Loss of NR2E3 represses AHR by LSD1 reprogramming, is associated with poor prognosis in liver cancer

The aryl hydrocarbon receptor (AHR) plays crucial roles in inflammation, metabolic disorder, and cancer. However, the molecular mechanisms regulating AHR expression remain unknown. Here, we found that an orphan nuclear NR2E3 maintains AHR expression, and forms an active transcriptional complex with transcription factor Sp1 and coactivator GRIP1 in MCF-7 human breast and HepG2 liver cancer cell lines. NR2E3 loss promotes the recruitment of LSD1, a histone demethylase of histone 3 lysine 4 di-methylation (H3K4me2), to the AHR gene promoter region, resulting in repression of AHR expression. AHR expression and responsiveness along with H3K4me2 were significantly reduced in the livers of Nr2e3rd7 (Rd7) mice that express low NR2E3 relative to the livers of wild-type mice. SP2509, an LSD1 inhibitor, fully restored AHR expression and H3K4me2 levels in Rd7 mice. Lastly, we demonstrated that both AHR and NR2E3 are significantly associated with good clinical outcomes in liver cancer. Together, our results reveal a novel link between NR2E3, AHR, and liver cancer via LSD1-mediated H3K4me2 histone modification in liver cancer development.

Lysine-specific demethylase-1 (LSD1) is a flavin-dependent amine oxidase, which in general, functions as histone demethylase by removing the methyl group from mono-and dimethylated histone H3 at lysine 4 (H3K4), leading to suppression of the downstream target genes 19 . The genetic depletion of LSD1 in mice is known to cause embryonic lethality, and LSD1-depleted embryonic stem cells exhibit markedly reduced viability 20,21 , indicating its crucial role in cell functions and survival. Many studies have shown that LSD1 overexpression increases cancer cell proliferation, invasion, and metastasis [22][23][24][25][26][27][28] . Furthermore, LSD1 overexpression is strongly correlated with poor clinical outcomes in many cancers, including liver cancer [29][30][31] . The inhibition of LSD1 activity with a small chemical inhibitor markedly decreased aggressive cancer cell phenotype and stem cell features, as a result of which LSD1 inhibition has attracted considerable attention as a novel therapeutic strategy for cancer treatment [32][33][34][35] . Although the LSD1-dependent H3K4me2 status change plays an important role in normal cell physiology and cancer progression, the epigenetic factor that modulates the distribution and function of LSD1 in maintaining normal epigenome remains to be identified.
Here, we reveal that NR2E3 is a novel upstream regulator of AHR. The presence of NR2E3 facilitates the formation of a transcriptionally active complex with specificity protein 1 (Sp1) and glucocorticoid receptor-interacting protein1 (GRIP1) in the proximal promoter region of the AHR gene whereas NR2E3 depletion markedly decreased active dimethyl-histone H3 lysine 4 (H3K4me2) marks by enhancing recruitment of the LSD1-associated repressor complex. This event decreased AHR expression and responsiveness. We further demonstrated that higher expression of NR2E3 or AHR in liver cancer patients is strongly correlated with good clinical outcomes. These findings indicated that interaction between NR2E3 and LSD1 plays a critical role in maintaining the normal epigenome and gene expression and that disruption of this interaction is associated with increased susceptibility and progression of liver cancer development.

Results
Association of NR2E3 with AHR-related gene networks. To identify crucial biological gene networks associated with NR2E3, we first obtained the NR2E3-dependent gene signature by performing RNA-seq analysis using small hairpin control (shCT) and NR2E3-depleted (shNR2E3) human liver HepG2 cell RNA lysates. A total of 3973 genes were differentially expressed by NR2E3 depletion (2,006 upregulated and 1,967 downregulated genes, with an adjusted P value [FDR q-value] < 0.05). GSEA analysis was carried out based on a ranked gene list and the results revealed that NR2E3 gene networks were significantly associated with AHR-related gene networks, including drug metabolism (cytochrome p450) and metabolism of xenobiotics (cytochrome p450; Fig. 1a and Supplementary Tables 1 and 2). These results are in line with the previous result suggesting that NR2E3 gene networks were closely related to AHR signaling pathways (GSE18431) 6 . Together with the previously obtained differential gene expression data derived from the control and NR2E3 knockout MCF-7 cell lysates, we identified the common gene set regulated by NR2E3 across two different cell types: ESR1-positive MCF-7 cells and ESR1-negative HepG2 cells. As seen in the Venn diagram, there are 252 commonly regulated genes (Fig. 1b  and Supplementary Table 3) between the two groups. The heatmap indicated that AHR expression was commonly decreased in both NR2E3-depleted HepG2 and MCF-7 cell signatures (Fig. 1c). By using the common gene set, we also performed the WikiPathway analysis to identify the common biological pathways (Supplementary  Tables 4 and 5), and the results consistently indicated that the NR2E3 gene networks are commonly highly associated with AHR signaling pathways (Fig. 1d).

NR2E3 regulates AHR expression and responsiveness in vitro.
