Research Article

Cap-specific terminal N6-methylation of RNA by an RNA polymerase II–associated methyltransferase

See allHide authors and affiliations

Science  11 Jan 2019:
Vol. 363, Issue 6423, eaav0080
DOI: 10.1126/science.aav0080

A cap-specific m6A writer

N6,2′-O-dimethyladenosine (m6Am) is present at the transcription start nucleotide of capped mRNAs in vertebrates. Akichika et al. quantified the abundance of this modification in the transcriptome and identified the writer protein, cap-specific adenosine methyltransferase (CAPAM), needed for this modification. CAPAM contains a unique structure that recognizes cap-specific N6-methyladenosine (m6A) as the substrate. The protein interacts with RNA polymerase II, suggesting that the modification occurs cotranscriptionally. The m6Am promotes the translation of capped mRNAs in a eIF4E-independent fashion.

Science, this issue p. eaav0080

Structured Abstract

INTRODUCTION

N6-methyladenosine (m6A), an abundant modification in eukaryotic mRNAs and long-noncoding RNAs, has been recognized as a major epitranscriptome mark that plays critical roles in RNA metabolism and function. In addition to the internal m6A, N6, 2′-O-dimethyladenosine (m6Am) is present at the transcription start site of capped mRNAs in vertebrates. Previous studies reported that an eraser protein, FTO, demethylates N6-methyl group of m6Am and destabilizes a subset of mRNAs, suggesting a possible function of m6Am in stabilizing A-starting capped mRNAs. However, the biogenesis and functional role of m6Am remain elusive.

RATIONALE

To reveal the functional and physiological roles of m6Am, it is necessary to identify a writer protein for N6-methylation of m6Am. We first established a highly sensitive method to analyze 5′-terminal chemical structures of the capped mRNAs using mass spectrometry (RNA-MS), and then measured m6Am methylation status accurately. We employed RNA-MS to identify the writer gene by a reverse genetic approach. We chose several candidates of uncharacterized methyltransferases (MTases) that are conserved in vertebrates, but not in yeast, which does not have m6Am. Each of the candidates was knocked out in human cells. If the target gene is disrupted, RNA-MS can detect the absence of m6Am in mRNAs prepared from the knockout cells.

RESULTS

RNA-MS analysis revealed that m6Am modification in human mRNAs is more abundant (92%) than previously estimated. We identified human PCIF1 as cap-specific adenosine-N6-MTase (CAPAM) responsible for N6-methylation of m6Am. Indeed, m6Am disappeared completely and converted to Am modification in mRNAs prepared from the CAPAM knockout (KO) cells. The CAPAM KO cells were viable, but sensitive to oxidative stress, implying the physiological importance of m6Am. We showed that CAPAM catalyzes N6-methylation of m6Am in the capped mRNAs in an S-adenosylmethionine (SAM)–dependent manner. A series of biochemical studies revealed that CAPAM specifically recognizes the 7-methylguanosine (m7G) cap structure and preferentially N6-methylates m7GpppAm rather than m7GpppA, indicating the importance of the 2′-O-methyl group of the target site for efficient m6Am formation. CAPAM has a N-terminal WW domain that specifically interacts with the Ser5-phosphorylated C-terminal domain (CTD) of RNA polymerase II (RNAPII), suggesting that the CAPAM is recruited to the early elongation complex of RNAPII and introduces m6Am in a nascent mRNA chain cotranscriptionally. We also solved the crystal structure of CAPAM complexed with the cap analog and SAM analog. The core region of CAPAM is composed of MTase and helical domains. The m7G cap is bound to a pocket formed by these two domains. The SAM analog is recognized by an active site with the characteristic NPPF motif in the MTase domain. This structure reveals the molecular basis of cap-specific m6A formation. RNA-sequencing analysis of the CAPAM KO cells revealed that loss of m6Am does not affect transcriptome alteration. This result does not support the proposed function of m6Am in stabilizing A-starting capped mRNAs. Instead, ribosome profiling of the CAPAM KO cells showed that N6-methylation of m6Am promotes the translation of capped mRNAs.

CONCLUSION

We identified PCIF1/CAPAM as a cap-specific m6A writer for vertebrate mRNAs. Structural analysis revealed the molecular basis of cap-specific m6A formation by CAPAM. The ribosome profiling experiment revealed that CAPAM-mediated m6Am formation promotes translation of A-starting mRNAs, rather than stabilization of mRNAs.

