Research ArticleMetabolism

Identification of entacapone as a chemical inhibitor of FTO mediating metabolic regulation through FOXO1

See allHide authors and affiliations

Science Translational Medicine  17 Apr 2019:
Vol. 11, Issue 488, eaau7116
DOI: 10.1126/scitranslmed.aau7116

The skinny on FTO

Although the fat mass and obesity-associated gene FTO has been linked to genetic risk of obesity, the mRNA demethylase that it encodes has proven difficult to therapeutically target. From a screen of approved drugs, Peng et al. identified entacapone, a catechol-O-methyltransferase inhibitor used in the treatment of Parkinson’s disease, as an inhibitor of FTO. In vivo administration of entacapone improved body weight and glucose tolerance and increased adipose thermogenesis in mice, which the authors tied to decreased FTO-catalyzed m6A demethylation of FOXO1 mRNA. Further studies will need to confirm the repurposing potential of entacapone for obesity or metabolic disease in humans.


Recent studies have established the involvement of the fat mass and obesity-associated gene (FTO) in metabolic disorders such as obesity and diabetes. However, the precise molecular mechanism by which FTO regulates metabolism remains unknown. Here, we used a structure-based virtual screening of U.S. Food and Drug Administration–approved drugs to identify entacapone as a potential FTO inhibitor. Using structural and biochemical studies, we showed that entacapone directly bound to FTO and inhibited FTO activity in vitro. Furthermore, entacapone administration reduced body weight and lowered fasting blood glucose concentrations in diet-induced obese mice. We identified the transcription factor forkhead box protein O1 (FOXO1) mRNA as a direct substrate of FTO, and demonstrated that entacapone elicited its effects on gluconeogenesis in the liver and thermogenesis in adipose tissues in mice by acting on an FTO-FOXO1 regulatory axis.


Obesity and diabetes, characterized by impaired energy homeostasis and disordered body metabolism, have become severe epidemics in modern society. Genome-wide association studies have identified the fat mass and obesity-associated gene (FTO) as a genetic factor related to obesity (15). A recent study suggested that “long-range functional connections” between FTO intronic single-nucleotide polymorphisms and the iroquois homeobox protein 3 (IRX3) may promote obesity (6, 7). There is also a substantial amount of evidence that FTO itself may regulate metabolic disorders (812). Studies in mouse models have shown that Fto overexpression leads to obesity, and its knockout or loss-of-function mutation results in reduced body weight (810). FTO expression in the brain has been implicated in the regulation of food intake (13). In adipocytes, FTO deficiency was reported to induce the expression of uncoupling protein 1 (UCP1), a mitochondrial inner membrane proton channel linked to thermogenesis (12). In the liver, correlations have been reported between Fto mRNA expression and glucose metabolism (14, 15).

The FTO protein belongs to the Fe2+ and α-ketoglutarate (α-KG)–dependent oxygenase family (16, 17). It demethylates N6-adenosine–modified (m6A) sites and N6,2′-O-dimethyladenosine–modified (m6Am) sites of mRNA (18, 19), thereby influencing multiple mRNA-related processes including transcript stability, alternative splicing, mRNA translocation, and protein translation (2025). Although efforts have been made to determine the physiological and pathological relevance of FTO, how FTO is involved in metabolic regulation remains unknown, mainly due to lack of information about its definitive mRNA substrates. Furthermore, although FTO is widely viewed as an attractive biological target, it is not yet clear whether a small-molecule inhibitor specifically targeting FTO could be developed for the treatment of metabolic disorders such as obesity and diabetes. Unfortunately, only a few FTO biochemical inhibitors have been reported to date (2631), and none have been critically evaluated for in vivo efficacy for the treatment of metabolic disorders. Thus, it is desirable to identify drug-like FTO inhibitors that may be useful in deciphering both the physiological function and potential clinical value of the FTO protein.

To uncover potential FTO inhibitors, we used a structure-based hierarchical virtual screening approach and comprehensively validated candidates using structural and biochemical experiments, in vivo experiments, and extensive transcriptome sequencing analyses. We identified entacapone, an inhibitor of catechol-O-methyltransferase (COMT) (32) used to treat Parkinson’s disease, as a potent FTO chemical inhibitor. Mechanistically, FTO demethylated m6A sites on forkhead box protein O1 (FOXO1) mRNA to up-regulate FOXO1 expression, thereby modulating gluconeogenesis and thermogenesis. Our studies support the potential repurposing of the U.S. Food and Drug Administration (FDA)–approved drug entacapone for the treatment of metabolic disorders such as obesity and diabetes.