Based on the strong biological association between NR2E3 and AHR, we studied the effects of NR2E3 depletion on the AHR expression level and responses. We silenced NR2E3 expression using shRNAs targeting NR2E3 in both MCF-7 and HepG2 cells. NR2E3 depletion markedly decreased both the AHR protein and mRNA expression levels in both MCF-7 and HepG2 cells ( Fig. 2a and b). The results indicated that NR2E3 plays a role in maintaining AHR expression at the transcriptional level. Next, control (shCT) and NR2E3-depleted (shNR2E3) MCF-7 and HepG2 cells were treated with 2,3,7,8-tetrachloro-p-dibenzodioxin (TCDD), a typical ligand for AHR activation, and the induction level of CYP1A1, a major target gene induced by AHR activation, was then determined. The induction levels of CYP1A1 in NR2E3-depleted cells (shNR2E3 KO I & II) were greatly reduced compared to the levels in cells transfected with control scramble shRNA (shCT) (Fig. 2c). We then examined whether NR2E3 overexpression affected AHR expression levels. In this experiment, both MCF-7 and HepG2 cells were transiently transfected either with empty plasmid or NR2E3-expressing plasmid. NR2E3 overexpression significantly increased the AHR protein and mRNA levels ( Fig. 2d and e). Consistently, the activity of reporter luciferase linked to DRE was greatly enhanced by NR2E3 overexpression when the cells were treated with TCDD (Fig. 2f). Taken together, our findings showed that NR2E3 maintains and positively regulates AHR expression at the transcriptional level, consequentially modulating its responsiveness.
NR2E3 mediates transcriptionally active complex formation for AHR expression. We used the GAL4 reporter luciferase system to identify which coactivator plays a positive role in NR2E3-mediated AHR expression. Cells were transfected with plasmid containing GAL4 DNA-binding domain linked to various coactivators, including SRC1, GRIP1, SRC3, PGC1α, and DRIP205 (Trap220) and GAL4-reporter luciferase with or without NR2E3 overexpression. We found out that GAL4 luciferase activity was significantly enhanced only when cells were co-transfected with GAL4-GRIP1 and NR2E3 and not any other combinations (Fig. 3a), suggesting that NR2E3 specifically enhanced the coactivator function of GRIP1. To further determine whether NR2E3 forms protein complexes with GRIP1, co-immunoprecipitation assay was performed using HepG2 cell lysates. The reciprocal interactions between NR2E3 and GRIP1 were detected by co-immunoprecipitation assays and followed by immunoblotting with opposite set of antibodies (Fig. 3b top and bottom), demonstrating that endogenous NR2E3 formed the protein complex with GRIP1. We observed similar results when we used MCF-7 cell lysate ( Supplementary Fig. 1a). To confirm whether anti-NR2E3 antibody pulled down NR2E3 protein, Co-immunoprecipitation assays were performed using both MCF-7 and HepG2 cell lysates ( Supplementary  Fig. 1b). In line with this result, immunocytochemical results showed that NR2E3 (green) and GRIP1 (red) were co-localized as multiple speckles within the nucleus (yellow) (Fig. 3c). In order to further define the role of GRIP1 in AHR expression, we examined whether GRIP1 depletion could affect AHR expression. MCF-7 and HepG2 cells were transfected with two types of siRNAs targeting different regions of the GRIP1 mRNA sequence (siGRIP1 I and II) or with scrambled control siRNA (siCT). GRIP1 depletion significantly decreased the AHR protein and mRNA levels ( Fig. 3d and e), indicating that NR2E3/GRIP1 transcriptionally active protein complex formation is crucial for maintaining AHR expression. Several reports have previously demonstrated that specificity protein 1 (Sp1) transcription factor plays a positive role in maintaining AHR expression 36,37 . Therefore, we examined whether NR2E3 also forms a protein complex with the Sp1 protein. The results of our study revealed that NR2E3 also interacts and co-localizes with Sp1 ( Supplementary Fig. 2a and 2b). These results suggested that NR2E3, GRIP1, and Sp1 together form a protein complex.
We used ChIP-PCR to further verify whether NR2E3, GRIP1, and Sp1 form a complex on the AHR gene promoter region. We used a primer set covering the proximal region of the human AHR gene promoter (−283 to −90), which contains GC-rich Sp1-binding sites that was previously identified as a positive regulator for AHR expression 38,39 . In contrast, the primer set that detects the distal region of the AHR promoter (−1969 to −1779) Figure 1. Identification of AHR as a novel target of NR2E3 by RNA sequencing (RNA-seq) and bioinformatics analyses. (a) Gene set enrichment analysis by employing RNA-seq data derived from small hairpin RNA control (shCT) and NR2E3-depleted (shNR2E3) HepG2 cell RNA lysate showed that NR2E3 gene networks are associated with drug metabolism involving cytochrome p450 and metabolism of xenobiotics by cytochrome p450 signaling pathways. (b) Venn diagram revealed 252 differentially expressed genes (DEGs) commonly regulated by NR2E3 depletion in both MCF-7 and HepG2 cells (p < 0.0005). (c) Heat map of the 253 common genes indicates that AHR expression (red) is markedly downregulated in both MCF-7 and HepG2 cells. (d) WikiPathway analysis using the common gene set showed the association of NR2E3-regulated genes with AHRrelated signaling pathways (FDR < 0.05).
