The m⁶A Reader ECT2 Controls Trichome Morphology by Affecting mRNA Stability in Arabidopsis

The YTH-domain family protein ECT2 is a reader of the epitranscriptomic mark m⁶A that affects trichome morphology and functions in mRNA stability and 3′ UTR processing in Arabidopsis.

Abstract

The epitranscriptomic mark N⁶-methyladenosine (m⁶A) can be written, read, and erased via the action of a complex network of proteins. m⁶A binding proteins read m⁶A marks and transduce their downstream regulatory effects by altering RNA metabolic processes. The characterization of m⁶A readers is an essential prerequisite for understanding the roles of m⁶A in plants, but the identities of m⁶A readers have been unclear. Here, we characterized the YTH-domain family protein ECT2 as an Arabidopsis thaliana m⁶A reader whose m⁶A binding function is required for normal trichome morphology. We developed the formaldehyde cross-linking and immunoprecipitation method to identify ECT2-RNA interaction sites at the transcriptome-wide level. This analysis demonstrated that ECT2 binding sites are strongly enriched in the 3′ untranslated regions (3′ UTRs) of target genes and led to the identification of a plant-specific m⁶A motif. Sequencing analysis suggested that ECT2 plays dual roles in regulating 3′ UTR processing in the nucleus and facilitating mRNA stability in the cytoplasm. Disruption of ECT2 accelerated the degradation of three ECT2 binding transcripts related to trichome morphogenesis, thereby affecting trichome branching. The results shed light on the underlying mechanisms of the roles of m⁶A in RNA metabolism, as well as plant development and physiology.

INTRODUCTION

The epitranscriptomic mark N⁶-methyladenosine (m⁶A) is the most prevalent dynamic internal mRNA modification in eukaryotes. In this process, analogous to DNA methylation, m⁶A can be dynamically written, erased, and read; these events are now known to function in gene regulation (Bokar et al., 1997; Jia et al., 2011; Zheng et al., 2013; Wang et al., 2014, 2015; Xiao et al., 2016). Mammals have at least two types of proteins that can write m⁶A modifications: methyltransferases like 16 (METTL16) and multiprotein writer complexes from which several proteins have been characterized (e.g., METTL3, METTL14, Wilms tumor 1-associating protein [WTAP], KIAA1429, and RNA binding motif protein 15) (Bokar et al., 1997; Liu et al., 2014; Ping et al., 2014; Schwartz et al., 2014; Patil et al., 2016; Pendleton et al., 2017). While less is known about m⁶A methylation in plants, two major transcriptome-wide sequencing studies in Arabidopsis thaliana (Can-0 and Hen-16) have demonstrated that m⁶A is a highly conserved and dynamic modification with functional roles in mRNA metabolism in plants (Luo et al., 2014). Several m⁶A methyltransferase subunits have been characterized in Arabidopsis (e.g., N⁶-adenosine-methyltransferase MT-A70-like [MTA], the plant homolog of human METTL3; Methyltransferase MT-A70 family protein, a homolog of human METTL14; FKBP12 interacting protein 37 [FIP37], a homolog of human WTAP; VIRILIZER, a homolog of human KIAA1429; and the E3 ubiquitin ligase HAKAI) (Zhong et al., 2008; Bodi et al., 2012; Shen et al., 2016; Růžička et al., 2017). m⁶A modification in plants is reversible and can be erased by two Arabidopsis homologs of the human m⁶A demethylase ALKBH5, ALKBH9B, and ALKBH10B (Duan et al., 2017; Martínez-Pérez et al., 2017). ALKBH9B has m⁶A demethylase activity and involved in defense against viral infection (Martínez-Pérez et al., 2017); ALKBH10B is an mRNA m⁶A demethylase affecting the floral transition and vegetative growth (Duan et al., 2017).

The characterization of m⁶A readers is thus important for deepening our understanding of how m⁶A functions in mRNA processing and thus functions in many biological processes. Fundamental studies have established that five human YTH (YT512-B Homology)-domain family proteins are m⁶A readers that function in or affect pre-mRNA splicing, nuclear export, mRNA degradation, and mRNA translation efficiency (Wang et al., 2014, 2015; Xiao et al., 2016; Hsu et al., 2017; Roundtree et al., 2017; Shi et al., 2017). Structural studies of the YTH domains from reader proteins in complex with m⁶A-containing RNA (e.g., YTHDF1 and YTHDC1) have revealed that all YTH domain proteins employ a highly conserved aromatic cage to recognize m⁶A (Xu et al., 2014, 2015). In conjunction with m⁶A readers, m⁶A modifications function in the regulation of stem cell fate, sex determination in Drosophila melanogaster, the maternal-to-zygotic transition in zebra fish (Danio rerio), and female oocyte maturation in mammals, among other roles (Aguilo et al., 2015; Geula et al., 2015; Lence et al., 2016; Ivanova et al., 2017; Kan et al., 2017; Zhao et al., 2017). To date, although a family of 13 proteins with putative YTH domains have been reported in Arabidopsis (Li et al., 2014), it remains unclear which plant proteins function as m⁶A readers. To explore and come to understand the roles of m⁶A in plants, the characterization of plant m⁶A readers is essential and will likely shed light on important regulatory mechanisms related to these extremely basic processes of RNA metabolism, including whether plants share similar m⁶A reading mechanisms with other eukaryotes or perhaps employ unique, plant-specific processes.

Here, we demonstrate that EVOLUTIONARILY CONSERVED C-TERMINAL REGION2 (ECT2) is an m⁶A reader protein. We establish that its m⁶A binding function is required for normal trichome morphology. Subcellular localization analysis showed that ECT2 is present in both the nucleus and the cytoplasm. We developed a method, formaldehyde cross-linking and immunoprecipitation (FA-CLIP), which allowed us to identify the binding sites of ECT2. Analysis of the positions of these sites not only showed dramatic 3′ untranslated region (UTR) enrichment, but also led to the identification of the m⁶A target motif, UGUA, which is consistently located in a position characteristic of so-called far upstream polyadenylation signals for alternative polyadenylation, suggesting that ECT2 might regulate 3′ UTR processing in the nucleus. We also show that the lack of ECT2 results in significantly reduced cellular populations of ECT2 target transcripts, indicating that, unlike the human m⁶A reader YTHDF2, the m⁶A reader ECT2 does not prepare m⁶A-modified transcripts for degradation in the cytoplasm, but it instead might facilitate mRNA stability in the cytoplasm. Disruption of ECT2 decreased the expression levels of three ECT2 binding transcripts related to trichome morphogenesis through accelerating their mRNA degradation, thereby affecting trichome branching. Collectively, our work demonstrates that the m⁶A binding function of ECT2 controls trichome morphology via affecting mRNA stability and suggests that ECT2 functions in basic RNA metabolism, specifically in 3′ UTR processing and mRNA stability.

RESULTS

The Ubiquitously Expressed YTH-Domain Family Protein ECT2 Is Strongly Expressed in Rapidly Developing Tissues and Has a Morphogenesis-Related Function

Exciting recent discoveries about how human YTH-domain family proteins mediate m⁶A modifications and thereby influence mRNA processing and biological processes motivated us to explore the potential regulatory roles of evolutionarily related proteins in plants (Wang et al., 2014, 2015; Xiao et al., 2016; Hsu et al., 2017; Roundtree et al., 2017; Shi et al., 2017). There are 13 YTH-domain family proteins in Arabidopsis (Li et al., 2014). qPCR analysis of RNA samples from Arabidopsis seedlings revealed that the mRNA abundance of ECT2 was higher than that of the other YTH-domain family genes (Supplemental Figure 1). We also detected ECT2 transcripts in most Arabidopsis vegetative and reproductive organs that we examined, highlighting the apparently ubiquitous transcription of this gene in Arabidopsis (Figure 1A). Histochemical assays based on a ECT2 promoter-driven GUS construct in transgenic Arabidopsis plants (ECT2_pro:GUS) indicated that ECT2 expression is strongest in rapidly growing tissues such as apical meristems, lateral root primordia, root tips, trichomes, and pollen (Figure 1B to 1G), suggesting that ECT2 functions in these actively developing tissues.

Figure 1.

ECT2 Is Ubiquitously Expressed, with Highest Expression Levels in Rapidly Developing Tissues.

(A) Relative gene expression was measured using qPCR with ACTIN2 as a reference gene, prior to normalization to ECT2 expression levels in 14-d-old seedlings. Data are represented as means ± se, n = 2 biological replicates × 3 technical replicates. Biological replicates are parallel measurements of biologically distinct samples, and technical replicates are repeated measurements of the same sample.

(B) to (G) GUS staining analysis of ECT2_pro:GUS transgenic Arabidopsis plants. Blue bar = 1 mm; purple bar = 500 μm; red bar = 100 μm.

(B) A 12-d-old seedling and its meristem (shown on the right).

(C) Stage VII lateral root of a 12-d-old seedling.

(D) Main root of 12-d-old seedling.

(E) Trichomes of a 12-d-old seedling.

(F) Three-week-old rosette leaves.

(G) A 5-week-old flower and its pollen (stamens with pollen shown on the right).