Virtual screening and hit validation

To identify potential FTO inhibitors from FDA-approved drugs, we used a structure-based hierarchical virtual screening approach that we developed in previous studies (Fig. 1A) (3337). We found that entacapone inhibited FTO demethylation activity at a median inhibitory concentration (IC50) of 3.5 μM (Fig. 1B). Entacapone competed with both m6A-containing oligonucleotide substrates and the FTO cofactor α-KG (Fig. 1B and fig. S1A) and bound to FTO with a disassociation constant of 234 nM in the presence of Fe2+ or 1072 nM without Fe2+ (fig. S1B). Consistent with the effect of FTO knockdown, entacapone treatment also enhanced the amount of m6A on mRNA in a variety of human cell lines, including Hep-G2 (Fig. 1C and fig. S1C). Entacapone did not show any inhibitory effect on the enzymatic activity of the RNA m6A demethylase AlkB homolog 5 (ALKBH5) or the ten-eleven translocation methylcytosine dioxygenase 1 (TET1), nor did it alter the DNA methylation or histone methylation patterns in entacapone-treated Hep-G2 cells (fig. S1, D to G).

Fig. 1 Structure-based hierarchical virtual screening and hit validation.

(A) Flowchart of structure-based virtual screening. (B) Enzymatic inhibition of FTO by entacapone (ENT), measured by changes in relative m6A amount (three independent replicates). (C) Relative m6A amount on mRNA in wild-type and FTO knockdown Hep-G2 cells with or without entacapone treatment (50 μM, 48 hours; n = 5 for each). scr, scramble; DMSO, dimethyl sulfoxide; shFTO, FTO knockdown by short hairpin RNA. (D) Left: The global structure of the FTO-entacapone complex (pink ribbons). Top right: The electrodensity map (blue) of entacapone (sticks; carbon atoms colored in green) in the FTO active site. Bottom right: The FTO-entacapone interactions, where carbon atoms of binding site residues are colored in cyan and the hydrogen bonds are illustrated with yellow dashed lines. (E) Structure-activity relationship of entacapone derivatives (three independent replicates). R groups are different chemical substitutions. Data are presented as means ± SEM. cpd, compound. *P < 0.05.

The structural basis of FTO inhibition

Entacapone is structurally distinct from any known inhibitors of FTO (fig. S2, A and B). To determine the structural basis of entacapone in FTO inhibition, we determined the crystal structure of the FTO protein bound with entacapone in the presence of a Zn2+ ion [Protein Data Bank (PDB) code, 6AK4; Fig. 1D and table S1]. The docking pose for entacapone agreed well with the crystallographically determined binding pose (fig. S2B). Consistent with the results of our enzyme kinetic observations, entacapone occupied both the cofactor and the substrate binding sites of FTO (Fig. 1D). The metaposition hydroxyl group on the nitrocatechol ring formed hydrogen bound to Arg322 and Tyr106 at the substrate binding site. The nitrile group formed a chelation interaction with the Zn2+ ion, which was also evident in a recently published protein-ligand complex structure of histone demethylase (38). The carbonyl group formed a hydrogen bond with Asn205, whereas the flexible diethyl-propanamide tail was deeply buried inside the cofactor binding site.

We performed a proof-of-concept structure-activity relationship exploration, the results of which accorded with the crystallographically determined binding mode of entacapone (Fig. 1E). Specifically, removing either of the hydroxyl groups (cpd-1 or cpd-2) diminished entacapone’s enzymatic inhibitory activity, and substitution of the methoxyl group for the ortho-hydroxyl group (cpd-3) reduced inhibitory activity (IC50 value of 10.4 μM). Replacing the nitro group with a benzoyl-amide group (cpd-4) did not reduce the inhibitory activity (IC50 value of 5.7 μM), consistent with the solvent-exposed nitro group lacking direct interactions with the protein in the crystal structure. Inside the cofactor binding site, the addition of hydrophobic methyl groups on the diethyl moiety (cpd-5) reduced inhibitory activity (IC50 value of 15.3 μM), and rigidifying the flexible diethyl tail to alicyclic groups (cpd-6 and cpd-7) improved inhibitory activity (IC50 values of 1.2 and 0.7 μM, respectively). We further obtained the crystal structure of cpd-7 bound with FTO (PDB code, 6AEJ), which revealed highly similar binding poses for both cpd-7 and entacapone (fig. S2C and table S1).

The metabolic effects of entacapone

The assumed association between FTO and metabolic disorders, as well as the extensive safe history of entacapone in long-term clinical use, prompted us to evaluate the possibility of repurposing entacapone for the treatment of metabolic disorders such as obesity and diabetes. We examined the dose-response effect of entacapone in reducing body weight in a high-fat diet-induced obese (DIO) mouse model for 5 weeks (fig. S3). An effective dosing regimen (600 mg/kg of body weight blended in with food) was used throughout the rest of the animal studies. After 3 weeks of entacapone treatment, mouse body weight decreased by 10.1% compared to controls (Fig. 2A), although both groups showed similar food intake (Fig. 2B). Both fat mass and fat mass ratio were also reduced after entacapone treatment (Fig. 2C). We conducted respiration calorimetry analysis to measure the effects of entacapone on energy metabolism. Entacapone treatment increased the energy expenditure of mice (Fig. 2D). Analysis of plasma showed that entacapone caused reductions in total cholesterol (17.6%), low-density lipoprotein cholesterol (31.0%), and triglycerides (10.2%; Fig. 2E). Further, we observed that the temperature of the skin surrounding inguinal white adipose tissue (iWAT) and interscapular brown adipose tissue (BAT) was higher in entacapone-treated mice versus controls after only 3 weeks of entacapone treatment, which suggested that thermogenesis in adipose tissue was a possible cause of the increased energy expenditure observed in the entacapone-treated mice (Fig. 2, F and G).

