Arabidopsis U2AF65 Regulates Flowering Time and the Growth of Pollen Tubes

During pre-mRNA splicing, U2 small nuclear ribonucleoprotein auxiliary factor 65 (U2AF65) interacts with U2AF35 and splicing factor 1 (SF1), allowing for the recognition of the 3′-splice site by the ternary complex. The functional characterization of U2AF65 homologs has not been performed in Arabidopsis thaliana yet. Here, we show that normal plant development, including floral transition, and male gametophyte development, requires two Arabidopsis U2AF65 isoforms (AtU2AF65a and AtU2AF65b). Loss-of-function mutants of these two isoforms displayed opposite flowering phenotypes: atu2af65a mutants showed late flowering, whereas atu2af65b mutants were characterized by slightly early flowering, as compared to that in the wild-type (Col-0) plants. These abnormal flowering phenotypes were well-correlated with the expression patterns of the flowering time genes such as FLOWERING LOCUS C (FLC) and FLOWERING LOCUS T (FT). However, the two atu2af65 mutants did not display any morphological abnormalities or alterations in abiotic stress tests. Double mutation of the AtU2AF65a and AtU2AF65b genes resulted in non-viable seeds due to defective male gametophyte. In vitro pollen germination test revealed that mutations in both AtU2AF65a and AtU2AF65b genes significantly impaired pollen tube growth. Collectively, our findings suggest that two protein isoforms of AtU2AF65 are differentially involved in regulating flowering time and display a redundant role in pollen tube growth.


INTRODUCTION
In eukaryotic cells, pre-messenger RNA (pre-mRNA) splicing is important for processes such as expression of intron-containing genes, remodeling of protein-protein interaction networks, and regulation of transcript abundance (Nilsen and Graveley, 2010). Spliceosomes, consisting of five small nuclear ribonucleoproteins (snRNPs) and numerous non-snRNPs, mediate these key cellular functions (Wahl et al., 2009). The U2 snRNP auxiliary factor (U2AF) is one such non-snRNP component. In mammalian cells, U2AF is composed of two subunits, U2AF65 and U2AF35, representing the large and the small subunits, respectively. During assembly of the spliceosome, U2AF65 directly binds to the polypyrimidine (Py) tracts upon interaction with SF1 through its third RNA recognition motif (RRM) (Rain et al., 1998;Thickman et al., 2006). Meanwhile, U2AF35 binds to the 3 -splice sites and may also promote the binding of U2AF65 to the Py tract through its interaction with the U2AF65 and serine-arginine (SR)-rich proteins (Zuo and Maniatis, 1996). These molecular interactions are essential for the initial recognition of the intronic 5and 3 -splice sites by the spliceosome. Although plant introns contain neither conserved branch-point sequences nor Py tracts (Luehrsen et al., 1994;Simpson and Filipowicz, 1996), similar interactions found in mammals may also be present in plants. Two U2AF65 homologs purified from wild tobacco (Nicotiana plumbaginifolia) can rescue the in vitro splicing of adenovirus pre-mRNA in the extracts of U2AF factor-deficient HeLa cells (Domon et al., 1998). In addition, the Arabidopsis genome has U2AF35, U2AF65, and SF1 homologs (Wang and Brendel, 2006;Jang et al., 2014), and Arabidopsis U2AF65 (AtU2AF65) and SF1 (AtSF1) proteins interact with their rice homologs (Jang et al., 2014;Park et al., 2017). These results suggest that the interactions among splicing factors during intron 3 -splice site recognition is conserved in plants; however, plant introns do not contain conserved regions such as the branch point sequences and Py tracts at the 3 -splice proximal sites found in mammalians.
