To verify the gene structure and check the expression patterns of the atU2AF³⁵a and atU2AF³⁵b genes, specific primers were designed from the 5′- and 3′-UTRs of both genes. Reverse transcription (RT)-PCR was conducted using RNAs extracted from Arabidopsis 7-d seedlings, leaf before flowering (LeafBF), leaf after flowering (LeafAF), meristem after flowering (MeriAF), root after flowering (RootAF), stem, flower, and silique tissues. As shown in Figure 1A, both atU2AF³⁵a and atU2AF³⁵b genes express in all these tissues. No clear intron retention product can be identified for either atU2AF³⁵a or atU2AF³⁵b, indicating that the 5′-UTR intron is spliced efficiently. Intriguingly, atU2AF³⁵a seems to have noncanonical introns in addition to the 5′-UTR intron. Several smaller bands in addition to the main product were observed in the RT-PCR for atU2AF³⁵a. Sequencing results revealed that two additional segments in the 3′-end could be removed from the main transcript. These additional segments are possible introns and named alternative intron (AltIntron) 1 and AltIntron2. AltIntron1 (287 nt) and AltIntron2 (345 nt) overlap with each other. The position of these introns is shown in Figure 1B. Neither of the AltIntrons is canonical; both have a repeat region flanking the intron-exon junction (AltIntron1, AGGAGCA; AltIntron2, AAAAC). Thus their real borders are difficult to determine. Independent RT-PCRs using different transcriptase and RNA preparations confirmed the existence of additional products. Splicing of AltIntron1 and AltIntron2 removes the coding sequences for the C terminus of atU2AF³⁵a protein. The truncated Arabidopsis U2AF³⁵a proteins retain the conserved N-terminal domains and a shortened SR domain. This domain structure is similar to U2AF²⁶, a duplicated copy of U2AF³⁵ in human and mouse (Shepard et al., 2002). Whether the truncated atU2AF³⁵a proteins have physiological roles is unknown.

As pre-mRNA splicing takes place in the nucleus, the atU2AF³⁵a and atU2AF³⁵b gene products should have nuclear localization if they are indeed splicing factors. atU2AF³⁵a and atU2AF³⁵b open reading frame (ORF) sequences were fused in frame downstream of the green fluorescent protein (GFP) coding sequence driven by the CaMV 35S promoter in gateway vector pMDC43 (Curtis and Grossniklaus, 2003). The unmodified pMDC43 was used as a control vector, with fluorescence detected in both the cell nucleus and the cytoplasm (Wang, 2005). Both atU2AF³⁵a and atU2AF³⁵b proteins are clearly enriched in nuclei and no differences were detected between them. As shown in Figure 4, strong green fluorescence was detected in the nuclei of root cells (Fig. 4C), leaf cells, guard cells (Fig. 4, D and E), and trichomes (Fig. 4F). Detailed study using confocal microscopy revealed that the distribution of atU2AF³⁵a and atU2AF³⁵b protein in the nucleus is not even (Fig. 4, A and B). They both are likely organized in nuclear speckles, a pattern similar to known SR proteins (Ali et al., 2003; Docquier et al., 2004; Lorkovic and Barta, 2004; Lorkovic et al., 2004; Tillemans et al., 2005).

