N6-Methyladenosine (m6A) is the most abundant modification of mammalian mRNAs. RNA methylation fine tunes RNA stability and translation, altering cell fate. The fat mass- and obesity-associated protein (FTO) is an m6A demethylase with oncogenic properties in leukemia. Here, we show that FTO expression is suppressed in ovarian tumors and cancer stem cells (CSC). FTO inhibited the self-renewal of ovarian CSC and suppressed tumorigenesis in vivo, both of which required FTO demethylase activity. Integrative RNA sequencing and m6A mapping analysis revealed significant transcriptomic changes associated with FTO overexpression and m6A loss involving stem cell signaling, RNA transcription, and mRNA splicing pathways. By reducing m6A levels at the 3′UTR and the mRNA stability of two phosphodiesterase genes (PDE1C and PDE4B), FTO augmented second messenger 3′, 5′-cyclic adenosine monophosphate (cAMP) signaling and suppressed stemness features of ovarian cancer cells. Our results reveal a previously unappreciated tumor suppressor function of FTO in ovarian CSC mediated through inhibition of cAMP signaling.

Significance:

A new tumor suppressor function of the RNA demethylase FTO implicates m6A RNA modifications in the regulation of cyclic AMP signaling involved in stemness and tumor initiation.

RNA modifications have been implicated in dysregulated gene expression under physiologic and pathologic conditions (1–3). N6-methyladenosine (m6A), the most prevalent internal modification in eukaryotic mRNA, is distributed widely across the transcriptome and alters mRNA stability and translation efficiency (1, 2). The effects of m6A on mRNA fate rely on specific RNA-binding proteins (“readers”; refs. 3, 4), which mediate enhanced mRNA decay or stability, and/or enhanced mRNA translation (5–7). The deposition of m6A is catalyzed by methyltransferase like 3 (METTL3), 14 (METTL14), and Wilms’ Tumor 1 Associating Protein (WTAP) and is removed by RNA demethylases, the fat mass and obesity-associated protein (FTO) and the alkylation repair homolog proteins 5 (ALKBH5; ref. 2). The functions of RNA marks are context-dependent and remain incompletely elucidated.

The link between m6A and oncogenesis was first described in acute myelogenous leukemia (AML), where overexpression of methyltransferases leading to increased m6A deposition inhibited stem cell differentiation and promoted leukemogenesis (8–10). Mutations of key regulatory genes METTL3, METTL14, and WTAP were found in AML and associated with poor clinical outcomes (8, 9, 11). The functions of epitranscriptomic marks regulating the fate of methylated transcripts in solid tumors are just emerging. In lung cancer, METTL3 enhances translation of EGFR and TAZ, leading to increased proliferation (12). ALKBH5 was shown to stimulate the growth of glioblastoma cells by stabilizing nascent RNA transcripts for the transcription factor FOXM1 (13). In contrast, reduced m6A methylation due to hotspot mutations in the METTL14 complex or due to decreased METTL3 expression, promoted endometrial carcinogenesis through an AKT-dependent mechanism (14). The role of the m6A regulatory machinery in ovarian cancer remains unexplored. A single previous report implicated m6A modifications in stabilization of the transcript for the receptor FZD10 in ovarian cancer cells resistant to PARP inhibitors (15), suggesting a potential role regulating DNA damage response mechanisms. Here, we sought to determine the function of the demethylase FTO in ovarian cancer.

FTO, a gene linked to a predisposition for childhood and adult obesity (16), regulates energy homeostasis by controlling food intake and fine tuning of nutrient sensing at the cellular level (17). FTO was the first recognized nucleic acid demethylase (18), physiologically targeting m6A residues in mRNA (19). RNA demethylation by FTO was detected in the midbrain of FTO-deficient mice and linked to altered dopaminergic transmission (20). FTO was found to be highly expressed in AML where it promoted cell transformation and proliferation through post-transcription regulation of ASB2, RARA, MYC, and CEBPA (21, 22).

High-grade serous ovarian cancer (HGSOC) is initially highly responsive to platinum-based therapy and has a propensity for early metastasis (23). Cancer stem cells (CSC) are characterized by the ability to self-renew, grow as spheres, differentiate, and generate tumors and have been implicated in HGSOC initiation, metastasis, and recurrence after chemotherapy (24, 25). CD133 and ALDH have been recognized as robust ovarian CSCs selection markers for (26, 27) and were used in these experiments.

We report that FTO expression is suppressed in HGSOC compared with fallopian tube epithelium. FTO overexpression inhibited colony and sphere formation in vitro and tumor initiation in vivo. Mechanistically, FTO induced higher levels of the second messenger cAMP by inhibiting its hydrolysis mediated by PDE4B and PDE1C through reduction of m6A levels and stability of mRNA transcripts. Our findings point to a tumor-suppressive function of FTO in HGSOC.

Cell culture

SKOV3 and HEK293T cells were purchased from ATCC. COV362 and OVCAR5 cells were provided by Dr. Kenneth Nephew (Indiana University, Bloomington, IN). OVCAR5 and COV362 cells were maintained in high glucose (4.5 g/L) DMEM medium (Corning) containing 10% FBS, 1% penicillin/streptomycin, and 0.1× Non-Essential Amino Acid. SKOV3 and primary cells recovered from human malignant ascites associated with ovarian cancer were cultured in 1:1 MCDB 105 and Medium 199 (Cellgro) supplemented with 10% FBS (Cellgro) and 100 U/mL penicillin and 100 μg/mL streptomycin.

Cell lines were tested to be pathogen and Mycoplasma negative (Charles River Research Animal Diagnostic Services) and periodically by using the Universal Mycoplasma Detection Kit (ATCC). Low passage number was used during experiments.

Human specimens

Advanced high-grade serous ovarian tumors or malignant ascites associated with ovarian cancer were collected fresh under Northwestern University (Chicago, IL)–approved protocol IRB#: STU00202468 (n = 5), processed immediately through enzymatic and mechanical disassociation into single cell suspension and cultured under stem cell conditions, as described previously (28, 29) and used for experiments. Cryopreserved HGSOC specimens (n = 10) and normal ovarian epithelial tissue (n = 6) were obtained from the Indiana University Simon Cancer Center Tissue Bank repository. Human fallopian tube epithelial cells were provided by Dr. Theresa Woodruff of Northwestern University and were obtained from routine hysterectomies, from consenting patients, as described previously (30). Informed written consent was obtained for tissue collection, which was approved by the institutional IRB in accordance with the Declaration of Helsinki.

Plasmid construction

FTO was PCR-amplified and subcloned into plenti-GIII-CMV-GFP-2A-puro lentivirus-based vector (Applied Biological Materials Inc.) by using Nhel and EcoRV digestion. The wild-type FTO and mutant FTO were cloned into lentivector-based pMIRNA1 (SBI) using the following primers: forward 5′-AGAGCTCTAGAACCACCATGGATTACAAAGATGAC-3′ and reverse 5′-CTAAGATTGCGGCCGCCTAGGGTTTTGCTTCCAGAAGC-3′. The lentiviral vector based shRNAs targeting human PDE1C (shPDE1C-1, shPDE1C-2) and human PDE4B (shPDE4B-1, shPDE4B-2) and nontargeting control shRNA were purchased from Applied Biological Materials. The lentiviral vector based shRNAs targeting human FTO (shFTO-1, shFTO-2) and nontargeting control shRNA were purchased from Origene Technologies. Lentiviral particles were produced in 293T cells by cotransfecting the lentiviral construct and the packaging genes VSV-G and Pspax2. The viral particles were harvested 72 hours after transfection and added to cells in the presence of polybrene (8 μg/mL) for 48 hours.

RNA extraction, quantitative RT-PCR analysis, colony-forming assay, and Western blotting methods are included in the Supplementary Material. Sequences of primers for FTO, ALDH1A1, SOX2, NANOG, and OCT4 are included in Supplementary Table S1.

m6A dot blot assay

mRNA was isolated from total RNA by using the Dynabeads mRNA Purification Kit (Ambion) according to the manufacturer's instructions and quantified by NanoDrop 2000 spectrophotometry. The assay was performed following a published protocol (www.bio-protocol.org/e2095) with some modifications. Briefly, the mRNA samples were loaded onto an Amersham Hybond-N+ membrane (GE Healthcare) and UV-crosslinked. The membranes were blocked with 5% nonfat dry milk and incubated with an anti-m6A antibody (Synaptic Systems) overnight at 4°C followed by incubation with HRP-conjugated goat anti-rabbit IgG-HRP (Santa Cruz Biotechnology) for 1 hour at room temperature. The membrane was developed with a chemiluminescent substrate (Thermo Fisher Scientific) before imaging.

m6A RNA methylation quantification

N6-methyladenosine RNA methylation was quantified by using the EpiQuik m6A RNA Methylation Quantification Kit (Epigentek Group Inc.). Briefly, 300 ng total RNA was bound to wells, incubated with capture antibody for 60 minutes, followed by detection antibody for 30 minutes, and enhancer solution for 30 minutes at room temperature. The reaction signal was read with an xMark microplate spectrophotometer (Bio-Rad) at 450 nm.

IHC

A tissue microarray (TMA; OVC1021) containing 92 ovarian cancer specimens (45 serous, 2 clear cell, 27 endometrioid, 18 mucinous) was obtained from Pantomics Inc. Sections from 6 normal fallopian tubes were obtained from the Tissue Pathology Core of the Lurie Cancer Center (Chicago, IL). The FTO antibody (ab92821) was from Abcam Inc. Detailed methods are in the Supplementary Material. H-scores were calculated for each specimen based on the intensity (0–3+) and percentage of cells staining (0%–100%), using the formula: intensity × percentage/100.

