FTO suppresses glycolysis and growth of papillary thyroid cancer via decreasing stability of APOE mRNA in an N6-methyladenosine-dependent manner

A total of 150 paired PTC tissue samples and their adjacent non-cancerous thyroid tissue samples were obtained from the Thyroid Surgery Department of the First Hospital of China Medical University between 2017 and 2018. One hundred pairs of PTC tissues and adjacent non-cancerous tissues (Cohort 1) were used in quantitative real-time reverse transcription PCR (qRT-PCR), 30 pairs (Cohort 2) were used for immunohistochemistry and 20 pairs (Cohort 3) were used for western blotting. Immunohistochemistry was performed on formalin-fixed and paraffin-embedded (FFPE) tissue specimens, while qRT-PCR and western blotting were performed on fresh frozen tissue specimens. All tissue specimens were independently confirmed by two pathologists. Written informed consent was obtained from all study participants. This study was approved by the Ethics Committee of the First Hospital of China Medical University. Human thyroid follicular epithelial cell line Nthy-ori3–1 and PTC cell lines including TPC1, K1, IHH4 and BCPAP were used in the current study. The source and culture method of these cell lines are described in additional file 2: supplementary materials and methods.

FTO, APOE, IGF2BP1, IGF2BP2, IGF2BP3 and NC siRNAs were obtained from Gene Pharma (Suzhou, China), their sequences are presented in Additional file 1 (Table S2). Transfections were performed using Lipofectamine 3000 (Invitrogen, Waltham, MA, USA) according to the manufacturer’s instructions. Recombinant lentiviruses containing sh-FTO or sh-APOE were constructed to stably knockdown FTO or APOE stably. Non-targeting shRNA (sh-NC) was used as negative control. FTO or APOE were overexpressed using recombinant lentiviruses containing complete coding sequences of these genes. In addition, we cloned the amplified mutant FTO-CDS from pcDNA3.1_mutFTO (carrying H231A and D233A, which disrupt the demethylase activity of FTO) into lentivector-based pMIRNA1 according to the methods of Li et al. [31] and Jia et al. [32]. Empty lentivirus (Vector) was used as negative control. All lentivirus vectors were constructed by Obio Technology (Shanghai, China). Infected cells were selected by puromycin. Transfection and infection efficiency were determined using qRT-PCR and western blotting.

Extra cellular acidification rate (ECAR) and oxygen consumption rate (OCR) were determined using Seahorse XF Glycolysis Stress Test Kit and Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, Palo Alto, CA), respectively. Transfected PTC cells were seeded into Seahorse dedicated 96-well cell culture plates overnight. Glucose, oligomycin and 2-deoxy-glucose (2-DG) were sequentially added to the corresponding well on the sensor cartridge for ECAR measurement. Oligomycin, FCCP and Antimycin A & Rotenone were sequentially added to the corresponding well for OCR determination. The prepared cell plate was then analyzed by Seahorse XFe96 Analyzer and data were analyzed using Seahorse Wave software.

Glucose uptake was determined using Glucose Colorimetric Assay Kit (BioVision, USA) according to the manufacturer’s protocols. Transfected PTC cells were seeded into 6-well plates and incubated for 48 h. Glucose enzyme mix specifically oxidizes glucose, to generate a product which reacts with a dye to generate color (OD 570 nm). Glucose content in the cell culture medium was determined and cells in each well were counted to normalize glucose concentration. Glucose uptake was determined indirectly by determining the remained glucose content in cell culture medium.

Lactate production was determined using Lactate Colorimetric Assay Kit (BioVision, USA) according to the manufacturer’s protocols. Lactate specifically reacts with an enzyme mix to generate a product, which interacts with lactate probe to produce color (OD 570 nm). Lactate production in the cell culture medium was detected and cells in each well were counted to normalize lactate concentration.

