SKIP controls flowering time via the alternative splicing of SEF pre-mRNA in Arabidopsis

Background Similar to other eukaryotes, splicing is emerging as an important process affecting development and stress tolerance in plants. Ski-interacting protein (SKIP), a splicing factor, is essential for circadian clock function and abiotic stress tolerance; however, the mechanisms whereby it regulates flowering time are unknown. Results In this study, we found that SKIP is required for the splicing of serrated leaves and early flowering (SEF) pre-messenger RNA (mRNA), which encodes a component of the ATP-dependent SWR1 chromatin remodeling complex (SWR1-C). Defects in the splicing of SEF pre-mRNA reduced H2A.Z enrichment at FLC, MAF4, and MAF5, suppressed the expression of these genes, and produced an early flowering phenotype in skip-1 plants. Conclusions Our findings indicate that SKIP regulates SWR1-C function via alternative splicing to control the floral transition in Arabidopsis thaliana. Electronic supplementary material The online version of this article (doi:10.1186/s12915-017-0422-2) contains supplementary material, which is available to authorized users.


Background
Correct timing of the vegetative to reproductive phase transition, or floral transition, is essential for seed setting in higher plants. Studies have shown that factors involved in the floral transition, including the polymeraseassociated factor (PAF) and FRIGIDA (FRI) complexes, are integrated into such major pathways as the autonomous, vernalization, and photoperiod pathways [1,2]. FLOWERING LOCUS C (FLC) encodes a MCMI, agamous, deficiens and serum response family (MADS)-box transcription factor that acts as a central repressor of the floral transition, and five homologs, MADS AFFECTING FLOWERING (MAF) 1 to MAF5, exist in Arabidopsis (Arabidopsis thaliana) [3][4][5]. The repression of FLC by factors involved in the vernalization and autonomous pathways activates the expression of SUPPRESSOR OF CONSTANS 1 (SOC1) and FLOWERING LOCUS T (FT) to accelerate flowering. Vernalization, the acceleration of flowering by prolonged cold, epigenetically silences FLC through the activity of polycomb repressive complex 2, which deposits the repressive histone mark trimethylation of lysine 27 on histone 3 (H3K27me3) [6]. The autonomous pathway includes a series of activities that promote the epigenetic modification and RNA-mediated chromatin silencing of FLC [7,8]. The antisense transcript of FLC affects the expression of the sense transcript, thereby influencing flowering time in Arabidopsis [9,10].
Precursor (pre)-messenger RNA (mRNA) splicing removes non-coding sequences (introns) from pre-mRNAs and joins the coding sequences (exons) together to generate mature mRNAs; thus, splicing plays important roles in regulating gene expression and increasing protein diversity in eukaryotes [26]. Splicing is catalyzed by the spliceosome, a large flexible RNA-protein complex consisting of five small nuclear ribonucleoprotein particles (snRNPs) and more than 180 types of proteins [27]. In eukaryotes, two distinct spliceosomes catalyze pre-mRNA splicing: the major spliceosome (consisting of the snRNPs U1, U2, U5, and U4/U6) catalyzes the splicing of U2-type introns with 5' splice site (5'ss) GT and 3' splice site (3'ss) AG sequences, while the minor spliceosome (consisting of the snRNPs U11, U12, U4atac, and U6atac) removes U12-type introns with AT-AC (5'ss-3'ss) termini [27,28]. U12-type introns are rare, constituting less than 1% of all introns found in animals and plants [28,29]. Alternative splicing is a process whereby the coding and non-coding fragments of a gene are rearranged in various ways by the spliceosome at different splice sites, giving rise to several mRNA transcripts from a single pre-mRNA. This process regulates transcriptome and proteome diversity, and it controls gene expression and function [30,31]. As in other eukaryotes, splicing is emerging as an important process affecting plant development and tolerance to biotic or abiotic stress [31].
