orthotopically grafted with a low number of E0771-luc mammary cancer cells. After allowing one day for tumor establishment, daily treatment with placebo or DSHN commenced. Immune checkpoint inhibitor treatment (αPDL1) was initiated 3 days post-graft, and continued every 2 days for six total treatments. 18d postgraft, all treatments were stopped and mice were monitored for subsequent tumor out-growth. Time from graft to developing a tumor burden of >2000 mm³ is depicted. Black dash represents censored mouse that had no palpable tumor 437d post graft (N = 10/group, Log-rank test, * indicates statistical significance, ns: not significant). (C) DSHN increases efficacy of αPDL1 therapy on orthotopic 4T1 tumors. 4T1 cells were orthotopically grafted into mice and allowed to establish for 5d prior to treatment start. Mice were euthanized on day 16 (waterfall plot shown with each bar representative of one mouse, with the dashed line indicating the threshold used to calculate the ratio used in the Fisher’s exact test). (D) DSHN reduces metastatic outgrowth of 4T1 tumors and enhances efficacy of αPDL1 compared to placebo (p) (2-way ANOVA followed by Šidák’s multiple comparison test for day 16). 4T1 cells were grafted intravenously, and lung metastatic lesions allowed to establish for 3d before treatment commenced. At the end of the study, lungs were removed and imaged ex vivo, quantified data indicated (N = 10, 1-way ANOVA followed by Newman-Keuls multiple comparison test). Representative luciferase images shown to the right. * or different letters denote P < 0.05 using indicated statistical test. (E) mRNA for NR0B2 and FoxP3 was assessed in the 4T1 metastatic lungs from (D), by qPCR (N = 15, 1-way ANOVA followed by Newman-Keuls multiple comparison test). (F) Nanostring analysis of isolated CD11B+; CD11C + cells from 4T1 metastatic lungs in mice treated with either placebo or DSHN. Principal component analysis shown here, when all assessed gene expressions were input. (G) Fold change of select genes implicated in the inflammatory pathway significantly downregulated by DSHN in the 4T1 metastatic lungs of mice.

We next assessed the effects of DSHN combined with αPDL1 on growth of primary 4T1 tumors. 4T1 cells were grafted orthotopically and allowed to establish tumors for 5d, at which point treatment was initiated and continued until day 16. Combined DSHN and αPDL1 treatment significantly reduced tumor growth compared to placebo with αPDL1 (Fig. 4C). Thus in two independent models, DSHN displayed some efficacy in slowing tumor growth when combined with immune checkpoint blockade. Next, it was important to evaluate whether DSHN had efficacy against metastatic lesions. For this, we grafted mice intravenously with 4T1 cells and waited 3d for overt lesions to form in common metastatic tissues (primarily lung and bone for the 4T1 model). In this model, DSHN alone slowed the growth of metastatic lesions (Fig. 4D). Combining DSHN with αPDL1 reduced outgrowth even further than DSHN alone (Fig. 4D), indicating that simultaneously targeting the innate and adaptive arms of the immune system will be beneficial for the treatment of immune-suppressive tumors such as breast cancer. qPCR analysis of the lungs from that experiment indicated that DSHN did not significantly alter NR0B2 gene expression, but did decrease FoxP3 expression (Fig. 4E). In a separate experiment, 4T1 metastatic lesions were allowed to form prior to acute treatment with placebo or DSHN. Lungs were then harvested and CD11B⁺ and CD11C⁺ cells were enriched for and subjected to Nanostring analysis of a myeloid cell panel. Principal component analysis revealed distinct transcriptional clustering between placebo and DSHN treated mice (Fig. 4F). Several genes previously implicated in the inflammasome response were downregulated by DSHN (Fig. 4G); the suppression of the inflammasome being a putative mechanism by which NR0B2 in myeloid cells attenuates T_reg expansion [https://doi.org/10.1016/j.canlet.2024.217086].

In summary, the tool compound DSHN provides proof-of-principle that the NR0B2-myeloid cell-T_reg axis can be targeted for the treatment of solid tumors. However, with the exception of 4T1 metastatic lesions, the effects of DSHN on tumor growth were not large. Furthermore, our in vitro studies required large doses of DSHN to observe significant regulation. Therefore, it was necessary to develop more efficacious NR0B2 agonists.

2.6. Development of a more efficacious NR0B2 agonist, DSHN-OMe

While DSHN was a useful proof-of-concept small molecule, BMDMs required high doses and chronic exposure for robust influences on subsequent T_regs expansion, potentially limiting its in vivo translation. Therefore, we synthesized several derivatives of DSHN and screened them using a reporter assay which makes use of the endogenous cholesterol homeostatic feedback loop, where FXR activation upregulates NR0B2, which then inhibits LXR and its normal induction of ABCA1 (Fig. 5A). Human hepatocellular carcinoma HEPG2 cells were used for the primary screen. HEPG2 cells were transfected with a reporter where luciferase expression was driven by the ABCA1 promoter. As expected, the LXR agonist GW3965 increased luciferase activity (Fig. 5B). The assay was validated by testing two known FXR agonists which would be expected to upregulate NR0B2: GW4064 a synthetic agonist, and obeticholic acid, a semisynthetic bile acid [27,28]. Both FXR agonists reduced luciferase activity in a dose-related manner.

Fig. 5. DSHN-OMe identified as NR0B2 agonist in physiologically relevant screen.

(A) Overview of screen used. Bile acids other FXR agonists upregulate NR0B2, which then inhibits LXR activity including the induction of ABCA1. For the screen, LXR was activated with the agonist GW3965 and the readout was reduction in ABCA1-luciferase activity. (B) The screen effectively identifies that FXR ligands obetecholic acid and GW4064 reduce ABCA1-luciferase induction by LXR agonist GW3965, in a dose related manner. Data was normalized to 100 μM GW4064 (0 %) and GW3965 alone (100 %). This normalized scale indicates % maximum response (%) which corresponds to treatment with GW3965. A linear regression model was applied to the data from each group, and whether the slope of the line was significantly different from 0 was determined (Wald test using the simple logistic model β1, testing departure from odds ratio of 1.0). A P-value of <0.05 is indicated by an asterisk (*), beside the legend. (C) Overview of screening DSHN derivatives, using strict cutoffs with respect to DSHN for passing each metric. Corresponding data is in Supplementary Fig. 4.

