Upregulation of CYP1B1 by hypoxia is mediated by ERα activation in breast cancer cells

Background: Endocrine therapy for breast cancer often leads to drug resistance and tumor recurrence; tumor hypoxia is also associated with mortality and tumor relapse. Cytochrome P450 1B1 (CYP1B1) regulates estrogen metabolism in breast cells and is known to be overexpressed in breast cancer tissue. Although the individual association of hypoxia-induced hypoxia-inducible factor-1α (HIF-1α) and CYP1B1 with tumorigenesis is well known, the association between HIF-1α and CYP1B1 leading to tumorigenesis has not been investigated. Here, we investigated the correlation between hypoxia and CYP1B1 expression in breast cancer cells for tumorigenesis-related mechanisms. Methods: Hypoxia was induced in the human breast cancer cell lines MCF-7 (Er-positive) and MDA-MB-231 (triple-negative) and the normal breast epithelial cell line MCF10A, and then subjected to immunoblotting, transient transfection and luciferase assays, gene silencing using small interfering RNA, PCR analysis, and chromatin immunoprecipitation, co-immunoprecipitation, and mammalian two-hybrid assays. Furthermore, immunouorescence analysis of tumor microarrays was performed and the pub2015 and The cancer genome atlas patient datasets were analyzed. Results: HIF-1α expression in response to hypoxia occurred in both normal and breast cancer cells, whereas CYP1B1 was induced only in estrogen receptor α (ERα)-positive breast cancer cells under hypoxia. HIF-1α activated ERα by direct binding as well as in a ligand-independent manner to promote CYP1B1 expression. Conclusions: Therefore, we established the mechanism by which hypoxia and ER-positivity orchestrate breast cancer relapse.


Background
Breast cancer is one of the most prevalent cancers in women, and malignant breast cancer is known to be the leading cancer-related deaths in women worldwide [1]. Breast cancer is classi ed depending on the expression of estrogen receptor α (ERα), epidermal growth factor receptor 2, and progesterone receptor; the lack of expression of all three receptors is called triple-negative breast cancer [2]. These classi cations drive the therapeutic approach for breast cancer patients in clinical practice [3,4].
Approximately 70% of breast cancers are ERα positive and are usually treated with anti-hormonal therapy [5]. Several molecules have been developed for hormone neutralization in women with ERα-positive cancers, including ER modulators such as tamoxifen, ER degraders, and aromatase inhibitors [6,7].
However, many patients develop drug resistance to anti-hormonal therapy, and approximately 50% of patients with malignant breast cancer do not respond to ER modulators in the initial setting [7]. Moreover, the underlying mechanisms of acquiring resistances to these therapeutics are not well understood and there is a lack of therapeutic targets to overcome tumor relapse.
Hypoxia-induced hypoxia-inducible factor-1α (HIF-1α) is one of the factors determining resistance to antihormone therapy and is a therapeutic target to overcome tumor relapse in breast cancer [8]. Intracellular signaling in hypoxia is mediated by the HIF-1α, which regulates the expression of several essential genes [9]. Under normoxic conditions, HIF-1α protein is ubiquitinated and degraded, but under hypoxia, it is stabilized and translocated to the nucleus by HIF-1β dimerization [10][11][12]. Several studies have reported that the overexpression of HIF-1α is associated with the initiation, progression, and recurrence of breast cancer [8, [13][14][15].
Several studies have shown that metabolites of 17β-estradiol (E2) induce breast tumor development, therefore, studies have been conducted to better understand the mechanisms involved in estrogen metabolism and estrogen metabolites [16,17]. In breast cancer cells, E2 binds to ERα and the complex translocates to the nucleus, where it binds the estrogen response element (ERE) site and induces CYP1B1 transcription [18]. CYP1B1, a monooxygenase expressed in endocrine regulatory tissues such as the breast, uterus, and ovary, is a major enzyme involved in E2 metabolism [19,20]. CYP1B1 is also known to be involved in hormone-dependent carcinogenesis by induction of metabolites of intracellular E2 and environmental carcinogens [16, [21][22][23]. It metabolizes E2 to 4-hydroxy-E2 to generate free radicals and form DNA adducts, which cause intracellular DNA damage and carcinogenesis in tissues, including breast tissue [17,18,[24][25][26].
Although the association of hypoxia with CYP1B1 has been reported, that between HIF-1α and CYP1B1 has not yet been studied. In this study, we assessed the role of hypoxia-induced HIF-1α in regulating CYP1B1 in breast cancer as well as the mechanism underlying this interaction. Our ndings suggest that HIF-1α is involved in the regulation of CYP1B1 in breast cancer cells and that suppressing HIF-1α-induced CYP1B1 expression in patients with recurrent cancer can be utilized as a novel therapeutic strategy for breast cancer.
Antibodies against PKA/p-PKA, Akt/p-Akt, ERK/p-ERK, SRC/p-SRC, p-ERα, and horseradish peroxidaselinked anti-mouse and rabbit IgG were obtained from Cell Signaling Technologies (Danvers, MA, USA). Antibodies against β-actin and CYP1B1 were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). The HIF-1α antibody was obtained from Abnova (Taipei City, Taiwan), and the ERα antibody was obtained from Abcam (Cambridge, UK). The antibody against p-ERa (Ser305) was obtained from Bethyl Laboratories (Montgomery, Texas, USA). PE-conjugated goat anti-mouse IgG and Alexa Fluor 488conjugated goat anti-rabbit IgG were purchased from Thermo Fisher Scienti c (Waltham, MA, USA).
Oligonucleotide primers for polymerase chain reaction (PCR) were custom synthesized by Bioneer (South Korea). All chemicals used were of the highest commercially available grade.