Fragments from APOE containing the wild-type m⁶A motifs (APOE-Wt) and mutant m⁶A motifs (m⁶A was replaced by C, APOE-Mut) were subcloned into pMIR-REPORT firefly luciferase vector (Obio Technology, Shanghai, China). Wild type and mutant APOE luciferase plasmids were co-transfected with empty, sh-FTO and FTO plasmids, using Renilla luciferase expressing vector pRL-TK (Promega, Madison, WI, USA). Relative luciferase activity was determined using dual-luciferase reporter assay system (Promega, Madison, USA). Relative firefly luciferase activity normalized to Renilla luciferase activity was then determined.

RNA-seq and MeRIP-seq were performed by Cloud-Seq Biotech (Shanghai, China). rRNAs were eliminated from the total RNA before performing RNA-seq and RNA libraries were constructed. Quality control and quantification of the library was conducted and RNA-seq was performed on Illumina HiSeq instrument using 150 bp paired-ended mode. For m⁶A-sequencing, MeRIP was performed using GenSeqTM m⁶A-MeRIP Kit (GenSeq, China) following the manufacturer’s instructions. Both the input samples without immunoprecipitation and the m⁶A IP samples were used for construction of the RNA-seq library. m⁶A-sequencing was performed on Illumina HiSeq instrument using 150 bp paired-ended mode after evaluating quality of the library. We extracted the sequences of the first 1000 peaks with the largest enrichment multiples for each group (50 bp on both sides of the vertex) from MeRIP-seq data and then scanned them with dreme software to identify meaningful motif sequences [10, 33].

Total RNA was extracted by RNAiso Plus (Takara, Japan), and DNAse was added to remove DNA. MeRIP assays were performed using the m⁶A RNA Enrichment Kit (Epigentek, USA) according to the manufacturer’s protocols. Target m⁶A-containing fragments were pulled down using a beads-bound m⁶A capture antibody, and RNA sequences in both ends of the m⁶A-containing regions were cleaved using Cleavage Enzyme Mix. The enriched RNA was then released, purified and eluted. QRT-PCR was performed following MeRIP to quantify changes in target gene m⁶A methylation.

RNA immunoprecipitation (RIP) assays were conducted using Magna RIP Kit (Millipore, New Bedford, MA) according to the manufacturer’s protocols. Cells were lysed with appropriate amount of complete RIP Lysis Buffer. RNA-binding protein were immunoprecipitated using anti-IGF2BP2 antibody (Proteintech, USA) and normal rabbit IgG. The co-precipitated RNAs were purified and dissolved in RNase-free water. Binding RNA targets were analyzed using qRT-PCR.

Five weeks old female athymic BALB/c nude mice were purchased from Beijing HFK bioscience Co., Ltd. (Beijing, China) and used to construct the xenograft tumor model. Transfected K1 or TPC1 cells were injected into the subcutaneous tissue of the mouse. After 4 weeks, the mice were euthanized, and tumors were excised and weighed. Tumor volume was calculated using the formula: Tumor volume (mm³) = longer diameter × shorter diameter²/2. Mice were fasted for 8 h and administered with approximately 200-300 μCi of ¹⁸F-FDG through the lateral tail vein after being anesthetized by 1% pentobarbital sodium a day before they were euthanized. Mice were maintained in a cage at room temperature for 1 h, then placed in the prone position on the examination bed for micro-PET and micro-CT imaging. ¹⁸F-FDG uptake was quantified by drawing region of interest (ROI) using Metis Viewer software and plotting maximum standardized uptake values (SUVmax). All animal experiments were conducted in accordance with the principles and procedures outlined in the guidelines of the Institutional Animal Care and Use Committee of China Medical University.

Detailed materials and methods are presented in Additional file 2: supplementary materials and methods.

