Transgenic reporter system visualizes the developmentally regulated mutually exclusive alternative splicing of the let-2 gene in vivo

In order to monitor the selection of the let-2 mutually exclusive exons (Fig. 1A) in vivo, we constructed a pair of reporter minigenes, let-2-9G and let-2-10R (Fig. 1B). The minigenes carry the same let-2 genomic DNA fragment spanning from exon 8 to exon 11 connected in-frame to cDNAs for fluorescent proteins, and termination codons are artificially introduced into exon 10 of let-2-9G and exon 9 of let-2-10R (Fig. 1B). We expected that, from the let-2-9G minigene, selection of exon 9 would lead to expression of an mRNA encoding a GFP fusion protein (E9-GFP), while selection of exon 10 would result in a nonproductive mRNA (E10x) due to the termination codon (Fig. 1B). In the same way, selection of exon 10 would lead to expression of an mRNA encoding an RFP fusion protein (E10-RFP) and selection of exon 9 would result in a nonproductive mRNA (E9x) from the let-2-10R minigene (Fig. 1B).

The let-2 reporter successfully visualized the alternative exon usage. We drove expression of the let-2 reporter minigenes under the body wall muscle-specific myo-3 promoter, since the endogenous let-2 is primarily expressed in the body wall muscles (Graham et al. 1997). As expected, expression of the let-2 reporter in the body wall muscles gradually and almost completely switched from GFP to RFP along with the development; embryos exclusively express E9-GFP and elder worms express E10-RFP (Fig. 1C). RT–PCR analyses of mRNA isoforms derived from the minigenes confirmed that the alternative exons are selected mutually exclusively to produce the E9-GFP and E10x mRNA isoforms from the let-2-9G minigene, and the E9x and E10-RFP isoforms from the let-2-10R minigene (data not shown). Consistent with the color switch of the let-2 reporter worms, the composition of the reporter mRNAs gradually switched from the exon 9 forms to the exon 10 forms in parallel with that of the endogenous let-2 mRNAs along with the developmental stage (Fig. 1D). These results indicate that the let-2 reporter visualizes the developmental switching of the let-2 mutually exclusive exons in vivo.

asd-2 mutants are defective in the developmental switching of alternative splicing of both the let-2 reporter and the endogenous let-2 gene

To identify factors regulating the let-2 alternative splicing in vivo, we screened for mutants defective in the developmental switching of the let-2 alternative splicing reporter expression in the body wall muscles. We mutagenized the transgenic reporter allele ybIs1371 (Fig. 1C) and screened ∼5 × 10⁴ genomes for F2 progeny that remain green until adulthood or that turn red as younger larvae. We isolated several independent mutant alleles, all of which remain green during larval development but eventually turn red in the senile stage (Fig. 2A; data not shown). RT–PCR analysis of the let-2 reporter confirmed that the mutants are defective in the switching to the exon 10 forms in the late larval stage (Fig. 2B, top panel) in parallel with the defect in the color switch. Furthermore, the mutants are also defective in the exon switching of the endogenous let-2 mRNAs (Fig. 2B, bottom panel), indicating that the let-2 reporter reflected the expression of the endogenous let-2 gene and allowed the isolation of alternative splicing defective mutants.

Single nucleotide polymorphism (SNP)-based chromosome mapping and genomic DNA sequencing suggested that the mutants are alleles of a single uncharacterized gene, T21G5.5 (see below). RNAi of the gene reproduced the color phenotype (Fig. 2C), confirming that the defect in the switching of the let-2 reporter expression resulted from loss or reduction of function of the gene. We therefore named it alternative splicing defective-2 (asd-2). Figure 2D represents a schematic structure of the asd-2 gene and summarizes the mutations identified in the asd-2 mutant alleles. Expressed sequence tag (EST) analysis has predicted that the asd-2 gene produces two protein isoforms, ASD-2a and ASD-2b, that differ in the N terminus sequence due to alternative use of the first exons (Fig. 2D), and we confirmed expression of both of the isoforms by cloning RT–PCR products (deposited to GenBank under accession nos. EF630625 and EF630626).

