Yeast two-hybrid assays.

The cDNA fragments for full-length RNPS1 and RNPS1-ΔRS/P (Table ​(Table1)1) were fused in-frame to the LexA DNA-binding domain of the pLexA vector to obtain pLexA-RNPS1 and pLexA-RNPS1-ΔRS/P, respectively. A human HeLa cell cDNA library subcloned into the B42 activation vector pB42AD (Clontech) was screened as described in the manufacturer's protocol. Briefly, Saccharomyces cerevisiae EGY48 harboring the β-galactosidase reporter plasmid p8op-lacZ (EGY48[p8op-lacZ]) was cotransformed with the pB42AD-human HeLa cDNA library and either pLexA-RNPS1 or pLexA-RNPS1-ΔRS/P. Positive clones were identified by galactose-dependent growth in medium lacking leucine and confirmed by β-galactosidase activity of the colonies. Duplicate clones were eliminated by comparing the lengths of their cDNA inserts amplified by PCR with two primers, pB42AD-F (5′-CCAGCCTCTTGCTGAGTGGAGATG-3′) and pB42AD-R (5′-GGCAAGGTAGACAAGCCGACAACC-3′), and by their Sau3AI-digested fragments.

To eliminate false-positive clones, plasmids from putative positive clones were transformed and tested again for galactose-dependent growth in medium lacking leucine and for β-galactosidase activity. pLexA-Lam (which encodes human lamin C_66-230 in pLexA; Clontech) and the pLexA vector were used as negative controls. The cDNA inserts in the selected positive plasmids were sequenced with the pB42AD-F primer. Independent transformants (~4 × 10⁶) were screened by their ability to grow in the absence of leucine as well as their β-galactosidase activity. A total of 157 colonies were isolated from the initial screening. After false-positive and duplicate clones were eliminated, 20 clones were shown to activate the reporters in combination with LexA-RNPS1 or LexA-RNPS1-ΔRS, but not with LexA alone or LexA-lamin. The sequences of each clone (~200 bp) were used to search sequence databases.

Preparation of recombinant proteins.

Expression and purification of glutathione S-transferase (GST) and GST fusion proteins were performed with Escherichia coli strain BL21-CodonPlus (DE3)-RP (Stratagene) following the manufacturer's instructions (Amersham). The E. coli strain JM101 containing pGEX-hTra2β was a gift from J. L. Manley, and expression and purification of the protein were performed as described (78).

Cell culture and transient transfections.

Adherent HeLa cells were cultured as described (38). For transient transfection, cells were plated at density of ~3.0 × 10⁵ cells per 3.5-cm dish or ~2.0 × 10⁶ cells per 10-cm dish the day before transfection. Each expression plasmid was transfected with Lipofectamine 2000 (Invitrogen) in serum-free Opti-MEM (Invitrogen) as described in the manufacturer's protocol. After 4 h incubation, Opti-MEM with 20% fetal bovine serum (FBS) was added to the medium to a final 10% FBS concentration. Approximately 85% of HeLa cells were transfected, as determined by expression of transfected pFLAG-GFP under a fluorescent microscope. Ehrlich ascites tumor cells (suspension culture) were transiently transfected by electroporation as described (74).

Preparation of whole HeLa cell extracts.

HeLa cells were transiently transfected with 15 μg of either pFLAG-p54, pFLAG-hTra2β, or pFLAG-GFP plasmids in 10-cm dishes. At 24 h after transfection, cells were lysed in 150 μl of lysis buffer [20 mM HEPES-NaOH (pH 7.3), 3.2 mM MgCl_2, 150 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 1 mM dithiothreitol (DTT), 1 mM phenylmethylsulfonyl fluoride] by freezing and thawing. The lysate was centrifuged at 20,000 × g, and the supernatant was used for immunoprecipitation assays.

In vitro-coupled transcription-translation.

In vitro translation was performed with the TNT T7 coupled wheat germ extract (see Fig. ​Fig.1A),1A), and reticulocyte lysate (see Fig. ​Fig.1C)1C) systems (Promega). Briefly, 0.5 to 1 μg of the E. coli expression plasmid pSBETc-RNPS1 (59) and the control luciferase T7 plasmid were incubated (in 50 μl) with TNT master mixture with 20 μCi of [³⁵S]methionine (Amersham) for 2 h at 30°C. The translation efficiency was checked by analyzing 2 μl of lysate on a 10% SDS-PAGE gel.