was used as a negative control (Fig. 3f, top). The results of the ChIP-PCR assays indicated that NR2E3 forms an active transcriptional complex with GRIP1 and Sp1 accompanied by RNA pol II recruitment (Fig. 3f, bottom) in the proximal region (Primer I) but not in the distal region (Primer II). Similar interactions were observed between NR2E3, Sp1, GRIP1, and RNA pol II in MCF-7 cells (Supplementary Fig. 2c). A Re-ChIP-PCR assay was performed to further confirm the NR2E3/GRIP1 or NR2E3/Sp1 protein complex formation. Formaldehyde-fixed chromatin was first immunoprecipitated with anti-NR2E3 antibodies followed by immunoprecipitation with anti-GRIP1 or anti-Sp1 antibodies. The corresponding immunoprecipitated lysate was used as a nonspecific control. The strong protein complex formed between NR2E3 and GRIP1 or NR2E3 and Sp1 in the proximal region of the AHR gene promoter region was then detected (Fig. 3g). Similarly, NR2E3/Sp1 and NR2E3/GRIP1 interactions were demonstrated by Re-ChIP-PCR assay using MCF-7 cell lysate ( Supplementary Fig. 2d). These results  . Formation of transcriptionally active complexes between NR2E3, Sp1, and GRIP1 on the proximal region of the AHR gene promoter. (a) Effects of NR2E3 overexpression on the GAL4 reporter luciferase activity in cells co-transfected with pM vector containing GAL4 DBD-linked coactivator, including SRC1, GRIP1 (SRC2), SRC3, PGC1α, and DRIP205 (TRAP220). (b) Co-IP assays confirmed the interactions between NR2E3 and GRIP1. HepG2 cells were lysed and Co-Immunoprecipitation (IP) was performed with anti-GRIP1 antibody or control non-immune IgG (IgG), followed by immunoblot analysis (IB) with NR2E3 antibody (Top). A reciprocal Co-IP assay with opposite set of antibodies was also performed (Bottom). (c) Co-localization of NR2E3 with GRIP1 in the nucleus of HepG2 cells. (d,e) Effect of GRIP1 depletion by using small interfering RNAs (siRNAs) targeting GRIP1 (siGRIP1 I and II) and scrambled control (siCT) on AHR and GRIP1 protein expression and mRNA levels in MCF-7 and HepG2 cells. (f) Binding of NR2E3, Sp1, and GRIP1 and RNA polymerase II on the AHR promoter regions, distal (−1969 to −1779) and proximal (−283 to −90) in HepG2 cells, as determined by ChIP-PCR. (g) Re-ChIP analysis of the complex formation between NR2E3 and Sp1 or NR2E3 and GRIP1 on the AHR proximal promoter region in HepG2 cells. (h,i) Effects of NR2E3 loss on the transcriptional and epigenetic status of the AHR proximal promoter. Binding of Sp1, GRIP1, RNA pol II, and H3K4me2 were reduced whereas demonstrated that NR2E3 is essential for the formation of active transcriptional complexes containing GRIP1 and Sp1 that are required in the maintenance of AHR expression.
Next, we investigated the effects of NR2E3 depletion on the epigenetic and transcriptional status of the AHR gene promoter, a ChIP-PCR assay was performed using NR2E3-depleted (shNR2E3 knockout I or II) and control HepG2 (shCT) cell lysates. NR2E3 depletion disrupted active protein complex formation of Sp1, GRIP1, and RNA pol II, and reduced their binding to this promoter region (−283 to −90; Fig. 3h). In contrast, binding of a corepressor SIN3A and a histone demethylase LSD1 to the proximal gene promoter of AHR was markedly increased, accompanied by reduced H3K4me2, which is an LSD1 substrate (Fig. 3i). Thus, NR2E3 loss not only disrupted the active transcriptional complex but also facilitated LSD1/Sin3A repressor complex recruitment. These results are in line with our previous report demonstrating that NR2E3 loss facilitates LSD1 recruitment and consequently decreased H3K4me2 on the proximal promoter region of the ESR1 7 .

Effects of Nr2e3 loss in vivo on Ahr expression and responsiveness.
To determine the role of NR2E3 in AHR status and AHR-mediated responsiveness in vivo, we employed Nr2e3 Rd7/Rd7 (retinal degeneration 7, Rd7) mutant mice (Jackson laboratory, Bar Harbor, MI) expressing non-detectable levels of Nr2e3 in the retina 40,41 . In Rd7 mice, the insertional mutation of LINE-1 transposon in the exon5 of Nr2e3 gene inhibits proper Nr2e3 mRNA processing and translation 41 . In fact, Rd7 mice have been primarily used to investigate the development of retinal diseases [1][2][3][4][40][41][42] . As the liver is a major organ with high AHR activity for detoxification and metabolic processes, we examined whether the Nr2e3 and Ahr expression levels were altered in the livers of Rd7 mice relative to the levels in the WT mice. We first identified that Nr2e3 protein and mRNAs were expressed in the liver of WT mice ( Fig. 4a and Supplementary Fig. 4c), unlike previous report 43 . Correspondingly, we observed significant decreases of both Nr2e3 and Ahr protein and mRNA levels in the liver of Rd7 mice (Fig. 4a, Supplementary  Figs 3a and 4a). In addition, with two more primer sets that amplify the region between exon 6 and exon 8 of mouse Nr2e3 gene (mouse Nr2e3 Exon 6-8 I & II), we further confirmed lower Nr2e3 mRNA expressions in the liver of Rd7 mice ( Supplementary Fig. 3a). In parallel, we investigated Nr2e3 protein expression levels in the liver and retina of Rd7 mice by performing additional immunoblotting assay with another Nr2e3 antibody (Abcam, Cat #172542). Results indicated that Nr2e3 protein levels were low in the liver of Rd7 mice (Supplementary Fig. 4a and 4c) but not detectable in the retina (Supplementary Fig. 4b and d). Consistent with these findings, immunostaining results revealed considerably lower Nr2e3 and Ahr levels in the livers of Rd7 mice (Fig. 4b). Intriguingly, we detected a much lower expression of Esr1, which is a previously identified target of Nr2e3 6, 7 , in Rd7 mouse livers, but no change in the LSD1 level (Supplementary Fig. 3b and c), suggesting that Nr2e3 plays a role in maintaining the Ahr and Esr1 expression levels. In addition, we examined whether other Ahr downstream target genes were downregulated in Rd7 mice, and found out that the expression level of Ppara, an Ahr target gene that is downregulated in the livers of Ahr knockout mice, was unchanged 44 (Supplementary Fig. 3d). Also, results from alanine aminotransferase (ALT) activity that typically detects liver injury showed no apparent liver damage in Rd7 mice ( Supplementary Fig. 3e). Nr2e3 likely regulates a different subset of gene networks that does not overlap with Ahr gene networks; moreover, Nr2e3 loss did not cause any spontaneous liver injury in vivo ( Supplementary  Fig. 3e).
