Regulation of Plant Developmental Processes by a Novel Splicing Factor

Serine/arginine-rich (SR) proteins play important roles in constitutive and alternative splicing and other aspects of mRNA metabolism. We have previously isolated a unique plant SR protein (SR45) with atypical domain organization. However, the biological and molecular functions of this novel SR protein are not known. Here, we report biological and molecular functions of this protein. Using an in vitro splicing complementation assay, we showed that SR45 functions as an essential splicing factor. Furthermore, the alternative splicing pattern of transcripts of several other SR genes was altered in a mutant, sr45-1, suggesting that the observed phenotypic abnormalities in sr45-1 are likely due to altered levels of SR protein isoforms, which in turn modulate splicing of other pre-mRNAs. sr45-1 exhibited developmental abnormalities, including delayed flowering, narrow leaves and altered number of petals and stamens. The late flowering phenotype was observed under both long days and short days and was rescued by vernalization. FLC, a key flowering repressor, is up-regulated in sr45-1 demonstrating that SR45 influences the autonomous flowering pathway. Changes in the alternative splicing of SR genes and the phenotypic defects in the mutant were rescued by SR45 cDNA, further confirming that the observed defects in the mutant are due to the lack of SR45. These results indicate that SR45 is a novel plant-specific splicing factor that plays a crucial role in regulating developmental processes.


INTRODUCTION
SR proteins constitute a highly conserved family of structurally and functionally related non-snRNP proteins with multiple roles in pre-mRNA splicing and other aspects of RNA metabolism [1][2][3][4][5]. These proteins have a modular domain structure with one or two N-terminal RNA recognition motifs (RRMs) and a C-terminal arginine/serine-rich (RS) domain. The RRM, which confers RNA-binding specificity, binds to specific regulatory sequences in pre-mRNA, and the RS domain mediates protein-protein and protein-RNA interactions in the splicing machinery [3,6]. An important feature of SR proteins is that any one of them can complement splicing-deficient S100 extracts in splicing of pre-mRNA substrates with consensus splice sites [7]. Based on this property, SR proteins have been called essential or general splicing factors. SR proteins in animals function as essential splicing factors in constitutive pre-mRNA splicing and also regulate alternative splicing (AS) by influencing splice site selection in a concentrationdependent manner [3,7]. During formation of the spliceosomal Ecomplex, ASF (alternative splicing factor)/SF2 (splicing factor 2), one of the SR proteins, recruits U1 snRNP to the 59 splice site by interacting simultaneously with the pre-mRNA and the U1-70K protein [3,6]. SR proteins (e.g., SC35 and ASF/SF2) are also involved in bridging 59 and 39 splice sites by interacting concurrently with U1-70K and U2AF 35 [8]. Furthermore, SR proteins facilitate incorporation of the tri-snRNP complex (U4/U6.U5 snRNP) into the spliceosome and promote base pairing between U2 and U6 snRNA [3,7]. Some animal SR proteins that shuttle between the nucleus and cytoplasm function in mRNA export, mRNA stability and/or translation [1]. These studies underscore the importance of this family of proteins in RNA metabolism. Although, SR proteins show functional redundancy in in vitro splicing assays, in vivo studies with non-plant systems indicate that some SR proteins are redundant whereas others are not [4].
In animals there are 11 SR splicing factors whereas this family of proteins is considerably expanded in plants with 19 in Arabidopsis [9][10][11] and 23 in rice [12,13]. Part of this expansion is attributed to differences in splice site recognition between plants and animals [10]. In support of this notion, several SR proteins have been shown to interact with U1-70K, a U1 snRNP specific protein that plays important roles in constitutive and regulated splicing [1,[14][15][16][17][18]. Some plant SR proteins appear to be orthologs of metazoan SR proteins whereas others are unique to plants with novel structural features. Seven of the 19 Arabidopsis SR proteins have no counterparts in animal systems [10,11].
SR45 was isolated as a U1-70K interacting protein in a yeast two-hybrid screen [17]. Unlike typical SR proteins, which contain a single RS domain at the C-terminus, SR45 has two distinct RS domains, one on either side of the RRM ( Figure 1A). Using a GFP-SR45 fusion protein, we showed that SR45 localizes to nuclear speckles and its mobility is regulated by ATP and transcription [19,20]. However, the molecular and biological functions of SR45 protein are not known. In this study, using an in vitro splicing complementation assay, we show that SR45 is an essential splicing factor. Further, the splicing patterns of several other SR genes is affected in a T-DNA insertion mutant, sr45-1, suggesting that, in addition to its role in constitutive splicing, it also plays a role in regulated splicing. Phenotypic analyses of sr45-1 plants indicate an important role for SR45 in regulating multiple plant-specific developmental processes including plant size, flowering time and morphology of organs. The work reported here is significant because it shows not only that SR45 affects several developmental processes but also that a plant-specific SR protein with novel domain organization is a bonafide SR splicing factor.

