Identification of a DNA Damage–Induced Alternative Splicing Pathway That Regulates p53 and Cellular Senescence Markers

The integrity of nuclear DNA in mammalian cells is constantly challenged by exposure to exogenous damaging agents (e.g., ionizing or UV irradiation, various chemicals), by the generation of reactive intracellular metabolites (e.g., oxygen radicals), and by DNA alterations caused by shortening or aberrations of chromosomal ends or abnormal DNA replication intermediates (1). Suboptimal cellular responses to DNA damage contribute to cancer development and to aging-associated pathogenesis (2). Conversely, many cancer therapeutics work by damaging DNA, thus making the cellular responses to damage important determinants of outcomes of cancer therapies.

The p53 tumor suppressor is a key player in cellular responses to DNA damage and other stresses. Levels of p53 protein increase after DNA damage and typically lead to cell-cycle arrest or programmed cell death (1). This stress-induced protein induction is controlled in part by an increased half-life of p53 protein resulting from inhibition of HDM2-mediated proteasomal degradation (3, 4). Increased translation of p53 mRNA after DNA damage also contributes to p53 induction, a process regulated by the damage-induced binding of ribosomal protein RPL26 to a double-strand RNA structure formed by base-pairing between the sequences in the 5′ untranslated region (UTR) and 3′ UTR of TP53 mRNA (5). Disruption of this translational control blunts p53 induction and p53-mediated growth arrest and apoptosis (5, 6).

Alternative splicing (AS) of pre-mRNA products is increasingly appreciated as an additional mechanism to expand RNA and protein diversity (7, 8). Mutations that perturb AS patterns are reported in a variety of human hereditary diseases and cancer (9, 10). For example, AS isoforms of apoptotic gene products, including members of the BCL2 family, are generated after DNA damage and appear to influence cell death propensities (11). Alternatively spliced forms of p53 mRNA and protein have been noted and have been suggested to contribute to the versatility of p53-mediated cellular responses (12). Expression of these different forms of p53 appears to vary in different tissues, and abnormal expression of these variants has been described in some cancers and linked to patients' clinical outcomes. Nine different alternatively spliced p53 mRNA products, generated by internal promoter usage or intron retention, have been reported (13, 14). For example, retention of part of intron 9 (referred to as “exon 9β” in this article) generates the isoform of TP53, which encodes a C-terminal truncated p53 protein due to introduction of a premature termination codon (PTC). This truncated TP53 RNA is typically degraded in cells via nonsense mediated RNA decay (NMD), but levels of TP53β RNA and protein appear to increase in some cells as they exhaust their replicative capacity and become senescent (15, 16). Although overexpression of p53β has been reported to trigger replicative senescence or apoptosis in different settings, little is known about the cellular or extracellular factors that regulate this alternative splicing.

As we conducted studies of ionizing irradiation (IR)–induced translational regulation of p53, we noted an increase in the generation of the TP53β RNA. Characterization of this unexpected finding uncovered a novel DNA damage–induced alternative splicing pathway that is required for increases in the levels of p53β as well as other alternatively spliced gene products. Generation of p53β after IR was dependent on inhibition of the SMG1 kinase, the binding of RPL26 to TP53 pre-mRNA, and the recruitment of the SRSF7 splicing factor. Similar signaling steps and induction of p53β were seen after cellular exposure to the base-damaging agent, methylmethanesulfonate (MMS). Interestingly, we noted that UV irradiation, which introduces bulky DNA adducts, also induces RPL26-dependent alternative splicing of TP53 mRNA, but in this case the ΔTP53 isoform, rather than the TP53β isoform, is induced. Characterization of the functional relevance of p53β induction demonstrated that overexpression of this p53 isoform induces a cellular senescent phenotype, as previously reported (17–19), and that this cellular outcome was dependent on the ability of the spliced protein product to bind to DNA and regulate transcription. Supporting a critical role for p53β and for this pathway in IR-induced cellular senescence, TP53β knockout or modulation of the other pathway regulators reduces IR-induced markers of cellular senescence. Thus, these studies demonstrate the existence of a previously unreported component of the DNA damage response (DDR) pathway whereby DNA damage induces alternative splicing of p53 and other gene products, and the induction of p53β appears to be involved in IR-induced cellular senescence.

We previously demonstrated that IR of cells causes RPL26 to bind to a double-strand RNA structure created by base-pairing of sequences from the 5′ and 3′ UTRs of TP53 mRNA and enhances p53 translation after DNA damage (5, 6). IR promotes the binding of RPL26 to p53 mRNA by inhibition of HDM2, which both mediates degradation of RPL26 and blocks binding of RPL26 to p53 mRNA (20). While characterizing the transcriptome bound by RPL26 after IR, we detected an abundance of the p53β splice variant in the RPL26 immunocomplex (Supplementary Fig. S1A). In unstressed cells, TP53β mRNA is rapidly degraded by the NMD pathway due to a premature termination codon inside the retained exon 9β. Because NMD takes place mainly in the cytosol (21, 22) and could complicate assessments of alternative splicing of TP53 mRNA, we quantitated the levels of TP53 pre-mRNA (unspliced message), total spliced TP53 mRNA, and TP53β variant mRNA (spliced messages) specifically in the nucleus (Fig. 1A). Based on current information of all identified p53 splice variants (reviewed in ref. 23), no alternative splicing events have been observed at introns 7 and 8. Splicing of intron 7 can serve as a measure of total spliced TP53 mRNA, whereas retention of intron 8 is seen only in TP53 pre-mRNA. We designed specific primers that cover these two regions to measure levels of total spliced (TS) TP53 mRNA and unspliced TP53 pre-mRNA, respectively. The ratio of spliced/unspliced message is a quantitative assessment of the relative amount of spliced mRNA.