To examine whether Nr2e3 loss could induce similar epigenetic and transcriptional status changes in the mouse Ahr gene promoter region, we performed in vivo ChIP assay using liver lysates of WT and Rd7 mice with the primer set covering the proximal promoter region (−138 to + 141). As expected, NR2E3 depletion resulted in significant disruption in the protein complex interactions between Sp1, GRIP1, and RNA pol II in this region (Fig. 4d) while Sin3A and LSD1 recruitment were enhanced, leading to decreased H3K4me2 levels (Fig. 4e). The patterns of epigenetic and transcriptional status changes induced by NR2E3 depletion were similar both in vivo and in vitro.
In order to determine whether ligand-activated Ahr responsiveness and downstream target gene expressions were altered by Nr2e3 depletion, both WT and Rd7 mice were treated with TCDD, and the induction levels of Cyp1a1 and Cyp1a2, which are the main target genes of ligand-activated Ahr, were then determined. The Cyp1a1 and Cyp1a2 mRNA levels were significantly lower in Rd7 mice than in WT mice (Fig. 4f). Consistently, significantly lower induction of Cyp1b1, another Ahr target gene, was observed ( Supplementary Fig. 5). Correspondingly, significantly lower 7-ethoxy-resorufin-O-deethylase (EROD) activity, which is an indicator of Cyp1a1 enzymatic activity, was detected in the livers of Rd7 mice relative to that in WT mice (Fig. 4f). These data demonstrated that Nr2e3 depletion in vivo reduced Ahr expression by inducing repressive epigenetic and transcriptional status of the Ahr gene promoter in an LSD1-dependent manner via H3K4me2 modification. Consequently, we observed markedly reduced induction levels of Cyp1a1 and Cyp1a2 mRNAs and reduced enzymatic activities of these two enzymes.

Restoration of AHR expression and H3K4me2 status by LSD1 inhibition in vivo. Our results
indicated that NR2E3 depletion in cells and mice facilitated the recruitment of LSD1, which greatly reduced the H3K4me2 histone marks in the AHR gene promoter (Figs 3 and 4). Therefore, to determine whether the histone demethylase activity of LSD1 plays a critical role in repressing AHR expression, we examined whether LSD1 inhibition could restore Ahr expression in vivo. We treated Rd7 mice with a chemical inhibitor of LSD1, SP2509 34,45 . This treatment (i.p., 15 mg/kg) restored Ahr protein and mRNA expression in the livers of Rd7 mice ( Fig. 5a and b); binding of the corepressors Sin3A and LSD1 was increased. Results are means ± SE for at least two or three independent experiments with three replicates per experiment, and significantly (P < 0.05) increased (*) or reduced (**) responses are indicated.
SCiENtifiC RepoRts | 7: 10662 | DOI:10.1038/s41598-017-11106-2 this result was verified by immunostaining analysis (Fig. 5c). Moreover, increased Esr1 but not LSD1 expression was observed in Rd7 mice treated with SP2509 ( Supplementary Fig. 6a and 6b). Both the Ahr and Esr1 genes were repressed by the LSD1 redistribution induced by Nr2e3 depletion (Figs 3I and 4e, Supplementary Fig. 3a-c).  However, SP2509 treatment did not result in any changes in the AHR mRNA and protein levels in WT mice ( Supplementary Fig. 7), suggesting that these genes are not normally repressed by LSD1.
A ChIP assay was performed to further validate the effects of SP2509 treatment on the epigenetic status change of the Ahr gene promoter. The active histone mark H3K4me2 and RNA pol II recruitment were increased while Sin3A binding was reduced with minimal change in the LSD1 binding status (Fig. 5d). Correspondingly, the overall H3K4me2 levels were significantly increased following SP2509 treatment; this increase was confirmed by measuring the H3K4me2 levels and by immunoblotting analysis (Fig. 5f). These results clearly demonstrated that the repression of Ahr expression is dependent on the histone demethylase of LSD1. We proposed a model to illustrate how NR2E3 loss alters the active epigenetic and transcriptional status of the AHR gene promoter to repressive state (Fig. 5g).
