Dear Editor,

Protein coding mRNAs and non-coding RNAs are extensively modified by chemical modifications,1 including methylation of the N6 position of adenosine (m6A).2 While mRNA internal m6A methylation is mediated by METTL3 and METTL14, which are members of the MT-A70 family,2 N6-methylation of cap-adjacent N6,2′-O-dimethyladenosine (m6Am) is catalyzed by PCIF1.3,4,5,6 Non-coding RNAs are also m6Am methylated internally,7 but the corresponding enzyme remains unknown. Here we provide evidence that METTL4 mediates internal m6Am methylation of U2 small nuclear RNA (snRNA).

METTL4, like its paralogs METTL3/14, is conserved from yeast to human (Fig. 1a).8 C. elegans and mouse METTL4 homologs have been shown to regulate DNA 6mA methylation in vivo,9,10,11 but are only considered as candidate DNA methyltransferases because of lack of biochemical evidence demonstrating that METTL4 has intrinsic DNA 6mA methyltransferase activity. To further investigate METTL4, we first characterized the localization of METTL4 and found that METTL4 carries a well-conserved, putative nuclear localization signal (NLS) (Supplementary information, Fig. S1a). Consistently, we found mainly nuclear localization of an exogenously introduced METTL4 (Supplementary information, Fig. S1b). Next, we generated METTL4 knockout (KO) 293T cell lines (Supplementary information, Fig. S1c) and found that METTL4 KO significantly reduced internal m6Am levels in the total RNA (Fig. 1b). Importantly, the internal m6Am can be readily rescued by wild-type but not the catalytically compromised METTL4 (METTL4APPA) (Fig. 1c; Supplementary information, Fig. S1d and e). In contrast, METTL4 KO or re-expression of an exogenously introduced METTL4 has no overt impact on 6mA level in the genomic DNA (Supplementary information, Fig. S2a). Furthermore, we failed to identify appreciable levels of 6mA in mitochondrial DNA or m6Am in mitochondrial RNA in 293T cells, respectively (Supplementary information, Fig. S2b and c).

Fig. 1: METTL4 is an internal m6Am methyltransferase regulating RNA splicing.
figure 1

a Phylogenetic analysis of MT-A70 family proteins indicating that METTL4 is likely an evolutionarily conserved N6-methylation-related enzyme from yeast to human. b Changes of distinct types of RNA modification in total RNA upon METTL4 KO in 293T cells, measured by quantitative LC–MS/MS. KO#3 and KO#6 denote two different METTL4 knockout clones. Data were analyzed by Student’s t-test. Error bars indicate mean ± SD (n = 3). c LC–MS/MS chromatograms of total A and m6Am in total RNA extracted from indicated cell lines. d m6Am level changes were determined by m6A Northwestern blot in WT, KO, and rescued cell lines. RNU6B was included as an internal loading control and a small RNA oligo containing m6A was used as a positive control. The black arrow indicates the migrating position of the most significantly decreased RNA(s). e Pie chart of relative percentage of each identified RNA in the gel region showing a decreased m6Am signal. f LC–MS/MS results showing that internal m6Am modification was only readily detected in U2 but not in 5.8S rRNA, SNORA73 or U1. n.d., not detected. g The activity of recombinant METTL4 towards different RNA substrates in vitro. h Michaelis–Menten kinetics of recombinant METTL4 was determined using the Am30 probe as a substrate in vitro. i Genome-wide alternative splicing changes upon METTL4 KO were identified by rMATS. Red indicates upregulation while blue indicates downregulation. j A working model for METTL4 methylating U2 snRNA. METTL4 is a novel internal m6Am methyltransferase, which can mediate N6-methylation of m6Am30 on U2 snRNA within an AAG motif. Loss of METTL4 induces global splicing changes. m7G: 7-methylguanosine, TMG: 2,2,7-trimethylguanosine.

Previous studies demonstrated that PCIF1 is responsible for N6-methylation of the cap-adjacent m6Am, which is only sensitive to nuclease P1 cleavage when pre-treated with the RNA 5′ pyrophosphohydrolase, RppH.3,4,5,6 We treated total RNA with or without RppH before nuclease P1 digestion and monitored the changes of m6Am levels (Supplementary information, Fig. S3a). As shown in Supplementary information, Fig. S3b and c, METTL4 depletion induces a drastic decrease of internal m6Am but does not affect the level of the cap-adjacent m6Am. Conversely, KO of the cap-adjacent m6Am methyltransferase, PCIF1, has no impact on the level of the internal m6Am. These findings suggest that METTL4 and PCIF1 represent distinct m6Am methyltransferases targeting different RNA types/regions.

We next set out to identify the m6Am substrate(s) for METTL4. Interestingly, m6Am signal is significantly enriched in RNAs less than 200 nucleotides (nt) in length but not in RNAs longer than 200 nt, including polyA(+) RNA and ribosomal RNAs (18S and 28S) (Supplementary information, Fig. S4a). To further identify the type of small RNA that is m6Am methylated, we separated small RNAs by electrophoresis and performed Northwestern with m6A-specific antibodies. We found that a 150-nt band is absent upon METTL4 KO, and importantly this signal can be restored by re-expression of an exogenously introduced METTL4 (Fig. 1d, band indicated by an arrow), suggesting that m6Am methylation of the yet-to-be-identified RNA species is dependent on METTL4. Consistently, LC–MS/MS analysis showed that RNAs in the length range of 125–225 nt are m6Am but not m6A methylated (Supplementary information, Fig. S4b and c). As a control, a prominent band around 70 nt is unaffected by METTL4 depletion (Fig. 1d, marked by an asterisk). Together, these findings support the hypothesis that METTL4 can mediate internal m6Am methylation of certain small RNAs.