Levels of nuclear TP53β mRNA in MCF-7 breast cancer cells rapidly and transiently increased after IR, whereas the levels of TP53 pre-mRNA and total spliced TP53 message remained unchanged (Fig. 1A, bottom left panel; Supplementary Fig. S1B). A significantly increased ratio of the spliced/unspliced mRNA after IR suggests a preferential RNA splicing to generate the β variant after damage (Fig. 1A, bottom right panel). Concomitantly, p53β protein levels also gradually increased and peaked around 30 minutes after IR (Fig. 1B). In contrast, full-length p53 protein usually achieves its maximum accumulation around 2 hours after damage. The induction of TP53β mRNA after IR was observed in a variety of cell types, including tumor lines such as HCT116 colorectal cells (data not shown) and SY5Y neuroblastoma cells (Supplementary Fig. S1A), and human primary adult fibroblasts (Fig. 1C).

The ATM protein kinase is a central mediator of many cellular responses to IR, including induction and phosphorylation of p53 protein (24). To assess the potential role of ATM in mediating IR-induced increases in TP53β mRNA, we treated MCF-7 cells with a small-molecule inhibitor of the ATM kinase, CP466722 (CP; ref. 25). As expected, CP treatment effectively blunted IR induction of full-length p53 protein (Fig. 2A, top band), an ATM-dependent induction. In sharp contrast, CP treatment not only failed to block IR-induced generation of TP53β mRNA, it instead robustly and sustainably increased TP53β mRNA (Supplementary Fig. S2A) and protein levels (Fig. 2B, Supplementary Fig. S2A) in the absence of IR and further enhanced p53β protein production after IR in primary fibroblasts (Fig. 2A). Similar to the observation with IR (data not shown), TP53β siRNA blunted the induction of p53β by CP at both the mRNA and protein levels (Supplementary Fig. S2B) and had no effect on the levels of full-length p53, thus confirming a specific impact on the alternatively spliced form of p53.

These results suggested that inhibition, rather than activation, of ATM could be contributing to generation of the p53β isoform. Because all kinase inhibitors affect multiple kinases, we attempted to more specifically explore the impact of the ATM kinase on p53β generation by using cells treated with ATM siRNA and cells in which the ATM gene had been deleted by use of CRISPR/Cas9 genomic manipulation. Unexpectedly, neither short-term (ATM siRNA) nor long-term (ATM CRISPR knockout) loss of ATM increased p53β levels nor affected the induction of p53β by CP (Fig. 2C; Supplementary Fig. S2C). These results suggested that the CP-mediated induction of p53β was occurring via inhibition of some other kinase. ATM protein is a serine/threonine protein kinase and a member of the phosphoinositide 3-kinase-related protein kinase (PIKK) family, which includes ATR, DNA-PK, and SMG1. Like ATM, all three of these kinases are genotoxic stress-responsive kinases (1, 26). In addition, SMG1 is a key kinase in the NMD pathway, where it phosphorylates the UPF1 protein (27). The NMD pathway is an RNA surveillance mechanism that targets RNAs with premature termination codons (such as TP53β) for degradation. Phosphorylated UPF1 facilitates the formation of NMD complex and executes the degradation process. Previous characterization of CP had demonstrated that it does not inhibit either ATR or DNA-PK (25), and the DNA-PK inhibitor NU 7026 failed to induce p53β (Supplementary Fig. S2D), thus raising the possibility that SMG1 could be the relevant kinase involved in p53β induction.

Treatment of MCF-7 cells with SMG1 siRNA increased cellular levels of the p53β isoform (Fig. 2D; Supplementary Fig. S2E and S2F), thus implicating SMG1 in this regulation. Because TP53β contains a premature stop codon and is vulnerable to NMD degradation, knocking down SMG1 could have induced p53β by blocking the NMD pathway. To examine this possibility, we used an alternative approach to blunt the NMD pathway, by knocking down UPF1, a downstream target of SMG1 and an effector in the NMD complex that degrades RNA. UPF1 siRNA treatment minimally increased levels of TP53β mRNA and protein levels in the cytoplasm, but had no significant effect on TP53β mRNA levels in the nuclear fraction (data not shown). Thus, the increase of the nuclear TP53β mRNA level occurring following reduction of the SMG1 level appears to be due to enhanced alternative splicing, not decreased degradation of the TP53β message by the NMD pathway.