Genomic data show that about 42% of the introncontaining genes involved in development and biotic or abiotic stress responses in Arabidopsis are alternatively spliced [32,33]. Alternative splicing of the tobacco (Nicotiana tabacum) N gene contributes to the control of plant disease resistance [34]. Regulator of C-repeat binding factor (CBF) gene expression 1, a cold-inducible DEAD (Asp-Glu-Ala-Asp) box RNA helicase, plays essential roles in pre-mRNA splicing, especially under cold stress conditions [35]. Regulator of abscisic acid (ABA) response 1/ RNA-binding protein 25, a proline-tryptophan-isoleucine (PWI) motif-and RNA recognition motif (RRM)-containing protein, affects ABA signaling by regulating the splicing of hypersensitive to ABA 1, a protein phosphatase 2C gene [36]. Alternative splicing is also an important mechanism in circadian clock regulation. Recently, Ski-interacting protein (SKIP), spliceosomal timekeeper locus 1, protein arginine methyltransferase 5, and two CIRCADIAN CLOCK-ASSOCIATED 1 (CCA1) transcripts (CCA1α and CCA1β) were shown to regulate the circadian clock through alternative splicing in Arabidopsis [37][38][39][40][41][42]. Root initiation defective 1, a DEAH-box RNA helicase associated with pre-mRNA splicing, plays important roles in meristem maintenance, leaf morphogenesis, and root morphogenesis [43]. In addition, evidence indicates that alternative splicing is involved in flowering time control in Arabidopsis. Alternative splicing of FLOWERING TIME CONTROL LOCUS A (FCA) pre-mRNA is involved in the control of the floral transition [44,45]. RZ-1B and RZ-1C, nuclear-localized RNAbinding proteins, are involved in pre-mRNA splicing and flowering time control via interactions with serine/ arginine-rich (SR) proteins [46]. AtSF1, the homolog of mammalian splicing factor 1 in Arabidopsis, interacts with U2 snRNP auxiliary factor 65a/b (U2AF65a/b) and acts as a splicing factor to regulate flowering time and ABA signaling [47].
SKIP, an SNW domain-containing protein, is evolutionarily conserved, with close homologs in yeast and mammals. In mammals, SKIP (also termed SNW1 and NCoA62) plays important roles as a transcriptional coregulator and splicing factor [48,49]. SKIP specifically regulates the alternative splicing and expression of p21 by interacting with the 3'ss recognition factor U2AF65 and recruiting it to p21 mRNA in vivo [49]. In yeast, the SKIP homolog Prp45 is a component of the Prp19related complex (or nineteen complex), which functions as a splicing factor [50,51]. Weak mutations in prp45 cause defects in the splicing of ACTIN and other genes, resulting in temperature-sensitive growth, while strong prp45 alleles are lethal [50,52]. OsSKIPa, which was able to rescue the defects in prp45 mutant yeast, regulates cell viability and stress tolerance in rice (Oryza sativa) [53]. In Arabidopsis, SKIP regulates cytokinin-associated leaf growth [54]. SKIP expression is induced by salt, mannitol, and ABA. Overexpression of SKIP confers tolerance to abiotic stress. By contrast, the down-regulation of SKIP causes reduced tolerance to abiotic stress during germination. SKIP activated the transcription of a reporter gene in yeast, suggesting that it regulates gene expression as a transcription factor [55]. SKIP is a splicing factor that interacts physically with the plant-specific SR protein SR45 to regulate circadian clock function. The skip-1 mutation disrupts this clock, creating a lengthened clock period phenotype by altering the alternative splicing of PSEUDO-RESPONSE REGULATOR 7 (PRR7) and PRR9, two genes in the morning loop of the oscillator [40]. Genome-wide defects in splicing have been observed in skip-1 plants through RNA high-throughput sequencing, suggesting that SKIP is a splicing factor [40]. In addition to defects in the circadian clock, skip-1 plants show pleiotropic phenotypes, including early flowering. However, the molecular and biochemical mechanisms whereby SKIP represses the floral transition remain obscure.