RNA sequencing

The RNA sequencing (RNA-seq) libraries were prepared using the NEBNext Ultra II RNA library prep kit from Illumina (New England Biolabs Inc.) and mRNA isolated from 1 μg of total RNA. After first-strand cDNA synthesis, second-strand cDNA synthesis, end repair of cDNA library, dA-tailing of cDNA library, Adaptor ligation, and PCR enrichment, the RNA-seq libraries were checked by using a BioAnalyzer, and sequenced on Illumina NextSeq500 with single-end, 75-bp read length. Each group was sequenced in triplicate. Data analysis is described in Supplementary Methods. Reads of RNA-seq shown in Supplementary Table S2 and RNA-seq QC is included in Supplementary Fig. S10.

MeRIP-seq

To examine m6A modifications across the genome, the Magna MeRIP m6A Kit (Millipore) was used. Briefly, 5 μg of mRNA was sheared to approximately 100 nt in length by metal-ion induced fragmentation, purified, and incubated with m6A antibody (Synaptic Systems) or mouse IgG (control)–conjugated magnetic protein A/G beads in 500 μL 1× IP buffer supplemented with RNase inhibitors at 4°C overnight. Methylated RNAs were immunoprecipitated with beads, eluted by competition with free m6A, and recovered with the RNeasy kit (QIAGEN). One tenth of fragmented RNA was saved as input control. MeRIP libraries were prepared by using the NEBNext Ultra II RNA library prep kit for Illumina (New England Biolabs Inc.), starting from the first-strand cDNA synthesis step. The libraries were sequenced on Illumina NextSeq500 with single-end 75-bp read length. Sequencing was performed in triplicates for GFP and FTO samples, along with their respective controls, resulting in 12 raw fastq files of average read depth 30.6 million reads. Quality check and analysis are described in Supplemental Methods. The accession numbers for RNA-Seq and MeRIP-Seq data reported in this paper are GSE130349 and GSE130350, respectively. Reads of MeRIP-seq show in Supplementary Table S2 and MeRIP-seq QC is included in Supplementary Fig. S11.

Gene-specific m6A qPCR

mRNA was purified from DNase I treated total RNA by using the Dynabeads mRNA Purification Kit (Life Technologies). Five micrograms of mRNA was sheared to approximately 100 nt in length by metal-ion induced fragmentation. One tenth of fragmented RNA was saved as input control, and then the MeRIPed RNAs was obtained as described above. Relative enrichment of m6A was quantified by the Verso SYBR Green 1-step qRT-PCR (Thermo Fisher Scientific) using the primers listed in Supplementary Table S1. The relative mRNA expression in each sample was calculated by the value of Cq in m6A IP portion divided by the value of Cq from the input (CqIP/CqInput). Gene-specific m6A levels were controlled to the relative mRNA expression levels between FTO-overexpressing and control samples (ΔCqtargetCqcontrol).

RNA stability assay

FTO and FTO-mut–overexpressing cell lines, shFTO cell lines, and their control cell lines were treated with the transcription inhibitor actinomycin D (5 μg/mL) at 0, 3, and 6 hours before cell collection. Total RNA was isolated by using TRIzol and purified with RNeasy Kit (QIAGEN). RT-PCR quantified the relative levels of target mRNAs by using the Bio-Rad iTaq universal SYBR green system. The degradation rate of mRNA (Kdecay) was estimated by using the equation: In(C/C0) = -Kdecayt; where C0 is the concentration of mRNA at time 0 hour, before the decay starts, t represents the transcription inhibition time, while C is the mRNA concentration at the time t. The Kdecay can be derived from calculating the exponential decay fitting of C/C0 versus time t. The half-time (t1/2) representing C/C0 = 50%/100% = 1/2 is calculated by using the following equation: ln(1/2) = -Kdecayt1/2. The mRNA half-life time value was estimated as t1/2 = ln2/Kdecay.

Preprocessing and analysis of The Cancer Genome Atlas ovarian cancer exon-array database

Raw Affymetrix Exon-array data for 569 high grade serous ovarian tumors and 8 control samples (fallopian tube epithelium) from The Cancer Genome Atlas (TCGA) and GTEx were processed with Multi-Mapping Bayesian Gene expression for Affymetrix whole-transcript arrays (MMBGX) to obtain gene- and isoform-level expression estimates. Ensembl database (version 56) was used as the reference genome. All expression estimates were normalized across the samples using locally weighted scatter plot smoothing (loess) algorithm. Gene-level expression between ovarian cancer specimens and control samples were compared with the two-sample t test with unequal variances and statistical significance was declared at P = 0.05 (31, 32).

Spheroid formation assay

One hundred ovarian cancer cells, were seeded into 96-well ultra-low attachment plates (Corning) in stem cell media (Stemcell Technologies) with 10% Mammocult proliferation supplements, (Stemcell Technologies) 4 μg/mL heparin (Stemcell technologies), 0.48 μg/mL hydrocortisone (Stemcell technologies), and 1% penicillin/streptomycin solution (100✗, Corning). Media was refreshed every 3 days. The numbers of spheroids per well were counted under an inverted microscope (5 fields/well) on day 14 of culture. Each experiment was performed in triplicate.

Extreme limited dilution analysis

A serial of dilution of cells (5, 10, 50, 100, 500, 1,000, 2,000, and 5,000) were FACS sorted directly into 96-well low-attachment plates and cultured in stem cell medium as described above for 14 days. Each dilution included 12 replicates. The numbers of spheroids per well were counted for each cell dilution. The CSC frequency for the two conditions was calculated by using the extreme limited dilution analysis (ELDA) software (http://bioinf.wehi.edu.au/software/elda/). The same approach was applied to cells dissociated from xenografts using serial dilution numbers of 1,250, 2,500, 5,000, and 10,000 cells.

ALDEFLUOR assay and FACS

Dissociated ovarian cancer single cells were centrifuged at 1,500 rpm for 5 minutes, resuspended in ALDEFLUOR assay buffer (Stemcell Technologies) containing the ALDH substrate, BODIPY-aminoacetaldehyde at 1.5 mmol/L. After incubation for 45 minutes at 37°C, cells were centrifuged at 1,500 rpm for 5 minutes at 4°C, washed with cold ALDEFLUOR assay buffer, and incubated in the same buffer supplemented with 0.5% BSA and CD133/1-PE-Vio615 (2 μL/million cells, Miltenyi Biotec) for 40 minutes on ice. The ALDH+/CD133+ population was gated using control cells incubated under identical conditions in the presence of the ALDH inhibitor (DEAB), and anti-mouse IgG1-PE-Vio615 (1 μL/million cells Miltenyi Biotec) isotype control. Flow cytometry and sorting used the FACSAria II flow cytometer (BD Biosciences) under sterile conditions.

In vivo xenograft experiments

Animal studies were conducted according to a protocol approved by the Institutional Animal Care and Use Committee of Northwestern University (# IS00003060). Female nude, athymic, BALB/c-nu/nu mice (6–7 weeks old; Harlan) were injected subcutaneously (s.c.) with serially diluted numbers (5,000, 10,000, or 20,000) of OVCAR5_GFP or OVCAR5_FTO cells mixed with Matrigel (Thermo Fisher Scientific). Mice were monitored biweekly and time to tumor initiation and tumor growth were recorded. Tumor length (l), width (w), and height (h) were measured with digital calipers and tumor volumes (v) were calculated as v = ½ × l × w × h. CSC frequency in single-cell suspensions dissociated from harvested xenografts and statistical significance were determined by using the ELDA software (http://bioinf.wehi.edu.au/software/elda/).

Isolation of tumor cells used previously described methods (29, 33) and is described in Supplementary Material.

Cyclic AMP XP assay

Cyclic AMP (cAMP) activity was measured with an AMP XP Assay Kit (Cell Signaling Technology) according to the manufacturer's instructions. In brief, 6 to 10 × 104 cells plated into 96-well plates under standard cell culture conditions were lysed in 100 μL of 1× lysis buffer for 15 minutes on ice. Thereafter, HRP-linked cAMP solution and cell lysates were incubated at room temperature for 3 hours on a plate shaker. Wells were washed 4 times before incubating in TMB substrate. The absorbance was measured at 450 nm and the percentage of cAMP activity was calculated as 100 × (1−A/A0), where A was the sample's absorbance and A0 was the absorbance at minimal stimulation (no cAMP).

Statistical analysis

Data were analyzed and presented as means ± SD. Two-tailed Student t test compared means between two groups, and one-way ANOVA was used for multiple group comparisons; P < 0.05 was considered significant.

FTO is expressed at lower levels in ovarian cancer cells, tumors, and in ovarian cancer stem cells

To understand whether RNA methylation is involved in the process of ovarian cancer initiation or progression, we compared the expression levels of m6A writers and erasers between normal fallopian tube epithelium (FTE), normal ovarian epithelium and ovarian tumors. No significant differences were observed between HGSOC specimens (n = 10), normal FTE (n = 4) and normal ovarian tissue (n = 5) for METTL3, METTL14, WTAP, or ALKBH5 (Fig. 1A). In contrast, the mRNA expression levels of the demethylase FTO were significantly lower in ovarian tumors compared with FTE and normal ovary tissue (Fig. 1A, P = 0.0485). An analysis of the TCGA Affymetrix Exon-array data including 569 high-grade serous and 8 control FTE specimens for the five genes revealed that FTO and ALKBH5, but not METTL3, METTl14, or WTAP, were significantly downregulated in HGSOC relative to FTE (P < 0.001; Fig. 1B; Supplementary Fig. S1). Among the 569 tumor samples, 497 (87%) and 566 (99%) displayed downregulated FTO and ALKBH5 expression levels, respectively, relative to normal controls. Because FTO expression levels were found to be significantly suppressed in both the TCGA database and the in-house dataset, we focused on FTO for further studies. FTO protein expression was measured by IHC in ovarian cancer specimens (n = 92; all histologic subtypes) included in a TMA and FT (n = 6). FTO expression quantified as H score was down regulated in ovarian cancer compared with FTE (P < 0.05), for the serous, endometrioid, and mucinous groups (Fig. 1C and D). Likewise, significantly lower FTO mRNA and protein expression levels were detected in ovarian cancer cell lines compared with primary FTE cell cultures (Fig. 1E and F, P < 0.01). Collectively, the data support that FTO expression is significantly downregulated in ovarian cancer cells and tumors.

