|Home | About | Journals | Submit | Contact Us | Français|
RBM25 has been shown to associate with splicing cofactors SRm160/300 and assembled splicing complexes, but little is known about its splicing regulation. Here, we characterize the functional role of RBM25 in alternative pre-mRNA splicing. Increased RBM25 expression correlated with increased apoptosis and specifically affected the expression of Bcl-x isoforms. RBM25 stimulated proapoptotic Bcl-xS 5′ splice site (5′ ss) selection in a dose-dependent manner, whereas its depletion caused the accumulation of antiapoptotic Bcl-xL. Furthermore, RBM25 specifically bound to Bcl-x RNA through a CGGGCA sequence located within exon 2. Mutation in this element abolished the ability of RBM25 to enhance Bcl-xS 5′ ss selection, leading to decreased Bcl-xS isoform expression. Binding of RBM25 was shown to promote the recruitment of the U1 small nuclear ribonucleoprotein particle (snRNP) to the weak 5′ ss; however, it was not required when a strong consensus 5′ ss was present. In support of a role for RBM25 in modulating the selection of a 5′ ss, we demonstrated that RBM25 associated selectively with the human homolog of yeast U1 snRNP-associated factor hLuc7A. These data suggest a novel mode for Bcl-xS 5′ ss activation in which binding of RBM25 with exonic element CGGGCA may stabilize the pre-mRNA-U1 snRNP through interactions with hLuc7A.
Alternative splicing is a regulatory mechanism that allows eukaryotes to generate numerous protein isoforms, often with diverse biological functions, from a single gene (2, 26, 49). Pre-mRNA splicing takes place within the spliceosome, which is assembled stepwise by the addition of small nuclear ribonucleoprotein particles (snRNP) and numerous accessory non-snRNP splicing factors (23, 31). The excision of introns and the joining of exons depend on the recognition and usage of 5′ splice sites (5′ ss) and 3′ ss by the splicing machinery (21, 33). The commitment complex forms when the 5′ ss is recognized by U1 snRNA base pairing and stabilized by U1 snRNP while the 3′ ss is recognized by the U2 auxiliary factor (U2AF) through U2 snRNA base pairing with the branch point. Subsequently, the U4/5/6 tri-snRNP is incorporated into the complex and the U1 snRNA base paired at the 5′ ss is replaced by U6 snRNA. These processes result in a fully assembled spliceosome that supports a series of rearrangements via RNA-RNA and RNA-protein interactions and activates the catalytic steps of cleavage, exon joining, and intron release (2, 26).
However, the splice site signals that define the 5′ ss and 3′ ss are often degenerate. How and when they are used are believed to be modulated by a combinational interplay of positive (splicing enhancers) and negative (splicing silencers) cis elements and trans-acting factors (2, 26), forming the basis of alternative splicing. The splicing regulatory (SR) proteins (18, 41) and the heterogeneous nuclear ribonucleoproteins (hnRNPs) (11, 46) bind with specificity to pre-mRNA (20). The SR proteins generally bind to enhancer elements through the RNA-binding domain and activate splicing at the nearby sites by recruiting spliceosomal components via protein-protein interactions mediated by the arginine-serine-rich (RS) domain (18, 41). The hnRNPs mostly bind to splicing silencers and inhibit the use of the nearby splice site, either by antagonizing positive regulators or by recruiting factors that impede the splicing machinery (11, 57).
A number of other protein families that consist of one or more RNA recognition motifs (RRM) (27) and confer tighter RNA-binding specificity have become visible as splicing regulators, e.g., PurH (39), Sam68 (35), TIA-1 (10), SAP155 (28), and Fox-2 (54). Some are composed of arginine-aspartic acid (RD), arginine-glutamic acid (RE), arginine-glycine-rich motifs, and/or other domains (19, 44, 50) that can mediate interactions between protein and RNA or protein molecules. By associating with different types of protein domains, the RRM domain can modulate its RNA-binding affinity and specificity in alternative splicing (19, 44, 50).
Even though splicing regulation occurs at every step throughout the assembly pathway, many factors primarily affect the efficiency of commitment complex formation (3, 47). Binding of SF2/ASF to an exonic enhancer of rat β-tropomyosin pre-mRNA can stimulate U2AF or U2 binding to a weak 3′ ss and U1 snRNP binding to a 5′ ss (43). Fox proteins interact with the upstream UGCAUG elements in a manner that blocks U2AF65 binding to the 3′ ss upstream of exon 4 and prevents exon inclusion in calcitonin/CGRP pre-mRNA (55). Interestingly, human homologs of yeast U1 snRNP-associated proteins, TIA-1 and hLuc7A, have emerged as important players that affect splice site selection. TIA-1 promotes U1 snRNP binding to a weak 5′ ss, followed by uridine-rich sequences through the glutamine-rich domain of TIA-1 and the U1-specific protein U1-C (14). Interaction of yeast yLuc7p with the upstream exon stabilizes the pre-mRNA-U1 snRNP interaction (15). Similarly, human hLuc7A affects 5′ ss selection possibly via the formation of a network of protein-RNA interactions that stabilize pre-mRNA-U1 snRNP interaction and thus formation of the commitment complex (36).
Alternative splicing is often tightly regulated in a cell type- or developmental-stage-specific manner. Splice site selection must therefore depend not only on the cis element's identity but also on the amount and/or activities of cellular splicing factors (25, 29). We have shown earlier that expression of RRM-containing splicing factors is regulated during erythroid differentiation (53, 54); RBM25 expression was downregulated during differentiation. RBM25 belongs to a family of RNA-binding proteins whose members share the RE/RD-rich (ER) central region and C-terminal proline-tryptophan-isoleucine (PWI) motif (16). RBM25 localizes to the nuclear speckles and associates with multiple splicing components such as splicing cofactors SRm160/300, U snRNAs, assembled splicing complexes, and spliced mRNAs (16). The characterization of RBM25 strongly suggests that it functions in pre-mRNA processing. However, immunodepletion of RBM25 did not alter the splicing activity of several tested pre-mRNAs, including dsx, Ftz, and Msl-2 (16), suggesting that RBM25 is not a general splicing factor.
