The Arabidopsis lncRNA ASCO modulates the transcriptome through interaction with splicing factors

Abstract

Alternative splicing (AS) is a major source of transcriptome diversity. Long noncoding RNAs (lncRNAs) have emerged as regulators of AS through different molecular mechanisms. In Arabidopsis thaliana, the AS regulators NSRs interact with the ALTERNATIVE SPLICING COMPETITOR (ASCO) lncRNA. Here, we analyze the effect of the knock‐down and overexpression of ASCO at the genome‐wide level and find a large number of deregulated and differentially spliced genes related to flagellin responses and biotic stress. In agreement, ASCO‐silenced plants are more sensitive to flagellin. However, only a minor subset of deregulated genes overlaps with the AS defects of the nsra/b double mutant, suggesting an alternative way of action for ASCO. Using biotin‐labeled oligonucleotides for RNA‐mediated ribonucleoprotein purification, we show that ASCO binds to the highly conserved spliceosome component PRP8a. ASCO overaccumulation impairs the recognition of specific flagellin‐related transcripts by PRP8a. We further show that ASCO also binds to another spliceosome component, SmD1b, indicating that it interacts with multiple splicing factors. Hence, lncRNAs may integrate a dynamic network including spliceosome core proteins, to modulate transcriptome reprogramming in eukaryotes.

The Arabidopsis lncRNA ASCO is bound by multiple splicing factors, including spliceosome core components. ASCO deregulation modulates alternative splicing of specific genes, thus shaping the global transcriptome and plant response to flagellin.

Introduction

Alternative splicing (AS) of pre‐mRNAs represents a major mechanism boosting eukaryotic transcriptome and proteome complexity 1. In recent years, the advent of novel sequencing technologies allowed us to analyze entire genomes and complete pools of transcripts, leading to the identification of a wide variety of mRNA isoforms in higher organisms. More than 90% of intron‐containing genes in humans and over 60% in plants are alternatively spliced 2, 3, 4. The significant diversity in the number of transcripts compared to the number of genes suggests that a complex regulation occurs at transcriptional and post‐transcriptional levels 5. Many mRNA isoforms derived from the same DNA locus are tissue‐specific or are accumulated under particular conditions 6. In humans, numerous studies suggest that the misregulation of RNA splicing is associated with several diseases 7, 8, 9, 10. In plants, AS plays an important role in the control of gene expression for an adequate response to stress conditions 11, 12, 13, 14, 15, 16, 17, 18, 19. Alternative splicing modulates gene expression mainly by (i) increasing gene‐coding capacity, thus proteome complexity, through the generation of a subset of mRNA isoforms derived from a single locus and/or (ii) triggering mRNA degradation through the introduction of a premature termination codon in specific isoforms that would lead to nonsense‐mediated decay (NMD). Besides the finding of an increasing number of AS events on mRNAs, next‐generation sequencing technologies led to the identification of thousands of RNAs with no or low coding potential (the so‐called noncoding RNAs, ncRNAs), which are classified by their size and location with respect to coding genes 20. The long ncRNAs (lncRNAs, over 200 nt) act directly in a long form or may lead to the production of small ncRNAs (smRNAs) acting through base pairing recognition of their mRNA targets. There is growing evidence that large amounts of lncRNAs accumulate in particular developmental conditions or during diseases, suggesting that they participate in a wide range of biological processes. In recent years, several lncRNAs from higher organisms have been characterized as modulators of virtually every step of gene expression through interaction with proteins involved in chromatin remodeling, transcriptional control, co‐ and post‐transcriptional regulation, miRNA processing, and protein stability during various developmental processes 20, 21, 22. In particular, a growing number of lncRNAs have been linked to the modulation of AS in both plants and animals 23. The main mechanisms involving lncRNAs in AS modulation have been classified as follows: (i) lncRNAs interacting with splicing factors 24, 25, 26, 27, 28; (ii) lncRNAs forming RNA–RNA duplexes with pre‐mRNA molecules 29, 30; and (iii) lncRNAs affecting chromatin remodeling of alternatively spliced target genes 31, 32.

In Arabidopsis thaliana, the lncRNA ASCO (ALTERNATIVE SPLICING COMPETITOR; AT1G67105) is recognized in vivo by the plant‐specific NUCLEAR SPECKLE RNA‐BINDING PROTEINS (NSRs), involved in splicing 24. Interestingly, there is no evidence that ASCO undergoes splicing, although it is recognized by splicing factors. The analysis of a transcriptomic dataset of the nsra/b mutant compared to wild‐type (WT) plants revealed an important number of intron retention events and differential 5′ start or 3′ end in a subset of genes, notably in response to auxin 33. Indeed, the nsra/b mutant exhibits diminished auxin sensitivity, e.g., lower lateral root (LR) number than WT plants in response to auxin treatment. This phenotype was related to the one observed for ASCO overexpressing lines. Interestingly, the splicing of a high number of auxin‐related genes was perturbed in nsra/b mutants and several of them behaved accordingly in the ASCO overexpressing lines. The ASCO‐NSR interaction was then proposed to regulate AS during auxin responses in roots 24. More recently, an RNA immunoprecipitation assay followed by RNA‐seq (RIP‐seq) served to identify genome‐wide RNAs bound in vivo by NSRa 34. Long ncRNAs transpired to be privileged direct targets of NSRs in addition to specific NSR‐dependent alternatively spliced mRNAs, suggesting that other lncRNAs than ASCO may interact with NSRs to modulate AS 34.

In this work, we thoroughly characterize ASCO knocked‐down plants and present its general role in AS regulation, not only in response to auxin treatment. A transcriptomic analysis of ASCO knocked‐down seedlings revealed a misregulation of immune response genes and, accordingly, ASCO RNAi‐silenced plants exhibited enhanced root growth sensitivity to flagellin 22 (flg22). The transcriptomic analysis of the ASCO overexpressing versus ASCO knocked‐down seedlings revealed distinct and overlapping effects on the entire mRNA population. Assessing the genome‐wide impact of ASCO function on AS, we found many flg22‐response regulatory genes to be differentially alternatively spliced in ASCO‐deregulated lines. Surprisingly, the effect of ASCO knock‐down on AS was clearly distinct from the defects of the nsra/b double mutant, suggesting that ASCO impacts AS through a different interaction with the splicing machinery. Searching for ASCO‐interacting proteins, we found SmD1b and PRP8a, two core components of the spliceosome that recognize subsets of AS‐regulated flg22‐regulatory genes, also differentially spliced in prp8‐7 35 and smd1b mutants. Furthermore, ASCO overexpression competes for PRP8a binding to particular mRNA targets. Hence, lncRNAs may interact with key conserved components of the spliceosome to integrate a dynamic splicing network that modulates transcriptome diversity in eukaryotes.

Results

The ASCO lncRNA participates in lateral root formation

It was previously shown that ASCO overexpression results in a lower number of LRs in response to auxin treatment, a phenotype related to that of the nsra/b mutant, suggesting that increasing ASCO expression may lead to a titration of NSR activity in splicing 34. To understand the role of ASCO in plant development, we generated independent RNAi lines to downregulate the levels of ASCO expression (RNAi‐ASCO1 and RNAi‐ASCO2, Fig EV1A and B). Under control growth conditions, RNAi‐ASCO plants do not exhibit significant changes in primary root growth when compared to Col‐0 WT (Fig EV1C), whereas both independent lines showed an enhanced LR density in response to auxin treatment (Fig EV1D), the opposite phenotype to the one displayed by the ASCO overexpressing lines 24. Furthermore, we transformed A. thaliana with a construct bearing 2,631 bp of the ASCO promoter region controlling the expression of the fusion reporter genes GFP‐GUS (proASCO::GFP‐GUS). The proASCO construct includes the full intergenic region upstream of ASCO in addition to the first fifteen nucleotides from the transcription start site of the ASCO locus (position of the first ATG found in the locus) fused to the reporter genes (Fig EV1E). In roots, proASCO::GFP‐GUS was active very early in LR development, in pericycle cells undergoing the first division (Fig EV1F), whereas activity was then restricted to the vasculature adjacent to the LR primordium between stages II and VIII of LR development 36. Thus, ASCO expression pattern is in agreement with the LR‐related phenotype of RNAi‐ASCO plants.

Figure EV1. The response to auxin is altered in RNAi‐ASCO lines.

