A GFP splicing reporter in a coilin mutant background reveals links between alternative splicing, siRNAs, and coilin function in Arabidopsis thaliana

Abstract Coilin is a scaffold protein essential for the structure of Cajal bodies, which are nucleolar-associated, nonmembranous organelles that coordinate the assembly of nuclear ribonucleoproteins (RNPs) including spliceosomal snRNPs. To study coilin function in plants, we conducted a genetic suppressor screen using a coilin (coi1) mutant in Arabidopsis thaliana and performed an immunoprecipitation-mass spectrometry analysis on coilin protein. The coi1 mutations modify alternative splicing of a GFP reporter gene, resulting in a hyper-GFP phenotype in young coi1 seedlings relative to the intermediate wild-type level. As shown here, this hyper-GFP phenotype is extinguished in older coi1 seedlings by posttranscriptional gene silencing triggered by siRNAs derived from aberrant splice variants of GFP pre-mRNA. In the coi1 suppressor screen, we identified suppressor mutations in WRAP53, a putative coilin–interacting protein; SMU2, a predicted splicing factor; and ZCH1, an incompletely characterized zinc finger protein. These suppressor mutations return the hyper-GFP fluorescence of young coi1 seedlings to the intermediate wild-type level. Additionally, coi1 zch1 mutants display more extensive GFP silencing and elevated levels of GFP siRNAs, suggesting the involvement of wild-type ZCH1 in siRNA biogenesis or stability. The immunoprecipitation-mass spectrometry analysis reinforced the roles of coilin in pre-mRNA splicing, nucleolar chromatin structure, and rRNA processing. The participation of coilin in these processes, at least some of which incorporate small RNAs, supports the hypothesis that coilin provides a chaperone for small RNA trafficking. Our study demonstrates the usefulness of the GFP splicing reporter for investigating alternative splicing, ribosome biogenesis, and siRNA-mediated silencing in the context of coilin function.


Introduction
Coilin is an intriguing protein that is present in most multicellular eukaryotes.Even after decades of research, the full range of coilin's functions and modes of action are not fully understood (Machyna et al. 2015).Coilin is best known as an essential structural constituent and prominent marker of Cajal bodies (CBs).CBs are nonmembranous nuclear organelles that are typically situated adjacent to the nucleolus, a second membrane-free nuclear organelle that also contains detectable amounts of coilin (Hebert 2013;Trinkle-Mulcahy and Sleeman 2017).CBs are major sites for the maturation, processing, and quality control of small nuclear RNAs (snRNAs), including those that are present in small nuclear ribonucleoprotein particles (snRNPs) required for pre-mRNA splicing, and small nucleolar RNAs (snoRNAs), which guide processing and maturation of rRNAs and tRNAs in the nucleolus (Machyna et al. 2015;Taliansky et al. 2023).
Even though coilin appears to be particularly enriched in CBs and is required for their formation and maintenance, a large fraction of coilin protein in animal cells is distributed throughout the nucleoplasm (Lam et al. 2002) where its roles are less clear.Through dynamic associations with chromatin, nucleoplasmic coilin in animals is thought to contribute to a number of processes including not only pre-mRNA splicing but also chromatin organization, DNA repair, and telomere maintenance (Machyna et al. 2015;Taliansky et al. 2023).Whether coilin also performs these multiple functions in plants is not yet clear.
The ability of coilin to facilitate CB assembly depends on selfassociations and interactions with RNA and other CB proteins (Makarov et al. 2013;Machyna et al. 2015).Conserved regions of coilin protein in plants and animals include an N-terminal selfassociation domain; a C-terminal atypical Tudor domain that mediates interactions with several snRNPs (Hebert et al. 2001;Xu et al. 2005); and a central disordered domain (Makarov et al. 2013).Although coilin lacks conventional RNA binding motifs, analysis of the secondary structure of Arabidopsis coilin revealed several degenerate RNA recognition motifs in the N-terminal region and central disordered domain (Makarov et al. 2013;Machyna et al. 2015).The reported ability of coilin to bind RNAs, particularly small noncoding RNAs such as snRNAs and snoRNAs (Machyna et al. 2014), led to the proposal that coilin may serve as a chaperone for the trafficking of nuclear noncoding small RNAs (Machyna et al. 2015).
The possible involvement of CBs and coilin in small interfering RNA (siRNA)-induced gene silencing (RNA silencing) has been explored in plants (Li et al. 2006;Pontes and Pikaard 2008;Pontes et al. 2013), where 21-24-nt siRNAs generated by dicer cleavage of double-stranded RNA (dsRNA) precursors mediate both posttranscriptional and transcriptional gene silencing (PTGS and TGS, respectively).DsRNAs can be produced by RNA-dependent RNA polymerases (RDR) acting on single-stranded RNAs that are perceived by the cell as aberrant in some way.Aberrant RNAs can include "foreign" RNAs derived from viruses and transgenes as well as endogenous RNAs that are untranslatable due to premature termination codons resulting from single nucleotide mutations, mis-splicing, or lack of a poly-A tail.PTGS involves sequencespecific degradation of target RNAs triggered by complementary 21-22-nt siRNAs, which are produced in Arabidopsis by DICER-LIKE4 (DCL4) and DCL2, respectively.TGS is associated with DNA methylation of promoter segments guided by DCL3-dependent 24-nt siRNAs in a process referred to as RNA-directed DNA methylation (RdDM) (Uslu and Wassenegger 2020;El-Sappah et al. 2021;Kryovrysanaki et al 2022).
In Arabidopsis, factors required for PTGS and TGS, such as DCL enzymes and ARGONAUTE (AGO) silencing effector proteins, have been detected in CBs by immunofluorescence microscopy (Li et al. 2006;Pontes and Pikaard 2008;Pontes et al. 2013).However, as shown by experiments using an Arabidopsis ncb (no cajal body) mutant (Collier et al. 2006), intact CBs do not seem to be required for siRNA-mediated gene silencing to occur.Neither RdDM nor the production of siRNAs at several tested loci was substantially affected by the loss of CB integrity in ncb mutants, although the efficiency of RdDM at these sites may have been somewhat reduced (Li et al. 2006).
Our interest in coilin was stimulated by the retrieval of multiple coilin (coi1) mutants in a forward genetic screen based on an alternatively spliced GFP reporter gene in Arabidopsis (Kanno et al. 2016(Kanno et al. , 2020)).In this screen, coi1 mutants were initially identified in young seedlings (shortly after germination) by their hyper-GFP phenotype relative to wild-type seedlings, which display an intermediate level of GFP fluorescence (Kanno et al. 2016).Although the basis of the hyper-GFP phenotype in young coi1 seedlings is not yet completely understood, it can be attributed at least in part to modified splicing of GFP pre-mRNA, leading to elevated amounts of the only GFP splice variant that can be translated into protein (Kanno et al. 2016(Kanno et al. , 2020)).
After publishing the identification of the hyper-GFP coi1 mutants (Kanno et al. 2016), we later observed that in many older coi1 seedlings (typically at the second to fourth true leaf stage), complete silencing of the GFP reporter gene occurs.The silencing pattern, featuring dark red leaves emerging from the shoot apex of a hyper-GFP plantlet, superficially resembled virus recovery, in which asymptomatic leaves appear during the growth of an infected plant.Recovery from virus infection is known to involve PTGS that is provoked by virus-derived siRNAs (Ghoshal and Sanfacon 2015;Uslu and Wassenegger 2020).
Here we report the results of experiments designed to investigate the delayed silencing phenomenon in older coi1 mutants.We also describe findings from genetic and biochemical analyses that were carried out to identify proteins that may be functionally linked to coilin and potentially illuminate new coilin-dependent pathways.

