RNA N6-methyladenosine demethylase FTO promotes breast tumor progression through inhibiting BNIP3

The breast tumor and normal tissues were obtained at the Third Affiliated Hospital of Sun Yat-Sen University and were approved by the institutional review board of the hospital. The study is compliant with all relevant ethical regulations regarding research involving human participants. For fresh tissues, breast tumors and adjacent normal tissues were separately dissected at the time of surgery and immediately transferred to RNAlater (R0901, Sigma). The paraffin-embedded specimens were collected from at the Eighth Affiliated Hospital of Sun Yat-Sen University. Breast cancer cell lines MDA-MB-231, MCF-7 and 4 T1 were obtained from American Type Culture Collection (ATCC) with authentication. These cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Corning, USA) with 10% fetal bovine serum (Gibco, USA) and antibiotics (Gibco, USA). Cells were grown in a 5% CO2 cell culture incubator at 37 °C.

Stable knockdown of target genes was achieved by lentiviral-based short-hairpin RNA (shRNA) delivery. PLKO.1 vector with anti-puromycin or anti-hygromycin plasmid was constructed by using the primer sequences listed in

Additional file 1: Table S1. For YTHDF2 overexpressing system, YTHDF2 cDNA (NM_016258↗) was cloned into pCDH puro lentiviral vector (CD510B-1, System Biosciences), For shRNA knockdown and overexpression, pLKO.1 and pCDH constructs together with packing and helper plasmids PAX2 and MD2G were co-transfected into 293 T cells by Calcium Phosphate Transfection Kit (CAPHOS-1KT, Sigma). Viruses were collected, filtered, and titrated before infecting target cells with 8 mg/ml Polybrene (TR-1003, Sigma). The infected cells were screened by puromycin or hygromycin accordingly.

For immunohistochemistry (IHC) analysis, breast cancer specimens tissue slides were deparaffinized, rehydrated through an alcohol series followed by antigen retrieval with sodium citrate buffer. Tumor sections were blocked with 5% normal goat serum (Vector) with 0.1% Triton X-100 and 3% H₂O₂ in PBS for 60 min at room temperature and then incubated with appropriate primary antibodies 4 °C overnight. IHC staining was performed with horseradish peroxidase (HRP) conjugates using DAB detection. Nuclei were counterstained with Hoechst. Images were taken with Nikon microscopy.

Cells in 6-well plate or 60 mm² dishes were lysed in 80–100 ul modified RIPA buffer (P0013B, Beyotime, China) containing the complete cocktail of protease inhibitors (#11836153001, Roche, Switzerland). Protein concentrations were determined with the BCA protein assay kit (P0011, Beyotime, China). Proteins were separated by 12 or 15% SDS-PAGE and transferred to nitrocellulose filters, and blotted with related antibodies in 4 °C overnight. Secondary antibodies pre-labeled in room temperature for 1 h. The nitrocellulose filters with target protein were exposed in visualizer (4600, Tanon, China). Antibodies were purchased from the following: anti-FTO (ab124892, 1:1000, Abcam), anti-BNIP3 (ab109362, 1:1000, Abcam), anti-Bcl2 (A0208, 1:1000, ABclonal), anti-β-actin (AF1700, 1:1000, R&D system), anti-tubulin (13E5, 1:1000, Cell Signaling) and anti-caspase 3 (A2156, 1:1000, ABclonal).

Total RNAs were extracted by Trizol (ThermoFisher, USA) following the manufacturer’s instruction. Complementary DNA was synthesized using the SuperScript™ III First-Strand Synthesis System DNA using the PrimeScript RT reagent Kit (RR036A, Takara, Japan). Real-time reverse-transcription PCR was carried out by SYBR-Green Master mix (RR820B, Takara, Japan) in 7500 apparatus (Applied Biosystems). GAPDH was used as an internal control for the normalization. All primers used in this study are listed in Additional file: Table S1. 1

Total RNAs were isolated by TRIzol (ThermoFisher, USA) according to the manufacturer’s instructions. RNA quality was analyzed by NanoDrop3000. The EpiQuik m6A RNA Methylation Quantification Kit (Colorimetric) (P-9005, Epigentek, USA) was used to measure the m6A content in total RNAs. Briefly, 200 ng RNAs were coated on assay wells. Capture antibody solution and detection antibody solution were then added to assay wells separately in a suitable diluted concentration following the manufacturer’s instructions. The m6A levels were quantified colorimetrically by reading the absorbance of each well at a wavelength of 450 nm, and calculations were performed based on the standard curve.