To pursue this idea, we obtained two independent homozygous T-DNA insertion lines for the ECT2 gene, ect2-1 (SALK_002225) and ect2-2 (SAIL_11_D07) (Supplemental Figures 2A and 2B) (Alonso et al., 2003). The T-DNA insertion is located in intron 3 of the ECT2 locus in ect2-1 and in exon 6 in ect2-2. RT-PCR showed that full-length ECT2 transcript was not expressed in ect2-1, whereas ect2-2 retained ∼26% of wild-type levels of full-length ECT2 transcript (Supplemental Figures 2C and 2H). Examination of ect2-1 and ect2-2 plants showed an obvious trichome morphological phenotype. Cryo-scanning electron microscopy (cryo-SEM) revealed that the trichomes of both ect2-1 and ect2-2 plants were more extensively branched than wild-type trichomes (Figure 2A). Whereas 83% of wild-type trichomes had three branches, and none of these plants had trichomes with more than four branches (8% had four branches), ∼50% of ect2 trichomes (50.6% in ect2-1 and 43.6% in ect2-2) had four branches, and 1 to 2% of ect2 trichomes had five branches (Figure 2B). This phenotype, which was exhibited in two independent homozygous ect2 mutants, strongly suggests that ECT2 has a function related to trichome morphogenesis.

Figure 2.

ECT2 Is an m⁶A Reader Protein Whose m⁶A Binding Function Is Required for Normal Trichome Morphology.

(A) Cryo-SEM analysis of trichome morphology in wild-type, ect2-1, ect2-2, ECT2_pro:ECT2-Flag/ect2-1, and ECT2_pro:ECT2m-Flag/ect2-1 plants. Trichomes are from the third and fourth leaves of 3-week-old Arabidopsis plants. Bar = 100 μm.

(B) Statistical analysis of trichome branch number in wild-type, ect2-1, ect2-2, ECT2_pro:ECT2-Flag/ect2-1, and ECT2_pro:ECT2m-Flag/ect2-1 plants. Three hundred trichomes from the third and fourth leaves of 3-week-old Arabidopsis plants were analyzed.

(C) In vitro RIP-LC-MS/MS assay showing that m⁶A modification is enriched in GST-ECT2-bound mRNA compared with the flow-through and input samples. GST-ECT2 was expressed in E. coli, and mRNA was isolated from 14-d-old wild-type Arabidopsis seedlings. Data are presented as means ± se, n = 3 biological replicates × 2 technical replicates. ****P < 0.0001 by t test (two-sided).

(D) In vivo FA-RIP-LC-MS/MS showing that m⁶A modification is enriched in ECT2-Flag-bound RNA compared with IgG-bound RNA. Fourteen-day-old ECT2_pro:ECT2-Flag/ect2-1 seedlings were used for the experiment. Data are presented as means ± se, n = 2 biological replicates × 3 technical replicates. **P < 0.01 by t test (two-sided).

ECT2 Is an m⁶A Reader Protein Whose m⁶A Binding Function Is Required for Normal Trichome Morphology

ECT2 is a homolog of human YTHDF1/YTHDF2/YTHDF3 proteins (each containing a YTH domain) (Supplemental Figure 3). We used a combination of in vitro, bioinformatics, and in vivo methods to examine the potential m⁶A binding activity of ECT2 and to evaluate any possible relationship between this activity and the observed morphogenesis-related phenotypes. We expressed and purified recombinant GST-tagged ECT2 protein from Escherichia coli (Supplemental Figure 4A) and used it to perform the in vitro RNA immunoprecipitation LC-MS/MS assays that are typically used in studies of m⁶A readers (Wang et al., 2014). We isolated poly(A)-tailed RNA from 14-d-old Arabidopsis seedlings and incubated it with purified GST-ECT2 protein. We used LC-MS/MS to examine the m⁶A/A ratios of RNA molecules that were immunoprecipitated by ECT2 (separated based on GST affinity beads), RNA molecules in the flow-through eluate from this separation, and RNA molecules from the input (i.e., unincubated) samples. The ECT2-bound RNAs were highly enriched for m⁶A modifications compared with RNA molecules in the flow-through and input samples, suggesting that ECT2 binds to m⁶A sites (Figure 2C). To verify that ECT2 recognizes m⁶A sites in planta, we generated transgenic Arabidopsis plants (ECT2_pro:ECT2-Flag/ect2-1) expressing an ECT2-Flag fusion protein driven by the ECT2 promoter in the ect2-1 mutant background and performed in vivo formaldehyde RNA immunoprecipitation (FA-RIP)-LC-MS/MS assays. The proportion of RNA molecules with m⁶A modifications was dramatically higher among the ECT2-Flag immunoprecipitation (IP) RNAs compared with control IgG IP RNAs (Figure 2D). To further confirm that the m⁶A binding function of ECT2 depends on the presence of the YTH domain, we designed a putative binding function-abolished form of ECT2 W521A/W534A (named ECT2m, which contains two mutations: Trp-521 to Ala and Trp-534 to Ala) based on sequence alignments between five human YTH-domain proteins (among these, YTHDF1/YTHDF2/YTHDC1, with m⁶A binding function ligands that have been identified in crystal structures) and 13 Arabidopsis YTH-domain family proteins, which strongly suggested that three tryptophan residues in the YTH domain of ECT2 (Trp-464, Trp-521, and Trp-534) are likely directly involved in the binding of ECT2 to m⁶A sites (Supplemental Figure 3). We expressed and purified recombinant GST-ECT2 and GST-ECT2m protein from E. coli (Supplemental Figure 4A) and performed electrophoretic mobility shift assay (EMSA) with synthetic 42-mer RNAs containing either m⁶A or A. The EMSA analysis showed that GST-ECT2m had a completely abolished m⁶A binding function compared with wild-type GST-ECT2 (Supplemental Figure 4B). Collectively, these results indicate that ECT2 binds to RNA transcripts that harbor m⁶A sites, establishing that ECT2, like human YTH-domain family proteins, is an m⁶A reader protein.

Having established that ECT2 is an m⁶A reader, we performed genetic complementation to determine whether ECT2’s m⁶A binding activity, per se, functions in morphogenesis. Here, we used two complementation lines in the ect2-1 mutant background. The ECT2_pro:ECT2m-Flag/ect2-1 line expressed an m⁶A binding function-abolished form of ECT2, W521A/W534A, while ECT2_pro:ECT2-Flag/ect2-1 plants expressed a FLAG-tagged but otherwise unaltered ECT2 protein (FLAG tag) (Supplemental Figure 2I). In both lines, expression was driven by the native ECT2 promoter. Consistent with the conclusion that the m⁶A binding activity of ECT2 is required for normal morphogenesis, the ect2-1 mutant phenotype was restored by the expression of wild-type ECT2, but not by the expression of the m⁶A binding function-abolished form of ECT2, W521A/W534A (Figures 2A and 2B). Thus, the m⁶A site binding activity of the m⁶A reader protein ECT2 is required for normal trichome morphogenesis in Arabidopsis.

ECT2 Is Present in Both the Nucleus and Cytoplasm

As the subcellular localization of m⁶A readers is known to influence the types of regulatory mechanisms they utilize (Wang et al., 2014, 2015; Xiao et al., 2016; Roundtree et al., 2017), we characterized the subcellular localization of ECT2. Analysis of the ECT2 sequence using the TargetP (Emanuelsson et al., 2007) identified no obvious localization tags. We generated transgenic Arabidopsis plants (ECT2_pro:ECT2-eGFP/ect2-1) and analyzed cells in the root tip using confocal microscopy, finding that ECT2 is present in both the nucleus and cytoplasm. Overlay of ECT2-eGFP and 4′,6-diamidino-2-phenylindole-staining images indicated colocalization of the ECT2 fusion signal with the nucleus, but there was also extensive eGFP signal in the cytoplasm (Figure 3). We also used wild tobacco (Nicotiana benthamiana) for the transient expression of the ECT2-eGFP fusion protein and again found that ECT2 was present in both the nucleus and cytoplasm (Supplemental Figure 5). The subcellular localization of ECT2 suggested that ECT2 might play dual roles in regulating pre-mRNA processing in the nucleus and other aspects of mRNA metabolism involving mature transcripts in the cytoplasm.

Figure 3.

ECT2 Is Expressed in Both the Nucleus and Cytoplasm.

Confocal microscopy showing the subcellular localization of ECT2 in ECT2_pro:ECT2-eGFP/ect2-1 transgenic Arabidopsis root tips. ECT2 is found in both the nucleus and cytoplasm. Magnified images of cells in the white boxes in (A) are shown in (B), and further magnifications are shown in (C). Bars = 10 μm.

