Differential m⁶A, m⁶A_m, and m¹A Demethylation Mediated by FTO in Cell Nucleus and Cytoplasm

Summary

FTO, the first RNA demethylase discovered, mediates the demethylation of internal N⁶-methyladenosine (m⁶A) and N⁶, 2-O-dimethyladenosine (m⁶A_m) at the +1 position from the 5’ cap in mRNA. Here, we demonstrate that the cellular distribution of FTO is distinct among different cell lines, affecting the access of FTO to different RNA substrates. We find that FTO binds multiple RNA species, including mRNA, snRNA, and tRNA, and can demethylate internal m⁶A and cap m⁶A in mRNA, internal m⁶m A in U6 RNA, internal and cap m⁶A_m in snRNAs, and N¹-methyladenosine (m¹A) in tRNA. FTO-mediated demethylation shows a greater impact on the transcript levels of mRNAs possessing internal m⁶A than the ones with cap m⁶A_m in the tested cells. We also show that FTO can directly repress translation by catalyzing m¹A tRNA demethylation. Collectively, FTO-mediated RNA demethylation occurs to m⁶A and m⁶A_m in mRNA and snRNA as well as m¹A in tRNA.

In Brief

Wei et al. show that FTO mediates internal m⁶A and cap m⁶A_m demethylation of polyadenylated RNA with differential substrate preferences in nucleus versus cytoplasm, in which the internal m⁶A demethylation correlates with transcript level changes. FTO also affects snRNA m⁶A and m⁶A_m levels and mediates tRNA m¹A demethylation.

Introduction

FTO was suggested to be associated with human obesity by GWAS (Fawcett and Barroso, 2010). A genetic variant of FTO has been shown to be associated with increased food intake (Cecil et al., 2008) while loss-of-function mutations in FTO cause severe growth retardation (Boissel et al., 2009) and CNS defects (Ho et al., 2010), which were recapitulated in murine model studies (Church et al., 2010; Fischer et al., 2009; Hess et al., 2013).

Due to these intriguing phenotypes, extensive efforts have been devoted to identify the substrate(s) and to understand the biological function of FTO. FTO was first reported to catalyze the demethylation of 3-methylthymine (3meT) in single-stranded DNA (Gerken et al., 2007) and 3-methyluracil (3meU) in RNA (Jia et al., 2008) as an iron(II)- and αKG-dependent dioxygenase. Later, FTO was identified as the first RNA demethylase that catalyzes reversal of the N⁶-methyladenosine (m⁶A) methylation in mRNA in vitro and inside cells (Fu et al., 2013; He, 2010; Jia et al., 2011). m⁶A is the most abundant internal modification in mammalian mRNAs (Fu et al., 2014). Adjacent to the 5’ cap, the second base in many mRNAs can be 2’-O-methylated (Adams and Cory, 1975; Wei et al., 1975b), with a portion of these bases also bearing m⁶A methylation to form m⁶A_m (Wei et al., 1975a), deposited by a yet to be identified methyltransferase. This modification was confirmed by transcriptome-wide m⁶A-seq and exists in considerably lower overall abundance compared with m⁶A (Linder et al., 2015; Molinie et al., 2016). The m⁶A portion of m⁶A_m is known to be an in vitro substrate of FTO (Fu, 2012), with a recent study showing that m⁶A_m stabilizes mRNA by preventing DCP2-mediated decapping and microRNA-mediated mRNA degradation (Mauer et al., 2017). However, the functional relevance of m⁶A_m removal by FTO has yet to be fully explored.

Surprisingly, this study also suggested that internal mRNA m⁶As may not be relevant substrates of FTO (Mauer et al., 2017), despite reports of a range of biological processes affected by the demethylation of internal m⁶A in the last several years: i) FTO-mediated m⁶A demethylation is critical in DNA UV damage response, with the m⁶A methyltransferase complex METTL3/14 exhibiting the opposite function (Xiang et al., 2017); ii) FTO plays noticeable roles in Flaviviridae family viruses infection via demethylating viral RNA m⁶A in host cells (Gokhale et al., 2016); certain members such as hepatitis C virus do not possess cap nor cap m⁶A_m but still respond to FTO; iii) FTO-catalyzed demethylation significantly affects glioblastoma stem cell differentiation, again with METTL3/14 showing opposing functions (Cui et al., 2017); iv) FTO mediates nuclear demethylation of m⁶A at the 5’ UTR in heat shock response (Zhou et al., 2015); v) internal m⁶A demethylation by FTO plays an oncogenic role in a subset of acute myeloid leukemia (Li et al., 2017), and the effect of 2-HG inhibition can be explained through internal m⁶A methylation changes to oncogenic transcript targets of FTO, such as MYC (Su et al., 2018); vi) using liver-specific Fto-transgenic mice Fto was shown to mediate both m⁶A and cap m⁶A_m demethylation; however, internal m⁶A demethylation controls the alternative translation (Zhou et al., 2018); vii) demethylation of internal m⁶A on a specific mRNA by FTO facilitates the stability and translation of the target transcript in neuronal developments (Yu et al., 2018); and viii) a recent report showing that FTO binds extensively across mRNA and intron regions in premRNA (Bartosovic et al., 2017).

In our biochemical studies, we confirmed that FTO could effectively demethylate both m⁶A and cap m⁶A_m from purified polyadenylated RNA. We were intrigued by how FTO demethylates m⁶A and/or m⁶A_m on target polyadenylated RNAs inside cells and the different functional consequences of the demethylation. We found that FTO localization within the nucleus and cytoplasm varies between cell types and that FTO has distinct substrate repertoires inside the cell nucleus and cytoplasm. We further identified additional RNA substrates of FTO, including m¹A in tRNA, m⁶A in U6 RNA, internal and cap m⁶A_m in snRNAs. This study presents the most comprehensive picture of the FTO-mediated RNA demethylation to date. It reveals a previously unappreciated spatial regulation imparted by nuclear versus cytoplasmic demethylation mediated by FTO, which exerts distinct effects on target RNAs.

Results

FTO Mediates Demethylation of Both m⁶A and m⁶A_m in vitro and inside Cells

FTO has been shown to demethylate both m⁶A and m⁶A_m (Fu, 2012; Fu et al., 2013; Jia et al., 2011; Mauer et al., 2017) (Figure 1A). To further validate these previous reports with biologically relevant substrates, we isolated polyadenylated RNAs possessing both naturally occurring m⁶A and m⁶A_m from HEK293T cells and performed in vitro demethylation using purified FTO (Figure S1A). The substrate polyadenylated RNAs were digested to single nucleosides after FTO treatment. Ultra-High-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) was utilized to quantify levels of both m⁶A and m⁶A_m from digested RNAs using carefully calibrated standard curves generated using synthetic nucleoside standards (Figure S1B).

Figure 1. FTO Demethylates both m⁶A and m⁶A_m inside Cells.

See also Figure S1.

(A) A sketch that FTO mediates demethylation of m⁶A and cap m⁶A_m in polyadenylated RNA.

(B) Quantification of the m⁶A/A and cap m⁶A_m/A ratio in polyadenylated RNA by LC-MS/MS. In comparison to controls, significant increases in the m⁶A/A ratio were consistently observed among HeLa, HEK293T, and 3T3-L1 cells upon transient knockdown of FTO (blue bars). Increases of the m⁶A_m/A ratio were consistently observed among HeLa, HEK293T, and 3T3-L1 cells upon transient knockdown of FTO (green bars). P values were determined using Student’s unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 4 experiments in (B).

As shown in Figure S1C, among HeLa, HEK293T, and 3T3-L1 cells, the levels of m⁶A_m in polyadenylated RNAs are about 1/10 to 1/15 of that of m⁶A. FTO (2 μM in 20 ul) is capable of almost complete demethylation of all m⁶A_m and 81.4% of m⁶A in 200 ng polyadenylated RNAs within 1 hour under reported demethylation conditions (Figure S1D) (Jia et al., 2011; Mauer et al., 2017). The use of EDTA in the same reaction, which chelates the essential iron cofactor, serves as a control. We further lowered the FTO concentration to achieve incomplete demethylation of cap m⁶A_m in isolated polyadenylated RNAs and compared the global changes of m⁶A and m⁶A_m. As shown in Figure S1E, 0.2 μM of FTO (20 ul) could demethylate 81.9% of m⁶A and 20.3% m⁶ m A in the purified polyadenylated RNAs. Our results unequivocally show that FTO demethylates both internal m⁶A and cap m⁶A_m in purified polyadenylated RNAs in vitro. FTO exhibits higher activity towards m⁶A_m; however, it should be noted that because the total amount of m⁶A is around 10-fold to 15-fold higher than that of m⁶A_m in polyadenylated RNAs (Figure S1C), more m⁶As are demethylated compared to m⁶A_m in purified polyadenylated RNAs in terms of the number of modifications that can be reversed by FTO in vitro.