Here, we report that RBM25 associates with a U1 snRNP-associated factor, hLuc7A (36), and activates proapoptotic Bcl-xS 5′ ss via its interaction with the exonic splicing enhancer, CGGGCA. Furthermore, the intracellular RBM25 levels affect the ratio of proapoptotic Bcl-xS to antiapoptotic Bcl-xL mRNA and correlate with apoptotic cell death. These data suggest a novel mechanism of Bcl-xS 5′ ss activation in which binding of RBM25 to exonic element CGGGCA may stabilize the pre-mRNA-U1 snRNP through interaction with hLuc7A. Thus, RBM25 is one of the RNA-binding regulators that direct the alternative splicing of apoptotic factors.
All DNA constructs were made by using standard cloning procedures and confirmed by sequencing. RBM25 (GenBank accession number NP_067062.1) was amplified from human skeletal muscle cDNA (Origene) with primers 5′-TCTTTTCCACCTCATTTGAATCGC-3′ and 5′-CTTCACAAGACCAATTTTCTTGGCTTC-3′ and cloned into the expressing vector pCDNA3.1-HA (Invitrogen) to generate pCDNA3.1-HA-RBM25. Full-length RBM25 (amino acids [aa] 1 to 843) and its individual RRM (aa 11 to 284), ER (aa 285 to 644), and PWI (aa 645 to 843) domains were cloned into pEGFP-C2 to generate RBM25-FL/EGFP (where EGFP is enhanced green fluorescent protein), RBM25-RRM/EGFP, RBM25-ER/EGFP, and RBM25-PWI/EGFP, respectively. Full-length RBM25 and its CTD (C-terminal domain; aa 641 to 830) were cloned into pGEX-6p1 to generate RBM25-FL/GST and RBM25-CTD/GST. Full-length RBM25 was cloned into the pcDNA3.1-T7 tag vector (Novagen) to generate RBM25/T7. RBM25 shRNA plasmids with the targeted nucleotide sequences GCTCCAGAGGATGGAACAAGAGGCTGAGA for sh81, CTCCATCTGTTTCCTCTGCCAGTGGCAAT for sh82, CCACTGTGTCTATGGTTGGAAAGCATTTG for sh83, and CTTCGGATTCTGTGAGTACAAGGAGCCAG for sh84 were purchased from Origene.
The Bcl-x minigene was a gift from C. E. Chalfant (Virginia Commonwealth University, Richmond). The Bcl-x exon 2 mutation and Bcl-xS 5′ ss mutation minigenes were constructed with a QuikChange II XL site-directed mutagenesis kit (Stratagene).
pCGT7-hLuc7A (GenBank accession number NM_01624.3) was kindly provided by O. Puig (European Molecular Biology Laboratory, Germany). The full length (aa 1 to 442), N half (aa 1 to 224), and C half (aa 225 to 442) of hLuc7A were subcloned into pGEX-6p1 (Amersham Biosciences) to produce hLuc7A-FL/GST, hLuc7A-N/GST, and hLuc7A-C/GST, respectively. The E1A minigene (pCEP4-E1A) was a gift from W. Y. Tarn (Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan).
HeLa and HEK293 cells were obtained from the American Type Culture Collection and cultured in Dulbecco modified Eagle medium supplemented with 0.1 mM nonessential amino acids, 1.0 mM sodium pyruvate, 10% fetal calf serum (Invitrogen), and penicillin/streptomycin (5,000 U/ml; Invitrogen). For transcription inhibition, 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole (DRB) was added at 100 μM and the culture was incubated for 2 h at 37°C prior to fixation and immunolabeling. Cells were transfected with Lipofectamine (Invitrogen) according to the manufacturer's instructions. Apoptosis was induced by addition of 500 nM staurosporine (Sigma) in medium and incubation for 0, 24, or 48 h.
Semiquantitative reverse transcription (RT)-PCR analysis of splicing products was performed as described previously (53). RNAs were reverse transcribed with pcDNA3.1/BGH primer for Bcl-x minigenes. Amplification of spliced products was performed with a pcDNA 3.1 sequence (5′-CTGATCAGCGGTTTAAACTTA-3′) and a primer specific to Bcl-x exon 2 (5′-GGA GCT GGT GGT TGA CTT TCT-3′). For each construct, two transfections were performed in each experiment. Each experiment was repeated three times, and standard deviations (SDs) were determined.
For E1A splicing analyses, RT was performed with primer P1 (5′-GGTCTTGCAGGCTCCGGTTCTGGC) and PCR was performed with primers P2 (5′-GCAAGCTTGAGTGCCAGCGAGTAG) and P3 (5′-CTCAGGCTCAGGTTCAGACACAGG).
Primer sets flanking exon 2 of Mcl1 (RT, 5′-ATTAGATATGCCAAACCAGC-3′; PCR, 5′-TGGAGATTATCTCTCGGTACCT-3′ and 5′-CAACACCTGCAAAAGCCAGCA-3′), exon 6 of caspase 3 (RT, 5′-GTTGCCACCTTTCGGTTAAC-3′; PCR, 5′-TCAGAGGGGATCGTTGTAGAAG-3′ and 5′-ATGTGCATAAATTCAAGCTTGTCG-3′), exon 2 of Bcl-x (RT, 5′-AGGGTTGCACCAATCAGGTAG-3′; PCR, 5′-ATGTCTCAGAGCAACCGGGAGCTG-3′ and 5′-TCATTTCCGACTGAAGAGTGAGCC-3′), and exon 6 of Fas (RT, 5′-AGGATTTAAAGTTGGAGAT-3′; PCR, 5′-ATGGAATCATCAAGGAATGCACA-3′ and 5′-GAGAACCTTGGTTTTCCTTTCTG-3′) were used to analyze the endogenous splicing patterns of its respective genes.
HeLa and HEK293 cells were transfected with either the pEGFP vector or RBM25/EGFP constructs and subjected to immunofluorescence staining with anti-SC35 antibody (Ab), anti-active caspase 3 Ab (Sigma), or anti-T7 Tag Ab (Novagen) as described previously (54). Endogenous RBM25 was stained with an anti-RBM25 Ab. All samples were counterstained with 4′,6′-diamidino-2-phenylindole (DAPI). The samples were viewed with a Zeiss Axiovert 200M inverted microscope (Zeiss, Inc.). The images were collected with SlideBook4 software and processed with Photoshop software (Adobe Systems, Inc.).
