Salt responsive alternative splicing of a RING finger E3 ligase modulates the salt stress tolerance by fine-tuning the balance of COP9 signalosome subunit 5A

Increasing evidence points to the tight relationship between alternative splicing (AS) and the salt stress response in plants. However, the mechanisms linking these two phenomena remain unclear. In this study, we have found that Salt-Responsive Alternatively Spliced gene 1 (SRAS1), encoding a RING-Type E3 ligase, generates two splicing variants: SRAS1.1 and SRAS1.2, which exhibit opposing responses to salt stress. The salt stress-responsive AS event resulted in greater accumulation of SRAS1.1 and a lower level of SRAS1.2. Comprehensive phenotype analysis showed that overexpression of SRAS1.1 made the plants more tolerant to salt stress, whereas overexpression of SRAS1.2 made them more sensitive. In addition, we successfully identified the COP9 signalosome 5A (CSN5A) as the target of SRAS1. CSN5A is an essential player in the regulation of plant development and stress. The full-length SRAS1.1 promoted degradation of CSN5A by the 26S proteasome. By contrast, SRAS1.2 protected CSN5A by competing with SRAS1.1 on the same binding site. Thus, the salt stress-triggered AS controls the ratio of SRAS1.1/SRAS1.2 and switches on and off the degradation of CSN5A to balance the plant development and salt tolerance. Together, these results provide insights that salt-responsive AS acts as post-transcriptional regulation in mediating the function of E3 ligase.

In recent years, a vast amount of data about AS have been produced by genome-wide studies in various organisms. However, despite the extensive AS data gathered, thus far, there is a gap between the large amounts of sequencing data and the functional identification of AS variants [24]. Consequently, the role of AS in regulating salt signaling and plant stress adaptations remains elusive. We are still in the early stage of understanding the potential mechanism of AS in response to environmental stresses [24].
In this study, we investigated the function of a RING-type E3 ligase gene, SRAS1, which generated two transcript isoforms that had opposing responses and divergent functions under salt stress. Our results provide novel insights into the AS-mediated post-transcriptional regulation of E3 ligase.

Two isoforms of RING-type E3 ligase SRAS1 exhibit opposite responses to salt stress
Recent studies have demonstrated that plant stress-related genes are particularly prone to AS events, which often modulate the ratio between active and non-active isoforms in response to abiotic stress [32,33]. To investigate the AS control of salt stress responses, we analyzed publicly available expression data through the Genevestigator mRNA-Seq platform (https://www. genevestigator.ethz.ch). This analysis identified the gene AT5g66070 as a salt-responsive gene (S1A Fig). AT5g66070 has been named Arabidopsis Tóxicos en Levadura 27 (ATL27), because it was originally assigned as a member of the E3 ubiquitin-protein ligase ATL subfamily [34]. And AT5g66070 has also been named ABA-related RING-type E3 ligase (AtARRE), as a negative regulator of ABA signaling [35]. In addition, the AT5g66070 transcripts are reported by TAIR (http://www.arabidopsis.org/) to undergo several AS events. Based on its salt inducibility, as well as on the capability of its transcript to undergo AS, we propose to rename this gene Salt-Responsive Alternatively Spliced gene 1 (SRAS1). SRAS1 consists of 5 exons and 4 introns, and according to the TAIR database could generate five splice variants. In contrast to the documented information, by reverse transcription-PCR (RT-PCR) we only detected two distinctive transcript isoforms of SRAS1 under salt stress. Intriguingly, the two AS transcripts exhibited opposite responses to salt stress ( Fig 1A). Sequencing of the PCR products revealed that the bottom band was SRAS1.1, consisting of 666 nucleotides encoding the full-length SRAS protein with the RING domain, whereas the upper band was SRAS1.2, which encodes a truncated isoform generated from an intron retention (IR) splicing event involving the first intron ( Fig  1A). SRAS1.2 was a novel transcript and not documented in TAIR website. We identify SRAS1.2 as AT5G66070. 6. This AS event introduced a premature termination codon (PTC) so that SRAS1.2 encoded a small sized protein with 59 amino acids lacking the C-terminal RING finger domain (Fig 1A). To investigate the relationship between SRAS1.1 transcript levels and salt stress in Arabidopsis, we performed RT-PCR analysis. Interestingly, levels of the SRAS1.1 transcripts were elevated after 0.5 h of NaCl treatment and continued to increase with prolonged treatment, whereas the levels of SRAS1.2 transcripts decreased dramatically as the duration of salt treatment increased (Fig 1B). Quantitative real-time PCR analyses were consistent with the above result, indicating that the SRAS1 transcripts had opposite responses to salt stress ( Fig 1C).

