mRNA adenosine methylase (MTA) deposits m⁶A on pri-miRNAs to modulate miRNA biogenesis in Arabidopsis thaliana

Lack of m⁶A Results in Impaired miRNA Biogenesis.

In order to assess possible effects of m⁶A on miRNA biogenesis, we performed small RNA (sRNA) sequencing on rosette leaves from 4-wk-old Arabidopsis plants of both WT Col-0 and a mutant line with severely reduced m⁶A levels (mta mutant). More specifically, the mta mutant line we used contains MTA cDNA under the ABI3 promoter in a homozygous MTA T-DNA insertion mutant background (mta ABI3:MTA). The ABI3 promoter drives strong embryo expression of MTA, enabling the plant embryo lethality to be bypassed. However, this promoter drives a very low level of expression postgermination, giving rise to plants with 80 to 90% less m⁶A modification compared to their WT counterparts (26). Our sequencing data showed a decrease in the overall abundance of mature miRNAs in mta mutant as compared to WT plants (SI Appendix, Fig. S1). Specifically, we identified 60 differentially expressed miRNAs that had high confidence scores (probability ≥ 0.9 or false-discovery rate ≤ 0.1), and 51 of these 60 miRNAs were found to be down-regulated, while only 9 miRNAs were up-regulated (Fig. 1A and SI Appendix, Table S1). This down-regulation of miRNAs in mta plants was confirmed by quantitative real-time PCR (RT-qPCR) for six arbitrarily selected miRNAs (miR159b, miR169a, miR319b, miR396b, mir399a, and miR850) (Fig. 1B).

Next, we went on to investigate the levels of primary miRNAs (pri-miRNAs) in the low methylation plants (mta mutants) by using the mirEX² platform (www.combio.pl/mirex2) developed by our group (30, 31). This platform utilizes a repository of 298 primer pairs, specifically for Arabidopsis pri-miRNA quantification by RT-qPCR. Of the 298 pri-miRNAs tested, 68 pri-miRNAs were at undetectable levels in our samples, leaving 230 for further analysis. The results of RT-qPCR revealed that 85 pri-miRNAs had statistically significant (P ≤ 0.05, n = 3) changes in their accumulation (WT vs. mta); 56 of these (∼66%) were found to be up-regulated, while 29 were down-regulated (Fig. 1C and SI Appendix, Table S1). Upon comparison of the RT-qPCR results with sRNA sequencing data, we identified 20 cognate pri-miRNA/miRNA pairs that showed higher accumulation of pri-miRNAs and lower levels of mature miRNAs (Fig. 1D). These results point toward an impaired processing of these pri-miRNAs in the absence of MTA.

Arabidopsis pri-miRNAs Are m⁶A-Methylated by MTA.

The general trend, showing an accumulation of pri-miRNAs and decreased levels of miRNAs in the conditions of reduced m⁶A methylation, raised the possibility that the presence of m⁶A is necessary for proper processing of some pri-miRNAs. Therefore, we tested the methylation status of the pri-miRNAs using m⁶A-RNA immunoprecipitation followed by sequencing (m⁶A-IP Seq). This method is not an equivalent of the commonly used MeRIPSeq (32), as we avoided the fragmentation of polyA RNA. We validated this protocol using spiked-in methylated and nonmethylated controls (SI Appendix, Fig. S2). We further tested the robustness of our data by comparing them to previously published data by Shen et al. (19) and Anderson et al. (33). In our dataset, we were able to identify 14,870 genes whose transcripts (15,562) are depleted in the mta mutant, indicating that these gene transcripts are methylated. Upon comparison of our data to the published datasets, we found an overlap of 2,868 (80%) and 3,100 genes (65%) between our data and data from Shen et al. (19) and Anderson et al. (33), respectively; 930 genes were common among all three datasets. A Venn diagram showing these overlaps is presented in SI Appendix, Fig. S3.

miRNAs can be encoded within intergenic regions as independent transcriptional units or can be embedded within protein-coding genes sharing transcriptional units with their host protein genes (34). For our analysis from the m⁶A-IP Seq data we selected transcripts of only those miRNA genes that are independent transcriptional units. Using this approach, we identified transcripts of 11 MIR genes that were enriched more than 1.5-fold in WT vs. mta, indicating that they are m⁶A-methylated in WT plants (hence can be immunoprecipitated with m⁶A antibody), and lack this methylation in the absence of MTA (Fig. 2A and SI Appendix, Fig. S4). All of the identified pri-miRNAs that passed the enrichment threshold were either absent or had significantly reduced abundance in the mta IP sample relative to WT IP.

