The STAR/GSG Family Protein rSLM-2 Regulates the Selection of Alternative Splice Sites*

We identified the rat Sam68-like mammalian protein (rSLM-2), a member of the STAR (signal transduction and activation of RNA) protein family as a novel splicing regulatory protein. Using the yeast two-hybrid system, coimmunoprecipitations, and pull-down assays, we demonstrate that rSLM-2 interacts with various proteins involved in the regulation of alternative splicing, among them the serine/arginine-rich protein SRp30c, the splicing-associated factor YT521-B and the scaffold attachment factor B. rSLM-2 can influence the splicing pattern of the CD44v5, human transformer-2β and tau minigenes in cotransfection experiments. This effect can be reversed by rSLM-2-interacting proteins. Employing rSLM-2 deletion variants, gel mobility shift assays, and linker scan mutations of the CD44 minigene, we show that the rSLM-2-dependent inclusion of exon v5 of the CD44 pre-mRNA is dependent on a short purine-rich sequence. Because the related protein of rSLM-2, Sam68, is believed to play a role as an adapter protein during signal transduction, we postulate that rSLM-2 is a link between signal transduction pathways and pre-mRNA processing.

the model of a large RNA processing unit (8,9) termed RNA factory (3). Pre-mRNA splicing is characterized by a high fidelity and can be modulated in a cell type-or development-specific way to use exons alternatively. Although the exact mechanisms governing splice site selection are still not fully understood, recent results indicate that loosely defined signals on the pre-mRNA known as splicing enhancers/silencers, play a crucial role in splice site selection (10 -12). An important class of proteins that recognize splicing enhancers/silencers is the serine/arginine-rich (SR) and SR-related protein family that is involved in both constitutive and alternative splicing (13,14). In addition, it has also been shown that an increasing number of heterogenous nuclear ribonucleoproteins (hnRNPs) are involved in the regulation of alternative splicing. For example, splicing regulation of the neuron-specific exon N1 or the src pre-mRNA is under the control of the hnRNPs hnRNP I (polypyrimidine tract binding protein), hnRNP F, and hnRNP H (15,16).
SR proteins and hnRNPs can change alternative splicing patterns in a concentration-dependent manner both in vivo and in vitro (for review, see Refs. 13 and 14). Because the relative expression levels of SR proteins and hnRNPs show tissue-dependent variations (17)(18)(19), one hypothesis is that tissue-specific splicing is the result of concentration differences of ubiquitously expressed proteins that regulate the usage of a given splice site.
Another way of regulating alternative splicing decisions could be the presence of factors specific for a tissue type or developmental stage. A well known example is expression of the sex-specific SR-related protein transformer (20). In addition, tissue-specific proteins such as ELAV (embryonic lethal abnormal vision) (21) and NOVA (22) were described, and we previously found tissue-specific variants of the splice factor tra2-␤ (23) which have a different effect on a given splice site (24). Finally, several proteins involved in pre-mRNA splicing, judged to be ubiquitously expressed by Northern blot analysis, show cell type-specific expression when examined by histochemical methods. These include tra2-␤1 (23,25), hnRNP L (19), and YT521-B (26).
The nuclear RNA-binding protein Sam68 was first described as a target of the tyrosine kinase Src during mitosis (27,28). Sam68 is a member of the STAR (signal transduction and activation of RNA) protein family (29), also called GSG protein family (GRP33, Sam68, GLD-1) (30,31), whose members share a slightly extended hnRNP K homology RNA binding domain, called maxi-KH domain. Because Sam68 binds to various pro-teins involved in signal transduction such as phospholipase C␥ or p85 phosphatidylinositide 3-kinase, this protein has been proposed to play a role as an adapter protein in fibroblast and lymphocyte signaling (32)(33)(34)(35). In addition, there is increasing evidence that Sam68 is important for cell cycle progression (36 -38). Recently, two nuclear Sam68-like mammalian proteins (SLM-1 and SLM-2, also called T-STAR/etoile/Salp) have been described (39 -41). However, the exact function of Sam68 and its relatives in the nucleus still remains elusive.
