Mammalian eIF4E2-GSK3β maintains basal phosphorylation of p53 to resist senescence under hypoxia

Hypoxia modulates senescence, but their physiological link remains unclear. Here, we found that eIF4E2, a hypoxia-activated translation initiation factor, interacted with GSK3β to maintain phosphorylation of p53, thus resisting senescence under hypoxia. RNA-binding protein RBM38 interacted with eIF4E to inhibit the translation of p53, but GSK3β-mediated Ser195 phosphorylation disrupted the RBM38-eIF4E interaction. Through investigation of RBM38 phosphorylation, we found that the eIF4E2-GSK3β pathway specifically regulated proline-directed serine/threonine phosphorylation (S/T-P). Importantly, peptides e2-I or G3-I that blocking eIF4E2-GSK3β interaction can inhibit the basal S/T-P phosphorylation of p53 at multiple sites, therby inducing senescence through transcriptional inhibition. Additionally, a nanobody was screened via the domain where eIF4E2 bound to GSK3β, and this nanobody inhibited S/T-P phosphorylation to promote senescence. Furthermore, hypoxia inhibited eIF4E2-GSK3β pathway by mediating S-Nitrosylation of GSK3β. Blocking eIF4E2-GSK3β interaction promoted liver senescence under hypoxia, thus leading to liver fibrosis, eventually accelerating N, N-diethylnitrosamine (DEN)-induced tumorigenesis. Interestingly, eIF4E2 isoforms with GSK3β-binding motif exclusively exist in mammals, which protect zebrafish heart against hypoxia. Together, this study reveals a mammalian eIF4E2-GSK3β pathway that prevents senescence by maintaining basal S/T-P phosphorylation of p53, which underlies hypoxia adaptation of tissues.


INTRODUCTION
Hypoxia plays a pivotal role in the pathogenesis of multiple human diseases [1]. Hypoxia-inducible factors (HIFs) mediates cellular adaptation to hypoxia by transcriptionally inducing a robust set of genes [1]. Adaptive protein synthesis is an alternative mechanism underlying hypoxia adaptation [2]. Hypoxia represses eukaryotic translation initiation factor 4E (eIF4E)-mediated translation, while its homolog eIF4E2 is activated and served as a translation initiation factor during hypoxia [3][4][5]. eIF4E2 specifically regulates translation by interacting with RNA-binding proteins [5,6].
Glycogen synthase kinase-3β (GSK3β), as a serine/threonine kinase, is highly active in resting cells and at the center of cellular signaling, which plays a key role in several physiological processes under hypoxia [14,15]. GSK3β appears to both promote and oppose senescence under different circumstances [16][17][18]. GSK3β prefers primed substrates that require pre-phosphorylation (primed motif), but it is also considered as a proline-directed protein kinase [14,19]. The mechanism by which these two kinase activities of GSK3β are distinguished is still unknown.
The RNA-binding protein RBM38 inhibits p53 translation by interacting with eIF4E that prevents eIF4E from binding to p53 mRNA [28]. GSK3β-mediated serine 195 (-Pro196) phosphorylation of RBM38 activates p53 translation by disrupting the RBM38-eIF4E interaction [29]. In this study, we found that RBM38 directly interacted with eIF4E2. The interrogation of RBM38 revealed the eIF4E2-GSK3β pathway that maintains the proline-directed phosphorylation of p53 at multiple serine/threonine sites (S/T-P), which resisting senescence.

Phosphoproteomic analysis using iTRAQ LC-MS/MS
HCT116 were treated with 5 μM e2-I or the scrambled peptide for 24 h and whole cell protein lysates were collected. The cell lysates were reduced with 10 mM DTT at 56°C for 30 min, alkylated with 50 mM iodoacetamide (IAM) at room temperature for 30 min in the dark. The proteolytic solution was subjected to phosphopeptides enrichment using an Immobilized Metal Affinity Chromatography method [30]. Then, the peptides were labeled with iTRAQ Reagent-8 plex Multiplex Kit (AB Sciex U.K. Limited) according to the manufacturer's protocol. Samples were iTRAQ labeled as following: ctrl_1,113; e2I_1, 114; ctrl_2, 115; e2I_2, 116. The enriched phosphopeptides were then dissolved in 2% acetonitrile/0.1% formic acid and analyzed using TripleTOF 5600+ mass spectrometer coupled with the Eksigent nanoLC System (AB SCIEX, USA). The original MS/MS file data were submitted to ProteinPilot Software v4.5 for data analysis. For protein identification, the Paragon algorithm which was integrated into ProteinPilot was employed against uniprot/swissprot-Homo database for database searching [31]. And proteins with a fold change larger than 1.5 and p-value < 0.05 were considered to be significantly differentially expressed. The Motif-x algorithm (http://motifx.med.harvard.edu) was used to extract motifs, and the significance threshold was set to P < 1e −6 .
