https://orcid.org/0000-0001-7532-0138
Abstract
Throughout all kingdoms of life, ribonucleotides are marked with covalent chemical modifications that change the structure and binding properties of modified RNA molecules. These marks are deposited by ‘writer’ proteins, recognized by ‘readers’, and removed by ‘erasers’, thus forming an epitranscriptomic system of marks and binding proteins directly analogous to the epigenome. Recent advances in marrying classical biochemical techniques with high-throughput sequencing have enabled detailed mapping of plant epitranscriptomic marks, which in turn yielded insights into how these marks regulate a host of biological processes, from shoot stem cell fate to floral transition and from leaf development to viral activity. In this review, we highlight recent developments in the study of plant epitranscriptomics, with an emphasis on N6-methyladenosine (m6A) and 5-methylcytosine (m5C). These studies have advanced the field beyond descriptive mapping or isolated genetic studies, and produced a more nuanced understanding of how components of the epitranscriptome and their binding proteins directly regulate critical aspects of plant biology.
Introduction
Over 100 types of covalent, post-transcriptional chemical modifications are known to decorate RNAs across all kingdoms of life and viruses (Limbach et al., 1994; Dunin-Horkawicz et al., 2006; Cantara et al., 2011; Machnicka et al., 2013). These marks alter the shape and sometimes the charge of their respective nucleotides and can thus alter base pairing and elicit changes in how a modified RNA molecule interacts with RNA-binding proteins (RBPs). As a result, modifications within mRNA are potent post-transcriptional regulators that affect multiple points of a transcript’s life cycle, including splicing and maturation (Zhao et al., 2014; Haussmann et al., 2016; Xiao et al., 2016), export into the cytoplasm (Fustin et al., 2013; Zheng et al., 2013), translation (Meyer et al., 2015; Wang et al., 2015; Choi et al., 2016), and RNA stability (Du et al., 2016; Shen et al., 2016; Mauer et al., 2017; Zhao et al., 2017b; Wei et al., 2018).
Here, we review recent studies in plants that link modification-induced transcript regulation to important biological outcomes, including viral infection (Martínez-Pérez et al., 2017), leaf morphogenesis (Arribas-Hernández et al., 2018; Scutenaire et al., 2018; Wei et al., 2018), shoot apical meristem (SAM) maintenance (Shen et al., 2016), floral transition (Duan et al., 2017), and root development (Cui et al., 2017; David et al., 2017) (Box 1). Importantly, these studies move beyond descriptive mapping by linking changes in modifications at specific transcripts (e.g. regulators of trichome development) to specific outcomes (e.g. trichome spike number), thus building a more nuanced view of the precise mechanisms by which RNA modifications function.
Known regulatory functions and corresponding target mRNAs of readers, writers, and erasers of the m6A and m5C epitranscriptome marks are plotted in relation to an adult, flowering Arabidopsis plant. Furthermore, the resulting plant phenotypes that have been observed upon the genetic ablation of these factors required for binding, deposition, and removal of these RNA covalent modifications are described. See Box 2 for citations of supporting studies. The image is adapted from the Database Center for Life Science (DBCLS), (https://commons.wikimedia.org/wiki/File:Arabidopsis_thaliana_by_togopic.png).
The epitranscriptome and m6A
Like their epigenetic counterparts in DNA, RNA modifications are an additional layer of information that is deposited by writer proteins and recognized by reader proteins that effect downstream functions. In some cases, RNA modifications can also be removed by eraser proteins, or alternatively may be diluted out through RNA decay. Thus, RNA chemical modifications form an ‘epitranscriptomic’ (Meyer et al., 2012; Saletore et al., 2012) regulatory network that has only recently been elucidated.