Next, we evaluated the effects of FTO perturbation on the cellular m⁶A and m⁶A_m levels of polyadenylated RNAs. We selected HeLa, HEK293T, and 3T3-L1 cell lines and measured the levels of m⁶A and m⁶A_m upon transient knockdown of FTO in comparison to the knockdown control by using UHPLC-MS/MS. A 19.2%, 11.8%, and 8.8% increase of the m⁶A/A ratio was observed in the polyadenylated RNAs in the knockdown samples compared to the knockdown control in HeLa, HEK293T, and 3T3-L1 cell lines, respectively (Figure 1B, blue bars). For m⁶A_m, a 13.7%, 20.6%, and 30.1% increase was consistently observed in the FTO knockdown samples compared to the knockdown control in HeLa, HEK293T, and 3T3-L1 cells, respectively (Figure 1B, green bars). The knockdown efficiency was validated by western blots (Figure S1F). These results revealed that the demethylation percentage of m⁶A inside cells by FTO is more profound in HeLa cells in comparison to HEK293T and 3T3-L1, while the FTO-mediated demethylation of m⁶A_m is more evident in 3T3-L1 cells compared to the other two cell lines. It again demonstrated both m⁶A and m⁶A_m demethylation by FTO inside different cell lines. Because the level of internal m⁶A is 10-fold higher than that of cap m⁶A_m in these cells, 5–10 times more internal m⁶A are subjected to FTO-mediated demethylation compared with m⁶A_m inside cells.

FTO Preferentially Targets Nuclear m⁶A and Both Cytoplasmic m⁶A and m⁶ m A in mRNA

We sought to investigate how FTO mediates differential m⁶A versus m⁶A_m demethylation in different cell lines and why FTO exhibits much higher m⁶A demethylation (over ~10-fold) versus m⁶A_m inside certain cells such as HeLa despite m⁶A_m being a preferred in vitro substrate. We reasoned that as almost all mRNA caps are bound by cap-binding proteins inside the cell nucleus (Muller-McNicoll and Neugebauer, 2013), the lack of m⁶A_m demethylation in certain cells might reflect substrate accessibility by FTO. FTO was first reported to be a nuclear protein, but it was later shown to exist in both cell nucleus and cytoplasm in different mammalian cell lines (Aas et al., 2017; Gulati et al., 2014). We asked whether the spatial regulation of FTO plays a crucial role in gaining access to its physiologically relevant substrate(s). We conducted confocal immunofluorescence (IF) imaging with an anti-FTO specific antibody to investigate the cellular localization of FTO in five mammalian cell lines: HeLa, Mel624, HepG2, HEK293T, and 3T3-L1 cell lines. As shown in Figure 2A, the relative proportion of FTO within the nucleus and cytoplasm varies across cell lines. In HeLa, Mel624, and HepG2 cells, FTO predominantly accumulates in the cell nucleus. However, FTO is located in both cell nucleus and cytoplasm in HEK293T cells. The presence of FTO in the cytoplasm is even more notable in 3T3-L1 cells than HEK293T cells. More immunofluorescent images for HeLa, HEK293T, and 3T3-L1 were collected and five representative pictures from each cell line were shown in the Figure S2A. Calculated by ImageJ, the colocalization of the endogenous FTO with the nucleus was 97.7%, 83.5%, and 61.8% in HeLa, HEK293T, and 3T3-L1 cells, respectively. Consistent with our hypothesis, the increasing proportions of cytoplasmic FTO from HeLa, HEK293T, to 3T3-L1 cells, correlate with the corresponding increased cellular demethylation percentage towards m⁶A_m and reduced m⁶A demethylation percentage in these cell lines (Figure 1B).

Figure 2. Subcellular Distribution of FTO Differs among Cell Lines and the Spatial Localization of FTO Influences its Demethylation of Target Substrates.

See also Figure S2.

(A) Subcellular localization of FTO (green) among HeLa, Mel624, HepG2, HEK293T, and 3T3-L1 cell lines. The nucleus was stained with DAPI (blue). Scale bar = 10 μm. Representative images are selected from three independent experiments.

(B) Quantification of the m⁶A/A and cap m⁶A_m/A ratios in polyadenylated RNA from HEK293T cells by LC-MS/MS. Transient knockdown of FTO led to a significant increase of nuclear polyadenylated RNA m⁶A/A ratio but not the cytoplasmic polyadenylated RNA m⁶A/A ratio (blue bars). Transient knockdown of FTO led to a significant increase of cytoplasmic polyadenylated RNA m⁶A_m/A ratio but no noticeable change of the nuclear polyadenylated RNA m⁶A_m/A ratio (green bars).

(C) Quantification of the m⁶A/A and cap m⁶A_m/A ratios in polyadenylated RNA from NB4 cells (left) and MONOMAC-6 cells (right) by LC-MS/MS. Stable knockdown of FTO led to both increased m⁶A/A and cap m⁶A_m/A ratios in total polyadenylated RNA; P values were determined using Student’s unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 6 experiments in (B); for n = 4 experiments in (C)

We next expressed an FTO mutant lacking the nuclear localization signal (ΔNLS-FTO) (Aas et al., 2017; Sanchez-Pulido and Andrade-Navarro, 2007) in HEK293T cell lines. The LCMS/MS results showed that overexpression of ΔNLS-FTO, which is directed to the cytoplasm, decreases m⁶A_m but not m⁶A in polyadenylated RNA HEK293T cells. This result contrasts with the effects caused by wild-type FTO overexpression mostly within nucleus (Figure S2B). Next, we separated the cell nuclear and cytoplasmic fractions in HEK293T cells and then measured FTO-mediated demethylation of m⁶A and m⁶A_m in polyadenylated RNAs in each fraction. As shown in Figure 2B (blue bars) knockdown of FTO led to a 37.9% increase in the m⁶A level in the cell nucleus but only a slight increase in the m⁶A level in cytoplasmic polyadenylated RNAs. In contrast, m⁶A_m in cytoplasmic but not nuclear polyadenylated RNAs is significantly increased upon FTO knockdown (Figure 2B, green bars). Furthermore, in HEK293T cells, overexpression of wild-type FTO but not ΔNLS-FTO led to a significant decrease of m⁶A in polyadenylated RNAs in the cell nucleus, while the nuclear m⁶A_m level in polyadenylated RNAs was not notably affected (Figure S2C). In the cytoplasm, m⁶A in polyadenylated RNA showed a small percentage change upon FTO overexpression, but m⁶A_m showed a larger decrease upon overexpression of ΔNLS-FTO than wild-type FTO (Figure S2D). The separation of nucleus and cytoplasm for HEK293T cells was validated by western blot (Figure S2E) and qRT-PCR (Figure S2F). The result further confirmed that the FTO-mediated demethylation of m⁶A is prominent in the cell nucleus, whereas cap m⁶A_m is more a target of FTO in the cytoplasm.

Because the m⁶A level is at least 10-fold higher than that of cap m⁶A_m in mRNA, even in the cytoplasm FTO still mediates effective m⁶A and cap m⁶A_m demethylation, in absolute terms. In certain AML cell lines in which we showed proliferation inhibition with FTO knockdown previously (Li et al., 2017), stable knockdown of FTO led to increased m⁶A in both nucleus and cytoplasm, whereas m⁶A_m only increased in the cytoplasmic but not nuclear fraction (Figure 2C, and Figure S2G-S2J). We found extensive cytoplasmic expression of FTO in these AML cell lines (Figure S2K), and quite dramatic cytoplasmic m⁶A demethylation (around 40% increase of the m⁶A level with FTO knockdown), suggesting the critical cytoplasmic m⁶A demethylation roles of FTO in these cells. It is noteworthy that unlike the ratio of 1/10 to 1/15 between m⁶A_m and m⁶A in HeLa, HEK293T, and 3T3-L1 cells (Figure S1C), the m⁶A_m to m⁶A ratio is among 1/25 to 1/30 in the NB4, MONOMAC-6 cells (Figure 2C), and several other AML cells (Li et al., 2017; Su et al., 2018).