DSHN derivatives were first screened using this ABCA1-reporter assay, where two cutoffs were applied: (1) the compound had to reduce basal luciferase activity to below 50 % of the vehicle (DMSO), and (2) the compound had to reduce the GW3965 induction of luciferase activity by 85 % (Fig. 5C and Supplementary Fig. 4A). We then followed this screen with several other sub-screens with stringent cutoffs: effect on endogenous ABCA1 and IL1β mRNA expression in RAW 264.7 myeloid cells, reduction of T_reg expansion in either BMDM or splenic CD11C⁺ co-cultures, and low cellular toxicity in an assay measuring reduction of resazurin to resorufin (Fig. 5C and Supplementary Fig. 4).

The methyl ester of DSHN (DSHN-OMe) emerged as the best performer across all these endpoints (Fig. 5C; preparation and characterization in Supplementary Data). We next performed dose response curves assessing the ability to suppress the induction of ABCA1 mRNA in murine BMDMs. The parental DSHN had very little effect until higher doses, with an estimated inhibitory concentration 50 % (IC₅₀) for DSHN was 56.2 μM (Fig. 6A). In comparison, DSHN-OMe had an estimated IC₅₀ of only 3.8 μM (Fig. 6B). In an independent experiment, a concentration of 50 μM DSHN-OMe was able to inhibit the induction of ABCA1 by GW3965, while the same concentration of the parental DSHN was not (Fig. 6C). DSHN-OMe showed a modest shift in NR0B2 thermal melt temperature, as did DSHN (Fig. 6D). The T_M for apo/unliganded was 46.74 ± 0.22, DSHN-OMe at 46.92 ± 0.17 and DSHN at 46.34 ± 0.51°C (Fig. 6D). Both molecules similarly reduced the magnitude of the transition suggesting that they are likely binding to the protein but not significantly affecting thermal stability. In further support of DSHN-OMe working through NR0B2, are the findings that the ability of DSHN-OMe to attenuate the induction of ABCA1 expression by the LXR agonist GW3965 was lost when NR0B2 was knocked down in BMDMs (Fig. 6E).

Fig. 6.

DSHN-OMe attenuates LXR induction of ABCA1 in an NR0B2 dependent manner.

(A) Dose response assay in BMDMs pre-treated with increasing doses of DSHN and then treated with 1 μM GW3965, using ABCA1 mRNA as an endpoint. Data was normalized to 0.01 μM DSHN (0 %) and GW3965 alone (100 %). DSHN-OMe (20 μM) is indicated for comparison. Data was fit to a 4-parameter variable slope model, and IC₅₀ estimated at 56.2 μM. (B) Dose response assay in BMDMs pre-treated with increasing doses of DSHN-OMe and then treated with 1 μM GW3965, using ABCA1 mRNA as an endpoint. Data was normalized to 0.01 μM DSHN-OMe (0 %) and GW3965 alone (100 %). DSHN (100 μM) is indicated for comparison. Data was fit to a 4-parameter variable slope model, and IC₅₀ estimated at 3.8 μM. (C) At a dose of 50 μM, DSHN-OMe effectively reduces induction of LXR target gene ABCA1 mRNA expression by GW3965 while the parental compound DSHN does not. Lines and asterisks denote statistical difference between GW3965 treated groups (N = 3–4/group, 1-way ANOVA followed by Šidák test). (D) Thermal shift assay using recombinant protein in presence of a fluorescent dye indicates that DSHN-OMe binds to NR0B2. Plot of raw data is above with a plot of the first derivative below. Note the modest shift in melt temperature for both DSHN and DSHN-OMe. (E) DSHN-OMe is ineffective at regulating ABCA1 mRNA when NR0B2 is knocked down with siRNA in BMDMs. Lines and asterisks denote statistically significant differences between treatment groups (P < 0.05, 1-way ANOVA followed by Šidák test).

2.7. DSHN-OMe does not affect BMDM viability but significantly attenuates BMDM-regulated T_reg expansion

In order to better assess the potential cellular toxicity of DSHN-OMe, we treated BMDMs with increasing doses and measured their ability to reduce resazurin to resorufin (similar to the MTT viability assay). Importantly, no significant difference was observed in viability up to 100 μM for either DSHN-OMe or its parental compound, DSHN (Fig. 7A). However, compared to DSHN, DSHN-OMe treated BMDMs elicited a reduction in T_reg expansion with a lower dose, and had a larger biological maximum (Fig. 7B). This was also evident in splenic CD11C⁺ (DC-enriched) cells, where 50 μM DSHN-OMe treatment resulted in significantly fewer T_regs compared to twice the dose of DSHN (100 μM) (Fig. 7C). Importantly, T cells resulting from expansion in co-culture with BMDMs that were pretreated with either DSHN-OMe or DSHN were less suppressive when subsequently cultured with a second set of naïve T cells, specifically assessing CD8⁺ expansion (Fig. 7D). An overview of the experimental procedures for Fig. 7B–D is depicted in Supplementary Fig. 5.

Fig. 7. DSHN-OMe treated BMDMs or splenic CD11C þ cells effectively reduce subsequent T_reg expansion.

