EGF receptor and PKCδ kinase activate DNA damage–induced pro-survival and pro-apoptotic signaling via biphasic activation of ERK and MSK1 kinases

DNA damage–mediated activation of extracellular signal–regulated kinase (ERK) can regulate both cell survival and cell death. We show here that ERK activation in this context is biphasic and that early and late activation events are mediated by distinct upstream signals that drive cell survival and apoptosis, respectively. We identified the nuclear kinase mitogen-sensitive kinase 1 (MSK1) as a downstream target of both early and late ERK activation. We also observed that activation of ERK→MSK1 up to 4 h after DNA damage depends on epidermal growth factor receptor (EGFR), as EGFR or mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK)/ERK inhibitors or short hairpin RNA–mediated MSK1 depletion enhanced cell death. This prosurvival response was partially mediated through enhanced DNA repair, as EGFR or MEK/ERK inhibitors delayed DNA damage resolution. In contrast, the second phase of ERK→MSK1 activation drove apoptosis and required protein kinase Cδ (PKCδ) but not EGFR. Genetic disruption of PKCδ reduced ERK activation in an in vivo irradiation model, as did short hairpin RNA–mediated depletion of PKCδ in vitro. In both models, PKCδ inhibition preferentially suppressed late activation of ERK. We have shown previously that nuclear localization of PKCδ is necessary and sufficient for apoptosis. Here we identified a nuclear PKCδ→ERK→MSK1 signaling module that regulates apoptosis. We also show that expression of nuclear PKCδ activates ERK and MSK1, that ERK activation is required for MSK1 activation, and that both ERK and MSK1 activation are required for apoptosis. Our findings suggest that location-specific activation by distinct upstream regulators may enable distinct functional outputs from common signaling pathways.

Irradiation (IR) 3 and chemotherapy are widely used for the treatment of cancer; however, adjacent healthy tissues can also be damaged, including the mucosal tissues of the gut and oral cavity. Such collateral tissue damage can result in significant comorbidities and, in some cases, limit the course of therapy (1). As IR and chemotherapies work in part by inducing apoptosis, modulation of apoptotic mediators specifically in nontumor cells could offer protection against collateral tissue damage, resulting in improved quality of life and, in some cases, better tumor eradication (2)(3)(4). PKC␦ is a ubiquitously expressed serine-threonine kinase that controls a wide variety of cell functions, including proliferation, cell survival, invasion and migration, and cell death (5). PKC␦ has been identified as an important mediator of DNA damage-induced apoptosis, and our laboratory and others have shown that inhibition of PKC␦ activation can reduce IR and chemotherapy-induced apoptosis in vitro and toxicity to nontumor tissues in vivo (6 -9). Remarkably, inhibition of PKC␦ has been shown to preserve salivary gland function in mice exposed to head and neck IR but did not impact treatment of the tumor (8). This supports previous data from our laboratory that suggests that, in contrast to normal cells, in some tumor cells PKC␦ does not regulate apoptosis but may instead have a prosurvival role (10 -12).
Studies from our laboratory indicate that nuclear PKC␦ is required for apoptosis, suggesting that PKC␦ function may be dictated in part by its subcellular localization (13)(14)(15)(16)(17)(18). Further, most studies suggest that PKC␦ does not directly regulate the apoptotic machinery but may instead integrate upstream signals to regulate cell fate decisions in response to cell distress or damage (2,17,19,20). In this regard, PKC␦ has been shown to regulate signaling through the mitogen-activated protein kinase (MAPK) pathways (extracellular signal-regulated kinases (ERK), c-Jun N-terminal kinases, and p38 MAPKs), primarily downstream of growth factor receptors (21)(22)(23) but also in response to DNA damage (24).
The MEK/ERK pathway has well-established roles in proliferation and survival and regulates cell cycle arrest and apoptosis in response to DNA damage (21,22,(25)(26)(27). In damaged cells, the duration, magnitude, and subcellular localization of ERK1/2 activation may be critical in determining whether the outcome is prosurvival or pro-apoptotic (21,25). We show that activation of ERK in response to DNA damage agents is bipha-sic, consisting of an early prosurvival phase and a later pro-apoptotic phase. These phases are mediated by distinct upstream regulators, with EGFR activating ERK in the early phase and PKC␦ activating late-phase ERK. Furthermore, we identify a unique ERK3 MSK1 signaling module regulated by nuclear PKC␦ that is essential for apoptosis. Our study shows that DNA damage induces temporally distinct prosurvival and pro-apoptotic signaling pathways and suggest that the functional output of ERK3 MSK1 activation in response to DNA damage is regulated, at least in part, by the upstream activator.