Development of the FA-CLIP Method Enables the Identification of Transcriptome-Wide ECT2-RNA Interaction Sites, Revealing an m⁶A Motif and 3′ UTR Enrichment

UV-cross-linking and immunoprecipitation (UV-CLIP) and photoactivatable ribonucleoside-enhanced cross-linking and immunoprecipitation (PAR-CLIP) methods have been widely used to map protein-RNA interaction sites in human cells (Ule et al., 2003; Hafner et al., 2010). Unlike the UV cross-linking used in UV-CLIP, PAR-CLIP involves the incorporation of 4-thiouridine into RNA for cross-linking. However, these methods have been ineffective in plants due to technical issues relating to cross-linking efficiency (Meyer et al., 2017). Formaldehyde cross-linking has been widely used to examine protein-DNA interactions in chromatin immunoprecipitation, protein-RNA interactions in plant FA-RIP, protein-protein interaction, and chromatin structure (Terzi and Simpson, 2009; Hoffman et al., 2015). Traditional FA-RIP used in plants can only identify the binding of transcripts to proteins and not protein-RNA interaction sites (Terzi and Simpson, 2009; Xing et al., 2015; Meyer et al., 2017). To enable the identification of the binding sites on transcripts targeted by ECT2, we developed a convenient FA-CLIP method that combines FA-RIP and CLIP. A conceptual diagram of the FA-CLIP method is presented in Figure 4A. The key innovation of FA-CLIP is the use of formaldehyde to cross-link proteins and RNA; this facilitates the chopping up of transcripts into protein binding regions and enables us to induce mutations using reverse transcription. In order to minimize the bias of formaldehyde cross-linking and to ensure the accuracy of our identified binding sites, our method uses IP enrichment and mutation to identify binding sites; by contrast, UV-CLIP and PAR-CLIP identify binding sites based only on mutation. We performed FA-CLIP to identify ECT2 binding sites (i.e., ECT2 binding peaks, termed FA-CLIP peaks) on transcripts using 14-d-old ECT2_pro:ECT2-Flag/ect2-1 and wild-type seedlings (termed FA-CLIP-ECT2 and Mock, respectively). Total RNA with rRNA depletion from ECT2_pro:ECT2-Flag/ect2-1 seedlings was used for RNA-seq (termed “Input”). The enrichment peaks were filtered using the criteria of enrichment fold (FA-CLIP-ECT2 [or Mock]/Input) ≥ 2 and P value < 0.01. IP enrichment-based ECT2 binding peaks (termed FA-CLIP enrichment peak) were obtained by subtracting the enrichment peaks in Mock (Mock versus Input) from the enrichment peaks in FA-CLIP-ECT2 (FA-CLIP-ECT2 versus Input). Mutation peaks were identified using PARAlyzer (Corcoran et al., 2011). The mutation-based ECT2 binding peaks (termed FA-CLIP mutation peaks) were identified as the mutation peaks in the FA-CLIP-ECT2 results while excluding the mutation peaks in Mock. ECT2 binding peaks (termed FA-CLIP peaks) were defined as overlapping peaks of FA-CLIP mutation peaks and FA-CLIP enrichment peaks.

Figure 4.

Transcriptome-Wide ECT2-RNA Interaction Sites Reveal a Unique Binding Motif and 3′ UTR Enrichment.

(A) Schematic diagram of the FA-CLIP method. FA-CLIP procedure: A cell extract from formaldehyde-cross-linked 14-d-old ECT2_pro:ECT2-Flag/ect2-1 seedlings was subjected to immunoprecipitation using anti-Flag M2 magnetic beads. Two on-bead digestions using RNase T1 and Protease K were performed sequentially: ECT2 protein-bound transcripts were digested into protein binding regions (∼30 nucleotides) using RNase T1, and ECT2 protein was digested by Protease K to leave an amino acid tag (that is, cross-linked residue) on ECT2-bound RNA fragments, which will induce mutation by reverse transcription. ECT2-bound RNA fragments were isolated, converted to cDNA (harboring mutations), and sequenced (termed FA-CLIP-ECT2). FA-CLIP of wild-type Col-0 (termed Mock) was also performed as described for FA-CLIP-ECT2. Total RNA with rRNA depletion from ECT2_pro: ECT2-Flag/ect2-1 seedlings was used for RNA-seq (termed Input). IP enrichment-based ECT2 binding peaks (termed FA-CLIP enrichment peaks) were obtained by subtracting the enrichment peaks in Mock (Mock versus Input) from the enrichment peaks in FA-CLIP-ECT2 (FA-CLIP-ECT2 versus Input). The mutation-based ECT2 binding peaks (termed FA-CLIP mutation peaks) are determined by subtracting the mutation peaks in Mock from the mutation peaks in FA-CLIP-ECT2. ECT2 binding peaks (termed “FA-CLIP peaks”) are defined as overlapping FA-CLIP mutation peaks and FA-CLIP enrichment peaks.

(B) Overlap of two biological replicates of the number of FA-CLIP peaks identifying ∼6000 high-confidence ECT2 binding sites (i.e., ECT2 targets) corresponding to 3680 unique transcripts/genes. Biological replicates are parallel measurements of biologically distinct samples.

(C) Pie chart presenting RNA types (that is, transcript species) of ECT2-bound transcripts.

(D) Overlap of the identified ECT2 binding peaks (FA-CLIP peaks) and m⁶A peaks.

(E) Pie chart presenting the fraction of ECT2 binding peaks in each of the three non-overlapping transcript segments (5′ UTR, coding sequence [CDS], and 3′ UTR).

(F) Metagene profiles of the distribution of ECT2 binding peaks along a normalized transcript composed of three rescaled nonoverlapping segments listed on the x axis.

(G) Binding motif identified by HOMER based on all identified ECT2 binding peaks (6047 peaks; P = 1 × 10⁻²¹⁵).

We conducted two biological replications (repeat experiments) of this FA-CLIP analysis and retained the FA-CLIP peaks common to both replications as high-confidence ECT2 binding peaks for subsequent analysis. Compared with the mutation peaks (6562 and 7111) in duplicate Mock samples, the mutation-based ECT2 binding peaks (termed “FA-CLIP mutation peak”) contained 37,614 and 37,694 peaks in two biological replications (Supplemental Figures 6A and 6B). By subtracting the enrichment peaks in Mock (20,794) from the enrichment peaks in FA-CLIP-ECT2 (34,679), 20,040 peaks were identified as IP enrichment-based ECT2 binding peaks (termed FA-CLIP enrichment peak) (Supplemental Figure 6C). Overlapping FA-CLIP mutation peaks with FA-CLIP enrichment peaks, the first replication identified ∼8700 FA-CLIP peaks and the second identified ∼10,400; ∼6000 peaks were common to both. These ∼6000 high-confidence ECT2 binding sites corresponded to 3680 unique transcripts/genes (Figure 4B; Supplemental Data Set 1). Almost all of the transcripts targeted by ECT2 were mRNA molecules, but ∼0.7% were noncoding RNA or other RNAs (Figure 4C). Moreover, when we looked at the overlap of the FA-CLIP-identified ECT2 targeted sites with the m⁶A sites identified from a separate m⁶A-seq analysis of poly(A)-tailed RNA from wild-type Col-0 seedlings (Duan et al., 2017), we found that 49.9% (3016 out of 6047) of ECT2-targeted peaks were modified with m⁶A (Figure 4D; Supplemental Data Set 2). Analysis of the distance between m⁶A and ECT2 binding sites revealed that m⁶A frequently overlaps with the binding site of ECT2 (Supplemental Figure 6D). These results confirm that ECT2 recognizes m⁶A on RNA transcripts in plants.

We then looked for other trends among the ECT2 targeted transcripts and found that the majority (90%) of ECT2-targeted sites occur within 3′ UTR of RNA transcripts (Figure 4E). We also systematically examined the distribution of the ECT2 binding sites along transcripts, revealing strong 3′ UTR enrichment of ECT2 target sites (Figure 4F). Clustering all ECT2 binding peaks using HOMER (Hypergeometric Optimization of Motif Enrichment) did not identify any previously reported signature m⁶A motifs in our data set (e.g., GGACU, and so on), but we did identify a strongly conserved motif among the ECT2 binding peaks: URUAY (R=G>A, Y=U>A, where the majority [over 90%] is UGUAY) (Figure 4G). Importantly, this motif was also clearly identified in an analysis of the ECT2 binding peaks common to both the m⁶A-seq and FA-CLIP data sets (Supplemental Figure 7). Collectively, we identified 3680 unique transcripts/genes containing over 6000 ECT2 binding sites, which were mainly located within 3′ UTRs and were clustered into the distinct binding motif, URUAY (R=G>A, Y=U>A, where the majority [over 90%] is UGUAY).

The URUAY Motif Is a Plant-Specific m⁶A Motif

Our FA-CLIP results showed that the ECT2 m⁶A reader recognizes a highly conserved motif, URUAY (R=G>A, Y=U>A, where the majority [over 90%] is UGUAY), that is distinct from the RRACH m⁶A motif that was characterized from other eukaryotes (Dominissini et al., 2012). To demonstrate that this URUAY motif is a genuine m⁶A site (that is, it can be methylated by an endogenous m⁶A writer), we performed in vitro methylation assays using cell extracts (nuclear fraction proteins) prepared from Arabidopsis seedlings. First, we randomly chose two m⁶A peaks that contained the UGUA motif from our m⁶A sequencing data to use as templates to build two RNA oligos: using splint ligation (Moore and Query, 2000), we introduced [γ-³²P] into the UGUA motifs, ensuring uniform distribution of [γ-³²P]-adenosine in these two RNA oligos (oligo 1 and oligo 2, see figure legends for sequences). Subsequently, we incubated these uniformly adenosine-labeled oligo substrates either with or without nuclear protein extracts for in vitro methylation reactions. Finally, we monitored the m⁶A modification status using nuclease P1 treatment and thin-layer chromatography (TLC) (Figure 5A).

Figure 5.

ECT2 Recognizes URUAY, a Plant-Specific 3′ UTR m⁶A Motif That Can Be Methylated by Arabidopsis Endogenous m⁶A Writer Proteins.

