Missplicing suppressor alleles of Arabidopsis PRE-MRNA PROCESSING FACTOR 8 increase splicing fidelity by reducing the use of novel splice sites.

Efficient splicing requires a balance between high-fidelity splice-site (SS) selection and speed. In Saccharomyces cerevisiae, Pre-mRNA processing factor 8 (Prp8) helps to balance precise SS selection and rapid, efficient intron excision and exon joining. argonaute1-52 (ago1-52) and incurvata13 (icu13) are hypomorphic alleles of the Arabidopsis thaliana genes ARGONAUTE1 (AGO1) and AUXIN RESISTANT6 (AXR6) that harbor point mutations creating a novel 3'SS and 5'SS, respectively. The spliceosome recognizes these novel SSs, as well as the intact genuine SSs, producing a mixture of wild-type and aberrant mature mRNAs. Here, we characterized five novel mutant alleles of PRP8 (one of the two Arabidopsis co-orthologs of yeast Prp8), naming these alleles morphology of ago1-52 suppressed5 (mas5). In the mas5-1 background, the spliceosome preferentially recognizes the intact genuine 3'SS of ago1-52 and 5'SS of icu13. Since point mutations that damage genuine SSs make the spliceosome prone to recognizing cryptic SSs, we also tested alleles of four genes carrying damaged genuine SSs, finding that mas5-1 did not suppress their missplicing. The mas5-1 and mas5-3 mutations represent a novel class of missplicing suppressors that increase splicing fidelity by hampering the use of novel SSs, but do not alter general pre-mRNA splicing.

INTRODUCTION

Pre-mRNA splicing is a co-transcriptional, high-fidelity process consisting of two sequential transesterifications carried out by the spliceosome, resulting in the excision of an intron and the ligation of its flanking exons. These reactions depend on conserved intronic and exonic sequences: the branch-point sequence (BPS) and the 5′ splice site (5′SS) and 3′SS (reviewed in 1–3). Mutations that damage genuine (authentic) SSs or create novel SSs frequently produce a mixture of wild-type and aberrant mature mRNAs from a single pre-mRNA; these mutations are a major cause of several rare human diseases, including inherited mental disorders (4).

Pre-mRNA processing factor 8 (named Prp8 in the yeast Saccharomyces cerevisiae and PRPF8 in humans, and referred to across species as PRP8) is a core component of the spliceosome, as well as its largest (>2,000 amino acids) and most highly conserved protein. PRP8 is an essential gene whose loss of function causes global splicing defects in all organisms studied. The S. cerevisiae Prp8 gene was identified based on its conditional lethal alleles, and human PRPF8 was identified based on the retinopathies caused by its loss-of-function alleles (1).

Efficient splicing requires a balance between high-fidelity sequence selection and speed. Structural analysis has revealed two alternative conformations of PRP8: one promotes fidelity over catalytic efficiency and the other promotes efficient, error-prone splicing (5). Indeed, some missense alleles of PRP8 suppress missplicing of pre-mRNAs carrying mutations that damage SSs, which are not efficiently recognized by the spliceosome. Such extragenic suppressor alleles of PRP8 reduce the frequency of selection of cryptic SSs by the spliceosome in S. cerevisiae, humans, Caenorhabditis elegans and Arabidopsis thaliana (hereafter Arabidopsis) (1,6).

ARGONAUTE1 (AGO1) is a key factor of microRNA (miRNA) pathways in Arabidopsis. The ago1-52 mutation creates a novel 3′SS in the last intron of AGO1 and the spliceosome uses this novel 3′SS more frequently than the genuine 3′SS, causing missplicing of the AGO1 pre-mRNA (7,8). In a genetic screen for second-site suppressors, we previously mutagenized homozygous ago1-52 plants using ethyl methanesulfonate (EMS) and identified 22 lines carrying extragenic suppressor mutations of its morphological phenotype (9). We named the suppressor genes MORPHOLOGY OF AGO1-52 SUPPRESSED (MAS). Eleven of the suppressed lines carried mutant alleles of the gene we named MAS2, which encodes the Arabidopsis ortholog of human NF-κB activating protein (NKAP) (10). The mas2 mutations are predicted to cause amino acid substitutions, and the corresponding mutated MAS2 proteins act as dominant informational suppressors that partially suppress the missplicing of ago1-52 by an unknown mechanism (8).

Five of the suppressor lines carried novel alleles of PRP8 (AT1G80070), a gene that we initially dubbed MAS5. The Arabidopsis genome has two PRP8 co-orthologs (11), but only AT1G80070 has been well studied, and traditionally named PRP8 (its mutant alleles are named prp8). Due to the lethal effects of its loss-of-function mutations, other studies have named PRP8 as ABNORMAL SUSPENSOR 2 (SUS2), EMBRYO DEFECTIVE 14 (EMB14), EMB33 and EMB177 (12). The other PRP8 co-ortholog is AT4G38780, whose loss-of-function alleles do not seem to cause any phenotypic effect (11). The mas5 alleles of PRP8 may represent a novel class of missplicing suppressors that promote splicing fidelity by disfavoring the use of novel 3′SSs and 5′SSs created by mutation.

MATERIALS AND METHODS

Plant material and growth conditions

Arabidopsis thaliana Landsberg erecta (Ler), Columbia-0 (Col-0), Wassilewskija (Ws-2), and Enkheim-2 (En-2) wild-type accessions were obtained from the Nottingham Arabidopsis Stock Center (NASC) and propagated in our laboratory for further analysis. Seeds of prp8-6 (in the Ler genetic background) and prp8-7 (Col-0) mutants were provided by C. Dean and M. Matzke, respectively; ago1-2 (Ws-2) by C. Benning; ago1-25 and ago1-27 (Col-0) by H. Vaucheret; and sca3-1, anu4-1, and ang1-2 (Ler) by J.L. Micol. Seeds of icu13 (En-2; N349), sar1-4 (Col-0; SALK_126801) and atprmt5-1 (Col-0; SALK_065814) were provided by NASC. The ago1-51, ago1-52 and mas5 mutants were isolated in our laboratory (7,10). Seed sterilization and sowing, plant culture and crosses were performed as previously described (13,14).

Positional cloning of MAS5 and genotyping of single and double mutants

Genomic DNA extraction from mas5-1 plants and mapping of the mas5-1 mutation by iterative linkage analysis to molecular markers were performed as previously described (15,16). The PCR primers used for fine mapping are listed in Supplementary Table S1. The mas5 point mutations were identified by Sanger sequencing using the primers described in Supplementary Table S2. At least two M3 plants carrying each mas5 allele were backcrossed twice. Plants that were phenotypically wild type but genotypically AGO1/AGO1;mas5/mas5 (AGO1 being the wild-type allele of the AGO1 gene) were selected from the F2 progeny. The mas5 homozygous plants of the F2 progeny derived from the second backcross were used for all further studies described here. The ago1 mas5 double mutants studied in this work were also reconstructed from the mas5 lines isolated after two backcrosses.

The single and double mutants carrying point mutations were genotyped by Sanger sequencing (ago1-51, ago1-52, ago1-25, ago1-27, mas5-1, mas5-2, mas5-3, mas5-4, mas5-5, mas5-6, prp8-6, anu4-1, ang1-2, sca3-1 and icu13) or by restriction with MboII (prp8-7). The sar1-4 and atprmt5-1 insertional mutants were genotyped by PCR amplification. The primers used are listed in Supplementary Table S2. Most Sanger sequencing reactions and electrophoreses were carried out in our laboratory with ABI PRISM BigDye Terminator Cycle Sequencing kits and an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems). Some Sanger sequencing reactions were carried out at Stab Vida (Caparica, Portugal).

