The phylogeny of YTHDF proteins comprises several clades in land plants

The adaptation of plants to terrestrial life was accompanied by the acquisition of morphological, physiological, and genomic complexity [53–55]. Knowing how diversification of YTHDF proteins came about during the course of plant evolution is relevant to understand their distinct functions and may hint at roles they might have played in this process. However, the species included in the so far most detailed phylogenetic study on plant YTHDF proteins jump from bryophytes (mosses, liverworts and hornworts [54,56–60]) to angiosperms (flowering plants) [26]. To include the YTHDF repertoires of intermediately diverging clades of embryophytes such as lycophytes, ferns, and gymnosperms, we performed phylogenetic analyses using YTHDF proteins from 32 species widely spread across land plant evolution. We also included YTHDFs from 8 species of green algae (chlorophytes and charophytes) and, as outgroup, 3 microalgal species from eukaryotic taxa evolutionarily close to plants (Fig 1A). In the resulting phylogenetic tree, all embryophyte YTHDFs branch from a single stem that connects this monophyletic group to the more divergent green algal orthologs (Figs 1B and S1–S3), whose peculiarities are further considered in the discussion. Regarding land plants, the tree largely agrees with the DF-A/B/C clades defined on the basis of angiosperm YTHDF sequences [26], with some minor differences. We consider that the former DF-C group comprising Ath ECT6/7/8/11 [26] can be subdivided into two groups that diverged early during YTHDF radiation (Figs 1B and S1). Thus, we introduced the name ‘DF-D’ for the clade defined by one Amborella trichopoda (a basal angiosperm) YTHDF protein (Atr DFD) and Ath ECT6/7 (Figs 1B and S1). Furthermore, since the group comprising bryophyte YTHDFs did not receive a designation in previous studies, we named it DF-E (mnemotechnic for ‘Early’). It includes YTHDF proteins from bryophytes and some lycophyte orthologs that branched close to the moss subgroup (Figs 1B and S1). Finally, an additional group composed of YTHDF proteins from six species of ferns and two gymnosperms was named DF-F, as they did not fall into any of the other clades (Figs 1B and S1). Interestingly, the alignment revealed independent gains and losses of molecular traits during plant YTHDF evolution. For example, an aspartate-to-asparagine substitution in helix α1 that increases affinity for m⁶A [61], characteristic of DC and plant DF-B YTH domain proteins [26, 61], is also present in the fern proteins of the DF-F clade, but absent in DF-Es, DF-As and gymnosperm DF-Fs (S4 Fig), suggesting independent losses or acquisitions. Taking these observations together, we conclude that land plant YTHDFs are phylogenetically more diverse than previously appreciated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Phylogenetic analysis of YTHDF proteins in plants.

(A) Schematic representation of Viridiplantae evolution and relative position of the species used in this study. The architecture of the diagram and the age (million years, My) indicated on some nodes are a simplified version of the trees from Su et al. [59] and Bowman [62]. The length of the branches has been adjusted for illustrative purposes. (B) Phylogenetic tree of YTHDF proteins in plants (S1–S4 Datasets), with color-coding and abbreviations of species names as in A. Three YTHDF proteins from taxa outside but evolutionarily close to Viridiplantae are included as outgroup. Arabidopsis thaliana (Ath) YTHDF proteins are named after the original nomenclature for proteins containing an Evolutionarily Conserved C-Terminal Region (ECT) established by Ok et al. [40]. Proteins from other plant species adhere to the nomenclature established by Scutenaire et al. [26], with small variations reflecting additional clades (DF-E, -F) and the split of the original DF-C clade into -C and -D. Statistical support is calculated as approximate likelihood-ratio test (aLRT), and indicated by grayscale-coded spheres and values on the most relevant nodes. For simplicity, only a subset of proteins is labelled–those from one representative species of the main taxa highlighted in bold in A–, but a comprehensive representation of the same tree with all protein names and aLRT values is available in S1 Fig. The length of the branches represents the evolutionary distance in number of amino acid substitutions per site relative to the scale bar. Double lines crossing the longest branches indicate that the branch has been collapsed due to space problems, but a schematic representation of their relative lengths is shown on the lower-left corner, and not-collapsed branches can be found in S1 Fig.

https://doi.org/10.1371/journal.pgen.1010980.g001

YTHDF protein diversification occurred early during land plant evolution

Our phylogenetic analysis shows that plant YTHDF diversification started at least before the radiation of Euphyllophytes (plants with true leaves comprising ferns, gymnosperms and angiosperms) more than 410 Mya [59,63]. This is because bryophytes possess one or two YTHDFs in the ‘early clade’ (DF-E), while all six fern species analyzed here have extended sets of DF proteins (4–9 paralogs) that spread into clades DF-F, DF-D, and the branch sustaining the DF-A and DF-B groups (Fig 1B). Furthermore, because one DF-protein from the lycophyte Isoetes taiwanensis branches from the common stem between clades DF-C and DF-D (Ita DF-CD), it is likely that a first diversification event started in the ancestral vascular plants. Thus, our analysis reveals that YTHDF radiation started early in land plant evolution and coincided with the acquisition of morphological complexity and the adaptation to diverse environments.

A functional complementation assay for plant YTHDF proteins

To address the degree of functional specialization among plant YTHDF proteins in a simple manner, we set up a functional complementation assay that scores the ability of each of the eleven arabidopsis YTHDF proteins (Fig 2A) to perform the molecular functions of ECT2/3/4 required for rapid cellular proliferation in leaf primordia. Initial attempts at using the ECT2 promoter for ectopic ECT expression in primordial cells were unsuccessful, as pilot experiments with a genomic ECT4 fragment revealed that its expression was substantially lower than that of a similar ECT2 genomic fragment when driven by the ECT2 promoter (S5 Fig), perhaps pointing to the presence of internal cis-regulatory elements. We therefore turned to the use of ECT cDNAs (cECTs) under the control of the promoter of the ribosomal protein-encoding gene RPS5A/uS7B/US7Y (At3g11940 [64,65]), henceforth US7Yp, active in dividing cells [66] in a pattern that resembles the ECT2/3/4 expression domain [16,17]. Transformation of the US7Yp:cECT2-mCherry construct in te234 plants resulted in complementation frequencies of ~40–50% among primary transformants, a percentage slightly lower but comparable to that obtained with the genomic ECT2-mCherry construct under the control of ECT2 promoter and terminator regions (ECT2p:gECT2-mCherry) [16] (S6A Fig). Free mCherry expressed from a US7Yp:mCherry transgene showed no complementation (S6A Fig), as expected. Importantly, the leaf morphology and the pattern of mCherry fluorescence in US7Yp:cECT2-mCherry lines was indistinguishable from that of ECT2p:gECT2-mCherry (S6B Fig). Hence, we proceeded with expression of cDNAs encoding all of ECT1-11 fused to C-terminal mCherry under the control of the US7Y promoter (henceforth ECT1-11) in te234 mutants (Fig 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Most but not all arabidopsis YTHDF proteins can replace ECT2/3/4-function.