ASD-2 is a novel member of the evolutionarily conserved STAR family

ASD-2 belongs to the evolutionarily conserved STAR family of RNA-binding proteins (Vernet and Artzt 1997). The family members share a highly conserved tripartite GSG (for GRP33, Sam68, GLD-1)/STAR domain (Jones and Schedl 1995; Vernet and Artzt 1997) composed of a single KH-type RNA-binding domain and two flanking domains called QUA1 and QUA2 (Fig. 3A). QUA1 is required and sufficient for homodimer formation of C. elegans GLD-1 and mouse QKI (Chen and Richard 1998; Ryder et al. 2004). QUA2 is involved in the RNA binding together with the KH domain for GLD-1 and human SF1 (Liu et al. 2001; Ryder et al. 2004; Lehmann-Blount and Williamson 2005). Figure 3B shows the phylogenetic tree of the GSG/STAR domains of the STAR family proteins from C. elegans and mammals and Drosophila Held Out Wings (HOW). ASD-2 GSG/STAR domain has 68%, 66%, and 74% identity with those of GLD-1, QKI, and HOW, respectively, while the ASD-2 KH-QUA2 domains have only 37% identity to those of human SF1. These sequence analyses indicated that ASD-2 belongs to the GLD-1 subfamily that includes QKI and HOW.

Mutations identified in the asd-2 mutants affect conserved residues in the GSG/STAR domain or the entire structure of the ASD-2b protein. Five missense mutations were assigned to the conserved residues within the GSG/STAR domain (Fig. 3A). yb1422 causes G125E substitution within the QUA1 domain. yb1415, yb1470, and yb1419 cause R154C, H193Y, and L202F substitutions within the KH RNA-binding domain, respectively. yb1423 causes E244K substitution within the QUA2 domain. Putative asd-2b-specific null mutations were identified in the other three alleles. yb1540 causes nonsense mutation in the first exon of the asd-2b isoform (Fig. 2D). yb1421 and yb1542 have base substitutions at the splice donor site in the first intron of the asd-2b isoform (Fig. 2D). These results suggested that the RNA-binding property of the ASD-2b isoform is responsible for the developmental switching of the let-2 mutually exclusive exons in the body wall muscles.

ASD-2b regulates the switching of the let-2 reporter in the body wall muscles

We then analyzed the expression profiles of the asd-2a and asd-2b isoforms. We constructed transcriptional fusion minigenes, asd-2a∷GFP and asd-2b∷RFP (Fig. 4A). These reporters showed differential tissue-specific expression profiles throughout development (Fig. 4B). Expression of asd-2a∷GFP was detected in hypodermis and pharyngeal muscles. On the other hand, expression of asd-2b∷RFP was observed in the body wall muscles and the pharyngeal muscles, consistent with the idea that asd-2b is responsible for the regulation of the let-2 alternative splicing in the body wall muscles.

Next, we asked whether expression of ASD-2b in the body wall muscle rescues the color phenotype of the asd-2 mutant. Overexpression of ASD-2b under the body wall muscle-specific myo-3 promoter overrescued the asd-2 (yb1423) mutant, and enforced E10-RFP expression even in early developmental stages (Fig. 4C). Figure 4D summarizes the effects of the asd-2 expression on the let-2 reporter. We concluded that the ASD-2b isoform regulates the let-2 alternative splicing reporter in the body wall muscles.

ASD-2b regulates the switching of the let-2 alternative exons via a short element in intron 10