FTO-mediated m⁶A Demethylation in mRNA Influences Transcript Level of Target mRNA

To understand the functional impact of m⁶A versus m⁶A_m demethylation by FTO in mRNA, we analyzed transcript level changes of m⁶A and m⁶A_m-containing transcripts upon FTO knockdown in HEK293T cells. The m⁶A sites were adopted from a previously published meRIP sequencing database (Meyer et al., 2012), and the transcripts starting with m⁶A_m were obtained from the previously published study (Mauer et al., 2017). Almost all of these annotated m⁶A_m sites are found within 500 nt from the transcription start site (TSS) (Figure S3A, top panel). To determine the individual effects of cap m⁶A_m and internal m⁶A and to avoid cross-effects between these two modifications, we classified transcripts into four groups. The first group of transcripts contains only m⁶A_m but not m⁶A, the second group of transcripts contains only m⁶A but not m⁶A_m, the third group of transcripts contains both modifications, and the rest of transcripts were considered as the fourth control group to compare against (Figure S3A bottom panel and Table S1). As shown in Figure 3A, using this classification scheme, we found that the global levels of transcripts containing only m⁶A_m don’t appear to statistically change upon FTO knockdown in HEK293T cells, while the global transcript levels of the other three groups show statistically noticeable increases.

Figure 3. FTO-mediated Demethylation of m⁶A but not m⁶A_m in Polyadenylated RNA More Affects Transcript Expression Level Changes.

See also Figure S3 and Table S1.

(A) mRNA transcripts are classified into four groups: transcripts containing only m⁶A_m, containing only internal m⁶A, containing both m⁶A and m⁶A_m (overlap), and transcripts that contain neither m⁶A nor m⁶A_m (labeled as the rest). The global expression level of m⁶A_m-alone transcripts did not significantly change upon FTO knockdown while the global transcript level of other three groups increased significantly.

(B) The cumulative fraction of gene expression shows that compared to the transcripts containing only m⁶A, containing both m⁶A and m⁶A_m, and containing no m⁶A or m⁶A_m, the transcripts containing only m⁶A_m but not m⁶A showed significantly less transcript level change with FTO knockdown compared to the control (in A, P values were determined using the Wilcoxon test; in B, P values were determined using Mann-Whitney test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 3987, m⁶A only; n = 2016, m⁶A only; n = 1307, m⁶ m A and m⁶A_m; n = 5598, the rest).

(C) The cumulative fraction of gene expression shows that FTO CLIP targets showed significantly increased expression level upon transient knockdown of FTO but not non-targets for transcripts containing only m⁶A compared to the rest genes. (In C, P values were determined using Mann-Whitney test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 1895, m⁶A genes overlapped with FTO targets; n = 2092, non-targets; n = 8896, the rest).

Transient knockdown of FTO led to increased transcript levels of the transcripts harboring only m⁶A compared to the transcripts without m⁶A or m⁶A_m in HEK293T cells (Figure 3B). It has previously been observed that increased m⁶A leads to increased mRNA transcript levels in HepG2 and leukemia cell lines (Huang et al., 2017a; Li et al., 2017). This effect was thought to be mediated by a group of newly identified m⁶A reader proteins IGF2BP1–3 (Huang et al., 2017a), which bind internal m⁶A and stabilize methylated mRNAs, or through other yet to be discovered reading mechanisms.

The presence of m⁶A_m adjacent to the 5’ cap was previously reported to stabilize mRNA transcripts by preventing DCP2-mediated cap cleavage and microRNA-mediated mRNA degradation in HEK293T cells (Mauer et al., 2017). However, compared to the other three groups of transcripts (Figure 3B), transcripts containing only m⁶A_m but not m⁶A showed significantly lower transcript level changes upon FTO knockdown. FTO knockdown should lead to elevated m⁶A_m but this does not appear to correlate with higher levels of these transcripts, despite the proposed transcript-stabilizing role of m⁶A_m at the cap (Mauer et al., 2017). This effect is opposite from those observed for transcripts containing both m⁶A_m and m⁶A or transcripts containing only m⁶A. Both of the latter two groups showed significantly larger increases in transcript levels upon FTO knockdown in comparison to the control group. Lastly, grouping just transcripts containing m⁶A versus non-m⁶A without considering m⁶A_m, and transcripts containing cap m⁶A_m versus non-m⁶A_m without considering m⁶A, transcripts containing m⁶A showed increased transcript levels upon FTO knockdown (Figure S3B), while transcripts containing m⁶A_m showed reduced levels compared with the rest upon FTO knockdown (Figure S3C).

These results suggest that though m⁶A_m could be associated with a higher global transcript level and higher transcript stability, the levels of transcripts bearing cap m⁶A_m dos not appear to change with the knockdown of FTO (Figure 3A). The level change of these transcripts was significantly less than the transcripts containing neither m⁶A_m nor m⁶A upon FTO knockdown, whereas transcripts containing m⁶A or both m⁶A and m⁶A_m had significantly larger increases in their transcript levels than transcripts containing neither or just cap m⁶A_m upon FTO knockdown (Figure 3B). We further cross-compared the list of transcripts contain only m⁶A but not m⁶A_m with the published FTO CLIP-seq results (Bartosovic et al., 2017) by separating transcripts contains only m⁶A but not m⁶A_m plus FTO binding into FTO targets and non-targets. Compared to the rest genes (genes except transcripts containing only m⁶A), we clearly observed in Figure 3C a statistically significant right-shift of the cumulative curve representing the increased transcript levels of the m⁶A targets (bound by FTO), with no significant change for non-targets possessing m⁶A but not bound by FTO. qRT-PCR was used to validate the relative transcript levels of three genes containing only m⁶A or m⁶A_m. The transcript level changes of these genes upon FTO knockdown were consistent with the RNA-seq data (Figure S3D).

FTO Demethylates m⁶A in U6 RNA and m⁶A_m in snRNAs

Two recent studies have surveyed FTO-bound RNA species through eCLIP-seq (Van Nostrand et al., 2016) or CLIP-seq of FLAG-tagged FTO (Bartosovic et al., 2017). These studies and our CLIP-seq results in 3T3-L1 cells using FLAG-tagged FTO showed that FTO is able to bind mRNA, tRNA, snRNA, snoRNA (Figure 4A, bar), and rRNA, among which mRNA is enriched (Figure 4A, donut). The purity of anti-FLAG-tagged pulldown was validated in Figure S4A (left) in our studies, and the ³²P labeled protein gel of CLIP is shown in Figure S4A (right), showing the formation of covalently linked FTO-nucleic acids complexes. We were interested in non-mRNA CLIP-targeted RNA species and examined the potential activity of FTO on those RNA species. We first assessed the effect of FTO knockdown or overexpression on the total m⁶A levels in mature rRNA (Figure S4B) in HeLa, HEK293T, and 3T3-L1 cell lines, respectively. No significant changes were observed for m⁶A in purified mature rRNAs, including 28S rRNA (Figure S4C) and 18S rRNA (Figure S4D). We also compared the m⁶A levels of top snRNA targets revealed by CLIP-seq, including U1 RNA (a complex band of U1 RNA and 5.8S, Figure S4E), U2 RNA, and U6 RNA, in Fto homozygous knockout (Fto^−/−) MEF versus the wild-type MEF cells. Interestingly, we observed increased m⁶A in the U6 RNA (Figure 4B) but not other snRNAs (Figure S4F) with the depletion of Fto in the Fto^−/− compared with the wild-type MEF cells, suggesting that U6 m⁶A could be demethylated by FTO.

Figure 4. FTO-mediated Internal m⁶A, Internal and Cap m⁶A_m Demethylation in snRNAs.

See also Figure S4.

(A) Bar graph summarizing the reads distribution of the FTO target RNA species in the 3T3-L1 cells; donut graph showing the percentage of the FTO-bound mRNA in all FTO CLIP-seq targets in the 3T3-L1 cell line.

(B) An increased m⁶A/A ration in the U6 RNA was observed by LC-MS/MS in Fto^−/− MEF cells compared to the control wild-type MEF cells.

(C) Quantification of the m⁶A_m/A ratio of the U1/5.8s RNA and U2 RNA by LC-MS/MS after decapping, showing increased m⁶A_m/A ratios in the Fto^−/− MEF cells compared to the wild-type control.

(D) Quantification of the m⁶A_m/A ratio in U1/5.8s RNA and U2 RNA by LC-MS/MS without decapping, showing increased internal m⁶A_m/A ratios in Fto^−/− MEF cells compared to the wild-type control. P values were determined using Student’s unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 6 experiments in (B) to (D).