(A) TLC results from an in vitro methylation assay using nuclear extracts with site specific labeled substrates. pA indicates [γ-³²P]-labeled adenosine. Oligo 1, 5′-CUCGAUCCUUUUUGUpAGUUUCCGAC-3′; Oligo 2, 5′-UAUGCGUCUACUGUpACGGUUGAAUUU-3′.

(B) TLC results from an in vitro methylation assay using nuclear extracts with site-specific labeled oligo RNA substrates. The RNA probes were as follows, and pA indicates [γ-³²P]-labeled adenosine. UGUpA, 5′-CUCGAUCCUUUUUGUpAGUUUCCGAC-3′; GGpACU, 5′-CUCGAUCCUUUUGGpACUGUUUCCGAC-3′; CUpAUG, 5′-CUCGAUCCUUUUCUpAUGGUUUCCGAC-3′.

(C) Quantification of m⁶A/A ratios in (B), as calculated by densitometry using Image J.

(D) EMSA measuring the dissociation constant (K_d, nM) of GST-ECT2 with methylated and unmethylated RNA probes. The 4 nmol RNA probe was labeled with [γ-³²P], and GST-ECT2 concentration ranged from 10 to 2000 nM. Oligo RNA, 5′-AUGGGCCGUUCAUCUGCUAAAA(GGXCU/UGUXA/CUXUG) GCUUUUGGGGCUU*G*U-3′, X = A/m⁶A. The asterisk indicates that thiol-protected bases were used for the experiment.

These in vitro methylation assays with plant nuclear proteins clearly showed that an endogenous plant m⁶A writer(s) could methylate the UGUA sites of the two transcripts we had selected randomly from among the ECT2 binding sites common to both the m⁶A-seq and FA-CLIP data sets. To explore the in vitro methylation activity in greater detail, we built three additional RNA molecules containing uniformly [γ-³²P]-adenosine-labeled motifs: (1) UGUA, (2) GGACU, and (3) the randomly generated sequence CUAUG. Recall that the GGACU motif is the known m⁶A consensus sequence in humans and other eukaryotic organisms. Assays with these three RNA substrates revealed that the endogenous Arabidopsis m⁶A writers present in nuclear protein extracts have more efficient methylation activity for the UGUA sequence than for GGACU or the random sequence CUAUG (Figures 5B and 5C). Collectively, these results demonstrate that the UGUA sequence is a unique m⁶A motif that can be modified by endogenous Arabidopsis m⁶A writer(s).

Given that human YTH-domain family proteins can read m⁶A modified RRACH sites (especially the GGAC sequence) (Wang et al., 2014, 2015; Xiao et al., 2016), we next compared ECT2’s binding affinity toward the UGUA and GGACU motifs. We used EMSA assays with synthetic 42-mer RNAs to analyze the binding affinities of the ECT2 m⁶A reader protein with the URUAY motif, the previously characterized GGACU m⁶A motif, and the randomly generated sequence CUAUG. A total of six 42-mer RNAs were synthesized: all shared the same sequence, except for a central 5-nucleotide region comprising UGUXA, GGXCU, or the random CUXUG, where X represents an adenosine with or without the m⁶A modification. When we exposed these 42-mer RNAs to a range of concentrations of recombinant GST-ECT2 protein, we found that ECT2 only bound to the 42-mer RNAs harboring the m⁶A modification. ECT2 was able to recognize both methylated UGUAA and methylated GGACU with similar binding affinities; ECT2’s binding affinity for the m⁶A-modified random sequence was ∼3-fold lower than that for the two m⁶A target motifs (Figure 5D). Human YTH-domain proteins read the m⁶A-modified GGAC sequence; we found that the Arabidopsis YTH-domain family protein ECT2 can bind to the m⁶A-modified UGUA motif. ECT2’s binding affinity for m⁶A-modified UGUA is similar to human YTH-domain family proteins’ binding affinity for m⁶A-modified GGACU (Wang et al., 2014), suggesting that ECT2 may function through specific recognition of m⁶A-methylated UGUA in plant cells. Collectively, our results demonstrate that ECT2 reads the plant-specific m⁶A motif, URUAY, which is present in the 3′ UTRs of mRNA molecules in Arabidopsis.

ECT2 Is Involved in 3′ UTR Processing and mRNA Stability

For context, polyadenylation—the addition of a poly(A) tail to a mRNA molecule—is initiated by so-called poly(A) signals, which are divided into four different groups: cleavage elements at the poly(A) site, near upstream elements located at positions ∼20 to 30 nucleotides upstream of the poly(A) site, far upstream elements (FUEs) that occur in a window from ∼40 to 150 nucleotides upstream of the poly(A) site, and downstream elements located 20 to 40 nucleotides beyond the poly(A) site (Shen et al., 2008). Having established that ECT2 binds to the URUAY m⁶A motif (R=G>A, Y=U>A, where the majority [over 90%] is UGUAY), we noted that this UGUA sequence has been reported to be a poly(A) signal for polyadenylation (Shen et al., 2008; Yang et al., 2011; Elkon et al., 2013; Masamha et al., 2014). In Arabidopsis, this sequence typically occurs in FUEs [∼40–150 nucleotides upstream of the poly(A) site] (Loke et al., 2005; Shen et al., 2008).

To investigate whether the UGUA sites that are bound by the ECT2 reader are somehow related to poly(A) signal(s), we calculated the distance between poly(A) site and either the ECT2 binding sites (FA-CLIP peaks) or mutation peaks in Mock. Unlike in the Mock results, the positions of the ECT2 binding sites are consistently located ∼30 to 150 nucleotides upstream of the poly(A) sites in a region characteristic of FUE polyadenylation signals (Supplemental Figure 8A). Subcellular localization assays showed that ECT2 occurs in the nucleus and cytoplasm (Figure 3). We thus propose that ECT2 binds to the m⁶A-modifed poly(A) signal sequence UGUA in the FUEs of its target transcripts and thereby promotes their polyadenylation, suggesting that ECT2 might regulate alternative polyadenylation and 3′ UTR processing in the nucleus. ECT2 may selectively bind m⁶A-containing poly(A) signal FUEs, thereby recruiting the polyadenylation machinery to undertake alternative polyadenylation and further 3′ UTR processing (Supplemental Figure 8C).

Given that ECT2 is localized to both the nucleus and cytoplasm and that the cytoplasmic m⁶A reader YTHDF2 degrades m⁶A-modified transcripts in human, we thought it might be informative to examine whether ECT2 influences mRNA stability; we therefore performed mRNA sequencing (mRNA-seq) in wild-type Col-0 and ect2-1 seedlings (Supplemental Data Set 3). The transcripts in the data set were analyzed as reads per kilobase per million reads (RPKM), and any transcript with an RPKM value <1 was excluded. The filtered transcripts were classified into separate groups defined based on our previous experimental results from this study: the ECT2 FA-CLIP targets (i.e., ECT2 binding transcripts, 3554 genes), the FA-CLIP peaks that overlapped with m⁶A peaks from our m⁶A methylome analysis (FA-CLIP+m⁶A targets, 2007 genes), and the 1126 remaining genes that had RPKM values >1 but were not among the ECT2 binding transcripts from the FA-CLIP analysis (termed “non-targets”). Analysis of transcript accumulation in ect2-1 and wild-type plants revealed that the mutant plants had significantly reduced accumulation of the majority of ECT2 binding transcripts relative to non-targets. A similar decrease was observed among the majority of ECT2 binding transcripts containing m⁶A modifications (Supplemental Figure 8B). These exciting results directly suggest that ECT2 does not function as the same type of reader as human YTHDF2, which has been shown to promote mRNA degradation. Rather, it indicates that ECT2 might function in facilitating m⁶A-mediated mRNA stability in the cytoplasm (Supplemental Figure 8C). Given that the length of the 3′ UTR of a transcript affects the binding of microRNA and RNA binding proteins for mRNA degradation (Tian and Manley, 2017), we speculate that this observed increased ECT2 target mRNA abundance (or stability) could also be a consequence of ECT2’s role in promoting m⁶A-mediated alternative polyadenylation of 3′ UTRs (Supplemental Figure 8C). The detailed mechanisms remain to be investigated in the future.

ECT2 Functions in Important Biological Pathways, Including Trichome Morphogenesis

Our mRNA-seq data set (Supplemental Data Set 1) shows that there were 105 significantly upregulated genes and 92 significantly downregulated genes in the ect2-1 mutant (cutoff criteria of FPKM fold change ≥ 2 and P value < 0.05) (Supplemental Data Sets 4 and 5). To explore which cellular processes and signaling pathways ECT2 may be involved in, we used DAVID to performed Gene Ontology (GO) analysis of the ECT2-targeted genes identified by FA-CLIP analysis and the 197 differentially expressed genes identified by mRNA-seq analysis (Supplemental Figure 9). GO analysis revealed that the ECT2-targeted genes were positively enriched in several pathways, including mRNA processing, phosphorylation, responses to temperature stimulus, and trichome morphogenesis. GO analysis of differentially expressed genes in ect2-1 based on mRNA-seq revealed strong enrichment for terms including, among others, response to temperature stimulus and response to endogenous stimulus. These predictions about gene function were informative when viewed in the context of the ect2 mutant phenotypes.