RNA-sequencing, Senescence-associated β-galactosidase (SAβ-Gal) assay The methods were described in experimental procedures or supplementary materials.

Statistical analysis
All quantitative data were shown as mean ± S.D. The difference between two groups of variables was compared by the two-tailed, paired or unpaired Student's t-test. For comparisons of more than two groups, oneway analysis of variance (ANOVA) was employed for normal distributions. P-value of <0.05 was considered as significant (*P < 0.05, **P < 0.01, and ***P < 0.001). Grayscale analysis of WB and quantification of cardiac or liver fibrosis or liver and cellular senescence were performed using ImageJ. All statistical analyses were performed using Graphpad Prism 8.0 for data analysis and imaging.
GST pull-down assays showed a direct interaction between eIF4E2 and GSK3β (Supplementary Fig. 1F, G), and immunoprecipitation (IP) assays confirmed their interaction in vivo (Supplementary Fig. 1H, I). Mapping found that eIF4E2 mutant (Δ231-242), lacking amino acid from 231 to 242, did not bind to GSK3β (Fig. 1C). Coincidentally, this absent sequence (231-242 of eIF4E2) was highly homologous to a known GSK3β-binding motif of FRAT (Fig. 1D) [33]. Corresponding to this sequence, peptide e2-I fused with cell-penetrating peptide was synthesized with scrambled e2-S as a control. As expected, e2-I inhibited the eIF4E2-GSK3β interaction in GST pull-down assay ( Supplementary  Fig. 1J). Importantly, e2-I inhibited the phosphorylation of RBM38-Ser195 in a dose-dependent manner in various cell lines, but did not affect the protein expression of RBM38 (Fig. 1E). As a control, e2-S had effect on neither ( Supplementary Fig. 1K). of note, the Ser195 was a proline-directed serine site.
The eIF4E2 in human has seven isoforms with different termini, of which isoform A, E, and G have the GSK3β-binding motif (designated as eIF4E2-withGβ) ( Supplementary Fig. 1P). Knockout cell line (eIF4E2-KO HCT116) was generated by CRISPR/Cas9 technology specially targeting the GSK3β-binding motif (Supplementary Fig. 1Q). Importantly, the level of phosphorylation of Fig. 1 eIF4E2 regulates GSK3β proline-directed kinase activity. A, B Depletion of eIF4E2 inhibits the phosphorylation of RBM38-Ser195. eIF4E2 siRNA#1 was transfected into HCT116 cells (A) or siRNA#2 was transfected into A549 or MCF7 cells (B) for 72 h, followed by WB with indicated antibodies. Antibody p-RBM38#1 or p-RBM38#2 was used to detect p-RBM38(Ser195) as indicated. C Identifying the GSK3β binding motif of eIF4E2 by GST pull-down assay. GSK3β directly interacts with eIF4E2, but not with eIF4E2 mutant lacking amino acid from 231 to 242 (Δ231-242). D The GSK3β binding motif of eIF4E2 or FRAT (Frat1 and Frat2) were aligned. E e2-I inhibits the phosphorylation of RBM38-Ser195. Cells were treated with different concentrations of e2-I (2, 5, 10 μM) or scrambled e2-S (10 μM) for 24 h, followed by WB. F CABS-dock showed the binding of e2-I to GSK3β. The crystallographic structure of GSK3β (PDB ID: 1H8F) was used and the interaction interface was highlighted in red. G The contact diagram of GSK3β with e2-I, proposed by CABS-dock web server. H G3-I inhibits the phosphorylation of RBM38-Ser195. I The most over-represented motif is proline-directed serine/threonine regulated by eIF4E2-GSK3β