Regulation of oncogenic C-terminal truncated p53β protein isoform expression by SRSF3–UPF1 splicing and surveillance axis

Abstract

Dysregulated expression of tumor suppressor genes can impair their functions, even promoting oncogenesis. Expression of p53β mRNA can be regulated by alternative splicing and RNA surveillance, otherwise translated to a C-terminal truncated p53 protein with a unique neoepitope. Here, we identified that p53 introns bear the binding sites for serine and arginine-rich splicing factor 3 (SRSF3). SRSF3 binding to p53 intron 9 facilitates upstream frameshift 1 (UPF1) recruitment and regulates production of p53β mRNA. We also demonstrated that this ternary ribonucleoprotein complex forms cotranscriptionally in chromatin. SRSF3 depletion disrupts SRSF3-UPF1 axis, elevating the levels of p53β mRNA encoding a C-terminal-truncated isoform. Intriguitly, p53β protein isoform lacks tumor-suppressive activity and promotes oncogenic epithelial–mesenchymal transition through enhanced cell migration and invasion. We define the coordinated roles of the splicing factor SRSF3 and the RNA surveillance factor UPF1 in regulating p53 mRNA isoform expression, thereby linking splicing fidelity with RNA surveillance during transcription. Our findings highlight the SRSF3-UPF1 axis for preventing oncogenic p53β protein isoform accumulation, offering a potential therapeutic target to restore p53 function and impede cancer progression.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13578-026-01556-5.

Introduction

Disruptions in pre-mRNA splicing and RNA surveillance can produce altered isoforms of tumor suppressor proteins, impairing their functions or enabling oncogenic pathways. The tumor suppressor p53 protein orchestrates antiproliferative transcriptional programs and genomic integrity maintenance during oncogenesis [1, 2]. Expression of p53 protein is regulated at multiple posttranscriptional layers, including alternative promoter, alternative splicing and alternative translation initiation [3–6]. The human TP53 gene (11 exons and 10 introns) generates many transcript variants, underscoring prominent alternative RNA processing of p53 mRNAs [7]. Among these variants, the canonical p53α mRNA encodes a full-length p53 with tumor-suppressive activity through proper intron 9 excision. In contrast, the transcripts p53β and p53γ arise from intron 9 that includes cryptic exons harboring premature termination codons (PTCs) [8, 9]. These variant transcripts are assumed to be subject to nonsense-mediated mRNA decay (NMD) that limits the expression of truncated, potentially deleterious proteins. However, the mechanisms that determine the fate of p53 mRNAs and and the expression of their protein isoforms, remain incompletely understood.

Serine and arginine-rich (SR) family proteins are splicing regulators and facilitate multiple mRNA processing [10–12]. Serine/arginine-rich splicing factor 3 (SRSF3; formerly SRp20) is the shortest SR protein, containing a single RNA recognition motif [13] and a short SR domain. SRSF3 participates in diverse RNA processing steps, including splicing, stability, nuclear export, and translation [12, 14–17]. SRSF3 may act in either oncogenic or tumor-suppressive pathways depending on the cell context and RNA targets [18–22]. SRSF3 downregulation has been associated with increased production of p53β, which contributes to cell senescence [23, 24]. The C-terminal domain (CTD) of p53 protein is involved in tetramerization essential for transactivation, and interactions with other proteins through multiple posttranslational modifications [25]. CTD-truncated p53 protein (ΔCp53) impairs sequence-specific DNA binding and p53 target-gene subset expression, thereby promoting neuroblastoma development [26, 27]. Similarly, a p53β mutant lacking N-terminal 133 amino acids (Δ133p53β) causes tumor aggressiveness, epithelial–mesenchymal transition (EMT), and invasion [28–30]. However, despite structural similarities between p53β and ΔCp53 protein isoforms, the mechanisms underlying p53β protein isoform expression and its precise functional role in cancer remain poorly understood. Given the central role of p53 mRNAs in cell-level stress responses, isoform-specific regulation of p53 transcripts warrants further investigation [1, 31, 32].

Upstream frameshift 1 (UPF1), a central NMD effector, is a cytoplasmic RNA helicase that degrades aberrant mRNAs in a translation-dependent manner [33, 34]. Loss or inactivation of UPF1 contributes to cancer progression by allowing the accumulation of NMD-targeted transcripts [35]. Beyond canonical NMD, UPF1 plays broader roles in RNA quality control and ribonucleoprotein (RNP) remodeling, including functional cooperation with splicing regulators such as SRSF1 [36]. Recent studies have expanded the roles of UPF1 to the nucleus, including genome stability maintenance and DNA damage response [33]. Nevertheless, the precise functions of nuclear UPF1, particularly transcription-coupled RNA surveillance, remain largely underdefined.

In this study, we examined the posttranscriptional regulation of p53 transcript variants, focusing on the alternative splicing–mRNA surveillance interplay. We identified a p53 intron-retained (p53-IR) transcript that accumulates upon UPF1 depletion; it retains an intron adjacent to a cryptic exon, a region highly enriched for SRSF3 binding. Our findings show that SRSF3 and UPF1 cooperate during RNA surveillance to suppress p53β mRNA expression via splicing-linked mechanisms. Notably, UPF1 interacts with SRSF3 in a manner enhanced by transcription elongation inhibition, suggesting a cotranscriptional surveillance machinery. Depletion of SRSF3 elevates p53β mRNA and isoform protein levels, which could promote EMT. These findings define a novel splicing surveillance axis governed by the SRSF3–UPF1 axis that controls p53 expression and may serve as a regulatory node in tumor progression.

Materials and methods

Plasmids

The plasmids pCMV-Myc-UPF1 and pcDNA-Flag-UPF1 wild-type (WT) were kindly provided by Dr. Yoon Ki Kim (Korea Advanced Institute of Science and Technology, Daejeon, South Korea). To construct 3xFlag-CMV10-SRSF3, SRSF3 cDNA was amplified from HeLa (human cervical carcinoma) cell-derived cDNA using Pfu DNA polymerase (Thermo Fisher #EP0501) and inserted into the 3xFLAG-CMV10 vector via EcoRI and BamHI restriction sites.

To generate the Dup4-1-p53 (Dup-p53) minigene construct, a genomic fragment of the human TP53—spanning 182 nucleotides (nt) of intron 9a (I9a), the full 195 nt of exon 9b (E9b), and 839 nt of intron 9b (I9b)—was amplified using Pfu DNA polymerase and cloned into the Dup4-1 plasmid (a gift from Dr. Douglas Black, University of California, Los Angeles, CA, USA) via ApaI and BglII restriction sites. The Dup-p53 plasmid was subjected to site-directed mutagenesis to either mutate a PTC or to predict SRSF3 binding motifs. The PTC mutant construct was referred to as GO, and SRSF3 binding-site mutants were referred to as Mutant 1 (Mut 1), Mut 2, Mut 3, and Mut 4, corresponding to mutations in I9a, E9b, I9b, or both I9a and E9b, respectively. Primers used for mutagenesis are listed in Supplementary Table 1.

For the MS2 tethering assay, Dup-p53-MS2 constructs were generated by inserting an AgeI restriction site into I9a or I9b of the Dup-p53 backbone via site-directed mutagenesis, and then an MS2 hairpin sequence, which was amplified from the pSL-MS2-12X plasmid. To construct the MS2 binding protein-fused SRSF3 expression vector, SRSF3 cDNA amplified from HeLa cell-derived cDNA was inserted into the pcCMS2 vector (a gift from Dr. Lykke-Andersen, University of California, San Diego, CA, USA). Primers used for mutagenesis are listed in Supplementary Table 1.