(A) DSHN-OMe has minimal effects on BMDM viability (N = 4–6/group). Primary BMDMs were treated at indicated concentrations of DSHN or DSHN-OMe for 24h. Subsequent reduction of resazurin to resorufin (as measured by fluorescence [560 nm excitation and 590 nm emission]). (B) DSHN-OMe treated BMDMs reduce T_reg expansion in a dose related manner, with an increased reduction compared to DSHN (data normalized to mean vehicle, which was set at 100 %. Non-linear three parameter regression shown, N = 4/point). Asteriscs denote statistically significant differences at indicated doses (Two-Way ANOVA followed by Šidák test, P < 0.05). An overview of experiments shown in panels B–D is depicted in Supplementary Fig. 5. (C) DSHN-OMe treated splenic derived CD11C + cells result in decreased T_reg expansion compared to twice the dose of DSHN. 50 μM DSHN-OMe treated CD11C + cells primed with OVA results in decreased T_reg expansion of OTI T cells compared to vehicle (DMSO) or 100 μM DSHN. Lines and asterisks denote P < 0.05 (1-way ANOVA followed by Šidák test). (D) T cells expanded in the presence of DSHN- or DSHN-OMe-treated BMDMs were less suppressive than those expanded in the presence of vehicle treated BMDMs, in terms of a subsequent co-culture of naïve T cells. The percentage of divided CD8⁺ T cells in the second round was quantified. (N = 3–5/group, lines and asterisks denote statistical significance, 1-way ANOVA followed by Tukey). (E) BMDMs pretreated with DSHN-OMe for 24h prior to washout and subsequent co-culture results in decreased T_reg expansion, while DSHN does not. Vehicle (DMSO) treated BMDMs are indicated on the right for comparison. Asterisks denote statistically significant differences at indicated doses (2-Way ANOVA followed by by Šidák’s post-hocm P < 0.05). (F) Cellular uptake of DSHN-OMe greatly exceeds that of DSHN though time. RAW264.7 cells were treated with 50 μM of vehicle (DMSO) or indicated compound for 4h, 8h or 24h. Compound was washed off, cells were lysed and resulting DSHN and DSHN-OMe contents were assessed by LC-MS/MS. Note different scales on y-axes. For DSHN treated cells, no DSHN-OMe was detected (not shown). For DSHN-OMe treated cells, some DSHN was detected. ND signifies when compounds were not detected. (N = 3/group. 4h timepoint was run in an experiment independent of the 8 and 24hr timepoints).

In a separate experiment, BMDMs were pretreated with compound and washed with PBS several times before incubation with expanding T cells. Here, DSHN was not effective at reducing T_reg expansion, while DSHN-OMe was (Fig. 7E). Collectively, these data supporting the superiority of DSHN-OMe compared to its parental compound suggested that DSHN-OMe either had better cellular penetration or more durable effects on myeloid cells compared to DSHN.

Conversion of a carboxylic acid to an ester is often associated with better cellular uptake [29]. Therefore, we explored whether the kinetics of uptake differed between the two compounds, using RAW 264.7 cells as a model for myeloid immune cells. RAW 264.7 cells were treated with 50 μM of DSHN or DSHN-OME for 4, 8 or 24h. At the appropriate time-point cells were washed and subjected to LC-MS/MS to quantify intracellular compound concentration. Surprisingly, DSHN had very little uptake (Fig. 7F). At 4hrs, DSHN was below detection limits. At 8h, 12.6 nM DSHN was detected. By 24h, no intracellular DSHN was detected. On the other hand, DSHN-OMe was detected at 7,481 nM by 4h, increasing to 19,279 nM by 8h, which only dropped slightly to 11, 089 nM at 24h (Fig. 7F). Interestingly, in DSHN-OMe treated cells, we were also able to detect small quantities of DSHN (<130 nM), higher than DSHN treated cells. This indicated that some DSHN-OMe was being metabolized back to its parental compound upon cellular entry (Fig. 7F). Given these results, we cannot rule out the possibility that the activity of DSHN-OMe is due to conversion to DSHN, although the shear difference in intracellular concentration (>2 log) would suggest that DSHN-OMe is likely the active modulator of NR0B2 (Fig. 7F).

2.8. DSHN-OMe has significant efficacy against murine models of breast and lung cancer

Given the enhanced cellular uptake and ability to attenuate myeloid cell-modulated T_reg expansion, it was next important to evaluate the efficacy of DSHN-OMe in in vivo models of mammary cancer. For these studies we made use of the syngeneic E0771 model, allowing us to also test the compound in mice lacking myeloid expression of NR0B2 (NR0B2^fl/fl;LysMCre⁺ mice). Of note, and as expected based on other cancer models [https://doi.org/10.1016/j.canlet.2024.217086], E0771 tumors grew at an accelerated rate in NR0B2^fl/fl;LysMCre⁺ mice compared to control NR0B2-replete mice (NR0B2^+/+;LysMCre⁺ mice Fig. 8A and B).

Fig. 8. DSHN-OMe reduces tumor and metastatic growth in murine models, in a myeloid cell-NR0B2 dependent manner.

(A) E0771 cells were grafted into control mice (NR0B2^+/+;LysMCre⁺) or mice where NR0B2 was knocked out in myeloid immune cells (NR0B2^fl/fl;LysMCre⁺). 8 days post-graft, daily treatment with placebo or DSHN-OMe was initiated and subsequent tumor growth followed through time. Lines and asterisks (*) denote statistically significant differences on the final measurement (day 21, 2-Way ANOVA followed by Šidák’s post-hoc, N = 5–8). (B) Waterfall plot illustration of data in (A) showing tumor volume at final day of measurement. Asterisk indicates significant difference (Chi-squared test, dashed line indicating threshold, P < 0.05). (C) Lewis Lung cells were grafted into control mice (NR0B2^+/+;LysMCre⁺) or mice where NR0B2 was knocked out in myeloid immune cells (NR0B2^fl/fl;LysMCre⁺). 10 days post-graft, daily treatment with placebo or DSHN-OMe was initiated and subsequent tumor growth followed through time. Lines and asterisks (*) denote statistically significant differences on the final measurement (day 21, 2-Way ANOVA followed by Šidák’s post-hoc, N = 8–13). (D) Waterfall plot illustration of data in (C) showing tumor volume at final day of measurement. Asterisk indicates significant difference (Chi-squared test, dashed line indicating threshold, P < 0.05). (E) 4T1 murine mammary cancer cells were grafted into wildtype BALB/C mice. 3 days post-graft, daily treatment with placebo or DSHN-OMe was initiated and subsequent tumor growth followed through time. Asterisks denote statistically significant differences between the two groups at the time-point indicated (2-Way ANOVA followed by Šidák’s post-hoc, N = 10–12). (F) EMT6 murine mammary cancer cells were grafted into wildtype BALB/C mice. 3 days post-graft, daily treatment with placebo or DSHN-OMe was initiated and subsequent tumor growth followed through time. Asterisks denote statistically significant differences between the two groups at the time-point indicated (2-Way ANOVA followed by Šidák’s post-hoc, N = 12). (G) Four tumors from placebo and DSHN-OMe treated groups were selected based on similar size (from E & F). Flow cytometry was performed to quantify Tregs (CD4⁺; FOXP3+). A binomial test was performed with the null hypothesis that of less than 0.2 fold difference between placebo and matched DSHN-OMe Treg quantity (P < 0.05). (H) DSHN-OMe is an effective treatment against metastatic 4T1 mammary cancer growth. 4T1 cells were grafted intravenously, and lung metastatic lesions allowed to establish for 3d before treatment commenced. Metastatic burden through time is presented to the left of metastatic burden on the final day of imaging in the middle panel (N = 12–17/group, mixed-effects model followed by Tukey’s multiple comparison test). Representative luciferase images in the right panel. (I) DSHN-OMe is an effective treatment against metastatic EMT6 mammary cancer growth. EMT6 cells were grafted intravenously, and lung metastatic lesions allowed to establish for 3d. DSHN-OMe was administered on days 4, 5, 8, 9 and 10 post-graft. 11d post-graft lungs were imaged ex vivo for luciferase (N = 15/group). (J) EMT6 metastatic lungs were assessed for Ki67 mRNA and (K) FOXP3 mRNA. (L) EMT6 metastatic lungs were assessed for CD8⁺ cytotoxic T cells (CD8⁺). For panels I–L, a Student’s t-test was used to test differences between groups. Asterisks indicate statistically significant differences (P < 0.05).