Then, we searched for a cis-acting element for the ASD-2b function in the let-2 pre-mRNA processing. Kabat et al. (2006) have recently reported that evolutionarily conserved intronic stretches are likely to be involved in the regulation of alternative pre-mRNA splicing in nematodes. Furthermore, they demonstrated that deletion of short stretches within let-2 intron 10 affected the alternative splicing regulation (Kabat et al. 2006). We therefore constructed a pair of modified let-2 reporter minigenes (M1) with a deletion of a short sequence in intron 10 (Fig. 5A), and observed that the M1 mutant reporter continued expressing E9-GFP in the late larval stages (Fig. 5B). Since the sequence deleted in the M1 minigenes contained a half-site of the QKI response element (QRE, NACUAAY-N_1–20-UAAY), a consensus sequence of mouse QKI (Galarneau and Richard 2005), we modified two residues in the short stretch, CUAACUCUAAC, in another pair of modified let-2 reporter minigenes (M2) (Fig. 5A), and found that disruption of the stretch also phenocopied the asd-2 mutants (Fig. 5C). Furthermore, the M2 mutant reporter no longer responded to the overexpression of the ASD-2b cDNA (Fig. 5D). We also constructed a pair of minigenes (M3) in which a conserved UGCAUG element is disrupted (Fig. 5A), and found that the M3 mutant reporter responded to ASD-2b overexpression (Fig. 5E) as did the wild-type (WT) reporter (Fig. 4C). These results indicated that the QRE-like stretch is essential for ASD-2b to enforce expression of E10-RFP.

To examine whether ASD-2 directly binds to the QRE-like stretch in vitro, we performed an electrophoretic mobility shift assay (EMSA) with RNA probes harboring the intact stretch or the mutation as in the M2 mutant reporter (Fig. 5F). The recombinant protein of the ASD-2 GSG/STAR domain (ASD-2 STAR) shifted the mobility of the wild-type RNA probe (WT) in a dose-dependent manner (Fig. 5F, lanes 2–6), but not of the M2 mutant probe (Fig. 5F, lanes 8–11), demonstrating the direct and specific binding of ASD-2 to the QRE-like stretch in vitro.

asd-2 promotes removal of let-2 intron 10 in the late larval stages

Next, we examined the processing steps of the endogenous let-2 pre-mRNA. When we look at exon 8 to exon 11 of the let-2 pre-mRNA, mutually exclusive inclusion of exon 9 or exon 10 requires sequential removal of two flanking introns. As intron 9 (30 nucleotides) is too short to be excised due to steric interference between the donor and the possible branch site (Smith et al. 1993; Southby et al. 1999), we can expect four introns to be removed in the first splicing, and the subsequent second splicing to produce either of the two mature let-2 mRNAs (Fig. 6A). To gain insight into the processing course of the endogenous let-2 pre-mRNA, we tried to detect and compare the amounts of the four possible processing intermediates or partially spliced RNAs (Fig. 6A) in the synchronized wild-type and asd-2 mutant worms.

To quantify the amounts of the let-2 processing intermediates, we prepared two pairs of PCR primers to amplify the complementary DNA fragments for the let-2 RNAs retaining intron(s). One primer set was designed on exon 8 and intron 10 to amplify partially spliced RNAs in which either of the upstream flanking introns was removed (Fig. 6B, left panel). The primer set amplified the unspliced pre-mRNA (8–9–10–11) and partially spliced RNAs in which intron 8 was excised (8/9–10–11) or the intron between exon 8 and exon 10 was excised (8/10–11). The other primer set was designed on intron 8 and exon 11 to amplify partially spliced RNAs in which either of the downstream flanking introns was removed (Fig. 6B, right panel). The primer set amplified the pre-mRNA (8–9–10–11) as well as partially spliced RNAs in which intron 10 was removed (8–9–10/11), or the intron between exon 9 and exon 11 was removed (8–9/11).

The normalized amounts of the four partially spliced RNAs are shown in Figure 6C. In both the wild-type and the asd-2 mutant embryos, the 8–9/11 form, which is committed to the exon 9 mRNA isoform, was almost exclusively detected (Fig. 6B [lanes 1,2,5,6], C), consistent with the exclusive expression of the exon 9 mRNA in the embryos. In parallel with the developmental switching to the exon 10 mRNA isoform in the wild-type L4 larvae, we observed switching in the major form of the partially spliced RNAs from the 8–9/11 form to the 8–9–10/11 form, one of two possible intermediates leading to the exon 10 mRNA (Fig. 6B [lane 7], C). On the other hand, the amount of the 8–9/11 form still predominated over that of the 8–9–10/11 form in the L4 larvae of the asd-2 mutant (Fig. 6B [lane 8], C), consistent with its switching-defective phenotype. These results indicated that asd-2 is required for selective removal of intron 10 in this stage. We also observed substantial increase in the amount of the 8/10–11 form in the L4 larvae independent of the asd-2 functions (Fig. 6B [lanes 3,4], C), which may contribute to the fraction of the exon 10 mRNA in the adult stage (Fig. 2B) and the color change of the let-2 reporter in the asd-2 senile adults.