The truncated isoform SRAS1.2 is not an active E3 ligase
To detect spatial expression patterns of SRAS1 transcripts, we performed quantitative realtime RT-PCR (qRT-PCR) analyses of RNA extracts from diverse plant tissues. The results revealed similar expression patterns of SRAS1.1 and SRAS1.2. Expression levels were strongest in flowers and stems; moderate in cauline leaves, rosette leaves, and roots; and weakest in siliques and seeds (Fig 2A). These results confirm that SRAS1 is constitutively expressed in various tissues; however, the expression levels of SRAS1.2 are approximately 2-fold higher than SRAS1.1 in flowers and stems, suggesting that the isoforms might have tissue-specific functions in different tissues.
Next, we examined the subcellular localization of the SRAS1.1 and SRAS1.2 proteins by transiently expressing SRAS1.1-GFP and SRAS1.2-GFP fusion proteins in Nicotiana benthamiana leaf cells. The fluorescence signals of SRAS1.1 and SRAS1.2 were detected in both cytoplasm and nucleus. The RING-finger domain was also localized in both the cytoplasm and nucleus ( Fig 2B). These observations confirm that SRAS1.1 and SRAS1.2 have similar localizations (Fig 2B). Considering that SRAS1.2 is severely truncated and its GFP fusion protein might be diffusible, we also fused SRAS1.2 with a GFP dimer to increase its molecular weight and the subcellular localization remained the same (S1B  Values are shown as the mean ± SE from three biological repeats. Statistically significant differences were identified between pairs of measurements using Student's t-test ( �� P < 0.01, ��� P < 0.001).
https://doi.org/10.1371/journal.pgen.1009898.g001 amino acid residues, including a conserved RING domain in the C-terminus. RING-type proteins always function as E3 ubiquitin ligases [17]. By contrast, SRAS1.2, truncated at the C-terminus, carries no RING domain, leading us to investigate whether the SRAS1.2 still had E3 ligase activity or not. To answer this question, we fused SRAS1.1 and SRAS1.2 with the GSTtag to create GST-SRAS1.1 and GST-SRAS1.2, and subjected these proteins to an in vitro autoubiquitination assay (Fig 2C). In the presence of E1, E2 and Myc-ubiquitin, the GST-SRAS1.1 protein exhibited clear autoubiquitination, implying that SRAS1.1 exhibits E3 ligase activity in vitro. On the other hand, we detected no ubiquitination of GST-SRAS1.2 ( Fig 2C). Taken together, these results indicate that SRAS1.1 is an active E3 ligase, but SRAS1.2 is not.