Besides m⁶A, we also tested FTO demethylation of m⁶A_m in these U RNAs. After decapping, increased m⁶A_m was observed in both U1/5.8s RNA and U2 RNA (Figure 4C). The quantification results showed a reasonable level of m⁶A_m in U2 RNA, with close to one m⁶A_m per U2 RNA molecule (187 bp long). To determine if the m⁶A_m is near the cap or internal, we measured the internal m⁶A_m level without decapping; decapping is required to remove the cap structure and expose cap m⁶A_m before nuclease P1 and alkaline phosphatase digestion. Internal m⁶A_m was also observed in both U2 and U1/5.8S RNA (Figure 4D), with increased internal m⁶A_m observed in U1/5.8s RNA in the Fto^−/− compared to the wild-type MEF cells. However, considering the average length of mammalian U1 snRNA and 5.8s rRNA (around 160 bp), the 0.1% ratio of internal m⁶A_m/A indicates less than 0.05 m⁶A_m per molecule of U1 snRNA. In U2 RNA both the cap and internal m⁶A_m modifications are more abundant, approaching to 0.5 per molecule of U2 RNA, respectively. The Fto knockout led to a significant increase of the total m⁶A_m level, most likely at the cap. Given the role of snRNPs in the splicing of mRNA (Maniatis and Reed, 1987), and previous work suggesting that FTO deficiency promotes splicing (Zhao et al., 2014), it will be interesting to investigate this potential link between FTO-induced modification changes in snRNAs and mRNA splicing in the future.

FTO Mediates m¹A Demethylation in tRNA

In the CLIP-seq results, we observed that FTO associates with tRNA and the distribution of CLIP reads in tRNA is significantly enriched compared to input (Table S2 and Table S3). We found that the knockdown of FTO led to the increase of 14.7%, 26.5%, and 38.1% of the total m¹A level in tRNA (Figure 5A) compared with the knockdown control in HeLa, HEK293T, and 3T3-L1 cells, respectively. The transcript levels of ALKBH1 and ALKBH3, which were previously shown to mediate tRNA m¹A demethylation (Liu et al., 2016; Ueda et al., 2017), remain unchanged upon the knockdown of FTO (Figure S5A) (Mauer et al., 2017), The overexpression of ΔNLS-FTO led to the opposite effects with about 19.6%, 23.1%, and 20.4% decrease of m¹A in tRNA observed (Figure 5B) when compared with controls in HeLa, HEK293T, and 3T3-L1 cells, respectively. We then separated the nuclear and cytoplasmic fractions of HEK293T cells with FTO knocked down versus the control. Both nuclear tRNA m¹A and cytoplasmic tRNA m¹A increased upon FTO knockdown, with cytoplasmic tRNA m¹A showing a larger increase (Figure 5C).

Figure 5. FTO-mediated tRNA m¹A Demethylation in vitro, inside Cells, and in Brain Tissues.

See also Figure S5, Table S2, and Table S3.

(A) Quantification of the m¹A/G ratio in total tRNA purified from HeLa, HEK293T, and 3T3-L1 cells by LC-MS/MS. The transient knockdown led to an increase of at least ~15% of the m¹A/G ratio in total tRNA compared to control cells.

(B) A consistent decrease (~20%) of the total tRNA m¹A/G ratio was observed upon overexpression of ΔNLS-FTO compared to the control.

(C) Transient knockdown of FTO led to a significant increase of both nuclear and cytoplasmic total tRNA m¹A/G ratio in HEK293T cells.

(D) A significant increase of m¹A in tRNA but not m⁶A or m⁶A_m in the polyadenylated RNA was observed in the Fto^−/− MEF cells compared to the wild-type MEF cells.

(E) Knockout of Fto led to significant increases of both nuclear and cytoplasmic tRNA m¹A levels in tRNA in MEF cells.

(F) A significant increase of the total tRNA m¹A but not polyadenylated RNA m⁶A or m⁶A_m could be observed in Fto^−/− mouse brain tissues compared to wild-type mouse brains.

(G) Demethylation of m¹A in tRNA isolated from HEK293T cells by recombinant FTO. EDTA chelates cofactor iron and inactivates FTO. The ratio of m¹A/G was shown. P values were determined using Student’s unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 4 experiments in (A) to (E); for n = 5 experiments in (F); for n = 3 experiments in (G).

Interestingly, the m¹A level in tRNA is dramatically affected by FTO-mediated demethylation in comparison to m⁶A and cap m⁶A_m in polyadenylated RNAs in Fto knockout mouse cells and tissues. In Fto^−/− MEF cells, we observed a significant increase of the total m¹A in tRNA with no noticeable differences of m⁶A or m⁶A_m in polyadenylated RNAs observed (Figure 5D). Both nuclear tRNA m¹A and cytoplasmic tRNA m¹A showed noticeable increases in Fto^−/− MEF cells compared with the wild-type MEF cells (Figure 5E). The separation of nucleus and cytoplasm for MEF cells was validated by western blot (Figure S5B) and qRT-PCR (Figure S5C). We further tested the changes of the m¹A levels in tRNA as well as m⁶A and cap m⁶A_m in polyadenylated RNAs in the brain tissue of Fto^−/− mice (Figure S5D). While the total m⁶A and m⁶A_m showed slight increases in five pairs of Fto^−/− versus the wild-type mouse brains, the total m¹A level in tRNA showed the largest increases in the brain of the Fto^−/− mice (Figure 5F). It is possible that the surviving Fto knockout mice have adapted to the loss of Fto with compensated changes of the total m⁶A or cap m⁶A_m levels in polyadenylated RNA. Many of these knockout mice still die early (several weeks after birth) in our hands, suggesting the lack of demethylation of individual mRNA or other RNA species still exhibits a noticeable effect on these mice. We also tested the level changes of m¹A in tRNA in two cell lines related to human leukemia. Stable knockdown of FTO leads to significantly increased tRNA m¹A in NB4 and MONOMAC-6 cells (Figure S5E).

We examined whether FTO exhibits biochemical demethylation activity towards m¹A in tRNA by treating the purified total tRNA from HEK293T cells with the FTO protein under reported demethylation conditions. The levels of abundant tRNA methylations, including m¹A, m¹G, m⁷G, and m⁵C, m²G, m²₂G, A_m, and m³C, were measured in the presence or absence of EDTA. Incubation of purified tRNA with FTO led to a dramatic decrease of the m¹A level (Figure 5G), and a slight decrease of m³C, but not levels of other modifications (Figure S5F), in comparison to the control samples, confirming that m¹A in tRNA is a biochemically viable substrate of FTO in addition to m⁶A and m⁶A_m in the polyadenylated RNAs. We then compared the demethylation activity of FTO towards tRNA m¹A and mRNA m⁶A by incubating similar numbers of m¹A and m⁶A together with FTO in the presence or absence of EDTA. The demethylation activity of FTO towards tRNA m¹A is lower than that of polyadenylated RNA m⁶A (Figure S5G). However, considering the abundances of tRNA and polyadenylated RNA inside mammalian cells, the demethylation towards tRNA m¹A could still be functionally important.

m¹A58 is the most common m¹A site in tRNA, residing in the loop region of a TΨC stem-loop motif of tRNA (Hopper and Phizicky, 2003; Phizicky and Hopper, 2010), and can be reversed through demethylation in HeLa and MEF cells (Liu et al., 2016). We reasoned that FTO may preferentially mediate m¹A demethylation in tRNAs with a selectivity towards a stem-loop structure, as previous studies suggested low activities of FTO towards m¹A in single-stranded nucleic acid substrates (Jia et al., 2008). We synthesized two ssRNA probes, one with m¹A in a stem-loop sequence mimicking the TΨC loops in tRNA^His(GUG) (contains m¹A58) and the other one with m¹A in an unstructured sequence (Figure S5H). Indeed, FTO exhibited low activity towards m¹A in the unstructured probe but a high preference for the stem-loop probe (Figure S5I). This biochemical result, combined with the quantifications of modification changes in various cells and tissues, support m¹A in tRNA as another substrate of FTO.

We asked why FTO could have distinct RNA species as demethylation substrates especially the preference for the loop structure. We examined the published protein structures and realized the unique structural features of FTO that allow it to gain access to the modification sites in multiple RNA species. We noticed that the overall structure of FTO has similarities with NSUN6, a tRNA m⁵C methyltransferase (Haag et al., 2015). In the overlapped structures of FTO and NSUN6 with bound tRNA, we observed that the FTO active site could gain access to tRNA m¹A58 (Figure S5J, left: overlapped structures of FTO (raspberry) with the complex of NSUN6 (purple) and tRNA (orange), with the iron center depicted in a yellow ball). However, another known mRNA m⁶A demethylase, ALKBH5, does not carry the similar structural fold as FTO does, which may explain the more constrained substrate specificity of ALKBH5 on mRNA m⁶A (Figure S5J, right: overlapped structures of ALKBH5 (green) with the complex of NSUN6 (blue) and tRNA (orange)).

