Introduction

Pre-mRNA splicing is an essential and tightly regulated process in eukaryotic systems. This process requires the removal of introns catalyzed by the spliceosome, consists of five major small nuclear ribonucleoproteins (U1, U2, U4, U5, and U6 snRNPs) and many more non-snRNP protein splicing factors (Jurica and Moore 2003). In mammalian systems, spliceosome assembly proceeds from the E′ complex to the C2 complex; in the spliceosomal E′ complex, U1 snRNP is recruited to the 5′ splice site (SS) and SF1/mBBP (also known as mammalian branchpoint binding protein) specifically recognizes the intron branch point sequence and interacts with the U2 snRNP auxiliary factor heterodimer (U2AF65 and U2AF35). Then, U2 snRNP binds to the polypyrimidine tract and 3′ SS, respectively, converting the E′ complex into the ATP-independent E′ complex (Chen and Manley 2009; Will and Luhrmann 2011).

The branchpoint binding protein (BBP) in yeast (Saccharomyces cerevisiae) and its mammalian ortholog SF1/mBBP is characterized by the presence of an N-terminal maxi-KH domain/KH-QUA-2 region, zinc knuckle (or zinc finger), and C-terminal proline-rich region (Arning et al. 1996; Abovich and Rosbash 1997). The maxi-KH domain of SF1/mBBP specifically recognizes and binds to branch point sequences (Liu et al. 2001), while the zinc knuckle domain has been implicated in RNA binding and raises the overall binding affinity (Berglund et al. 1997, 1998; Garrey et al. 2006). The domain architecture of these SF1 proteins is also largely conserved in plants (Lorkvic and Barta 2002; Wang and Brendel 2006; Schwartz et al. 2008), although plant SF1 proteins contain an addition RRM domain between the zinc finger and the proline-rich region when compared with their yeast and metazoan counterparts.

SF1 protein is important for viability in yeast (S. cerevisiae), roundworm (Caenorhabditis elegans), mice (Mus musculus), and human (Abovich and Rosbash 1997; Shitashige et al. 2007; Tanackovic and Krämer 2005). Depletion of SF1 from human cell lines or yeast compromises their viability and a knockout of sf1 in mice was lethal. Moreover, the heterozygote Sf1+/− mouse has been used to show the effects of changes in the splicing pattern of certain pre-mRNAs (Shitashige et al. 2007). Recently, we identified a homolog of splicing factor SF1 that is essential for plant development in Arabidopsis thaliana (Jang et al. 2014). A homozygous T-DNA mutant of AtSF1, atsf1-2, had several developmental defects including early flowering, dwarfism, and abscisic acid (ABA) hypersensitivity. Moreover, we have shown that AtSF1 is involved in the alternative splicing of pre-mRNA of some genes. These results suggest that AtSF1 is not essential for viability in Arabidopsis, unlike its yeast and metazoan counterparts.

Pre-mRNA splicing requires a large number of RNA binding proteins that possess one or two RNA recognition motifs (RRMs) (Alba and Pages 1998; Lorkovic et al. 2000; Lorkvic and Barta 2002). RRMs with approximately 80–90 amino acids form a four-stranded β-sheet packed against two α-helices (Birney et al. 1993; Maris et al. 2005). The most conserved sequences of the RRM, which are located in β1 and β3, consist of eight and six amino acids, named RNP1 and RNP2, respectively; these sequences are crucial for RNA binding (Bentley and Keene 1991; Birney et al. 1993). In Arabidopsis, 196 of RRM containing proteins have previously been identified through in silico searching for the RRM motif (Lorkvic and Barta 2002). There are some reports that the RRM domain is involved in RNA binding; protein–protein interactions; and protein targeting implicated in all aspects of RNA metabolism including pre-mRNA splicing, polyadenylation, mRNA transport, mRNA stability, and translation (Adam et al. 1986; Sachs et al. 1986; Glisovic et al. 2008; Lorkovic 2009; Tao et al. 2013).

In human SF1 and its yeast counterpart MSL5, the KH-QUA2 region is necessary and sufficient for the recognition of their branch point sequences in vitro (Liu et al. 2001; Jacewicz et al. 2015). Therefore, zinc fingers and other domains are presumably not necessary for branch point recognition in vivo. However, the other domains could have roles in other functions such as nuclear retention in yeast MSL5 (Rutz and Séraphin 2000). The aim of this study was to investigate the role of the RRM domain unique to plant SF1 proteins. As the RRM domain of plant SF1 proteins is well conserved in the plant kingdom, characterization of its function will provide us with useful information on plant splicing mechanisms.

Here, we investigated the role of the RRM domain in AtSF1 by expressing AtSF1 lacking the RRM domain or with a deletion of the RNP1/2 motif under its own promoter in the atsf1-2 mutant line (pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2; pAtSF1 964bp ::AtSF1∆ RNP1/2:GUS). We showed that the RRM domain of AtSF1 was necessary for the full restoration of atsf1-2 mutant phenotypes such as early flowering, dwarfism, and heat stress tolerance. Furthermore, the restored flowering phenotype and heat stress response in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants correlated with changes in flowering time and heat stress-related genes, respectively. Our results suggest that the RRM domain of AtSF1 is involved in pre-mRNA splicing of genes involved in flowering time control and heat stress responses.

