METTL14 suppresses the expression of YAP1 and the stemness of triple-negative breast cancer Authors: Xupeng Bai, Jiarui Liu, Shujie Zhou, Lingzhi Wu, Xiaojie Feng, Pumin Zhang Published: 2024 ABSTRACT BackgroundTriple-negative breast cancer (TNBC) has pronounced stemness that is associated with relapse. N6-methyladenosine (m6A) plays a crucial role in shaping cellular behavior by modulating transcript expression. However, the role of m6A in TNBC stemness, as well as the mechanisms governing its abundance, has yet to be elucidated.MethodsWe analyzed proteomic and transcriptomic data derived from breast cancer cohorts, with an emphasis on m6A regulators. To unravel the role of m6A in TNBC, we employed RNA sequencing, methylated RNA immunoprecipitation sequencing, RNA immunoprecipitation, chromatin immunoprecipitation, and luciferase reporter assays with mesenchymal stem-like (MSL) TNBC models. The clinical relevance was validated using human tissue microarrays and publicly accessible databases.ResultsOur findings indicate that the global level of m6A modification in MSL TNBC is downregulated primarily due to the loss of methyltransferase-like 14 (METTL14). The diminished m6A modification is crucial for the maintenance of TNBC stemness, as it increases the expression of yes-associated protein 1 (YAP1) by blocking YTH domain-containing family protein 2 (YTHDF2)-mediated transcript decay, thereby promoting the activation of Hippo-independent YAP1 signaling. YAP1 is essential for sustaining the stemness regulated by METTL14. Furthermore, we demonstrated that the loss of METTL14 expression results from lysine-specific demethylase 1 (LSD1)-mediated removal of histone H3 lysine 4 methylation at the promoter region, which is critical for LSD1-driven stemness in TNBC.ConclusionThese findings present an epi-transcriptional mechanism that maintains Hippo-independent YAP1 signaling and plays a role in preserving the undifferentiated state of TNBC, which indicates the potential for targeting the LSD1-METTL14 axis to address TNBC stemness.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13046-024-03225-2. Background Triple-negative breast cancer (TNBC) is defined as the absence of estrogen receptor (ER), progesterone receptor (PR), and epidermal growth factor receptor 2 (HER2). While defects in homologous recombination render TNBC more responsive to chemotherapy [1, 2], the risk of early relapse remains higher than that of other subtypes [3], and conventional modalities cannot address the malignancy [4]. Studies based on BRCA1-mutant breast epithelial lineages have suggested that TNBC may develop from abnormal differentiation and mesenchymal transition of luminal progenitor cells [2, 5]. Consistently, TNBC is histologically poorly differentiated and displays enhanced stemness compared to other breast cancer subtypes, as evidenced by a higher proportion of CD44+CD24− and ALDH+ breast cancer stem cells (BCSCs) [6–8]. These cells remain dormant but are equipped with robust self-renewal capabilities and heightened resistance to therapy-induced chromatin damage, ultimately leading to treatment failure [9, 10]. The mesenchymal stem -like (MSL) TNBC subtype, whose transcriptome highly resembles BCSC features, has the worst treatment prognosis [11]. Thus, understanding the regulation of stemness in TNBC is critical for overcoming treatment difficulties, particularly given its reliance on chemo-based regimens. The Hippo signaling pathway exerts a significant suppressive effect on stem cell differentiation and oncogenesis by regulating YAP1 [12, 13]. Normally, the MST/LATS kinase in the Hippo pathway inactivates YAP1 by promoting its phosphorylation and ubiquitination in response to tissue homeostasis and matrix stiffness signals [14], thereby maintaining the normal size and differentiation status of tissues [15]. However, in TNBC, loss of the Hippo pathway results in hyperactivation of YAP1 signaling, which is essential for maintaining basal-like and stem-like properties [16]. This is in stark contrast to ER+ breast cancers, in which YAP1 impairs ER-dependent growth [17, 18], underscoring the regulatory specificity of YAP1 in poorly differentiated breast cancers. Previous studies have investigated the upstream cues that trigger YAP1 hyperactivation [19–22]; however, these efforts focused on the mechanisms that silence the Hippo pathway, with the epigenetic control of YAP1 in cancer being largely unknown. N6-methyladenosine (m6A) is the predominant form of modification in mammalian mRNA. It is a dynamic and reversible process, governed by the 'writer' complex and demethylases [23, 24]. Various readers recognize the m6A mark and exert control over mRNA transcription [25], splicing [26], subcellular distribution [27], stability [28], and translation [29], thereby affecting gene expression and biological processes including carcinogenesis [30]. Methyltransferase-like 14 (METTL14) assumes the role of a navigator in the heterodimer complex with METTL3, responsible for identifying structure-specific m6A motifs and guiding METTL3 to accomplish the N6-methylation of adenosine [31]. Several clinical studies have highlighted the correlation between low intratumoral METTL14 expression and unfavorable prognosis in patients with TNBC [32, 33]; however, the exact role and regulatory mechanism have not been addressed. Here, we unveil the role of m6A in steering the stemness of TNBC. Our findings show that the global level of m6A is markedly downregulated in TNBC primarily due to the loss of METTL14. The enforced METTL14 expression suppresses TNBC stemness both in vitro and in vivo. Mechanistically, the decreased m6A impedes YTHDF2-mediated degradation of YAP1 transcript, leading to increased expression and transactivation of YAP1, independent of the Hippo pathway. Additionally, we identified histone demethylation-mediated transcriptional repression as a key mechanism underlying the loss of METTL14 in TNBC. Materials and methods Cell lines Breast cancer cell lines used in this study were obtained from the ATCC. MDA-MB-231 and Hs578T cells were grown in Dulbecco's modified Eagle's high-glucose medium (Gibco, China) supplemented with 10% fetal bovine serum (FBS, Vazyme, China). MDA-MB-436, SUM159, MCF7, T47D, BT474, and ZR-75–1 cells were cultured in RPMI-1640 medium (Gibco) supplemented with 10% FBS. All cell lines were maintained in a humidified environment at 37 °C with 5% CO2. Mammosphere formation assay Single cells were seeded in ultra-low attachment 96-well or 24-well plates (Corning, USA) at a density of either 1 × 103 or 8 × 103 cells per well using MammoCult™ Human Medium (STEMCELL Technologies, Canada) supplemented with 4 µg/mL heparin (STEMCELL Technologies), and 0.48 µg/mL hydrocortisone (STEMCELL Technologies). Cells were cultured for 4–5 days to form primary spheres. Primary spheres (> 70 μm) were collected and dissociated using 0.25% trypsin–EDTA (Biosharp) for secondary sphere formation. The spheres were counted and photographed using a light field microscope (Olympus, Japan) with a phase-contrast module. Mammosphere formation efficiency (MFE) was calculated as the number of spheres per well divided by the number of cells seeded per well multiplied by 100%. Short interference RNA (siRNA) transfection The siRNAs used in this study were obtained from GenePharma (Suzhou, China). Cells were transfected with siRNAs using Lipofectamine™ RNAiMAX Transfection Reagent (Invitrogen, USA) and Opti-MEM (Gibco) for a period of 48–72 h. Knockdown efficiency was verified by immunoblotting. Detailed information regarding the target sequences of the siRNAs is provided in Suppl. Table 1. Lentiviral transduction Stable depletion of METTL14 was accomplished using lentiviruses carrying DNA oligos encoding specific short hairpin RNAs (shRNAs). These lentiviruses were generated through the co-transfection of HEK293T cells with pMD2.G, psPAX2, and pLKO.1 (shMETTL14, shYAP1, or shLSD1) plasmids, and concentrated using a Universal Virus Concentration Kit (Beyotime, China). The pLKO.1-shcontrol (shNC) served as the negative control. Detailed information regarding the target sequences of the shRNAs is provided in Suppl. Table 1. Stable expression of METTL14 was achieved using lentiviruses harboring either the wild-type or mutant METTL14 coding sequence (CDS). These lentiviruses were produced through the co-transfection of HEK293T cells with pMD2.G, psPAX2, pCDH-M14-WT (or pCDH-M14-R298P) plasmids and concentrated using a Universal Virus Concentration Kit. The pCDH vector was used as the empty control. Cells were transduced with lentivirus in a culture medium supplemented with 8 µg/mL polybrene (Beyotime). After 48 h, the transduced cells were cultured in virus-free medium containing 4 μg/mL puromycin (Beyotime) for 48-h selection and then maintained in medium containing 2 μg/mL puromycin. Knockdown efficiency was verified by immunoblotting. Plasmid transfection The CDS of the gene responsible for expressing a specific protein was synthesized by Sangon Biotech (Shanghai, China) and inserted into the multiple cloning site (MCS) of the pCDH vector. Ligated plasmids were obtained by agarose gel extraction and subjected to antibiotic selection. The endotoxin-free plasmid was introduced into cells using Lipofectamine 2000 Transfection Reagent (Invitrogen) and Opti-MEM (Gibco) and incubated for a period of 48–72 h. Protein expression was confirmed by immunoblotting. Immunoblotting Protein lysates were extracted using RIPA Lysis and Extraction Buffer (Thermo Fisher Scientific) supplemented with 1 × Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific). Nuclear and cytoplasmic proteins were extracted using a Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime). Protein samples were quantified using the Pierce BCA Protein Assay Kit (Thermo Scientific) and denatured in 1 × SDS-PAGE Sample Loading Buffer (Beyotime). For immunoblot analysis, equal amounts of protein (10–20 μg) were separated on a 10% SDS-PAGE gel, transferred to a 0.45 μm Immobilon polyvinylidene fluoride membrane (Millipore, USA), and then blocked with 5% skim milk. After overnight incubation with the primary antibody at 4 °C, the membrane was incubated with horseradish peroxidase (HRP) -conjugated secondary antibody for 1 h at room temperature (RT). The protein lanes were detected using the SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific) and captured using an ImageQuant 800 Western blot CCD imager (Cytiva, Amersham, UK). GAPDH was used as the internal reference for total or cytoplasmic protein, while Lamin B was used for nuclear protein. Information regarding the antibodies used is provided in Suppl. Table 