The poly (A) RNAs (300 ng) were spotted onto a nylon membrane (GE Healthcare). The membranes were then UV cross-linked (254 nm), blocked, incubated with m6A antibody (ABE572, 1:1000, Merck Millipore) in 4° Covernight. Antibodies pre-labeled in room temperature for 1 h. The nylon membrane with m6A dots were exposed in visualizer (4600, Tanon, China). The same 300 ng poly (A) RNAs were spotted on the membrane, stained with 0.2% methylene blue in 0.3 M sodium acetate (pH 5.2) for 2 h, and washed with ribonuclease-free water for 1 h.

Total RNAs were extracted by TRizol (ThermoFisher, USA) from stable shFTO MDA-MB-231 cells and the controls. Chemically fragmented RNA (100 nucleotides) was incubated with m6A antibody for immunoprecipitation according to the standard protocol of Magna methylated RNA immune-precipitation (MeRIP) m6A Kit (#17–10,499, Merck Millipore, USA). Enrichment of m6A containing mRNA was analyzed either by qRT-PCR with the primers listed in Additional file 1: Table S1 or by high-throughput RNA sequencing. For high-throughput sequencing, purified RNA fragments from m6A-MeRIP were used for library construction with the NEBNext Ultra RNA library Prep kit for Illumina (E7530S, NEB, USA) and were sequenced by Illumina HiSeq 2000. Library preparation and high-throughput sequencing were performed by Novogene (Guangzhou, China). Sequencing reads were aligned to the human genome GRCh37/hg19 by Bowtie2, and the m6A peaks were detected by magnetic cell sorting as described [35].

For CCK8 assay, cells were seeded at 1000 cells per well in 96-well plates with fresh medium. Cell viability was assayed using Cell Counting Kit-8 (CK04, Dojindo, Japan) at the time in 0, 48, 72, 96,120,144 h. The microplates were incubated at 37 °C for additional 4 h. Absorbance was read at 450 nm using a microplate reader (ThermoFisher, USA) and the results were expressed as a ratio of the treated over untreated cells (as 100%).

For EdU (5-Ethynyl-2′-deoxyuridine) assay, logarithmic growth stage cells were seeded in 6-well plate with corresponding concentration of EDU reagent for 3 h. Cells were washed with PBS for 5 min twice, before incubating with 4% Paraformaldehyde for 30 min. After washing with PBS for 5 min twice, samples were permeated with 0.3% TritonX-100 in PBS, and dyed with reaction solution (C0075S, Beyotime, China).The images collected with 20× and 40× visions in Nikon microscopy.

For cell apoptosis assays, cells were performed using Annexin V-PI Apoptosis Detection Kit I (WLA001a, Wanleibio, China) according to the manufacturer’s instruction, and followed by flow cytometry analysis (Beckman, USA).

For mammosphere formation assay, cells were seeded into ultralow attachment plates (Corning) at a density of 20,000 viable cells/mL in a serum-free DMEM-F12 supplemented with 100× insulin, 20 ng/mL epidermal growth factor and 20 ng/mL basic fibroblast growth factor (Sigma), and 0.4% bovine serum albumin (Sigma), 100 × penicillin &streptomycin for two weeks until the mammosphere became visible.

For clonogenic assay, cells were seeded onto 35 mm² dishes at a density of 5000 viable cells/mL in a serum-free DMEM-F12 supplemented with 100× insulin, 20 ng/mL epidermal growth factor and 20 ng/mL basic fibroblast growth factor (Sigma), and 0.4% bovine serum albumin (Sigma), 100 × penicillin &streptomycin for two weeks until the colonies became visible and stained with Crystal Violet Staining Solution.