RNA immunoprecipitation (RIP) was performed as described previously (32). Briefly, 2 × 108 HeLa cells were cross-linked with formaldehyde and resuspended in FA lysis buffer (50 mM HEPES-KOH, 140 mM NaCl, 1 mM EDTA, 0.1% sodium deoxycholate, 1.0% Triton X-100, pH 7.5) containing protease inhibitors and RNase inhibitors. Lysates were sonicated and treated with DNase. Soluble extracts were incubated with anti-RBM25 or rabbit immunoglobulin G (IgG), washed with lysis buffer, digested with RNase-free DNase (Roche), and finally treated with proteinase K. The RNA segments retrieved and the input saved were analyzed by RT-PCR for the presence of Bcl-x or Mcl1 with a random primer for RT. PCR was performed with primer sets Bcl-x1-S (5′-ATGTCTCAGAGCAACCGGGAGCTG-3′) and Bcl-x1-As (5′-CTGTTGGGGATCTCTGACCAGA-3′) for Bcl-x and hMcl1-Ex2-S (5′-AAGACGATGTGAAATCGTTGTCTC-3′) and hMcl1-Intr2-As (5′-TCCTTCATTCTACTTCCACTCCA-3′) for Mcl1.
Gel mobility shift assays were performed as described previously (24), with modification. The biotinylated wild-type (WT) UACCGGCGGGCAUUCAGUGA and mutated UACCGGUUAGAGUUCAGUGA RNA substrates were synthesized by Invitrogen. Binding of biotinylated RNA to RBM25 was achieved by incubating 0.2 nM RNA and variable amounts of protein for 30 min at 4°C in 20 μl of binding buffer (10 mM Tris-HCl, 10 mM HEPES, 100 mM NaCl, 0.1% Triton X-100, 2 mM MgCl2, 1.5 mM dithiothreitol [DTT] [pH 7.5], 7% glycerol). For competition assays, a molar excess of unlabeled competitor RNAs at various levels was added to the preincubation reaction mixture. Samples were fractionated in a native 5% polyacrylamide gel, transferred to Hybond-N+ nylon membrane (Amersham Biosciences), and detected with a LightShift chemiluminescent electrophoretic mobility shift assay kit (Pierce) by following the manufacturer's protocol. RBM25 proteins were purified from nuclear extracts of RBM25/T7-transfected HeLa cells with a T7 · Tag affinity purification kit (Novagen). Psoralen cross-linking reactions were carried out as described previously (56), with transcripts spanning the region from 92 nucleotides (nt) upstream to 33 nt downstream of the WT, mu4, or consensus Bcl-xS 5′ ss in the presence or absence of 300 ng purified RBM25. The cross-linked RNAs were analyzed on 5% polyacrylamide gels containing 8.3 M urea.
Coimmunoprecipitation (co-IP) of RBM25 and hLuc7A was performed with HeLa cell nuclear extracts prepared as described previously (12). Nuclear extracts were diluted in co-IP buffer (30 mM Tris-HCl, 100 mM NaCl, 10 mM MgCl2, 1 mM DTT, 0.5% Triton X-100 [pH 7.5], protease inhibitors) and precleared with rabbit IgG. Subsequently, the supernatant was immunoprecipitated with preimmune rabbit IgG or anti-RBM25 Ab. The immunoprecipitated samples were treated with RNase and examined for the presence of RBM25 with an anti-RBM25 Ab or for the presence of hLuc7A with an anti-Luc7A Ab (Novus).
An anti-RBM25 polyclonal Ab was raised against RBM25-CTD/GST fusion proteins. RBM25 proteins were detected with an antihemagglutinin (anti-HA) Ab (Roche Diagnostics) for HA-tagged RBM25 or with an anti-RBM25 Ab for endogenous RBM25 with the ECL detection kit (Amersham Pharmacia). An anti-β-actin Ab (Sigma) served as a loading control.
Glutathione S-transferase (GST) alone or Luc7A/GST fusion proteins were affinity purified via coupling to glutathione-Sepharose beads (GE Healthcare) according to the manufacturer's procedure. Equal amounts of GST fusion proteins were incubated with 500 μg of HeLa cell nuclear extracts in the binding buffer (30 mM Tris-HCl, 100 mM NaCl, 10 mM MgCl2, 1 mM DTT, 0.5% Triton X-100 [pH 7.5], protease inhibitors). The presence of RBM25 bound to Luc7A was detected by Western blotting with an anti-RBM25 Ab.
Transfected cells grown on 35-mm plates were harvested and processed for annexin V staining or for Western blot analysis. For annexin V staining, cells were washed in phosphate-buffered saline and stained with annexin V-phycoerythrin (BD Biosciences) according to the manufacturer's instructions. GFP-positive cells were then analyzed with a FACScalibur flow cytometer (Becton Dickinson).
RBM25 consists of a proline-rich region and an RRM domain at the amino-terminal end, an RE/RD-rich (ER) domain in the central region, and a PWI domain at the carboxyl-terminal end (Fig. (Fig.1A).1A). RBM25 is known to localize to nuclear speckles (16). Our Ab against the C-terminal domain of RBM25 confirmed that endogenous RBM25 localized predominantly in the nuclear speckles as a punctate structure (Fig. (Fig.1B,1B, −DRB, RBM25). As previously reported, SC35 also localized to the nuclear speckles (Fig. (Fig.1B,1B, −DRB, SC35). Superimposition of RBM25 and SC35 revealed that both proteins are intensely colocalized in the same region (Fig. (Fig.1B,1B, −DRB, SC35 + RBM25). Upon treatment with DRB, an inhibitor of RNA polymerase II-dependent transcription, both proteins relocated to enlarged and more round nuclear speckles (Fig. (Fig.1B,1B, +DRB, SC35, and RBM25). Speckles function as storage compartments that can supply splicing factors to active transcription sites. Splicing factors are recruited from speckles to sites of transcription (30); conversely, splicing factors accumulate in enlarged and rounded speckles when transcription is inhibited (30, 48). The colocalization and co-redistribution of RBM25 and SC35 suggest that RBM25 is organized in a pattern very similar to that of other splicing factors during states of transcriptional activity and quiescence.