• A
  ASCO transcript levels in two independent 14‐day‐old RNAi‐ASCO lines compared to WT. Error bars represent standard deviation. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates).
• B
  RNA‐seq read coverage on ASCO lncRNA in RNAi‐ASCO and WT seedlings RNA‐seq data. The region cloned to generate dsRNA is indicated on the gene structure. Coverage plots were made with IGV software 115 using normalized (read per million) bigwig files. The same scale (0–200) was used for each track.
• C, D
  Primary root length (C) and lateral root density (D) of WT and two independent RNAi‐ASCO lines grown on media with or without 100 nM NAA and measured 7 days after germination. The asterisk (*) indicates a significant difference as determined by Mann–Whitney's U‐test (P < 0.05, n = 16 biological replicates). Errors bars show mean value ± standard deviation.
• E
  Scheme of the transcriptional fusion to GFP:GUS used to analyze the ASCO promoter activity.
• F
  Activity of proASCO during LR development. Scale bar corresponds to 20 μm.

Deregulation of ASCO expression triggers a transcriptional response to biotic stress

In order to decipher the role of ASCO in the regulation of gene expression at a genome‐wide level, we performed RNA‐seq with A. thaliana 14‐day‐old seedlings RNAi‐ASCO1 versus WT Columbia (Col‐0) accession in standard growth conditions. Overall, more genes were upregulated (321) than downregulated (178) in ASCO‐silenced plants (Fig 1A). Over 90% of deregulated transcripts correspond to protein‐coding genes, according to Araport11 gene annotation (Fig 1B; Table EV1). To extend our understanding on the genome‐wide role of ASCO in the control of gene expression, we searched the putative function of differentially expressed genes using Gene Ontology (GO). This analysis revealed a clear enrichment of deregulated genes involved in immune and defense responses (FDR < 8e‐4), as well as related pathways such as “response to chitin” and “glucosinolate metabolic pathways” (Fig 1C). Interestingly, related pathways were also partially observed in nsra/b mutants in response to auxin 34. The upregulation of biotic stress‐related genes was validated by RT–qPCR in both RNAi‐ASCO lines compared to WT for a subset of 6 chosen transcription factors (TFs) which have been linked to the response to pathogens (Fig EV2A): STZ/ZAT10 (AT1G27730) encodes for a Zn–finger TF involved in the response to oxidative stress 37 and acts as a negative regulator of methyl jasmonate (MeJA) biosynthesis 38, MYB29 (AT5G07690) positively regulates the biosynthesis of aliphatic glucosinolate (AGSL), an essential defense secondary metabolite in A. thaliana 39, WRKY33 (AT2G38470) controls the ABA biosynthetic pathway in response to the necrotrophic fungi Botrytis cinerea 40, ERF6 (AT4G17490) is a positive regulator of the MeJA and ethylene‐mediated defense against B. cinerea 41, ERF104 (AT5G61600) participates in the ethylene‐dependent response to flg22 42, and ERF105 (AT5G51190) was shown to be strongly regulated in response to chitin 43 and to bind to the GCC‐box pathogenesis‐related promoter element 44. Remarkably, all of these pathogen‐related TFs are transcriptionally overaccumulated in control conditions in the RNAi‐ASCO plants (Fig EV2A), indicating that the deregulation of ASCO expression triggers molecular defense responses likely through the induction of pathogen‐related TFs.

Figure 1. ASCO modulates steady‐state levels of transcripts involved in plant immune responses affecting the sensitivity to flg22 peptide.

1. Number of differentially up‐ and downregulated genes (DEG) in RNAi‐ASCO1 seedlings as compared to wild type (WT) according to the RNA‐seq data (FDR < 0.01, |log2FC| ≥ 0.75).
2. Fraction of DEG found in each transcript class as defined in the Araport11 gene annotation. AS stands for antisense.
3. Enriched Gene Ontology (GO) of DEG in RNAi‐ASCO1 seedlings as compared to WT, x‐axis represents the ‐log10 FDR for the enrichment of each GO category over genome frequency.
4. Representative picture of 14‐day‐old plants grown 9 days in liquid 1/2MS supplemented with 1 μM flg22. The scale bar representing 0.6 cm is included in the picture.
5. Lateral root density of WT and two independent RNAi‐ASCO lines 9 days after transfer in 1/2MS supplemented with 0.1 or 1 μM flg22.
6. Representative picture of root apical meristems after cell wall staining, in response to flg22. TZ: transition zone; QC: quiescent center.
7. Root apical meristem size of WT and RNAi‐ASCO1 (e.g distance from QC to TZ in μm).

Data information: Error bars indicate the standard error. The asterisk (*) indicates a significant difference as determined by Mann–Whitney's U‐test (P < 0.05, n = 18 biological replicates).

Figure EV2. Expression of immunity‐related TF in RNAi‐ASCO lines and ASCO expression in roots during flg22 treatment.

• A
  Transcript levels of stress‐related DEG identified in the RNA‐seq data in the WT and the two independent RNAi‐ASCO lines, measured by RT–qPCR. RNAs were extracted from 14‐day‐old plants. Data are mean of 3 independent biological replicates. Error bars represent standard deviation. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates).
• B, C
  Time–course analysis of ASCO (B) and CYP81F2 (C, AT5G57220) expression levels after treatment with media supplemented or not with 1 μM flg22. RNAs were extracted from 10‐day‐old plants. Data are mean of 3 independent biological replicates. Error bars represent standard deviation.

It is known that peptides corresponding to the most conserved domains of eubacterial flagellins (flg) act as potent elicitors in A. thaliana. Notably, flg22 causes callose deposition, induction of genes encoding for pathogenesis‐related proteins, and a strong inhibition of growth including root development 45, 46, 47. Thus, we first assessed the transcriptional accumulation over time of ASCO in response to flg22. As shown in Fig EV2B, ASCO accumulation in roots was not significantly affected by flg22, compared to CYP81F2 used as a positive control (Fig EV2C) 48. Then, we characterized the physiological response of both ASCO RNAi‐silenced lines to the exogenous treatment with flg22. Five‐day‐old plantlets were treated or not for 9 additional days with 0.1 or 1 μM flg22. Strikingly, the roots of RNAi‐ASCO1 and 2 plants exhibited a normal development in control conditions (Fig EV3A and B), whereas they were more sensitive to flg22 treatment, exhibiting a significantly shorter primary root (Figs 1D and EV3C) but a minor reduction in the number of total LRs, resulting in a higher density of LRs (Fig 1E). Cell wall staining and microscopic observation allowed us to quantify meristem size and determine that RNAi‐ASCO plants show a reduction of the meristematic zone in response to flg22, e.g., shorter distance between the quiescent center and the beginning of the transition zone (Fig 1F and G). This reduction in size is the result of a significantly lower number of cells forming the root meristematic zone (Fig EV3D). Together with the physiological phenotype, we characterized the molecular response to this elicitor. A small subset of flg22‐responsive genes was chosen 49, 50, 51 to assess putative expression changes due to ASCO knock‐down. In mock conditions, RNAi‐ASCO lines exhibited an increased expression for certain flg22‐responsive genes tested (Fig EV3E). Interestingly, this subset of genes suffered an overall lower induction after 3 h of flg22 in RNAi‐ASCO plants (Fig EV3F), in agreement with the previously observed altered sensitivity to flg22 of RNAi‐ASCO roots.

Figure EV3. The response to flg22 is altered in RNAi‐ASCO lines.

1. Representative picture of 14‐day‐old plants grown 9 days in liquid 1/2MS corresponding to Mock condition. The scale bar representing 0.6 cm is included in the picture.
2. Representative picture of root apical meristems after cell wall staining, in mock condition. TZ: transition zone; QC: quiescent center.
3. Primary root length of WT and two independent RNAi‐ASCO lines 9 days after transfer in &frac12; MS supplemented with 0.1 or 1 μM flg22. The asterisk (*) indicates a significant difference as determined by Mann–Whitney's U‐test (P < 0.05, n = 18).
4. Meristematic cell number in WT and RNAi‐ASCO1 primary root apex, treated or not with 1 μM flg22. The asterisk (*) indicates a significant difference as determined by Mann–Whitney's U‐test (P < 0.05, n = 18 biological replicates).
5. Relative transcript levels of a subset of flg22 responsive genes in control conditions (AGIs are indicated in Dataset EV5). RNAs were extracted from 10‐day‐old plants. Data are mean of three independent biological replicates. Error bars represent standard deviation. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates).
6. Relative modulation of the same genes as in (E), in response to flg22. Results are normalized against the response in WT plants, taken as 100%. Error bars represent standard deviation. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates).

To further demonstrate the link between ASCO and the response to flg22, we searched for additional independent Arabidopsis lines exhibiting a deregulation in ASCO accumulation. We characterized two insertional mutants located at the 5′ region (asco‐1) and the 3′ region of the locus (asco‐2; Fig EV4A). The first line, asco‐1, resulted in an overexpressor of a truncated ASCO version (lacking a minor portion of the 5′ region), whereas the asco‐2 T‐DNA line shows minor changes in ASCO expression (Fig EV4B and C). Interestingly, the nearly 50‐fold overaccumulation of ASCO RNA in asco‐1 plants do not yield any significant root growth phenotype, possibly a slight reduction in LR density (Fig EV4D and E). Accordingly, when we assessed two independent 35S:ASCO overexpressing lines, reaching an overaccumulation of 1,000–2,500‐fold RNA levels (Fig EV4B), plants exhibit a longer main root and a lower density of LRs in response to flg22 (Fig EV4F and G). Therefore, ASCO participates in the regulation of biotic stress‐related genes, shaping root architecture in response to flg22.

