Downregulation of splicing factor SRSF3 induces p53β, an alternatively spliced isoform of p53 that promotes cellular senescence

Abstract

Most human pre-mRNA transcripts are alternatively spliced, but the significance and fine-tuning of alternative splicing in different biological processes is only starting to be understood. SRSF3 (SRp20) is a member of a highly conserved family of splicing factors that have critical roles in key biological processes, including tumor progression. Here, we show that SRSF3 regulates cellular senescence, a p53-mediated process to suppress tumorigenesis, through TP53 alternative splicing. Downregulation of SRSF3 was observed in normal human fibroblasts undergoing replicative senescence, and was associated with the upregulation of p53β, an alternatively spliced isoform of p53 that promotes p53-mediated senescence. Knockdown of SRSF3 by short interfering RNA (siRNA) in early-passage fibroblasts induced senescence, which was associated with elevated expression of p53β at mRNA and protein levels. Knockdown of p53 partially rescued SRSF3-knockdown-induced senescence, suggesting that SRSF3 acts on p53-mediated cellular senescence. RNA pulldown assays demonstrated that SRSF3 binds to an alternatively spliced exon uniquely included in p53β mRNA through the consensus SRSF3-binding sequences. RNA crosslinking and immunoprécipitation assays (CLIP) also showed that SRSF3 in vivo binds to endogenous p53 pre-mRNA at the region containing the p53P-unique exon. Splicing assays using a transfected TP53 minigene in combination with siRNA knockdown of SRSF3 showed that SRSF3 functions to inhibit the inclusion of the p53β-unique exon in splicing of p53 pre-mRNA. These data suggest that downregulation of SRSF3 represents an endogenous mechanism for cellular senescence that directly regulates the TP53 alternative splicing to generate p53β. This study uncovers the role for general splicing machinery in tumorigenesis, and suggests that SRSF3 is a direct regulator of p53.

INTRODUCTION

Alternative RNA splicing (AS) is the alternative selection of splicing sites within precursor mRNAs to produce multiple mature RNA isoforms from a single primary transcript. AS promotes proteomic diversity in temporal and tissue-specific manners,^1–2 and most human genes undergoes AS.^2–4 In some cases, the protein products of splicing variants cooperate with full-length proteins.⁵ In others, they may show different or even opposing functions, thus offering plasticity and a fast response to cellular physiological function and stress.^5,6 Aberrant AS may also have a key role in tumorigenesis,^7–14 pluripotency¹⁵ and cell fate determination.^16–18 To date, the role of AS in the regulation of oncogenes and tumor suppressors has also attracted considerable interest.^19,20 This evidence also suggests that functionally coherent genes may be co-regulated by one or a few RNA-binding splicing factors to promote a specific biological process, and may explain the pro-oncogenic behavior of some splicing factors.^21–28

Cellular senescence, an irreversible proliferation arrest, is a physiological response to various types of stress, including telomere shortening, DNA damage and oncogene activation.^29–33 Cellular senescence is also an endogenous mechanism to suppress tumorigenesis in vivo, working alongside cell death programs.^34–37 Few studies to date have focused on AS that occurs during cellular senescence, and only a few of genes have been revealed to undergo AS with functional consequences to cellular senescence.^5,37–39 The p53 signaling pathway has a critical role in the activation and regulation of cellular senescence. The human TP53 gene generates at least 12 different protein isoforms.^40–45 We previously reported that the p53β isoform generated by AS of TP53 pre-mRNA was expressed during replicative cellular senescence, and induced cellular senescence when overexpressed.⁵ However, the mechanism of TP53 pre-mRNA splicing to produce the p53β isoform during cellular senescence is unknown.