Clinical association of AHR and NR2E3 with liver cancer development. A previous study reported that the Ahr expression was decreased during liver carcinogenesis induced by diethylnitrosamine (DEN) in mice and that Ahr knockout mice exhibited enhanced development of DEN-induced liver tumor formation 15 . This finding indicates that Ahr may function as a tumor suppressor gene in liver cancer. However, the clinical association of AHR with human liver cancer has not yet been firmly established. Thus, we examined whether AHR levels are significantly correlated with prognosis in human liver cancer development. For this analysis, we employed a publicly available database (GEO in the National Center for Biotechnology Information) to retrieve gene expression data of liver cancer patients. The results of the Kaplan-Meir survival analysis using two different clinical datasets (GEO10143 and GEO10186) 46,47 showed that the overall survival (OS) was significantly better in patients with higher AHR expression than in patients with lower AHR expression ( Fig. 6a and b, top), indicating that patients expressing high levels of AHR had good clinical outcomes. Correspondingly, the hazard ratios from each analysis result were also low ( Fig. 6a and b, bottom).
We next investigated whether NR2E3 levels are altered in precancerous liver diseases such as cirrhosis and liver tumor by comparing the protein levels between diseased and normal liver tissues. Immunohistochemical staining of NR2E3 was carried out using tissue microarray analysis of human liver tissue. The NR2E3 expression levels were low in cirrhotic tissue and even lower in liver tumor tissue (Fig. 6c). Figure 6d and Supplementary  Fig. 6 contain representative images showing the immunohistochemical staining of NR2E3 in normal, cirrhotic, and liver tumor tissues ( Fig. 6d and Supplementary Fig. 8). Of the 24 cirrhotic liver tissues, 10 showed moderate (33%, N = 14) and low (8%, N = 2) levels of NR2E3 staining (Fig. 6, left). Furthermore, liver tumor tissue samples that were immunostained with NR2E3 antibody exhibited high (36%), moderate (14%, N = 7) low (26%) and non-detectable (24%) staining levels (Fig. 6 right). Supplementary Tables 6-8 present the clinical information of the tissue samples analyzed (Supplementary Tables 6-8). Consistently, the results obtained from a publically available database of cancer expression profiles (ww.oncomine.org) showed decreased NR2E3 mRNA expression in human cirrhotic and hepatocellular carcinoma relative to normal liver tissues ( Supplementary Fig. 9).
By further employing the other independent human liver tumor gene expression data from The Cancer Genome Atlas (TCGA) which contains patient clinical information, we determined the clinical significance of NR2E3 in terms of the OS of liver cancer patients by performing Kaplan-Meier survival analysis. The results consistently indicated that higher NR2E3 levels in the patients were significantly associated with good clinical outcomes (Fig. 6e). Collectively, these results showed for the first time that both AHR and NR2E3 are good prognostic indicators and may function as tumor suppressor genes during the development of liver cancer.

Discussion
AHR plays important roles in a wide range of cellular events to maintain cellular homeostasis, including cell growth, differentiation, transformation, and death 11,12 . Thus, proper control of AHR levels in target tissues in response to various endogenous signals and xenobiotic chemical exposures is vital. However, our understanding of the underlying mechanism that regulates and maintains AHR expression remains largely unclear. Based on our initial indication that NR2E3-mediated signaling pathways were likely associated with AHR, our study is the first to discover a novel connection between NR2E3 and AHR and establish that NR2E3 is a novel upstream regulator of AHR. Our data demonstrated that NR2E3 specifically binds to the proximal promoter region of the AHR gene and thereby, acts as a basis for the formation of transcriptional activator complex. Several reports have shown that the AHR gene promoter lack TATA boxes and instead contains GC-rich sites for Sp1 transcription factor binding 38,39 . By using the binding of Sp1 to the proximal region of the AHR gene as a positive control, we confirmed that NR2E3 forms a protein complex with Sp1 and a coactivator GRIP1, leading to the formation of a transcriptionally active NR2E3/GRIP1/Sp1 complex with RNA pol II ( Fig. 3f and g). Consistently, the depletion of NR2E3 or GRIP1 using siRNA targeting NR2E3 or GRIP1 markedly decreased the AHR expression in both MCF-7 and HepG2 cells (Figs 2a,b and 3d,e), demonstrating that NR2E3 and GRIP1 are essential components for maintaining AHR expression. Correspondingly, the results of the ChIP assay indicated that depletion of NR2E3 disrupted the transcriptionally active complex consisting of NR2E3, GRIP1, and Sp1, leading to AHR repression ( Fig. 3h and i). In NR2E3-depleted cells, the induction level of CYP1A1, a key downstream target of ligand-activated AHR, was also markedly reduced (Fig. 2c). Taken together, these results demonstrated that NR2E3 status is a key player for maintaining AHR expression and responsiveness.
