eIF2α phosphorylation bypasses premature senescence caused by oxidative stress and pro-oxidant antitumor therapies

Inactivation of either eIF2αP or PERK results in the induction of premature senescence

We were interested to examine the effects of impaired eIF2αP on the proliferation of primary mouse and human cells. We observed that primary mouse embryonic fibroblasts (MEFs) bearing a substitution of serine 51 to alanine (S51A) in both eIF2α alleles (herein referred to as eIF2αA/A cells) exhibited a substantially lower proliferation capacity than the isogenic primary MEFs with a wild-type eIF2α allele (i.e. eIF2α S/S cells) (Fig. 1A). The eIF2αA/A MEFs ceased proliferating in early passages and displayed an enlarged, flattened morphology characteristic of senescent cells as indicated by increased staining for senescence associated β-galactosidase (SA-βGal) (Fig.1B). Similar observations were made with human diploid IMR-90 fibroblasts engineered to express an HA-tagged form of eIF2αS51A under conditions of which endogenous eIF2α was down regulated by shRNA [herein referred to as knock-in (KI) cells; Figs. 1C, D]. We saw that IMR-90 KI cells proliferated slower and exhibited increased staining for SA-β-Gal compared to control IMR-90 cells with wild-type (WT) eIF2α (Fig. 1E). These data suggested that impaired eIF2αP inhibits proliferation and induces premature senescence in mouse and human primary fibroblasts.

To better address the anti-proliferative properties of eIF2αP-deficient cells, we focused on mouse cells and examined the levels of DNA damage, which is a hallmark of senescence. We looked at the phosphorylation of H2AX at serine 139 [herein referred to as gamma (γ) H2AX], which is an established marker of the DNA damage response. We found that eIF2αA/A MEFs displayed a substantially increased level of γ-H2AX staining compared to eIF2αS/S MEFs (Fig. 2A). We also examined the cell cycle distribution of MEFs and found that eIF2αA/A MEFs displayed a decrease in G₀/G₁ and increase in G₂/M compared to eIF2αS/S MEFs (Fig. 2B). Increased G₂/M was an unexpected result because mouse fibroblasts undergoing senescence usually arrest in G₀/G₁ due to activation of p53 and/or Rb [13]. Consistent with the down regulation of cells in G₀/G₁, we found that p53 together with its transcriptional targets p21 and Mdm2 were substantially decreased in eIF2αA/A MEFs compared to eIF2αS/S MEFs (Fig. 2C). We also observed that the hyper- and hypo-phosphorylated forms of Rb, which can be distinguished by their different migration in polyacrylamide gels, were decreased in eIF2αA/A MEFs compared to eIF2αS/S MEFs (Fig. 2C). Furthermore, we noticed that eIF2αA/A MEFs contained lower levels of the cyclin-dependent kinase 1 (Cdk1), an event that could account for the increased G₂/M of the cells (Figs. 2B,C).

PERK plays a major role in the adaptation of cells to stress via the induction of eIF2αP. Consistent with this notion, we observed that genetic loss of PERK significantly reduced the proliferative capacity of primary MEFs (i.e. PERK^−/− MEFs) compared to isogenic primary PERK^+/+ MEFs (Fig. 3A). Inhibition of PERK^−/− cell proliferation was accompanied by an induction of senescence as indicated by increased staining of cells for SA-β-Gal (Fig. 3B). Moreover, PERK^−/−MEFs displayed increased levels of DNA damage and noticeable differences in cell cycle distribution as indicated by a decrease in G₀/G₁ and increase in G₂/M compared to isogenic PERK^+/+ MEFs (Figs. 3C,D). Furthermore, Cdk1 as well as p53 together with Mdm2 and p21 were down regulated in PERK^−/− MEFs compared to PERK^+/+ MEFs (Fig. 3E). Moreover, eIF2αP was substantially decreased in PERK^−/− MEFs compared to PERK^+/+ MEFs (Fig. 3E). The significant similarities between PERK^−/− and eIF2αA/A MEFs as well as the substantial loss of eIF2αP in PERK^−/− MEFs supported the interpretation that PERK plays a major role in the inhibition of premature senescence mediated by eIF2αP.

Induction of premature senescence by impaired eIF2αP is caused by increased intracellular ROS levels

Cells deficient in eIF2αP display increased rates of general protein synthesis, which are associated with ER overload and accumulation of improperly folded proteins [14]. Protein misfolding in the ER can lead to increased ROS production [15], a process that is significantly facilitated by defects in either eIF2αP or PERK [4,5,14]. Consistent with these findings, we observed that eIF2αA/A or PERK^−/− MEFs contained increased ROS levels compared to their isogenic control MEFs (Suppl. Fig. 1). Since ROS significantly contributes to DNA damage and induction of senescence [16], we hypothesized that ROS was a cause of senescence in MEFs deficient in either PERK or eIF2αP. We addressed this matter in primary eIF2αA/A MEFs by assessing their proliferative capacity in the absence or presence of Trolox, a derivative of vitamin E with anti-oxidant function [17]. We observed that Trolox substantially improved the proliferation rates of eIF2αA/A MEFs and restored them to levels that were nearly identical to eIF2αS/S MEFs (Fig. 4A). Also, treatment with Trolox significantly decreased the DNA damage response in eIF2αA/A MEFs compared to eIF2αS/S MEFs confirming its anti-oxidant function and indicating that increased ROS production contributes to increased DNA damage (Fig. 4B).