Plant morphometry and histology

Rosette pictures were taken with a Nikon SMZ1500 stereomicroscope equipped with a Nikon DXM1200F digital camera. For large rosettes, several pictures from the same plant were assembled with the Photomerge tool of Adobe Photoshop CS3 software. The backgrounds of the rosette pictures were homogenized using the Adobe Photoshop CS3 software without modifying the rosette images.

For cell size measurements, first-node leaves, collected 21 days after stratification (das), were cleared with ethanol and chloral hydrate, and mounted on slides. The samples were photographed under a Leica DMRB microscope equipped with a Nikon DXM1200 digital camera. The micrographs were transformed into diagrams by drawing the cell margins on a Cintiq 18SX Interactive Pen Display (Wacom) and using Adobe Photoshop CS3 software. Whole rosette area and palisade mesophyll cell size were measured with NIS Elements AR 3.1 (Nikon) software, as previously described (17).

RT-qPCR analysis

RNA isolation was performed using TRIzol Reagent (Invitrogen). Following cDNA synthesis, RT-qPCR analyses were performed in a Step-One Real-Time PCR System (Applied Biosystems) as previously described (7). Three biological replicates were used, each consisting of a mixture of three rosettes collected 15 das. Three technical replicates were used per biological replicate. The ACTIN2 (ACT2) housekeeping gene was used as an internal control for relative quantification.

Immunoblot analysis

Immunoblot analyses were performed as previously described (18). The anti-AGO1 (AS09 527; Agrisera), anti-CUL1 (kindly provided by J.C. del Pozo), and anti-RbcL (AS03 037; Agrisera) primary antibodies were used at 1:10,000, 1:3,000 and 1:2,500 dilutions, respectively. WesternSure HRP Goat anti-Rabbit IgG (LI-COR) secondary antibody was used at 1:50,000 dilution. Detection was performed using the WesternSure PREMIUM Chemiluminescent Substrate (LI-COR) and a C-Digit Blot Scanner (LI-COR). The Image Studio Analysis (LI-COR) software was used for band quantification.

RNA-seq and splicing analysis

Three biological replicates, each consisting of 5 μg of total RNA (isolated with TRIzol Reagent [Invitrogen]) from three rosettes collected 15 das, were sent to Novogene (Cambridge, United Kingdom) for high-throughput sequencing. cDNA libraries were produced with the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs) and sequenced on a NovaSeq 6000 Illumina platform using a S4 Flow Cell and a 2 × 150 bp paired-end run. More than 98.5 M non-stranded 150 bp paired-end reads, equivalent to 14.8 Gbp of raw data, were generated from each library (Supplementary Table S3). All the FASTQ files were submitted to the Sequence Read Archive (SRA) database of the National Center for Biotechnology Information (NCBI) under the BioProject accession PRJNA787038 (https://www.ncbi.nlm.nih.gov/sra/PRJNA787038).

Splicing analysis was carried out at the Bioinformatics for Genomics and Proteomics Unit of the Centro Nacional de Biotecnología (CNB, Madrid). Quality and purity of raw reads were assessed with FastQC 0.11.9 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and FastQ Screen 0.14.1 (19), respectively. Reads were aligned against the Arabidopsis Ler genome (NCBI accession GCA_001651475.1), using STAR 2.7.9a (20) with default parameters, except for –alignIntronMax and –alignMatesGapMax, which were set to 15,000. Potential optical duplicates and secondary alignments were identified and removed using the Picard Toolkit (https://broadinstitute.github.io/picard/) to get the effective reads from the aligned reads (Supplementary Table S3). Finally, differential splicing events were determined for each group pair (mas5-1 or mas5-3 versus Ler) by applying the standard pipeline defined for the ASpli 2.4.0 R package (21), and indicating a minimum read length of 100 bp and a maximum intron size of 14,334 bp, which corresponds to that of the longest intron in the reference genome. Briefly, multiexonic genes were divided into bins, which were then classified as exclusively exonic (including the external exons, defined as the first or last exon of a transcript), exclusively intronic, original intron (Ios, defined as introns before splitting, resulting from the retention/inclusion of two or more sub-bins), or annotated alternative splicing bins. Bins (excluding the external exons and Ios) were subjected to differential splicing analysis if genes with which they were associated were expressed above a minimum threshold of 10 reads in both genotypes compared, and if bins had >5 reads in at least one genotype. Finally, reads at the bin level were normalized to the read counts of their corresponding gene, and the differential bin usage was estimated. Only those bin-based splicing events with a false discovery rate (FDR) <0.1 and an absolute Delta (percent spliced-in, PSI) or Delta (percent intron retention, PIR) >5% were considered statistically significant.

RNA fluorescence in situ hybridization

Tissue preparation and RNA fluorescence in situ hybridization (RNA-FISH) were performed as previously described (22,23), using a 40-mer fluorescein-labeled oligo(dT) probe (Eurofins Genomics) at a concentration of 0.5 μg/ml in PerfectHyb Plus Hybridization Buffer (Sigma-Aldrich). Fluorescein was excited at 488 nm and its emission collected at 515/30 nm, maintaining the same detector gain to allow direct comparisons of fluorescence intensity between samples.

Accession numbers

Sequence data from this article can be found at The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org) under the following accession numbers: PRP8 (AT1G80070), AGO1 (AT1G48410), SCA3 (AT2G24120), ANU4 (AT1G02280), ANG1 (AT2G27530), AXR6 (AT4G02570; also called ICU13), SAR1 (AT1G33410) and ATPRMT5 (AT4G31120).

RESULTS

Isolation of dominant mutant alleles of PRP8 that suppress the morphological phenotype of ago1-52

ago1-52, a recessive and hypomorphic allele of AGO1, causes a pleiotropic phenotype (7) that is easily distinguishable from that of its wild type Ler at any developmental stage (Figure 1A and B). We performed a second-site mutagenesis screen for suppressors of the morphological phenotype of ago1-52 (10). The suppressor mutation carried by the M3 progeny of an M2 plant (P8 25.1) was named mas5-1 (Figure 1C) and crossed to Col-0 to obtain an F2 mapping population. Two phenotypic classes were defined: plants exhibiting the phenotype of ago1-52 and plants similar to a wild-type Col-0/Ler hybrid. We genotyped plants from each class for 32 molecular markers known to be polymorphic between Col-0 and Ler, as well as for the presence of ago1-52 and its AGO1 wild-type allele.

Figure 1.