Sequential and cotranscriptional m6Am formation mediated by CAPAM.

CAPAM is recruited to the early elongation stage of RNAPII through specific interaction between the WW domain and Ser5-phosphorylated CTD. The m7G cap MTase (RNMT) complexed with the capping enzyme (RNGTT) and 2′-O-MTase (CMTR1) are also recruited to this complex, indicating a hierarchical formation of m7Gpppm6Am—pppA, GpppA, m7GpppA, m7GpppAm, and m7Gpppm6Am.

Abstract

N6-methyladenosine (m6A), a major modification of messenger RNAs (mRNAs), plays critical roles in RNA metabolism and function. In addition to the internal m6A, N6, 2′-O-dimethyladenosine (m6Am) is present at the transcription start nucleotide of capped mRNAs in vertebrates. However, its biogenesis and functional role remain elusive. Using a reverse genetics approach, we identified PCIF1, a factor that interacts with the serine-5–phosphorylated carboxyl-terminal domain of RNA polymerase II, as a cap-specific adenosine methyltransferase (CAPAM) responsible for N6-methylation of m6Am. The crystal structure of CAPAM in complex with substrates revealed the molecular basis of cap-specific m6A formation. A transcriptome-wide analysis revealed that N6-methylation of m6Am promotes the translation of capped mRNAs. Thus, a cap-specific m6A writer promotes translation of mRNAs starting from m6Am.

RNA molecules are enzymatically modified after transcription. Indeed, more than 160 chemical modifications have been found in various RNAs across all domains of life (1, 2). Recent studies using deep sequencing methods detected several species of modifications in eukaryotic mRNAs and noncoding RNAs in a transcriptome-wide manner (35). These findings raise the concept of “epitranscriptome” and highlight the importance of RNA modifications as regulatory elements in gene expression.

N6-methyladenosine (m6A) is an abundant modification in mRNAs and plays a key regulatory role in various biological events, including meiosis (6, 7), cell differentiation (810), neuronal function (11, 12), cancer proliferation (13, 14), circadian rhythm (15), sex determination (16), and chromosomal silencing (17). The biogenesis and dynamics of m6A have been studied extensively; the modification is introduced by the writer complex METTL3-METTL14-WTAP (18) and METTL16 (19), and it is demethylated by eraser proteins (ALKBH5 and FTO) (18). Internal m6As are decoded differently by several reader proteins, including YTH family proteins, hnRNP C, and eIF3 (20), thereby leading to diverse fates of mRNAs.

The 7-methylguanosine (m7G) cap is a characteristic 5′-terminal structure of eukaryotic mRNAs (Fig. 1A) (21). In the nucleus, this modification not only stabilizes mRNAs, but also promotes their transcription, splicing, polyadenylation, and nuclear export (22). In the cytoplasm, the m7G cap is required for translation of the majority of mRNAs. The m7G cap is introduced at the 5′ terminus of nascent mRNAs via 5′ to 5′ linkage at the initial stage of transcription, following the recruitment of the capping enzyme complex (RNGTT and RNMT) to the Ser5-phosphorylated C-terminal domain (CTD) of RNA polymerase II (RNAPII) in the early elongation complex (23). After the m7G cap formation, the 2′ hydroxyl group of the transcription start nucleotide is cotranscriptionally methylated by CMTR1 (24, 25) and that of the second nucleotide is methylated by CMTR2 (26) (Fig. 1A). The interferon-induced factor, IFIT1, recognizes hypomethylated viral RNAs and prevents their translation; the 2′-O-methylation at the first nucleotide of mRNAs antagonizes IFIT1, allowing them to escape the innate immune system (27, 28). In vertebrate mRNAs, if the transcription start nucleoside is adenosine, its N6 position is methylated to form N6, 2′-O-dimethyladenosine (m6Am) (Fig. 1A) (29, 30). Recent studies reported FTO-mediated demethylation of m6Am and its association with RNA metabolism (31, 32). The biogenesis and physiological importance of this cap-specific m6A modification have not been fully understood.

Fig. 1 Comprehensive analysis of mRNA 5′-terminal modification.

(A) Chemical structure of the 5′-terminus of a typical human mRNA. (B) Overview of the mass spectrometric analysis of 5′-capped fragments of human mRNAs. (C) Bubble chart showing mass spectrometric profiling of 5′-capped fragments (red circles) of mRNAs. Black circles indicate the noncapped fragments. The bubble sizes are in proportion to the square root of their intensity (red) or the log10 (black). (D) Extracted-ion chromatograms (XICs) for 5′-capped RNase T1-digested fragments of mRNAs. The sequence, mass/charge ratio (m/z), and charge state are shown on the right. n.d., not detected. (E and F) CID spectra of m7Gpppm6AmGp and m7GpppGmGp. The product ions were assigned as indicated. Asterisks represent 7-methylguanine dissociation.