The RNA-binding protein SRSF3 (previously named as SRp20) is the smallest member of the highly conserved serine/arginine-rich splicing factor family. It is characterized by one RNA recognition motif at the N-terminus and one signature serine/arginine or arginine/serine dipeptide repeat (RS domain) at the C terminus.^46–50 SRSF3 modulates the splicing of numerous genes.^{28,46,51–55} Recent studies proposed pro-oncogenic activity of SRSF3 on the basis of upregulation in various types of human cancer^28,56 and its ability to promote cell neoplastic transformation when overexpressed,²⁸ indicating an important function of SRSF3 in tumorigenesis. Here, we investigate the role of the splicing factor SRSF3 in the regulation of TP53 pre-mRNA splicing in the context of cellular senescence. Our data reveal that downregulation of SRSF3 drives TP53 pre-mRNA splicing towards the p53β isoform, and represents an endogenous mechanism to promote cellular senescence.

RESULTS AND DISCUSSION

Endogenous expression of SRSF3 mRNA and protein were examined in normal human fibroblast strains (MRC-5 and WI-38)at early passages (MRC-5 at population doublings (PDL) 40 and WI-38 at PDL 34, Y in Figure 1) and at replicative senescence (MRC- 5 at PDL 66 and WI-38 at PDL 48, S in Figure 1). Cells undergoing replicative senescence were characterized by the hallmark activity of senescence-associated β-galactosidase (SA-β-gal) (Supplementary Figure S1a). The senescent human fibroblasts showed accumulation of p53β (Figure 1b) without a significant change in the amounts of full-length p53 protein.^5,57 SRSF3 mRNA and protein were reduced in replicatively senescent cells (Figures 1a and b). Similarly, SRSF3 was downregulated in replicatively senescent T lymphocytes isolated from healthy blood donors (Supplementary Figure S1b). On the contrary, premature senescence in fibroblasts induced by oncogenic Ras did not affect the expression of SRSF3 (Supplementary Figure S1c) or p53β.⁵ These data suggest that decreased SRSF3 expression and increased p53β are inversely correlated phenotypes related to replicative cellular senescence.

Figure 1.

Replicative senescence-associated changes in expression of endogenous SRSF3. (a) Downregulation of SRSF3 mRNA during replicative senescence. SYBR green qRT-PCR analysis of SRSF3 were performed in early-passage (Y) and replicatively senescent (S) human fibroblast strains MRC-5 and WI-38 using primers 5’-AGC TGATGC AGT CCG AGA G-3’ and 5’-GGT GGG CCA CGATTT CTA C-3’. Cells were examined at PDL 40 (Y) and 66 (S) for MRC-5; and 34 (Y) and 48 (S) for WI-38. The qRT-PCR data were normalized to control GAPDH mRNA (TaqMan control assay 4352934E) and shown as the relative SRSF3 mRNA level (mean ± s.d. from triplicate samples). **P< 0.001. (b) Downregulation of SRSF3 protein and upregulation of p53β protein during replicative senescence. Thirty microgram of protein lysates from the same set of cells as in (a) were electrophoresed and immunoblotted with antibodies for SRSF3 (clone 7B4 from American Type Culture Collection, Manassas, VA, USA) and p53β (TLQ40^5,40). β-actin was a sample loading control. SRSF3 and p53β protein levels were normalized to β-actin levels and were shown as relative values below each band. Note that full- length p53 protein levels were not significantly changed during replicative senescence in these fibroblasts.^5,57

To examine the role of endogenous SRSF3 in cellular senescence, short interference RNA (siRNA) was used to specifically knockdown endogenous expression of SRSF3 in early-passage MRC-5 (Figure 2). Two siRNA oligonucleotides (siSRSF3 no. 1 and siSRSF3 no. 2), targeting two different regions of SRSF3 mRNA, efficiently knocked down endogenous SRSF3 mRNA expression by B 90% and 70%, respectively (Figure 2a). The knockdown of SRSF3 by the two siRNAs was also verified at protein level (Figure 2b). The cells transfected with siSRSF3 no. 1 and siSRSF3 no. 2, but not those with a control siRNA (siControl), underwent a senescent growth arrest rapidly (within 7 days) with concomitant induction of the SA-β-gal activity (Figures 2c and d. Supplementary Figures S2a and b). SRSF3-knockdown-induced senescence exhibited p53 phosphorylation at serine 15 (Figure 2b), a posttranslational modification associated with p53 activation,^5,58–62 without a significant increase in p53 protein amounts (Figure 2b). Knockdown of SRSF3 also increased p53P protein level (Figure 2b), reminiscent of its accumulation in replicative senescence as in Figure 1. When the alternative splicing patterns of TP53 (Figure 2e) were examined by conventional RT-PCR and qRT-PCR, mRNA levels of p53β were remarkably upregulated upon SRSF3 knockdown, while the levels of full- length p53 mRNA were moderately decreased (Figures 2f and g). Of note, p53γ mRNA fell below the detection limit of our RT-PCR assay amplifying exons 7 through 10 (p53, Figure 2f), suggesting that p53γ alternative splicing is not a major event in normal human fibroblasts. Conversely, overexpression of SRSF3 reduced p53β mRNA levels (Figure 2h).