pathway. Phosphorylation motifs were extracted by using Motif-x algorithm and threshold for significance was set to P < 0.000001. J, K e2-I inhibits the proline-directed phosphorylation, including Tau-Ser396 (J) and HIF1α-Ser589 (K) but does not affect the phosphorylation of Creb-Ser129 that is a serine site within priming motif. L Depletion of eIF4E2 inhibits the Tau-Ser396 and HIF1α-Ser589. eIF4E2 siRNA#2 was transfected into A549 cells for 72 h, followed by WB with indicated antibodies. M Knockout of eIF4E2 isoforms with GSK3β binding motif (eIF4E2-withGβ) inhibits the RBM38-Ser195 and HIF1α-Ser589. Lysis of eIF4E2-KO HCT116 (KO) and isogenic wild-type HCT116 (WT) cells, followed by WB with indicated antibodies. Fig. 2 eIF4E2-GSK3β maintains p53 phosphorylation at multiple S/T-P sites. A HCT116 cells were treated with e2-I for 24 h followed by treatment with cycloheximide (CHX) for the indicated times. The intensity of p53 expression for each time point was quantified by grayscale analysis and plotted against time. B e2-I promotes nuclear localization of p53. Cells were treated with 5 μM e2-I or scrambled e2-S for 24 h, then the nuclear fraction was separated from the cytoplasmic fraction, followed by WB with antibodies against p53, Tubulin (markers for cytoplasmic), and Histone (markers for nuclear). C e2-I inhibits the CPT-induced phosphorylation of p53 at multi-S-P. Cells were treated with peptide as in (B), along with mock-treated or treated with 200 nM CPT for 24 h, followed by WB. D e2-I inhibits the nocodazole-induced phosphorylation of p53-Thr81. The experiment was done as described in (C), except that cell were mock-treated or treated with nocodazole (50 ng/ml). E e2-I inhibits multi-S/T-P phosphorylation of p53 at basal conditions. F Depletion of eIF4E2 inhibits the phosphorylation of p53-Ser33 and p53-Ser315. eIF4E2 siRNA#2 was transfected into HCT116 cells for 72 h, followed by WB with indicated antibodies. G Knockout of eIF4E2 inhibits the phosphorylation of p53-Ser33 and p53-Ser315. Lysis of eIF4E2-KO HCT116 (KO) and isogenic wild-type HCT116 (WT) cells, followed by WB with indicated antibodies. H Mutant p53-6A is preferentially located in the nuclear. Vectors expressing FLAG tagged p53, p53-6A or p53-6D, were transfected into p53-null HCT116 cells for 48 h. The nuclear and cytoplasmic fractions were separated, followed by WB. I Green fluorescent signals showed that mutant p53-6A is preferentially located in nuclear. Vectors expressing GFP-fused p53, p53-6A or p53-6D, were transfected into p53-null HCT116 cells for 24 h. GFP fluorescence and DAPI staining were observed using confocal microscopy. Scale bars, 5 μm.
S/T-P dephosphorylation might suppress cytoplasmic degradation of p53 by regulating its cellular distribution. We generated vectors expressing p53 mutant 6 A (S/T-P mutated to A-P, A is alanine) and 6D (D is aspartic acid), for mimicking unphosphorylated and phosphorylated p53 respectively. We found that p53-6A was preferentially expressed in nucleus, whereas p53-6D in cytoplasm (Fig. 2H). GFP fluorescence fused with p53 showed similar cellular localization of 6 A and 6D (Fig. 2I). Overall, our results suggested that eIF4E2-GSK3β pathway maintained the multiple S/T-P phosphorylation of p53, which regulates the cellular localization of p53, thus modulating its stability.