Here, we demonstrate for the first time the involvement of the Sam68-like mammalian protein rSLM-2 in the regulation of alternative splicing. This protein interacts with splicing regulatory proteins in vivo and in vitro and influences the splice site selection of three different minigenes in a concentrationdependent manner. Using a CD44 minigene, we show that the rSLM-2-dependent inclusion of exon v5 depends on a purinerich sequence to which rSLM-2 binds in vitro.

MATERIALS AND METHODS
Two-hybrid Screening and Cloning-A yeast two-hybrid screen and interaction experiments were performed as described (26,42). Using rSAF-B as a bait in pGBT9, 200,000 colonies of a rat brain embryonic day E16 library (Stratagene, pAdGal4-cDNA as prey) were screened. The DNA of six lacZ-positive clones able to grow on selective medium containing 10 mM 3-aminotriazole was sequenced as described (26). An amino-terminal FLAG tag was introduced to the full-length form or rSLM-2 by polymerase chain reaction followed by subcloning into pcDNA3.1.
Antiserum Production and Purification-Peptides specific for rSLM-2 VVTGKSTLRTRGVTCG and PRARGVPPTGYRPCG were coupled to keyhole limpet hemocyanin and used to immunize rabbits. After 121 days, serum was purified by affinity chromatography, employing a mixture of recombinant GST-rSLM-2 and the two peptides following the manufacturers instructions (Pierce). Dilution for Western blot was 1:1,000 and for immunohistochemistry 1:100. Preabsorption of 10 l of anti rSLM-2 antibody at a concentration of 0.07 mg/ml with 80 g rSLM-2 peptides for 30 min at room temperature abolished the specific signal.
In Vitro Protein Interaction Assay-The cDNA of potential interactors of rSLM-2 was cloned into pCR3.1 (Invitrogen) and used for an in vitro reticulocyte lysate transcription/translation (TNT, coupled reticulocyte lysate system, Promega) to obtain the corresponding 35 S-labeled proteins. For the binding experiments, 2 l of the reactions were incubated with 1 g of GST or GST-rSLM-2 coupled to glutathione-Sepharose 4B (Amersham Pharmacia Biotech) in the presence of 200 l of HNTG buffer and 0.1% Triton X-100 (43) for 2 h at 4°C. Washing and detection were as described (43).
RNA Gel Shift Assay-RNA gel shift assays were performed as described (24). End labeling of the RNA oligonucleotides was performed with T4 polynucleotide kinase (New England Biolabs) and [␥ 32 P]ATP followed by purification using the nucleotide removal kit (Qiagen). Indicated amounts of recombinant GST-rSLM-2 were incubated with 30 pmol of the 32 P-labeled RNA oligonucleotides for 15 min at 30°C. The oligonucleotide sequences were: r-rich, GAGGAGGAAAGAGGAGAGAGAAAGGAGGAA; y-rich, CUCCUCCUUUCUCCUCUCUCUUUCCUCCUU; v5, AGUAUCAGGAUGAAGAGGAGACCCCACAUGCUACAAGCACAA; v5ls9, AGUAUCAGGcgacgcgucgGACCCCACAUGCUACAAGCACAA.
For the CD44v5 and clathrin light chain B minigenes, polymerase chain reaction conditions were as described (24,44). For the htra2-␤ minigene, 20-s denaturation at 94°C, 20-s annealing at 65°C, and 40-s extension at 72°C for 33 cycles were used followed by a final extension at 72°C for 20 min.
For the tau minigenes, polymerase chain reaction was carried out for 22 cycles with 94°C for 1 min, annealing at 65°C for 1 min, and extension at 72°C for 1 min. The resulting splicing pattern was quantified using the Herolab EASY system.

Molecular
Cloning and Sequence of rSLM-2-SAF-B was initially cloned because of its strong interaction with scaffold/ matrix attachment regions (47). Recently, we described the interaction of SAF-B with RNA polymerase II as well as with splicing factors (7). In addition, SAF-B changed the adenovirus E1A alternative splicing pattern in a concentration-dependent manner (7).