Figure EV4. ASCO insertion mutants do not exhibit drastic changes in RNA accumulation.

• A
  Scheme of T‐DNA insertions along ASCO locus in asco‐1 and asco‐2 mutants. Red arrows indicate probes used for qPCR expression analysis.
• B
  Relative ASCO transcript levels in two independent 14‐day‐old 35‐ASCO lines, asco‐1, and asco‐2 compared to WT. Error bars represent standard deviation. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates).
• C
  Quantification of ASCO transcript levels targeting different regions of the locus by RT–qPCR in WT and asco mutants. A, B, C, and D probe positions are indicated in (A). The error bars represent standard deviation between 3 biological replicates, and (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates).
• D, E
  Primary root length (D) and lateral root density (E) of WT, asco‐1, and asco‐2 plants 9 days after transfer in 1/2MS supplemented with 1 μM flg22. The asterisk (*) indicates a significant difference as determined by Mann–Whitney's U‐test (P < 0.05, n = 24 biological replicates). Errors bars show mean value ± standard deviation.
• F, G
  Primary root length (F) and lateral root density (G) of WT, two RNAi‐ASCO, and two 35S:ASCO independent lines 9 days after transfer in 1/2MS supplemented with 1 μM flg22. Error bars indicate the standard error. The asterisk (*) indicates a significant difference as compared to the WT, determined by Mann–Whitney's U‐test (P < 0.05, n = 24 biological replicates). Error bars show mean value ± standard deviation.

ASCO modulates the alternative splicing of a subset of pathogen‐related mRNAs with unaltered accumulation

Considering that overexpression of ASCO affected the AS of NSR mRNA targets 24, we searched for mis‐spliced genes potentially explaining the global physiological impact of ASCO deregulation. To this end, we used two complementary approaches to detect both differential AS based on annotated isoforms (Reference Transcript Dataset for A. thaliana, AtRTD2) 52 and potentially nonannotated differential RNA processing events using RNAprof 33, 52. Based on RNAprof, a total of 303 differential RNA processing events in 281 distinct genes were identified comparing RNAi‐ASCO with WT plants in control growth conditions (Dataset EV2), whereas the SUPPA2 method 53 identified 205 genes with evidence of differential AS in the AtRTD2 database. Comparison of the two analyses with differentially expressed genes (DEG) in RNAi‐ASCO lines revealed that most differentially alternatively spliced (DAS) genes are not differentially accumulated (Fig 2A). In addition, our analyses showed the complementarity between the two approaches since only 24 common DAS genes were identified by both methods. Classification of the location and the relative isoform accumulation (up or down) of these events revealed that the majority of them were located in introns and had higher read coverage in RNAi‐ASCO plants, suggesting that ASCO inhibited proper intron splicing on these genes (Fig 2B). Nevertheless, we also identified differential events located within 5′UTR, CDS, or 3′UTR suggesting that other RNA processing events, in addition to intron retention, such as alternative transcription start sites or polyadenylation sites, are affected by ASCO expression levels. Analysis of differential AS events with SUPPA revealed 317 significant DAS events (|dPSI| > 0.1, P < 0.01) on 205 unique genes from the AtRTD2 transcript annotation database (Dataset EV3). Similarly to the analysis with RNAprof, most of these events corresponded to intron retention (62%) but we also identified a significant number of alternative 3′ splice site and alternative 5′ splice site selection modulated in ASCO knock‐down lines (Fig 2C). To determine the most significant impact of the AS events, we sought to identify isoform switching events (i.e., co‐ordinated variations in abundance of two isoforms) using the IsoformSwitchAnalyzeR package (Dataset EV4) 54. Strikingly, isoform switching events were detected for 52 genes, out of which 12 and 34 were common cases detected by RNAprof and AtRTD2‐SUPPA, respectively (Fig 2D). In silico analysis of the protein sequences derived from switching isoforms indicated that the AS events may lead to (i) change of ORF length, (ii) gain or loss of conserved PFAM protein domain and signal peptides, and (iii) change of the coding potential and the sensitivity to NMD (Fig 2E). Since AS can often trigger NMD, an important mechanism of plant gene expression regulation 55, we compared DAS genes to those transcripts overaccumulated in the double mutant of the NMD factor homologs UP FRAMESHIFT1 (UPF1) and UPF3, upf1‐upf3 56. As shown in Fig 2F, 66 and 29 genes regulated by NMD were reported by RNAprof and AtRTD2‐SUPPA as alternatively spliced in RNAi‐ASCO plants, respectively. Hence, the majority of AS events controlled by ASCO seem to be independent of the UPF1‐UPF3‐mediated RNA quality control machinery, at least in the conditions previously assessed.

Figure 2. ASCO modulates alternative splicing.

1. Comparison of differentially processed transcripts (RNAprof) and differential AS genes (AtRTD2‐SUPPA) with differentially expressed genes (DEG).
2. Number of genes containing at least one differential RNA processing event (as defined by RNAprof P _adj < 0.001) in CDS, introns, 5′UTR and 3′UTR between RNAi‐ASCO1, and WT. Up and down fractions correspond to increase or decrease, respectively, of RNA‐seq coverage in RNAi‐ASCO1 for each specified gene feature.
3. Proportion of DAS events identified by AtRTD2‐SUPPA in RNAi‐ASCO1 compared to WT; alternative 3′ site (A3′), alternative 5′ site (A5′), intron retention (IR), exon skipping (ES).
4. Comparison of differentially processed transcripts (RNAprof) and differentially AS genes (AtRTD2‐SUPPA) with genes showing significant isoform switch events (Isoswitch).
5. Summary of the predicted consequence of the isoform switch events as shown by the feature acquired by the upregulated isoform. ncRNA stands for noncoding RNA, and NMD stands for nonsense‐mediated decay.
6. Comparison of differentially processed transcripts (RNAprof), differentially AS genes (AtRTD2‐SUPPA), and differentially expressed genes (DEG) with genes significantly upregulated in the upf1‐upf3 mutant 56, indicating genes potentially regulated by NMD.

Furthermore, we performed RNA‐seq with 14‐day‐old seedlings 35S:ASCO1 versus Col‐0 WT plants in standard growth conditions. Interestingly, there is a minimal overlap between DEG and DAS genes in WT versus RNAi‐ASCO1 and 35S:ASCO1. Strikingly, the up‐ and down‐deregulation of ASCO resulted in alternative subsets of DAS genes, including only 120 common DAS between RNAi‐ASCO1 and 35S:ASCO1, compared to 227 and 137 excluding events, respectively (Fig 3A). Further comparison of DEG fold change revealed a global correlation of gene expression changes in 35S:ASCO1 and RNAi‐ASCO1 as compared to WT. However, we show that particular subsets of genes responded to the down‐ or upregulation of ASCO (Fig 3B). Similarly, in these lines we compared the dPSI (difference of Percent Spliced In), which represents the change of each AS event. The analysis revealed that the group of 120 common AS events are positively correlated between the two lines as compared to wild type (Fig 3C). In addition, this also revealed that dPSI of AS events significantly regulated in response to either overexpression or silencing of ASCO, respectively, was not correlated between the two lines (Fig 3C). Overall, the effect of ASCO silencing was more extensive on DAS, compared to its overexpression.

Figure 3. RNAi‐ASCO and 35S:ASCO lines share common and distinct subsets of DEG and DAS targets.

1. Overlap between differentially expressed (DEG) and spliced (DAS) genes in RNAi‐ASCO and 35S:ASCO as compared to WT.
2. Scatter plot showing the respective gene expression fold change in RNAi‐ASCO and 35S:ASCO lines as compared to WT. Genes showing significant changes in RNAi‐ASCO and 35S:ASCO are highlighted as yellow and blue dots, respectively.
3. Scatter plots showing the respective percent spliced in difference (dPSI) in RNAi‐ASCO and 35S:ASCO lines as compared to WT. Genes showing significant changes in RNAi‐ASCO, 35S:ASCO, or in both lines are highlighted as red, green, or blue dot, respectively. Gray dots represent all AS events.