To further validated the transcription repression activity of the inhibition of eIF4E2-GSK3β pathway, more targets were checked according to the transcriptome results. The RT-PCR results showed that peptide e2-I treatment (24-h), as well as overexpression of p53-6A (24-h) inhibited the mRNA expression of HNRNPD, EEF2, GAMT, NEK9 ( Supplementary Fig. 4A, B). The examination of more known transcriptional targets of p53 revealed that e2-I treatment (24-h) or overexpression of p53-6A (24-h) inhibited the mRNA expressions of BCL2 and BAD, both of which have effect on apoptosis with opposite effect [47] (Supplementary Fig. 4A, B). Importantly, Western blotting confirmed that 24-h e2-I treatment inhibited TOPBP1, TRX1, BCL2 or BAD in p53-dependent manner ( Supplementary Fig. 4C), but not p21 and MDM2. These results suggested that the activation of p21 upon 96-h e2-I treatment is an indirect and late event responsible for the development of senescence. Consistently, we found that e2-I treatment (24-h) and expression of p53-6A (24-h) had no effect on either VEGF or GSN ( Supplementary Fig. 4A, B), two known genes regulated by senescence [48,49]. These data suggested that the transcriptional repression resulted from inhibition of eIF4E2-GSK3β pathway is a general and direct effect, which is a cause rather than a consequence of senescence.
Single-domain antibodies (nanobodies) targeting the interaction interface can be used to block protein-protein binding [51]. We isolated nanobodies to disrupt the eIF4E2-GSK3β interaction by screening a yeast surface-displayed library of synthetic nanobody sequences. GST pull-down assay demonstrated that . Followed by WB with indicated antibodies (lower panel). Scale bars, 25 μm. B Heat map showed the differential expression genes associated with senescence in HCT116 and p53-null HCT116 cells, respectively. Differential expression genes (DEGs) were identified upon e2-I treatment by RNA-Seq. C, D e2-I inhibits mRNA expression of TOPBP1 or TRX1 depending on p53. HCT116 or HCT116-p53 null cells were treated with 5 μM e2-I or scrambled e2-S for 24 h. Total RNA was extracted and real-time quantification reverse transcriptase polymerase chain reaction (QRT-PCR) was performed. E Mutant p53-6A suppresses the expression of TOPBP1. Vectors expressing FLAG tagged p53, p53-6A or p53-6D, were mock-transfected or transfected into p53-null HCT116 cells for 48 h. Then, the nuclear fraction was separated from the cytoplasmic fraction, followed by WB. F SA-β-Gal staining of p53-null HCT116 cells expressing FLAG tagged (p53-WT, p53-6A, p53-6D, and mock) for 96 h (left panel), and quantitative analysis of the percentages of SA-β-Gal positive cells (n = 3, right panel). Scale bars, 25 μm. G Expression of nanobody Nb-28A1 inhibits the phosphorylation of RBM38-Ser195 and p53-Ser315. Vector pcDNA3-HA-Nb-28A1 or control vector pcDNA3-HA-Nb-BV025 was transfected into HCT116 cells for 48 h, followed by WB with indicated antibodies. H Nanobody Nb-28A1 induced cell senescence depending on p53. SA-β-Gal staining of HCT116 or p53-null HCT116 cells expressing HA-Nb-28A1 or HA-Nb-BV025 for 96 h (left panel), and quantitative analysis percentages of SA-β-Gal positive cells (n = 3, right panel). Scale bars, 25 μm. I Expression of nanobody HA-Nb-28A1 for 24 h downregulates the expression of TOPBP1, TRX1, BCL2 or BAD depending on p53. Vector pcDNA3-HA-Nb-28A1 or control vector pcDNA3-HA-Nb-BV025 was transfected into HCT116 or p53-null HCT116 cells for 24 h, followed by WB with indicated antibodies.

Hypoxia inhibits the eIF4E2-GSK3β pathway
Considering that eIF4E2 was activated under hypoxia, we inferred that eIF4E2-GSK3β pathway might play a role under hypoxia. We observed that hypoxia significantly inhibited phosphorylation of RBM38-Ser195 in different cell lines (Fig. 4A, Supplementary Fig.  6A). eIF4E2 was detected in the anti-GSK3β immune complex under normal conditions, whereas less eIF4E2 was detected under hypoxia (Fig. 4B). The phase separation-based protein interaction reporter (SSPIER) analysis indicated that eIF4E2-GSK3β interaction led to the phase separation, thus prompting the formation of EGFP droplets, but 2-h hypoxia exposure demolished EGFP droplets ( Supplementary Fig. 6B) [53,54].