Cell culture and siRNA transfection

HeLa cells were cultured in DMEM high-glucose medium (Hyclone). Colorectal adenocarcinoma cell lines—SW480 and HCT116—were maintained in RPMI-1640 medium (Hyclone). All media were supplemented with 10% fetal bovine serum (FBS; Hyclone) and 1% penicillin–streptomycin (Hyclone). Plasmid DNA was transfected using Lipofectamine™ 2000 (Invitrogen), and siRNAs were transfected utilizing Lipofectamine™ RNAiMAX (Invitrogen), according to the manufacturer’s instructions. The sequences of siRNAs targeting SRSF3, SRSF1, UPF1, and the siGFP control are provided in Supplementary Table 2.

RT-PCR and RT-qPCR

Total RNA was extracted employing the Quick-RNA MiniPrep Plus kit (Zymo Research) per manufacturer’s instructions. For cDNA synthesis, 2.0 μg of total RNA was reverse transcribed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher) with random hexamer primers (Thermo Fisher). RT-PCR or quantitative PCR (qPCR) was performed utilizing Premix Taq DNA polymerase (Tech & Innovation) or PowerUp™ SYBR™ Green Master Mix (Thermo Fisher) on a StepOnePlus Real-Time PCR System (Applied Biosystems), respectively. Sequences of the primers used for RT-PCR and RT-qPCR are listed in Supplementary Table 3.

Western blots

Cells were lysed in RIPA lysis buffer containing 25 mM Tris–HCl at pH 7.6, 150 mM NaCl, 1% IGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% SDS, and protease inhibitor cocktail (Sigma Aldrich). Samples were separated on 10% or 12% SDS-PAGE gels and then transferred to PVDF membranes (Millipore). The antibodies used were as follows- anti-UPF1 (Cell Signaling, #12,040), anti-SRSF3 (Invitrogen, #33–4200), anti-FLAG-M2 (Sigma, #F3165), anti-SRSF1 (Thermo Fisher, #32–4500), anti-Histone H3 (Abcam, #ab1791), anti-α-tubulin (Santa Cruz, #sc-8035), anti-β-actin (Abcam, #ab6276), anti-Myc (Sigma, #05-724MG), anti-p53 (Santa Cruz, DO-1 #sc-126; Pab1801 #sc-98; BIORAD, DO-11 #MCA1704), and E-cadherin (Biosciences, #610,181). Horseradish peroxidase-conjugated goat antirabbit or mouse IgG (H + L) secondary antibodies (Thermo Fisher, #31,460 or #31,430) were used.

Immunofluorescence

HeLa cells were cultured on coverslips, rinsed with phosphate-buffered saline (PBS), fixed with 3.7% formaldehyde in PBS for 10 min at 4 °C, and permeabilized with methanol for 5 min at 4 °C. Cells were then rinsed three times in PBS containing 0.5% Tween-20 and blocked with 5% bovine serum albumin in PBS for 1 h at room temperature. After blocking, the cells were rinsed again and incubated overnight at 4 °C in a humidified chamber with primary antibodies: anti-UPF1 (Cell Signaling, #12,040), anti-SRSF3 (Invitrogen, #33–4200), antiserine 2-phosphorylated RNA polymerase II (Abcam, #ab5095), anti-trimethylation of histone H3 lysine 4 (Abcam, #ab8580), and anti-SRSF7 (MyBioSource, #MBS5316251). After three washes in PBS with 0.5% Tween-20, the coverslips were incubated with Alexa Fluor 488 or 568-conjugated secondary antibody (Invitrogen, #A327231 and #A21090). Cells were then washed three times with PBS for 5 min each and mounted onto glass slides using VECTASHIELD mounting medium with 4′,6-diamidino-2-phenylinodole (DAPI) for nuclear staining. Images were acquired using confocal fluorescence microscopy (Fluoview, FV3000; Olympus, Melville, NY, USA).

Subcellular fractionation

Subcellular fractionation was per a previously published protocol [21] with slight modifications. To isolate the chromatin fraction, cells (2 × 10⁷ cells/mL) were resuspended in hypertonic buffer—10 mM Tris–HCl (pH 7.5), 10 mM KCl, 1.5 mM MgCl₂, 0.5 mM DTT, and a protease–phosphatase inhibitor cocktail (PPI; Sigma, #PPC1010) and incubated on ice for 15 min. Cells were then centrifuged at 430 × g for 10 min at 4 °C. The pellet was resuspended in 1 mL of hypotonic buffer—50 mM Tris–HCl at pH 7.5, 150 mM NaCl, 2 mM MgCl₂, 0.3% IGEPAL CA-630 (Sigma, # I8896), and PPI—incubated on ice for 10 min, and centrifuged again at 950 × g for 10 min at 4 °C. The supernatant was collected as the cytoplasmic fraction. The pellet was then resuspended in 1 mL of a lysis buffer—50 mM Tris–HCl pH 7.5, 150 mM NaCl, 2 mM MgCl₂, 0.5% IGEPAL CA-630 and PPI—incubated on ice for 10 min, and centrifuged at 950 × g for 10 min at 4 °C. After discarding the supernatant, the nuclear pellet was further fractionated to separate the nucleoplasm and chromatin. The nuclear pellet was resuspended in 100 μL of buffer 1: 50% glycerol (v/v), 20 mM Tris–HCl at pH 7.9, 75 mM NaCl, 0.5 mM EDTA, 0.85 mM DTT, PPI, and 100 U of RNasin (Promega, #N2111). Next, 900 μL of buffer 2A—20 mM HEPES at pH 7.6, 300 mM NaCl, 0.2 mM EDTA, 1 mM DTT, 7.5 mM MgCl₂, 1 M urea, 1% IGEPAL CA-630, PPI, and 100 U of RNasin—was added. The sample was vortexed for 10 s, incubated on ice for 10 min, and centrifuged at 15,000 × g for 5 min at 4 °C; the supernatant was collected as the nucleoplasmic fraction. To isolate the chromatin fraction, 100 μL of buffer 1 was added to the remaining pellet, followed by 900 μL of buffer 2B (same composition as buffer 2 A, except for 1.5% but not 1% IGEPAL CA-630). After vortexing for 10 s and incubating on ice for 10 min, the sample was centrifuged at 15,000 × g for 5 min at 4 °C. The supernatant was discarded, and the pellet was washed twice with 600 μL of buffer 2A. Finally, the chromatin fraction was sedimented by centrifugation at 15,000 × g for 5 min at 4 °C and resuspended in 1 mL of buffer 3: 50 mM Tris–HCl at pH 7.4, 100 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, and 100 U RNasin. Before use, the nucleoplasmic and chromatin fractions were sonicated at 40% amplitude for 3 min—on 30 s/off 30 s, repeated three times—to shear genomic DNA.