In order to link inhibition of SMG1 kinase activity to the regulation of p53β seen after IR or CP treatment, we examined the phosphorylation of the endogenous substrate of SMG1, UPF1, after these treatments. The PIKKs, including SMG1, phosphorylate target proteins at S/TQ sites (27–29) and an antibody that specifically recognized phosphorylated S/TQ residues bound to immunoprecipitated UPF1 (Fig. 2E and Supplementary Fig. S2G), consistent with basal phosphorylation of UPF1 by a PIKK. Treatment of cells with CP (Fig. 2E, left), irradiation of cells (Fig. 2E, right), or knockdown of SMG1 with siRNA (Supplementary Fig. S2G) all attenuated UPF1 S/TQ phosphorylation. These results suggest that treatment of cells with IR or CP inhibits the kinase activity of SMG1 and promotes the alternative splicing of p53β.

To explore the mechanism of how SMG1 interferes with TP53β splicing, we investigated interactions between SMG1 and TP53 pre-mRNA. No differences in total SMG1 binding to TP53 pre-mRNA were noted before and after CP treatment using immunoprecipitation of SMG1 and RT-PCR for bound TP53 mRNA (data not shown). However, changes in site-specific binding of SMG1 associated with SMG1 inhibition were found (Fig. 2F). Using an RNase-mediated protein footprinting technique (30, 31) to fine-map the binding regions of SMG1 on TP53 pre-mRNA, we observed a CP-induced selective eviction of SMG1 from TP53 pre-mRNA at sites between introns 8 and 10 (covering exon 9 and exon 9β). Strikingly, this is exactly where the alternative splicing of p53β occurs (regions B, C, and F in Fig. 2F). The release of SMG1 from TP53 pre-mRNA after CP treatment is regulated locally because neither the binding of SMG1 to more distal regions (regions A and J in Fig. 2F) nor the global binding of SMG1 to the bulk TP53 pre-mRNA (data not shown) were affected. These results suggest that kinase-competent SMG1 occupies certain regulatory regions of precursor mRNA and that inactivation of the kinase exposes selected region(s) for factors promoting p53β splicing.

We previously implicated RPL26 protein in the regulation of p53 mRNA translation after IR via its direct binding to a 5′–3′ UTR interaction region of human TP53 mRNA in the nucleus (5, 6). Because TP53β and full-length TP53 mRNA are transcribed from the same promoter and share the same UTR regions, RPL26 should also have the capacity to bind TP53β mRNA. In fact, the increased levels of TP53β mRNA after IR were first noted in an evaluation of the transcriptome bound to RPL26 following IR (Supplementary Fig. S1A), suggesting direct binding of RPL26 to TP53β mRNA.

Supporting a role of RPL26 in the induced alternative splicing of the p53β variant, downregulation of RPL26 strongly suppressed the induced levels of TP53β mRNA and protein (Fig. 3A; Supplementary Fig. S3A and S3B). With primers designed to detect unspliced TP53 pre-mRNA (Fig. 1A), the direct binding of RPL26 protein to TP53 pre-mRNA was also confirmed and linked to this regulatory pathway by the observation that downregulation of SMG1 increased the amount of both unspliced TP53 pre-mRNA and spliced TP53β mRNA in the nuclear anti-RPL26 immunocomplex (Fig. 3B; Supplementary Fig. S3C and S3D). Thus, RPL26 contributes to IR-induced and CP-induced increases in p53β levels and directly binds to TP53 pre-mRNA in the nucleus in a manner dependent on inhibition of SMG1 and its associated selective release from the exon 9 region of TP53 pre-mRNA.

Though both SMG1 and RPL26 appear to regulate alternative splicing of TP53 mRNA, neither are splicing factors. Although SMG1 and RPL26 appear to be transducers of the DNA-damage signal to elicit the alternative splicing of p53β, there must be a downstream executor of the actual splicing process. Removal of introns from precursor mRNAs is carried out by the spliceosome, a highly dynamic protein–RNA complex (32). The assembly of spliceosomes on precursor mRNAs and the splicing events themselves are tightly regulated. Two families of proteins have been recognized as classic trans-acting splicing regulators: serine/arginine-rich splicing family proteins and heterogeneous nuclear ribonucleoproteins (hnRNPs). Among 12 SR proteins, SRSF3 (SRp20), the smallest family member, has been reported to be a negative regulator of p53β splicing and participates in p53β-dependent replicative senescence (18). In a recent genome-wide siRNA screen, SRSF7 stands out as an important regulator of the alternative splicing of exon 6 of FAS/CD95 and other AS events in genes involved in apoptosis (33). This SR family member has been shown to cooperate with SRSF3 for the alternative exon splicing of CD44 and transport of viral RNAs (34), and altered expression of SRSF7 is associated with renal cancer (35). Knockdown of SRSF3 in MCF-7 cells actually increased basal p53β levels (Supplementary Fig. S4A), whereas SRSF7 knockdown reduced both basal and induced p53β levels (Fig. 4A; Supplementary Fig. S4B). Thus, SRSF7 appears to be the splicing factor involved in damage induction of p53β. Because SRSF7 siRNA had no effect on the basal splicing of the full-length TP53 message (Supplementary Fig. S4C), it seems that the alternative splicing of p53β is a specific process mediated by SRSF7.