In this study, we revealed that SKIP plays an essential role in regulating flowering time of Arabidopsis. SKIP was found to regulate the splicing of SEF pre-mRNA and suppress flowering by activating the expression of FLC. FLC expression was reduced in skip-1, leading to an early flowering phenotype under long-day (16 h of light/8 h of darkness, LD) and short-day (8 h of light/16 h of darkness, SD) conditions. SKIP was also required for the normal splicing of SEF pre-mRNA (which encodes a component of the SWR1-C) through direct binding. Splicing defects in SEF were found to contribute to the early flowering phenotype of skip-1. Further, H2A.Z enrichment at FLC chromatin was reduced in skip-1. Our findings indicate a role for alternative splicing in regulating SWR1-C function to control the floral transition in Arabidopsis.

SKIP is required for the floral transition and normal development in Arabidopsis
Previous studies showed that the skip-1 mutation lengthened the period of the circadian clock by impairing the alternative splicing of PRR7 and PRR9 in Arabidopsis [40]. The skip-1 mutation confers an early flowering phenotype under LD and SD conditions [40]; however, the underlying molecular mechanisms are unknown. A complementation test using SKIP genomic DNA and skip-1 plants revealed that the skip-1 mutation is responsible for the observed early flowering phenotype of the mutant (Fig. 1a-d; Additional file 1: Table S1 and Additional file 2: Table S2).
Pleiotropic phenotypes were observed in skip-1 compared to the wild type (WT), including reduced root growth ( Fig. 1e and f; Additional file 3: Table S3); smaller sepals, petals, and flowers ( Fig. 1h-j); shorter stigmas, stamens, and siliques ( Fig. 1k-m). In the present study, we focused on dissecting the mechanisms whereby SKIP controls flowering time in Arabidopsis.

SKIP activates the expression of FLC and its homologs to repress flowering
Because the skip-1 mutation is insensitive to photoperiod, we examined the expression of CONSTANS (CO), a key regulator of photoperiodic flowering, in the skip-1 mutant [56]. CO expression in the skip-1 mutant was reduced or similar to that in WT Columbia (Col-0) plants under LD and SD conditions (Additional file 4: Figure S1a and b). To explore the molecular mechanisms whereby SKIP controls flowering time, we analyzed the expression of FLC and its homologs MAF1 to MAF5, which are major suppressors of flowering, in skip-1. The expression of FLC, MAF1, MAF4, and MAF5 was significantly suppressed by the skip-1 mutation, leading to an early flowering phenotype under LD and SD conditions. Genomic DNA corresponding to the SKIP gene was able to recover the suppression of FLC by the skip-1 mutation (Fig. 2a-d).
Down-regulation of FLC is usually accompanied by the up-regulation of downstream genes to accelerate flowering. SOC1 and FT are known to accelerate the vegetative to reproductive phase transition in Arabidopsis downstream of FLC [2]. Twin sister of FT (TSF), an FT homolog, promotes the floral transition [57,58]. FD, a transcription factor, interacts with FT to promote the floral transition in the shoot apical meristem [59]. Both FT and FD are required for floral meristem formation [59,60]. As expected, the skip-1 mutation activated the expression of SOC1, FT, and TSF, but not of FD, and it promoted flowering under LD and SD conditions ( Fig. 2e-j; Additional file 4: Figure S1c). These results suggest that skip-1 represses the expression of FLC and activates the expression of downstream flowering time integrators (e.g., SOC1, FT, and TSF) to accelerate flowering.
To confirm that FLC repression is the major mechanism that promotes flowering in skip-1, skip-1 plants were transformed with a 35S:FLC construct. Flowering time in the 35S:FLC transgenic lines was similar to that in WT plants because of the overexpression of FLC, indicating that FLC was able to recover the early flowering phenotype of skip-1 ( Fig. 3b and c; Additional file 6: Table S5 and Additional file 7: Table S6).
Together, these genetic and molecular data demonstrate that SKIP suppresses the floral transition in Arabidopsis in an FLC-dependent manner. SKIP promotes FLC expression by controlling the alternative splicing of SEF pre-mRNA We next investigated the mechanisms whereby SKIP regulates FLC expression in detail. There are three possible explanations for the silencing of FLC by skip-1. First, the skip-1 mutation may cause the abnormal splicing of FLC pre-mRNA and decrease the accumulation of functional FLC mRNA. Second, the skip-1 mutation may cause the overexpression of genes in the autonomous or vernalization pathway to reduce FLC expression. Third, the skip-1 mutation may repress the expression of flowering time suppressors, including genes encoding members of the PAF1 complex, the SWR1-C, and so forth.