As an orthogonal approach to verify that these pri-miRNAs were bona fide targets for MTA directed methylation, we carried out RNA IP (RIP) using an anti-GFP antibody and the GFP-tagged MTA (35S:MTA-GFP) plant line. As a control, the plants expressing only GFP in the WT Col-0 background were used. RIP was performed on the nuclear fraction and followed by RT-qPCR on oligo(dT) primed cDNA. Eighteen pri-miRNAs (10 from our m⁶A-IP Seq data and 8 randomly chosen) were selected for verification by RT-qPCR, and of these all but 4 showed a statistically significant enrichment (P < 0.05) in the MTA–GFP sample when compared to the GFP control (Fig. 2B and SI Appendix, Fig. S5). This indicates that pri-miRNAs are frequently bound by MTA. Taken together, these data suggest that MTA binds and methylates at least a set of pri-miRNAs, and that this binding promotes their processing to mature miRNAs.

Lack of m⁶A Affects Stem–Loop Structure of pri-miRNAs and Their Ability to Bind HYL1.

The effect of m⁶A on secondary structure of RNA has been well documented. Such alterations in RNA structure induced by m⁶A are known as “m⁶A switches” (35–37). These m⁶A switches influence the ability of RNA-binding proteins to recognize the RNA molecules, which further determines the substrate RNA fate. We used protein interaction profile sequencing (PIP-seq) in order to determine whether such structural changes could be observed in miRNA precursors and hence influence miRNA biogenesis. Using this approach, double-stranded RNAs and single-stranded RNAs can be detected after single- and double-stranded specific RNase treatment, respectively (38–40), and thus we were able to identify and distinguish between the structured and unstructured regions of the miRNA precursors. We noticed that the nucleotide accessibility for double- or single-stranded RNase(s) is altered in mta mutants. Our data show a significant reduction in the level of pri-miRNA stem–loop structure in the mta mutant as compared to WT plants (Fig. 3A, two box plots on the left). We observed even stronger differences when we focused only on those pri-miRNAs that we identified as m⁶A-methylated (Fig. 3A, two box plots on the right). By constraining a folding algorithm with PIP-seq–derived structure scores, we obtained 2D structural models for a few examples among the m⁶A-methylated pri-miRNAs (models for pri-miR160a, pri-miR163a, pri-miR166a, and pri-miR830 are presented in SI Appendix, Fig. S6). Presented models represent the most frequently interrogated structural conformation among the potentially multiple existing folding patterns for each RNA. These data-based structural models suggest that the pri-miRNA regions containing the miRNA/miRNA* stem regions tend to be in an unpaired conformation in the absence of the m⁶A mark (mta mutants) as compared to the methylated pri-miRNAs (WT plants) (SI Appendix, Fig. S6).

In light of these data, we reasoned that these changes in structure could in principle influence the binding of HYL1 to these precursors. HYL1 is a double-stranded RNA binding protein and is an important player in miRNA biogenesis (41, 42). Hence, we studied the binding of HYL1 to pri-miRNAs in WT and mta mutant plants using RIP (HYL1-RIP) followed by RT-qPCR. We found that 7 of the 10 pri-miRNAs tested were significantly less abundant in HYL1-IP samples from the mta mutants as compared to WT (Fig. 3B). A decrease in the binding efficiency of HYL1 that is a double-stranded RNA binding protein, to pri-miRNAs in the mta mutant is an additional indication that a double-stranded conformation of miRNA/miRNA* containing stem–loop regions of pri-miRNAs are formed less frequently in low methylation (mta) than in WT plants.

MTA Interacts with RNA Pol II In Situ.