SR45 Is Present in Flowering Plants but Not in Algae and Animals
To identify Arabidopsis SR45 homologs in other organism we searched all the available plant, fungal and animal genomes and EST sequence databases (see Materials and Methods ). These searches revealed the presence of SR45 in rice, maize and in many other flowering plants, but not in algae (Chlamydomonas reinhardtii; Cyanidioschyzon merolae), a diatom (Thalassiosira pseudonana) and animals. These results indicate that SR45 has appeared later in evolution only in the flowering plants clade. Interestingly, there are two SR45 genes in rice (accessions AK070420 and AK063761) as opposed to one in Arabidopsis, indicating that SR45 may have undergone duplication after the divergence of monocots and dicots. A comparison of the amino acid sequences of Arabidopsis SR45 with homologs from rice and maize (accession BT016650) showed similar domain organization and considerable sequence conservation (51-80% similarity) ( Figure 1A). The absence of SR45 in animals and algae and its presence in flowering plants suggest that SR45 may have evolved to perform functions that are specific to flowering plants.

SR45 Protein Expressed in Insect Cells Complements
Splicing-Deficient S100 Cell Extract An important property of SR proteins is that each one is sufficient to complement splicing-deficient HeLa cell S100 extract in splicing of at least some pre-mRNA substrates. Because its domain Figure 1. Arabidopsis SR45 is a splicing factor. (A) Alignment of SR45 amino acid sequences from Arabidopsis (AtSR45, At1g16610), rice (RiceSR45_1, Accession AK070420; RiceSR45_2, Accession AK063761) and maize (MaizeSR45, Accession BT016650). The RRM domain is underscored and the RS domains at the N-and C-terminus are indicated by dashed underlines. Identical amino acids are shown by reverse lettering. Dashes indicate gaps in alignment. (B) In vitro splicing of b-globin pre-mRNA in S100 cell extract supplemented with either 25 ng of ASF/SF2 (ASF) or increasing amounts (10, 30 and 90 ng) of purified SR45 (SR45) expressed in insect cells. The positions of pre-mRNA, spliced mRNA and 59 exon are indicated to the right of the blot. Boxes and a line indicate exons and an intron, respectively. doi:10.1371/journal.pone.0000471.g001 structure is different from typical SR proteins, SR45 is not classified as an SR protein in a recent report [21]. Furthermore, animal proteins with similar domain organization such as human Tra2a and Trab cannot substitute for the essential splicing function of SR proteins [22]. To determine if SR45 functions as an essential splicing factor, we expressed SR45 as a His.tag fusion in insects cells and used the purified protein to analyze its splicing activity in S100 extract with a b-globin pre-mRNA substrate. Remarkably, the purified SR45 was able to activate splicing of bglobin pre-mRNA in a concentration-dependent manner and at a level comparable to recombinant human ASF/SF2 ( Figure 1B). These results provide evidence that SR45, despite its unique domain organization, is a bonafide SR protein and functions as an essential splicing factor.