The dual-luciferase vector pmiGLO was purchased from Promega (C838A, Promega, USA). The 3’UTR of BNIP3 was amplified by PCR using the genomic DNA from MDA-MB-231 cells as a template. A clone whose sequence was identical to the NCBI reference sequence NM_004052↗ was used to clone into pmiGLO vector with SacI and SalI restriction sites. Three putative m6A recognition sites were identified in 3’UTR. Mutagenesis from A to T was generated by QuikChange II Site-Directed Mutagenesis Kit (200,523, Agilent, USA) according to the instruction. Luciferase activity was measured by Dual Luciferase Reporter Gene Assay Kit (RG028, Beyotime, China) in GM2000 (Promega). Experiments were performed in triplicates. The firefly luciferase activity values were normalized to the Renilla luciferase activity values that reflect expression efficiency. Data are presented as mean values (± s.d.).

Mice were housed at five mice per cage under pathogen-free conditions. All animal care and experiments were approved by the Institutional Animal Care and Use Committee of Sun Yat-Sen University, Guangzhou, China, and the study is compliant with all relevant ethical regulations regarding animal research. Mice were euthanized when they met the institutional euthanasia criteria for tumor size and overall health condition.

For the subcutaneous implantation model, 5 4-week-old female Balb/c mice were randomly grouped and injected with 1 × 10⁶ shCtrl, shFTO or shFTO/shBNIP3 KD 4 T1 cells. Tumors were measured with a caliper every 4 days to analyze tumor growth. Tumor volume was calculated by the formula V = ab2/2, where a and b are the tumor’s length and width, respectively. At the experimental endpoint, tumors tissues were harvested and fixed with 4% PFA for paraffin-embedded section.

For tumor metastasis mouse model, 5 4-week-old female Balb/c mice were randomly grouped and injected with 1 × 10⁶ shCtrl, shFTO or shFTO/shBNIP3 KD 4 T1 cells via tail vein. To detect lung metastasis, mice were sacrificed 3 weeks after tumor cells injection. Lung tissues were harvested and fixed with 4% PFA for paraffin-embedded section and lung metastases were detected with the Nikon microscopy.

For orthotopic xenograft mouse model, 5 4-week-old female NOD/SCID mice were randomly grouped. After NOD/SCID were anaesthetized and the skin was incised, shCtrl or shFTO MDA-MB-231-luciferase cells (1 × 10⁶) in 50 ul Hanks solution were orthotopically injected into mammary fat pads using a 1-ml Hamilton microliter syringe, and then the incision was closed using surgery suture threads with needle. Mice tumors were monitored by the IVIS system after luciferin injection for 15 min.

The gene expression profile dataset GSE9014↗, GSE11812↗ and GSE3188↗ was downloaded from GEO database. Data from GEO or RNA-Seq were analyzed by R (V3.3, http://www.bioconductor.org↗) with edgeR package. Fold-change (FC) of gene expression was calculated with a threshold criteria of log2FC ≥ 1.5 and P value< 0.01. KEGG pathway enrichment analysis was performed to investigate the processes of the candidate genes, by applying online tools of the KOBAS 3.0 (https://david.ncifcrf.gov/↗). The Search Tool for the Retrieval of Interacting Genes (STRING) database (V10.5, https://string-db.org/↗) was recruited to predict the potential interaction between BNIP3 and apoptosis genes at protein level. The online database of R2: Genomics Analysis and Visualization Platform (https://hgserver1.amc.nl↗) was applied to determine the clinical survival of the candidate genes. The relative expression of FTO was computed in breast tumor cohort (e.g. IHC samples) compared to the normal cohort, by which the value indicated the number of standard deviations away from the mean of expression in the normal population. High expression: > 1; Low expression: <− 1 (log2).

Means, SD and SEM were analyzed using Graphpad prism 7.0. Two-tailed Student’s t-test, were used to compare the statistical difference between indicated groups. Pearson analysis was used to analyze correlation between genes. Statistical significance was accepted for P-values of < 0.05.