The S-Nitrosylation at cysteine can inhibit proline-directed kinase activity of GSK3β [19]. Interestingly, Cys317 was located in the domain where GSK3β bound to eIF4E2. L-Arginine is used as a substrate for the production of nitric oxide to induce S-Nitrosylation [55]. The SSPIER analysis revealed that L-Arginine inhibited eIF4E2-GSK3β interaction (Fig. 4C). Western blot showed that L-Arginine inhibited phosphorylation of RBM38-Ser195 (Fig.  4D), and that L-NAME treatment partially restored the phosphorylation of RBM38-Ser195 under hypoxia (Supplementary Fig. 6C). N G -nitro-L-Arginine methylester (L-NAME), as a NOS inhibitor, can inhibit S-Nitrosylation by blocking NO generation [56]. Furtherly, we found that hypoxia inhibited S/T-P phosphorylation of p53 at Fig. 4 Hypoxia inhibits the eIF4E2-GSK3β pathway. A Hypoxia inhibits the phosphorylation of RBM38-Ser195. A549 cells were exposure to hypoxia (1% O 2 ) for 18, 24 h, or normoxia for 24 h, followed by WB. B Hypoxia inhibits eIF4E2-GSK3β binding. Cells were exposure to normal or hypoxia (1% O 2 ) condition for 24 h. Then, cell extracts were subjected to Co-IP with anti-GSK3β antibody or IgG, followed by WB. C SPPIER assay showed L-Arginine inhibits eIF4E2-GSK3β interaction. Cells transiently expressed EGFP-eIF4E2-HOTag3-T2A-GSK3β-HOTag6, and then 2mM L-Arginine was added to the cells for 1 h. Scale bars, 1 μm. D L-Arginine inhibits the phosphorylation of RBM38-Ser195. Cells were mocktreated or treated with different concentrations of L-Arginine (0.5, 1, 2 mM) for 24 h and subjected to WB. E Hypoxia inhibits the phosphorylation of p53-Ser33, Ser46, Ser315 and the expression of TOPBP1. HCT116 (left) or p53-null HCT116 (right) cells were exposure to normoxia or hypoxia (1% O 2 ) condition for 18 h, followed by WB with indicated antibodies. F Expression of eIF4E2 isoform A activates the phosphorylation of Rbm38-Ser195 in eIF4E2-KO HCT116 cells. Vectors expressing eIF4E2 isoform A with HA-tag were mock-transfected or transfected into eIF4E2-KO HCT116 cells for 48 h respectively, along with treatment with 5 μM G3-I or scramble peptide as indicated for 24 h, followed by WB.
To determine the role of eIF4E2-GSK3β pathway under hypoxia, overexpressed eIF4E2 isoform A was introduced into eIF4E2-KO HCT116 cells. We found that the expression of eIF4E2 isoform A activated RBM38-Ser195 phosphorylation, which can be blocked by peptide G3-I (Fig. 4F). It takes a long period of time for hypoxia to induce senescence, during which hypoxia inhibited the growth of eIF4E2-KO HCT116, making senescence difficult to be detected. However, expression of eIF4E2 isoform A preserved the normal growth of eIF4E2-KO HCT116 under hypoxia ( Supplementary Fig.  6D, E). These results suggested that hypoxia inhibited eIF4E2-GSK3β activity by inducing S-Nitrosylation of GSK3β.

Blocking eIF4E2-GSK3β interaction promotes liver senescence under hypoxia
To explore the physiological role of the eIF4E2-GSK3β pathway, peptides were intraperitoneally injected to mice. We found that e2-I injection effectively inhibited the phosphorylation of RBM38-Ser195 in liver, compared with scrambled e2-S. We used CasRx system to validate this e2-I effect is reached through eIF4E2-GSK3β pathway [57]. After hydrodynamic injection with CasRx and sg-eIF4E2 mixture, eIF4E2 mRNA levels were decreased in GFP + hepatocytes, along with the decreased phosphorylation of RBM38-Ser195 and p53-Ser315. However, e2-I had no further effect on their phosphorylation in eIF4E2 knockdown hepatocytes (Fig. 5A).