To explore the silencing mechanisms of FLC, MAF1, MAF4, and MAF5 in the skip-1 mutant, we first tested for defects in the splicing of FLC, MAF1, MAF4, and MAF5 pre-mRNA in the skip-1 mutant; except for MAF1, no splicing defects were detected (Additional file 8: Figure S2a; Additional file 9: Figure S6b, d, and f). Next, we assessed the mRNA expression of genes in both the autonomous and vernalization pathways, including FCA, FY, FLK, FLD, FPA, FVE, LD, VRN1, VRN2, and VIN3 [24]. The expression  Figure S2b and c). Finally, we analyzed the expression of floral suppressors, which were chosen according to a review [24]. Our results revealed aberrant alternative splicing and reduced levels of mature SEF mRNA, which encodes a component of the SWR1-C (Additional file 8: Figure S2d).
The expression and abnormal alternative splicing of SEF caused by the skip-1 mutation were confirmed by reverse transcription-PCR (RT-PCR) and quantitative reverse transcription-PCR (qRT-PCR), respectively ( Fig. 4; Additional file 10: Figure S3). Elevated accumulation of alternatively spliced SEF isoforms, including those showing intron retention, compared to the WT was detected in skip-1 (Fig. 4). However, the level of mature SEF mRNA encoding functional SEF protein was reduced by (See figure on previous page.) Fig. 2 SKIP is required for the transcriptional activation of FLC and its homologs to repress flowering. Ten-day-old seedlings were used to examine flowering time-related gene expression by quantitative reverse transcription (qRT)-PCR. a, c FLC expression in WT plants, skip-1, and three of the skip-1 complemented transgenic lines described in Fig. 1 under LD (a) and SD (c) conditions. b, d MAF expression in WT and skip-1 plants under LD (b) and SD (d) conditions. e, f, h, i SOC1 and FT expression in WT plants, skip-1, and the three complemented transgenic lines under LD (e and f) and SD (h, i) conditions. g, j TSF and FD expression in WT and skip-1 plants under LD (g) and SD (j) conditions. ACTIN 2 (ACT2) was used as an endogenous control. The values are the mean ± s.d. One-way analysis of variance (ANOVA; Tukey's multiple comparison test) was performed for data in a, c, e, f, h, and i, and two-tailed unpaired t test was performed for data in b, d, g, and j. Statistically significant differences are indicated by different lowercase letters (P < 0.05). There are statistically significant differences between all non-identical letters skip-1 ( Fig. 4; Additional file 10: Figure S3). SEF has been characterized as a suppressor of flowering time and an SWR1-C component [20]. Therefore, we hypothesized that the down-regulation of FLC in skip-1 mutants is caused by the abnormal splicing of SEF pre-mRNA.
To test this hypothesis, we generated three SEF complementary DNA (cDNA) constructs: one carrying the fulllength SEF coding sequence (CDS), the second carrying the CDS plus the first intron (WT SEF IR , wtSEF IR ), and the third carrying the CDS plus a mutated version of the first intron (mutated SEF IR , mSEF IR ), driven by the 35S promoter. The 35S:wtSEF IR and 35S:mSEF IR constructs were identical except for the 5'ss and 3'ss in the first intron, which were mutated in mSEF IR to produce a transcript that included the first intron (Fig. 5a). All three constructs were transformed into sef-2 plants [20].
As expected, RT-PCR showed that, unlike wtSEF IR , which was efficiently spliced, mSEF IR plants accumulated unspliced SEF IR transcripts, which mimicked the large transcript in skip-1 mutant plants (Fig. 5b-d).
Transgenic plants expressing full-length SEF cDNA and wtSEF IR , but not mSEF IR , partially rescued the early flowering phenotype of sef-2, indicating that wtSEF IR has a similar function to full-length SEF cDNA. In contrast, the unspliced form (mSEF IR ) could not activate FLC expression and hence failed to delay flowering under LD and SD conditions ( Fig. 5b and e; Fig. 6; Additional file 11: Table S7; Additional file 12: Table S8).