To find new proteins involved in the regulation of alternative pre-mRNA splicing, we used the yeast two-hybrid system to screen an embryonic (E16) rat brain library with rSAF-B (7) as an interacting partner. From about 200,000 colonies screened, six lacZ-positive clones were able to grow in the presence of 5 mM 3-aminotriazole. Two clones contained an open reading frame bearing 67% homology to human Sam68 (27). The predicted protein shares the typical domain structure with members of the STAR family (29) and shows 96% sequence identity to the recently identified mouse proteins etoile/Sam68-like mammalian protein SLM-2 (39 -41). It was therefore named rSLM-2. The protein has a maxi-KH homology RNA binding motif (shadowed box, Fig. 1). Similar to Sam68, rSLM-2 has a central region containing multiple arginine/glycine (RG) dipeptides, followed by a stretch of six prolines which matches the class I consensus SH3 domain binding site (48). The carboxyl- terminal part of rSLM-2 is rich in tyrosine residues and represents a potential SH2 domain binding site (48,49).
Identification of Novel Interactors of rSLM-2 Using the Yeast Two-hybrid Screen-To obtain information on the function of rSLM-2, we performed yeast two-hybrid screens with rSLM-2 as an interacting partner in an embryonic rat brain library. It has been shown previously that SLM-2 interacts with the highly homologous protein Sam68 (39), with itself as well as with hnRNP G family members (40). Our two-hybrid results with rSLM-2 confirmed these interactions because we isolated the rat homologs of Sam68 and hnRNP G. In addition, among the lacZ-positive yeast colonies that were able to grow on selective medium containing 5 mM 3-aminotriazol, we identified four novel interactors of rSLM-2: rSAF-B, which has initially been used to clone rSLM-2; the novel splicing-associated protein YT521-B (26,50); hnRNP L (51); and the SR-protein SRp30c (52) ( Fig. 2A). In addition, we tested its interaction with another KH domain containing RNA-binding protein, SF1 (53). Although Sam68 was shown to bind to other KH domaincontaining proteins such as Bicaudal C or Grp33 (54), rSLM-2 did not interact with the splicing factor SF1, indicating the specificity of the observed interactions. . Plus signs indicate growth; minus signs indicate no growth on His Ϫ plates containing 5 mM 3-aminotriazole. None of the constructs showed self-activation. Panel B, coimmunoprecipitations with rSLM-2. EGFP fusion proteins were overexpressed in HEK293 cells and immunoprecipitated using a monoclonal anti-GFP antibody. The immunoprecipitated protein is indicated by a superscript IP. EGFP alone was used as a negative control in each experiment. The immunoprecipitates were analyzed with a polyclonal anti-EGFP antibody (left column) to verify successful immunoprecipitation. The immunoprecipitates were also analyzed with interactor-specific antibodies (middle column) that are listed in the right column. Arrows indicate the protein signals observed. IG, immunoglobulin signal. Immunoprecipitated rSLM-2 interacts with endogenous SAF-B (top row). Immunoprecipitated EGFP-hnRNP L (second row), EGFP-YT521-B (third row), and EGFP-SRp30c (fourth row) bind to FLAG-rSLM-2. Precipitated EGFP-SF1 did not interact with FLAG-rSLM-2 (fifth row, middle column). The ratio of immunoprecipitation to lysate loaded on the gel was 20:1 in each case.
These novel two-hybrid interactions were tested subsequently using coimmunoprecipitations (Fig. 2B). We generated a specific antibody against rSLM-2 which did not cross-react with the related proteins Sam68 or rSLM-1. 2 To exclude nucleic acid-mediated interactions, benzonase was present in all experiments. In addition, the RNA-binding protein SF1 (53) was used as a negative control (Fig. 2B, bottom row). We expressed EGFP-tagged proteins in HEK293 cells and precipitated them using a monoclonal anti-GFP antibody. Overexpressed EGFP was used as a negative control. The efficiency of the immunoprecipitation was analyzed in Western blots with a polyclonal anti-GFP antibody (Fig. 2B, left column). The coimmunoprecipitating proteins and cell lysates were analyzed by Western blot (Fig. 2B, middle column) using the appropriate antibody (right column). Using this method, we found endogenous hSAF-B to bind to rSLM-2 (Fig. 2B, top row). In agreement with previous data (39,40) and our data obtained in yeast, cotransfected hnRNP G and endogenous Sam68 interacted with rSLM-2 (data not shown). Finally, we confirmed the interaction of rSLM-2 with hnRNP L (second row), with the splicing associated protein YT521-B (third row), and the interaction of rSLM-2 with the splicing factor SRp30c (fourth row).