SRAS1.1 is a regulator in salt stress signaling
To investigate the biological roles of SRAS1 under salt stress, we ordered a sras1-1 (Salk_034426) knockout mutant in which a T-DNA insertion in the 5' untranslated region of the gene almost completely disrupted SRAS1.1 and SRAS1.2 expression (S2A Fig). Relative to wild-type (WT) plants, sras1-1 mutants were more sensitive to salt stress (Fig 3A-3F). We also constructed overexpression transgenic lines in the Col-0 ecotype background. Two 35S::   1 seeds germinated, whereas 57.3% of WT, 86.9% of SRAS1.1 and 43.33% of SRAS1.2 seeds germinated (Figs 3A-3C and S2D and S2E). After the seedlings grew in 1/2 MS medium containing 200 mM NaCl for 14 days, the fresh weight of sras1-1 and SRAS1.2 overexpression lines was lower than that of WT, whereas the fresh weight of SRAS1.1 overexpression lines was greater (Figs 3D and S2F and S2G). Seedlings grown in normal 1/2 MS medium for 3 d were transferred onto 1/2 MS medium with or without 100 mM NaCl and grown for another 7 d. Seedling root growth of sras1-1 and SRAS1.2 overexpression lines was reduced after salt treatment. And sras1-1 mutants showed about 72% reduction in root growth relative to the untreated counterparts. Under the same salt conditions, 35S::SRAS1.1 lines grew better than the sras1-1 mutant, the 35S::SRAS1.2 lines and the WT (Fig 3E and 3F). To achieve SRAS1.2 overexpression with full-length cDNA, we also created point mutations at the 5' splice site (from GU to AT) in the second intron to produce SRAS1. To test whether SRAS1.1 and SRAS1.2 could rescue the mutant salt-sensitive phenotype, we generated sras1-1/35S::SRAS1.1 and sras1-1/35S::SRAS1.2 transgenic plants overexpressing SRAS1.1 and SRAS1.2 proteins in the sras1-1 mutant background. When germinated on medium with salt stress, both sras1-1 and the sras1-1/35S::SRAS1.2 were more sensitive to salt stress than the WT in terms of both the germination rate, seedling fresh weight and root growth (Fig 3A-3F). In sharp contrast, sras1-1/35S::SRAS1.1 plants were more tolerant to salt stress. In the presence of 200 mM NaCl, the fresh weight of sras1-1 was reduced by~64% relative to WT, whereas the fresh weight of sras1-1/35S::SRAS1.2 plants was even lower than for sras1-1, indicating that the sras1-1 mutant phenotype was only rescued by overexpressing SRAS1.1, not SRAS1.2. These results showed that overexpression of SRAS1.1 made the plants more tolerant to salt stress, suggesting that SRAS1.1 plays a positive role in regulating plant resistance to salt stress.

SRAS1 is involved in mediating salt-responsive gene expression
To understand the transcriptional changes in the SRAS1.1-overexpressing lines, we conducted a transcriptome analysis using 14-day seedlings of the 35S::SRAS1.1 transgenic lines and the Col-0 controls. Three biological replicates were sequenced for each growth condition. The average read length aligned to the Arabidopsis reference genome (TAIR) was approximately 101 bp. Each library contained 23-24 million reads. Approximately 90% of reads mapped to unique loci, and 190 genes in 35S::SRAS1.1 transgenic plants were significantly differentially expressed relative to the WT (130 up-regulated and 60 down-regulated; criteria: P value < 0.001, fold change > 2; the details of each gene are provided in S1 Table) (Fig 4A-4C). We found that more genes were upregulated than downregulated. To evaluate the relationship between the genome-wide expression profiles, we performed linkage hierarchy clustering of differential expression between the WT sample and SRAS1.1 samples.
Genomic analysis of the gene list showed that the genes induced by SRAS1.1 were coexpressed with genes induced by ferric chelate reductase activity, oxidoreductase activity, and abiotic stress. Gene Ontology (http://www.geneontology.org/), Kyoto Encyclopedia of Genes and Genomes (http://www.genome.jp/kegg/) analyses revealed that upregulated genes were closely related to salt stress (GO0009607) and light stress (GO0016723). These genes can be divided into four categories according to their biological roles: molecular chaperones, energy factors, signal transmission, and metabolic factors (Figs 4D and S3A-S3C). We hypothesize that SRAS1.1 regulates the expression of particular stress-resistance genes and light-related genes under high-salinity conditions. To validate the RNA-seq analysis, we checked the  (Fig 4G and 4H). Taken together, these findings suggest that SRAS1.1 plays a role in salt stress signaling.