The strong tendency toward accumulation of pri-miRNAs and low abundance of miRNAs in mta mutants is reminiscent of various Arabidopsis mutants of genes encoding the proteins that participate in miRNA biogenesis, like Dicer-like 1 (DCL1), HYL1 (42–44), Serrate (SE) (45), and TGH (46). These proteins are involved in early stages of miRNA biogenesis and are needed for efficient processing of pri-miRNAs to miRNAs. These observations led us to suggest that MTA might act at early stages of miRNA biogenesis, probably cotranscriptionally. The deposition of m⁶A is usually assumed to be cotranscriptional; this assumption is supported by the fact that m⁶A can influence other processing events, such as selection of polyadenylation sites or pre-mRNA splicing, which are themselves cotranscriptional (47–49). However, a direct association between METTL3 and RNA Pol II has only been shown in mammalian cells when the rate of transcription was artificially slowed down (50). We confirmed colocalization of MTA and RNA Pol II in Arabidopsis using immunolocalization (SI Appendix, Fig. S7). Furthermore, we used the proximity ligation assay (PLA) to confirm in vivo interactions of MTA with RNA Pol II in plants under physiological conditions. Our results show direct interactions between MTA–GFP and RNA Pol II in cell nuclei with MTA–GFP expression but not in the GFP control (Fig. 4). These results were obtained for RNA Pol II phosphorylated at both Serine 5 and Serine 2. Thus, we show that MTA is associated with the RNA Pol II from an early stage of transcription and likely methylates pri-miRNAs (among other RNA Pol II transcripts) cotranscriptionally.

MTA Acts at Early Stages of miRNA Biogenesis.

While MTA interacting with RNA Pol II is an indication of it methylating RNA Pol II transcripts cotranscriptionally in general, we investigated whether specific proteins involved in miRNA biogenesis could interact with MTA. To address this possibility, we performed a screen on possible MTA interactors involved in early steps of miRNA biogenesis, like HYL1, Cap binding protein 20 (CBP20), SE, TGH, and Dawdle (DDL1), using a microscopic approach. We performed FRET analyzed by fluorescence lifetime imaging microscopy (FLIM). As FRET can occur only when two proteins are within nanometers of each other, FRET–FLIM helps to quantify direct protein–protein interactions. This FRET–FLIM analysis confirmed the close association and interactions of MTA and TGH (Fig. 5). No such interactions could be seen in the case of any other tested proteins (SI Appendix, Fig. S8). Thus, we identified the miRNA biogenesis factor TGH as an MTA partner.

Lack of m⁶A/MTA Impairs Microprocessor Complex Assembly.

Having identified TGH as an interactor of MTA, we focused on the m⁶A methylation status of pri-miRNAs in the tgh mutant. m⁶A-IP followed by RT-qPCR revealed that except for two pri-miRNAs, there is no overall significant reduction of m⁶A methylation in pri-miRNAs in the tgh mutant (SI Appendix, Fig. S9). This result indicates that TGH is not required for m⁶A methylation, and MTA acts upstream of TGH in miRNA biogenesis.

It is known that during miRNA biogenesis, TGH facilitates DCL1 action and pri-miRNA processing via promoting the pri-miRNA–HYL1 interaction (46). We then tested whether a lack of MTA, and hence loss of the MTA–TGH interaction, could also affect Microprocessor assembly. Using coimmunolocalization, we determined that in the tgh mutant DCL1 colocalization with RNA Pol II (phosphorylated at Serine 2) is significantly reduced. We found a similar level of reduction in DCL1 colocalization with RNA Pol II in the mta mutant (Fig. 6 A and B). Similarly, HYL1 colocalization with RNA Pol II is also reduced in mta mutant plants similar to the tgh mutant (Fig. 6 C and D). The observation that the lack of either TGH or MTA hampers colocalization of RNA Pol II with DCL1 and HYL1 suggests that MTA facilitates Microprocessor assembly either by attracting TGH via direct MTA–TGH interaction or with help of an additional m⁶A reader protein (unknown so far).

MTA Regulates the Level of miR393b which Is Involved in Auxin Response.