One line with T-DNA inserted into the 5′-UTR intron (SALK_050678; Alonso et al., 2003) was identified for atU2AF³⁵a. No insertion line was found for the atU2AF³⁵b gene. To knock down the atU2AF³⁵b gene, antisense and RNA interference (RNAi) vectors were constructed and transformed into Arabidopsis (vector constructions shown in Supplemental Fig. 3). Real-time RT-PCR revealed that the expression level of atU2AF³⁵a is down-regulated by 2.5-fold in atU2AF³⁵a T-DNA insertion plants (Fig. 5F). The atU2AF³⁵b RNAi plants have the atU2AF³⁵b gene knocked down to 7-fold lower expression. The atU2AF³⁵b antisense plants surprisingly have a normal atU2AF³⁵b expression level, but the atU2AF³⁵a gene is up-regulated to 2-fold. The atU2AF³⁵b gene is down-regulated relative to atU2AF³⁵a in antisense plants. Pleiotropic phenotypes were observed in the mutants and transgenic lines. The homozygous atU2AF³⁵a T-DNA insertion plants and atU2AF³⁵b transgenic plants are all late flowering under both long- and short-day conditions. As shown in Figure 5A, wild-type Arabidopsis (Col-CS60000) flowers at 20 to 25 d with 11 to 12 leaves under continuous light conditions. The atU2AF³⁵a T-DNA plants flower at 25 to 30 d with 12 to 15 leaves. The atU2AF³⁵b RNAi and antisense plants flower at 26 to 32 d with 13 to 15 leaves. In addition to the late-flowering phenotype, all three mutants have shorter flowers with enlarged bottoms compared with wild type (Fig. 5B). A few atU2AF³⁵a and atU2AF³⁵b plants show an arrested primary stem at a certain stage (Fig. 5C, indicated by a red arrow), with branches below the terminus growing normally. Distinct silique shape was observed in atU2AF³⁵a T-DNA plants. As shown in Figure 5, D and E, normal siliques are cylindrical, tapering at the ends. In contrast, the siliques in atU2AF³⁵a T-DNA plants are flattened and widened at the distal end. Additionally, silique number is increased at the terminus of the stem in atU2AF³⁵a mutants. Leaf morphology changes were also observed in atU2AF³⁵a/b plants, with atU2AF³⁵a mutants showing larger, yellowish, and flatter leaves and atU2AF³⁵b transgenic plants showing smaller, dark green, and serrated leaves (Wang, 2005).

Because atU2AF³⁵a and atU2AF³⁵b presumably function in splicing, RT-PCR was carried out using total RNA from wild type, atU2AF³⁵a mutant, and atU2AF³⁵b transgenic lines to probe expression of 16 genes, including 12 genes predicted to be alternatively spliced, three genes (FLOWERING LOCUS C [FLC], FCA, and FPA) involved in the flowering pathway (Macknight et al., 1997; Sheldon et al., 1999; Schomburg et al., 2001), and AtDBR1 (At4g31770) as a constitutively spliced multiple-intron single-copy gene (Wang et al., 2004). No significant differences were observed for the 12 alternatively spliced genes and FPA among these plants. FLC expression levels are much higher in atU2AF³⁵b antisense plants and several minor PCR products were observed in AtDBR1 mutant and transgenic plants (Wang, 2005). The increased levels of FLC transcripts are consistent with the late-flowering phenotype in atU2AF³⁵b antisense plants. For FCA, four transcript isoforms (α, β, γ, and δ) are known (see Fig. 6B), of which FCA-γ is the only isoform producing full-length protein (Macknight et al., 1997). The γ and δ isoforms were detected in all lines using primers designed from the UTRs, although the δ isoform was clearly visible only in the atU2AF³⁵b antisense (column BA in Fig. 6A) plants. As shown in Figure 6A, extra bands different from the known isoforms were produced from all mutant and transgenic lines. Three of these were cloned and sequenced, including two bands (FCA-m1 [1.4 kb] and FCA-m2 [1.2 kb]) from atU2AF³⁵b antisense and one band (FCA-m3 [547 bp]) from atU2AF³⁵b RNAi plants. Sequence alignment of the three isoforms against the genome revealed four novel introns (SI1–4) shown in Figure 6, B and C. All these introns have noncanonical intron borders, with repeated segments on both intron-exon junctions. The sequences of the repeated junctions are distinct for the different introns. Multiple exons were skipped in all novel sequenced FCA isoforms by splicing of a rather long (>4.5 kb) noncanonical intron. FCA-m1 also retained the sixteenth intron. These novel FCA isoforms demonstrate that the splicing pattern of some genes can be changed in plants with altered atU2AF³⁵a/b expression levels.