SRAS1 physically interacts with CSN5A
In general, RING-type E3 ligases function by ubiquitinating target proteins and triggering their degradation via the 26S proteasome [17,36,37]. To elucidate the regulatory mechanism of SRAS1 in response to salt stress, we performed a yeast two-hybrid (Y2H) screen to identify possible interacting proteins (S4 Fig). Based on this screen, we discovered that CSN5A was a putative interacting partner of SRAS1.1. CSN5A is an essential subunit of CSN5, which is an important unit of the COP9 signalosome (CSN), an evolutionarily conserved multiprotein complex that mediates light-regulated development and stress-linked resistance in plants and mammals [38,39]. There are two CSN5 subunits in Arabidopsis, CSN5A and CSN5B. The two subunits are assembled into distinct CSN complexes in vivo, which are present in different abundances; CSN5A appears to be predominant [40,41]. We found that the CSN family member CSN5A interacted with both SRAS1.1 and SRAS1.2 ( Fig 5A). Due to the small size of the SRAS1.2 protein, we speculated that the binding of SRAS1.1 and SRAS1.2 to CSN5A was stimulated by the presence of the short N-terminal motif ( Fig 5B). To test this hypothesis, we estimated the binding activity in a yeast two-hybrid system. The results revealed that amino acids 1-19 in the N-terminus of SRAS1 were critical for the interaction between SRAS1 and CSN5A ( Fig 5B). To corroborate the interaction between SRAS1 and CSN5A, we performed a bimolecular fluorescence complementation (BiFC) assay. Reconstituted fluorescence was detected in the SRAS1.1-cYFP/nYFP-CSN5A and SRAS1.2-cYFP/nYFP-CSN5A samples, indicating that SRAS1.1 and SRAS1.2 interacted with CSN5A in vivo ( Fig 5C). Next, we confirmed the physical interaction between SRAS1 and CSN5A by in vitro pull-down assay. CSN5A-His interacted with both GST-SRAS1.1 and GST-SRAS1.2 ( Fig 5D). We also performed an in vivo coimmunoprecipitation (Co-IP) assay to check the interaction of CSN5A with SRAS1.1 and SRAS1.2 in plants. In these experiments, the 35S::SRAS1.1-GFP and 35S::SRAS1.2-GFP constructs were expressed in N. benthamiana through Agrobacterium tumefaciens-mediated infiltration. Total proteins were immunoprecipitated with an anti-GFP affinity gel matrix, and the bound proteins were eluted and subjected to immunoblot analysis with anti-GFP and anti-CSN5 antibodies. The results revealed that both SRAS1.1 and SRAS1.2 interacted with CSN5A ( Fig 5E). In addition, an in planta luciferase complementation imaging (LCI) assay showed that coexpression of SRAS1.1 and SRAS1.2 with CSN5A generated strong luminescence signals that were not detected in the control pairs (Fig 5F and 5G). Collectively, these results indicate that SRAS1 physically interacts with CSN5A.
Some recent studies showed that AS-generated splicing isoforms from one gene could interact with each other in the form of dimers [28,32]. To test this possibility, we performed Y2H and BiFC assays, both of which confirmed that SRAS1.1 interacted with SRAS1.2 (S5A and S5B Fig).

SRAS1.1 promotes CSN5A degradation by the 26S proteasome under salt stress
To confirm our speculation that CSN5A is a substrate of SRAS1.1, we conducted an in vitro ubiquitination assay. CSN5A was expressed with a His tag in E. coli, and SRAS1.1 was expressed with a GST tag. The results revealed that CSN5A was ubiquitinated by SRAS1.1 in the presence of E1, E2 and ubiquitin. When any of these essential proteins was omitted from the reaction, the ubiquitinated form of CSN5A was not detected. These data directly demonstrate that CSN5A is a substrate of SRAS1.1 (Fig 6A).
Next, we performed western blotting to detect CSN5A protein levels in the WT, SRAS1.1overexpressing, SRAS1.2-overexpressing and sras1-1 lines. The results revealed that the levels of CSN5A protein were lower in the SRAS1.1-overexpressing lines than in the WT, sras1-1 mutant and SRAS1.2-overexpressing lines (Fig 6B). Because SRAS1.1 expression levels were very low under normal growth conditions, and this gene could be induced by NaCl treatment, we then evaluated the stability of CSN5A with or without NaCl treatment in the SRAS1.1-overexpressing, sras1-1 mutant and SRAS1.2-overexpressing lines. A cell-free protein degradation assay showed that the levels of CSN5A protein were slightly lower in the WT in the absence of salt treatment, but decreased more rapidly in the SRAS1.1-overexpressing transgenic lines under the same condition, especially after 10 min (Fig 6C and 6D). By contrast, the CSN5A