Auxin response defects have been reported for hypomorphic mutants of the known plant m⁶A methyltransferase complex (20). We found this to be the case for the auxin-responsive DR5pro:GUS reporter construct when introduced into both WT and the mta mutant background. Upon induction of 14-d-old seedlings with 2,4-dichlorophenoxyacetic acid (2,4-D), the mta plants showed much less GUS expression (Fig. 7A).

miR393b has been described to be involved in auxin response regulation, and in mutants lacking miR393b, auxin signaling is reduced (51). In our sRNA sequencing data, we found that miR393b is down-regulated in mta mutants, although we could not detect its precursor in our m⁶A-IP Seq. However, because of its recognized importance in regulating auxin responses, we further investigated the requirement of m⁶A for miR393b accumulation. We confirmed that pri-miR393b is m⁶A-methylated using m⁶A-IP followed by RT-qPCR. IP of RNA bound to MTA also confirmed MTA binding to pri-miR393b (Fig. 7B). Interestingly, our PIP-seq analysis revealed that pri-miR393b is less structured in mta than in WT plants. Moreover, the 2D model of this precursor revealed that its stem–loop region that contains mature miR393b sequence is formed more efficiently when pri-miR393b contains m⁶A (WT) in comparison to unmethylated precursors (mta mutant), similar to the other miRNAs showing underaccumulation in mta mutant plants in comparison to WT (SI Appendix, Fig. S10).

To further demonstrate that the low expression level of miR393b is in fact caused by the lack of MTA, we designed a transient expression assay in Nicotiana benthamiana. We used constructs designed to express pri-miR393b, MTA, or a catalytically inactive version of MTA, ΔMTA (D482A). ΔMTA was prepared by primer-induced point mutation resulting in the following change: Aspartic acid at position 482 to alanine (D482A) at the catalytic DPPW motif (52), making the protein catalytically inactive. These were introduced individually or in combination into Nicotiana leaves by agroinfiltration. We then assessed mature miR393b levels by Northern blot and found that the pri-miRNA393b transgene produces ∼2.3 times more miR393b when it is coexpressed with MTA, while this effect was abolished to a large extent when MTA was replaced by its catalytically inactive version (ΔMTA) (Fig. 7C). Thus, we show that the reduced auxin response in mta mutant plants is partially caused by the regulatory defects of miR393b biogenesis due to the lack of MTA activity.

Plant Material.

A. thaliana WT Col-0, mta mutants (26), MTA-GFP (p35S:MTA-GFP), and GFP plants were sown on Jiffy pots and stratified for 2 d in dark at 4 °C. Thereafter, the plants were grown in plant growth chambers at 22 °C with 16-h light and 8-h dark cycles (70% humidity, 150- to 200-µmol m⁻² s⁻¹ photon flux density). Rosette leaves from 4-wk-old plants were harvested, immediately flash-frozen in liquid nitrogen, and used or stored at −80 °C until further use.

sRNA Sequencing and Analysis.

sRNA fraction from 4-wk-old A. thaliana WT Col-0 and mta mutant plants was used to prepare libraries that were sequenced at Fasteris on the HiSeq 4000 platform. A detailed description of the library preparation and data analysis can be found in SI Appendix, Supplementary Materials and Methods.

m⁶A-RNA IP of pri-miRNAs and Sequencing.

PolyA enriched RNA from 4-wk-old A. thaliana WT Col-0 and mta mutant plants was used to perform m6A IP. Libraries were prepared from RNA before IP (input) and after (IP). A detailed description of the library preparation and data analysis can be found in SI Appendix, Supplementary Materials and Methods.

Data Submission and Availability.

The data obtained in this study (sRNA and m⁶A-IP Seq) has been deposited under the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) accession no. GSE122528. The PIP-seq data used for the secondary structure analysis of miRNA transcripts can be found under NCBI GEO accession no. GSE108852.

RIP.

Transgenic Arabidopsis line p35S:MTA-GFP and mta mutants were grown along with the GFP and WT Arabidopsis lines. RIP was performed as described by Raczynska et al. (63). A detailed protocol for RIP and data analysis can be found in SI Appendix, Supplementary Materials and Methods.

PIP-Seq Analysis to Access RNA Structure Scores of miRNA Processors.