N6-Methyladenosine (m6A) is the most abundant and physiologically relevant mRNA modification and is currently the best example of a complete epitranscriptomic system with known writers, readers, and eraser proteins in both plants and mammals (Liu and Pan, 2016; Roundtree and He, 2016; Zhao et al., 2017a; Patil et al., 2018). m6A was first detected in populations of animal mRNAs using chromatographic methods (Dubin and Taylor, 1975; Perry et al., 1975), and subsequent studies have shown it to be a ubiquitous mark in viruses (Beemon and Keith, 1977; Canaani et al., 1979), bacteria (Deng et al., 2015), fungi (Bodi et al., 2010), and plants, including Arabidopsis, oat, wheat, and maize (Kennedy and Lane, 1979; Nichols, 1979; Haugland and Cline, 1980; Zhong et al., 2008). Recent advances in pairing antibody pulldowns with high-throughput sequencing have enabled mapping of m6A at peak (~100 nucleotides) (Dominissini et al., 2012; Meyer et al., 2012; Luo et al., 2014) and single nucleotide resolution (Liu et al., 2013; Linder et al., 2015), leading to rapid advances in our understanding of m6A-mediated post-transcriptional regulation. Within both Arabidopsis and mammals, these studies revealed m6A enrichment over mRNA stop codons and 3'-untranslated regions (UTRs), suggesting a role in modulating transcript stability and translation (Dominissini et al., 2012; Meyer et al., 2012; Luo et al., 2014). Luo and colleagues also observed plant-specific enrichment over the start codon, though work in our lab (data not shown) and that of others (Shen et al., 2016) has not been able to replicate these observations. m6A mapping has yet to be applied to crop species, and thus our review will primarily cover m6A in Arabidopsis, followed by a brief discussion of other modifications detected using 5-methylcytosine (m5C) mapping (Cui et al., 2017; David et al., 2017) and in silico methods (Vandivier et al., 2015).
Studies of m6A writers reveal regulation of plant development
m6A is deposited by a large, multicomponent complex that was first characterized in mammals and includes METTL14 and METTL3 methyltransferases, as well as WTAP and KIAA1429 cofactors (Liu et al., 2014; Ping et al., 2014; Schwartz et al., 2014,b). Arabidopsis contains functional orthologs including MTB and MTA (orthologs of METTL14 and METTL3, respectively) (Zhong et al., 2008; Bodi et al., 2012), FIP37 (ortholog of WTAP) (Shen et al., 2016), and VIRILIZER (ortholog of KIAA1429) (Růžička et al., 2017). The m6A writer complex in Arabidopsis also contains the E3 ubiquitin ligase HAKAI, which to date appears to be plant specific and raises the intriguing possibility that mRNA modification and protein decay are linked (Růžička et al., 2017).
Even before the development of m6A sequencing approaches, genetic studies clearly demonstrated a role for m6A writer proteins in important processes in plant cells. Specifically, complete loss of MTA leads to early embryonic lethality (Zhong et al., 2008) and, consistent with its necessity for development, MTA expression is high within shoot meristems (including the SAM), lateral roots, and floral tissue (Zhong et al., 2008). Consistently, strong developmental phenotypes have been observed in hypomorphs of other writer complex proteins. For example, complete loss of FIP37 is embryonic lethal, while partial loss leads to overproliferation of the SAM (Shen et al., 2016). Likewise, partial loss of VIRILIZER leads to defects in lateral root emergence, gravitropism, and cotyledon development (Růžička et al., 2017). The robust, pleiotropic phenotypes of m6A writer protein mutants mirror the host of developmental defects observed in animal and yeast writer mutants (Hongay and Orr-Weaver, 2011; Schwartz et al., 2013; Batista et al., 2014; Geula et al., 2015) and strongly indicate that m6A is an important developmental regulator.
Combining genetic studies with m6A sequencing pushed the field forward by providing one of the first examples in plants of a detailed molecular mechanism by which this regulation can occur. Shen and colleagues identified transcripts encoding two major regulators of the SAM, WUSCHEL (WUS) and SHOOTMERISTEMLESS (STM), as containing FIP37-dependent m6A foci (Shen et al., 2016). In the fip37 hypomorph, these two regulators are both more stable and more abundant, suggesting that loss of methylation writers may trigger SAM overproliferation via stabilization of these important regulators. Furthermore, transient up-regulation of FIP37 leads to re-methylation and clearance of excess WUS and STM, and mutants of wus and stm both suppress the fip37 phenotype (Shen et al., 2016), providing clear evidence that methylation of the transcripts encoding key SAM regulators is responsible for correct plant development (Box 1).
m6A readers regulate leaf and trichome morphology
m6A marks modulate both secondary structure and RBP interaction, and are read either ‘directly’ by reader proteins that contain methyl-binding aromatic pockets (Luo and Tong, 2014; Xu et al., 2014), or ‘indirectly’ by relaxing secondary structure to reveal binding sites for single-stranded RBPs (Liu et al., 2015). Most work in Arabidopsis has focused on direct reading through YTH-domain containing proteins (YTHDs), though recent work in our lab suggests that m6A-induced structural rearrangements are in fact a major regulator of mRNA stability (data not shown).