In response to DNA damage, biphasic activation of ERK drives survival and apoptosis
ParC5 rat parotid acinar cells provide a useful model to study DNA damage-induced cell death, as their response to irradiation is similar to that observed in salivary acinar cells in vivo (3). In parC5 cells treated with etoposide, ERK activation is biphasic, with an initial peak at around 2 h and a second peak at 6 -8 h (Fig. 1A). Similar results were seen when cells were treated with IR (Fig. 2B). Quantification of pERK/ERK across multiple time points from three independent experiments is shown in Fig. 1B. To determine whether one or both of peaks of ERK activity are required for apoptosis, we assayed activation of caspase-3 in cells pretreated with PD98059, which inhibits MEK, the upstream regulator of ERK. When apoptosis was assayed 2-4 h after addition of etoposide, pretreatment with PD98059 resulted in an up to 4-fold increase in apoptosis, demonstrating that early activation of ERK conveys a survival signal (Fig. 1C). However, at 8 h, PD98059 slightly suppresses apoptosis. Consistent with this, when parC5 cells were pretreated with PD98059 or a second MEK inhibitor, U0126, and then treated with etoposide for 18 h, apoptosis was significantly reduced compared with cells treated with etoposide alone (Fig. 1D). Similar results were seen with pretreatment using MEK inhibitors followed by IR (Fig. 1E). These data demonstrate that ERK activation in response to DNA damage can drive both survival and apoptosis and that these distinct functional outputs may be dictated by the kinetics of activation.

Activation of EGFR in response to DNA damage promotes cell survival
Activation of EGFR occurs rapidly but transiently in response to DNA damage, with kinetics similar to the early phase of ERK activation ( Fig. 2A, left). Early activation of ERK requires EGFR activity, as it is abolished in cells pretreated with gefitinib ( Fig. 2A, right). Notably, EGFR inhibition only blocked early activation of ERK (up to 4 h), with no effect on late activation of ERK (Fig. 2, A and B). However, both EGFR inhibitors increased apoptosis up to 8 h after addition of etoposide (Fig.  2C), similar to that seen with the MEK inhibitor PD98059 (Fig. 1C).

DNA damage-induced pro-survival and pro-apoptotic signaling
Previous studies have suggested a role for EGFR and ERK in regulation of DNA double-stranded break repair, which could in part explain the prosurvival phenotype we observed (29 -33). To test this, we assayed DNA damage using a comet assay in cells treated with IR and pretreated with gefitinib or PD98059. Although untreated cells and inhibitor-treated cells had a similar amount of DNA damage at 15 min, DNA damage was increased in inhibitor-treated cells 30 min and 1 h after IR compared with the control, indicating a delay in DNA repair with inhibition of ERK or EGFR (Fig. 2D). Our study shows that prosurvival signaling through EGFR3 ERK correlates with enhanced DNA repair in the first few hours after DNA damage. This could occur via direct regulation of DNA repair by EGFR3 ERK or through an indirect mechanism, such as prolonged activation of cell cycle checkpoints to allow repair.