C12-1, a transgenic line expressing a green fluorescent protein (GFP)-SKIP fusion protein (from SKIP:GFP-SKIP), recovered the early flowering phenotype of the skip-1, indicating that SKIP was functional in this line (Fig. 7a). We next used a commercial antibody against GFP to immunoprecipitate GFP-SKIP. The GFP-SKIP immunoprecipitates were then reverse-transcribed into cDNA and amplified by PCR with SEF-specific primers. Four sets of samples were prepared: no immunoprecipitation (Input; positive control), immunoprecipitation without anti-GFP antibodies (NA-IP; negative control), and PCR following RNA-IP against GFP with (RT+) or without (RT-) reverse transcription (Fig. 7b).
SEF pre-mRNA was detected in the Input and RT+ samples from line C12-1 and confirmed by sequencing (Fig. 7b). These results suggest a role for the SKIP-containing spliceosome in the splicing of SEF pre-mRNA.
SKIP activates FLC by regulating the splicing of SEF pre-mRNA by replacing H2A with H2A.Z at FLC chromatin SEF is a component of the SWR1-C, which is involved in the deposition of H2A.Z to target chromatin to activate gene expression [20]. To determine whether the reduced expression of SEF in skip-1 plants influences the deposition of H2A.Z on a global scale, nucleoproteins were extracted from WT, skip-1, sef-2, and arp6-5 plants and Western blot analyses were performed using anti-H2A.Z antibodies. Compared to the WT, the H2A.Z enrichment on a global scale was significantly reduced by skip-1, sef-2, and arp6-5 mutations, indicating that SKIP may be involved in chromatin remodeling through the splicing of SEF in vivo (Fig. 7c). To verify whether the reduced transcript level of SEF in skip-1 affects the deposition of H2A.Z at FLC, we performed a chromatin immunoprecipitation (ChIP) assay to determine the level of H2A.Z across the entire FLC locus (Fig. 7d and e). In the WT, H2A.Z was predominantly enriched around the translation start site and first intron (regions B5, B6, and B9), stop site, and 3' untranslated region (regions B13, B14, and B15), consistent with previous data [21] (Fig. 7d). The abundance of H2A.Z across FLC chromatin was lower in skip-1 than in the WT, except in regions B8, B11, and B12 ( Fig. 7d and e).
To analyze whether the reduced level of SEF in skip-1 was responsible for the down-regulation of MAF4 and MAF5, the abundance of H2A.Z across MAF4 and MAF5 was examined in WT, skip-1, sef-2, and arp6-5 plants. The H2A.Z levels across most regions of MAF4 and MAF5 were obviously decreased in skip-1 ( Fig. 7f and g).
These results indicate that the reduced expression of SEF in skip-1 decreased the deposition of H2A.Z at FLC, MAF4, and MAF5 in vivo to suppress their expression and promote flowering ( Fig. 7d and f ).

Discussion
RNA splicing is an essential posttranscriptional process that controls gene expression and increases transcriptomic and proteomic diversity in eukaryotes. The accuracy and efficiency of pre-mRNA splicing, controlled by cis-acting elements and trans-acting factors, play important roles in regulating the expression and normal cellular function of genes [26]. The identification of splicing factors is a key step in dissecting the posttranscriptional regulation of plant development. SKIP, an SNW domain-containing protein, has been reported to be a splicing factor [40]. In this study, the mechanisms whereby SKIP controls flowering time through the alternative splicing were investigated.

SKIP is essential for normal plant development
According to the present and previous studies, SKIP is a nuclear splicing factor that controls the circadian clock through alternative splicing in Arabidopsis [40,55] (Additional file 13: Figure S4). Mutations in SKIP globally affect pre-mRNA splicing under normal growth conditions [40]. This finding is consistent with the pleiotropic phenotypes of skip-1 observed in this study. The loss-of-function skip-1 mutant confers early flowering phenotype under LD and SD conditions, which is consistent with the previous results [40,64] (Fig. 1a-d, Additional files 1 and 2: Tables S1 and S2). In addition, the skip-1 mutant exhibits multiple developmental defects, including short roots, stamens, pistils, and siliques, and small flowers, sepals, and petals, indicating that SKIP plays essential roles in plant development (Fig. 1e-m; Additional file 3: Table S3). However, the mechanisms of SKIP that regulate these biological processes remain unknown.