Importantly, DSHN-OMe significantly reduced E0771 tumor growth (growth through time in Fig. 8A and waterfall plot in Fig. 8B). However, the growth-reducing effects of DSHN-OMe were only observed in mice with myeloid cell expression of NR0B2, with no significant impact being observed in NR0B2^fl/fl;LysMCre⁺ mice (Fig. 8A and B). Therefore, DSHN-OMe had significant efficacy in a myeloid cell NR0B2 dependent manner, as in the observed effects were being mediated through the anticipated target.

Since T_regs are considered a problem common to many solid tumor types, we reasoned that DSHN-OMe should have broad efficacy. To test this, we evaluated the ability of DSHN-OMe to influence the growth of murine Lewis Lung tumors grown in the flank. Similar to our E0771 results, DSHN-OMe significantly reduced tumor growth through time compared to placebo-treated mice (Fig. 8C and D). Again, these anti-tumor effects were only observed in mice with replete myeloid cell expression of NR0B2 and not in NR0B2^fl/fl;LysMCre⁺ mice (Fig. 8C and D).

We next evaluated whether DSHN-OMe altered primary tumor growth of two additional murine models reflective of TNBC. After allowing for primary tumors to establish for 3 days following orthotopic graft into the mammary fat pad, treatment with either placebo or DSHN-OMe was initiated. The volume of 4T1 primary tumors in mice treated with DSHN-OMe were significantly reduced and even started to show regression (Fig. 8E). Likewise, tumor growth was attenuated in the syngeneic EMT6 model (Fig. 8F). When myeloid cells (CD11B+) were isolated from these tumors, we found that DSHN-OMe had no significant influence on NR0B2 expression in 4T1 tumors, but did increase expression in EMT6 tumors (Supplementary Figs. 6A–B). The therapeutic relevance of this regulation and differences between tumor types is not clear at this point. Tumor size itself has been associated with T_reg abundance in tumors reviewed in Refs. [30,31]. Therefore, we used the natural variance within treatment groups to find tumors of similar sizes and compared their T_reg abundance. As expected, DSHN-OMe treated tumors had significantly reduced T_regs in all four of the matched cases, for both 4T1 and EMT6 tumors (Fig. 8G). These findings were supported by findings that FoxP3 mRNA was decreased in tumors from DSHN-OMe treated mice (Supplementary Fig. 6C). Extensive analysis of different myeloid cell populations by flow cytometry of dissociated cells from the tumors in Fig. 8E&F found only modest changes in Ly6C⁻ ;Ly6G⁺ and Ly6C⁺;Ly6G⁻ populations (sometimes referred to as inflammatory granulocytes or inflammatory monocytes respectively, or myeloid derived suppressor cells) in EMT6 tumors with no significant differences observed in 4T1 tumors (Supplementary Fig. 6D). DSHN-OMe treatment did not significantly alter body weight (Supplementary Figs. 7A–F). DSHN-OMe did slightly increase the expression of ALT, a liver function enzyme, in the primary tumor bearing mice (Supplementary Fig. 7G, corresponding to mice in Fig. 8E and F). However, this was not observed in mice with lung metastases (Supplementary Fig. 7H, corresponding to Fig. 8G). Thus, this increase may be due to the loss of tumor mass itself. Other liver function enzymes assayed were not altered (AST,ALP, GGT). Given the role of NR0B2 in regulation of cholesterol homeostasis, DSHN-OMe may alter plasma cholesterol concentrations. Indeed, the FXR agonist, obeticholic acid, increases LDL-cholesterol and decreases HDL-cholesterol [32]. However, we did not observe any statistically significant changes in plasma cholesterol concentrations (Supplementary Fig. 7I, corresponding to Fig. 8E and F). Collectively, these data suggest that DSHN-OMe is well tolerated, although more comprehensive toxicity testing needs to be completed.

Metastatic breast cancer urgently requires new therapeutic approaches. Therefore, we tested the ability of DSHN-OMe to reduce the subsequent growth of established metastases. 4T1 cells were introduced intravenously and lesions allowed to establish for 3d. Since ICB has been approved for a subset of TNBC patients, we also included groups treated with αPDL1. DSHN-OMe significantly reduced the outgrowth of established 4T1 metastatic lesions as a single agent (Fig. 8H). Addition of αPDL1 did not significantly influence the effects of DSHN-OMe in this assay, although the short time frame (due to humane endpoint of placebo groups) may not have captured the contribution of checkpoint inhibition. We confirmed these findings using a different model of murine TNBC, EMT6. DSHN-OMe significantly reduced metastatic burden in EMT6-grafted mice (Fig. 8I), indicating that DSHN-OMe has broad efficacy against different mammary cancer models. Ki67 mRNA was also decreased in EMT6 metastatic lungs of DSHN-OMe treated mice, suggesting a decreased number of proliferating cancer cells (Fig. 8J). As expected, metastatic lungs from DSHN-OMe treated mice had decreased Foxp3 expression (Fig. 8K), and a corresponding increase in the relative abundance of CD8⁺ cytotoxic T cells (Fig. 8L). Therefore, DSHN-OMe represents an improved NR0B2 agonist with significant anti-tumor effects, both against primary tumor growth and also metastatic outgrowth.