Materials and methods

Plant materials and growth conditions

Wild-type or transgenic Arabidopsis plants [ecotype Columbia (Col-0)] were grown in Sunshine Mix 5 (Sun Gro Horticulture, Agawarm, MA, USA) or Murashige and Skoog (MS) medium at 23 °C under LD conditions (16:8 h light:dark photoperiod) at a light intensity of 120 μmol m−2 s−1. atsf1-2 mutants (Jang et al. 2014) were used for plant transformation.

For heat treatments, 1-week-old Arabidopsis seedlings grown on MS plates were covered with aluminum foil to expose the plants to homogeneous heat conditions in darkness. The plates were transferred to a heat chamber set at 45 °C and incubated for 70 min. After heat treatments, the seedlings recovered at 23 °C for 3 days with white light illumination (Lee et al. 2015).

For ABA treatments, seeds were plated on medium containing 1% sucrose. At least 50 seeds per genotype were stratified at 4 °C for 3 days, and the seedlings with green cotyledons were scored after incubation for 6 days at 23 °C (Fujii et al. 2009).

Plasmid construction

To generate the pAtSF1 2.4kb ::AtSF1∆RRM:GUS and pAtSF1 2.4kb ::AtSF1:GUS constructs, the coding sequences of AtSF1 with or without the RRM domain were amplified using specific primers and cloned into the pBI101 vector. An AtSF1 construct without RNPs (pAtSF1 964bp ::AtSF1∆RRM:GUS) was also generated using the same approach. To make the 35S::AtSF1∆RRM:GUS and 35S::AtSF1:GUS constructs, amplified AtSF1 coding sequences with or without the RRM domain were cloned into pBA-Myc vectors harboring the 35S promoter. To produce constructs containing the full-length sequence of AtU2AF65a or AtU2AF65b containing only two RRM domains (RRM2/RRM3), the amplified coding sequences were cloned into the pGEX-5X-1 vector (GE Healthcare, Chicago, USA). The resulting recombinant plasmid was sequenced to verify the absence of PCR errors during amplification. Oligonucleotide sequences used in this study are listed in Supplementary Table S1.

Generation of transgenic plants and measurement of flowering time and leaf area

Transgenic plants were generated using the floral dip method with minor modifications (Weigel and Glazebrook 2006). Agrobacterium tumefaciens strain GV3101 harboring the gene constructs was infiltrated into the atsf1-2 mutant background that showed silencing of NPTII gene for kanamycin resistance. Transgenic seedlings were first selected using kanamycin for pBI101 vector or BASTA for pBA-Myc vector and then verified by PCR-based genotyping. At least 20–30 T1 seedlings were analyzed for each construct.

To score flowering time, the total numbers of rosette and cauline leaves of at least five or six independent transgenic lines (at least 16 individual plants per independent transgenic line) were counted in the T2 or T3 generation. To determine whether the flowering time of the transgenic plants differed significantly compared with the atsf1-2 mutants, the data were analyzed using the SPSS software version 12.0 (IBM SPSS Statistics).

To measure leaf area, the rosette leaf area of at least 16 independent transgenic lines was analyzed using the ImageJ software.

Expression analysis

For RNA expression analysis, total RNA was extracted from whole seedlings using a Trizol reagent (Invitrogen, Carlsbad, CA, USA). Samples for RT-PCR or qRT-PCR were harvested at Zeitgeber time 16, unless stated otherwise, and were frozen immediately in liquid nitrogen before being stored at −80 °C until use. RNA quality was determined with a Nanodrop ND-2000 spectrophotometer (Nanodrop Technologies), and only high-quality RNA samples (A 260/A 230 > 2.0 and A 260/A 280 > 1.8) were used for subsequent experiments. cDNA synthesis was conducted according to the manufacturer’s instructions and 1 µg of RNA was used (Roche Applied Science, Madison, WI, USA). The RT-PCR analysis was performed as described previously (Jang et al. 2014). The qRT-PCR analysis was carried out in 384-well plates with a LightCycler 480 (Roche Applied Science) using Roche SYBR Green Master mixture (Roche Applied Science). One stably expressed gene (PP2AA3) was used as a reference gene. All qRT-PCR experiments were carried out in three biological replicates (independently harvested samples) with three technical replicates each. For determination of relative abundance of transcripts, the detailed procedures were previously described (Lee et al. 2013). Oligonucleotide sequences used in this study are listed in Supplementary Table S1.