PLOS GENETICS
Functional analysis of SRAS1 variants in Arabidopsis protein levels remained steady in the sras1-1 mutant and SRAS1.2-overexpressing lines ( Fig  6C). Under salt stress conditions, the degradation of CSN5A was significantly faster in SRAS1.1-overexpressing lines and the WT, but slower in the SRAS1.2-overexpressing lines (Fig 6C). CSN5A-His was much faster and had a shorter half-life in 35S::SRAS1.1#1, with or without salt treatment (Fig 6C and 6D). However, there was no significant difference at the transcription level (S6A and S6B Fig). Taken together, these results demonstrated that the degradation of CSN5A by SRAS1.1 was accelerated under salt stress (Fig 6C and 6D). Hence, we hypothesized that the reduction in CSN5A levels in transgenic lines was due to degradation by 26S proteasome. To test this hypothesis, we investigated whether treating cells with the MG132 would attenuate the degradation of CSN5A. WT, SRAS1.1-overexpressing, SRAS1.2overexpressing and sras1-1 mutant lines were treated with the inhibitor, and total proteins were extracted. Nearly 78.5% of CSN5A disappeared after 1 hour. MG132 almost completely blocked degradation of CSN5A in the SRAS1.1-overexpressing transgenic lines, indicating that CSN5A was degraded by the 26S proteasome ( Fig 6E). These data directly demonstrate that CSN5A degradation promoted by SRAS1.1 is dependent on the 26S proteasome pathway.

SRAS1.2 suppresses the degradation of CSN5A by competing with SRAS1.1
The results described above raised an important question about the function of the truncated isoform SRAS1.2. The E3 ligase activity of SRAS1.1 is essential for its ability to promote CSN5A degradation under salt stress; however, SRAS1.2 also interacted with CSN5A even though it lacked E3 ligase activity. Therefore, we sought to determine what role SRAS1.2 plays in response to salt stress. We hypothesized that SRAS1.2 acted as a competitor that interferes with CSN5A degradation mediated by SRAS1.1. To test this idea, we performed a dose-dependent in vitro competitive pull-down experiment in which we investigated whether SRAS1.2 affected the interaction between SRAS1.1 and CSN5A. As shown in Fig 7A, increasing amounts of GST-SRAS1.2 decreased the amount of GST-SRAS1.1 bound to CSN5A-His, indicating that the interaction between CSN5A and SRAS1.1 was significantly weakened when SRAS1.2 was co-expressed with the other two proteins (Fig 7A). We performed western blotting to determine whether CSN5A protein levels are affected in response to salt stress ( Fig 7B). Indeed, the levels of CNS5A gradually decreased during NaCl treatment (Figs 7B and S6C). Moreover, LCI assay showed that co-expression of SRAS1.2 and SRAS1.1 with CSN5A decrease the luminescence signal, implying SRAS1.2 affects the interaction between SRAS1.1 and CSN5A (Fig 5E and 5F). In Y2H assays, SRAS1.1 and SRAS1.2 share the same binding site of CSN5A (Fig 5B). For this reason, we propose that SRAS1.2 inhibits the degradation of CSN5A by competing with SRAS1.1.