Cell culture and treatment
The human breast cancer cell lines MCF-7 (Er-positive) and MDA-MB-231 (triple-negative) were obtained from the Korea Cell Line Bank (KCLB, Seoul, South Korea) and the normal breast epithelial cell line MCF10A was obtained from the Laboratory Animal Resource Center (Korea Research Institute of Bioscience and Biotechnology, South Korea). MCF-7 and MDA-MB-231 cells were cultured in DMEM from HyClone (Thermo Fisher Scienti c) supplemented with 10% FBS in a humidi ed 5% CO 2 incubator at 37°C . MCF10A cells were cultured in Human Mammary Epithelial Cell Systems (HMEM) BulletKit medium from Lonza (Basel, Switzerland) and only cells up to passage 15 were used in the experiments. To induce hypoxia, these cells were cultured in a hypoxia chamber (Billups-Rothenberg, San Diego, CA, USA) containing a 1% O 2 gas mixture. Cell lines were DNA ngerprinted by the Korea Cell Line Bank (KCLB, Seoul, South Korea). The stock solution of the inhibitor was diluted with dimethylsulfoxide (DMSO) and added directly to the culture medium. Control cells were treated with DMSO only, and the nal DMSO concentration was always <0.2%.

Immunoblot
Protein lysates were prepared from cells cultured under hypoxia or normoxic conditions, and proteins were quanti ed using the Bradford protein assay (Bio-Rad, Irvine, CA, USA). The lysates were electroblotted on polyvinylidene di uoride membranes and reacted with the primary antibodies and secondary antibodies. The membranes were visualized with enhanced chemiluminescence (ECL) solution and analyzed using a ChemiDoc Imager (Bio-Rad).
The CYP1B1-Luc FL and deletion plasmids (−910 to +25 and −91 to +25) were used to investigate the hypoxia-mediated CYP1B1 promoter binding sites. After transfection, cells were incubated for 24 h under normoxic or hypoxic conditions, and cell lysates for luciferase reporter assay were prepared. Cell lysates were mixed with an equal volume of luciferase assay substrate reagent and analyzed using a luminometer (SpectraMax M5; Molecular Devices).
Gene silencing using small interfering RNA (siRNA) The expressions of HIF-1α and ERα were knocked down using siRNA, and the results were analyzed by real-time qPCR, chromatin immunoprecipitation, and immunoblot analysis. siRNAs targeting HIF-1A and ESR1 mRNA and non-targeting siRNAs were purchased from OriGene (OriGene Inc, Rockville, MD, USA).