3. Discussion

NR0B2 is a non-canonical nuclear receptor in that it does not have a DNA binding domain. As such, its mechanisms of action are believed to be the result of direct interaction and modulation of other proteins including other nuclear receptors such as LXR and LRH-1. However, NR0B2 does have a well-defined ligand binding domain, making it amenable to modulation by small molecules. While its roles in cholesterol homeostasis have been best described in the liver, we have previously demonstrated that NR0B2 has extra-hepatic expression including in myeloid immune cells [13]. NR0B2 within BMDMs was found to influence T cell expansion where T_regs were decreased [13]. In a paper co-published in this issue, we demonstrate that this NR0B2 activity is conserved in several different myeloid immune cell types. Mechanistically, NR0B2 regulates several aspects of the inflammasome in myeloid cells, resulting in decreased IL1β secretion and the subsequent T_reg expansion [https://doi.org/10.1016/j.canlet.2024.217086]. T_regs are associated with poor response to therapy and poor prognosis, yet have proven to be therapeutically intractable. Therefore, NR0B2 presented as a unique opportunity, especially for metastatic breast cancer and other solid tumors where ICB has been disappointing.

In order to develop NR0B2 as a potential therapeutic target, it was important to first evaluate its role in human disease. In this regard, within breast and ovarian tumors, we found an inverse correlation between NR0B2 and FoxP3, a marker of T_regs. In a separate cohort, our IHC analysis of serially sectioned breast tumors corroborated this inverse correlation between NR0B2 and FOXP3 staining (Fig. 1). Notably, elevated NR0B2 expression within several different tumor types was associated with increased progression-free survival time. Importantly, NR0B2 was also prognostic for survival in cancer patients receiving ICB (Fig. 2). Furthermore, genes that were upregulated in BMDMs after treatment with the putative NR0B2 agonist DSHN were associated with improved overall survival (Fig. 3). Therefore, NR0B2 appears to be having a favorable effect within human tumors, which is consistent with our pre-clinical data [13] [https://doi.org/10.1016/j.canlet.2024.217086].

DSHN had previously been reported to be an NR0B2 agonist [21]. However, before now, it has not been evaluated for in vivo efficacy against any tumor model. When we evaluated this molecule in murine models of mammary tumors, we found that it had modest efficacy in reducing tumor growth when combined with ICB, and as a single agent in a model of metastatic breast cancer (Fig. 4). In vitro, DSHN required high doses and continuous exposure in order to alter T_reg expansion. Thus, we determined that the pharmaceutical properties of DSHN would hinder its translation. Therefore, we developed a series of DSHN derivatives and screened them for their ability to inhibit the LXR-mediated induction of ABCA1. Several derivatives that had efficacy in this assay were then screened in secondary screens with stringent cutoffs (Fig. 5).

The DSHN derivative, DSHN-OMe emerged as being superior. Compared to the parental DSHN, DSHN-OMe had a significantly improved IC₅₀ in terms of inhibiting the LXR-mediated induction of ABCA1 mRNA in BMDMs. These effects were dependent on NR0B2 (Fig. 6). We found that DSHN-OMe had far higher cellular uptake when compared to DSHN, which likely explains its superior efficacy (Fig. 7). Importantly, DSHN-OMe had robust anti-growth effects on primary tumors and established metastatic lesions (Fig. 8). These effects were dependent on the myeloid immune expression of NR0B2; data strongly supporting the critical role of NR0B2 in controlling tumoral T_regs and subsequent growth. Other reports have indicated that NR0B2 can work directly on cancer cells to impair proliferation [16,33,34]. DSHN was found to inhibit hepatocellular carcinoma migration presumably by decreasing CCL2 secretion [21]. These cell-intrinsic effects may be important for certain cancers with high NR0B2 expression such as those of the liver. Our data from murine models of mammary and lung cancer indicate that DSHN-OMe works through myeloid immune cells, as no response was observed when NR0B2 was knocked out of the myeloid lineage.

To identify DSHN-OMe, we made use of the endogenous cholesterol homeostasis feedback loop, whereby NR0B2 inhibits LXR. The activation of the bile acid receptor, FXR, up-regulates NR0B2. Thus, besides directly targeting NR0B2, it may be possible to utilize FXR. In this regard, the semi-synthetic bile acid, obeticholic acid is FDA approved for the treatment of primary biliary cholangitis, and may represent a rapidly translatable avenue for this axis. Unfortunately, obeticholic acid is associated with several side effects including pruritus and altered circulating cholesterol concentrations, resulting in low patient compliance [32,35]. This provides rationale for the continued development of downstream targets such as NR0B2; FXR itself having both NR0B2-dependent and -independent effects. Furthermore, the pharmacology of FXR is complex, whereby agonists as determined by simple receptor-reporter assays may not be reflective of their ability to alter myeloid cell function; as in, the FXR is selectively modulated, similar to the LXR, estrogen receptors and several other nuclear receptors [15,36]. However, acute use of obeticholic acid to re-educate the tumor microenvironment may be one way to limit undesirable side effects. Collectively, our data provide proof of concept that mediators of downstream cholesterol homeostasis can be leveraged for the treatment of solid tumors.

4. Methods

Methods are similar to [https://doi.org/10.1016/j.canlet.2024.217086] and some sections are directly quoted from that text. Statistical analyses were performed with GraphPad Prism, version 6.0 or higher.