The splicing activity of SRSF7 is subject to iron regulation through its Zinc Knuckle domain (33). Intracellular iron inhibits the binding of SRSF7 to FAS/CD95 mRNA and results in the exon 6 inclusion. Consistent with a role for this iron-dependent enzyme in alternative splicing of p53, decreasing intracellular iron levels by treatment with the iron chelator deferoxamine (DFO) suppressed p53β splicing, and elevating iron levels by treatment of cell with Hemin enhanced it (Fig. 4B; Supplementary Fig. S4D). IR treatment had no effect on SRSF7 protein levels or intracellular distribution (data not shown), but IR treatment did increase the binding of SRSF7 to TP53 pre-mRNA (Fig. 4C). Further, a fraction of SRSF7 protein binds to RPL26 protein independent of RNA (Supplementary Fig. S4E). As expected, this interaction is more easily detected in the nucleus than in the cytoplasm (Fig. 4D) because the majority of SRSF7 locates in the nuclei. No interactions between RPL26 and SRSF3 were noted (data not shown). Importantly, siRNA knockdown of RPL26 blunted the binding of SRSF7 to precursor TP53 mRNA (Fig. 4E), demonstrating a functional role for RPL26 in the recruitment of SRSF7 to TP53 mRNA (Supplementary Fig. S4F).

To assess whether DNA damaging agents other than IR could also activate this alternative splicing pathway, we treated cells with either MMS or UV irradiation. Similar to the cellular responses seen with IR treatment, exposure to the base-damaging agent, MMS, induced a rapid induction of both TP53β mRNA and protein, and this induction was dependent on both RPL26 and SRSF7 (Supplementary Fig. S4G). In contrast, although UV irradiation also induced RPL26-dependent alternative splicing of TP53, the Δ isoform of p53 was induced, not the β isoform (Supplementary Fig. S4H and S4I). Thus, different types of DNA damage appear to regulate splicing differently.

Previous reports have suggested that the β isoform of p53 contributes to replicative senescence in lymphocytes and fibroblasts (15, 16). A genetically manipulated mouse expressing a p53 protein lacking the carboxyl terminus, similar to the human p53β protein, exhibits deregulated telomere metabolism and premature senescence (36), and overexpression of human p53β induces apoptosis or cellular senescence in different systems (15, 17, 37). Consistent with these observations, we found that overexpression of human p53β increased markers of cellular senescence in a variety of human tumor cell lines (MCF-7, A549, and HCT116), including enlarged cell size, decreased cell proliferation, and elevated senescence-associated β-galactosidase (SA-β-gal) activity (Fig. 5A; Supplementary Fig. S5A), but had no effect on apoptosis (Supplementary Fig. S5B). Introduction of an R175H hotspot mutation, which impairs the ability of full-length p53 protein to bind DNA (38), in the TP53β cDNA abolished p53β-dependent growth arrest and SA-β-gal activity (Fig. 5A), suggesting that the induction of senescence characteristics by p53β requires its ability to bind to DNA, and presumably its ability to affect transcription. Notably, overexpression of p53β failed to induce growth arrest, senescence-associated β-galactosidase activity, or apoptosis in cells lacking endogenous p53 genes (p53-null HCT116 and H1299; Supplementary Fig. S5B and S5C and data not shown). However, we could not detect significant binding between p53β and full-length p53 proteins (Supplementary Fig. S5D), suggesting that any functional interaction of p53β and full-length p53 proteins is not through a physical interaction, but rather via collaborative effects of these two proteins on transcriptional regulation. Consistent with this suggestion, characterization of p53 target gene profiles regulated by p53β expression revealed three groups of differentially regulated genes: group I, regulated by p53β independent of full-length p53 status (e.g., p21, TP53INPI, etc.); group II, regulated by p53β, but dependent on the presence of full-length p53 (e.g., GDF15 and ADAMTS7); and group III, not regulated by p53β (Supplementary Fig. S5E). Thus, the induction of alternatively spliced p53β has the potential to interfere with regulation of the p53 network and alter cellular responses to DNA damage and other stresses.

In order to more effectively investigate the functional roles of p53β in response to DNA damage, we utilized CRISPR/Cas9 gene editing to specifically knock out p53β without affecting full-length p53 expression and induction. We established two p53β CRISPR clones that lost expression of TP53β mRNA due to a deletion of the branch site adenine (A) upstream of the polypyrimidine tract and the 3′ acceptor site within the intron 9 region of the TP53 gene required for β splicing (Fig. 5B; Supplementary Fig. S5F). Although treatment of these mutated cells with IR, CP, or SMG1 siRNA could no longer induce p53β in these cells (Fig. 5B), neither cellular proliferation (Supplementary Fig. S5G, left) nor the induction of full-length p53 and p53-dependent transactivation of p21 by IR (Supplementary Fig. S5G, right) were affected. In contrast, IR induction of the cellular senescence markers, SA-β-galactosidase activity (Fig. 5C, Supplementary Fig. S6A) as well as markers of the senescence associated secretory phenotype (SASP; Supplementary Fig. S6B), were significantly reduced in the p53β-knockout cells and p53β knockdown cells (Supplementary Fig. S6A, right graph). Further, modulation of the “SMG1–RPL26–SRSF7” pathway required for IR-induced TP53β alternative splicing also affected cellular senescence markers as well. Knockdown of SMG1, which activates this splicing pathway, increased the percentage of SA-β-galactosidase–positive cells (Fig. 5D, Supplementary Fig. S6C, left graph), and knockdown of SRSF7, a positive splicing factor for p53β, suppressed IR-induced SA-β-galactosidase induction (Fig. 5E; Supplementary Fig. S6C, right graph). Thus, modulation of multiple steps in this alternative splicing pathway, SMG1, SRSF7, and p53β, all affect IR induction of cellular senescence markers.