(A) Diagram showing the relative length of the N-terminal IDRs of Arabidopsis thaliana (Ath) YTHDF proteins (ECT1-ECT11) together with Marchantia polymorpha (Mpo) DFE, Homo sapiens (Hs) YTHDF2, and Drosophila melanogaster (Dm) YTHDF. Numbers indicate the length of the proteins in amino acids (aa). (B) Strategy followed for the functional assay. US7Yp:cECT(X)-mCherry constructs are introduced in ect2-1/ect3-1/ect4-2 (te234) plants, and complementation rates are estimated by the percentage of primary transformants (T1) whose first true leaves have a size (s) exceeding 0.5 mm after 10 days of growth. The construct US7Yp:mCherry is used as negative control. Examples of T1 seedlings expressing control and ECT2 constructs are shown. Dashed outlines are magnified below to show mCherry fluorescence in emerging leaves. (C) Weighed averages of the complementation percentages observed for each US7Yp:cECT(X)-mCherry construct in 2–5 independent transformations (I.T.). Letters between parenthesis indicate the DF- clade assigned to each protein. Colouring highlights ECT2 (reference) and proteins with low or negligible complementation capacity (ECT1/9/11) for fast identification along the manuscript. Low transformation efficiency (T.E.) for ECT9 can be inferred by the low number of transformants (n) highlighted in red, but detailed T.E. and raw complementation rates in each transformation can be found in S9 and S7A Figs respectively. Additional data regarding this assay is shown in S8 Fig. (D) Same as C on a single transformation using the rdr6-12/te234 background. (E-H) Rosette phenotype of 32-day-old T1s bearing US7Yp:cECT(X)-mCherry constructs in the rdr6-12/te234 background, sorted according to their degree of complementation in three panels: F (high), G (residual) and H (non-complementation and enhancement). Different genetic backgrounds and control transformants (mCherry) are shown in E as a reference. The scalebar (5 cm) applies to all plants in the four panels.

https://doi.org/10.1371/journal.pgen.1010980.g002

Percentage of complementation among primary transformants as a readout for functionality

To design the experimental approach in detail, we considered the fact that expression of transgenes involves severe variations in levels and patterns among transformants. This is due to positional effects of the T-DNA insertion and the propensity of transgenes to trigger silencing in plants [67], and explains why only a fraction of ECT2-mCherry lines rescues loss of ECT2 function (S6A Fig, [16]). Hence, many independent lines need to be analyzed before choosing stable and representative lines for further studies. Taking this into account, we decided to perform systematic and unbiased comparisons of ECT functionality by counting the fraction of primary transformants (henceforth T1s) able to complement the late leaf emergence of te234 plants for each construct. We set a size threshold for the rapid assessment of large numbers of transformants considering that leaves longer than 0.5 mm after 10 days of growth constitute unambiguous sign of complementation, because this is the minimum size never reached by te234 or control lines (S2B Fig). Hence, we calculated complementation percentages as the number of T1s with leaf size (s) larger than 0.5 mm relative to the total number of transformants, and averaged that score over several independent transformations (Figs 2C and S7A, S8A).

Most arabidopsis YTHDF paralogs can perform the molecular function required for leaf development

The results of the comparative analysis reveal a high degree of functional overlap within the arabidopsis YTHDF family, because eight out of the eleven YTHDF proteins, namely ECT2/3/4/5/6/7/8/10, complement the delayed leaf emergence of ect2/3/4 mutants with frequencies ranging from 14% (ECT6) to 39% (ECT2) when expressed in the same cells (Fig 2C). Importantly, complementation is indirect proof of m⁶A-binding ability, because the proteins must bind to the targets whose m⁶A-dependent regulation is necessary for correct leaf organogenesis in order to restore ECT2/3/4 function [16]. This conclusion is not trivial, as some eukaryotes lacking canonical m⁶A-methyltransferases encode YTH domain proteins [68] and, indeed, the YTH domain of fission yeast Mmi1 does not bind to m⁶A [69]. Thus, most, but not all, arabidopsis YTHDF proteins retain the molecular functions required to stimulate proliferation of primed stem cells. Of the three remaining proteins, ECT1 showed only residual complementation capacity, while no complementation was observed for ECT11 at this stage (Fig 2C). Finally, we could not conclude at this point whether ECT9 could replace ECT2/3/4 function or not, because this transgene produced very few transformants in the te234 background with barely visible fluorescence (Figs 2C and S7–S9). All such transformants underwent complete silencing in the following generation (T2).

Differences in complementation scores are not explained by a bias in expression levels

Our T1 scoring method relies on the assumption that the US7Y promoter drives similar ranges of expression of the different ECT cDNAs. However, different silencing propensities between the ECT cDNAs might introduce bias in the results. To take this possibility into account, we first performed western blotting from three lines per transgene selected as the best-complementing in their group (S8B Fig). The results showed that the higher complementation scores of e.g. ECT2, ECT3 and ECT8 are not explained by higher expression. Rather, the reverse was true. For example, while ECT2 could complement te234 plants with the lowest expression among all ECTs, the few lines expressing ECT1 able to pass the complementation threshold had the highest expression of all ECTs, suggesting that ECT1 can only partially compensate for loss of ECT2/3/4 when overexpressed.