Molecular nature and effects of the mas5 alleles of PRP8 examined in this study. (A–H) Suppression of the morphological phenotype of ago1-52 by the mas5 mutations. Rosettes of (A) the wild-type Ler, (B) the ago1-52 single mutant, and the (C) ago1-52 mas5-1, (D) ago1-52 mas5-2, (E) ago1-52 mas5-3, (F) ago1-52 mas5-4, (G) ago1-52 mas5-5, and (H) ago1-52 mas5-6 double mutants. The plants shown in (C–H) belong to the M3 generation of the genetic screen described in (10) and still had not been backcrossed to Ler. Photographs were taken 21 days after stratification (das). Scale bars: 4 mm. (I) Schematic representation of the PRP8 gene, indicating the nature and positions of the mas5 mutations and their predicted effects on the PRP8 protein. Empty and filled boxes represent untranslated and coding exonic regions, respectively. Lines between boxes represent introns, and red arrows indicate the positions of point mutations. Mutated nucleotides are shown in red. (J) Schematic representation of the PRP8 domains. The same colors have been used to highlight the regions of the PRP8 gene encoding the corresponding domains (in I), and those domains in the PRP8 protein (in J). Sequence and domain information about PRP8 was obtained from TAIR10 (https://www.arabidopsis.org/) and (3). (K) Prediction of the Arabidopsis PRP8 3D structure with indication of the residues altered by the mas5 mutations. The structure was downloaded from AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk/; PDB: AF-Q9SSD2-F1) and visualized with the ChimeraX 1.2.5 software (https://www.rbvi.ucsf.edu/chimerax/). PRP8 domain colors are the same than those used in (J), and residues altered by the mas5 mutations are shown in red. The close-up view of PRP8 surface, with focus on the region containing the mas5 mutations, has been shaded to highlight the protein cavities and pockets.

Iterative linkage analysis of these F2 plants with additional molecular markers, allowed us to define a 446-kb candidate interval flanked by cer461530 and cer470312 (Supplementary Table S1) that harbored the mas5-1 suppressor mutation. Among the genes within the above interval, we considered AT1G80070 (PRP8) to be the best candidate gene for MAS5, since it is one of the two co-orthologs encoding PRP8, the largest factor of the core spliceosome. Indeed, Sanger sequencing of AT1G80070 in ago1-52 mas5-1 plants revealed a G→A mutation that is predicted to cause an E1769K substitution (Figure 1I). We sequenced AT1G80070 in all lines carrying putative extragenic suppressors of ago1-52 that we had isolated in our screen and identified 10 that carried mas5 alleles with mutations in PRP8 (Supplementary Figure S1). Six of these mas5 alleles are unequivocally of different origins and some alleles were isolated twice in lines originating from the same parental group subjected to mutagenesis.

The mas5-1 to mas5-6 alleles of PRP8 caused different degrees of suppression in the M3 generation (Figure 1C–H). The mas5-1 and mas5-2 mutations are identical even though they were derived from different parental groups and therefore originated from independent mutational events (Supplementary Figure S1). However, the ago1-52 mas5-1 and ago1-52 mas5-2 plants of the M3 generation had different morphological phenotypes (Figure 1C and D), likely due to the presence of other mutations resulting from EMS mutagenesis. For example, the ago1-52 mas5-2 mutant also carried mas2-7, an allele of MAS2 that also suppresses ago1-52 (8). The mutational burden caused by EMS is also clearly evidenced by the chlorotic phenotype of the ago1-52 mas5-6 M3 plant shown in Figure 1H, as also observed in all ago1-52 mas5 lines shown in Figure 1. mas5-1 and mas5-3 F2 plants, from second backcrosses to Ler, were used for all further studies in this work.

The finding of six allelic mas5 mutations of independent origin in a single genetic screen strongly supports the hypothesis that PRP8 is the causal gene for the suppression of the ago1-52 phenotype in the ago1-52 mas5-1 to ago1-52 mas5-6 double mutants. The known functional role of PRP8 as a core component of the spliceosome also suggests that the mas5 alleles act as informational suppressors of the aberrant splicing of ago1-52.

Mutations in yeast Prp8 that act as missplicing suppressors map to the same regions that harbor mas5 mutations in Arabidopsis PRP8

Crystallographic structural analyses revealed the existence of five functional domains in yeast Prp8 (Figure 1J and K, and Supplementary Figure S2): the Reverse transcriptase-like, Linker, Endonuclease-like (24), RNase H-like (25) and C-terminal Jab1/MPN (26) domains. Crystallography also revealed the existence of a cavity formed by the Reverse transcriptase thumb (one of the three subdomains of the Reverse transcriptase-like domain [amino acids 1257–1375]), the Endonuclease-like domain (amino acids 1652–1821), and the RNase H-like domain (amino acids 1836–2091; Figure 1J and K, and Supplementary Figure S2). This cavity is involved in the interaction of PRP8 with the 5′SS, 3′SS and BPS of any intron (Supplementary Figures S3 and S4), contributing to the fidelity of the two splicing steps, and is where most missplicing suppressor mutations that have been mapped in yeast and humans are located (3,24,27,28).

To find the residues in yeast Prp8 that are homologous to those that the mas5-1 to mas5-6 mutations affect in Arabidopsis PRP8, we performed a multiple alignment of the amino acid sequences of PRP8 orthologs. We focused on species with the highest number of previously described mutations, including informational missplicing suppressor alleles from humans, S. cerevisiae, and C. elegans (Supplementary Figure S2). Then, using the cryo-electron microscopy (cryo-EM) structures of the yeast spliceosomal and post-spliceosomal complexes assembled on a single-intron pre-mRNA from the Protein Data Bank (https://www.rcsb.org) and the ChimeraX 1.2.5 software for their visualization, we located the homologous residues in Prp8 (Supplementary Figures S3 and S4). Specifically, we located the homologous residues in the most recently determined cryo-EM structures for five of the eight major functional states of the spliceosome: activated complex (Bact; PDB: 7DCO, at 2.5 Å resolution; 29), catalytically activated complex (B*; PDB: 6J6Q, at 3.7 Å resolution; 30), catalytic step I complex (C; PDB: 7B9V, at 2.8 Å resolution; 31), catalytically activated step II complex (C*; PDB: 5WSG, at 4.0 Å resolution; 32), and post-catalytic complex (P; PDB: 6BK8, at 3.3 Å resolution; 33).

Most amino acids affected by the mas5 mutations in Arabidopsis PRP8 are conserved across all PRP8 orthologs and five of these mutations affect residues of the cavity mentioned above, including the Linker region (Figure 1I-K and Supplementary Figures S2-S4). Three mas5 mutations causing the highest levels of suppression are located close to each other, affecting the end of the Endonuclease-like domain (mas5-1 and mas5-2) and the start of the RNase H-like domain (mas5-5). The mas5-3 and mas5-6 mutations damage the Linker region (amino acids 1375–1648), where several missplicing suppressor mutations also occur in yeast Prp8 (Figure 1I-K, Supplementary Figures S2-S4, and Supplementary Table S4).

The identical mas5-1 and mas5-2 mutations (Figure 1I–K) are predicted to cause an E1769K change, which substitutes a basic amino acid by an acidic one; their most similar mutation affecting the homologous residue of yeast is D-135, which causes an E1817G change, substituting a non-polar amino acid by an acidic one (Supplementary Figure S2 and Supplementary Table S4). This D-135 mutation suppresses missplicing caused by mutations in position 2 of the 5′SS of a reporter gene, which is used as a cryptic 5′SS (34). The mas5-5 mutation causes a D1894N change, which corresponds to D1942 of yeast Prp8. D and N are polar amino acids, but the substitution caused by mas5-5 adds an amino group and removes the negative charge of D (Supplementary Figure S2). In the B* to P yeast spliceosomal complexes, the homologous residues to those of Arabidopsis PRP8 affected by the mas5-1 (mas5-2) and mas5-5 mutations are very close to each other (Supplementary Figures S3 and S4).