To explore the role of the m⁶A modification in PTC, differential expression of m⁶A modification enzymes was determined in PTC tissues and normal thyroid tissues using public clinical databases including The Cancer Genome Atlas (TCGA) and GEO. Analysis using TCGA dataset showed that mRNA of the methylases METTL3, METTL14 and WTAP, as well as the demethylases FTO and ALKBH5 were significantly downregulated in PTC tissues compared with the expression level in normal tissues (Fig. 1a). In addition, the findings showed that expression of FTO was significantly downregulated in the 3 GEO datasets (GSE60542, GSE27155 and GSE3467) of PTC tissues compared with the expression in normal tissues (Fig. 1b). Expression level of m⁶A modification enzymes in the 30 pairs of PTC tissue samples in the current study was determined by qRT-PCR and the findings showed that the mRNA expression of FTO was downregulated in the PTC tissues (Fig. 1c). Consistent with our findings, the overall level of m⁶A was higher in PTC tissues compared with the level in the paired non-cancerous tissues (Fig. 1d). To further explore downregulation of FTO in PTC tissues, the tissue sample size was increased to 100 pairs and qRT-PCR analysis showed that FTO mRNA was downregulated in PTC tissues compared with the expression level in paired non-cancerous tissues (Fig. 1e). Analysis of the relationship between FTO expression and clinicopathological features showed that expression of FTO was significantly negatively correlated with extrathyroidal extension (p = 0.004) and lymph node metastasis (p = 0.035) (Table 1). In addition, we found that FTO expression was significantly negatively correlated with tumor size (p = 0.031) in TCGA database (Additional file 1: Fig. S4). Further, protein expression of FTO was determined in 30 paired PTC tissue samples by immunohistochemistry. The findings showed that FTO protein level was significantly downregulated in PTC tissues and mainly localized in the nuclei of PTC cells (Fig. 1f). Moreover, analysis by western blotting showed downregulation of protein expression of FTO in PTC tissues (Fig. 1g). These findings indicate that FTO is significantly downregulated in PTC and may play a role in regulation of PTC progression through m⁶A modification.

Relative expression of FTO was determined in a normal human thyroid follicular epithelial cell line Nthy-ori3–1 and four PTC cell lines (K1, BCPAP, IHH4, and TPC1). The findings showed that FTO expression was significantly lower in TPC1 and K1 cell lines compared with the level in the Nthy-ori3–1 cell line. However, FTO expression in IHH4 and BCPAP was slightly lower compared with the level in the Nthy-ori3–1 cell line (Fig. 2a). Therefore, KI and TPC1 cell lines which showed consistency with the expression trend of FTO in PTC tissues for subsequent experiments. Effect of FTO on cell proliferation in vitro and in vivo was explored to determine the role of FTO in PTC. Efficiency of FTO knockdown and overexpression of FTO was verified by qRT-PCR and western blotting (Fig. 2b-c and Additional file 3: Fig. S1a-b). The findings showed that knockdown or overexpression of FTO increased or decreased the global m⁶A modification level in PTC cells, respectively (Fig. 2d and Additional file 3: Fig. S1c). CCK-8 assays showed that FTO knockdown significantly promoted proliferation of PTC cells, and similar findings were obtained using colony formation assays (Fig. 2e, f and Additional file 3: Fig. S1f, g). Furthermore, it showed that overexpression of FTO inhibited PTC cell proliferation, whereas FTO-mut TPC1 cells with deletion of the demethylation domain showed no effect on proliferation (Fig. 2g, h and Additional file 3: Fig. S1d, e). These findings indicate that FTO-induced PTC proliferation depends on its demethylases function. In addition, we found that the proportion of cells decreased significantly in G0/G1 phase and increased significantly in S phase after FTO knockdown through cell cycle analysis, whereas FTO overexpression showed the opposite effects (Additional file 3: Fig. S1h, i). This result shows that FTO could block the transition of cell cycle from G0/G1 phase to S phase, thereby inhibiting its proliferation. Subsequently, we constructed sh-NC and sh-FTO stable cell lines and performed nude mice xenograft model experiments after confirming that FTO expression in sh-FTO group was significantly lower than that in sh-NC group (Additional file 3: Fig. S1j). Findings from in vivo experiments showed that knockdown of FTO significantly increased K1 tumor growth (Fig. 2i, j) and tumor weight (Fig. 2k) in xenograft mouse models. Furthermore, overexpression of FTO significantly reduced TPC1 tumor growth (Fig. 2l, m) and tumor weight (Fig. 2n) in xenograft mouse models. However, effect of FTO on tumor growth was not observed in vivo after overexpression of FTO-mut (Fig. 2l-n). These findings indicate that FTO inhibits tumor growth in PTC through its demethylation domain.