The same functional classes of ECTs emerge from complementation of silencing-deficient te234

Next, we tested whether the same results could be reproduced in plants lacking RNA-DEPENDENT RNA POLYMERASE 6 (RDR6), a gene essential for the biogenesis of transgene-derived short interfering RNAs (siRNAs) [70, 71]. Because siRNA-guided transgene silencing is impaired in rdr6 mutants, higher and more homogenous ranges of expression among different transgenes can be expected in this background. Hence, we constructed rdr6-12/te234 mutants that exhibited both the narrow, downward-curling leaf phenotype of rdr6 mutants [72] and the characteristic rosette morphology of te234 [16] (S10A Fig). Importantly, the delay in leaf emergence was identical in te234 and rdr6-12/te234 mutants (S10A Fig) and the complementation rates of ECT2 transgenes were much higher in the silencing-deficient plants (S6A Fig), indicating that the <100% complementation frequencies in te234 were indeed due to RDR6-dependent silencing. In the rdr6/te234 background, complementation scores of the eight functionally equivalent ECTs (ECT2/3/4/5/6/7/8/10) ranged from 59% (ECT6) to 94% (ECT2), and the differences between these ECTs were less pronounced, but followed a trend largely unchanged compared to the te234 background (Fig 2C and 2D). On the other hand, the difference between the eight ECTs with ECT2/3/4-like function and the group of ECTs with negligible complementation capacity (ECT1/9/11) was highlighted by the rdr6 mutation: ECT1 and ECT11 did not increase their scores (Fig 2D), and absence of silencing resulted in a number of ECT9 transformants with visible mCherry fluorescence, yet no complementation activity (Figs 2D and S10B), thus verifying the lack of ECT2/3/4-like function of ECT9.

Rosette phenotypes corroborate the complementation scores

To verify the results of the complementation assay in the adult phenotype, we characterized rosettes of two independent lines for each transgene (Fig 2E–2H). As expected, rdr6-12/te234 plants expressing ECTs with high complementation scores resembled single rdr6-12 mutants (Fig 2E and 2F), while the ECT1 and ECT11 transformants with the biggest first true leaves at 10 DAG produced plants only slightly bigger than the rdr6-12/te234 background (Fig 2E and 2G), indicating low but residual ECT2/3/4-like function for these two proteins. Strikingly, rdr6-12/te234 plants expressing ECT9 exhibited strong developmental phenotypes (Figs 2H and S10B) that resembled those of severely m⁶A-depleted plants [51,52]. Such an effect was not observed for any other Ath ECT. This result suggests that ECT9 can bind to m⁶A targets, either exerting opposite molecular actions compared to ECT2/3/4, or simply precluding their possible regulation by remaining ECTs such as ECT5/6/7/8/10 by competitive binding. We tested whether ECT9 can outcompete endogenous ECT2/3/4 in the proliferating cells of the apex by expressing the ECT9 transgene in single rdr6-12 mutant plants. We found no evidence for this, because the transformation produced many fluorescent T1s without obvious defects (S11 Fig). Altogether, our results indicate that arabidopsis YTHDF proteins may be divided into three functional classes, i) ECT2/3/4/5/6/7/8/10 with the molecular functionality of ECT2/3/4 required for correct and timely leaf formation; ii) ECT1/11, with only residual ECT2/3/4 functions, and iii) ECT9, unable to perform such functions and toxic when present in the expression domain of ECT2/3/4 in their absence.

Validation of the functional assay by loss-of-function genetic analysis

One of the predictions on genetic interactions between Ath ECT genes that emerges from our results is particularly unexpected and, therefore, suitable for the validation of our complementation assay by classical genetic analysis: Mutation of ECT1 should not exacerbate the developmental defects of te234, even if it is endogenously expressed in leaf primordia. ECT1 lends itself well to validation for two reasons. First, ECT1 is the closest paralog of ECT2/3/4 (Fig 1B), intuitively suggesting some degree of genetic redundancy. Second, ECT1 promoter-reporter fusions [40] and mRNA-seq data ([73], S12 Fig) suggest that expression levels and tissue specificity of ECT1 are comparable to that of ECT4, whose activity is easily revealed as an enhancer mutant of ect2/ect3 [16].

ECT1 and ECT2/3/4 are cytoplasmic and are expressed in the same cell types

We first assessed the expression pattern and subcellular localization of ECT1 using stable transgenic lines expressing TFP translationally fused to the C-terminus of ECT1 (ECT1-TFP, S13A Fig). The pattern of turquoise fluorescence was strongly reminiscent of that of fluorescent ECT2/3/4 fusions [16] with signal at the base of young leaves, in leaf primordia, main root tips and lateral root primordia at different stages (Fig 3A–3D). In addition, confocal microscopy of meristematic root cells showed cytoplasmic ECT1-TFP signal with heterogenous texture similar to ECT2/3/4 [16] (Fig 3E). We noticed, however, that most root tips had a few cells containing distinct ECT1-TFP foci (Fig 3F) in the absence of stress. We found no trace of free TFP that could be causing imaging artifacts (S13A Fig) but note that the ECT1-TFP transgenic lines selected on the basis of visible and reproducible pattern of fluorescence may overexpress ECT1 (S13B Fig). We conclude from the similar expression pattern and subcellular localization of ECT1 and ECT2/3/4 that assessment of the possible exacerbation of te234 phenotypes by additional loss of ECT1 function is a meaningful test of our functional assay.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Endogenous ECT1 has negligible ECT2/3/4-like function, despite similar expression pattern and subcellular localization.

(A-D) Expression pattern of ECT1p:gECT1-TFP (gDNA) in aerial organs (A), main root (B), lateral root primordia (C) and emerging lateral roots (D) of 10-day-old seedlings. The expression mimics the pattern observed for ECT2/3/4 (Arribas-Hernández et al., 2020). (E-F) Intracellular localization of ECT1-TFP in meristematic cells of root tips in unchallenged conditions. Although the cytoplasmic signal is largely homogenous (E), sporadic foci (F) are frequently observed. S13A Fig shows the integrity of the fluorescently tagged protein in independent transgenic lines assessed by protein blot. (G) Schematic representation of the ECT1 locus. Exons are represented as boxes and introns as lines. The positions and identifiers of the T-DNA insertions assigned to the ect1-1, ect1-2 and ect1-3 alleles are marked, and the location of qPCR amplicons and hybridization probes (P) for analyses is indicated below. Genotyping of ect alleles is shown in S13C–S13F Fig. (H) Expression analysis of ECT1 mRNA in wild type and T-DNA insertion lines by qPCR. *P<0.01, **P<0.001, ***P<0.0001, and ****P<0.00001 in T-test pairwise comparisons (supporting data in S9 Dataset). Northern blot using the probe (P) marked in G detects ECT1-TFP mRNA, but the endogenous ECT1 transcript is below detection limit (S13B Fig). (I) Morphological appearance of seedlings with or without ECT1 in the different backgrounds indicated. Nomenclature and additional phenotyping of these and alternative allele combinations can be found in S14 Fig. DAG, days after germination. Scale bars are: 1 cm in A, 1 mm in B, 100 μm in C, 10 μm in D-F, 1 mm in upper panels of I, and 1 cm in the lower panels.