Molecular Characterization of the sr45-1 Mutant
To analyze the in vivo function of SR45, we characterized a T-DNA insertion mutant of SR45 (Figure 2A). Genomic PCR analyses with SR45-specific primers (LP+RP) that flanked the T-DNA insertion site amplified the expected size SR45 product from wild-type (WT) but not from the mutant indicating that SR45 is disrupted ( Figure 2B). In contrast PCR with the T-DNA-specific primer (LBb1) and SR45-specific primer (LP) amplified a product from the mutant but not from WT ( Figure 2B), confirming that the insertion is in SR45. Insertion of T-DNA at a single locus was verified by Southern analysis ( Figure 2C). RT-PCR with SR45specific forward and reverse primers (1F+414R) amplified the expected size product from WT but not from the mutant ( Figure 2D). To test if a truncated transcript upstream of the insertion is produced, PCR was performed with forward primer, 1F and a primer that is complementary to SR45 before the T-DNA (172R). The expected size product was amplified indicating that a truncated SR45 transcript is produced in the mutant ( Figure 2D). However, the level of the truncated transcript is only 8% of the full-length transcript in WT, indicating that it may be unstable. This mutant was named sr45-1 and was used in the experiments reported below.
The Alternative Splicing Pattern of Several Related SR Genes is Altered in the sr45-1 mutant To gain insight into the molecular mechanism of SR45, we analyzed the expression pattern of the alternative transcripts of all 19 SR genes in different organs in sr45-1 and WT. Figure 3 shows that the AS patterns of SR30, RS31, RS31a, SR34 and SR34b were different between sr45-1 and WT. Sequencing of the alternative transcripts showed that, except for SR34b, the smallest transcript codes for a functional protein, whereas the rest code for truncated proteins ( Figure 3). In the vegetative tissues of sr45-1, the amounts of SR30 mRNA isoform 1 and 3 diminished to negligible levels, whereas the level of isoform 5 increased by approximately 2 fold (Figure 3, SR30). In inflorescence, isoform 1 and 3 showed reduced abundance relative to isoform 5. Similarly, in sr45-1, the smaller size transcripts of RS31 (isoform 1 and 2) and RS31a (isoform 1, 3 and 4) were also diminished. In some cases, this reduction was accompanied by a moderate increase in the longer size products in some genes (e.g. SR30). In the case of SR34, there was an overall decrease in all transcripts in sr45-1. The abundance of the two smaller transcripts (isoform 1 and 3) in all tissues decreased. In stems and leaves, but not in roots and inflorescence, this reduction was accompanied by an increase in isoform 5. The abundance of the longest isoforms significantly diminished but this may have resulted due to an overall decrease in transcript level in sr45-1 relative to WT. In all tissues, the abundance of the intermediate size transcript (isoform 6) of SR34b, which codes for fulllength protein, remained similar between sr45-1 and WT. In vegetative tissues, the abundance of the longer transcripts (isoform 7 and 8) decreased and the smaller size transcripts (isofrom 1 and 2) increased in sr45-1. In inflorescence of sr45-1, however, isoform 7 and 8 decreased but there was no significant change in the levels of isoform 1 and 2. The pattern of alternative transcripts of the rest of the SR genes, except for very subtle changes in a few transcripts, remained similar between sr45-1 and WT (RSZ32 and data not shown). Sequence analyses of the affected transcripts show that the longest intron is either partially retained or excluded in the alternative transcripts. Since mutation in SR45 caused a decrease in the smaller size products and an increase in the longer size product, SR45 likely functions by influencing splicing of the longest intron to maintain a balance in the relative abundance of alternative transcripts. Overall these analyses indicate that SR45 affects the AS of a subset of transcripts of SR protein-encoding genes, which in turn may regulate the splicing of other transcripts and eventually control the different phenotypes observed in sr45-1plants.

sr45-1 Plants Show Delayed Flowering and Altered Leaf Morphology
For phenotypic characterization, sr45-1 and WT plants were grown under identical conditions. At the seedling stage, the leaves of the sr45-1 plants were elongated and curly ( Figure 4A and B). In general sr45-1 plants were smaller (about one-third) than WT throughout their life cycle ( Figure 4A-D). The mutant showed delay in transition to reproductive phase ( Figure 4C and D). Root growth was also slower in the mutant as compared to the WT RT-PCR expression analyses of SR45 in WT and sr45-1 plants using SR45-specific primers (1F and 414R, which will amplify full-length SR45; 1F and 172R will amplify a truncated SR45 transcript before the T-DNA insertion). Cyc, cyclophilin product was amplified to show equal amount of cDNA template in PCR. PCR product in WT and sr45-1 was normalized to cyclophilin expression in WT and sr45-1, respectively. Numbers below each panel indicate the level of transcript in WT and sr45-1 plants. The level of SR45 transcript in WT is considered as 1. doi:10.1371/journal.pone.0000471.g002