In order to better understand the impact of ASCO deregulation on the plant response to flg22, we focused on the transcriptional accumulation of specific genes. Strikingly, several pathogen‐related genes appeared differentially spliced in the RNAi‐ASCO1 line although they were not affected in their global expression levels (Fig 3A). These AS events include two members from the NB‐LRR disease resistance genes: RPP4 57 and RLM3 58, as well as the splicing regulatory serine‐rich protein‐coding gene SR34 (AT1G02840), needed for accurate response to pathogens 59. The splicing of the SR34 own pre‐mRNA is auto‐regulated and depends on the activity of immune response factors 60. Other relevant AS targets are SNC4 (AT1G66980) which encodes a receptor‐like kinase that participates in the activation of the defense response, and its AS is impaired in defense‐related mutants, affecting the response to pathogens 60; SEN1 (AT4G35770), a senescence marker gene primarily regulated by salicylic acid (SA)‐ and MeJA‐dependent signaling pathways 61 and NUDIX HYDROLASE7 (NUDT7, AT4G12720) which regulates defense and cell death against biotrophic pathogens 62. Another interesting target is EPITHIOSPECIFIER PROTEIN (ESP, AT1G54040) a gene involved in plant defense to insects which is differentially spliced in response to MeJA 63 although this gene was also differentially expressed in RNAi‐ASCO plants. We also included in the following analysis a NAD(P)‐binding Rossmann‐fold protein family gene (NRG, AT2G29290), exhibiting drastically altered AS upon ASCO knock‐down. DAS events in the chosen genes mentioned above, first identified in silico (Fig 4A and D, Appendix Fig S1), were validated by RT–PCR and polyacrylamide gel electrophoresis by calculating the ratio between alternatively spliced and fully spliced isoforms (isoform ratio, Figs 4B and C, and 4E and F, Appendix Fig S2). All events tested excepting RLM3 displayed significant changes in isoform ratio. Most events led to changes in conserved protein domains (Fig 4 and Appendix Fig S2). For instance, retention of the last intron in RPP4 gave place to a protein predicted with a lower number of LRR repeat as previously observed in other R‐genes such as RPS4 64. Splicing events in SR34 and ESP were further validated by quantitative RT–qPCR where each differential event was normalized with respect to an internal gene probe (called INPUT) which corresponds to a common exon. This allowed for the calculation of the splicing index (defined in the methods section, Appendix Fig S3). Splicing index was not calculated for the other chosen genes due to technical difficulties in primer designing. Altogether, our results indicate that the knock‐down of ASCO expression affects the AS of a subset of genes whose isoforms distribution may modulate the pathogen‐related transcriptome and affect the response to flg22.

Figure 4. Validation of AS events in RNAi‐ASCO lines.

• A–F
  (A, D) Differential RNA processing events of SR34 (AT1G02840) and NUDT7 (AT4G12720) transcripts detected by RNAprof from the comparison of RNA‐seq libraries of 14‐day‐old WT (red) and RNAi‐ASCO1 (blue) plants. Three biological replicates were used. Vertical purple lines and P‐values indicate significant differential processing events. Structure of SR34 (A) and NUDT7 (D) RNA isoforms. Large black boxes indicate exons, narrow black boxes indicate UTRs, and black lines indicate introns. Colored boxes indicate protein domains affected by an AS event. RS domain: Arg/Ser‐rich domain. Red arrows indicate probes used for gel electrophoresis. Protein domains were retrieved from Uniprot database (https://www.uniprot.org). (B, E) Analyses of RT–PCR products of SR34 (B) and NUDT7 (E) transcripts on 8% acrylamide gel. (C, F) Quantification of the ratio of SR34 (C) and NUDT7 (D) isoforms detected in the gel in (B) and (E), respectively. RNAs were extracted from WT and RNAi‐ASCO1 14‐day‐old plants. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates). Error bars show mean ± standard deviation.

ASCO interacts with the spliceosome core components PRP8a and SmD1b

The fact that ASCO interacts with NSRs strongly suggested that its deregulation would affect a large subset of NSR‐targeted AS events. Surprisingly, DAS genes in the RNAi‐ASCO and 35S:ASCO plants only partially coincide with those in nsra/b double mutants (in response to auxin or not). In all, out of the 589 DAS events identified in nsra/b compared to WT, only 140 (32%) are common with RNAi‐ASCO1 and 109 (33%) with 35S:ASCO, representing 24 and 19% of all DAS events in the nsra/b mutant, respectively (Fig EV5A). Furthermore, nsra/b plants do not respond to flg22 in the same way as RNAi‐ASCO plants (Fig EV5B and C), indicating that ASCO further modulates AS in an NSR‐independent manner by an unknown mechanism, notably affecting the response to biotic stress. Therefore, in order to decipher the AS‐related complexes implicating ASCO, we performed an antisense oligonucleotide‐based pull‐down method, related to the chromatin isolation by RNA Purification (ChIRP) 65, 66 using nuclear extracts to purify ASCO‐containing RNPs. Eighteen biotinylated probes matching ASCO were used in independent sets called EVEN and ODD, respectively (Dataset EV5). ASCO‐ChiRP was performed in 4 biological replicates for each set of probes: ODD, EVEN, and an additional set matching the LacZ RNA, used as a negative control. ASCO enrichment was corroborated by ChiRP followed by RNA purification and RT–qPCR (Fig 5A). We then performed a mass spectrometry on the proteins from the purified ASCO‐containing RNP to identify potential ASCO protein partners. Strikingly, among the RNA‐related proteins identified in the EVEN and ODD samples, but not in the LacZ, we found the pre‐mRNA‐processing‐splicing factor 8A, PRP8a. PRP8 is a core component of the spliceosome and is highly conserved in higher organisms; null mutations generally result in embryonic lethality 67. In Arabidopsis, a PRP8a leaky mutation was found to also affect the AS of the COOLAIR lncRNA 68 and results in a high number of intron retention events 35. Therefore, we developed specific antibodies against PRP8a and we tested them in immunolocalization experiments that revealed a nuclear localization pattern (Appendix Fig S4) similar to what was previously observed in Drosophila 69. In order to validate the interaction with ASCO in vivo, we performed a RNA immunoprecipitation assay followed by qPCR (RIP–qPCR) from nuclear extracts. We show that PRP8a can recognize the spliceosomal U5 RNA 70 taken as a positive control, as well as the ASCO lncRNA (Fig 5B). The efficiency of the PRP8a immunoprecipitation (IP) was assessed by Western blot comparing nuclei input samples, against the unbound fraction after IP, as well as the anti‐PRP8a IP and the anti‐IgG IP (Fig 5C). We then assessed the binding of PRP8a to the pathogen‐related mRNAs differentially spliced in the RNAi‐ASCO lines. PRP8a was indeed able to interact with 4 of these ASCO‐related DAS genes. Furthermore, their binding was impaired upon the overexpression of ASCO (Fig 5D), hinting at an ASCO‐mediated competition of these mRNAs inside the PRP8a‐containing spliceosome complex. Interestingly, ASCO is overaccumulated in the prp8‐7 mutant allele 35 (Fig 6A), as it occurs in the nsra/b mutant plants 24. Remarkably, similar AS defects were shown between the prp8‐7 mutant and RNAi‐ASCO lines for pathogen‐related genes (Fig 6B and C, Appendix Fig S5), indicating that the flg22 differential phenotype of RNAi‐ASCO plants may be related to the interaction with the spliceosome components. Recently, we identified another core component of the spliceosome, SmD1b, linked both to AS and the recognition of aberrant ncRNAs to trigger gene silencing 71. Interestingly, ASCO expression levels are also increased in the smd1b mutant (Fig 6D), exhibiting the same transcript accumulation as in prp8‐7 and nsra/b mutants. Hence, we wondered whether this other core component of the spliceosome also interacts with ASCO lncRNA. Using pUBI:SmD1b‐GFP plants (smd1b mutant background) 71, we performed a RIP assay and found that SmD1b also recognizes ASCO in vivo (Fig 6E) as well as the U6 RNA taken as a positive control. Furthermore, SmD1b recognizes the four pathogen‐related transcripts assessed (Fig 6F) although only two out of three pathogen‐related transcripts assessed were DAS in smd1b mutants: ESP, SR34, but not NUDT7 (Fig 6G and H, Appendix Fig S5). SNC4 total transcript levels were dramatically reduced in the smd1b mutant, hindering the analysis of relative isoforms accumulation (Appendix Fig S5J). Altogether, our results indicate that ASCO, an apparently intron‐less lncRNA, interacts with PRP8a and SmD1b, two core components of the spliceosome, contributing to determine the dynamic ratio between hundreds of alternatively spliced mRNAs, notably pathogen‐related genes. Hence, lncRNAs appear as possible dynamic interactors of multiple core components of the splicing machinery, likely modulating the splicing patterns of particular subsets of mRNAs.

Figure EV5. nsra/b affects root responses to flg22 and splicing of subsets of genes partially overlapping with RNAi‐ASCO and 35S:ASCO .