Figure 2.

Knockdown of endogenous SRSF3 induces cellular senescence and p53β expression, (a) siRNA knockdown of SRSF3. Two predesigned silencer select siRNA duplex oligoribonudeotides targeting SRSF3 mRNA coding region (siSRSF3 no. 1, si 2732,5’-AGA GCU AGA UGG AAG AAC ATT-3’; and siSRSF3 no. 2, s12733, 5’-GCA ACA AGA CGG AAU UGG ATT-3’), and a nonspecific negative control siRNA (siControl, 4390843) were purchased from Applied Biosystems (Foster City, CA, USA). Early-passage MRC-5 fibroblasts (at PDL 34) were transfected twice at day 1 and day 4 with siSRSF3 no. 1, siSRSF3 no. 2 and siControl at a final concentration of 2, 10 and 10nM, respectively, by using the Lipofectamine RNAiMAX transfection reagent (Invitrogen, Grand Island, NY, USA) according to the supplier’s protocol. Total RNA samples were prepared at day 7 and were analyzed by qRT-PCR. Data (mean ± s.d from triplicate samples) are shown as the relative values to control cells (siControl). (b) Knockdown of SRSF3 induces p53β protein and p53 phosphorylation. Protein expression levels of SRSF3 (7B4), p53β (TLQ40), full-length p53 (DO-1) and p53 phosphorylated at serine 15 (p53-pS15; Cell Signaling, Danvers, MA, USA, no. 9284) were examined. β-actin was a loading control, (c) Representative pictures of SA-β-gal staining. Senescence β-Galactosidase Staining Kit (Cell Signaling, no. 9860) was used according to the supplier’s protocol, (d) Quantitative analysis of SA-β-gal staining. At least 200 cells from 4–6 random fields (randomly chosen microscopic fields containing at least 20 cells per field) were observed and counted using a standard light microscope. Perinuclear blue staining beyond the background level in a vast majority of control cells was regarded as positive. The data (mean ± s.d.) were from three independent experiments. **P< 0.001 obtained from Student’s t-test. The results are consistent with those from an image analysis-based quantification (Supplementary Figure S2a) and from visual counting by an independent observer (Supplementary Figure S2b). (e-g) Increased expression of p53p mRNA in SRSF3-knockdown cells, (e) Schematic diagram of TP53 gene structure from exon 7 to exon 10. Three alternative splicing patterns (α, β and γ) are indicated along with the primer pair specific for p53β (red arrows; 5’-CTT TGA GGT GCG TGT TTG TGC-3’ (forward) and 5’-TTG AAA GCT GGT CTG GTC CTG A-3’ (reverse, encompassing exons ¡9 and 9)) and the primer pair for all three splicing patterns (labeled as ‘p53’, black arrows; 5’-CTC ACC ATC ATC ACA CTG GAA-3’ (forward) and 5’-TCATTC AGC TCT CGG AAC ATC-3’(reverse)).’^s The red dot in exon i9 indicates the stop codon for p53β translation, (f) Conventional RT-PCR analysis using the primer pairs shown in (e). Biological triplicates of siRNA-transfected cells (siSRSF3 no. 1, siSRSF3 no. 2 and siControl) were examined. GAPDH served as a control ((5’- CCA TCT TCC AGG AGC GAG A-3’ (forward) and 5’-TGT CAT ACC AGG AAA TGA GC-3’ (reverse)).⁵ Note that p53γ splicing was below the detection limit in this assay (middle panel), (g) p53β mRNA was quantified by SYBR Green-based qRT-PCR using the p53β-specific primer pair in (e). Overall p53 level was quantified by qRT-PCR using TaqMan Gene Expression Assay (Applied Biosystems HS00153340_m1). The expression levels of p53β and p53 were normalized with GAPDH (Applied Biosystems 4333764F) and expressed as the relative values to siControl (defined as 1.0). *P<0.01; **P<0.001. P-values were obtained from Student’s t-test. (h) Overexpression of SRSF3 reduces p53β mRNA. Early-passage MRC-5 fibroblasts were transduced with a lentiviral vector driving SRSF3 or control vector. Cells were collected 72 h post-transduction and analyzed by the p53β-specific qRT-PCR as in (g). **P< 0.001. Cell death of currently unknown origin was observed 7 days post-transduction and the replicative lifespan of the cells could not be evaluated.