ECT2’s m⁶A Binding Function Stabilizes Trichome Morphogenesis-Related Genes

Next, we further explored the mechanism underlying the abnormal trichome branching phenotype of ect2 plants. In the GO term analysis of ECT2 FA-CLIP binding transcripts, we found that 12 ECT2-targeted genes were related to trichome morphogenesis. We selected three such genes (TRANSPARENT TESTA GLABRA1 [TTG1], IRREGULAR TRICHOME BRANCH1 [ITB1], and DISTORTED TRICHOME2 [DIS2]), each of which contained overlapping ECT2 binding sites and m⁶A sites in their 3′ UTRs, as characterized in our FA-CLIP and m⁶A-seq data sets (Figure 6A). We performed FA-RIP-qPCR and m⁶A-IP-qPCR assays using 14-d-old seedlings to verify that these three genes did indeed (1) contain m⁶A sites in their 3′ UTRs and (2) were bound by ECT2 (Figures 6B and 6C), supporting the notion that our FA-CLIP and m⁶A-seq data were both accurate and robust. We then measured the expression levels of these three transcripts in 14-d-old seedlings. The expression levels of TTG1, ITB1, and DIS2 were significantly reduced in ect2-1 compared with the wild type (Figure 6D), which is consistent with the abnormal trichome morphology phenotype observed in the mutants. Taking into account the proposed functions of ECT2 (Supplemental Figure 8C), we reasoned that ECT2’s m⁶A binding function stabilizes TTG1, ITB1, and DIS2 transcripts. To investigate this possibility, we measured the lifetimes of these three transcripts by blocking transcription with actinomycin D. Transcription inhibition assays showed that TTG1, ITB1, and DIS2 transcripts were degraded more rapidly in ect2-1 plants compared with wild-type plants, whereas the mRNA lifetimes of negative control gene AT2G07689 in ect2-1 and the wild type were similar (Figure 6E). Collectively, we demonstrated that ECT2’s m⁶A binding function stabilizes TTG1, ITB1, and DIS2 transcripts, thereby affecting trichome morphogenesis (Figure 6F). The subcellular localization of ECT2 and high-throughput sequencing data analysis suggested that ECT2 might play roles in regulating 3′ UTR processing in the nucleus and directly facilitating mRNA stability in the cytoplasm (Supplemental Figure 8). The stabilization of TTG1, ITB1, and DIS2 transcripts may be a consequence of one of the aforementioned dual functions of ECT2. The exact pathway underlying how mRNA stability is affected by ECT2’s m⁶A binding function remains to be further investigated.

Figure 6.

ECT2’s m⁶A Binding Function Increases mRNA Stability of Three Trichome Morphogenesis-Related Genes.

(A) Three trichome morphogenesis-related genes are m⁶A methylated and are ECT2 targets.

(B) FA-RIP-qRT-PCR validation of the binding affinity of ECT2 to TTG1, ITB1, and DIS2. AT2G07689 is the internal control gene. Data are represented as means ± se, n = 2 biological replicates × 3 technical replicates.

(C) m⁶A-IP-qRT-PCR validation of the m⁶A peaks in TTG1, ITB1, and DIS2. AT2G07689 was used as the internal control gene. Data are represented as means ± se, n = 2 biological replicates × 3 technical replicates.

(D) Relative mRNA levels of TTG1, ITB1, and DIS2 in 14-d-old wild-type and ect2-1 Arabidopsis seedlings, with ACTIN2 as a reference gene. Data are represented as means ± se, n = 3 biological replicates × 2 technical replicates. **P < 0.01 and ***P < 0.001 by t test (two-sided).

(E) The mRNA lifetime of TTG1, ITB1, and DIS2 in the wild type and ect2-1. ECT2 non-target AT2G07689 was used as the negative control. Seven-day-old seedlings treated with actinomycin D were used for the transcription inhibition assays with 18S as the internal control gene. Data are represented as means ± se, n = 2 biological replicates × 3 technical replicates. TI, transcription inhibition.

(F) Proposed model describing how the m⁶A binding protein ECT2 regulates Arabidopsis trichome morphology.

DISCUSSION

Previous studies of m⁶A writers and erasers in plants revealed that the m⁶A modification plays fundamental roles in plant development (Zhong et al., 2008; Shen et al., 2016; Duan et al., 2017; Martínez-Pérez et al., 2017; Růžička et al., 2017). The characterization of m⁶A readers is an essential prerequisite for exploring and coming to understand the roles of m⁶A in plants; however, the identities of plant m⁶A readers have been unclear. Here, we showed that ECT2 is an m⁶A reader in Arabidopsis whose m⁶A binding function is required for normal trichome morphology (Figure 2). Previous studies have reported increased trichome branching in conditional knockout MTA plants and in plants overexpressing FIP37, suggesting that m⁶A is involved in regulating trichome morphology (Vespa et al., 2004; Bodi et al., 2012). This study deepens our understanding of how m⁶A affects trichome cell development by demonstrating the functional impact of the m⁶A reader activity of ECT2 on trichome branching.

We developed the convenient FA-CLIP method, which allowed us to identify ECT2-RNA interaction sites at the transcriptome-wide level (Figure 4A). Using these methods, we characterized transcriptome-wide ECT2 binding sites; our findings support the notion that ECT2 recognizes m⁶A-modified mRNA in plant cells (Figure 4D). Transcriptome-wide characterization of ECT2 binding sites revealed that ECT2 binding sites are dramatically enriched within the 3′ UTRs of mRNA transcripts and feature a highly conserved sequence motif: URUAY (R=G>A, Y=U>A, where the majority [over 90%] is UGUAY) (Figures 4E to 4G), which is distinct from the target motif of human YTH-domain family proteins (YTHDF1/YTHDF2/YTHDC1) (Wang et al., 2014, 2015; Xiao et al., 2016). Our results demonstrate that this UGUA motif can be methylated by endogenous Arabidopsis m⁶A writers and can be recognized by ECT2. ECT2’s binding affinity for m⁶A-modified UGUA is similar to the binding affinity of human YTH-domain family proteins for m⁶A-modified GGACU, suggesting that ECT2 may function through the specific recognition of m⁶A-methylated UGUA in plant cells. Our experimental evidence supports the notion that the m⁶A reader ECT2 binding motif UGUA sequence we identified in Arabidopsis is a plant-specific m⁶A consensus motif (Figure 5).

The subcellular localization of ECT2 in the nucleus and cytoplasm indicates that ECT2 has at least two functions. The UGUA sequence has been confirmed as a poly(A) signal for polyadenylation in many organisms (Graber et al., 1999). In Arabidopsis, the UGUA signal is located in FUEs that occur in a window from ∼40 to 150 nucleotides upstream of the poly(A) site (Loke et al., 2005). Our study identified the ECT2 binding motif UGUA and found the ECT2 binding sites are located ∼30 to 150 nucleotides upstream of poly(A) sites, strongly suggesting that ECT2 binds to poly(A) signals in FUEs and functions in polyadenylation (Supplemental Figure 8A). Analysis of the human m⁶A methylome and m⁶A writer subunit suggested that m⁶A functions in polyadenylation (Ke et al., 2015; Molinie et al., 2016; Yue et al., 2018), but the underlying mechanism and the alternative polyadenylation function-related m⁶A reader protein have not been clarified. Our experimental data suggest that the Arabidopsis m⁶A reader ECT2 functions in polyadenylation and 3′ UTR processing in the nucleus. The molecular mechanism of m⁶A-meditaed polyadenylation in plants may be unique and distinct from that in human or other eukaryotes. We proposed a model for a mechanism wherein ECT2 selectively binds to m⁶A-containing poly(A) signal FUEs, thereby recruiting the polyadenylation machinery to undertake alternative polyadenylation and further 3′ UTR processing (Supplemental Figure 8C). Furthermore, the finding that plants with a functioning ECT2 reader exhibit increased accumulation of ECT2-targeted transcripts suggests that ECT2 in the cytoplasm might promote mRNA stability (Supplemental Figure 8B). On the other hand, considering that alternative polyadenylation of 3′ UTRs is known to affect mRNA stability by affecting the binding of miRNAs and 3′ UTR binding proteins (Kedde et al., 2007; Chen, 2009; Tian and Manley, 2017), ECT2 in the nucleus might also influence mRNA stability through its possible role as a mediator of alternative polyadenylation and 3′ UTR processing (Supplemental Figure 8C).

Our mechanistic studies confirmed that mRNA stability affected by the m⁶A binding function of ECT2 controls trichome morphology. ECT2 bound to three trichome morphogenesis-related transcripts, TTG1, ITB1, and DIS2, containing m⁶A modifications. Disruption of ECT2 accelerated the mRNA degradation of TTG1, ITB1, and DIS2, thereby affecting trichome branching (Figure 6).

In some ways similar to the demonstrated role of histone methylation in the epigenetic regulation of the floral transition in Arabidopsis (Henderson and Dean, 2004), the epitranscriptomic modification m⁶A has been found to control embryo development, flowering time, and trichome morphology, among other processes (Zhong et al., 2008; Bodi et al., 2012; Duan et al., 2017). Our characterization of the function of the Arabidopsis m⁶A reader ECT2 in mRNA processing sheds light on the underlying mechanisms through which m⁶A functions in RNA metabolism specifically and more broadly in plant development and physiology.

