Fig 2: m6Aboost predicts 36,556 m6A sites from HEK293T miCLIP2 data. (A) Predicted m6A sites are located around the stop codon. Visualization as in Figure 2F. (B) Selected m6A sites were validated by SELECT with HEK293T cells treated with METTL3 inhibitor (STM2457) or DMSO control. Visualization as in Figure 2H. Neighboring unmodified A nucleotides as control are given in Supplementary Figure S5C. *** P value < 0.001, ** P < 0.01, two-sided Student's t-test, n = 3. (C) Predicted m6A sites overlap with YTHDF1 binding sites. Genome browser view of the gene CCDC85B shows crosslink events from published YTHDF1 iCLIP data for HEK293T together with miCLIP2 signal (merge of four replicates) and m6Aboost-predicted m6A site (green arrowhead) from our HEK293T miCLIP2 data. (D) YTHDF1 precisely binds at the predicted m6A sites. Percentage of m6A sites (position 0) with YTHDF1 binding sites (y-axis) in a 21-nt window are given for predicted m6A sites at DRACH (yellow) and non-DRACH (green), as well as predicted unmodified sites (grey). (E) Predicted m6A sites from HEK293T miCLIP2 overlap with published m6A data. Venn diagram shows single-nucleotide overlap with miCLIP and m6ACE-seq data (m6A antibody by Synaptic Systems and Abcam, respectively). Note that m6A sites in Boulias et al., 2019 had been filtered for DRACH motifs. (F, G) Analysis of m6A sites that are unique to one of the three datasets compared in (E). (F) Unique m6A sites accumulate around stop codons. Visualization as in Figure 2F. (G) Unique m6A sites are enriched in YTHDF1 binding sites. Visualization as in (D). (H) Most m6A sites are shared between two different human cell lines. Venn diagram shows overlap of predicted m6A sites in expressed genes (TPM = 20, n = 3,298) from HEK293T and C643 cells. A Venn diagram without expression filter is shown in Supplementary Figure S5E. (I) More m6A sites are shared between two human cell lines in higher expressed genes. TPM threshold representing the gene expression (x-axis) against the Jaccard index (y-axis). Numbers of overlapping m6A sites are shown as comparison (blue).

Fig 3: Optimization of signal-to-noise for m6A-ELISA. (A) Testing of m6A antibody specificity. Various ratios of in vitro transcribed (IVT) RNA unmodified adenosine (A), and IVT m6A containing RNA (m6A) were used (1%, 0.5%, 0.1%, or 0% m6A IVT RNA). In total, 50 ng IVT RNA was used for each sample. Displayed are the mean and individual OD450 values of n = 2 technical replicates from a representative experiment. (B) Primary antibody and blocking reagent testing. Primary antibody (ABClonal-A19841) was incubated with or without 0.5 µg/mL of total RNA to reduce background binding to unmodified RNA. We used 50 ng in vitro transcribed (IVT) RNA generated with unmodified adenosine (A), and 0.01 ng of IVT m6A containing RNA (m6A). Additionally, we used poly(A)-selected mRNA samples from wild-type (WT) (FW1511) diploid cells and cells harboring gene deletion in IME4 (ime4?) (FW7030) that were induced to enter meiosis. Displayed are the mean and individual OD450 values of n = 2 technical replicates. (C) Similar analysis as in B. We used 50 ng poly(A)-selected mRNA samples from WT and ime4? cells that were induced to enter meiosis, and normalized samples to the WT signal. Displayed are n = 4 technical replicates across n = 2 independent experiments (unpaired t-test of P = 0.0004). (D) Primary antibody testing of samples using m6A-ELISA. Two m6A antibodies (Invitrogen-RM362 and Synaptic systems 202003) were tested in the m6A-ELISA. Displayed are the mean and individual OD450 values of n = 2 technical replicates from a representative experiment. (E) Similar analysis as in A, except that isolated total RNA was used for the analysis for WT and ime4? cells induced to enter meiosis. Displayed are the mean and individual OD450 values of n = 2 technical replicates from a representative experiment. (F) Comparison of m6A with a commercial m6A ELISA (EpiQuik). Like in A, the m6A signal for WT and ime4? cells were compared. For each experiment, the ime4? OD450 signals were normalized to WT OD450 signal (set to 1). The mean and individual values of n = 4 independent experiments are shown.

Fig 4: FIO1-dependent m6A modification of poly(A)+mRNA is rare or absent, but FIO1 is required for U6 snRNA methylation.(A) Liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis. Boxplots showing that the ratio of m6A to A in poly(A)+mRNA is only modestly reduced in fio1-1 mutants compared to Col-0. In contrast, m6A levels are significantly reduced in the fip37-4 mutant. (B) The intersection of m6A modification sites detected by nanopore direct RNA sequencing of poly(A)+RNA purified from Col-0, fip37-4, and fio1-1, visualised using an upset plot. All sites shown have significant differences in modification level in a three-way comparison between fip37-4, fio1-1, and Col-0. Bars show size of intersections between sites which are significant in each two-way comparison. Total intersection sizes displayed in black above each bar. A comparison to previously identified m6A sites using the orthogonal technique, miCLIP is included: orange and blue bar fractions show the number of sites within each set intersection that have or do not have an miCLIP peak (Parker et al., 2020) within 5 nt, respectively. Percentage of intersections with miCLIP support is displayed in orange above each bar. A small number of sites are significant in the three-way comparison, but in neither two-way comparison (far left bar). (C–D) Detection of RNAs immunoprecipitated from Col-0 and fio1 mutant alleles with anti-m6A antibodies using RT-qPCR analysis. The data are presented as strip-plots with mean and 95% CIs, showing the enrichment of U6 and U2 snRNAs over input using (C) Synaptic systems #202 003 anti-m6A antibody and RNA purified from Col-0 and fio1-1, and (D) Millipore ABE572 anti-m6A antibody and RNA purified from Col-0, fio1-1, fio1-3, and fio1-4. Y axes show -?Ct (m6A-IP — input) corrected for input dilution factor. Strip-plots show mean values for three or four independent RT-qPCR amplifications on each biological replicate immunopurification experiment. Figure 2—source data 1.LC:MS/MS data for Col-0, fio1-1, and fip37-4 poly(A)+RNA. Figure 2—source data 2.Differential modification sites detected from a three-way omparison of Col-0, fip37-4, and fio1-1 mutants using nanopore direct RNA sequencing (DRS) data. Figure 2—source data 3.m6A-IP quatitative PCR (qPCR) data for U6 and U2 snRNAs in Col-0, fio1-1, fio1-3, and fio1-4 mutants. Figure 2—source data 4.Differential poly(A) site usage results for fip37-4 and fio1-1 identified using nanopore direct RNA sequencing (DRS) data.