Plant material and growth conditions
Wild-type and mutant plants used in this study were in the ecotype Col-0 background and cultivated under long-day conditions (22-23°C, 16 hours light, 8 hours dark).Gene names are written in italics and proteins in upright letters.Mutant names are indicated in lowercase letters; wild-type names are in uppercase letters.To test for the occurrence of PTGS, rdr6-14 and dcl4-12 alleles were introduced into the wild-type line (WT T) and other mutants by intercrossing.For assessments of GFP fluorescence intensity in seedlings, seeds of the desired line were surface sterilized, germinated on solid Murashige and Skoog (MS) medium in plastic Petri dishes, and cultivated in a growth incubator under a 16-hour light/8-hour dark cycle at 24°C.Seedlings were observed daily for fluorescence levels under a Leica fluorescence stereomicroscope over a period of 4-6 weeks.
Seeds of the WT T line and coilin suppressor mutants as well as other mutants used in this study are available from the Arabidopsis Biological Research Center (ABRC) under the ABRC seed stock numbers listed in Supplementary Table 1.Accession numbers for data from RNA sequencing, small RNA sequencing, and whole genome DNA resequencing presented in this manuscript are also listed in Supplementary Table 1.

Nomenclature of GFP phenotypes
Five major GFP phenotypes-as assessed by the strength of GFP fluorescence, which correlated with the level of GFP protein determined by Western blotting-were observed in various genotypes.
GFP-intermediate (relative to GFP-weak and hyper-GFP phenotypes) is observed in WT T seedlings in the shoot apex, hypocotyl, and root, conforming to the expression pattern of the viral enhancer driving GFP expression (Fig. 1; Daxinger et al. 2009).The strength of GFP fluorescence in these regions is enhanced in a rdr6 mutant background.
Hyper-GFP refers to GFP fluorescence that is considerably stronger than that observed in WT T plants, and that is usually visible upon germination and maintained thereafter throughout the lifetime of the plant (seen in coi1 and cwc16a mutants in this study).
GFP-weak refers to GFP fluorescence that is visibly lower than that observed in the WT T line.
GFP-weak/intermediate refers to microscopically visible GFP fluorescence that is greatly reduced from that of the coi1 single mutant to approximately the intermediate WT T level (seen in coi suppressor mutants).Note that the GFP weak and GFP weak/ intermediate phenotypes could sometimes be masked by the accumulation of chlorophyll during plant growth.In these cases, it could be difficult to assess GFP silencing occurring at later stages of development.
GFP-negative (neg) refers to no visible GFP fluorescence.Seedlings appear dark red owing to the autofluorescence of chlorophyll at the excitation wavelength for GFP (seen in gfp mutants and aerial parts of coi1 mutants undergoing delayed sense PTGS (S-PTGS).
The GFP phenotypes were visualized using stereo-fluorescence microscopy in several hundred seedlings from multiple generations of each genotype.The GFP phenotypes were consistent with GFP levels determined in selected genotypes by Western blotting.For nearly all genotypes, 100% of the seedlings displayed the GFP fluorescence and silencing phenotypes reported herein.The only exception was delayed S-PTGS in the hyper-GFP coi1 mutant, which typically occurred in 40-90% of seedlings at the 2nd-4th true leaf stage.

Forward genetic screen and identification of coilin suppressor mutants
The protocol for forward genetic screens based on an alternatively spliced GFP reporter gene in Arabidopsis [T line; referred to herein as "wild-type" (WT) T ] (Fig. 1)] has been described previously (Kanno et al. 2016(Kanno et al. , 2018;;Kanno, Lin, Fu, Matzke, et al. 2017;Kanno, Lin, Fu, Chang, et al. 2017).In the present coilin suppressor screen, approximately 40,000 Arabidopsis seeds of a homozygous coi1-8 mutant (also homozygous for the T locus) were treated with ethyl methane sulfonate (EMS) and sown on soil (M1 generation).From approximately 20,000 M1 plants that grew to maturity and produced self-fertilized seeds, 52 batches of M2 seeds (the first generation when a recessive mutation can be homozygous) were harvested.M2 seeds were surface sterilized, germinated on solid MS medium in plastic Petri dishes, and cultivated in a growth incubator as described above.
Putative suppressor mutants were identified by examining M2 seedlings under a Leica fluorescence stereo-microscope approximately 7 d after germination for reduced GFP fluorescence relative to the hyper-GFP phenotype of the coi1-8 single mutant.Around 100 putative suppressor mutants were recovered and grouped into 2 main categories: GFP-weak/intermediate and GFP-neg.Sequencing of the GFP reporter gene in the GFP-neg mutants revealed that 5 harbored a previously unidentified loss-of-function mutation in the GFP coding sequence (CDS) (Supplementary Table 1; described in Supplementary Fig. 1; Fu et al. 2015).
GFP-weak/intermediate mutants that contained a wild-type GFP nucleotide sequence were further tested to confirm: (1) stable meiotic inheritance of the GFP-weak/intermediate phenotype; (2) the presence of the expected coi1-8 mutation in the endogenous coi1 gene; and (3) the absence of any second-site intragenic suppressor mutations induced by the EMS treatment into the coi1-8 gene sequence.GFP-weak mutants that passed these additional The 3 black arrowheads at the 5′ end of the enhancer represent a short tandem repeat, which becomes methylated in the presence of homologous siRNAs (Kanno et al. 2008).Opposing blue arrowheads denote the positions of primers used to detect DNA methylation (Supplementary Table 3).Enhancer region, 1277 bp; minimal promoter, 91 bp; expected GFP CDS, 720 bp, Total length, 2088bp.Transcripts detected: Although transcription was predicted to start downstream of the TATA minimal promoter, this promoter, and the downstream ATG start codon remain unused (gray letters).Instead, transcription initiates at a cryptic promoter upstream in the tandem repeats (black arrow).Three polyadenylated GFP transcripts can be detected: an unspliced form (GFP pre-mRNA) and 2 spliced variants, which contain canonical (GU-AG) and noncanonical (AU-AC; GFP mRNA) splice sites, respectively (Kanno et al. 2008).Owing to the presence of stop codons (asterisks) in the unspliced and GU-AG transcripts, only the AU-AC transcript can be translated into GFP protein (green bar).Variations in the proportions of the 3 splice variants by alternative splicing modify the level of GPF fluorescence (Kanno et al. 2020).Translation initiates from a new upstream in-frame ATG start codon (black letters), which results in a 27 amino acid N-terminal extension to standard GFP protein (short green bar) (Fu et al. 2015).
Splicing, siRNAs, and coilin function | 3 tests were then subjected to next-generation mapping (NGM) (James et al. 2013;Kanno et al. 2020) and/or candidate gene sequencing to identify the respective causative suppressor mutations.Mutations were confirmed by the identification of multiple alleles and/or complementation analyses.All mutations reported here are recessive.

Western blotting using a GFP antibody
Western blotting to determine the relative levels of GFP protein level in the coi1 wrap53, coi1 smu2, and coi1 zch1 double mutants compared to the coi1 single mutant and WT T line was performed as described in a previous publication (Fu et al. 2015).For loading control, a duplicate gel containing the same samples was run and stained with Coomassie Brilliant blue.