METHODS

Plant Materials and Growth Conditions

The Arabidopsis thaliana genotypes used in this study include wild-type (Col-0) and two T-DNA insertion mutant lines, ect2-1 (SALK_002225) and ect2-2 (SAIL_11_D07), which were obtained from the Arabidopsis Biological Resource Center and confirmed as homozygotes (Supplemental Figures 2A and 2B). Transgenic plants (ECT2_pro:ECT2-Flag/ect2-1, ECT2_pro:ECT2m-Flag/ect2-1, ECT2_pro:ECT2-eGFP/ect2-1, and ECT2_pro:GUS) were obtained by transforming plasmids into either ect2-1 or wild-type (Col-0) plants. All Arabidopsis plants were grown on 0.5× Murashige and Skoog (1/2 MS) nutrient agar plates (PhytoTechnology Laboratories) for 14 d and then the seedlings were transfer to soil. The plants were grown under the following conditions: 16 h light/8 h dark at 22°C, and light intensities of 90 to 120 μE m⁻² s⁻¹ (provided by fluorescent tubes with a white 4100K spectrum purchased from Bainuo).

Plasmid Construction

Total RNA was extracted using TRIzol reagent (Thermo Scientific) and reverse transcribed into cDNA using SuperScript III reverse transcriptase (Thermo Scientific). The full-length ECT2 cDNA was amplified via PCR and cloned into pGEX-6p-1 between the BamHI and XhoI sites for protein expression and purification. Two residue sites were mutated (Supplemental Figures 3 and 4) in the YTH domain of ECT2 to generate a putative binding function-abolished form of ECT2, ECT2m (ECT2 W521A/W534A). Full-length ECT2 and ECT2m were cloned into a modified pCAMBIA1305 vector between the SalI and PstI sites. The pCAMBIA1305 vector was modified by inserting the ECT2 native promoter (2 kb upstream of ECT2) between the EcoRI and XbaI sites and by adding a 3×FLAG or eGFP tag sequence between the HindIII and BstEII sites. Thus, the P_ECT2-ECT2-Flag, P_ECT2-ECT2m-Flag, and P_ECT2-ECT2-eGFP constructs were obtained, and an ECT2 promoter GUS construct (P_ECT2-GUS) was generated to analyze ECT2 expression. ECT2 was cloned into pCAMBIA1300 (C-terminal eGFP between the HindIII and KpnI sites) between the SacI and HindIII sites to generate P_35S-ECT2-eGFP. Schematic representations of the constructs are shown in Supplemental Figures 2D to 2G. All constructs were confirmed by Sanger sequencing, and the primers used in their generation are shown in Supplemental Table 1.

Protein Expression and Purification

GST-ECT2 and GST-ECT2m proteins were purified using the same method. Plasmids were transformed into Escherichia coli strain BL-21 Gold competent cells. The E. coli cells were grown at 37°C to an OD₆₀₀ of 0.6 to 0.8, and recombinant protein expression was then induced with 0.3 mM IPTG. After 20 h of incubation at 16°C, the pellet from each 2-liter culture was collected, resuspended in 30 mL of lysis buffer (10 mM Tris-HCl, pH 8.0, 500 mM NaCl, 1 mM PMSF, 3 mM DTT, and 5% glycerol), and sonicated for 10 min. The sample was centrifuged at 13,000 rpm for 30 min, and the supernatant was filtered through a 0.45-μm filter membrane. The filtered supernatant was loaded onto a GST affinity column (GE Healthcare) that had been balanced with equilibrium buffer (10 mM Tris-HCl, pH 8.0, 500 mM NaCl, and 3 mM DTT). After washing the column with 30 mL of equilibrium buffer, the sample was eluted using 15 mL of elution buffer (10 mM Tris-HCl, pH 8.0, 500 mM NaCl, 10 mM reduced glutathione, and 3 mM DTT). The crude samples were further purified using a Superdex 75 size exclusion column (GE Healthcare) with storage buffer (10 mM Tris-HCl, pH 7.5, 200 mM NaCl, 3 mM DTT, and 5% glycerol). The purified protein was stored at −80°C.

Plant Transformation

Transformation of Arabidopsis was performed using the floral dip method (Clough and Bent, 1998) with Agrobacterium tumefaciens strain GV3101. The transgenic plants were screened on 1/2 MS medium containing hygromycin B and further identified by qPCR and protein immunoblotting. ECT2_pro:GUS transgenic plants were obtained by transforming the P_ECT2-GUS construct into 6-week-old wild-type Col-0 plants. The complementation lines ECT2_pro:ECT2-Flag/ect2-1, ECT2_pro:ECT2m-Flag/ect2-1, and ECT2_pro:ECT2-eGFP/ect2-1 were generated by transforming the P_ECT2-ECT2-Flag, P_ECT2-ECT2m-Flag, and P_ECT2-ECT2-eGFP constructs, respectively, into 6-week-old ect2-1 plants.

GUS Staining Assay

The GUS assay was performed using a staining kit (Coolaber). ECT2_pro:GUS transgenic plant tissues were incubated overnight at 37°C in staining buffer. After terminating the reaction, the tissues were cleared with 70% ethanol and imaged.

Gene Expression Analysis by RT-qPCR

Total RNA was extracted from 14-d-old Arabidopsis seedlings using TRIzol reagent, and the total RNA was treated with DNaseI (NEB). First-strand cDNA was synthesized using SuperScriptIII (Thermo Scientific) and random primers. qPCR was performed using Ultra SYBR Mixture with ROX (CWBIO) on a ViiA 7 Dx instrument (Applied Biosystems). All samples were analyzed in duplicate, and the relative expression levels were determined based on ACTIN2 as the internal control. The 2^−ΔΔCT method was used to calculate the gene expression levels. The qPCR primers used for ECT2 and ACTIN2 are listed in Supplemental Table 1; the qPCR primers used for the 12 other YTH-domain family genes were previously reported (Li et al., 2014) and are also listed in Supplemental Table 1.

Subcellular Localization

The root tips of 14-d-old transgenic Arabidopsis ECT2_pro:ECT2-eGFP/ect2-1 plants and the leaves of 3-week-old Nicotiana benthamiana plants transiently expressing P_35S-ECT2-eGFP (72 h after infiltration) were used to examine the subcellular localization of ECT2. The P_35S-ECT2-eGFP plasmid was transformed into Agrobacterium GV1301 cells. Cultures were grown at 28°C to OD₆₀₀ of ∼1.0. The cells were resuspended in infiltration buffer (10 mM MES, pH 5.6, 10 mM MgCl₂, and 100 μM acetosyringone) to adjust the OD₆₀₀ to 0.6 and infiltrated into 3-week-old N. benthamiana leaves. Confocal images were taken on an LSM 700 confocal laser scanning microscope (Zeiss).

Cryo-SEM Analysis of Arabidopsis Trichomes

Cryo-SEM was used to study the trichome branches of 3-week-old wild-type, ect2-1, ect2-2 ECT2_pro:ECT2/ect2-1, and ECT2_pro:ECT2m/ect2-1 Arabidopsis leaves (the third and the fourth leaves). The analyses were performed as previously described (Esch et al., 2004) with minor modifications. The equipment included the FEI Helios NanoLab G3 UC scanning electron microscope (Thermo Scientific) and the Quorum PP3010T workstation (Quorum Technologies), which had a cryo preparation chamber connected directly to the microscope. The Arabidopsis leaves were frozen in subcooled liquid nitrogen (−210°C) and transferred in vacuo to the cold stage of the chamber, where sublimation (−90°C, 5 min) and sputter coating (10 mA, 30 s) with platinum were conducted. Finally, the samples were transferred to another cold stage in the scanning electron microscope and imaged.

FA-CLIP

Fourteen-day-old wild-type and ECT2_pro:ECT2-Flag/ect2-1 seedlings grown on 1/2 MS plates were harvested and subjected to FA-CLIP to obtain the Mock and FA-CLIP-ECT2 sequencing data sets. Total RNA isolated from ECT2_pro:ECT2-Flag/ect2-1 seedlings using TRIzol reagent was subjected to rRNA removal with a RiboMinus Plant Kit for RNA-seq (Thermo Scientific). The total RNA with rRNA depletion (50 ng) was used to generate a library using a NEBNext Ultra RNA Library Prep Kit for Illumina kit (NEB) to produce the Input.

The FA-CLIP procedure was performed as follows: (1) Formaldehyde fixation and cross-linking. Col-0 (Mock) and ECT2_pro:ECT2-Flag/ect2-1 seedlings (FA-CLIP-ECT2) were fixed on ice for 30 min in 1% formaldehyde supplemented with PMSF under a vacuum. After removing the 1% formaldehyde solution, 125 mM glycine solution was used to quench the reaction for 5 min under a vacuum. The tissues were washed three times with precooled water, and all of the water was then removed by blotting the tissue with a paper towel. The tissues were frozen in liquid nitrogen and stored at −80°C or used directly for the next step in the protocol.