Further analysis was conducted to explore the mechanism of action of FTO in modulating growth of PTC. FTO can maintain energy homeostasis by controlling energy consumption [34]. However, studies have not fully elucidated whether FTO plays a role in energy metabolism in PTC. To explore the regulatory effect of FTO on PTC energy metabolism, the effect of FTO on the glycolytic metabolism of PTC was determined through ECAR of Seahorse metabolic analysis. The findings showed that knockdown of FTO significantly increased ECAR levels in K1 and TPC1 cells (Fig. 3a and Additional file 3: Fig. S2a), implying that it can inhibit glycolysis in PTC. Moreover, analysis showed that OCR level was decreased after FTO knockdown, implying that it promoted mitochondrial oxidative phosphorylation in PTC, which also shows its inhibitory effect on glycolysis from the side (Additional file 3: Fig. S2b, c). Cancer cells increase glucose uptake to compensate for the lower efficiency of ATP produced by glycolysis compared with mitochondrial oxidative phosphorylation thus promoting uncontrolled cell growth and division. Glycolysis results in high amounts of lactate which can be utilized as the main energy source by other cells, and glycolytic intermediates may be diverted into various biosynthetic pathways [23]. Extracellular glucose analysis showed that glucose uptake was significantly increased after FTO knockdown in PTC cells (Fig. 3b and Additional file 3: Fig. S2d). Extracellular lactate level was determined as an indicator of glycolysis and the findings showed significantly increased lactate production in PTC cells after FTO knockdown (Fig. 3c and Additional file 3: Fig. S2e). The process of aerobic glycolysis involves several enzymes. Further analysis of regulatory mechanism of glycolysis was performed by determining expression levels of key enzymes in glycolysis after FTO knockdown. QRT-PCR analysis showed that GLUT1, HK-II and LDHA mRNA level were significantly upregulated after FTO knockdown in K1 cells (Fig. 3d). Western blotting showed that protein expression level of GLUT1, HK-II and LDHA were significantly upregulated after FTO knockdown (Fig. 3e and Additional file 3: Fig. S2f), further confirming the effect of FTO on glycolytic enzymes. PTC cells were treated with glycolysis inhibitor, 2-DG, and the findings showed that 2-DG inhibited proliferation and colony formation caused by FTO knockdown (Fig. 3f, g and Additional file 3: Fig. S2g, h). This finding indicated that FTO affects cell proliferation by modulating glycolysis. Further, effect of FTO on glycolysis in vivo was determined using ¹⁸F-FDG PET ([¹⁸F]-fluoro-2-deoxyglucose positron emission tomography) to determine glucose uptake in mice tumor. PET-CT findings showed that glucose uptake in xenograft mouse tumor model was significantly increased after FTO knockdown compared with the control (Fig. 3h).

Gain of function assays showed that overexpression of FTO significantly decreased ECAR levels, glucose uptake and lactate production in PTC cells (Fig. 3i-k and Additional file 3: Fig. S2i-k). Furthermore, OCR levels were significantly increased in PTC cells (Additional file 3: Fig. S2l, m). Moreover, GLUT1, HK-II and LDHA protein expression were significantly downregulated after FTO overexpression (Fig. 3l and Additional file 3: Fig. S2n). Overexpression of FTO significantly reduced glucose uptake in mice (Fig. 3m). However, the effect on glycolysis was not observed both in vitro and in vivo after overexpression of FTO-mut (Fig. 3i-m and Additional file 1: Fig. S2m), indicating that FTO modulates glycolysis through its demethylation domain. In summary, these findings indicate that FTO inhibits PTC progression by modulating glycolysis.