RNA and small RNA sequencing (RNAseq and sRNAseq)
Total RNA was prepared using a Plant Total RNA Miniprep kit (GMbiolab, Taiwan) from 2-wk-old seedlings of the WT T line and the coi1 wrap53, coi1 smu2, and coi1 zch1 double mutants cultivated on solid MS medium as described above.RNA concentrations were assessed by NanoDrop (ND-1000 spectrophotometer).Library preparation, RNAseq and sRNAseq were performed (biological triplicates for each sample) as described previously by an in-house Genomic Technology Core facility (Kanno et al. 2016).Whole genome resequencing of the coi1 wrap53, coi1 smu2, and coi1 zch1 double mutants was carried out to identify any remaining EMS-induced second-site mutations that alter splice sites.These mutations were then excluded from the analysis of alternative splicing (AS).

Analysis of RNAseq and sRNAseq data
Pair-ended RNAseq reads were mapped to the Arabidopsis TAIR10 genome in two steps.The first step was to map reads to a combined transcriptome database of Araport11 (Cheng et al. 2017) and Atrtd2 (Zhang et al. 2017) using Bowtie2 (Langmead and Salzberg 2012), where only ungapped alignments of read pairs been were mapped to the same transcripts were accepted.Coordinates of these transcriptome-mapping alignments were transferred to the genome.The second step was to map the rest reads to the TAIR10 genome using BLAT (Kent 2002) directly.
To gain alignments of small RNA reads to the GFP-containing vector pAJM-EPRV (GenBank accession: HE582394.1)as precise as possible, the following prioritized approach was adopted: (1) reads were mapped to the canonical GFP transcript, only perfect matches were accepted, (2) for splicing junctions supported by RNAseq data, reads were mapped to junction-spanning fragments and only perfect matches were accepted, (3) repeat above 2 steps by allowing 1 mismatch, and (4) rest reads were mapped to pAJM-EPRV using BLAT with a set of fine scanning parameters (-minMatch = 1 -minScore = 14 -tileSize = 9) and allowing one mismatch.For comparing abundances of GFP siRNAs of different lengths between samples, corresponding read counts were normalized into read count per million reads, as well as read coverage per million reads along the GFP region in pAJM-EPRV.
AS detection for intron-retention (IR), exon-skipping (ES), and alternative donor/acceptor (altDA) was done using a similar method as described in our previous publications (Kanno, Lin, Fu, Chang, et al. 2017;Kanno, Lin, Fu, Matzke, et al. 2017;Kanno et al. 2018).For each kind of AS event, read-based counts that are supporting the AS event from the control and treatment samples were considered as signals, whereas read-based counts that are related but not supporting the event were considered as backgrounds.In practice, intron read depths, counts of splicing read skipping exon(s), and counts of reads exactly supporting one donor-acceptor pair were taken as the signals for IR, ES, and altDA events, respectively.For backgrounds, read depths of neighboring exons, counts of splicing reads involving one skipped exon, and counts of reads not supporting the donor-acceptor pair were taken for IR, ES, and altDA events, respectively.Signals from the control and treatment samples were then compared to the backgrounds using χ 2 test for goodness-of-fit for discovering the differential preference of AS events in the 2 samples.of-fit for discovering differential preference of AS events in the 2 samples.In this study, an AS event was considered significant if its P-value is below 0.05.Note that these AS comparisons on RNAseq data were made by accumulating numbers from three biological replicates, i.e. numbers of merged samples.This is because read numbers on AS events are usually smaller than those for differentially expressed gene discovery.Doing so would give us a higher statistical power when applying the χ 2 -test for goodness-of-fit than separate biological replicates.Also, note that P-values here were not corrected because our aim is to discover general AS differences among samples under the same comparison standard.

Protein immunoprecipitation of epitope-tagged coilin proteins
Total proteins were extracted from 2-wk-old seedlings of coilin-mRed, FLAG-coilin, and coilin-FLAG transgenic plants.Protein extraction and immunoprecipitation were carried out according to a published protocol (Huang et al. 2016).Briefly, 5 grams of seedlings were ground to a powder in liquid nitrogen.After grinding, 14 ml of SPII + buffer [100 mM Na-phosphate, pH 8.0, 150 mM NaCl, 5 mM EDTA, 5 mM EGTA, 0.1% NP-40, 1 mM PMSF, 5μM MG132, 1× Phosphatase inhibitor cocktail 2 (Sigma-Aldrich), 1× Phosphatase cocktail 3 (Sigma-Aldrich), and a protease inhibitor (cOmplet EDTA-free Protease Inhibitor Cocktail, Roche) were added to the ground sample.The solution was sonicated followed by centrifugation (2 times at 20,000g for 10 min at 4°C) to remove insoluble cell debris.The supernatant constituted the protein extract that was used for immunoprecipitation.
One hundred microliters of Anti-FLAG M2 Magnetic Beads (Millipore) or RFP-trap magnetic beads (ChromoTek) were washed 3 times with 1 ml of SII buffer and then added to the protein extract.The whole solution was incubated at 4°C for 1 hour with 20 rpm agitation.Following this step, the beads were collected using a magnetic stand and washed 3 times with 10 ml of SPII buffer.The bound proteins were eluted by adding elution buffer [100 mM Na-phosphate, pH 8.0, 150 mM NaCl, 0.05% Triton X-100, 500μg/ ml 3× FLAG peptide (Millipore)] for FLAG-immunoprecipitation (IP) samples or, for mRed-IP sample, according to the manufacturer's instruction, incubating the beads with 0.2 M glycine pH 2.5 solution followed by adding 1 M Tris base pH 10.4 for neutralization.

LC-MS/MS analysis and data processing
Digested peptides were injected into a Thermo Scientific Ultimate 3000 coupled with a Thermo Scientific Q-Exactive Mass Spectrometer.The samples were run in a linear 90 min gradient at a flow rate of 300 nL/min.The raw files were searched against the Araport_11 database using Proteome Discoverer (version 2.2).The peptide precursor mass tolerance was set at 10 ppm, and MS/MS tolerance was set at 0.02 Da.The false discovery rate (FDR) at protein and peptide levels was set at 1%.The variable modification was set as oxidation on Methionine resides, and the cysteine carbamidomethylation was set as a static modification.

Characteristics of delayed GFP silencing in coi1 single mutants
The coi1 mutants were identified in a previous forward genetic screen based on a wild-type" T line (WT T) containing an alternatively spliced GFP reporter gene in Arabidopsis (Fig. 1) (Kanno et al. 2016(Kanno et al. , 2020)).WT T seedlings display an intermediate level of GFP fluorescence relative to 2 possible extremes: GFP-weak and hyper-GFP (Kanno et al. 2020).The coi1 mutants could be distinguished from wild-type in young seedlings by their hyper-GFP phenotype [Fig.2, a-c; 8 days after germination (DAG)] (Kanno et al. 2016).
After publication of these results (Kanno et al. 2016), we subsequently found that later in development-typically the second to fourth true leaf stage (referred to here as "older" seedlings, possibly coincident with the juvenile-to-adult transition; Bui et al. 2020)-40-90% of coi1 seedlings spontaneously undergoes a GFP silencing phenomenon characterized by the abrupt emergence of dark red leaves, indicative of GFP gene silencing, from a hyper-GFP shoot.This silencing does not infiltrate below the point of initiation to lower parts of the seedling, which stay hyper-GFP (Fig. 2, b and c, older seedlings).Once triggered, delayed GFP silencing in older coi1 seedlings persists in adult plants through the reproductive stage (Fig. 2, b and c, floral buds).