https://doi.org/10.1371/journal.pgen.1010980.g003

Mutation of ECT1 does not exacerbate the phenotype of ect2/ect3/ect4 plants

We isolated three homozygous knockout lines caused by T-DNA insertions in the ECT1 gene, ect1-1, ect1-2, and ect1-3 (Figs 3G and 3H and S13C–S13F), and crossed two of them to different ect2/ect3/ect4 triple mutants [16,17] to obtain two independent allele combinations of quadruple ect mutants, qe1234 and Gqe1234 (see S14A Fig for nomenclature). In both cases, the quadruple mutant plants were indistinguishable from the corresponding te234 and Gte234 triple mutant parentals (Figs 3I and S14B). Similarly, triple ect1/ect2/ect3 mutants were identical to double ect2/ect3 mutants, and single ect1 mutants did not show any obvious defects in ect2/3-associated traits such as leaf or root organogenesis and trichome branching (Figs 3I and S14B–S14D). Thus, despite their similar expression pattern, there is no indication of redundancy between ECT1 and the other three Ath DF-A paralogs. This unexpected conclusion agrees with the poor ECT2/3/4-like activity of ECT1 in our functional assay, thereby validating the approach.

The divergent function of ECT1 and ECT11 is caused mainly by properties of their IDRs

Having established that different molecular functions are represented in the arabidopsis YTHDF family, we went on to map the molecular regions responsible for this functional divergence. First, we investigated whether lack of te234 complementation by ECT1/9/11 is due to altered target binding by the YTH domain, IDR-related effector functions, or a combination of both. Thus, we built chimeric IDR/YTH constructs between ECT2 and ECT1/9/11 (Fig 4A). We also tested an ECT8_IDR/ECT2_YTH chimera (C-8/2) to assess whether hybrid constructs can at all be functional, choosing ECT8 as a positive control due to its robust complementation of seedling and adult phenotypes (Fig 2C–2D, 2F). Expression of the constructs in te234 mutants showed that the C-8/2 chimera was fully functional (Figs 4B and S7B, S15), thus validating the approach. The ECT1- and ECT11-derived hybrids showed that their YTH domains retain m⁶A-binding ability, because the complementation frequencies for the C-2/1 and C-2/11 chimeras were not zero (Fig 4B), contrary to an m⁶A-binding deficient point mutant of ECT2 [16]. They also proved, however, that both the IDR and the YTH domain of these two proteins only retain a reduced degree of function compared to the equivalent regions in ECT2. In particular, the IDRs of both proteins scored lower than their YTH domains when combined with the other half from ECT2 (Fig 4B), pointing to divergence in the IDR as the main cause for different functionality. In wild type ECT1 and ECT11 proteins, the combination of the two poorly-performing parts is likely the cause of the minute ECT2/3/4-like activity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Dissection of ECT1/9/11 functional regions.

(A) Schematic representation of the strategy followed to express chimeras with N-terminal IDR and YTH domains of different ECT proteins. (B) Weighed averages of the complementation rates observed for each chimeric construct measured as in Fig 2B and 2C. The raw complementation rates and transformation efficiencies of each independent transformation (I.T.) can be found in S7B and S9 Figs respectively, and photographs of the scored 10-day-old T1 seedlings with mCherry fluorescence are shown in S15 Fig. (C) 10-day-old primary transformants of rdr6-12/te234 with the indicated transgenes. Dashed outlines are magnified below to show mCherry fluorescence. (D) Same genotypes as in C at 26 days after germination. Additional independent transgenic lines, developmental stages and controls for C and D can be found in S10 Fig. All transgenes are expressed from the US7Y promoter and contain mCherry fused to the C-terminus.

https://doi.org/10.1371/journal.pgen.1010980.g004

The IDR of ECT9 is incapable of performing ECT2/3/4 molecular functions, but its YTH domain retains some ECT2/3/4-like function

The ECT2_IDR/ECT9_YTH chimera (C-2/9) yielded normal transformation efficiencies and up to 17% of T1s had signs of complementation compared to 56% for ECT2 (Figs 4B and S7B, S9, S15), suggesting that the YTH domain of ECT9 retains at least partial functionality, including m⁶A-binding activity. However, the reverse construct (ECT9_IDR/ECT2_YTH, C-9/2) recapitulated the results obtained for ECT9: low transformation efficiency and low fluorescence in the few T1s recovered, and complete silencing in the next generation (Figs 4B and S7B, S9, S15). Hence, we introduced the C-9/2 chimera into the rdr6-12/te234 background and observed plants with stronger fluorescence exhibiting aberrant morphology, again similar to what was obtained with ECT9 (Figs 4C, 4D and S7C, S10B). These results confirm that ECT9 is not functionally equivalent to ECT2/3/4, and point to the IDR as the main site of functional divergence of the protein.

The molecular requirements for m⁶A-binding are conserved in the YTH domains of ECT1/9/11

To pinpoint molecular characteristics of ECT1/9/11 responsible for their different functionality, we performed a series of comparative analyses between ECT1/9/11 and other ECT proteins. We first analyzed the YTH domains, taking ECT2 as a reference and using the structurally well-characterized YTH domain of Homo sapiens (Hs) YTHDF1 in complex with RNA GGm⁶ACU [61] as a model. Amino acid sequence alignment of the YTH domains of all Ath ECTs (S16 Fig), AlphaFold-predictions [74, 75] of the apo structures and homology models [76] of the m⁶A-bound state (S17 Fig) confirmed a high degree of conservation of the key residues required for m⁶A-RNA binding and overall fold of the YTH domain of ECT1/9/11, in agreement with the partial functionality observed in our chimeric constructs (Fig 4B). However, we also detected a few non-conservative amino acid substitutions specific for ECT1/9/11 (Figs 5A and S16), preferentially located on the surface of the protein but far from the RNA-binding groove (S17 Fig). These non-conservative substitutions are prime candidates to explain the minor differences in functionality, a scenario that implies biological functions of YTH domains other than mere m⁶A-binding capacity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Analysis of ECT1/9/11 functional regions.