YTHDs have been studied extensively in animals, and have been shown to direct a wide array of regulatory functions including associating with methylated Xist mRNA to direct X inactivation (Patil et al., 2016), destabilizing transcripts by recruiting the CCR-NOT deadenylase (Du et al., 2016), directing transcripts to P-bodies (Wang et al., 2014), and facilitating translation by recruiting initiation factors (Wang et al., 2015). Arabidopsis YTH homologs were first characterized by their highly conserved C-termini, and were thus named EVOLUTIONARILY CONSERVED C-TERMINAL REGIONs (ECTs) (Ok et al., 2005), though only recently have the functions of YTHDs been elucidated in plants.
Three concurrent publications in The Plant Cell demonstrate a role for ECTs in regulating leaf morphology and trichome development (Arribas-Hernández et al., 2018; Scutenaire et al., 2018; Wei et al., 2018) (Box 2). Taken together, these studies revealed that the loss of ECT2 (Arribas-Hernández et al., 2018; Scutenaire et al., 2018; Wei et al., 2018) as well as the paralog ECT3 (Arribas-Hernández et al., 2018) led to defects in leaf development and an increase in the number of trichome spikes (Box 1). Notably, trichome spike number is linked to ploidy and, in agreement, ect2 mutants display increased numbers of endoreduplication events (mitosis without cytokinesis) (Scutenaire et al., 2018), raising the possibility that m6A is an important regulator of cell cycle progression in plants just as has been demonstrated in yeast (Schwartz et al., 2013) and mammals (Horiuchi et al., 2006; Zhao et al., 2017b). Moreover, endoreduplication plays an important role in root hair cell formation, raising the question of whether m6A and its readers regulate root hair density and nutrient uptake.
m6A-binding proteins are required for proper trichome development
Three recent studies (Arribas-Hernández et al., 2018; Scutenaire et al., 2018; Wei et al., 2018) have demonstrated that two ECT proteins, ECT2 and ECT3, read m6A, including on transcripts involved in trichome development and endoreduplication. Loss of ECT2/3 or ablation of their m6A recognition domain leads to defects in leaf development and endoreduplication. Wei et al. (2018) further show that ECT2 binds a plant-specific URUAY motif.
m6A writer proteins affect plant developmental processes
Růžička et al. (2017) identified VIRILIZER and HAKAI as components of the m6A methyltransferase complex. vir mutants have strong, pleiotropic defects, while hakai mutants have a milder phenotype. Additionally, Shen et al. (2016) revealed that FIP37 is required for methylation and destabilization of transcripts encoding the shoot apical meristem regulators WUS and STM, preventing the SAM from overproliferating.
m6A eraser proteins regulate plant flowering time
Duan et al. (2017) found that ALKBH10B demethylates numerous mRNAs, which can result in their stabilization. This study found that ALKBH10B demethylation target mRNAs include those encoding the master flowering regulator FT as well as SPL3 and SPL9, preventing delayed flowering.
m5C is required for proper plant root development
Cui et al. (2017) provided a transcriptome-wide map of m5C sites across the Arabidopsis transcriptome. Additionally, this study revealed that TRM4B-dependent m5C stabilizes transcripts encoding the root regulators SHY2 and IAA16, which probably explains why the loss of this m5C methyltransferase (trm4b mutant plants) leads to defects in root elongation.
Numerous epitranscriptome marks are found in plant mRNAs
Vandivier et al. (2015) found that novel epitranscriptome marks including m3C and modified internal guanosines are associated with unstable transcripts in the Arabidopsis transcriptome.
Since ECT proteins are direct readers, all three studies utilize mutations in the ECT2/3 aromatic cage in order to show m6A binding dependence of leaf and trichome phenotypes (Arribas-Hernández et al., 2018; Scutenaire et al., 2018; Wei et al., 2018). Wei and colleagues go a step further and identify a novel URUAY motif that is both plant specific and associated with transcript stabilization, in contrast to observed m6A-mediated destabilization in animals (Wei et al., 2018). Notably, three transcripts encoding proteins involved in trichome development (TTG1, ITB1, and DIS2) contain methylated versions of this motif and are destabilized upon loss of ECT2 (Wei et al., 2018), providing a direct mechanism for the observed trichome phenotypes.
m6A erasers target viruses and regulate floral transition
Like DNA methylation, RNA m6A can be removed by active demethylases (erasers) or by passive ‘dilution’ (rounds of RNA decay and transcription). m6A erasers in animals include AlkB and AlkB-homology (AlkBH) family proteins such as FTO (Jia et al., 2011) and ALKBH5 (Zheng et al., 2013), and both appear to function primarily in the nucleus. Relatedly, genetic ablation of m6A erasers leads to altered RNA splicing and export (Zheng et al., 2013; Zhao et al., 2014).