To explore the downstream potential targets of FTO in PTC, RNA-seq was performed to determined differentially expressed genes (DEGs) of the si-NC group and si-FTO group using K1 cells. The findings showed that 490 genes were significantly downregulated, whereas 675 genes were significantly upregulated after FTO knockdown (Fig. 4a). To verify whether changes in gene expression were attributed to FTO-mediated m⁶A modification, MeRIP-seq was performed to identify significantly elevated m⁶A modification after FTO knockdown in K1 cells. MeRIP-seq findings showed 2553 unique m⁶A peaks in si-NC group, 1989 unique m⁶A peaks in si-FTO group and 8496 shared m⁶A peaks in both groups (Fig. 4b). Consistent with previous studies, analysis showed that m⁶A peaks were highly enriched around the stop codon in si-NC and si-FTO groups (Fig. 4c). Mapping of m⁶A methylomes in K1 cells showed that ‘GGACU’ which was highly enriched in the m⁶A peaks was the most common m⁶A motif (Fig. 4d). The average peak lengths of si-NC and si-FTO group are 1002.86 and 1037.31 respectively. Gene ontology (GO) analysis showed that the differentially expressed transcripts were significantly enriched in gene sets involved in glucose transmembrane transporter activity, indicating that FTO plays an important role in glycolysis (Additional file 3: Fig. S3a). Pathway analysis showed that ‘Pathways in cancer’ was significantly correlated with FTO-mediated m⁶A modification (Additional file 3: Fig. S3b). Then, we combined RNA-seq and MeRIP-seq data according to the items with the same gene name and screened out mRNAs that were differentially expressed and had significantly higher m⁶A modification levels following FTO knockdown according to the criteria of fold change > 2 and p-value < 1.0e-5. After obtaining 356 DEGs (353 upregulated genes and 3 downregulated genes) encompassing 628 increased m⁶A peaks from sequencing data, we queried their expression in PTC tissues and normal thyroid tissues using public databases such as TCGA database. Subsequently, we employed qRT-PCR to identify the expression of DEGs in PTC, which were predicted from the public database in our own paired PTC tissue samples. Finally, we identified seven genes containing 13 m⁶A peaks for further functional experiments and discovered that APOE might be the key gene for FTO-mediated m⁶A modification as it could modulate glycolytic metabolism of PTC. (Fig. 4e). Analysis of MeRIP-seq data using Integrative Genomics Viewer (IGV) software showed that APOE mRNA had one m⁶A peak, which was detected near the stop codon and was significantly higher after FTO knockdown (Fig. 4f). However, expression and biological function of APOE in PTC should be further elucidated. Therefore, APOE was selected as the optimal target for FTO-mediated m⁶A modification for subsequent analysis.

Consistent with sequencing analysis, the findings showed that knockdown or overexpression of FTO upregulated or downregulated expression of APOE at mRNA and protein levels (Fig. 4g-j and Additional file 3: Fig. S3c-f). However, APOE expression showed no difference after overexpression of FTO-mut (Fig. 4h, j). To explore m⁶A modification of APOE by FTO, 3-Deazaadenosine (DAA, an inhibitor of methylation) was used and the findings showed that DAA significantly decreased expression of APOE that was increased by FTO knockdown (Fig. 4k and Additional file 3: Fig. S3g). FTO-mediated m⁶A modification of APOE mRNA was preliminarily verified. To further explore this effect, m⁶A-seq data was validated by performing MeRIP-PCR. The findings showed that anti-m⁶A antibody significantly enriched APOE mRNA level in K1 cells (Fig. 4l). Moreover, MeRIP-qPCR showed that m⁶A levels of APOE mRNA were significantly increased after FTO knockdown (Fig. 4m). Subsequently, luciferase reporters containing either the wild-type or mutant APOE were constructed to explore the effect of m⁶A modification on APOE expression. Findings from luciferase reporter assays showed that luciferase activity of wild-type APOE was significantly increased after FTO knockdown whereas overexpression of APOE reduced luciferase activity. However, knocking down or overexpressing FTO showed no effect on the luciferase activity of mut-type APOE (Fig. 4n, o and Additional file 3: Fig. S3h, i). These findings indicate that FTO plays a role in m⁶A modification of APOE.