Open in a separate window

FIG. 1.

In vitro binding assays of RNPS1 to pinin, hTra2β, and p54 proteins. (A) The indicated recombinant GST-tagged fusion proteins coupled to glutathione Sepharose beads were incubated with in vitro-translated ³⁵S-labeled RNPS1 or luciferase (Luc) as a control, and the pelleted beads were washed and analyzed by 10% SDS-PAGE (GST pull-down assays). The upper panel shows the autoradiograph. A portion (one fifth) of each ³⁵S-labeled protein used in the pull-down reactions was loaded as the input control (Input). Marker protein sizes are indicated on the left. The lower panel shows a 10% SDS-PAGE gel (stained with Coomassie blue) that was used for autoradiography. The corresponding proteins are indicated with arrowheads. (B) Expression of FLAG-tagged constructs (GFP, p54, and hTra2β) in HeLa whole-cell extracts. The extracts (5 μl each) were separated by 10% SDS-PAGE and detected by immunoblotting with anti-FLAG antibody. (C) Immunoprecipitation analyses of ³⁵S-labeled RNPS1 and luciferase (Luc) as a control with the indicated expressed FLAG-fused proteins (GFP as a control). Anti-FLAG antibody immobilized on protein G-Sepharose was used for the immunoprecipitation. Immunoprecipitated ³⁵S-labeled RNPS1 protein was detected by autoradiography. A portion (one fifth) of each ³⁵S-labeled protein used in the immunoprecipitation was loaded as control (Input).

GST pull-down and immunoprecipitation assays.

GST-tagged fusion proteins (pinin-C23, hTra2β, or GST alone for a control) coupled to glutathione Sepharose beads (Amersham) were incubated for 120 min at 4°C with 1 μl of in vitro-translated RNPS1 (or luciferase as a control) in a total volume of 300 μl with binding buffer A (20 mM HEPES-KOH [pH 7.3], 3.2 mM MgCl_2, 150 mM NaCl, 1 mM EDTA, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 33.3 U of DNase I [Promega] per ml, 10.7 μg of RNase A [Worthington] per ml, and 0.5 mg of bovine serum albumin [Sigma] per ml). After incubation, the beads were washed four to five times with binding buffer A. The pelleted beads were heated at 92°C for 5 min with 15 μl of SDS sample buffer. After removing insoluble material by microcentrifugation, the proteins were separated on a 10% SDS-PAGE gel.

FLAG-tagged cDNA constructs (p54, hTra2β, or GFP as a control) were transiently transfected into HeLa cells and their expression was checked by immunoblotting. Immunoblotting was performed as described (59) with anti-FLAG M2 monoclonal antibody (Sigma) and anti-mouse IgG horseradish peroxidase-linked whole antibody (Amersham). The immunoreactive proteins were detected by chemiluminescence according to the manufacturer's protocol (ECL detection reagents; Amersham). A washed protein G-Sepharose slurry (20 μl = 10 μl of beads; Amersham) was first added to the FLAG-tagged expressed-protein extracts and rocked at 4°C for 1 h to preadsorb nonspecific protein G binding proteins. After brief microcentrifugation, the supernatants were recovered and 3 μg of anti-FLAG M2 monoclonal antibody, 3 μl of in vitro-translated RNPS1 (or luciferase as a control), and another 20 μl of fresh protein G-Sepharose beads mixture were added in a total volume of 300 μl with binding buffer A. Incubation was continued at 4°C with rocking for 2 h. The protein G-Sepharose beads were washed four times in 500 μl of binding buffer A, and the bound proteins were analyzed on a 10% SDS-PAGE gel as described above.

The ³⁵S-labeled RNPS1 in the SDS-polyacrylamide gel was analyzed with a phosphorimager (Storm 840; Molecular Dynamics) or a Bio-imaging analyzer (BAS2000; Fujifilm).

In vivo splicing and RT-PCR analyses.