To investigate the role of m6A modification in breast cancers, we systematically analyzed the transcriptomic profiles of 111 breast tumors and 12 non-tumorous (NT) breast tissues (GSE9014↗, Additional file 2: Figure S1A), and identified that FTO, the core m6A demethylase, was significantly up-regulated in breast tumors compared with normal tissue (Fig. 1a and b). We further confirmed the up-regulation of FTO in the group of DNBC (ER−/PR−/Her2+) and late stages (GRADE II and III) three clinical stages of breast cancer (Fig. 1c), suggesting that FTO may play a predominant role in mediating m6A modification in breast cancer. We also found that FTO was higher expressed in breast cancer cell lines than other cancer cell lines (GSE11612, Additional file 2: Figure S1B). To validate the up-regulated RNA level of FTO, we performed the immunohistochemistry (IHC) staining assay to detect the protein expression level of FTO in 36 clinical human breast tumor tissues and 12 corresponding NT adjunct breast tissues (Fig. 1d and Additional file 2: Figure S1C). Consistently, FTO protein was significantly overexpressed in breast tumor tissues compared to their adjunct tissues according to the quantification of IHC results (Fig. 1e), which supported our initial observation of FTO up-regulation in breast cancer. Next, we detected the global m6A level in 2 fresh human breast tumors and their corresponding adjunct NT tissues by the RNA dot-blotting assay (Fig. 1f) and 5 fresh human breast tumors and 3 normal breast tissues by the m6A colorimetric analysis (Fig. 1g). In line with initial observation, a notable decrease of global m6A abundance was detected in breast tumors. Moreover, with clinical outcome analysis, we found that up-regulation of FTO was significantly associated with lower survival rates in patients with advanced stage of breast cancer (Fig. 1h) and patients with ER negative breast cancer (Fig. 1i). It indicates that up-regulation of FTO may be implicated in breast cancer initiation and progression.

To determine whether FTO was critical to breast cancer cell growth, we generated two stable FTO-knockdown models in human MDA-MB-231 and MCF-7 cell lines, by infecting two distinct shRNA lentivirus (shFTO#1 and shFTO#2). The FTO-knockdown effects were confirmed in both RNA expression level and protein expression level (Additional file 3: Figure S2A and B). Knockdown of FTO significant inhibited cell growth in MDA-MB-231 and MCF-7 cells (Fig. 2a and b). We further validated the suppressed effect of FTO knockdown in cell proliferation by the EDU staining assay (Fig. 2c and d). Furthermore, we performed the mammosphere formation assay by seeding control and FTO-knockdown cells in mamosphere culture medium. As shown in Fig. 2e, ~ 1% of total MDA-MB-231 cells and ~ 3% of total MCF-7 cells formed mammospheres containing ~ 900 cells and ~ 2800 cells respectively after 12 days of cultivation in non-adherent dishes. Knockdown of FTO dramatically suppressed mammosphere formation in both MDA-MB-231 cells and MCF-7 cells (Fig. 2e). Similarly, depleting FTO inhibited breast cancer cell colony-forming abilities (Fig. 2f). We next examined the function of FTO in cell survival by flow cytometry with Annexin V/PI staining. FTO depletion resulted in significant cell apoptosis in both MDA-MB-231 and MCF-7 cells (Fig. 2g). Taken together, our results suggest that FTO plays an important role in controlling breast cancer cell growth, colony formation and cell death.

To further verify the oncogenic role of FTO in breast cancer, we performed a subcutaneous implantation experiment in BALB/c mice to examine the effect of FTO-knockdown in breast cancer tumorigenicity. Stable FTO-knockdown 4 T1 cells (mouse-derived breast cancer cell line) were constructed by using FTO shRNA (Fig. 3a), and subcutaneously injected into 4-week-old female BALB/c mice. We observed that loss of FTO effectively inhibited breast tumor growth in mice as reflected by the significant reduction of tumor size and tumor weight comparing to the controls (Fig. 3c-e). Alternatively, Rhein, a potent FTO inhibitor [36], was used to examine the inhibitory effect of FTO (Fig. 3b). We intraperitoneally injected Rhein (10 mg/kg) into mice 3 times per week until the tumor reached 10 mm³. Consistently, breast tumor retardation was observed in mice treated with Rhein by comparing to the control treated with DMSO (Fig. 3c-e).