SKIP is involved in the transcriptional activation of FLC and its homologs
The mutation in skip-1 suppressed FLC, MAF1, MAF4, and MAF5 expression, accelerated flowering by activating the expression of downstream flowering time integrators (including SOC1, FT, and TSF), and promoted early flowering ( Fig. 2; Additional file 4: Figure S1). Consistent with this, the inactivation of FLC in skip-1 partially recovered the late flowering phenotype of the FLC-activated mutants FRI, fve, and flk ( Fig. 3a; Additional file 5: Table S4). Overexpression of FLC rescued the early flowering phenotype of skip-1 (Fig. 3b and c; Additional file 6: Table S5 and Additional file 7: Table S6). These results provide solid evidence that the suppression of FLC expression is responsible for the early flowering phenotype of skip-1, suggesting that SKIP is required for the activation of FLC expression. However, the results of our pre-mRNA splicing assay and the previous report show that SKIP did not affect the splicing of FLC pre-mRNA, implying that SKIP indirectly regulates FLC expression to control flowering [64] (Additional file 8: Figure S2a).   Fig. 7 SKIP is required for the pre-mRNA splicing of SEF through direct binding, and it regulates the deposition of H2A.Z at FLC, MAF4, and MAF5 chromatin. a Phenotypes of the plant materials used for RNA-IP. b RNA-IP analysis of WT and C12-1 plants. SKIP-associated SEF pre-mRNA was detected by RT-PCR. C12-1 is a skip-1 complemented transgenic line harboring the SKIP:GFP-SKIP construct. ACT2 was used as a negative control. Input, positive control; NA-IP, no antibody was used during immunoprecipitation as a negative control; GFP-IP (RT+) or GFP-IP (RT-), PCR was performed following RNA-IP against GFP with (RT+) or without (RT-) reverse transcription. c Global changes in H2A.Z deposition in WT, skip-1, sef-2, and arp6-5 plants as detected by Western blot using anti-H2A.Z antibodies. Tubulin was used as a loading control. The relative level of H2A.Z to tubulin was quantified by Image J software. One-way analysis of variance (ANOVA; Tukey's multiple comparison test) was performed. Statistically significant differences are indicated by different lowercase letters (P < 0.05). d, f H2A.Z enrichment at FLC (d), MAF4, and MAF5 (f) chromatin in WT, skip-1, sef-2, and arp6-5 plants as measured by ChIP with qPCR using anti-H2A.Z antibodies. The data are the mean ± s.d. (n = 3) in d and f. e, g Diagram of FLC (e), MAF4, and MAF5 (g) with exons indicated as black boxes, introns indicated as gray boxes, and the promoter indicated as a black line. The PCR primer sets used are shown as black lines below the diagram. The names of the primer sets correspond to the numbers on the x-axis of the graphs in d and f. kb kilobases The skip-1 mutation did not impact the levels of mature mRNA produced from genes encoding FLC suppressors and activators of the autonomous and vernalization pathways [24], except for SEF (Additional file 8: Figure S2b-d).