(2) Immunoprecipitation. Three grams of seedling material was ground into a fine powder in liquid N₂. Immediately after the samples turned dark green, they were combined with 3 mL of lysis buffer (150 mM KCl, 50 mM HEPES, pH 7.5, 2 mM EDTA, 0.5% Nonidet P-40 [v/v] 0.5 mM DTT, 2 mM EDTA, 1× cocktail protease inhibitor [Roche], and 40 units/mL RiboLock RNase Inhibitor [Thermo Scientific]). The samples were mixed well, incubated with rotation at 4°C for 20 min, and centrifuged at 15,000 rpm for 30 min at 4°C. The supernatant was filtered through a 0.22-μm membrane syringe. Agarose beads (100 μL; Promega) were washed twice with NT2 buffer (200 mM NaCl, 50 mM HEPES, pH 7.5, 2 mM EDTA, 0.05% Nonidet P-40, 0.5 mM DTT, 40 units/mL RiboLock RNase Inhibitor, and 1× cocktail protease inhibitor), added to the sample lysate, and incubated for 1 h at 4°C for prehybridization. The supernatant was collected as a precleared sample. The OD₂₆₀ of the precleared sample was measured by diluting 1 μL lysate in 500 μL ammonium acetate buffer. The total absorbance units were equal to 500-fold OD₂₆₀ multiplied by the total volume in milliliters. Turbo DNase (2000 units; Thermo Scientific) and 1000 units of RNase T1 (Thermo Scientific) were added per 25 absorbance units of the sample to partially digest the sample at 22°C for 15 min, and the samples were then placed on ice for 2 min. Anti-Flag M2 magnetic beads (100 μL; Sigma-Aldrich) were prewashed (4×) with 600 μL NT2 buffer. Col-0 and ECT2_pro:ECT2-Flag/ect2-1 samples were incubated with the washed beads separately and rotated continuously at 4°C for 4 h. The beads were collected and washed five times with 1 mL of ice-cold NT2 buffer and rotated continuously at 4°C for 5 min each time. The sample was resuspended into 400 μL NT2 buffer, followed by the addition of 10 units/μL RNase T1. The sample was incubated at 22°C for 15 min to digest the RNA to a size of ∼30 nucleotides. After incubation on ice for 5 min, the beads were washed (4×) with 500 μL of high-salt buffer (500 mM KCl, 50 mM HEPES, pH 7.5, 0.05% Nonidet P-40 [v/v], 0.5 mM DTT, 200 units/mL RNase inhibitor, and 1× cocktail protease inhibitor). The beads were resuspended into 400 μL 1×T4 PNK buffer prior to the next step.

(3) End repair. After adding 5 μL of T4 PNK (NEB), the reaction was incubated at 37°C for 1 h. Following the addition of 4 μL ATP (NEB) and 5 μL T4 PNK, the sample was incubated for an additional 30 min. The beads were washed (3×) with 800 μL of 1× T4 PNK buffer.

(4) Protease K digestion. The beads were resuspended into 200 μL of 1× protease K buffer (50 mM Tris-HCl, pH 7.5, 75 mM NaCl, 6 mM EDTA, and 1% [w/v] SDS), followed by the addition of 2 mg/mL protease K (NEB) and incubation at 50°C for 30 min.

(5) RNA recovery. The RNA was recovered by phenol/chloroform extraction followed by ethanol precipitation.

(6) Library construction and sequencing. The concentration of the recovered RNA was measured using a Qubit kit (Thermo Scientific). Fifty nanograms of RNA was used to generate the library using an NEB Next Multiplex Small RNA Library Prep Set for Illumina (NEB).

(7) After quality control and quantification, the libraries were sequenced on the Illumina HiSeq 2500 SR50 platform.

EMSA

EMSA was performed following a previously reported method (Wang et al., 2014) with minor modifications. (1) To assay the binding affinity of GST-ECT2 and GST-ECT2m, the following fluorescently labeled RNA oligonucleotides were used: 5′-FAM-UCUUUUGUXAGACUUGUACUCUUUA-3′, where X indicates either an A or an A with m⁶A modification. The RNA probe concentration was 4 nmol. The concentration of GST-ECT2 and GST-ECT2m ranged from 0 to 2000 nM. (2) For the motif recognition assay, [γ-³²P]-radiolabeled RNA probes were used; RNA with different motifs was synthesized with the sequence 5′-AUGGGCCGUUCAUCUGCUAAAA(GGXCU/UGUXA/CUXUG)GCUUUUGGGGCUUGU-3′, where X indicates either an A or an A with m⁶A modification. The final concentrations of GST-ECT2 used for EMSA ranged from 10 to 2000 nM for the UGUA and CUAUG oligonucleotides and from 10 to 500 nM for the GGACU oligonucleotides.

Preparation of Nuclear Extracts from Arabidopsis Seedlings

Nuclear fractions were extracted from 14-d-old Col-0 Arabidopsis seedlings as follows: Protoplasts were isolated using a previously reported method (Zhai et al., 2009) and separated into the cytoplasmic and nuclear fractions using a nuclear extraction kit (Abcam).

Splint Ligation to Generate a Site-Specific [γ-³²P]-Radiolabeled RNA Substrate

Splint ligation was performed to generate site-specific radiolabeled RNA substrates for in vitro methylation assays using a previously reported method (Moore and Query, 2000). Briefly, the 5′ end of a donor RNA oligo was phosphorylated with [γ-³²P]ATP in a 5-μL reaction containing 0.9 μL [γ-³²P]ATP (10 μCi/μL), 0.5 μL 10 PNK buffer, 1 μL 20 μM donor RNA, 1 μL T4 PNK (NEB), and 1.6 μL diethyl pyrocarbonate (DEPC) water. The reaction was incubated at 37°C for 2 h and heated at 75°C for 15 min to inactivate the kinase. Next, 1.5 μL of 10 ligation buffer, 1 μL of 20 μM acceptor, 1 μL of 20 µM DNA bridge, and 3.5 μL DEPC water were added, and the reaction was heated at 75°C for 2 min. The samples were cooled slowly to room temperature. One microliter of 10 mM cold ATP and 2 μL of T4 DNA ligase (NEB) were then added to the samples, which were incubated for an additional 4 h at 30°C. The DNA bridge was digested with 1 μL of RQ1 RNase-free DNase (Promega) at 37°C for 30 min. Finally, the samples were incubated at 75°C for 20 min to inactivate the DNase. The resulting site-specific radiolabeled RNA substrates were used in the in vitro methylation assays.

In Vitro Methylation Assay

Each 50-μL methylation reaction contained 10 μL 5× reaction buffer (75 mM HEPES, pH 7.5, 20% glycerol, 250 mM KCl, 250 mM NaCl, 5 mM MgCl₂, 2.5 mM DTT, and 100 μM ATP), 5 μL 32 mM SAM (NEB), 1 μL RNase inhibitor (40 units/µL), 15 μL nuclear extract (1.1 mg/mL), 1 μL site-specific radiolabeled RNA substrates, and 16.5 μL DEPC water. The reaction was incubated at 37°C for various times ranging from 5 min to 1 h. The reactions were quenched at 75°C for 20 min. Ten microliters of methylation mixture (as well as a 5′ radiolabeled m⁶A control RNA [9-mer RNA oligo with a m⁶A at the 5′ terminus]) was added to the samples, along with 1 μL Nuclease P1 (Wako) and 1 μL 0.1 M NH₄Ac. The samples were digested at 42°C for 2 h. After the digestion, 0.5-μL sample aliquots were loaded on cellulose TLC plates (Merck) and separated for 28 h in a solution of isopropanol:HCl:water (70:15:15 [v/v]) . The plate was air-dried in a fume hood at room temperature, and radioactive signal detection and quantification were performed with a Typhoon FLA 9500 phosphor imager (GE Healthcare).

In Vitro RIP-LC/MS/MS

In vitro pull-down assays were performed as previously reported (Wang et al., 2014), with the following modifications: RNA was extracted from 14-d-old Arabidopsis seedlings, and 0.2 μg of mRNA was used as the Input sample. mRNA (0.8 μg) and GST-ECT2 (final concentration 500 nM) were diluted in 200 μL binding buffer (150 mM NaCl, 0.1% Nonidet P-40, 10 mM Tris-HCl, pH 7.4, 40 units/mL RNase inhibitor, and 0.5 mM DTT). The protein-RNA solution was mixed and rotated at 4°C for 2 h. Ten microliters of GST-affinity magnetic beads (Pierce) were resuspended in 50 μL of binding buffer after being washed four times with 200 μL of binding buffer. The solution was combined with the beads and rotated for an additional 2 h at 4°C. The aqueous phase was collected by ethanol precipitation and the pellet was dissolved in 50 μL RNase-free water. The recovered fraction was saved as the flow-through sample. The beads were washed three times with 200 μL of binding buffer, and 500 μL of TRIzol reagent was then added. The purified RNA was saved as the ECT2-bound sample. LC-MS/MS was used to measure the level of m⁶A in the input, flow-through, and ECT2-bound samples.

LC-MS/MS for m⁶A Quantification

RNA (100 ng) was digested with 1 unit of Nuclease P1 in 50 μL of buffer containing 10% 0.1 M ammonium acetate (pH 5.3) at 42°C for more than 3 h, followed by the addition of 1 unit of shrimp alkaline phosphatase (NEB) and 10% Cutsmart buffer. The mixture was incubated at 37°C for an additional 3 h. The samples were then centrifuged at 15,000 rpm for 30 min and the aqueous phase was injected into an LC-MS/MS system. Nucleosides were separated using a UPLC pump (Shimadzu) with a ZORBAX SB-Aq column (Agilent) and analyzed by MS/MS using a Triple Quad 5500 mass spectrometer (AB SCIEX) running in positive ion mode and the multiple reaction monitoring feature. MS parameters were optimized for m⁶A detection. Nucleosides were quantified using the nucleoside-to-base ion mass transitions of m/z 268.0 to 136.0 (A), m/z 282.0 to 150.1 (m⁶A), m/z 244.0 to 112.0 (C), m/z 284.0 to 152.0 (G), and m/z 245.0 to 113.1 (U). Standard curves were generated using a concentration series of pure commercial nucleosides (Sigma-Aldrich) analyzed using the same method. Concentrations of nucleosides and m⁶A/A ratio in samples were calculated by fitting the signal intensities to the standard curves (Wang et al., 2014; Duan et al., 2017).