Arabidopsis contains multiple ALKBH family proteins, two of which have been shown to be physiologically relevant. Martínez-Pérez and colleagues show that the Arabidopsis homolog of ALKBH9B interacts with viral RNAs, and facilitates viral infection through a yet to be elucidated mechanism (Martínez-Pérez et al., 2017). Duan and colleagues show that ALKBH10B primarily demethylates mRNA in vivo, and plants that lack this demethylase (alkbh10b mutants) display delayed flowering. This phenotype can be explained by destabilization of the transcript encoding the master flowering regulator FLOWERING LOCUS T (FT) as well as SPL3 and SPL9 in alkbh10b mutant plants (Duan et al., 2017) (Boxes 1, 2). This indicates that m6A can also destabilize mRNAs, which is consistent with observations in our lab (data not shown) and from studies of ECT2 (Wei et al., 2018) suggesting that m6A can have opposing effects on mRNA stability depending on the specific context in which this mark is found in the parent transcript. More specifically, the exact mechanism(s) by which certain m6A marks are stabilizing and others are destabilizing remains an open question, though one logical model for these differential effects may depend on the specific subtranscript localization of the m6A mark in parent transcripts. In this model, mRNA coding sequence (CDS) marks could trigger ribosome stalling and subsequent RNA degradation, thereby destabilizing the parent transcript, while 3'-UTR marks could relieve secondary structure and prevent transcript cleavage by double-stranded endonucleases. This model will need further testing in future studies.
An early glimpse into m5C and root development
m5C has been extensively studied in the context of DNA, and thus has well established methods such as bisulfite conversion followed by sequencing (bisulfite-seq) for quantitative mapping at single nucleotide resolution that are readily applicable to RNA (Schaefer et al., 2009; Hussain et al., 2013). In plants, sequence context-based prediction of m5C has also been successfully applied to mapping m5C (Song et al., 2018). Despite the abundance of mapping techniques, assigning functional relevance to mRNA m5C has been challenging given that m5C writers such as the yeast tRNA:m5C methyltransferase (Trm4) and homologs in animals and plants generally target both mRNA and tRNAs and thus their mutation produces highly pleiotropic phenotypes (Brzezicha et al., 2006; Goll et al., 2006; Schaefer et al., 2009). That being said, epitranscriptomic mapping of m5C has revealed intriguing links between m5C and root development (Box 1).
For instance, David and colleagues applied bisulfite-seq to map m5C sites in Arabidopsis and identify modified transcripts in a variety of tissue types including roots, shoots, and siliques (David et al., 2017). They observe destabilization of tRNAs, hypersensitivity to oxidative stress, and impaired root elongation in trm4b mutants, though they are unable to determine if any observed phenotypes are due to mRNA m5C specifically (David et al., 2017) (Box 2). Cui and colleagues, in contrast, use an antibody pull-down method based on the assertion that bisulfite conversion could be prone to non-m5C artifacts, such as m3C or hydroxy-m5C (hm5C). They observe strong enrichment in mRNA CDSs and around stop codons, and are able to link mRNA m5C with increased stability of transcripts encoding key regulators of root development such as SHORT HYPOCOTYL 2 (SHY2) and INDOLEACETIC ACID-INDUCED PROTEIN 16 (IAA16), strongly suggesting a possible mechanism for m5C-mediated regulation of root development (Cui et al., 2017) (Box 2).
Consistent with the efficacy of sequence-based m5C prediction (Song et al., 2018), both studies also observe context dependence of m5C. For instance, David and colleagues observed that a 50-mer sequence containing an identified methylated C site is sufficient to confer methylation when added onto a heterologous reporter transcript (David et al., 2017), while Cui and colleagues were able to observe enriched sequence motifs around m5C sites (Cui et al., 2017), perhaps due to the increased specificity of their method. Intriguingly, Cui and colleagues also observe reduced ribosome occupancy in m5C-marked mRNAs, consistent with the expectation that CDS-localized modifications would probably interfere with translation. Thus, m5C seems to be another important epitranscriptome mark in plants and animals that will warrant future research efforts.