Here we identified PCIF1 as a cap-specific adenosine N6-methyltransferase (CAPAM) responsible for N6-methylation of m6Am. CAPAM interacts with the Ser5-phosphorylated CTD of RNAPII, resulting in the formation of m6Am at the early stage of the transcription cycle. The crystal structures of CAPAM complexed with substrates revealed the cap-specific m6A formation mediated by a helical domain of CAPAM. A ribosome profiling experiment showed that m6Am promotes cap-dependent translation. Our results highlight CAPAM as an m6A writer for mRNAs.

Mass spectrometric analysis of m6Am in capped mRNAs

To accurately measure m6Am frequency, we first conducted a direct analysis of capped mRNAs using RNA mass spectrometry (Fig. 1B). Polyadenylated [poly(A)+] RNAs from human embryonic kidney 293T (HEK293T) cells were partially fragmented with Zn2+ and immunoprecipitated with an antibody against m7G to isolate the 5′-capped fragments. The fragments were digested with ribonuclease (RNase) T1 and analyzed by capillary liquid chromatography (LC)–nano-electrospray ionization (ESI)–mass spectrometry (RNA-MS) (33). We explored the m7G-capped dimers (m7GpppN1Gp) to pentamers (m7GpppN1N2N3N4Gp) containing 0 to 3 methyl groups and detected 52 species of 5′-capped fragments (Fig. 1CD and table S1). Each fragment was further probed and sequenced by collision-induced dissociation (CID) (Fig. 1, E and F). Using this approach, we detected 15 species of 5′-capped fragments bearing m6Am at the first position (table S1). Notably, m7Gpppm6AmGp was detected as a major species, and m7GpppAmGp was detected as a minor species in the mass chromatograms (Fig. 1D). Neither m7GpppAGp nor m7Gpppm6AGp was detected, suggesting that CMTR1-mediated 2′-O-methylation was efficiently introduced prior to m6A formation at the first position. The result showed that 5′-capped mRNAs contain 92% m6Am and 8% Am (Fig. 1D). Consistently, m7GpppGmGp, but not m7GpppGp, was efficiently detected (Fig. 1D). This finding suggested that m6Am modification is more dominant than previously estimated; 67% in HeLa S3 (30) and 48 to 75% in HEK293T cells (31).

To estimate the m7G immunoprecipitation (IP) experiment, we analyzed the flow-through fraction and detected no m7G-capped RNA fragments, suggesting that a large majority of the capped mRNAs were immunoprecipitated (fig. S1). Although this antibody has a specificity to N2, N2, 7-trimethylguanosine (TMG)–capped RNAs, we could not detect any 5′ termini of U-snRNAs (small nuclear RNAs) in the elution fraction (fig. S1), indicating that they were removed during the poly(A)+ preparation.

Identification of cap-specific adenosine N6-methyltransferase (CAPAM)

To identify the factor responsible for N6-methylation of m6Am at the first position of capped mRNAs, we used a reverse genetics approach coupled with RNA-MS (33). First, we chose 15 previously uncharacterized methyltransferase (MTase) genes that are not conserved in budding yeast, because m6Am is not present in fungi. Among them, we focused on PCIF1, which was originally identified as a factor that interacts with the phosphorylated CTD of RNAPII (34, 35) and has an uncharacterized domain similar to that of DNA m6A MTases M.EcoKI and M.TaqI (36). This led us to speculate that PCIF1 is a factor responsible for m6Am formation on the nascent transcript by association with RNAPII. Therefore, we knocked out this gene in HEK293T cells using the CRISPR-Cas9 system (Fig. 2A) and obtained two knockout (KO) cell lines (KO#1 and KO#2) in which both alleles had frameshift mutations. Western blotting confirmed the absence of endogenous PCIF1 in these KO cell lines (Fig. 2B). The 5′-capped fragments of mRNAs were isolated from PCIF1 KO#1 and subjected to RNA-MS (Fig. 2C). Notably, m7Gpppm6AmGp disappeared completely; instead, m7GpppAmGp was accumulated in PCIF1 KO#1. When the KO cells were rescued by plasmid-encoded PCIF1, m6Am was efficiently restored, demonstrating that PCIF1 is responsible for the conversion of Am to m6Am at the first position of capped mRNAs (Fig. 2C). Because PCIF1 has an uncharacterized m6A MTase domain with the conserved NPPF motif (fig. S5C) (36), we constructed a mutant of PCIF1 with Asn553→Ala (N553A) substitution in the NPPF motif. The N553A mutant did not efficiently restore m6Am in the PCIF KO cells (Fig. 2C), indicating the importance of Asn553 for m6Am formation in cells. To examine whether PCIF1 is also involved in the internal m6A formation, we analyzed the nucleoside composition of poly(A)+ RNAs using LC-MS and found no substantial decrease in internal m6A in PCIF1 KO#1 relative to wild-type (WT) cells (fig. S2). Hence, we renamed PCIF1 as cap-specific adenosine-N6-methyltransferase, or CAPAM.