RNA extraction and digital PCR analysis
Under hypoxia or normoxia, total cellular RNA was isolated using the RNeasy Mini Kit (Qiagen, Germany) and reverse-transcribed using RNA-to-cDNA EcoDry reagent (TaKaRa, Japan). PCR products were directly monitored during PCR assays using QuantStudio 3 real-time PCR system (Applied Biosystems) and detected via SYBR Gene reporter dye enhancement. The levels of CYP1B1 and β-actin mRNA in MCF7 cells were compared with those in control cells using the comparative cycle threshold (Ct) method.
Chromatin immunoprecipitation (ChIP) assay Cells cultured under normoxic or hypoxic conditions were treated with formaldehyde to cross-link protein and DNA. The cells were then sonicated and chromatin-DNA complexes were precipitated using an antibody against ERα or non-speci c mouse IgG using an EZ CHIP kit (Upstate, Lake Placid, NY, USA). PCR was performed using puri ed DNA, oligonucleotide PCR primers, and Taq DNA polymerase (TaKaRa). PCR products were analyzed on an agarose gel using the SYBR Safe DNA Gel Stain kit.
Co-immunoprecipitation (Co-IP) assay Cells were cultured under hypoxic or normoxic conditions and lysed with IP buffer (Roche). Protein lysates were pre-removed by incubation with Protein A beads (GE Healthcare) and then incubated with beads and ERα antibodies. Antibody-bound beads were washed with PBS, electrophoresed on SDS-PAGE, and then immunoblotted with HIF-1α.

Immuno uorescence analysis of tumor microarrays (TMAs)
A human breast cancer tissue microarray (TMA, #BC081120e) containing 110 cases was purchased from Biomax (Rockville, MD). The immuno uorescent image was quanti ed using a quantitative analysis system (Thunder, Leica). Image analysis was performed using the LAS-X software (Leica) by acquiring uorescence images of DAPI, CYP1B1, and HIF-1α for each TMA spot and scoring the sum of the target pixel intensities. Tumors were divided into high and low groups according to the ER expression score (according to the patient information sheet). The correlation of CYP1B1 and HIF-1α in ERα-positive breast tumors was assessed using linear (Pearson) and nonparametric (Spearman) correlation coe cients.
The cancer genome atlas (TCGA) data and cBioPortal Ethical approval for the study was not required due to the retrospective nature of this study using only publicly available data. Transcriptome analysis data and pathological data of human breast cancer biopsy samples were obtained from the following dataset in cBioPortal platform: BRCA-TCGA-pub2015 [817 cases with mRNA data (RNA Seq V2 RSEM)], BRCA-TCGA [1100 cases], BRCA-METABRIC [1904 cases with mRNA data (microarray)]. Gene expression levels were normalized (Z-score) and scaled (logtransformed). Co-expression, overall survival, and volcano plots were calculated according to cBioPortal's online instructions and analyzed using the GraphPad Prism 9.00 software. The BRCA-METABRIC dataset was analyzed using the cBioPortal online tool for exploration and comparison between the two patient groups, and 304 genes differentially expressed in the ERα-positive high pro le/HIF1α-CYP1B1 high group was ltered (q-value < 0.01, |Log 2 (FC)| > 0.58) and subjected to gene enrichment analysis using the Metascape platform (http://metascape.org) for gene annotation. The ltered data were subjected to hierarchical clustering and visualized as a heatmap.

Statistical analysis
Data are presented as ± standard error of the mean (SEM). All statistical analyses were performed using Student's t-test or the Mann-Whitney U-test when data were not normally distributed. Statistical signi cance was set at P < 0.05.