Chromatin immunoprecipitation

Cells were exposed to ultraviolet (UV) radiation at 200 mJ/cm², washed with PBS, and harvested. Cross-linked cells were lysed in lysis buffer 1—50 mM HEPES–KOH at pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% IGEPAL CA-630, 0.25% Triton X-100, 1 mM PMSF, and PPI—and incubated for 5 min at 4 °C. Lysates were centrifuged at 2,000 rpm for 3 min at 4 °C. Pellets were resuspended in lysis buffer 2—10 mM Tris–HCl at pH 8.0, 200 mM NaCl, 1 mM EDTA, 1 mM PMSF, and PPI—and centrifuged at 2,000 rpm for 3 min at 4 °C. Nuclei were lysed in lysis buffer 3: 10 mM Tris–HCl at pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.1% Na-deoxycholate, 0.5% SDS, 1 mM PMSF, and PPI. Chromatin was fragmented by sonication—40% amplitude, on 30 s/off 30 s, repeated 5-times—on ice. The lysates were cleared by centrifugation at 13,000 rpm for 15 min at 4 °C. Supernatants were diluted fivefold with immunoprecipitation (IP) dilution buffer—20 mM Tris–HCl at pH 8.0, 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100—and incubated overnight at 4 °C with anti-FLAG-M2 affinity gel (Sigma, #A2220). Beads were sequentially washed, once with low-salt buffer (20 mM Tris–HCl at pH 8.0, 150 mM NaCl, 2 mM EDTA at pH 8.0, 1% Triton X-100, and 0.1% SDS), twice with high-salt buffer (20 mM Tris–HCl at pH 8.0, 500 mM NaCl, 2 mM EDTA at pH 8.0, 1% Triton X-100, and 0.1% SDS), twice with LiCl buffer (50 mM HEPES–KOH at pH 7.5, 500 mM LiCl, 1 mM EDTA at pH 8.0, 1% NP-40, and 0.7% sodium deoxycholate), and once with TE buffer (10 mM Tris–HCl at pH 7.5 and 1 mM EDTA at pH 8.0). Immunocomplexes were eluted from beads using elution buffer—1% SDS and 100 mM NaHCO₃—at 65 °C for 30 min. For removing cross-links, the eluates were incubated overnight at 65 °C with 200 mM NaCl. Samples were then treated with 400 μg/mL RNase A (Thermo Fisher, #EN0531) at 37 °C for 45 min, followed by protein digestion with 200 μg/mL protease K—50 mM Tris–HCl at pH 7.5 and 10 mM EDTA—at 65 °C for 1 h. DNA was purified using ChIP DNA Clean & Concentrator (Zymo Research, #D5205). Subsequently, reverse transcription was performed with Superscript™ IV Reverse Transcriptase (Invitrogen, #18,090,010), and qPCR was performed on a StepOne™ Real-time PCR system (Applied Biosystems). Primers used for the ChIP-qPCR analysis of p53 pre-mRNA are listed in Supplementary Table 4.

UV-RNA-immunoprecipitation

Cells were transfected for 48 h with either 3xFlag-CMV10 or 3xFlag-CMV10-SRSF3 expression plasmids. After transfection, adherent cells were cross-linked by UV irradiation at 400 mJ/cm², washed with PBS, and lysed in RIP lysis buffer: 10 mM Tris–HCl at pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.1% sodium deoxycholate, 0.5% SDS, and PPI. The lysate was incubated with 400 U of RNase-free DNase I (NEB, #M0303) at 37 °C for 20 min with spinning at 700 rpm. Samples were then sonicated—40% amplitude, on 3 s/off 3 s, repeated 3-times—and centrifuged to remove insoluble material. Of the resulting supernatant, 10% was stored at–80 °C as an input control. The remaining was diluted fivefold in IP dilution buffer—20 mM Tris–HCl at pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, and PPI—and incubated overnight at 4 °C with anti-Flag-M2 affinity gel (Sigma, #A2220). Immunoprecipitants were washed sequentially: once with low-salt buffer, twice with high-salt buffer, twice with LiCl buffer, and once with TE buffer. Beads were then eluted for RNA in elution buffer at 65 °C for 30 min with shaking, followed by centrifugation at 3,000 rpm for 3 min at 4 °C. The RNA samples were treated with 200 μg/mL protease K—50 mM Tris–HCl at pH 7.5 and 10 mM EDTA—at 65 °C for 1 h to remove the proteins. RNA was extracted using a solution comprising 250 μL of Trizol and 50 μL of chloroform, followed by vigorous vortexing and centrifugation at 13,000 rpm for 15 min at 4 °C. The aqueous phase was treated with 20 U/μL DNase I at 37 °C for 15 min to degrade any residual DNA. RNA was purified and concentrated employing the RNA Clean & Concentrator kit (Zymo Research, #R1014). Before reverse transcription with Superscript™ IV Reverse Transcriptase, the RNA was again treated with DNase I to eliminate any remaining genomic DNA. RT-qPCR was performed on a StepOne™ Real-Time PCR system. Primers used for RIP analysis of p53 pre-mRNA are listed in Supplementary Table 4.

Knockout of SRSF3 using the CRISPR/Cas system

SRSF3 was inactivated with a CRISPR/Cas system. The guide RNA sequence targeting the human SRSF3 exon 3—5′- TCGACAGTTCCACTCTTACACGG-3′—was cloned into a pRG vector. SW480 cells were then cotransfected with the gRNA- and SpCas9-encoding plasmids.

Targeted deep sequencing

Genomic DNA was isolated from CRISPR/Cas plasmid-transfected cells with a DNeasy Blood & Tissue kit (Qiagen, Cat. no. 69581) 48 h posttransfection. SRSF3 loci were amplified using 100 ng of genomic DNA, and deep-sequencing libraries were generated by employing PCR. TruSeq HT Dual Index primers were used to label each sample. Pooled libraries were paired-end sequenced on a MiniSeq platform (Illumina).

Scratch wound migration assay

Confluent monolayers of cells stably expressing Flag-p53β were scratched with a sterile 200-µL pipette tip to generate a cell-free gap between two adjoining areas. Loose cells were removed by washing twice with PBS. Wound healing was monitored under phase-contrast microscopy immediately and after 24 h. The area covered by the migrating cells was randomly photographed at three sites along the length of the scratch.

Transwell migration and invasion assay

Cells stably expressing Flag-p53β were serum-starved with 0.1% FBS for 18 h, trypsinized, and resuspended in serum-free RPMI1640 medium at 2 × 10⁵ cells/mL. Then, 250 µl of the cell suspension was applied to the upper chamber of a transwell insert (8-μm pore size, Corning). The cells were allowed to migrate for 24 h toward the bottom chamber containing complete medium added with the chemoattractant. Non-migrating cells present on the upper surface were removed, and those that migrated to the bottom of the insert were fixed in methanol and stained with hematoxylin and eosin for cell counting. The migrated cells in five randomly selected fields per filter were counted. The results represent the average and SD of three independent experiments.

Proximity ligation assay

The UPF1–SRSF3 interaction was investigated using the Duolink® In Situ Red Starter Kit Mouse/Rabbit (DUO92101, Sigma) following the manufacturer’s instructions. Cells were fixed and incubated with anti-UPF1 and anti-SRSF3 primary antibodies. Samples were mounted on glass slides in VECTASTHIELD mounting medium containing DAPI for nuclear staining. Images were acquired using confocal fluorescence microscopy (Fluoview, FV3000; Olympus, Melville, NY) under identical acquisition settings for all treatment conditions. proximity ligation assay (PLA) signals (red puncta) were quantified using ImageJ software and normalized to the nuclear area. Statistical significance was assessed using one-way ANOVA, with significance thresholds set at p < 0.05 and p < 0.01.

Results

1Expression of p53 mRNAs are moderaly regulated by UPF1

TP53 generates alternatively spliced isoforms, p53α mRNA and p53β mRNA, depending on the exclusion and inclusion of cryptic exon 9b (E9b) within intron 9, respectively (Fig. 1A and Supplementary Fig. 1). As a PTC triggers UPF1 recruitment and initiates translation-dependent NMD, PTC-containing p53β mRNA is typically considered an NMD target (Fig. 1A). To test whether UPF1-mediated NMD is strictly dependent on PTC presence in p53 transcripts, we constructed p53 minigenes (DUP reporter with intron 9 of p53 cloned in between β-globin exons). These harbored either a WT PTC (UAA; referred to as “STOP”) or a mutant sequence (UUA; referred to as “GO,” without a stop codon) (Supplementary Fig. 2 A). Using such p53 minigenes, we tested the preferential binding of UPF1 to PTC-containing transcripts via UV-RNA immunoprecipitation (RIP) analysis in Flag-UPF1-expressing HeLa cells (Supplementary Fig. 2B). UPF1 was associated only with Dup-β transcripts (mimicking endogenous p53β mRNA) from the STOP minigene, but not with Dup-α transcripts (mimicking endogenous p53α mRNA) (Fig. 1B and Supplementary Fig. 2 C). Notably, the DUP-p53 minigene was alternatively spliced to only the Dup-α or Dup-β derived from the STOP minigene, but not those from the GO minigene (Supplementary Fig. 1B). PTC-dependent alternative splicing and specific binding of UPF1 to PTC-containing Dup-β transcripts suggest that PTC within p53 intron 9 is related to alternative splicing of p53 minigenes; however, the underlying mechanisms are unknown.

