The m6A pathway protects the transcriptome integrity by restricting RNA chimera formation in plants

Abstract

Global, segmental, and gene duplication-related processes are driving genome size and complexity in plants. Despite their evolutionary potentials, those processes can also have adverse effects on genome regulation, thus implying the existence of specialized corrective mechanisms. Here, we report that an N6-methyladenosine (m⁶A)-assisted polyadenylation (m-ASP) pathway ensures transcriptome integrity in Arabidopsis thaliana Efficient m-ASP pathway activity requires the m⁶A methyltransferase-associated factor FIP37 and CPSF30L, an m⁶A reader corresponding to an YT512-B Homology Domain-containing protein (YTHDC)-type domain containing isoform of the 30-kD subunit of cleavage and polyadenylation specificity factor. Targets of the m-ASP pathway are enriched in recently rearranged gene pairs, displayed an atypical chromatin signature, and showed transcriptional readthrough and mRNA chimera formation in FIP37- and CPSF30L-deficient plants. Furthermore, we showed that the m-ASP pathway can also restrict the formation of chimeric gene/transposable-element transcript, suggesting a possible implication of this pathway in the control of transposable elements at specific locus. Taken together, our results point to selective recognition of 3'-UTR m⁶A as a safeguard mechanism ensuring transcriptome integrity at rearranged genomic loci in plants.

Figure S1.. NERD controls chimeric RNA formation at psORF/AT5TE50260 loci.

(A) Depicted are individual RNA-seq (including spliced reads) reads from WT, nerd-1, and complemented nerd+T plants mapping to the psORF-AT5TE50260 locus. Exons are shown with colored thick bars and UTRs with colored thin bars. The nerd-dependent cryptic intergenic intron indicated as thin diagonal red lines results in the partial splicing of the annotated exons and UTR regions colored with red-hatched bars in nerd-1 mutant. Primers used in the semi-qRT–PCR assays are indicated. Genomic regions analyzed in quantitative RT–PCR are labeled P1 to P3. (B) Semiquantitative RT–PCR analyses of chimeric psORF-AT5TE50260 transcripts in WT, nerd-1, and nerd+T plants. Unspliced (u) and spliced (s) forms of RT–PCR products are indicated. EF1-4α was used as loading control. Minus RT (−RT) reactions are controls for genomic DNA contamination. (C) Quantitative RT–PCR analysis of chimeric psORF-AT5TE50260 transcripts in WT, nerd-1, and nerd+T plants. Relative transcript levels were normalized to actin using the ΔΔCt method. Data represent the means of four independent experiments and error bars the corresponding SD values.

Figure S2.. NERD affects chimeric mRNA formation in Arabidopsis.

(A) Shown are normalized density profiles of RNA-seq and corresponding sashimi plots from two nerd-1 up-regulated GENE2 (AT4G30570 and AT1G71330) whose overexpression results from transcription readthrough of an upstream GENE1 (AT4G30580 and AT1G71340). Exons are shown with colored thick bars, UTRs with colored thin bars, introns with black dashed lines, and nerd-dependent cryptic intergenic introns with dashed diagonal red lines. The annotated exons and UTR regions partially spliced in nerd-1 are indicated by cross-hatched bars. The unannotated exons or UTR regions newly appearing in nerd-1 are shown with red bars. The numbered horizontal lines indicate the chromatin regions analyzed by qPCR in chromatin immunoprecipitation assays. (B) Semiquantitative RT–PCR analyses of chimeric AT4G30580/70 and AT4G71340/30 mRNAs in WT, nerd-1, and nerd+T plants. EF1-4α was used as loading control. Minus RT (−RT) reactions are controls for genomic DNA contamination. F1, F2 and R1 refer to the primers indicated on panel A as arrowheads. (C) RNA gel blot analysis of AT4G30580 and AT4G71340 mRNAs in WT, nerd-1, and nerd+T plants. The chimeric mRNA is indicated by an arrow. The probes used are indicated by N in panel (A). AT2G43410 is used as a control. (D) H3K4me3 chromatin immunoprecipitation analysis of the AT4G30580/70 and AT4G71340/30 gene pairs in WT, nerd-1, and nerd+T plants by qPCR. The location of the primer pairs is shown in panel (A). Values are means ± SD from two independent experiments.

Figure 1.. NERD restricts chimeric mRNA formation in Arabidopsis.