Fig. 2 Identification of CAPAM.

(A) Schematic illustration of the human CAPAM locus and the indels (insertions and deletions) (red) introduced by the CRISPR-Cas9 system targeted by two sgRNAs in the CAPAM KO strains. The protospacer and protospacer adjacent motif (PAM) on the sense strand are underlined and boxed, respectively. The arrowheads in exons 3 and 4 indicate the target sites of sgB-2 (KO#1) and sg7-5 (KO#2), respectively. (B) Western blot analysis showing the absence of CAPAM in KO cells. (C) XICs for the 5′-capped RNase T1-digested fragments (with A at the first position) of mRNAs obtained from WT (left), KO#1 (left-middle), KO#1 rescued by plasmid-encoded human WT CAPAM (right-middle), and the N553A mutant (right). The fragment information is the same as that shown in Fig. 1D. n.d., not detected. (D) In vitro methylation assay of human CAPAM in the presence (+) or absence (-) of SAM. XICs of the 5′-capped RNase A–digested fragments without (upper panels) or with (bottom panels) m6A. n.d., not detected.

We next measured in vitro methylation activity of purified CAPAM protein toward a 5′-capped mRNA substrate and clearly observed m6A formation in the presence of both CAPAM and S-adenosylmethionine (SAM) (Fig. 2D). A small amount of m6A was detected even in the absence of SAM, indicating that some endogenous SAM was bound to the recombinant CAPAM. CID analysis of the methylated fragment confirmed that m6A indeed occurred at the first position of the capped fragment (fig. S3). Little activity of N6-methylation was observed in both G-capped mRNA (GpppA) and noncapped mRNA (pppA) (Fig. 3A), suggesting that CAPAM specifically recognizes the m7G cap structure. We then compared m6A-forming efficiency of the capped mRNA substrate with or without 2′-O-methylation at the first position and found that N6-methylation of m7GpppAm is faster than that of m7GpppA (Fig. 3A and fig. S4A). The Km (Michaelis constant) values of m7GpppAm and m7GpppA were determined to be 3.5 and 28 μM, respectively (Fig. 3B), demonstrating that CAPAM preferentially recognizes the capped mRNAs with Am modification. Consistently, the efficient N6-methylation of m7GpppAm was reported in a previous study using HeLa cell lysate (30). Our results indicated the hierarchical formation of m7Gpppm6Am. To examine the substrate specificity, we compared m6A methylation activities of CAPAM toward 10 RNA substrates with different sequences at positions 2 and 3 (fig. S4B). CAPAM showed some preference for the 5′-terminal sequence of mRNAs, but did not have a strong sequence specificity. In addition, we measured the activities of CAPAM toward a series of capped RNA substrates with different lengths. CAPAM did not efficiently introduce m6A on the 3- to 5-nucleotide (mer) substrates, whereas CAPAM methylated the 6-mer substrate as efficiently as the 110-mer substrate (Fig. 3C). Thus, 6-mer is the minimum substrate for the CAPAM-mediated m6A formation.

Fig. 3 Biochemical characterization of CAPAM.