Coi1 mutants and WT T seedlings contain siRNAs likely derived from aberrant but not translatable GPF transcripts
To understand the delayed GFP silencing phenomenon in older coi1 mutants, we tested whether it was associated with GFP small RNAs.We sequenced small RNAs in coi1 mutant seedlings and as a control, in WT T seedlings, which we assumed would lack GFP small RNAs because there was no prior evidence for PTGS in this line.GFP small RNAs of both sense and antisense orientations were indeed detected, not only in the coi1 single mutant (Fig. 3a) but also, unexpectedly, in WT T plants displaying intermediate levels of GFP fluorescence (Fig. 4a).
In both the coi1 mutant (Fig. 3a) and WT T plants (Fig. 4a), the GFP small RNAs were predominantly 21-nt in length, with lesser amounts of 22 and 24-nt small RNAs.Although the siRNAs were primarily derived from the GFP protein CDS, much lower amounts of mainly 21-nt GFP small RNAs could be reproducibly detected upstream of the GFP CDS, particularly in an exon sequence present in the spliced GU-AG and unspliced transcripts but not in the translatable AC-AU variant (Supplementary Fig. 2, blue box).Neither of the former two transcripts is translatable owing to the presence of numerous stop codons (Fig. 1), which could potentially channel them into a pathway of siRNA production and PTGS (Liu and Chen 2016).However, the relative paucity of siRNAs from the unique central region of the unspliced transcript (Supplementary Fig. 2, red dotted box) suggests that the spliced GU-AG transcript is the primary source of upstream GFP small RNAs.The substantial quantitative reduction in GFP small RNAs in the upstream sequence compared to the GFP CDS may reflect the decreased activity of RDR6 as it progresses 3′ to 5′ along the substrate RNA (Moissiard et al. 2007; Supplementary Fig. 2).Additional evidence that the abundant small RNAs originating from the GFP CDS are also likely to originate from aberrant GFP transcripts and not the translatable AU-AC RNA is provided by the cwc16a mutant (Fig. 2d).This mutant was identified by its hyper-GFP phenotype in the same splicing screen as the coi1 mutant (Kanno, Lin, Fu, Matzke, et al. 2017;Kanno et al. 2020).The cwc16a mutant, which is defective in a putative step 2 splicing factor, produces almost exclusively the spliced, translatable AU-AC transcript (Kanno, Lin, Fu, Matzke, et al. 2017), and-in agreement with the hypothesis that GFP small RNAs are not derived from this transcript-the cwc16a mutant is essentially devoid of GFP small RNAs from both the GFP CDS and upstream regions (Fig. 5a; Supplementary Fig. 3).Consistent with the lack of GFP siRNAs in the cwc16a mutant, delayed GFP silencing has not yet been observed in hundreds of cwc16a seedlings examined, which remain hyper-GFP during the lifetime of the plant (Fig. 2d).
The collective findings indicate that the translatable AU-AC transcript is largely excluded from the small RNA biogenesis pathway.Thus, among the three GFP splice variants detected in the GFP splicing reporter system (Fig. 1), the untranslatable GU-AG transcript, and perhaps to a lesser extent, the unspliced transcript, represent the most credible sources of all GFP small RNAs (Supplementary Fig. 2).

GFP small RNAs are bona fide siRNAs that trigger delayed GFP silencing in coi1 mutants
To determine whether the GFP small RNAs are authentic siRNAs capable of triggering PTGS-of the GFP reporter gene, we crossed a coi1 mutant and WT T plants with mutants defective in the PTGS factors RNA-DEPENDENT RNA POLYMERASE 6 (RDR6), which copies aberrant RNAs to produce dsRNAs, and DCL4, which cleaves dsRNAs into 21-nt siRNAs (Lopez-Gomollon and Baulcombe 2022).As described below, these experiments not only implicated classical sense (S)-PTGS in delayed GFP silencing in older coi1 mutants, but they also revealed a previously unsuspected, moderate level of S-PTGS in WT T seedlings.
Relative to the single coi1 mutant (Fig. 3a), coi1 rdr6 seedlings displayed increased GFP fluorescence and the loss of virtually all size classes and orientations of GFP small RNAs (Fig. 3b).Notably, delayed GFP silencing similar to that observed in older coi1 single mutants (Figs. 2, b and c, 3a) was never observed in coi1 rdr6 seedlings, which remained hyper-GFP throughout growth and development (Fig. 3b).
Conversely, coi1 dcl4 seedlings showed substantially reduced GFP fluorescence compared to the coi1 single mutant (Fig. 3c).However, unlike delayed silencing in older coi1 seedlings, which does not infiltrate below the site of initiation (Figs.2, b and c hypocotyl (compare Fig. 3, a and c, white arrows).The more extensive GFP silencing in coi1 dcl4 seedings was associated with a substantial reduction in 21-nt GFP siRNAs and a massive increase in 22-nt GFP siRNAs (Fig. 3c; Supplementary Fig. 3c).Previous work has shown that a DCL4 deficiency results in elevated DCL2 activity on double-stranded RNA substrates, resulting in increased production of 22-nt siRNAs that can enhance silencing (Mlotshwa et al. 2008;Chen et al. 2010;Parent et al. 2015;Taochy et al. 2017).Similar changes were seen when rdr6 and dcl4 mutations were introduced into the WT T line: rdr6 mutations enhanced GFP fluorescence and eliminated GFP siRNAs; dcl4 mutations reduced GFP fluorescence and increased the accumulation of 22-nt siRNAs (Fig. 4, b and c, respectively).
Greatly elevated levels of 22-nt GFP siRNAs and extensive GFP silencing were also observed in cwc16a dcl4 double mutants (Fig. 5c).This finding indicates that the cwc16a single mutant, which normally does not produce GFP siRNAs or exhibit S-PTGS, is nevertheless capable of undergoing S-PTGS provided sufficient numbers of siRNAs, such as the elevated level of 22-nt siRNAs in a cwc16a dcl4 double mutant, are available.

Summary and limitations of our analysis of delayed S-PTGS in coi1 mutants and GFP siRNAs
The results of experiments with the rdr6 and dcl4 mutants demonstrated that the abrupt, nonspreading delayed silencing phenotype observed in older coi1 single mutants is due to canonical S-PTGS triggered by 21-22-nt GFP siRNAs, which likely originate largely from the aberrant GU-AG GFP splice variant.What remains unclear is why delayed S-PTGS is provoked suddenly at a relatively late period of seedling development and not earlier in young coi1 seedlings.Despite the presence of detectable amounts of 21-22-nt GFP siRNAs in young (2-wk-old) coi1 seedlings, they consistently display a hyper-GFP phenotype before the sudden appearance of delayed S-PTGS in older coi1 seedlings (Fig. 2).
One possibility is that the threshold number of siRNAs required to effectively induce S-PTGS is only surpassed in older coi1 seedlings, perhaps due to enhanced transcription of aberrant GFP splice variants at that time.It is currently difficult to evaluate this hypothesis because RNA-seq data are currently available only from 2-wk-old seedlings.The proportions of the 21-22-nt GFP siRNAs (Supplementary Fig. 3b) and the 3 GFP splice variants (Supplementary Fig. 4) differ to some extent between the WT T line and the coi1 mutant at 2 wks.However, the effects of these differences in GFP siRNAs are difficult to assess, since the threshold level of siRNAs needed to induce S-PTGS is not known.Moreover, the differences in GFP splice variants affect primarily the translatable AU-AC splice variant (Supplementary Fig. 4), which is not a major source of GFP siRNAs (Supplementary Fig. 2).Whether these quantitative variations in GFP siRNAs and splice variants observed in 2-wk-old seedlings are elevated during development to eventually trigger the sharp onset of delayed S-PTGS in older coi1 seedlings is not yet known.