(A) Topology of the YTH domain of Homo sapiens (Hs) YTHDF1 with secondary structure elements drawn in pink [61]. The two grey helices represent the only structural elements outside the YTH domain as defined by Stoilov et al. [8]: α0 immediately upstream, and α4 at the C-terminus of the protein. Superimposed on the diagrams are the non-conservative amino acid substitutions found in Ath ECT1, ECT9 and ECT11 compared to the majority of Ath ECTs (see alignment in S16 Fig). The localization of these amino acids in the predicted 3D structures, predominantly on the surface, and the spatial integrity of the aromatic cage in the homology models of ECT1/9/11, can be found in S17 Fig. (B) Slab simulations of the IDRs of the indicated proteins showing their tendency to remain in condensed phase or disperse in solution. Source videos of ECT2, ECT1 and ECT11 simulations are provided as S1–S3 Movies (C) Time-averaged density profiles for a subset of the proteins in B. The analyses of the remaining Ath ECTs are shown in S19A Fig. (D) Calculated free energy change for the transition between aqueous solution and condensed phase (ΔG_trans = RT ln [c_dilute / c_dense]) for the IDRs of the indicated proteins. (E) Cumulative charge distribution of the IDRs of the indicated proteins. The length in amino acids (aa) is at scale. The same analysis for all Ath ECTs can be found in S19D Fig.

https://doi.org/10.1371/journal.pgen.1010980.g005

The IDRs of ECT1/9/11 have biophysical properties distinct from ECT2/3/4/5/6/7/8/10

We next focused on the N-terminal regions of Ath ECT1/9/11. Because protein disorder predictions using MobiDB [77] confirmed that they are all disordered (S18 Fig), we used approaches suitable for the study of IDRs. Prediction of short linear motifs (SLiMs) that may mediate interaction with effector proteins gave many possibilities that did not follow a clear pattern when compared between the different ECT classes (S18 Fig). Hence, we decided to study other properties known to be important in IDRs and focused on the propensity to form liquid condensates, the amino acid composition, and the distribution of charge. To determine whether the distinct functional classes of ECTs differed in their propensity to phase separate, we employed a recently developed coarse-grained model of IDRs with residue-specific parameters estimated from experimental observations of many IDRs [78] to simulate their tendency to remain in the condensed phase or disperse in solution. As control, we used the well-studied IDR of the human RNA-binding protein Fused in Sarcoma (FUS) [79–82]. The results showed that the IDR of ECT2 has a clear propensity to remain in the condensed phase (Fig 5B–5C and S1 Movie) with a highly favorable Gibb’s free energy change for the transition from dilute aqueous solution to condensed phase (Fig 5D), comparable to, albeit not as pronounced as for the IDR of Hs FUS (Fig 5B–5D). While the IDRs of most Ath ECT proteins behaved comparably to ECT2, the IDRs of ECT1 and ECT11 were clear outliers (Figs 5B–5D and S19A). Compared to all other ECTs, the IDR of ECT1 showed a much stronger propensity to phase separate, consistent with the spontaneous granule formation of ECT1-TFP in vivo (Figs 3F and 5B–5D, and S2 Movie). On the contrary, the IDR of ECT11 dispersed into solution much more readily (Fig 5B–5D and S3 Movie). Thus, the IDRs of ECT1 and ECT11 have markedly different biophysical properties from those of ECT2/3/4/5/6/7/8/10, perhaps contributing to their different functionality in vivo. In contrast, the IDR of ECT9 did not stand out from the rest in the simulations of phase separation propensity. We also note that the different phase separation propensity among the ECTs is driven, in the main, by their different content of sticky residues as defined by the CALVADOS model [78] (S19B Fig). In this context, it is potentially of interest that ECT1, ECT9 and ECT11 all have a higher content of Phe residues than the nearly Phe-depleted IDRs of ECT2/3/4/5/6/7/8/10 (S19C Fig). For ECT1, whose Tyr and Trp enrichment is comparable to that of ECT2/3/4/5/6/7/8/10 (S19C Fig), the additional enrichment of the similarly sticky Phe may contribute to its more pronounced phase separation propensity.

We finally analyzed the charge distribution, as this may dictate properties such as compactness and shape of IDR conformations [83, 84]. We found that the IDRs of ECT2/3/4/5/6/7/8/10 show a recurrent pattern with accumulation of net negative charge towards the N-terminus, and recovery to near-neutrality closer to the YTH domain, resulting in creation of electric dipoles (Figs 5E and S19D). The IDRs of ECT1, ECT9 and ECT11 differ from this pattern, but in distinct ways: ECT1 does not show a pronounced patterning of the charged residues, while ECT9 increasingly accumulates negative charge (Fig 5E) and ECT11 increasingly accumulates positive charge (S19D Fig). It is an interesting possibility, therefore, that charge separation in the IDR is a requirement to fulfill the molecular functions of ECTs to accelerate growth in primordial cells.

The ability to complement ect2/ect3/ect4 does not simply follow affiliation with phylogenetic clades

We next compared the functional groups of YTHDF proteins defined by our complementation assays with the YTHDF phylogeny. Strikingly, the ability to rescue defective timing of leaf emergence does not follow the defined phylogenetic clades, because both high and low/null complementation capacities are present in the same clades. For example, ECT2, ECT5 and ECT8 in clades DF-A, -B and -C respectively present high ECT2/3/4-like functionality, while ECT1, ECT9 and ECT11 in these same clades do not (Figs 1B and 2C–2H). It is noteworthy, however, that there is a clear trend for non-complementors to be amongst the most highly divergent proteins in each group, as revealed by the length of the branches in the phylogenetic tree (Fig 1B). Indeed, some of the longest branches in both our tree and the one in the study by Scutenaire et al. [26] correspond to ECT1, ECT9, and ECT11. Accordingly, the three proteins have the highest number of amino acid substitutions compared to the only YTHDF protein of the liverwort Marchantia polymorpha (Mpo DFE) [85] (Fig 6A). These results suggest that the common ancestor of YTHDF proteins in land plants may have had a molecular function related to that of the modern Ath ECT2/3/4, and that this function has been conserved in the least-divergent members of the DF-A/B/C/D clades. Subsequent duplication and neofunctionalization may have driven the rapid evolution of highly-divergent YTHDF proteins (ECT11, ECT1, and ECT9 in arabidopsis) that have lost their primary function and perhaps fulfill different molecular roles in the plant m⁶A pathway.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Heterologous expression of a bryophyte DFE rescues loss of ECT2/3/4 function in arabidopsis, while metazoan YTHDF proteins aggravates it.