In silico methods identify other modifications in Arabidopsis
In addition to m6A and m5C, there are probably a variety of other less abundant modifications in mRNAs. For instance, pseudouridine (Ψ) has been found to stabilize RNA secondary structure and correlate with transcript stability (Schwartz et al., 2014,a), and has been mapped with single nucleotide resolution in both mammals and yeast (Carlile et al., 2014; Lovejoy et al., 2014; Schwartz et al., 2014,a; Li et al., 2016,a), though it has yet to be mapped in plants. Nonetheless, Arabidopsis possesses homologs of Ψ writer proteins, and is known to contain modified uracils (Vandivier et al., 2015), suggesting that Ψ profiling of plants could yield new insights. Work in mammals has also uncovered N1-methyladenosine (m1A) (Dominissini et al., 2016; Li et al., 2016,b; Safra et al., 2017), though at much lower frequency than originally proposed (Safra et al., 2017).
In silico methods for mapping mRNA modifications exploit the tendency of certain modifications along the Watson–Crick base pairing edge (unfortunately m6A does not occupy this edge) to interfere with proper base recognition during generation of cDNA libraries when using reverse transcriptase for first-strand synthesis. This interference manifests as base substitutions in sequencing reads, which induce mismatches during genome mapping, and thus can be used to infer sites of RNA modification (Ryvkin et al., 2013). For example, applying the high-throughput annotation of modified ribonucleotides (HAMR) method in Arabidopsis has indicated the presence of modified cytosines and guanosines in mRNA molecules and non-coding RNAs (ncRNAs) (Vandivier et al., 2015), suggesting that the full complement of mRNA modifications has yet to be fully characterized. These observations are supported by antibody pull-down of 3-methylcytosine (m3C) in plants (Vandivier et al., 2015) and by the recent discovery of mRNA m3C writers in mammals (Xu et al., 2017). From our in silico analyses, we have observed an over-representation of modifications among uncapped, actively degrading RNAs, as well as correlation of modifications with transcript instability, suggesting that most of these HAMR-identified modifications destabilize the parent transcript (Vandivier et al., 2015) (Box 2). Moreover, we have observed enrichment of stress response functions among these unstable modified transcripts (Vandivier et al., 2015), suggesting that these marks could play a role in repressing stress-responsive transcripts under normal growth conditions. It is imperative to test this hypothesis in future studies to better understand the regulation of the plant stress response transcriptome, and how it can be modulated for crop improvement purposes.
Conclusions
The advent of sequencing-based epitranscriptome mapping has rapidly deepened our understanding of how chemical modifications, in particular m6A and m5C, are key regulators of development across multiple kingdoms of life. Multiple recent studies have identified homologous m6A readers, writers, and erasers in Arabidopsis, and used a combination of genetic ablation and sequencing-based mapping to demonstrate a clear role for the plant epitranscriptome in vivo in regulating a host of development processes including root development, meristem identity, leaf and trichome morphology, and floral transition. In particular, epitranscriptome mapping has revealed which key developmental regulator transcripts contain m6A, pointing to clear mechanisms by which the epitranscriptome functions. While m6A is the most abundant and physiologically relevant mRNA modification in plants, bisulfite-seq and m5C RNA immunoprecipitation (RIP) sequencing (m5C-seq) have indicated the presence and importance of m5C, and in silico identification combined with antibody pull-down has also indicated the presence of m3C and additional modified guanosines with the ability to modulate mRNA stability in the plant transcriptome. Further studies of these modifications, and identification and mapping of other mRNA covalent modifications and identifying their readers, writers, and erasers will be critical to furthering our understanding of the plant epitranscriptome.
Additionally, the effects of these post-transcriptional regulatory marks on plant biotic and abiotic stress responses will be an important avenue of future inquiry. As both m6A and other epitranscriptome marks have been linked to these responses (Bodi et al., 2012; Vandivier et al., 2015), the mechanisms by which these marks respond to and effect the plant response to various environmental conditions and stresses is currently mostly unknown. Furthermore, these initial links between the epitranscriptome and plant stress responses suggest that covalent RNA modifications may provide powerful and easy-to-use markers for breeding more stress-tolerant crop plants. Thus, the links between epitranscriptome marks and plant stress response are an extremely important area of inquiry for future research endeavors.
Acknowledgements
We thank members of the Gregory lab for their helpful discussions and comments on the manuscript. This work was funded by the NSF (MCB-1623887 and IOS-1444490 to BDG). The funder had no role in review design, decision to publish, or preparation of the manuscript.
Abbreviations:
- HAMR
high-throughput annotation of modified ribonucleotides
- m1A
N1-methyladenosine
- m6A
N6-methyladenosine
- m3C
3-methylcytosine
- m5C
5-methylcytosine
- ncRNA
non-coding RNA
- RBP
RNA-binding protein
- SAM
shoot apical meristem
- UTR
untranslated region
- Ψ
pseudouridine
Comments