The reporter plasmids expressing pre-mRNA substrates were as follows; pHD1012-DUP51 (a gift from R. Kole; 30) for duplicated human β-globin pre-mRNA (DUP51), pCR3.1-hTra2-β1 (a gift from J. L. Manley) (66) for human transformer-2-β1 minigene (hTra2β), and pCMV-F₁γEx8-10 (38) for human mitochondrial ATP synthase γ-subunit minigene (F₁γ). HeLa cells were transiently transfected in a 3.5-cm dish with 2 μg of the reporter plasmids and various amounts of protein expression plasmids; pFLAG-p54 (0 or 2 μg), pFLAG-hTra2β (0 or 2 μg), pCMV-3×FLAG-RNPS1 (0, 0.5, or 2 μg), pCMV-3×FLAG-RNPS1-ΔS (0, 0.5, or 2 μg), pCMV-3×FLAG-RNPS1-ΔRS/P (0, 0.5, or 2 μg), pCMV-3×FLAG-RNPS1-mtRRM (0, 0.5, or 2 μg), and pCMV-3×FLAG-RNPS1-RS/P (0, 0.5, or 2 μg). The final amount of transfected DNA was kept constant at 6 μg by addition of control pFLAG-CMV-2 vector. At 24 h after transfection, cells were harvested and total RNA was isolated with Trizol reagent (Invitrogen) as described in the manufacturer's protocol. The isolated total RNA was treated with DNase I for 1 h, denatured at 65°C for 10 min, and immediately chilled on ice prior to reverse transcription (RT).

RT-PCR was used to detect in vivo splicing products. First strand cDNA was synthesized (in 10 μl) by Superscript II (50 units; Invitrogen) for 50 min at 42°C with 400 ng of total RNA in 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl₂, 10 mM DTT, 0.5 mM each of each deoxyribonucleoside-5′-triphosphate (dNTP), and either 0.1 mM oligo(dT)₁₆ primer or 50 nM specific RT primer. The RT primers used were Glo-E2 (5′-CAGGTGAGCCAGGCCATCACTAAAG-3′) for DUP51 substrate, pCR3.1RT (5′-GCCCTCTAGACTCGAGCTCGA-3′) for hTra2β, and oligo(dT)₁₆ primer for F₁γ substrate. PCR amplification was performed (in 25 μl) with 1 μl of first-strand cDNA mixture, and 1 unit of Taq DNA polymerase (Takara) in 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl₂, 0.2 mM each of each dNTP, and 0.2 μM each of the forward and reverse PCR primers. The PCR primers used were DUP-F (5′-CACTAGCAACCTCAAACAGACACCATGC-3′) and DUP-R (5′-ATGAGCCTTCACCTTAGGGTTGCCC-3′) for DUP51 substrate, MGTraBam (5′-GGGGATCCGACCGGCGCGTCGTGCGGGGCT-3′) and MGTraXho (5′-GGGCTCGAGTACCCGATTCCCAACATGACG-3′) for the hTra2β substrate, and the sense and antisense primers (in SP6 promoter and exon 10, respectively) for the F₁γ substrate were previously described (38). The PCR conditions for each cycle were 94°C for 20 s, 62°C for 30 s, and 72°C for 1 min, and a total of 30 cycles were performed. The amplified PCR products were analyzed by electrophoresis on a 3% agarose gel and visualized by ethidium bromide staining. The essential RT-PCR experiments were repeated at least three times, and all these data gave consistent results.

RNPS1 binds to hTra2β, pinin, and p54 in vitro.

To confirm physical binding of RNPS1 to pinin and hTra2β, in vitro translated ³⁵S-labeled RNPS1 was subjected to pull-down assays with recombinant GST-pinin and GST-hTra2β fusion proteins in the presence of DNase and RNase. Since full-length recombinant GST-pinin could not be expressed in E. coli, we used the second and third coiled-coil domains of pinin (pinin-C23), which include the minimal region shared by all of five selected pinin cDNA clones (Table ​(Table2).2). We found that GST-pinin-C23 and GST-hTra2β coprecipitated with RNPS1 (Fig. ​(Fig.1A,1A, lanes 4 and 6), but not with the control luciferase (lanes 5 and 7), or GST alone (lane 3). Since recombinant GST-p54 could not be expressed in E. coli, we performed immunoprecipitations with extracts from HeLa cells in which FLAG-tagged p54 was overexpressed (Fig. ​(Fig.1B).1B). After addition of ³⁵S-labeled RNPS1 to the HeLa cell extract, p54 was immunoprecipitated with an anti-FLAG antibody in the presence of DNase and RNase. The coprecipitated RNPS1 was analyzed by SDS-PAGE followed by autoradiography. RNPS1, but not control luciferase, coimmunoprecipitated with p54, but not with GFP as a negative control (Fig. ​(Fig.1C,1C, lanes 3 to 5). Consistent results, as in the GST pull-down assays, were obtained with HeLa cell extracts expressing FLAG-tagged hTra2β (lanes 6 and 7). We conclude that RNPS1 can bind to p54, hTra2β and the coiled-coil domain of pinin via protein-protein interactions in vitro.

p54, hTra2β, and pinin interact with different domains of RNPS1.