To explored the pathological role of the eIF4E2-GSK3β pathway, chronic intermittent hypoxia (IH) condition was established [58]. We intraperitoneally injected e2-S/e2-I to mice followed by IH. Two months later, e2-I, rather than e2-S, induced senescence ( Supplementary Fig. 8A, B) and excessive senescence-associated secretory phenotype (SASP), as shown by increased secretion of proinflammatory cytokines such as IL-6, IL-8 and IL-1β (Fig. 5F). Subsequently, e2-I promoted liver fibrosis under IH (Fig. 5G). Persistent senescence accompanied by excessive SASP might promote tumorigenesis. The neonatal mice were treated with diethylnitrosamine (DEN) to initiate tumorigenesis, and then treated with peptide (e2-S/e2-I), followed by IH. Since senolytics drugs (dasatinib and quercetin, D + Q) can significantly blunt liver tumor progression with few liver lesions through elimination of senescent cells [59], we treated mice with D + Q after peptide/IH treatment (Fig. 5H, upper panel). We found that as early as 24 weeks, hepatocellular carcinoma (HCC) developed upon e2-I/IH treatment (Fig.  5H, I) or G3-I/IH treatment ( Supplementary Fig. 8C, D), but not upon scrambled peptides/IH treatment. Impressively, senolytics drug blunted tumor progression accelerated by e2-I/G3-I treatment under IH (Fig. 5H, I, Supplementary Fig. 8C, D). The possible reason might lie in that senolytics drug efficiently eliminated SA-β-Gal + cells induced by peptide/IH treatment (Fig. 5J). In contrast, e2-I/G3-I treatment did not advance HCC under normoxic condition (Supplementary Fig. 8E). These results suggested the protective roles of eIF4E2-GSK3β in tissues under physiological hypoxia.
Mammalian eIF4E2 protects heart of zebrafish According to NCBI GenBank, eIF4E2 isoforms with GSK3β-binding motif (eIF4E2-withGβ) appears only in mammals (Fig. 6A). We further confirmed that non-mammalians such as Oryzias latipes, Danio rerio, Xenopus tropicalis and Gallus gallus did not express eIF4E2-withGβ isoforms (Fig. 6B). We speculated that eIF4E2-withGβ expression might exert protective role in non-mammalian by interacting with their own GSK3β since GSK3β is highly conserved. We introduced human eIF4E2 isoform A into zebrafish Tg (cmlc2: eGFP) embryos. The expression of eIF4E2 isoform A rescued abnormal heart loop caused by hypoxia stress (Fig. 6C), and G3-I hindered this rescue effect (Fig. 6C). RT-PCR results showed that eIF4E2 expression increased the expression of TOPBP1 and TRX1 (Fig. 6D, E). In addition, expression of zebrafish p53-4A (S/T-P to A-P) significantly inhibited their expression in p53-null HCT116 (Fig. 6F, G).
Recently, tannic acid modification (TANNylated) was established to deliver peptide to the heart [60]. After intravenous injection with TANNylated e2-I/e2-S to mice and IH treatment, we found that TANNylated e2-I rather than e2-S induced cardiac fibrosis (Fig.  6H). TANNylated e2-I inhibited the phosphorylation of RBM38-Ser193 and p53-Ser315 in heart under physiological hypoxia, indicating the conservative function of eIF4E2-GSK3β pathway in different tissues (Fig. 6I).  5 Blocking eIF4E2-GSK3β interaction promotes liver senescence under hypoxia. A Plasmids expressing CasRx-GFP, sgRNAs targeting eIF4E2 (sg-eIF4E2) or non-targeting (NT) guide RNAs were delivered to mouse livers by hydrodynamic tail-vein injection. 96 hr later, GFP + hepatocytes were sorted for the quantification of mRNA or protein expression levels. B, C e2-I promoted liver senescence under intermittent hypoxia (IH). After intraperitoneal injection of 35 mg/kg e2-I or scrambled e2-S, mice were housed at normoxia or physiological hypoxia (12% O 2 ) conditions and on the 5th day, liver tissue blocks were stained for SA-β-Gal activity (B), then sectioned and counterstained with nuclear fast red (C, left panel), and SA-β-Gal staining results were quantified by the ratio of SA-β-Gal + area to the total image area for three fields (C, n = 3, right panel). Scale bars, 25 μm. D, E Nanobody Nb-28A1 promoted liver senescence under intermittent hypoxia (IH). AAV8-Nb-28A1 or AAV8-Nb-BV025 (1×10 11 viral genomes in 100 μl