For protein expression analysis, whole seedlings were ground to a powder in liquid nitrogen and the powder was suspended in a buffered solution of 50 mM Tris–HCl (pH 8.0), 150 mM NaCl, 10% glycerol, 0.5% Triton X-100, 2 mM phenylmethanesulfonyl fluoride, and complete protease inhibitor cocktail (Roche Applied Science). The protein concentration was determined using Bradford solution (Bio-Rad, Hercules, CA, USA). Subsequently, proteins were separated using 10% SDS-PAGE, transferred onto PVDF membrane (Bio-Rad), as described previously (Kim et al. 2008). PVDF membranes were then probed with mouse monoclonal c-Myc antibody (9E10) diluted to 1:500 (Santa Cruz Biotechnology, Dallas, TX, USA) and subsequently incubated with secondary antibody diluted to 1:2000 (Enzo Life Sciences, UK). The membrane was treated using ECL detection reagent (Innotech, Daejon, Republic of Korea).

For GUS histochemical analysis, the procedure was performed as described by Chen et al. (1998).

In vitro pull-down analysis

In vitro pull-down analyses were conducted as described previously by Jang et al. (2014). Briefly, glutathione S-transferase (GST) fusion recombinant proteins were mixed with in vitro translation products synthesized using the T7 TNT-coupled Transcription/Translation System (Promega, Madison, WI, USA) and the mixtures were gently rotated for 2 h at 4 °C. Subsequently, they were washed three times with the washing buffer and eluted with 10 mM reduced glutathione in 100 mM NaCl and 20 mM Tris–HCl (pH 7.2). Finally, the eluted protein samples were analyzed by 12.5% SDS-PAGE and visualized by autoradiography.

Results

Deletion of the RRM domain from AtSF1 does not affect the stability of AtSF1 or its interaction with AtU2AF65 proteins

To determine whether the RRM domain of AtSF1 is important for AtSF1 activity, we generated transgenic plants overexpressing AtSF1 lacking the RRM domain and fused to a 6x-Myc tag at the N-terminal region in the atsf1-2 mutant background (35S::cMyc:AtSF1∆RRM atsf1-2) (Fig. 1a). As the deletion of a large portion such as the RRM domain may result in mRNA instability or degradation of the truncated protein because of changes in its conformation (Severing et al. 2012), we first analyzed RNA and protein expression of the transgene in 35S::cMyc:AtSF1∆RRM atsf1-2 plants. RT-PCR analysis showed that the full or truncated transgene was normally expressed in 35S::cMyc:AtSF1∆RRM atsf1-2 and 35S::cMyc:AtSF1 atsf1-2 plants (Fig. S1). Western blot analysis confirmed that the overexpression of truncated or full AtSF1 proteins was also successfully detected in these plants (Fig. 1b), indicating that deletion of the RRM domain from AtSF1 does not affect the RNA or protein expression of the introduced transgene. However, the band intensity of AtSF1 signals was weaker in 35S::cMyc:AtSF1∆RRM atsf1-2 plants than in 35S::cMyc:AtSF1 atsf1-2 plants, suggesting that the RRM domain may affect the protein stability of AtSF1 proteins.

Fig. 1
figure 1

Expression and binding affinity of AtSF1∆RRM. a Schematic diagram of the transformation constructs used in this study. The AtSF1 constructs with or without the RRM domain fused to the 6xcMyc tag at the N-terminus (35S::cMyc:AtSF1 and 35S::cMyc:AtSF1∆RRM) were transformed into atsf1-2 mutants. b Western blot analysis using anti-cMyc antibody. The recombinant cMyc-AtSF1 protein was detected in 35S::cMyc:AtSF1∆RRM atsf1-2 and 35S::cMyc:AtSF1 atsf1-2 plants. Ponceau S-stained Rubisco large subunit (rbcL) was used as a loading control. c In vitro interaction between AtSF1∆RRM and AtU2AF65 proteins. Full length of AtU2AF65a (RRM1/2/3) or truncated versions of AtU2AF65b (RRM2/3) proteins were fused to glutathione S-transferase (GST). GST or GST-tagged proteins were incubated with in vitro translated (IVT) AtSF1 proteins (35S-labeled). The bands indicate the eluted AtSF1 proteins visualized by autoradiography. The input lanes contain 10% of the 35S-labeled proteins. Coomassie blue-stained bands indicated by asterisks show the amount and quality of the GST or GST fusion proteins used in this assay

Full size image

The N-terminal ULM motif of SF1 is known to interact with the UHM (RRM3) domain of U2AF65 proteins (Selenko et al. 2003; Webb et al. 2005). Although we expected that truncated AtSF1 lacking the RRM domain would still interact with AtU2AF65, we examined their interaction to exclude the possibility that the truncated form of AtSF1 affected the interaction with AtU2AF65. In vitro pull-down assays revealed that truncated AtSF1 lacking the RRM domain still interacted with AtU2AF65a and AtU2AF65b to a similar extent as complete AtSF1 (Fig. 1c). This result indicates that the binding affinity of full or truncated AtSF1 to AtU2AF65 proteins is similar.

Taken together, these results suggest that truncated AtSF1 lacking the RRM domain could be functional in transgenic plants.