To further characterize the domain in RBM25 responsible for speckle localization, we expressed full-length RBM25, as well as its individual RRM, ER, and PWI domains, fused with EGFP in HeLa cells and analyzed its localization relative to splicing factor SC35. EGFP was detected throughout the cell when transfected with the vector alone (Fig. (Fig.1C,1C, GFP vector, GFP). The localization of full-length RBM25 was similar to that of endogenous RBM25; it colocalized with SC35 in nuclear speckles (Fig. (Fig.1C,1C, RBM25/FL, RBM25 + SC35). A different labeling pattern was observed for the RRM- and PWI-EGFP fusions. Both domains diffusely localized throughout the nucleus without any apparent accumulation on the speckles (Fig. (Fig.1C,1C, RBM25/RRM and RBM25/PWI, GFP). The ER domain produced a speckle pattern within the nucleus that coincided with that of SC35 (Fig. (Fig.1C,1C, RBM25/ER, RBM25 + SC35). Nuclear speckle localization of the ER domain was not cell type specific, as similar localization was seen in transfected murine erythroleukemia cells (data not shown). Thus, the ER domain of RBM25 is responsible for targeting RBM25 to the nuclear speckle.
Although transient transfection of RBM25 localized to the nuclear speckles, our attempts to achieve stable RBM25 expression lines failed. These cells began to undergo apoptosis between 48 and 72 h after RBM25 introduction.
To gain more insight into the nature of these events, we first transfected HEK293 cells with EGFP or RBM25/EGFP and subsequently determined the extent of cell death by annexin V binding, cleavage of caspase 3, and nuclear fragmentation assays. Cells harvested 48 h after transfection showed a threefold increase in annexin V-positive cells in RBM25/EGFP-transfected cells compared with that of EGFP cells (Fig. (Fig.2A).2A). Increased expression of RBM25 also significantly increased the percentage of activated-caspase 3-stained, GFP-positive populations (Fig. (Fig.2B).2B). Furthermore, a significant fraction of RBM25-overexpressing cells displayed typical nuclear fragmentation associated with apoptotic cell death (Fig. (Fig.2C).2C). These results suggest that an elevation in RBM25 expression caused an increase in apoptotic cell death, indicating that a correlation exists between RBM25 activity and induction of apoptosis.
We then documented the time course of this effect as revealed by nuclear fragmentation in approximately 500 RBM25/GFP-expreesing cells at 24, 48, and 72 h after transfection. At 24 h, different degrees of RBM25/GFP expression, as judged by the intensity/brightness of green fluorescence, were observed. In cells exhibiting “average” brightness, which applied to approximately 95% of the cells, RBM25/GFP localized to nuclear speckles. Of interest, nuclear fragmentation occurred in ~5% of the cells; these cells exhibited above-average brightness. The cells in Fig. 1B and C expressed an average intensity of RBM25/GFP and nuclear speckle localization and are thus indicative of the vast majority of the cells observed.
At 48 h posttransfection, RBM25/GFP accumulated in large aggregates in the nuclei of approximately 25% of the expressing cells (Fig. (Fig.2D,2D, RBM25-GFP, GFP, arrow). These cells displayed various degrees of nuclear fragmentation (Fig. (Fig.2D,2D, RBM25-GFP, DAPI, arrow). The number of cells exhibiting nuclear fragmentation increased to 50% as transfected cells proceeded to 72 h. We were unable to select stable lines carrying RBM25 due to cell death caused by its overexpression.
We further analyzed the average exogenously expressed RBM25 in the transfected cells by Western blot analysis with an anti-RBM25 Ab. At 48 h posttransfection, an ~1.3- to 1.5-fold increase in RBM25/GFP expression (Fig. (Fig.2E,2E, RBM25-GFP, RBM25-GFP) over that of endogenous RBM25 (Fig. (Fig.2E,2E, RBM25-GFP, RBM25) was detected in total cell lysates. Taking into account the facts that every cell did not express equal amounts of RBM25/GFP and that only approximately 80% transfection efficiency was achieved, we estimate that the average RBM25/GFP expression is approximately 1.6- to 2.0-fold over that of endogenous RBM25. This amount of overexpression is sufficient to induce apoptosis.
To examine whether RBM25 expression and localization changed during apoptosis, HEK293 cells were treated with staurosporine for 0, 24, and 48 h and analyzed. Nuclear speckle localization of RBM25 was observed in untreated cells (Fig. (Fig.3A,3A, 0 h, RBM25). At 24 h, in the majority of the cells, RBM25 accumulated in large aggregates in the nucleus; these aggregates appeared to reside in subnuclear locations containing less condensed DNAs. Nuclear fragmentation was observed in a small population of cells. As apoptosis proceeded to 48 h, RBM25 accumulated in the subnuclear compartment with less condensed DNAs in some cells (Fig. (Fig.3A,3A, 48 h, RBM25, DAPI, merge, arrow) and a diminished RBM25 immunostaining signal was present in the majority of the cells with nuclear fragmentation (Fig. (Fig.3A,3A, 48 h, RBM25, DAPI, merge, arrowhead).
We then examined whether RBM25 expression changed during the process. Equal amounts of cell lysates from both induced and uninduced cells were blotted with an anti-RBM25 Ab. An almost equally expressed 120-kDa band was detected in all samples (Fig. (Fig.3B,3B, RBM25, lanes 0, 24, 48, and 72). An additional higher-molecular-weight band was also detected in time zero and uninduced 24-h lysates (Fig. (Fig.3B,3B, RBM25, lanes 0 and −24). These results suggest that RBM25 expression changes could result from either posttranslational modification or expression of different isoforms. Western blot analysis also clearly demonstrated that the disappearance of RBM25 at 48 h in cells in which apoptosis had been induced was not due to protein degradation because RBM25 remained as intact and abundant in Western blot assays of apoptotic cells as it appeared at time zero (Fig. (Fig.3B,3B, lanes 0 and 48). These results imply a possible conformational change occurring in RBM25 that modifies the antigenic epitope recognized by the anti-RBM25 Ab.