HIF-1α is associated with CYP1B1 expression in breast cancer cells
To investigate the effect of hypoxia on CYP1B1 expression in breast cell lines, HIF-1α and CYP1B1 expression were analyzed by immunoblotting. HIF-1α protein level was induced by hypoxia at 6, 12, and 24 h but CYP1B1 expression was not altered in MCF10A normal breast epithelial cells (Fig. 1a) or ERαnegative MDA-MB-321 cells (Fig. 1b). In contrast, both HIF-1α and CYP1B1 expression increased due to hypoxia in MCF-7 cells, which are ERα-positive (Fig. 1c). To verify that hypoxia-induced CYP1B1 expression is associated with HIF-1α, ERα-positive MCF-7 cells were cultured under hypoxic conditions after transient transfection with HIF-1α-siRNA. The increase in CYP1B1 expression due to hypoxia was suppressed by HIF-1α knockdown (by HIF-1α-siRNA) at 24 h, suggesting that CYP1B1 expression under hypoxia is induced by HIF-1α (Fig. 1d). As a negative control, the use of scrambled siRNA instead of HIF-1α-siRNA lead to no decrease in CYP1B1 expression under hypoxic conditions (Fig. 1d). To con rm the role of HIF-1α and hypoxia on CYP1B1 expression, cells were treated with a YC-1, an HIF-1α inhibitor developed for the treatment of circulatory disorders, and immunoblot was performed. YC-1 inhibited CYP1B1 expression under hypoxic conditions (Fig. 1e). Next, to determine whether CYP1B1 activity is regulated by hypoxia, MCF-7 cells were transiently transfected with CYP1B1-Luc and HIF-1α-siRNA under hypoxic conditions. Hypoxia-induced CYP1B1 luciferase activity, which was reduced by HIF-1α knockdown with HIF-1α siRNA (Fig. 1f). These results suggest that CYP1B1 expression under hypoxia is mediated by HIF1-α.
The estrogen response element (ERE) of the CYP1B1 promoter is su cient for hypoxia-induced upregulation of CYP1B1 CYP1B1 expression and luciferase activity were induced in ERα-positive MCF-7 cells by hypoxia but not in ERα-negative MCF10A or MDA-MB-231 cells, suggesting that CYP1B1 regulation by hypoxia is ERαdependent ( Fig. 2a and 2b). Based on these ndings, we used the CYP1B1 promoter vector to verify whether ERα is involved in regulating CYP1B1 expression in hypoxia. The human CYP1B1 promoter full length (−910/+25 bp; Fig. 2c) contains a xenobiotic response element (XRE; −853/-824 bp), an active protein 1 binding site (AP-1; −149/−129 bp), and the E2 response element (ERE; −84/−49 bp), which is important for regulating CYP1B1 transcription [18, 26, 28]. We found that CYP1B1 promoter activity was increased even in the construct in which XRE and AP-1 were deleted. In addition, hypoxia-induced CYP1B1 activity was increased even in −91/+25 containing only the ERE site, con rming that the ERE site is essential for hypoxia-mediated CYP1B1 regulation (Fig. 2c). Moreover, it was established that hypoxia directly regulates CYP1B1 through the luciferase activity of the ERE-Luc vector in MCF-7 cells (Fig. 2d).
ERα-mediated hypoxia-induced CYP1B1 expression was further veri ed by immunoblot analysis in MCF-7 cells treated with ERα-siRNA. ERα-siRNA treatment suppressed CYP1B1 expression in MCF-7 cells under hypoxia (Fig. 2e). Next, ChIP analysis con rmed the binding of ERα to ERE in the CYP1B1 promoter region under hypoxia (Fig. 2f). These results suggest that hypoxia-activated ERα can induce CYP1B1 expression by binding the CYP1B1 promoter in ERα-positive MCF-7 cells.