Expression of AtSF1 lacking its RRM domain only partially rescues the effects of the atsf1 mutation on flowering time and dwarfism

Previously, we have shown that a mutation in AtSF1 results in pleiotropic developmental defects including slightly early flowering and dwarfism under normal growth conditions (Jang et al. 2014). To examine whether the RRM domain of AtSF1 is needed to rescue these developmental defects observed in atsf1-2 mutants, we analyzed T3 35S::cMyc:AtSF1∆RRM atsf1-2 plants. As shown in Fig. S2A and Supplementary Table S2, five independent 35S::cMyc:AtSF1∆RRM atsf1-2 plants flowered with approximately 12 leaves under long-day (LD) conditions at 23 °C, indicating that the overexpression of AtSF1∆RRM partially rescued the early flowering phenotype of the atsf1 mutant. Conversely, five independent 35S::cMyc:AtSF1 atsf1-2 plants flowered with approximately 14 leaves. Thus, the flowering of 35S::cMyc:AtSF1 atsf1-2 plants was more similar to that of wild-type plants. Furthermore, both 35S::cMyc:AtSF1∆RRM atsf1-2 and 35S:c:Myc:AtSF1 atsf1-2 plants had normal plant size and inflorescence phyllotaxy, similar to wild-type plants (Fig. S2B and C). These results indicate that the introduction of 35S::cMyc:AtSF1∆RRM partially rescues the flowering defects of atsf1-2 mutants.

As the ectopic expression of AtSF1∆RRM under the control of the 35S promoter may cause potential side effects, we introduced the AtSF1 construct lacking the RRM domain expressed under its own 2.4 kb promoter (pAtSF1 2.4kb ::AtSF1∆RRM:GUS) into atsf1-2 mutants (Fig. 2a) and analyzed five independent pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants in the T3 generation. As shown in Fig. 2b, Fig. S3A, and Supplementary Table S2, the respective pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants partially rescued the early flowering phenotype of atsf1-2 mutants, flowering at the approximately 13-leaf stage under LD conditions at 23 °C. Furthermore, the leaf size of pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants was approximately 650 mm2, smaller than in the wild-type plants, where it was approximately 820 mm2 (Fig. 2c; Fig. S3B). Meanwhile, pAtSF1 2.4kb ::AtSF1 atsf1-2 plants used as a control completely complemented the effect of the atsf1 mutation on flowering time and leaf size. However, the height of pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2, pAtSF1 2.4kb ::AtSF1 atsf1-2, and wild-type plants was the same (Fig. 2d). These results indicate that the expression of AtSF1∆RRM under its own promoter only partially restores some of the developmental defects of atsf1-2 mutants.

Fig. 2
figure 2

Phenotypic analysis of pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 and pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants. a Schematic diagram of the transformation constructs used in this study. The AtSF1 constructs without the RRM domain (pAtSF1 2.4kb ::AtSF1∆RRM:GUS) or RNP1/2 (pAtSF1 964bp ::AtSF1∆RNP1/2:GUS) under its own promoter were transformed into atsf1-2 mutants. b Flowering time, c leaf area, and d plant height of independent pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 transgenic lines (T3 generation) at 23 °C under long-day (LD) conditions. As a control, independent pAtSF1 2.4kb ::AtSF1 atsf1-2 transgenic lines were used. Rosette leaf area was analyzed using the ImageJ software. Error bars indicate the standard error of the mean (SEM) of three biological replicates. Asterisks indicate a significant difference in phenotypes of transgenic plants compared with those of wild-type (Col-0) plants (Student’s t test, *P < 0.05, **P < 0.01). e Flowering time of independent pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 transgenic lines (T2 generation) at 23 °C under LD conditions. As a control, independent pAtSF1 964bp ::AtSF1:GUS atsf1-2 transgenic lines were used. Asterisks indicate a significant difference in phenotype of transgenic plants compared with that of atsf1-2 mutants (upper bracket) or wild-type plants (lower bracket) (Student’s t test, *P < 0.05, **P < 0.01)

Full size image

The deletion of conserved signature sequence motifs (RNP1 and RNP2) within the RRM domain is known to abolish the ability of the RRM domain to bind RNA (Burd and Dreyfuss 1994). Therefore, we also generated transgenic plants in which both RNP1 and RNP2 were deleted from AtSF1 (pAtSF1 964bp ::AtSF1∆RNP1/2:GUS) (Fig. 2a). We used the shorter version of the endogenous AtSF1 promoter (964 bp), because it was also effective for this experiment (unpublished, Wang EJ, and Kim J-K) and was technically easier to work with when making the mutant AtSF1 constructs. In vivo production of recombinant proteins was confirmed by histochemical GUS staining of transgenic plants (Fig. S4A). Six independent pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants showed similar rescued phenotypes (Fig. 2e; Fig. S4B and C); these phenotypes were also observed in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants (Fig. 2b–d; Fig. S3, and Supplementary Table S2). Meanwhile, pAtSF1 964bp ::AtSF1:GUS atsf1-2 plants exhibited fully complemented phenotypes. This result indicates that, like AtSF1∆RRM, the expression of AtSF1∆RNP1/2 also only partially recovers the effect of the atsf1-2 mutation.