A total of two biological replicates of PIP-seq, including single-stranded RNA-seq and double-stranded RNA-seq libraries, for leaves five to nine from 4-wk-old WT and mta mutant plants were constructed as previously described (37, 39). The raw data has been deposited under the GEO accession no. GSE108852. Refer to SI Appendix, Supplementary Materials and Methods for the detailed protocol and data analysis pipeline.

Immunolocalization.

The immunolabeling experiments using the Duolink PLA fluorescence protocol were performed on isolated nuclei of 4-wk-old A. thaliana leaves (WT, plants with MTA-GFP expression or GFP expression, mta and tgh mutants). Before isolation, the leaves were fixed in 4% paraformaldehyde in PBS, pH 7.2, for 1 h, washed three times in PBS at room temperature, and then the nuclei isolation protocol according to the method of Pontvianne et al. (64) was followed.

Double Immunodetection of RNA Pol II with MTA, DCL1, and HYL1.

A detailed description of double-immunodetection protocol and the accompanying statistical analysis can be found in SI Appendix, Supplementary Materials and Methods.

PLA.

PLA was performed on isolated nuclei from leaves of MTA-GFP and GFP (control) lines. In situ PLA detection was carried out using the appropriate Duolink in situ Orange Kit Goat/Rabbit (Sigma-Aldrich) according to the protocol of the manufacturer. A detailed protocol for PLA can be found in SI Appendix, Supplementary Materials and Methods.

FRET–FLIM.

FRET–FLIM was performed using protoplasts were isolated from 3- to 4-wk-old WT Arabidopsis leaves using the method described in Knop et al. (65). For the detailed protocol, please refer to SI Appendix, Supplementary Materials and Methods.

GUS Staining for Auxin Response Analyses.

Fourteen-day-old seedlings of WT and mta mutants with the DR5pro:GUS reporter gene construct were incubated in 1/2 MS medium containing 10 µM 2,4-D in ethanol or in ethanol alone as a control for 8 h. After GUS staining the pictures were taken using the Leica M60 stereo microscope.

pri-miR393b, MTA, and ΔMTA Constructs.

pri-miR393b (sequence from mirEX², www.combio.pl/mirex2) and MTA cDNA (AT4G10760) were amplified and cloned into pENTR/D-TOPO vectors (Thermo Fischer Scientific). ΔMTA was prepared by a primer-induced point mutation resulting in the following change: Aspartic acid at position 482 was changed to alanine (D482A) at the catalytic DPPW motif. For transient expression in N. benthamiana, these sequences were then cloned into pMDC32 using the Gateway (Thermo Fisher Scientific) cloning system.

N. benthamiana Transient Expression.

pri-miR393b, MTA, and ΔMTA constructs were transformed into Agrobacterium tumefaciens (AGL1) using electroporation. After verification of constructs in Agrobacterium by sequencing, tobacco leaves were transformed as described by Bielewicz et al. (56). Leaves were transformed either with pri-miR393b, MTA, and ΔMTA alone or in combinations (pri-miR393b + MTA, pri-miR393b + ΔMTA); leaves were harvested after 72 h and RNA isolation (as described above) was followed by Northern blotting.

Northern Blotting.

Northern blotting was performed as described in Kruszka et al. (66). Briefly: 30 µg of RNA (per sample) isolated from transfected tobacco leaves was loaded on 8 M denaturing urea polyacrylamide gel (15%) in TBE buffer (0.089 M Tris, 0.089 M boric acid, and 0.002 M EDTA, pH 8.0). RNA was then transferred onto the Amersham Hybond-NX nitrocellulose membrane (GE Healthcare) using a Trans-Blot Electrophoretic Transfer Cell (Bio-Rad) and fixed using CL-1000 UV Crosslinker (UVP). Prehybridization and hybridization were performed in hybridization buffer (3.5% SDS, 0.375 M sodium phosphate dibasic, 0.125 M sodium phosphate monobasic) at 42 °C with DNA oligo probes (Sigma) labeled at their 5′ ends with γ³²P ATP (Hartmann Analytic). U6 was used as a loading control. After washing, the blots were exposed for up to 3 d to a phosphorimaging screen (Fujifilm) and the results were visualized with the Fujifilm FLA5100 reader (Fujifilm) and quantified using Multi Gauge V2.2 (Fujifilm).