(A) Boxplot showing the read counts in gene body and corresponding downstream region for genes showing enhanced transcription readthrough in nerd-1. ***P < 0.001; t test. (B) Evolutionary analysis of GENE1/GENE2 conservation in 325 angiosperm genomes. The heat map shows the relative percentage of four different scenarios for each gene pair: adjacent: GENE1 and GENE2 have a contiguous conserved position; distant: GENE1 and GENE2 are separated on the same or different chromosomes. GENE1 only and GENE2 only: correspond to situations where only one gene of the tandem pairs was found in the genome. (C) Semi-quantitative RT–PCR analysis of extended/chimeric ITN mRNAs in WT, nerd-1, nerd-1+T, nerd-3, and nerd-3+T plants (left panel). Exons are shown with colored thick bars, UTRs with colored thin bars, introns with black dashed lines, and nerd-dependent cryptic intergenic introns with dashed diagonal red lines. The annotated exons and UTR regions partially spliced in nerd-1 are indicated by cross-hatched bars. The unannotated exons or UTR regions newly appearing in nerd-1 are shown with red bars. Primers used in the RT–PCR experiments are shown as arrowheads on the right panel. Transcript regions analyzed in quantitative RT–PCR are labeled as bars. Unspliced (u) or spliced (s) forms of RT–PCR products corresponding to chimeric or extended mRNA are indicated. EF1-4α was used as loading control. Minus RT (−RT) reactions are controls for DNA contamination. (right panel) (D) Expression score of GENE1 and GENE2 at ITN loci in different tissues using already published RNA-seq data (Araport11; Krishnakumar et al, 2015) and expressed in TPM.

Figure S3.. NERD affects chimeric mRNA formation in Arabidopsis.

(A) Shown are sashimi plots from loci forming chimeric mRNA transcripts in nerd-1 mutant. The transcript structures and annotations are the same as in Fig S2A. (B) Shown are sashimi plots from two loci forming extended mRNA in nerd-1 mutant. The transcript structures and annotations are the same as in Fig S2A. (C) qRT–PCR analysis of ITN GENE1/2 mRNAs in WT, nerd-1, and nerd+T plants. Relative transcript levels were normalized to actin and WT using the ΔΔCt method. Data represent the means of four experiments and error bars the corresponding SD values. AT4G26410 is used as a control.

Figure S4.. Evidence for concerted and specific roles for NERD and FPA in mRNA chimera control in Arabidopsis.

(A) Overlap between loci showing chimeric mRNA formation in fpa-7 and nerd-1 mutants. (B) Semi-quantitative RT–PCR analyses of ITN- and FPA-dependent loci in WT, nerd-1, fca-9, and fpa-7 plants. The unspliced (u) or spliced (s) forms of RT–PCR products of extended/chimeric mRNAs are indicated. Minus RT (−RT) reactions are controls for genomic DNA contamination. EF1-4α was used as loading control. (C) Expression score of GENE1 and GENE2 at FPA-dependent chimeric loci in different tissues using already published RNA-seq data (Araport11) and expressed in TPM. (D) Distribution of GENE1 and GENE2 from FPA-dependent chimeric loci among the nine CSs defined in A. thaliana. The percentage of genes associated with each CS is indicated.

Figure 2.. Recently rearranged gene pairs are significantly overrepresented among ITN loci.

(A) Relative percentage of GENE1/GENE2 duplication categories compared with total gene set in A. thaliana. (B) Examples of GENE2 translocation that occurred recently in A. thaliana (top), after the separation from the A. lyrata (middle), and C. rubella (bottom) lineages. (C) Boxplot of Ka/Ks ratios of GENE2, pseudogenes, and total genes using closest paralogs in A. thaliana and orthologs in A. lyrata. In (A) and (C), P-values were determined using Kruskal–Wallis statistical test between each gene category. ns: not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. (D) Distribution of GENE1 and GENE2 from ITN loci among the nine CSs defined in A. thaliana. The percentage of genes associated with each CS is indicated.

Figure S5.. Recently rearranged gene pairs are significantly overrepresented among ITN loci.

(A) Example of a GENE2 translocation that occurred in the last common ancestor of A. thaliana and A. lyrata, after the separation from the C. rubella lineage. (B) Diagram showing the ITN reporter constructs developed to assess GENE2 protein expression in vivo. Westerns blots showing cMyc-tagged GENE2 protein expression in independent transformants are shown together with a Coomassie staining showing equal loading (left). (C) AT4G30570 GENE2 encodes for an N-terminally truncated version of VTC1 gene product. The AT4G30570 ORF truncation results from a missed start codon that shifts the predicted initiation codon to the next in frame methionine at position 22. No in-frame ATG codon is present upstream to annotated ATG codon. Sequence comparison of conserved N-terminal domains of VTC1-type proteins in eukaryotes. Hs, Homo sapiens; Sp, Schizosaccharomyces pombe; Zm, Zea mays.