Fig. 2 Metabolic effects of entacapone treatment or Fto knockout.

(A to C) The effects of entacapone on the (A) body weight, (B) food intake amounts, and (C) fat mass and fat mass ratio of DIO mice (n = 10 per group) after entacapone treatment (600 mg/kg per day) for 9 weeks. (D) The energy expenditure of DIO mice (n = 8 per group) after entacapone treatment for 9 weeks. (E) Total cholesterol (CHO), low-density lipoprotein cholesterol (LDL-c), and triglycerides (TG) in the serum of DIO mice (n = 8 for each) after entacapone treatment for 9 weeks. (F and G) The body temperature of the skin surrounding iWAT (F) and interscapular BAT (G) of DIO mice (n = 10 for each) after entacapone treatment for 3 weeks. (H) The fasting blood glucose concentrations of DIO mice (n = 10 for each) after entacapone treatment for 3 weeks. (I and J) The effects of entacapone on (I) the fasting blood glucose and (J) glucose tolerance test (GTT) of db/db mice (n = 11 for each) after entacapone treatment for 3 weeks. (K and L) The (K) fasting blood glucose and (L) glucose tolerance test of the hepatic Fto knockout (KO) mice (n = 10 for each). Data are presented as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.005.

We also found that entacapone treatment decreased fasting blood glucose in DIO mice (Fig. 2H). We subsequently used a diabetic db/db mouse model to evaluate these effects. As expected, entacapone-treated db/db mice had decreased fasting blood glucose (Fig. 2I) and improved glucose tolerance (Fig. 2J) compared to the control mice. These observations suggested that FTO inhibition may interfere with gluconeogenesis. To test this hypothesis, we generated liver-specific Fto knockout mice (fig. S4, A to D), which, compared to controls, showed no difference in body weight, hepatic glycogen concentration, or the concentration of two liver transaminases in serum commonly used to assess liver injury (fig. S4, E to G). However, the fasting blood glucose in liver-specific Fto knockout mice was lower than that of controls. We also observed the improved glucose tolerance (Fig. 2, K and L, and fig. S4H). In particular, liver-specific Fto knockdown mice did not show altered serum insulin concentrations, although they had reduced fasting blood glucose and an improved pyruvate tolerance (fig. S5, A to C). Thus, we provided evidence that hepatic FTO regulates gluconeogenesis in the liver.

FOXO1 mRNA as a direct substrate of FTO

To explore the molecular mechanism responsible for hepatic FTO-regulated gluconeogenesis, we analyzed transcriptome profiles in FTO knockdown Hep-G2 hepatic cells by RNA sequencing (RNA-seq). Genes differentially expressed between control and FTO knockdown samples were identified with fold-change cutoff value of 1.5 and the false discovery rate cutoff value of 0.001. Among the genes down-regulated after FTO knockdown, pathways associated with glucose metabolism, including the glycolysis/gluconeogenesis pathways, were enriched (Fig. 3A and fig. S6A). We confirmed that two critical gluconeogenic genes, glucose 6-phosphatase (G6PC) and phosphoenol-pyruvate carboxykinase 1 (PCK1), were down-regulated after FTO knockdown (Fig. 3B and fig. S6B). Consistent with the above findings, G6PC was also down-regulated in entacapone-treated Hep-G2 cells (fig. S6C). Expression of these two gluconeogenic genes has been reported to be regulated by a common transcription factor, FOXO1 (39).

Fig. 3 Identification of FOXO1 mRNA as a direct substrate of FTO.

(A) The top five ranked pathways based on Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of down-regulated genes after FTO knockdown in the Hep-G2 cell line. PPAR, peroxisome proliferator–activated receptor. (B) Quantitative polymerase chain reaction analysis of the expression of FTO, G6PC, and PCK1 after FTO knockdown (n = 3 for each), relative to actin. (C) The enrichment of four mammalian-conserved m6A sites in FOXO1 mRNA using RNA immunoprecipitation in cells (n = 3 for each) relative to input and the demethylation activity of FTO on the synthetic mRNA fragments containing enriched sites (three independent replicates). IgG, immunoglobulin G. (D) FOXO1 expression in the FTO knockout and FTO rescue cells. GAPDH, glyceraldehyde phosphate dehydrogenase. (E) Hepatic Foxo1 expression in adenovirally hepatic Fto knockdown db/db mice (Ad-shFto) and scramble (Ad-scr) as control (n = 5 for each). Data are presented as means ± SEM. *P < 0.05 and ***P < 0.005.