To begin to clarify the mechanisms by which p53β contributes to IR-induced senescence, we utilized a human whole transcriptome array to compare the global gene expression profiles of control and p53β CRISPR knockout clones before and after IR. Control cells and p53β CRISPR knockout cells differentially expressed 72 genes and 103 genes (with a fold change ≥2), respectively, 4 hours after IR (Supplementary Fig. S6D). Among these genes, 43 genes changed exclusively in the irradiated p53β CRISPR knockout clones (mainly upregulated) and 14 unique genes changed in the control group (Supplementary Fig. S6D). Gene ontology analyses (Supplementary Table S1) revealed common IR-induced pathways in the control versus p53β CRISPR knockout cells, such as “thymocyte apoptotic process (GO: 0070242)” and “neutrophil activation involved in immune response (GO: 0002283).” On the other hand, the “negative regulation of cell aging (GO: 0090344)” pathway with BCL6 and SIRT1 was suppressed in control cells after IR, but not in the p53β CRISPR knockout cells (Fig. 5F, top). The failure of the p53β CRISPR knockout cells to repress these gene products after IR was specifically validated in separate experiments (Fig. 5F, bottom). These observations suggest that p53β may regulate IR-induced senescence by repressing negative regulators of aging. Together, these data imply that p53β may link more closely to DNA damage–induced senescence than other p53-dependent phenotypes, such as damage-induced apoptosis and growth arrest.

To ask whether the “SMG1–RPL26–SRSF7” pathway induced alternative splicing of gene products other than p53, we analyzed the microarray data at an exon level for IR-induced alternatively spliced RNAs. A total of 607 alternative splice events were noted in IR-treated cells, with no significant change of the transcription of those genes. The top and bottom 20 alternative splice events were ranked by the value of their splicing index (Supplementary Table S2), and we chose to validate two of these events, CDYL and SMAD3 (Fig. 6A). In contrast to the intron retention and premature stop codon required to generate p53β, the CDYL splice variants were generated by differential exon usage. The CDYL970 variant shares a 3′ sequence with CDYL4824 (major form), but includes an alternative first exon. To confirm that the “SMG1–RPL26–SRSF7” pathway was involved in IR-induced alternative splicing of CDYL, we treated cells with siRNAs of SMG1 or SRSF7. Reduction of SMG1 specifically increased levels of CDYL970 isoform (Fig. 6B), thus recapitulating the splicing of CDYL observed in IR-treated cells (Fig. 6A, left). Reduction of SRSF7 expression mildly repressed the basal level of CDYL970 splice variant and significantly repressed the IR-induced alternative splice form (Fig. 6C). This was a specific effect on this splice variant, as another alternative splice form, CDYL4824, was not affected in the same ways.

We identified a new signaling pathway that regulates the alternative splicing of p53 and other mRNA species in response to genotoxic damage and participates in the regulation of radiation-induced cellular senescence markers (Fig. 6D; Supplementary Fig. S4F). IR inhibits the activity of the PIKK family kinase SMG1, releasing it from the exon 9 region of TP53 pre-mRNA, which leads to the binding of RPL26 protein to TP53 pre-mRNA and recruitment of the RNA splicing factor SRSF7. All of these steps are required for increased expression of TP53β mRNA and protein after IR and also appear to regulate the alternative splicing of other gene products as well. Similar cellular responses are seen after cellular exposure to the base-damaging agent MMS. Inhibition of these pathway steps, including specific knockout of the β isoform of p53, significantly blunts cellular senescence markers induced by irradiation. The regulation of cellular senescence by p53β requires its ability to bind to DNA, appears to depend on functional interplay between the alternatively spliced p53 protein and full-length p53 protein, and is associated with transcriptional repression of select gene products involved in regulation of cellular aging, such as BCL6 and SIRT1. Therapeutic irradiation has numerous side effects, including induction of tissue fibrosis, cellular senescence, and aging phenotypes (39–41). The model pathway reported here suggests that suppression of p53β induction or activity could mitigate some of these toxicities. Future studies in vivo could explore these possibilities.