The 8–9/11 form, the only intermediate detected in the embryos, will further be processed exclusively into the exon 9 mRNA (Fig. 6A). However, the 8–9–10/11 form, the major intermediate in the wild-type L4 larvae, has two remaining splice acceptors, for exon 9 and exon 10, and may not necessarily be processed into the exon 10 mRNA. We therefore analyzed the fate of a pre-mRNA from a modified let-2 reporter minigene that carried only exon 8 to exon 10, and observed a selective use of the exon 10 splice acceptor (Fig. 7A [top panel], B [lane 1]). We reasoned that the selective use of the exon 10 acceptor (TTTTTAG) was due to the relative weakness of the exon 9 splice acceptor (ATTCTAG), since the exon 9 acceptor became predominantly selected (Fig. 7A [bottom panel], B [lane 2]) when the reporter was further modified, so that the exon 9 acceptor matched the consensus of U2AF65 binding for C. elegans (TTTTCAG) (Hollins et al. 2005). These results are schematically summarized in Figure 7C.

Course of mutually exclusive alternative splicing of let-2 pre-mRNA and the switching mechanism by ASD-2

The unique property of the developmental switching and the isolation of the mutants of the let-2 alternative splicing allowed us to analyze the relative amounts of the partially spliced RNAs (Fig. 6). In the case of Nova-2-regulated cassette exons in mice, Ule et al. (2006) observed biased accumulation of a partially spliced RNA in which either of the upstream or the downstream flanking introns was removed in the exon-including condition, either in the wild-type or the Nova-2 knock-out brain. In general, the steady-state amount of the processing intermediate is determined by the rate of production (removal of an intron in the first splicing) and the rate of degradation (mainly the subsequent splicing), and therefore it does not necessarily reflect the rate of production of the mature mRNA isoform produced from the intermediate. Nevertheless, we found a dramatic switching in the major species of the partially spliced let-2 RNAs during larval development in parallel with the switching from the exon 9 to the exon 10 isoforms of the mature let-2 mRNA (Fig. 6C), suggesting that the major species of the partially spliced RNAs is the processing intermediate of the major course of the pre-mRNA processing. Furthermore, we observed a defect in the switching in the major species of the partially spliced let-2 RNAs in L4 larvae of the asd-2 mutant (Fig. 6C), consistent with the defect of the asd-2 mutant to produce the exon 10 mRNA. Assuming that asd-2 mutation does not affect the rate of the second splicing, the result indicates that asd-2 is required for the production of the major 8–9–10/11 form in the wild-type L4 larvae, and strongly suggests that the 8–9–10/11 form is the intermediate in the course of the let-2 pre-mRNA processing by which most of the exon 10 mRNA is produced in the later stages.

Figure 8 illustrates a model of the processing courses of the let-2 pre-mRNA and the mode of switching regulation by ASD-2 raised by the present study. The unique feature of the let-2 alternative splicing is that either of the two introns downstream from the mutually exclusive exons is removed in the first splicing out of four possible flanking introns (Fig. 6A). In the embryonic stage, or when asd-2 is inactive (Fig. 8, left panel), the splice donor of exon 9 (Fig. 8, green circle, GTACTG) is preferentially selected to produce the 8–9/11 intermediate, since the splice donor of exon 10 (Fig. 8, green triangle, GCAAGA) is noncanonical and weaker (Farrer et al. 2002; Kabat et al. 2006), leading to the production of the exon 9 mRNA. On the other hand, when asd-2 is active in the late larval stages (Fig. 8, right panel), ASD-2 binds to the QRE-like cis-acting element in intron 10 to promote the selection of the weak exon 10 splice donor, leading to the production of the 8–9–10/11 intermediate. Thereafter, the splice acceptor of exon 10 (Fig. 8, red circle, TTTTTAG) is preferentially selected, as that of exon 9 (Fig. 8, red triangle, ATTCTAG) is weaker.