To map the domains of RNPS1 that are responsible for its interaction with p54, hTra2β, and pinin, respectively, deletion mutants of RNPS1 were used in the yeast two-hybrid assays (Fig. ​(Fig.2A).2A). RNPS1-NT+S (Table ​(Table1)1) and RNPS1-ΔRS/P (Table ​(Table1),1), both of which possess the S domain, interacted with p54 (Fig. ​(Fig.2A).2A). In contrast, RNPS1-RRM+RS/P (Table ​(Table1)1) and RNPS1-RS/P (Table ​(Table1),1), which share the RS/P domain, interacted with hTra2β. RNPS1-ΔRS/P and RNPS1-RRM+RS/P, which share the RRM, interacted with pinin. These data indicate that distinct domains of RNPS1 are responsible for the interactions with the different factors (Fig. ​(Fig.2B).2B). Previously, the N-terminal region of RNPS1, which includes the S domain, was shown to be necessary for the interaction with PITSLRE p110 kinase and a tumor-rejection antigen, SART3 (37, 56).

Open in a separate window

FIG. 2.

Protein interaction domains of RNPS1 analyzed by yeast two-hybrid assays. (A) Yeast host strain EGY48 was cotransformed with the indicated combinations of RNPS1 deletion constructs (in pLexA; see Table ​Table1)1) and selected clones (p54, hTra2β, and pinin in pB42AD). As a negative interaction control, the empty vector pB42AD (vec.) was used. Protein interactions were detected by β-galactosidase activity and auxotrophy for leucine. (B) A model of protein interactions of RNPS1 with p54, pinin, and hTra2β is shown.

RNPS1 localizes to nuclear speckles in living cells.

It was previously shown that human RNPS1 often localizes to enlarged (or mega) nuclear speckles when transiently overexpressed, and its distribution partially overlaps that of SC35 (an SR protein) in a typical nuclear speckle pattern (56). Here, we examined the expression pattern of mouse RNPS1 (99% identical to human RNPS1) in living Ehrlich ascites cells (Fig. ​(Fig.3A).3A). EGFP-RNPS1 exhibited exclusive nuclear localization with a canonical speckled pattern, while the control EGFP distributed throughout the whole cell (data not shown). In agreement with this observation, mouse RNPS1 was shown to be enriched in biochemically purified interchromatin granule clusters (part of the structural component of the nuclear speckles), which also contain many known splicing factors, including snRNPs and SR proteins (62). Continued observation of the transfected living cells showed dynamic intranuclear movements and redistribution of the speckled particles (data not shown).

Open in a separate window

FIG. 3.

Subcellular localization of WT RNPS1 and deletion mutants, p54, and hTra2β. (A) EGFP-fused mouse RNPS1 was transiently transfected into living Ehrlich ascites tumor cells and analyzed by fluorescence microscopy (24 h after transfection). A phase contrast micrograph taken at the same time is shown below. (B) The GFP fusion constructs with either NT or RS/P domain (GFP alone as a control) were transiently transfected into HeLa cells. The cells were fixed and permeabilized at 12 h after transfection and analyzed by fluorescence microscopy. (C) The DsRed fusion proteins of human RNPS1 (WT) and two deletion mutants, RNPS1-ΔS and RNPS1-ΔRS/P, were transiently cotransfected into HeLa cells with FLAG-tagged p54 and FLAG-tagged hTra2β. The cells were fixed and permeabilized at 15 h after transfection, treated with an anti-FLAG antibody, and analyzed by confocal microscopy. Both images were also superimposed, and the yellow color indicates colocalization of the two proteins in nuclear speckles (Merge). Overexposed images (black and white) are also shown to highlight the cytoplasmic localization that is specifically observed in RNPS1-ΔRS/P protein. Bar, 10 μm.