In unstressed cells, TP53β mRNA levels are kept low by two mechanisms: One is tightly regulated alternative splicing of TP53β by kinase-competent SMG1, which occupies the regions of TP53 pre-mRNA required for p53β splicing and prevents the access of positive splicing regulators (i.e., RPL26 and SRSF7) to the unspliced message (Figs. 2–4, model in Supplementary Fig. S4F). In addition to limited production of TP53β mRNA due to suppression of its splicing, the retention of exon9β in TP53β mRNA introduces a premature stop codon that would activate the NMD pathway for its rapid degradation. Because SMG1 is also a critical kinase for the NMD pathway, these two mechanisms both rely on the kinase activity of SMG1, but differ in location. Although splicing events occur inside nuclei and most RNA decay pathways act in the cytosol, there are some reports suggesting that NMD degradation can take place in the nuclei, depending on the nature of the mRNA (42, 43). We noted that nuclear TP53β mRNA levels are around 5-fold higher than its levels in the cytoplasm (data not shown), suggesting that any NMD that might occur in the nucleus is likely far less efficient at degrading TP53β message than cytoplasmic NMD and supports the focus on interactions of the various pathway components in the nuclear fraction of cells.

In contrast to the activation of other PIKKs, including ATM, ATR, and DNA-PK, that occurs after genotoxic stress (1, 26, 44), SMG1 kinase activity appears to be blunted at early time points after IR (Fig. 2E). The inhibition of SMG1 activity after IR or CP treatment leads to local eviction of SMG1 protein from TP53 precursor mRNA and allows binding of RPL26 protein to TP53 precursor mRNA, an interaction that facilitates recruitment of SRSF7 to promote the alternative splicing of TP53. Both RPL26 and SRSF7 bind to TP53 pre-mRNA, which is present only in the nucleus (Figs. 3B, 4C and 4E), similar to our previous observations demonstrating the role of RPL26 in regulating p53 translation after DNA damage by binding to mature TP53 mRNA in the nucleus (5, 6). RPL26 is also an integral ribosomal protein involved in rRNA processing in the nucleoli and general translation in the cytoplasm. However, because a pioneer round of translation is required for eliciting NMD of mRNAs containing premature termination codons (21, 22), if RPL26 were affecting levels of TP53β mRNA through regulation of this initial translation step as a component of ribosomes, then knocking down RPL26 should stabilize p53β mRNA instead of inhibiting p53β synthesis. Thus, like regulation of DNA damage–induced p53 protein translation (5, 6, 45), multiple lines of evidence suggest that the regulation of p53β levels is an extraribosomal activity of RPL26.

Although we have not yet fully characterized the molecular mechanisms involved in the TP53 RNA splicing events, the iron-dependent SRSF7 splicing regulator is clearly involved. We demonstrated that SRSF7 interacts with RPL26 protein, that the binding of SRSF7 to TP53 pre-mRNA requires RPL26, and that SRSF7 downregulation blunts CP and IR induction of p53β (Fig. 4; Supplementary Fig. S4). In contrast, SRSF3 downregulation has the opposite effect and actually increases levels of p53β (Supplementary Fig. S4A), which is consistent with previous reports (18). SRSF7 is an iron-dependent splicing factor, and p53β expression was reduced by iron chelation and enhanced by iron elevation (Fig. 4B; Supplementary Fig. S4D). We suspect that the general RNA binding activity of SRSF7 is not directly altered by iron because the enzyme contains a highly conserved RNA Recognition Motif (RRM) whose function is independent of iron (33), but rather that iron is altering the binding preferences of SRSF7 to a different subset of mRNAs and facilitating select alternative splicing events.

Over 90% of human genes have at least 2 to 3 alternative splice variants (7, 46). Nine different transcripts are derived by splicing TP53 precursor mRNA, encoding for 12 different proteins. p53 is a central mediator of cellular stress responses, but stress regulation of p53 is largely posttranscriptional (47), and cellular effects have largely been attributed to functions of the full-length protein. However, there have been some suggestions that temporal or spatial expression of p53 protein variants might also contribute to cellular outcomes during p53-dependent stress responses. Consistent with prior suggestions that p53β may participate in the senescence process caused by either exhaustion of replication capacity (replicative senescence) or stress-induced premature senescence (15, 16), we found that the induction of p53β contributes to IR-induced cellular senescence. Although specific deletion of p53β does not interfere with cell growth, DNA damage signaling (Supplementary Fig. S5G), or p53-dependent induction of genes involved in apoptosis or inflammatory responses (Supplementary Fig. S5B and Supplementary Table S1), its deletion does relieve both IR induction of senescence markers and IR repression of certain negative regulators of cell aging (Fig. 5F). We validated two such p53β targets, BCL6 and SIRT1, both of which are epigenetic transcriptional repressors, suggesting that p53β may influence IR-induced senescence via chromatin effects. Interestingly, in contrast to the BCL6 repression seen with p53β expression, full-length p53 promotes BCL6 transcription (48). Because we noted that p53β induced senescence only in the presence of an endogenous p53 allele (Supplementary Fig. S5C), these observations further support the concept that p53β induction is altering the p53 target profile.