(A) Evolutionary divergence of Arabidopsis thaliana (Ath) ECT proteins estimated by pairwise comparisons between these and the only DF protein in Marchantia polymorpha (Mpo DFE) according to the alignment of YTHDFs from bryophytes and angiosperms in S7–S8 Datasets (see Methods). (B-C) Weighed averages of the complementation rates observed for Mpo DFE (S/L refer to the Short and Long splice forms found in meristematic tissues–apical notches and gemma cups–of M. polymorpha thalli), Homo sapiens (Hs) YTHDF2 and Drosophila melanogaster (Dm) YTHDF in the te234 (B) or rdr6-12/te234 (C) backgrounds, calculated as in Fig 2C–2D. n, total number of transformants; I.T. number of independent transformations (only one in C). Raw T1 counts, complementation rates and transformation efficiencies can be found in the S6 Dataset, and S7 and S9 Figs. (D) 32-day-old primary transformants expressing the transgenes analysed in C. Additional controls and developmental stages are shown in S10B Fig. (E) Slab simulations of the IDRs of the indicated proteins as in Fig 5B. (F) Time-averaged density profiles of the slab simulations in E. (G) Calculated free energy change for the IDRs of the indicated proteins as in Fig 5D. (H) Cumulative charge distribution of the IDRs of the indicated proteins, as in Fig 5E. The same analysis for the two metazoan YTHDFs can be found in S19D Fig.

https://doi.org/10.1371/journal.pgen.1010980.g006

The molecular function of ECT2/3/4 was present in the first land plants

To test whether the common ancestor of plant YTHDF proteins indeed had a molecular function similar to that of the modern Ath ECT2/3/4, we subjected Mpo DFE (Fig 2A) to our functional assay, as bryophytes are the earliest-diverging taxon of land plants (Fig 1A, [54,56–60]). The result showed a partial, but clear capacity of two different splice forms of Mpo DFE to complement the delay in leaf emergence and morphology defects of arabidopsis te234 plants (Figs 6B and S7D, S20A–S20B). Although the complementation scores were low (Fig 6B), a thorough inspection of the fluorescence intensity in all primary transformants revealed that the transgene was only expressed very weakly in a reduced subpopulation of plants, possibly due to the tendency to trigger silencing upon heterologous expression of a long cDNA from a distant plant species [86]. We therefore introduced the best-complementing isoform into rdr6-12/te234 mutants. This approach yielded a much higher complementation score of 52% (Fig 6C), and plants very similar to those expressing ECT2 in the same background (Figs 6D and S6B). Thus, the Ath ECT2/3/4 activity needed to stimulate proliferation of primed stem cells was present in the first embryophytes, suggesting that this activity is an ancestral molecular function of YTHDF proteins in land plants. Gratifyingly, the simulations of the behavior of the IDR of Mpo DFE showed a dipole-like charge pattern and a phase separation propensity similar to those of the ECT2/3/4/5/6/7/8/10 group, further substantiating that these IDR characteristics correlate with the ability to fulfill ECT2/3/4 function in leaf formation (Fig 6E–6H). We investigated this interesting possibility more deeply by analysing the IDRs of YTHDF proteins from species representing the main clades of land plants that diverged prior to flower evolution, a moss, a lycophyte, a fern and a gymnosperm, and from the sole species in the sister group to all flowering plants, A. trichopoda [87, 88]. This analysis showed that at least one YTHDF protein with ECT2-like charge distribution and phase separation propensity is present in all these taxa (S21A Fig). In particular, this included the only YTHDF protein in the moss Ceratodum purpureus. Interestingly, when the proteins from the different taxa were analysed together, a general trend for increased phase separation propensity that again correlated with higher ‘stickiness’ (λ) was apparent in proteins of clades -A, -B and -E compared to those of -C, -D and -F (S21B Fig). We also noticed that in the species that encode multiple YTHDF proteins, some paralogs exhibit divergent IDR features similar to those seen in Ath ECT1/9/11, but this phenomenon was not clade-specific (S21 Fig). This result suggests that the IDR-based functional diversification observed in arabidopsis may apply more generally. We conclude that the molecular function of YTHDF proteins required for stimulation of primordial cell proliferation is deeply conserved in land plants, and that the biophysical properties of the IDRs of Mpo DFE and the Ath DF proteins capable of performing this function are recurrent in all the major plant taxa.

Heterologous expression of animal YTHDFs enhances the phenotype caused by loss of ECT2/3/4 in Arabidopsis

Finally, we subjected Hs YTHDF2 and Drosophila melanogaster (Dm) YTHDF (Fig 2A) to our functional assay to test whether they could function molecularly like plant YTHDFs. While we did not observe complementation of the delayed leaf emergence of te234 plants in any of the transformants (Figs 6B and S7D), several Hs YTHDF2 lines showed more acute developmental defects compared to te234 (S20B Fig). Although these defects were not observed among Dm YTHDF lines, no expression could be detected in these plants either (S7D Fig). Hence, we repeated the transformation in the rdr6-12/te234 background, this time obtaining numerous transformants of both Dm YTHDF and Hs YTHDF2 with strong fluorescence (Figs 6C and S10B). These plants showed aberrant morphology comparable to the effect caused by expression of ECT9 and C-9/2, with resemblance to arabidopsis mutants with severe depletion of m⁶A [51,52] (Figs 6D and S10B). Thus, expression of metazoan YTHDFs in arabidopsis ECT2/3/4-deficient plants exacerbates their developmental phenotype. This result is in agreement with the generally accepted idea that m⁶A-YTHDF2 destabilizes mRNA in mammals [9,89–92] while the m⁶A/ECT2/3/4 axis does the opposite in plants [23,51,93–97]. However, an analogous molecular mechanism between plant and animal YTHDFs cannot be completely ruled out, because a different cellular context may be hampering the molecular activity of metazoan YTHDFs in plant cells, and simple competition for targets with remaining ECTs could be the cause of the enhanced phenotype. Nonetheless, we note that the IDRs of Hs YTHDF2 and Dm YTHDF had much weaker phase separation propensity (Fig 6E–6G) and different charge distribution (S19D Fig) compared to ECT2, perhaps contributing to the lack of ECT2/3-like function in planta.