2. Immunohistochemical (IHC) The human TNBC tissue microarray was obtained from Weiao Biotechnology (Shanghai, China). Human tissue microarray for luminal breast cancer was obtained from the tissue bank of the Department of Gynecologic Oncology, Affiliated Cancer Hospital of Zhengzhou University. The slides were deparaffinized with xylene and dehydrated with ethanol. After antigen retrieval with 0.01 M citrate buffer, the sections were blocked with goat serum and incubated with primary antibody at 4 °C overnight. Next, the sections were incubated with HRP-conjugated anti-rabbit secondary antibody for 1 h at RT and stained with the DAB HRP Color Development Kit (Beyotime). The IHC score was determined as the product of the percentage of positive cells and staining intensity (weak = 1; moderate = 2; strong = 3). Information regarding the antibodies used is provided in Suppl. Table 2. Quantitative real-time PCR (qRT-PCR) Total RNA was extracted using the EZ-press RNA Purification Kit (EZBioscience, USA). RNA quality was measured using a NanoDrop One Microvolume UV–Vis Spectrophotometer (Thermo Fisher Scientific). Reverse transcription was performed with 1 μg total RNA using the Script cDNA Synthesis Kit (Bio-Rad, USA), and qRT-PCR was performed with 100 ng cDNA using the SYBR Green Supermix (Bio-Rad) on a CFX96 Real-Time PCR Detection System (Bio-Rad). GAPDH was used as the internal reference. Relative gene expression was calculated using the 2−ΔΔCT method. Primers used for qRT-PCR are listed in Suppl. Table 3. Dot blotting The relative level of mRNA m6A modification was measured by dot blotting. Total mRNA was prepared using the Mag-MK mRNA Purification Kit (Sangon Biotech) and boiled for 3 min at 95 °C. The samples (200, 400, or 800 ng) were then loaded onto nitrocellulose membranes, cross-linked with UV light, and blocked with 5% skim milk. After overnight incubation with a primary antibody against N6-methyladenosine (#56593, CST, USA) at 4 °C, the membrane was incubated with HRP-conjugated anti-rabbit secondary antibody (#7074, CST) for 1 h at RT and stained with 1 × methylene blue solution (Sangon Biotech) for 2 h at RT. The membrane was detected using the SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific) and captured using an ImageQuant 800 Western blot CCD imager. Flow cytometry analysis Cell suspensions were filtered through a 37-μm filter and resuspended in flow cytometry staining buffer (Proteintech, China). Cells were stained with BODIPY-aminoacetaldehyde (STEMCELL Technologies) or fluorescence-conjugated antibodies against BCSC surface markers and analyzed using a CytoFLEX LX Flow Cytometer (Beckman, USA). Information regarding the antibodies used is provided in Suppl. Table 2. Data were processed using FlowJo v10.8.1 software (BD Biosciences, USA). Chemo-/radio-sensitivity assay The chemo-/radiosensitivity of cells was assessed by colony formation assay. Cells were seeded in 6-well plates and treated with different concentrations of doxorubicin or paclitaxel for 48 h, or with a single dose of irradiation. Then these cells were cultured for 7–9 days and stained with a 0.5% crystal violet solution. The colonies (> 50 cells) were counted and photographed. RNA immunoprecipitation (RIP) Cells were collected in Pierce IP lysis buffer (Thermo Fisher Scientific) containing 1 × Halt Protease and Phosphatase Inhibitor Cocktail and 0.1 U/μL RNase inhibitor (Beyotime). Cell lysates were incubated with 4 μg anti-YTHDF2 (24,744–1-AP, Proteintech) or IgG control antibody (30,000–0-AP, Proteintech) at 4 °C overnight in RIP buffer (150 mM KCl, 25 mM Tris, 5 mM EDTA, 0.5 mM DTT, 0.2% IGEPAL CA-630, 1 × Halt Protease and Phosphatase Inhibitor Cocktail and 0.1U/μL RNase inhibitor) with rotation. The samples were then mixed with 50 μL Pierce Protein A/G Magnetic Beads (Thermo Fisher Scientific) at 4 °C for 2 h to capture the antibody-protein-RNA complex. The beads were resuspended in 150 μL RIP buffer containing 0.1% SDS and incubated with 30 μg proteinase K (Beyotime) at 55 °C for 30 min to remove proteins. The supernatant was subjected to RNA extraction and qRT-PCR. Primers used for RIP-qPCR are listed in Suppl. Table 3. RNA stability assay Cells were treated with 5 μg/mL Actinomycin D (Act.D, GlpBio, USA) for 0, 2, 4, 6, and 8 h, and total RNA was extracted using the EZ-press RNA Purification Kit and used for qRT-PCR. The percentage of remaining mRNA was determined as the average Cq at each time point relative to the average Cq at 0 h. The decay rate of the mRNA is shown as an exponential fit. Protein stability assay Cells were treated with 100 μg/mL cycloheximide (CHX; CST, USA) for 0, 1, 3, 6, and 9 h, and total protein was extracted and subjected to immunoblotting. RNA sequencing (RNA-seq) and methylated RIP-sequencing (MeRIP-seq) For RNA-seq, poly(A) mRNA was extracted using the Dynabeads® mRNA DIRECT™ Kit (Thermo Fisher Scientific), and library preparation was performed using the NEBNext® Ultra™ RNA Library Prep Kit (New England Biolabs, USA). Each group was sequenced in duplicate by Illumina NovaSeq 6000 with pair-end 75 cycles. Raw reads were obtained from the sequencing machine and cleaned using fastp v0.18.0 to remove adapters and low-quality bases. Clean reads were aligned to the human reference genome GRCh37/hg19 using HISAT v2.2.4. Differentially expressed genes (DEGs) were analyzed using DESeq2 v1.40.2 with a threshold (p < 0.05 and fold change > 1.5). For MeRIP-seq, poly(A) mRNA was extracted and fragmented using Ambion® RNA Fragmentation Reagents (Thermo Fisher Scientific). m6A immunoprecipitation was performed using fragment RNA and m6A monoclonal antibody (#68,055–1-Ig, Proteintech) according to the RIP procedure. RNA from the input and m6A IP samples was purified using the BeyoMag™ mRNA Purification Kit (Beyotime) and used for library preparation using the NEBNext® Ultra™ RNA Library Prep Kit. The input and m6A IP libraries were sequenced by Illumina HiSeq 4000 with 150 bp paired-end reads. Sequencing data were deposited in the NCBI Gene Expression Omnibus under accession code GSE245282. Raw reads were cleaned and aligned to the human reference genome GRCh37/hg19. Peak calling and annotation were performed using MACS2 v2.1.2 (q < 0.05) and exomePeak2 v1.12.0 (P < 0.00001). The base frequency matrix and motif search were accomplished with Homer v3.0. DiffBind v2.8 software was used to merge peaks among groups, and exomePeak2 was used to obtain the levels of merged peaks in each sample. Differential peaks were identified with FDR < 0.05 and fold change > 2. Sequencing data were deposited in the NCBI Gene Expression Omnibus (GEO) database under accession number GSE245282. Genes of interest were visualized using Integrative Genomics Viewer v2.16.2. Pathway enrichment analysis was performed with the KEGG database. Gene set enrichment analysis was performed with GSEA v4.3.2 and molecular signature databases (https://www.gsea-msigdb.org). Dual-luciferase reporter assay Transcriptional activity of YAP1 was measured by dual-luciferase reporter assay. To this end, the promoter region (-800/+100) of YAP1 was cloned upstream of the modified Firefly luciferase (F-luc) gene into the pGL3-basic vector (Promega, USA) and co-transfected with pRL-TK expressing Renilla luciferase (R-luc). Relative light unit (RLU) was detected 48 h after transfection using a Dual-Luciferase Reporter Gene Assay Kit (Beyotime) on a microplate reader (BioTek Synergy 2, USA). Transcriptional activity was determined by the RLU ratio of firefly to renilla. The translational efficiency of YAP1 mRNA was measured using the pmirGLO vector (Promega) by inserting the exon 10 and 3’UTR of the YAP1 gene downstream of the F-luc gene. The ratio of F-luc to R-luc (reporter protein expression) was determined 24 h after transfection using a Dual-Luciferase Reporter Gene Assay Kit on a multifunctional microplate reader, while the mRNA expression ratio of F-luc to R-luc (reporter mRNA expression) was determined by qRT-PCR. The primers used for reporter mRNA analysis are listed in Suppl. Table 3. Translation efficiency is calculated by the ratio of reporter protein level to mRNA level. The m6A modification sites were investigated using the pmirGLO vector by inserting the exon 10 region of YAP1 containing the wild-type or mutant m6A motif (A was replaced by T) downstream of the F-luc gene. The effect of the point mutation on reporter protein expression was determined by luminescence detection 24 h after transfection. Chromatin immunoprecipitation (CHIP) The CHIP assay was performed using the BeyoChIP Enzymatic ChIP Assay Kit (Beyotime). Cellular genomes and proteins were cross-linked with 1% formaldehyde and collected in 1 × PBS supplemented with 1 × Halt Protease and Phosphatase Inhibitor Cocktail. The genome was digested and fragmented in CHIP buffer containing micrococcal nuclease, and the fragmentation effect was validated by agarose gel electrophoresis. Fragmented samples were mixed with 4 µg anti-H3K4me2 (91,322, Proteintech) or IgG control antibody (30,000–0-AP, Proteintech) overnight at 4 °C with slow rotation, followed by incubation with 30 µL Protein A/G Magnetic Beads/Salmon Sperm DNA with gentle rotation for 3 h at 4 °C. The magnetic beads were eluted with an Elution Buffer to collect the supernatant. The supernatant was heated at 65 °C for 2 h to remove cross-linking and incubated with 0.15 mg/mL proteinase K at 45 °C for 60 min to remove the protein for subsequent RNA extraction and qRT-PCR. Primers used for CHIP-qPCR are listed in Suppl. Table 3. Animal experiments Female athymic (nu/nu) BALB/C mice (5-week old, 18–20 g) were purchased from Gempharmatech Co. Ltd. (Jiangsu, China). The mice were housed in a specific-pathogen-free (SPF) animal facility at the First Affiliated Hospital of Zhejiang University School of Medicine and allowed free access to SPF-grade food and water. All animal procedures were approved by the Animal Care and Use Committee of the First Affiliated Hospital of Zhejiang University School of Medicine. After one-week adaptation, the mice were randomly divided into the indicated groups with five mice per group. Gene-modified cells in the logarithmic growth phase were collected and subcutaneously injected into the right flank of mice at the indicated density in 0.1 mL 1 × PBS to perform the limiting dilution assay (LDA). Tumor size was measured using a caliper every four days, and tumor volume was calculated according to the following formula (length × width2 × 0.5). Tumor incidence was recorded every 4 days, and the frequency of CSCs/CICs in tumors was analyzed using the L-Calc™ software (STEMCELL Technologies). Database analysis Intratumoral