Implications of pre-existing GFP siRNAs in the WT T line
Small RNA sequencing and experiments with rdr6 and dcl4 mutants also revealed that the WT T line contains an endogenous and formerly unsuspected Given the new finding that the WT T line contains an endogenous pool of GFP siRNAs that can provoke a moderate degree of S-PTGS, it is conceivable that some of the mutants we identified in the previous screen for splicing factors (Kanno et al. 2020) could also modulate S-PTGS.Indeed, a recent study reported that PRP39 (identified in our splicing screen by its hyper-GFP phenotype; Kanno, Lin, Fu, Chang, et al. 2017b), has distinct roles in not only splicing but also in S-PTGS (Bazin et al. 2023).
Note that the relationship between moderate S-PTGS in WT T and delayed S-PTGS in coi1 mutants is not entirely clear.Our data indicate that the two types of S-PTGS are phenotypically distinct with respect to the timing of silencing, the strength and frequency of silencing, and the GFP expression and silencing pattern (Table 1).In WT T seedlings, moderate S-PTGS occurs continuously during plant growth and development and affects 100% of seedlings.The pattern of the resulting intermediate GFP expression reflects the activity of the viral enhancer upstream of the GFP protein coding region (shoot apex, hypocotyl, and root; Fig. 2a; Table 1).This pattern is maintained when GFP expression is increased (that is, S-PTGS is suppressed) in an rdr6 mutant (Fig. 4,  a and b; Table 1).By contrast, the coi1 mutant shows a pervasive pattern of hyper-GFP fluorescence that is visible upon germination (Kanno et al. 2016) and maintained during the lifetime of the plant, provided delayed S-PTGS does not occur (Figs.4b, 6b).When delayed S-PTGS does occur in 40-90% of older coi1 seedlings, complete and nonspreading GFP silencing is observed in all aerial parts of the plant above the point of initiation (Fig. 2b; Table 1).We think this description of the 2 types of S-PTGS in the GFP splicing reporter system best fits our data.However, until this system is fully understood, we cannot completely rule out possible alternative interpretations.

Future perspective
An analysis of the time course of GFP pre-mRNA synthesis and alternative splicing patterns, as well as the accumulation of GFP siRNAs and GFP protein, starting with young coi1 and WT T seedlings and extending into later stages of development, would help to address remaining uncertainties about delayed S-PTGS in older coi1 seedlings.Investigation of these questions in our well-defined transgene system is likely to reveal more general principles relevant to developmentally associated changes in alternative splicing patterns that can potentially affect siRNA production and expression of endogenous genes.

Identifying proteins that interact with coilin
To learn more about components of coilin-dependent pathways in plants and how coilin mutations foster the hyper-GFP phenotype in young coi1 seedlings, we performed a genetic suppressor screen using a coi1 mutant and conducted an IP-mass spectrometry (MS) analysis using epitope-tagged coilin proteins.

Genetic suppressor screen
EMS mutagenesis and initial tests for putative suppressor mutants were performed as described in the Methods section using seeds of the hyper-GFP coi1-8 mutant (Supplementary Fig. 5).Of the GFP-weak/intermediate suppressor mutants that passed the initial tests and were subsequently carried on for further analysis (Methods), 5 contained a mutation in a gene retrieved in the former forward screen using the WT T line to identify splicing factors: prp4k, sac3a, cbp80, prp8a, and prp18a (Supplementary Table 1; Kanno et al. 2020).The finding of these 5 genes in both the splicing screen and the coi1 suppressor screen suggests that the involvement of the respective factors in GFP pre-mRNA splicing is independent of coilin function.However, 5 GFP-weak suppressor mutants contained mutations in novel factors that were identified only in the present coi1-8 suppressor screen: wrap53, smu2, zch1, mad1, and sde2-like (Supplementary Table 1).The failure to identify these 5 factors in the former splicing screen suggests their detection in our system is related in some way to the coilin deficiency in the coi1 mutant.Here we discuss wrap53, smu2, and zch1.In a coi1 mutant background, these mutations reduce the intensity of GFP fluorescence and the levels of GFP protein to approximately the intermediate state of the WT T line (Fig. 6a-e, 8 DAG; Supplementary Fig. 6).In addition, coi1 zch1 mutants eventually display enhanced GFP silencing that extends into the shoot apex and top of the hypocotyl (Fig. 6e).

WRAP53 (AT4g21520)
WRAP53/TCAB1 (WD40 encoding antisense to P53/telomerase Cajal body protein 1) is an evolutionarily conserved WD40 repeat protein that is widely distributed in the plant, fungal and animal kingdoms (Fig. 7a; Supplementary Fig. 7a).To our knowledge, WRAP53 has not yet been investigated in detail in plants nor have wrap53 mutations been recovered in any prior genetic screen.
In mammalian cells, WRAP53 has been shown to interact with coilin (Machyna et al. 2015) and to be essential for CB cohesion (Henriksson and Farnebo 2015).By contrast, we found that a wrap53 mutation-unlike a coi1 mutation-does not disrupt CB structure in Arabidopsis, at least in the cell type tested (trichomes) (Supplementary Fig. 8).Furthermore, we did not identify ATWRAP53 as a coilin-interacting protein in a comprehensive IP-MS analysis (Supplementary Table 2).These results suggest that WRAP 53 may act differently in plants than in mammals and not partner directly with coilin to support CB architecture.Additional work in the future is needed to confirm this conjecture and to determine the mechanism by which wrap53 mutations attenuate the hyper-GFP phenotype in young coi1 wrap53 seedlings.

SMU2 (suppressors of mec8 and unc52) (At2g26460)
SMU2 is a splicing factor that is conserved in most plant and animal species that carry out alternative splicing (examples in Fig. 7b; Supplementary Fig. 7b).Previous studies in plants found that SMU2 is important for constitutive and alternative splicing in maize (Chung et al. 2009) and for magnesium homeostasis in Arabidopsis (Feng et al. 2020).
Table 1.Summary of genotypes, GFP phenotypes, GFP siRNAs, and silencing patterns in the mutants described in this study.Hyper-GFP at all stages of development (Fig. 2d)