Fig. 1.

Moderate regulation of alternatively spliced p53β mRNA by UPF1. A Alternative splicing (AS) and nonsense-mediated decay (NMD) involving exon 9b (E9b) within p53 intron 9. Inclusion or skipping of E9b (dark gray box) generates p53β or p53α mRNA, respectively. E9b is expected to recruit UPF1 and can be degraded. B UV-crosslinking RNA immunoprecipitation (UV-RIP) assays. HeLa cells were cotransfected with Flag-UPF1 and Dup-p53 minigenes. Lysates were immunoprecipitated using anti-Flag M2 agarose beads. Enrichment of the E9b-included isoform (Dup-β) was assessed using primers spanning the E9b–β-globin exon 2 junction of p53. Data are represented as mean ± SEM (n ≥ 3). *p < 0.05. C Splicing analysis of Dup-p53 minigene-transfected HeLa cells. STOP and GO reporters were used for NMD analysis with or without a premature termination codon (PTC). Primers targeting β-globin exon 1 (forward) and exon 2 (reverse) were used. The upper band corresponds to the E9b-included isoform (β), and the lower band to the E9b-skipped isoform (α). β-actin serves as a loading control. D Splicing analysis from Dup-p53 (STOP) minigene in siUPF1-transfected HeLa cells compared with siGFP-transfected control cells. The upper band corresponds to the E9b-included isoform (β), and the lower band to the E9b-skipped isoform (α). RT-PCR was performed employing β-globin exon 1 (forward) and exon 2 (reverse) primers. UPF1 knockdown was confirmed by Western blot assays (bottom). β-actin was used as a loading control. E RT-qPCR analysis for endogenous p53 mRNA isoforms—p53α and p53β—in UPF1-depleted HeLa cells. F RT-qPCR analysis for exon 1 of endogenous p53 in UPF1-depleted HeLa cells, corresponding to the 5′ UTR. (G, H) Decay rates of p53α mRNA G and p53β mRNA H in siUPF1- and siGFP-transfected cells following transcription inhibition using actinomycin D (ActD). The mean half-lives of p53α or p53β mRNAs are indicated by a dashed line. Data are represented as mean ± SEM (n ≥ 3)

Next, we asked whether the expression of PTC-containing Dup-β transcripts is regulated by UPF1 via the use of UPF1 knockdown (KD) and the Dup-p53 minigene reporter (Fig. 1C). UPF1 depletion enhanced the expression of Dup-α as well as Dup-β transcripts derived from the STOP minigene. Consistent with the findings of the minigene analysis, the levels of endogenous p53α and p53β mRNAs were also modestly elevated upon UPF1 KD (Fig. 1D and Supplementary Fig. 2 C). Additionally, we observed an increase in the 5′ UTR-containing p53 transcripts using p53 exon 1-detecting primers (Fig. 1D), suggesting that UPF1 may influence transcription and splicing rather than acting solely through post-transcriptional RNA decay.

To directly evaluate the RNA stability, we measured the half-life of p53 transcripts in UPF1-depeleted versus control HeLa cells following transcriptional inhibition with actinomycin D (ActD) (Fig. 1E, F). If p53β mRNAs were primarily degraded via UPF1-mediated NMD, then half-life would increase substantially upon UPF1 KD. However, this was not the case, indicating that UPF1 is not the key regulator in p53β mRNA decay. Likewise, the stability of p53α mRNA lacking the PTC-containing E9b remained unchanged upon UPF1 KD.

As a control for NMD efficiency, we analyzed HRAS variant 2 (V2) mRNA, alternative splicing-derived NMD substrate carrying a PTC within its exon 5 (Supplementary Fig. 3 A). Upon UPF1 KD, HRAS V2 mRNA level was increased, and the half-life was significantly extended from 1 to 3 h (Supplementary Fig. 3 A, B). Thus, the increased transcription and stabilization of HRAS V2 mRNA confirm effective NMD suppression under the UPF1 KD condition we used in the experiments.

Intron-retained p53 transcript associates with chromatin

Since p53 gene express diverse mRNAs, we next sought for more p53 varinat transcripts generated from the alternative splicing site. The E9b of TP53 is generally flanked by two introns: intron 9a (I9a), which is upstream, and intron 9b (I9b) downstream (Fig. 2A). These introns are sequentially spliced, with I9a excised before I9b. However, non-ordered splicing also occurs, in which I9b is removed before I9a. We hypothesized that if I9b is spliced first, the resulting transcript would retain I9a along with E9b, thereby harboring three predictable PTCs, two within I9a and one within E9b. To investigate this possibility, we designed primer pairs that specifically detect transcripts retaining either I9a or I9b (Supplementary Fig. 4 A and Fig. 2B). RT-PCR detected I9a-retained transcripts using both primer sets: a and b, but not I9b-retained ones (Supplementary Fig. 4 A). We denoted this previously uncharacterized I9a-retained transcript as p53-IR transcript. Moreover, when compared with intron 2 retention, a well-characterized transcript that encodes an N-terminally truncated p53 isoform, p53-IR transcript was more abundantly expressed (Fig. 2B), while the I9b-retained transcript was not detected at all.

Fig. 2.

Identification and localization of the intron-retained p53 transcript. A Diagram of sequential (seq) or non-sequential (non-seq) splicing of p53 intron 9. I9a and I9b indicate the upstream and downstream introns flanking E9b, respectively. Predicted premature termination codons (PTCs) are shown in red. B RT-qPCR analysis of intron retention in p53 transcripts under basal conditions in HeLa cells. I2, intron2; N.D., not detected. C Decay rates of p53-IR transcript in siUPF1- and siGFP-transfected HeLa cells following transcription inhibition with ActD. The mean half-lives of p53-IR transcript are indicated by a dashed line. Data are represented as mean ± SEM (n ≥ 3). D RT-qPCR-based quantification of p53α, p53β, and p53-IR transcript following cycloheximide (CHX) treatment in HeLa cells. Data are represented as mean ± SEM (n ≥ 3). **p < 0.01. E, F Ratios of p53α, p53β, and p53-IR transcript levels altered by CHX treatment in the nuclear and cytoplasmic fractions of HeLa cells, compared with DMSO-treated controls. Data are presented as mean ± SEM (n ≥ 3). ***p < 0.001. G Subcellular localization of p53 transcripts in HeLa cells. Cytoplasmic (Cy), Nucleoplasmic (Np), and chromatin-associated (Chr) fractions are indicated. GAPDH pre-mRNA and NEAT1 long noncoding RNA were used as nuclear-retained RNA markers, and 18S ribosomal RNA was used as a cytoplasmic RNA marker

Given that p53-IR transcript contains three predicted PTCs (Fig. 2A), we first examined whether it is subjected to NMD-mediated regulation. We measured the half-life of p53-IR transcript; its stability did not increase upon UPF1 depletion (Fig. 2C). We next inhibited the NMD pathway using cycloheximide (CHX) and quantified each p53 mRNA isoform as well as p53-IR transcript (Fig. 2D). As expected, levels of p53β mRNA and HRAS V2 mRNA were enhanced markedly upon CHX treatment, whereas those of p53-IR transcript remained unchanged (Fig. 2D and Supplementary Fig. 4B).