It seemed unlikely that the “SMG1–RPL26–SRSF7” pathway exists to regulate the alternative splicing of just p53. In fact, changes in hundreds of alternative splice products were noted after IR exposure (Supplementary Table S2). We validated IR-induced alternative splice changes in two of these gene products, CDYL and SMAD3, and demonstrated that CDYL regulation is also modulated by SMG1 and SRSF7 (Fig. 6). The functional roles of these alternative gene products in IR responses have not yet been elucidated. Thus, this pathway appears to be a general modulator of alternative splicing after IR or MMS treatment. Kornblihtt and colleagues previously reported changes in alternative splicing of BCL-X induced by UV irradiation in a p53-independent manner, a process that was dependent on inhibition of RNA polymerase II elongation (11). Although we did not observe induction of p53β after UV irradiation, induction of the Δ isoform of p53 was noted after UV exposure (Supplementary Fig. S4H). Thus, DNA damage induction of alternative splicing events may be a general phenomenon, but the exact mechanisms involved and which specific alternative gene products are induced appears to vary between different types of damage or stress. Future studies will be needed to characterize splicing pathways induced by different types of damage or stress and the functional roles of the alternatively spliced products. A virtually unlimited repertoire of novel gene products can be created by alternative splicing pathways. Thus, it now appears clear that DNA damage signaling pathways have the capacity to modulate gene expression beyond regulation of translation or transcription.

MCF-7, A549, H1299, and HCT116 cells were maintained in DMEM plus 10% FBS and 1% antibiotics. SY5Y cells, both control and ATM CRISPR knockout, were maintained in RPMI supplemented with 15% FBS plus 1% antibiotics. Human adult fibroblast-PCS-201-012 was purchased from the ATCC and cultured in the same medium. MCF-7, A549, H1299, HCT116WT, HCT116 p53^−/−, and HCT116 p21^−/− cell lines stably expressing FLAG-tagged p53β, p53βR175H, or GFP-tagged p53β were generated by infection of lentiviruses and selected with 2 μg/mL of puromycin. After selection, stable lines were maintained in DMEM plus 10% FBS supplemented with 1 μg/mL of puromycin. siRNA duplexes (4 nmol/L) were transfected into cells using Lipofectamine RNAiMAX (Invitrogen). MCF-7 (date of purchase, 4/18/2006), SY5Y (2/12/1999), A549 (2/7/2006), H1299 (3/24/2000), HCT116WT (2/1/2003), HCT116 p53^−/− (5/1/2009), and HCT116 p21^−/− (11/14/2005) were purchased from the ATCC between the years 1999 and 2009 and authenticated on August 29, 2016, when the manuscript was under review, using certified human cell line authentication (CLA) analysis provided by the Duke University DNA analysis facility.

Plasmids used in this study for viral packaging and generation of stable lines include pLenti-FLAG p53β, pLenti-FLAGp53γ, pLenti-FLAG p53βR175H, and pLenti-GFP-p53β. CP466722 was synthesized in the Chemical Biology Department at St. Jude Children's Research Hospital (49). DFO (Cat#D9533) and Hemin (Cat#H9039) were purchased from Sigma. All chemicals were dissolved in DMSO. siRNA duplexes were designed and synthesized by Sigma. Their sequences are listed in Supplementary Table S3. ATM siRNA (human ATM ON-TARGET plus SMARTpool) and CHK2 siRNA (human CHEK2 siGENOME SMARTpool) were purchased from Thermo Scientific Dharmacom.

The RNase-mediated protein footprint protocol was adopted from refs. 30 and 31. In brief, MCF-7 cells in 15-cm dish with 80% confluence were fixed with 1% final concentration of formaldehyde at room temperature for 10 minutes with shaking. A final concentration of 125 mmol/L glycine was then added drop-wise for an additional 5 minutes of incubation. Then, cells were washed twice with ice-cold PBS and collected. Cell pellets were lysed in lysis buffer as described in ref. 6. SMG1 protein was immunoprecipitated from cell lysate using appropriate antibody (Bethyl Laboratories), washed with NT2 buffer (6), and resuspended in 100 μL 1× RNaseONE digestion buffer followed by 100 U/mL RNaseONE digestion (Promega) with 200 μg/mL BSA for 1 hour at room temperature. The proteins were then denatured and digested by 1% SDS and 0.1 mg/mL proteinase K (Roche) for 15 minutes at room temperature before 2 hours of incubation at 65°C for reverse cross-linking. After reverse cross-linking, digested RNA was extracted by phenol-chloroform for real-time RT-PCR as described below. IP-RT-PCR reaction was performed without crosslink and RNase digestion as described in ref. 6.

Cell lysates were prepared by a freeze–thaw, followed by incubation in RIPA buffer for 30 minutes on ice, and the supernatants were analyzed by immunoblot analysis or immunoprecipitation. For immunoblot analysis, 20- to 50-μg protein samples were denatured in equal volume of SDS sample buffer (BioRad), separated by 4% to 12% SDS-PAGE and transferred to nitrocellulose membrane. The blots were probed with primary antibody (at concentrations recommended by the manufacturers) against p53 (DO-1; Santa Cruz Biotechnology), nucleolin (MS-3; Santa Cruz Biotechnology), PARP, pSQ/TQ (Cell Signaling Technology), RPL26, UPF1, SMG1, SRSF7 (Bethyl Laboratories), LDH (Abcam), ATM (25), CHK2 (Santa Cruz Biotechnology), SRSF3 (Life Technology), and p21 (BD Pharmingen). Primary antibody binding was detected by incubating with horseradish peroxidase (HRP)–conjugated anti-rabbit, anti-mouse secondary antibody (1:1,000 dilution; Rockland Inc.) followed by enhanced chemiluminescent visualization (ECL) system (Amersham Biosciences). For immunoprecipitation of UPF1, 1 mg whole-cell extract in RIPA buffer (6) was cleared by protein A/G-PLUS agarose beads (Calbiochem) and rabbit IgG (Sigma-Aldrich). Precleared lysates were incubated with anti-UPF1 antibody (Bethyl Laboratories). Immunoprecipitated proteins were then washed extensively with lysis buffer and subjected to western blot analysis as described above. For coimmunoprecipitations, cells were lysed in TGN buffer (6), and the rest of the procedure followed the immunoprecipitation protocol.