The altered epigenetic status of the AHR gene promoter region was further analyzed in vivo using Rd7 mice. Rd7 mice have been primarily used to investigate its role in the development of retinal diseases but not to study other tissues and diseases 1-4, 40, 41 . Interestingly, we found out that Rd7 mice showed a partial phenotype; Nr2e3 proteins were not expressed in the retina but low in the liver of Rd7 mice ( Fig. 4 and Supplementary Fig. 4C and D) and this was possibly due to tissue-specific differences in LINE-1 activity, splicing variation or low mRNA stability, etc 44,45,48 . Nonetheless, our results demonstrated that NR2E3 depletion in vivo induced low Ahr (Fig. 4a-c), and Esr1 (mouse estrogen receptor α, another NR2E3 target 6, 7 ) expressions in the livers of Rd7 mice, suggesting its positive regulative role in Ahr and Esr1 expressions ( Supplementary Fig. 2a,b, and Fig. 4). However, the expression of Pparα, which is a downstream target gene of AHR, remained unchanged between Rd7 and WT mouse livers 49 (Supplementary Fig. 2c), indicating that NR2E3 likely regulates different sets of genes independent of AHR. We also determined whether NR2E3 depletion caused any liver damage in Rd7 mice, but no damage was observed ( Supplementary Fig. 3e). Nr2e3 depletion in the livers of Rd7 mice altered the epigenetic status of the AHR gene promoter, which was similar to the changes detected at the cellular level. Together, these findings indicated that NR2E3 depletion decreased AHR expression by disrupting the active NR2E3/GRIP/Sp1 complex (Fig. 4d). Correspondingly, the induction levels and activities of CYP1A1 and CYP1A2 detoxification enzymes in the liver were significantly reduced by TCDD exposure in Rd7 mice relative to that in WT mice (Fig. 4f). Taken together, these results indicate that NR2E3 modulates the expression levels of AHR and ESR1, which are crucial molecular effectors that function in maintaining normal liver physiology, in response to environmental exposure and various endogenous signals.
LSD1 is an essential epigenetic regulator that can demethylate H3K4me2, which is related to active transcription status. This histone demethylation results in the repression of target genes 20,21 . In our previous study, we demonstrated that NR2E3 depletion reduced H3K4me2 levels by redirecting LSD1 to the proximal promoter region of the ER gene, turning off ER gene transcription 7 . We observed a similar event where NR2E3 depletion induced repressive epigenetic status by facilitating recruitment of the corepressors Sin3A and LSD1 to the proximal promoter region of the AHR gene, accompanied by reduced H3K4me2 levels and disruption of active transcriptional complex formation (Fig. 3h and i). To further determine whether Ahr repression induced by Nr2e3 depletion in vivo is dependent on the histone demethylase activity of LSD1, we administered SP2509, a chemical inhibitor of LSD1, into Rd7 mice 34,50 . Surprisingly, this treatment fully restored Ahr and Esr1 expression (Fig. 5a-c and Supplementary Fig. 4a and b), suggesting that the repression of these NR2E3 target genes is dependent on histone demethylase activity of LSD1. It has become increasingly evident that LSD1 is highly oncogenic correlated with the progression of various human cancers. Indeed, the inhibition of LSD1 enzymatic activity impaired cancer cell growth, metastasis, and even cancer stem cell properties, as a result of which LSD1 is an attractive molecular target for treating various types of cancers [32][33][34][35] . Intriguingly, a significant decrease in the H3K4me2 level was observed during human liver cancer progression whereas substantially increased expression of LSD1 was previously reported; patients with higher expression of LSD1 or lower H3k4me2 marks exhibited poor clinical outcomes [28][29][30][31][32] . These data are highly indicative of the likely involvement of NR2E3 depletion or loss in the increased susceptibility of liver cancer development, which occurs partly via regulating LSD1 distribution and activity. Our data suggested that NR2E3 loss caused, which would probably promote susceptibility to and progression of liver cancer, in part, via modulating LSD1 redistribution and activity.
A previous study on Ahr knockout mice revealed significant increase in DEN-induced liver tumor formation in the knockout mice than in WT mice however without DEN exposure Ahr knockout mice did not exhibit any sign of damage or form liver tumor spontaneously 17 . This result suggested that AHR may exert tumor suppressive activity independent of its ligand-activated function. Interestingly, the tumor suppressive role of AHR has been reported in several cancers. AHR loss increased intestinal carcinogenesis in mice containing adenomatous polyposis coli gene mutation 18 . In the development of prostate cancer in transgenic adenocarcinoma of the mouse prostate (TRAMP) mice, AHR loss enhanced prostate carcinogenesis 20 . Furthermore, the good prognostic activity of AHR has been previously reported in breast and pancreatic cancers [51][52][53] . In addition to these, the results from our survival analysis using two independent clinical datasets clearly demonstrated that high AHR expression levels were strongly associated with good clinical outcomes in patients with liver cancer (Fig. 6a). Given that AHR can function as either tumor suppressor gene or oncogene 15,16 , it is still possible that AHR may act selectively according to ligand type, etiology, subtype, and stage of liver cancer. We also observed that the NR2E3 proteins levels were decreased in some cirrhotic tissues and more reduced in liver tumors (Fig. 6c). Correspondingly, liver cancer patients who expressed high levels NR2E3 are significantly correlated with good prognosis (Fig. 6e). Overall, our results provide evidence that disruption of the NR2E3-LSD1-AHR signaling axis may play an important role in the development of precancerous liver diseases and liver cancer. The tumor suppressive role of NR2E3 during liver cancer development is currently under investigation.
Interestingly, several recent reports have demonstrated that loss of AHR caused development of retinal degenerative diseases similar to the retinal diseases caused by NR2E3 depletion. AHR deficiency in mice has been shown to cause age-related macular degeneration with atrophy in the retinal pigment epithelium 54, 55 , implicating potential crosstalk between the NR2E3 and AHR gene networks, although this possibility needs more investigation.