• A
  Overlap between differentially spliced events in RNAi‐ASCO, 35S:ASCO and WT and nsra/b mutant, and WT treated with NAA 24.
• B, C
  Primary root length (B) and lateral root density (C) of WT and nsra/b double mutants 9 days after transfer in 1/2MS supplemented with 1 μM flg22. Error bars show mean value ± standard deviation. The asterisk (*) indicates a significant difference as determined by Mann–Whitney's U‐test (P < 0.05, n = 16 biological replicates).

Figure 5. ASCO modulates the activity of the spliceosome component PRP8a.

1. Analysis of ASCO enrichment by ChIRP using two sets of independent biotinylated probes ODD and EVEN compared to negative control with probes designed against the LacZ RNA. The fold enrichment was calculated between ODD or EVEN samples against LacZ. These samples were used for protein precipitation and mass spectrometry analyses (ChIRP‐MS), from which PRP8a was identified as a potential ASCO partner.
2. Validation of PRP8a‐ASCO interaction by PRP8a‐RIP. U5 RNA was used as a positive control.
3. Immunoblot analysis was performed during PRP8a immunoprecipitation (IP). The same volume of input, unbound fraction was loaded as well as 20% of the eluted IP fraction. α‐PRP8a and α‐IgG : IP performed with anti‐PRP8a or control rabbit IgG antibody, respectively.
4. PRP8a recognition of a subset of DAS RNAs is impaired in the 35S:ASCO plants. A housekeeping gene (PP2A, AT1G13320) RNA was used as a negative control.

Data information: In (B and D), the results are expressed as a percentage of the input for PRP8a RIP followed by RT–qPCR, and IgG RIP was used as a negative control. Error bars represent the standard error of 3 biological replicates. N/D stands for nondetectable. The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3).

Figure 6. PRP8a and SmD1b regulate AS of ASCO mRNA targets.

• A
  ASCO transcript levels in WT and prp8‐7 mutants.
• B, C
  prp8‐7 leaky mutant displays similar AS events as observed in RNAi‐ASCO. Quantification of SR34 (B) and ESP (C) isoforms splicing index by RT–qPCR.
• D
  ASCO transcript levels in smd1b mutant. In A and D, RNAs were extracted from WT, prp8‐7, and smd1b 14‐day‐old plants.
• E
  SmD1b can bind ASCO in vivo. U6 RNA was used as a positive control and a housekeeping gene (HKG2, AT4G26410) RNA as a negative control. The results were expressed as % INPUT in SmD1b‐GFP RIP and IgG RIP used as a negative control.
• F
  SmD1b recognizes in vivo the RNAs of 4 genes regulated by ASCO. The results were expressed as % INPUT in SmD1b‐GFP RIP and IgG RIP used as a negative control.
• G, H
  smd1b mutant displays similar AS events as observed in RNAi‐ASCO. Quantification of SR34 (G) and ESP (H) isoforms splicing index by RT–qPCR.

Data information: The asterisk (*) indicates a significant difference as determined by Student's t‐test (P < 0.05, n = 3 biological replicates). Error bars show mean ± standard error.

Discussion

Long noncoding RNAs modulate splicing regulatory networks

We show here that reducing ASCO expression has a major effect on AS at the genome‐wide level in plants. In animals, different splicing factors can recognize lncRNAs in vivo 23, e.g., Y‐BOX BINDING PROTEIN 1 (YBX1) 72, POLY(RC) BINDING PROTEINS 1 and 2 (PCBP1/2), FOX proteins 73, and the serine‐rich splicing factors, such as SRSF6 74, among others. The lncRNA GOMAFU, for example, is recognized through a tandem array of UACUAAC motifs by the splicing factor SF1, which participates in the early stages of spliceosome assembly 75. Furthermore, GOMAFU was found to directly interact with the splicing factors QUAKING homolog QKI and SRSF1 25. In adult mice, GOMAFU is expressed in a specific group of neurons and has been implicated in retinal cell development 76, 77, brain development 78, and post‐mitotic neuronal function 79. GOMAFU's downregulation leads to aberrant AS patterns of typically schizophrenia‐associated genes 25. Other lncRNAs recognized by splicing factors are NUCLEAR PARASPECKLE ASSEMBLY TRANSCRIPT 1 (NEAT1) and NEAT2 (also known as METASTASIS ASSOCIATED LUNG ADENOCARCINOMA TRANSCRIPT 1; MALAT1) 80. RNA FISH analyses revealed an intimate association of NEAT1 and MALAT1 with the SC35 splicing factor containing nuclear speckles in both human and mouse cells, suggesting their participation in pre‐mRNA splicing. Indeed, the ASCO lncRNA also interacts with NSRs and SmD1b both localized in nuclear speckles 24, 71, whereas we show here that PRP8a seems to have nuclear localization in Arabidopsis. It was shown that NEAT1 localizes to the speckles periphery, whereas MALAT1 is part of the polyadenylated component of nuclear speckles 80. MALAT1 acts as an oncogene transcript, and its aberrant expression is involved in the development and progression of many types of cancers 81, 82, 83. MALAT1 can promote metastasis by interacting with the proline‐ and glutamine‐rich splicing factor SFPQ, blocking its tumor suppression activity 26. In plants, little is known about the interaction between splicing factors and lncRNAs 20, 23. NSRs are a family of RNA‐binding proteins that act as regulators of AS and auxin‐regulated developmental processes such as lateral root formation in A. thaliana. These proteins were first shown to interact with some of their alternatively spliced pre‐mRNA targets and ASCO lncRNA 24. More recently, a RIP‐seq approach on an NSRa fusion protein in A. thaliana mutant background allowed the identification of genome‐wide NSR targets, e.g., specific alternatively spliced mRNAs as well as a plethora of lncRNAs, including ASCO 34. Strikingly, ASCO was detected albeit not among the most abundant NSRa‐interacting lncRNA, suggesting the existence of an intricate network of multiple lncRNAs and splicing factors interactions. In fact, we showed here that the impact of ASCO deregulation on AS at a genome‐wide level barely overlaps with the defects observed in the nsra/b mutant background (with or without auxin), indicating that ASCO and NSRs participate in common as well as in independent molecular mechanisms related to AS.

The ASCO lncRNA knocked‐down plants show altered sensitivity to flagellin

The comparison of the transcriptome of RNAi‐ASCO and 35S:ASCO plants revealed common and specific subsets of DAS genes. This dual effect caused by the up‐ or downregulation of ASCO accumulation hints to the potential relevance of a stoichiometric factor impacting the action of ASCO within the spliceosome. ASCO‐silenced plants exhibit an enhanced sensitivity to flg22, in contrast to 35S:ASCO and nsra/b plants. In agreement, overexpressing ASCO plants and nsra/b mutants behave similarly in response to auxin 24. Interestingly, auxin signaling is known to control the balance between growth and immunity 84. The auxin response was recently identified as a major component of the root transcriptional response to beneficial and pathogenic bacteria elicitors and is thought to mediate the observed reshaping of the root system in response to bacterial defense elicitors 85. Our results suggest that ASCO has a wider function than the simple titration of NSR activity. Remarkably, we now determined that ASCO is recognized by additional splicing factors: the spliceosome core components PRP8a and SmD1b. Accordingly, RNAi‐ASCO lines and a prp8‐7 leaky mutant exhibit similar AS defects of flg22‐regulated genes. However, smd1b mutants resulted in a deregulated ratio of isoforms of only ESP and SR34, but not NUDT7. The milder effect of ASCO‐related SmD1b over the subset of pathogen‐related genes may be due to a compensatory role of SmD1a in the smd1b background. Core splicing factors null mutants usually give very severe phenotypes or embryo lethality, and smd1b mutation was proposed to be partially compensated by SmD1a 71. Thus, the prp8‐7 leaky allele and the smd1b compensated mutant both exhibit partial effects on global constitutive splicing. Altogether, our results indicate that a complex network of lncRNAs and splicing factors involving ASCO, PRP8a, SmD1b, and NSRs dynamically shapes transcriptome diversity, integrating developmental, and environmental cues, thus conditioning the response to biotic stress (Fig 7).

Figure 7. The interaction of ASCO lncRNA and the spliceosome components PRP8a and SmD1b shapes the transcriptional response to flg22 modulating alternative splicing.

Proposed mechanism of ASCO lncRNA action. ASCO hijacks NSR proteins to modulate the population of alternatively spliced transcripts. Additionally, ASCO is recognized by PRP8a and SmD1b, two core components of the spliceosome, conditioning the SmD1b/PRP8a‐dependent transcriptome diversity in response to flagellin.