PKC␦ activation of ERK in response to DNA damage drives apoptosis
PKC␦ is a pro-apoptotic kinase that also regulates ERK signaling in some biological contexts (5). We investigated the kinetics and magnitude of ERK activation in parC5 cells depleted of PKC␦ using shRNA (Fig. 3A) and in salivary gland tissue lysates from PKC␦ ϩ/ϩ and PKC␦ Ϫ/Ϫ mice that were treated with IR to the head and neck (Fig. 3B). In PKC␦ ϩ/ϩ mice, the pattern of ERK activation by IR was similar to that observed in cells treated with etoposide, with two phases of ERK activation at 2 and 6 h (Fig. 3B). In PKC␦ Ϫ/Ϫ mice, both phases of ERK activation were reduced; quantitation of pERK/ ERK by densitometry shows that the second peak (6 h) was more dramatically reduced than the first peak (2 h) (45% versus 80% reduction, respectively) (Fig. 3B). This suggests that depletion of PKC␦ primarily inhibits the "pro-apoptotic" phase of Gefitinib + Etoposide Etoposide Figure 2. EGFR activation of ERK promotes cell survival. A, ParC5 cells were pretreated with the DMSO control or 600 nM gefitinib prior to addition of 50 M etoposide for the indicated times (hours). Cells were harvested and immunoblotted for the indicated proteins. B, ParC5 cells were pretreated with the DMSO control or 300 nM gefitinib prior to treatment with 10 Gy of IR. Cells were harvested at the indicated times (hours) post-IR and immunoblotted for the indicated proteins. C, ParC5 cells were pretreated with the DMSO control or 300 nM gefitinib or afatinib prior to addition of 50 M etoposide. Cells were harvested for measurement of caspase-3 activity at the indicated times (hours). D, ParC5 cells were pretreated with DMSO, 300 nM gefitinib, or 20 M PD98059 prior to irradiation with 10 Gy of IR. Following IR, cells were harvested at various time points (hours) and prepared using a neutral comet assay according to the Trevigen comet assay protocol. Data shown are representative experiments. Statistics represent two-way ANOVA for comparison of time point or treatment with the corresponding control. **, p Ͻ 0.001; *, p Ͻ 0.05.

DNA damage-induced pro-survival and pro-apoptotic signaling
ERK activation. To confirm this, we assayed ERK activation in parC5 cells depleted of PKC␦ with shRNA (␦561) or a scrambled shRNA (SCR). Similar to PKC␦ Ϫ/Ϫ mice, depletion of PKC␦ resulted in a small decrease in the early phase of ERK activation by etoposide at 2 h and a more dramatic effect on later activation of ERK at 6 h (Fig. 3A). Our data suggest that PKC␦ primarily regulates the second phase of ERK activation to drive apoptosis, independent of prosurvival signaling by ERK. Depletion of PKC␦ also slightly decreased activation of EGFR in cells treated with etoposide ( Fig. 3A), consistent with partial inhibition of the early phase of ERK activation (Fig. 3, A and B). To directly ask whether PKC␦ contributes to cell survival in response to DNA damage, we analyzed caspase-3 activation in parC5 cells depleted of PKC␦. Depletion of PKC␦ using three independent shRNAs suppressed apoptosis at all time points, indicating that PKC␦ drives apoptosis and does not play a role in the early prosurvival effect through EGFR3 ERK (Fig. 3C). Our study demonstrates that, in response to DNA damage, PKC␦ activation of ERK appears to solely drive cell death signaling.
Nuclear translocation of PKC␦ is required for DNA damageinduced apoptosis, and direct targeting of PKC␦ to the nucleus induces apoptosis (13,16). Therefore, we investigated the hypothesis that the pro-apoptotic phase of ERK activation is mediated through nuclear PKC␦. ParC5 cells were transduced withAd-GFP,Ad-GFP-PKC␦WT,orAd-GFP-PKC␦NLS,aconstruct where we added an SV40 NLS to PKC␦, resulting in direct targeting of PKC␦ to the nucleus (15,16). Remarkably, expression of either Ad-GFP-PKC␦WT or Ad-GFP-PKC␦NLS resulted in increased ERK activation (Fig. 3D). PKC␦CF is a caspase-cleaved form of PKC␦ generated in response to DNA damage and is a potent inducer of apoptosis (28). PKC␦CF is constitutively nuclear because of loss of constraints imposed by the regulatory domain of the kinase (16). As shown in Fig. 3E, expression of pEGFPN1-PKC␦CF robustly induces pERK, further supporting our conclusion that nuclear localization of PKC␦ is sufficient to activate ERK. To determine whether ERK activation is required for apoptosis induced by nuclearly targeted PKC␦, parC5 cells transduced with Ad-GFP-PKC␦NLS were treated with the MEK inhibitors PD98059 and U0126.

DNA damage-induced pro-survival and pro-apoptotic signaling
Expression of Ad-GFP-PKC␦NLS induced caspase-3 activation, as shown previously, and this could be suppressed by both MEK inhibitors, indicating a requirement for ERK activation downstream of PKC␦ for induction of apoptosis (Fig. 3F).