DSHN was synthesized by Sai Life (Hyderabad, India). GW3965, PMA, Ionomycin, Obeticholic acid and GW4064 were obtained from Cayman (Ann Arbor, MI). DSHN-OMe was synthesized in house. Anti-CD4, anti-Foxp3, anti-CD25, anti-CD8, anti-CD3, and anti-CD45 were from BD Biosciences. Carboxy fluoroscein succinimidyl ester (CFSE) and Cell Trace Violet (CTV) were from BioLegend (San Diego, CA) and Invitrogen (Thermo Fisher Scientific, USA). LIVE/DEAD fixable viability kits were from Invitrogen. Foxp3/transcription factor staining buffer set (00–5523-00) was purchased from eBioscience. IL-1β (432604) and Cathepsin B (EA100429) enzyme-linked immunosorbent assay (ELISA) kits were from BioLegend (San Diego, CA) and Origene (Rockville, MD) respectively. Fetal bovine serum (FBS) was from Cytiva HyClone. Non-essential amino acids, sodium pyruvate, penicillin/streptomycin, DPBS without Ca²⁺ and Mg²⁺, cell stripper, trypsin and RPM1–1640 were purchased from Corning. GlutaMAX, ACK lysis buffer, β-mercaptoethanol and Type II Collagenase were from Gibco (Thermo Fisher Scientific, USA). LDL-cholesterol and HDL-cholesterol were measured using a cholesterol oxidase assay (Crystal Chem High Performance Assays, USA).

HepG2 cells were a gift from Sayeepriyadarshini Anakk (University of Illinois at Urbana-Champaign). EMT6-luc cells were a gift from Hasan Korkaya (Augusta University). E0771 and 4T1-luc cells were a gift from Mark Dewhirst (Duke University). RAW 264.7, Lewis lung carcinoma and B16 cells were purchased from American Type Culture Collection. Cell lines were not cultured longer than 2 months after thawing or after passage 20. Cell lines were routinely tested for mycoplasma.

4.1. Survival and correlational analysis of human tumors

Survival analysis in Fig. 1 was performed the Kaplan-Meier Plotter webtool (https://kmplot.com/analysis/) [37] and cBioPortal (https://www.cbioportal.org/) [38–40]. The Kaplan-Meier Plotter webtool uses aggregated data from GEO, EGA, and TCGA. Differentially upregulated genes between vehicle and DSHN treated BMDMs from RNA-seq analysis (fold change threshold of 2 fold, and FDR<0.01), were used to create a non-weighted signature of 22 genes. This signature was then used to probe the METABRIC dataset as obtained through cBioPortal. Tumors were then parsed into upper and lower quartiles and used to examine survival data using Kaplan-Meier analysis, similar to Ref. [41].

4.2. Animal tissue and in vivo studies

All protocols involving animals were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Illinois Urbana-Champaign. Female wild-type C57BL/6 and BALB/C mice were purchased from Charles River Laboratory. Mice were 8–12 weeks old at the start of the experiment. Founder OT-II mice and MMTV-PyMT were purchased from Jackson Laboratory and bred in-house. Founder NR0B2^fl/fl mice were a kind gift from John Auwerx and Kristina Shoonjans (Ecole Polytechnique de Lausanne). The mice were matched for age in each experiment. Bioluminescent, in vivo imaging was performed on a PerkinElmer IVIS Spectrum CT.

4.3. Preparation of bone marrow-derived macrophages (BMDMs) and splenic CD11C⁺ cells

Bone marrow cells were collected from mouse tibia and femur. Cells were passed through 70-μm cell strainer and subsequently cultured 10 mL complete RPMI media supplemented with 20 ng/mL of recombinant murine M-CSF (576406; BioLegend, 315–02 PeproTech). On day 3, additional 5 ml of the complete media was added. On day 7, media was replaced. BMDMs were harvested on day 10 using cell stripper solution for further experiments. Dendritic cells were isolated from the spleens of wildtype, using mouse CD11c UltraPure MicroBeads (130–125-835; Miltenyi Biotec) according to the manufacturer’s instructions) and cultured in RPMI medium supplemented with 10 % heat-inactivated charcoal-stripped FBS, 50 μM β-mercaptoethanol, 1 % sodium pyruvate, 1 % nonessential amino acids, % penicillin/streptomycin, and 1 % Glutamax.

4.4. Cell isolation from tissues

Fresh tumors, and lungs were collected separately from mice, and digested in DMEM/F12 supplemented with 2 mg/ml type II collagenase and 1 % penicillin/streptomycin for 45 min at 37 °C while shaking. Subsequently, cells were passed through a 70-μm filter into single-cell suspension, washed with FACS buffer, incubated with ACK lysis buffer for 1 min, and washed with FACS buffer before antibody staining for flow cytometry. Single cell suspensions of spleens were obtained by mechanical dissociation through a 70-μm filter in isolation buffer (DPBS supplemented with 0.5 % BSA and 2 mM EDTA). Subsequently, cells were washed with isolation buffer, incubated with ACK lysis buffer, and washed with isolation buffer before antibody staining for flow cytometry or immune cell isolation. Cd11b+ (130–126-725) and Ly6G+ (130–120-337) were isolated from single cell suspensions using UltraPure MicroBeads from Miltenyi B.

4.5. RNA extraction, reverse transcription and quantitative real-time PCR

RNA was extracted using columns or the TRIzol method as previously described [42,43]. cDNA was generated through reverse transcription using random hexamers as previously described [6,44]. Quantitative, real time PCR using the relative quantitation method was performed as previously described [41,44,45]. A list of primers is included in Supplemental Table 1.

4.6. In vitro T_reg expansion assays

Assays were performed similar to previously described [13,14]. T cells were cultured in RPMI medium supplemented with 10 % heat-inactivated charcoal-stripped FBS, 50 μM β-mercaptoethanol, 1 % sodium pyruvate, 1 % nonessential amino acids, 1 % penicillin/streptomycin, and 1 % Glutamax. Naïve CD4⁺ T cells were isolated from the spleens of wildtype or OT-II mice, using Naive CD4⁺ T Cell Isolation Kit (130–104-453; Miltenyi Biotec) according to the manufacturer’s instructions. T cells were labeled with the vital dye 2.5 μM CFSE or 5 μM Cell Trace Violet according to the manufacturer’s instructions. For wildtype T_reg expansion in presence of antigen presenting cells (APCs), naïve CD4⁺ T cells were co-cultured with APCs at indicated ratio in the presence of 0.5 μg/mL (anti-CD3 BioLegend), 0.5 ng/mL TGFβ (BioLegend) and 0.5 ng/mL IL-2 (BioLegend) for 72 h at 37 °C. For OT-II Treg expansion, OTII naïve CD4⁺ T cells from OT-II mice were activated by antigen presentation by culturing with BMDMs primed with 10 μg/mL OVA_323–339 (Bachem) and 0.5 μg/mL LPS (Sigma) or dendritic cells primed with 10 μg/mL OVA_323–339. To deplete IL1β (clone B122; BE0246; BioXCell) and IL18 (clone YIGIF74–1G7; BE0237; BioXCell), isotype control or respective neutralizing antibodies were added at 5 μg/mL to T cell cocultures. The proliferation of T cells was assessed by flow cytometry 72h after co-culture at indicated ratio.