In Arabidopsis, the lncRNA ELF18‐INDUCED LONG NONCODING RNA 1 (ELENA1) is regulated by the perception of the translation elongation factor Tu (elf18) and it was identified as a factor enhancing resistance against Pseudomonas syringae. It was shown that ELENA1 directly interacts with Mediator subunit 19a (MED19a), modulating the enrichment of MED19a on the PATHOGENESIS‐RELATED GENE1 (PR1) promoter 86. Several other examples of lncRNAs mediating the environmental control of gene expression illustrate the relevance of the noncoding transcriptome as a key integration factor between developmental and external cues 68, 87, 88, 89. The sensitivity to pathogens has been shown to be affected in spliceosome‐related mutants. For instance, it was recently reported that the prp40c mutants display an enhanced tolerance to Pseudomonas syringae 90. On the other hand, several other splicing‐related genes have been identified as positive regulators of plant immunity against Pseudomonas 91, 92, 93. Therefore, the modulation of the expression and activity of splicing‐related components appear to be important for the proper response to pathogens.

LncRNAs as highly variable components of the conserved spliceosomal machinery

Here, we show that the highly structured lncRNA ASCO, which does not seem to contain introns, is capable to interact with PRP8a and modulate PRP8a binding to ASCO‐related AS targets. The spliceosome is a large complex composed of five different small nuclear ribonucleoprotein complexes subunits (snRNPs). Each subunit includes noncoding and nonpolyadenylated small nuclear uridine (U)‐rich RNAs (U snRNAs) and core spliceosomal proteins, along with more than 200 non‐snRNPs splicing factors 94. PRP8a is one of the largest and most highly conserved proteins in the nucleus of eukaryotic organisms. It occupies a central position in the catalytic core of the spliceosome and has been implicated in several crucial molecular rearrangements 67. In Arabidopsis, analysis of PRP8a leaky mutation suggests that PRP8a recognizes the lncRNA COOLAIR in vivo to modulate its AS 68 hinting at an interaction with lncRNAs. COOLAIR designates a set of transcripts expressed in antisense orientation of the locus encoding the floral repressor FLC 95. Two main classes of COOLAIR lncRNAs are produced by AS and polyadenylation of antisense transcripts generated from the FLC locus. One uses a proximal splice site and a polyadenylation site located in intron 6 of FLC, whereas the distal one results from the use of a distal splice and polyadenylation sites located in the FLC promoter 95. Notably, prp8‐7 partial loss of function leads to a reduced usage of COOLAIR proximal polyadenylation site and an increase of FLC transcription which is associated with late‐flowering phenotypes 68, 95. Interestingly, the FLC/COOLAIR module is strongly deregulated in the nsra/b double mutant and NSRa was linked to flowering time further supporting multiple interactions of lncRNAs and the splicing machinery 34. Although NSRa‐COOLAIR interaction seems not to occur, it was proposed that the control of NSRa over COOLAIR involves the direct interaction and processing of the polyadenylation regulatory gene FPA 34. In the model legume Medicago truncatula, the NSRs closest homolog, RNA‐BINDING PROTEIN 1 (RBP1), is localized in nuclear speckles where many components of the splicing machinery are hosted in plant cells. Remarkably, RBP1 interacts with a highly structured lncRNA, EARLY NODULIN 40 (ENOD40), which participates in root symbiotic nodule organogenesis 96, 97, 98. ENOD40 is highly conserved among legumes and was also found in other species such as rice (Oryza sativa) 99, but shows no homology to ASCO lncRNA 24. In contrast to the nuclear localization of Arabidopsis ASCO, ENOD40 was found both in the nucleus and in the cytoplasm, and it is able to relocalize RBP1 from nuclear speckles into cytoplasmic granules during nodulation 98. These observations suggested a role of the lncRNA ENOD40 in nucleocytoplasmic trafficking, potentially modulating RBP1‐dependent splicing and further supporting the multiple interactions of lncRNAs with splicing regulators. A major result shown here is that ASCO is recognized by PRP8a and SmD1b, two central regulators of splicing and not only by the NSR proteins which are plant‐specific “peripheral” regulators of splicing. Indeed, the nsra/b null double mutants did not display major phenotypes in contrast to null PRP or SmD components. The identification of how the ASCO lncRNA interacts with PRP8a will certainly contribute to understanding the intricate network of lncRNA‐mediated regulation of core splicing factors, thus opening wide perspectives for the use of lncRNAs in the modulation of the dynamic population of alternatively spliced mRNAs in higher organisms. Interestingly, a search for ASCO homologs across the Brassicaceae family reveals that 9 additional copies of ASCO exist in A. thaliana and related sequences are also present in other Brassicaceae species, including A. halleri and A. lyrata, and the more distant species Capsella rubella and Capsella grandiflora (Appendix Fig S6A). However, none of the four detectable A. thaliana ASCO‐like homologs suffered any significant alteration in RNAi‐ASCO and 35S:ASCO lines (Appendix Fig S6B), suggesting that none of them seem to compensate for the absence or overaccumulation of the ASCO lncRNA. The existence of ASCO‐like sequences in other species suggests that conserved lncRNA‐mediated mechanisms of AS regulation may occur through the interaction with highly conserved splicing factors. As PRP8a and SmD1b as well as the snRNAs are highly conserved spliceosomal components in contrast to the outstanding variability of lncRNA sequences along evolution, our results hint at a yet undiscovered evolutionary layer in the fine‐tuning of AS in specific cell types and different environmental conditions without affecting essential splicing activity. Structure and short sequences inside lncRNAs may contribute to the evolution of splicing regulatory networks in eukaryotes.

Materials and Methods

Plant material and growth conditions

All the lines used in this study were in the A. thaliana Columbia‐0 (Col‐0) background. We used the nsra/b double mutant and the ASCO overexpressing lines from [Ref. 24]. The insertion lines WiscDsLoxHs110_08A (asco‐1) and SAIL_812_C08 (asco‐2) were obtained from the T‐DNA mutant collection at the Salk Institute Genomics Analysis Laboratory (SIGnAL, http://signal.salk.edu/cgi-bin/tdnaexpress) via NASC (http://arabidopsis.info/). Seeds from prp8‐7 in the T line Col‐0 background 35 and smd1b 71 mutants were provided by H. Vaucheret. The pUBQ10:SmD1b‐GFP line was used for RNA immunoprecipitation assays. Plants were grown at 20°C with a 16‐h light/8‐h dark photoperiod (long days) on solid half‐strength Murashige and Skoog (1/2MS) medium.

Generation of transgenic lines

ProASCO::GUS transgenic lines

The promoter region of ASCO (2631‐bp upstream of the transcription start) was amplified from A. thaliana genomic DNA using gene‐specific primers listed in Dataset EV5. The amplicon was subcloned into the pENTR/D‐TOPO vector and recombined in a pKGWFS7 binary destination vector, upstream of the GFP, and GUS sequences. ProASCO::GUS constructs were transferred into A. thaliana by standard Agrobacterium‐mediated protocol 100. Three lines were selected based on 3:1 segregation for the transgene (single insertion) and brought to T3 generation where the transgene was in a homozygous state. All lines behave similarly as for GUS expression.

RNAi‐ASCO knocked‐down lines

The first 233‐bp of the ASCO transcript was amplified from A. thaliana genomic DNA using gene‐specific primers listed in Dataset EV5. Amplicons were subcloned into the pENTR/D‐TOPO vector and recombined in a pFRN binary destination vector 101 to target the ASCO RNA by long dsRNA hairpin formation.

Root growth analysis

For analysis of auxin impact on root architecture, plants were grown as described in [Ref. 24]. Briefly, seeds were sterilized and directly sown on plates containing 1/2MS medium supplemented or not with 100 nM NAA. Plantlet root architecture was analyzed using the RootNav software after 7 days of growth 102. For analysis of flagellin impact on root architecture, plants were previously grown 5 days on solid 1/2MS medium + 1% sucrose and then transferred for additional 9 days in liquid 1/2MS media + 1% sucrose supplemented or not with 0.1 μM or 1 μM of synthetic flg22 peptide (GeneCust). For each plantlet, lateral roots were counted, and the primary root length was measured using the RootNav software. Experiments were done at least two times, with a minimum of 16 plants per genotype and condition. Statistical tests were performed using the Mann–Whitney's U‐test (P < 0.05) using wild‐type values as reference.

Root meristem measurements

Plants were grown as for root growth analysis in response to flg22. The treated plants were stained with SCRI Renaissance 2200 (Renaissance Chemicals) as described in [Ref. 103]. Images were obtained with LSM880 (Zeiss) confocal microscope. The SR2200 fluorescence was excited with a 405 nm laser line and emission recorded between 410 and 686 nm (405/410–686). Cell counting and primary root meristem measurements were performed using ImageJ package (https://imagej.nih.gov/ij/). Experiments were done two times, with a minimum of 18 plants per genotype and condition. Statistical tests were performed using the Student's t‐test (P < 0.05).