Figure 1.

FTO is downregulated in ovarian cancer cells and ovarian tumors. A,mRNA expression levels measured by qRT-PCR for the RNA methylation regulatory genes METTL3, METTL14, WTAP, FTO, and ALKBH5 between normal fallopian tube epithelium (FT; n = 4), ovarian epithelium (n = 5), and human ovarian cancer specimens (n = 10). B, Expression levels of FTO and ALKBH5 in HGSOC (n = 569) compared with normal fallopian tube epithelium (n = 8) extracted from the TCGA database. C, Representative FTO IHC staining in fallopian tube and ovarian cancer specimens (magnification, ×100). Scale bars, 200 μm. D, H-scores compared FTO expression in fallopian tube (n = 6) compared with endometrioid (n = 27; P = 0.037), mucinous (n = 18; P = 0.019), serous (n = 45, P = 0.036), and clear cell (n = 2, NS) subtypes. E,FTO mRNA expression levels in primary human fallopian tube epithelial cells (n = 2; FEM121, FEM122) and ovarian cancer cell lines (SKOV3, OV90, HEY, COV362, OVCAR3, OVCAR4, OVCAR5, and OVCAR8). F, FTO protein expression levels were measured by Western blotting in human immortalized fallopian tube epithelial cells (FT; FT-190) and ovarian cancer cells lines (SKOV3, OV90, HEY, COV362, OVCAR3, OVCAR4, OVCAR5, and OVCAR8). G,FTO mRNA expression levels in ALDH+ and ALDH FACS sorted cells from OVCAR5 and COV362 cell lines. H,FTO mRNA expression levels in ALDH+CD133+ (DP) and ALDH+ compared with ALDHCD133 (DN) FACS-sorted cells derived from HGSOC. Bars display mean fold changes ± SD (n = 5). I, Quantification of m6A levels in OVCAR3, COV362, and OVCAR5 cells cultured as monolayers or spheroids. J, M6A levels in FACS-sorted ALDH and ALDH+ cells from ovarian cancer cell lines and from human tumors. Bars, means ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Figure 1.

FTO is downregulated in ovarian cancer cells and ovarian tumors. A,mRNA expression levels measured by qRT-PCR for the RNA methylation regulatory genes METTL3, METTL14, WTAP, FTO, and ALKBH5 between normal fallopian tube epithelium (FT; n = 4), ovarian epithelium (n = 5), and human ovarian cancer specimens (n = 10). B, Expression levels of FTO and ALKBH5 in HGSOC (n = 569) compared with normal fallopian tube epithelium (n = 8) extracted from the TCGA database. C, Representative FTO IHC staining in fallopian tube and ovarian cancer specimens (magnification, ×100). Scale bars, 200 μm. D, H-scores compared FTO expression in fallopian tube (n = 6) compared with endometrioid (n = 27; P = 0.037), mucinous (n = 18; P = 0.019), serous (n = 45, P = 0.036), and clear cell (n = 2, NS) subtypes. E,FTO mRNA expression levels in primary human fallopian tube epithelial cells (n = 2; FEM121, FEM122) and ovarian cancer cell lines (SKOV3, OV90, HEY, COV362, OVCAR3, OVCAR4, OVCAR5, and OVCAR8). F, FTO protein expression levels were measured by Western blotting in human immortalized fallopian tube epithelial cells (FT; FT-190) and ovarian cancer cells lines (SKOV3, OV90, HEY, COV362, OVCAR3, OVCAR4, OVCAR5, and OVCAR8). G,FTO mRNA expression levels in ALDH+ and ALDH FACS sorted cells from OVCAR5 and COV362 cell lines. H,FTO mRNA expression levels in ALDH+CD133+ (DP) and ALDH+ compared with ALDHCD133 (DN) FACS-sorted cells derived from HGSOC. Bars display mean fold changes ± SD (n = 5). I, Quantification of m6A levels in OVCAR3, COV362, and OVCAR5 cells cultured as monolayers or spheroids. J, M6A levels in FACS-sorted ALDH and ALDH+ cells from ovarian cancer cell lines and from human tumors. Bars, means ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Close modal

Given the previously reported role of m6A modifications in the regulation of normal and leukemic stem cell populations (8, 9), FTO expression was measured in cell subpopulations derived from cell lines or human ovarian tumors. Cells characterized by high ALDH activity possess tumor initiating capacity, self-renewal ability, effectively form spheres, express “stemness” associated transcription factors, and are quiescent (26, 34, 35). FTO expression levels were significantly lower in the ALDH+ population compared to ALDH cells derived from OVCAR5 and COV362 cell lines (Fig. 1G, P < 0.01). Similarly, FTO expression levels were lower in ALDH+ and ALDH+/CD133+ cells compared with the ALDH/CD133 cell populations derived from human HGSOC specimens (n = 5; Fig. 1H, P < 0.01). FACS-based separation of ALDH+/CD133+ cells is shown in Supplementary Fig. S2.

To determine the effects of FTO overexpression on m6A levels, an m6A RNA quantification assay was used. This assay requires >105 cells and is not applicable to rare cell populations, such as CSCs, therefore, cultures were enriched under sphere conditions (28, 29). Global m6A mRNA levels were significantly increased in spheroids derived from OVCAR3, OVCAR5 and COV362 cells compared to monolayer cultures (Fig. 1I, P < 0.05). Global m6A levels were significantly higher in ALDH+ versus ALDH cells (Fig. 1J, P < 0.05), FACS sorted from HGSOC cell lines or human tumors, and propagated under nondifferentiating culture conditions, as spheres. These results show increased global m6A RNA levels, accompanied by FTO downregulation, in ovarian cancer stem cells (OCSC) relative to non-CSCs.

FTO overexpression inhibits ovarian CSC proliferation/self-renewal and tumor initiation capacity

To investigate the functions of FTO in OCSCs, we developed stable cell lines overexpressing FTO. FTO mRNA and protein expression levels were increased in OVCAR5, COV362, and SKOV3 cells stably transduced with FTO-carrying lentivirus compared with vector control (Fig. 2A and B; Supplementary Fig. S3A–S3C). Accordingly, the global m6A levels on mRNA were decreased in FTO-overexpressing OVCAR5, COV362, and SKOV3 cells (Supplementary Fig. S3D and S3E). Notably, m6A levels were significantly higher in OVCAR5-FTO cells grown under nondifferentiating conditions as spheroids (OVCAR5-FTO-Sph) compared with monolayer cultures (Supplementary Fig. S3E, P < 0.05), consistent with the concept that RNA methylation would be increased in cultures enriched with CSCs, associated with FTO downregulation.

Figure 2.

FTO overexpression inhibits colony forming ability of ovarian cancer cells and cancer stem cell characteristics. A and B,FTO overexpression in OVCAR5 cells (A) and COV362 cells (B) was verified by using qRT-PCR (mRNA) and Western blot analysis (protein). C and D, Numbers of colonies generated by OVCAR5 (C) and COV362 (D) cells transduced with FTO or control vector (n = 3). E, Numbers of spheroids formed by OVCAR5 cells transduced with control or FTO during 14-day culture (n = 3). F,ALDH1A1 mRNA expression levels in SKOV3 and OVCAR5 cells ± FTO. G,mRNA expression levels for the stemness-associated transcription factors OCT4, NANOG, and SOX2 in OVCAR5 cells transduced with control or FTO; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Figure 2.

FTO overexpression inhibits colony forming ability of ovarian cancer cells and cancer stem cell characteristics. A and B,FTO overexpression in OVCAR5 cells (A) and COV362 cells (B) was verified by using qRT-PCR (mRNA) and Western blot analysis (protein). C and D, Numbers of colonies generated by OVCAR5 (C) and COV362 (D) cells transduced with FTO or control vector (n = 3). E, Numbers of spheroids formed by OVCAR5 cells transduced with control or FTO during 14-day culture (n = 3). F,ALDH1A1 mRNA expression levels in SKOV3 and OVCAR5 cells ± FTO. G,mRNA expression levels for the stemness-associated transcription factors OCT4, NANOG, and SOX2 in OVCAR5 cells transduced with control or FTO; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Close modal

Colony-forming assays showed that FTO-overexpressing OVCAR5, COV362 and SKOV3 cells formed significantly fewer colonies compared to (GFP-expressing) control cells (Fig. 2C and D; P < 0.01; Supplementary Fig. S3F). Similarly, much fewer spheres were formed by FTO-overexpressing OVCAR5 versus control cells (Fig. 2E, P < 0.01). The morphology of spheres derived from FTO-overexpressing cells was less well structured, as such cells formed loose aggregates, without a peripheral ring, as compared to control cells, which formed classic spheres (Fig. 2E). To further characterize the effects of FTO on stem cell characteristics, the percentage of ALDH+ cells and the expression of stemness associated transcription factors (NANOG, SOX2, and OCT4) were assessed. The percentage of ALDH+ cells was significantly lower in FTO-overexpressing OVCAR5 versus control cells (Supplementary Fig. S3G, P < 0.05). Moreover, expression levels of ALDH1A1 (Fig. 2F, P < 0.001) and of transcription factors (NANOG, SOX2, and OCT4) were significantly lower in FTO-overexpressing OVCAR5 and SKOV3 versus control cells (Fig. 2G, P < 0.05 and Supplementary Fig. S3H).

