Y14 governs p53 expression and modulates DNA damage sensitivity

Y14 is a core component of the exon junction complex (EJC), while it also exerts cellular functions independent of the EJC. Depletion of Y14 causes G2/M arrest, DNA damage and apoptosis. Here we show that knockdown of Y14 induces the expression of an alternative spliced isoform of p53, namely p53β, in human cells. Y14, in the context of the EJC, inhibited aberrant exon inclusion during the splicing of p53 pre-mRNA, and thus prevent p53β expression. The anti-cancer agent camptothecin specifically suppressed p53β induction. Intriguingly, both depletion and overexpression of Y14 increased overall p53 protein levels, suggesting that Y14 governs the quality and quantity control of p53. Moreover, Y14 depletion unexpectedly reduced p21 protein levels, which in conjunction with aberrant p53 expression accordingly increased cell sensitivity to genotoxic agents. This study establishes a direct link between Y14 and p53 expression and suggests a function for Y14 in DNA damage signaling.


Results
Depletion of Y14 induces a p53 isoform. To evaluate the cellular function of Y14, we depleted Y14 in HeLa cells by transient transfection of a short interfering RNA (siRNA) that targets Y14. The level of Y14 and its protein partner Magoh was significantly reduced by Y14 siRNA (Fig. 1A). As reported 9,10 , we observed that Y14 depletion caused cell-cycle arrest at G2/M phase and increased the sub-G1 population and apoptosis in HeLa cells ( Supplementary Fig. 1). While examining the expression of cell-cycle and apoptotic factors, we fortuitously found that Y14 depletion induced a p53 band of smaller molecular size (Fig. 1A). This band was detected by a monoclonal antibody that recognizes an N-terminal peptide of p53 (DO-1), but not by a C-terminal peptide antibody (C- 19), suggesting that the band represented a C-terminal deletion variant of p53 (Fig. 1A).
Using reverse transcription (RT)-PCR with different sets of primers (Fig. 1B), we confirmed that Y14 depletion induced the expression of the p53β isoform that includes exon i9, an aberrant exon residing in intron 9, and encodes a truncated protein with a C-terminal sequence distinct from that of p53α (hereafter p53α refers to full-length p53) (Fig. 1C). No product corresponding to p53γ (208 bp) was detected (Fig. 1C, P3-P6). It has been reported that p53β can be induced by depletion of the splicing regulator SRSF3 or by treating cells with caffeine or digoxin 14,15 . This result suggested a role of Y14 in regulating p53 splicing. Moreover, Y14 depletion induced Bcl-xS, as previously reported 6 , confirming the effect of Y14 on splicing regulation. Notably, a mild increase of p53β mRNA did not reflect its protein level, which was almost comparable to that of p53α (Fig. 1A); this could be explained by the higher protein stability of p53β than p53α 16 .
Next, we depleted Y14 in a variety of cancer cell lines and examined p53 expression. Y14 depletion induced p53β protein significantly in breast cancer MCF7 cells and minimally in colon cancer HCT116 cells (Fig. 1D, lanes 4, 6). Because the TP53 allele in head and neck squamous carcinoma SAS cells contains a nonsense mutation in exon 10, Y14 depletion-induced p53β was indistinguishable from the truncated mutant p53 in SAS cells (lanes 7, 8). Nevertheless, we observed an increase in p53β mRNA expression, albeit to different extents, in all cell lines examined (Fig. 1D, RT-PCR). Additionally, we noted that the level of total p53 proteins increased after Y14 depletion, which was consistent among different cell lines (Fig. 1A,D). To examine how Y14 depletion causes p53 stabilization, we co-expressed FLAG-tagged p53 and HA-tagged ubiquitin in HeLa cells, and observed that anti-FLAG precipitated p53 was ubiquitinated after treatment of the cells with the proteasome inhibitor MG132 ( Supplementary Fig. 2, lane 2). Ubiquitinated p53 was not detected in cells expressing the mutant ubiquitin (lanes 3,4) or in Y14-depleted cells (lanes 5-8). Thus, Y14 depletion might prevent p53 ubiquitination. Moreover, the p53 protein level increase was not observed upon depletion of eIF4A3 (see below for Fig. 2), suggesting a specific role for Y14 in p53 protein biogenesis.