Histochemical GUS staining

Histochemical GUS staining was performed according to [Ref. 104] Briefly, 10‐day‐old plantlets grown in standard conditions were fixed in cold 90% acetone and incubated overnight at 37°C in the GUS staining buffer. Roots were subsequently fixed in 4% paraformaldehyde for 1 h and washed several times in 70% ethanol before a final wash in 10% glycerol prior observation. Images were acquired using an AxioImagerZ2 microscope (Zeiss).

RNA extraction and RT–PCR analyses

Total RNA was extracted using Quick‐RNA Kit (ZYMO RESEARCH), and DNase treatment was performed according to the manufacturer's protocol. One μg of DNase‐free RNA was reverse transcribed using Maxima H Minus Reverse Transcriptase (Thermo Scientific). cDNA was then amplified in RT–qPCRs using LightCycler 480 SYBR Green I Master (Roche) and transcript‐specific primers on a Roche LightCycler 480 thermocycler following standard protocol (45 cycles, 60°C annealing). Experiments were done in biological triplicates with at least three technical replicates. Expression was normalized to 2 constitutive genes (AT1G13320 and AT4G26410) 105. For analysis of flg22 impact on gene expression, plants were previously grown 9 days on solid 1/2MS medium + 1% sucrose and then transferred for additional 24 h in liquid 1/2MS medium + 1% sucrose before adding or not 1 μM of flg22. Five plantlets were pooled for each replicate. The fold induction of expression after flg22 treatment was normalized to the WT response considered as 100%. For analysis of gene expression after a flg22 kinetic, roots from 8 plants were pooled for each replicate. For AS analysis, isoform‐specific primers were designed for each differential event and the signal was normalized with respect to an internal gene probe (called INPUT) corresponding to a common exon for each group of transcripts. This allows differentiating the change of each isoforms independently of the expression level of the studied gene (splicing index) in each sample 33. The splicing index was calculated following this equation: splicing index = 2^{[∆Ct(specific isoform) − (∆Ct(INPUT)]}. Error bars on qRT–PCR experiments represent standard deviations, and significant differences were determined using Student's t‐test (P ≤ 0.05, n ≥ 3 biological replicates). All the used primers are listed in Dataset EV5.

For RT–PCR analysis, the amplification was performed using Phusion High‐Fidelity DNA Polymerase and transcript‐specific primers as manufacturer's protocol. PCR products were separated on an 8% polyacrylamide gel stained with SYBR Gold (Thermo Fischer Scientific) and revealed using a ChemiDoc MP Imaging System (Bio‐Rad). Band intensity was quantified using ImageJ package (https://imagej.nih.gov/ij/). Isoform ratio was calculated as the ratio of intensity of the two bands corresponding either to the alternatively spliced or to spliced transcript isoforms, respectively.

Transcriptome studies

Total RNA was extracted using RNeasy Plant Mini Kit (Qiagen) from whole 14‐day‐old Col‐0, 35S:ASCO1, and RNAi‐ASCO1 plants grown on 1/2MS medium. Three independent biological replicates were produced per genotype. For each biological repetition and each point, RNA samples were obtained by pooling RNA from more than 200 plants. After RNA extraction, polyA RNAs were purified using Dynabeads mRNA DIRECT Micro Kit (Ambion). Libraries were constructed using the Truseq Stranded mRNA Sample Prep Kit (Illumina^®). Sequencing was carried out at the POPS Transcriptomic Platform, Institute of Plant Sciences Paris‐Saclay in Orsay, France. The Illumina HiSeq2000 technology was used to perform paired‐end 100‐bp sequencing. A minimum of 30 million of paired‐end reads by sample were generated. RNA‐seq preprocessing included trimming library adapters and quality controls with Trimmomatic 106. Paired‐end reads with Phred Quality Score Qscore > 20 and read length > 30 bases were kept, and ribosomal RNA sequences were removed with SortMeRNA 107. Processed reads were aligned using Tophat2 with the following arguments: –max‐multihits 1 ‐i 20 –min‐segment‐intron 20 –min‐coverage‐intron 20 –library‐type fr‐firststrand –microexon‐search ‐I 1,000 –max‐segment‐intron 1,000 –max‐coverage‐intron 1,000 –b2‐very‐sensitive. Reads overlapping exons per genes were counted using the FeatureCounts function of the Rsubreads package using the GTF annotation files from the Araport11 repository (https://www.araport.org/downloads/Araport11_Release_201606/annotation/Araport11_GFF3_genes_transposons.201606.gff.gz). Significance of differential gene expression was estimated using DEseq2 108, and the FDR correction of the P‐value was used during pairwise comparison between genotypes. A gene was declared differentially expressed if its adjusted P‐value (FDR) was ≤ 0.01 and its absolute fold change was ≥ 1.5.

Gene Ontology analysis

Gene Ontology enrichment analysis was done using AgriGO (http://bioinfo.cau.edu.cn/agriGO) and default parameters.

AS analysis

The RNA profile analysis was performed using the RNAprof software (v1.2.6) according to [Ref. 33]. Briefly, RNAprof software allows detection of differential RNA processing events from the comparison of nucleotide level RNA‐seq coverage normalized for change in gene expression between conditions. Here, the RNAprof analysis compared RNA‐seq data from biological triplicates of WT, RNAi‐ASCO1, and 35S:ASCO1 lines. Differentially processed regions genes were filtered as follows: fold change > 2 and P < 0.001. Overlap between gene features and differentially processed regions was done using in‐house R scripts (https://github.com/JBazinIPS2/Bioinfo/blob/master/RNAprof_events_selection.Rrst). Only regions fully included in a gene features were kept for further analysis. The RNAprof software archive, including documentation and test sets, is available at the following address: http://rna.igmors.u-psud.fr/Software/rnaprof.php. Transcript level quantification was performed using pseudo‐alignment counts with kallisto 109 on AtRTD2 transcripts sequences (https://ics.hutton.ac.uk/atRTD/RTD2/AtRTDv2_QUASI_19April2016.fa) with a K‐mer size of 31‐nt. Differential AS events in the AtRTD2 database were detected using SUPPA2 with default parameters 53. Only events with an adjusted P < 0.01 were kept for further analysis. Isoforms switch identification was performed with the IsoformSwitchAnalyzeR package 54 according to 34.

Whole‐mount immunolocalization

Specific rabbit polyclonal antibodies were developed against PRP8a using the peptide TNKEKRERKVYDDED (Li International). Five‐day‐old seedlings were fixed with 4% paraformaldehyde in microtubule stabilization buffer (MTSB) 110 for 1 h and rinsed once in glycine 0.1 M and twice with MTSB. Cell walls were partially digested for 45 min at 37°C in cellulase R10 1% w/v (Onozuka), pectolyase 1% w/v, and cytohelicase 0.5% w/v (Sigma) solution. After two PBS washes, root tissues were squashed on polylysine‐treated glass slides (VWR International) and dipped in liquid nitrogen. The coverslip was then removed, and the slides were left to dry. After 2 rinses with PBS and 2 with PBS‐0.1% Triton, they were treated with BSA 3% in PBS‐Triton buffer for 1 h and incubated with the anti‐PRP8a antibody (dilution 1:400) for 16 h at 4°C in a humid chamber. After incubation, slides were rinsed 8–10 times with PBS‐Triton and incubated for 1 h at 37°C with the secondary antibody (anti‐Rabbit IgG coupled to Alexa Fluor^® 594, dilution 1:500; Thermo Scientific) and rinsed 10 times with PBS‐Triton and once with PBS. Slides were mounted in Vectashield© containing DAPI (VECTOR Laboratories). Images were obtained with LSM880 (Zeiss) confocal microscope equipped with Plan‐Apochromat 63×/NA1.40 Oil M27 lens. Dapi and Alexa 594 fluorescences were, respectively, excited with 405 nm and 561 nm diodes and recorded between 410–500 nm and 570–695 nm.

LncRNA‐bound nuclear protein isolation by RNA purification

A method adapted from the ChIRP protocol 65, 66, 111 was developed to allow identification of nuclear proteins bound to specific lncRNAs. Briefly, plants were in vivo crosslinked, and nuclei of cells purified and extracted through sonication. The resulting supernatant was hybridized against biotinylated complementary oligonucleotides that tile the lncRNA of interest, and putative lncRNA‐containing protein complexes were isolated using magnetic streptavidin beads. Co‐purified ribonucleoprotein complexes were eluted and used to purify RNA or proteins, which were later subject to downstream assays for identification and quantification.

Probe design

Antisense 20‐nt oligonucleotide probes were designed against the ASCO full‐length sequence (AT1G67105) using an online designer at http://singlemoleculefish.com/. All probes were compared with the A. thaliana genome using the BLAST tool at the NCBI, and probes returning noticeable homology to non‐ASCO targets were discarded. Eighteen probes were finally generated and split into two sets based on their relative positions along the ASCO sequence, such as EVEN‐numbered and ODD‐numbered probes were separately pooled. A symmetrical set of probes against LacZ RNA 66 was also used as the mock control. All probes were ordered biotinylated at the 3′ end (Invitrogen).