Although the detailed splicing mechanism is poorly understood in plants, some components of the plant spliceosome play essential roles in plant growth and development, such as in the flowering time. Mutations in AtU2AF35a or AtU2AF35b result in pleiotropic phenotypes including altered flowering time and abnormal morphological defects in leaves, flowers, and siliques (Wang and Brendel, 2006). A loss of AtSF1 activity also causes pleiotropic developmental defects such as early flowering and hypersensitivity to abscisic acid (ABA) (Jang et al., 2014). Furthermore, these mutations in spliceosomal components result in altered splicing patterns of specific genes in plants. For instance, altered AtU2AF35 expression generates novel splicing isoforms of the FCA pre-mRNA due to the splicing of non-canonical introns (Wang and Brendel, 2006). The mutation of AtSF1 affects the alternative splicing pattern of the heat shock transcription factor HsfA2 (Jang et al., 2014;Lee et al., 2017). These results suggest that lesions in spliceosomal components affect alternative splicing of a specific set of genes. Recent reports have also shown that ambient temperature shifts lead to the alternative splicing of splicing-related genes (Verhage et al., 2017) and the cyclin-dependent kinase G1 (CDKG1) affects the alternative splicing of AtU2AF65a in response to the changes in ambient temperature (Cavallari et al., 2018). These results suggest that the components of the spliceosome may be the target of ambient temperature-induced alternative splicing, thereby coordinating plant growth and development under continuously fluctuating temperature conditions. The mutations in spliceosomal components also result in defects in female-or male-gametophytes. Loss-of-function mutations in genes encoding components of snRNPs such as the U2 snRNP-specific proteins ATROPOS (ATO)/Prp9 (Moll et al., 2008) and SF3b130 (Aki et al., 2011), the U5 snRNP-specific protein GAMETOPHYTIC FACTOR 1 (GFA1)/CLOTHO (CLO)/VAJRA (VAJ)/Snu114 (Moll et al., 2008;Yagi et al., 2009), and the U4/U6 snRNP-specific protein LACHESIS (LIS)/Prp4 (Gross-Hardt et al., 2007;Volz et al., 2012) lead to abolished or impaired genetic transmission through the female gametophyte. Mutations in the genes for the DEAH-box RNA-dependent ATPases Prp22/ROOT INITIATION DEFECTIVE1 (RID1) and Prp16/CLUMSY VEIN (CUV) cause defects in the female and male gametophyte, respectively (Ohtani et al., 2013;Tsugeki et al., 2015). Notably, mutants of ATO, SF3b130, GFA1/CLO/VAJ, and RID1 show less severe defects in male gametophyte development. However, a lesion in the SHOOT REDIFFERENTIATION DEFECTIVE 2 (SRD2) gene, causing defective expression of all the snRNAs, does not result in abnormal transmission through the male gametophyte (Onouchi et al., 2000;Ohtani and Sugiyama, 2005;Onodera et al., 2008). These results suggest that loss-of-activity mutations of spliceosomal components differentially affect male and female gametophyte development.
Although U2AF65 genes have been identified in plants such as Arabidopsis, rice, and tobacco (Gniadkowski et al., 1996;Domon et al., 1998), the function of U2AF65 in plants has not yet been reported. Four U2AF65 homologs are predicted in the Arabidopsis genome and only two of these (hereafter defined as AtU2AF65a and AtU2AF65b) have a U2AF homology motif (UHM), i.e., the third RNA recognition motif (RRM3) (Wang and Brendel, 2006;Jang et al., 2014). Here we show that defects in AtU2AF65a and AtU2AF65b may cause abnormal flowering time and pollen tube growth during the male gametophyte development.

Plant Materials and Growth Conditions
All mutants used in this study were Arabidopsis plants in the Columbia (Col-0) background, except for fy-1 (Ler) (Simpson et al., 2003;Henderson et al., 2005). Various T-DNA alleles of AtU2AF65a (SALK_075828, SALK_128420C, SALK_142561, and SALK_144790) and AtU2AF65b (SALK_055049C and SALK_104128) were obtained from the Arabidopsis Biological Resource Center (ABRC) (Alonso et al., 2003). The T-DNA homozygous lines and the double mutants used in this study were verified by DNA-based PCR genotyping using gene-specific and T-DNA specific primers. The oligonucleotide sequences used for genotyping are provided in Supplementary Table S1

Plasmid Construction