ERK activates MSK1 in response to DNA damage
Our study shows that nuclear PKC␦ induces apoptosis via activation of ERK; however, downstream targets of ERK that drive apoptosis are largely unknown. MSK1/2 kinases are activated downstream of ERK and p38 and predominantly localize to the nucleus, where they phosphorylate transcription factors and regulate chromatin remodeling (29 -31). In response to DNA damage, MSK1 is activated in a biphasic manner with kinetics similar to what we have shown previously for ERK (Figs. 4A and 1A). Activation of MSK1 in ␦561 and ␦844 cells is suppressed compared with SCR cells (Fig. 4A). Similar to ERK (Fig. 3, A and B), loss of PKC␦ suppresses late MSK1 activation (6 to 8 h) to a greater extent than early activation (2 to 4 h) (Fig.  4A), whereas pretreatment with gefitinib only suppresses early activation of MSK1 (Fig. 4B).
Our observation that MSK1 and ERK activation are co-regulated in response to DNA damage suggests that activation of MSK1 may contribute to prosurvival signaling through EGFR3 ERK and pro-apoptotic signaling through PKC␦3 ERK. Indeed, depletion of MSK1 using three unique shRNAs results in increased apoptosis up to 8 h after addition of etoposide, indicating that, like ERK, early activation of MSK1 is prosurvival (Fig. 4C). In contrast, depletion of MSK1 dramatically decreases cell death 24 -48 h after IR (Fig. 4D), analogous to what we described previously when MEK/ERK was inhibited (Fig. 1E). Our data support a model whereby, in response to DNA damage, EGFR activation of ERK and MSK1 promotes cell survival in the first few hours after DNA damage, whereas late activation of MSK1 downstream of PKC␦3 ERK may be required for apoptosis.

PKC␦ drives DNA damage-induced apoptosis via ERK3 MSK1
As targeting PKC␦ to the nucleus activates ERK (Fig. 3D), we asked whether nuclearly targeted PKC␦ also activates MSK1 as evidenced by phosphorylation on Ser-376. As seen in Fig. 5A, transduction of parC5 cells with Ad-GFP-PKC␦NLS results in activation of MSK1. MSK1 activation by Ad-GFP-PKC␦NLS could be suppressed by pretreatment with the MEK/ERK inhibitor PD98059 but not the p38MAPK inhibitor SB20358 (Fig.   5B). Unexpectantly, inhibition of p38MAPK with SB20358 increased pERK in both Ad-GFP and Ad-GFP-PKC␦NLS transduced cells; however, pMSK1 was only increased in cells that were also transduced with Ad-PKC␦NLS (Fig. 5B). Notably, p38MAPK inhibition has no effect on etoposide induced apoptosis (Fig. S1), indicating that this pathway does not contribute to apoptotic signaling through PKC␦.

DNA damage-induced pro-survival and pro-apoptotic signaling
To verify a pro-apoptotic function for MSK1 downstream of nuclear PKC␦, we analyzed the ability of Ad-GFP-PKC␦NLS to induce apoptosis in parC5 cells in which MSK1 was depleted with shRNA. Apoptosis driven by expression of Ad-GFP-PKC␦NLS was suppressed in cells depleted of MSK1, and the degree of suppression was correlated with the amount of MSK1 depletion (Figs. 5C and 4C). Finally, we asked whether MSK1 is required for apoptosis induced by PKC␦CF. Expression of pEG-FPN1-PKC␦CF potently induced apoptosis in control parC5 cells, but apoptosis was severely reduced in cells depleted of MSK1 (Fig. 5D).
Our study defines a common ERK3 MSK1 pathway that, when activated by EGFR in response to DNA damage, drives cell survival but when activated by nuclear PKC␦ drives apoptosis (Fig. 6). Moreover, we define a nuclear specific function for PKC␦ in driving apoptosis through ERK3 MSK1. Activation of ERK3 MSK1 for survival versus apoptosis is temporally distinct, which may reflect access to specific upstream regulators.

Discussion
We show that, in the context of DNA damage, activation of ERK can have a dramatically different biological outcome depending on when, how, and where in the cell it is activated. We identify temporally distinct prosurvival and pro-apoptotic signaling pathways that are both mediated through ERK3 MSK1. Prosurvival signaling through EGFR3 ERK enhances DNA repair in the first few hours after DNA damage through a PKC␦-independent pathway. In contrast, pro-apoptotic signaling through ERK occurs 6 -8 h after DNA damage and is independent of EGFR but dependent on PKC␦. Furthermore, our study identifies a unique ERK3 MSK1 signaling module regulated by nuclear PKC␦ that is essential for DNA damage-induced apoptosis. Our study suggests that location-specific activation by distinct upstream regulators may enable common signaling pathways to have unique functional outputs.
The MEK/ERK pathway has well established roles in proliferation and survival and paradoxically regulates cell cycle arrest  Our study defines a common ERK3 MSK1 pathway that, when activated by EGFR in response to DNA damage, drives cell survival but when activated by nuclear PKC␦ drives apoptosis.