In summary, our work revealed the novel role of NR2E3 as a positive upstream transcriptional regulator of AHR. Loss of NR2E3 caused repression of AHR by epigenetic reprogramming, which altered the active H3K4me2 status by modulating LSD1 distribution and activity. Furthermore, our results revealed that the NR2E3 and AHR levels are significantly associated with good clinical outcomes in terms of survival in patients with liver cancer and illustrated that Rd7 mice could be used as a novel animal model for investigating the underlying molecular mechanisms associated with development of precancerous liver diseases and liver cancer. Furthermore, our results strongly suggested that decrease or loss of NR2E3 could trigger epigenetic reprogramming promoting susceptibility to and development of liver cancer such that inhibition of LSD1 or modulation of NR2E3 function can be a novel prevention or therapeutic strategy in liver cancer development, in part, by regulating expression of tumor suppressive genes. Further studies are required to examine these possibilities in different experimental settings.

Methods
Cell lines, reagents, plasmids, and tissue microarrays. HepG2 and MCF-7 cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA) and were not further tested or authenticated by the authors. These cell lines were maintained at 37 °C in the presence of 5% CO 2 in minimal essential medium-α (MEMα) or RPMI medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin solution (Sigma Aldrich, St Louis). The cells were seeded at a concentration of 2.5 × 10 4 per well, and the FBS content of the medium was reduced to 5%. Antibodies against β-actin, ER alpha (Cat #: sc-543), AHR (cat#: sc-5579), SRC2 (Cat#: sc-135931), and RNA pol II (Cat#: sc-9001) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). NR2E3 antibodies were obtained from Aviva Systems Biology (Cat#: ARP39069 and ARP39070), Santa Cruz Biotechnology (Cat #: sc-374513 and sc-292264), Proteintech (Cat#: 14246-1-AP) and Abcam (Cat #: ab172542). Sin3A (Cat #: 8056) was purchased from Cell Signaling (Danvers, MA). H4Ac (Cat #: 39925), Sp1 (Cat #: 39058), LSD1 (Cat #: 39186) and H3K4me2 (Cat #: 39679) were procured from Active Motif (Carlsbad, CA). 2, 3, 7, 8-Tetrachlorodibenzo-p-dioxin (TCDD) was dissolved in DMSO, and the stock solutions were directly added to the culture media. Control cells were treated with DMSO alone. Small interfering RNAs (siRNAs) targeting GRIP1 (siRNA ID: ASI_Hs02_00341797 and SASI_Hs01_00110929) and negative control siRNAs were purchased from Sigma Aldrich. Small hairpin RNAs targeting NR2E3 were previously described 6,7 . SP2509 was purchased from Cayman Chemical (Ann Arbor, MI). Tissue microarrays for NR2E3 staining of liver tumor and cirrhotic tissues (NBP2-30221 and NBP2-30276) were purchased from Novus Biologicals, LLC (Littleton, CO). For Kaplan-Meier survival analysis of NR2E3, liver cancer tissue array (Hliv-HCC180Sur-03) was obtained from US Biomax, Inc. (Rockville, MD). GAL4 DNA-binding domain constructs linked to various coactivators, including GAL4-SRC1, GAL4-SRC2 (Grip1), GAL4-SRC3, GAL4-PGC-1α and GAL4-Drip205 (Trap220) and a reporter luciferase construct linked to consensus dioxin response element (DRE) were previously used and verified 56 . NR2E3 expression plasmid (pcDNA4-NR2E3) was also used previously 6, 7 . Animals and treatment. C57BL/6 J, wild-type (WT), and homozygous Nr2e3rd7 (Rd7) mice (6-8 weeks old) were obtained from Jackson Laboratory (Bar Harbor, ME). All animal experiments were carried out in compliance with the guidelines established by the Animal Care Committee of University of Cincinnati. The animals were acclimated to temperature-and humidity-controlled rooms with a 12-h light/ dark cycle for 1 week prior to use. The mice had access to laboratory chow and tap water ad libitum. The mice were allocated for treatment with olive oil (n = 4) or TCDD (1 and 10 µg/kg; n = 4) dissolved in olive oil. For TCDD treatment, mice were intraperitoneally (i.p.) injected with 1 or 10 µg/kg TCDD dissolved in olive oil (2%, v/v). Twenty-four hours after administration, the animals were anesthetized with CO2 and liver tissues were obtained and frozen immediately in liquid nitrogen. Then, the mice were injected i.p. with the LSD1 inhibitor SP2509 (25 mg/kg) twice at 24-h intervals and then after 48 h from the last injection. Thereafter, the mice were sacrificed and their liver tissues were collected. All the animal experiments were approved by the Institutional Animal Care and Use Committee Real-time quantitative RT-qPCR. Total RNA was extracted from cells or liver tissue, and the cDNAs were amplified for quantitative real-time PCR, which was performed as previously described 6,7 . The obtained data were normalized to GAPDH or β-actin. The primer set sequences used: human AHR Luciferase assay. For one hybrid assay, GAL4 DNA-binding domain constructs linked to various coactivators, including GAL4-SRC1, GAL4-SRC2 (Grip1), GAL4-SRC3, GAL4-PGC-1α and GAL4-Drip205 (Trap220), were used 56 . Briefly, cells were transfected using Lipofectamine 2000 (Invitrogen) with the appropriate GAL4 plasmid and GAL4 reporter luciferase. The next day, cells were lysed and luciferase assays were performed using a Dual-Luciferase ® Reporter Assay System (Promega, Madison, WI). The luciferase activity was normalized to that of renilla luciferase activity as previously described 9 . For measuring effects of NR2E3 expression on the transcriptional activation function of AHR, a reporter luciferase construct linked to consensus dioxin response element (DRE) was employed 56 . Cells were co-transfected either with equal amount of empty plasmid pcDNA4.0 (empty) as negative control or NR2E3 expression plasmid (pcDNA4-NR2E3) 6,7 and DRE reporter luciferase, and cells were treated with TCDD on the next day. After 16 h, the cells were lysed and luciferase assays were performed as previously descirbed 6, 7, 56 .