In addition, we performed an orthotopic xenograft experiment in NOD/SCID mice. The MDA-MB-231 cells were engineered to stably express luciferase for in vivo imaging. Control and FTO-knockdown MDA-MB-231 cells were orthotopically injected into mouse mammary gland fat pads, and xenograft tumor growth was monitored accordingly. At 30 days after injection, all mice bearing FTO-knockdown cells barely had detected tumors whereas mice bearing control cells showed visible tumor growth as reflected by the luciferase signals in vivo (Fig. 3f and g). To investigate the effect on tumor metastasis, FTO-knockdown 4 T1 cells and control cells intravenously injected into BALB/c mice. As expected, silencing FTO attenuated lung metastasis in mice (Fig. 3h-j). Taken together, our results indicated that FTO plays a critical role in promoting breast tumor growth and metastasis in vivo.

To investigate the molecular mechanism of FTO and identify its downstream targets in breast cancer, we performed transcriptome sequencing to examine the expression changes in our stable FTO-knockdown MDA-MB-231 cells (Additional file 4: Table S2) and MCF-7 cells treated with DMOG, a FTO inhibitor (GSE3188↗, Additional file 5: Table S3). According to the standard GEO2R analysis and quantile normalization, 55 genes with significant changes were overlapped in these two experiments (Fig. 4a). KOBAS 3.0 was adopted to conduct the gene enrichment pathway in both MDA-MB-231 and MCF-7 cells [37–39]. KEGG analysis revealed that inhibiting FTO could dramatically affect the differentially expressed genes enriched in the signal pathways involved in cell proliferation, cell cycle and apoptosis (Fig. 4b). We also performed the m6A-Seq to map the m6A modification in MDA-MB-231 cells, and found that most m6A signal was enriched around the stop codon of mRNAs. Among these, we selected BNIP3, a pro-apoptosis gene in the FoxO signaling pathway as a candidate target of FTO-mediated m6A modification for further investigation (Fig. 4c and d). BNIP3 was significantly up-regulated in FTO-knockdown MDA-MB-231 cells and FTO-inhibition MCF-7 cells (Fig. 4e and f). The co-expression analysis by the string showed that BNIP3 was closely related with the apoptosis genes such as Bcl-2 and Caspase 3 (Fig. 4g).

To verify BNIP3 as a FTO downstream target, we detected both BNIP3 mRNA expression level and protein expression level in breast cancer cells. In agreement with our RNA-seq data, BNIP3 was remarkably up-regulated in stable FTO-knockdown MDA-MB-231 and MCF-7 cells (Fig. 5a and b). Meanwhile, silencing FTO promoted the cleaving of Caspase 3 and decreased the expression of Bcl2 in MDA-MB-231 and MCF-7 cells (Fig. 5c and d). These results showed that FTO inhibited cell apoptosis via down-regulating BNIP3. To validate BNIP3 as a bona fide target of FTO for the m6A modification, we performed the m6A-RNA immunoprecipitation assay and analyzed with qRT-PCR. As expected, knockdown of FTO dramatically promoted the m6A level of BNIP3 mRNA (Fig. 5e). From the m6A RNA-seq, 3 potential m6A sites (RRACH) were identified in the 3′ UTR near to the termination codon of BNIP3 (Fig. 4c). To further prove the effect of m6A modification on BNIP3 expression, we cloned the BNIP3 3’UTR portion containing 3 potential m6A sites into a dual luciferase reporter construct pmirGLO and generated 3 mutant BNIP3 3’UTR reporter vectors, respectively. For the mutant form of BNIP3, we replaced the adenosine base in m6A consensus sequences with thymine to abolish the m6A modification (Fig. 5f). Relative normalized luciferase activities of the wild-type and 3 mutant BNIP3 3’UTR reporter vectors were compared in control and FTO-knockdown MDA-MB-231 cells. A significant induction (~ 1.8 fold) in the luciferase activity was observed in the wild-type 3’UTR of BNIP3 in FTO-knockdown cells compared with control cells, while only mutant #1 almost abolished this induction, indicating the modulation of BNIP3 expression was under the control of FTO-associated m6A modification on site #1 (Fig. 5g). Furthermore, we next investigated the potential mechanism by which m6A methylation regulates the expression of BNIP3. As YTHDF2 is recognized as the main m6A reader in the 3’UTR sites of target genes [40, 41], we measured its role in mediating BNIP3 expression level by m6A. Overexpression of YTHDF2 showed no effects in reducing the mRNA expression level of BNIP3 in FTO-silenced breast cancer cells (Additional file 6: Figure S3A-B), suggesting that FTO mediated-m6A modification decreased BNIP3 expression in the YTHDF2-independent manner.