To further evaluate the localization of p53-IR transcript relative to p53β mRNA under NMD inhibition condition, we assessed the distribution of each isoform after CHX treatment. The nucleus–cytoplasm ratio of p53α mRNA remained unchanged, while p53β mRNA and HRAS V2 mRNA showed increased cytoplasmic accumulation as expected to be NMD target transcripts (Fig. 2E and Supplementary Fig. 5 A). In the case of p53-IR transcript, its distribution remained nuclear even after CHX treatment, suggesting it as a non-canonical transcript (Fig. 2F). Subcellular fractionation was verified using Lamin B1 and β-tubulin as nuclear and cytoplasmic protein markers, respectively: NEAT1 lncRNA and MT-CYB (mitochondrial cytochrome B) mRNA served as nuclear and cytoplasmic RNA controls, respectively (Supplementary Fig. 5B–D).

Since p53-IR transcript remains nuclear and is not targeted by NMD, we investigated where alternatively spliced p53 mRNAs are localized (Fig. 2G). p53α mRNA was widely detected in all three fractions (chromatin, nucleus and cytoplasm), consistent with its multilayered regulation as a canonical transcript. In contrast, p53β mRNA was restricted mainly to the cytoplasm and chromatin fractions, consistent with efficient splicing followed by cytoplasmic export. Notably, was strictly nuclear in location, with robust enrichment in the chromatin and nucleoplasm fractions, similar to GAPDH pre-mRNA. Included as fractionation controls, NEAT1 lncRNA was restricted to the nucleus, whereas 18S ribosomal RNA (rRNA) demonstrated predominantly cytoplasmic with minor nucleoplasmic presence. Together, these findings indicate that p53-IR transcript is a nuclear-detained transcript, likely escaping NMD due to its restricted subcellular localization, whereas p53β mRNA appears to be exported to the cytoplasm, allowing for subsequent translation.

Chromatin-associated UPF1 interacts with SRSF3 and TP53 gene

We had previously reported that UPF1 associates with chromatin for transcription-coupled RNA surveillance [37]. Given that p53β mRNA expression is induced upon SRSF3 downregulation [23], we hypothesized that SRSF3 may act as a coregulator of UPF1 during splicing regulation. Subcellular fractionation revealed UPF1 and SRSF3 are localized in all cell compartments, including the chromatin-associated fraction (Fig. 3A). Although UPF1 was mainly located in the cytoplasm, it was also detected in the nucleoplasmic and chromatin-associated fractions, consistent with previous findings [37]. SRSF3 was most abundant in the chromatin-associated fraction, followed by the cytoplasm and nucleoplasm. To confirm equal loading during fractionation, we used markers such as α-tubulin (cytoplasm), SRSF1 (nucleoplasm), and Histone H3 (chromatin). Immunofluorescence assays further revealed that SRSF3 is predominantly localized in the nucleus and prominently colocalizes with serine 2-phosphorylated RNA polymerase II (Pol II-pS2), a transcription-elongating polymerase marker; trimethylated histone H3 lysine 4 (H3K4me3), an active transcription histone marker; and another splicing factor, SRSF7 (Supplementary Fig. 6 A). Pearson’s correlation coefficients confirmed robust colocalization, with median values of 0.82 for Pol II-pS2, 0.89 for H3K4me3, and 0.86 for SRSF7 (Supplementary Fig. 6B). These results support that SRSF3 interacts with elongating Pol II and participates in splicing-linked processes.

Fig. 3.

Chromatin association of SRSF3 and UPF1 at the TP53 gene. A Subcellular localization of SRSF3 and UPF1 in HeLa cells. α-tubulin and Histone H3 are markers for the cytoplasmic and chromatin-associated fractions, respectively. Regarding SRSF1, hyperphosphorylated forms localize to the nucleoplasm, while hyperphosphorylated forms localize to the cytoplasm. B UPF1 and SRSF3 chromatin immunoprecipitation (ChIP) analysis of TP53 genomic regions. UPF1 ChIP was performed using an anti-UPF1 antibody compared with an anti-IgG control in HeLa cells, whereas SRSF3 ChIP was performed using an anti-Flag antibody in Flag-SRSF3-expressing HeLa cells compared with Flag-empty vector (Flag-Vec) control cells. Pro, promoter; I4, intron 4; I9, intron 9; E11, exon 11. Data are presented as mean ± SEM (n ≥ 3). *p < 0.05; **p < 0.01. C Coimmunoprecipitation (co-IP) showing the interaction between UPF1 and SRSF3. HeLa cells were cotransfected with Myc-UPF1 and Flag-SRSF3, and immunoprecipitation was performed utilizing anti-Flag antibody. D Coimmunofluorescence (co-IF) of UPF1 and SRSF3 after treatment with transcription or translation inhibitors, which are actinomycin D (ActD) and flavopiridol (Flavo), or cycloheximide (CHX), respectively. UPF1 was detected using Cy3-conjugated secondary antibody (red), and SRSF3 was identified employing the Alexa Fluor 488-conjugated one (green). Nuclei were counterstained with DAPI (blue). Scale bars, 5 μm. E Quantification of UPF1 and SRSF3 colocalization by Pearson’s correlation coefficient. Data are presented as mean ± SEM (n ≥ 30). ***p < 0.001. F Representative images of in situ proximity ligation assay (PLA) between UPF1 and SRSF3 in HeLa cells. Each cell was treated with DMSO (control), ActD, Flavo, or CHX. PLA signals are shown as red puncta, and nuclei are counterstained with DAPI (blue). Scale bars, 5 μm. G Quantification of the mean of nuclear PLA signals per nucleus. Data are presented as mean ± SEM (n ≥ 30). Statistical significance was determined by one-way ANOVA. *p < 0.05; ***p < 0.001

To explore whether the nuclear retention of p53-IR transcript is related to chromatin-associated UPF1, we preformed chromatin immunoprecipitation using an antibody against endogenous UPF1 (Fig. 3B) [38]. UPF1 was enriched across the TP53 locus, including the promoter, introns 4 and 9, as well as exon 11 (the 3′ UTR), suggesting its broad association of UPF1 with the transcriptionally active TP53 gene. Flag-SRSF3-ChIP analysis revealed that SRSF3 binds more specifically to introns 4 and 9 of TP53 compared to UPF1, with a wider binding tendency (Fig. 3B). Notably, SRSF3 and UPF1 were markedly enriched at intron 9, a region for determining the criticality of I9a retention versus E9b inclusion. To examine any physical interactions between UPF1 and SRSF3, we performed coimmunoprecipitation (co-IP) in HeLa cells overexpressing Flag-SRSF3 and Myc-UPF1 (Fig. 3C). UPF1 coprecipitated with SRSF3, suggesting their physical interaction in vitro. AlphaFold3-based prediction exhibited moderately confident folding of individual proteins, stable UPF1-SRSF3 interface hindered by intrinsically disordered regions of SRSF3 (Supplementary Fig. 7 A, B). Taken together, UPF1 and SRSF3 interaction is mediated by p53 -IR transcript in a chromatin-bound state.

Transcription elongation blockade promotes nuclear UPF1 and SRSF3 co-localization

Given the nucleus-localized UPF1–SRSF3 interaction during transcription-coupled pre-mRNA splicing and RNA surveillance, we hypothesized that such an association might be influenced by transcriptional activity. To investigate if the interaction ceases in response to transcriptional or translational inhibition, we treated HeLa cells with respective inhibitors and analyzed UPF1 and SRSF3 localization through immunofluorescence (Fig. 3D).

Posttreatment with the transcription initiation inhibitor, ActD, UPF1 remained primarily cytoplasmic, while SRSF3 was restricted to the nucleus compared to control cells. Pearson’s correlation coefficient for UPF1 and SRSF3 colocalization was unchanged relative to the control (Fig. 3E), indicating limited nuclear overlap under transcription initiation blockade. In contrast, suppression of transcription elongation with the CDK9 inhibitor flavopiridol (Flavo) or translation with CHX caused marked nuclear accumulation of UPF1 (Fig. 3D). Under Flavo treatment, SRSF3 retained a diffused nuclear pattern, but its colocalization with UPF1 was remarkably increased (Fig. 3E). Although CHX also induced robust nuclear accumulation of UPF1 with prominent nuclear puncta, SRSF3 localized largely apart from UPF1, with no significant changes in their colocalization compared to the control (Fig. 3D, E). These findings indicate that transcription elongation blockade, rather than inhibition of transcriptional initiation or translation, promotes nuclear colocalization of UPF1 and SRSF3.