C-terminal but not N-terminal NLS of RNPS1 may be functional.

Previously, a putative simian virus 40 T antigen-type nuclear localization signal (NLS) (amino acids 278 to 287) was reported in RNPS1 (2), and if it is extended to amino acid 296 it matches the consensus bipartite-type NLS (reviewed in references 28 and 29). We also found a bipartite type-NLS in the N-terminal region (amino acids 19 to 47) of RNPS1, which is predicted by PROSITE motif search (32).

To test the function of these NLS candidates, two RNPS1 domain-GFP fusion constructs were transiently expressed in HeLa cells and their subcellular localization was examined (Fig. ​(Fig.3B).3B). The NT-GFP fusion protein (Table ​(Table1),1), which has the N-terminal putative NLS, showed both nuclear and cytoplasmic localization, as observed with the control GFP protein. In contrast, RS/P-GFP protein (Table ​(Table1),1), which includes the C-terminal putative NLS, reveals exclusive nuclear localization. These data suggest that the C-terminal, but not N-terminal, NLS (amino acids 278 to 287 or 296) functions as an active nuclear localization signal.

p54 and hTra2β colocalize with RNPS1 in nuclear speckles.

To investigate the colocalization of RNPS1-associated factors with RNPS1, p54 and hTra2β cDNA expression plasmids were transiently transfected into HeLa cells (we did not detect expression of the full-length pinin cDNA). The localization of DsRed-RNPS1 and epitope-tagged proteins FLAG-p54 and FLAG-hTra2β, was analyzed by confocal laser microscopy with a red fluorescent protein and an antibody against the FLAG tag. All three fusion proteins localized in nuclear speckles (Fig. ​(Fig.3C,3C, column WT) and these distribution patterns are consistent with previous immunofluorescence studies (21, 59, 66). While RNPS1 and p54 showed relatively little signal in the nucleoplasm, hTra2β was found somewhat more prominently in the nucleoplasm. The merged images revealed that RNPS1 colocalizes with both p54 and hTra2β in nuclear speckles.

We examined the role of two specific RNPS1-domains in the subcellular localization of the protein. To this end, we constructed two deletion mutants; lacking either the S domain (RNPS1-ΔS; Table ​Table1)1) or the RS/P domain (RNPS1-ΔRS/P; Table ​Table1),1), which were fused to the C terminus of DsRed, and transiently coexpressed together with either FLAG-p54 or FLAG-hTra2β in HeLa cells. The RNPS1-ΔS protein localized mostly in the nuclear speckles. However, additional numerous small foci that did not necessarily colocalize with p54 (see red spots in the merged image) were found dispersed throughout the nucleoplasm (Fig. ​(Fig.3C,3C, column ΔS). In contrast, cells coexpressing hTra2β displayed colocalization of this protein with RNPS1-ΔS in both speckles and small nucleoplasmic foci (see the merged image). Overexpressed RNPS1-ΔRS/P was predominantly found in enlarged nuclear speckles (column ΔRS/P), however, some of the large spots were also detected in nucleoli (see the merged image). Consistent with the presence of C-terminal functional NLS in the RS/P domain, a small amount of the ΔRS/P protein was often observed in the cytoplasm, which was never observed with the wild-type RNPS1 and RNPS1-ΔS proteins (see overexposed). RNPS1-ΔRS/P did not colocalize well with hTra2β (see red spots in the merged image). These observations are generally consistent with our yeast two-hybrid results obtained with deletion mutants of RNPS1 (Fig. ​(Fig.2B2B).

RNPS1 promotes exon skipping in vivo.

RNPS1 and SF2/ASF synergistically stimulate both constitutive and alternative splicing in vitro, and thus RNPS1 was characterized as a general splicing activator (59). As RNPS1 was purified based on the activation of distal alternative 3′ splice sites (SF7 activity), we assumed that factor(s) associated with RNPS1 may be necessary for the SF7 activity that was observed in partially purified fractions. Here, we examined the activity of RNPS1 with its associated factors in the regulation of alternative splicing in vivo.