To investigate whether SRSF3 knockdown-induced senescence is p53-dependent, endogenous SRSF3 expression was knocked down in early-passage MRC-5 fibroblasts by siRNA (siSRSF3) with or without the cotransfection of a p53 siRNA (siP53). The knockdown effects on SRSF3 and p53 were confirmed by qRT- PCR assays (Figures 3a and b). We found that knockdown of p53 led to partial rescue of SRSF3 knockdown-induced senescence in MRC-5 fibroblasts, as indicated by decreased SA-β-gal staining (Figures 3c and d, and Supplementary Figures S2c and d). Similarly, knockdown of SRSF3 was performed in hTERT/NHF and hTERT/NHF-p53shRNA cells, an hTERT (human telomerase reverse transcriptase)-immortalized human fibroblast cell line transduced with control vector or p53 shRNA. We observed that SRSF3- induced senescence was decreased in hTERT/NHF-p53shRNA cells compared with hTERT/NHF cells (Supplementary Figures S2e and f). These results indicate that loss of SRSF3 leads to senescence at least in part through a p53-dependent mechanism, consistent with the upregulation of p53P (Figure 2). Altogether, our findings suggest that endogenous expression of SRSF3 is critical to the replicative potential of normal human fibroblasts, and that downregulation of SRSF3 has a physiological role in the induction of cellular senescence.

Figure 3.

SRSF3 knockdown-induced senescence is p53-dependent. SRSF3 expression was knocked down in early-passage MRC-5 fibroblasts, as described in Figure 2, except that mixture of siSRSF3 no. 1 and siSRSF3 no. 2 was used as siSRSF3 in this experiment. Knockdown of p53 was carried out by an siRNA oligonucleotide siTP53, 5’-GAC UCC AGU GGU AAU CUA Ctt-3’, which targets amino-acid residues 259–265 of full-length p53 and p53β. (a) Knockdown of SRSF3 was confirmed by qRT-PCR as in Figures 1a. (b) Knockdown of p53 was confirmed by qRT-PCR as in Figure 2g. (c) Representative pictures of SA-β-gal staining, (d) Quantitative analysis of SA-β-gal staining was carried out as in Figure 2d. The data (mean ± s.d.) were from three independent experiments. *P<0.01 was obtained from was obtained from Student’s t-test. The results are consistent with those from an image analysis-based quantification (Supplementary Figure S2c) and from visual counting by an independent observer (Supplementary Figure S2d).