To further validate the biochemical interaction between UPF1 and SRSF3 beyond colocalization at the cell-level, we next examined if these proteins formed proximity complexes in situ (Fig. 3F). Having established transcription-dependent changes in UPF1–SRSF3 colocalization, we employed an in situ PLA, which enables visualization of proteins located within < 40 nm of one another, thereby revealing their spatial association in fixed cells. Under basal conditions, multiple PLA signals were emitted, indicating UPF1–SRSF3 interactions in HeLa cells. Notably, Flavo markedly increased the number of PLA signals (Fig. 3G), consistent with the enhancement in colocalization observed upon transcription elongation blockade. Similarly, although more modest, an elevation in PLA signal was also detected under CHX or other conditions, suggesting that transcriptional dynamics modulate the extent of UPF1–SRSF3 association. Collectively, these findings support a model in which transcription elongation promotes nuclear proximity and functional interplay between UPF1 and SRSF3, reinforcing their role as cooperative regulators within the cotranscriptional RNA surveillance and splicing machinery.

SRSF3 -bound p53-IR transcript recruits UPF1 for p53β mRNA surveillance

As UPF1 binding at I9a may contribute to intron retention (IR) by interfering with or modulating splicing efficiency, we investigated how SRSF3 and UPF1 regulate the splicing decision between p53-IR transcript and p53β mRNA. We first performed ultraviolet crosslinking-RNA immunoprecipitation (UV-RIP) in HeLa cells overexpressing Flag-SRSF3, and compared them to control cells transfected with Flag-vector (Fig. 4A and Supplementary Fig. 8 A). Using primers specific to various regions of the pre-mRNA, we mapped SRSF3 binding and found that SRSF3 was predominantly bound to E9b and accumulated at I9a (Fig. 4A), consistent with a previous report [23].

Fig. 4.

Binding of SRSF3 at the retained intron of p53 transcript. A UV-RNA immunoprecipitation (UV-RIP) analysis of endogenous p53 (pre-)mRNA. Flag-SRSF3-expressing and Flag-Vec control cells were subjected to immunoprecipitation using an anti-Flag antibody. Data are presented as mean ± SEM (n ≥ 3). *p < 0.05; **p < 0.01. B Schematic diagram of Dup-p53 minigene reporters harboring mutations in SRSF3-binding motifs, indicated by black circles, and mutated motifs by white circles. The WT contains all intact SRSF3-binding motifs. Mutant reporters (Mut1–Mut4) contain individual or combinations of mutated SRSF3-binding motifs, as indicated. C UV-RIP analysis of Dup-p53 minigene reporters. Flag-SRSF3-expressing cells and Flag-Vec control cells were subjected to immunoprecipitation using an anti-Flag antibody. Data are presented as mean ± SEM (n ≥ 3). **p < 0.01. D Splicing analysis of Dup-p53 minigene reporters in transfected HeLa cells. Relative Dup-β transcript level was measured by RT-qPCR. Data are presented as mean ± SEM (n ≥ 3). *p < 0.05. E SRSF3 tethering analysis in combination with UPF1 knockdown. MS2 RNA hairpins were inserted into I9a of the Dup-p53 minigene, and MS2-binding protein (MBP)-fused SRSF3 was co-transfected to tether SRSF3 to the MS2 RNA. RT-qPCR was used to quantify Dup-α and Dup-β isoforms. Data are presented as mean ± SEM (n ≥ 3). **p < 0.01. F Proposed model illustrating the splicing-linked RNA surveillance mechanism. p53-IR transcript interacts with SRSF3, which subsequently recruits UPF1. Formation of p53-IR–SRSF3–UPF1 ternary complex suppresses E9b inclusion, thereby preventing production of the p53β mRNA

To identify functional SRSF3-binding motifs involved in E9b excision, we analyzed I9a of TP53 for canonical SRSF3 binding motifs such as “(A/U)C(A/U)(A/U)C” or “(A/U)CNUC” (Supplementary Fig. 8B) [39–41]. Based on the findings of this in silico analysis, we constructed a series of WT and mutant p53 minigenes, each harboring specific mutations: Mut 1 (I9a), Mut 2 (E9b), Mut 3 (I9b), and Mut 4 (both I9a and E9b) (Fig. 4B and Supplementary Fig. 8 C). We then performed UV-RIP using these Dup-p53 minigenes to assess SRSF3 binding (Fig. 4C). SRSF3 binding was markedly reduced in Mut 1 and Mut 2, consistent with the endogenous p53 pre-mRNA binding pattern presented in Fig. 4A. However, SRSF3 binding was not remarkably decreased in Mut 3, suggesting that I9b may not be a primary SRSF3-binding site and likely plays a minimal role in regulating exon 9b inclusion. We next analyzed the splicing patterns of these Dup-p53 minigenes to determine how SRSF3 binding influences splicing (Fig. 4D). Interestingly, only with the Mut 4 construct, which lacks SRSF3 binding motifs in I9a and E9b, production of Dup-β isoform was elevated, but not with Mut 1, Mut 2, and Mut 3. We think that SRSF3 enrichment in I9a and E9b promotes the retention of I9b and represses E9b inclusion, thereby inhibiting p53β formation.

We hypothesized that SRSF3 binding to p53-IR transcript facilitates a surveillance mechanism targeting p53β mRNA. To test this, we designed a tethering assay using Dup-p53 minigenes with MS2 binding sequences inserted into I9a (MS2–I9a) or I9b (MS2–I9b). These were coexpressed with MS2 binding protein (MBP)-fused SRSF3 (Fig. 4E and Supplementary Fig.  9 A). Tethering SRSF3 to I9a strongly suppressed E9b inclusion (Dup-β) without affecting Dup-α levels (Fig. 4E). In comparison, SRSF3 tethered to I9b did not alter Dup-α or Dup-β contents (Supplementary Fig. 9 A), indicating a potential role of SRSF3 as a splicing factor during I9b excision. Next, we examined whether SRSF3 binding to I9a serves as a platform for UPF1 recruitment as a part of an RNA surveillance mechanism independent of classical NMD. We combined the SRSF3 tethering assay with UPF1 KD (Supplementary Fig. 9B). Strikingly, the repression of Dup-β by SRSF3–I9a was dramatically reversed upon UPF1 depletion, increasing Dup-β levels by 8.3-fold, while Dup-α remained unaffected (Fig. 4E). In contrast, SRSF3–I9b led to only a modest increase of 1.5-fold in Dup-β upon UPF1 KD (Supplementary Fig. 9 A), as if the stabilization of endogenous p53β transcripts upon UPF1 depletion was due to suppression of UPF1-mediated decay.

Collectively, these results demonstrate that SRSF3 binding to I9a of the p53-IR transcript promotes UPF1-dependent decay of p53β mRNA (Fig. 4F). We propose that the SRSF3–UPF1 complex acts as a posttranscriptional checkpoint to suppress the accumulation of aberrant p53 isoforms like p53β mRNA by generating p53-IR transcript for degradation.