The observations that RBM25 colocalized with splicing factors in the nuclear speckles and that its overexpression resulted in increased apoptotic cell death suggest a possible role for RBM25 in splicing regulation involved in the apoptotic pathway. This prompted us to search for targets of RBM25 with a focus on apoptotic factors. A large number of these factors are regulated via alternative splicing, a process that allows for the production of discrete protein isoforms with distinct apoptotic functions (42).
We analyzed a panel of apoptotic factors, Mcl1 (1), caspase-3 (22), Bcl-x (4), and Fas (13), for the expression of their isoforms in response to RBM25 overexpression (Fig. (Fig.4A).4A). Expression of RBM25 was validated by Western blotting with anti-HA Abs (Fig. (Fig.4A,4A, anti-HA). Among the tested factors, increased Bcl-xS and reduced Bcl-xL were noted in the presence of RBM25 (Fig. (Fig.4A,4A, Bcl-x), implying that RMB25 may regulate Bcl-x alternative splicing. Bcl-x pre-mRNA uses an alternative 5′ ss to produce the antiapoptotic Bcl-xL or the proapoptotic Bcl-xS isoform. RBM25 exerted its effect on more efficient utilization of the Bcl-xS 5′ ss, resulting in an increased ratio of Bcl-xS to Bcl-xL. On the other hand, no effects on capase-3, Mcl1, and Fas were detected (Fig. (Fig.4A,4A, casp-3, Mcl1, and Fas). Since the inclusions of the alternative exons in these transcripts are already vastly predominant, their response to RBM25 was then further evaluated in an RBM25 knockdown background (Fig. (Fig.4B).4B). Treatment of cells with an RBM25 shRNA depleted the endogenous RBM25 by 85% (Fig. (Fig.4B,4B, anti-RBM25, lanes non- and sh-82); nevertheless, no changes in the splicing pattern were detected. The alternative splicing of caspase-3, Mcl1, and Fas was not affected by RBM25 (Fig. (Fig.4B,4B, casp-3, Mcl1, and Fas). Thus, the effect of RBM25 on Bcl-x pre-mRNA processing was specific and not attributable to a generalized effect on the RNA splicing machinery.
To investigate the regulation of Bcl-x splicing by RBM25, we used a Bcl-x minigene (a gift from C. E. Chalfant, Virginia Commonwealth University, Richmond) that spans the entire alternatively spliced region from exon 1 to exon 3, with a shortened intron 2 (Fig. (Fig.5A).5A). The Bcl-x minigene produced the same Bcl-x splicing pattern as seen with endogenous Bcl-x and produced two splice variants of the Bcl-x transcripts with a ratio of Bcl-xS to Bcl-xL of 0.31 in HeLa cells (Fig. (Fig.5B,5B, lane 0). Cotransfection of pcDNA3.1-HA-RBM25 with this reporter activated splicing pathways which led to increased short variant Bcl-xS with respect to transfection of pcDNA3.1-HA alone (Fig. (Fig.5B,5B, lanes 0, 0.25, 0.5, 0.75, 1.00, and 1.50). The selection of the upstream 5′ ss in exon 2, resulting in the production of Bcl-xS, was stimulated in an RBM25 dose-dependent manner (Fig. (Fig.5B).5B). The addition of 0.25 μg RBM25 increased the ratio of Bcl-xS to Bcl-xL to 0.64, while the ratio increased to 2.3 when 1.5 μg of RBM25 was introduced. The expression of RBM25 was validated by a Western blot assay with anti-HA Abs (Fig. (Fig.5B,5B, anti-HA). These results suggest that the Bcl-x reporter gene responded to RBM25 by enhancing Bcl-xS 5′ ss usage.
The observation that increased expression of RBM25 could enhance the usage of the Bcl-xS 5′ ss prompted us to examine whether reduction of RBM25 expression would block the usage of the same 5′ ss. Several RBM25 shRNA constructs (Fig. (Fig.6A),6A), individually or in combination, reduced endogenous RBM25 expression in HeLa cells (Fig. (Fig.6B,6B, RBM25, lanes sh81, sh82, sh83, sh84, and all). A nonfunctional control nonsilencing shRNA served as a control (Fig. (Fig.6B,6B, RBM25, lane non).
We analyzed Bcl-x splicing patterns in RBM25 shRNA-treated cells. The nonsilencing control did not affect the splicing patterns (Fig. (Fig.6B,6B, lane non) compared with that of untreated cells (Fig. (Fig.5B,5B, lane 0). Reduction in RBM25 clearly affected the Bcl-x splicing pattern, in which a discernible decrease in Bcl-xS 5′ ss usage was observed in RBM25 shRNA-treated cells (Fig. (Fig.6B,6B, lanes non, sh81, sh82, sh83, sh84, and all). Bcl-xS production was reduced by 3- to 10-fold. The decrease corresponded to the potency of the sh-RBM25 used to knock down RBM25 (Fig. (Fig.6B,6B, expression, lanes non, sh81, sh82, sh83, sh84, and all). These results further suggest an involvement of RBM25 in the determination of 5′ ss selection in the Bcl-x gene.
To examine whether RBM25 exerted its activity through an interaction with a cis-acting element, we first identified RNA sequences that have an effect on 5′ ss selection in response to RBM25 expression levels. We generated mutant forms of Bcl-x minigenes in which we replaced several sequences spanning exon 2 (Fig. (Fig.7A).7A). Except for Mu1 and Mu5 (Fig. (Fig.7B,7B, −RBM25, lanes Mu1 and Mu5), the replacements affected the splicing pattern with either increased (Fig. (Fig.7B,7B, −RBM25, lane Mu2) or decreased ratios of Bcl-xS to Bcl-xL (Fig. (Fig.7B,7B, −RBM25, lanes Mu3 and Mu4) compared with that of the WT (Fig. (Fig.7B,7B, −RBM25, lane WT). These results suggest that the exonic sequences could activate or repress Bcl-xS 5′ ss usage.