In summary, EGFP-rSLM-2 was able to homomultimerize and to bind to a subset of splicing and splicing-associated factors in the yeast two-hybrid system as well as in coimmunoprecipitations.
rSLM-2 Forms a Complex with SR Proteins in Vivo-It has been shown previously that the rSLM-2 interactor SAF-B binds to htra2-␤1, SRp30c, SF2/ASF, and hnRNP A1 and can change alternative splice sites in vivo (7,55). Therefore, we asked whether rSLM-2 is present in a complex with splicing factors in vivo. Screening of several cell lines revealed that SLM-2 is expressed in HN10 cells. These cells are derived from hippocampal neurons (56) that express SLM-2 (Fig. 4A, Lysate).
We performed an immunoprecipitation in HN10 cell lysates with the pan SR antibody mAb104 (57) in the presence of benzonase and found rSLM-2 to be present in this complex (Fig.  4A, IP). One-third of the corresponding immunoprecipitates were analyzed with the mAb104 antibody ( Fig. 4B) to verify the successful immunoprecipitation. Furthermore, mAb104 did not cross-react with rSLM-2, even when we overexpressed EGFP-rSLM-2 (Fig. 4C). We conclude that endogenous rSLM-2 forms an in vivo complex with factors implicated in pre-mRNA splicing.
rSLM-2 Can Change Alternative Splicing Patterns-The association of rSLM-2 with splicing factors as well with the splicing-associated protein YT521-B and hnRNPs points to a role in pre-mRNA processing. Therefore, we asked whether rSLM-2 could modulate splice site selection in different cell lines in a concentration-dependent manner, analogous to several proteins involved in splicing, such as SR proteins (14,58), hnRNPs (58), SAF-B (7), and YT521-B (26). To investigate this possibility, we employed a CD44 reporter gene, which contains the alternative exon v5 (44). This minigene was transfected with increasing amounts of EGFP-rSLM-2 in HEK293 cells. Vector DNA (pEGFP) was added to ensure that comparable amounts of DNA were transfected in each experiment (59). Increasing the amount of transfected pEGFP-rSLM-2 resulted in an increased incorporation of exon v5, as shown in Fig. 5A, suggesting a rSLM-2 concentration-dependent modulation of splice site selection. Western blots with lysates from the transfections confirmed an increase in EGFP-rSLM-2 expression (data not shown).
Next, we determined the role of various rSLM-2 domains and tested deletion variants for their influence on splice site selection. As shown in Fig. 5B, deletion of either the RG-rich region and the potential SH2 and SH3 binding domains (rSLM-2⌬2), the tyrosine-rich SH2 domain binding site (rSLM-2⌬3), or the KH domain (rSLM-2⌬4) abolished the influence on splice site selection. All deletion variants were expressed at similar levels in the transfected cells (Fig. 5C). We conclude that the ability of rSLM-2 to bind to RNA is necessary for splice site selection. However, in addition to RNA binding, the potential protein interaction sites are also necessary for the splice site switch. This suggests that rSLM-2 acts in a complex with other proteins on pre-mRNA splicing.
We therefore wondered whether proteins binding to rSLM-2 can influence the regulation of alternative splicing of the CD44 minigene and analyzed the interactors SRp30c, hnRNP G, and SAF-B. First, we tested them separately with the CD44 minigene and found that SRp30c slightly repressed exon v5, whereas hnRNP G and SAF-B had no effect on exon v5 usage  6 -10). GST alone was used as a negative control (lanes 11-15). 4. Interaction of endogenous rSLM-2 with splicing factors. Panel A, interaction of endogenous rSLM-2 with SR proteins. Immunoprecipitation using the pan SR protein antibody mAb104 (IP) was performed with lysates from HN10 cells. rSLM-2 was detected with the rSLM-2 antiserum. The ratio of immunoprecipitate to lysate loaded on the gel was 20:1. Panel B, verification of the immunoprecipitation analyzed in panel A. One-third of the mAb104 immunoprecipitate was analyzed with the mAB104 antibody (IP) and compared with the HN10 lysate. Panel C, mAb104 does not cross-react with overexpressed EGFP-rSLM-2. HEK293 cells were transfected with EGFP-rSLM-2. Lysates were probed with the rSLM-2 antiserum (left) or mAb104 (right), indicating that mAB104 does not cross-react with rSLM-2. (Fig. 5D). Then, we analyzed these interactors in the presence of rSLM-2 and found that all of them led to a significant decrease of the exon v5 stimulation by rSLM-2 (Fig. 5E). These data provide evidence that the rSLM-2-dependent inclusion of exon v5 can be antagonized by SRp30c, hnRNP G, and SAF-B.