CSN5A acts downstream of SRAS1 to modulate salt stress response
CSN5A acts as a molecular switch between the stress response and development in Arabidopsis [42]. To validate the genetic hierarchy between SRAS1 and CSN5A, we crossed the sras1-1 and csn5a-2 mutants and analyzed the phenotype under salt stress (Fig 7C-7F). In the presence of 100 mM NaCl, the germination rate of csn5a-2 was higher than those of WT and sras1-1 ( Fig  7D and 7F). The germination percentage of sras1-1 csn5a-2 seeds was comparable to that of csn5a-2 seeds (Fig 7D and 7F). The sras1-1 mutant is sensitive to salt stress, and sras1-1 csn5a-2 double mutant rescues the sras1-1 salt-sensitive phenotype (Fig 7D and 7F).
The transcriptome analysis showed that the expression levels of genes involved in oxidoreductase activity changed (Fig 4D). Thus, we detected the H 2 O 2 level in WT, SRAS1.1-overexpressing, SRAS1.2-overexpressing, sras1-1, csn5a-2 and sras1-1 csn5a-2 lines under salt stress (Fig 7G and 7H). The cellular level of H 2 O 2 was higher in sras1-1 mutant seedlings than in WT seedlings, whereas the H 2 O 2 level was lower in csn5a-2 mutant seedlings. And the H 2 O 2 level of sras1-1 csn5a-2 seedlings was comparable to that of csn5a-2, indicating that CSN5A acted downstream of SRAS1 (Fig 7G and 7H). Based on our findings, we propose a model illustrating the different roles of the AS-generated SRAS1 transcripts to balance the CSN5A at the crossroads of growth and stress responses (Fig 8).

Crucial roles of CSN5 in balancing plant development and salt stress tolerance
The CSN is the photomorphogenic complex regulating the activity of cullin-RING E3 ubiquitin ligases [43,44]. Among the eight CSN subunits, CSN5 subunit harbors the activity center of the CSN, binding numerous regulators and differentially affects the stability of each [41,45]. CSN5, encoded by CSN5A and CSN5B, is a key factor in plant development. The Arabidopsis csn5a csn5b double mutant has a lethal phenotype. The null mutant csn5a phenotype is less severe than the phenotype of csn5a csn5b double mutant, whereas the phenotype of csn5b mutant is similar with WT, suggesting that CSN5A is the predominantly functional subunit in plant development and growth [46][47][48]. The multiple functions of CSN5 have been extensively explored. The csn5b mutant is more tolerant to salt stress. CSN5B interacts with a GDP-mannose pyrophosphorylase VTC1 regulating ascorbic acid (AsA) synthesis in response to salt stress [46]. This investigation shows that the seeds of csn5a mutant are tolerant to salt stress than WT seeds, and the H 2 O 2 level is lower in csn5a mutant seedlings (Fig 7C-7H). The detailed mechanism on how CSN5A controls ROS accumulation is still not clear. However, we speculate that this may be related to the role of CSN5A in ABA signaling. The previous study showed that CSN5A regulated seed germination by facilitating degradation of the RGL2 and ABI5 proteins [42]. Together, these observations revealed that CSN5A tightly modulates plant growth and salt stress responses. It should be noted that the csn5b mutant is more tolerant to salt stress than csn5a, implying that CSN5B plays the predominantly role in salt resistance.
This raises the question of how CSN5A is regulated between plant growth and stress resistance. In this study, we have found that the salt-responsive AS of the E3 ligase SRAS1.1 finetunes the CSN5A degradation. The salt stress mediates the ratio of SRAS1.1/SRAS1.2 to switch on and off the degradation of CSN5A (Fig 8).
To test whether SRAS1.1 interacts with CSN5B, Y2H and Co-IP assays were performed. We found that SRAS1.1 interacted with CSN5B in yeast, but SRAS1.1 did not interact with CSN5B in plants (S7A and S7B Fig). We speculated that CSN5A and CSN5B share high sequence similarity. A cell-free protein degradation assay showed that the CSN5B protein levels remained steady in SRAS1.1-overexpressing lines, indicating that SRAS1.1 did not affect the CSN5B degradation (S7C and S7D Fig).