SRSF3 suppresses the expression of the C-terminal truncated p53β protein isoform

To explore the generality of SRSF3-mediated suppression of p53β mRNA, we measured the expression of p53 mRNA levels upon SRSF3 KD in various cell lines (Fig. 5A and Supplementary Fig. 10 A) [23]. In colorectal cancer cell lines (HCT116 and SW480), p53α mRNA levels are not much changed, but p53β mRNA levels are significantly increased in all cell lines (Fig. 5A). We next sought to test whether evading SRSF3-based surveillance could affect translation of p53β mRNA. Specific role of SRSF3, but not SRSF1, to the translation of p53β protein isoform was demonstrated by SRSF3 KD and SRSF1 KD followed by western blot analysis (Fig. 5B). To further study the role of SRSF3 on p53β protein expression, we established SRSF3 knockout (KO) using CRISPR/Cas9 system in SW480 (Fig. 5C and Supplementary Fig. 10B, C). We utilized the panel of anti-p53 antibodies (Pab1801 for 11–25 residues; DO-11 for 181–190 residues) and observed increased expression of p53β protein isoform in the SRSF3 KO SW480 cell line. Intriguingly, UPF1 KD did not increase p53β protein isoform, consistent with a modest role of UPF1 in p53β mRNA expression (Fig. 1D and 5D). The specific antibody targeting p53β protein isoform was generated using distinctive p53β C-terminal epitope to quantify the expression of p53β protein isoform following SRSF3 KO in SW480 (Fig. 5C).

Fig. 5.

Expression of C-terminal truncated p53 protein upon SRSF3 depletion. A RT-qPCR analysis of p53α and p53β mRNA levels in HEK293, HeLa, HCT116, and SW480 cells transfected with siRNA targeted to SRSF3 (siSRSF3) compared to control siRNA (siGFP). Exon junction-specific primers used for isoform detection are shown (left). Data are presented as mean ± SEM (n ≥ 3). *p < 0.05; **p < 0.01. B Western blot analysis of p53 protein isoforms in SW480 cells transfected with siRNAs targeting SRSF3 and SRSF1. p53 isoforms were detected using anti-p53 (DO-1) antibody. The upper band corresponds to p53α protein (~ 53 kDa), and the lower band to p53β protein isoform (~ 47 kDa). β-actin serves as a loading control. C Western blot analysis of p53 protein isoforms in SRSF3 knockout (KO) SW480 cells obtained by employing CRISPR/Cas9. Total cell lysates were analyzed using anti-p53 (Pab1801 and DO-11) and a custom anti-p53β-specific antibody. β-actin serves as a loading control. D Western blot analysis of p53 protein isoforms in UPF1-depleted SW480 cells (siUPF1) compared to control cells (siGFP). The anti-p53 (DO-1) antibody used in (B) was also employed to detect p53α and p53β protein isoforms. β-actin serves as a loading control. E Quantification of p53α and p53β protein isoform levels detected by anti-p53 (DO-11) antibody. Data are presented as mean ± SEM (n = 3). **p < 0.01. F Proposed model explaining the production of the C-terminal truncated p53β protein isoform

TP53 gene generates many p53 protein isoforms, including oncogenesis-promoting Δ133p53β protein isoform [25, 28, 30]. We examined the expression of Δ133p53β protein isoform with DO-11 antibody by triple biological replicates of SRSF3 KO in SW480 cells (Fig. 5C and Supplementary Fig. 11 A and C). Specificity of p53 antibodies was examined by overexpressing Δ133p53α clone followed by western blot analysis (Supplementary Fig. 11B). As expected, Δ133p53α protein isoform was readily detected by DO-11, but not by DO-1. Importantly, Δ133p53 protein isoform was not induced in SRSF3 KO cells, in contrast to prominent induction of p53β protein isoform upon SRSF3 KO (Fig. 5C, 5E and Supplementary Fig. 11 C). p53β mRNA is translated into the C-terminal truncated p53β protein isoform as found in colorectal cancer cells, which can be strongly suppressed by SRSF3 (Fig. 5F).

p53β protein isoform induces epithelial-mesnechymal transition of cancer cells

A schematic representation of the p53α and p53β proteins illustrates a lack of the C-terminal oligomerization domain in the p53β protein, corresponding to residues 326–356 of p53α (Supplementary Fig. 11 A). To address the function of p53β in cancer cells, we established stable colorectal cancer cell lines—HCT116^WT, HCT116^p53−/−, DLD-1 (p53 S241F)—and a lung cancer cell line, A549, harboring the p53β transgene (Supplementary Fig. 12A, B). Although the expression levels of p53β mRNA have been linked to differentiation and senescence in fibroblasts and T lymphocytes, its role in EMT and metastasis has not been well studied [23, 42]. We observed that p53β expression induced a mesenchyme-like morphology, including spindle-shaped cells and loss of contact inhibition, in all cell lines examined (Fig. 6A). Such a morphological shift indicates that p53β promotes EMT-like phenotypic conversion [43]. Consistent with these alterations, the expression of genes encoding EMT-associated transcription factors, ZEB1 and TWIST1, was markedly upregulated in all p53β-expressing, HCT116 (p53 WT and null), DLD-1, and A549 cells (Fig. 6B and Supplementary Fig. 13 A). E-cadherin, a key epithelial cell marker, is deregulated during EMT and is associated with destabilized cell–cell junctions, increased invasiveness, and metastasis [44]. p53β expression disrupted the assembly of E-cadherin at cell–cell junctions, suggesting that p53β plays a vital role in inducing an EMT phenotype (Supplementary Fig. 13B).

Fig. 6.

Oncogenic functions of the C-terminal truncated p53β protein isoform. A Cell morphology of colorectal cancer cell lines (HCT116^WT and HCT116^p53−/−, DLD-1) and a lung cancer cell line (A549) upon Flag-p53β overexpression. B RT-qPCR analysis of EMT markers, ZEB1 and TWIST mRNAs, in stably Flag-p53β- expressing HCT116^WT and HCT116^p53−/−, and A549 cells. Data are presented as mean ± SEM (n ≥ 3). *p < 0.05; **p < 0.01. C, D Wound closure assay in stably Flag-p53β-expressing HCT116^WT and HCT116^p53−/−, and A549 cells. Wound closure was measured 48 h after scratching. Data are presented as mean ± SEM (n ≥ 3). *p < 0.05. E, F Matrigel invasion assay using stably Flag-p53β-expressing HCT116^WT and HCT116.^p53−/−, and A549 cells. Cells were seeded in the upper chamber of Matrigel-coated transwells. Invaded cells on the bottom surface were counted after 48 h. Data are presented as mean ± SEM (n ≥ 3). *p < 0.05; ***p < 0.001

As Δ133p53β promotes cancer progression [28–30], we examined whether the EMT-promoting effects observed upon Flag-p53β expression were attributable to p53β protein isoform, rather than to an unintended generation of the Δ133p53β isoform through alternative start codon usage. Codon frame prediction indicated that Δ133p53β could be translated from p53β mRNA. To assess the generation of Δ133p53β from the Flag-p53β expression clone, we performed western blot analysis with anti-p53 DO-11 antibody (recognizes residues 181–190) in HCT116^WT, DLD-1, and A549 cells (Supplementary Fig. 12B). Δ133p53β was not detected in any cell line; only Flag-p53β protein isoform and endogenous p53α proteins were observed by panel of anti-p53 antibodies.

To determine whether these EMT-like changes enhance cell motility, we performed wound-healing and transwell migration assays. p53β-expressing HCT116 (p53 WT and null), DLD-1, and A549 cells closed wound gaps substantially faster and exhibited greater transwell migration than Flag-vector-expressing control cells (Flag-Vec) (Fig. 6C-6D, and Supplementary Fig. 14 A). Notably, HCT116^p53−/− cells expressing Flag-p53β almost completely covered the wound (0.4 ± 0.1 μm) within 48 h after scratching (Fig. 6C). These observations were confirmed by a Boyden chamber transwell migration assay. p53β expression markedly enhanced the number of migrating HCT116 (p53 WT and null), DLD-1, and A549 cells relative to Flag-Vec (Fig. 6E-6D, and Supplementary Fig. 14B). Therefore, we suggest that p53β functions as an oncogenic factor, particularly in promoting EMT and metastasis across colon and lung cancer cells.