The rSLM-2-dependent Exon Inclusion Depends on a Purinerich Exonic Splicing Enhancer-Pull-down assays on immobilized RNA with HEK293 cell lysates containing overexpressed rSLM-2 showed that rSLM-2 selectively binds to purine-rich RNA (39). 3 Previously, a purine-rich splice enhancer has been characterized in exon v5 of the CD44 gene using linker-scan mutations (44). Because the analysis of rSLM-2 deletion variants demonstrated the necessity of its RNA binding domain in splice site regulation (Fig. 5B), we were interested in determining the target sequence of the CD44 pre-mRNA. We compared the influence of rSLM-2 on three different CD44 minigene mutants ls8 -ls10 (Fig. 6A). In the ls9 variant, the purine-rich enhancer sequence is replaced. Similar to the wild type minigene, EGFP-rSLM-2 was able to induce exon v5 inclusion with the ls8 and ls10 minigene constructs in HEK293 cells (Fig. 6B).
To determine whether recombinant GST-rSLM-2 binds directly to CD44 exon v5 RNA, we performed gel mobility shift assays with several RNA oligomers. First, we determined that GST-rSLM-2 binds to the purine-rich oligonucleotide r-rich. With increasing amounts of rSLM-2, a super shift is visible which is probably due to the ability of rSLM-2 to multimerize (Fig. 6C). This is consistent with our two-hybrid and immunoprecipitation data and the observation that GST-rSLM-2 migrates as a complex of about 450 kDa upon gel filtration (data not shown). Next, we employed the oligonucleotide v5, which contains the regulatory sequence of the CD44 exon v5 and observed binding of rSLM-2 to it (Fig. 6D, lanes 1-3). Then we used an oligonucleotide (v5ls9) containing the mutation ls9, which abolished the effect of rSLM-2 on CD44 exon v5 regulation in vivo (Fig. 6B). As shown in Fig. 6D, lanes 4 -6, no binding is observed with the v5Rls9 oligonucleotide. Furthermore, we observed binding to the purine-rich oligonucleotide 3 (44). An increasing amount of pEGFP-rSLM-2 was cotransfected with 2 g of the CD44v5 minigene. In each titration point, pEGFP was added to keep the total amount of DNA constant. Panel B, the CD44v5 minigene was cotransfected with 2 g of pEGFP, pEGFP-rSLM-2, or its deletion variants (shown on the left). Panel C, Western blot analysis demonstrating equal amounts of pEGFP-rSLM-2 and its deletion variants in cellular lysates of the transfections. 20 l of the lysates was analyzed with an anti-GFP antibody. Panel D, the CD44v5 minigene was cotransfected with 2 g of the rSLM-2 interactors SRp30c, htra2-␤1, and SAF-B to determine their influence on exon v5 regulation. Panel E, the CD44v5 minigene was cotransfected with 2 g of rSLM-2 in the presence of 2 g of its interactors. r-rich (Fig. 6D, lanes 7-9), but no specific binding to the pyrimidine-rich oligonucleotide y-rich could be detected (Fig. 6D,  lanes 10 -12). No binding of these oligonucleotides was observed when GST was used (Fig. 6D, lanes 13-16). We conclude that rSLM-2 regulates inclusion of CD44 exon v5 by binding to a purine-rich exonic enhancer (see Fig. 8).