sr45-1 Plants Flowered Later than WT in Both Long-day and Short-day Photoperiods
Genetic analysis of flowering time in Arabidopsis has uncovered four major flowering pathways-the photoperiodic, the vernalization, the autonomous and the hormonal pathways [23,24]. To determine which of these pathways is affected in sr45-1, we quantified flowering time under short-day (SD, 8 h:16 h light:dark) and long-day (LD, 16 h:8 h light:dark) conditions. Flowering time was measured as days to bolting and number of rosette leaves at the first appearance of a flower. Under LD a majority of WT plants flowered around 29 days after germination ( Figure 5A). In contrast, sr45-1 plants flowered around 40 days, reflecting, on average, 11 days of delay in flowering compared to the WT ( Figure 5A). Similarly, at the appearance of the first flower, sr45-1 had 17 leaves whereas WT had only 12 leaves. Under SD for sr45-1, the average number of days to bolting was 102, which was about twice the time required for WT to start bolting ( Figure 5A). Likewise, the number of leaves at flowering was 97 in sr45-1 as compared to 36 in WT. Flowering time under a 12-h photoperiod was very similar to LD ( Figure 5A). Together, these data indicate that sr45-1 exhibits delayed flowering under both LD and SD, although the delay is more prominent under SD.
The Late Flowering Phenotype of sr45-1 Was Rescued by Vernalization but Was Insensitive to Photoperiod, Placing it in The Autonomous Pathway of Flowering To determine the effect of vernalization on the flowering of sr45-1, we vernalized sr45-1 and WT seeds for 40 days and grew them under LD and SD conditions. Under LD, in WT vernalization marginally decreased days to bolting by one day and number of leaves at flowering from 12 to 10. On the other hand, vernalization of sr45-1 reduced the number of days to bolting and leaves at flowering close to WT ( Figure 5B). Similarly under  SD, vernalization also reverted flowering time closer to WT. However, unlike under LD, vernalization was not fully effective. Vernalized sr45-1 plants flowered 31 days earlier than unvernalized plants, however, they were still 34 days later in flowering than vernalized WT plants, indicating a slight interaction of photoperiod with vernalization on flowering. In summary, the SR45-dependent late flowering phenotype was rescued by vernalization, demonstrating that SR45 affects the autonomous pathway.
To gain insight into the molecular effects of SR45 on flowering, we performed microarray experiments consisting of three biological replicates of sr45-1 and WT using the Affymetrix Arabidopsis 22K gene chip and analyzed the data for flowering time genes (Table S1 and Figure S1). One of the prominent genes that displayed increased expression in sr45-1 was the FLOWERING LOCUS C (FLC) ( Figure 5C, Figure S1B and Table S1), which is a MADS-box transcription factor and was originally identified as a mutation that was early flowering under LD [25,26]. The FLC gene is a convergence point for several flowering pathways [23,24]. Close to a dozen genetic loci that affect the expression of FLC have been characterized phenotypically and at the molecular level [23,24]. Seven of these genes, FCA [27], FY [28], FVE [29], FLD [30], LD [31], FLK [32,33] and FPA [34] belong to the autonomous pathway and three (VRN1, VRN2, and VIN3) belong to the vernalization pathway ( Figure 5D). Expression analyses of these genes in the microarray expriment combined with verification by RT-PCR showed that these genes were not differentially expressed in sr45-1 as compared to WT ( Figure 5D and Figure S1 and Table S1). VRN2 [35] and FCA, which is known to regulate flowering by producing multiple transcripts [27], did not reveal any distinguishable alteration in their expression level or AS pattern between sr45-1 and WT ( Figure 5C and Figure S1). These analyses, however, do not rule out the possibility that subtle changes in the AS, limited to only a few nucleotides and indistinguishable by RT-PCR, are affected by SR45. Similarly, the AS pattern of FLC gene was also unchanged in sr45-1. FLC represses the downstream target genes SOC1 [36] and FT [37]. RT-PCR analyses showed no differential expression of FT, whereas there was some repression of SOC1 in sr45-1 compared to WT ( Figure 5C and D), suggesting that the elevated level of FLC in sr45-1 is causing late-flowering phenotype by affecting the activity of some other downstream components. Microarray and RT-PCR analyses with the CONSTANS (CO) gene (photoperiod pathway) [38], showed no alteration of expression in sr45-1 indicating that it does not affect the photoperiod flowering pathway. Similarly, the expression of several genes that positively regulate FLC, such as the members of the PAF1-complex, were also not changed ( Figure S1 and Table S1). Overall, these analyses indicate that SR45 affects flowering in an FLC-dependent manner.