Discussion

This study uncovers RNA-based quality control mechanism in which the splicing factor SRSF3 and the RNA surveillance factor UPF1 collaborate to prevent the accumulation of oncogenic p53β protein isoform (Fig. 7). By demonstrating how disrupted splicing-linked surveillance drives EMT via the processing of aberrant p53 transcripts, our findings reveal a previously unrecognized checkpoint in posttranscriptional gene regulation. These insights highlight a promising avenue for restoring p53 function in tumors via a dysregulation of splicing or decay machinery.

Fig. 7.

Biogenesis of p53 mRNAs and functions of p53 protein isoforms. The scheme represents the role of a newly identified intron-retained p53 transcript (p53-IR transcript) in preventing the generation of the C-terminal truncated and pro-metastatic p53β protein. During transcription elongation, SRSF3 associates with the p53-IR transcript and recruits UPF1 forming a transcript surveillance complex. It specifically monitors p53 pre-mRNA and suppress p53β mRNA expression and C-terminal-truncated p53β protein, which promote epithelial–mesenchymal transition (EMT) and metastasis

We propose a novel posttranscriptional regulatory mechanism in which SRSF3 and UPF1 coordinate splicing dynamics and control the production of certain p53 isoforms. Specifically, our data show that a surveillance complex—composed of SRSF3, UPF1, and intron-retained p53 transcripts—targets p53β for repression. Through mutagenesis and UV-RIP analysis using p53 minigene constructs, we identified I9a upstream of E9b as a crucial SRSF3-binding site. This interaction facilitates UPF1 recruitment and selectively suppresses p53β mRNA expression; SRSF3 loss impairs this surveillance axis, allowing aberrantly spliced mRNAs to escape degradation. As a result, p53β mRNA is translated into p53β protein isoform that lacks a C-terminal domain but contains a unique extension derived from E9b, promoting EMT and enhancing cell migration in colorectal cancer. Our findings provide mechanistic insights into the processes by which the p53-IR–UPF1–SRSF3 complex maintains transcriptomic fidelity by linking splicing regulation to mRNA surveillance.

Recent studies have expanded the canonical role of UPF1 in cytoplasmic NMD, revealing its critical nuclear functions. For instance, in Schizosaccharomyces pombe, UPF1 is associated with actively transcribing RNA polymerase II target genes in an RNA-dependent manner [45]. Similarly, genome-wide analysis in Drosophila has shown that UPF1 binds cotranscriptionally to nascent RNAs, particularly accumulating at exonic regions proximal to splice sites, suggesting a role in splicing-coupled RNA quality control [46]. In cancer cell line systems, chromatin-associated UPF1 might have a role in transcription-coupled RNA surveillance [37]. Additionally, UPF1 promotes the formation of R-loops to facilitate double-strand break repair [47], underscoring its multifaceted role in linking RNA processing with genome stability. Our data support this expanding paradigm by revealing that UPF1, recruited to p53 I9a via SRSF3, contributes to nuclear RNA surveillance and suppresses the production of oncogenic p53β isoforms.

IR is increasingly recognized as a central layer of gene regulation in response to stress, cel differentiation, and tumor suppression [48, 49]. For example, IR-mediated suppression of splicing factor expression plays a role in granulopoiesis [50]. Factors regulating IR include transcription elongation regulators and splicing factors that recruit exosomal machinery to degrade inefficiently spliced transcripts [51, 52]. In our model of p53 alternative splicing, a proximal intron containing a PTC serves as a platform for SRSF3–UPF1-mediated surveillance. Further transcriptome-wide studies are needed to establish the prevalence of this mechanism in regulating other IR-containing alternative transcripts.

Beyond their classical role in splicing, SR proteins function in transcription, RNA export, translation, and other RNA metabolism processes [53, 54]. They act as coordinators and integrators of posttranscriptional gene regulation [11]. The previously undescribed recruitment of UPF1 by SRSF3 to a retained intron within p53 mRNA, as reported here, adds to this complexity. Notably, while our manuscript was under preparation, a study was published reporting that SRSF1 can stimulate NMD via UPF1 recruitment [36], supporting the broader role of SR proteins in RNA surveillance. SRSF3, traditionally considered an oncogene due to its overexpression in various cancers, is paradoxically downregulated in certain cancer contexts [55–57], suggesting a complex interplay with surveillance factors, including UPF1, in oncognesis.

Despite the mechanistic insights provided by our study, a genome-wide generalization of the SRSF3–UPF1 surveillance axis remains to be established. Although public CLIP-seq and RNA-seq datasets provide valuable information on SRSF3 and UPF1 RNA-binding landscapes, UPF1 datasets have focused on its role in the NMD of PTC-containing transcripts and translation-regulation via 3′UTR binding. Furthermore, SRSF3 datasets have been obtained in mouse models in which human p53 alternative spliced isoforms are not expressed, limiting their relevance to the mechanism described here. Moreover, low-abundance and chromatin-associated intron-retained transcripts such as p53-IR transcript are often underrepresented in conventional CLIP-seq and bulk RNA-seq datasets, particularly under primary conditions. Therefore, future studies must focus on the genome-wide identification of SRSF3–UPF1 coregulated intron-retained RNA targets under transcription-perturbed conditions [58–60]. Our findings suggest that the functional outcomes of SRSF3 depend more on its RNA-binding activity and RNP assembly with specific target transcripts than on expression levels. Thus, the context-specific role of SRSF3 in monitoring the production of aberrant spliced mRNA isoforms may be a key determinant of its oncogenic or tumor-suppressive properties.

Although posttranslational modifications of p53 are well studied, much less is known about the regulation of its transcript isoforms [3, 61]. Understanding the role of SRSF3 in controlling p53 mRNA splicing is essential for several reasons. First, recent evidence suggests that p53 mRNA itself contributes to the stress-responsive p53 protein complex [31, 32]. Second, when p53β mRNA escapes surveillance and is exported, it encodes a truncated protein with a unique neoepitope [3, 62], promoting EMT phenotypes in colorectal cancer. Third, the accumulation of mutations in TP53 and splice variants in human tumors suggests that the misregulation of mRNA isoforms adds another layer of pathogenic complexity [63]. Other SR proteins—SRSF1 and SRSF7—and the exon junction complex—Y14—might also be involved and potentially modulate stress responses related to p53β mRNA [64]. These findings open new therapeutic possibilities centered on p53 mRNA isoforms as targets for cancer intervention.

In summary, we have uncovered a surveillance mechanism involving SRSF3 and UPF1 that targets intron-retained p53 transcripts to prevent the expression of oncogenic p53β protein isoform. We propose that the interaction between SRSF3 (a trans-acting factor) and retained intronic elements (cis-regulatory cues) represents a molecular switch that governs isoform selection and mRNA fate. This work reveals how splicing and RNA quality control converge to regulate transcript diversity and suggests that specifically targeting such splicing surveillance networks could offer novel strategies in cancer therapy.

Abbreviations

PTC

Premature termination codon

NMD

Nonsense-mediated decay

IR

Intron retention

RIP

RNA Immunoprecipitation

EMT

Epithelial–mesenchymal transition

Author contributions

J.J., D.H., T.P., and J.H. performed experiments and analyzed the data. J.J. and D.H. wrote the manuscript. T.K. established the SRSF3-targeting CRISPR-Cas9 system. S.J. supervised the project. J.J. and D.H. contributed equally to this work. All authors reviewed and approved the final version of manuscript.

Funding

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2023–00208597 and RS-2025–02214538) to S.J. (RS-2019-NR041293) to D.H.,and by the Samsung Science & Technology Foundation (SSTF-BA1601-16) to S.J.

Data availability

All data supporting the findings of this study are included within the article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no conflict of interest.