FTO-mediated tRNA Demethylation Affects Translation

We further validated the m¹A demethylation activity of FTO on selected target tRNAs (Figure 6A). Biotin-labeled DNA probes complementary to tRNA species were used to specifically isolate the corresponding tRNA from 3T3-L1 cells. The m¹A level of the purified specific tRNA was then measured using LC-MS/MS. The results demonstrated that the knockdown of FTO leads to consistent increases of the m¹A levels in most top confident tRNAs targets bound to FTO, including tRNA^Glu(CUC), tRNA^His(GUG), tRNA^Gly(GCC), tRNA^Gly(ACC), tRNA^Asp(GUC), tRNA^Lys(CUU), tRNA^Gln(CUG), and tRNA^Leu(CAA) (Figure 6B). In addition, we also tested several tRNAs that were not FTO targets according to our FTO CLIP-seq data, including tRNA^Ala(AGC), tRNA^Asn(GUU), tRNA^Gly(UCC), and tRNA^Phe(GAA). Results revealed that FTO does not exhibit enzymatic activity on these non-targeted tRNAs (Figure S6A). Therefore, the majority of tRNAs identified by CLIP-seq are substrates of FTO.

Figure 6. FTO Mediates Specific tRNA m¹A Demethylation and Suppresses Translation.

See also Figure S6.

(A) Pie chart showing the top FTO-bound tRNAs identified by CLIP-seq with FLAG-tagged FTO in 3T3-L1 cells.

(B) Quantification of the m¹A/G ratio in tRNA targets bound to FTO by LC-MS/MS. In comparison to the control, m¹A/G ratios of tRNA^Glu(CUC), tRNA^His(GUG), tRNA^Gly(GCC), tRNA^Gly(ACC), tRNA^Asp(GUC), tRNA^Lys(CUU), tRNA^Gln(CUG), and tRNA^Leu(CAA) noticeably increased upon transient knockdown of FTO in 3T3-L1 cell.

(C) Quantification of total protein synthesis in MEF cells by flow cytometry. HPG incorporation into wild-type MEF cell was recorded in gray and the Fto^−/− MEF cells in red.

(D) Supplementation of total tRNA purified from Fto^−/− MEF cell leads to increased translation efficiency in vitro compared to using the total tRNA from wild-type MEF cells.

(E) Cartoon illustration of the reporter assays was shown: firefly luciferase (F-luc) was used as the reporter and Renilla luciferase (R-luc) on the same plasmid was used as the internal transfection control. 6 × GUG(Glu)-coding sequences (recognized by tRNAGlu(CAC)) and 6 × CAC(His)-coding sequences (recognized by tRNAHis(GUG)) were inserted after the PLK promoter region of F-luc as the positive reporter, respectively. The effects of tRNA^Glu(CUC) and tRNA^His(GUG) were revealed by the reporter assay. The reporter assay showed significant increases of protein synthesis in the Fto^−/− MEF cell compared to the wild-type MEF cell when the control reporter and test reporters (with inserted Glu(CUC)-coding sequence or His(GUG)-coding sequence) are transiently expressed in the wild-type and Fto^−/− MEF cell, respectively. P value was determined using Student’s unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. means not significant. Error bars, mean ± s.d. for n = 3 experiments in (B) and (C); n = 8 experiments in (D) and (E).

We then used an alkyne-modified glycine analog, L-homopropargylglycine (HPG), to metabolically label newly synthesized proteins in Fto^−/− and wild-type MEF cells. The HPG-labeled cells were then fluorescently labeled and analyzed by flow cytometry. The results showed a notable increase in the protein synthesis in the Fto^−/− cells compared to the wild-type cells (Figure 6C and Figure S6B).

To separate the effects of the FTO-mediated tRNA m¹A demethylation from the effects of the FTO-mediated adenosine demethylation in mRNA on translation, we performed an in vitro translation assay by using a rabbit reticulocyte-based luciferase reporter system (Promega). tRNAs extracted from Fto^−/− and wild-type MEF cells were subjected to the in vitro luciferase mRNA translation assay using rabbit reticulocyte lysate, respectively. The activity of the translated luciferase protein was measured in the incubation mixture by a luminometer. As shown in Figure 6D, the incubation of tRNAs extracted from Fto^−/− cells induced significantly higher luciferase signals than tRNAs extracted from the wild-type MEF cells, indicating that the FTO-mediated tRNA demethylation led to increased protein translation efficiency.

We further designed a reporter assay with six repeated codon sequences (6×GAG(Glu) for tRNA^Glu(CUC) target and 6×CAC(His) for tRNA^His(GUG) target) added to the 5’ of firefly luciferase (F-luc) (Figure 6E, left); a second Renilla luciferase (R-luc) encoded in the same plasmid was used to normalize the expression efficiency. For both reporters, the significantly elevated translation was observed in the Fto^−/− MEF cells compared to the wild-type MEF cells (Figure 6E, right). We also applied a reporter assay with six repeated codon sequences (6×GGA(Gly)) for non-target tRNA^Gly(UCC) to the same reporter system mentioned above as the negative control. As expected, there was no noticeable change of translation between Fto^−/− MEF cells and the wild-type MEF cells (Figure S6C). These results further indicate that the FTO-mediated tRNA m¹A demethylation negatively affects translation, which is consistent with the previous discovery that m¹A-methylated tRNAs are preferentially recognized and delivered to translation-active polysomes (Liu et al., 2016).

Discussion

FTO harbors two structural domains, an N- and a C-terminal domain. The active site of FTO resides in its N-terminal domain and is exposed to the protein surface (Han et al., 2010). Unlike ALKBH1, which possesses a distinct tRNA-binding domain (Liu et al., 2016), FTO contains an L1-loop surrounding its active site across to the Jelly-roll motif on the other side of the active site. The L1-loop facilitates FTO’s selectivity towards single-stranded RNA substrates. The additional C-terminal domain, which is absent in the other AlkB family proteins (Yang et al., 2008; Yu et al., 2006), may endow FTO with the ability to associate with potential protein partners. These structural features may allow FTO to accomplish demethylation of different RNA substrates via interacting with diverse partner proteins depending on the cellular context.

Demethylation of m⁶A and Cap m⁶A_m in mRNA by FTO in the Nucleus and Cytoplasm

Both internal m⁶A and cap m⁶A_m exist in polyadenylated RNA in mammalian cells. The levels of internal m⁶A are about 10–20 folds higher than those of cap m⁶A_m in different cell lines (Wei et al., 1975a; Wei et al., 1975b). Our cellular results and biochemical assays using FTO and purified mRNA unequivocally show demethylation of both internal m⁶A and cap m⁶A_m by FTO. While FTO catalyzes more potent demethylation on cap m⁶A_m over m⁶A in vitro, the extent of demethylation inside cells is opposite with 5–10 folds more m⁶A demethylated by FTO compared with m⁶A_m in cell lines tested, in absolute terms. The primary polyadenylated RNA substrate of FTO in the cell nucleus appears to be m⁶A, with cap m⁶A_m more a target of FTO in the cytoplasm (Figure 7).

Figure 7. A Model for FTO-mediated RNA Demethylation.

Nuclear m⁶A is the main substrate of FTO in the cell nucleus. FTO also mediates m¹A tRNA demethylation as well as m⁶A (U6) and m⁶A_m (U2 in particular but also U1) snRNA demethylation in the nucleus. FTO can localize to the cytoplasm and mediate mRNA m⁶A and cap m⁶A_m demethylation as well as tRNA m¹A demethylation in the cytoplasm.

Nuclear mRNAs consist of only a portion of total cellular mRNAs. A notable fraction of nuclear mRNAs or other polyadenylated RNAs are substrates of FTO-mediated demethylation. In HeLa cells, although FTO could impact a significant fraction of nuclear polyadenylated m⁶A, this effect is “diluted” when examining the total mRNA pool. It should be noted that when FTO shuttles to the cytoplasm, it may still demethylate a portion of m⁶A in cytoplasmic mRNAs as it could gain access to both internal m⁶A and cap m⁶A_m on mRNA (Figure 7). In certain AML cells, the cytoplasmic m⁶A demethylation by FTO can affect almost 40% of the total m⁶A (Figure S2H and Figure S2J). Because the total m⁶A level is considerably higher than that of m⁶A_m in mRNA, a change of the similar percentage of m⁶A versus m⁶A_m manifests as a much larger number of m⁶A changes in mRNA.