HIF-1α interacts with ERα in MCF-7 cells
To further con rm that ERα regulates CYP1B1 expression by hypoxia, we evaluated the effect of hypoxia in two types of breast cancer cells, namely, MDA-MB-231 and MCF-7 with different ERα expression. HIF-1α expression in MDA-MB-231 cells was induced by hypoxia but not in MCF-7 cells with high ERα expression (Fig. 3a). Also, there was no change in ERα expression when MCF-7 cells were incubated for 0-60 min under hypoxia (Fig. 3b). Therefore, it was established that the ERα expression is not regulated by hypoxia.
As ERα expression is not regulated by HIF-1α, we investigated the association between HIF-1α and ERα by a co-immunoprecipitation assay using cell lysates of MCF-7 cells cultured under hypoxic conditions. Cell lysates under normoxic and hypoxic conditions were immunoprecipitated using ERα antibodies, and the immunoprecipitates were subjected to immunoblot analysis using HIF-1α antibodies. HIF-1α and ERα were bound in MCF-7 cells under hypoxia (Fig. 3c). To identify the binding domain of ERα required for interaction with HIF-1α, M2H assay was performed using ERα truncated mutants. ERα has three functional domains-activation domain 1 (AF1; 1-180 aa), DNA binding domain (DBD; 180-302 aa), and activation domain 2 (AF2; 302-595 aa) (Fig. 3d). To determine the sites required for binding between ERα and HIF-1α, a deletion construct of these regions (pBind-ERα deletion construct) was designed. Luciferase activity by the interaction of HIF-1α with ERα AF2 (LBD) was approximately 2.5-fold higher than that using vectors containing AF1 or DBD (Fig. 3d). LBD is a region with hormone binding sites, homo-and hetero-dimerization interfaces, and ligand-dependent co-regulator binding [29]. These ndings suggest that HIF-1α and ERα physically interact with each other under hypoxia in MCF-7 cells via the LBD of ERα (Fig. 3e).

HIF-1α induces ERα activation in a ligand-independent manner
Phosphorylation of phosphatidylinositol-3-kinase (PI3K) [29,30] and extracellular signal-regulated kinase (ERK) [31,32] are known to module the HIF-1α signaling pathway. A recent study showed that the cAMPdependent protein kinase A (PKA) phosphorylates HIF-1α and increases its activity in cancer cells [33]. Although ERα can be activated by genomic pathways [34], non-genomic pathways mediated by phosphorylation by various kinases are also important in the regulation of ERα activity [35]. Figure 4a shows the phosphorylation sites of ERα protein. Serine 118, 167, and 305 are activated by the kinases ERK, Akt, and PKA, respectively [36-38], and tyrosine 537 is activated by Src [39]. Each of these phosphorylation sites is responsible for E2-independent activation and non-genomic pathways [40]. To investigate whether ERα is activated by hypoxia-induced HIF-1α protein signaling cascade, we analyzed the degree of ERα phosphorylation in lysates of cells cultured under hypoxia by immunoblotting. When MCF-7 cells were incubated under hypoxia lead to phosphorylation of ERα at S118, 167, and 305 in a time-dependent manner (Fig. 4b). These kinases showed increased activation in a time-dependent manner in MCF-7 cells under hypoxia for up to 60 min (Fig. 4c). To further validation of the upstream signaling pathways involved in hypoxia, we evaluated the effects of several kinase inhibitors on CYP1B1 protein expression. PD, LY, and H-89, inhibitors of ERK, Akt, and PKA, respectively, decreased hypoxiainduced CYP1B1 expression (Fig. 4d). Moreover, these inhibitors reduced hypoxia-mediated CYP1B1 mRNA expression and CYP1B1 promoter luciferase activity in breast cancer cells under hypoxia Thus, hypoxia induces ligand-independent ERα activation as a non-genomic mechanism, as well as the binding of ERα and HIF-1α (Fig. 4f). PI3K and ERK are known to be hypoxia-induced modulators of HIF-1α signaling pathways.