Expression of APOE increased significantly with increase in the level of m⁶A modification, implying that its m⁶A modification may increase stability of APOE mRNA [7]. Subsequently, RNA stability assays were conducted and RNA stability curves showed that half-life of APOE mRNA was significantly prolonged after FTO knockdown in K1 and TPC1 cells (Fig. 4p and Additional file 3: Fig. S3j). Therefore, upregulation of APOE induced by FTO can be attributed to increase in stability of APOE mRNA caused by m⁶A modification. Although m⁶A modification can be modulated by ‘writers’ and ‘erasers’, its functional outcomes are depended on selective recognition of m⁶A sites by ‘readers’. IGF2BP1–3 modulates stability of mRNA plays a role as a m⁶A reader [16]. Analysis showed that knocking down IGF2BP2, but not IGF2BP1/3 significantly downregulated APOE mRNA in K1 and TPC1 cells (Fig. 4q and Additional file 3: Fig. S3k). Moreover, half-life of APOE mRNA was significantly decreased after IGF2BP2 knockdown in K1 and TPC1 cells (Fig. 4r and Additional file 3: Fig. S3l). Analysis of the mechanism using RIP assay showed that IGF2BP2 directly bound to the m⁶A sites of APOE (Fig. 4s). Further qRT-PCR analysis for RIP assay showed that knockdown of FTO significantly increased binding efficiency of IGF2BP2 to APOE mRNA (Fig. 4t). In summary, these findings indicate that FTO inhibits APOE mRNA stability and expression through m⁶A modification mediated by IGF2BP2.