Taken together, these results suggest that the RRM domain of AtSF1 is important for the full function of the AtSF1 protein in plant development processes such as flowering time and plant size.

The RRM domain of AtSF1 affects alternative splicing of only specific subsets of transcripts like FLM pre-mRNA

To examine whether the partial restoration of flowering time seen in AtSF1 transgenic plants lacking the entire RRM domain or the RNP1/2 motifs correlated with changes in flowering time gene expression, we performed quantitative reverse transcription PCR (qRT-PCR) analysis of 8-day-old whole seedlings grown at 23 °C. SHORT VEGETATIVE PHASE (SVP) and FLOWERING LOCUS M (FLM) expression decreased significantly in atsf1-2 mutants, whereas their expression in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants was similar to their expression in pAtSF1 964bp ::AtSF1:GUS atsf1-2 and wild-type plants (Fig. 3a). RT-PCR analysis of pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plant also showed similar gene expression pattern (Fig. S6A). However, the expression levels of FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1 (SOC1), which are major downstream targets of FLM and SVP (Lee et al. 2013; Pose et al. 2013), were reduced in atsf1-2 mutants, even though their expression was recovered in the rescued transgenic plants (Fig. S5). Because FT and SOC1 expression could not explain the early flowering phenotype shown in the atsf1-2 mutants (Fig. 2b, e; Fig. S2A, and Supplementary Table S2), we assessed the expression of downstream floral integrator genes such as APETALA1 (AP1) and LEAFY (LFY). As shown in Fig. 3a, LFY expression was increased marginally in atsf1-2 mutants, whereas its expression was similar or slightly higher in the rescued transgenic plants as compared with WT. However, AP1 expression was nearly similar in all examined plants (Fig. S5). These results suggest that the altered expression levels of FLM, SVP, and LFY could explain the flowering phenotype of atsf1-2 and the rescued transgenic plants.

Fig. 3
figure 3

Expression of flowering time genes in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants. a SVP, FLM, and LFY expression in independent pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 lines grown at 23 °C under LD conditions. Total RNA was isolated from 8-day-old seedlings. As a control, independent pAtSF1 964bp ::AtSF1 atsf1-2 transgenic lines were used. The expression levels in wild-type plants were set to 1.0. Error bars indicate SEM of three biological replicates. Asterisks indicate a significant difference in the phenotype of transgenic plants compared with the atsf1-2 mutants (upper bracket) or wild-type plants (lower bracket) (Student’s t test, *P < 0.05, **P < 0.01). b Expression of two FLM splicing forms (FLM-β and FLM-δ) in independent pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 lines

Full size image

We next investigated the expression levels of two FLM splicing forms (FLM-β and FLM-δ) in atsf1-2, pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2, and pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants. In atsf1-2 mutants, FLM-β transcript levels decreased, whereas those of FLM-δ marginally increased (Fig. 3b; Fig. S6B). Interestingly, the level of FLM-β transcripts in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 and pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants increased only marginally when compared with atsf1-2 mutants, whereas FLM-δ transcript levels in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 and pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants were similar to wild-type plants (Fig. 3b; Fig. S6B). This result suggests that the RRM domain of AtSF1 may affect the alternative splicing of FLM to produce the FLM-β splicing isoform.

Because splicing-defective mutants such as protein arginine methyltransferase 5 (prmt5) and spliceosomal timekeeper locus1 (stip1) globally affect the splicing patterns of circadian clock genes (Hong et al. 2010; Jones et al. 2012), we also examined alternative splicing isoforms of LATE ELONGATED HYPOCOTYL (LHY), CIRCADIAN CLOCK ASSOCIATED1 (CCA1), and TIMING OF CAB EXPRESSION1 (TOC1) in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants. As shown in Fig. S6C, the expression levels of major and minor spliced transcripts (arrows) of LHY, CCA1, and TOC1 differed in atsf1-2 mutants compared with wild-type plants. Conversely, the major or minor transcript levels of these genes were restored in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 and pAtSF1 2.4kb ::AtSF1 atsf1-2 plants, indicating that the deletion of the RRM domain in the AtSF1 does not affect alternative splicing of LHY, CCA1, and TOC1.

Taken together, these results suggest that the RRM domain of AtSF1 affects the alternative splicing of only specific subsets of transcripts.

Expression of AtSF1 lacking its RRM domain only marginally restores the hyposensitivity of atsf1 mutants to heat stress

Our previous findings have shown that many heat shock protein-encoding genes are highly expressed in atsf1-2 mutants at 23 °C under LD conditions (Jang et al. 2014). Therefore, to examine whether AtSF1 is involved in modulating the heat stress response, we first tested the heat stress response of atsf1-2 mutants. Under heat stress conditions (45 °C, 70 min), atsf1-2 mutants exhibited increased resistance to heat (78% survival rate), whereas wild-type plants did not (55% survival rate) (Fig. 4a, b), suggesting that the AtSF1 functions as a negative regulator of the heat stress response. Under the same conditions, 35S::FCA and fca-9 plants as controls showed increased and reduced resistance to heat, respectively (Fig. 4a, c) (Lee et al. 2015).