Effect of rSLM-2 on Other Minigenes-Next, we tested the effect of rSLM-2 on the alternative splicing of the htra2-␤1 gene in HN10 cells (23). Transfection of pEGFP alone leads to 40% inclusion of alternative exon 2. Increasing amounts of rSLM-2 caused a decrease in the ␤4 isoform and led to an increase in the htra2-␤1 isoform (Fig. 7A).
Next, we analyzed the effect of rSLM-2 expression on a tau minigene in COS cells. The tau minigene contains the alternatively spliced tau exon 3 (SV⌬2/3) inserted in an insulin expression vector (45). Exons 2 and 3 exhibit the rare incremental combinatorial splicing pattern because exon 3 never appears independently of exon 2. Exons 2 and 3 are adult-specific in the central nervous system (60) but seem to be constitutive in the peripheral nervous system (61,62). Upon cotransfection, ex- The locations of the 10-base pair linker scan mutations ls8, ls9, and ls10 within exon v5 are indicated (44). In the mutant ls9 the purine-rich sequence ATGAAGAGGA is replaced by CGACGCGTCG. Panel B, the effect of EGFP-rSLM-2 on three different linker scan mutations (ls8 -10) in exon v5 of the CD44 minigene has been compared with the effect of EGFP. With the ls8 and ls10 minigenes, 2 g of transfected EGFP-rSLM-2 increased exon v5 inclusion similar to the wild type minigene. In contrast, only a minor effect of EGFP-rSLM-2 was detected with the ls9 construct. HEK293 cells were used for transient transfections. Panel C, recombinant rSLM-2 binds to RNA in vitro. An increasing amount of GST-rSLM-2 was incubated with a 32 P-labeled polypurine RNA oligonucleotide (r-rich) and separated on a nondenaturing gel. The RNA⅐rSLM-2 complex is indicated by a closed arrow, a larger complex observed at higher rSLM-2 concentration with an open arrow. Panel D, recombinant rSLM-2 binds the purine-rich exon v5 enhancer in vitro. 30 pmol of GST-rSLM-2 was incubated with the RNA oligonucleotides v5, v5ls9, r-rich, and y-rich. Each oligonucleotide was competed with the 20-and 40-fold excess of unlabeled v5 oligonucleotide (lanes 2 and 3; 5 and 6; 8 and 9; 11 and 12). Upon longer exposure, binding of the v5ls9 oligonucleotide is observed only in the absence of competitor. The dot indicates an unspecific complex seen with the y-rich oligonucleotide that is competed by excess of r-rich oligonucleotide. No shift is observed with GST (lanes [13][14][15][16] or without protein (lanes 17-20).
pression of rSLM-2 stimulates exon 3 usage, whereas EGFP alone has no effect (Fig. 7B). Interestingly, a sequence stretch in exon 2 of htra2-␤ and a sequence stretch in exon 3 of the tau transcript show a significant homology to the purine-rich sequence of exon v5 of the CD44 minigene, which is necessary for the rSLM-2-dependent exon inclusion.
Finally, overexpression of rSLM-2 did not stimulate inclusion of the pyrimidine-rich neuron-specific exon EN of the clathrin light chain B transcript (data not shown) and did not repress exon EN usage when this exon is activated by improving its 5Ј-splice site (Fig. 7C). In contrast, htra2-␤1 can regulate splice site usage of this construct (Fig. 7C), which emphasizes the specificity of rSLM-2-mediated splice site selection. Taken together, this indicates that rSLM-2 can change the alternative splicing pattern of specific substrates in a concentration-dependent manner by binding to purine-rich sequences. DISCUSSION

rSLM-2 Interacts with Proteins Involved in Splice Site
Selection-SAF-B has been shown to associate with the splicing factors SRp30c, ASF/SF2, and htra2-␤1 as well as with the carboxyl-terminal domain of the largest subunit of RNA polymerase II, which has also been shown to stimulate pre-mRNA splicing (7,63). In addition, SAF-B binds to matrix attachment regions and stimulates the generation of the 10S E1A transcript in a concentration-dependent manner (7).