saline) were intravenous injected. Two weeks later, mice were housed at normoxia or physiological hypoxia (12% O 2 ) conditions and on the 5th day, liver tissue blocks were stained for SA-β-Gal activity (D, upper panel), equal amounts of liver tissue were subjected to WB for checking the expression of nanobody (D, lower panel). Then sectioned and counterstained with nuclear fast red (E, left panel), and SA-β-Gal staining results were quantified by the ratio of SA-β-Gal + area to the total image area for three fields (n = 3) (E, right panel). Scale bars, 25 μm. F e2-I induced excessive senescence-associated secretory phenotype (SASP). Mice (6 weeks old) were injected intraperitoneally with 35 mg/kg e2-I or scrambled e2-S twice a week for 6 weeks. After each injection, the mice were exposed to physiological hypoxia (12% O 2 ) conditions for 8 h. At 8 weeks, the levels of IL-6, IL-8 and IL-1β in serum of mice were detected by using ELISA kits. Data are presented as mean SD from each group (n = 3). G e2-I leads to liver fibrosis. The liver fibrosis of mice from experiment (F) was evaluated by Masson's trichrome staining and representative results were shown (left panel). Scale bars, 50 μm.

DISCUSSION
Senescence is a key component of normal physiology contributing to tissue homeostasis. However, senescent cells may adversely affect tissue microenvironment through senescence-associated secretory phenotype (SASP), which promote tumor progression or age-related diseases. Hypoxia plays a critical role in the pathogenesis of diseases, potentially by inducing persistent senescence and SASP [61]. However, the relationship between hypoxia and senescence in physiological context remains unclear [11]. In this study, we found that the eIF4E2-GSK3β pathway prevented cellular senescence under physiological hypoxia by maintaining proline-directed serine/threonine phosphorylation of p53 (Fig. 7). eIF4E2, as a homolog of eIF4E, is a well-known translational activator under hypoxia, which is essential for hypoxia adaptation [62]. Unexpectedly, our results indicated that eIF4E2-GSK3β interaction via a conserved motif regulated proline-directed kinase activity of GSK3β. Other proteins might regulate GSK3β by similar mechanism. A study showed that FRAT-GSK3β interaction regulates the proline-directed phosphorylation of FOXK, and FRAT shares the conserved GSK3β-binding motif with eIF4E2 [63,64]. of note, a specific nanobody against the GSK3β-binding motif of eIF4E2 efficiently inhibits eIF4E2-GSK3β pathway, informing on the design of a more specific GSK3β inhibitor.
Although p53 is a canonical inducer of cellular senescence, it can also suppress senescence [65]. p53 is pivotal for establishing senescence after its activation by stress. However, whether or how physiological p53 activity regulates senescence is not quite clear [22,65]. In current study, we found that eIF4E2-GSK3β pathway maintained the basal S/T-P phosphorylation of p53, which enabled p53 to prevent senescence under physiological condition. Multiple site S/T-P dephosphorylations confered p53 with the transcriptional repression activity such as inhibiting the expression of TOPBP1 or TRX1 (Fig. 7). Further, p53-6A expression promoted senescence. These data suggested that basal phosphorylation of p53 was required for prevention of senescence in physiological context. However, p53-6D also promoted senescence, indicating that S/T-P phosphorylation of p53 would trigger senescence in stress context. Stress-induced phosphorylation of p53 at Ser 15 and Thr18 is known to activated senescence [66]. The perturbation of Ser18 (mouse) phosphorylation resulted in p53 physiological functional defects, thus leading to senescence [67]. Thus, the relationship between p53 phosphorylation and senescence is complex, and our observations of S/T-P phosphorylation suggested that extent of p53 phosphorylation and its cellular context further determined p53 function in modulating senescence, which warrants further investigation. In addition, we found that hypoxia inhibited the eIF4E2-GSK3β pathway to suppress the S/T-P phosphorylation of