Emerging roles of truncated splicing isoforms in environmental stress adaptation
The ubiquitin-proteasome system (UPS), including ubiquitin (Ub), E1, E2, E3, 26S proteasome, is a rapid mechanism for selective protein degradation and plays crucial roles in plant development and stress tolerance [14,15]. The E3 ligase genes are stress-responsive, resulting in the altered ubiquitination and degradation of target proteins [36,49]. With more than 500 predicted members, the RING E3s constitute the largest ligases family [33]. The RING E3s are controlled by transcriptional regulation and post-translational modifications [17]. Here, we demonstrated that stress-responsive AS acted as a post-transcriptional regulation modulating the function of E3 ligases in two ways. First, the truncated isoform losing RING domain could not degrade the target protein, instead, it played the opposite role by interacting and protecting the target protein. We propose that the truncated isoform competes with the full length E3 through the same binding site. Second, AS changed the ratio of two transcripts on demand in response to environmental stresses. Immunoblotting analysis showed that the protein levels of SRAS1.1 and SRAS1.2 remained steady under salt stress (S8A and S8B Fig). We propose that post-transcriptional regulation is a major factor affecting the expression of SRAS1.1 and SRAS1.2.
Notably, the recent studies showed that several truncated isoforms generated by AS lost key domains and exhibited diverse and powerful functions under stresses [50][51][52]. For example, a truncated splice form produced by heat-induced AS activates the HsfA2 promoter and positively autoregulates its own transcription [28]. HAB1.2, a truncated HAB1 PP2C isoform that lacks all phosphatase activity, is unable to transduce the stress signal via the ABA pathway [32]. However, little is known about functional splicing variants produced in response to salt stress. To determine how many salt-responsive AS genes exist, we analyzed and found the transcripts of 237 reported genes strongly responded to salt stress in Arabidopsis. About 28% of them (67 genes) had splicing variants annotated in the TAIR database. 41 SRAS genes were identified including SRAS1 in this study (S2 Table). The functions of AS-triggered truncated proteins will be investigated in our future work.

How salt stress triggered and affected AS
The recent studies showed that AS was a significant regulatory mechanism in response to salt stress [24,26,53,54]. However, exactly how salt stress controls AS isoform ratios and the timing of the mechanism in response to environmental signals remains elusive. In mammals, increasing evidence points to epigenetic and epitranscriptome changes, such as chromatin structure, DNA methylation, histone modifications and transcription elongation rate [55,56]. In plants, the role of epigenetic modifications in regulating transcription rate and mRNA abundance under stress is beginning to emerge [24,56]. Ullah et al. indicated that the chromatin structure was more open in retained introns, suggesting that the open chromatin architecture enhanced the pol II elongation rate, which led to skipping of splice sites [57]. Floral initiator Shk1 kinase binding protein1 (SKB1) associated with chromatin and thereby increases H4R3sme2 (histone 4 arginine 3 symmetric demethylation) levels to confers high salt tolerance by regulating transcription and pre-mRNA splicing [58]. On the other hand, emerging evidence also indicates that splicing factors play key roles in stress-responsive pre-mRNA splicing and AS, unveiling a novel regulatory layer in plant stress tolerance [59]. Genome-wide analysis and genetic results revealed that SKIP, a novel splicing factor, was required for the AS and mRNA maturation of lots of salt tolerance genes [60,61]. SR45a, a member of the conserved SR family, directly interacted with the cap-binding complex to promote salt stress responsive gene regulation [62].
Elucidation of the precise manner in salt-triggered AS will be investigated in the future work.
In this study, an example is provided that salt responsive AS could be a regulatory mechanism modulating the levels of E3 ligases. It would be interesting to investigate how widespread this mechanism is, and the molecular dissection of the AS-generated SRASs. Such studies will provide insights and details into the stress-responsive AS regulation.

Plant material and growth conditions
Arabidopsis (Arabidopsis thaliana) plants used in this study were in the Columbia-0 background. The T-DNA insertion lines sras1-1 (SALK_034426) and csn5a-2 (SALK_027705), were obtained from the Arabidopsis Biological Resource Center (https://abrc.osu.edu/). All the mutants were confirmed by RT-PCR (primers were listed in S3 Table). Seeds were sterilized and then plated on 1/2 MS medium (pH 5.8; Sigma-Aldrich, St. Louis, MO, USA) containing 1.5% agar. After stratification in the dark at 4˚C for 3 days, the plates were transferred to a growth chamber with long-day conditions (LDs; 16-h light/8-h dark cycles) at 22˚C. After 7 days, the seedlings were potted in soil and placed in a growth chamber.