Characterization of Leaf Morphology
To quantify the change in leaf shape, we analyzed the length to width ratio of leaves in WT and sr45-1. Significant differences in the leaf length/width ratio were observed between WT and sr45-1, which were more obvious in juvenile (leaf 1 to 4) than in mature leaves (leaf 5 to 7) ( Figure 6A and B). On average sr45-1 leaves were about two-fold more elongated than in WT. Since floral organs are modified leaves, we also compared their morphology in fully opened flowers of sr45-1 and WT. Like vegetative leaves, petals were also elongated in sr45-1 ( Figure 6C). Together, these data indicate that SR45 is necessary for the normal expansion of organs.
To better characterize the leaf morphology, we compared epidermal cell-shape in the leaves of sr45-1 to WT with scanning electron microscopy (SEM). Compared to WT, the abaxial surfaces of sr45-1 leaves had some pavement cells that were abnormal in shape ( Figure 6D). These cells were significantly more elongated along the length of the leaf and did not display clear indentations and lobes as in WT cells. The presence of these cells probably interferes with normal expansion of leaf cells resulting in the curly phenotype of sr45-1 leaves. Comparison of the adaxial surfaces of sr45-1 to WT revealed two distinct phenotypes. First, several pavement cells of sr45-1 displayed abnormal expansion along all axes. In these cells, the indentations and lobes, although detectable, were less prominent as compared to WT making them look more elongated and arranged length-wise along the length of leaf ( Figure 6D). Second, contrary to regularly spaced stomata in WT leaves, in sr45-1 they were clustered together. To quantify the contribution of cell size and cell number to the elongated phenotype of leaves, we compared the number of pavement cells and stomata in sr45-1 to WT. Stomata were approximately 30% more in sr45-1 than in WT ( Figure 6E). The number of pavement cells per unit area in sr45-1 was slightly more than in WT (65/mm 2 in sr45-1 to 61/mm 2 in WT), however this difference was statistically not significant (p = 0.55) ( Figure 6E). The fact that there are more stomata in the mutant suggests that cell division may have been affected early in leaf development. Cell size and the orientation of expansion on the contrary were dramatically changed in sr45-1. Although, the means of pavement cells size of sr45-1 and WT were not significantly different (p = 0.3), visual examination revealed obvious differences in the distribution of their cell sizes. Quantification of the cell size data is displayed as box plots in Figure 6F, which show that the cell size distribution in sr45-1 is more varied than in WT (size range was 6347 mm 2 in sr45-1 compared to 4199 mm 2 in WT). Additionally, the scatter of cell sizes in the 4 th quartile was about two-fold greater for sr45-1 than for WT. This shows that a substantial proportion of sr45-1 cells were bigger than the biggest cells in WT. Together, these data indicate that the elongated leaf shape phenotype is probably caused by abnormal cell expansion and by subtle changes in cell division.

Flowers of the sr45-1 Plants Have Abnormal Petal and Stamen Numbers
The mutant plants grown under short day conditions had higher occurrence of flowers with altered number of petals and stamens ( Figure 7). This phenotype was more prominent in the early flowers and was rarely observed in later flowers. This phenotype is different from the well-studied phenotype of mutants of floral homeotic genes, where whole whorls of floral organs change from one type to another. Instead, sr45-1 flowers have all four whorls but the number of petals and stamens varied from 3 to 8 in several combinations ( Figure 7B-I). The number of sepals and carpels remained unchanged. In the same inflorescence several flowers had the WT organ numbers ( Figure 7B and C). However, most of the inflorescences had at least one flower with altered petal and stamen numbers. Overall, of the 67 flowers counted, 11 had the abnormal petal and stamen number, which corresponds to a penetrance of approximately 15%. SR45 thus partially affects the petal and stamen number in a development-specific manner.

Full-length SR45 Rescued sr45-1
To demonstrate that the lack of SR45 was responsible for the sr45-1 phenotype, we complemented the mutant with a GFP-tagged SR45 (GFP-SR45). The GFP-SR45 construct was described earlier and was shown to be functional [20]. These plants exhibited a characteristic speckled pattern of GFP-SR45 in the nuclei indicating that it was expressed and localized correctly ( Figure 8A,  Figure 8A, B and C). Consistent with the flowering time of the rescued GFP-SR45/sr45-1 plants, the FLC levels were also reverted back to WT ( Figure 8D, FLC), further confirming that the mutant phenotypes were due to SR45 mutation. Similarly, RT-PCR analysis of several independent transformants showed the expected size SR45 product ( Figure 8D, SR45). In addition, the splicing pattern of SR30, SR34, SR34b, RS31 and RS31a was also rescued in three independent GFP-SR45/sr45-1 transgenic lines ( Figure 8D).