The amino acids affected by mas5-3 (R1547K) and mas5-6 (G1524R) do not correspond to those of yeast suppressor mutations identified in the same region (W1609R, W1575R, E1576V, and T1565A; Supplementary Figure S2) (35). This region forms a disordered loop that interacts with the BPS (24). However, the mas5-3 and mas5-6 missense mutations cause opposite changes. On the one hand, R and K are similar positively charged amino acids, and the R1547K change only eliminates two amino groups. On the other hand, the G1524R substitution in the mas5-6 mutant adds three amino groups and a positive charge. We found that the yeast Prp8 residue homologous to the PRP8 residue affected by the Arabidopsis mas5-3 mutation (R1595) forms part of the so-called 1585-loop of this protein (amino acids 1585–1598 [32] or 1576–1599 [33]), which interacts directly with the intron lariat–3′ exon in the C* spliceosomal complex, stabilizing the 3′SS for the second transesterification (Supplementary Figure S4). However, the yeast Prp8 residue homologous to the PRP8 residue affected by the Arabidopsis mas5-6 mutation (G1572) was not located near the pre-mRNA, snRNAs, of any of the conserved residues affected by the other mas5 mutations, in all yeast spliceosomal complexes (Supplementary Figures S3 and S4).

The only mas5 mutation that affects the N-terminal domain of PRP8 is mas5-4 (T555I), but we observed a clear interaction of the homologous residue in yeast Prp8 with the 5′SS in all spliceosomal complexes (Supplementary Figures S3 and S4). An az50 semidominant mutation affecting the corresponding residue in C. elegans (T524S) was previously found in a screen for factors that modify the frequency of cryptic splicing, but does not produce overall changes in splicing (6).

mas5-1 does not suppress the morphological phenotypes of ago1-25 or ago1-27

For further analysis, we selected the mas5-1 and mas5-3 mutations, which affect two different regions: the Endonuclease-like domain and the Linker region of PRP8, respectively (Figure 1I-K). We backcrossed the ago1-52 mas5-1 (P8 25.1; Figure 1C) and ago1-52 mas5-3 (P7 24.1; Figure 1E) M3 lines twice to Ler. After these backcrosses, mas5-3 plants exhibited moderately larger rosettes compared to Ler and mas5-1 but were otherwise very similar to Ler throughout vegetative and reproductive development (Figure 2A, C, E, G and H). We then crossed the backcrossed mas5 mutants to ago1-52, to confirm the suppression, and to ago1-25 and ago1-27, to determine the specificity of the suppression. ago1-25 and ago1-27 carry EMS-induced point mutations in the Col-0 genetic background that cause single amino acid substitutions but do not alter their own pre-mRNA splicing (36).

Figure 2.

Suppression of the morphological and molecular phenotypes of ago1-52 by mas5-1 and mas5-3. (A–F) Rosettes of (A) Ler, (B) ago1-52, (C) mas5-1, (D) ago1-52 mas5-1, (E) mas5-3 and (F) ago1-52 mas5-3 plants. (G) From left to right, adult plants of Ler, mas5-1, mas5-3, ago1-52, ago1-52 mas5-1, and ago1-52 mas5-3. Photographs were taken (A–F) 21 and (G) 52 das. Scale bars: (A–F) 4 mm, and (G) 5 cm. (H) Boxplot showing the distribution of rosette areas in plants of the genotypes shown on the X-axis. Boxes are delimited by the first (Q1, lower hinge) and third (Q3, upper hinge) quartiles. Whiskers represent the most extreme data points that are no more than Q3 + 1.5 × IQR or no less than Q1 − 1.5 × IQR, where the interquartile range (IQR) is Q3 − Q1. ♦: Mean. —: Median. Asterisks indicate significant differences from the corresponding parental lines (indicated by color) in a Student's t-test (*P <0.05 and **P<0.0001). At least 15 rosettes per genotype were measured from plants collected 21 das. (I) Schematic representation of the AGO1 gene and molecular effects of the ago1-52 mutation. Gene structure is represented as described in the legend of Figure 1. gDNA and cDNA indicate genomic and complementary DNA, respectively. The molecular changes in mutant cDNAs and proteins are shown as red letters. The genuine and novel 3′SSs are boxed in blue and yellow, respectively. (J) RT-qPCR analysis of the expression of total (tAGO1), wild-type (wAGO1), and mutant (ago1-52) mRNA splice variants in plants of the genotypes shown. (K) Percentage of wAGO1 and ago1-52 mRNA splice variants. Error bars in (J, K) indicate standard deviation. (L) Detection of AGO1 protein isoforms by immunoblot analysis using a primary antibody against AGO1 (α-AGO1). Asterisks indicate the wild-type AGO1 (*) and mutant AGO1-52 (**) proteins. Detection of the RuBisCO large subunit with α-RbcL was used as a loading control. (M) Relative quantification of the wAGO1 and AGO1-52 proteins shown in (L), using the Image Studio Analysis software (LI-COR). Total RNA and proteins were extracted from plants collected 15 das.

In ago1-52 mas5-1 plants, and to a lesser extent in ago1-52 mas5-3, rosette size and whole plant height were partially restored to the Ler values (Figure 2A-H). The ago1-52 mas5-1 plants were similar to Ler, but ago1-52 mas5-3 resembled ago1-52, with dark green rosettes harboring two large, spatulate first leaves, like those of ago1-52 (Figure 2A-F). However, the main stem heights of ago1-52 mas5-1 and ago1-52 mas5-3 were closer to those of Ler, mas5-1 and mas5-3 (Figure 2G). We also obtained histological evidence for suppression: the small palisade mesophyll cell size of ago1-52 was normalized in ago1-52 mas5-1 plants, and to a lesser extent in ago1-52 mas5-3 (Supplementary Figure S5). We did not find any evidence of morphological suppression during vegetative or reproductive development in ago1-25 mas5-1, ago1-27 mas5-1, ago1-25 mas5-3 or ago1-27 mas5-3 plants (Supplementary Figure S6).

Taken together, these results indicate that mas5-1, and to a lesser extent mas5-3, appears to specifically suppress the ago1-52 allele. It is likely that the remaining mas5 alleles also act specifically on the ago1-52 mutation, but this has yet to be demonstrated.

The mas5 mutations modify the ratios of mRNA splice variants and protein isoforms produced by ago1-52

The novel 3′SS of ago1-52 seems to be used by the spliceosome more frequently than the genuine one, as shown by the level of the mRNA variant containing 10 nt of the 21st intron, which is more abundant than the wild-type variant. Translation of the misspliced ago1-52 mRNA variant produces a truncated protein (AGO1-52), which is 55 amino acids shorter than the wild-type AGO1 (wAGO1) and includes 15 amino acids not present in wAGO1 at its C-terminus (Figure 2I) (7).

To study the suppression of ago1-52 by mas5-1 and mas5-3 at the molecular level, we amplified total (tAGO1) and wild-type (wAGO1) mRNA splicing variants by RT-qPCR using specific primers (Supplementary Table S2) (8). In agreement with the different levels of morphological suppression in both double mutants, the ratio of wAGO1/ago1-52 mRNAs was higher in ago1-52 mas5-1 than in ago1-52 mas5-3 (Figure 2I-K). Therefore, the suppression of ago1-52 by mas5-1 could be due to the almost 10-fold increase in wAGO1 mRNA levels in ago1-52 mas5-1 plants, the reduced levels of aberrant ago1-52 mRNA in the double mutant, or both.