Salt stress treatment
For the germination assays, 30 seeds for each line were sterilized and plated on 1/2 MS medium or 1/2 MS medium containing 200 mM or 100mM NaCl after 3 days of stratification. For each germination assay, biological triplicates were performed. For the fresh weight assay, seedlings were grown on 1/2 MS medium or 1/2 MS medium containing 200 mM NaCl after 14 days, seedlings were collected and weighed. For the root growth, seedlings grown in normal 1/2 MS medium for 3 d were transferred onto 1/2 MS medium with or without 100 mM NaCl and grown for another 7 days.

Plasmid construction and plant transformation
The CDS of SRAS1.1 and SRAS1.2 were amplified by PCR from WT cDNA, then subcloned into the plasmid pBI121 (S4 Table). After verification by sequencing, the binary vector was introduced into Agrobacterium tumefaciens strain GV3101. The strains carrying different constructs were used to transform WT or sras1-1 plants using the floral dip method. For selection of transgenic plants, the T0 seeds were sterilized and germinated on agar medium containing 50 mg/L kanamycin or 25 mg/L hygromycin. T3 homozygous lines were used for further study.

RNA extraction and RT-PCR, qRT-PCR
Total RNA was extracted from 7-day-old seedlings of WT and transgenic plants before and after treatment with 200 mM NaCl or specific organs of WT using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The RT-PCR and qRT-PCR assays were performed as described previously [63]. The UBQ10 reference gene was used as an internal control. Intensity of gel bands were quantified using ImageJ software (http://imagej.nih.gov/ij/).

RNA-seq analysis
Total RNA was isolated with TRIzol (Invitrogen) from 2 week old seedlings of the WT and 35S::SRAS1.1 overexpression lines grown on 1/2 MS medium. The transcriptome analysis was performed by CapitaBio Technology with three biological replicates. Library construction was performed according to illumina standard instructions. Reads were aligned to the Arabidopsis genome using TopHat2. Genes with adjusted P < 0.01 were considered to be differentially expressed. We uploaded the transcriptome data in the National Center for Biotechnology Information Sequence Read Archive (PRJ NA735616).

Subcellular localization
For transient expression in N. benthamiana leaf cells, the SRAS1.1 and SRAS1.2 coding regions and an individual RING domain (175-218 aa) were fused in the frame to the coding region for the C-terminal GFP driven by the CaMV35S promoter. For the larger GFP fusion protein, we constructed plasmids 2×GFP which express 2 copies in tandem of GFP, under the cytomegalovirus (CMV) promoter, then SRAS1.2 coding regions were fused in the frame. The GFP dimer vector was constructed using the method described in references [64]. The constructs were infiltrated into N. benthamiana plant leaves. After 3 days, fluorescence was observed with a Zeiss LSM880 confocal microscope (Zeiss, Germany) at 488 nm. The amphiphilic styryl dye FM4-64 was used as a PM marker.

Y2H assays
To examine SRAS1.1 and SRAS1.2 interaction with CSN5A and CSN5B, the expression constructs were co-transformed into yeast Y2H Gold Yeast Strain cells and transformed cells were selected by growth on SD-Leu-Trp (DDO) medium and SD-L-W-H-Ade (QDO).

In vitro ubiquitination assays
The in vitro ubiquitination assays were performed as described previously [66]. In brief, A 100 ng quantity of wheat (Triticum aestivum) E1, 200 ng of purified E2, 5 mg of Myc-tagged ubiquitin (Boston Biochemicals, Cambridge, MA, USA), 1 mg of purified GST-SRAS1.1, GST-SRAS1.2 and CSN5A-His or CSN5B-His were added to 30 μl of ubiquitination reaction buffer (50 mM Tris-HCl pH 7.5, 2 mM ATP, 5 mM MgCl 2 , 2 mM DTT). After 24 h at 30˚C the reactions were stopped by adding 5×loading buffer, the samples were then boiled at 100˚C for 5 min. The products were electrophoresed on a 15% SDS polyacrylamide gel electrophoresis (PAGE) gel and detected with anti-His (CWBIO, Beijing, China) and anti-Myc (CWBIO, Beijing, China) antibodies by western blotting.