We then examined whether the mutated minigenes would still be responsive to RBM25 by cotransfection of RBM25 and the minigenes into HeLa cells. All but one responded to the stimulation of RBM25 by increasing the ratio of Bcl-xS to Bcl-xL (Fig. (Fig.7B,7B, +RBM25, lanes WT, Mu1, Mu2, Mu3, and Mu5). Mutation of an element, CGGGCA, located 64 to 69 nt upstream of the Bcl-xS 5′ ss drastically reduced Bcl-xS expression and shifted the ratio of Bcl-xS to Bcl-xL from 0.33 to 0.02 (Fig. (Fig.7B,7B, −RBM25, lane Mu4). Furthermore, mutation of this element abolished the ability of RBM25 to affect 5′ ss selection (Fig. (Fig.7B,7B, +RBM25, lane Mu4). These results suggest that RBM25 might exert its activity through its binding to the CGGGCA element.
We then investigated whether the endogenous Bcl-x RNA was associated with RBM25 by using an RIP assay with an anti-RBM25 Ab. RNA templates retrieved by RIP were analyzed by RT-PCR with primer sets to detect Bcl-x RNA. As shown in Fig. Fig.8A,8A, Bcl-x RNA was detected in both input and anti-RBM25 precipitates (Fig. (Fig.8A,8A, input, α-RBM25, lanes +RT). The specificity of RNA IP is evidenced by the fact that IgG did not precipitate Bcl-x RNA (Fig. (Fig.8A,8A, Bcl-x, IgG, lane +RT). Furthermore, no products were detected in the absence of RT (Fig. (Fig.8A,8A, Bcl-x, α-RBM25, lane −RT). These results suggest that Bcl-x RNA associates with RBM25. To ensure that anti-RBM25 Ab did not pull down any nonspecific RNA sequences, we amplified isolated RNA for the presence of Mcl1 with its specific primer sets. While Mcl1 RNA was detected in the input, it was not found in anti-RBM25 precipitates (Fig. (Fig.8A,8A, Mcl1, input and α-RBM25, lanes +RT). These results are consistent with the notion that Mcl1 is not a substrate of RBM25 and validate the specificity of the RIP assay.
To further examine whether RBM25 binds to CGGGCA directly, we performed an electrophoretic mobility shift analysis with CGGGCA-containing RNA and purified RBM25 protein. We examined the specific interaction between RBM25 and CGGGCA with a probe derived from the WT consisting of the sequences flanked by 6 nt upstream and 8 nt downstream (Fig. (Fig.8B,8B, WT). The mutant sequence TTAGAG, flanked by identical upstream and downstream sequences (Fig. (Fig.8B,8B, Mut), served as a control. The intensity of the retarded band increased when the WT probe was incubated with increasing amounts of RBM25 (Fig. (Fig.8C,8C, WT, lanes 0, 0.5, 1.0, 2.0, and 4.0 μM). GST alone did not bind to the RNAs (Fig. (Fig.8C,8C, WT, lane GST). No retarded bands were observed when increased amounts of RBM25 were incubated with the mutant probe (Fig. (Fig.8C,8C, Mut, lanes 0, 0.5, 1.0, 2.0, and 4.0 μM). To assess the specificity of this WT RNA-protein interaction, a binding competition assay was performed with a 1-, 5-, or 20-fold molar excess of unlabeled WT or mutant RNA. The RNA-protein complexes were partially competed away by a 5-fold excess, and completely competed away by a 20-fold excess, of WT (Fig. (Fig.8D,8D, WT) but not mutant RNA (Fig. (Fig.8D,8D, Mut), suggesting that the observed WT RNA-RBM25 interaction is CGGGCA sequence specific.
To pinpoint whether CGGGCA conferred RBM25 sensitivity, we examined the element in the context of the E1A reporter gene. A naturally occurring CGGGCA element is located 88 to 94 nt upstream of the 13S 5′ ss in E1A (Fig. (Fig.9A).9A). Transfection of an E1A minigene alone produced three major classes of mRNAs with ~46% 13S, 38% 12S, and 16% 9S through the use of alternative 5′ ss (Fig. (Fig.9B,9B, non, lane −RBM25). The 13S 5′ ss is much stronger than the 12S or 9S 5′ ss. Unexpectedly, the presence of cotransfected RBM25 did not change the overall splicing pattern (Fig. (Fig.9B,9B, non, lane +RBM25). We then asked whether the effect of CGGGCA would be more potent in the context of a weak 9S 5′ ss. We created E1A+WT and E1A+Mu4 constructs in which the WT (UGGGCA) or the Mu4 (TTAGAG) sequence was inserted at positions 64 to 69 upstream of the 9S 5′ ss, respectively. The presence of UGGGCA resulted in a shift in splicing favoring the use of the 9S 5′ ss; an elevated 9S accounted for ~25% and a concomitant reduction of 12S to ~30% (Fig. (Fig.9B,9B, WT, lane −RBM25). Cotransfection of RBM25 promoted ~25% more 9S splicing (Fig. (Fig.9B,9B, WT, lane +RBM25). The insertion of Mu4 sequence exerted no significant effect on splicing in either the absence or the presence of RBM25 (Fig. (Fig.9B,9B, Mu4, lanes −RBM25 and +RBM25). These results suggest that UGGGCA could enhance the utilization of the weak 9S 5′ ss but not that of a strong 13S 5′ ss through its interaction with RBM25.
The observation that CGGGCA exerted its effect on a weak 5′ ss prompted us to examine how RBM25 activated a 5′ ss with Bcl-x. We first analyzed the strength of Bcl-x 5′ ss with the Analyzer Splice Tool (http://ast.bioinfo.tau.ac.il/SpliceSiteFrame.htm), which uses an algorithm based on the work of Shapiro and Senapathy (45) to calculate the scores of donor and acceptor sequences. Bcl-xS possesses a weak 5′ ss with a score of 72.66, while Bcl-xL 5′ ss is relatively stronger at 81.51; the reference consensus sequence CAG/GTAAGT has a score of 100 (Fig. 10A).
Early recognition of the 5′ ss involves base-pairing interaction with the 5′ end of U1 snRNA. We performed psoralen-mediated UV cross-linking assays to examine whether U1 snRNP is recruited to transcripts containing the Bcl-xS 5′ ss (Fig. 10B, WT). A psoralen-dependent cross-linked U1 snRNA/Bcl-x was detected upon incubation of a radioactively labeled WT sequences in HeLa cell nuclear extracts (Fig. 10B, WT, lane +/−). This interaction was significantly enhanced when ~300 ng of purified RBM25 was added (Fig. 10B, WT, lane +/+), suggesting that the presence of RBM25 facilitates U1 snRNP binding.