4.7. In vitro T cell suppression assays

BMDMs were cocultured with naïve CD4⁺ T cells and differentiated into T_regs in the presence 0.5 μg/mL (anti-CD3 BioLegend), 1 ng/mL TGFβ (BioLegend) and 1 ng/mL IL-2 (BioLegend) for 72 h. Next, expanded CD4⁺ T cells were cocultured at different ratios with CFSE/CTV labeled CD8⁺ T cells and stimulated with Mouse T-Activator CD3/CD28 Dynabeads at 2 cells per bead ratio and 0.5 ng/mL IL-2 (BioLegend). Cell culture media consisted with RPMI 1640, supplemented with 10 % heat-inactivated charcoal-stripped FBS, 50 μM β-mercaptoethanol, 1 % sodium pyruvate, 1 % nonessential amino acids, 1 % penicillin/streptomycin, and 1 % Glutamax. CD8⁺ T cells were isolated from wildtype mice spleens using CD8a + T Cell Isolation Kit (130–104-075) from Miltenyi Biotec according to the manufacturer’s instructions. Flow cytometric analysis of suppressive cultures was preformed after 72 h. Beads were removed using DynaMag magnet (Invitrogen) prior to flow cytometry staining and analysis.

4.8. Flow cytometry staining and analysis

For surface staining, cells were stained with LIVE/DEAD fixable viability dye diluted in PBS (1:1000) for 30 min at 4 °C in the dark, washed and subsequently stained with fluorochrome-conjugated antibodies for cell surface antigens in FACS buffer (DPBS supplemented with 2 % FBS and 1 % penicillin/streptomycin) at 1:100 for 30 min at 4 °C in the dark. Next, cells were fixed with 4 % Formalin and incubated at 4 °C in the dark until analysis. For intracellular staining, cells were stained with intracellular cytokines according to the manufacturer’s protocol using the ebioscience Foxp3/transcription factor staining buffer set. In brief, after surface staining, live cells were fixed and permeabilized in 1 × fix/perm buffer for 30 min at room temperature or overnight at 4 °C in the dark. Fluorochrome antibodies against intracellular antigens were diluted in 1 × permeabilization buffer at 1:50 and stained for 30 min at room temperature. Samples were washed serially with 1 × permeabilization buffer and FACS buffer and resuspended in FACS buffer for analysis. Cytometry data were acquired on either BD LSRII, BD Fortessa, BD Symphony A1, BD Accuri C6 and Thermo Attune NxT.

4.9. Gene silencing and overexpression

For gene knockdown small interfering RNA was delivered using HiPerFect transfection reagent (Qiagen). BMDMs were transfected with either negative control or Caspase-1/IL-1β-complementary siRNA (SMARTpool, ON-TARGETplus siRNA, Dharmacon) at 50 nM. NR0B2 siRNA was from Sigma. Expression of the target genes after the knockdown was assessed using qPCR. For NR0B2 overexpression, NR0B2 overexpressing plasmid was delivered using Effectine transfection reagent (Qiagen). All transfections were conducted in accordance with the manufacturer’s protocol. Media was changed 12 h post-transfection and cells were harvested for subsequent experiments 48 h post transfection.

4.10. RNA-sequencing and analysis

Bulk RNA-sequencing was performed on an SP lane with 1×100nt reads and produced over 480 million reads. Data has been deposited into GSE241918 (specifically GSE241916. After sequencing, raw reads were mapped to GRCm39 reference genome with STAR (2.7.6a). featureCounts was used to quantify the number of reads assigned to a feature for each sample. Limma-trend (limma 3.60.0) approach was used for normalization on the read counts and testing for differentially expressed genes (DEGs). Normalized data were clustered on a multidimensional scaling (MDS) plot, and the DEG list was used to plot a heatmap using heatmap.2 package.

Weighted gene co-expression network analysis (WGCNA 1.72–5) is a data mining method developed to discover the co-expressed gene clusters (modules) and detect the core genes (hub genes) in each module. To perform WGCNA, we re-fit the STAR output to glmQLFit model, and filtered out DEGs with FDR >0.4. The soft thresholding power β was set to 9 to ensure a correlation coefficient close to 0.8. 15 modules were detected. Module eigengenes (ME) were calculated and visualized to represent the gene expression signatures of each

4.11. Screening to identify NR0B2 agonists

HepG2 cells were co-transfected with ABCA1 renilla reporter plasmid and TK-control firefly luciferase vector at 20:1 using Lipofectamine 3000 reagent according to manufacturer’s instructions for 24hrs. Afterwards, 7000 cells were seeded in to 384 well f-bottom, white plates (Greiner Bio-One). Test compounds were added to the plates in a final working volume of 30 μL (for a final concentration of 50 μM) in presence of DMSO or 1 μM GW3965 and incubated for 20–24hrs at 37°C. Luciferase activity was measured using the Luc-Pair Duo-Luciferase HT Assay kit (GeneCopoeia, Rockville, MD, USA) by adding luciferase substrates sequentially following manufacturer instructions. Luminescence was measured with 2 s integration times in a microplate reader (BioTek Cytation 5), 15 min after adding each substrate. The addition of reagents to the microplates was timed at each step to match the reading time delays and reading sequence of the microplate reader.

4.12. Resazurin cell viability assay

20,000 cells were seeded into 96 well f-bottom, black plates (Corning Costar) and cell viability was measured using 100 μL resazurin per well at 0.3 mg/mL (Acros Organics). The fluorescence signal was quantified by plate reader (560 nm excitation/590 nm emission) at 2–3 h.