Histone extraction and quantification of the H3K4me2 histone modification level. For measuring the global H3K4 di-methylation levels, the EpiQuik ™ Global Di-Methyl Histone H3-K4 Quantification Kit (EpiGenTek, Farmingdale, NY) was employed. The histone extraction and colorimetric assay for measuring global H3K4me2 levels were both carried out according to the manufacturer's protocol.
Immunohistochemical staining. After the sections were deparaffinized in xylene and rehydrated through a series of graded ethanol solutions, antigenic retrieval was performed by immersing the sections in 0.01 mM sodium citrate (pH 6.0) and heating in an microwave oven (100 °C) for 9 min. Deparaffinized sections were incubated with peroxidase-blocking reagent (Biogenex, CA, USA) for 9 min in a humidified chamber to block endogenous peroxidase activity. After blocking nonspecific binding sites with nonspecific staining blocking reagent (Vector Laboratories, Burlingame, CA, USA) for 30 min, the sections were incubated with primary antibodies at 4 °C overnight. Subsequently, peroxidase-conjugated secondary antibodies (Vector Laboratories) and 3,3-diaminobenzine-tetrachloride (DAB; Vector Laboratories) were used according to the manufacturer's instructions. The sections were counterstained with hematoxylin and observed under a microscope. Differential gene expression. Differential gene expression between NR2E3 knockout and control MCF7 cells: Bioconductor/R limma package 58 was used to predict differentially expressed genes (DEGs) between the NR2E3 knockout (n = 3) and control (n = 3) groups from the previously published MCF7 dataset (GSE18431). Genes were considered differentially expressed when they passed the cutoff criterion of p < 0.0005. Differential gene expression between NR2E3 knockout and control HepG2 cells: To identify the set of DEGs between NR2E3 knockout Hep2G cells (n = 2) and control HepG2 cells (n = 2), we used single-end RNA-seq data. These RNA-seq data were deposited in the Gene Expression Omnibus (GEO) website (GSE79463). Briefly, single-end reads were aligned to the mm10 genome from UCSC (downloaded from Illumina's iGenomes repository, https://support.illumina.com/sequencing/sequencing_software/ igenome.html) using TopHat 59 , then read counts were computed using FeatureCounts 60 . Differential gene expression analyses were carried out using the DESeq. 2 method 61 . Genes were considered differentially expressed when they passed the cutoff criterion of p < 0.0005. These RNA-seq data were deposited in the Gene Expression Omnibus (GEO) website (GSE79463).
Common gene sets. A total of 252 common genes were collected by intersecting two DEG sets described above (microarray and RNA-seq). Heatmaps were generated using the R/pheatmap package. The log2 expression values of these common genes were subtracted by the average log2-expression values of the control to visualize log2-fold changes in heatmaps. Genes were clustered using the hierarchical clustering method. Figure 1b presents two clusters, including AHR.
Gene set enrichment analysis. We used WebGestalt41 for gene set enrichment analysis for 252 common genes 62 . Differentially expressed gene sets were queried against the WikiPathway gene sets and an FDR cutoff of <0.05 was applied to select significantly enriched gene sets/pathways. Figure 1c presents the gene sets associated with AHR signaling.
Pre-ranked GSEA analysis. GSEA analysis 63 was performed using a pre-ranked option (all genes) because conventional GSEA is not optimized for RNA-seq data. The ranked list was generated according to the inverse of P values with +/− (i.e. up/down) signs and classic weighting method was applied 64 . Gene sets defined in the KEGG database were used for reference.
Determination of changes in the NR2E3 mRNA level and prognostic activity using human clinical liver datasets. We mined data using Oncomine (https://www.oncomine.org/). Two datasets (GSE14520 and GSE6764) were used for the analysis (Supplementary Figure 6) 36,37,65 . The Kaplan-Meir survival analysis was performed using two clinical datasets (GSE10143 and GSE10186) 46,47 . The TCGA RNAseq and clinical data of 369 liver cancer cases were downloaded from Genomic Data Commons Data Portal at National Cancer Institute (https://gdc-portal.nci.nih.gov/). Only upper quartile normalized fragments per kilobase of transcript per million mapped reads (UQ-FPKM) were extracted for survival analysis. The cohort was then stratified into two high or low expression of NR2E3 based on their expression value. Overall survival was determined by days to last follow up and days to death.

Statistical analysis.
Results are representative of at least 2-3 independent experiments with three replicates per experiment. A one-way analysis of variance (ANOVA) was used to determine the significance of the differences between treatment groups. The Newman-Keuls test was used for multi-group comparisons. Statistical significance was set at a P value of <0.05.