Immortalized and tumor cells deficient in eIF2αP are tolerant to increased levels of intracellular ROS

Immortalized and tumorigenic cell lines are generally thought to be refractory to the anti-proliferative effects of intracellular ROS [16]. As such, we wished to examine whether impaired eIF2αP had an effect on intrinsic ROS levels and proliferation of immortalized cells. We employed primary MEFs from mice engineered to express a floxed/floxed WT eIF2α transgene in eIF2αA/A background (herein referred to as fTg/eIF2αA/A cells) [4]. Expression of Cre recombinase in fTg/eIF2αA/A MEFs causes deletion of eIF2α WT transgene leading to expression of the non-phosphorylatable eIF2αS51 under the control of the endogenous eIF2α gene promoter [4]. Primary fTg/eIF2αA/A MEFs were first immortalized by the expression of SV40 large T antigen followed by either the constitutive expression of Cre (Fig. 5A) or the expression of a Cre chimera protein with the hormone binding domain of the estrogen receptor (Cre-ERT2) (Fig. 5B). We noticed that downregulation of eIF2αP by both approaches resulted in increased ROS production, which was proportional to the degree of impaired eIF2αP (Figs. 5C, D). Also, increased ROS production was associated with an inhibition of cell proliferation compared to control cells (Figs. 5E,F). However, the differences in proliferation between eIF2αP-proficient and deficient MEFs were maintained after Trolox treatment indicating that the diminished proliferative capacity of eIF2αP-deficient cells was not caused by the upregulation of intracellular ROS (Figs. 5E,F).

We made similar observations in tumor cell lines, that is, in human fibrosarcoma HT1080 cells and lung adeno-carcinoma A549 cells, both of which were engineered to be deficient in eIF2αP (i.e. KI cells) (Fig. 6A; Suppl. Fig. 2A). Specifically, we observed that KI tumor cells displayed reduced proliferation and increased ROS production compared to WT cells (Figs. 6B,C; Suppl Fig. 2B). However, treatment with Trolox increased the proliferation capacity of both WT and KI cells but did not restore the proliferation of KI cells to the same levels as of WT tumor cells (Fig.6D; Suppl. Fig. 2C). This finding suggested that eIF2αP-deficient tumor cells were equally sensitive to the anti-proliferative effects of intracellular ROS as eIF2αP-proficient tumor cells.

Tumor cells deficient in eIF2αP are increasingly susceptible to ROS-mediated induction of senescence in response to doxorubicin

Tumor cells can tolerate ROS up to a certain level above which they become sensitive to the anti-proliferative effects of ROS [18]. This has been considered an important property of tumor cells, which can be exploited in anti-tumor therapies with pro-oxidant drugs [18]. Thus, we were interested to examine whether eIF2αP could determine the sensitivity of tumor cells to excessive oxidative stress. Doxorubicin is a widely prescribed anti-tumor drug that induces ROS synthesis and acts as an inducer of senescence when applied in sub-lethal amounts [19]. We observed that treatment of HT1080 with 30 nM doxorubicin did not cause cell death but rather resulted in deregulation of cell cycle as indicated by an increased accumulation of tumor cells in G₂/M, which was more noticeable for KI than WT tumor cells (Suppl Fig. 3). We also observed that treatment with the sub-lethal concentration of doxorubicin resulted in the induction of SA-β-Gal staining in HT1080 cells, which was more evident for KI than in WT tumor cells (Fig. 7A). We further noticed that doxorubicin treatment increased eIF2αP in WT tumor cells but not in KI tumor cells (Fig. 7B) and resulted in a higher amount of ROS synthesis in KI than in WT tumor cells (Fig. 7C; compare it with Fig. 6C). This data indicated that induction of senescence by doxorubicin was enhanced by the loss of eIF2αP and coincided with increased ROS production in eIF2αP-deficient cells compared to eIF2αP-proficient cells. To address the role of ROS in the induction of senescence, HT1080 and A549 cells were subjected to doxorubicin treatment in the absence or presence of Trolox. We found that incubation with the anti-oxidant decreased the background levels of SA-β-gal in both tumor cell types and resulted in a substantial reduction of SA-β-Gal staining in doxorubicin treated tumor cells (Fig. 8). Interestingly, Trolox treatment caused a higher reduction in SA-β-Gal staining in doxorubicin-treated KI tumor cells than in WT tumor cells indicating that increased ROS synthesis in KI cells contributes to a higher induction of senescence than in WT tumor cells (Fig. 8). Collectively, these data suggested that ROS production contributes to doxorubicin-mediated senescence of tumor cells, a process that is substantially enhanced by the loss of eIF2αP.