Figure S6.. AGO hook motifs are not required for the termination-promoting activity of NERD.

(A) Western blot analyses of nerd-1 plants expressing flag-tagged WT (T) and ago-hook mutant (tag) versions of NERD were performed using the anti-flag antibody. UGPase antibody was used as control (*). (B) Semiquantitative RT–PCR analysis of extended/chimeric ITN mRNAs in WT, nerd-1, nerd+T, and nerd+Tag plants. The positions of unspliced (u) or spliced (s) RT–PCR products are indicated. Minus RT (−RT) reactions are controls for genomic DNA contamination. EF1-4α was used as loading control. (C) qRT–PCR analysis of extended/chimeric ITN mRNAs in WT, nerd-1, nerd+T, and nerd+Tag plants. Relative transcript levels were normalized to ACTIN and WT using the ΔΔCt method. Data represent the means of three independent experiments and error bars the corresponding SD values.

Figure 3.. FIP37-dependent m⁶A modification controls mRNA chimera formation at ITN loci.

(A) Shown are sashimi plots of two ITN loci forming chimeric mRNA transcripts in fip37L mutant. The transcript structures and annotations are the same as in Fig 1C. (B) Semi-quantitative RT–PCR analyses at ITN loci in nerd-1, fip37L, and their corresponding WT controls (top part) and at FPA-only loci in fpa-7, fip37L, and their corresponding controls (bottom part). EF1-4α was used as loading control. Minus RT (−RT) reactions are controls for genomic DNA contamination. (C) Overlap between loci showing an increase in downstream read counts in nerd-1 and fip37L plants. **** denotes P-value < 0.0001 for enrichment in the overlap, hypergeometric statistical test. (D) qRT–PCR analysis of extended/chimeric ITN mRNAs in WT, nerd-1, fip37L, and nerd-1/fip37L plants. WT and nerd-1 lines used here correspond to segregants harboring the LEC-FIP37 transgene. Relative transcript levels were normalized to actin using the ΔΔCt method. Data represent the means of four replicate experiments and error bars correspond to SD values. AT4G26410 is used as a control. (E) LC-MS/MS quantification of the m⁶A/A ratio in mRNA isolated from WT, fip37L, and nerd-1 plants. *P < 0.05; ***P < 0.001; t test. (F) ITN GENE1 and methylated control mRNAs are m⁶A-methylated in an FIP37- and NERD-dependent manner. m⁶A-IP-qPCR analysis was performed on several ITN and methylated control loci in WT, nerd-1, and fip37L seedlings. Error bars, mean ± SD; n = 3 biological replicates. (G) Scatter plot representation of m⁶A enrichment score (on the Y-axis) and FC in downstream reads between WT and fip37L conditions (X-axis).

Figure S7.. FIP37 restricts chimeric/extended mRNA formation at ITN, but not FPA-only loci in Arabidopsis.

(A) Normalized reads and junctions mapping at ITN loci in WT and fip37L plants. The transcript structures and annotations are the same as in Fig S2. (B) Normalized reads and junctions mapping to FPA-only genomic loci in WT and fip37L plants. The transcript structures and annotations are the same as in Fig S1. The mRNA chimera formed upon the depletion of FPA is indicated with red dashed lines according to Duc et al (2013).

Figure S8.. Extended/chimeric mRNA formation at ITN loci is coupled with their loss of m⁶A in fip37L.

(A) ITN GENE1s are m⁶A methylated in an FIP37-dependent manner. Upper panel shows m⁶A peaks as revealed by m⁶A-seq performed in WT and fip37L plants (Shen et al, 2016). Light blue color represents input reads, whereas red color represents IP reads. The representation of ITN loci is the same as Fig S1. (B) Loss of m⁶A methylation upon FIP37 depletion does not trigger extended/chimeric mRNA formation at non-ITN genes. Upper panel shows m⁶A peaks as revealed by m⁶A-seq performed in WT and fip37L plants (Shen et al, 2016). Light blue color represents input reads, whereas red color represents IP. Normalized reads mapping to the methylated control loci are shown in WT and fip37L backgrounds (bottom panel). The FC in read counts in the region downstream of control genes is <2.