SR45 Is a Plant-Specific Splicing Factor
SR45 was identified as a U1-70k interacting protein in a yeast twohybrid screen [17] and verified to co-localize with U1-70K and another splicing factor, SR1, in the nucleus as speckles [20], a hallmark feature of all SR proteins both in plants and metazoans [10,39]. These observations, together with its unique domain organization (an N-terminal and a C-terminal RS domain separated by an RRM domain), suggested that SR45 plays a role in splicing (Figure 2). To be classified as a bonafide splicing factor, a protein is usually tested for its ability to complement splicing in a splicing-deficient extract. We showed that SR45 was able to splice a pre-mRNA substrate in a splicing-deficient extract, suggesting that it is a splicing factor ( Figure 1B). This finding is further supported by the fact that a mutation in SR45 led to changes in the splicing pattern of several other SR genes (discussed below).

SR45 Is Involved in Plant-Specific Developmental Processes
Homologs of SR45 are found only in higher plants, which suggests, that SR45 performs a plant-specific function. This notion is supported by the functional characterization of SR45 at the phenotypic and molecular levels. The most prominent plantspecific function affected by SR45 was flowering. The sr45-1 plants were late flowering in both SD and LD conditions, essentially placing it in the autonomous pathway. This was validated by physiological, molecular and genetic analyses. First, the late-flowering phenotype was overcome by vernalization under LD ( Figure 5) consistent with the late-flowering characteristics of other autonomous pathway mutants such as fve, fld, fca, fy, fpa, ld and flk, all of which are late flowering under both SD and LD and are rescued by vernalization [28][29][30][31][32][33][34]. Second, sr45-1 had an elevated level of FLC ( Figure 5C, Figure S1B), whose expression is elevated in all autonomous pathway mutants. Third, expression of CO, which is a photoperiod pathway gene, was not changed in the sr45-1 mutant, indicating that SR45 does not operate in the photoperiod pathway.
RT-PCR analyses showed that a truncated transcript of SR45 is produced in sr45-1 plants, which may code for a truncated version of SR45 that would lack the last one-third of the protein. This may result in a non-functional or a protein with abnormal function ultimately resulting in impaired activity in splicing. However, the level of the truncated transcript in the mutant is very low (about 8% of WT level, Fig. 2D). In support of this conclusion, the heterozygous plants were similar to WT (data not shown). Hence, the observed phenotypes in sr45-1 plants are not likely due to a hypomorphic allele or a dominant negative allele of SR45.
Arabidopsis leaf development requires the establishment of proximodistal, adaxio-abaxio and mediolateral asymmetry, which is a complex process controlled by several genes that are involved in polar cell expansion and differentiation [40,41]. The leaf morphology of sr45-1 bears resemblance to those mutants that are defective in the leaf length-width expansion. Analysis of our microarray and RT-PCR data for several known genes involved in leaf morphology such as ANGUSTIFOLIA (AN), ROT3 [40], AN3 [42], CURLY LEAF (CLF) [43], DRL [44], AS1 [45], AS2 [46] and ATHB13 [47], however, did not reveal a significant change in their expression level between the mutant and WT ( Figure S2 and Table S2). This indicates that these genes are not affected by SR45. Comparison of cell shapes and number in sr45-1 to WT showed that SR45 likely affects both cell expansion and division. Several genes involved in diverse cellular processes affect cell expansion. For example, wall-associated kinase2 (WAK2) antisense plants had smaller cell sizes than control plants [48]. Interestingly, in sr45-1 leaves, which had more expanded cells, the expression of this gene (259560_at) together with several other WAK-like genes such as, WAK1 (259561_at) and WAK3 (259559_at) were moderately (2 to 3-fold over WT) induced (Table S2, and RT-PCR data not shown) suggesting a positive correlation between the expression of these genes with cell expansion. The stomatal density phenotype in sr45-1 showed resemblance to the stomatal density mutants such as tmm [49] sdd [50] yda [51] and a synergistically interacting group of receptor-like kinases, er erl1 erl2 [52]. However, an analysis of the expression of these genes in the microarray data did not reveal significant differences between sr45-1 and WT plants, indicating that SR45 acts on yet to be identified genes for controlling stomatal density. Our work revealed the importance of splicing in development and provides a framework for investigating the role of splicing in the development of leaves and other organs.
In addition to late-flowering phenotypes, sr45-1 plants also had defects in the number of petals and stamens in a significant proportion of plants (Figure 7). The abnormal organ number observed in sr45-1 did not follow the well-studied characteristic ABC model, which specify organ identity in each of the four whorls of a flower rather than their number in each whorl [53]. Therefore, SR45 likely controls the activities of gene(s) involved in specifying the number of floral organs. Altered organ numbers in the Arabidopsis flowers is observed in lesions in several genes, such as WIGGUM/ERA1 [54], ULTRAPETALA [55], PLURIPETALA [56], PIE1 [57], CLV1 [58], CLV2 [59] and CLV3 [60]. None of these genes were, however, differentially expressed ( Figure S2 and Table S2), suggesting that other genes, which may or may not be related to the reported genes, may be affecting the organ numbers in an SR45-dpenedent manner.