protein expression among the different breast cancer subtypes was analyzed using The Cancer Genome Atlas (TCGA) Breast Cancer Proteome (PDC000173) and Clinical Proteomic Tumor Analysis Consortium (CPTAC) Prospective Breast BI Proteome (PDC000120) from the NCI Proteomic Data Commons. Intratumoral gene expression among different breast cancer subtypes was analyzed using TCGA-BRCA dataset from the NCI GDC Data Portal. The correlation between METTL14 expression and the tumor stemness index was evaluated by Pearson’s correlation analysis using the uniformly normalized TCGA pan-cancer datasets from the UCSC Xena and the tumor stemness indices calculated by DNA methylation signatures for each sample from the NCI Genomic Data Commons [34]. Statistical analysis Data are presented as the mean ± SD derived from three biological replicates. Statistical analyses were performed using the Prism 9 software (GraphPad, USA). The differences between two groups were assessed using an unpaired two-tailed Student’s t-test. For multiple comparisons, one-way or two-way analysis of variance (ANOVA) accompanied by Šídák post hoc testing was employed. Statistical significance was set at P < 0.05. Role of funders The funder provided financial support for the study reagents, materials, and salaries of assistant personnel; however, they did not participate in the implementation of the study or dissemination of findings. Results Loss of METTL14 correlates with TNBC stemness We found that MSL TNBC cell lines displayed lower global m6A levels than luminally differentiated ER+ breast cancer cell lines (MCF7 and T47D) and ER+HER2+ breast cancer cell lines (ZR-75–1 and BT474) (Fig. 1a). We reasoned that the diminished m6A might be associated with poor differentiation status of breast cancer. CD44+CD24− or ALDH+ breast cancer stem cells (BCSCs) isolated from multiple breast cancer cell lines confirmed these findings (Fig. 1b and Suppl. Figure 1a). Multivariate Cox regression analysis of prognostic factors for TNBC identified the intratumoral m6A level as an independent factor for overall survival (OS) (HR = 0.29, 95% CI 0.09–0.92, P = 0.04) (Fig. 1c). To determine the underlying mechanism, we analyzed the gene expression of known m6A regulators in breast cancer tissues from The Cancer Genome Atlas (TCGA) and found that METTL14 expression was much lower in TNBC than in normal adjacent and non-TNBC tissues, as compared to other regulators (Suppl. Figure 1b). Similar results were also obtained from GSE21653 [35] (Fig. 1d). We next consulted proteomics data from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) and TCGA-BRCA and found that METTL14 was significantly downregulated in TNBC tissues compared with non-TNBC tissues (Fig. 1e and f). Consistently, the protein expression of METTL14 was diminished in multiple MSL TNBC cell lines compared with that in luminal breast cancer cells (Fig. 1g). Low expression of METTL14 correlated with a higher Scarff-Bloom-Richardson grade (Fig. 1h). Pan-cancer analysis of the tumor stemness index showed that METTL14 expression exhibited the most significant negative correlation with the stemness of breast cancer (r = -0.14, P < 0.0001) [34] (Suppl. Figure 1c). Immunoblotting of METTL14 in CD44+CD24− or ALDH+ BCSCs confirmed this correlation (Fig. 1i). Furthermore, gene set enrichment analysis using the transcriptome from TCGA-BRCA cohort presented a notable enrichment of stem/basal gene signatures in the METTL14 low-expression group [36–39] (Fig. 1j). Survival analysis according to IHC scores of tumor tissues showed that patients with low levels of METTL14 or m6A had a shorter OS duration (Fig. 1k), and this prognostic correlation was also demonstrated by KMplot analysis (Suppl. Figure 1d). These results suggest that the loss of METTL14 correlates with poor differentiation status of breast cancer and may play a role in maintaining the stemness of TNBC.Fig. 1Loss of METTL14 correlates with TNBC stemness. a, b Global m6A levels of mRNA in representative cell lines of human luminal breast cancer (BC), mesenchymal stem cell-like (MSL) TNBC, ALDH+ or CD44+CD24− breast cancer stem cells (BCSCs), and non-BCSCs were determined using dot blotting. Methylene blue (MB) staining was used as the loading control. c Multivariable Cox regression analysis was conducted to determine the hazard ratio for each prognostic factor of TNBC (n = 80 from the human tissue microarray). d Expression of METTL14 in normal adjacent (n = 29), luminal A (n = 89), luminal B (n = 49), HER2+ (n = 24), and basal (n = 75) tumor tissues was obtained from the GSE21653 cohort. Data are shown as the mean ± SD of the normalized microarray expression. e, f Protein expression of METTL3, METTL14, METTL6, WTAP, VIRMA, ZC3H13, RBM15, RBM15B, FTO, and ALKBH5 in normal adjacent (n = 15), non-TNBC (n = 84), TNBC (n = 35), luminal breast cancer (n = 70), and TNBC (n = 23) tissues was obtained from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) and the Cancer Genome Atlas (TCGA). Data are shown as the mean ± SD of normalized protein expression. g Protein expression of METTL14 in representative cell lines of luminal BC and MSL TNBC was determined using immunoblotting. GAPDH was used as the loading control. h Expression of METTL14 in Scarff-Bloom-Richardson grade1 (n = 45), grade2 (n = 89), and grade3 (n = 125) tissues was obtained from the GSE21653 cohort. Data are shown as the mean ± SD of the normalized microarray expression. i Protein expression of METTL14 in BCSCs and non-BCSCs was determined using immunoblotting. GAPDH was used as the loading control. j Enrichment plots show the gene set enrichment analysis of the whole transcriptome between METTL14 low expression (bottom 25%, n = 272) and high expression (top 25%, n = 272) groups in TCGA-BRCA cohort using various stemness-related gene signatures. k Representative images of immunohistochemistry (IHC) staining for METTL14 and m6A in human TNBC tissues. Scale bars: 625 μm. Survival analysis of TNBC patients (n = 80, from the tissue microarray) stratified by METTL14 or m6A IHC score (median) was conducted using the log-rank test Loss of METTL14 correlates with TNBC stemness. a, b Global m6A levels of mRNA in representative cell lines of human luminal breast cancer (BC), mesenchymal stem cell-like (MSL) TNBC, ALDH+ or CD44+CD24− breast cancer stem cells (BCSCs), and non-BCSCs were determined using dot blotting. Methylene blue (MB) staining was used as the loading control. c Multivariable Cox regression analysis was conducted to determine the hazard ratio for each prognostic factor of TNBC (n = 80 from the human tissue microarray). d Expression of METTL14 in normal adjacent (n = 29), luminal A (n = 89), luminal B (n = 49), HER2+ (n = 24), and basal (n = 75) tumor tissues was obtained from the GSE21653 cohort. Data are shown as the mean ± SD of the normalized microarray expression. e, f Protein expression of METTL3, METTL14, METTL6, WTAP, VIRMA, ZC3H13, RBM15, RBM15B, FTO, and ALKBH5 in normal adjacent (n = 15), non-TNBC (n = 84), TNBC (n = 35), luminal breast cancer (n = 70), and TNBC (n = 23) tissues was obtained from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) and the Cancer Genome Atlas (TCGA). Data are shown as the mean ± SD of normalized protein expression. g Protein expression of METTL14 in representative cell lines of luminal BC and MSL TNBC was determined using immunoblotting. GAPDH was used as the loading control. h Expression of METTL14 in Scarff-Bloom-Richardson grade1 (n = 45), grade2 (n = 89), and grade3 (n = 125) tissues was obtained from the GSE21653 cohort. Data are shown as the mean ± SD of the normalized microarray expression. i Protein expression of METTL14 in BCSCs and non-BCSCs was determined using immunoblotting. GAPDH was used as the loading control. j Enrichment plots show the gene set enrichment analysis of the whole transcriptome between METTL14 low expression (bottom 25%, n = 272) and high expression (top 25%, n = 272) groups in TCGA-BRCA cohort using various stemness-related gene signatures. k Representative images of immunohistochemistry (IHC) staining for METTL14 and m6A in human TNBC tissues. Scale bars: 625 μm. Survival analysis of TNBC patients (n = 80, from the tissue microarray) stratified by METTL14 or m6A IHC score (median) was conducted using the log-rank test METTL14-mediated m6A modification suppresses the stem-like trait of TNBC To determine the role of METTL14, we established METTL14-depleted MCF7, MDA-MB-231, and SUM159 cell lines using a lentivirus-mediated hairpin RNA (shRNA) expression system with two different shRNAs targeting METTL14. These stable cell lines displayed efficient depletion of METTL14 (Fig. 4a and b) and global m6A compared to cells transduced with the negative control lentivirus (Suppl. Figure 2a). Mammosphere formation under ultralow attachment culture conditions showed that METTL14 depletion significantly promoted both primary and secondary sphere formation in SUM159 and MDA-MB-231 cells (Fig. 2a). The depletion increased the percentage of ALDH+ BCSCs in SUM159 and MDA-MB-231 cells (Fig. 2b). Chemoresistance is an important hallmark of BCSCs. METTL14 depletion not only enhanced cell resistance to paclitaxel (Fig. 2c) and doxorubicin (Suppl. Figure 2b) but also conferred radioprotective effects on these cells (Suppl. Figure 2c). Furthermore, the proportion of CD49f+ basal-like cells, which mark a BCSC population associated with taxane resistance in TNBC [40], was also increased by METTL14 knockdown (Fig. 2d). We next inoculated immunodeficient mice with decreasing numbers of stable shNC and shM14 cells and evaluated tumor incidence using limiting dilution analysis (LDA) [41]. The results showed that as few as 4 × 104 METTL14-depleted TNBC cells were capable of forming tumors in 2/5 immunodeficient mice compared to 0/5 for control cells (Fig. 2e), suggesting that METTL14 depletion could enrich CSCs in TNBC.Fig. 2METTL14-mediated m6A modification suppresses the stem-like trait of TNBC. a Representative images of the secondary mammospheres of TNBC cells with or without METTL14 depletion. Scale bar: 200 μm (upper panel) and 500 μm (lower panel). b Flow cytometric analysis of ALDH+ cell population in SUM159 and MDA-MB-231 cells with or without METTL14 depletion. c Representative images of colonies formed by SUM159 shNC and shM14 cells treated with paclitaxel (PTX) for 48 h. d Flow cytometric analysis of CD49f+ cell population in SUM159 and MDA-MB-231 cells with or without METTL14 knockdown. e The image shows the number of tumor-bearing mice in the indicated group (n = 5 per group) after inoculating BALB/C mice with shM14 or shNC MDA-MB-231 cells. The cancer-initiating cell (CIC) frequency was quantified using the L-Calc software. f Flow cytometric analysis of ALDH+ cell population in MDA-MB-231 cells overexpressing wild-type (WT) or mutated METTL14. g Representative images of the secondary mammospheres of MDA-MB-231 cells overexpressing WT or mutated METTL14. h The image shows the number of tumor-bearing mice in the indicated group (n = 5 per group) after inoculation of BALB/C mice with MDA-MB-231 cells overexpressing WT or mutated METTL14. Quantification of CIC frequency was performed using L-Calc software. i Representative images of the secondary mammospheres of METTL14-depleted SUM159 cells expressing shRNA-resistant WT or mutated METTL14. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant METTL14-mediated m6A modification suppresses the stem-like trait of TNBC. a Representative images of the secondary mammospheres of TNBC cells with or without METTL14 depletion. Scale bar: 200 μm (upper panel) and 500 μm (lower panel). b Flow cytometric analysis of ALDH+ cell population in SUM159 and MDA-MB-231 cells with or without METTL14 depletion. c Representative images of colonies formed by SUM159 shNC and shM14 cells treated with paclitaxel (PTX) for 48 h. d Flow cytometric analysis of CD49f+ cell population in SUM159 and MDA-MB-231 cells with or without METTL14 knockdown. e The image shows the number of tumor-bearing mice in the indicated group (n = 5 per group) after inoculating BALB/C mice with shM14 or shNC MDA-MB-231 cells. The cancer-initiating cell (CIC) frequency was quantified using the L-Calc software. f Flow cytometric analysis of ALDH+ cell population in MDA-MB-231 cells overexpressing wild-type (WT) or mutated METTL14. g Representative images of the secondary mammospheres of MDA-MB-231 cells overexpressing WT or mutated METTL14. h The image shows the number of tumor-bearing mice in the indicated group (n = 5 per group) after inoculation of BALB/C mice with MDA-MB-231 cells overexpressing WT or mutated METTL14. Quantification of CIC frequency was performed using L-Calc software. i Representative images of the secondary mammospheres of METTL14-depleted SUM159 cells expressing shRNA-resistant WT or mutated METTL14. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant METTL14 is the core component of m6A-installing methyltransferase heterodimer. To investigate whether the role of METTL14 in TNBC is associated with its canonical function, we overexpressed wild-type METTL14 (M14-WT) or a loss-of-function mutant (M14-Mut, R298P) in MDA-MB-231 cells (Suppl. Figure 2d and 4c). The inactive mutant robustly increased the proportion of ALDH+ cells, whereas the wild-type METTL14 reduced this proportion (Fig. 2f). Apparently, the mutant acted in a dominant-negative manner. Similarly, wild-type METTL14 could effectively suppress primary and secondary sphere formation in MDA-MB-231 cells, whereas the mutant enhanced sphere formation (Fig. 2g). The in vivo LDA also demonstrated that the overexpression of wild-type METTL14 decreased, while the inactive METTL14 mutant increased, the tumor incidence (Fig. 2h). In addition, we restored METTL14 expression in METTL14-depleted cells using shRNA-resistant METTL14 (resM14-WT or resM14-Mut) (Fig. 4b). The re-expression of resM14-WT, but not resM14-Mut, restored global m6A level (Suppl. Figure 2a) and reduced the sphere-forming ability of METTL14-depleted cells (Fig. 2i), suggesting that shRNAs targeting METTL14 were on target. These results demonstrate that METTL14 suppresses the stemness of TNBC via its writer function. Identification of the m6A-regulated transcripts involved in TNBC stemness To explore the mechanisms by which METTL14-mediated m6A modification regulates breast cancer stemness, we performed methylated RNA immunoprecipitation (MeRIP) sequencing using MCF7 cells, in which METTL14 is highly expressed. We also retrieved transcriptome sequencing data from MDA-MB-231 cells transfected with short interfering RNAs (siRNAs) targeting METTL14 or control siRNAs from the GSE81164 dataset [42]. A total of 116 upregulated and 60 downregulated genes (P < 0.05, fold change ≥ 1.5) shared by MCF7 and MDA-MB-231 cells were identified after METTL14 depletion (Fig. 3a and Suppl. Data 1). Pathway analysis of these differentially expressed genes using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database showed that METTL14-regulated genes were associated with the Hippo signaling pathway (P = 0.002) (Fig. 3b and Suppl. Data 1).Fig. 3Identification of m6A-regulated transcripts involved in TNBC stemness. a Venn diagrams show the number of significantly upregulated or downregulated genes in MCF7 and MDA-MB-231 cell lines (P < 0.05, fold change ≥ 1.5) upon METTL14 depletion. Intersection shows the number of shared differentially expressed genes (DEGs). b Bar chart shows the top five upregulated or downregulated Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enriched in 176 DEGs. c The predominant consensus m6A motif identified by the HOMER software in the MeRIP-seq library of MCF7 cells. d Metagene plot shows the distribution of m6A peaks in different functional regions of mRNA in MCF7 shNC and shM14 cell lines. e Volcano plot shows the statistical distribution of m6A-hypo genes (diff P < 0. 05, fold change > 2) with decreased mRNA expression (hypo-down) and increased mRNA expression (hypo-up) in MCF7 cells upon METTL14 depletion. f Enrichment circle plot shows the top ten KEGG pathways enriched in m6A-hypo genes. g Venn diagrams show the overlay analysis of shared DEGs with m6A-hypo genes upon METTL14 depletion. Intersection shows the number of shared m6A-hypo DEGs. The heatmap shows the mRNA expression of selected genes across all samples for RNA sequencing. h Venn diagrams show the overlay analysis of shared m6A-hypo DEGs with the core embryonic stem cell (ESC)-like gene module. i Enrichment plots show gene set enrichment analysis of the whole transcriptome in MCF7 and MDA-MB-231 cell lines upon METTL14 depletion using the YAP1-induced target gene set. j Volcano plot shows the expression of YAP1-induced target genes in MCF7 and MDA-MB-231 cells upon METTL14 depletion Identification of m6A-regulated transcripts involved in TNBC stemness. a Venn diagrams show the number of significantly upregulated or downregulated genes in MCF7 and MDA-MB-231 cell lines (P < 0.05, fold change ≥ 1.5) upon METTL14 depletion. Intersection shows the number of shared differentially expressed genes (DEGs). b Bar chart shows the top five upregulated or downregulated Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enriched in 176 DEGs. c The predominant consensus m6A motif identified by the HOMER software in the MeRIP-seq library of MCF7 cells. d Metagene plot shows the distribution of m6A peaks in different functional regions of mRNA in MCF7 shNC and shM14 cell lines. e Volcano plot shows the statistical distribution of m6A-hypo genes (diff P < 0. 05, fold change > 2) with decreased mRNA expression (hypo-down) and increased mRNA expression (hypo-up) in MCF7 cells upon METTL14 depletion. f Enrichment circle plot shows the top ten KEGG pathways enriched in m6A-hypo genes. g Venn diagrams show the overlay analysis of shared DEGs with m6A-hypo genes upon METTL14 depletion. Intersection shows the number of shared m6A-hypo DEGs. The heatmap shows the mRNA expression of selected genes across all samples for RNA sequencing. h Venn diagrams show the overlay analysis of shared m6A-hypo DEGs with the core embryonic stem cell (ESC)-like gene module. i Enrichment plots show gene set enrichment analysis of the whole transcriptome in MCF7 and MDA-MB-231 cell lines upon METTL14 depletion using the YAP1-induced target gene set. j Volcano plot shows the expression of YAP1-induced target genes in MCF7 and MDA-MB-231 cells upon METTL14 depletion In addition, we identified 41,427 m6A peaks in the MeRIP sequencing library of MCF7 cells (Suppl. Figure 3a). These m6A peaks were significantly enriched in the most common RRACH consensus motif (R, purine; A, methylable adenosine; C, cytosine; H, non-guanine base) (Fig. 3c) and were mainly distributed around the stop codon site of the mRNAs (Fig. 3d and Suppl. Figure 3b). Genes with one or more authentic m6A-hypo peaks (diff P < 0.05, fold change > 2) after METTL14 knockdown were termed as m6A-hypo genes. A total of 472 m6A-hypo genes were identified in METTL14-depleted MCF7 cells (Suppl. Data 2). Most of these genes were associated with upregulated mRNA expression following METTL14 depletion (Suppl. Figure 3c), and YAP1 was the most significantly upregulated gene in METTL14-depleted MCF7 cells (Fig. 3e). KEGG pathway analysis of these m6A-hypo genes revealed that the Hippo signaling pathway was greatly affected by m6A alterations mediated by METTL14 (P = 0.025) (Fig. 3f and Suppl. Data 2), which is in line with the pathway analysis of shared differentially expressed genes. Overlap analysis of the m6A-hypo genes and shared differentially expressed genes showed that there were 12 genes potentially regulated by METTL14-mediated m6A modification (Fig. 3g). Thereinto, the expression of heparan sulfate proteoglycan 2 (HSPG2), protocadherin 7 (PCDH7), adrenoceptor β1 (ADRB1), and YAP1 was consistently upregulated in METTL14-depleted MCF7 and MDA-MB-231 cells (Fig. 3g). To determine the dominant gene regulating TNBC stemness, we overlapped these genes with the core embryonic stem cell-like gene module and found that YAP1 was the principal effector of stemness regulated by METTL14-mediated m6A modification [37] (Fig. 3h). Consistently, gene set enrichment analysis of the whole transcriptome from MCF7 and MDA-MB-231 cells revealed that YAP1-induced genes were remarkably enriched in the METTL14-depleted groups [43] (Fig. 3i), and most of these genes were upregulated by METTL14 depletion (Fig. 3j). These data suggest that YAP1 is likely to be regulated by m6A. METTL14 regulates YAP1 expression independently of the Hippo signaling Consistent with the RNA-seq data, depletion of METTL14 expression caused upregulation of YAP1 mRNA and protein (Fig. 4a and b), and this upregulation was abolished by reconstituted expression of shRNA-resistant wild-type METTL14, but not the inactive mutant, in SUM159 cells (Fig. 4b). Results from six breast cancer cell lines transfected with METTL14 siRNAs also showed a consistent upregulation of YAP1 protein (Suppl. Figure 4a). In addition, IHC staining of MDA-MB-231-derived xenografts and analysis of the CPTAC dataset demonstrated a negative correlation between METTL14 and YAP1 (Fig. 4c and d). METTL14 knockdown had no significant effect on the other m6A regulators (Suppl. Figure 4b).Fig. 4METTL14 regulates YAP1 expression