RNA-Seq

Fourteen-day-old Col-0 and ect2-1 seedlings were collected and ground in liquid nitrogen. Total RNA was extracted by adding TRIzol to the ground samples. mRNA was then isolated using Dynabead oligos (dT)₂₅ (Thermo Scientific). One hundred nanograms of the extracted mRNA was used as the template for library construction using a NEBNext Ultra RNA Library Prep Kit for Illumina kit (NEB).

In Vivo FA-RIP-LC/MS/MS and in Vivo FA-RIP-qPCR

For in vivo FA-RIP, most of the procedures were same as those used for the FA-CLIP method, but without the RNase T1 digestion step. The resulting input and IP RNA were separated into two parts: one part was subjected to LC-MS/MS to measure the m⁶A level and the other was reverse transcribed into cDNA and analyzed via qPCR to measure the enrichment levels. AT2G07689 (encoding NADH-Ubiquinone/plastoquinone [complex I] protein) was used as an internal control, since (1) AT2G07689 mRNA did not show any obvious m⁶A peak from m⁶A-seq data; (2) AT2G07689 mRNA was not enriched by ECT2 from the FA-CLIP data; (3) AT2G07689 showed relatively invariant expression levels between Col-0 and ect2 plants; and (4) AT2G07689 is considered to be a housekeeping gene. Samples were performed in 2 biological replicates × 3 technical replicates.

m⁶A-IP-qPCR

m⁶A-IP-qPCR was performed as previously reported (Duan et al., 2017). Briefly, 14-d-old wild-type seedlings were used for the IP assays. Five micrograms of mRNA was used for each m⁶A-IP sample. mRNA was fragmented into ∼200-nucleotide molecules with RNA fragmentation reagents (NEB) and incubated with 5 μg m⁶A antibody (Synaptic Systems) for 4 h at 4°C. The m⁶A-containing fragments were immunoprecipitated with preblocked Protein A Dynabeads (Thermo Scientific) and eluted with 7 mM m⁶A nucleoside-containing solution. After ethanol precipitation, the m⁶A-bound fraction RNA, as well as input mRNA, were subjected to reverse transcription and qPCR assays using AT2G07689 as the internal control gene. Samples were performed in 2 biological replicates × 3 technical replicates.

mRNA Stability Measurements

An mRNA stability measurement assay was performed as previously described (Duan et al., 2017) with minor modification. Briefly, 7-d-old wild-type and ect2-1 Arabidopsis seedlings grown on 1/2 MS medium were transferred to 10-cm Petri dishes containing 10 mL 1/2 MS liquid medium. After 30 min incubation, 0.2 mM actinomycin D was added to the buffer. The tissues were collected at 30 min after the transcription inhibitor was added; these samples are referred to as 0 h samples. The 3, 6, and 8 h samples were collected and immediately frozen in liquid nitrogen. The tissues were stored at −80°C or subjected to total RNA extraction. cDNAs were generated with SuperScript IV reverse transcriptase (Thermo Scientific) using the oligo d(T) primer. mRNA levels were quantified by RT-qPCR with gene-specific qPCR primers (Supplemental Table 1). 18S RNA was used as the internal control, and the primers for 18S were designed as previously reported (Cao et al., 2016). Samples were performed in 2 biological replicates × 3 technical replicates.

Sequencing Data Processing

RNA-Seq Data Processing

After adapter trimming with Cutadapt (Martin, 2012), the reads were mapped to TAIR10 (Lamesch et al., 2012) using TopHat (Trapnell et al., 2009), and RPKM values were calculated using Cufflink (v2.2.1) (Trapnell et al., 2010). The differentially expressed genes between ect2-1 and the wild type were defined based on a cutoff criterion of FPKM fold change ≥ 2 and P value < 0.05. Gene annotations were downloaded from the Ensembl plants database (TAIR10 release 31).

FA-CLIP Data Analysis

After adapter trimming with Cutadapt, the reads were mapped to TAIR10 using Bowtie2 (Langmead and Salzberg, 2012). ECT2 binding peaks (termed FA-CLIP peaks) were defined as overlapping peaks of FA-CLIP mutation peaks and FA-CLIP enrichment peaks.

Mutation Peaks

Mutation peaks were calculated using PARalyzer v1.1 with default settings, except that a mutation peak contained at least two mutation sites instead of only one mutation site and all types of mutations were taken into account rather than only T-to-C conversion (Corcoran et al., 2011). The mutation-based ECT2 binding peaks (termed FA-CLIP mutation peaks) were obtained by subtracting the mutation peaks in Mock from the mutation peaks in FA-CLIP-ECT2.

PARalyzer is commonly used to generate a high-resolution map of high-confidence RNA-protein interaction sites from CLIP deep-sequencing databased on patterns of nucleotide mutations coupled with read density. The parameters used are as follows: bandwidth = 3; minimum_read_count_per_group = 10; minimum_read_count_per_cluster = 5; minimum_read_count_for_KDE = 5; minimum_cluster_size = 10; minimum_conversion_locations_for_cluster = 2; minimum_conversion_count_for_cluster = 1; minimum_read_count_for_cluster_inclusion = 5; minimum_read_length = 13; minimum_number_of_non_conversion_mismatches = 1; and additional_nucleotides_beyond_signal = 5.

Enrichment-Based Peak

Peak calling was performed as previously reported using the same criteria with enrichment fold (FA-CLIP-ECT2 [or Mock]/Input) ≥ 2 and FDR < 0.01, and the only difference was that the aligned reads were not extended due to the small average fragments size (Ma et al., 2017). IP enrichment-based ECT2 binding peaks (termed FA-CLIP enrichment peak) were obtained by subtracting the enrichment peaks in Mock (Mock versus Input) from the enrichment peaks in FA-CLIP-ECT2 (FA-CLIP-ECT2 versus Input).

Gene annotations were downloaded from the Ensembl plants database (TAIR10 release 31). HOMER (Heinz et al., 2010) was used for motif identification. DAVID (Huang et al., 2007) was used to performed GO analysis of the ECT2-targeted genes that were identified in the FA-CLIP analysis and the 197 differentially expressed genes identified in the mRNA-seq analysis. The top 40 GO items for the ECT2-targeted genes were selected for visualization using the interactive graph function of REVIGO (Supek et al., 2011).

Statistical Analysis

Statistical analysis was performed by one-way ANOVA followed by LSD post-hoc tests using SPSS 20.0 (IBM) (see Supplemental File 1 for ANOVA tables).

Accession Numbers

Sequence data in this study can be found under the following accession numbers: ECT1, AT3G03950; ECT2, AT3g13460; ECT3 At5g61020; ECT4, At1g55500; ECT5, At3g13060; ECT6, At3g17330; ECT7, At1g48110; ECT8, AT1G79270; ECT9, At1g27960; ECT10, AT5G58190; ECT11, At1g09810; ECT12, At4g11970; CPSF, At1g30460; TTG1, AT5G24520; ITB1, AT2G38440; DIS2, AT1G30825; and NADH-Ubiquinone/plastoquinone (Complex I) protein, AT2G07689. All raw high-throughput sequencing data have been submitted to Gene Expression Omnibus under accession number GSE108119.

Supplemental Data

• Supplemental Figure 1. Relative expression levels of YTH domain family genes in Arabidopsis seedlings.

• Supplemental Figure 2. ect2 mutants and transgenic lines.

• Supplemental Figure 3. Sequence alignment of YTH domain family proteins in Arabidopsis and human.

• Supplemental Figure 4. Binding affinity of GST-ECT2 and GST-ECT2m to m⁶A methylated RNA probe.

• Supplemental Figure 5. Subcellular localization of ECT2 based on transient expression of 35S_pro:ECT2-eGFP in Nicotiana benthamiana.

• Supplemental Figure 6. ECT2 binding peaks identified by FA-CLIP.

• Supplemental Figure 7. ECT2 binding motif identified by HOMER.

• Supplemental Figure 8. ECT2 is involved in 3′ UTR processing and mRNA stability.

• Supplemental Figure 9. GO analysis of ECT2-related transcripts.

• Supplemental Table 1. Primers and oligonucleotide probes used in this study.

• Supplemental Data Set 1. FA-CLIP binding sites of ECT2.

• Supplemental Data Set 2. FA-CLIP peaks of ECT2 overlapping with those identified by m⁶A-seq.

• Supplemental Data Set 3. RNA-seq of wild-type Col-0 and ect2-1.

• Supplemental Data Set 4. Downregulated genes in ect2-1 compared with the wild type revealed by RNA-seq.

• Supplemental Data Set 5. Upregulated genes in ect2-1 compared with the wild type revealed by RNA-seq.

• Supplemental File 1. ANOVA tables.