RT-PCR on atU2AF³⁵a genes revealed two extra bands in addition to the constitutively spliced product. Noncanonical alternative splicing events were identified by sequencing the extra products, which in each case removes a segment with repeated borders from the second exon of the atU2AF³⁵a gene. We also found that three extra FCA isoforms were produced by excision of noncanonical introns with repeated junctions in our atU2AF³⁵ mutant and transgenic lines. Very likely the change of atU2AF³⁵ levels will lead to the usage of noncanonical sites in some pre-mRNAs. It is interesting to note that all these noncanonical introns have repeated junctions. It may be possible that some protein homodimer is involved in symmetric ss recognition when U2AF³⁵ levels are abnormal. As the expression level of atU2AF³⁵a and atU2AF³⁵b changes dynamically during growth and development, it is likely that atU2AF³⁵a can autoregulate the level of full functional protein by the noncanonical alternative splicing events. In vertebrates, the U2AF1 gene can also be alternatively spliced by inclusion of an additional exon, producing an isoform with seven amino acid differences in the ψRRM (Pacheco et al., 2004). Two maize SR proteins (ZmSRp31A and ZmSRp31B) also show noncanonical alternatively spliced introns (Gupta et al., 2005), strongly suggesting that many splicing factors can posttranscriptionally regulate their expression by noncanonical alternative splicing.

Sequence alignment and phylogenetic analysis demonstrated that the U2AF³⁵ gene exists in the ancestor of eukaryotic organisms. The ancient U2AF³⁵ contained at least a ψRRM and two CCCH zinc fingers. It may also contain a run of Glys because the animal homologs and some monocot plant homologs have this motif. But this motif was lost in some homologs. In the plant kingdom, a C-terminal motif (SERE) was acquired after the divergence of green algae and probably before the divergence of seed plants. The SERE motif may have plant-specific functions, such as recognizing plant-specific splicing signals or interacting with plant-specific SR proteins. There might be another ancient U2AF³⁵ gene in plants, as suggested by the separation of the dicot II clade from the dicot I-monocot-ptU2AF³⁵a group, which was probably lost in monocot and some dicot (Arabidopsis) lineages. After the divergence of monocot and dicot plants, individual duplications of U2AF³⁵ genes happened in the ancestor of monocot plants, nightshade family (potato and tomato), Triticeae (wheat and barley), and Arabidopsis. The ancient plant U2AF³⁵ gene may have had introns in the 5′-UTR, as suggested by the fact that the U2AF³⁵ homologous genes in Arabidopsis, maize, and rice all have introns in the 5′-UTR. None of them have introns in either coding region or the 3′-UTR. Interestingly, the 5′-UTR intron is alternatively spliced in maize and rice homologs and the alternative splicing patterns and some intron segments are well conserved between the two monocot species (Wang, 2005), suggesting important biological functions of the 5′-UTR intron and the alternative splicing events in gene regulation.

Plant introns have neither conserved branch sequences nor a Py tract. The 3′-ss recognition in plants will probably rely more on U2AF than in mammals. What the exact roles of plant U2AF³⁵ homologs are and how they achieve their functions are challenging questions. Similar to mammalian introns, some plant introns may be AG independent and may not require U2AF³⁵ for correct splicing. Multiple copies of U2AF³⁵ exist in nearly all higher plant genomes. Different U2AF³⁵ homologs may preferably interact with different U2AF large subunits to form different U2AF heterodimers, as suggested by variations in the interacting regions of U2AF³⁵ from the same species as well as different species (Wang, 2005). Also the RS domain and its surrounding regions of U2AF³⁵ contain many variations, indicating the flexibility of interaction between U2AF³⁵ and variable SR proteins. It is likely that some SR proteins may preferably interact with one of the U2AF³⁵ proteins, which in turn interacts with specific U2AF large subunits. The balance among different U2AF-SR protein complexes on pre-mRNAs may be important to the splicing of particular introns. Disruption of this balance will produce abnormal splicing isoforms and reduce the abundance of normal splicing isoforms, as demonstrated in FCA genes in our mutant and transgenic lines. Thus we would expect to see similar phenotypes independent of which U2AF³⁵ gene was disrupted, as observed in our mutant and transgenic lines.

Multiple sequence alignments of the U2AF³⁵ proteins were generated with ClustalW using default parameters (Thompson et al., 1994). The alignments were visualized using the BioEdit program (version 5.0.9; http://www.mbio.ncsu.edu/BioEdit/bioedit.html). Phylogenetic analysis of the sequences was conducted using MEGA software (version 2.1; http://www.megasoftware.net; Kumar et al., 2001). The phylogenetic trees were constructed using the neighbor-joining method with a bootstrap test. The distance model used was a Kimura 2-parameter. All other parameters used were default.