HIF-1α positively correlates with CYP1B1 expression in breast cancer patients
To clinically evaluate the correlation between HIF-1α and CYP1B1, we performed immuno uorescence analysis using a breast cancer patient TMA. In the TMA speci cation sheet, 10 normal breast tissues and 100 breast cancer tissues were classi ed according to the immunohistochemical (IHC) notation of the ER ("−" indicates ER-negative and "+" indicates ER-positive; Supple Fig. 1a, b and Supple Table 1).
Fluorescence intensity results of CYP1B1 and HIF-1α in 100 breast cancer tissue samples from TMA (Supple Fig. 1c, d, Supple Tables 2, 3) were scored using the ImageJ software (National Institutes of Health, Bethesda, MD, USA). The uorescence intensity of CYP1B1 and HIF-1α in ER-positive breast cancer tissue was higher than that in ER-negative breast cancer tissue (Fig. 5a). Pearson's correlation analysis was performed on ER-positive TMA samples (N = 64), and there was a signi cant positive correlation between CYP1B1 and HIF-1α levels (p < 0.0001, r² = 0.26) (Fig. 5b). When ER-positive breast cancer tissues were divided into low (IHC score: 1+, 2+) and high (IHC score: 3+) expression groups according to ER expression status (TMA speci cation sheet), the expression of HIF-1α was higher in benign tissues (Supple Fig. 1e). Next, after separating groups using the 25th or 75th percentile of HIF-1α as a cutoff, the expression of CYP1B1 was established to be signi cantly increased in ER-positive tissues with high HIF-1α expression (Fig. 5c).
To validate the clinical signi cance of HIF-1α and CYP1B1 expression in ER-positive breast cancer patients, further analysis was performed using the breast cancer patient transcriptional public data set (cBioPortal study IDs: brca_metabric, pub2015 and TCGA). After classifying ER-positive patients in all three data sets, a positive correlation between HIF1a and CYP1B1 expression was con rmed (Supple. Fig.  2a, B, C). Additional analysis was performed based on the brca_metabric dataset with the largest number of patient data. Analysis of 1459 ER-positive patients from the brca_metabric dataset showed a statistically signi cant positive correlation between the expression of CYP1B1 and HIF-1α (Fig. 6a). In this dataset (metabric), patients with high CYP1B1 and HIF-1α expression were classi ed as ER high (N = 26, top 25%) and ER low (N = 33, bottom 25%) based on ER expression. Of the 18 042 genes expressed in this patient group, 304 genes (DEG; q-value < 0.01, 0.58-|Log2(FC)|) that changed in the ER low versus ER high group were identi ed (red spots, increase = 136 and blue spots, decrease = 168) (Fig. 6b). Gene ontology (GO) and enrichment analysis (metascape) were performed based on the differentially expressed genes (DEG) of the ER/HIF1α/CYP1B1 triple high group. Genes associated with tumor recurrence were identi ed at the top and visualized as a percentage (% of number) and −Log 10 (q-value) of shared genes (Fig. 6c, Supple Table 4 and Supple Fig. 3a); 43 such genes, including cell cycle-related genes such as aurora kinase, associated with tumor recurrence were analyzed by hierarchical clustering and visualized as heat maps (Supple. Fig. 3b, Supple. Table 5). We also found that the survival rate is signi cantly lower in patients with high ER/HIF1α/CYP1B1 breast cancer (Fig. 6d). The pub2015 and TCGA patient datasets, as well as the metablic dataset, showed a remarkable correlation between high ER/HIF1α/CYP1B1 and lower survival rates and cancer recurrence in breast cancer patients (Supple. Fig.  4a and b). Taken together, our results suggest that the overexpression of HIF-1α and CYP1B1 in ERpositive breast cancer patients is associated with negative clinical outcomes.