Finally, the effects of FTO on cancer stemness were assessed through limiting dilution assays. In vitro, spheroid formation was measured from serially diluted FTO-overexpressing or control OVCAR5 cells (Supplementary Table S3; Supplementary Fig. S3I). ELDA calculations (36) demonstrated that the frequency of CSCs was significantly decreased in FTO-overexpressing OVCAR5 cells (1 in 677.36) relative to the control cells (1 in 7.19; P = 1.99e−43, Supplementary Table S3). Tumor initiation capacity (TIC) was assessed by injecting subcutaneously serially diluted GFP- or FTO-overexpressing OVCAR5 cells (5,000, 10,000, and 20,000 cells) into female nude mice. Tumor size derived from FTO-expressing cells was significantly smaller at all dilutions compared with tumors derived from the control GFP-expressing cells (Fig. 3AC, P < 0.05). Time to tumor initiation was prolonged for FTO-overexpressing versus vector-transduced OVCAR5 cells (41 days vs. 23 days, P < 0.05; Supplementary Fig. S4A). ELDA calculations (36) indicated that the CSCs' frequency was significantly lower in FTO-expressing OVCAR5 (1 in 119,931) versus control cells (1 in 10,941; P = 0.00491; Fig. 3D). Furthermore, FACS demonstrated that the population of ALDH+ cells was significantly lower in tumors from mice xenografted with FTO-overexpressing OVCAR5 cells than those from control xenografts (Fig. 3E and F, P < 0.01). Increased FTO expression in xenografts was verified by qRT-PCR (Fig. 3G, P < 0.01). Expression levels of stemness-associated factors (ALDH1A1, NANOG, SOX2, and OCT4) were significantly lower in FTO-expressing tumor cells vs. control cells (Fig. 3H; P < 0.05). Cells dissociated from FTO-overexpressing xenografts had diminished sphere forming ability and contained fewer CSCs compared with cells dissociated from control xenografts (Supplementary Fig. S4B and S4C, P = 4.08e−5). Overall, these data suggest that forced expression of FTO reduced the population of CSCs and inhibited their specific functions.

Figure 3.

FTO inhibits tumor initiation capacity. A–C, Representative images (top) and tumor volumes (means ± SD; bottom) of subcutaneous xenografts formed by 5,000 (A), 10,000 (B), or 20,000 (C) OVCAR5_GFP and OVCAR5_FTO cells (n = 3 or 5). D, Numbers of mice harboring detectable tumors for each cell dilution. CSC frequency was calculated using the ELDA software (P = 0.00491). Plot indicates the log fraction of mice that did not form tumors versus the number of injected cells. E, Side scatter shows the percentages of ALDH+ cells in single-cell suspensions derived from OVCAR5-GFP or OVCAR5-FTO xenografts. F, Percentages of ALDH+ cells from single-cell suspensions derived from OVCAR5-GFP or OVCAR5-FTO xenografts (n = 3). G, qRT-PCR measured FTO mRNA expression levels in harvested OVCAR5-GFP and OVCAR5-FTO xenografts. H, Fold changes (means ± SD) for ALDH1A1, NANOG, SOX2, and OCT4 mRNA levels in OVCAR5-FTO versus OVCAR5-GFP–derived xenografts. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Figure 3.

FTO inhibits tumor initiation capacity. A–C, Representative images (top) and tumor volumes (means ± SD; bottom) of subcutaneous xenografts formed by 5,000 (A), 10,000 (B), or 20,000 (C) OVCAR5_GFP and OVCAR5_FTO cells (n = 3 or 5). D, Numbers of mice harboring detectable tumors for each cell dilution. CSC frequency was calculated using the ELDA software (P = 0.00491). Plot indicates the log fraction of mice that did not form tumors versus the number of injected cells. E, Side scatter shows the percentages of ALDH+ cells in single-cell suspensions derived from OVCAR5-GFP or OVCAR5-FTO xenografts. F, Percentages of ALDH+ cells from single-cell suspensions derived from OVCAR5-GFP or OVCAR5-FTO xenografts (n = 3). G, qRT-PCR measured FTO mRNA expression levels in harvested OVCAR5-GFP and OVCAR5-FTO xenografts. H, Fold changes (means ± SD) for ALDH1A1, NANOG, SOX2, and OCT4 mRNA levels in OVCAR5-FTO versus OVCAR5-GFP–derived xenografts. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Close modal

FTO knockdown promotes CSC self-renewal

Loss-of-function studies used OVCAR5 cells stably transduced with individual shRNAs targeting FTO (shFTO-1/2) or a control (nontargeting) shRNA (shCTL). The downregulation of FTO at both RNA and protein levels mediated by shFTO was confirmed (Fig. 4A) along with higher levels of m6A measured in FTO-depleted compared with control cells (Fig. 4B; Supplementary Fig. S5A). Depletion of FTO by shFTO-1 or shFTO-2 resulted in substantial upregulation of ALDH1A1 (Fig. 4A, P < 0.001). Depletion of FTO by shFTO-2 potently induced the expression of all stem cell factors (SOX2, OCT4, and NANOG), while transduction of shFTO-1 only induced OCT4 expression (Fig. 4C, P < 0.001). FTO knockdown resulted in significantly increased numbers of colonies (Fig. 4D, P < 0.05) and spheres (Fig. 4E; P < 0.05), relative to the control group. FTO knockdown led to an increased population of ALDH+ cells compared with vector-transduced cells (Fig. 4F, P < 0.01). Tumor initiation capacity (TIC) also was assessed by injecting subcutaneously serially diluted shCTL or shFTO cells (500, 1,000, and 2,000 cells) into female nude mice. Time to tumor initiation was prolonged for shFTO vs. shCTL cells (9 days vs. 14 days, P = 0.0491; Supplementary Fig. S5B). ELDA calculations indicated that the CSCs frequency was higher in shFTO (1 in 298) versus shCTL (1 in 1,106; P = 0.0320; Supplementary Fig. S5C). Overall, depletion of FTO in ovarian cancer cells caused an increased level of m6A, induced cancer cell proliferation as spheres, induced colony formation, and promoted a CSC phenotype.

Figure 4.

FTO knockdown promotes ovarian cancer cell stemness. A,FTO and ALDH1A1 mRNA expression levels in OVCAR5 cells transduced with nontargeting shRNA (shCTL) or shRNA targeting FTO (shFTO-1 and shFTO-2). Western blotting assessed FTO protein expression in shCTL-, shFTO-1–, and shFTO-2– transduced cells. B, m6A levels were quantified in shCTL- and shFTO-transduced OVCAR5 cells by using the EpiQuik m6A RNA Methylation Kit. C,mRNA expression levels of the stemness markers NANOG, SOX2, and OCT4 were determined by qRT-PCR in FTO knockdown versus control cells (n = 3). D, Colonies formed by shCTL-, shFTO-1–, and shFTO-2– transduced OVCAR5 cells (top) and average numbers of colonies formed by shCTL-, shFTO-1–, and shFTO-2–transduced OVCAR5 cells (bottom). E, Average numbers of spheroids formed by shCTL-, shFTO-1–, and shFTO-2–transduced OVCAR5 cells (n = 3). F, FACS quantification of ALDH+ cells in shCTL- and shFTO-2–transduced OVCAR5 cells (n = 3). G, FTO protein and mRNA expression levels after stable transduction of FTO and FTO-mutant (FTO-mut) in OVCAR5 cells. H, Numbers of colonies generated from OVCAR5 cells stably transduced with vector control, FTO, or FTO mutant. I, Number of spheres generated from OVCAR5 cells stably transduced with vector control, FTO, or FTO mutant. J, qRT-PCR measured mRNA expression levels for NANOG, OCT4, and SOX2 in vector control, FTO, or FTO mutant-transduced ovarian cancer cells. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Figure 4.

FTO knockdown promotes ovarian cancer cell stemness. A,FTO and ALDH1A1 mRNA expression levels in OVCAR5 cells transduced with nontargeting shRNA (shCTL) or shRNA targeting FTO (shFTO-1 and shFTO-2). Western blotting assessed FTO protein expression in shCTL-, shFTO-1–, and shFTO-2– transduced cells. B, m6A levels were quantified in shCTL- and shFTO-transduced OVCAR5 cells by using the EpiQuik m6A RNA Methylation Kit. C,mRNA expression levels of the stemness markers NANOG, SOX2, and OCT4 were determined by qRT-PCR in FTO knockdown versus control cells (n = 3). D, Colonies formed by shCTL-, shFTO-1–, and shFTO-2– transduced OVCAR5 cells (top) and average numbers of colonies formed by shCTL-, shFTO-1–, and shFTO-2–transduced OVCAR5 cells (bottom). E, Average numbers of spheroids formed by shCTL-, shFTO-1–, and shFTO-2–transduced OVCAR5 cells (n = 3). F, FACS quantification of ALDH+ cells in shCTL- and shFTO-2–transduced OVCAR5 cells (n = 3). G, FTO protein and mRNA expression levels after stable transduction of FTO and FTO-mutant (FTO-mut) in OVCAR5 cells. H, Numbers of colonies generated from OVCAR5 cells stably transduced with vector control, FTO, or FTO mutant. I, Number of spheres generated from OVCAR5 cells stably transduced with vector control, FTO, or FTO mutant. J, qRT-PCR measured mRNA expression levels for NANOG, OCT4, and SOX2 in vector control, FTO, or FTO mutant-transduced ovarian cancer cells. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Close modal

The demethylase function of FTO regulates cancer stemness phenotype

To determine whether the effects of FTO are due to its demethylase function, we used a mutant carrying two point mutations H231A and D233A, which disrupt its enzymatic function (20). FTO and FTO-Mut were stably overexpressed in OVCAR5 cells (Fig. 4G). To confirm the mutant's loss of demethylase function, global m6A mRNA levels were measured by ELISA in total RNA from FTO and FTO-Mut transduced cells and found to significantly decreased in FTO-overexpressing OVCAR5 cells compared with control cells, but not altered in FTO-Mut transduced cells (Supplementary Fig. S5D). Although wild-type FTO overexpression reduced colony (Fig. 4H) and sphere formation (Fig. 4I) compared with control, forced expression of the mutant did not significantly alter the phenotype. Similarly, OCT4 and SOX2 mRNA expression levels were downregulated in cells expressing wild-type FTO, but not in cells transduced with the FTO mutant (Fig. 4J, P < 0.05). Collectively, these data support that the demethylase activity of FTO is required for its suppressing function against the OCSC phenotype.