Figure 4.. m⁶A-assisted polyadenylation controls mRNA chimera formation at ITN loci.

(A) 3′ RACE gel images of ITN loci in WT and fip37L plants. RACE assays were performed with total RNA isolated from 9-d-old seedlings. The polyadenylated sites corresponding to the ITN GENE1 (pAx) are indicated at the bottom of the gel, whereas the polyadenylated sites corresponding to the mRNA chimera (pAc) migrate at the top. The FIP37-dependent GENE1 polyadenylation sites are marked in red. (B) The primers used in the 3′ RACE are indicated with arrowheads and are localized on the corresponding sequences in (B). The oligo (dT) anchor primer is indicated with a broken arrow. (B) PolyA sites detected at ITN loci. The position of the polyA sites is indicated at the X-axis relative from ITN GENE1 STOP codon. The respective positions of ITN GENE1 and GENE2 loci are indicated with green and blue lines. Primers used in RACE assays are indicated with arrowheads and broken arrow. (C) 3′ RACE gel images of m⁶A methylated control loci in WT and fip37L plants. RACE assays were performed with total RNA isolated from 9-d-old seedlings. The position of the polyA sites is indicated at the X-axis relative from ITN GENE1 STOP codon.

Figure 5.. CPSF30L links m⁶A methylation to site-specific polyadenylation at ITN loci.

(A) Structure of the CPSF30 gene and mRNA transcripts. Exons and UTRs are shown with thick and thin grey bars, respectively. Introns are shown with black-dashed lines. APA at Intron-2 produces CPSF30S mRNA, whereas alternative splicing of Intron-2 generates CPSF30L mRNA. Location of the YTH domain in CPSF30L is indicated with a brown rectangle. The positions of T-DNA insertions in the cpsf30 mutants are indicated. Primers used in the qRT–PCR experiments are indicated. (B) Overlap between loci showing a significant increase in downstream read counts in cpsf30-1, cpsf30-3, and fip37L plants. **** denotes P-value < 0.0001 for enrichment in the overlaps, hypergeometric statistical test. (C) Boxplot showing the read counts in the 500-bp downstream region for FIP37-dependent and FIP37-independent CPSF30-responsive loci in WT, cpsf30-1, cpsf30-3, and fip37L plants. ns, not significant; ***P < 0.001; t test. (D) Relative contribution of CPSF30S and CPSF30L isoforms to mRNA chimera formation control at ITN- and FIP37-independent CPSF30-responsive loci. Data were calculated from the qRT–PCR analysis of ITN- and FIP37-independent CPSF30 chimeric/extended transcripts presented in Fig S9E. Data represent the means of three independent experiments and error bars corresponds to SD values. (E) Iso. Cont. to read. stands for isoform contribution to readthrough and is expressed in % (E) 3′ RACE gel images of ITN loci in WT, fip37L, and cpsf30-3 9-d-old seedlings. RACE assays were performed as previously described. (F) Western blot analyses of cpsf30-1 mutant plants expressing Myc-tagged WT (30L) and yth mutant (30Lyth) versions of CPSF30L. The corresponding Coomassie staining is represented on the left part. The anti-Myc antibody was used to visualize the tagged proteins (right part). (G) qRT–PCR analysis of ITN- and FIP37-independent CPSF30-responsive mRNAs accumulation in WT, cpsf30-1, cpsf30-1+30L, and cpsf30-1+30Lyth plants. Relative transcript levels were normalized to actin using the ΔΔCt method. Data represent the means of four independent experiments and error bars correspond to SD values.

Figure S9.. CPSF30L isoform controls extended/chimeric mRNA formation at ITN loci.