Among the m6A sites identified in the human and mouse transcriptomes, four common sites were found in FOXO1 mRNA (20). Because FTO is an m6A demethylase, we measured the m6A amounts of these four putative sites in FTO pulldown mRNA from Hep-G2 cells and found that two sites (m6A-1 and m6A-2) in FOXO1 mRNA were strongly enriched compared to control (Fig. 3C). Consistently, synthetic mRNA fragments containing these two m6A sites with flanking sequences were efficiently demethylated by FTO (Fig. 3C). An m6A pulldown assay showed that m6A amounts at these two m6A sites in FOXO1 mRNA were higher in FTO knockout cells than that in FTO wild-type cells (fig. S6D). Reconstitution of FTO expression in FTO knockout cells led to the recovery of FOXO1 expression, validating the finding that FOXO1 is a downstream target of FTO (Fig. 3D). In line with the finding that m6A modification participates in the regulation of gene translation (22, 24), we observed that FTO deficiency caused by either FTO knockdown or treatment with entacapone led to reduced expression of FOXO1 (fig. S6E). However, FOXO1 protein expression was not affected by knocking down another m6A demethylase family member, ALKBH5 (fig. S6F). These observations support the view that FTO specifically regulates FOXO1 expression by demethylating m6A sites on FOXO1 mRNA in hepatic cells.

An FTO-FOXO1 axis regulating gluconeogenesis

Considering the central role of FOXO1 in regulating hepatic gluconeogenesis in the fasting state (3943), we speculated that FOXO1 is involved in hepatic FTO-regulated gluconeogenesis in vivo. We evaluated the regulatory role of the FTO-FOXO1 axis in diabetic db/db mice. We found that db/db mice with Fto expression knocked down specifically in the liver manifested decreased fasting blood glucose concentrations and decreased expression of both Foxo1 and G6pc compared to control mice (Fig. 3E and fig. S7, A and B). The cyclic adenosine 3′,5′-monophosphate response element–binding protein (Creb) has been reported to transcriptionally regulate Foxo1 (39). However, we observed no changes in Creb expression after Fto knockdown (fig. S7C), indicating that it is not involved in the regulation of the FTO-FOXO1 signaling pathway. Nor did we find any difference in the expression of Irx3 (fig. S7D), suggesting that Irx3 does not play a critical role in Fto-associated gluconeogenesis.

We used a live imaging system with an adenovirus-encoded G6pc luciferase reporter to study the FTO-FOXO1 axis in regulating gluconeogenesis in the liver (40, 44). We observed that Fto deficiency in mice caused by specific Fto knockout in the liver, expression of a catalytically dead FTO mutant (R96A), or inhibition of Fto activity by entacapone led to a reduction in the G6pc luciferase signal. We observed a similar phenomenon in mice with a liver-specific Foxo1 knockout (Fig. 4A and fig. S7E). We next determined whether entacapone regulated liver gluconeogenesis via the FTO-FOXO1 axis in an m6A-dependent manner. In entacapone-treated hepatic Foxo1 knockout mice, only mice adenovirally reexpressing wild-type Foxo1 (liver-Foxo1-KO + Ad-WT-Foxo1), but not synonymously m6A-mutated Foxo1 (liver-Foxo1-KO + Ad-mut-Foxo1), showed a reduction in both fasting blood glucose and liver-specific G6pc luciferase signal (Fig. 4B). These findings confirmed that the FTO-FOXO1 axis regulates hepatic gluconeogenesis and that entacapone regulates fasting blood glucose through its direct effect on the hepatic FTO-FOXO1 signaling axis.

Fig. 4 The FTO-FOXO1-G6PC signaling cascade and gluconeogenesis in liver.

(A) The effects on G6pc–luciferase (Luc) signal of hepatic Fto knockout (liver-Fto-KO and control; n = 6 for each), FTO loss-of-function mutation (R96A) (Ad-Fto-Mut and Ad-Fto-WT; n = 10 for each), Foxo1 knockout (liver-Foxo1-KO; n = 6 for each), or pharmacological attenuation of Fto activity via administration of entacapone (600 mg/kg per day) for 11 days (vehicle; n = 6 for each). (B) The effects of entacapone treatment on the G6pc:Luciferase signal and the fasting blood glucose in hepatic Foxo1 knockout mice expressing wild-type Foxo1 (n = 6 for each) and in hepatic Foxo1 knockout mice expressing synonymous mutant Foxo1 (n = 5 for each). Data are presented as means ± SEM. *P < 0.05 and ***P < 0.005.