Signaling pathways affected by the FTO/m6A axis in ovarian cancer cells

To decipher the mechanisms by which FTO-mediated m6A modifications altered stemness, RNA-sequencing (RNA-Seq) and methylated RNA immunoprecipitation-sequencing (MeRIP-seq) compared OVCAR5 cells overexpressing FTO or vector. A total of 2,406 significant differential m6A peaks were distributed in the 3′ UTR (18%), 5′UTR (7%), introns (29%), and coding sequence (CDS, 32%; Fig. 5A). A similar pattern was observed for total m6A peaks (22,703 peaks; 18% were in the 3′UTR, 10% in 5′ UTR, 27% in intronic regions, and 28 % in CDS; Supplementary Fig. S6A). Motif analysis of the peaks region using DREME (37) showed that the common m6A motif GGAC was significantly enriched (Supplementary Fig. S6B; p-value of 3.5e-042). There were 4,550 differentially expressed genes observed between FTO and vector-transfected ovarian cancer cells, as determined by RNA-seq (FDR < 0.05, P < 0.05). Integration of the differentially expressed (n = 4,550 genes) and differentially m6A marked transcripts (n = 1,272 RNA transcripts) yielded 731 overlapping genes (Fig. 5B), of which, 93.4% were demethylated in FTO-expressing cells (P < 0.05; fold-change ≥ 1.2). Heatmap analysis indicated a clear separation between the gene expression profiles of GFP- and FTO-expressing cells (Fig. 5C). Among the 731 genes, 246 were hypomethylated and downregulated, 437 were hypomethylated and upregulated, 28 were hypermethylated and downregulated, and 20 were hypermethylated and upregulated (Fig. 5D).

Figure 5.

Genome wide m6A modifications and transcriptomic changes induced by FTO. A, The distribution of differential m6A peaks (n = 2,406) in the 3′UTR, 5′UTR, intron, noncoding, stop codon, intergenic, or CDS regions. B, Venn diagram illustrates the numbers of genes with significant expression or m6A markings changes (FDR < 0.05; P < 0.05) as measured by RNA-seq and MeRIP-seq. C, Heatmap obtained through unsupervised complete linkage hierarchical clustering based on Euclidean distance and scaling across rows demonstrates differential gene expression between FTO- and vector-expressing OVCAR5 cells (n = 3). Differentially expressed and differentially m6A marked genes (P < 0.05 and FDR < 0.05) were included (n = 731 genes). D, Genes displaying a significant change in m6A marks and expression levels in FTO versus control OVCAR5 cells were grouped in four categories: hypomethylated/downregulated (n = 246; bottom left quadrant); hypomethylated/upregulated (n = 437; bottom right quadrant); hypermethylated/downregulated (n = 28; top left quadrant); hypermethylated/upregulated (n = 20; top right quadrant). E, GSEA reveals significantly upregulated (red) and downregulated (blue) pathways (P ≤ 0.05; FDR ≤ 0.05) in FTO- versus control-transduced OVCAR-5 cells. Enrichment Map and Auto Annotate applications in Cytoscape were used to construct the visual representation. Nodes represent different gene sets from the GSEA database and edges represent the similarity coefficients between two linked nodes (gene sets). This enrichment map was manually curated after removing uninformative individual gene sets. F, Top downregulated or upregulated transcripts in FTO-overexpressing cells. Bars represent fold change, extracted from RNA-seq analysis. G, Interaction network of relevant biological pathways affected by FTO overexpression using genes selected from the hypomethylated/downregulated group.

Figure 5.

Genome wide m6A modifications and transcriptomic changes induced by FTO. A, The distribution of differential m6A peaks (n = 2,406) in the 3′UTR, 5′UTR, intron, noncoding, stop codon, intergenic, or CDS regions. B, Venn diagram illustrates the numbers of genes with significant expression or m6A markings changes (FDR < 0.05; P < 0.05) as measured by RNA-seq and MeRIP-seq. C, Heatmap obtained through unsupervised complete linkage hierarchical clustering based on Euclidean distance and scaling across rows demonstrates differential gene expression between FTO- and vector-expressing OVCAR5 cells (n = 3). Differentially expressed and differentially m6A marked genes (P < 0.05 and FDR < 0.05) were included (n = 731 genes). D, Genes displaying a significant change in m6A marks and expression levels in FTO versus control OVCAR5 cells were grouped in four categories: hypomethylated/downregulated (n = 246; bottom left quadrant); hypomethylated/upregulated (n = 437; bottom right quadrant); hypermethylated/downregulated (n = 28; top left quadrant); hypermethylated/upregulated (n = 20; top right quadrant). E, GSEA reveals significantly upregulated (red) and downregulated (blue) pathways (P ≤ 0.05; FDR ≤ 0.05) in FTO- versus control-transduced OVCAR-5 cells. Enrichment Map and Auto Annotate applications in Cytoscape were used to construct the visual representation. Nodes represent different gene sets from the GSEA database and edges represent the similarity coefficients between two linked nodes (gene sets). This enrichment map was manually curated after removing uninformative individual gene sets. F, Top downregulated or upregulated transcripts in FTO-overexpressing cells. Bars represent fold change, extracted from RNA-seq analysis. G, Interaction network of relevant biological pathways affected by FTO overexpression using genes selected from the hypomethylated/downregulated group.

Close modal

Gene set enrichment analysis (GSEA) on genes preranked by log2 fold change and log P values yielded “embryonic stem cells associated with H3K27me3” [normalized enrichment score (NES) of −5.12, FDR < 0.0001; Supplementary Fig. S6C] as being enriched in GFP-expressing cells and “downregulated hematopoietic stem cells” as being enriched in FTO-expressing cells (NES = 2.26, FDR = 0.005; Supplementary Fig. S6C); supporting FTO's role inhibiting stemness. Other gene sets represented in FTO cells vs. controls include “downregulated genes in metastasis” (NES = 4.07) and “downregulated genes in neoplastic transformation” (NES 3.32; Supplementary Fig. S6C), supporting a tumor suppressor function. An enrichment map was built by using significant gene sets (P < 0.05 and FDR < 0.05). FTO overexpression led to upregulated gene sets linked to immune response and RNA transcription pathways and downregulated DNA repair pathways, stem cell signaling, and mRNA splicing (Fig. 5E). The top hypomethylated and differentially expressed genes are included in Supplementary Tables S4–S6 and include the adhesion molecule ICAM1, the phosphodiesterases PDE1C and PDE4B, the stem cell related factor Noggin, the signaling molecule TACSTD2, and others (Fig. 5F). Focusing on genes related to signaling, selected from the “hypomethylated-downregulated” group, and analyzed by using Ingenuity Pathway Analysis, we identified gene networks related to signal transduction, cAMP signaling, RNA transcription, and various cancer-related pathways (Fig. 5G; Supplementary Fig. S6D). Importantly, genes related to cAMP hydrolysis (PDE1C and PDE4B) and signaling were represented in these networks and studied further.

M6A-modified mRNAs regulate the expression of two phosphodiesterases

Among the top five differentially m6A modified and expressed genes, the phosphodiesterases 4B (PDE4B) and 1C (PDE1C) were hypomethylated and downregulated in FTO-overexpressing cells compared with controls (Supplementary Table S4). The differentially m6A enriched regions were included in the 3′ UTR for PDE4B and spanned the last three exons of the PDE1C transcript, including the last exon with STOP codon and 3′UTR, and are shown in Fig. 6A. Validation studies showed that FTO knockdown significantly upregulated expression of both PDE4B and PDE1C in OVCAR5 cells (Fig. 6B, P < 0.01). Conversely, expression of PDE4B and PDE1C was significantly downregulated upon forced expression of wild-type FTO (but less by the FTO mutant) in OVCAR5, COV362, and SKOV3 cells (Fig. 6C, P < 0.01; Supplementary Fig. S7A and S7B, P < 0.05). Furthermore, expression levels of PDE4B and PDE1C were higher in ALDH+/CD133+ cells compared to the ALDH/CD133 cell populations derived from human HGSOC specimens (n = 3; Fig. 6D, P < 0.001). Significantly reduced m6A levels at the 3′ UTR of PDE4B and near the 3′UTR for PDE1C were validated in cells overexpressing wild-type FTO (but not in those overexpressing FTO mutant) by gene-specific m6A qRT-PCR using m6A immunoprecipitated total RNA (Fig. 6E), confirming the findings of MeRIP-sequencing. To further demonstrate that m6A modifications at the 3′UTR of these transcripts alter mRNA stability, we treated cells overexpressing FTO or FTO-mutant with actinomycin D, a transcription inhibitor, and measured the half-lives of PDE1C and PDE4B transcripts. Indeed, wild-type FTO overexpression led to significantly shortened half-life of PDE1C (5 to 2.9 h, Fig. 6F) and PDE4B (7.87 to 2.87 h, Fig. 6G), while overexpression of FTO mutant did not significantly alter PDE1C (5 to 4.5 h, Fig. 6F) and only mildly affected PDE4B half-life (7.87 to 4.91 h, Fig. 6G). Collectively, these data suggest that FTO directly affects the mRNA stability of the two phosphodiesterase genes. Because recent studies have reported binding of the insulin growth factor 2 mRNA-binding proteins 2 and 3 (IGF2BP 2/3) to m6A modified regions, leading to enhanced mRNA stability (7), we further checked the possible binding sites of IGF2BP2 and IGF2BP3 within these differentially enriched regions, using RBPmap with maximum stringency (P < 0.001) and conservation filter. The bioinformatic tools detected possible binding sites in a 10,000 bp region: for PDE4B, five sites for IGF2BP2 and four for IGF2BP3 and for PDE1C, fifty-three sites for IGF2BP2 and forty-two for IGF2BP3 (average Z-score >3; Supplementary Fig. S8).