RNA Targets of SR45 Include Other SR Genes
SR45 is a splicing factor and therefore would affect a phenotypic outcome by affecting splicing of other genes. An effect on constitutive splicing could lead to the accumulation of unspliced pre-mRNAs, which may encode truncated proteins and/or have reduced stability. Transcripts of plant SR genes themselves are the targets of AS regulation in a tissue-and development-specific manner [61][62][63][64]. To understand the mechanism of SR45 in splicing, we analyzed the AS pattern of all SR genes in sr45-1 and WT plants. Analyses of SR genes showed changes in the AS pattern of several SR genes, which in turn are likely to regulate the splicing of other genes. In affected SR genes, the longest intron was alternatively spliced. In most cases, the general pattern is a reduction in the usage of the distal 39 splice sites that generate a smaller transcript and an increase in the usage of the proximal 39 splice sites in sr45-1 compared to WT. This indicates that the WT SR45 protein favors the usage of the distal 39 AS sites and is responsible for a specific balance of alternative transcripts. Although the effect of depleting a specific plant SR protein on the splicing pattern of other genes has not been reported so far, studies with overexpression of two different types of plant SR proteins, SR30 and RSZ33, resulted in abnormalities in development and morphology, which were accompanied by a marked change in the splicing pattern of SR30 and SR1 [62,64]. These studies indicate that the ratio of the splice variants is critical for normal plant development. SR45 also changed the splicing pattern of transcripts from several other SR genes. This, together with the fact that SR45 complements S100 cell extract, suggests that it is involved in both constitutive and regulated splicing. Based on our observation that the AS pattern of several other SR genes is differentially modulated by SR45, it is likely that the splice variants of SR proteins modulate the splicing and/or other RNA processing activities of genes involved in the developmental processes affected in the sr45-1 plants.

Plant Material and Growth Conditions
Seeds of T-DNA insertion lines in the SR45 (At1g16610) in Columbia (Col) background were obtained from the SALK collection (http://signal.salk.edu/; Salk_004132) at the ABRC.
The T-DNA insert was verified by genomic PCR with SR45specific and the T-DNA-specific primers. WT  uk/InterProScan/) and the presence of RS domain manually. The amino acid sequences of Arabidopsis (At1g16610), rice (AK070420 and AK063761) and maize (translated from BT016650) SR45 proteins were aligned with ClustalX using default parameters [65].

In Vitro Splicing Assay
The coding region of SR45 was amplified from a cDNA clone with forward (59CGGGATCCGCGAAACCAAGTCGTGGC39) and reverse primers (59CCGCTCGAGTTAAGTTTTACGAGGTG-GAG39) containing BamHI and XhoI sites (indicated in bold), respectively. The amplified product was digested with BamHI and XhoI and ligated into pFASTBAC vector and verified by sequencing. The recombinant baculovirus was prepared as recommended by the manufacturer (Invitrogen). Recombinant virus was used to infect Sf21 cells and the recombinant protein fused to His.tag was purified using Ni 2+ affinity chromatography [66]. ASF/SF2, which was also fused to His.tag and expressed in insect cells, was purified essentially as SR45. For in vitro S100 complementation splicing assays, 10, 30 and 90 ng of SR45 was used with the b-globin pre-mRNA as a substrate. Twenty-five ng of His-tagged ASF/SF2 was used as a positive control. Splicing assays were carried out for 2 hours at 30uC. Splicing reactions were deproteinized and precipitated with ethanol. Spliced products were fractionated in a 6% denaturing PAGE and visualized by autoradiography.