An FTO-FOXO1 axis modulating thermogenesis

We wished to determine the molecular mechanisms by which FTO-FOXO1 regulated thermogenesis in adipose tissues of entacapone-treated DIO mice. Histological analysis revealed that entacapone treatment decreased cell size in murine iWAT (Fig. 5A). RNA-seq analysis of iWAT from entacapone-treated and control mice showed that entacapone treatment resulted in a >10-fold increase in the expression of Ucp1 (Fig. 5B), a well-known biomarker of the thermogenesic state of adipose tissues (4548). The oxidative phosphorylation pathway was strongly enriched among the genes up-regulated by entacapone (Fig. 5C), including many components of complexes I to V of the electron transport chain (fig. S8A). These observations suggested that the antiobesity effect of Fto inhibition by entacapone in DIO mice may have resulted from increased thermogenesis in adipose tissue.

Fig. 5 The FTO-FOXO1-UCP1 regulation of thermogenesis in adipose tissue.

(A) Adipocyte cell sizes of iWAT in DIO mice (hematoxylin and eosin staining, 200×). (B) Scatter plot of RNA-seq analysis of iWAT (entacapone versus vehicle). RPKM, reads per kilo base of exon model per million mapped reads. (C) Representative KEGG pathways enriched in the up-regulated genes were identified using RNA-seq analysis. HIF-1, hypoxia-inducible factor–1. (D) Relative m6A amounts on mRNA of iWAT in DIO mice (n = 3 for each). (E) Distribution of m6A peaks across the length of mRNA transcripts. 5′ untranslated region (5′UTRs), CDSs, and 3′UTRs of mRNAs were individually binned into regions spanning 1% of their total length, and the percentage of m6A peaks that fall within each bin was determined. The moving averages of peak percentage in iWAT of vehicle-treated and entacapone-treated mice are shown. (F) The Integrative Genomics Viewer tracks displaying RNA-seq and m6A sequencing (m6A-seq) read distribution for the Foxo1 mRNA of iWAT. IP, immunoprecipitation. (G) Enrichment of m6A-1 and m6A-2 sites on Foxo1 mRNA in the iWAT of DIO mice (n = 3 for each). (H) Western blot analysis of Foxo1, Ucp1, and Fto expression in iWAT of DIO mice. (I) Ucp1:Luciferase signals of entacapone treatment and Fto and Mettl3 knockdown by small interfering RNA (siFto and siMettl3, respectively) in 293T cells (n = 2 for each). Data are presented as means ± SEM. ***P < 0.005.

We found that the overall extent of m6A methylation in the iWAT of entacapone-treated mice was 21.0% higher than in control mice (Fig. 5D). We established the m6A methylomes of iWAT from control and entacapone-treated mice using an m6A antibody pulldown assay combined with m6A seq. In both control and entacapone-treated samples, m6A peaks were enriched in an RGACH motif (fig. S8B), which mostly occurred in high abundance in mRNA coding sequences, especially near stop codons, and 3′UTRs (Fig. 5E and fig. S8, C and D). However, entacapone treatment led to both enhanced and newly appearing m6A peaks that were enriched in the Foxo pathway (fig. S8, E and F). Specifically, m6A-1, the first m6A site located in the coding sequence of FOXO1 mRNA in Hep-G2 cells, was uniquely present in the iWAT of entacapone-treated DIO mice. We also observed an increased m6A peak at m6A-2 site of Foxo1 mRNA, the second m6A site identified in the FOXO1 coding sequence (CDS) of Hep-G2 cells (Fig. 5F). Both m6A peaks were enriched upon Fto inhibition by entacapone, which we confirmed by m6A antibody pulldown combined with quantitative polymerase chain reaction assays (Fig. 5G). m6A peaks were not detected on Ucp1 mRNA in iWAT of either entacapone-treated or control mouse samples, indicating that Ucp1 mRNA is not a direct substrate of Fto (fig. S8G).

Because FOXO1 has been shown to repress the UCP1 gene transcription (49, 50), we analyzed the effect of entacapone on Fto, Foxo1, and Ucp1 expression in iWAT of DIO mice. In accordance with suppressed Fto activity after entacapone treatment, we observed decreased Foxo1 expression and increased Ucp1 expression in iWAT (Fig. 5H); similar effects were also evident in interscapular BAT, with decreased Foxo1 and increased Ucp1 expression observed after entacapone treatment (fig. S8, H and I). We then used a cell-based Ucp1-luciferase reporter assay to examine the role of FTO-FOXO1 signaling in controlling Ucp1 gene transcription. Fto inhibition by either entacapone treatment or knockdown enhanced Ucp1 promoter activity. Conversely, knockdown of Mettl3, a methyltransferase with the opposite role of Fto, markedly suppressed Ucp1 promoter activity (Fig. 5I). Thus, the FTO-FOXO1 axis is implicated in the regulation of thermogenesis in the adipose tissues through control of transcription of FOXO1 downstream targets such as UCP1.