Fig. 4
figure 4

Response of pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants to heat stress treatments. a Survival rate and b, c photographs of 1-week-old transgenic plants under heat stress conditions. The plants grown on MS plates were exposed to heat (45 °C for 70 min) and allowed to recover at 23 °C for 3 days. As a control, independent pAtSF1 2.4kb ::AtSF1 atsf1-2 transgenic lines were used. The phenotypes of 35S::FCA and fca-9 in response to heat stress were consistent with a previous report (Lee et al. 2015). Survival rates of three biological replicates were measured. Asterisks indicate a significant differences in the response of transgenic plants compared with atsf1-2 mutants (upper bracket) or wild-type plants (lower bracket) (Student’s t test, *P < 0.05, **P < 0.01). Error bars indicate SEM of three biological replicates

Full size image

As the AtSF1 and FCA mutant plants have opposite responses to heat stress conditions (Fig. 4) and AtSF1 affects alternative splicing of some pre-mRNAs (Jang et al. 2014), we investigated the expression levels of alternatively spliced FCA transcripts. RT-PCR data revealed that the expression levels of FCA- γ, FCA-β, and FCA-α were not significantly altered in atsf1-2 mutants compared with wild-type plants (Fig. S7). This result suggests that the increased tolerance of atsf1-2 mutants to heat stress is not associated with the alternative splicing patterns of FCA.

To investigate whether the RRM domain of AtSF1 affects the response of atsf1-2 mutants to heat stress, we next treated pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants at 45 °C for 70 min. pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants showed increased resistance to heat, with a comparable survival rate to the atsf1-2 mutants (71% surviving) (Fig. 4a, b). Meanwhile, pAtSF1 2.4kb ::AtSF1 atsf1-2 showed a similar heat response to that of wild-type plants (59% survival rate). This result suggests that the expression of AtSF1∆RRM suppresses the thermotolerance shown in atsf1-2 mutants only marginally, if at all.

The transcript levels of a heat shock transcription factor (HsfA2, AT2G26150) and five heat shock genes that are direct downstream targets of HsfA2 were up-regulated in the atsf1-2 mutant under non-heat stress conditions (Jang et al. 2014). In addition, after heat stress treatment (45 °C for 70 min), the differences in HsfA2 expression between wild-type and atsf1-2 mutants were nearly identical (Fig. S8). Thus, we examined the expression levels of HsfA2 and seven heat shock genes in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants grown at 23 °C under LD conditions. Two HSFA2 transcripts (HsfA2 and HsfA2-II) in wild-type plants were detected, whereas only HsfA2 transcripts known as a functional form of HsfA2 were detected in atsf1-2 mutants (Fig. 5; Fig. S8). In addition, the expression levels of direct targets of HsfA2 were higher in atsf1-2 mutants than those in wild-type plants under the same conditions. These results suggest that already enhanced accumulation of these heat shock response genes may be responsible for increased resistance to heat treatments in atsf1-2 mutants (Fig. 4).

Fig. 5
figure 5

Expression of heat stress-related genes in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants. a Schematic diagrams of two spliced variants (HSFA2 and HSFA2-II) of HsfA2. The letters ‘a’ and ‘b’ adjacent to the gel image indicate HSFA2 and HSFA2-II, respectively. The black boxes, gray box, and lines indicate the exons, alternative exon, and introns, respectively. The arrowheads above the exons indicate the start and stop codons, respectively. The black arrows below the gene structure indicate the positions of the forward and reverse primers. b Expression of HsfA2 and seven HSP gene expressions in independent pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 lines grown at 23 °C under LD conditions. Total RNA was isolated from 8-day-old seedlings. As a control, independent pAtSF1 964bp ::AtSF1 atsf1-2 transgenic lines were used. The numbers below the gel panels represent the average band intensities of three biological replicates. The signal in wild-type plants was set to 1.0. The band intensity was analyzed using the Quantity One program (Biorad). UBQ10 gene was used as an internal control

Full size image

Consistent with the partially rescued phenotype of pAtSF1 2.4kb ::AtSF1 atsf1-2 transgenic plants under heat stress conditions (Fig. 4), RT-PCR data showed that the expression levels of HsfA2, AT4G10250, AT5G12030, AT1G52560, AT5G37670, and AT1G53540 were nearly restored to wild-type levels, whereas the expression levels of AT5G12020 and AT4G25200 were not (Fig. 5b). Of the seven heat shock genes examined, only AT5G12020 and AT4G25200 are not direct targets of HsfA2 (Nishizawa et al. 2006), potentially accounting for their unaltered expression patterns in transgenic plants. Interestingly, alternatively spliced HsfA2 transcripts (HSFA2-II) as a signature of the cytosolic protein response in Arabidopsis (Sugio et al. 2009) were still detected in pAtSF1 964bp :: AtSF1∆RNP1/2:GUS atsf1-2 plants (Fig. 5a, b), suggesting that the RRM domain may affect exon-skipping type alternative splicing.