To check the β-glucuronidase (GUS) expression patterns driven under AtU2AF35 and AtU2AF65 promoter, we generated the pAtU2AF35a 0.9 kb ::GUS, pAtU2AF35b 0.7 kb ::GUS, pAtU2AF65a 1.3 kb ::GUS, and pAtU2AF65b 1.5 kb ::GUS constructs. For AtU2AF35a, the genomic region from 876 nucleotides (nt) before the ATG start codon was cloned, as previously described (Wang and Brendel, 2006). For AtU2AF35b, the genomic region from 700 nt before the start codon was newly cloned, because the previous report has shown that the promoter region of different lengths (555 and 982 nt) have similar GUS staining patterns (Wang and Brendel, 2006). For AtU2AF65a, the cloned region is from 1,300 nt before the start codon and includes the 5 -untranslated region (UTR) region of AtU2AF65a, and the 3 -UTR region and part of the exon of At4G36700. For AtU2AF65b, we had tried to clone constructs with the promoter region of different lengths, which begin 1,500 nt before the start codon, because this gene has an intergenic region (approximately 6,800 nt). However, we were not able to detect GUS stainingpositive clones from any of the promoter-GUS constructs. We are going to characterize this unusual trait of the AtU2AF65b promoter in the near future. We amplified these promoter regions of each gene using specific primers and cloned them into the pGWB433 vector (gift from Tsuyoshi Nakagawa at the Research Institute of Molecular Genetics, Shimane University, Japan). To make the pAtU2AF65a 890 bp ::AtU2AF65a construct, the full-length coding sequence (CDS) of AtU2AF65a and the genomic region from 890 nt before the start codon including the 5 -UTR region of AtU2AF65a and the 3 -UTR region of At4G36700 were amplified and cloned into a pCHF3 vector. The reason we used the shorter AtU2AF65a promoter for the complementation test is that we wanted to find out the shorter promoter required for AtU2AF65a's function. The construct with 890 nt promoter construct was our first and single one in order to check its usefulness for the complementation test. To produce the p35S::AtU2AF65b construct, the full-length CDS of AtU2AF65b was amplified and cloned into a pCHF3 vector. The resulting recombinant plasmid was sequenced to verify the absence of PCR errors during amplification. The oligonucleotide sequences used in this study are listed in Supplementary Table S1.

Generation of Transgenic Plants and Measurement of Flowering Time
Transgenic plants were generated using the floral dip method with minor modifications (Weigel and Glazebrook, 2002). The Agrobacterium tumefaciens strain GV3101, harboring the gene constructs, was infiltrated into the AtU2AF65 mutant backgrounds. Transgenic seedlings were first selected for pCHF3 vector content using kanamycin and then verified by PCR-based genotyping. A number of T 1 seedlings between 20 and 30 were analyzed for each construct.
To assess the 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 transgenic line) were counted in the T 2 or T 3 generation (when we checked flowering time in T 3 generation, all experiments were repeated three times). To determine whether the flowering time of the transgenic plants significantly differed from that of the wild-type (Col-0) plants, the data were analyzed using the SPSS software version 12.0 (Michaels, 2009). The error bars indicate standard error of the mean (SEM) of three biological replicates consisting of independently harvested samples. The leaf number ratio (16 • C/23 • C, LNR) under LD conditions was used as an indicator of temperature-responsive flowering (Lee et al., 2007). A hypothetical temperature-sensitive plant produces a different total number of leaves at 23 and 16 • C; thus, its LNR is near 2.0.

Germination Assays Under Hormone Treatment or Abiotic Stress Conditions
The sensitivity of the plants to hormones and abiotic treatments was determined as previously described (Hugouvieux et al., 2001). Seeds were sown on half-strength MS medium containing 1% sucrose, and the MS medium was maintained at 4 • C for 2 days in the dark, after which it was transferred to a normal growth chamber kept at 23 • C under LD conditions at a light intensity of 120 µmol m −2 s −1 . Seed germination was analyzed in MS medium supplemented with various concentrations of ABA, NaCl, and mannitol. The germination rate of the seeds was measured at the indicated days. All experiments were repeated three times.