RNA (1 μg) prepared using TRIzol reagent (Invitrogen) was treated with DNase I (Invitrogen) and then reverse transcribed following the high-capacity cDNA reverse transcription kit instruction manual (ABI). Real-time PCR was performed using the StepOne plus real-time PCR system (ABI). The reaction was performed in triplicate using FastStart Universal SYBR Green Master reagents (Roche). PCR condition was 95°C 10′ (denature) followed by 40 cycles of 95°C 3″, 60°C or 56°C 30″ for amplification and 95°C 15″, 60°C 1′ for the melting curve. The mRNA level was calculated using the ΔΔCt method and normalized to the GAPDH expression level. Primer sequences for qPCR are listed in Supplementary Table S3.

Flow SA-β-galactosidase assay was adopted from ref. 50. In brief, subconfluent cells were first treated with 100 nmol/L bafilomycin A1 (Sigma) for 1 hour in fresh DMEM plus 10% FBS and 1% antibiotics medium at 37°C, 5% CO₂. A final concentration of 33 μmol/L C₁₂FDG (Setareh Biotech) was added to the medium for an extra 2 hours of incubation. The cell monolayer was then washed twice with room-temperature PBS and harvested by trypsinization followed by centrifugation at 200 × g for 10′ at 4^oC. Cells were resuspended in ice-cold PBS at a concentration of 1 × 10⁶ cells/mL and run immediately on BDCalibar. Data were collected and analyzed using CellQuest following the instructions of the manufacturer's manual. SA-β-galactosidase staining was performed following the protocol provided by the Senescence β-galactosidase cell staining kit (Cell Signaling Technology). Stained cells were maintained in an appropriate amount of 70% glycerol and subjected to microscopy.

To knock out p53β, guide RNA (gRNA) targeting p53β (Forward: 5′CACCGATATATATTATGGTATAAGT; Reverse: 5′AAACACTTATACCATAATATATATC) was designed and cloned into lentiCRISPRv2 plasmid for viral packaging. MCF-7 cells were infected with the viruses and selected by 2 μg/mL puromycin for 2 days, followed by 14-day maintenance in 1 μg/mL puromycin. The stable p53β CRISPR knockout clones undergo serial dilution for single-cell separation. The individual p53β CRISPR knockout cells were expanded in puromycin-free DMEM supplemented with 10% FBS and 1% antibiotics. Finally, the DNA was extracted from these clones and the region surrounding the Cas9 cutting site was PCR amplified for sequencing.

Total RNAs of two control cells and two p53β CRISPR knockout clones in triplicate without any treatment or 4 hours after 20 gy irradiation were extracted using an RNeasy mini kit (Qiagen) and treated with DNAse I (Life Technology). The mRNA samples were assayed using the Affymetrix GeneChip Human Transcriptome Array 2.0 platform at the Sequencing and Genomic Technologies Shared Resource of the Duke Cancer Institute. The resulting arrays were preprocessed using the Robust Multichip Average (RMA) algorithm (51), followed by quantile normalization at the probe (n = 925,032) and gene (n = 70,523) levels using the RMA function in the R (52) extension package affy (53). The annotation and library files were downloaded from the Affymetrix website (release 35.2). The Transcriptome Analysis Console (TAC) 3.0 software was used for alternative splicing analysis using the Affymetrix recommended filter criteria and algorithm options (54). The alternative splicing index was calculated as follows:

for gene i and exon j without IR (condition (0)) or with IR (condition (1)). To account for multiple testing, FDR adjusted P values were computed using the Benjamini–Hochberg method (55). Gene pathway analyses were conducted using a resampling-based method implemented by the R extension package safe (56) using Gene Ontology categories (57). Microarray data have been deposited to the GEO database with accession number GSE84813.

No potential conflicts of interest were disclosed.

Conception and design: J. Chen, M.B. Kastan

Development of methodology: J. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Chen, J. Crutchley

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Chen, J. Crutchley, D. Zhang, K. Owzar, M.B. Kastan

Writing, review, and/or revision of the manuscript: J. Chen, J. Crutchley, K. Owzar, M.B. Kastan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Chen, J. Crutchley

Study supervision: M.B. Kastan

We would like to thank the Duke Sequencing and Genomic Technologies Core facility (a Duke Cancer Institute CCSG and a Duke Genomic and Computational Biology shared resource facility) for their technical support, microarray data management, and feedback on the generation of the microarray data reported in this manuscript. We also thank Dr. Donald Fleenor for outstanding technical support, and members of the Kastan lab for ideas and manuscript review.

This work was supported by grants R01ES005777 and P30CA014236 from the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.