The model described above indicates that the fate of the pre-mRNA, whether to be the exon 9 or exon 10 mRNA, is primarily determined by the removal of a flanking intron in the first splicing. The model also indicates that the mutually exclusive nature of the let-2 alternative exons is ensured by the ordered removal of the flanking introns and the asymmetric strength of the splice donors and the acceptors of the two alternative exons, as well as by the steric interference of the two exons. In the present study, we observed that endogenous asd-2b was expressed in embryos (Fig. 4B) and that the let-2 exon 10 mRNA isoform was eventually produced in the adult stage, even in the asd-2 mutant background. These observations suggest a possibility that besides ASD-2, other regulators, such as modulators of the ASD-2 function and/or other trans-factors, are involved in the let-2 alternative splicing. Regulation of the order of the intron removal may also be involved in the developmental switching as implicated by the substantial increase in the 8/10–11 intermediate in the L4 stage (Fig. 6C). It would be an interesting challenge to elucidate all of the factors involved in the precise control of developmental switching of the let-2 mutually exclusive alternative splicing in vivo.

STAR family RNA-binding protein ASD-2b regulates alternative splicing in C. elegans

STAR family proteins are known as key regulators of the various aspects of biological processes through multiple functions in the RNA metabolism. In the STAR family, the GLD-1 subfamily has been shown to play multiple roles through direct binding to the target RNAs. The Quaking gene generates three isoforms of proteins by alternative splicing. The nuclear isoform QKI-5 regulates alternative splicing of the myelin-associated glycoprotein (MAG) gene (Wu et al. 2002), while the balance of the nuclear and cytoplasmic isoforms QKI-6 and QKI-7 regulates the nuclear export of the myelin basic protein (MBP) mRNA (Larocque et al. 2002). Drosophila how encodes two alternative isoforms and the balance between the opposing activities of the two isoforms regulates tendon cell differentiation through the level of the transcription factor Stripe (Nabel-Rosen et al. 1999, 2002). Unlike vertebrates and insects, C. elegans has two members of the GLD-1 subfamily of genes. The cytoplasmic protein GLD-1 (Jones et al. 1996) regulates the germ cell development and sexual identity by regulating mRNA translation and stability (Jan et al. 1999; Lee and Schedl 2001, 2004; Schumacher et al. 2005). Our genetic studies demonstrate that ASD-2b regulates alternative splicing in the body wall muscles. Although the molecular function of the ASD-2a isoform should be characterized further, ASD-2a may play essential roles in the nematode development, since the asd-2 deletion allele ok483, in which common exon 3 and exon 4 are deleted, is lethal (WormBase, http://www.wormbase.org), while the asd-2b-specific mutants look otherwise normal and only show the color phenotype. The multiple nuclear and cytoplasmic isoforms produced by alternative splicing and/or gene duplication may have conferred the multiple functions in mRNA metabolisms to the subfamily.

Sam68 and its paralogs, SLM1 and SLM2 of mammals constitute of the Sam68/SLM subfamily in the STAR family (Fig. 3B; Di Fruscio et al. 1999). The subfamily has been shown to mediate the signal transduction to RNA metabolism by modulating the RNA-binding properties upon phosphorylation (Stoss et al. 2001; Matter et al. 2002; Haegebarth et al. 2004; Paronetto et al. 2007). Although signals regulating the GLD-1 subfamily are to be elucidated, regulation of the RNA-binding property may be involved in the regulation of the let-2 alternative splicing by ASD-2b in vivo.