independently of the Hippo signaling. a The mRNA expression of YAP1 and METTL14 in breast cancer (BC) cell lines with or without METTL14 depletion was detected using qPCR. b Protein expression of METTL14, YAP1, and NANOG in BC cells with or without METTL14 depletion or rescue of WT METTL14 or the inactive mutant was determined by immunoblotting. GAPDH was used as the loading control. c Representative images of immunohistochemistry (IHC) staining for METTL14 and YAP1 in MDA-MB-231-derived xenografts. Scale bars: 25 μm. d Correlation between METTL14 and YAP1 protein expression in breast tumors from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) dataset was determined using Pearson’s correlation coefficient (n = 134). e, f BC cell lines with or without METTL14 depletion or overexpression of WT METTL14 or the inactive mutant were treated with 5 μg/mL Act.D for the indicated time, and the stability of YAP1 mRNA was detected using qPCR. g Nuclear and cytoplasmic protein expression of YAP1 in BC cells with or without METTL14 depletion was determined by immunoblotting. GAPDH was used as a loading control for cytoplasmic proteins, while Lamin B was used as a loading control for nuclear proteins. h, i YAP1 enrichment at the promoter of YAP1-induced target genes was determined using ChIP-qPCR analysis. j BC cell lines with or without METTL14 depletion, overexpression of WT METTL14, or the inactive mutant were treated with 5 μg/mL CHX for the indicated time, and the stability of YAP1 was determined by immunoblotting. GAPDH was used as the loading control. k Protein expression of p-LATS1, LATS1, active YAP1, and YAP1 in TNBC cells with or without METTL14 depletion was determined by immunoblotting. GAPDH was used as the loading control. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant METTL14 regulates YAP1 expression independently of the Hippo signaling. a The mRNA expression of YAP1 and METTL14 in breast cancer (BC) cell lines with or without METTL14 depletion was detected using qPCR. b Protein expression of METTL14, YAP1, and NANOG in BC cells with or without METTL14 depletion or rescue of WT METTL14 or the inactive mutant was determined by immunoblotting. GAPDH was used as the loading control. c Representative images of immunohistochemistry (IHC) staining for METTL14 and YAP1 in MDA-MB-231-derived xenografts. Scale bars: 25 μm. d Correlation between METTL14 and YAP1 protein expression in breast tumors from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) dataset was determined using Pearson’s correlation coefficient (n = 134). e, f BC cell lines with or without METTL14 depletion or overexpression of WT METTL14 or the inactive mutant were treated with 5 μg/mL Act.D for the indicated time, and the stability of YAP1 mRNA was detected using qPCR. g Nuclear and cytoplasmic protein expression of YAP1 in BC cells with or without METTL14 depletion was determined by immunoblotting. GAPDH was used as a loading control for cytoplasmic proteins, while Lamin B was used as a loading control for nuclear proteins. h, i YAP1 enrichment at the promoter of YAP1-induced target genes was determined using ChIP-qPCR analysis. j BC cell lines with or without METTL14 depletion, overexpression of WT METTL14, or the inactive mutant were treated with 5 μg/mL CHX for the indicated time, and the stability of YAP1 was determined by immunoblotting. GAPDH was used as the loading control. k Protein expression of p-LATS1, LATS1, active YAP1, and YAP1 in TNBC cells with or without METTL14 depletion was determined by immunoblotting. GAPDH was used as the loading control. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant Given that MTTL14 is known to methylate mRNA, we reasoned that the increased YAP1 expression in METTL14-depleted cells most likely resulted from improved mRNA stability. To determine if this was true, we treated breast cancer cells with actinomycin D (Act.D) and examined the decay rate of YAP1 mRNA. The results of qPCR showed that the deletion of METTL14 significantly prolonged the decay time of YAP1 mRNA (Fig. 4e), whereas overexpression of the wild-type, but not the inactive METTL14 mutant, accelerated the decay (Fig. 4f) and decreased YAP1 expression level (Suppl. Figure 4c and d). In addition, we constructed a luciferase reporter plasmid by placing the promoter region (-800 to +100) of YAP1 into the pGL3-Basic vector. Through luciferase reporter assay, we found that METTL14 depletion slightly inhibited YAP1 promoter activity (Suppl. Figure 4e), suggesting that METTL14 did not promote YAP1 transcription. Additionally, METTL14 depletion did not affect the distribution of YAP1 mRNA in the nucleoplasm (Suppl. Figure 4f). However, it cannot be excluded that the translational efficiency of YAP1 mRNA was promoted by METTL14 depletion [44]. To determine the effect of METTL14 on YAP1 mRNA translation, we constructed a pmirGLO reporter with m6A-modified regions (exon 10 and 3’UTR) of YAP1 [29]. Luciferase reporter assay and qPCR of the luciferase gene showed that METTL14 depletion slightly decreased the translational efficiency of YAP1 mRNA (Suppl. Figure 4 g). These results suggest that METTL14 depletion upregulates YAP1 expression by stabilizing its mRNA. We next asked if METTL14 depletion-induced YAP1 is functioning and thus examined the transactivity of YAP1. As shown in Fig. 4g, both cytoplasmic and nuclear fractions of YAP1 increased in METTL14-depleted cells. More importantly, not only did nuclear translocation increase upon METTL14 depletion, but also its transcriptional activity, as chromatin immunoprecipitation (ChIP) analysis showed that YAP1 was significantly enriched at the promoters of target genes [43], such as IL6, CTGF, CYR61, PXCDH7, and FST (Fig. 4h), which is consistent with the increased expression of these genes identified by RNA-seq (Suppl. Data 1). In contrast, overexpression of wild-type METTL14 led to a significant reduction in YAP1 occupancy at these genes; however, the inactive METTL14 mutant failed to generate similar results (Fig. 4i). The Hippo pathway is the most important signal regulating YAP1 protein expression. Kinases in this pathway are often lost or inactivated in TNBC, leading to hyperactivated YAP1 signaling (Suppl. Figure 4 h) [19]. However, we found that the protein stability of YAP1 was not affected by METTL14 depletion (Fig. 4j). METTL14 depletion resulted in an increase in both total YAP1 and active YAP1 expression without influencing the activity of LATS1 (Fig. 4k), suggesting that the role of METTL14-mediated m6A modification in modulating YAP1 expression is independent of Hippo signaling. The m6A-modified 3’UTR triggers YTHDF2-dependent YAP1 mRNA decay We next sought to investigate the exact m6A site and reader that determines the stability of YAP1 mRNA. Indeed, MeRIP sequencing data from MCF7 and MDA-MB-231 cells showed a specific m6A signal around the stop codon of YAP1 (Fig. 5a and Suppl. Figure 5a). In contrast, no m6A signal was detected on the NANOG gene (Suppl. Figure 5b). Notably, there was a loss of m6A peak on the 3’UTR of mRNA, relative to the m6A peak on exon 10 in MDA-MB-231 cells, which is distinct from MCF7 shNC cells but similar to METTL14-depleted cells (Fig. 5b). Also, YAP1 mRNA expression and stability in MDA-MB-231 cells were significantly higher than those in MCF7 cells (Suppl. Figure 5c and d), implying that this loss might play a role in determining YAP1 expression. We next identified three potentially methylated adenosine bases within the specific m6A peak using a sequence-based site predictor [45] (Fig. 5c and Suppl. Data 3). To determine the effects of these m6A sites on YAP1 expression, we constructed a luciferase gene reporter with the exon 10 and 3’UTR regions of YAP1 containing wild-type or mutated m6A motifs (Fig. 5d). The reporter assay showed that METTL14 overexpression significantly reduced luciferase activity with the wild-type, Mut1, and Mut2 fragments, whereas the Mut3 fragment conferred resistance to METTL14 overexpression (Fig. 5e). These findings suggest that m6A-modified 3’UTR suppresses YAP1 expression.Fig. 5The m6A-modified 3’UTR triggers YTHDF2-dependent YAP1 mRNA decay. a IGV shows the m6A signal in YAP1 of MCF7 cells with or without METTL14 depletion. b IGV shows the specific m6A peaks of YAP1 in MCF7 and MDA-MB-231 cells. The gray square indicates the m6A peak downregulated by METTL14 loss. c Mapping of m6A modification sites on YAP1 mRNA by SRAMP. Red arrows indicate m6A sites identified by MeRIP sequencing data. d Scheme shows m6A motif positions within YAP1 mRNA and their artificial mutations for luciferase reporter assays. e The pmirGLO reporter with WT or mutated exon 10–3’UTR region of YAP1 was transfected into MDA-MB-231 cells with or without METTL14 overexpression for 24 h, and the relative F-luc/R-luc ratio was determined using luciferase reporter assays. f Protein expression of YTHDF2, METTL14, and YAP1 in MCF7 and METTL14-overexpressing MDA-MB-231 cells, with or without YTHDF2 knockdown, was detected by immunoblotting. GAPDH was used as the loading control. g IGV shows the YTHDF2 binding signals in YAP1 of MDA-MB-231, MCF7, and SK-BR3 cells from the GSE137258 dataset. The square indicates the YTHDF2 binding peaks overlapped with the specific m6A peaks. h The mRNA expression of YAP1 in MCF7 and METTL14-overexpressing MDA-MB-231 cells, with or without YTHDF2 knockdown, was detected using qPCR. i MCF7 and METTL14-overexpressing MDA-MB-231 cells with or without YTHDF2 knockdown were treated with 5 μg/mL Act.D for the indicated time. The stability of YAP1 mRNA was determined by qPCR. j RIP was performed using anti-IgG control or anti-YTHDF2 antibodies in MDA-MB-231 cells with or without METTL14 overexpression, followed by qPCR analysis of YAP1 mRNA. k RIP was performed with anti-IgG control or anti-YTHDF2 antibodies in METTL14-overexpressing MDA-MB-231 cells with WT or mutated YAP1 overexpression, followed by qPCR analysis of YAP1 mRNA. l The pmirGLO reporter with WT or mutated exon10-3’UTR region of YAP1 was transfected into MDA-MB-231 cells with or without YTHDF2 overexpression for 24 h, and the relative F-luc/R-luc ratio was determined using a dual-luciferase reporter assay. m The mRNA expression of YAP1 in MCF7 and MDA-MB-231 cells with or without YTHDF2 overexpression was detected by qPCR. n MCF7 and MDA-MB-231 cells with or without YTHDF2 overexpression were treated with 5 μg/mL Act.D for the indicated time. The stability of YAP1 mRNA was determined by qPCR. o Protein expression of YTHDF2 and YAP1 in MCF7 and MDA-MB-231 cells with or without YTHDF2 overexpression was detected using immunoblotting. GAPDH was used as the loading control. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant The m6A-modified 3’UTR triggers YTHDF2-dependent YAP1 mRNA decay. a IGV shows the m6A signal in YAP1 of MCF7 cells with or without METTL14 depletion. b IGV shows the specific m6A peaks of YAP1 in MCF7 and MDA-MB-231 cells. The gray square indicates the m6A peak downregulated by METTL14 loss. c Mapping of m6A modification sites on YAP1 mRNA by SRAMP. Red arrows indicate m6A sites identified by MeRIP sequencing data. d Scheme shows m6A motif positions within YAP1 mRNA and their artificial mutations for luciferase reporter assays. e The pmirGLO reporter with WT or mutated exon 10–3’UTR region of YAP1 was transfected into MDA-MB-231 cells with or without METTL14 overexpression for 24 h, and the relative F-luc/R-luc ratio was determined using luciferase reporter assays. f Protein expression of YTHDF2, METTL14, and YAP1 in MCF7 and METTL14-overexpressing MDA-MB-231 cells, with or without YTHDF2 knockdown, was detected by immunoblotting. GAPDH was used as the loading control. g IGV shows the YTHDF2 binding signals in YAP1 of MDA-MB-231, MCF7, and SK-BR3 cells from the GSE137258 dataset. The square indicates the YTHDF2 binding peaks overlapped with the specific m6A peaks. h The mRNA expression of YAP1 in MCF7 and METTL14-overexpressing MDA-MB-231 cells, with or without YTHDF2 knockdown, was detected using qPCR. i MCF7 and METTL14-overexpressing MDA-MB-231 cells with or without YTHDF2 knockdown were treated with 5 μg/mL Act.D for the indicated time. The stability of YAP1 mRNA was determined by qPCR. j RIP was performed using anti-IgG control or anti-YTHDF2 antibodies in MDA-MB-231 cells with or without METTL14 overexpression, followed by qPCR analysis of YAP1 mRNA. k RIP was performed with anti-IgG control or anti-YTHDF2 antibodies in METTL14-overexpressing MDA-MB-231 cells with WT or mutated YAP1 overexpression, followed by qPCR analysis of YAP1 mRNA. l The pmirGLO reporter with WT or mutated exon10-3’UTR region of YAP1 was transfected into MDA-MB-231 cells with or without YTHDF2 overexpression for 24 h, and the relative F-luc/R-luc ratio was determined using a dual-luciferase reporter assay. m The mRNA expression of YAP1 in MCF7 and MDA-MB-231 cells with or without YTHDF2 overexpression was detected by qPCR. n MCF7 and MDA-MB-231 cells with or without YTHDF2 overexpression were treated with 5 μg/mL Act.D for the indicated time. The stability of YAP1 mRNA was determined by qPCR. o Protein expression of YTHDF2 and YAP1 in MCF7 and MDA-MB-231 cells with or without YTHDF2 overexpression was detected using immunoblotting. GAPDH was used as the loading control. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant YT521-B homology (YTH) domain-containing family proteins, including YTHDF1, YTHDF2, and YTHDF3, are m6A readers with the ability to degrade mRNA. We speculated that these readers regulate YAP1 expression by recognizing m6A marks. As expected, YTHDF2 depletion notably increased YAP1 protein expression in MCF7 and METTL14-overexpressing MDA-MB-231 cells, without affecting METTL14 (Fig. 5f and Suppl. Figure 5e). Analysis of YTHDF2 cross-linking and immunoprecipitation datasets across multiple breast cancer cell lines showed that the binding signals of YTHDF2 covered the m6A-specific peak around the stop codon of YAP1 [46] (Fig. 5g). Consistently, YTHDF2 depletion led to a robust increase in the expression and stability of YAP1 mRNA (Fig. 5h and 5i). Furthermore, METTL14 overexpression increased YTHDF2 binding to YAP1 mRNA (Fig. 5j), whereas overexpression of mutated YAP1 decreased this binding (Fig. 5k). YTHDF2 overexpression reduced luciferase activity driven by the wild-type exon 10–3’UTR fragment, while the mutated fragment rendered luciferase activity resistant to overexpression (Fig. 5l). In addition, YTHDF2 overexpression reduced the expression and stability of YAP1 mRNA (Fig. 5m and 5n) and led to a decrease in YAP1 protein (Fig. 5o). These results suggest that YTHDF2 is responsible for degrading YAP1 mRNA by recognizing m6A-modified 3’UTR. YAP1 underlies the m6A-regulated stemness in TNBC YAP1 is a specific regulator of stem-like properties in basal-like breast cancer cells [43]. YAP1 knockdown resulted in decreased NANOG protein level (Suppl. Figure 6) and impaired the sphere-forming ability of MDA-MB-231 and SUM159 cells with or without METTL14 depletion (Fig. 6a and b). To determine the essentiality of YAP1 as an m6A-modified target in maintaining TNBC stemness, we used siRNAs to knock down YAP1 expression in METTL14-depleted TNBC cells to a level comparable to that in shNC cells (Fig. 6c). The knockdown reduced NANOG protein expression (Fig. 6c) and stemness of METTL14-depleted TNBC cells to levels comparable to shNC cells (Fig. 6d and e). To examine whether the essentiality is a consequence of the transforming property of stable METTL14 depletion, we generated stable SUM159 and MDA-MB-231 cell lines expressing shRNAs targeting YAP1 (shYAP1). Consistent with YAP1 knockdown by siRNA transfection (Suppl. Figure 6), stable YAP1 depletion modestly decreased the protein expression of NANOG in SUM159 and MDA-MB-231 cells (Fig. 6f) and disrupted their potential to form spheres (Fig. 6g). Knockdown of METTL14 in YAP1-depleted SUM159 and MDA-MB-231 cells failed to improve their sphere-forming capability (Fig. 6g), though an increase in NANOG protein level was observed (Fig. 6f). These results indicate that YAP1 is essential for METTL14-regulated stemness in TNBC.Fig. 6YAP1 underlies the m6A-regulated stemness in TNBC. a Representative images of the primary mammospheres of SUM159 and MDA-MB-231 shNC cells with or without YAP1 knockdown. Scale bar: 500 μm. Data are shown as the mean ± SD of twelve replicate samples. b Representative images of primary mammospheres of METTL14-depleted SUM159 and MDA-MB-231 cells with or without YAP1 knockdown. Scale bar: 500 μm. Data are shown as the mean ± SD of twelve replicate samples. c Protein expression of METTL14, YAP1, and NANOG in METTL14-depleted TNBC cells with or without YAP1 knockdown was determined by immunoblotting. GAPDH was used as the loading control. d Representative images of the primary mammospheres of METTL14-depleted TNBC cell lines with or without YAP1 knockdown. Scale bar: 500 μm. e Flow cytometric analysis of ALDH+ cells in METTL14-depleted TNBC cell lines with or without YAP1 knockdown. f Protein expression of METTL14, YAP1, and NANOG in YAP1-depleted TNBC cells with or without METTL14 knockdown was determined by immunoblotting. GAPDH was used as the loading control. g The stemness of YAP1-depleted TNBC cells with or without METTL14 knockdown was detected using in vitro limiting dilution assays. h Protein expression of METTL14, YAP1, and NANOG in METTL14-overexpressed MDA-MB-231 cells with or without ectopic YAP1 expression was determined by immunoblotting. GAPDH was used as the loading control. i Representative images of primary mammospheres of METTL14-overexpressing and control MDA-MB-231 cells with or without ectopic YAP1 expression. Scale bar: 500 μm. j Flow cytometric analysis of ALDH+ cells in METTL14-overexpressing and control MDA-MB-231 cells with or without ectopic YAP1 expression. Unless otherwise stated, data are presented as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant YAP1 underlies the m6A-regulated stemness in TNBC. a Representative images of the primary mammospheres of SUM159 and MDA-MB-231 shNC cells with or without YAP1 knockdown. Scale bar: 500 μm. Data are shown as the mean ± SD of twelve replicate samples. b Representative images of primary mammospheres of METTL14-depleted SUM159 and MDA-MB-231 cells with or without YAP1 knockdown. Scale bar: 500 μm. Data are shown as the mean ± SD of twelve replicate samples. c Protein expression of METTL14, YAP1, and NANOG in METTL14-depleted TNBC cells with or without YAP1 knockdown was determined by immunoblotting. GAPDH was used as the loading control. d Representative images of the primary mammospheres of METTL14-depleted TNBC cell lines with or without YAP1 knockdown. Scale bar: 500 μm. e Flow cytometric analysis of ALDH+ cells in METTL14-depleted TNBC cell lines with or without YAP1 knockdown. f Protein expression of METTL14, YAP1, and NANOG in YAP1-depleted TNBC cells with or without METTL14 knockdown was determined by immunoblotting. GAPDH was used as the loading control. g The stemness of YAP1-depleted TNBC cells with or without METTL14 knockdown was detected using in vitro limiting dilution assays. h Protein expression of METTL14, YAP1, and NANOG in METTL14-overexpressed MDA-MB-231 cells with or without ectopic YAP1 expression was determined by immunoblotting. GAPDH was used as the loading control. i Representative images of primary mammospheres of METTL14-overexpressing and control MDA-MB-231 cells with or without ectopic YAP1 expression. Scale bar: 500 μm. j Flow cytometric analysis of ALDH+ cells in METTL14-overexpressing and control MDA-MB-231 cells with or without ectopic YAP1 expression. Unless otherwise stated, data are presented as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant We next investigated whether restoring YAP1 was sufficient to reverse METTL14-induced inhibition of stemness by transiently expressing YAP1 in METTL14-overexpressing MDA-MB-231 cells. Immunoblotting analysis showed that ectopic expression of YAP1 failed to rescue the protein expression of NANOG (Fig. 6h), suggesting that NANOG is not a target of YAP1. Since NANOG is not m6A-modified either, it is likely a marker associated with METTL14-regulated TNBC stemness. Furthermore, ectopic expression of YAP1 failed to restore the stemness of METTL14-overexpressing MDA-MB-231 cells (Fig. 6i and j). These findings indicate that YAP1 is necessary, but not sufficient, for m6A-regulated TNBC stemness. The expression of METTL14 is transcriptionally suppressed by LSD1 METTL14 expression was significantly lower in TNBC than in luminal cancer (Suppl. Figure 1b). Similar results were also observed in BCSCs (Fig. 7a). Hence, we speculated that low expression of METTL14 might result from transcriptional repression. To identify the factors that govern the transcription of METTL14, we consulted the CHIP-seq data accessible in the CHIP-Atlas data-mining suite and found that lysine-specific demethylase 1 (LSD1, also known as KDM1A) displayed the highest binding abundance at the promoter region of METTL14 [47] (Fig. 7b and Suppl. Data 4). In contrast to the expression pattern of METTL14 (Fig. 1f), LSD1 was highly expressed in ALDH+ or CD44+CD24− BCSCs (Fig. 7c). Furthermore, LSD1 knockdown by two different siRNAs resulted in a marked elevation of METTL14 mRNA in MDA-MB-231 and SUM159 cells (Fig. 