rdr6 coi1
Hyper-GFP throughout the plant (Fig. 3b) No GFP silencing, even in "older" seedlings (Fig. 3b) All GFP siRNAs were eliminated (Fig. 3b) GFP-weak: GFP fluorescence that is visibly lower than that observed in the WT T line.Observed in some splicing mutants (Kanno et al. 2020).GFP-weak/intermediate: GFP fluorescence that is greatly reduced from that of the coi1 single mutant to approximately the intermediate WT T level) (coi suppressor mutants; this study).GFP-negative (neg): whole seedlings (or parts of seedlings) exhibit no visible GFP fluorescence and appear dark red owing to the autofluorescence of chlorophyll at the excitation wavelength for GFP.Observed with gfp loss-of-function mutations (Fu et al. 2015;this study) and emerging leaves of hyper-GFP coi1 mutants undergoing delayed S-PTGS (this study).Delayed S-PTGS refers to abrupt and complete GFP silencing observed at the 2nd−4th true leaf stage in 40-90% of coi1 seedlings.Silencing does not spread below the site of initiation and is maintained throughout the lifetime of the plant.
Splicing, siRNAs, and coilin function | 9 In mammalian cells, the SMU2 ortholog, termed RED (Fig. 7b), interacts with the protein SMU1 to form a complex required for spliceosome activation (Keiper et al. 2019).Although interactions between SMU1 and SMU2 proteins have been reported in plants (Chung et al. 2009;Feng et al 2020), we were unable to demonstrate such an interaction using yeast 2-hybrid assays, possibly because the SMU1 protein was unstable in yeast cells under our experimental conditions.Interestingly, however, we did retrieve smu1 mutants in the previous screen using the GFP splicing reporter to identify splicing factors (Kanno, Lin, Fu, Matzke, et al. 2017;Kanno et al. 2020).Whereas a smu2 mutant (in a coi1 background) exhibits a GFP-weak/intermediate phenotype approaching that of the WT T line (Fig. 6d), a smu1 mutant (in a wild-type COI1 background) displays a hyper-GFP phenotype (Kanno, Lin, Fu, Matzke, et al. 2017).The opposite effects of smu1 and smu2 mutations on the same GFP splicing reporter gene probably reflect differing impacts on the splicing of the 2 mutations in specific genetic backgrounds.In GFP-weak/intermediate coi1 smu2 double mutants, the unspliced, untranslatable GFP pre-mRNA predominates (Supplementary Fig. 4), whereas, in hyper-GFP COI1 smu1 mutants, the translatable AU-AG transcript is the major splice variant (Kanno, Lin, Fu, Matzke, et al. 2017).A further interesting but currently unexplained feature of the splicing defects in coi1 smu2 mutants is the exceptionally high number of IR events, and to a lesser extent exon skipping (ES) and alternative 5′-splice site events compared to coi1 wrap53 and coi1 zch1 mutants (Supplementary Table 4).Further work is required to understand the mechanism of SMU2 in alternative splicing and the suppression of the hyper-GFP phenotype in coi1 seedlings.

ZCH1 (At1G48950)
ZCH1 is a zinc finger protein containing a zf-C3HC domain (Fig. 7c).Although this domain is widely distributed throughout the eukaryotic kingdom, the ZCH1 protein we identified is specific to plants (Supplementary Fig. S7C).We identified 5 zch1 alleles in our screen (Fig. 7c), all of which exhibit reduced GFP fluorescence in a coi1 mutant background (Fig. 6e, 8 DAG).Around 1-to 2-wks after germination, however, the initially GFP-weak/intermediate phenotype of coi1 zch1 mutants becomes even weaker and infiltrates downward into the hypocotyl (Fig. 6e).Although this silencing pattern is reminiscent of that seen in dcl4 backgrounds (compare Fig. 3d with Fig. 4c), the predominant size class of GFP siRNA associated with silencing differs in each case: 22-nt in the case of dcl4 (Figs. 3c,4c,5b) and 21-nt in the case of coi1 zch1 mutants (Supplementary Figs.3a-c).The increased level of 21-nt siRNAs in coi1 zch1 relative to a coi1 single mutant and other coi1 suppressor mutants (Supplementary Fig. 3a-c) hints that the wildtype ZCH1 protein may promote biogenesis and/or accumulation of GFP 21-nt siRNAs in a manner that remains to be determined.
During the course of our study, ZCH1 was also identified as MEM1 (methylation elevated mutant1) in Arabidopsis using a reverse genetics approach to identify new proteins involved in DNA demethylation (Lu et al. 2020).MEM1 was found to act similarly to ROS3 in a ROS1-mediated DNA demethylation pathway to remove methylation from transposons and transgene promoters (Lu et al. 2020).MEM1 has also been reported to safeguard against DNA damage (Wang et al. 2022).In our system, we did not detect changes in DNA methylation in the upstream enhancer of the GFP reporter gene in a coi1 zch1 mutant (Supplementary Table 3; Fig. 1).Therefore, the zch1 mutations we identified are not likely to reduce GFP expression in coi1 mutants through a mechanism that involves DNA methylation/demethylation of transcriptional regulatory regions.Conceivably, MEM1/ZCH1 could play roles in multiple processes including PTGS, DNA demethylation, and DNA repair depending on the physiological context in which it acts. .Silencing observed in coi1 zch1 double mutants is also distinct from the delayed silencing phenotype in coi1 single mutants, which occurs later during seedling growth and does not spread below the site of initiation (Fig. 3a).Moreover, the premature and extended silencing occurs in 100% of coi1 zch1 seedlings whereas delayed silencing occurs in only 40-90% of coi1 seedlings.Despite the differences in the frequency, timing, and pattern of coi1 zch1 silencing vs coi1 delayed silencing, both are associated with 21-nt GFP siRNAs derived from the GFP CDS (Fig. 3, a and d).
(a) WRAP53  S7a-c.a) ATWRAP53 (At2g21520) is 425 amino acids in length and encoded by a single copy gene in Arabidopsis.ATWRAP53 protein contains six WD40 repeat domains.In our screen, we recovered four wrap53 alleles, all of which contain a mutation in or around 1 of the 6 WD40 domains.b) ATSMU2 (At2g26460) is 585 amino acids in length and encoded by a single copy gene in Arabidopsis.In metazoans, SMU2 is referred to as RED, named after a region rich in Arg (R)/Glu (E) or Asp/Asp (D) repeats, which can also be identified in ATSMU2 close to the N-and C-termini.Our screen recovered 2 smu2 alleles, both of which contain mutations that are located in a recognizable RED domain.c) ATZCH1 (At1g48950) is a zinc finger protein that is 594 amino acids in length and encoded by a single copy gene in Arabidopsis.ATZCH1 contains one zf-C3Hc domain.Our screen recovered five independent zch1 alleles.

IP-mass spectrometry analysis (MS)
To identify putative coilin-interacting proteins, we performed an IP-MS analysis using three epitope-tagged coilin proteins: C-terminal mRED (coilin-mRed); N-terminal FLAG (FLAG-coilin); and C-terminal FLAG (coilin-FLAG) (Supplementary Table 2).This analysis identified several major functional categories of putative coilin-interacting proteins: (1) splicing factors, including many Sm proteins that constitute the SMN complex needed for biogenesis of spliceosomal snRNPs (Hebert et al. 2001: Xu et al. 2005); (2) histone proteins and histone acetylases/deacetylases, many of which are located in the nucleolus; and (3) nucleolar proteins involved primarily in rRNA processing or modification (Supplementary Table 2).These results are consistent with prominent roles for plant coilin in splicing, nucleolar chromatin structure, and ribosome function (Machyna et al 2015).Moreover, they are compatible with the observation that coilin is present in both CBs and nucleoli (Trinkle-Mulcahy and Sleeman 2017).