Functional Relevance of Cap m⁶A_m and Internal m⁶ A Demethylation by FTO

A previous study showed that the presence of m⁶ A_m adjacent to the 5’ cap prevents decapping and stabilizes mRNA transcripts in HEK293T cells (Mauer et al., 2017). In order to further probe the effect of m⁶A_m, we categorized the previously identified transcripts (Mauer et al., 2017) into four classes: cap m⁶ A_m -only transcripts, m⁶ A_m-only transcripts, transcripts containing both modifications, and transcripts with neither modification. In HEK293T cells, m⁶A-only transcripts showed increased transcript levels upon FTO knockdown (Figure 3A and 3B), suggesting the presence of mechanisms that stabilize m⁶A-modification transcripts. The YTHDF2-mediated decay of methylated transcripts operates in HeLa cells (Ke et al., 2017; Zhao et al., 2014), but may not be the predominant pathway affecting the transcript levels of m⁶A-containing transcripts in HEK293T and AML cells (Li et al., 2017).

To our surprise, when we analyzed cap m⁶A_m-only transcripts, we found that FTO knockdown does not noticeably affect mRNA transcript levels of cap m⁶A_m-only genes in reported HEK293T cells. A careful analysis indicated that the previous analysis combined cap m⁶A_m-only transcripts with transcripts that contain both cap m⁶A_m and m⁶A (Mauer et al., 2017).

The transcript stabilizing effect is most likely derived from the internal m⁶As that co-exist with cap m⁶A_m in transcripts that contain both modifications. So far, all functional relevant reports of FTO have been ascribed to internal m⁶A in mRNA, which include a wide range of different systems (Cui et al., 2017; Gokhale et al., 2016; Su et al., 2018; Xiang et al., 2017; Yu et al., 2018; Zhou et al., 2015; Zhou et al., 2018; Zou et al., 2016). A recently published FTO CLIP-seq result revealed a peak of FTO binding closely around known m⁶A sites and in intronic regions in HEK cells, suggesting that FTO mediates intronic m⁶A demethylation (Bartosovic et al., 2017).

The discoveries of the cap m⁶A_m in stabilizing mRNA and its demethylation by FTO are interesting and could potentially open new research directions (Mauer et al., 2017); however, the effects of FTO on cap m⁶A_m of these mRNA transcripts require future investigations. The methyltransferase will need to be identified and the phenotypes of the methyltransferase knockout cells or mice need to be investigated and compared to severe phenotypes observed in Fto knockout mice. The role of cap m⁶A_m demethylation in RNA species other than mRNA merits consideration. Appropriate and sensitive detection method and procedures are critical to investigate the FTO-mediate demethylation in the future research of the community.

tRNA m¹A and Small Nuclear RNA Demethylation by FTO

CLIP-seq results of FTO followed by LC-MS/MS tests revealed additional substrates of FTO, including m⁶A in U6 RNA (Figure 4B) as well as internal and cap m⁶A_m in snRNA (Figure 4C and 4D). The demethylation towards modifications in these U RNAs may affect mRNA splicing. However, cautions must be taken as the ratio of m⁶A/A or m⁶A_m/A for these small nuclear RNAs ranges from 0.05–2%. For short RNA species such as U1 and U6, the very low modified A/A ratio (0.05–0.5%) indicated far below stoichiometric methylation with 0.02–0.2 m⁶A_m or m⁶A per snRNA molecule. Close to one m⁶A_m, either at the cap or internal, exists per U2 RNA molecule, and the cap U2 m⁶A_m demethylation by FTO can have functional consequences in splicing, as previously suggested (Maniatis and Reed, 1987; Zhao et al., 2014).

When comparing the modification levels in the Fto^−/− mouse MEF cells and brain tissues with the wild-type control we observed a large change of m¹A in tRNA. m¹A methylation in tRNA is important for the cell viability and fitness (Oerum et al., 2017), with m¹A/A ratio ranging around 1–3%. Previous studies have shown that m¹A58 is installed by the methyltransferase complex Trmt6/61 (Anderson et al., 1998) and can be reversed by ALKBH1 (Liu et al., 2016). The dynamic regulation of tRNA m¹A58 could impact translation initiation and elongation. In the current study, our CLIP sequencing data showed that FTO binds m¹A58-containing tRNAs. Further biochemical characterization revealed that FTO possesses effective tRNA m¹A demethylation activity in vitro and inside cells (both nucleus and cytoplasm). In fact, FTO was originally shown to be able to reverse DNA 1mA damage in 2007 (Gerken et al., 2007), most likely on ssDNA (Jia et al., 2008). The reporter assay confirmed that the FTO-mediated tRNA demethylation affects translation, which is reminiscent of the mechanism used by ALKBH1 in regulating translation elongation. FTO mediates both nuclear and cytoplasmic m¹A demethylation in tRNA. The mRNA cap m⁶A_m demethylation (Mauer et al., 2017) and tRNA m¹A demethylation by FTO could collectively repress translation, with cap demethylation perhaps destabilizing mRNA in certain cells or growth conditions and tRNA demethylation reducing translation. The m¹A tRNA demethylation by FTO may also impact tRNA fragments during certain development processes, which merits further investigations.

In summary, we uncover a previously unrecognized spatial regulation of RNA demethylation through FTO with distinct roles in the cell nucleus versus cytoplasm. Internal m⁶A is the primary target of nuclear FTO in most cells. FTO can be localized to the cytoplasm and mediate both m⁶A and cap m⁶A_m demethylation in polyadenylated RNA (Figure 7). In certain AML cells, we observed dramatic (up to 40%) m⁶A mRNA demethylation by FTO in the cytoplasm. FTO also mediates demethylation of m¹A in tRNA in both cell nucleus and cytoplasm, which may tune translation together with cytoplasmic cap m⁶A_m mRNA demethylation. Some of the phenotypes observed for FTO in brain tissues may be partially derived from tuning translation through tRNA demethylation. We also identified additional RNA substrates including m⁶A and internal and cap m⁶A_m in snRNAs (Figure 7). The snRNA demethylation, in particular, m⁶A_m in U2 RNA, may contribute to effects on splicing (Zhao et al., 2014). Lastly, the nuclear-centralized m⁶A demethylation and the ability of FTO to mediate m¹A demethylation suggest potential nuclear roles of the m⁶A or m¹A demethylation, on both polyadenylated RNA and other nuclear RNA species, in affecting gene expression, which will be investigated in the future.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests may be directed to and will be fulfilled by the lead contact corresponding author Chuan He (chuanhe@uchicago.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell culture

Wild-type MEF (mouse embryonic fibroblasts) and Fto^−/− MEF cell lines were obtained from Prof. Arne Klungland. Spontaneously immortalized clonal MEF cell lines were established from individual E13.5-E15.5 days-old embryos obtained from heterozygous Fto^+/− crossings. Limbs were removed and remaining tissue chopped into small pieces and homogenized into single-cell suspension using a pipette. MEFs are grown in DMEM with 10% serum, 2 mM glutamine and penicillin/streptomycin. Cells grown for 4–7 days were trypsinized and frozen in complete medium with 10% DMSO as stock. Cells are grown in complete medium for 2–4 days before using them for following analysis. Human HeLa, HepG2, HEK293T, Mel624, and 3T3-L1 cell lines were purchased from ATCC. The aforementioned cells were maintained in DMEM (Gibco, #11965) with 10% fetal bovine serum (FBS, Gibco, 10438–026) and 1× Pen/Strep (Gibco, 15140) at 37°C with 5% CO₂. FTO stable knockdown and the knockdown control AML cells were obtained from Prof. Jianjun Chen and cultured as previously reported (Li et al., 2017).

Tissues

Wild-type and Fto^−/− mouse brains were obtained from Prof. Arne Klungland. Littermate animals were used as Fto^+/+ controls (wild-type). The mice were bred and housed in a 12-hour light/dark cycle at the Comparative Medicine, Oslo University Hospital, Rikshospitalet, Norway, with a diet of pellets and water ad libitum. Brains were carefully removed from decapitated mice and quickly frozen on dry ice and stored at −80°C.

METHOD DETAILS

Plasmid construction

FLAG-tagged human FTO in the pcDNA 3.0 vector was obtained as previously reported (Jia et al., 2011). FLAG-tagged mouse FTO in the pCMV6-Entry vector was purchased from OriGene (MR208064). FLAG-tagged human ΔNLS-FTO mutant in pCMV6-Entry was generated by the deletion of 17 amino acids from the N-terminal of FTO full-length cDNA. High-purity plasmids used for mammalian cell transfection were prepared using HiSpeed Plasmid Maxi Kit (QIAGEN).

siRNA knockdown, and plasmid transfection

AllStars negative control siRNA (Qiagen, 1027281) was used as control siRNA in knockdown experiments. siRNA directed against human FTO (used in HEK293T and HeLa cells) was ordered from Qiagen (SI04293625, Hs_FTO_7 FlexiTube siRNA). Silencer^® Select against mouse FTO (used in 3T3-L1 cells) was ordered from ThermoFisher (assay ID 166879). Cells were resuspended 16 hours prior to the transfection and maintained at 50% confluency for siRNA transfection or 80% confluency for plasmid transfection. Transfection was achieved by using Lipofectamine RNAiMAX (Invitrogen) for siRNA and Lipofectamine 2000 (Invitrogen) for DNA plasmids, following the manufacturer’s protocols, respectively. The cells were harvested after 48 hours post siRNA transfection or 24 hours post plasmid transfection for subsequent experiments.