SEF is a component of the SWR1-C, which exchanges histone H2A for H2A.Z, producing variant nucleosomes. The SWR1-C is required for H2A.Z deposition at three loci: FLC, MAF4, and MAF5. The enrichment of H2A.Z deposition at FLC, MAF4, and MAF5 chromatin promotes their transcription and delays flowering [21]. However, the regulatory mechanisms (especially posttranscriptional regulation) affecting the functions of the SWR1-C are unknown. In this study, we found that SKIP bound SEF pre-mRNA and regulated its splicing ( Fig. 4 and Fig. 7b; Additional file 8: Figure S2). Our results show that the skip-1 mutation confers defects in SEF pre-mRNA splicing, leading to a reduced level of mature SEF mRNA and elevated levels of alternatively spliced SEF isoforms due to intron retention ( Fig. 3; Additional file 8: Figure S2d; Additional file 10: Figure S3). These isoforms of SEF do not function as mature SEF mRNAs because they cannot complement the early flowering defects in sef-2. Our findings show that the reduced level of mature SEF mRNA caused by abnormal splicing in skip-1 inactivated the expression of FLC, MAF4, and MAF5 and accelerated flowering. These results are consistent with studies showing that mutations in PIE1, ARP6, and SEF (i.e., SWR1-C component genes) conferred early flowering phenotypes in Arabidopsis by silencing FLC, MAF4, and MAF5 expression [19,20,25,65,66]. We were unable to produce a skip-1/sef-2 double mutant because of lethality, indicating that SKIP and SEF are crucial for normal plant development.
In mammals, SKIP acts as a splicing factor and transcriptional co-regulator by interacting with other proteins [48,49]. That is the case for the ortholog of SKIP in Arabidopsis. SKIP has also been demonstrated to regulate the expression of FLC and MAFs by interacting with ELF7, a component of the PAF1 complex [64]. The expression of FLC, MAF1, and MAF4 in skip-1 was significantly lower than that in sef-2, which suggests that besides SEF, other regulators, such as ELF7, exist in skip-1 to control the expression of FLC, MAF1, and MAF4 (Additional file 14: Figure S5; Additional file 9: Figure S6c and e). These results reveal that SKIP plays dual roles, acting as splicing factor and transcriptional co-activator, in regulating the expression of FLC and MAFs to control flowering time in Arabidopsis.
The expression of MAF1 in the sef-2 mutant is similar to that in the WT, but significantly higher than that in the skip-1 mutant; however, the splicing isoforms of MAF1 in the skip-1 mutant is higher than that in the WT (Additional file 9: Figure S6a and b). These results indicate that SKIP directly regulates the expression of MAF1 through alternative splicing and the PAF1 complex [64].
SKIP regulates the H2A.Z enrichment at the chromatin of FLC and its homologs to delay flowering through the splicing of SEF pre-mRNA Epigenetic modifications, including DNA methylation, histone methylation, acetylation, monoubiquitylation, and chromatin remodeling, of FLC and MAF have been extensively explored [2]. Previous studies showed reduced deposition of H2A.Z at FLC, MAF4, and MAF5 in arp6-1 and pie1-5, causing chromatin configuration changes and the suppression of these genes [21]. In the present study, we found that H2A.Z enrichment at FLC, MAF4, and MAF5 was decreased in sef-2 plants. Similarly, H2A.Z enrichment at FLC, MAF4, and MAF5 was down-regulated by the skip-1 mutation, resulting in the inactivation of their expression and the promotion of flowering in skip-1 plants ( Fig. 7d and f ). Our findings reveal a novel mechanism whereby SKIP controls flowering time through pre-mRNA splicing and posttranscriptional regulation of the SWR1-C via ATP-dependent chromatin remodeling processes (Fig. 8). SKIP is required for the alternative splicing of the SEF pre-mRNA; the splicing of SEF pre-mRNA affects the deposition of H2A.Z at FLC, MAF4, and MAF5 chromatin; H2A.Z deposition activates the expression of these genes to delay flowering in Arabidopsis.
It should be indicated that the levels of mature SEF mRNA and presumably its protein were reduced through alternative splicing but they were not knocked out in skip-1 (e.g., in the sef-2 mutant). It is reasonable to expect that the deposition of H2A.Z in the skip-1 mutant would be intermediate between WT and sef-2; however, the level of H2A.Z at some regions of MAF4 chromatin in skip-1 was lower than that in sef-2 mutant. It is possible that SEF targets substrates with differential affinities or that other unknown factors regulating the deposition of H2A.Z were inactivated in skip-1. This will be confirmed in future experiments.