Figure 6.

PDE1C and PDE4B are FTO target genes in ovarian cancer cells. A, Differences in m6A peaks distributions as detected by MeRIP-seq in the 3′UTR of PDE4B and the last three exons of PDE1C mRNA transcripts in control or FTO-transduced OVCAR5 cells. B and C,mRNA expression levels of PDE4B and PDE1C in shCTL- or shFTO-transduced OVCAR5 cells (B) or in control, FTO, or FTO mutant–transduced OVCAR5 cells (C). D,PDE4B and PDE1C mRNA expression levels in human HGSOC–specimens derived ALDH/CD133 double-positive (DP+) cells compared with ALDH/CD133 double-negative cells (DN−). E, Gene-specific m6A PCR measured changes in m6A levels in the 3′UTR of PDE4B and associated with PDE1C in FTO, FTO-mutant, versus control-transduced OVCAR5 cells. F and G,PDE1C (F) and PDE4B (G) mRNA transcript half-lives in ovarian cancer cells transduced with control, FTO, or FTO-mutant. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Figure 6.

PDE1C and PDE4B are FTO target genes in ovarian cancer cells. A, Differences in m6A peaks distributions as detected by MeRIP-seq in the 3′UTR of PDE4B and the last three exons of PDE1C mRNA transcripts in control or FTO-transduced OVCAR5 cells. B and C,mRNA expression levels of PDE4B and PDE1C in shCTL- or shFTO-transduced OVCAR5 cells (B) or in control, FTO, or FTO mutant–transduced OVCAR5 cells (C). D,PDE4B and PDE1C mRNA expression levels in human HGSOC–specimens derived ALDH/CD133 double-positive (DP+) cells compared with ALDH/CD133 double-negative cells (DN−). E, Gene-specific m6A PCR measured changes in m6A levels in the 3′UTR of PDE4B and associated with PDE1C in FTO, FTO-mutant, versus control-transduced OVCAR5 cells. F and G,PDE1C (F) and PDE4B (G) mRNA transcript half-lives in ovarian cancer cells transduced with control, FTO, or FTO-mutant. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Close modal

FTO/m6A regulated phosphodiesterases are involved in maintaining stemness

To determine whether the two phosphodiesterase genes play a role in maintaining cancer stemness, we used shRNA mediated knockdown and small molecule inhibitors. Knockdown of either PDE1C or PDE4B (Fig. 7A) downregulated the expression of stemness-associated factors (NANOG, OCT4, and SOX2; Fig. 7B and C) and inhibited spheroid formation (Fig. 7D) consistent with a stemness-inhibitory effect induced by FTO overexpression. Furthermore, PDE1C knockdown could at least in part reverse the effects of FTO knockdown on promoting expression of ALDH1A1, NANOG, OCT4 and SOX2 in OVCAR5 cells (Fig. 7EI), suggesting that these genes are critical targets of FTO and are required for maintaining the stemness phenotype in ovarian cancer cells.

Figure 7.

FTO-regulated PDE4B and PDE1C maintain the ovarian cancer stemness phenotype. A,PDE1C and PDE4B mRNA levels in OVCAR5 cells transduced with control or PDE1C or PDE4B targeting shRNA. B and C,NANOG, OCT4, and SOX2 mRNA levels in OVCAR5 cells transduced with control or PDE1C (B) or PDE4B (C) targeting shRNA. D, Sphere formation derived from OVCAR5 cells transduced with control or PDE1C or PDE4B targeting shRNA. E,PDE1C mRNA levels in sh-control or sh-FTO–transduced OVCAR5 cells, subsequently transduced with control or PDE1C shRNA (two sequences). F–I,ALDH1A1 (F), NANOG (G), OCT4 (H), and SOX2 (I) mRNA levels in sh-control or sh-FTO–transduced OVCAR5 cells, subsequently transduced with control or PDE1C shRNA. J, cAMP activity in GFP- or FTO-transduced OVCAR5 cells. K, Cyclic AMP activity was detected in rolipram (1 μmol/L)- and vinpocetine (50 μmol/L)-treated shCTL or shFTO stably transduced OVCAR5 cells. L, Cyclic AMP activity was determined in ALDH+ cells compared with ALDH cells FACS-sorted from OVCAR5 and COV362 cells. M, Proposed mechanism by which FTO regulates tumor initiation and ovarian cancer stem cells. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Figure 7.

FTO-regulated PDE4B and PDE1C maintain the ovarian cancer stemness phenotype. A,PDE1C and PDE4B mRNA levels in OVCAR5 cells transduced with control or PDE1C or PDE4B targeting shRNA. B and C,NANOG, OCT4, and SOX2 mRNA levels in OVCAR5 cells transduced with control or PDE1C (B) or PDE4B (C) targeting shRNA. D, Sphere formation derived from OVCAR5 cells transduced with control or PDE1C or PDE4B targeting shRNA. E,PDE1C mRNA levels in sh-control or sh-FTO–transduced OVCAR5 cells, subsequently transduced with control or PDE1C shRNA (two sequences). F–I,ALDH1A1 (F), NANOG (G), OCT4 (H), and SOX2 (I) mRNA levels in sh-control or sh-FTO–transduced OVCAR5 cells, subsequently transduced with control or PDE1C shRNA. J, cAMP activity in GFP- or FTO-transduced OVCAR5 cells. K, Cyclic AMP activity was detected in rolipram (1 μmol/L)- and vinpocetine (50 μmol/L)-treated shCTL or shFTO stably transduced OVCAR5 cells. L, Cyclic AMP activity was determined in ALDH+ cells compared with ALDH cells FACS-sorted from OVCAR5 and COV362 cells. M, Proposed mechanism by which FTO regulates tumor initiation and ovarian cancer stem cells. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Close modal

Similar observations were made using small-molecule inhibitors. Vinpocetine, an alkaloid derived from the Vinca plant, inhibits PDE1C, increasing levels of both cAMP and cGMP (38), which in turn, can activate protein kinase A (PKA) and alter transcription. Rolipram, inhibits cAMP hydrolysis by blocking PDE4B (39). Inhibition of both phosphodiesterases, especially of PDE1C, blocked spheroid formation (Supplementary Fig. S9A, P < 0.05) and expression of stem cells markers, ALDH, OCT4, NANOG, and SOX2 (Supplementary Fig. S9B and S9C, P < 0.05), supporting that FTO⊣PDE signaling plays a role in cancer stemness regulation. To demonstrate that the effects of phosphodiesterase inhibitors on stemness are dependent on FTO, we measured ALDH1A1 and SOX2 mRNA expression levels in response to the PDE1C and PDE4B inhibitors in stably transfected OVCAR5 cells with shRNA control or shRNA targeting FTO. The reduction in ALDH1A1 and SOX2 mRNA expression levels induced by the PDE1C inhibitor vinpocetine was partially rescued in sh-FTO compared with control cells (Supplementary Fig. S9D and S9E, P < 0.001). Similarly, the effects of the PDE4B inhibitor rolipram on ALDH1 and SOX2 mRNA expression levels were partially rescued in sh-FTO compared with control cells (Supplementary Fig. S9F and S9G, P < 0.05).

As the main function of phosphodiesterases is to regulate cAMP and cGMP balance through hydrolysis, we next measured whether modulation of FTO levels alters cAMP levels. Increased cAMP levels were found in FTO-overexpressing cells compared to control cells, consistent with downregulation of phosphodiesterase expression levels in FTO-overexpressing ovarian cancer cells (Fig. 7J, P < 0.05). Furthermore, cAMP levels augmented by treatment with PDE1C and PDE4B inhibitors were diminished in ovarian cancer cells, in which FTO was knocked down by shRNA (Fig. 7K), further supporting the role of this demethylase regulating the cAMP balance. Finally, we measured cAMP in CSCs versus non-CSCs sorted from ovarian cancer cell lines. Significantly decreased cAMP levels were found in ALDH+/CD133+ cells compared with ALDH/CD133 cells FACS-sorted from OVCAR5 and COV362 cell lines (Fig. 7L, P < 0.01), suggesting that cAMP signaling is significantly inhibited in CSCs. Collectively, these data support that m6A modified PDE1C and PDE4B regulate cancer stemness by modulating the intracellular cAMP balance, which is decreased in ovarian CSCs versus non-CSCs, as proposed in the model shown in Fig. 7M.