To identify novel proteins involved in mRNA processing, we performed a yeast two-hybrid screen using SAF-B as a bait. This led to the identification of rSLM-2 as a novel interactor of SAF-B.
rSLM-2 and its close relatives Sam68 and rSLM-1 are members of the STAR protein family (29), also called GSG protein family (30). These nuclear proteins combine a slightly modified FIG. 7. The influence of rSLM-2 on the alternative splicing of the htra2-␤, tau, and clathrin light chain B pre-mRNA. The structure of the minigenes is shown on the top of each panel. A statistical evaluation of three independent experiments is shown below representative ethidium bromide gels. Error bars indicate the S.D. Panel A, an increasing ratio of pEGFP-rSLM-2 to pEGFP was cotransfected with 2 g of the htra2-␤ minigene in HEK293 cells. This increase leads to exon II skipping. Panel B, the influence of rSLM-2 concentration on tau minigenes was tested in COS7 cells. 4 g of the tau minigenes was cotransfected with indicated amounts (g) of pEGFP-rSLM-2. rSLM-2 leads to an increase in exon 3 inclusion. X, phiX HaeIII digest; M, 100-bp ladder. Panel C, up to 4 g of pEGFP-rSLM-2 did not promote exon skipping of clathrin light chain B exon EN in the construct pJS85. In pJS85, exon EN is activated in fibroblasts by an optimized 5Ј-splice site (24). In contrast, this exon is regulated by htra2-␤1 (right). hnRNP K homology RNA binding domain, called maxi-KH domain, with binding sites for proteins containing phosphotyrosine or proline binding domains like SH2, SH3, or WW domains. To investigate the function of rSLM-2, we searched for novel interactors using yeast two-hybrid screens and further confirmed these interactions with coimmunoprecipitations and in vitro interaction assays using recombinant rSLM-2 protein.
Interestingly, all of the rSLM-2-interacting proteins found within these three independent systems are implicated in the regulation of alternative splicing. In addition to SAF-B, which has initially been used to isolate rSLM-2, the splicing-associated protein YT521-B has been found to interact with rSLM-2. Using the SRp20 and htra2-␤ minigenes, we recently demonstrated the influence of YT521-B on alternative splicing (26). Like rSLM-2, YT521-B interacts with Sam68 and is tyrosinephosphorylated upon overexpression of the tyosine kinases Src and Fyn (26). It remains to be determined whether rSLM-2 is tyrosine-phosphorylated by nuclear tyrosine kinases, such as SIK/BRK, which was shown to phosphorylate Sam68 (64,65).
It is notable that rSLM-2, SAF-B, and YT521-B all interact with the splicing factor SRp30c. The physiological significance of this complex formation of rSLM-2 has been demonstrated by endogenous coimmunoprecipitation using the pan anti-SR protein antibody mAb104 (Fig. 4). However, in contrast to SAF-B and YT521-B, no direct binding could be detected between rSLM-2 and the splicing factors htra2-␤1 ( Figs. 2A and 3) or SF1 (Fig. 2). This demonstrates the specificity of the analyzed interactions, which have been carried out in the presence of benzonase to avoid RNA-dependent protein interactions.
Finally, a growing number of hnRNPs have been characterized as molecular players in the regulation of alternative splicing, among them hnRNP A1 (58), hnRNP F, hnRNP H (15), hnRNP I (polypyrimidine tract binding) (16), hnRNP L (66), and a testis-specific member of the hnRNP G family, RNAbinding motif (RBM) (67,68). Using a two-hybrid screen with RBM as a bait, Venables et al. were previously able to isolate the human homologue of rSLM-2, T-STAR (40). Now we provide evidence that rSLM-2 also interacts with the ubiquitously expressed hnRNP G and with hnRNP L.
rSLM-2 Regulates Alternative Splice Site Selection-The binding properties of rSLM-2 indicate a role in pre-mRNA processing. We tested this hypothesis by performing transient transfection assays with several reporter minigenes that contained purine-rich sequences in their alternative exons. As a control, a pyrimidine-rich exon of clathrin light chain B was employed. We found that rSLM-2 changes the splicing patterns of several alternatively spliced exons in a concentration-dependent manner. It induced inclusion of exon v5 of CD44 (44) (Fig. 5A), of exon 3 of neurofilament tau (Fig. 7B) (69), but caused exon II skipping of the htra2-␤ exon II (23) (Fig. 7A). The opposing effect of rSLM-2 on different minigenes is reminiscent of the effects of splicing factors on natural minigenes in vivo. For example, SF2/ASF promotes exon EN inclusion when tested with the clathrin light chain B minigene (58) but causes exon 4 skipping with the SRp20 minigene (70).