Summary of MS-IP analysis and coi1 suppressor screen
We identified a number of functionally diverse factors in the coi1 suppressor screen and the MS-IP analysis on coilin protein.These findings reflect the proposed multifarious roles of coilin in plants.Although we found no strict overlap between the proteins identified in the MS-IP analysis and coi1 suppressor screen, this is not unexpected because the former proteins presumably interact physically with coilin protein whereas those in the latter category are likely to involve functional relationships (for example, acting in the same pathway) that do not necessarily entail a physical interaction.Nevertheless, there was some functional overlap in proteins identified by the two methods.The findings of the smu2 mutant in the coi suppressor screen and numerous Sm proteins and other splicing-related factors in the MS-IP analysis support a prominent role for coilin and CBs in pre-mRNA splicing.In addition, the discovery of the zch1 mutant in the suppressor screen and its proposed influence on siRNA accumulation is consistent with the roles of coilin in various small RNA-mediated processes as revealed by the IP-MS analysis.

Conclusions
In this study, we used a GFP splicing reporter gene in a coi1 mutant background to carry out experiments designed to investigate an unusual gene silencing phenomenon appearing in older coi1 seedlings and to learn more about coilin functions and components of coilin pathways in plants.Our results indicate that delayed GFP silencing in older coi1 seedlings follows a canonical S-PTGS pathway that involves 21-22-nt siRNAs emanating largely from an aberrant splice variant of GFP pre-mRNA.The extent to which coi1 mutations enhance delayed S-PTGS is not yet known.So far, coi1 is the only hyper-GFP mutant we have identified (Kanno et al. 2020) that frequently displays such a striking delayed silencing phenotype, suggesting that the coi1 mutation somehow sensitizes the GFP reporter gene to S-PTGS.We also uncovered the existence of a previously unknown population of endogenous 21-24 nt GFP siRNAs in the WT T line.These siRNAs are able to induce a moderate level of S-PTGS leading to intermediate GFP fluorescence midway between GFP-weak and hyper-GFP phenotypes.The trans-generational stability of intermediate GFP fluorescence in the WT T line renders it a valuable tool for identifying mutations that either enhance or suppress GFP expression, and are thus likely to impair factors acting in alternative splicing, RNA silencing and, in a coi1 mutant background, general coilin functions in plants.
In addition to TGS/RdDM (Kanno et al. 2008;Eun et al. 2012) and alternative splicing (Kanno et al. 2020), the new findings of S-PTGS involvement add a third dimension to processes that can influence the activity of the GFP splicing reporter.Our detailed analysis of this system has revealed its versatility and usefulness for studying the contributions of different expression mechanisms and their coordinated influence on the activity of a complex genetic locus.
The coi1 suppressor screen identified several new proteins that may functionally interact with coilin.We provide foundational information for three of these proteins-WRAP53, SMU2, and ZCH1 -all of which remain to be fully characterized in plants.Future study of these proteins is not only of inherent interest but is also potentially valuable for discovering expanded functions of coilin in plants.Seeds of the WT T line and the mutants as well as the high-throughput sequencing data we generated are publicly available and provide many resources for the plant scientific community (Supplementary Table 1).
An IP-MS analysis substantiated major roles for plant coilin in splicing, nucleolar chromatin structure, and rRNA maturation.These processes require in some cases small RNAs.(Preuss et al. 2000;Kishore et al. 2010;Wang et al. 2015;Wang and Chekanova 2016;Park et al. 2019;Huang et al. 2022).The identification of zch1 mutations in the coi1 suppressor screen and their enhancement of GFP siRNA accumulation provide at least an indirect link between coilin function and siRNA biogenesis and/or stability.
Our cumulative results highlight the involvement of coilin in multiple processes, some of which can be unified through a requirement for small RNAs.These findings lend support to previous proposals suggesting that coilin may facilitate RNP biogenesis by acting as a chaperone for small nuclear noncoding RNAs (Machyna et al. 2015).Regarding coilin's participation in alternative splicing, our prior forward genetic screen to identify splicing factors (Kanno et al. 2020) represents, to our knowledge, the first in any organism based on an alternatively spliced gene and it is the only forward screen, apart from "no cajal bodies" (ncb), which retrieved coi1 mutants (Collier et al. 2006).These findings suggest that coilin protein may carry out a special role in specifically alternative splicing in a way that is worthy of further study.

Fig. 1 .
Fig. 1.Alternative splicing of GFP reporter gene.T-DNA construct: One complete copy of the indicated T-DNA construct is integrated into the genome of WT T plants.An endogenous pararetrovirus enhancer is upstream of a minimal 35S promoter (TATA) and the anticipated GFP protein CDS (long green bar)The 3 black arrowheads at the 5′ end of the enhancer represent a short tandem repeat, which becomes methylated in the presence of homologous siRNAs(Kanno et al. 2008).Opposing blue arrowheads denote the positions of primers used to detect DNA methylation (Supplementary Table3).Enhancer region, 1277 bp; minimal promoter, 91 bp; expected GFP CDS, 720 bp, Total length, 2088bp.Transcripts detected: Although transcription was predicted to start downstream of the TATA minimal promoter, this promoter, and the downstream ATG start codon remain unused (gray letters).Instead, transcription initiates at a cryptic promoter upstream in the tandem repeats (black arrow).Three polyadenylated GFP transcripts can be detected: an unspliced form (GFP pre-mRNA) and 2 spliced variants, which contain canonical (GU-AG) and noncanonical (AU-AC; GFP mRNA) splice sites, respectively(Kanno et al. 2008).Owing to the presence of stop codons (asterisks) in the unspliced and GU-AG transcripts, only the AU-AC transcript can be translated into GFP protein (green bar).Variations in the proportions of the 3 splice variants by alternative splicing modify the level of GPF fluorescence(Kanno et al. 2020).Translation initiates from a new upstream in-frame ATG start codon (black letters), which results in a 27 amino acid N-terminal extension to standard GFP protein (short green bar)(Fu et al. 2015).

Fig. 2 .
Fig. 2. Delayed GFP silencing in coi1 single mutants.a-c) Row 1: "Young" seedlings (8 DAG) of coi1 single mutants (2 alleles: coi1-7 and coi1-8; Supplementary Fig. 5) display hyper-GFP phenotypes relative to the WT T line, which has an intermediate level of GFP fluorescence.Rows 2, 3: In "older" coi1 seedlings (2nd-4th true leaf stage), abruptly delayed silencing, typified by dark red leaves emerging from a hyper-GFP stem, can occur.Similar delayed and abrupt silencing was also observed with a GUS transgene in Arabidopsis line L1 (Elmayan et al 1998) Row 4: Delayed silencing persists into the floral stage.A) all rows: Most WT T plants also continue to show faint but visible accumulation of GFP fluorescence into the floral stage.A, all rows).D, all rows): The hyper-GFP cwc16a mutant (Kanno et al. 2017a, 2020) does not undergo delayed silencing but remains hyper-GFP during the entire lifetime of the plant.

