Methylation of Structured RNA by the m6A Writer METTL16 Is Essential for Mouse Embryonic Development

Abstract

Internal modification of RNAs with N⁶-methyladenosine (m⁶A) is a highly conserved means of gene expression control. While the METTL3/METTL14 heterodimer adds this mark on thousands of transcripts in a single-stranded context, the substrate requirements and physiological roles of the second m⁶A writer METTL16 remain unknown. Here we describe the crystal structure of human METTL16 to reveal a methyltransferase domain furnished with an extra N-terminal module, which together form a deep-cut groove that is essential for RNA binding. When presented with a random pool of RNAs, METTL16 selects for methylation-structured RNAs where the critical adenosine is present in a bulge. Mouse 16-cell embryos lacking Mettl16 display reduced mRNA levels of its methylation target, the SAM synthetase Mat2a. The consequence is massive transcriptome dysregulation in ∼64-cell blastocysts that are unfit for further development. This highlights the role of an m⁶A RNA methyltransferase in facilitating early development via regulation of SAM availability.

Keywords: METTL16; Mat2a; SAM availability; SAM synthetase; U6 snRNA; blastocysts; crystal structure; m(6)A; morula; splicing.

Graphical abstract

Figure 1

Structure of Human METTL16 Reveals an N-Terminal Module Essential for Activity (A) In vitro methylation assays of indicated full-length (FL) human m⁶A methyltransferases with ¹⁴C-SAM and different RNA substrates (right). Predicted structure of a short hairpin RNA (RNA6) derived from the longer MAT2A hairpin 1 (Pendleton et al., 2017) and its mutant (RNA6-mut) with A→U mutation of the methylated adenosine are shown. The MET1 RNA has the consensus site for methylation by the METTL3+METTL14 complex, while the MET2 RNA has a point mutation (C→U) of a conserved residue in the methylation consensus site (see Table S1). Single-stranded RNA markers (length in nucleotides, nt) are ³²P-end-labeled. See also Figure S1D. (B) Domain architecture of human METTL16. RBD, RNA-binding domain (1–78 aa); MTase, methyltransferase domain; VCR, vertebrate conserved region. Boundaries of the two protein constructs crystallized in this study are indicated (in green). The ΔN version has an N-terminal deletion. (C) Schematic view of the MTase domain. Cylinders represent α helices, and arrows represent β strands. Regions shaded in red (α1-2 and β1-2) are seen only in the METTL16-core structure and together with α3 form a separate N-terminal module. (D) Model of the METTL16-ΔN construct (PDB 6GFK). Two-sulfate (SO₄^2-) ions visualized in the crystal structure are circled. A disordered loop between α8 and β6 is connected by a dotted line. SAH, S-adenosyl homocysteine. (E) Model of the METTL16-core construct (PDB 6GFN). The additional regions at the N terminus seen in this structure are shown in red. See also Figure S1G. (F) A zoom of the catalytic pocket in the METTL16-ΔN structure showing coordination of SAH. Catalytic residues N184, P185, P186, and F187 and position of a sulfate (SO₄^2-) ion are indicated. See also Figure S2A. (G) In vitro methylation assays showing that METTL16-ΔN protein is inactive. The METTL16-Core protein was used as untagged or tagged (SUMO) versions. See also Figure S2C. Quality of proteins used is shown on the right. Protein molecular weight markers (in kilo Daltons, kDa) are indicated.