(A) qRT–PCR analysis of AtCPSF30S and AtCPSF30L encoding transcripts in WT, cpsf30-1, and cpsf30-3 seedlings. Relative transcript levels were normalized to actin using the ΔΔCt method. Data represent the means of four experiments and error bars the corresponding SD values. The primers used are shown as arrowheads on Fig 5A; (A) and (B) amplify AtCPSF30S transcripts, whereas (A) and (C) amplify AtCPSF30L ones. (B) Normalized reads and junctions mapping at ITN loci in WT, cpsf30-1, cpsf30-3, and fip37L plants. The transcript structures and annotations are the same as in Fig S2A. (C) Normalized reads and junctions mapping at FIP37-independent CPSF30-dependent loci in WT, cpsf30-1, cpsf30-3, and fip37L plants. The transcript structures and annotations are the same as in Fig S2A. (D) Semiquantitative RT–PCR analysis of extended/chimeric ITN- and FIP37-independent CPSF30 loci in WT, fip37L, and cpsf30-1 seedlings. EF1α was used as loading control. Minus RT (−RT) reactions are controls for genomic DNA contamination. (E) qRT–PCR analysis of ITN- and FIP37-independent CPSF30 loci in WT, cpsf30-1, and cpsf30-3 plants. Relative transcript levels were normalized to actin using the ΔΔCt method. Data represent the means of at least three independent experiments and error bars the corresponding SD values. (F) Sequence comparison of CPSF30L YTH domain with other YTH motifs known to bind m⁶A. Asterisks identify residues that compose the conserved aromatic cage of the YTH domain that is involved in m⁶A recognition. The residues mutated of the CPSF30Lyth mutant are indicated in red. The three first sequences correspond to human YTH domains, whereas ZrMRB1 comes from Zygosaccharomyces rouxii MRB1 and the last one corresponds to AtCPSF30L YTH domain.

Figure 6.. m⁶A-assisted polyadenylation restricts chimeric GENE-TE transcript formation in Arabidopsis.

(A) An example of a LINE1-type retroelement translocation that has occurred recently in A. thaliana (top), after the separation from the A. lyrata (middle) and C. rubella (bottom) lineages. (B) Normalized reads mapping to the chimeric AT3G47890-AT3TE71565 loci are shown in WT, fip37L, cpsf30-1, and cpsf30-3 backgrounds. Upper panel shows m⁶A peaks as revealed by m⁶A-seq performed in WT and fip37L backgrounds (Shen et al, 2016). Light blue color represents input reads, whereas red color represents IP reads. The transcript structures and annotations are the same as in Fig 1C. Primers used in the RT–PCR and McrBC experiments are shown as green, red, and grey arrowheads, respectively. (C) 3′ RACE gel images of chimeric AT3G47890-AT3TE71565 loci in WT, fip37L, cpsf30-3, and cpsf30-1 plants (top panel). RACE assays were performed as previously described. pA2 corresponds to the AT3G47890 polyadenylated site that decreases in fip37L and cpsf30-3 plants, whereas pAc indicates the polyadenylated sites corresponding to the mRNA chimera. PolyA sites detected at AT3G47890/AT3TE71565 ITN loci are indicated relative to GENE1 STOP codon (bottom panel). The respective positions of ITN GENE1 and TE loci are indicated with green and blue lines. The primers used in the 3′ RACE are indicated with arrowheads. The oligo (dT) anchor primer is indicated with a broken arrow. (D) Model for m⁶A-assisted polyadenylation (m-ASP) at ITN loci. NERD and FIP37 are required to reach full m⁶A deposition at all loci including ITN GENE1s (1). The recognition of m⁶A by the YTHDC-type domain of CPSF30L promotes cleavage and polyadenylation at GENE1 3′-UTR (2), therefore restricting mRNA chimera formation (3).

Figure S10.. m⁶A-assisted polyadenylation can restrict chimeric GENE-TE transcript formation in Arabidopsis.

(A) Organization of AT3TE71565 LINE element. TSD means target site duplication. RT indicates reverse transcriptase domain. The polyA tail is shown (pA). (B) Snapshot from Jacobsen laboratory browser indicating that AT3TE71565 LINE is methylated on the three cytosine contexts and produced specific siRNAs in WT plants. (C) Semiquantitative RT–PCR analyses of chimeric AT3G47890/AT3TE71565 transcripts in fip37L and cpsf30-1 mutants along with their corresponding WT controls. EF1-4α was used as loading control. Minus RT (−RT) reactions are controls for genomic DNA contamination. The primers used are represented on Fig 6B as arrowheads. (D) Location of the AT3TE71565 LINE1 element relative to AT3G47890 3′-UTR and corresponding sequence in the A. lyrata ortholog. TAA indicates the stop codon. The locations of the main polyA sites are indicated with vertical arrowheads. The FIP37/CPSF30L-dependent polyA site is marked in red. (E) Analysis of DNA methylation at AT3TE71565 locus into WT, fip37L, cpsf30-1, and cpsf30-3 plants. Genomic DNA was digested with the McrBC methylation sensitive enzyme and used as a template for PCR. Undigested DNA at the AT3TE71565 locus was used as a loading control. The primers used are represented on Fig 6B as green and grey arrowheads.