RNA-binding specificity of ASD-2

We identified the essential QRE-like cis-element for ASD-2b to regulate let-2 alternative splicing. Several studies have identified the direct binding sites of the GLD-1 subfamily and addressed the consensus of the binding sequences. Ryder et al. (2004) demonstrated that GLD-1 preforms a dimer, and specifically binds to the UACU(C/A)A stretch(es) in the target mRNAs. They also showed that QKI recognizes the consensus NA(A>C)UAA by extensive mutagenesis of the GLD-1 consensus sequence (Ryder and Williamson 2004). Enrichment of the high-affinity RNA ligands using systematic evolution of ligands by exponential enrichment (SELEX) defined the QRE as a bipartite consensus sequence NACUAAY-N_1–20-UAAY, consisting of a core site (NACUAAY) and a tetra-nucleotide half site (UAAY) (Galarneau and Richard 2005). Recently, the primary sequence and the secondary structure requirement for Drosophila HOW response element (HRE) has been determined as the ACUAA stretch in the loop (Israeli et al. 2007). The results from these studies indicate that the GSG/STAR domains of the GLD-1 subfamily share not only the high amino acid sequence identity, but also the consensus binding sequence determined as QRE, although the sequence that binds with the highest affinity may differ among the GLD-1 subfamily members (Saccomanno et al. 1999; Ryder and Williamson 2004).

The cis-element we identified in this study consists of direct repeats of the penta-nucleotide stretch, CUAAC, which matches five out of six nucleotides of the QRE core site. The tetra-nucleotide QRE half site, which can be either upstream of or downstream from the core site, has been shown to be required for the high-affinity binding (Galarneau and Richard 2005) and the requirement of the half site in the proximity has also been suggested for GLD-1 (Ryder et al. 2004). It is likely that the direct repeats of the penta-nucleotide compensated the requirement of the complete core site for let-2 regulation by ASD-2 as in the case of QRE-1 in the 3′-untranslated region (UTR) of MBP mRNA, where multiple clusters of the half site constitute QRE (Galarneau and Richard 2005). QKI and GLD-1 form homodimers via the QUA1 domain (Chen and Richard 1998; Ryder et al. 2004) and the E48 residue of QKI, conserved among the GLD-1 subfamily proteins (Fig. 3A), is essential for the dimer formation and the function of the Quaking gene (Chen and Richard 1998). We identified a substitution of a conserved residue in the QUA1 domain in the asd-2 (yb1422) mutant (Fig. 3A), suggesting the functional importance of the QUA1 domain for the ASD-2 function. ASD-2 may also form a homodimer via the conserved QUA1 domain, and each subunit may contact one CUAAC stretch in binding to the let-2 pre-mRNA. Comparison of the dimer formation activities and the RNA-binding properties of the GSG/STAR domains derived from the various asd-2 mutants isolated in this study would lead to further understanding of the RNA-recognition by the STAR family proteins.

Recent global and comparative analyses of the alternatively spliced exons have demonstrated that the (A)CTAAC stretch, which resembles the QRE, is enriched in the introns flanking the alternative exons in both mammals and nematodes, suggesting that the stretch is an intronic alternative splicing regulator (Kabat et al. 2006; Sugnet et al. 2006; Das et al. 2007; Yeo et al. 2007). The GLD-1 subfamily may be involved in the regulation of these genes, although the requirement of the (A)CTAAC stretch and the function of the GLD-1 subfamily are to be experimentally addressed.

Visualization of the alternative splicing patterns in vivo

In the present study, we utilized a pair of symmetric minigenes, both of which contain four exons and three intervening introns for visualizing the expression profiles of the let-2 mutually exclusive alternative exons (Fig. 1B). In contrast, for visualizing the tissue-specific expression profiles of the egl-15 mutually exclusive exons, we utilized a single unified minigene, in which GFP and RFP cDNAs were introduced into the alternative exons (Kuroyanagi et al. 2006). With the unified reporter minigene, we demonstrated that mutually exclusive selection of the egl-15 alternative exons is regulated by the upstream-most intron, and that the downstream-most intron is dispensable for the regulation (Kuroyanagi et al. 2006). In general, the downstream flanking intron and the downstream constitutive exon should be included in the alternative splicing reporter minigenes, since the cis-element(s) in the intron downstream from the alternative exon(s) could be essential for the regulation, as demonstrated in this study. As the let-2 reporter minigenes are symmetric and contain the intact genomic sequence, except for a modification in one exon in each of the reporter minigenes, we can assume that expression of the reporter reflects the regulation of the endogenous gene, and that the simultaneous modification of the cis-element(s) in both of the minigene pairs should cause the same effect on the pre-mRNAs derived from these minigenes, as demonstrated in this study. We therefore conclude that a symmetric pair of minigenes is feasible for the visualization of the alternative splicing patterns and for the elucidation of the regulation mechanisms in C. elegans.