To determine whether the sequence CGGGCA is required for U1 interaction with the Bcl-xS 5′ ss, we repeated the psoralen cross-linking assay with a Mut4 mutant probe. Mutations in the RBM25 binding site reduced the cross-linking product to a small extent (Fig. 10B, Mu4, lane +/−) compared with that of the WT probe (Fig. 10B, WT, lane +/−). In contrast to the WT probe, the addition of RBM25 did not promote U1 cross-linking to the Mu4 transcript (Fig. 10B, Mu4, lane +/+). These results suggest that RBM25-dependent U1 recruitment requires CGGGCA.
Data from Fig. Fig.9B9B suggest that the presence of CGGGCA, followed by a strong 13S 5′ ss, had no effect on 13S splicing activity in the presence of RBM25. We thus tested whether an increased strength in Bcl-xS 5′ ss would attenuate the effects of RBM25 in a cross-linking assay with Bcl-xS and a consensus 5′ ss probe (Fig. 10A, 5′cons). The 5′cons transcripts cross-linked to U1 snRNA (Fig. 10B, 5′cons, lane +/−) much more strongly than did the WT in the absence of added RBM25 (Fig. 10B, WT, lane +/−). Addition of RBM25 did not increase 5′cons cross-linking to U1 snRNA (Fig. 10B, 5′cons, lane +/+). These data suggest that the contribution of RBM25 to U1 snRNP recruitment is particularly critical for a weak 5′ ss.
Subsequently, we analyzed the effect of the strength of Bcl-xS 5′ ss on splicing and its response to RBM25 in HeLa cells. Strengthening Bcl-xS 5′ ss to consensus sequence led to ~100% usage of the 5′ ss in the absence of RBM25 (Fig. 10C, 5′cons, lane −RBM25). Although RBM25 had an enhanced effect on the selection of the weak WT 5′ ss (Fig. 10C, WT, lane +RBM25), a strong 5′ ss did not require facilitation by RBM25 for the 5′ ss selection (Fig. 10C, 5′cons, lane +RBM25). These results suggest that RBM25 may facilitate WT 5′ ss recognition by improving the interaction between U1 snRNA and the weak 5′ ss.
Several positive regulators activate splicing of alternative exons with a weak 5′ ss by promoting interaction of U1 snRNP with the 5′ ss. U1 snRNP recognizes the 5′ ss and is among the first factors to interact with the pre-mRNA to form “complex E,” which commits the pre-mRNA to the splicing pathway (26). We explored the possibility that RBM25 associates with the U1 snRNP complex and plays a role in recruiting U1 snRNP to the weak 5′ ss. U1 snRNP is composed of a 165-nt RNA (U1 snRNA), seven different Sm proteins common to other snRNP, and three U1-specific proteins (U1-70K, U1-A, and U1-C) (38). To probe the interactions between RBM25 and U1 snRNP, RBM25 was immunoprecipitated from HeLa cell nuclear extracts and analyzed by Western blotting for the presence of U1-specific proteins. No U1-specific proteins were detected in the precipitates (data not shown).
Human Luc7A has recently been shown to associate with U1-A and U1 snRNA and stabilize pre-mRNA-U1 snRNP interaction (36). We thus examined whether RBM25 associated with hLuc7A in the immunoprecipitates. Since RBM25 contains RRM and PWI motifs, both of which can bind to RNA, we also examined whether the association is RNA mediated. As shown in Fig. 10D, hLuc7A was abundantly detected in anti-RBM25 immunoprecipitates in an RNase-independent manner (Fig. 10D, α-RBM25, lane p, +RNase). The association of hLuc7A with RBM25 is most likely through the carboxyl half of hLuc7A because both the full length and the C half of hLuc7A, but not the N half, pulled down RBM25 in a GST pull-down experiment (Fig. 10E). Furthermore, hLuc7A colocalized in the nuclear speckles with RBM25 (Fig. 10F, RBM25+Luc7A). These results suggest that binding of RBM25 to the CGGGCA sequence may help recruitment of U1 snRNP to the weak 5′ ss through its interaction with U1 snRNP-associated hLuc7A and provide a molecular mechanism for the function of this splicing regulator.
The present study identifies a novel splicing factor, RBM25, which modulates apoptosis through regulation of Bcl-x isoform expression. RBM25 associates with a U1 snRNP-associated factor, hLuc7A, and activates Bcl-xS 5′ ss via its interaction with the exonic splicing enhancer CGGGCA. Through both gain-of-function and loss-of-function experiments, we provide the first direct evidence that intracellular RBM25 levels affect the ratio of proapoptotic Bcl-xS to antiapoptotic Bcl-xL mRNA. Thus, we suggest that RBM25 is a member of a family of RNA-binding regulators that direct alternative splicing of apoptotic factor Bcl-x.
RBM25 has recently been proposed to regulate splicing; however, no direct pre-mRNA targets were identified (16). The correlation between RBM25 expression and the induction of apoptosis suggest that RBM25 either directly or indirectly targets proteins involved in apoptosis or apoptotic pathways. Here we show that regulation of apoptosis by RBM25 involves alternative splicing of its cellular target, Bcl-x. Bcl-x pre-mRNA employs alternative 5′ ss to produce the antiapoptotic Bcl-xL or the proapoptotic Bcl-xS isoforms (4). An increase in RBM25 has a significant effect in promoting the selection of the Bcl-xS 5′ ss, whereas reduction of RBM25 shifts the balance toward the antiapoptotic Bcl-xL) form.
Our results and those of others (16) showed that overexpression or depletion of RBM25 had no detectable in vivo effect on splicing of several tested endogenous genes. The effect of RBM25 in modulating the selection of Bcl-xS 5′ ss seems to strictly depend on its specific interaction with an exonic element, CGGGCA, situated 64 to 69 nt upstream of the Bcl-xS 5′ ss. Thus, we propose that RBM25 is a substrate-specific splicing regulator rather than an essential constitutive splicing factor. Database searches revealed that the sequence motif CGGGCA is present in many alternatively, as well as constitutively, spliced exons. Whether RBM25 exerts similar functional activity in the splicing of these exons is yet to be determined.