We generated Foxo1 mutant mice by synonymously mutating the m6A-2 site on Foxo1 mRNA (Foxo1-Mut) to further confirm the role of the FTO-FOXO1 axis in vivo (fig. S9, A to C). We first performed RNA-seq on the liver samples taken from fasted hepatic Fto knockout and Foxo1-Mut mice. Many of the commonly down-regulated genes in these mice mutants are known to be involved in metabolism-related pathways including gluconeogenesis (fig. S9, D and E). Entacapone treatment for 3 weeks led to a reduction in both body weight and fasting blood glucose concentration in wild-type mice but not in Foxo1-Mut mice (fig. S9, F and G). In addition, entacapone treatment did not influence the food intake of either Foxo1 wild-type or Foxo1-Mut mice (fig. S9H). These results demonstrated that entacapone regulates body weight and blood glucose through FTO-FOXO1 signaling in vivo (fig. S10).


Despite long-standing research efforts, the discovery of therapeutic interventions that can regulate metabolic homeostasis proceeds slowly. We selected the m6A demethylase FTO as a potential target and performed a computer-aided virtual screening to find small molecular inhibitors of FTO. We identified entacapone, an FDA-approved drug, as a selective inhibitor of FTO activity involved in the regulation of metabolic homeostasis. The pharmaceutical drug entacapone could serve as a useful “tool compound” to study the in vivo functions of FTO. Furthermore, the binding characteristics of our solved FTO-entacapone crystal complex structure and our proof-of-concept structure-activity relationship study establish a structural foundation for designing a new generation of drug-like FTO inhibitors. Our strategy moving from virtual screening to hit validation and from animal physiology to mechanistic elucidation represents an illustrative model potentially valuable for the next generation of translational medicine.

Entacapone was originally approved by the FDA for use as an adjunctive therapy in combination with levodopa and carbidopa for the treatment of Parkinson’s disease. Mechanistically, entacapone is believed to peripherally inhibit COMT activity, increasing the bioavailability of levodopa and extending its duration of action in the brain (32). We cannot completely exclude the effects of entacapone on COMT from its demonstrated effects on FTO. However, it was reported that there was no significant difference in mean body weight between COMT knockout and wild-type mice in different age groups (51, 52). In addition, COMT deficiencies in DIO mice (by use of a COMT inhibitor or small interfering RNA), or from a low-activity COMT allele in humans, have been implicated in exacerbated characteristics of type 2 diabetes and obesity (53, 54). Therefore, we reasoned that the metabolic effects of entacapone strongly implicate the engagement of FTO.

Our identification of FOXO1 mRNA as a substrate of FTO may aid the discovery of biomarkers for FTO-associated diseases. FOXO1 is a transcription factor, and FOXO1 has been traditionally considered to be a difficult target for manipulation through pharmacological approaches (43). Selectively silencing FOXO1 through the inhibition of FTO could potentially be used to treat dysregulated metabolic homeostasis.

Considering that FOXO1 mRNA is not the only direct substrate of FTO, it is likely that other FTO target genes may have functional roles related to obesity or diabetes. It was reported that an m6Am site at the 5′ cap of mRNA molecules is a substrate of FTO and FTO-mediated demethylation of this site reduces mRNA stability (19). However, our RNA-seq results only illustrated the presence of specific internal m6A sites within FOXO1 mRNA. FOXO1 mRNA stability is not changed by FTO expression knockdown or activity inhibition. In particular, synonymously mutating this specific m6A site effectively blocked the effects of FTO-FOXO1 axis. Therefore, the internal m6A site identified in our study is critical in mediating the regulatory role of FTO-FOXO1 axis both in thermogenesis in adipose tissues and in gluconeogenesis in the liver.

Our study has several limitations. We are aware that data obtained from rodent models cannot be extrapolated directly to humans, especially given the differences in adipose tissue metabolism between mice and humans (55, 56). Further studies will be needed to establish relevant efficacies in human cells and tissues. Another limitation was the potentially problematic high dose of entacapone required in our animal studies, on account of its low potency and its short plasma elimination half-life (about 0.5 hours); we thus conclude that it is desirable to develop entacapone analogs with improved pharmacokinetic and pharmacodynamic properties for the treatment of metabolic disorders.


Study design

The aims of this study were to identify potential FTO inhibitors from FDA-approved drugs by applying a structure-based hierarchical virtual screening strategy and to study the molecular mechanism of the regulatory role of FTO in metabolism. The enzymatic inhibitory activity of the selected compounds was measured using an FTO-catalyzed demethylation assay. The binding affinity of entacapone against FTO was measured, and the complex structures of FTO bound with inhibitors were determined to elucidate the protein-ligand interactions. We synthesized analogs of entacapone to explore a proof-of-concept structure-activity relationship. To evaluate the possibility of repurposing entacapone for the treatment of FTO-related metabolic disorders such as obesity and diabetes, we measured the therapeutic efficacy of entacapone in DIO and diabetic mouse models. Transcriptome sequencing analysis of FTO knockout hepatic cells and entacapone-treated adipocytes identified substrates of FTO. We confirmed the FTO-FOXO1 regulatory axis using live-imaging analysis. All procedures in this protocol were conducted with the approval of the Institutional Animal Care and Use Committee of the National Institute of Biological Sciences, Beijing in accordance with the governmental regulations of China. Animals were randomized into groups with similar body weights or similar blood glucose concentrations before study initiation. Investigators and data analyzers were blinded for the mouse studies. Outliers were excluded from a group if they were located ±2 SDs away from the group mean. Sample sizes were determined by internal pilot studies and are listed in the figure legends. All experiments were replicated more than once, as indicated in figure legends.