(A) In vitro methylation efficiency of mRNA substrates (110 mer) bearing different 5′-terminal structures (m7GpppAm-, m7GpppA-, GpppA-, and pppA-) by human CAPAM. The rate of m6A(m) formation (percentage) was measured as the mean ± SD (n = 4 independent experiments) of the molar ratio of the incorporated methyl group calculated from the 14C radioactivity to the substrate RNA at each time point. **P < 1.0 × 10−6 by Student’s t test. (B) Kinetic analysis of in vitro methylation of 5′-capped mRNA substrates (110 mer) with Am (red line) or A (black line) at the first nucleotide by human CAPAM. The initial velocity (Vi) was calculated as the mean ± SD (n = 3 independent experiments). The Km and Vmax values were calculated using Prism 7. (C) In vitro methylation of 5′-capped mRNA substrates with different lengths, as indicated. XICs for the RNA fragments with A (upper panels) or m6A (bottom panels). (D) The human CAPAM WW domain binds specifically to Ser5-phosphorylated CTD. Glutathione S-transferase–tagged WW domains derived from CAPAM (left) and Pin1 (right) were pulled down with four different CTD peptides (heptad repeats) immobilized to streptavidin beads. U: unphosphorylated peptide; pS2: Ser2-phosphorylated; pS5: Ser5-phosphorylated; (-): without peptide. (E) Immunoprecipitation of CAPAM from whole-cell extracts from HeLa cells using CAPAM-specific antibody and normal rabbit immunoglobulin G (control), followed by immunoblotting with antibodies against the indicated proteins.

CAPAM has a WW domain at its N-terminal region (fig. S5A) (34). This domain is homologous to the Pin1 WW domain, which binds to the phosphorylated Ser-Pro motif (37). Therefore, we examined the specificity of the CAPAM WW domain for RNAPII CTD peptides with phosphorylation at Ser2 or Ser5 (Fig. 3D). The CAPAM WW domain interacted specifically with the Ser5-phosphorylated CTD, whereas the Pin 1 WW domain interacted with both peptides (Fig. 3D). We next examined a specific interaction between CAPAM and RNAPII with the Ser5–phosphorylated CTD in HeLa cells. CAPAM and RNAPII with the Ser5-phosphorylated CTD were coimmunoprecipitated (Fig. 3E), indicating that CAPAM is recruited to the early elongation complex of RNAPII.

Molecular basis of CAPAM-mediated N6-methylation

To elucidate the mechanism of N6-methylation by CAPAM, we determined the crystal structure of the CAPAM complex containing m7G-capped RNA and a cofactor analog, S-adenosylhomocysteine (SAH). For crystallization, we constructed human and zebrafish CAPAMs (hCAPAM and zCAPAM, respectively) with truncation of the N-terminal WW domain and the C-terminal part (Fig. 4A and fig. S5A), because the WW domain did not affect the in vitro MTase activity of CAPAM (fig. S5B). We determined the crystal structures of hCAPAM (Apo and SAH-bound forms) and zCAPAM (Apo, SAH-bound, m7GpppA/SAH-bound, and m7GpppAmG/SAH-bound forms) at 1.8 to 2.9 Å resolutions (fig. S6 and table S2). Because these six overall structures are similar [root mean square deviation (RMSD) of <1.5 Å for aligned Cα atoms], we hereafter describe the structure of m7GpppA/SAH-bound zCAPAM unless otherwise stated.

Fig. 4 Crystal structure of CAPAM.

(A) Domain organization of zCAPAM. (B) Overall structure of zCAPAM in complex with m7GpppA and SAH. (C and D) Recognition of SAH and the m7G cap. Hydrogen bonds are shown as dashed lines. (E) Putative binding site of the target Am nucleotide. (F) In vitro MTase activities of zCAPAM mutants. The substrate used for this assay was 5′-capped mRNA (110 mer). The relative rate of m6Am formation was calculated as the mean ± SD (n = 4 independent experiments). **P < 1.0 × 10−6 by Student’s t test.