DNA damage-induced pro-survival and pro-apoptotic signaling
and apoptosis, particularly in response to DNA damage (21,22,(25)(26)(27). We show that biphasic activation of ERK in response to DNA-damaging agents sequentially activates prosurvival and pro-apoptotic signaling pathways. EGFR activation is required for the early, prosurvival phase of ERK activation in response to DNA damage. Notably, pretreatment of cells with gefitinib only suppressed the early phase of ERK activation, confirming that early and late ERK activation have distinct upstream regulators. The kinetics of activation of EGFR and early activation of ERK are similar to the kinetics of DNA repair, as assayed by comet, supporting published studies that show a role for EGFR in the regulation of IR-induced DNA double-stranded break repair (32)(33)(34)(35). In some cases, EGFR regulation of DNA repair is associated with nuclear translocation of EGFR, and EGFR has been shown to interact with and regulate DNA repair machinery, including DNAdependent protein kinase (32). EGFR can also regulate chromatin remodeling and DNA repair through tyrosine phosphorylation of histone H4 (36). The second, pro-apoptotic phase of ERK activation is largely dependent on PKC␦ and independent of EGFR. A pro-apoptotic function for PKC␦ is supported by extensive studies in vitro and in vivo that demonstrate a role for PKC␦ in apoptosis in response to many types of cell and tissue damage (3,4,10,17,18,20). Furthermore, our study confirms that PKC␦ does not contribute to cell survival signaling in response to DNA damage, as depletion of PKC␦ suppresses the apoptotic response at all time points (Fig. 3C). However, the fact that ERK activation is only partially suppressed in irradiated tissues from a ␦ knockout mouse supports data suggesting that other pathways, such as Akt and NF-B, can contribute to ERK activation in response to DNA damage (22). MSK1/2 kinases, members of the AGC kinase family, are activated by the ERK1/2 and p38 pathways in response to extracellular stimuli and predominantly localize to the nucleus (29 -31). Activation of MSK1/2 results in phosphorylation of transcription factors required for immediate-early gene expression and chromatin-associated proteins, including histone H3 (30,(37)(38)(39). In our study, MSK1 was activated by DNA damage with similar kinetics as ERK, and early MSK1 activation, like early ERK activation, was EGFR-dependent. We show that early activation of MSK1 through an EGFR3 ERK pathway promotes cell survival, consistent with studies that show a role in stress resistance (30) and enhanced apoptosis in an MSK1/2 knockout mouse (40). Given its known role in chromatin modification, it is interesting to speculate that MSK1 could contribute to the increase in EGFR-induced DNA repair we observed in the first few hours after DNA damage.
Notably, prosurvival signaling through EGFR3ERK3 MSK1 is transient; however, our study does not address how this signal is turned off. Nyati et al. (41) show that ataxia telangiectasiamutated (ATM) activation can down-regulate ERK through activation of dual specificity phosphatase-1 (DUSP1), also known as MKP-1. Likewise, nuclear localized DUSPs, such as DUSP5, are candidates for regulation of the nuclear level of activated ERK (42). Inhibition of ERK activation through DUSP6 has been shown to enhance the DNA damage response and increase sensitivity to EGFR inhibitors (43). Finally, it is possible that prosurvival ERK signaling does not need to be "turned off" for PKC␦ to induce pro-apoptotic ERK signaling, as sustained ERK activation has been shown to induce apoptosis (21,25).
As PKC␦ is ubiquitously expressed and regulates a wide variety of cell functions in addition to apoptosis, it is important to understand how apoptotic specific signaling by PKC␦ is mediated (5). We have shown that nuclear accumulation of PKC␦ is necessary for apoptosis in response to DNA-damaging agents and sufficient for apoptosis in many cell types (13)(14)(15)(16)(17)(18). This suggests that the subcellular localization of PKC␦ in the nucleus may dictate the apoptosis-specific functions of PKC␦. Conversely, it is tempting to speculate that localization of PKC␦ in other cellular compartments may dictate other functions, such as proliferation and cell migration. In this study, we use a constitutively nuclear targeted form of PKC␦ to address the role of nuclear PKC␦ in the regulation of pro-apoptotic signaling through ERK3 MSK1. Expression of PKC␦NLS potently activates ERK and MSK1, and activation of these kinases is necessary for PKC␦NLS to induce apoptosis (Fig. 5). Likewise, expression of ␦CF also induces ERK and apoptosis through an MSK1-dependent pathway (Figs. 3E and 5D). Although our study shows that nuclear targeting of PKC␦ is sufficient to activate ERK3 MSK1, it should be noted that PKC␦WT can also activate ERK (Fig. 3D). Thus we cannot rule out a role for PKC␦ in activating cytoplasmic ERK3 MSK1. Indeed, PKC␦ is a known effector of growth factor signaling and has been shown to activate ERK in this context (9).
MSK1 is largely associated with cell survival through regulation of transcription of stress-related genes (34). This is consistent with the prosurvival role we demonstrate for early activation of MSK1 (Fig. 4C). Early activation of MSK1, like ERK, occurs independently of PKC␦ but depends on EGFR (Fig. 4, A  and B). As MSK1 is predominantly a nuclear kinase, further studies are needed to clarify whether EGFR activates a cytoplasmic or nuclear pool of MSK1. Our study defines an additional role for MSK1 in the regulation of cell death and shows that it is a component of the signaling cascade regulated by nuclear PKC␦ that activates the apoptotic response. How PKC␦ regulates nuclear ERK3 MSK1 signaling is not known, although unpublished studies from our laboratory suggest that PKC␦ may regulate the abundance of nuclear and cytoplasmic DUSPs. Likewise, additional studies will be needed to decipher how the ERK3 MSK1 pathways induce apoptosis. It is tempting to speculate that, as a chromatin-modifying kinase, MSK1 could contribute to DNA repair by regulating the accessibility of damaged DNA to repair complexes.
Our identification of an ERK3 MSK1 signaling cascade that can regulate cell survival or apoptosis, depending on how and where it is activated, suggests a common mechanism that can be tuned according to the desired outcome. Conceivably, these pathways could also have shared targets that promote different functional outcomes dictated by the specific biological context. Clarification of prosurvival and pro-apoptotic targets of ERK3 MSK1 is essential to maximize the therapeutic potential of targeting these pathways.