1. Inferences on molecular properties of land plant YTHDF proteins

Our study provides at least four new insights into molecular properties of land plants YTHDF proteins, all of which are important for future, more detailed mechanistic studies of eukaryotic m⁶A-YTHDF function. First, the functional assays with Ath ECT proteins and the chimeras combining the YTH domain of ECT1/9/11 with the IDR of ECT2 allow us to infer that all 11 arabidopsis YTHDFs have m⁶A-binding capacity, and suggest that some, if not all, of their in vivo functions involve m⁶A binding. Second, the complementation analyses of the reverse chimeras show that the IDRs of Ath ECTs with divergent functions contribute substantially to their different molecular properties. This finding implies a key role of IDRs in defining the molecular activities of YTHDF proteins, and is in line with the conclusions of a recent study of mammalian YTHDFs [48], with the functional importance of the IDR of Hs YTHDF2 inferred from early tethering studies [9], and with the isolation of IDR-dependent interactors proposed to be important for Hs YTHDF2 and Ath ECT2 function [28–30, 96, 98]. Third, the ability to direct correct leaf formation correlates with a combination of two biophysical properties of the IDRs of plant YTHDFs: a clear propensity to phase separate, and an organization of charged residues that creates an electric dipole. Future studies are needed to address whether a causal relation underlies this correlation, and, if so, what its molecular basis is. Fourth, amino acid substitutions in YTH domains inferred to be capable of binding to m⁶A also contribute, to a lower extent, to functional specialization. Because most affected amino acids are surface-exposed but not predicted to contact RNA, this observation indicates that YTH domains should not be seen as inert adaptors for m⁶A association, as they may also participate in effector functions. It is an important goal of future studies to identify such functions and define how they contribute to functional diversification.

2. The ancestral function and evolution of YTHDF proteins

It is a key finding of our study that the sole M. polymorpha YTHDF protein possesses the molecular functions necessary for leaf formation in flowering plants. This remarkable result demonstrates deep functional conservation, across at least 450 My of evolution, and suggests that stimulating cell division in organ primordia may be the ancestral function of land plant YTHDFs. It is now a question of profound interest to determine how the molecular activity of the ancient YTHDF proteins evolved and diversified as the main eukaryotic linages came into existence. Our work provides several important new insights to guide the understanding of this process. In particular, since it is a valid question whether the main evolutionary event that gave rise to the ancestral function of land plant YTHDFs occurred prior to their terrestrialization, we open this section of the discussion with considerations on how the available information on algal YTHDF proteins informs the debate. We close with the analysis of Ath ECT1 divergence as a case study on evolution of specialized YTHDF function.

3. Functional versus biological redundancy and the need for many YTHDFs

Our study reveals that most YTHDF proteins in higher plants can perform, essentially, the same molecular functions that allow ECT2/3/4 to promote cell proliferation in organ primordia and control organogenesis. We anticipate that this molecular function may be used in distinct biological contexts by ECT5/6/7/8/10 due to different expression patterns and response to stimuli, thereby achieving biological specialization. Such variability in expression patterns can be inferred from available RNA-Seq data in the arabidopsis eFP Browser (BAR.toronto.ca) that show, for example, that ECT8 is primarily expressed during senescence-associated processes, perhaps in connection to re-growth after reversible proliferative arrest [105,106]. Clearly, however, higher plants do not only use many YTHDF-encoding genes to provide the same molecular function(s) in distinct expression domains, as shown by the specialized molecular activity of Ath ECT1, ECT11, and most prominently ECT9. Whether these divergent proteins act to counteract the ECT2/3/4-like function by competing for m⁶A-containing transcripts, perhaps halting growth upon stress, or whether they perform entirely independent functions in different tissues or cell types, is open for new investigations.

An outstanding question not yet resolved by our study is why higher plants seem to need many YTHDF proteins. Indeed, the number of paralogs in vascular plants well surpasses the three YTHDFs present in most vertebrates [6] and the even smaller sets in other eukaryotes (S2 Fig, [8,68]). The need for many YTHDFs of considerable variability is all the more puzzling given our demonstration here that the majority of them retains the same molecular functions. Robustness is a commonly given explanation for redundancy, but the same argument could apply to YTHDCs, typically encoded by 0–2 genes in most eukaryotes including land plants [8,68] (S2 Fig), or methyltransferase subunits, for which only single copies have been maintained during evolution [68]. Therefore, the answer should probably be sought elsewhere. It is clear now that the m⁶A/YTHDF axis controls cellular proliferation and organ growth in plants [17,97], and a major difference between land plants and other complex organisms in that regard is the extreme plasticity of plant development. Because growth and stress responses are also tightly intertwined, it is possible that plants use a plethora m⁶A-YTHDF growth-stimulators in different organs, responsive to different stresses and developmental cues, to shape plant architecture in response to stimuli. The answer to the paradox may be in the details: plant YTHDFs responsive to different environmental cues may fine-tune growth programs through slightly different performance and target-binding capacity in partly overlapping expression domains, where they are known to compete for targets [23].

Plant material and growth conditions

All the lines used in this study are in the A. thaliana Columbia-0 (Col-0) ecotype. The T-DNA insertion lines rdr6-12 [72], ect1-1 (SAIL_319_A08), ect1-2 (GK-547H06), and ect1-3 (SALK_059722) were obtained from Nottingham Arabidopsis Stock Centre. The mutants ect2-3, ect3-2, ect2-3/ect3-2 (Gde23), ect2-3/ect3-2/ect4-2 (Gte234) and ect2-1/ect3-1/ect4-2 (te234) used for genetic crosses or as background to produce transgenic lines, and the primers for genotyping the corresponding alleles, have been previously described [16,17]. The standard growing conditions used are detailed in S1 Appendix.