Cytoplasm and nucleus separation

Cytoplasm and nucleus were separated using NE-PER^® Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific, 78833) following the manufacturer’s protocol. The capability of this kit to efficiently separate cytoplasm and nucleus for HEK293T cell and MEFs cell has been reported previously (Huang et al., 2017b; McKenna and Wright, 2015). According to the published protocol (Conrad and Ørom, 2017), the separation efficiency was validated using qRT-PCR for total RNA purified from different cellular fractions as well as western blot.

Immunofluorescent imaging sample preparation

Immunofluorescent imaging was performed as reported previously (Zhao et al., 2014). In brief, the cells were cultured in an 8-well chamber (Lab-Tek) after treatment indicated in each experiment. The cells were washed once in phosphate buffered saline (PBS) and then fixed in 4% paraformaldehyde in PBST (PBS with 0.05% Tween-20) at the room temperature for 15 min. The fixing solution was removed, and −20 °C chilled methanol was immediately added to each chamber. After incubated for 10 min at the room temperature, the cells were rinsed once with PBS and blocked with blocking solution (1% BSA with PBST) for 1 hour at the room temperature. After that, the blocking solution was replaced with the primary antibody (ab126605 for FTO, 1:1000 dilution in the blocking solution, identical to the previous report (Aas et al., 2017)) and incubated for 1 hour at room temperature. After being washed 4 times with PBST (300 μl, 5–10 min for each wash), secondary antibody (1:300 dilution in PBST) was added to the mixture and incubated at room temperature for 1 hour. After washing 4 times with PBST (300 μl, 5–10 min for each wash), anti-fade reagent (Slowfade, Invitrogen) was added to mount the slides.

RNA purification

Total RNA was purified with TRIzol^® reagents (Thermofisher Scientific, #15596018) from cell lysate. Polyadenylated RNA was purified from total RNA with two rounds of polyA tail purification using Dynabeads^® mRNA DIRECT^™ kit (Thermofisher Scientific, #61006). The rRNAs was further removed using RiboMinus^™ Eukaryote kit (Thermofisher Scientific, A1083708). Small RNAs (< 200 nt) were separated from total RNA and then subjected to size selection using 15% TBE-urea gel (Thermofisher Scientific, EC6885BOX). tRNA was then sliced and recovered from the gel.

For small nuclear RNAs, the nuclei were first isolated using aforementioned NE-PER kit and the total nuclear RNA was purified using TRIzol^® reagents. Small nuclear RNAs (< 200 nt) were separated from total nuclear RNA and then subjected to size selection using 6% TBE-urea gel (Thermofisher Scientific, EC6865BOX). Individual small nuclear RNA was then sliced based on molecular weight and recovered from the gel.

Biotinylated single-stranded DNA probes:

As previously reported (Liu et al., 2016), DNA probes were designed to complement with the 3’ gene sequences of tRNA^Glu(CUC), tRNA^His(GUG), tRNA^Gly(GCC), tRNA^Gly(ACC), tRNA^Asp(GUC), tRNA^Lys(CUU), tRNA^Gln(CUG), and tRNA^Leu(CAA), tRNA^Ala(rGC), tRNA^Asn(GUU), tRNA^Gly(UCC), and tRNA^Phe(rAA). The probes used for tRNA selection are listed in Table S4.

Individual tRNA isolation

As previously reported (Liu et al., 2016), Streptavidin-conjugated C1 magnetic Dynabeads (Invitrogen) were used for individual tRNA target isolation. 20 μl of RNase-free beads were washed once with buffer A (10 mM Tris-HCl, pH 7.5, 2 mM EDTA, 2 M NaCl), and finally resuspended in 20 μl of buffer A. Subsequently, 200 μM of biotinylated oligonucleotides in 20 μl of water were mixed with resuspended Dynabeads in buffer A and incubated at room temperature for 30 min. After the incubation, the oligonucleotide-coated Dynabeads were then washed for four times with buffer B (5 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 M NaCl) and equilibrated in 6× SSC solution (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate, pH 7.0). Total tRNA in 6× SSC solutions and the oligonucleotide-coated Dynabeads and were heated for 10 min at 75°C separately. The tRNA and beads were mixed and incubated together at 75°C for another 10 min. Thereafter, the suspension was gently rotated at room temperature for 3 hours before washed, in succession, three times with 3× SSC, twice with 1× SSC, and several times with 0.1× SSC until the absorbance of the wash solution at 260 nm closed to zero. tRNA retained on the beads was eluted three times using RNase-free water.

Quantitative analysis of RNA modification levels

For the quantification of the m⁶A and cap m⁶A_m in polyadenylated RNA, purified polyadenylated RNA (described in the RNA purification section above) was further treated with RppH (NEB, M0356S) in NEB Thermopol buffer for 2 hours to remove the cap. 50 ng polyadenylated RNA after the removal of the cap, purified tRNA and purified snRNAs were digested by nuclease P1 (Sigma, N8630), respectively, in 20  l of buffer containing 25 mM NaCl and 2.5 mM ZnCl₂ for 1 h at 42 °C. Subsequently, 1 unit of alkaline phosphatase (Sigma, P5931) and NH₄HCO₃ (100 mM) were added and the sample was incubated for another 1 h at 37 °C. The samples were then filtered (0.22  m, Millipore) and injected into a C18 reverse phase column coupled online to Agilent 6460 LC-MS/MS spectrometer in positive electrospray ionization mode. The nucleosides were quantified by using retention time and the nucleoside to base ion mass transitions (268-to-136 for A; 296-to-150 for m⁶A_m, 282-to-150 for m⁶A and m¹A (retention times of 2.5 and 0.9 min, respectively)). Quantification was performed by comparing with the standard curve obtained from pure nucleoside standards running with the same batch of samples. The m⁶A and m⁶A_m levels were calculated as the ratio of m⁶A and m⁶A_m to A, respectively. The tRNA m¹A level was calculated as the ratio of m¹A to G.

Biochemistry assay of FTO activity in vitro

Similar to a previous report (Liu et al., 2016), the demethylation activity assay was performed in standard 20 μl of reaction buffer containing KCl (100 mM), MgCl₂ (2 mM), SUPERNase In (0.2 U/μl, life technology), L-ascorbic acid (2 mM), α-ketoglutarate (300 M), (NH₄)₂Fe(SO₄)₂·6H₂O (150 μM), and 50 mM of HEPES buffer (pH 6.5). For m⁶A and m⁶A_m in polyadenylated RNA, 200 ng polyadenylated RNA purified from HEK293T cells was incubated with 2 M FTO purified in mammalian cells (High concentration) or 0.2 M FTO (low concentration) in the above reaction buffer for 3 hours and then quenched by the addition of 5 mM of EDTA, respectively. For tRNA m¹A, 200 ng tRNA purified from HEK293T cell was incubated with 2 M FTO purified in mammalian cells in the aforementioned reaction buffer for 3 hours and then quenched by the addition of 1 mM of EDTA. Excessive amount of EDTA was added to control samples. To incubate similar amounts of m¹A and m⁶A with FTO, 100 ng purified mammalian polyadenylated RNA and 20 ng tRNA were pooled together and reacted under the aforementioned demethylation conditions; tRNA m¹A/A ratio is around 2–4% while polyadenylated RNA m⁶AA ratio is around 0.4–0.8%. The RNAs were then isolated with 100 μl TRIzol^® reagents (Thermofisher Scientific, #15596018) using standard protocol and subjected to RNA digestion prior to LC-MS/MS analysis.

Biochemical demethylation assays of FTO towards m¹A in single-stranded RNA (ssRNA) were conducted in 20 μl of reaction mixture together with 2 M FTO purified from the mammalian cells and 100 ng probes. The reaction was incubated at 37°C for 1 h and quenched by addition of 5 mM of EDTA. Excessive amount of EDTA was added to control samples. The samples were then centrifuged at 16000 g for 30 min at 4°C. The supernatant was collected for RNA digestion prior to LC-MS/MS analysis. The sequences of the probes were listed in the STAR Method.