4.13. Compound uptake assay

Cellular uptake of DSHN and DSHN-OMe was determined in RAW 264.7 cells cultured in 60 × 15mm tissue culture dishes. Cells were treated with vehicle (0.1 % DMSO), 50 μM DSHN or 50 μM DSHN-OMe for 4 h, 8 h, or 24 h. After incubation, cells were harvested, washed with PBS and cell pellets were incubated at −80 °C freezer. Next, cell pellets were resuspended in 200 μL 70:30 MeOH:H₂0 and sonicated to lyse cells. Debris from lysed cell suspension was removed by centrifugation. Resulting supernatant was analyzed by LC-MS/MS.

4.14. Protein expression, purification, and thermal shift assay

A gene containing a hexa-His-TEV fusion with full length NR0B2 in pET21(a)+ was transformed in E.coli BL21(DE3). A single colony was used to inoculate a 100 mL starter culture of LB broth containing 100 μg/mL ampicillin, which was allowed to grow overnight with shaking at 37°C. This starter culture was used to inoculate flasks containing autoinducing media, which were grown at 37°C with shaking until they reached an OD⁶⁰⁰ of 0.8. Subsequently, the temperature was reduced to 16°C and cells were allowed to grow for another 24 h. After harvesting by centrifugation, cells were resuspended at 15 mL/g cell paste in a buffer comprised of 25 mM HEPES pH 8.0, 250 mM NaCl, 20 mM imidazole pH 8.0, 5 % glycerol, 0.5 mM TCEP, and Roche EDTA-free protease inhibitor cocktail. Cells were lysed by sonication then centrifuged at 18,000×g for 30 min to remove insoluble material. The lysate was loaded onto a gravity flow Ni-NTA column, washed with 5 column volumes (CVs) of resuspension buffer, and protein was eluted using the resuspension buffer supplemented with 500 mM imidazole. Collected protein was dialyzed overnight in the resuspension buffer without imidazole then concentrated and purified on a Superdex 200 HiLoad 200 16/600 size exclusion column. A peak corresponding to ~29 kDa was collected and molecular weight was verified using SDS-PAGE. For thermal shift assays, vehicle (DMSO), 1 mM DSHN-OME, or 1 mM DSHN was incubated with 6 μM purified NR0B2 overnight at 4°C. We found that this concentration was the minimal required to give a reliable melt curve. The next morning the mixtures were centrifuged at 20,000×g for 30 min to remove any insoluble protein or ligand. The supernatant was mixed with SYPRO orange (Thermo) then placed in 96-well qPCR plates in triplicate. Melt curves were obtained with a Applied Biosystems qPCR machine between 25 and 95°C at a gradient of 0.15 °C/s. Melt curve data were fit using Thermo Protein Thermal Shift Software. These experiments were performed three independent times with three technical replicates each.

4.14.1. Human tumor specimens

Deidentified serial breast cancer sections and corresponding subtype information were obtained from archival collections at the Département de Biologie et de Pathologie des Tumeurs, Centre Georges-François Leclerc, Dijon, France. They were stained and analyzed by the Research Histology and Tissue Imaging Core, University of Illinois at Chicago, Illinois, USA. Three consecutive sections from each sample were stained with dual immunostain for FoxP3 and panCK, RNAscope assay with NR0B2 probe, and dual immunostain for CD8 and panCK. Staining for all targets was performed on BOND RX autostainer (Leica Biosystems, Deer Park, IL) using preset protocols. For dual immunostaining, the first and third slides were stained with either FoxP3 (1:100, #12653, Cell Signaling Technology, Danvers, MA) or CD8 (1:100, #ACI3160, Biocare Medical, Pacheco, CA) antibodies using BOND Polymer Refine Detection kit (#DS9800), followed by staining with panCK antibody (1:4000, #M351501, Agilent, Santa Clara, CA) using BOND Polymer Red Detection Kit (#DS9390). The middle slide was stained with RNAscope 2.5 LS Probe - Hs-NR0B2 assay (#877508, Advanced Cell Diagnostics, Newark, CA) using RNAscope 2.5 Leica assay-RED reagents for hybridization and detection. The standard ACD Red Rev BOND Rx protocol was modified to include BOND Epitope Retrieval Solution 2 (Leica Biosystems) pretreatment for 25min at 95 °C and to increase Amp5 step time to 30 min. Positive (Hs-PPIB, #313908) and negative (DapB, #3120) probe assays were included with each batch of samples. The stained slides were scanned at 40X magnification on PhenoIMager HT (Akoya Biosciences, Malborough, MA). All three slides were registered using HALO software (Indicalabs, Albuquergue, NM). HALO algorithms were used to segment tumor from stroma and count positive T cells and NR0B2 transcripts, at single-cell resolution. We computed Spearman correlation coefficients to evaluate associations between CD8 and FoxP3 density and NR0B2 expression in the tumor and stromal compartments. These correlations were also computed within strata defined by ER/PR, HER2 and triple negative status. CD8, FoxP3 and NROB2 expression was compared among the three subtypes using pairwise two-sided multiple comparison-adjusted analysis. All P values were two-sided with a 0.05 threshold for statistical significance.

4.15. Cellular proliferation assays

Cell proliferation was assessed by staining for total DNA with Hoechst 33342 dye, similar to previously reported [41,44].

Abbreviations

α

anti

DSHN

a small molecule agonist of NR0B2

DSHN-OMe

a methyl ester of DSHN

FXR

farnesoid × receptor

GW3965

an LXR agonist

GW4064

an FXR agonist

LXR

liver × receptor

NR0B2

small heterodimer partner

TNBC

triple negative breast cancer

T_reg

regulatory T cell

Data and materials availability

All of the reagents developed during the course of this work will be made available to colleagues without restriction. Upon request we will, at no charge (with the exception of shipping and handling costs), send animals, reagents and protocols to interested investigators. Transcriptomic data will be submitted to public repositories (NCBI-GEO; pending). Should any property be patented, we will ensure that the technology remains widely available to the research community in accordance with University of Illinois policies and the NIH Principles and Guidelines document.