cDNAs encoding these protein factors were transiently overexpressed in HeLa cells, and the splicing patterns of three reporter minigenes containing alternatively spliced exons were examined by RT-PCR (Fig. ​(Fig.4).4). We first tested the exon/inclusion-skipping model substrate DUP51, which consists of three exons and two identical introns of the human β-globin gene (30). Besides a majority of middle-exon inclusion product (271 nucleotides), we also observed a middle-exon skipping product (220 nucleotides) specifically upon overexpression of either RNPS1 or p54 (Fig. ​(Fig.4A,4A, lanes 1 to 3, 12). This p54 effect is consistent with the previous in vivo splicing results, i.e., p54 stimulates the use of the distal 5′ splice site in E1A pre-mRNA (86). The activity of RNPS1 in promoting exon skipping was verified with a natural, alternatively spliced substrate, human Tra2-β1 (hTra2β) pre-mRNA (1, 66). Overexpression of RNPS1 induced double-exon skipping in hTra2β pre-mRNA (Fig. ​(Fig.4B,4B, lanes 1 to 3; beta 3 product). In the human mitochondrial ATP synthase γ-subunit minigene (F₁γ) pre-mRNA (38), however, RNPS1 could not induce exon skipping (muscle type). Rather, we observed decreasing products of both exon skipping and exon inclusion (Fig. ​(Fig.4C,4C, lanes 1 to 3). In these transfections, we observed increasing unspliced pre-mRNA by RT-PCR with two primers between the 5′ end of exon 8 and the downstream intron (data not shown), and therefore the decreasing RT-PCR signal of spliced products indicate the actual repression of splicing. We also took internal control of the RT-PCR (under the same condition) by detecting endogenous β-actin mRNA, and we did not observe any changes of RT-PCR products (data not shown). We conclude that RNPS1 promotes exon skipping in a substrate-specific manner in vivo.

Open in a separate window

FIG. 4.

Effect of transiently overexpressed RNPS1, p54, and hTra2β on three kinds of alternatively spliced pre-mRNAs. (A) In vivo splicing of a model β-globin DUP51 pre-mRNA. (B) In vivo splicing of an hTra2β minigene pre-mRNA (C) In vivo splicing of a human F₁γ minigene pre-mRNA. The structures of the pre-mRNAs and the spliced products are schematically illustrated on the top of each panel (numbers denote length in nucleotides). HeLa cells were transfected with a constant amount of individual pre-mRNA reporter plasmid along with a constant (indicated by rectangles) or increasing (indicated by triangles) amounts of protein expression plasmids pCMV-3×FLAG-RNPS1 (wild type), pCMV-3×FLAG-RNPS1-ΔS (ΔS), pCMV-3×FLAG-RNPS1-ΔRS/P (ΔRS/P), pCMV-3×FLAG-RNPS1-mtRRM (mtRRM), pCMV-3×FLAG-RNPS1-RS/P (RS/P), pFLAG-p54 (p54), and pFLAG-hTra2β (hTra2β). An empty vector, pFLAG-CMV-2, was used to keep the total amount of transfected DNA constant. Spliced products were analyzed by RT-PCR with specific DNA primers (indicated by arrows), followed by agarose gel electrophoresis. The length (in nucleotides) of splicing products (A) or specific alternatively spliced isoforms (B and C) are indicated on the right of the panels. The bands above the beta 1 (lanes 4, 5, and 12 to 16 in B) and non-muscle type (* in C) were found as nonspecific amplified products unrelated to splicing (data not shown) (42).

RRM and RS/P domains of RNPS1 are responsible for exon-skipping activity.

To determine which of the RNPS1 domains mediate its exon-skipping activity, we analyzed in vivo splicing upon overexpression of various RNPS1 mutants (Fig. 4A and B, lanes 4 to 11). Previously, it was shown that mutation of two solvent-exposed aromatic residues in the RRM of SF2/ASF and hnRNP A1 completely abolishes their functions in constitutive splicing and alternative splicing, respectively (17, 61). Therefore, we introduced the corresponding mutations in the RRM of RNPS1 to inactivate it (Table ​(Table1).1). Deletion of the RS/P domain (ΔRS/P) and mutations in the RRM (mtRRM) completely abolished exon skipping in the DUP51 pre-mRNA (220 nucleotides; lanes 6 to 9) and double-exon skipping in the hTra2β pre-mRNA (beta 3; lanes 6 to 9), whereas deletion of the S domain (ΔS) had no effect on exon-skipping induction in both substrates (220-nucleotide and beta 3; lanes 4 and 5). However, the RS/P domain alone had no exon-skipping activity in both substrates (lanes 10 and 11). The domain requirement for the exon-skipping activity is essentially consistent with these two different substrates, DUP51 and hTra2β pre-mRNAs. We conclude that both the RRM and RS/P domains are necessary, while the S domain is dispensable, for the exon-skipping activity of RNPS1.