As a control, we performed immunoblot analysis using the ago1-2 null mutant (37), which does not produce any AGO1 protein (Figure 2L). The wAGO1 protein (∼130 kDa) was the only AGO1 protein detected in the Ler, mas5-1, and mas5-3 extracts. In agreement with our RT-qPCR results, we detected high levels of the mutant AGO1-52 protein (∼125 kDa) in ago1-52, along with low levels of wAGO1. We also detected two bands in ago1-52 mas5-1 and ago1-52 mas5-3, corresponding to the wAGO1 and AGO1-52 protein isoforms, as previously shown in ago1-52 mas2-1 plants (8). Therefore, the level of wAGO1 was higher than that of AGO1-52 in both ago1-52 mas5-1 and ago1-52 mas5-3 (Figure 2L and M). These results are in agreement with the stronger suppression of morphological defects in ago1-52 mas5-1 compared to ago1-52 mas5-3 (Figure 2A–H).

We repeated the RT-qPCR and immunoblot analyses with the original M4 lines harboring the mas5-4, mas5-5 and mas5-6 alleles (Figure 1F–H). We obtained similar results to those found with the ago1-52 mas5-1 double mutant with the mas5-6 allele, and less suppression with mas5-5, and in particular with mas5-4, as we observed with the mas5-3 allele in ago1-52 mas5-3 (Supplementary Figure S7). Therefore, the suppression of ago1-52 by the mas5 mutations may involve effects at the translational level; for example, the wAGO1 splice variant may be more translatable than the ago1-52 mRNA.

mas5-1 increases splicing fidelity in the icu13 allele of AXR6, which contains a novel 5′SS

To determine whether the suppression by mas5 alleles is specific to the AGO1 gene or whether it also occurs in other genes whose mutations eliminate or create novel SSs, we crossed mas5-1 to the incurvata13 (icu13), scabra3-1 (sca3-1), angulata4-1 (anu4-1) and angusta1-2 (ang1-2) mutants. We selected these four additional mutants because 1) they harbor the same type of transitions (G→A or C→T) but exhibit different types of missplicing, 2) they do not appear to be functionally related to each other or to AGO1 or PRP8, and 3) their morphological phenotypes are easily distinguishable by eye (Figure 3 and Supplementary Figure S8). We also crossed mas5-1 to the ago1-51 mutant, which is in a Ler background, like ago1-52. Unlike ago1-52 and icu13, the ago1-51, sca3-1, anu4-1 and ang1-2 point mutations damage genuine SSs, favorizing the recognition of nearby cryptic SSs by the spliceosome (Supplementary Figure S8).

Figure 3.

Suppression of the morphological and molecular phenotypes of icu13 by mas5-1. (A) AXR6 gene structure, mRNA splice variants, and CUL1 isoforms from the translation of icu13 mRNA transcripts, represented as described in the legend of Figure 2. A red arrow indicates the position of the icu13 mutation. (B–G) Rosettes of (B) En-2, (C) icu13 (in the En-2 genetic background), (D) icu13 (in the En-2/Ler hybrid genetic background), (E) Ler, (F) mas5-1, and (G) icu13 mas5-1. Photographs were taken 28 das. Scale bars: 1 cm. (H) RT-qPCR analysis of the expression of the total (tAXR6), wild-type (wAXR6) and mutant (icu13.2) mRNA splice variants in En-2, icu13, icu13 mas5-1, Ler, and mas5-1 plants. (I) Percentage of wAXR6 and icu13.2 mRNA splice variants. Error bars in (H, I) indicate standard deviations. (J) Immunoblot analysis of CUL1 proteins using a primary antibody against CUL1 (α-CUL1). Asterisks indicate RUB-modified CUL1 (*) and CUL1 (**). Detection of the RuBisCO large subunit with α-RbcL was used as a loading control. Total RNA and proteins were extracted from plants collected 15 das.

icu13 is a recessive hypomorphic allele of the AUXIN RESISTANT6 (AXR6) gene, which encodes CULLIN1 (CUL1), a component of the core SCF complex that catalyzes the ubiquitination of proteins for their degradation by the proteasome (38). A C→T transition in icu13 creates a novel 5′SS upstream of the genuine 5′SS of its 15th intron. The alternative use of both 5′SSs by the spliceosome generates two different splice variants from icu13. One of these mRNA variants (which we named icu13.1) has a synonymous mutation (GGC→GGU, both codons encoding glycine) and is produced when the genuine 5′SS is used by the spliceosome. The other mRNA variant (icu13.2) lacks the last 5 nt of the 15th exon, which causes a frameshift that generates a premature termination codon (PTC) due to the use of the novel 5′SS. Translation of the latter mRNA is predicted to produce a truncated protein (ICU13.2) with only 492 amino acids, instead of the 738 amino acids of the wild-type CUL1 (wCUL1) (Figure 3A).

We genotyped plants from all the phenotypic classes found in the F2 progeny of a mas5-1 × icu13 cross. The icu13/icu13;PRP8/PRP8 plants were identical to their icu13/icu13 parent, whereas icu13/icu13;PRP8/mas5-1 and icu13/icu13;mas5-1/mas5-1 plants were similar to En-2 (Figure 3B-G). These results indicate that mas5-1 acts as a dominant suppressor of the icu13 mutant phenotype, as it does for ago1-52. Accordingly, we analyzed the relative levels of the mRNA variants known to be produced by icu13 (18): wAXR6 (including the completely wild-type AXR6 variant and the icu13.1 variant, which carries a synonymous mutation) and icu13.2 (Figure 3A). Similar to previous findings, the levels of mature mRNAs produced by the icu13 allele of AXR6 were reduced 0.3-fold compared to wild type and less than 50% of these mRNAs were wAXR6 (including the icu13.1 variant). In the icu13 mas5-1 double mutant, however, the mRNA levels were higher, and wAXR6 became the major variant (Figure 3H and I).

icu13.2 might be targeted by the nonsense-mediated mRNA decay (NMD) pathway, as its mutation maps to the 15th of its 20 exons (Figure 3A) and produces a PTC at the 16th exon. NMD is the major RNA surveillance pathway and is universal among eukaryotes; NMD recognizes and elicits the degradation of unproductive mRNA variants with PTCs, thereby preventing their translation (39). This would explain the low levels of mRNAs produced by the icu13 allele, since its major variant icu13.2 is likely to be degraded by NMD. However, mas5-1 partially restored the use of the novel 5′SS of the icu13 pre-mRNA, thereby decreasing the icu13.2/wAXR6 ratio (Figure 3H and I). These findings explain why the total amounts of mature RNAs produced by icu13 in the icu13 mas5-1 double mutant were higher than those of the icu13 single mutant.

We also examined the protein products of icu13 in icu13 mas5-1 plants by performing an immunoblot assay with a polyclonal antibody against CUL1. CUL1 was more abundant in icu13 mas5-1 plants than in icu13 (Figure 3J). These findings confirm (at the protein level) the suppression of icu13 by mas5-1 that we observed at the morphological and mRNA levels. Similar to a previous report (18), we did not detect the predicted truncated CUL1 isoform (ICU13.2) in icu13 or icu13 mas5-1 plants, reinforcing the notion that the NMD pathway degrades the icu13.2 mRNA.

ago1-51 and sca3-1 carry transitions that damage a 5′SS of the AGO1 and SCA3 genes, respectively. In the case of ago1-51, three detectable mature mRNAs were produced, which include very low amounts of the wild-type variant (dubbed here as wAGO1) (Supplementary Figure S8A) (7,8). As in ago1-51, the cryptic 5′SS in sca3-1 appears to be stronger than the damaged genuine one, as shown by the very low levels of wild-type SCA3 mRNA (sca3-1.1 in Supplementary Figure S8B) compared to those of the sca3-1.2 variant, whose translation should result in a wild-type and a truncated protein, respectively (40). The anu4-1 and ang1-2 mutations damage a 3′SS that the spliceosome does not seem to recognize. Splicing of the anu4-1 and ang1-2 pre-mRNAs generates three different mRNA variants that suffer frameshifts (Supplementary Figure S8C and S8D), and in consequence do not produce detectable wild-type ANU4 and RPL10aB proteins, respectively (41). We performed Sanger sequencing to genotype plants from all the phenotypic classes in the different F2 populations that we obtained. The ago1-51 mas5-1, sca3-1 mas5-1, anu4-1 mas5-1 and ang1-2 mas5-1 double mutant plants were indistinguishable from their respective ago1-51, sca3-1, anu4-1 and ang1-2 single mutant F2 siblings (Supplementary Figure S8E-N), suggesting that in these mutants, mas5-1 does not reduce the frequency of the selection of cryptic SSs by the spliceosome.