Phylogenetic analysis

Amino acid sequences of all YTHDF proteins found in 40 species distributed among the main taxa of Viridiplantae were downloaded from UniProt (Apweiler et al., 2004), TAIR (www.arabidopsis.org), Phytozome [107], GinkgoDB [108], FernBase [109] and PhycoCosm [110,111] (S1 Dataset). As outgroup, we selected three species from the eukaryotic taxa Glaucophyta (Cyanophora paradoxa, Cpa), Haptophyta (Emiliania huxleyi, Ehu) and Rhizaria (Paulinella microscpora, Pmi), as these are evolutionarily closer to Viridiplantae than e.g. Opisthokonta such as animals and fungi [99, 100]. All sequences were aligned with MUSCLE [112,113] using default parameters (S2 Dataset). The alignment was trimmed from both ends to discard regions with poor sequence conservation according to the quality and conservation scores provided by Jalview [114] (S3 Dataset and S23 Fig). The retained part of the alignment, comparable to the region used for phylogenetic analysis in a previous study [26], contains the entire canonical YTH domain [8] and, on both sides, short stretches with high conservation scores (S23 Fig). Evolutionary analyses were conducted using the maximum likelihood algorithm IQ-Tree 2 [115,116] from the CIPRES Science Gateway v. 3.3 (https://www.phylo.org/index.php/) to infer a phylogenetic tree. The model based on Q matrix [117] estimated for plants [118], ‘Q.plant+R8’, was determined as best-fit using the built-in tool ModelFinder [119] according to Bayesian Information Criterion (BIC) scores, and used for the analysis. Measures of branch supports by approximate Likelihood Ratio Test [aLRT] (SH-aLRT, %) [120] were also calculated. The resulting tree (S4 Dataset) was rooted on Pmi YTHDF, annotated adding SH-aLRT values, and color-coded using FigTree (https://github.com/rambaut/figtree/). An additional alignment performed with a subset of these proteins with the purpose to illustrate the conservation of the YTH domain in green algae and non-flowering plants was calculated and trimmed as described above (S5 Dataset). The graphical representations of this alignment (S3 and S4 Figs) were obtained using Jalview [114].

Estimates of evolutionary divergence between A. thaliana and M. polymorpha YTHDF proteins were obtained computing pairwise distances from the full length alignment of homologs in a subset of land plant species (bryophytes, basal angiosperms, magnoliids and dicots) (S6 Dataset). These taxa were chosen based on the observation that inclusion of divergent proteins from green algae, ferns, gymnosperms and monocots causes loss of information due to more frequent gaps, while proteins from species closely related to M. polymorpha (bryophytes) and A. thaliana (dicots and basal angiosperms) maximize the accuracy of the alignment. The analysis was done as described above, and we used the Equal Input model [121] in MEGA11 [122] to calculate the distances. All positions containing gaps and missing data were eliminated (complete deletion option), resulting a total of 171 positions in the final dataset. The complete matrix can be found in S7 Dataset.

Cloning, plant transformation and line selection

We employed the scar-free USER cloning method [123] to produce US7Yp:cYTHDF(X)-mCherry-OCSter and ECT1p:gECT1-FLAG-TFP-ECT1ter constructs by gluing PCR-amplified DNA fragments into the pCAMBIA3300-U plasmid [124] that provides resistance to glufosinate in plants. Cloning of the trial ECT2p:gECT1-mCherry-ECT2ter and ECT2p:gECT4-mCherry-ECT2ter constructs was done with GreenGate [125]. The detailed cloning, transformation and line selection strategy for each construct, as well as the specific details of T1 selections for the complementation assay, are explained in S1 Appendix.

IDR modeling and simulations

We carried out coarse-grained simulations of protein IDRs using the implicit-solvent CALVADOS 2 forcefield [78]. In CALVADOS, amino-acid residues are mapped to single beads corresponding to their amino-acid type. Neighboring residues are linked using harmonic bonds with an equilibrium distance of 0.38 nm and force constant of 8033 kJ mol^-1nm^-2. Nonbonded interactions are modeled with a hydropathy amino-acid “stickiness” model using λ-parameters from CALVADOS 2 [78,126]. Salt-screened electrostatic interactions are accounted for with a truncated (r_c = 4 nm) and shifted Debye-Hückel potential; non-electrostatic interactions were truncated at 2 nm [126]. We used a Langevin integrator with friction coefficient γ = 0.01 ps^-1 to propagate the simulation in 0.01 ps timesteps. All simulations were carried out at T = 293 K.

We performed slab simulations of IDRs of the following proteins: Ath ECT1-11, Dm YTHDF1, Hs YTHDF2 and FUS, Mpo DFE, Cpu DFE, Ita DFE1-3, Cri DFF1-3, DFAB and DFD1-2, Gbi DFF1-2, DFA1-2, DFB, DFC1-2 and DFD, and Atr DFA1, DFB, DFC1-2 and DFD, using a previously described procedure [126,127]. IDR sequences of all proteins are detailed in S1 Dataset. Briefly, we simulated 100 copies of each IDR (one type of IDR at a time) in a simulation slab of dimensions 17x17x300 nm for Ath ECT1/3/6-11, Gbi DFB and DFF1 and Atr DFC2, 15x15x150 nm for Hs FUS, or 25x25x300 nm for all other proteins, accounting for the different IDR sequence lengths. We simulated each system for 7500 ns and recorded configurations at 1-ns intervals. The first 2000 ns of each simulation were considered as equilibration and not used for analysis. We centered the protein condensates in the z-dimension of the simulation box during post-processing. The concentrations of the dense phase and dilute phase were calculated by fitting the concentration profiles to a hyperbolic function, as described by Tesei et al. [78]. Excess transfer free energies were calculated from the dilute and dense phase equilibrium concentrations via ∆G_trans = RT ln (c_dilute / c_dense) [128]. All simulation data and analysis scripts are available at the Zenodo repository (https://doi.org/10.5281/zenodo.8269719).

Charges (q) were added along the IDR sequences to obtain cumulative charge plots. Arg and Lys residues were assigned charges of q = +1; Glu and Asp residues were assigned charges of q = -1; other residue types, including His, were assigned charges of q = 0. The average ‘stickiness’ of a given IDR sequence was determined by calculating the mean hydrophobicity of the IDR amino acid composition, using the CALVADOS 2 hydrophobicity scale reported in [126].

Additional methods

S1 Appendix describes additional standard methods including DNA and cDNA obtention, genotyping (see S13C–S13F Fig and S2 Table for additional details), fluorescence microscopy, qPCR (see S9 Dataset for the raw data), RNA and protein blotting, and the characterization of YTH domains.