Luciferase reporter assay

The luciferase activity was measured by the luminometer (SYNERGY|HTX) following the manufacturing settings of luciferase reporter assay.

Similar to a previous report (Ueda et al., 2017), in vitro translation reaction was performed using the Flexi^® Rabbit Reticulocyte Lysate System (Promega, L4540). Total tRNA purified from wild-type MEF cell and Fto^−/− MEF cell was incubated for 30 min at 30°C with the reaction mixture following the manufacturing protocol.

For luciferase reporter inside cells, tRNA^His(GUG) reporter was obtained as previously reported (Liu et al., 2016). Similarly, F-luc-6 × Glu(CAC) and F-luc-6 × Gly(UCC) reporter plasmid were obtained by inserting the sequence of GAGGAGGAGGAGGAGGAG and GGAGGAGGAGGAGGAGGA, respectively, before the F-luc coding region. To normalize translation differences between Fto^−/− and wild-type MEF cells introduced by any other factors including initiation (i.e. variations of cellular levels of tRNA^iMet and other tRNAs), a control reporter (F-luc plus R-luc) devoid of these 6× sequences was also transfected. The normalization factor from this control reporter was applied to signals obtained from the 6×GAG(Glu)-reporter, 6×CAC(His)-reporter, and 6×UCC(Gly)-reporter. General protocol: The wild-type and Fto^−/− MEF cells were maintained in the 6-well plates at 80% confluency. 500 ng of reporter plasmids (pmirGlo empty vector as control or pmirGlo-specific tRNA anti-codon inserted vector) were transfected into wild-type and Fto^−/− MEF cells following the aforementioned plasmid transfection protocol. After 6 hours, each well was each well was trypsin-digested, extensively washed with PBS, and re-seeded into a 96-well plate. After 18 hours from re-seeding, the cells in the 96-well plate were assayed using Dual-Glo^® Luciferase Assay System (Promega, E2920). Renilla Luciferase (R-luc) was used to normalize firefly luciferase (F-luc) activity to evaluate the translation efficiency of the reporter.

Measurement of protein synthesis rate

HPG (Life Technologies; 50 nM final concentration) was added to the culture medium and incubated for 1 h. Cells were then lifted from plates and washed twice with phosphate buffered saline (PBS) and then fixed in 0.5 ml of 1% paraformaldehyde (Affymetrix) in PBS for 15 min on ice followed by permeabilizing in 200 μl PBS supplemented with 3% FBS (Sigma) and 0.1% saponin (Sigma) for 5 min at room temperature. The azide-alkyne reaction was performed with the Click-iT Cell Reaction Buffer Kit (Life Technologies) and azide-modified Alexa Fluor 488 (Life Technologies) at 5 mM final concentration. After a 30-minute reaction, the cells were washed twice with 3% BSA. All cells were filtered through a 40-μm cell strainer to obtain single cell suspensions. For flow cytometric analysis, all sorted fractions were double sorted to ensure high purity. Data were analyzed by FlowJo (Tree Star) software.

CLIP-seq

Similar to a previous report (Liu et al., 2016), in brief, for each cell line, 5 plates of cells in 15 cm style dish were grown until 80% confluency (~10⁷ cells). The medium was aspirated and the cells were washed with PBS once and irradiated once with 400 mJ/cm2 at 254 nm in Stratalinker on ice. After irradiation, cells were harvested with a scraper, transferred to a microtube, pelleted at 4000 rpm for 3 min at 4°C, and then lysed in RIPA lysis buffer (Sigma, R2078) with cOmplete^™, Mini, EDTA-free Protease Inhibitor Cocktail and SUPERNase In at 4°C for 4 hours. Subsequently, the lysis mixture was centrifuged at 17,000 g at 4°C for 30 min and the supernatant was carefully collected. Meanwhile, add 100 μl Anti-FLAG^® M2 Magnetic Beads to a fresh microtube, wash beads twice with 1 ml lysis buffer. Change to a new tube, leave the beads in last wash until IP experiment. 100 μl of Anti-FLAG^® M2 Magnetic Beads was added into the supernatant and mixed gently at 4°C for 4 hours. The supernatant was then discarded and the beads were washed for 3 times with 1 ml of high salt buffer and another 8 times with 1 ml of wash buffer (Wang et al., 2009). The RNAs were then eluted by proteinase K digestion following the manufacturer’s protocol. Subsequently, RNAs were recovered by phenol/chloroform extraction and then subjected to library preparation using NEBNext^® Multiplex Small RNA Library Prep Set for Illumina^® (NEB, E7300S).

CLIP samples were pooled and sequenced with Hiseq-2000. Raw reads from CLIP samples were first trimmed according to recommended settings (Bolger et al., 2014). Gene structure annotations were downloaded from UCSC mm10 RefSeq/Repeatmasker. For CLIP-seq, we used a modified version of PARalyzer (Corcoran et al., 2011) to allow for all mutations, rather than just T to C mutations. For tRNA, the reads in each tRNA gene were counted. Similar to the previous report (Bartosovic et al., 2017), RNA-seq with rRNA depletion for 3T3-L1 was reanalyzed and used as the input for CLIP-seq (Schick et al., 2015). The reads distribution in tRNA of input (RNA-seq) in 3T3-L1 cells were shown in Table S2. The CLIP-seq in HEK293T cells was reanalyzed (Bartosovic et al., 2017). The reads distribution in tRNA of CLIP-seq and input (RNA-seq) in HEK293T cells were shown in Table S3.

Classification of the m⁶A and m⁶A_m only genes

The m⁶A_m list was obtained from supplementary table 1 in the previous study (Mauer et al., 2017). The m⁶A list was obtained from table S6 in the previous report (Meyer et al., 2012). In the reported m⁶A list, the peak site was annotated. The genes containing both m⁶A and m⁶A_m were first classified by the overlap of two lists followed by deducting the genes that have m⁶A in 5’UTR (though there must be some over deduction in this step as definitely not all the 5’UTR m⁶As are m⁶A_m, but just to be on the safe side). m⁶A only genes were classified by deducting genes containing both m⁶A and m⁶A_m mentioned above from the genes in 2012 list. Similarly, cap m⁶A_m only genes were classified by deducting genes containing both m⁶A and m⁶A_m mentioned above from the genes in 2017 list. The list of three groups of genes was provided in Table S1.

Replication

For the quantification of modification changes using LC-MS-MS: 6 replicates (RNAs were purified from three plates of cells each time and two rounds of experiments were performed) were tested for most samples of transient knockdown or overexpression and controls; 4 replicates (RNAs were purified from two plates of cells each time and two rounds of experiments were performed) were tested for samples with Fto knockout, wild-type or FTO stable knockdown and controls; 5 replicates (RNAs from five Fto^−/− and wild type brains, respectively) were tested for the mouse brain samples of Fto knockout and wild-type; and 3 replicates were tested for the individual tRNAs purified from Fto^−/− and wild-type MEF cells. For the reporter assay experiments: in vitro, 8 replicates were performed (tRNAs were purified from four plates of cells each time and two rounds of experiments were performed); inside cells, 8 replicates were performed (cells were seeded in eight wells parallelly in one 96-well plate). For the HPG incorporation experiment, 3 replicates (three plates of cells) were tested.

QUANTIFICATION AND STATISTICAL ANALYSIS

P values were determined using the two-tailed Student’s unpaired t-test for unpaired samples in LC-MS/MS results, protein synthesis results, and reporter results. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Error bars represent mean ± s.d.. n.s. means not significant, n = 3 for experiments in the Figure 5G, 6B, 6C, S4C, S4D, S5F, S5G, S5I, and S6A; n = 4 for experiments in the Figure 1B, 2C, 5A to 5E, S2G to J, and S5E; n = 5 for experiments in Figure 5F; n = 6 for experiments in Figure 2B, 4B to 4D, S1C to S1E, S2B to S2D, S3D, and S4F; n = 8 for experiments in Figure 6D, 6E and S6C. P values were determined using the Wilcoxon test in expression level analysis and using Mann-Whitney test in cumulative fraction results. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Error bars represent mean ± s.d.. n.s. means not significant, for n = transcript numbers in Figure 3 and Figure S3.

DATA AND SOFTWARE AVAILABILITY

Data resources

Sequencing data have been deposited into the Gene Expression Omnibus (GEO) under the accession number GSE106395.

Highlights.

1. FTO mediates both internal m⁶A and cap m⁶A_m demethylation of polyadenylated RNA

2. FTO exhibits differential substrate preferences in nucleus versus cytoplasm

3. The internal mRNA m⁶A demethylation by FTO correlates with transcript level changes

4. FTO affects snRNA m⁶A and m⁶A_m levels and mediates tRNA m¹A demethylation