Our results indicate that mas5-1 partially suppresses the missplicing caused by the preferential use of novel 5′SS (in icu13) or 3′SS (in ago1-52). Our results also suggest that mas5-1 (and probably the other mas5 mutations) does not have global effects on splicing. This would explain why the mas5-1 single mutant is similar to the wild type, as has been shown for several missplicing suppressor alleles of S. cerevisiae Prp8 and C. elegans prp-8 (1,6).

ago1-52 synergistically interacts with hypomorphic alleles of PRP8

To compare the functional nature of the mas5 alleles with other prp8 alleles previously studied, we crossed ago1 plants to prp8-6 and prp8-7 plants, which carry hypomorphic alleles of PRP8 (Figure 4 and Supplementary Figure S9). The prp8-6 mutant is in the Ler genetic background (42), as are ago1-51 and ago1-52, whereas prp8-7 is in the Col-0 genetic background (43), as are ago1-25 and ago1-27. Under our growth conditions, the rosette leaves of prp8-6 and Ler were very similar (Figure 4B), whereas those of prp8-7 were slightly pointed, serrated, and pale (Figure 4C).

Figure 4.

Genetic interactions between prp8 hypomorphic alleles and ago1-52. (A–F) Rosettes of (A) Col-0, (B) prp8-6, (C) prp8-7, (D) ago1-52, (E) ago1-52 prp8-6, and (F) ago1-52 prp8-7. Photographs were taken 21 das. Scale bars: 4 mm. (G) RT-qPCR analysis of the expression of total (tAGO1), wild-type (wAGO1), and mutant (ago1-52) AGO1 mRNA splice variants. (H) Percentage of wAGO1 and ago1-52 splice variants. Error bars in (G, H) indicate standard deviation. (I) Detection of AGO1 protein isoforms by immunoblot analysis using a primary antibody against AGO1 (α-AGO1). Asterisks indicate wild-type AGO1 (*) and mutant AGO1-52 (**). Detection of the RuBisCO large subunit with α-RbcL was used as a loading control. (J) Relative quantification of the wAGO1 and AGO1-52 proteins shown in (I), using the Image Studio Analysis software (LI-COR). Total RNA and proteins were extracted from plants collected 15 das.

The residue altered by the prp8-6 missense mutation (G1891E) in Arabidopsis is conserved with human PRPF8 (G1867) but not with yeast Prp8 (A1939; Supplementary Figure S2); PRP8 protein levels are similar in prp8-6 and the wild type (42). The Arabidopsis hypomorphic prp8-7 mutation causes a G1820E substitution in a 17-amino-acid extension within the RNase H-like domain of PRP8 (residues 1860–1875 in yeast, which correspond to residues 1812–1827 in Arabidopsis; Supplementary Figure S2). Cryo-EM analyses of yeast Prp8 revealed that this protein undergoes conformational rearrangements during pre-mRNA splicing and that the 17-amino-acid region can exist as a β-hairpin or a disordered loop, depending of the splicing step (44). Some missense prp8 alleles affecting residues of this 17-amino-acid region stabilize the disordered loop conformation, which in turn provides high efficiency but low fidelity to the splicing of pre-mRNAs from reporter constructs. The growth of these yeast mutants with error-prone splicing resembles that of the wild type. By contrast, other prp8 alleles harbor missense mutations affecting the same 17-amino-acid region that stabilize the β-hairpin conformation and cause low efficiency but high-fidelity splicing; therefore, the growth of yeast harboring these alleles is worse than the wild type (5). The global effect of the highly efficient but error-prone splicing caused by prp8-7 is the retention of a low amount (6.7%) of introns (43).

We believe that the different combinations of Col-0 and Ler genetic backgrounds that are present in the ago1-25 prp8-6, ago1-27 prp8-6, ago1-25 mas5-1, ago1-27 mas5-1, ago1-25 mas5-3 and ago1-27 mas5-3 double mutants contribute to their phenotypes (Supplementary Figures S6 and S9), making difficult to interpret the phenotypes of these double mutants. However, the ago1-52 prp8-6 and ago1-52 prp8-7 plants displayed a more severe morphological phenotype than ago1-52 and were completely sterile; in addition, the phyllotaxy of ago1-52 prp8-7 was strongly affected (Figure 4D–F). However, RT-qPCR and immunoblot analyses allowed us to conclude that the prp8-6 and prp8-7 mutations do not seem to modify ago1-52 pre-mRNA splicing, and that the synergistic morphological phenotypes of ago1-52 prp8-6 and ago1-52 prp8-7 plants cannot be explained by an increase in ago1-52 missplicing (Figure 4G-J). Nevertheless, these results clearly reveal that the mas5 alleles are not hypomorphic.

The mas5 mutations do not alter global pre-mRNA splicing, but modify the ratio of the proximal/distal 3′SS use in NAGNAG motifs

Some prp8 suppressor mutations modify the genetic interactions among mutant alleles of the genes encoding different spliceosome factors and cofactors, alleles that cause global missplicing (1). This is the case for prp8-8 and prp8-9; these alleles were isolated in a genetic screen for suppressors of the phenotype of the Arabidopsis atprmt5-1 mutant, which carries a T-DNA insertion in the 21st exon of PROTEIN ARGININE METHYLTRANSFERASE 5 (PRMT5). Arabidopsis PRMT5 regulates constitutive and alternative pre-mRNA splicing by promoting spliceosome assembly and activation (45–47). The atprmt5-1 mutation causes an increase in intron retention (IR) events, a global splicing alteration that is suppressed by the prp8-8 (P347S) and prp8-9 (P1141S) mutations, which are both considered neomorphic, since the loss of function of Arabidopsis PRP8 did not suppress the splicing defects of atprmt5-1 (48).

To shed further light on the functional nature of the mas5 alleles, we generated the atprmt5-1 mas5-1 double mutant, which was indistinguishable from atprmt5-1 (Supplementary Figure S10). Therefore, whereas prp8-8 and prp8-9 (which are dominant, like mas5-1 and mas5-3) suppressed atprmt5-1, mas5-1 did not. These results strongly suggest that these different alleles of PRP8 alter different PRP8 protein activities, as expected from the different locations of the amino acids changed by the mas5-1, prp8-8 and prp8-9 mutations in PRP8 (Supplementary Figure S2).