Our deletion analysis showed that binding of rSLM-2 to RNA is not sufficient to influence splice site selection. The deletion mutant lacking a functional KH domain rSLM-2(⌬4) is exclusively nuclear (data not shown). In contrast, mutants lacking parts of the carboxyl terminus (rSLM-2⌬2 and ⌬3) are also present in the cytosol (data not shown). This indicates that potential protein binding sites such as the tyrosine-rich carboxyl terminus or the proline-rich region are also necessary for activity and proper localization of rSLM-2. Interestingly, the rSLM-2-interacting proteins SAF-B, hnRNP G, and SRp30c inhibited exon v5 inclusion mediated by rSLM-2 (Fig. 5D), although these three interactors alone exerted only a minor or no effect on the CD44 minigene. Because SAF-B, hnRNP G, and SRp30c are ubiquitously expressed but rSLM-2 is predominantly expressed in muscle, brain, and testis (40 and data not shown), this suggests that cell type-specific combinations of rSLM-2 and some of its interactors may contribute to different splicing patterns in different cell types. This supports the model that different concentrations of antagonizing factors govern splice site selection.
Binding of rSLM-2 to a purine-rich RNA sequence is necessary to regulate splice site selection because either deleting its RNA binding domain or changing the purine-rich sequence abolishes an effect on usage of exon v5 (Figs. 5B and 6B).
Because rSLM-2 binds directly to purine-rich RNA (Fig. 6, C and D), it is likely that it regulates splice site selection through interaction with purine-rich sequences in vivo. The in vitro binding of rSLM-2 to exon v5 RNA, but not to the mutant v5ls9 ( Fig. 6D) is in complete agreement with the regulation of CD44 exon v5, but not CD44 exon v5ls9 by rSLM-2 (Fig. 6B). In addition, rSLM-2 cannot regulate the pyrimidine-rich exon EN of clathrin light chain B (Fig. 7C), even when this exon is activated by a 5Ј-splice site improvement. A sequence comparison between the regulated exons of CD44, tra2-␤1 and tau (Fig. 8) also suggests a purine-rich motif as the likely site of action.
Our data provide for the first time evidence for a role of a STAR protein in the regulation of alternative splicing. In addition, there is increasing evidence that the STAR protein Sam68 is also involved in alternative splice site selection. First, Sam68 has been found to cross-link to an intronic regulatory RNA sequence of the tropomyosin pre-mRNA (71). Second, Bedford et al. recently demonstrated the binding of Sam68 to the spliceosome-associated protein FBP21 (72). Finally, like rSLM-2, Sam68 interacts with the testis-specific splicing factor RBM (67). However, a direct effect of Sam68 on alternative splice site selection remains to be shown.
In addition to the role of Sam68 in viral replication (73,74), two possible functions of rSLM-2 and Sam68 have been investigated to date. First, there is evidence that rSLM-2 and Sam68 are important for cell cycle progression (27,28,36,41), but a direct molecular link is still missing. We speculate that STAR proteins influence the cell cycle by regulating mRNAs necessary for its progression.
Furthermore, Sam68 has been proposed to act as an adapter protein within signal transduction pathways. Upon T cell or insulin receptor stimulation, Sam68 is tyrosine-phosphorylated (33)(34)(35). Phosphorylation changes its binding affinities to phospholipase C␥1, to the regulatory p85 subunit of phosphatidylinositide 3-kinase (32), to itself, and to RNA (54,75). The p85 phosphatidylinositide 3-kinase also binds to the human homolog of rSLM-2 (41). There are numerous examples for the regulation of alternative splicing by extracellular stimuli (for review, see Ref. 76). Because this list of stimuli which includes insulin (77), nerve growth factor (78), cytokines (79), or neuronal activity (25) is rapidly growing, we are now investigating whether rSLM-2 and similar adapter proteins are part of a signal transduction cascade from receptors toward the spliceosome.
Acknowledgments-We thank Claudia Cap for sequencing and James Chalcroft for artwork.