Fig. 3 .
Fig. 3. Delayed GFP silencing in coi1 mutants is due to siRNA-mediated s-PTGS.a) Top: delayed S-PTGS in coi1 single mutant.Hyper-GFP fluorescence is extinguished at the 2nd-4th true leaf stage and does not infiltrate below the site of initiation (white arrow).Bottom: GFP siRNAs, predominantly 21-nt in length (blue lines), are distributed throughout the GFP CDS (green bar) and more sparsely, in the upstream region flanked by AU and GU splice sites.b) Top: coi1 rdr6 double mutants remain hyper-GFP throughout the plant's lifetime and GFP siRNAs are essentially eliminated (Bottom).c) Top: coi1 dcl4 double mutants display strong GFP silencing in aerial parts of a seedling including, the primary true leaves (yellow arrow), which remain mostly hyper-GFP in the single coi1 mutant a, yellow arrow).Unlike the more spatially restricted delayed S-PTGS in coi1 single mutants, silencing in coi1 dcl4 double mutants extends into the shoot apex and the top of the hypocotyl (compare a and c, white arrows), coincident with a large increase of 22-nt GFP siRNAs (orange lines, Bottom).The more extensive GFP silencing observed in a coi1 dcl4 background potentially masks any abrupt GFP silencing that could occur later in older coi dcl4 double mutants.d) coi1 zch1 double mutants exhibit a pervasive silencing phenotype similar to that observed in coi1 dcl4 but the size class of the predominant GFP siRNA differs: 22 nt in coi1 dcl4 vs 21 nt in coi1 zch1.Seedlings were photographed at approximately 21 DAG.Length and distribution of GFP siRNAs along the GFP CDS and upstream region are indicated.Y-axes show read coverage per million reads (scale differs in each graph).Quantitative comparative differences are shown in Supplementary Fig. 3. AU-AC splice sites are shown in green letters, indicating that the resultant RNA is translated into GFP protein.GU-AG splice sites are shown in red letters indicating that the resultant RNA is untranslatable (see Fig. 1).

Fig. 4 .
Fig. 4. Pre-existing intermediate PTGS of GFP in WT t line.a) Top: The WT T line shows intermediate GFP fluorescence that is highest in the shoot apex, hypocotyl, and portions of the root (Daxinger et al. 2009).Bottom: GFP siRNAs in WT T are predominantly 21-nt in length (blue lines) and likely originate from the spliced, untranslatable GU-AG transcript (Supplementary Fig. 2).b) Top: an rdr6 mutation introduced into the WT T line increases GFP fluorescence in the shoot apex and hypocotyl and eliminates nearly all GFP siRNAs (Bottom).c) Top: a dcl4 mutation results in diminished GFP fluorescence that gradually spreads into the shoot apex and hypocotyl.Bottom: 22-nt siRNAs (red lines) are substantially increased in a dcl4 mutant background while 21-nt siRNAs are greatly reduced.Seedlings were photographed at 10 and 21 DAG.Length and distribution of GFP siRNAs along the GFP CDS and upstream region are indicated.Y-axes show read coverage per million reads (scale differs in each graph).Quantitative comparative differences are shown in Supplementary Fig. 3. AU-AC splice sites are shown in green letters, indicating that the resultant RNA is translated into GFP protein.GU-AG splice sites are shown in red letters indicating that the resultant RNA is untranslatable (see Fig. 1).

Fig. 5 .
Fig. 5. Delayed GFP silencing does not occur in a single cwc16a mutant, which lacks GFP siRNAs.a) Top: cwc16 mutants do not undergo delayed silencing, remaining hyper-GFP throughout the plant's lifetime, and (Bottom) contain negligible amounts of GFP siRNAs.b) Top: cwc16a dcl4 double mutants undergo silencing that extends into the shoot apex and top of hypocotyl, and they produce large quantities of 22-nt GFP siRNAs (Bottom) owing to strong DCL2 activity (presumably acting on dsRNA synthesized from minute amounts of aberrant GFP transcripts) in the absence of DCL4.Seedlings were photographed at approximately 21 DAG.Length and distribution of GFP siRNAs along the GFP CDS and upstream region are indicated.Y-axes show read coverage per million reads (scale differs in each graph).Quantitative comparative differences are shown in Supplementary Fig. 3. AU-AC splice sites are shown in green letters, indicating that the resultant RNA is translated into GFP protein.GU-AG splice sites are shown in red letters indicating that the resultant RNA is untranslatable (see Fig. 1).
refers to the level of GFP fluorescence in WT T relative to the GFP-weak and hyper-GFP phenotypes (observed in various mutants).GFP fluorescence in WT T is visible in the shoot apex, hypocotyl, and root, reflecting the expression pattern of the viral enhancer upstream of the GFP CDS (Fig.1).Fluorescence in WT T can be either enhanced or reduced, respectively, in rdr6 and dcl4 mutant backgrounds (this study).Hyper-GFP: GFP fluorescence observed in some splicing mutants of the WT T line (e.g.coi1, cwc16a;Kanno et al. 2020) that is considerably stronger than the intermediate GFP levels in WT T. The hyper-GFP phenotype is visible upon germination and throughout most of the seedling during growth and development.

Fig. 6 .
Fig. 6.GFP phenotypes in coilin suppressor screen mutants.a, b) Intermediate GFP fluorescence in the WT T line compared to the hyper-GFP phenotype in a nonsilenced coi1 single mutant, which was derived from the WT T line(Kanno et al. 2016), at 8 and 14 DAG.The coi1 suppressor screen identified mutations that attenuate the hyper-GFP phenotype of the single coi1 mutant.c-e) Reduced GFP expression in the coi1 wrap53, coi1 smu2, and coi1 zch1 double mutants at 8 and 14 DAG.The coi1 zch1 double mutant seedlings e) show shortly after germination a strong reduction of GFP fluorescence that gradually extends into the shoot apex and hypocotyl.This resembles the silencing pattern observed when dcl4 mutations are introduced into the WT T line, although the size class of siRNA that accumulates differs [21-nt in coi1 zch1 vs 22-nt in dcl4 (Fig.3, c and d)].Silencing observed in coi1 zch1 double mutants is also distinct from the delayed silencing phenotype in coi1 single mutants, which occurs later during seedling growth and does not spread below the site of initiation (Fig.3a).Moreover, the premature and extended silencing occurs in 100% of coi1 zch1 seedlings whereas delayed silencing occurs in only 40-90% of coi1 seedlings.Despite the differences in the frequency, timing, and pattern of coi1 zch1 silencing vs coi1 delayed silencing, both are associated with 21-nt GFP siRNAs derived from the GFP CDS (Fig.3, a and d).

Fig. 7 .
Fig.7.Gene structures, protein domains, and positions of mutations in Arabidopsis thaliana WRAP53, SMU2, and ZCH1.Intron-exon structures of the indicated genes and nucleotide sequence changes are shown at the top of each section; changes in amino acid sequences are at the bottom (asterisks indicate premature termination codons).Recognizable protein domains are indicated by red boxes.Amino acid sequence alignments of ATWRAP53, ATSMU2, and ATZCH1 proteins with the corresponding putative orthologs in other plant species and model organisms are shown in Supplementary Fig.S7a-c.a) ATWRAP53 (At2g21520) is 425 amino acids in length and encoded by a single copy gene in Arabidopsis.ATWRAP53 protein contains six WD40 repeat domains.In our screen, we recovered four wrap53 alleles, all of which contain a mutation in or around 1 of the 6 WD40 domains.b) ATSMU2 (At2g26460) is 585 amino acids in length and encoded by a single copy gene in Arabidopsis.In metazoans, SMU2 is referred to as RED, named after a region rich in Arg (R)/Glu (E) or Asp/Asp (D) repeats, which can also be identified in ATSMU2 close to the N-and C-termini.Our screen recovered 2 smu2 alleles, both of which contain mutations that are located in a recognizable RED domain.c) ATZCH1 (At1g48950) is a zinc finger protein that is 594 amino acids in length and encoded by a single copy gene in Arabidopsis.ATZCH1 contains one zf-C3Hc domain.Our screen recovered five independent zch1 alleles.