The utilization of dichromatic alternative splicing reporters has recently been described in other organisms. Carstens and colleagues (Newman et al. 2006) utilized a pair of reporter minigenes for the FGFR2 mutually exclusive exons IIIb and IIIc with a similar structure to the let-2 reporter. They demonstrated that the reporter and a fluorescence-activated cell sorter (FACS) facilitated the analysis of the function of various trans-factors in the selection of the mutually exclusive exons. Cooper and colleagues (Orengo et al. 2006) utilized another type of a single dichromatic reporter minigene, in which GFP and RFP cDNAs were fused in the different frame so that mutually exclusive expression of GFP and RFP from the single minigene reflected the expression of the alternative isoforms. Garcia-Blanco and colleagues (Bonano et al. 2006) generated knock-in mice that visualized the repression of the FGFR2 exon IIIb in vivo. The mice carry an RFP-minigene to monitor efficiency of constitutive splicing and a GFP-minigene to monitor repression of FGFR2 exon IIIb at a single-cell resolution in vivo. The advantage of these dichromatic fluorescent alternative splicing reporters, including those in our study, is that relative abundance of the two alternative isoforms in each cell can be reliably analyzed with a minimal effect of other aspects of gene expression such as a transcription level. Application of these alternative splicing reporters will facilitate expression profiling and experimental identification of trans-factors and cis-elements of various alternative splicing events.

In conclusion, visualization of the expression profiles of the alternative exons enabled us to isolate the splicing mutants and identify trans-acting factors and cis-acting elements. Isolation of the splicing mutants enabled us to detect and analyze the partially spliced pre-mRNAs or processing intermediates. These experimental data led to a model of how only one out of two exons is exclusively included in each mRNA and how the factor and elements regulate the fate of the exons. Experimental approaches of visualization of the splicing patterns, genetic analyses of splicing mutants, and detection of the processing intermediates in vivo, together with the comprehensive bioinformatic approaches, would open the way to further elucidate the spatiotemporally regulated fate of pre-mRNAs, or deciphering the splicing codes in metazoans.

Plasmid construction

We constructed the reporter minigenes essentially as described in the text. To construct the reporter cassettes for let-2-9G, let-2-10R, and let-2-10/R-WT, we amplified the let-2 genomic DNA fragments spanning from exon 8 to either exon 11 or exon 10 and cloned them into Gateway Entry vectors (Invitrogen) carrying either EGFP (Clontech) or mRFP1 (Campbell et al. 2002) cDNAs using the In-Fusion system (BD Biosciences). We introduced termination codons by using QuickChange (Stratagene). To construct the ASD-2b Entry vector, we amplified the ASD-2b cDNA fragment and cloned it in pENTR-D/TOPO (Invitrogen). We constructed the Expression vectors by homologous recombination between the reporter cassettes and the myo-3 promoter vector pDEST-myo-3p (Kuroyanagi et al. 2006) by utilizing the Gateway system (Invitrogen). We generated M1, M2, and M3 mutant minigenes from myo-3∷let-2-9G and myo-3∷let-2-10R, and myo-3∷let-2-10/R-Mut from myo-3∷let-2-10/R-WT by mutagenesis utilizing QuickChange (Stratagene). We amplified the genomic fragments of the asd-2a and asd-2b promoter regions and cloned them into promoterless Expression vectors to construct asd-2a∷GFP and asd-2b∷RFP. We confirmed all of the constructs by sequence analysis. Sequences of the primers used in the construction are available in the Supplemental Material.