The contribution of RBM25 becomes negligible when CGGGCA is associated with a strong 13S 5′ ss in an E1A reporter or when base-pairing complementarity between Bcl-xS 5′ ss and U1 snRNA is improved. Therefore, the necessity of RBM25 to activate the weak 5′ ss implies that binding to CGGGCA facilitates 5′ ss recognition by U1 snRNP. RBM25 has been shown to associate with U1 snRNAs (16). Our immunoprecipitation and GST-pull down results suggest that these associations are most likely not through the association of RBM25 with U1-specific proteins (U1A, U1C, or U1-70K) but via interactions with hLuc7A, the human homolog of yeast U1 snRNP component yLuc7p. RBM25-containing speckles are also precisely colocalized with hLuc7A nuclear structures, which further suggests cross talk between these proteins. Several lines of evidence support a role for Luc7 in splicing regulation in both yeast and mammalian cells. Yeast yLuc7p is required for commitment complex formation in vitro and is necessary for weak 5′ ss recognition by U1 snRNP (15). Mammalian hLuc7A is loosely associated with the human U1 snRNP and affects 5′ ss selection by stabilizing U1 snRNP-pre-mRNA interactions, as well as being involved in the formation of the E complex (36). Moreover, hLuc7A has recently been identified as a component of the supraspliceosome (9) and interacts with splicing factors RNPS1 (40) and SRrp53 (6).
Our identification of the association between hLuc7A and RBM25 provides insight into the mechanisms of RBM25 splicing activation. Binding of RBM25 to the exonic sequence CGGGCA could activate a weak 5′ ss through its interaction with hLuc7A, which directly binds to the U1 snRNP necessary for 5′ ss recognition. This activity would be reminiscent of TIA-1, which binds to U-rich sequences downstream of the weak 5′ ss and facilitates recruitment of U1 snRNP and production of the proapoptotic Fas receptor isoform (13). Future experiments are required to define in further detail the nature and chronology of the interactions among RBM25, hLuc7A, U1 snRNP, and the 5′ ss.
The molecular structure of RBM25 suggests that it contains domains essential for splicing regulation. The nuclear localization signal prediction program (http://cubic.bioc.columbia.edu/cgi/var/nair/resonline.pl) identifies three putative signals situated within the RRM domain (aa 233 to 238), RE/RD domain (aa 414 to 426), and PWI domain (aa 706 to 711). We found that each signal governs the nuclear localization of its individual domain. In addition, the RE/RD-rich domain is responsible for the nuclear speckle localization of RBM25. It has been shown that replacement of the RS domain of SF2/ASF with an RE or RD domain also mediates proper nuclear speckle localization, albeit to a lesser extent (7). The repeats in RBM25 may thus serve as a targeting signal to nuclear speckles.
Our studies also show that RBM25 and SC35 coredistribute to the enlarged speckles upon inhibition of transcription. As with other splicing factors, RBM25 can lose its interchromatin granule cluster connections in transcriptionally active nuclei and localize to the enlarged unconnected speckle domains in transcriptionally inactive nuclei during states of transcription inhibition (5, 34).
The alternation of positive and negative charges of the amino acid residues (arginine-aspartates in RD and arginine-glutamates in RE) is reminiscent of phosphorylated RS (arginine-phosphorylated serine) domains. Thus, the RE/RD-rich domain may also mimic the structural and functional properties of a natural phosphorylated RS domain that mediates protein-protein interactions. It has been shown that stable protein interactions between SAFB1/SAFB2 with Sam68/T-STAR are mediated primarily by their RE-rich regions and play an important role in in vivo selection of splice sites (44). We demonstrate that the carboxyl half of hLuc7A consists of multiple RE and RS repeats that interact with RBM25 in an RNA-independent manner. Whether RE/RD repeats of RBM25 are responsible for such interactions remains to be determined.
Alternative splicing plays a fundamental role in the control of apoptosis. Several pre-mRNAs for cell death factors are alternatively spliced, yielding isoforms with opposing functions during programmed cell death (42). The regulation of Bcl-x alternative splicing is of critical importance to the apoptotic process and is highly pertinent to the progression of cancer. The expression of Bcl-xL increases in a number of lymphomas (52), while Bcl-xS is downregulated in many types of cancer cells. Several cis elements and trans-acting factors have been shown to regulate Bcl-x isoform expression. hnRNP F/H (17) and Sam68 (35) induce Bcl-xS expression, whereas SAP155 (28) favors the selection of Bcl-xL. Furthermore, Bcl-x splicing is coupled to signal transduction, since ceramide (8, 28) and protein kinase C (37) also affect splicing regulation. The ratio of the Bcl-xS to the Bcl-xL isoform will likely be dictated by the coordinated contribution of both cis- and trans-acting factors. Even though RBM25 is abundantly expressed in many of the cell types examined, its expression was found to be regulated during erythroid differentiation. As a result, regulated expression of RBM25 may play a critical role in mediating temporal and/or spatial splicing decisions of its target genes within the cell. Moreover, changes in the expression of any of these factors may influence how cells respond to apoptotic signals.
A growing family of RRM-containing proteins has been attributed to modulate apoptosis (50). In this report, we not only identified a new exonic splicing enhancer element for Bcl-xS splicing but also uncovered an important splicing regulator, RBM25, which interacts with this exonic splicing enhancer element to enhance the weak 5′ ss selection, most likely through its interaction with a U1 snRNP-associated protein, hLuc7A. We suggest that RBM25 is an RNA-binding apoptosis regulator. The mechanism of alternative 5′ ss selection of Bcl-x pre-mRNA has emerged as a potential target for anticancer therapeutics (51). Our studies will lead to greater insights into the molecular mechanisms by which RBM25 acts as a modulator of apoptosis.
We thank O. Puig (European Molecular Biology Laboratory, Germany) for the hLuc7A plasmid, C. E. Chalfant (Virginia Commonwealth University, Richmond) for the Bcl-x minigene, and W. Y. Tarn (Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan) for the E1A minigene. We thank H. R. Widlund (Dana-Farber Cancer Institute, Boston, MA) for valuable suggestions throughout this project.
This work was supported by a Claudia Barr Award (S.C.H.) and NIH grant HL24385 (E.J.B.).
Published ahead of print on 28 July 2008.