Conclusions
Our study reveals the regulatory roles of SKIP in flowering time control of Arabidopsis. SKIP is a splicing factor. The skip-1 mutant exhibits an early flowering phenotype, indicating that SKIP is a suppressor of flowering time. SKIP promotes the expression of floral repressors, including FLC, MAF1, MAF4, and MAF5, and delays flowering time through regulating the pre-mRNA splicing of SEF. SKIP binds to the pre-mRNA of SEF to control its splicing and activates its expression. The elevated expression of SEF by SKIP accelerates the substitution of H2A by H2A.Z at FLC, MAF4, and MAF5 chromatin, leading to increased FLC, MAF4, and MAF5 expression and delayed flowering time in Arabidopsis.

Plant materials and growth conditions
All plant materials used in this study were of Arabidopsis ecotype Col-0. Seeds were sterilized and placed on Murashige and Skoog (MS) medium with 0.3% agar and 1% sucrose. After stratification in the dark at 4°C for 2 days, the plates were transferred to white light (70 μmol m -2 s -1 ) in a Percival CU36L5 growth chamber (Percival Scientific, Perry, IA, USA). Plants for flowering time determination were grown under various light-dark photocycles with cool white fluorescent light (100 μmol m -2 s -1 ) at 22°C during the day and 18°C at night.

Root elongation and fresh weight assays
Five-day-old seedlings grown in MS medium were transferred to fresh MS agar medium and allowed to grow for an additional 14 days before being harvested. The root lengths and shoot and root fresh weights were then measured.
To produce the 35S:FLC construct, the 591-bp CDS of FLC was amplified and cloned into pCambia1300 under the control of the 35S constitutive promoter. The resulting construct was transformed into skip-1 plants via the Agrobacterium-mediated floral dip method [67]. Transformants were selected on MS medium containing hygromycin. Single insertion lines were selected based on the segregation of antibiotic resistance. Flowering time and FLC mRNA enrichment were examined in T3 transgenic lines.
To produce 35:SEFc, 35S:wtSEF IR , and 35S:mSEF IR , the 516-bp CDS of SEF (SEFc), the SEF CDS plus the first intron (594 bp; wtSEF IR ), and a mutated version of SEF with the first intron that contained an altered 5'ss and 3'ss (594 bp; mSEF IR ) were amplified and cloned into pCambia1300 under the control of the 35S promoter. Next, the constructs were transformed into skip-1 via the Agrobacterium-mediated floral dip method [67]. Transformants were selected on MS medium containing hygromycin. Single insertion lines were selected based on the segregation of antibiotic resistance. Flowering time and the expression of SEF and FLC were examined in T3 transgenic lines.

Confocal microscopy and subcellular localization analysis
Protoplasts isolated from 3-or 4-week-old Arabidopsis leaves grown under 12 h of light/12 h of darkness as described by Asai et al. [68] and transiently transformed with SKIP:GFP-SKIP (constructed in pCambia1300) and roots from 7-day-old T3 transgenic seedlings harboring SKIP:GFP-SKIP were used for a subcellular localization analysis. Images were collected using a Zeiss LSM 510 Meta confocal laser scanning microscope (Carl Zeiss, Oberkochen, Germany) as described previously [69].

RNA-IP
RNA-IP was conducted as described previously [70,71] with slight modifications. Thirteen-day-old whole seedlings of transgenic complementation lines harboring SKIP:GFP-SKIP in a skip-1 background and grown under LD conditions were harvested to detect the association of SKIP with SEF pre-mRNA. After crosslinking in 1% formaldehyde, pre-immunoprecipitation treatment, and immunoprecipitation with anti-GFP antibodies (Abcam Cat. No. ab290, Bat. No. GR197631-1, RRID: AB_303395), the immunoprecipitation products were eluted with elution buffer. The associated RNAs were quantified by RT-PCR with primer pairs crossing the intron-exon junctions in SEF pre-mRNA after reversal of the crosslinks.

ChIP
ChIP was performed as described by Gendrel et al. [72] using 13-day-old seedlings grown on MS medium under LD conditions. Anti-H2A.Z antibodies were purchased from Abcam (Abcam Cat. No. ab4174, Bat. No. 185934-1, RRID: AB_304345). The primers used for qPCR are provided in Additional file 15: Table S9.