To test whether the mas5 mutations cause global alterations in pre-mRNA splicing, we carried out RNA-seq analyses of RNA extracted from Ler, mas5-1, and mas5-3. Using the ASpli software (21), 98,488 exons and 118,974 introns from 21,790 multiexonic genes were evaluated for each sample (Supplementary Dataset S1), which excluded external exons and Ios (see Materials and Methods). We filtered bin-based splicing events using an FDR <0.1 and an absolute Delta PSI or Delta PIR >5%. We only found 251 and 164 differential splicing events in mas5-1 and mas5-3, respectively, compared to the wild type, 33 of which were common to both mutants (Figure 5A and Supplementary Datasets S2–S4).

Figure 5.

Genome-wide analysis of pre-mRNA splicing in the mas5 mutants. (A) Percentage of differential splicing events identified in mas5-1 and mas5-3, and Venn diagrams showing the IR and Alt 3′SS events. IR: intron retention; ES: exon skipping; Alt 3′SS/5′SS: alternative 3′/5′ splicing site. (B, C) Plots of AT1G65520 and AT1G16010 aligned reads taken as representative examples of (B) IR and (C) Alt 3′SS events that were statistically different in the three biological samples of Ler (in grey), mas5-1 (in red), and mas5-3 (in blue). Only the intron and flanking exons (represented in gene structures by black lines and boxes, respectively) corresponding to the sites of the events are shown in plots obtained with the IGV software (http://software.broadinstitute.org/software/igv/). (D) DNA sequences corresponding to the statistically significant Alt 3′SS events identified in mas5-1 and mas5-3, with an absolute Delta PSI >20%. Intronic and exonic sequences are shown in lowercase and uppercase, respectively. The proximal and distal 3′SSs are boxed in blue and yellow, respectively. E0XX indicates the exon number of each gene.

Increased IR events were the most frequent: 201 (80.1% of the total missplicing events) and 117 (71.3%) in mas5-1 and mas5-3, respectively, with 25 common to both mutants (Figure 5A and Supplementary Datasets S2-S4). The second most frequent event found in both mutants was decreased Alt 3′SS: 31 (12.3%) and 25 (21.4%) in mas5-1 and mas5-3, respectively, with only 7 common to both mutants (Figure 5A and Supplementary Datasets S2-S4).

Using the IGV software, we confirmed the increased IR events in both mutants (Figure 5B) and found that 27 out of 31 (in mas5-1) and 21 out of 25 (in mas5-3) of the decreased Alt 3′SS events affected tandem 3′SSs, which were exactly 3 nt apart (NAGNAG). We also confirmed that both mutants used the proximal 3′SS more frequently than the distal one, compared to the wild type, which also uses both 3′SSs (Figures 5C and D, and Supplementary Figure S11). The presence of NAGNAG motifs in the 3′SS occurs widely in eukaryotic genomes, including the human and Arabidopsis genomes, in which 1,890 have been found by analyzing 435 RNA-seq datasets, with a mean number of 201 NAGNAG motifs with confirmed alternative use per sample (49). Because both 3′SSs are exactly 3 nt apart (in-frame), their alternative choice for the spliceosome would produce proteins differing in a single amino acid, which might not affect its function. Interestingly, in all cases the proximal 3′SS, which seems to be the strongest one because is more frequently chosen by the wild type, seemed to be more favored over the distal 3′SS in both mas5 mutants (Supplementary Datasets S2 and S3).

Comparing the number of IR events previously detected in prp8-7 (8,124 events affecting 6.7% of total introns; 43), we conclude that the mas5-1 and mas5-3 suppressor mutations do not alter global pre-mRNA splicing.

Misspliced mRNAs prevent the recruitment of mRNA export factors, causing nuclear accumulation of poly(A)+ RNAs (50), which should be evident in prp8-7, but not in mas5-1 or mas5-3, according to the RNA-seq results. To test this hypothesis, we carried out RNA-FISH assays with a fluorescently labeled oligo-dT probe against poly(A)+ RNAs. We used as a positive control the sar1-4 mutant, which carries a null allele of SUPPRESSOR OF AUXIN RESISTANCE1 (SAR1), encoding NUCLEOPORIN160 (NUP160), and shows elevated nuclear retention of poly(A)+ RNAs (22,50). We found nuclear accumulation of poly(A)+ RNAs within the nucleus of prp8-7 and sar1-4 leaf palisade mesophyll cells, but not in mas5-1 (Figure 6). These results further support our RNA-seq results.

Figure 6.

Detection of poly(A)+ RNAs in prp8-7 and mas5-1 leaf cells. (A–O) Poly(A)+ RNA-FISH assays in palisade mesophyll cells of (A, F, K) Col-0, (B, G, L) prp8-7, (C, H, M) sar1-4, (D, I, N) Ler and (E, J, O) mas5-1 leaves. Fluorescent signals correspond to (A–E) nuclear 4′,6-diamidino-2-phenylindole (DAPI) staining, (F–J) fluorescein from an oligo(dT) probe, and (K–O) their overlay. Confocal laser-scanning micrographs were taken from 10 leaves per genotype of plants collected 14 das. Scale bars: 50 μm.

DISCUSSION

The mas5 mutations belong to an unusual class of missplicing suppressors that increase splicing fidelity without causing a dramatic global increase in missplicing

Point mutations at the 5′SSs or 3′SSs may abolish or reduce their recognition by the spliceosome, which then recognizes a cryptic site close to the genuine, mutated SS. Moreover, point mutations in exonic or intronic sequences may create novel SSs. PRP8 encodes the core component of the spliceosome and is conserved across all eukaryotes. Numerous prp8 alleles have been described to act as missplicing suppressors, mainly in S. cerevisiae. Most of these mutations promote changes in SS choice by the spliceosome, increasing the frequency of proper, in-frame splicing of pre-mRNAs. These suppressor mutations reduce the frequency of use of cryptic SSs, which are already close to a mutated genuine SS (1).

We identified five different, allelic mas5 mutations in the Arabidopsis PRP8 gene in a second-site mutagenesis screen for extragenic suppressors of the morphological phenotype of ago1-52, a mutant allele of AGO1 that undergoes missplicing. Except mas5-4, all mas5 mutations affect residues of the cavity and the Linker of PRP8, where most yeast suppressor mutations are also located (Figure 1J and K, and Supplementary Figure S2). In the five mas5 alleles, the wild-type residue is predicted to be replaced by a positively charged amino acid (K in the identical mas5-1 and mas5-2 mutations, and mas5-3, and R in mas5-6), or a negatively charged amino acid is removed (D in mas5-5). The cavity interacts with the 5′SS, 3′SS and BPS, and is thought to play an important role in splicing fidelity. Our results suggest that regions that make up the cavity are also hotspots for missplicing suppressor mutations in Arabidopsis PRP8 that are likely involved in splicing fidelity. However, when we looked at the yeast residues homologous to those affected by the Arabidopsis mas5 mutations in five cryo-EM structures of spliceosomal complexes, we did not find interactions with mRNA, snRNAs, or other spliceosomal factors, except for the residues homologous to those affected by the mas5-3 and mas5-4 mutations; these residues could interact with the intron lariat–3′ exon (stabilizing the 3′SS for the second transesterification) and 5′SS of the single-intron pre-mRNA used in the models, respectively (Supplementary Figures S3 and S4). It is possible that the mas5 mutations modify the conformation of these spliceosomal complexes, but this remains to be tested.

Some dominant alleles are antimorphic (with a dominant negative effect); when they are heterozygous with a wild-type allele, they antagonize the function of the wild-type protein, thus leading to a loss of function. This mainly occurs in genes encoding subunits of multimeric complexes (51), as is the case of PRP8. In genetic screens, performed in S. cerevisiae, several dominant alleles of Prp8 have been identified that alter both SS choice by the spliceosome and alternative splicing efficiency (1). Many of these mutations do not have detrimental effects or visible phenotypes.