7d). METTL14, as well as global m6A levels in MDA-MB-231 and SUM159 cells, were also significantly upregulated by LSD1 knockdown (Fig. 7e). However, the other m6A regulators were not affected by LSD1 depletion (Suppl. Figure 7a). In addition, LSD1 overexpression in MDA-MB-231 and MCF7 cell lines effectively reduced the mRNA and protein expression of METTL14 as well as global m6A modification (Fig. 7f and g). Immunostaining of LSD1 in human TNBC tissues showed that the intratumoral protein level of LSD1 was inversely correlated with that of METTL14 (R = -0.3, P = 0.0068) (Fig. 7h). These results suggest that the loss of METTL14 in TNBC and BCSCs may be due to the high expression of LSD1.Fig. 7The expression of METTL14 is transcriptionally suppressed by LSD1. a The mRNA expression of METTL14 in breast cancer stem cells (BCSCs) and non-BCSCs was detected using qPCR. b Analysis of transcription and chromatin regulators of METTL14 in breast cells using the CHIP-Atlas data-mining suite. c Protein expression of LSD1 and METTL14 in BCSCs and non-BCSCs was determined using immunoblotting. GAPDH was used as the loading control. d The mRNA expression of METTL14 in MDA-MB-231 and SUM159 cells with or without LSD1 knockdown was detected by qPCR. e Protein expression of LSD1 and METTL14 in MDA-MB-231 and SUM159 cells with or without LSD1 knockdown was determined by immunoblotting. GAPDH was used as the loading control. The m6A levels of global mRNA in these cells were determined using dot blotting. Methylene blue (MB) staining was used as the loading control. f The mRNA expression of METTL14 in MDA-MB-231 and MCF7 cells with or without LSD1 overexpression was detected by qPCR. g Protein expression of LSD1 and METTL14 in MDA-MB-231 and MCF7 cells with or without LSD1 overexpression was detected by immunoblotting. GAPDH was used as the loading control. Global m6A levels of mRNA in these cells were determined using dot blotting. MB staining was used as a loading control. h Representative images of immunohistochemical (IHC) staining for METTL14 and LSD1 in human TNBC tissues (n = 80). Scale bars: 625 and 100 μm. IHC staining score was calculated, and the correlation was determined using the Pearson correlation coefficient. i The upper IGV visualization shows LSD1 and H3K4me2 signals on the METTL14 promoter of T47D cells with or without LSD1 knockout from the GSE168644 dataset. The lower IGV visualization shows LSD1 and H3K4me2 signals at the METTL14 promoter of MCF7 cells with or without LSD1 knockdown from the GSE112230 dataset. j The binding abundance of H3K4me2 at the METTL14 promoter in MDA-MB-231 and SUM159 cells with or without LSD1 knockdown was determined by ChIP-qPCR. k The binding abundance of H3K4me2 at the METTL14 promoter in MDA-MB-231 and MCF7 cells with or without LSD1 overexpression was determined by ChIP-qPCR. Data are presented as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant The expression of METTL14 is transcriptionally suppressed by LSD1. a The mRNA expression of METTL14 in breast cancer stem cells (BCSCs) and non-BCSCs was detected using qPCR. b Analysis of transcription and chromatin regulators of METTL14 in breast cells using the CHIP-Atlas data-mining suite. c Protein expression of LSD1 and METTL14 in BCSCs and non-BCSCs was determined using immunoblotting. GAPDH was used as the loading control. d The mRNA expression of METTL14 in MDA-MB-231 and SUM159 cells with or without LSD1 knockdown was detected by qPCR. e Protein expression of LSD1 and METTL14 in MDA-MB-231 and SUM159 cells with or without LSD1 knockdown was determined by immunoblotting. GAPDH was used as the loading control. The m6A levels of global mRNA in these cells were determined using dot blotting. Methylene blue (MB) staining was used as the loading control. f The mRNA expression of METTL14 in MDA-MB-231 and MCF7 cells with or without LSD1 overexpression was detected by qPCR. g Protein expression of LSD1 and METTL14 in MDA-MB-231 and MCF7 cells with or without LSD1 overexpression was detected by immunoblotting. GAPDH was used as the loading control. Global m6A levels of mRNA in these cells were determined using dot blotting. MB staining was used as a loading control. h Representative images of immunohistochemical (IHC) staining for METTL14 and LSD1 in human TNBC tissues (n = 80). Scale bars: 625 and 100 μm. IHC staining score was calculated, and the correlation was determined using the Pearson correlation coefficient. i The upper IGV visualization shows LSD1 and H3K4me2 signals on the METTL14 promoter of T47D cells with or without LSD1 knockout from the GSE168644 dataset. The lower IGV visualization shows LSD1 and H3K4me2 signals at the METTL14 promoter of MCF7 cells with or without LSD1 knockdown from the GSE112230 dataset. j The binding abundance of H3K4me2 at the METTL14 promoter in MDA-MB-231 and SUM159 cells with or without LSD1 knockdown was determined by ChIP-qPCR. k The binding abundance of H3K4me2 at the METTL14 promoter in MDA-MB-231 and MCF7 cells with or without LSD1 overexpression was determined by ChIP-qPCR. Data are presented as the mean ± SD for a representative of three independent experiments performed in triplicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant LSD1 is a well-known transcription repressor that demethylates mono- or di-methylated histone H3 lysine 4 (H3K4me1/2) [48]. Analysis of the CHIP-seq datasets (GSE168644 and GSE112230) derived from various breast cancer cell lines showed that LSD1 and H3K4me2 specifically bind to the transcription start site of METTL14 (Fig. 7i), whereas H3K4me1 was not found (Suppl. Figure 7b). Notably, when LSD1 was knocked down or knocked out in MCF7 and T47D cell lines, respectively, there was a consistent increase in the abundance of H3K4me2 at the METTL14 promoter (Fig. 7i), suggesting that LSD1 may control METTL14 expression by regulating H3K4me2 levels. To further confirm the role of LSD1, we performed CHIP-qPCR analysis in TNBC cells using specific primers targeting the H3K4me2 binding region within the METTL14 promoter. The results showed that LSD1 depletion led to a substantial increase in the abundance of H3K4me2 at the METTL14 promoter (Fig. 7j), whereas overexpression of LSD1 resulted in a decrease (Fig. 7k). These findings provide compelling evidence that LSD1 acts as a negative regulator of METTL14 expression in breast cancer. Loss of METTL14 is critical for LSD1-driven stemness in TNBC LSD1 plays a crucial role in the growth and metastasis of breast cancer [49, 50]. Analysis of the GSE21653 and TCGA breast cancer cohort showed that LSD1 was highly expressed in breast cancer with basal-like traits (Fig. 8a and b). Data from the CPTAC datasets also showed that TNBC exhibited much higher LSD1 protein expression than normal breast and luminal cancer tissues (Fig. 8c). Also, high expression of LSD1 was associated with a higher Scarff-Bloom-Richardson grade in breast cancer (Fig. 8d). The high intratumoral protein level of LSD1 correlated with a shorter OS time in patients with TNBC (Fig. 8e). These results support the association of LSD1 with a poorly differentiated state and worse prognosis in breast cancer.Fig. 8Loss of METTL14 is critical for LSD1-driven stemness in TNBC. a Expression of LSD1 in normal adjacent (n = 29), luminal A (n = 89), luminal B (n = 49), HER2+ (n = 24), and basal tumor (n = 75) tissues from the GSE21635 cohort. Data are shown as the mean ± SD of the normalized microarray expression. b Expression of LSD1 in normal adjacent (n = 113), non-TNBC (n = 860), HER2+ (n = 38), and TNBC (n = 192) tissues from The Cancer Genome Atlas (TCGA) cohort. Data are presented as the mean ± SD of the normalized read counts. c Protein expression of LSD1 in normal adjacent (n = 15), luminal (n = 75), HER2+ (n = 9), and TNBC (n = 15) tissues was obtained from the Clinical Proteomic Tumor Analysis Consortium (CPTAC). Data are shown as the mean ± SD of normalized protein expression. d Expression of LSD1 in Scarff-Bloom-Richardson grade1 (n = 45), grade2 (n = 89), and grade3 (n = 125) tissues from the GSE21635 cohort. Data are shown as the mean ± SD of the normalized microarray expression. e Representative images of immunohistochemical (IHC) staining for LSD1 in human TNBC tissues. Scale bars: 625 μm. Survival analysis of TNBC patients (n = 80, from the tissue microarray) stratified by LSD1 IHC score (median) was conducted using the log-rank test. f Protein expression of LSD1, METTL14, YAP1, and NANOG in TNBC cells with or without LSD1 overexpression and METTL14 rescue was detected by immunoblotting. GAPDH was used as the loading control. g Representative images of the primary mammospheres of TNBC cells with or without LSD1 overexpression and METTL14 rescue. Scale bar: 500 μm. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. ***P < 0.001. h Flow cytometric analysis of ALDH+ cell population in TNBC cells with or without LSD1 overexpression and METTL14 rescue. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. ***P < 0.001. i In vivo limiting dilution assay shows the frequency of cancer initiation cells (CICs) in MDA-MB-231 cells with or without LSD1 overexpression and METTL14 rescue (n = 5 per group). Quantification of CIC frequency was performed using L-Calc software. j Representative images of IHC staining for LSD1, METTL14, and YAP1 in MDA-MB-231-derived xenografts. Scale bars: 25 μm. k LSD1 exhibits elevated expression levels in TNBC and functions to repress the transcription of METTL14 by demethylating H3K4me2 at the promoter region. This demethylation leads to a decrease in METTL14 expression and subsequent reduction in global m6A levels. Loss of m6A prevents the recognition and degradation of YAP1 mRNA by YTHDF2, thereby sustaining the hyperactivation of YAP1 signaling, which is a key factor contributing to the pronounced stemness observed in TNBC Loss of METTL14 is critical for LSD1-driven stemness in TNBC. a Expression of LSD1 in normal adjacent (n = 29), luminal A (n = 89), luminal B (n = 49), HER2+ (n = 24), and basal tumor (n = 75) tissues from the GSE21635 cohort. Data are shown as the mean ± SD of the normalized microarray expression. b Expression of LSD1 in normal adjacent (n = 113), non-TNBC (n = 860), HER2+ (n = 38), and TNBC (n = 192) tissues from The Cancer Genome Atlas (TCGA) cohort. Data are presented as the mean ± SD of the normalized read counts. c Protein expression of LSD1 in normal adjacent (n = 15), luminal (n = 75), HER2+ (n = 9), and TNBC (n = 15) tissues was obtained from the Clinical Proteomic Tumor Analysis Consortium (CPTAC). Data are shown as the mean ± SD of normalized protein expression. d Expression of LSD1 in Scarff-Bloom-Richardson grade1 (n = 45), grade2 (n = 89), and grade3 (n = 125) tissues from the GSE21635 cohort. Data are shown as the mean ± SD of the normalized microarray expression. e Representative images of immunohistochemical (IHC) staining for LSD1 in human TNBC tissues. Scale bars: 625 μm. Survival analysis of TNBC patients (n = 80, from the tissue microarray) stratified by LSD1 IHC score (median) was conducted using the log-rank test. f Protein expression of LSD1, METTL14, YAP1, and NANOG in TNBC cells with or without LSD1 overexpression and METTL14 rescue was detected by immunoblotting. GAPDH was used as the loading control. g Representative images of the primary mammospheres of TNBC cells with or without LSD1 overexpression and METTL14 rescue. Scale bar: 500 μm. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. ***P < 0.001. h Flow cytometric analysis of ALDH+ cell population in TNBC cells with or without LSD1 overexpression and METTL14 rescue. Data are shown as the mean ± SD for a representative of three independent experiments performed in triplicate. ***P < 0.001. i In vivo limiting dilution assay shows the frequency of cancer initiation cells (CICs) in MDA-MB-231 cells with or without LSD1 overexpression and METTL14 rescue (n = 5 per group). Quantification of CIC frequency was performed using L-Calc software. j Representative images of IHC staining for LSD1, METTL14, and YAP1 in MDA-MB-231-derived xenografts. Scale bars: 25 μm. k LSD1 exhibits elevated expression levels in TNBC and functions to repress the transcription of METTL14 by demethylating H3K4me2 at the promoter region. This demethylation leads to a decrease in METTL14 expression and subsequent reduction in global m6A levels. Loss of m6A prevents the recognition and degradation of YAP1 mRNA by YTHDF2, thereby sustaining the hyperactivation of YAP1 signaling, which is a key factor contributing to the pronounced stemness observed in TNBC We next asked whether METTL14 was critical for LSD1-driven stemness. To this end, we overexpressed LSD1 with or without rescuing METTL14 expression in SUM159 and MDA-MB-231 cells. As expected, LSD1 overexpression reduced the endogenous protein expression of METTL14 but increased that of YAP1 and NANOG (Fig. 8f). Rescue of METTL14 in these cells abolished the inducible effects of LSD1 on YAP1 and NANOG (Fig. 8f). Mammosphere formation assays showed that the sphere-forming capacity of TNBC cells was significantly enhanced by LSD1 overexpression; however, co-expression with METTL14 eliminated this effect (Fig. 8g). Flow cytometric analysis of ALDH+ cells also demonstrated that METTL14 rescue compromised the inducible effect of LSD1 on stemness (Fig. 8h). Consistently, in vivo LDA showed that the frequency of CIC in METTL14 rescue group was remarkably decreased compared to LSD1 overexpression alone (Fig. 8i). IHC staining of xenografts derived from these cells confirmed the essential role of METTL14 loss in LSD1-induced YAP1 expression (Fig. 8j). These results suggest that the function of LSD1 in promoting TNBC stemness is largely mediated by repressing METTL14 expression. Discussion We show that the loss of METTL14, resulting from LSD1-mediated histone demethylation, favors the stemness of TNBC by interfering with YTHDF2-mediated degradation of YAP1 transcript (Fig. 8k). The role of METLL14 in cancer is multi-faceted. For example, in leukemia, METTL14 was shown to promote stemness by inducing MYC and MYB expression [51]. Similarly, METTL14 promotes tumorigenicity and chemoresistance in osteosarcoma by inducing MN1 expression [52]. In contrast to its oncogenic effect, METTL14 was found to trigger ferroptosis in endocervical cancer [53], impede the tumor growth of ocular melanoma [54], and promote genome-wide repair to prevent skin cancer [55], acting as a tumor suppressor. The diverse functions of METTL14 in these cancers may be attributed to tissue-specific expression or regulation [30]. However, the role of METTL14 in breast cancer remains controversial. While one study showed that suppressing METTL14 with siRNA impaired colony formation and invasion of TNBC cells [42], two recent clinical investigations indicated that low expression of METTL14 in breast cancer tissue is associated with shorter survival time of patients and is an independent unfavorable indicator for TNBC [32, 33]. Despite the fact that the cohorts in these studies are derived from various public datasets and regions and underwent distinct prognostic evaluations, the collective findings indicate that the loss of METTL14 is associated with an unfavorable prognosis in TNBC. Our analysis of proteomic data from breast cancer, sourced from CPTAC and TCGA, demonstrated that METTL14 is consistently and significantly downregulated in TNBC, aligning with the observed decrease in m6A levels. In addition, it was reported that METTL14 can induce H3K27me3 demethylation by enlisting KDM6B at the chromatin level [56]. This function is not dependent on methyltransferase activity but plays a role in the differentiation of mouse embryonic stem cells. Our results showed that the expression of METTL14 mutant with abrogated methyltransferase activity robustly promoted sphere formation, likely due to dominant-negative effects, suggesting a suppressive role of m6A in TNBC stemness. The hyperactivation of YAP1 in TNBC is considered due to the silencing of Hippo signaling. For instance, O-GlcNAc glycosylation of LATS2 blocks its phosphorylation by MST [19]. Similarly, the Cullin 2 E3 ubiquitin ligase subunit PRAMEF2 promotes the degradation of LATS1 by catalyzing its polyubiquitination [20]. Furthermore, non-coding RNAs can also inhibit LATS1 through various mechanisms [21, 22]. In this study, we identified YAP1 as an m6A-regulated target from both MCF7 and MDA-MB-231 cells. One of the primary reasons for selecting MCF7 for sequencing is its comparatively higher expression of METTL14 and m6A levels relative to MDA-MB-231, rendering it more susceptible to METTL14 knockdown, which facilitates the screening of DEGs associated with METTL14. While our initial focus was on the differences between luminal breast cancer and MSL TNBC, this does not mean that the role of METTL14 in determining YAP1 expression is opposite in these two breast cancer subtypes. METTL14 knockdown in MCF7 cells also resulted in the upregulation of YAP1, mirroring the response observed in TNBC, suggesting that the underlying molecular mechanism is consistent across various breast cancer subtypes, irrespective of the Hippo pathway activity. Notably, the responses of different breast cancer subtypes to YAP1 upregulation are quite different. A recent study has shown that high expression of YAP1 is detrimental to ER+ breast cancer [18]. Another investigation showed that an lncRNA promotes the progression of ER+ and HER2+ breast cancer by promoting METTL14-mediated m6A modification [57]. In alignment with these findings, our observations revealed that the knockdown of METTL14 resulted in growth inhibition in MCF7 cells while having no substantial impact on the growth of TNBC (data not shown). Therefore, it appears that high levels of METTL14 expression may be unfavorable in subtypes other than TNBC owing to the downregulation of YAP1, which contrasts with the behavior observed in TNBC. It was documented that YAP1 can activate the transcription of ornithine decarboxylase 1 and trigger the polyamine-dependent translation of LSD1 [58]. Thus, our data and those of others suggest a feedback loop for maintaining cell stemness. Although we demonstrated that YAP1 is necessary for m6A-regulated stemness of TNBC, rescue of YAP1 is not sufficient to reverse the suppressive effects of METTL14 on stemness. This suggests that activation cues other than YAP1 are also required for m6A-regulated stemness. It is noteworthy that the overexpression of METTL14 also resulted in a reduction of NANOG, one of the Yamanaka factors, although this effect occurred independently of m6A modification. It is difficult to determine whether this change is a consequence of weakened stemness or due to the function of METTL14 at the chromatin level. Further exploration of the mechanisms by which NANOG cooperates with METTL14 to maintain breast cancer stemness is necessary to answer this question. Nevertheless, YAP1 is considered one of the therapeutic targets for TNBC [59]. A recent investigation employed YAP1 inhibitors in the treatment of TNBC, yielding encouraging results from both in vitro and in vivo models [60]. Thus, our findings offer additional biomarkers for identifying patients who may benefit from YAP1 inhibitors and propose alternative strategies for targeting YAP1. LSD1 belongs to the flavin adenine dinucleotide-dependent monoamine oxidase family and is the first histone demethylase discovered. The main function of LSD1 is to reduce chromatin accessibility by demethylating H3K4me1/2 [48] and to promote the transcriptional activation of hormone receptors by demethylating H3K9me1/2 [61, 62]. In breast cancer, LSD1 is controlled by ubiquitin and is involved in the fine-tuning of gene expression programs associated with differentiation, similar to its function in embryonic stem cells [63]. For example, LSD1 is deubiquitinated and stabilized by USP28 and OTUD7B to support breast cancer stemness and metastasis [50, 64]. Furthermore, LSD1 was found to regulate immune-related genes, and inhibition of LSD1 improves the immunotherapy of TNBC [65]. Our results demonstrated that LSD1 can upregulate YAP1 expression and promote stem-like properties in TNBC largely by repressing the transcription of METTL14, suggesting that the loss of METTL14 is another mechanism critical for the function of LSD1 in determining the aggressiveness of cancer cells. Because LSD1 and METTL14 regulate gene expression at different levels via crosstalk, METTL14 expression level may serve as an indicator for the efficacy of LSD1-based cancer therapy. Conclusion This study indicates that LSD1 disrupts the transcription of METTL14 in TNBC by demethylating histone at the promoter region, which subsequently diminishes global m6A modification. The reduction in m6A modification prevents YTHDF2-mediated degradation of YAP1 transcript, resulting in hyperactivated YAP1 signaling that is essential for maintaining the stemness of TNBC. These results reveal an interplay of epigenetic modifications involved in the maintenance of TNBC stemness. Supplementary Information Supplementary Material 1.Supplementary Material 2.Supplementary Material 3.Supplementary Material 4.Supplementary Material 5.Supplementary Material 6. Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Supplementary Material 5. Supplementary Material 6.