Taken together, these results suggest that the partially rescued heat response shown in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants is correlated with the expression of HsfA2 and five HsfA2 target genes.

Discussion

AtSF1 contains at least three domains (a KH, a zinc finger, and an RRM); while the conserved domains, like the KH domain, are expected to have similar functions to their counterparts in yeast and metazoan, little research has been conducted on the role of these structural domains. In this study, we provide evidence that AtSF1 functions in the control of flowering time and the heat stress response and that the RRM domain is important for AtSF1 activity in these processes.

We have recently reported that a mutation in AtSF1 leads to slightly early flowering (Jang et al. 2014). Furthermore, several studies have revealed that misexpression of some splicing factors such as serine-arginine 45 (SR45) and PROTEIN ARGININE METHYL TRANSFERASE 5 (PRMT5)/Shk1 kinase binding protein1 (SKB1) alters flowering time (Ali et al. 2007; Zhang et al. 2011). Based on complementation studies and expression analyses (Figs. 2, 3; Figs. S2, S3, and S5, and Supplementary Table S2), we have here shown that the early flowering phenotype of atsf1-2 mutants was associated with up-regulation of LFY, and down-regulation of FLM and SVP. Our results suggest that AtSF1 may control flowering time by regulating LFY expression through an FLM and SVP-dependent pathway. This is consistent with evidence showing that the flowering time of 35S::LFY fve-1 plants was similar to that of 35S::LFY plants (Nilsson et al. 1998), that svp-32 fca-9 or svp-32 fve-3 plants flowered with similar leaf numbers as svp-32 plants (Lee et al. 2007), and that svp-32 flm-3 plants have additive flowering time (Lee et al. 2013). In addition, altered activity in other E′ complex components such as AtU2AF35a or AtU2AF35b results in altered flowering time (Wang and Brendel 2006). Thus, it is probable that AtSF1 regulates the expression of a subset of flowering time genes and thereby modulates flowering time.

Notably, reduced expression levels of FLM-β, but not FLM-δ, were observed in atsf1-2 mutants (Fig. 3b; Fig. S6B), suggesting that AtSF1 could affect alternative splicing of FLM pre-mRNA to produce FLM-β transcripts. The early flowering phenotype of flm-3 mutants (Lee et al. 2013; Pose et al. 2013) is consistent with the phenotype of atsf1-2 mutants (Fig. 2; Fig. S2, and Supplementary Table S2). Our evidence that pAtSF1 964bp :: AtSF1∆RNP1/2:GUS atsf1-2 and pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants had reduced FLM-β expression (Fig. 3b; Fig. S6B) could explain why pAtSF1 964bp :: AtSF1∆RNP1/2:GUS atsf1-2 or pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants could partially recover the effect of the atsf1-2 mutation on flowering time (Fig. 2; Supplementary Table S2). This result implies that the RRM domain of AtSF1 affects the alternative splicing of FLM to produce the alternatively spliced FLM-β isoform. The temperature-dependent alternative splicing of FLM produces antagonistic FLM-β and FLM-δ isoforms that affect ambient temperature-responsive flowering (Pose et al. 2013). Moreover, loss of AtSF1 activity leads to temperature-insensitive flowering phenotype (unpublished, Lee KC, Lee JH, and Kim J-K). Thus, the analysis of the flowering phenotype of the transgenic plants in which the RRM domain of AtSF1 is absent at different temperatures, and FLM-β and FLM-δ expression in atsf1-2 mutant at different temperatures will clarify the role of AtSF1 in ambient temperature-responsive flowering.

A recent report has shown that alternative splicing coupled with nonsense-mediated mRNA decay (AS-NMD) modulates FLM-mediated thermal induction of flowering (Sureshkumar et al. 2016). Moreover, alteration of AtSF1 function affects the patterns of alternative splicing of some specific transcripts, which produces aberrantly spliced transcripts (Jang et al. 2014). These results suggest that alternative splicing of FLM pre-mRNA by AtSF1 may interconnect with AS-NMD regulatory mechanism in the regulation of ambient temperature-responsive flowering. This hypothesis is supported by our observation that atsf1-2 mutants did not respond to changes in ambient temperature, as shown in upframeshift (upf) mutants (Sureshkumar et al. 2016). Furthermore, we observed non-canonical FLM transcripts produced from exon-skipping and intron retention events in atsf1-2 mutants (unpublished, Lee KC, Lee JH, and Kim J-K). Thus, further investigation is required to elucidate the regulatory mechanism of alternative splicing of FLM pre-mRNA through interaction between UPF and AtSF1 in the ambient temperature-responsive flowering.