Worm culture and microscopy

We cultured worms following standard methods (Brenner 1974). We prepared the transgenic lines essentially as described (Mello et al. 1991) using lin-15 (n765) as a host. We generated integrant lines by ultraviolet light irradiation (Mitani 1995). We treated ybIs1371 [myo-3∷let-2-9G myo-3∷let-2-10R lin-15(+)] animals with 47 mM ethyl methanesulfonate (EMS) for 4 h at room temperature, and we screened the F2 progeny for mutants with altered expression profiles of the GFP and RFP under a fluorescence stereoscope (MZ16FA; Leica) equipped with a dual-bandpass filter set. We captured the images of the let-2 reporter worms with CCD cameras (E300 and DP71; Olympus). We used a confocal microscope (Fluoview FV500; Olympus) for image scanning and processed the acquired images with Metamorph (Molecular Devices). We identified the asd-2 gene by SNP-based mapping essentially as described (Wicks et al. 2001). Briefly, we crossed the mutants with the Hawaiian wild-type strain, CB4856, which has many SNPs compared with the standard wild-type strain N2, and the SNPs of each F2 progeny with the green phenotype were analyzed by either restriction fragment length polymorphisms or direct sequencing of the PCR-amplified genomic fragments. Of ∼75 predicted genes within the defined 391-kb region, we amplified and sequenced 13 RNA-related genes. We performed RNAi by soaking L4 hermaphrodites as described previously (Kuroyanagi et al. 2000; Maeda et al. 2001) by using dsRNAs corresponding to nucleotides 164–841 of the ASD-2b ORF.

RT–PCR

We prepared total RNAs and performed RT–PCR analysis essentially as described previously (Kuroyanagi et al. 2006). We digested the RT–PCR products for the let-2 reporter and the endogenous let-2 mRNAs with BglII before electrophoresis separation. Sequences of the primers used in the RT–PCR assays are available in the Supplemental Material.

EMSA

We prepared the recombinant protein for the GSG/STAR domain of ASD-2 (ASD-2 STAR), corresponding to amino acids 55–280 of ASD-2b, as a His-tagged protein. We introduced the ASD-2 STAR cDNA fragment into the modified bacterial cold-induction vector (Kuroyanagi et al. 2007) and induced the protein production following the standard cold-induction procedure (Takara). We purified the His-ASD-2 STAR protein with Ni-NTA agarose (Qiagen). We prepared the ³²P-labeled wild-type and M2 RNA probes by in vitro transcription with [α³²P]UTP and T7 RNA polymerase (Takara). We incubated the gel-purified RNA probes alone or with increasing concentrations of His-ASD-2 STAR in 25 μL of RNA-binding buffer (20 mM HEPES-KOH at pH 7.9, 320 mM KCl, 5% glycerol, 1% Triton X-100, 1 mM DTT, 0.1 mM PMSF) supplemented with 130 ng/μL Escherichia coli tRNA and 50 ng/μL BSA for 30 min at 20°C. Each sample was separated on a nondenaturing polyacrylamide gel and visualized with a phosphorimage analyzer (FLA-3000G; Fuji Film).

Analysis of partially spliced let-2 RNAs

We performed RT with random hexamers. We performed PCR following standard protocols supplemented with [α³²P]dCTP, and separated the PCR products on 5% nondenaturing polyacrylamide gels and quantified the signal intensity of each band with a phosphorimage analyzer (FLA-3000G; Fuji Film). We calculated the amount of each processing intermediate by normalization to that of the pre-mRNA in the same reaction, taking their G + C contents into account. We confirmed the identity of each band by cloning and sequence analysis. We performed control PCRs using RT products without reverse transcriptase (shown in Supplemental Fig. 1) to calculate the contamination of genomic DNAs in the total RNA fraction. Sequences of the primers used in the PCRs are available in the Supplemental Material.

Database information

The gene assignment asd-2 has been approved by the CGC Genetic Map and Nomenclature Curator. The nucleotide sequences of the asd-2 cDNAs have been deposited in the GenBank/EMBL/DDBJ databases with the accession numbers EF630625 and EF630626.