The core region of CAPAM contains the helical and MTase domains (Fig. 4, A and B). The helical domain consists of three-helix bundles (α1-α6-α8 and α4-α5-α6), a four-helix bundle (α1-α2-α3-α6), and β sheets (β1-β2 and β3-β4-β5) (Fig. 4B and fig. S7). A Dali search (38) detected no structural similarities between the helical domain and any known protein structure. The MTase domain adopts a canonical Rossmann fold containing a conserved catalytic motif (residues 558 to 561) (Fig. 4B and fig. S7). SAH is bound to a catalytic pocket of CAPAM in a manner similar to that of class I MTases, such as DNA m6A MTase M.TaqI (Fig. 4C and figs. S8A and S9A). The m7G cap is bound to the “m7G site” located between the helical and MTase domains (Fig. 4, B and D, and fig. S8B). The ribose and guanine moieties of m7G are recognized by Arg239/Arg269/Glu563 and Glu563/Trp593/Pro596/Pro597, respectively (Fig. 4D). A mutational assay confirmed the importance of Arg239/Arg269/Glu563 for m7G cap recognition (Fig. 4F). The m7G cap, but not the target adenosine adjacent to the m7G cap, was visible in the electron density map (fig. S8B), suggesting that the target nucleotide is disordered in the crystal structure. Based on the reported structure of M.TaqI MTase bound to a DNA substrate (39), we modeled a 2′-O-methyladenosine (Am) at the active site of CAPAM (fig. S9, B and C). Although M.TaqI is bound to a double-stranded DNA, the active site of M.TaqI recognizes the adenine base flipped out from the double helix (fig. S9, A and C). In addition, the MTase domain of CAPAM is highly homologous to that of M.TaqI, and the characteristic (DNSH)PP(YFW) motif is conserved in all m6A MTases (36). Thus, we modeled the target Am in the active site of CAPAM, based on the M.TaqI complex structure. The model suggested that the adenine moiety of Am forms hydrogen-bonding interactions with Asn558/Pro559/Phe561 and π-stacking interactions with Phe561/Phe619, and the ribose moiety of Am forms van der Waals interactions with His621. A mutation study supported the importance of these residues for m6A formation (Fig. 4, E and F, and fig. S9B). Notably, the helical domain forms a positively charged groove (fig. S10, A and B) and is highly conserved in animals (fig. S10C), suggesting that the RNA strand following the 5′ cap binds to this positive groove. We mutated six basic residues in the helical domain of CAPAM and found that five of them reduced m6A-forming activities significantly (fig. S10D). The result suggests that the basic helical domain serves as the RNA-binding surface. Overall, our structural and mutagenesis data provide mechanistic insights into m7G-capped RNA recognition and m6A formation by CAPAM.

Physiological role of CAPAM

To investigate the biological role of CAPAM, we explored the phenotypic features of CAPAM KO cell lines. Although CAPAM KO cells grew well and showed a growth rate similar to that of WT cells under normal culture conditions, they showed defective growth under oxidative stress conditions (Fig. 5A). This finding indicates that CAPAM is involved in the cellular response to oxidative stress.

Fig. 5 Physiological importance and translation regulation by m6Am.

(A) Growth curves of WT, KO#1, and KO#2 cell lines cultured in normal medium (left) or medium containing 30 μM H2O2 (right). Fluorescence was calculated as the mean ± SD (n = 5 independent biological replicates). **P < 1.0 × 10−7 by Student’s t test. (B) Cumulative plot of the fold-changes in mRNA expression in KO#1 versus WT cells. The mRNAs were classified into five groups based on the first nucleotides (m6Am, Am, Gm, Cm, and Um), and fold-changes in the steady-state levels of the groups were determined. Each box in the inset panel shows the first quartile, median, and third quartile, and the whiskers represent the 1.5 × interquartile ranges. *P < 1.0 × 10−4 versus m6Am and Am/Gm/Cm/Um by Wilcoxon’s rank sum test. (C) Cumulative plot of the fold-change in TE in KO#1 versus WT cells. TE was calculated as the ratio of the normalized read counts obtained from RNA-seq and ribosome profiling. The mRNAs were classified as described for (B). ***P < 1.0 × 10−11 versus m6Am and Am/Gm/Cm/Um by Wilcoxon’s rank sum test. (D) Differential TEs between WT and KO#1. The log2 fold-change of each TE was plotted against the normalized read counts from RNA-seq (plots with P < 0.05 in red). (E) Stacked bar chart of the number of total and down-regulated mRNAs. The number in each box represents the number of classified mRNAs with different first nucleotides. *P < 1.0 × 10−3 versus total m6Am start mRNAs and down-regulated m6Am start mRNAs by a binomial test.

Next, we performed RNA sequencing (RNA-seq) to compare the transcriptomes of CAPAM KO and WT HEK293T cells. A comparison of the steady-state levels of all detected transcripts between CAPAM KO and WT cells revealed 25 up-regulated mRNAs and 36 down-regulated mRNAs [false discovery rate (FDR) <0.01] upon KO of CAPAM (fig. S11). A previous study reported that demethylation of m6Am by overexpression of the eraser protein FTO results in significant destabilization of a subset of mRNAs starting with m6Am (31). To confirm this result, we classified mRNAs into five groups according to their first nucleotides (m6Am, Am, Gm, Cm, and Um) based on the published miCLIP and CAGE data (31, 40), and calculated fold-changes in their steady-state levels upon KO of CAPAM (Fig. 5B and fig. S12A). In contrast to the previously reported effect of FTO overexpression (31), we observed a slight increase in the level of mRNAs with m6Am upon KO of CAPAM. This result does not support the proposed function of m6Am to stabilize A-starting capped mRNAs. Consistently, a recent report showed that FTO mainly affects the expression levels of mRNA containing internal m6A rather than mRNA starting with m6Am (32).