As a functional target of FTO, BNIP3 is required for cell apoptosis, suggesting its role as tumor suppressor in breast cancer. By analysis of the Oncomine database, we found that BNIP3 was frequently down-regulated in various human cancers including breast cancer comparing to the non-tumor adjunct tissues (Fig. 6a). Overexpression of BNIP3 in MCF-7 cells activated Caspase3 cleavage and inhibited the expression of Bcl-2 (Fig. 6b). Our analysis of R2 datasets revealed that BNIP3 levels are positively correlated with overall survival of patients with breast cancer treated with taxane-anthracycline (P-value = 0.035), and with Tamoxifen (P-value = 0.026) (Fig. 6c and d). By analyzing the data set (GSE9014↗), we detected the levels of BNIP3 and FTO in normal and breast tumor samples. A significant negative correlation between BNIP3 and FTO levels was observed in breast tumors (coefficient = − 0.3083, P = 0.0029) (Fig. 6e). To validate their correlation, we performed the immunohistochemistry assay to detect the protein levels of BNIP3 and FTO in our breast tumor sample cohort consisting of 36 primary breast tumor tissues. Our results showed that samples with higher FTO expression (vs average expression level of FTO in breast tumor tissues) was frequently associated with lower BNIP3 level and vice versa (Fig. 6f and Additional file 7: Figure S4). Consistently, a significant negative correlation between BNIP3 and FTO levels was observed in our clinical cohort (coefficient = − 0.3325, P = 0.0468) (Fig. 6g). Moreover, we also detected that BNIP3 was upregulated in FTO inhibiting or FTO silencing tumor samples in 4T1-hypodermic breast cancer models (Fig. 6h, i). Taken together, we concluded that BNIP3 was negatively correlated with FTO expression in clinical samples.

To investigate whether BNIP3 mediated FTO-dependent tumor growth and progression, we generated two distinct shRNAs targeting BNIP3 (shBNIP3–1 and shBNIP3–2) in FTO stable knockdown breast cancer cells (Additional file 8: Figure S5A-C). We found that suppression of BNIP3 alleviated the inhibitory effects on cell proliferation mediated by FTO (Fig. 7a and b). Consistently, we confirmed the similar effects of BNIP3 by the EDU staining assay in MDA-MB-231 and MCF-7 cells (Fig. 7c and d). To verify the role of BNIP3 in vivo, we performed subcutaneous implantation with stable FTO-knockdown 4 T1 cells and double knockdown cells (shFTO and shBNIP3) into the 4-week-old female Balb/c mice. Double knockdown of BNIP3 and FTO promoted breast tumor growth in mice as reflected by the significant increase of tumor size and tumor weight comparing to the stable FTO-knockdown cells (Fig. 7e-g). In addition, we detected the increase of lung metastasis in the double knockdown 4 T1 cells by tail vein injection compared to the stable FTO-knockdown 4 T1 cells (Fig. 7h-j). The above results indicated that silencing BNIP3 could significantly alleviate the FTO-dependent inhibitory effects on tumor growth and metastasis in vitro and in vivo.

This work was supported in part by grants from National Natural Science Foundation of China 31701114 (G. Wan), 81602972 (X. Xiong); Science and Technology Planning Project of Guangdong Province 2017A010105004 (X. Wu), 2017A010103009, 2017B020227009 (B. Wei); Fundamental Research Funds for University-Key Cultivation Project of Young Teacher in Sun Yat-Sen University 17ykzd11 (G. Wan), Fundamental Research Funds for the Central Universities 16ykjc23 (B. Wei). The funders were not involved in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Please contact the corresponding authors for all data requests.

All experiments were approved by the Ethics Committee of the Third Affiliated Hospital of Sun Yat-Sen University (No.²⁰¹⁶2–145) and by the Animal Ethics Committee of Sun Yat-Sen University (No.²⁰¹⁷03–07).

All authors give consent for the publication of the manuscript in Molecular Cancer.

The authors declare that they have no competing interests.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Please contact the corresponding authors for all data requests.