Leaf size and plant height are two important growth traits regulated by a variety of environmental and genetic factors. In this study, our data showed that atsf1 mutation resulted in the reduction of leaf size and plant height in atsf1-2 mutant (Fig. 2c, d; Fig. S2B and C, Fig. S3, and Fig. S4B and C). In addition, several studies have revealed that leaf size is determined by interconnection between cell division and expansion in Arabidopsis (Gonzalez et al. 2012; Gonzalez and Inzé 2015). Furthermore, plant height is associated with stem elongation mediated by multiple phytohormones including gibberellin (GA), brassinosteroid (BR), and auxin (Wang and Li 2008). It raises the possibility that AtSF1 may control leaf size and plant height by regulating a variety of genes. This notion is supported by our previous microarray data in atsf1-2 mutant (Jang et al. 2014) that down-regulation of the genes involved in cell expansion such as GROWTH-REGULATING FACTOR5 (GRF5), AINTEGUMENTA (ANT), and TARGET OF RAPAMYCIN (TOR) (Vanhaeren et al. 2014), and down-regulation of BR signaling pathway genes such as BRASSINOSTEROID-INSENSITIVE 2 (BIN2) and BRASSINAZOLE-RESISTANT 1 (BZR1) (Zhao et al. 2002; Wang et al. 2002). Moreover, we showed that pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants did not fully recover the effect of the atsf1 mutation on leaf size, but not plant height (Fig. 2c, d; Fig. S2B and C, Fig. S3, and Fig. S4B and C), suggesting that the RRM domain in AtSF1 may play a role in the leaf development.

Heat shock transcription factors are known to be the key players mediating plant responses to highly elevated temperature or under heat shock conditions (von Koskull-Döring et al. 2007). We have recently revealed that mutation of AtSF1 results in the up-regulation of HsfA2 and five heat shock genes that are direct downstream targets of HsfA2 (Jang et al. 2014). These results imply that AtSF1 may mediate the heat response. Consistent with this hypothesis, the results of the current study showed that atsf1-2 mutants exhibited increased tolerance to heat stress, whereas pAtSF1 2.4kb ::AtSF1 atsf1-2 plants showed reduced resistance to heat (Fig. 4). In addition, a functional form of HsfA2 transcripts was significantly increased in atsf1-2 mutants under non heat-stress conditions (Fig. 5; Fig. S8). Moreover, we showed that pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants did not fully suppress the thermotolerance shown in atsf1-2 mutants (Fig. 4), suggesting that the RRM domain in AtSF1 may play a role in the heat stress response. Give that the fact that several heat stress conditions affect alternative splicing of HsfA2, thereby differentially producing alternatively spliced HsfA2 isoforms (HsfA2-II and HsfA2-III) (Sugio et al. 2009; Liu et al. 2013), it is expected that AtSF1 is necessary for the alternative splicing of HsfA2 pre-mRNA under a variety of heat stress conditions. Thus, it will be informative to determine whether AtSF1 functions in wider heat stress treatments by the regulation of alternative splicing of HsfA2.

AtSF1 as a splicing factor negatively regulates heat stress (Fig. 4) and the overexpression of FCA, which is known to produce four different spliced transcripts (Macknight et al. 2002), causes resistant to heat stress (Lee et al. 2015); an important question is, therefore, whether AtSF1 controls thermotolerance by directly modulating the alternative splicing of FCA. The expression levels of three major FCA spliced isoforms were not significantly altered in atsf1-2 mutants (Fig. S7), however, suggesting that AtSF1 may require downstream genes other than FCA for the regulation of the heat stress response. Further investigation into the downstream targets of AtSF1 will provide a better understanding of the modulation of thermotolerance by AtSF1.

If the RRM domain were essential for AtSF1 function, it would be expected that all the phenotypes of atsf1-2 mutants would be restored in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants. However, pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants completely complemented the ABA hypersensitivity shown in atsf1-2 mutants, similar to pAtSF1 2.4kb ::AtSF1 atsf1-2 plants (Fig. S9). Moreover, the reduced expression of CYP707A2, which may be correlated with the ABA-hypersensitive phenotype of the atsf1-2 mutants, was fully recovered in pAtSF1 964bp ::AtSF1∆RNP1/2:GUS atsf1-2 plants (Fig. S10); this is an apparent inconsistency with other phenotypes observed in pAtSF1 2.4kb ::AtSF1∆RRM:GUS atsf1-2 plants (Figs. 2, 4; Figs. S2, S3, S4). These results suggest that AtSF1 protein lacking the RRM domain still has sufficient activity for pre-mRNA splicing of genes important for the ABA response.

The RRM domain functions not only as an RNA-binding motif but also a protein–protein interaction motif (Rain et al. 1998; Thickman et al. 2006; Loerch and Kielkopf 2016). For example, the RRM domain of the yeast U2AF23 (human U2AF35 counterpart) protein works as a frame, such that two zinc finger domains of yeast U2AF23 are arranged side by side on the RRM to bind the 3′ splice site of pre-mRNA (Yoshida et al. 2015).

Further study will provide new insight into the role of the RRM domain of AtSF1 in plant-specific pre-mRNA splicing processes.

Author contribution statement

J-KK conceived and designed the research. KCL conducted experiments. KCL, YHJ, S-KK, H-YP, MPT, JHL, and J-KK analyzed data. KCL, JHL, and J-KK wrote the manuscript. All authors read and approved the manuscript.