RNA Expression Analyses
Thirty micrograms of total RNA was used for northern blot analysis as described previously (Jang et al., 2014), and 32 P-labeled AtU2AF65a and AtU2AF65b cDNA, which are approximately 500 nt fragment of CDSs starting from the ATG start codon in both AtU2AF65a and AtU2AF65b, were used as probes due to the unique nucleotide sequences for each gene. For reverse transcription-polymerase chain reaction (RT-PCR) and real-time quantitative PCR (RT-qPCR) analyses, total RNA was extracted from a variety of tissues or from 8day-old whole seedlings of Arabidopsis plants using QIAzol Lysis reagent (Qiagen, Germany). The RNA quality was determined using a Nanodrop ND-2000 spectrophotometer (Nanodrop Technologies, Waltham, MA, United States), and only the high-quality RNA samples (A260/A230 > 2.0 and A260/A280 > 1.8) were used for subsequent experiments. The first-strand complementary DNA (cDNA) was synthesized from 1 µg of total RNA in accordance with the manufacturer's instructions (Invitrogen, United States). The RT-qPCR analysis was carried out in 384-well plates with a LightCycler 480 (Roche Applied Science, United States) using SYBR green. RT-qPCR experiments were carried out using Roche SYBR Green Master mixture (Roche Applied Science, United States). In accordance with the best established RT-qPCR practices (Hong et al., 2010), PP2AA3 was included as a reference gene due to its stable expression in Arabidopsis. All RT-PCR and RT-qPCR experiments were performed in three biological replicates consisting of independently harvested samples with three technical replicates each. The error bars indicate the SEM of three biological replicates consisting of independently harvested samples. The oligonucleotide sequences used for expression analysis are provided in Supplementary Table S1. Changes in gene expression were calculated via the CT method (Ramakers et al., 2003). For GUS expression analysis, GUS staining was carried out as described by Chen et al. (1998).

Characterization of Pollen Grains and Pollen Tubes
Pollen viability was assayed as described by Alexander (1969). For in vitro pollen tube growth assay, the pollen grains from mutant and wild-type flowers were plated onto the surface of agar plates and cultured at 23 • C with a relative humidity of 100% for 3 h and then examined using a light microscope (OLYMPUS model BX-51). The detailed procedure has been previously reported (Jiang et al., 2005). All experiments were repeated three times. To determine whether the proportion data such as pollen morphology and pollen tube length significantly differ between mutants and wild-type (Col-0) plants, the data were analyzed by χ 2 test.

Isolation and Characterization of the Single Mutants of AtU2AF65a and AtU2AF65b
We have recently revealed that AtSF1 and AtU2AF65 interact with each other and that the alteration of AtSF1 activity affects the flowering time (Jang et al., 2014;Lee et al., 2017). Therefore, we analyzed the phenotypes of the T-DNA alleles of AtU2AF65a and AtU2AF65b to study their biological function. We obtained four T-DNA alleles (SALK_128420C, SALK_142561, SALK_075828, and SALK_144790) for AtU2AF65a and two T-DNA alleles (SALK_055049C and SALK_104128) for AtU2AF65b to isolate homozygous T-DNA insertional mutants for each gene. PCR genotyping and sequencing analyses confirmed the T-DNA insertion sites of each T-DNA allele (Figures 1A,B). They were renamed atu2af65a-1 (SALK_128420C), atu2af65a-2 (SALK_142561), atu2af65a-3 (SALK_075828), atu2af65a-4 (SALK_144790), atu2af65b-1 (SALK_055049C), and atu2af65b-2 (SALK_104128), respectively. Northern blot analysis revealed that AtU2AF65a and AtU2AF65b expression did not occur in atu2af65a-4 and atu2af65b-1 mutants, respectively (Figures 1A,B), whereas the other AtU2AF65a and AtU2AF65b T-DNA alleles displayed significantly lower or similar expression than the wild-type (Col-0) plants, suggesting that they were knock-down mutants. We also performed RT-PCR analyses using primer sets that spanned the T-DNA insertion sites in three atu2af65a and two atu2af65b mutant alleles. PCR fragments of the expected size were detected in wild-type (Col-0) plants, but no amplification was observed for the regions spanning the T-DNA insertion sites in these mutant alleles ( Figure 1C). Based on these observations, we used the atu2af65a-4 and atu2af65b-1 knock-out mutants for further studies. As shown in Figure 2, the single atu2af65a and atu2af65b mutants did not show major defects in growth and development, except for flowering. The flowering time of atu2af65a-4 mutants was delayed at 23 • C under LD conditions as compared to that of wild-type (Col-0) plants, whereas that of atu2af65b-1 mutants was slightly accelerated [atu2af65a-4, atu2af65b-1, and wild-type (Col-0) plants = 23.7, 14.0, and 17.2 leaves, respectively)]. These flowering phenotypes of atu2af65a and atu2af65b mutants were consistent with the findings (Verhage et al., 2017;Xiong et al., 2019). Notably, the