Translational regulation by m6Am

Given that the 5′-cap structure plays a critical role in translation initiation, we examined whether the m6Am modification is involved in protein synthesis. To this end, we performed ribosome profiling and RNA-seq to compare translation efficiency (TE) profiles of CAPAM KO and WT HEK293T cells. To reveal the effect of m6Am modification on translation, we classified mRNAs as described above (m6Am, Am, Gm, Cm, and Um). We observed a significant decrease in translation of mRNAs with m6Am upon KO of CAPAM (Fig. 5C and fig. S12B). Indeed, we found a strong correlation between the TE change of m6Am-starting transcripts in the two CAPAM KO strains versus WT strains with high correlation coefficient (R = 0.81) (fig. S12E), suggesting that N6-methylation of m6Am up-regulates the translation. A comparison of the translation efficiency of all detected transcripts between CAPAM KO and WT cells revealed 3 up-regulated mRNAs and 75 down-regulated mRNAs (FDR <0.05) upon KO of CAPAM (Fig. 5D and fig. S12, C and D). The down-regulated genes in the CAPAM KO cells, but not up-regulated genes, showed significant enrichment of m6Am-starting transcripts (Fig. 5E and table S3). A Gene Ontology enrichment analysis revealed that the down-regulated genes are associated with mRNA transport and metabolic processes, and with translation (fig. S13, A and B). We carried out a metagene analysis of ribosomal protected fragments (RPFs) around start codons for m6Am-starting transcripts and other transcripts (fig. S14), but found no significant difference between WT and CAPAM KO cells, indicating that m6Am modification does not influence ribosome distribution in each mRNA. We also analyzed translation efficiencies of upstream open reading frames (uORFs) and found no obvious effect of m6Am on the uORF expression upon KO of CAPAM (fig. S15).

The eukaryotic translation initiation factor eIF4E directly recognizes the cap structure of mRNAs and is essential for cap-dependent translation initiation (41). In addition, the binding affinity of eIF4E for the cap structure is modulated by the first nucleotide of capped mRNAs (42). Hence, we examined the binding affinity of eIF4E for capped mRNAs with m7Gpppm6Am or m7GpppAm using an electrophoretic mobility shift assay (EMSA). We observed no significant effect on the binding affinity of eIF4E for mRNAs with or without N6-methylation of m7Gpppm6Am (fig. S16), suggesting that other factors and mechanisms independent of eIF4E are involved in the m6Am-mediated translational regulation.

Discussion

Using direct RNA-MS analysis of capped mRNAs from HEK293T cells, we found that 92% of A-starting mRNAs have the m6Am modification, and the remainder have the Am modification, suggesting that the m6Am frequencies observed in previous studies were underestimated. Because we did not detect m7GpppAGp and m7Gpppm6AGp in this study, it is likely that Am formation by CMTR1 precedes m6A formation by CAPAM. This finding is consistent with the observation that CAPAM preferentially N6-methylates Am rather than unmodified A (Fig. 3, A and B) (30). In our model structure, the target Am can be recognized in a pocket formed by N558, F561, F619, and H621 (Fig. 4E and fig. S9B). Hydrophobic interaction conferred by 2′-O-methyl group of Am might be involved in the strong binding to this pocket. Otherwise, C3′ endo ribose puckering of Am conferred by 2′-O-methylation might be a preferable conformation recognized by CAPAM. These speculations show how the Am is specifically recognized by the enzyme. We also showed that the CAPAM WW domain binds specifically to the Ser5-phosphorylated CTD, indicating that CAPAM is recruited to the early elongation complex of RNAPII and that N6-methylation of m6Am takes place cotranscriptionally. Our results suggested a hierarchical formation of m7Gpppm6Am—pppA, GpppA, m7GpppA, m7GpppAm, and m7Gpppm6Am in the nascent transcript at the early stage of transcription elongation by RNAPII.