Animal experiments

Wild-type mice were purchased from the Beijing Vital River Laboratory Animal Technology Co. Ltd. The db/db mice were purchased from the Nanjing Biomedical Research Institute of Nanjing University; these were introduced from the Jackson Laboratory. The albumin-cre mice and the Foxo1loxp/loxp mice were obtained from the Jackson Laboratory. Animals were maintained in a 12-hour light/dark photoperiod with food and water provided ad libitum.

Structure-based hierarchical virtual screening

We applied a structure-based hierarchical virtual screening strategy to identify potential FTO inhibitors from a library of FDA-approved drugs. Briefly, an automatic docking procedure was applied to generate ligand binding poses, as described previously (36). The binding site was defined as both the substrate site and the cofactor N-oxalylglycine site, as in the crystal structure of FTO (PDB code, 3LFM) (17). The Fe2+ ion was processed according to a previously published protocol (57), with reduced van der Waals interaction and partial atomic charge parameters. We docked 1323 FDA-approved drugs processed by ZINC (58) against FTO using DOCK 3.5.54 (5961) and outputted the top 500 docking poses for each compound. We filtered these docking poses based on three predefined structural descriptors according to the FTO crystal complex structure bound with N3-methylthymidine. First, the docked compounds should form close atomic contacts (i.e., distance cutoff of 4.5 Å) with residues in both the substrate binding site (including Arg322, Arg96, and Tyr106) and the cofactor binding pocket (including Tyr295, Arg316, Ser318, and Asn205). Second, the qualified ligands should form at least one hydrogen bond interaction (distance threshold of 4.5 Å) with the oxygen atom of the hydroxyl group of Tyr106 and the nitrogen atom of the amino group of Glu234. Third, the desired candidates should form chelation interactions with the Fe2+ (distance cutoff of 3 Å with acceptor heteroatoms). After filtering, 332 compounds remained. The docking poses were then minimized and rescored with a more sophisticated scoring method Molecular Mechanics/Generalized Born Surface Area using the PLOP program (34). The refined poses were filtered and visually checked, and 19 compounds were selected for further experimental validation (table S2).

Phenotypic measurement

Because entacapone is rapidly metabolized in vivo (32), we optimized the route of administration and dosage of entacapone. For the treatment group, entacapone (600 mg/kg of body weight) was orally administered to the mice by blending it with the food. The measured plasma concentration of entacapone in mice (averaged for an entire 24-hour period) was 6.4 ± 0.8 μM.

Counter-obesity experiments were performed in a DIO mice model. Eight-week-old male C57BL/6 mice were fed a high-fat diet (45% fat; OpenSource Diets) for 12 weeks. Obese mice with body weight 20% higher than that of mice fed a normal diet were selected for experiments. Entacapone was administered to 10 randomly selected obese mice, with vehicle-treated mice as controls. Body mass and food intake amount were measured each week. The body composition of live mice was determined using the magnetic resonance imaging analyzer (EchoMRI-700). Oxygen consumption, carbon dioxide production, and respiratory exchange ratio were measured by Physiocage (Oxylet, Panlab). After 24 hours of acclimatization, calorimetry parameters were measured for 24 hours. Body temperature of the skin surrounding iWAT and interscapular BAT was measured using infRec analyzer (NS9500 Lite, Avio/NEC) after entacapone treatment for 3 weeks. Sections of iWAT and interscapular BAT were counterstained with hematoxylin.

Blood samples were extracted from the tail and examined with an Aviva Nano blood glucose meter system (Accu-Chek). For the glucose tolerance tests and the pyruvate tolerance tests, mice were injected intraperitoneally with glucose (2 g/kg of body weight; Sigma-Aldrich) or pyruvate (2 g/kg of body weight; Sigma-Aldrich) after fasting for 6 hours. Blood glucose was measured 0, 15, 30, 60, 90, and 120 min after injection. Hepatic glycogen was measured with a Glycogen Assay Kit (Abcam). Cholesterol, low-density lipoprotein cholesterol, triglycerides, serum glutamic oxaloacetic transaminase (SGOT), and serum glutamic pyruvic transaminase (SGPT) in serum were measured with a Hitachi Biochemical Analyzer. Serum insulin was measured with an enzyme-linked immunosorbent assay kit (Millipore).

Statistical analysis

Data in figures are shown as means ± SEM. Normality tests were conducted before unpaired two-tailed Student’s t tests in all animal studies. P < 0.05 was considered statistically significant. *P < 0.05, **P < 0.01, and ***P < 0.005. All data values are reported in data file S1.