expression of the AtU2AF65a gene (pAtU2AF65a 890 bp ::AtU2AF65a) in the atu2af65a-4 mutant background rescued the mutant phenotype, whereas AtU2AF65b overexpression (p35S::AtU2AF65b) in the atu2af65b-1 mutant background was not able to restore the wild-type (Col-0) phenotype (Supplementary Figures S1, S2). In order to establish whether different light or temperature conditions affect the flowering time, we further examined the impact on this parameter of plant incubation either at 23 • C under SD or at 16 • C under LD conditions. The flowering time phenotypes of these mutants were not significantly affected by the change in light/temperature conditions and the relative ratios remained substantially unchanged. For instance, atu2af65a-4 plants produced more leaves at 16 • C (37.9 leaves) than at 23 • C (23.2 leaves) (LNR = 1.6). Under same conditions, wild-type (Col-0) plants exhibited a delayed flowering time [16 • C (28.7 leaves) FIGURE 2 | Flowering phenotypes of atu2af65a and atu2af65b mutants under different growth conditions. Plants were grown at 23 • C under long-day (LD) condition (23LD), at 23 • C under short-day (SD) condition (23SD) or at 16 • C under long-day (LD) condition (16LD). The number of total leaves (rosette plus cauline leaves) shown in the bar graphs represents the flowering phenotype of each plant. As a control, the wild-type (Col-0) plants were used. The photos were taken upon atu2af65b mutant flowering. The error bars indicate the standard error of the mean (SEM) of three biological replicates consisting of independent samples. The asterisks denote a significant difference in flowering time between transgenic and wild-type (Col-0) plants (Student's t-test, * P < 0.05, * * P < 0.01). Scale bars 3 cm.
vs. 23 • C (16.1 leaves); LNR = 1.8]. This indicated that the flowering of atu2af65a-4 plants was sensitive to differences in temperature. More specifically, at 23 • C under SD conditions, the leaf counts were 67.3, 45.6, and 54.9 for atu2af65a-4, atu2af65b-1, and wild-type (Col-0) plants, respectively. Furthermore, vernalization treatment showed that both the atu2af65 mutants and wild-type (Col-0) plants responded similarly to this condition (Supplementary Figure S3). However, the atu2af65a-4 and atu2af65b-1 single mutants did not respond to several abiotic stresses including exposure to various concentrations of ABA, NaCl, and mannitol (Supplementary Figure S4). These flowering time data indicated that AtU2AF65a and AtU2AF65b differentially affected the flowering time control in Arabidopsis.

Expression Patterns of AtU2AF65a and AtU2AF65b
We analyzed the expression patterns of endogenous AtU2AF65a and AtU2AF65b by RT-PCR analysis. The AtU2AF65a and AtU2AF65b transcripts were abundantly expressed in a variety of tissues (Figures 3A,B) and their expression levels were very similar in most tissues, such as vegetative tissues and flowers. However, the expression level of AtU2AF65a was approximately twofold higher than that of AtU2AF65b in siliques. These ubiquitous expression patterns were similar to those of other splicing factor genes such as AtSF1 and AtU2AF35 in Arabidopsis (Wang and Brendel, 2006;Jang et al., 2014).
In order to investigate the spatial expression patterns of AtU2AF65a and AtU2AF65b in more detail, we analyzed β-glucuronidase (GUS) activity in transgenic pAtU2AF65a 1.3 kb ::GUS and pAtU2AF65b 1.5 kb ::GUS plants, harboring regions of about 1.3 and 1.5 kb upstream of the ATG of AtU2AF65, respectively. We also made pAtU2AF35a 0.9 kb ::GUS and pAtU2AF35b 0.7 kb ::GUS plants harboring about 0.9 or 0.7 kb regions upstream of the AtU2AF35 ATG, to compare the GUS staining patterns between AtU2AF65 and AtU2AF35. We were able to detect GUS activities in all the transgenic plants except for the  (c)]. Scale bars 100 µm. Note that the GUS expression patterns of pAtU2AF35b 0.7 kb ::GUS were similar to those of pAtU2AF35b 550 bp ::GUS or pAtU2AF35b 982 bp ::GUS, as previously described (Wang and Brendel, 2006). pAtU2AF65b::GUS plants, because our attempt to clone the promoter of the AtU2AF65b gene was not successful. The spatial expression patterns of the GUS activities were very similar in most tissues. Strong GUS activities were detected in all parts of the young seedlings including shoot apex regions (Figures 3Ca,Cb,Da,Db,Ea,Eb). Older leaves also showed intense GUS staining in the pAtU2AF65a 1.3 kb ::GUS and pAtU2AF35b 0.7 kb ::GUS plants (Figures 3Dc,Dd,Ec,Ed). Particularly high GUS expression was observed in some parts of the flowers such as sepals, stamens, anthers, pollen, and stigma (Figures 3Ce,Cf,De,Df,Ee,Ef). Petals showed weak GUS expression. However, GUS activities were not observed in the tissues of the transgenic plants expressing only GUS constructs ( Figure 3F). In general, the GUS staining patterns were consistent with the results of RT-PCR analysis (Figures 3A,B) and with the findings in pAtSF1::GUS and pAtU2AF35::GUS plants (Wang and Brendel, 2006;Jang et al., 2014).