We next explored the regulatory role of SRSF3 in TP53 pre- mRNA alternative splicing at the ¡9 exon, which is uniquely included in p53p transcripts. By using Splicing Rainbow (EMBL-EBI Alternative Splicing Workbench, http://www.ebi.ac.uk/asd-srv/wb.cgi), a SELEX database-based tool that identifies binding sites for splicing factors on RNA sequence, we identified UUUCAAA, UACUUGAC and UACUUCCU in p53β-specific exon i9 sequence as three putative consensus binding sites (BS) for SRSF3. We carried out RNA pulldown assays using biotin-labeled RNA oligomers derived from the exon ¡9 with or without point mutations in the putative SRSF3 BS (Figure 4a and Supplementary Figure S3a). RNA oligomers rl, r4 and r5 containing a putative SRSF3 BS specifically pulled-down SRSF3 protein from nuclear extracts of HEK293 cells or total cell lysates from mouse embryonic fibroblasts (Supplementary Figure S3b). The three RNA oligomers (r2, r3 and r6) derived from other regions in exon i9 did not or only minimally bind to SRSF3 protein (Supplementary Figure S3b). In agreement with the previous reports showing that the phosphorylation of SR splicing factors is required for their interaction with RNA,^63–70 dephosphorylation by treatment with calf intestine phosphatase abrogated the ability of SRSF3 to bind to RNA oligomers r1, r4 and r5 (data not shown). Introduction of mutations into the SRSF3 BS in RNA oligomers r1, r4 and r5 (r1*, r4* and r5*_, Figure 4a) partially or fully abolished the binding of SRSF3 to these RNAs (Figure 4b). These data suggest the sequence-specific binding of SRSF3 to the p53β-unique exon i9 on p53 pre-mRNA.

Figure 4.

SRSF3 binds to exon i9 and prevents its inclusion during p53 pre-mRNA splicing, (a) Identification of putative SRSF3-binding sites (BS) in the i9 exon and their point mutations. Biotin-labeled RNA oligomers with (r1*, r4* and r5*) or without (r1, r4 and r5) the corresponding mutations were synthesized and HPLC (RNase-free) purified by Integrated DNA Technologies (Coralville, IA, USA) for RNA pulldown assay. See also Supplementary Figure S3a. (b) SRSF3 interacts with SRSF3 BS in the i9 exon. RNA pulldown assay was performed and analyzed as previously described.⁷⁵ Pulled-down proteins from HEK293 nuclear extracts (ProteinOne, Bethesda, MD, USA) were Immunoblotted with anti-SRSF3 antibody (7B4). r2 (Supplementary Figure S3a) was used as a negative control, (c, d) RNA-CLIP assay to demonstrate that SRSF3 binds to exon i9 of endogenous p53 pre-mRNA. SRSF3 protein with N- or C-terminal FLAG tag was expressed in HEK293T cells. After UV-crosslinking, cell lysates were prepared and used in immunoprecipitation with anti-FLAG antibody (M2, Sigma-Aldrich, St Louis, MO, USA). Cells with vector alone were included as a negative control, (c) Cell lysates before (Input) and after immunoprecipitation (CLIP-product) were analyzed in western blot with anti-FLAG antibody. The expression and immunoprecipitation of FLAG-tagged SRSF3 proteins were confirmed. β-actin was a loading control, (d) Total RNA samples were prepared from cell lysates before (Input) and after immunoprécipitation (CLIP-product) and subject to RT-PCR amplifying the exon i9- containing region of p53 pre-mRNA (using i9F7 and i9R5 primers in Supplementary Figure S4a). PCR products appeared only in the presence of reverse transcriptase (RT + ), but not in the absence (RT— ), excluding a possibility of amplification from contaminated genomic DNA. The upper bands were identified as specific by nested PCR (Supplementary Figure S4b) and direct DNA sequencing (Supplementary Figure S4c), while the lower bands were nonspecific. Full details of RNA-CLIP assay procedures are available in Supplementary information, (e) Diagram of human TP53 minigene expression vector containing sequences from exon 7 to exon 10. TP53 genome sequences from exon 7 to exon 10 were PCR-amplified from genomic DNA of MRC-5 and cloned into pCMV-tag 2B vector (Stratagene, La Jolla, CA, USA), in which the expression of cloned sequences is driven by CMV immediate early (IE) promoter. Black arrows indicate the primer pair to amplify all minigene-derived p53 transcripts (5’-CGG GCG GAT CCG TTG G-3’, forward primer ‘E7’, spanning the vector and exon 7 sequence; and 5’-CCC TCG AGC TGG AGT G-3’, reverse primer ‘El o; spanning the vector and exon 10 sequence). A red arrow indicates the reverse primer ‘Ei9’ to specifically detect p53p splicing (5’-TTG AAA GCT GGT CTG GTC CTG A-3’, encompassing exons ¡9 and 9). (f ) MRC-5 fibroblasts were transfected with the p53 minigene construct, together with SRSF3 siRNA (siSRSF3 no.1 in Figure 2) or control siRNA (siControl). qRT-PCR assays were performed using the above primer pairs (E7-E10 and E7-Ei9). The expression of neomycin-resistant gene (derived from pCMV-tag 2B vector) was also quantitated for normalization of transfection efficiencies (see Supplementary Figures S5a and b, and the figure legend). The levels of p53β (detected by E7-Ei9) in relative to all minigene-derived transcripts (detected by E7-E10) were calculated from the data shown in Supplementary Figures S5a and b. Data are mean ± s.d. from three independent experiments and shown as the relative value to siControl. **P< 0.001 obtained from Student’s t-test