We noted that the mouse p53 gene has a shorter intron 9, which exhibits limited sequence homology to the 3′ part of its human counterpart ( Supplementary Fig. 3A). A product equivalent to human p53β was not detected in Y14-depleted mouse neuroblastoma Neuro2a cells ( Supplementary Fig. 3B). However, we noted that the level of p53 protein was raised in Y14-depleted Neuro2a cells, which was reminiscent of p53 activation in Rbm8a halopinsufficiency mouse embryos 12 , suggesting a conserved role for Y14 in p53 protein stabilization. The EJC core prevents p53β expression. Next, we examined whether depletion of any other EJC factor or disruption of nonsense-mediated mRNA decay (NMD) also induces p53β . The result showed that p53β was induced by deletion of eIF4A3 ( Fig. 2A) or Magoh ( Supplementary Fig. 4A) but not by depletion of the NMD factor Upf1 ( Fig. 2A) or overexpression of the dominant-negative Upf1 ( Supplementary Fig. 4B). Therefore, p53β was likely induced by depletion of the EJC but not by blockage of the NMD pathway. Although the level of p53β mRNA was extremely low, overexpression of Y14 or eIF4A3 reduced basal p53β expression (Fig. 2B). Knockdown of another EJC factor (eIF4A3 or Magoh) greatly reversed the effect of Y14 overexpression (Fig. 2C, lanes 3, 4), as expected. This result may moreover imply the concerted action of the EJC core factors in preventing exon i9 inclusion.
We suspected that splicing of intron 10, which carries the strong 5′ and 3′ splice sites, may be faster than that of intron 9, and hence the EJC is recruited to prevent aberrant exon i9 inclusion. We therefore generated the . To further test our supposition that efficient intron 10 splicing suppresses exon i9 inclusion via recruiting an EJC, we generated a tethering reporter by replacing intron 10/exon 11 with a fragment containing six tandem repeats of the MS2 coat protein (MCP) binding sites (Fig. 2D, p53mini-MS2). The resulting reporter was co-transfected with the vector expressing an MCP-Y14 fusion into HeLa and MCF7 cells. RT-PCR analysis revealed that MCP-Y14 could reduce p53β expression from the minigene (Fig. 2G), indicating that Y14 association with a downstream exon-mimicking the anchor of an EJC-suppressed exon i9 inclusion. Hence, several lines of the data implied that EJC deposition modulates p53β expression ( Fig. 2B,C,F,G).
Camptothecin disrupts Y14 depletion-induced p53β expression. Because the transcriptional elongation rate affects exon selection 17 , we tested whether the RNA polymerase II inhibitor 5,6-dichloro-1-β -d-ribofuranosyl-benzimidazole (DRB) influences p53β expression. We also examined two anti-cancer and DNA-damaging agents, camptothecin (CPT) and doxorubicin, that respectively inhibit DNA topoisomerase (TOP) I and II. CPT creates TOP1-DNA adducts, which also interfere with transcription elongation 18 . DRB and CPT have differential impact on splicing [19][20][21] . When Y14-depleted cells were treated with either of these compounds, immunoblotting revealed that CPT particularly reduced p53β protein (Fig. 3A). RT-PCR confirmed that CPT blocked the expression of the basal and Y14-induced p53β mRNA isoform (Fig. 3B, cp. lanes 1 and 3 in P3-P4). Moreover, CPT also eliminated the effect of Y14 depletion on two other targets, namely Bcl-x and SF1 (Fig. 3B, lanes 1-4). Nonetheless, DRB or doxorubicin had no such effects (Fig. 3B, lanes 5-10). Moreover, we examined two CPT target transcripts, EIF2S2 and PNN 20 . Y14 depletion neither affected their splicing (Fig. 3C, lane 2) nor altered the effect of CPT (lane 4). Thus, Y14 depletion and CPT treatment affected the splicing of overlapping, but still distinct, sets of transcripts. More importantly, our result indicated that CPT particularly reversed the effect of Y14 depletion in alternative splicing.
Notably, CPT preferentially affects the splicing of splicing factors, including Y14 20 ( Supplementary Fig. 5). CPT treatment likely yields a loss-of-function isoform of Y14 and thus impacts on the splicing of Y14 targets. However, this cannot explain how CPT specifically blocked Y14 depletion-induced alternative splicing. This intriguing result must be a topic of future investigation. Additionally, CPT exposure causes co-transcriptional R-loop formation 18 . The possibility of whether such a structure recruits a factor(s) to facilitate the selection of EJC-sensitive exons also remains to be tested by future experiments.