Our results define a previously unappreciated tumor-suppressive role of FTO in HGSOC, which is opposite to the tumor-promoting function of FTO previously reported in other types of cancers. Through both gain- and loss-of-function studies, we demonstrated that FTO is a tumor suppressor and inhibits the stemness features of ovarian CSCs. Through transcriptome-wide RNA-seq and m6A mapping, we identified two phosphodiesterase genes (PDE4B and PDE1C) as FTO target transcripts that regulate cAMP signaling and play a critical role in maintaining the stemness of ovarian CSCs. Our findings are highly significant, as they define for the first time a tumor-suppressive role for FTO in solid tumors and identify cAMP signaling as a key pathway modulated by m6A mRNA modifications in this context.

Alterations of m6A levels have been previously linked to abnormal cellular differentiation states in cancer (40). As a key regulatory enzyme for m6A modifications, studies of FTO have been initiated in cancer; with initial reports pointing to pro-tumorigenic functions in glioblastoma and leukemia. Overexpression of FTO was detected in brain tumors and an FTO inhibitor suppressed tumor formation and progression in glioblastoma models (41). FTO was also found to be highly expressed in AML, where it promoted leukemic cell growth and transformation by regulating m6A deposition at the 3′ UTR of ankyrin repeat and SOCS box protein 2 (ASB2) and the retinoic acid receptor α (RARA) involved in cell differentiation (21). R2-hydroxyglutarate (R-2HG), an oncometabolite that accumulates in leukemias or gliomas associated with mutant isocitrate dehydrogenases, exhibited anti-tumor activity by suppressing FTO and increasing global levels of m6A (22). The R-2HG-FTO axis contributed to enrichment in m6A peaks in the 5′UTR and CDS regions of MYC and CEBPA transcripts, suppressing their expression and function, and exerting anti-leukemogenic effects (22). In contrast, our results show that overexpression of FTO associated with decreased m6A levels suppressed tumorigenicity and sphere-forming capacity of ovarian CSCs, while FTO depletion increased tumorigenicity and stemness characteristics, providing compelling evidence for a tumor suppressor function. Importantly, a mutant FTO construct, lacking the demethylase function, did not alter the stemness phenotype, suggesting that the role of FTO in ovarian CSCs relies on its m6A demethylase activity.

The context-dependent function of FTO may be accounted for by tissue-specific targets regulating unique circuitries in distinct cancers. We show that FTO had significant effects upon the transcriptome of ovarian cancer cells with over 700 transcripts being significantly deregulated as a consequence of m6A modifications. Pathways related to stem cell signaling, RNA transcription, mRNA splicing, and DNA repair were significantly altered by overexpression of FTO. Among potential targets, two phosphodiesterases, PDE4B and PDE1C, were hypomethylated and downregulated in response to FTO overexpression. The mRNA half-life of these two transcripts was significantly altered by FTO overexpression, but not by the mutant construct, supporting the direct effect of this demethylase on PDE4B and PDE1C. M6A methylation is typically recognized by reader proteins, which guide the subsequent processing of mRNAs. The classical pathway involves YT521-B homology domain containing proteins, which recognize m6A modifications and direct mRNAs toward decay, resulting in decreased gene expression (6). Another mechanism involves alterations in the local structure of RNA, resulting in alternative splicing (42). Emerging data indicate that mRNAs with reduced m6A alterations could be downregulated due to increased instability (22). The two targets described here PDE1C and PDE4B fall into this category, being downregulated in response to decreased m6A mRNA modifications. A recently proposed mechanism accounting for this phenomenon involves IGF2BP1-3, which recognize m6A modified targets and stabilize them, leading to increased gene expression (7).

This is the first report linking FTO to regulation of cellular processes mediated by the second messenger cAMP. Interestingly, recent studies have associated cAMP and cGMP signaling with maintenance of CSCs in breast cancer models and inhibitors of PDE were shown to deplete ALDH+ populations (43). The main effector of cAMP is PKA, which phosphorylates several metabolic enzymes and the cAMP-response element-binding protein CREB (44). Here we show that cAMP levels were significantly downregulated in ovarian CSCs versus non-CSCs and that FTO overexpression induced increased cAMP levels, while its knockdown decreased them. Interestingly, PKA activation induced by cAMP was shown to cause epigenetic reprogramming, leading to differentiation of tumor initiating mammary cells (45), further supporting the significance of this pathway to inhibiting stemness. In all, our findings identify the cAMP pathway as a new FTO target in CSCs, associated with tumor-inhibiting functions, and provide key new information on the role of m6A modifications regulated by FTO in the fine tuning of cellular fate in HGSOC.

M. Kandpal reports grants from NIH (funding from 5R01LM011297-06, PI: Ramana Davuluri) during the conduct of the study. J. Chen reports other grants from Genovel Biotech Corp (scientific founder of the company) outside the submitted work. R.V. Davuluri reports grants from NIH (R01LM011297 grant from NLM) during the conduct of the study. D. Matei reports personal fees from Radius Inc. (consultant), GSK (consultant), Astra Zeneca (consultant), and GOG Foundation (mentor/consultant) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

H. Huang: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. Y. Wang: Data curation, formal analysis, validation, investigation, methodology, writing-review and editing. M. Kandpal: Software, formal analysis, visualization, methodology, writing-review and editing. G. Zhao: Software, formal analysis, visualization, methodology, writing-review and editing. H. Cardenas: Data curation, formal analysis, writing-review and editing. Y. Ji: Software, formal analysis, visualization, writing-review and editing. A. Chaparala: Data curation, validation, writing-review and editing. E.J. Tanner: Resources, data curation, formal analysis, writing-review and editing. J. Chen: Resources, formal analysis, methodology, writing-review and editing. R.V. Davuluri: Resources, formal analysis, visualization, methodology, writing-review and editing. D. Matei: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, writing-original draft, project administration, writing-review and editing.