We performed an RNA-CLIP assay (crosslinking and immuno-precipitation of RNA-protein complexes)^71,72 to show in vivo binding of SRSF3 to endogenous p53 pre-mRNA (Figures 4c and d and Supplementary Figure S4). When SRSF3 protein tagged with an N- or C-terminal FLAG in HEK293T cells was immunoprecipi- tated with anti-FLAG antibody (Figure 4c) and co-immunopreci- pitated RNA samples were analyzed by RT-PCR amplifying the region containing exon i9 (¡9F7 and ¡9R5 primers. Supplementary Figure S4a), both FLAG-tagged SRSF3 proteins were found to specifically bind to the RNA region (Figure 4d). The specificity of the binding was further confirmed by nested PCR (Supplementary Figures S4a and b) and sequencing analysis of its products (Supplementary Figure S4c).

To directly examine how SRSF3 regulates alternative splicing of p53 pre-mRNA, a CMV immediate early (IE) promoter-driven minigene was constructed, which contains human TP53 gene sequences from exon 7 (E7) to exon 10 (E10) with corresponding intervening intron sequences (Figure 4e). Early-passage MRC-5 fibroblasts were transiently transfected with the minigene construct, in combination with siRNA against SRSF3 (siSRSF3) or control siRNA (siControl), and examined by qRT-PCR for alternatively spliced p53P transcripts (E7-Ei9 primer pair), all minigene-derived transcripts including those corresponding to full-length p53 mRNA and p53β mRNA (E7-E10 primer pair), and neomycin- resistant gene transcript as a control for transfection efficiency. While siSRSF3 reduced RNA levels detected by E7-E10 (Supplementary Figure S5a), it upregulated p53β transcripts detected by E7-Ei9 (Supplementary Figure S5b), resulting in approximately sixfold increase in the relative level of alternative splicing generating p53β (Figure 4f). The E7-E10 primer pair was also used in conventional RT-PCR, which showed siSRSF3-induced shift from full-length p53 to p53β transcripts (Supplementary Figure S5c). These results suggest that SRSF3 functions to prevent the inclusion of exon ¡9 during p53 pre-mRNA splicing, thereby inhibiting the production of p53β isoform.

In summary, this study provides the first evidence that splicing factor SRSF3 is critical to the replicative potential of normal human fibroblasts, and physiologically regulates cellular senescence by regulating the alternative splicing of p53 pre-mRNA. Down- regulation of SRSF3 in aging cells elevates p53β level and constitutes a physiological mechanism that promotes senescence. Our study is consistent with the hypothesis that AS is an important regulatory mechanism in tumor suppression and progression, and thus dysregulation of general splicing factors can have critical roles in tumorigenesis. Better understanding of SRSF3 and other splicing regulatory molecules would help us better understand aging and tumorigenesis, and hopefully aid us for cancer therapeutics. The underlying mechanism leading to reduced expression of SRSF3 in replicatively senescent cells remains to be understood. As SRSF3 is transactivated by E2F,⁴⁶ a transcription factor that is frequently inactivated during senescence,^60,73,74 we speculate that E2F inactivation in the early stage of senescence may result in SRSF3 downregulation. Further studies that characterize the possible link between downregulated SRSF3 and inactivated E2F in cellular senescence may reveal a new mechanism through which cellular senescence harnesses the AS program to promote its own progression.