Comet assay
Materials for the comet assay were purchased from Trevigen (Gaithersburg, MD). ParC5 cells were treated with DMSO, 300 nM gefitinib, or 20 M PD98059 and incubated at 37°C for 1 h prior to irradiation using a cesium-137 source (10 Gy). Following IR, cells were harvested at the indicated times for measurement of DNA damage using the neutral comet assay according to the Trevigen comet assay protocol. Images were analyzed by Trevigen Comet Analysis Software (version 1.3d). Tail moment was used to quantify DNA damage.

In vivo irradiation
C57Bl/6 PKC␦ ϩ/ϩ and PKC␦ Ϫ/Ϫ mice (45) were maintained at the University of Colorado Anschutz Medical Campus in accordance with laboratory animal care guidelines and protocols and with approval of the University of Colorado Denver Institutional Animal Use and Care Committee. Six-to eightweek-old female mice were anesthetized as described, and the head and neck regions were irradiated using a cesium-137 source, whereas the remainder of the body was shielded with lead. Salivary glands were harvested and tissue extracts prepared as described previously (3).

Analysis of apoptosis in vitro
Active caspase-3 was detected with the Caspase-3 Cellular Activity Assay Kit PLUS (Biomol, Farmingdale, NY; BML-ALK7030001), which uses N-acetyl-DEVD-p-nitroaniline as a substrate, according to the manufacturer's instructions. DNA fragmentation was assayed using the Cell Death Detection ELISA Plus Kit (Sigma-Aldrich) according to the manufacturer's instructions.

Statistics
Data shown in figures are representative experiments repeated a minimum of three times. Error bars indicate standard error. Caspase-3 activity, DNA fragmentation, and comet assays were designed with triplicate biological samples. Statistics were determined using GraphPad Prism 7 software. A twoway ANOVA (␣ ϭ 0.05) with Dunnett's multiple comparisons (**, p Ͻ 0.001; *, p Ͻ 0.05) was performed within each time point or treatment, comparing each column (gray or white) with the corresponding control (black).