However, null alleles of genes encoding components of the spliceosome, or its associated factors, can cause global missplicing and lethality, as is the case of PRP8. Our RNA-seq analysis showed that the mas5-1 and mas5-3 mutations, and probably the other mas5 mutations, do not cause global defects in splicing, since only a few missplicing events were detected. Most of these missplicing events are increased IR events, corresponding to 0.17% (201 events) and 0.1% (117 events) of the introns analyzed in mas5-1 and mas5-3, respectively (Figure 5A and Supplementary Figures S2 and S3), which are minimal compared to those found in the prp8-7 mutant (8,124 events, corresponding to 6.7% of total introns), which exhibits a very weak morphological phenotype (Figure 4C) (43). Our results suggest that the mas5-1 and mas5-3 mutations improve the choice of the strongest 3′SS (the proximal one) compared with the wild type, at least in cases where there are NAGNAG sequences. These findings are in line with our previous results that suggested an increase in splicing fidelity in the ago1-52 mas5-1, ago1-52 mas5-3 and icu13 mas5-1 double mutants, and explain why mas5-1 and mas5-3 plants exhibit a wild-type phenotype. Based on their suppression of the missplicing of ago1-52 and icu13, we propose that the mas5 mutations represent a class of novel and uncommon PRP8 alleles whose behavior differs from that of alleles that increase splicing fidelity by suppressing cryptic splicing.

In animals and land plants, around 25% of the alternative splicing events are due to the use of alternative 3′SSs and 5′SSs, and about half of these 3′SSs are present in a NAGNAG motif and thus are separated by only 3 nt (52–55). In most cases, NAG tandem repeats are in phase and their differential splicing events give rise to a protein with an insertion or a deletion of a single amino acid (52,53,56). There is evidence that both protein isoforms from hundreds of genes with NAGNAG 3′SSs exist in Arabidopsis, rice (Oryza sativa), and the moss Physcomitrella patens (53–55). For example, the alternative splicing of the 3′SS of intron 14 in the Arabidopsis ZINC-INDUCED FACILITATOR-LIKE1 gene produces two mRNA variants that differ by 2 nt. One of these mRNAs codes for a full-length protein that localizes to the plasma membrane and functions in auxin-regulated processes, whereas the second variant codes for a truncated protein that localizes to the tonoplast membrane and functions in drought tolerance (57). The use of an alternative 3′SS in a NAGNAG sequence also produces the two isoforms of Arabidopsis U1-35K, a factor involved in splicing of rare U12-type introns. The shorter isoform, which lacks a glutamine, exhibits altered binding affinity to different components of the spliceosome complex (58). These studies suggest the functional significance of alternative splicing, as a result of the presence of tandem 3′SSs in plants.

In addition, our mas5 alleles appear to differ from other Arabidopsis prp8 dominant alleles, such as prp8-8 and prp8-9 (48), because they do not suppress the morphological phenotype caused by mutations in ATPRMT5, which encodes another spliceosome-related factor (Supplementary Figure S10). These findings strongly suggest that these different alleles of PRP8 alter different activities of PRP8, as expected based on the different localizations of the mas5-1, prp8-8 and prp8-9 mutations (Supplementary Figure S2).

Effect of mutations that create novel SSs but do not alter genuine SSs in model species and humans

Base substitutions are the most frequent type of mutations induced by the chemical mutagens most widely used to study model organisms, and they represent the major form of spontaneous genetic polymorphisms found in many species, including humans (59). The identification of mutated genes that cause a phenotype of interest has traditionally relied on the use of iterative linkage analysis to identify candidate mutations. Such candidate mutations are commonly chosen by focusing mostly on nonsynonymous substitutions in exons or, to a lesser extent, on substitutions that disturb SSs (60). However, in not few cases, none of the candidate genes was ultimately found to be the causal gene for the phenotype under study, despite recent progress in whole-genome sequencing technologies. Some of these cloning failures could be due to mutations that remain unnoticed because they create a synonymous codon or occur in a deep intronic region that does not form part of a genuine SS. Nevertheless, the effects of these apparently silent mutations can be strong, since some create novel SSs that are favored by the spliceosome compared to the genuine SSs, even though these SSs are otherwise intact. ago1-52 and icu13 belong to this class of mutations. The morphological and molecular phenotypes of ago1-52 are caused by a point mutation in an intronic region that has no obvious functional role, whereas in icu13, these phenotypes appear to be caused by a synonymous change at the end of an exon. In both cases, however, the mutation creates a novel SS that causes missplicing.

Recent studies integrating DNA and RNA data from whole-genome exon sequencing and transcriptomic analysis revealed that human mutations in deep intronic regions or those that yield synonymous codons in coding regions are the causes of several hereditary disorders and have been associated with cancer. A computational genomic analysis of 235 individuals of the 1000 Genomes Project estimated that each genome contains an average of 10 intronic mutations in sequences other than SSs or BPS. These mutations are associated with disorders, since they generate novel SSs without damaging genuine SSs, which in turn cause missplicing and often introduce a PTC in the misspliced mRNA (60). In addition, computational analysis of 8,656 tumors from The Cancer Genome Atlas project discovered several hundred novel mutations in intronic sequences, which cause missplicing and might have an impact on cancer; some of these mutations damage key tumor suppressor genes, such as TP53, the key tumor suppressor gene that encodes P53, the so-called guardian of the genome (61,62). These mutations cannot be detected by sequencing exomes, which is the most frequently used method to identify mutations associated with human genetic disorders.

The study of missplicing suppressors may be useful for a better understanding of splicing, as well as for engineering SS selection by the spliceosome

Due to its relative simplicity and rapid growth, S. cerevisiae has traditionally been recognized as the best model organism to study several cross-kingdom conserved processes, including splicing. However, 97% of protein-coding genes of S. cerevisiae lack introns, and several splicing factors and cofactors that are present in multicellular organisms are not encoded by its genome, including those that participate in alternative splicing, an event that is rare in this yeast but common in plants and animals (63). Several animal species are used as models to better understand missplicing causing human diseases and to design strategies for suppressing missplicing (64). Our findings indicate that Arabidopsis, like other multicellular organisms, could be useful for analyzing human disorders involving highly conserved genes, such as PRP8. It might be possible to suppress the effects of some mutations that cause missplicing in mammalian and particularly human cells by obtaining mutations equivalent to the mas5 mutations that mutate amino acids that exhibit cross-kingdom conservation and do not impair Arabidopsis growth or development.

Our findings also suggest that mutants that show missplicing may be good candidates for investigating both missplicing suppression and splicing itself. Indeed, such an approach might be a better choice than using minigenes to recapitulate artificial exon skipping events, because mutations such as those in the mas5 lines, are present in their natural cellular and chromosomal context.

DATA AVAILABILITY

Sequence data from this article can be found at The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org) under the following accession numbers: PRP8 (AT1G80070), AGO1 (AT1G48410), SCA3 (AT2G24120), ANU4 (AT1G02280), ANG1 (AT2G27530), AXR6 (AT4G02570), SAR1 (AT1G33410), and ATPRMT5 (AT4G31120). All the FASTQ files were submitted to the Sequence Read Archive (SRA) database of the National Center for Biotechnology Information (NCBI) under the BioProject accession PRJNA787038 (https://www.ncbi.nlm.nih.gov/sra/PRJNA787038).

Click here for additional data file.