Expression of Flowering Time Genes in atu2af65 Mutants
To examine whether the observed changes in the flowering phenotype of the atu2af65a-4 and atu2af65b-1 mutants were correlated with altered expression of flowering time genes, we performed RT-PCR and RT-qPCR analyses of the 8-dayold whole seedlings grown at 23 • C under LD conditions. The expression of FLOWERING LOCUS C (FLC), SHORT VEGETATIVE PHASE (SVP), and FLOWERING LOCUS Mβ (FLM-β) significantly increased in the atu2af65a-4 mutants, whereas the expression of only FLC was selectively decreased in the atu2af65b-1 mutants ( Figure 4A). However, the expression of FLM-δ as another FLM spliced form was not altered in the atu2af65a-4 mutants. Moreover, the expression levels of FLOWERING LOCUS T (FT), which is a major downstream target of FLC, SVP, and FLM (Lee et al., 2013;Pose et al., 2013), were significantly altered ( Figure 4B). We also observed changes in the spliced or unspliced isoforms of the FLC transcripts in the atu2af65a-4 and atu2af65b-1 mutants (Supplementary Figure S5), indicating that the altered alternative splicing of these FLC transcripts affected the changes of major FLC transcripts in these mutants. The reduced expression levels and altered splicing patterns of FLC observed in atu2af65b-1 mutants were consistent with the recent report (Xiong et al., 2019). However, we did not find significant changes in the alternative splicing of FCA or SVP transcripts (Wang and Brendel, 2006). We next investigated the expression of negative and positive regulators of FLC in atu2af65 mutants. To this end, the transcriptional profiles of FLOWERING CONTROL LOCUS A (FCA), FY, FVE, LUMINIDEPENDENS (LD) (negative FLC regulators) as well as FRIGIDA (FRI) (a positive FLC regulator) were addressed (Ratcliffe et al., 2001). Although minor alterations in FY and FVE expression were observed in the atu2af65a-4 or atu2af65b-1 mutants (Figure 4C), these changes were too modest to explain the altered flowering phenotype shown by the atu2af65 mutants (Figure 2). Interestingly, FRI expression was significantly increased only in the atu2af65a-4 mutants. Because the genetic background of atu2af65a-4 mutants is Col-0, which includes a weak FRI allele (Gazzani et al., 2003), we further checked the expression levels of three other FLC-specific regulators (Choi et al., 2011), i.e., FRIGIDA LIKE1 (FRL1), SUPPRESSOR OF FRIGIDA4 (SUF4), and FLC EXPRESSOR (FLX). Their strong expression also observed in the atu2af65a-4 mutants ( Figure 4D). Collectively, these results suggest that the late flowering phenotype of atu2af65a-4 mutants is likely due to increased FLC activity in an FRI complex-dependent manner, whereas the early flowering phenotype of atu2af65b-1 mutants is likely caused by decreased FLC activity in an FRI complexindependent manner.

The atu2af65 Mutations Are Defective in Male Gametophyte Function
Because the atu2af65a and atu2af65b single mutants did not display noticeable defective phenotypes, except for flowering time (Figure 2), we decided to investigate the phenotype of atu2af65a atu2af65b double mutants. To generate these double mutants, the atu2af65a-4 homozygous mutants, either as male or female parents, were crossed with the atu2af65b-1 homozygous mutants. Genetic crossing was also performed using the atu2af65a-3 homozygous mutants, as a source of knock-down allele. The resulting F 1 progeny [+/atu2af65a-4;+/atu2af65b-1 (65A/65a;65B/65b) or +/atu2af65a-3;+/atu2af65b-1] was subsequently allowed to self-pollinate and the genotypes of the F 2 progeny were determined by PCR-based genotyping.