Figure 2

The N-Terminal Module of Human METTL16 Is Required for Substrate RNA Binding (A) Domain architecture of human METTL16. UV crosslinking assay (triplicate reactions) showing RNA-protein crosslinks (X-link) between human full-length (FL) METTL16 and MAT2A hairpin 1 RNA. See also Figure S2C for in vitro methylation with the same proteins and RNA. (B) Overview of the METTL16-core MTase domain. Key positively charged residues that create the RNA-binding groove are indicated. Note that residues K26 and K31 when mutated (MUT2) do not affect activity. SAH, S-adenosyl homocysteine; SO₄^2-, position of a sulfate ion as seen in METTL16-ΔN is shown. The disordered loop with catalytic residues is shown as a dotted line. (C) Surface charge representation of the METTL16-core MTase domain showing a positively charged (blue) groove (outlined) leading from the N terminus to the catalytic pocket. (D) Cartoon showing the N-terminal 40 amino acids of human METTL16, with the highlighted positively charged residues that were mutated (red, with asterisks). Gel shows the in vitro methylation assay with wild-type (WT) or indicated point mutant METTL16-core proteins. Quality of recombinant proteins used is shown below the gel. RNA7 was used as an RNA substrate (see Table S1). Single-stranded RNA markers (length in nucleotides, nt) are ³²P-end-labeled. See also Figure S2D. (E) In vitro methylation assay with RNAs indicated and mutants carrying multiple point mutations on the N-terminal module (see D). See also Figure S2E for additional mutations within the RNA binding groove. (F) UV crosslinking assay with METTL16 proteins indicated and ³²P-end-labeled RNA6. The positions of the free RNA and RNA-protein crosslinks (X-link) are shown. Control binding reactions are carried out without any protein (RNA alone) or with bovine serum albumin (BSA). (G) Sequence alignment of METTL16 orthologs showing the catalytic residues and disordered loop region. See also Figure S1A. Deletions and mutations introduced into the loop in the context of the METTL16-core construct are indicated. In vitro methylation assay with indicated proteins and RNAs is shown below. Quality of proteins used is shown in Figure S2F.

Figure 3

METTL16 Requires Structured RNA for m⁶A Methylation (A) In vitro methylation assay with RNAs carrying truncations of the stem region. See also Figure S3A. The predicted structures of RNAs used are shown below. (B) In vitro methylation assay. (C) In vitro methylation assay with RNAs carrying mutations in the nonamer consensus sequence (shown below). Short and long exposures of the gel are shown. (D) Scheme of an in vitro methylation experiment using a library of randomized (N = any of the four nucleotides) RNA oligos. (E) For each sequence, we predicted the minimum free energy (MFE) secondary structure using RNAfold (STAR Methods). A model structure is shown in dot bracket notation. (F) The representation of individual structures (corresponding to unique dot bracket notation) was compared between m⁶A-IP samples and input samples. Top ten IP-enriched structures are shown. The 15th position adenosine (A) that is in the consensus nonamer sequence is highlighted. (G) Frequency of structures forming stem, loop and other selected features at individual positions is shown. The IP-enriched structures have increased frequency of 15A (red arrowhead) in a bulge and surrounded by double-stranded regions (stems), pointing to specific structural requirements of RNA substrates for METTL16. (H) Structures enriched or depleted in m⁶A IP were compared to those that do not show such difference (between IP and input). While the IP-enriched structures have higher proportion of 15A (red arrowhead) forming a bulge or lying between two stems, the IP-depleted structures show the opposite trend, with lack of structures with 15A in a bulge or in between two stems. (I) The barplot shows the proportion of structural features in which the 15A was found. Note the high proportion of structures with 15A in the bulge and between two stems, among the m⁶A IP-enriched structures. See also Figure S3. (J) In vitro methylation assay with METTL16-core protein and RNAs (RNA21 and RNA23) selected from randomized library methylation experiment (D). This confirms the specific methylation of 15A which is in a 1 nt bulge (in RNA21, but not in mutant RNA22).