RNPS1 and p54 cooperatively modulate alternative splicing in vivo.

To test whether the identified RNPS1-associated factors, p54 and hTra2β, cooperate with RNPS1 in the regulation of alternative splicing in vivo, the splicing patterns of all three substrates were examined upon coexpression of these factors in HeLa cells (Fig. ​(Fig.4,4, lanes 12 and above). Unfortunately, transfected full-length pinin was not detectably expressed in HeLa cells.

With the DUP51 pre-mRNA, coexpression of both p54 and RNPS1 stimulated exon skipping (220-nucleotide; lanes 12 to 14, compare lanes 1 to 3). This stimulation was also observed with RNPS1-ΔS (lanes 15 and 16), which lacks p54-interacting S domain (Fig. ​(Fig.2B).2B). Thus, the stimulation of exon skipping is independent from p54 interaction to S domain.

With the hTra2β pre-mRNA, coexpression of these factors induced totally distinctive splicing patterns. Overexpression of p54 produced beta 1 (exon 2-skipping) as a major spliced product (lane 12), whereas hTra2β overexpression generated beta 4 (exon inclusion) almost exclusively (lane 17). Coexpression of either p54 or hTra2β together with RNPS1 induced the exon-skipping products beta 3 and beta 1, respectively (lanes 14 and 19). As in the DUP51 pre-mRNA, stimulated exon skipping, even stronger, was also observed with RNPS1-ΔS, which lacks a p54 interacting S domain (beta 3; lanes 15 and 16). These data, taken together, indicate that cotransfection of either p54 or hTradβ does not interfere with the exon-skipping activity of RNPS1, which was observed with both DUP51 and hTra2β substrates (Fig. 4A and B, lanes 1 to 3).

In contrast, we observed synergistic effects upon the coexpression with the F₁γ mini-gene pre-mRNA. Coexpression of RNPS1 with p54 promoted a shift from exon skipping (muscle type) to exon inclusion (non-muscle type; Fig. ​Fig.4C,4C, lanes 12 to 14), whereas the same level of RNPS1 overexpression alone repressed both splicing (lanes 1 to 3). Coexpression of RNPS1 with hTra2β also repressed both splicing (lanes 23 and 24). With the DUP51 and hTra2β pre-mRNAs, equivalent coexpression of RNPS1 and p54 rather stimulated exon skipping (Fig. 4A and B, lanes 12 to 14). Therefore, this modulation of alternative splicing under the p54 coexpression is F₁γ substrate specific. Observed negative activity of RNPS1 (Fig. ​(Fig.4C,4C, lanes 1 to 3) is not due to nonspecific inhibition, but rather reveals substrate-specific negative regulation, since the same level of RNPS1 overexpression did not cause comparable inhibition with other substrates (Fig. 4A and B, lanes 1 to 3). Moreover, splicing assays with RNPS1 mutants suggested that RS/P domain is necessary for the splicing repression effect of RNPS1, while either S domain or RRM is dispensable (Fig. ​(Fig.4C,4C, lanes 4 to 9). However, RS/P domain itself has no repression activity (lanes 10 and 11). Interestingly, this domain requirement is different from that for the exon-skipping activity (see above). Therefore, the negative effect appears to be a domain-specific intrinsic activity of RNPS1 to the particular substrate.

Nuclear speckle colocalization of RNPS1 with hTra2β and p54.

A two-step subcellular localization mechanism was suggested for the RS domain containing D. melanogaster splicing regulator Tra protein: the first step is transport across the nuclear membrane via the NLS, while the second step is localization in the speckles via the RS domain (39). A similar mechanism is conceivable for the nuclear speckle localization of RNPS1.