Arabidopsis seeds were sown in soil and grown at 4°C for 4 d, and then the plants were moved to a growth room and grown at 22°C with continuous light. Total plant RNA was isolated using either TRIzol reagent (Invitrogen) or the Plant RNeasy mini kit (Qiagen) from 0.1 to 0.2 g of different tissues. The manufacturer's protocol was followed. For Arabidopsis, root, leaf, meristem, stem, and flower tissues from wild-type ecotype Columbia were used. Total RNA was dissolved in 30 μL diethyl pyrocarbonate-treated water and saved at −20°C.

The primer sequences are described in Supplemental Table I. Total RNA was treated by RQ1 RNase free DNase according to the manufacturer's protocol (Promega). Two micrograms of treated RNA were then used for first-strand synthesis and PCR according to the manufacturer's protocol (Invitrogen). A mixture of treated RNAs was used as a no-RT control. For real-time RT-PCR, PRIMER EXPRESS version 2.0 software (Applied Biosystems) was used to design oligonucleotide primers. cDNAs were prepared as described above and diluted 600-fold for amplification of the 18S ribosome RNA gene and 3-fold for other genes. One microliter of diluted cDNA was used in a 25-μL reaction with SYBR Green master mix (Applied Biosystems). All reactions were performed in triplicate by using a Prism 5700 sequence detection system (Applied Biosystems). The experiments were replicated twice using different RNA samples. Primer efficiency was checked for each primer pair by constructing a standard curve using an equal mixture of all cDNAs (Applied Biosystems). The expression level of each gene was calculated based on 2^−ΔΔct method described in user bulletin number 2 (Applied Biosystems). The relative amount of calculated message was normalized to the level of the 18S rRNA gene.

Two potential promoter regions together with the 5′-UTR region were checked for both atU2AF³⁵a and atU2AF³⁵b. For atU2AF³⁵a, promoter 1 (PGaa1) is the genomic region from 876 nt before the ATG start codon, and promoter 2 (PGaa2) is from 1,358 nt before the start codon. For atU2AF³⁵b, promoter 1 (PGab1) is from 555 nt before the start codon, and promoter 2 (PGab2) is from 982 nt before the start codon. As shown in Supplemental Figure 2, both longer promoters for atU2AF³⁵a and atU2AF³⁵b include part of the first exon of the neighboring gene. These tentative promoters were amplified from Arabidopsis genome DNA. Primers are shown in Supplemental Table I. PCR products were purified and ligated to vector pCAMBIA1381z. The vectors were subjected to sequencing from both ends to make sure the insertions were correct. In addition to the two promoters for each gene, the CaMV 35S promoter was linked with the GUS gene and used as a strong promoter control (PGxx). The empty pCAMBIA1381z (no promoter) was used as a no-promoter control (negative control, PG₀₀). These constructs are all shown in Supplemental Figure 2. The right vectors were used for Arabidopsis transformation by methods described below. Three to five individual transgenic plants from each transformation were subjected to histochemical GUS assays, following the protocol described in Weigel and Glazebrook (2002).

GFP vectors were constructed based on vector pMDC43 (Curtis and Grossniklaus, 2003). The ORFs of atU2AF³⁵a and atU2AF³⁵b genes were amplified and cloned into the pENTR vector, then excised and transferred to the pMDC43 vector using Gateway technology (Invitrogen). Antisense vectors were constructed based on binary vector pCAMBIA1301. The vector diagrams are shown in Supplemental Figure 3. The ORF of atU2AF³⁵b was amplified by PCR, using primers described in Supplemental Table I. The PCR product was isolated, digested, and inserted into the downstream of the CaMV 35S promoter in the reverse direction. To construct an RNAi vector, the atU2AF³⁵b ORF was amplified using the primer set shown in Supplemental Table I. The PCR product and antisense vector were digested by MluI and BsrGI (New England Biolabs), then ligated by T4-DNA ligase (Promega). The RNAi construct uses the atU2AF³⁵b ORF to replace part of the antisense vector sequence in the sense direction. The resulting transcript forms a hairpin structure that triggers silencing of the endogene.