Discussion
Approximately 70% of breast cancers are ER-positive [41] and treated with anti-hormonal therapy; therefore, the recurrence of breast cancer is high and signi cantly negatively impacts patient survival. Although many studies have been conducted to nd biomarkers for breast cancer recurrence [42,43], recurrence and metastasis remain unresolved. Previous studies have shown that HIF-1α is overexpressed in advanced breast cancer and that the hypoxic tumor microenvironment [44] is involved in tumorigenesis, refractory cancer, and recurrence [13][14][15]45]. Therefore, several studies have developed drugs that target the HIF-1α signaling pathway. Inhibition of HIF-1α by YC-1 is known to decrease cell growth and metastasis in breast cancer [46]. In addition, antiangiogenic therapy that suppresses hypoxia is an effective treatment approach because hypoxia can induce tumor progression and metastasis [47][48][49].
CYP1B1 is an E2 hydroxylase enzyme involved in the production of DNA damage inducers through estrogen biosynthesis and metabolism [22]. Several studies have reported that CYP1B1 is overexpressed in malignancies of various tissue origins, including the breast, colon, lung, brain, skin, prostate, ovarian, and liver cancers [50][51][52][53][54]. As CYP1B1 expression is increased in tumor tissues compared to normal tissues [53, 55, 56], CYP1B1 has garnered considerable research attention as a potential therapeutic target in tumors [57].
Although the individual association of HIF-1a and CYP1B1 with tumorigenesis is well known, to the best of our knowledge, their correlation has not been studied in breast cancer. We investigated the relationship between CYP1B1 expression and HIF-1a in human ERα-positive breast cancer. We found that hypoxiainduced CYP1B1 expression occurred only in ERα-positive breast cancer MCF-7 cells. HIF-1α can bind ERα directly, and this complex translocates from the cytoplasm to the nucleus and binds the ERE site of the CYP1B1 promoter in MCF-7 cells. We also found that hypoxia activates ERα in a ligand-independent manner in the absence of E2 through the activation of signaling kinases (Fig. 7). Moreover, immuno uorescence analysis of tissue microarrays from ER-positive breast cancer patients con rmed a positive correlation between HIF-1α and CYP1B1. A positive correlation between cancer recurrence and mortality and high ERα, HIF-1a, and CYP1B1 expression was also con rmed in a dataset of breast cancer patients.
A limitation of this study is that we were unable to verify whether inhibition of HIF-1α and CYP1B1 inhibited cancer recurrence in recurrent ERα-positive breast cancer patients. Our ndings provide the foundation for further studies and promote the use of HIF-1α and CYP1B1 as therapeutic targets in the treatment of recurrent tumors.

Conclusions
We report the mechanism by which hypoxia-induced HIF-1α regulates CYP1B1 through the activation of ERα in breast cancer cells and the role of hypoxia and Erα positivity in tumor relapse. Our ndings con rm the role of hypoxia in the pathogenesis and progression of ERα-positive breast cancer and suggest that the regulation of CYP1B1 expression under hypoxic tumor conditions can be targeted to overcome cancer recurrence, especially as an alternative to conventional hormone therapy.

Abbreviations
ChIP, chromatin immunoprecipitation; Co-IP, Co-immunoprecipitation; Ct, cycle threshold; DMSO, dimethylsulfoxide; DEG, differentially expressed gene; ECL, enhanced chemiluminescence; FBS, fetal bovine serum; GO, Gene ontology; IHC, immunohistochemical; LBD, ligand-binding domain; M2H, mammalian two-hybrid; PCR, polymerase chain reaction; siRNA, small interfering RNA; SEM, standard error of the mean; TCGA, The cancer genome atlas; TMA, tissue microarray Declarations Firehose Legacy; https://www.cbioportal.org/study/summary?id=brca_tcga. The breast cancer TMA dataset analyzed in this study is available in the Supplementary information le. H&E image of Supple Fig. 1b          Mechanism of hypoxia-mediated CYP1B1 upregulation through ER activation. In the presence of estrogen as an ER ligand, estrogen directly binds ER or induces CYP1B1 expression through ER activation through