To better assess the biological significance of the findings, we examined the effects of doxorubicin on tumor growth in xenograft tumor assays in nude mice. We observed that doxorubicin treatment resulted in a substantial reduction of growth of HT1080 KI cells compared to control WT cells during the observation period of 40 days (Figs. 9A, B). When the tumor masses were measured at the end point of the experiment, it became evident that while tumor growth between WT and KI cells did not significantly differ, treatment with doxorubicin resulted in a significant reduction of KI tumor mass compared to WT tumor mass (Fig. 9C). When tumors were cut into pieces of equal size and subjected to SA-β Gal staining, we noticed that doxorubicin treated KI tumors displayed higher levels of senescence than WT tumors (Fig. 9D). Increased levels of senescence in doxorubicin treated KI tumors was also evident after staining of tumor sections for SA-β Gal (Fig. 9E) whereas staining for eIF2αP verified that the KI state was maintained during tumor growth in nude mice (Fig. 9F). These data indicated that loss of eIF2αP promotes the anti-tumor effects of doxorubicin in vivo via the induction of senescence.

Cell culture and treatments

Primary mouse embryonic fibroblasts (MEFs) were derived from PERK^+/− or eIF2αS/A mice as described [4,14,42]. MEFs were maintained in Dulbecco modified Eagle medium (DMEM; Wisent) supplemented with 10% fetal bovine serum (FBS; Gibco), 1x of essential and non-essential amino acids (Gibco) and antibiotics (100 U/ml of penicillin-streptomycin; Gibco). IMR-90, HT1080 and A549 cells were cultured in DMEM supplemented with 10% FBS and antibiotics. Immortalization of primary MEFs was performed by infection of pBABE-retroviruses bearing SV40 large T antigen (Addgene) and selection in 2.5 μg/ml puromycin (Sigma). Immortalized fTg/AA MEFs were transduced with either the pLKO.1-Cre lentiviral vector (Lv-Cre; Addgene) or pBabe-CreERT2-Hygro followed by selection in 200 μg/ml hygromycin (Sigma).

Generation of human cells deficient in eIF2αP

Cells were infected with pMSCV retroviruses containing an HA-tagged form of human eIF2αS51A cDNA and the gene encoding for the green fluorescence protein (GFP) under the control of an internal ribosome entry site (IRES). After infection, GFP-positive cells were sorted out by flow cytometry and re-infected with either pGIPZ insert-less lentiviruses (control) or pGIPZ lentiviruses expressing an shRNA specific for the 3′ UTR of human eIF2α mRNA (Open Biosystems). Polyclonal populations were established by selection of cells in 2.5 μg/ml puromycin (Sigma).

Senescence-associated β-galactosidase assay

Staining of cells and tumors for SA-β-Gal was performed as previously described [43,44].

Cell cycle analysis

Cells were subjected to propidium iodide (PI) staining and FACScan analysis based on a previously described protocol [45]. FACS was performed with BD FACScalibur and the data was analyzed using the FlowJo software (Tree Star inc).

ROS measurement and γ-H2AX staining

Endogenous ROS levels were quantified by incubating the cells with 5 μM 2′,7′-dichlorodihydrofluorescein diacetate (Molecular Probes) or 2.5 μM Cell-RoxTM Deep Red reagent (Molecular Probes) and FACscan analysis according to manufacturer's specifications. γ-H2AX staining was performed as described [46]. Cells were stained with 0.1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) for visualization of the nucleus.

Colony formation assays

Cells (2×10⁴) were seeded and maintained in 6-well plates for 7 days. Visualization of colonies was performed by crystal violet staining as described [47]. Scanning of the plates was done with an Oxford Optronix Scanner and quantification was done by using Gel Count software.

Western Blot analysis

Protein extraction and immunoblotting were performed as described [30]. The antibodies used were: rabbit monoclonal against phosphorylated eIF2α at S51 (Novus Biologicals), mouse monoclonal eIF2α (Cell Signaling), mouse monoclonal antibody against actin (Clone C4, ICN Biomedicals Inc), rabbit polyclonal against p53 (Novacastra), mouse monoclonal against Mdm2 [48], and mouse monoclonal against Rb or p21 (BD Biosciences). All antibodies were used at a final concentration of 0.1-1 μg/ml. Following incubation with the indicated primary antibodies and washes, membranes were probed with anti-mouse or anti-rabbit IgG antibodies conjugated to horseradish peroxidise (HRP) (Mandel Scientific). Proteins were visualized with the enhanced chemilumi-nescence (ECL) reagent (Thermo Scientific) detection system according to the manufacturer's instructions. Quantification of protein bands was performed by densitometry using Scion Image from the NIH.

Xenograft tumor assays

Injection of cells in female athymic nude mice (Charles River Inc.) and tumor monitoring were performed as described [49]. Mice were treated with 4mg/kg doxorubicin delivered by intraperitoneal injections twice per week. The animal studies were performed in accordance with approved protocols and regulations by the Animal Welfare Committee of McGill University (protocol #5754).

Statistical analysis

Error bars represent standard error as indicated and significance in differences between arrays of data tested was determined using two-tailed Student T test (Microsoft Excel).