To determine the identities of the m6Am-methylated RNA species, we carried out RNA-seq of the RNAs in the size range of 125–225 nt purified from the gel slice, and found that 5.8S rRNA, SNORA73 snoRNA and U2 snRNA were ranked as the top three most abundant RNAs (Fig. 1e). In order to determine which RNA(s) carries the internal m6Am, we purified 5.8S rRNA, SNORA73 snoRNA, U2 as well as U1 snRNAs (which was suggested to be m6Am methylated12) with biotin-labeled synthetic complementary DNA oligos, respectively, and quantified m6Am carried by these four different RNAs (detailed procedure in Supplementary information, Fig. S4d). As shown in Fig. 1f, a considerable amount of m6Am was detected only in U2 snRNA, and there is essentially no m6Am signal detectable in the 5.8S rRNA, SNORA73, or U1 snRNA, suggesting that U2 snRNA is modified by internal m6Am and that METTL4 may be the responsible enzyme in vivo. This finding is consistent with a four-decades-old finding that adenosine of U2 snRNA at position 30 is possibly decorated by m6Am modification,13 and a more recent transcriptomic single-nucleotide resolution mapping of m6Am suggesting that U2 snRNA may carry internal m6Am modification.7

To interrogate the hypothesis that METTL4 may mediate internal m6Am methylation on U2 snRNA, we investigated the enzymatic activity of the recombinant human METTL4 by carrying out in vitro assays using a 35-nt long RNA of human U2 snRNA as a substrate (nt 6–40). Recombinant METTL4 protein was incubated with multiple RNA substrates (detailed in Supplementary information, Materials and Methods) in the presence of stable isotope-labeled S-adenosylmethionine (d3-SAM) (Supplementary information, Fig. S5a and b). The activity of METTL4 for the U2 snRNA Am30 probe was approximately 100-fold higher than that for the A30 probe, suggesting that 2′-O-methylation is necessary for METTL4 to mediate methylation on the N6 position (compare probe A30 and Am30 in Fig. 1g; Supplementary information, Fig. S5b). Interestingly, this methylation event is drastically reduced when the neighboring A29 or G31 was changed to other bases (refer to probes AAmA and GAmG in Fig. 1g). In addition, only a low level of methylation was observed when DNA oligos, Cap-Am, or GACU RNA oligos (optimal substrates of METTL3/14) were used as substrates (Cap-Am and GACU probes in Fig. 1g; DNA probe in Supplementary information, Fig. S5c). Consistent with the in vivo results, this enzymatic activity is intrinsic to METTL4 as catalytic mutation abrogated its ability to methylate the U2 snRNA probe (Supplementary information, Fig. S5d). We next determined the Km (Michaelis constant) value of recombinant METTL4 methylating probe Am30 to be around 0.43 μM (Fig. 1h). These results suggest that U2 snRNA is a substrate of METTL4, which mediates U2 m6Am30 methylation, with a preferred sequence motif of AAG, and a requirement for pre-deposited 2′-O-methylation.

U2 snRNA is an essential component of the major spliceosome and is involved in branch point selection and catalysis during pre-mRNA splicing.14 Interestingly, m6Am30 is localized immediately upstream of the branch point recognition sequence (BPRS) (Supplementary information, Fig. S6a), suggesting that this modification may regulate splicing activity of the spliceosomal complex.14 To test this possibility, we performed RNA-seq and investigated splicing changes in WT and METTL4 KO cells using rMATS software. A total of 1402 significantly altered splicing events (delta PSI: 637 up-regulated vs 765 down-regulated) were identified after comparing the transcriptomes of WT and METTL4 KO cells (Fig. 1i; Supplementary information, Fig. S6b) and validated by RT-PCR for PPP3CB, TTC28, and TJP1 genes (Supplementary information, Fig. S6c and d), demonstrating that METTL4 is likely involved in splicing regulation.

Taken together, we demonstrated that the MT-A70 family member METTL4 is a novel internal m6Am methyltransferase, which mediates N6-methylation of Am30 on U2 snRNA in the context of an AAG motif in vivo and in vitro. Our findings support the notion that most if not all RNA internal m6Am modification is mediated by METTL4, but whether METTL4 has additional substrates besides U2 snRNA remains to be explored in the future. Additionally, analysis of RNA alternative splicing in WT and METTL4 KO cells shows that METTL4 loss impacts splicing (Fig. 1j). Our genetic rescue experiments suggested that the enzymatic activity of METTL4 and therefore likely U2 m6Am methylation is important for splicing regulation. However, we cannot formally exclude the possibility that METTL4 exerts its splicing regulation via regulation of different substrate(s) other than U2 snRNA, or in an enzymatic activity independent manner. Future experiments including transciptome-wide mapping of internal m6Am and CLIP-seq of METTL4 are necessary to address these questions further. Additionally, it will be interesting to investigate the enzymatic properties of METTL4 in different species as 2′-O-methylation, which is necessary for human METTL4, is absent in lower species such as yeast and fruitfly. Lastly, though we did not observe any significant defects (e.g., cell proliferation) of 293T cells when METTL4 is abrogated, HepG2 cell growth is compromised upon METTL4 knockdown (Supplementary information, Fig. S6e and f), suggesting that the role of METTL4 may be context-dependent. Interestingly, a previous study demonstrated that Mettl4-deficient mice display craniofacial dysmorphism and abnormalities in the immune system.10 Our discovery of METTL4 as a novel RNA m6Am methyltransferase not only identifies a potential epitranscriptomic mechanism to regulate RNA alternative splicing, but also provides a basis for future mechanistic investigations of biological functions of METTL4 in embryonic development and human diseases.