We found that the RS/P domain of RNPS1, which contains several RS or SR dipeptide sequences (though too short for a typical RS domain) and a putative simian virus 40 T antigen-type or bipartite-type NLS (amino acids 278 to 287 or 296), directs localization exclusively to the nucleus. Both the RS/P and the S domains of RNPS1 are required for proper nuclear speckle localization, since the mutant proteins deleted for these domains (ΔS and ΔRS/P) showed almost exclusively nuclear localization, but did not distribute in a typical speckled pattern. The RS domain of some, but not all, SR and SR-related proteins can serve as a targeting signal to nuclear speckles (19, 39, 54). We showed that the SR protein p54 interacts with the S domain of RNPS1, and that this protein-protein interaction may direct RNPS1 to the proper nuclear speckles. A similar mechanism has been proposed from the study of the D. melanogaster Tra protein. Tra mutant protein lacking an RS domain, which shows diffuse nucleoplasmic localization, can be directed to the nuclear speckles via protein-protein interaction with Tra2 (39).

Possible mechanisms of splicing activation by RNPS1.

We previously reported that RNPS1 and SR proteins cooperatively activate general splicing in vitro (59), however the underlying mechanism remained to be elucidated. Our discovery of associated factors provides insights into the mechanisms of splicing activation.

The splicing coactivator complex SRm160/300, which also possesses an S domain and an RS/P domain, promotes splicing in a purine-rich ESE-dependent manner (7, 31). The association of SRm160/300 and U2 snRNP with pre-mRNA requires both U1 snRNP and hTra2β, the latter of which binds to ESEs, suggesting that SRm160/300 may promote critical interactions between hTra2β and the snRNP machinery in the spliceosome (31). Since we showed that possible interactions of RNPS1 with both hTra2β and the putative U1 snRNP component hLuc7A, an analogous mechanism of splicing stimulation, involving RNPS1, hTra2β, and hLuc7A, is conceivable in pre-mRNAs that have a purine-rich ESE. In contrast with the case of SRm160/300, however, splicing activation by RNPS1 is not highly dependent upon the pre-mRNA substrates (59), and thus it is unlikely that the corresponding purine-rich ESEs exist in all the interactive substrates.

RNPS1 may activate ESE-independent splicing by a different mechanism that involves pinin and U2 snRNP and/or p54 and U2AF⁶⁵. A potential interaction between RNPS1 and pinin has recently been suggested by a yeast two-hybrid screen (53; P. Ouyang and W. Y. Tarn, unpublished data), and here we demonstrate that the RRM of RNPS1 can directly bind to the second and third coiled-coil region of pinin. Recently, the interaction of RNPS1 with the coiled-coil region of pinin has also been reported (53). Pinin is enriched in nuclear fractions containing the U2 snRNP proteins SF3a and SF3b (12), which are required for assembly of the splicing complex A and for stable association of U2 snRNP with the branch site region (5, 13, 14, 49). On the other hand, we found that RNPS1 can bind to p54 through its S domain, and p54 can directly interact with U2AF⁶⁵ (86). Based on all these observations, we propose that RNPS1 stimulates recruitment of U2 snRNP and U2AF⁶⁵ via interaction with pinin and p54, respectively, to activate purine-rich ESE-independent pre-mRNA splicing. In agreement with this hypothesis, our splicing complex analysis revealed that RNPS1 stimulated formation of the ATP-dependent splicing complex A in vitro (J. H. Trembley et al., submitted for publication).

We thank J. L. Manley for the pET14b-hTra2β and pGEX-hTra2β plasmids; W. W. Franke for the DRSP-MH22 plasmid; J. Y. Wu for the pcDNA3-p54 plasmid; S. Stamm for the hTra2-β1 minigene plasmid; R. Kole for the pHD1012-DUP51 plasmid; H. Sakamoto for the pCMX-hGFP-S65A/Y145F plasmid; and J. Brandner for anti-DRSP-M2 and M3 antibodies. We are grateful to M. P. Deutscher, A. R. Krainer, T. Misteli, J. F. Cáceres, C. Jain, and members of the Mayeda laboratory for helpful discussions and critical reading of the manuscript and to M. J. Moore and W. Y. Tarn for communicating results prior to publication.

This work was supported by funds awarded by the L. P. Markey Trust, an institutional research grant (IRG-98-277-04) from A.C.S., Florida Biomedical Research Program grant (BM031) from F.D.H. to A.M., and by grants-in-aid from the Ministry of Education, Science and Culture of Japan to H.E. E.S. was supported by a postdoctoral fellowship from the Uehara Memorial Biological Science Foundation and a grant from the Jichi Medical School Young Investigator Award in Japan. A.M. is a research member of the Sylvester Comprehensive Cancer Center.