Figure 4

Reduced Mat2a mRNA Levels and Embryonic Lethality around Implantation Stage in the Mettl16 Mutant Mice (A) Generation of a Mettl16 knockout (KO) allele. See also Figure S4A and STAR Methods. (B) Timeline of mouse embryogenesis. Embryonic day 2.5 (E2.5) embryos referring to 16-cell morula stage, E3.5 blastocysts, and E6.5 and E8.5 embryos were collected for genotyping. KO embryos were detected in expected Mendelian ratios till E3.5 (colored in green), but at sub-Mendelian ratios at E6.5 or none beyond (colored in red). See also Figures S4C–S4F. (C) Genotyping of E2.5 embryos from Mettl16^+/− x Mettl16^+/− crosses confirmed the expected Mendelian ratios among the genotypes. Scale bar in μm is indicated. (D) Transcriptome of individual isolated E2.5 embryos of Mettl16^−/− (KO), Mettl16^+/− (HET), and Mettl16^+/+ (WT) was sequenced and compared between the genotypes. The MA plots show a very limited number of differentially expressed genes (red dots, adjusted p ≤ 0.1). See also Figure S5. (E) Heatmap shows the expression of genes with significant differential expression between any two genotypes (adjusted p ≤ 0.1). Genes differentially expressed in Mettl16^−/−(KO) when compared to both Mettl16^+/− (HET) and Mettl16^+/+ (WT) are marked by red arrowhead. (F) The boxplots show the expected downregulation of the targeted gene (Mettl16) in KO samples, as well as the downregulation of Mat2a. Transcript levels of individual samples are shown as dots. See also Figure S5C. (G) Normalized read coverage along the Mat2a locus demonstrates the overall depletion in the KO. Note that the gene is on the Crick strand, so it goes from right to left. (H) Lack of METTL16 results in aberrant splicing of the last intron. The reads spanning the splice junction (SJ) of last Mat2a (ENSMUST00000059472.9) intron are significantly depleted in the KO even when normalized to overall Mat2a transcript levels. This is accompanied by slight increase for intron reads and increased usage of alternative 3ʹ splice-site characteristic for the ENSMUST00000206904.1 and ENSMUST00000206692.1 variants. See also Figure S5D.

Figure 5

E3.5 Mettl16^−/− Blastocysts Display Normal Morphology but Vast Transcriptome Dysregulation (A) E3.5 Mettl16^−/− KO embryos display normal morphology and their counts from Mettl16^+/− x Mettl16^+/− crosses correspond to expected Mendelian ratios among the genotypes. Scale bar in μm is indicated. (B) The boxplots show the expected downregulation of the targeted gene (Mettl16) in KO samples, as well as the downregulation of Mat2a. Transcript levels of individual samples are shown as dots. See also Figure S6A. (C) MA plots comparing the expression between the genotypes reveal that the vast number of genes are dysregulated in the Mettl16^−/− KO embryos. The genes with significantly different expression are shown as red dots (adjusted p ≤ 0.1). (D) Heatmap shows the expression of 5,166 genes with significant differential expression between any two genotypes (adjusted p ≤ 0.1). Note the massive dysregulation in the KO embryos. See also Figures S6B–S6D. (E) Venn diagrams compare the lists of dysregulated genes when Mettl16^−/− expression is compared to Mettl16^+/− or to Mettl16^+/+. (F) Comparison of proportion of reads encompassing splice junctions does not reveal a difference in splicing between individual genotypes. (G) Global transcription from exons, introns, and repeats is not affected in Mettl16^−/−. Error bars refer to SD.

Figure 6

A Model for METTL16 Function during Early Embryonic Development (A) Structural comparison of METTL16 core and METTL3/METTL14 complex. METTL3 (PDB 5IL2), colored purple, was superimposed on the METTL16-core-SAH (PDB 6GFN), colored in green (core) and red (N-terminal). Gate loops 1 and 2, and the interface loop of METTL3, are colored in blue, orange, and yellow, respectively. The disordered loop in our METTL16-core is shown as a dotted line. (B) Surface charge representation of human METTL16-core domain with a modeled tRNA (from PDB: 2ZZM). See STAR Methods. The SAH bound in the catalytic pocket is shown. (C) A methyl-acceptor adenosine (orange) was modeled into the SAH binding site of METTL16 core (PDB 6GFN) by superimposition of an m⁶A DNA MTase, EcoP151 (PDB: 4ZCF, chain B). The sulfate binding site (as in Figure 1F) overlaps with the adenosine base moiety. (D) A model summarizing the physiological role of METTL16 during early mouse development. The downregulation of the SAM synthetase Mat2a mRNA in Mettl16 KO E2.5 morula is potentially a trigger for subsequent massive alteration in gene expression in the E3.5 blastocysts. Such mutant embryos fail to proceed further in development (indicated in red).