p53, suggesting that hypoxia enabled p53 to function as a transrepressor, which was consistent with previous reports [68,69]. A peculiarity of GSK3β is its high activity in unstimulated resting cells. We also suggest that the eIF4E2-GSK3β pathway remained highly active under normal cellular conditions, which was necessary for the pivotal physiological roles of p53 by maintaining its basal S/T-P phosphorylation [20]. eIF4E2 isoforms that with GSK3β-binding motif only exist in mammals. eIF4E2 is highly correlated with hypoxia adaptation of mammals to Tibetan Plateau [70,71]. A 63-bp insertion between the exons responsible for distinguishing the different isoforms has been reported in eIF4E2 of Tibetan, which contributes to hypoxia adaptation [71]. Based on it, we speculated that 63-bp insertion might mediate alternative splicing associated with the abundance of different eIF4E2 isoforms, and that the eIF4E2-withGβ isoform might be more abundant in high-altitude Tibetans, which would facilitate adaptation of Tibetans to hypoxic surroundings. As a proof, we found that inhibiting eIF4E2-GSK3β pathway caused liver senescence, fibrosis and tumorigenesis under physiological hypoxia. eIF4E2-GSK3β pathway played a protective role in heart against hypoxia. In addition, the eIF4E2-GSK3β pathway could Fig. 6 Mammalian eIF4E2 protects heart of zebrafish. A Sequence alignment of c-termini of eIF4E2 from major species, including mammals, reptiles, amphibians, fish, and birds. B WB detects the protein expression of eIF4E2-withGβ isoforms in different non-mammals. C Human eIF4E2 isoform A prevents disorder of heart looping of zebrafish. Vectors expressing eIF4E2 isoform A were mock-injected or injected into embryos (n = 5) at the 1-cell stage. 24 h later, embryos were exposure to hypoxic (1% O 2 ) or normoxia for 24 h, and representative Tg (cmlc2: eGFP) heart were showed at 48 h. D, E Expression of eIF4E2 isoform A increases the mRNA expression of TOPBP1 or TRX1 in zebrafish embryo. Total RNA was extracted and reverse transcriptase polymerase chain reaction (RT-PCR) was performed (D). Real-time quantification reverse transcriptase polymerase chain reaction (QRT-PCR) was performed (E). F S/T-P sites of zebrafish p53 were indicated. G Zebrafish p53-4A mutant suppresses the expression of TOPBP1 or TRX1. Vectors expressing Zebrafish p53-4A, were mock-transfected or transfected into p53null HCT116 cells for 48 h, followed by WB. H TANNylated e2-I or e2-S was intravenously injected to mice twice a week for 6 weeks, followed by intermittent hypoxia (IH) exposure (12% O 2 for 8 h after injection). Cardiac fibrosis was evaluated by Masson's trichrome staining and representative macroscopic photographs of heart tissue was showed (n = 4) (left panel). Quantified data of the fibrosis area. Statistical significance was determined by one-way analysis of variance (n = 4) (right panel). I Mice were intravenously injected with TANNylated e2-I or e2-S, and housed at physiological hypoxia (12% O 2 ) conditions for 48 h. Then, heart tissues were analyzed by WB. Fig. 7 Model showing function of the eIF4E2-GSK3β. Left: eIF4E2-GSK3β maintains p53 phosphorylation at multiple S/T-P sites. Right: Peptides e2-I or G3-I blocking eIF4E2-GSK3β interaction can inhibit multiple S/T-P phosphorylation of p53, which subsequently transcriptional inhibiting the expression of TOPBP1 or TRX1. As a result, blocking eIF4E2-GSK3β interaction promotes liver senescence that accelerates DENinduced liver cancer under hypoxia.
regulate the S/T-P phosphorylation of Tau or HIF1α, which may underpin the effect of the eIF4E2-GSK3β pathway on hypoxia adaptation. In conclusion, we suggest that senescence modulation function of eIF4E2-GSK3β pathway is critical for hypoxia adaptation of tissues (Fig. 7).

DATA AVAILABILITY
All data needed to evaluate the conclusions of this study are present in the manuscript. Requests for additional information related to this paper should be directed to the corresponding author. The RNA-Seq data: Sequence Read Archive (SRA) PRJNA605921. The iTRAQ LC-MS/MS proteomics data: iProX IPX0002016000.