This research was supported by the U.S. Department of Veterans Affairs (I01 BX000792-06), NCI (R01-CA224275), and the Diana Princess of Wales endowed Professorship from the Robert H. Comprehensive Cancer Center (to D. Matei). Tumor specimens were procured through the Tissue Pathology Core, sequencing was performed in the NUSeq Core, and flow cytometry analyses were performed in the Flow Cytometry Core Facility supported by NCI CCSG P30 CA060553 awarded to the Robert H. Lurie Comprehensive Cancer Center. This research was supported in part through the computational resources and staff contributions provided for the Quest high performance computing facility at Northwestern University, which is jointly supported by the Office of the Provost, the Office for Research, and Northwestern University Information Technology. J. Chen is supported by NIH R01 CA214965. We thank Dr. Marcus Peter for valuable comments.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Fu
Y
,
Dominissini
D
,
Rechavi
G
,
He
C
. 
Gene expression regulation mediated through reversible m(6)A RNA methylation
.
Nat Rev Genet
2014
;
15
:
293
306
.
2.
Yue
Y
,
Liu
J
,
He
C
. 
RNA N6-methyladenosine methylation in post-transcriptional gene expression regulation
.
Genes Dev
2015
;
29
:
1343
55
.
3.
Deng
X
,
Su
R
,
Weng
H
,
Huang
H
,
Li
Z
,
Chen
J
. 
RNA N(6)-methyladenosine modification in cancers: current status and perspectives
.
Cell Res
2018
;
28
:
507
17
.
4.
Bartosovic
M
,
Molares
HC
,
Gregorova
P
,
Hrossova
D
,
Kudla
G
,
Vanacova
S
. 
N6-methyladenosine demethylase FTO targets pre-mRNAs and regulates alternative splicing and 3′-end processing
.
Nucleic Acids Res
2017
;
45
:
11356
70
.
5.
Wang
X
,
Lu
Z
,
Gomez
A
,
Hon
GC
,
Yue
Y
,
Han
D
, et al
N6-methyladenosine-dependent regulation of messenger RNA stability
.
Nature
2014
;
505
:
117
20
.
6.
Wang
X
,
Zhao
BS
,
Roundtree
IA
,
Lu
Z
,
Han
D
,
Ma
H
, et al
N(6)-methyladenosine modulates messenger RNA translation efficiency
.
Cell
2015
;
161
:
1388
99
.
7.
Huang
H
,
Weng
H
,
Sun
W
,
Qin
X
,
Shi
H
,
Wu
H
, et al
Recognition of RNA N(6)-methyladenosine by IGF2BP proteins enhances mRNA stability and translation
.
Nat Cell Biol
2018
;
20
:
285
95
.
8.
Weng
H
,
Huang
H
,
Wu
H
,
Qin
X
,
Zhao
BS
,
Dong
L
, et al
METTL14 inhibits hematopoietic stem/progenitor differentiation and promotes leukemogenesis via mRNA m(6)A modification
.
Cell Stem Cell
2018
;
22
:
191
205
.
9.
Vu
LP
,
Pickering
BF
,
Cheng
Y
,
Zaccara
S
,
Nguyen
D
,
Minuesa
G
, et al
The N(6)-methyladenosine (m(6)A)-forming enzyme METTL3 controls myeloid differentiation of normal hematopoietic and leukemia cells
.
Nat Med
2017
;
23
:
1369
76
.
10.
Barbieri
I
,
Tzelepis
K
,
Pandolfini
L
,
Shi
J
,
Millan-Zambrano
G
,
Robson
SC
, et al
Promoter-bound METTL3 maintains myeloid leukaemia by m(6)A-dependent translation control
.
Nature
2017
;
552
:
126
31
.
11.
Yao
QJ
,
Sang
L
,
Lin
M
,
Yin
X
,
Dong
W
,
Gong
Y
, et al
Mettl3-Mettl14 methyltransferase complex regulates the quiescence of adult hematopoietic stem cells
.
Cell Res
2018
;
28
:
952
4
.
12.
Lin
S
,
Choe
J
,
Du
P
,
Triboulet
R
,
Gregory
RI
. 
The m(6)A methyltransferase METTL3 promotes translation in human cancer cells
.
Mol Cell
2016
;
62
:
335
45
.
13.
Zhang
S
,
Zhao
BS
,
Zhou
A
,
Lin
K
,
Zheng
S
,
Lu
Z
, et al
m(6)A demethylase ALKBH5 maintains tumorigenicity of glioblastoma stem-like cells by sustaining FOXM1 expression and cell proliferation program
.
Cancer Cell
2017
;
31
:
591
606
.
14.
Liu
J
,
Eckert
MA
,
Harada
BT
,
Liu
SM
,
Lu
Z
,
Yu
K
, et al
m(6)A mRNA methylation regulates AKT activity to promote the proliferation and tumorigenicity of endometrial cancer
.
Nat Cell Biol
2018
;
20
:
1074
83
.
15.
Fukumoto
T
,
Zhu
H
,
Nacarelli
T
,
Karakashev
S
,
Fatkhutdinov
N
,
Wu
S
, et al
N(6)-methylation of adenosine of FZD10 mRNA contributes to PARP inhibitor resistance
.
Cancer Res
2019
;
79
:
2812
20
.
16.
Frayling
TM
,
Timpson
NJ
,
Weedon
MN
,
Zeggini
E
,
Freathy
RM
,
Lindgren
CM
, et al
A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity
.
Science
2007
;
316
:
889
94
.
17.
Gulati
P
,
Yeo
GS
. 
The biology of FTO: from nucleic acid demethylase to amino acid sensor
.
Diabetologia
2013
;
56
:
2113
21
.
18.
Gerken
T
,
Girard
CA
,
Tung
YC
,
Webby
CJ
,
Saudek
V
,
Hewitson
KS
, et al
The obesity-associated FTO gene encodes a 2-oxoglutarate-dependent nucleic acid demethylase
.
Science
2007
;
318
:
1469
72
.
19.
Jia
G
,
Fu
Y
,
Zhao
X
,
Dai
Q
,
Zheng
G
,
Yang
Y
, et al
N6-methyladenosine in nuclear RNA is a major substrate of the obesity-associated FTO
.
Nat Chem Biol
2011
;
7
:
885
7
.
20.
Hess
ME
,
Hess
S
,
Meyer
KD
,
Verhagen
LA
,
Koch
L
,
Bronneke
HS
, et al
The fat mass and obesity associated gene (Fto) regulates activity of the dopaminergic midbrain circuitry
.
Nat Neurosci
2013
;
16
:
1042
8
.
21.
Li
Z
,
Weng
H
,
Su
R
,
Weng
X
,
Zuo
Z
,
Li
C
, et al
FTO plays an oncogenic role in acute myeloid leukemia as a N(6)-methyladenosine RNA demethylase
.
Cancer Cell
2017
;
31
:
127
41
.
22.
Su
R
,
Dong
L
,
Li
C
,
Nachtergaele
S
,
Wunderlich
M
,
Qing
Y
, et al
R-2HG exhibits anti-tumor activity by targeting FTO/m(6)A/MYC/CEBPA signaling
.
Cell
2018
;
172
:
90
105
.
23.
Vaughan
S
,
Coward
JI
,
Bast
RC
 Jr
,
Berchuck
A
,
Berek
JS
,
Brenton
JD
, et al
Rethinking ovarian cancer: recommendations for improving outcomes
.
Nat Rev Cancer
2011
;
11
:
719
25
.
24.
Zhang
S
,
Balch
C
,
Chan
MW
,
Lai
HC
,
Matei
D
,
Schilder
JM
, et al
Identification and characterization of ovarian cancer-initiating cells from primary human tumors
.
Cancer Res
2008
;
68
:
4311
20
.
25.
Nguyen
LV
,
Vanner
R
,
Dirks
P
,
Eaves
CJ
. 
Cancer stem cells: an evolving concept
.
Nat Rev Cancer
2012
;
12
:
133
43
.
26.
Silva
IA
,
Bai
S
,
McLean
K
,
Yang
K
,
Griffith
K
,
Thomas
D
, et al
Aldehyde dehydrogenase in combination with CD133 defines angiogenic ovarian cancer stem cells that portend poor patient survival
.
Cancer Res
2011
;
71
:
3991
4001
.
27.
Chefetz
I
,
Grimley
E
,
Yang
K
,
Hong
L
,
Vinogradova
EV
,
Suciu
R
, et al
A pan-ALDH1A inhibitor induces necroptosis in ovarian cancer stem-like cells
.
Cell Rep
2019
;
26
:
3061
75
.
28.
Condello
S
,
Morgan
CA
,
Nagdas
S
,
Cao
L
,
Turek
J
,
Hurley
TD
, et al
beta-Catenin-regulated ALDH1A1 is a target in ovarian cancer spheroids
.
Oncogene
2015
;
34
:
2297
308
.
29.
Condello
S
,
Sima
L
,
Ivan
C
,
Cardenas
H
,
Schiltz
G
,
Mishra
RK
, et al
Tissue tranglutaminase regulates interactions between ovarian cancer stem cells and the tumor niche
.
Cancer Res
2018
;
78
:
2990
3001
.
30.
Zhu
J
,
Xu
Y
,
Rashedi
AS
,
Pavone
ME
,
Kim
JJ
,
Woodruff
TK
, et al
Human fallopian tube epithelium co-culture with murine ovarian follicles reveals crosstalk in the reproductive cycle
.
Mol Hum Reprod
2016
;
22
:
756
67
.
31.
Vivian
J
,
Rao
AA
,
Nothaft
FA
,
Ketchum
C
,
Armstrong
J
,
Novak
A
, et al
Toil enables reproducible, open source, big biomedical data analyses
.
Nat Biotechnol
2017
;
35
:
314
6
.
32.
The Cancer Genome Atlas Research Network
,
Weinstein
JN
,
Collisson
EA
,
Mills
GB
,
Shaw
KR
,
Ozenberger
BA
, et al
The Cancer Genome Atlas Pan-Cancer analysis project
.
Nat Genet
2013
;
45
:
1113
20
.
33.
Li
J
,
Condello
S
,
Thomes-Pepin
J
,
Ma
X
,
Xia
Y
,
Hurley
TD
, et al
Lipid desaturation is a metabolic marker and therapeutic target of ovarian cancer stem cells
.
Cell Stem Cell
2017
;
20
:
303
14
.
34.
Wang
Y
,
Cardenas
H
,
Fang
F
,
Condello
S
,
Taverna
P
,
Segar
M
, et al
Epigenetic targeting of ovarian cancer stem cells
.
Cancer Res
2014
;
74
:
4922
36
.
35.
Meng
E
,
Mitra
A
,
Tripathi
K
,
Finan
MA
,
Scalici
J
,
McClellan
S
, et al
ALDH1A1 maintains ovarian cancer stem cell-like properties by altered regulation of cell cycle checkpoint and DNA repair network signaling
.
PLoS One
2014
;
9
:
e107142
.
36.
Hu
Y
,
Smyth
GK
. 
ELDA: extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays
.
J Immunol Methods
2009
;
347
:
70
8
.
37.
Bailey
TL
,
Boden
M
,
Buske
FA
,
Frith
M
,
Grant
CE
,
Clementi
L
, et al
MEME SUITE: tools for motif discovery and searching
.
Nucleic Acids Res
2009
;
37
:
W202
8
.
38.
Medina
AE
. 
Vinpocetine as a potent antiinflammatory agent
.
Proc Natl Acad Sci U S A
2010
;
107
:
9921
2
.
39.
Kinoshita
T
,
Kobayashi
S
,
Ebara
S
,
Yoshimura
Y
,
Horiuchi
H
,
Tsutsumimoto
T
, et al
Phosphodiesterase inhibitors, pentoxifylline and rolipram, increase bone mass mainly by promoting bone formation in normal mice
.
Bone
2000
;
27
:
811
7
.
40.
He
L
,
Li
J
,
Wang
X
,
Ying
Y
,
Xie
H
,
Yan
H
, et al
The dual role of N6-methyladenosine modification of RNAs is involved in human cancers
.
J Cell Mol Med
2018
;
22
:
4630
9
.
41.
Cui
Q
,
Shi
H
,
Ye
P
,
Li
L
,
Qu
Q
,
Sun
G
, et al
m(6)A RNA methylation regulates the self-renewal and tumorigenesis of glioblastoma stem cells
.
Cell Rep
2017
;
18
:
2622
34
.
42.
Liu
N
,
Dai
Q
,
Zheng
G
,
He
C
,
Parisien
M
,
Pan
T
. 
N(6)-methyladenosine-dependent RNA structural switches regulate RNA-protein interactions
.
Nature
2015
;
518
:
560
4
.
43.
Klutzny
S
,
Anurin
A
,
Nicke
B
,
Regan
JL
,
Lange
M
,
Schulze
L
, et al
PDE5 inhibition eliminates cancer stem cells via induction of PKA signaling
.
Cell Death Dis
2018
;
9
:
192
.
44.
Sassone-Corsi
P
. 
The cyclic AMP pathway
.
Cold Spring Harb Perspect Biol
2012
;
4
:
a011148. DOI: 10.1101/cshperspect.a011148
.
45.
Pattabiraman
DR
,
Bierie
B
,
Kober
KI
,
Thiru
P
,
Krall
JA
,
Zill
C
, et al
Activation of PKA leads to mesenchymal-to-epithelial transition and loss of tumor-initiating ability
.
Science
2016
;
351
:
aad3680
. DOI: 10.1126/science.aad3680.

Supplementary data