The EJC ensures efficient translation of p53 mRNA. The above data showed that Y14 depletion induced p53β and may additionally increase p53 protein level, although by which mechanism the latter was achieved is unclear. Hence, it was necessary to examine the effect of Y14 overexpression on p53. To our surprise, we observed a minimal increase in the level of p53 protein. The level of p53 mRNA remained unchanged except  Fig. 6). This slight p53 increase was also observed upon overexpression of FLAG-eIF4A3, but not of the mRNA cap-binding mutant of Y14 (W73V) 4 (Fig. 4A, lanes 3, 4). Accordingly, immunoprecipitation and RT-PCR showed that the wild-type Y14, but not the W73V mutant, associated with the p53 mRNA (Fig. 4B). However, we were unable to detect a significant shift of p53 mRNA in polysome fractions upon overexpression of Y14; perhaps this was due to a very minimal effect of Y14 in increasing p53 protein level. Nevertheless, it would be of interest to examine whether cellular signaling may control p53 mRNA translation via Y14 or the EJC. Together with the above results, we assumed that slow or delayed splicing of intron 9 renders exon i9 inclusion, and the EJC, while loaded onto the adjacent ligated exons, can prevent such aberrant splicing and also participates in the subsequent translation of p53 mRNA (Fig. 4C).
Depletion of Y14 suppresses p21 protein expression. The observation that Y14 depletion altered p53 expression (Figs 1-3) and resulted in G2/M arrest and γ H2AX foci accumulation ( Supplementary Figs 1 and 7, respectively) prompted us to examine whether Y14 also modulates the expression of p53 targets as well as cellular response to DNA damaging agents. We selected three p53 targets (the cyclin-dependent kinase inhibitor p21 Cip1/ Walf1 , E3 ubiquitin-protein ligase Mdm2, and apoptoptic factor Bax) to examine their expression in Y14-depleted HeLa cells. RT-PCR showed that Y14 depletion indeed enhanced their mRNA expression (Fig. 5A), a possible result of p53 activation. Intriguingly, immunoblotting showed that Y14 depletion reduced the level of p21, but did not significantly alter the level of Mdm2 and Bax (Fig. 5B). Nevertheless, Y14 depletion increased the level of γ H2AX and phosphorylated histone 3 (Fig. 5B), reflecting G2/M arrest and induction of γ H2AX foci.
The above observation prompted us to examine whether p53β had any suppressive effect on p21 protein expression. FLAG-tagged p53α and p53β was individually expressed or co-expressed in HeLa cells. p53β exhibited as a weaker transactivator than p53α with respect to p21 gene expression (Fig. 5C, lanes 2, 3). Co-expression of p53α and p53β , which mimicked Y14 depletion, could still promote the expression of both p21 mRNA and protein (Fig. 5C, lane 4). A similar result was observed using a p53-responsive reporter in TP53-null H1299 cells ( Supplementary Fig. 8), indicating that p53β does not compromise the activity of p53α . We also observed that Y14 knockdown reduced p21 protein level in p53-null H1299 cells (Fig. 5D), indicating that Y14 depletion caused downregulation of p21 protein independently of p53. We have previously found that Y14 depletion selectively reduces protein synthesis 4 . To assess whether Y14 particularly modulates p21 translation, we metabolically labeled newly synthesized proteins in HeLa cells with azidohomoalanine. After the click chemistry reaction, biotinylated proteins were affinity purified for immunoblotting analysis. Figure 5E shows that the level of overall p21 protein was moderately reduced in the Y14-depleted cell lysate (lane 2), whereas newly synthesized p21 was almost not detectable (lane 4). This result suggested that Y14 depletion abolished p21 protein expression likely at the translational level.