DISCUSSION
Pre-mRNA splicing plays an essential role in various growth and developmental processes in plants, but as yet little is known about the function of many of the spliceosomal components. Here, we provided evidence that two isoforms of AtU2AF65 are important for flowering and for the growth of pollen tubes during male gametophyte development.
Since plant introns do not contain the conserved branch point sequences and Py tracts at the 3 -splice proximal sites found in mammalians (Wahl et al., 2009), the mechanism of 3 -splice site recognition in plants may differ. Our previous findings, however, showed that AtU2AF65 proteins physically interact with AtU2AF35 and AtSF1 proteins in vitro and in planta (Jang et al., 2014;Park et al., 2017), as already known for yeasts and metazoans. Therefore, the initial step consisting in 3 -splice site recognition may be functionally similar in mammalians and plants, despite the above mentioned structural differences in plant introns. This notion is supported by the finding that two tobacco U2AF65 homologs (NpU2AF65a and NpU2AF65b) can splice adenoviral pre-mRNA in HeLa cell extracts depleted of endogenous U2AF factors (Domon et al., 1998). However, some differences in the mechanism of intron recognition may occur in plants. Firstly, several copies of U2AF genes such as U2AF65 and U2AF35 exist in nearly all higher plant genomes (Wang and Brendel, 2006;Jang et al., 2014). Secondly, plant SF1 homologs have an RRM domain, which is not found in their fungal and metazoan counterparts (Worden et al., 2009). Thirdly, neither the branch point nor the 3 -splice site regions significantly contribute to the interaction between RNA and U2AF65 homologs, as is the case with tobacco (Domon et al., 1998). Finally, we observed differential binding affinities of AtU2AF65a and AtU2AF65b for AtU2AF35 proteins . These differences can account for the diversity and flexibility of molecular interactions between various components of the spliceosome in the initial recognition of the 3 -splice site in plants.
Several lines of evidence suggest that mutations in both AtU2AF65a and AtU2AF65b may result in the changes in flowering time. Our and other groups have previously reported that the AtU2AF65 and AtU2AF35 proteins interact  and both T-DNA mutation in AtU2AF35a as well as RNA interference against AtU2AF35b cause late flowering phenotypes (Wang and Brendel, 2006;Jang et al., 2014). Moreover, we found that the expression patterns of AtU2AF35 and AtU2AF65 were similar (Figure 3; Wang and Brendel, 2006). For instance, RT-PCR analyses showed that AtU2AF35 and AtU2AF65 expression was observed in all vegetative and reproductive tissues; however, the abundance of their transcripts was different in the tissues tested. The GUS assays also revealed that the staining patterns of AtU2AF65a promoter activity were similar to those of AtU2AF35a and AtU2AF35b in all the tissues analyzed; however, their GUS staining intensities were slightly different. However, T-DNA alleles of AtU2AF65a were associated with a late flowering phenotype, whereas T-DNA alleles of AtU2AF65b resulted in early flowering (Figure 2). Due to this partial inconsistency, the phenotype of atu2af65b mutants should be interpreted with caution, especially since the overexpression of AtU2AF65b in the atu2af65b mutant background was not able to rescue the mutant phenotype (Supplementary Figure S2). It is therefore difficult to infer the AtU2AF65b function from the results obtained with the T-DNA mutant alleles. One possible explanation is that AtU2AF65b does not affect the flowering time via a dominant negative effect when it is ectopically overexpressed, because overexpressed AtU2AF65b could interrupt the formation of functional complexes, which does not objectively reflect the flowering processes in wildtype plants. However, we cannot exclude the possibility that AtU2AF65b regulates flowering time under specific conditions such as abiotic or biotic stresses. A recent report showed FIGURE 6 | Pollen morphology in the F 2 progeny containing the atu2af65a or atu2af65b mutant alleles. (A) Pollen morphology from wild-type (Col-0, 65A/65A;65B/65B), atu2af65a-4/atu2af65a-4;+/+ (65a/65a;65B/65B), +/+; atu2af65b-1/atu2af65b-1 (65A/65A;65b/65b), atu2af65a-4/atu2af65a-4;+/atu2af65b-1 (65a/65a;65B/65b), +/atu2af65a-4/atu2af65b-1/atu2af65b-1 (65A/65a;65b/65b) plants. The atu2af65a mutant plants with the (65a/65a;65B/65b) genotype or without atu2af65b mutant alleles (65a/65a;65B/65B) and the AtU2AF65b mutant plants with (65A/65a;65b/65b) genotype or without AtU2AF65a mutant alleles (65A/65A;65b/65b) were selected from the F 2 progeny population. Scale bars 50 µm. (B) Shape of pollen grains shown in panel (A). The shape of pollen was categorized into three groups (normal, short, and burst). The average of three biological replicates is shown in the histogram. Numbers in parenthesis denote the total number of pollen grains in three biological replicates of each genotype used for this analysis (approximately 110 pollen grains was used for each biological replicate). The error bars indicate the SEM of three biological replicates consisting of independently harvested samples. The asterisks denote a significant difference in pollen morphology between mutants and wild-type (Col-0) plants (χ 2 test, * * * P < 0.001).