Materials and Methods

Fig. S1. Experimental validation of the FTO-entacapone interaction and entacapone target selectivity assessment.

Fig. S2. Crystal complex structures of FTO bound with different types of inhibitors.

Fig. S3. Dose-response effect of entacapone on the body weight of DIO mice.

Fig. S4. Generation and phenotypic measurements of liver-specific Fto knockout mice.

Fig. S5. Hepatic Fto regulates gluconeogenesis in the liver.

Fig. S6. FTO-FOXO1 regulatory axis modulates G6PC and PCK1 expression in a hepatic cell line.

Fig. S7. Hepatic Fto knockdown in db/db mice and enzymatic activity of FTO R96A mutant protein.

Fig. S8. Characterization of m6A modifications and changes in gene expression in mouse adipose tissues.

Fig. S9. Generation, RNA-seq analysis, and phenotypic analysis of Foxo1-Mut mice with the Foxo1 synonymous mutation at the Foxo1–m6A-2 site.

Fig. S10. Proposed FTO-FOXO1 regulatory axis in adipose tissue and liver.

Table S1. Statistics on data collection and structure refinement of FTO-inhibitor complex structures.

Table S2. Structures and FTO enzymatic inhibition activities of 19 selected compounds from virtual screening.

Data file S1. Summary of all reported data values.

References (6278)


Acknowledgments: We thank the staff at the Shanghai Synchrotron Radiation Facility (SSRF) for assistance in data collection. We thank M. Montminy (The Salk Institute for Biological Studies) for providing the G6PC:Luc plasmid, B. Zhu [Institute of Biophysics, Chinese Academy of Sciences (CAS)] for TET1 protein, and Z. Yuan for pGL4.23-Ucp1-luciferase and pGL4.23 control constructs and pRenilla-TK plasmid (Institute of Biophysics, CAS). Funding: This work was supported by Strategic Priority Research Program A of the CAS (XDA16000000 to Y.-G.Y.), the National Natural Science Foundation of China (NSFC) (31625016 to Y.-G.Y.), the National Basic Research Program of China (2014CB849802 to N. Huang and 2016YFC0900300 to Y.-G.Y.), the National Major Scientific and Technological Special Project for “Significant New Drugs Development” (2013ZX09509102 to N. Huang), the CAS Strategic Priority Research Program (XDB14030300/QYZDY-SSW-SMC027 to Y.-G.Y.), Youth Innovation Promotion Association (CAS 2016097 to B.S.), and the Chinese “Recruitment Program of Global Youth Experts” (to E.E.Z.). Computational support was provided by the Special Program for Applied Research on Super Computation of the NSFC–Guangdong Joint Fund (the second phase) under grant no. U1501501 (to N. Huang). Author contributions: N. Huang, E.E.Z., and Y.-G.Y. designed experiments. S.P. and N. Huang performed the computational screening and designed the structure-activity relationship studies. Yanli Wang and W.Z. purified proteins. S.P. performed the enzymatic activity assays and SPR binding assays. Yanli Wang, R.C., and W.Z. performed the crystallization and structural determination. D.J., W.X., and Q.L. performed cellular functional studies. H.Z, D.J., S.Z., P.L., N. Hou, Y.N., S.P., and G.Z. performed animal studies. N. Hou, G.Z., L.Z., and F.W. generated the FTO conditional knockout mice and FOXO1 mutant mice. B.S., W.X., C.G., and Q.L. performed RNA-seq and m6A-seq studies of adipose tissues and analyzed data. H.H. and T.C. performed RNA-seq studies and data analysis in Hep-G2 cells. Yankai Wang and S.H. synthesized DNA and RNA oligos. W.L., H.W., Y.M., and Z.Z. performed liquid chromatography–tandem mass spectrometry experiments. Z.M. contributed reagents and cell lines. S.P., D.J., W.X., B.S., Y.-L.Z., Y.-G.Y., E.E.Z., and N. Huang analyzed the data and wrote the manuscript. All authors discussed the results and commented on the manuscript. Competing interests: G.Z. and N. Huang have equity interest in RPXDs (Suzhou) Co. Ltd., which develops FTO inhibitors for the treatment of metabolic disorders. N. Huang, G.Z., S.P., and N. Hou hold an international patent (United States, 14147531; European, 13858101.2; Chinese, 201210497436.7 and 201380071479.1; Australia, 2013351676; Canada, 2892902) on the use of entacapone for prevention and treatment of obesity and related metabolic diseases. Data and materials availability: The coordinates and structure factors for the crystal complex structure of FTO-entacapone and FTO–cpd-7 were deposited in the PDB with PDB codes 6AK4 and 6AEJ, respectively. The m6A-seq and RNA-seq data have been deposited in the Gene Expression Omnibus (GEO) database under accession numbers GSE106998, GSE107411, GSE125785, and GSE125786 and have also been deposited in the GSA database ( under accession number PRJCA000645. All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article