Depletion of Y14 sensitizes cells to DNA damage. p21 primarily functions as a G1/S inhibitor via sup-
pressing the activity of the CDK2 complexes. Nonetheless, p21-deficient cells exhibit delayed G2/M progression and are more vulnerable to DNA damage 22 . p21 also confers anti-apoptotic activity under various cellular conditions 23 (and references therein). The observation that Y14 depletion reduced p21 protein expression prompted us to examine the genotoxic stress response of Y14-depleted cells. Y14 depletion resulted in a higher basal level of the sub-G1 population ( Fig. 5F and Supplementary Fig. 1A), as reported 9 . We treated cells with different doses of CPT or doxorubicin or X-irradiation (IR) and subsequently measured the sub-G1 fraction as an apoptotic index. The result showed that Y14 depletion caused a greater increase in cellular sensitivity to CPT and IR as compared with mock treatment, but it had less impact on the sensitivity to doxorubicin (Fig. 5F). Finally, we examined whether Y14 depletion affects DNA double-strand break repair capacity using immunoblotting of γ H2AX. The result showed that the level of γ H2AX was constantly higher in Y14-depleted cells than in control both before and  1 and 2, respectively) or co-transfected (lane 3) in HeLa cells. RT-PCR shows p21 and GAPDH mRNAs. Immunoblotting (IB) shows p21, transiently expressed p53 proteins, and α -tubulin. (D) H1299 cells were transfected with control or Y14 siRNA. Immunoblotting shows p21, Y14 and α -tubulin. (E) HeLa cells were transfected with the control or Y14 siRNA followed by metabolic labeling with L-azidohomoalanine. After biotinylation and affinity purification (AP) using streptavidin Sepharose, bound proteins were analyzed by immunoblotting using antibody against p21 and GAPDH. Protein and RNA in 1% lysates were also analyzed by RT-PCR and immunoblotting, respectively. The level of p21 protein in Y14depleted cells relative to that of control cells was indicated; the averages and standard deviations were obtained from three experiments. (F) Line graphs show sensitivity of control (open circle) or Y14-depleted (closed circle) cells to different doses of camptothecin, doxorubicin, or X-rays. The values represent the mean and standard deviation obtained from at least three independent experiments; p-values: * < 0.05, ** < 0.01, *** < 0.001. (G) Control or Y14-depleted cells were not treated (NT, lanes 1, 2) or irradiated with 2 Gy of X-rays, and cells were collected at the indicated time points post-irradiation (lanes 3-10). Immunoblotting was performed to detect phosphorylated H2AX (γ H2AX) and α -tubulin.
after IR treatment (Fig. 5F), suggesting that Y14 may impair DNA repair capacity. Moreover, the possibility that depletion of Y14 or other EJC factors induces co-transcriptional R-loops and subsequent DNA damage 24 remains to be investigated.

Discussion
We report for the first time that the EJC factors regulate p53β isoform expression in human cells. p53β is detected in most albeit not all human tissues 13 but is absent in mouse cells due to the lack of exon i9 ( Supplementary Fig. 3). The detailed mechanism for regulation of p53β expression is not fully understood. We postulated that differential strengths of the authentic and cryptic (+ 329) 5′ SSs of intron 9 account for exon i9 selection. The weak authentic 5′ SS site may not compete efficiently with the + 329 5′ SS so that exon i9 is activated (Fig. 4C). Moreover, the strong 5′ SS of intron 10 may render its faster splicing than intron 9, leading to deposit of an EJC onto exon 10, which thus prevents exon i9 inclusion or facilitate authentic 5′ SS utilization. Depletion of SRSF3 also activates the expression of p53β 14 . SRSF3 suppresses exon i9 inclusion via binding to its consensus binding elements in exon i9. Caffeine induces p53β by downregulating SRSF3 15 . It would be of interest to know whether the splicing activity of the EJC is modulated by any cellular signaling pathways. Moreover, depletion of the EJC factors impacts different types of alternative splicing 6,7 , and thus how the EJC generally selects the targets remains to be investigated.
Depletion of Y14 increased overall p53 protein levels in various human cells, albeit to different extents, likely by preventing p53 ubiquitination ( Fig. 1 and Supplementary Fig. 2). p53 stabilization was not observed in eIF4A3-depleted cells (Fig. 2), suggesting a specific role for Y14 in modulating p53 protein stability, but the underlying mechanism by which Y14 functions in protein stability control remains future investigation. It is notable that SRSF1 stabilizes p53 by abolishing Mdm2-mediated ubiquitination and degradation via its interactions with ribosomal proteins 25 . Thus, Y14 and SRSF1 likely exert opposite roles in modulating p53 stability. Y14 depletion also increased p53 protein level in mouse cells (Supplementary Fig. 3). In fact, Mao et al. first reported that Rbm8a haploinsufficiency impairs mouse cortical development and causes microcephaly due to imbalance between proliferation and differentiation of neural progenitors and severe apoptosis of neurons 11 . These phenotypes indeed reflect the effects of Y14 depletion in cultured cells 8,9 . Mao et al. further reported that haploinsufficiency of individual EJC core factors up-regulated the expression of p53 target transcripts, and that p53 activation accounts for microcephaly possibly by disrupting cell cycle and inducing apoptosis 12 . It is somewhat surprisingly that p53 ablation is sufficient to rescue microcephaly of Rbm8a haploinsufficiency mutants 12 . Y14 depletion indeed activated p53 targets at least at the transcriptional level (Fig. 5A), but it also yielded additional effects, including induction of p53β , aberrant splicing of mitotic and apoptotic factors, and reduction of p21 protein synthesis in human cells (Fig. 6). Depletion of p53 failed to restore p21 protein expression or reduce the sub-G1 population in Y14-depleted cells ( Supplementary Fig. 9), reflecting diverse effects of Y14 depletion in human cells, and may also indicate species-specific effect of Y14 depletion al least on alternative splicing of p53.