Crosslinking and ribonucleoprotein complexes purification

For protein extraction, approximately 250 g of 7‐day‐old Col‐0 plants grown on solid half‐strength MS medium was irradiated three times with UV using a CROSSLINKER^® CL‐508 (Uvitec) at 0.400 J/cm². For RNA extraction, 10 g of 7‐day‐old Col‐0 plants grown on solid half‐strength MS medium was crosslinked under vacuum for 15 min with 37 ml of 1% (v/v) formaldehyde. The reaction was stopped by adding 2.5 ml of 2 M glycine, and seedlings were rinsed with Milli‐Q purified water. For both crosslinking methods, 6 g of the fixed material was ground in liquid nitrogen and added to 50‐ml tubes with 25 ml of extraction buffer 1–15 ml of plant material ground to fine dust (the nuclei were prepared starting with 30 fifty‐milliliter tube; buffer 1: 10 mM Tris–HCl pH 8, 0.4 M sucrose, 10 mM MgCl₂, 5 mM β‐mercaptoethanol, 1 ml/30 g of sample powder Protease Inhibitor Sigma Plant P9599). The solution was then filtered through Miracloth membrane (Sefar) into a new tube, and 5 ml of extraction buffer 2 (10 mM Tris–HCl pH 8, 0.25 M sucrose, 10 mM MgCl₂, 5 mM β‐mercaptoethanol, 1% Triton X‐100, 50 μl protease inhibitor) was added. The solution was then centrifuged, the supernatant discarded and the pellet was resuspended in 500 μl of extraction buffer 3 (10 mM Tris–HCl pH 8, 1.7 M sucrose, 2 mM MgCl₂, 5 mM β‐mercaptoethanol, 0.15% Triton X‐100, 50 μl protease inhibitor) and layered on top of fresh extraction buffer 3 in a new tube. After centrifugation at 13,000 rpm for 2 min at 4°C to pellet nuclei, the supernatant was discarded and the pellet resuspended in 300 μl of nuclei lysis buffer (50 mM Tris–HCl pH 7, 1% SDS, 10 mM EDT, 1 mM DTT, 50 μl protease inhibitor, 10 μl RNAse inhibitor per tube) to degrade nuclear membranes. Samples were sonicated three times in refrigerated BIORUPTOR Plus (Diagenode), 10 cycles 30 s ON–30 sec OFF in a Diagenode TPX microtube M‐50001. After centrifugation, the supernatant was transferred to a new tube and diluted two times volume in hybridization buffer (50 mM Tris–HCl pH 7, 750 mM NaCl, 1% SDS, 15% formamide, 1 mM DTT, 50 μl protease inhibitor, 10 μl RNAse inhibitor). One hundred pmol of probes were added to samples and incubated 4 h at 50°C in a thermocycler. Samples were transferred to tubes containing Dynabeads‐Streptavidin C1 (Thermo Fisher Scientific) and incubated 1 h at 50°C. Then, samples were placed on a magnetic field and washed three times with 1 ml of wash buffer (2× SSC, 0.5% SDS, 1 mM DTT, 100 μl protease inhibitor).

Protein purification

Samples for protein extraction were DNase‐treated according to the manufacturer (Thermo Scientific). After addition of 1.8 ml of TCA‐acetone (5 ml 6.1 N TCA + 45 ml acetone + 35 μl β‐mercaptoethanol), samples were incubated overnight at −80°C. After centrifugation at 44,000 g for 20 min and 4°C, the supernatant was discarded and 1.8 ml of acetone wash buffer (120 ml acetone, 84 μl β‐mercaptoethanol) was added to the samples. Then, samples were incubated 1 h at −20°C and centrifuged again at 40,000 g for 20 min and 4°C. The supernatant was discarded, and the dry pellet was used for mass spectrometry analysis.

RNA purification

Samples for RNA extraction were boiled for 15 min after washing with 1 ml of wash buffer (2× SSC, 0.5% SDS, 1 mM DTT, 100 μl protease inhibitor). Beads were removed in a magnetic field, and TRIzol/chloroform RNA extraction was performed according to the manufacturer (Sigma). RNAs were precipitated using 2 volumes EtOH 100, 10% 3 M sodium acetate, and 1 μl glycogen and washed with EtOH 70%. RNAs were kept at −20°C before use for reverse transcription and RT–qPCR analysis.

Liquid Chromatography–Mass Spectrometry (LC‐MS) analysis

Proteins purified from ribonucleoprotein complexes were analyzed using the King Abdullah University of Science and Technology proteomic facilities. Dry pellets of samples purified with either ODD, EVEN, or LacZ probe sets were solubilized in trypsin buffer (Promega) for digestion into small peptides. The solubilized peptides were then injected into a Q Exactive™ HF hybrid quadrupole‐Orbitrap mass spectrometer (Thermo Scientific) with a Liquid Chromatography (LC) Acclaim PepMap C18 column (25 cm length × 75 μm I.D. × 3 μm particle size, 100 Å porosity, Dionex). Data were analyzed for each sample using Mascot software (Matrix Science), with a minimal sensitivity of 2 detected peptides per identified protein.

RNA immunoprecipitation

Eleven‐day‐old plants grown in Petri dishes were irradiated three times with UV using a CL‐508 cross‐linker (Uvitec) at 0.400 J/cm². Briefly, fixed material was ground in liquid nitrogen and homogenized and nuclei isolated and lysed according to [Ref. 112]. RNA immunoprecipitation was basically performed as described by [Ref. 113]. The nuclei extract (input) was used for immunoprecipitation with 50 μl of Dynabeads‐Protein A (Thermo Fisher Scientific) and 1 μg of anti‐GFP antibodies (Abcam ab290) or anti‐PRP8a (Li International), respectively. Beads were washed twice for 5 min at 4°C with wash buffer 1 (150 mM NaCl, 1% Triton X‐100, 0.5% Nonidet P‐40, 1 mM EDTA, and 20 mM Tris–HCl, pH 7.5) and twice with wash buffer 2 (20 mM Tris–HCl, pH 8) and finally resuspended in 100 μl Proteinase K buffer (100 mM Tris–HCl, pH 7.4, 50 mM NaCl, and 10 mM EDTA). Twenty microliters were saved for further immunoblot analysis. After Proteinase K (Ambion AM2546) treatment, beads were removed with a magnet, and the supernatants were transferred to a 2‐ml tube. RNA was extracted using TRI Reagent (Sigma‐Aldrich) as indicated by the manufacturer. Eighty microliters of nuclei extracts was used for input RNA extraction. The immunoprecipitation and input samples were treated with DNase, and random hexamers were used for subsequent RT. Quantitative real‐time PCR reactions were performed using specific primers (Dataset EV5). Results were expressed as a percentage of cDNA detected after immunoprecipitation, taking the input sample as 100%. For Western blot analysis, immunoprecipitated proteins were eluted by incubating 20 μl of beads in 20 μl of 2× SDS‐loading buffer without β‐mercaptoethanol (100 mM Tris–HCl pH 6.8, 20% glycerol, 12.5 mM EDTA, 0.02% bromophenol blue) at 50°C for 15 min. Input, unbound and eluted fractions were boiled in 2× SDS‐loading buffer with 1% β‐mercaptoethanol for 10 min, loaded onto a 4–20% Mini‐PROTEAN^® TGX™ Precast Protein Gels (Bio‐Rad), and transferred onto a PVDF membrane using the Mini Trans‐Blot^® Cell system (Bio‐Rad) for 3 h at 70 V. Membranes were blocked in 5% dry nonfat milk in PBST and probed using PRP8a antibody (1:500) and an HRP coupled anti‐Rabbit secondary antibody (Bio‐Rad, 1:10,000). All antibodies were diluted in 1% dry nonfat milk in PBST. Blots were revealed with the Clarity ECL substrate according the manufacturer instruction (Bio‐Rad) and imaged using the ChemiDoc system (Bio‐Rad).

Data availability

Data were deposited in CATdb database 114 (http://tools.ips2.u-psud.fr/CATdb/) with ProjectID NGS2016‐07‐ASCOncRNA. This project was submitted from CATdb into the international repository GEO (http://www.ncbi.nlm.nih.gov/geo) with accession number GSE135376.

Author contributions

RR, JB, NR‐B, MM, LL, AC, CC, and FA performed the experiments; SH was in charge of the RNA‐seq libraries preparation and sequencing; TB, RR, JB, CC, MB, MC, and FA analyzed the data; JB performed bioinformatic analyses; JB, MC, and FA designed the experiments; MC and FA conceived the study and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Dataset EV1

Dataset EV2

Dataset EV3

Dataset EV4

Dataset EV5

Review Process File

EMBO Reports (2020) 21: e48977