Although the vital role of FTO in PTC growth by glycolysis were confirmed, it was unclear whether the effect was specifically attributed to APOE, the target gene for m⁶A modification downstream of FTO. To explore the role of APOE in PTC, expression level of APOE in cohort 1 was determined by qRT-PCR. The findings showed that mRNA expression of APOE was significantly upregulated in PTC tissues compared with the level in paired non-cancerous tissues (Fig. 5a). Analysis of the relationship between APOE expression and its clinicopathological features, showed that expression of APOE was significantly positively correlated with tumor size (p = 0.005) and extrathyroidal invasion (p = 0.028) (Table 2). In addition, we found that APOE expression was significantly negatively linked to age (p < 0.001), extrathyroidal invasion (p = 0.012) and TNM stage (p = 0.035) in TCGA database (Additional file 1: Fig. S5). In cohort 1, expression of APOE in PTC was significantly negatively correlated with expression level of FTO (Fig. 5b). Consistent with our results, the expression of FTO and APOE in TCGA database also showed a significantly negative correlation (R = -0.259, p = 2.96e-09) (Fig. 5c). In addition, protein expression of APOE was determined in PTC paired tissue samples of cohort 2. The findings showed that APOE protein expression was significantly higher in PTC tissues compared with adjacent non-cancerous tissues (Fig. 5d). Subsequently, the biological function of APOE in PTC growth was explored. Efficiency of APOE knockdown was confirmed by qRT-PCR and western blotting (Additional file 3: Fig. S4a-d). To explore whether APOE plays an important role as the target gene of FTO in PTC growth, rescue experiments were conducted to explore if APOE knockdown can reverse the effects of FTO knockdown. The findings showed that downregulation of APOE expression significantly reduced high proliferation levels induced by si-FTO (Fig. 5e, f and Additional file 3: Fig. S4e, f). Moreover, knocking down APOE significantly reduced the increased ECAR level induced by si-FTO, implying that APOE can reverse increase in glycolysis rate caused by FTO (Fig. 5g and Additional file 3: Fig. S4g). Moreover, the reduced OCR level induced by si-FTO was rescued by si-APOE, indicating the role of APOE in promoting glycolysis from another point of view (Additional file 3: Fig. S4h, i). In addition, extracellular glucose and lactate analysis showed that APOE knockdown reversed increased glucose uptake and lactate production induced by si-FTO (Fig. 5h, i and Additional file 3: Fig. S4j, k). Analysis of the effect of APOE on expression of key glycolysis proteins showed that only GLUT1 protein expression was significantly downregulated after APOE knockdown, whereas expression levels of HK-II and LDHA were not affected (Fig. 5j and Additional file 3: Fig. S4l). This finding indicates that FTO and APOE mainly affects glycolysis by regulating glucose transport. To explore whether APOE affects cell proliferation through glycolysis, PTC cells overexpressing APOE were treated with 2-DG. Similar to FTO, 2-DG inhibited the increased proliferation caused by overexpression of APOE (Fig. 5k and Additional file 3: Fig. S4m-o), indicating that APOE affects proliferation through glycolysis. In vivo analysis showed that sh-APOE significantly impaired tumor growth (Fig. 5l, m) and tumor weight (Fig. 5n) in xenograft mouse models caused by sh-FTO. Furthermore, micro-PET analysis showed that sh-APOE significantly reduced glucose uptake induced by sh-FTO in xenograft mouse models (Fig. 5o). These findings indicate that APOE could promote tumor growth through glycolysis and is regulated by FTO-mediated m⁶A modification in PTC.

To explore the related signaling pathway downstream of FTO/APOE axis, Gene set enrichment analysis (GSEA) was performed to identify the enriched signature of FTO and APOE in PTC. GSEA analysis found that FTO and APOE have their own related signaling pathways, but the common signaling pathway of FTO/APOE downstream was only the IL-6/JAK/STAT3 signaling pathway. GSEA based on the RNA-seq data showed that IL-6/JAK/STAT3 signaling pathway was highly enriched in si-FTO group compared with the si-NC group (Fig. 6a). Moreover, GSEA based on transcriptome data of APOE expression in PTC from the TCGA database showed that IL-6/JAK/STAT3 signaling pathway was highly enriched in APOE high-expression group compared with the low-expression group (Fig. 6b). In addition, the APOE high-expression group were highly enriched in thyroid cancer (Additional file 3: Fig. S5a). These GSEA findings indicated that FTO and APOE may affect glycolysis of PTC by modulating IL6/JAK/STAT3 signaling pathway. Subsequently, effects of FTO and APOE on expression of related proteins in this pathway were explored by western blotting. The findings showed that the protein expression levels of phosphorylated JAK2 (pJAK2) and STAT3 (pSTAT3) in this pathway were significantly increased after FTO knockdown, whereas FTO overexpression showed the opposite effects. (Fig. 6c, d and Additional file 3: Fig. S5b, c). In addition, we found that the protein expression levels of these two phosphorylated protein were significantly decreased after APOE knockdown, whereas APOE overexpression showed the opposite effects. (Fig. 6e, f and Additional file 3: Fig. S5d, e). Furthermore, the findings showed that APOE knockdown significantly abrogated the increased pJAK2, pSTAT3 and GLUT1 protein expression level caused by si-FTO (Fig. 6g and Additional file 3: Fig. S5f). In summary, these findings indicate that APOE exerts its effect on glycolysis in PTC tissues modulating IL6/JAK2/STAT3 signaling pathway after m⁶A modification by FTO.