AtU2AF65a and AtU2AF65b appear to differentially control FLC expression, and this might result in the opposite flowering phenotypes observed. However, considering the expression patterns of flowering time-related genes, including positive and negative regulators of FLC, in the atu2af65a and atu2af65b mutants (Figure 4), and the effect of vernalization on the flowering time in these mutants (Supplementary Figure S3), it is likely that AtU2AF65a and AtU2AF65b control flowering time by modulating FLC expression in FRI complex-dependent andindependent manners, respectively. Similarly, the mutations in CAP-BINDING PROTEIN 80 (CBP80)/ABA HYPERSENSITIVE 1 (ABH1) affect flowering time via FRI-mediated FLC regulation (Bezerra et al., 2004). However, because the genetic background (Col-0) of the atu2af65a mutants includes a weak FRI allele (Gazzani et al., 2003), further investigations are needed to ascertain whether an altered AtU2AF65 function may have an impact on FLC expression in the FRI FLC background. Furthermore, genome-wide RNA-seq analysis in the fri FLC and FRI FLC backgrounds is required to identify the general or specific effect of atu2af65 mutation on splicing regulation in flowering. These information would result in a better understanding of the control exerted by AtU2AF65 on flowering time.
The importance of splicing factors during male gametophyte development has been reported by several research groups (Tsukaya et al., 2013). For instance, the mutations in ATO, SF3b130, GFA1/CLO/VAJ, and RID1 cause defects in male gametophyte development (Moll et al., 2008;Yagi et al., 2009;Ohtani et al., 2013). Since these genes are involved in cell wall synthesis, cytoskeletal functions, metabolism, cellular transport, signaling, and secretion, all important elements for pollen tube growth, one possible scenario is that AtU2AF65 is involved in the pre-mRNA splicing of a specific set of genes for pollen tube development. This notion is supported by previous evidence showing that mutations in genes involved in a number of processes, including potassium transport, membrane trafficking, carbohydrate metabolism, signaling cascades, and cell wall functions, affect pollen germination or pollen tube growth (Mouline et al., 2002;Johnson et al., 2004;Lalanne et al., 2004;Lobstein et al., 2004;Jiang et al., 2005). Notably, as the U2AF65associated protein (UAP56) plays an important role in the nuclear export of mRNAs in mammals (Fleckner et al., 1997;Gatfield et al., 2001), another possible scenario is that AtU2AF65 affects the export of nuclear mRNAs for pollen tube development. This hypothesis is supported by the observation that Arabidopsis UAP56 interacts with mRNA export factors such as the RNA and export factor-binding proteins (REF)/ALY2 and MODIFIER OF SNC1 (MOS11) (Kammel et al., 2013). Therefore, AtU2AF65 may affect the splicing or nuclear export of the pre-mRNAs of a specific set of genes, thereby ensuring normal pollen tube stability and growth during male gametophyte development. However, direct evidence in support of this hypothesis is lacking and further work is needed to definitely establish the role of the AtU2AF65 in male gametophyte development.

AUTHOR CONTRIBUTIONS
J-KK conceived and designed the research. H-YP and HL conducted the experiments. JL and J-KK analyzed the data. JL and J-KK wrote the manuscript. All authors read and approved the manuscript.