Structural comparison with other m6A writers, METTL3-METTL14 (43) and METTL16 (44), revealed their diverse mechanisms of RNA substrate recognition and m6A modification. Whereas these m6A writers share the common MTase domain with a Rossmann fold, they have additional domains or subunits that define the RNA substrate specificity (fig. S17). In CAPAM, the helical domain forms a positively charged groove that can bind 5′-capped single-stranded RNA. The METTL3-METTL14 complex has a positively charged surface near the active site, which may bind single-stranded RNA containing the consensus motif. In METTL16, the N-terminal domain forms a wide groove that accommodates its structured RNA substrates. These distinct structural features of the m6A writers likely contribute to their functional divergence in the RNA recognition and m6A modification.

Our finding that CAPAM KO cells grew well under normal conditions suggests that N6-methylation of m6Am is not required for cell viability. Nonetheless, CAPAM KO cells showed strong sensitivity to H2O2 treatment, indicating that m6Am plays a regulatory role in gene expression in response to oxidative stress. Among m6Am-starting transcripts, translation efficiency of the SOD1 mRNA, which encodes superoxide dismutase, was significantly decreased upon KO of CAPAM (table S3). This finding might partly explain why CAPAM KO strains are sensitive to oxidative stress. An RNA interference–based genetic screen identified CAPAM/PCIF1 as a putative tumor suppressor in a bladder cancer model (45), indicating that it might be involved in cell proliferation under certain conditions. Further studies seem necessary to unveil the physiological role of this gene. Here, we found that N6-methylation of m6Am has an ability to up-regulate cap-dependent translation; however, N6-methylation of m6Am did not modulate binding of eIF4E to the cap structure. Other cap-binding proteins might be involved in this process.

Methods summary

5′-capped RNA fragments of poly(A)+ RNAs from HEK293T cells were enriched by antibodies against m7G cap and subjected to capillary liquid chromatography coupled with nano-electron spray ionization mass spectrometry (33) to directly detect m7G-capped oligomers with m6Am at the transcription start site. CAPAM was knocked out in HEK293T cells by the CRISPR-Cas9 system using two single guide RNAs (sgRNAs) with different target sites. For in vitro reconstitution of m6A, His-SUMO–tagged CAPAM was recombinantly expressed in E. coli, and substrate RNAs were transcribed using T7 RNA polymerase in the presence of cap analogs. Polyclonal antibody against CAPAM was affinity purified from rabbit serum. CAPAM was crystallized with or without SAH and a cap analog, and x-ray diffraction data were collected on beamlines at SPring-8 and Swiss Light Source. The cDNA libraries for RNA-seq analysis were prepared according to the Illumina Truseq protocol, and ribosome profiling was performed as described (46).

Supplementary Materials

www.sciencemag.org/content/363/6423/eaav0080/suppl/DC1

Materials and Methods

Figs. S1 to S17

Tables S1 to S4

References (4766)

References and Notes

Acknowledgments: We are grateful to the lab members, especially to Y. Sakaguchi and S. Ito, for technical assistance. We thank C. K. Ho and S. Shuman for providing materials. Funding: This work was supported by Grants-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (MEXT); Japan Society for the Promotion of Science (JSPS) (26113003, 26220205, and 18H05272 to Ts.S., 21115003 and 18K19141 to Ta.S., 17K07282 to Y.H., JP17H05679 and JP17H04998 to S.I.), and the Basic Science and Platform Technology Program for Innovative Biological Medicine from the Japan Agency for Medical Research and Development (AMED) to O.N. This work used the Vincent J. Coates Genomics Sequencing Laboratory at the University of California Berkeley, supported by NIH S10 OD018174 Instrumentation Grant, and Bioinformatics Analysis Environment Service on RIKEN Cloud at RIKEN Advanced Center for Computing and Communications. Author contributions: S.A. performed all biochemical and genetic works assisted by Ta.S. S.H. performed structural studies assisted by H.N. and R.I. Y.S. performed ribosome profiling and data analysis supported by S.I. Y.H. and A.S. performed the IP experiment. S.A., S.H., H.N. and Ts.S. wrote this paper. All authors discussed the results and revised this paper. Ts.S. and O.N. designed and supervised all the work. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Ribosome profiling and RNA-seq data are deposited in NCBI (accession numbers: GSE122071). The coordinates of CAPAM structures are deposited in the Protein Data Bank (accession numbers 6IRV, 6IRW, 6IRX, 6IRY, 6IRZ, 6IS0).
View Abstract

Stay Connected to Science

Subjects

Navigate This Article