All together, our results provide biochemical evidence that Y14 impacts p53 signaling in cultured human cells. The details of how Y14 modulates multiple steps of p53 expression and the impact of its absence on cellular functions remain to be investigated.

Materials and Methods
Cell culture and transfection. Human cervical cancer HeLa cells, non-small cell lung carcinoma H1299 cells and head and neck squamous carcinoma SAS cells were cultured in Dulbecco's Modified Eagle's medium (Invitrogen), breast cancer MCF7 cells in RPMI 1640 medium (Gibco), and colon cancer HCT116 cells in McCoy's 5A medium (Gibco) at 37 °C and 5% CO 2 . All the mediums contained 10% fetal bovine serum and penicillin/streptomycin/glutamine (Invitrogen). Cell lines were kind gifts of T.-C. Lee, M.-H. Yang, Z.-F. Chang (Taipei). Transfection was performed using Lipofectamine 2000 (Invitrogen). For knockdown experiments, 1 × 10 5 HeLa cells on 6-well culture plates were transfected with 100 nM siRNA (low GC control, or targeting Y14, Upf1, eIF4AIII or p53). All siRNAs were purchased from ThermoFisher Scientific, and their sequences are listed in Table S1. Plasmids. The p53 minigene reporter (p53mini) contained a 4719-bp HeLa cell DNA fragment spanning the 3′ part of intron 6 to the 5′ part of exon 11 (NC_000017.10, CH17: 7584967-7589686) of the human TP53 gene. The corresponding DNA fragment was PCR-amplified with the specific primers (Table S2) using HeLa cell genomic DNA as template, and subcloned into the pGEMT vector (Promega). The CMV promoter from pcDNA3.1 (Invitrogen) was inserted into pGEMT-p53mini, resulting the p53mini reporter. The mutant minigenes (p53mini-i9 + 1 m, -i9 + 329 m, -i10weak, -i10dead) were generated by PCR-based site-directed mutagenesis using primers in Table S2. The p53mini-MS2 vector was generated by inserting a 6 × MS2-containing DNA fragment derived from the β 6MS2 vector 26 downstream of exon 10 of a previously reported p53 minigene (a gift of Z.-M. Zheng and C. C. Harris, Bethesda) that spanned exon 7 to 10 (NC_000017.10. CH17: 7584967-7588698) 14 . The expression vectors for FLAG-tagged Y14 and eIF4A3 and MCP-Y14-HA were previously described 3,26 . The pcDNA3.1-p53 expression vector was a gift of S.-Y. Shieh (Taipei). To generate the p53β expression vector, we exploited a two-step PCR strategy to replace the C-terminal 62 a.a. coding sequence of p53α with the 10 a.a. p53β -specific sequence. The sequence of all vectors has been confirmed.

Metabolic labeling and affinity purification of newly synthesized proteins. HeLa cells were
transfected with the control or Y14 siRNA and incubated 48 hrs post-transfection. Metabolic labeling of polypeptides with L-azidohomoalanine was subsequently performed using the Click-iT Protein Reaction Buffer kit (Invitrogen) as described 4 . The reactions were incubated with streptavidin Sepharose (GE Healthcare Life Science) overnight at 4 °C followed by extensive wash with 0.5% SDS-containing phosphate-buffered saline (PBS). Purified proteins were analyzed by immunoblotting.
Immunoprecipitation-RT-PCR. HeLa cells were transfected with a FLAG-tagged protein expression vector for 48 hrs. Cell lysates were prepared and incubated with anti-FLAG M2 beads (Sigma) in NET-2 buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.05% NP-40 at 4 °C for 2 hrs. After washing with NET-2 buffer, RNA was recovered using TRIzol reagent, and subjected to RT-PCR analysis using the primers shown in Table S3. DNA damage treatment and sub-G1 analysis. HeLa cells were transfected with control or Y14 siRNA as above. Thirty-two hrs post-transfection, cells were treated with CPT or doxorubicin for 16 h or irradiated with X-rays followed by incubation for 16 h before harvest. Subsequently, cells were trypsinized, washed with PBS and fixed in 70% ethanol overnight at − 20 °C. After fixation, cells were subjected to staining in a solution containing 1 mg/ml propidium iodide, 0.1% triton X-100 and 10 mg/ml RNase A in PBS. Samples were analyzed by using FACSCanto-6color (BD Biosciences).