Quantification of Leaf and Petal Morphology
Cotyledonary and rosette leaves from 3 week old WT and sr45-1 plants grown on MS plates were excised, arranged in the order of age and photographed together. Length and width of each leaf was determined using Adobe Photoshop and the length/width ratio was calculated. Similarly, fully-opened flowers from soil-grown WT and sr45-1 plants were photographed and the length/width ratio of petals was determined. Data were statistically analyzed for significant differences using the Students' t-test in Microsoft Excel.

Flowering Time Measurement and Vernalization Treatment
For vernalization, seeds were surface-sterilized, suspended in 0.1% Phytagar and stratified for 2 days at 4uC. Seeds were incubated at 22uC for 2 days and transferred back to 4uC for another 40 days. Unvernalized seeds were surface-sterilized and stratified for 2 days at 4uC. Vernalized and unvernalized seeds were germinated and grown under LD (16 h:8 h light:dark) or SD (8 h:16 h light:dark) conditions. Flowering time was measured both as days to bolting and the number of rosette leaves at flowering. Data for 48 to 72 plants were statistically analyzed for significant differences using the Students' t-test in Microsoft Excel.

RT-PCR Analyses of Flowering Genes and Analysis of Alternative Splicing of Pre-mRNAs of SR Proteins
For flowering time, leaf morphology and floral organ number, RT-PCR analyses were performed with cDNA prepared from 2-week old Arabidopsis seedlings with Takara EX Taq TM polymerase and gene specific primers (Sequences of the primers are provided in Table S3). PCR products were quantified with the NIH Image J software. The intensities of PCR products were normalized to cyclophilin. For determining the AS patterns of the 19 Arabidopsis SR genes, RT-PCRs with gene-specific primers were performed with total RNA isolated from root, stem, leaf and inflorescence of five-week old WT and sr45-1 plants. Equal amount of template in each assay was verified with cyclophilin primers. The sequences of SR gene primers, which in most cases corresponded to the first and last exons, are given in Table S3.

Microarray Experiments and Data Analyses
Total RNA was isolated from fifteen-day old seedlings grown on MS plates at 16h:8h light:dark cycle and 22uC according to the Trizol method (Invitrogen). The RNA samples were treated with DNAseI and purified using the Qiagen RNA isolation columns. Complementary RNA synthesis, hybridization to the Affymetrix Arabidopsis Genome ATH1 Array, data acquisition, processing and analyses were done according to the Affymetrix GeneChip Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA). Two additional experiments using similar conditions and GeneChip analyses as above were repeated several months apart as a biological replicate. These three experiments generated three control and three sr45-1 data sets. Each data set was initially analyzed separately with the Affymetrix MAS5.0 software, which yielded log ratios of sr45-1 signal to WT and an associated p-value indicating the significance of the change. For statistical comparison analyses, gene expression data were calculated using the modelbased expression index-perfect match (MBEI-PM) algorithm with invariant set normalization in the dChip software (http://www. dchip.org) [67]. The Gene expression data were analyzed for significant differential expression with a moderated paired t-test in the R/Bioconductor software (http://www.R-project.org). The resultant p-values were corrected for false discovery rate (FDR) using the Benjamini-Hochberg (HB) algorithm in R/Bioconductor [68]. Genes with at least two-fold change with an associated FDRcorrected p-value#0.05 were considered significantly changed in sr45-1 relative to WT.

Scanning electron microscopy
Scanning electron microscopy was conducted as described [69]. Cell sizes were measured with the NIH Image J software in at least three SEM images of WT and sr45-1 leaves. Figure S1 Expression analyses of flowering time genes in WT and sr45-1 plants. (A) RT-PCR was performed with total RNA isolated from two week old plants with gene-specific primers as described in the Experimental procedures. For full name of genes see the legend of Figure S1B below. (B) PCR bands were quantified with NIH Image J software (http://rsb.info.nih.gov/ij/ ). Each PCR product in WT and sr45-1 was normalized to cyclophilin transcript level in WT and sr45-1, respectively. Data shown are the percent of WT, with WT levels adjusted to 100 percent. VRN, The vernalization pathway; Autonomous, The autonomous flowering pathway; PAF1, RNA polymerase II (Pol II) Associated Factor 1-complex; VIP, vernalization independence; Integrator; Floral pathway integrators; Photoperiod, Photoperiod pathway genes; MIG, meristem identity genes.