Serum deprivation initiates adaptation and survival to oxidative stress in prostate cancer cells

Inadequate nutrient intake leads to oxidative stress disrupting homeostasis, activating signaling, and altering metabolism. Oxidative stress serves as a hallmark in developing prostate lesions, and an aggressive cancer phenotype activating mechanisms allowing cancer cells to adapt and survive. It is unclear how adaptation and survival are facilitated; however, literature across several organisms demonstrates that a reversible cellular growth arrest and the transcription factor, nuclear factor-kappaB (NF-κB), contribute to cancer cell survival and therapeutic resistance under oxidative stress. We examined adaptability and survival to oxidative stress following nutrient deprivation in three prostate cancer models displaying varying degrees of tumorigenicity. We observed that reducing serum (starved) induced reactive oxygen species which provided an early oxidative stress environment and allowed cells to confer adaptability to increased oxidative stress (H2O2). Measurement of cell viability demonstrated a low death profile in stressed cells (starved + H2O2), while cell proliferation was stagnant. Quantitative measurement of apoptosis showed no significant cell death in stressed cells suggesting an adaptive mechanism to tolerate oxidative stress. Stressed cells also presented a quiescent phenotype, correlating with NF-κB nuclear translocation, suggesting a mechanism of tolerance. Our data suggests that nutrient deprivation primes prostate cancer cells for adaptability to oxidative stress and/or a general survival mechanism to anti-tumorigenic agents.


Results
Serum deprivation prevented an apoptotic phenotype in prostate cancer cells. Serum deprivation in vitro reduces levels of growth factors in tumor cells that could que cells for death 32 . However, Martindale et al. demonstrated that serum deprivation primes cells to adapt to injury, stress, or death 33 . To study the role of serum deprivation in adaptive survival during oxidative stress in prostate cancer, we first visualized the phenotype of serum-containing and serum-deprived prostate cancer cell lines, and the phenotype of serum-deprived cell lines stimulated with hydrogen peroxide (H 2 O 2 ), our model of oxidative stress [34][35][36] . Serum-containing cells in each prostate cancer cell line (PC3, DU145 and LNCaP) were vulnerable to H 2 O 2 exposure and displayed significant cell death morphology as early as 4 h as denoted by black arrows (Fig. 1a). Additionally, we observed morphological hallmarks which accompany apoptosis such as rounding up of the cell, pyknosis (reduction in cellular volume), nuclear shrinkage, and retraction of pseudopodia 37 . However, over time (up to 8 h), serumdeprived PC3 and DU145 cells remained viable and developed a more flat morphology in addition to an increase in cell size and number-a phenotype that was more apparent with the addition of H 2 O 2 38 . In contrast, serumdeprived LNCaP cells (Fig. 1aiii) maintained a cell death morphology, and more so after H 2 O 2 stimulation, which was inhibited by the antioxidant, N-acetyl cysteine (NAC). Particularly for the PC3 and DU145 cell lines, this data suggests that serum deprivation may induce an adaptation to oxidative stress and promote long-term survival in more aggressive cancer cells.
Tumor cells can maintain higher ROS levels than normal cells, and are reported to confer more resistance to ROS-mediated death 39,40 . Serum deprivation, alone, induces oxidative stress 41 which we measured in each cell line with 2′,7′-dichlorofluorescin diacetate (DCFDA) for ROS generation. Serum-deprived PC3, DU145, and LNCaP cell lines displayed a significant increase in the generation of ROS compared to serum-containing control cells (overall p values: PC3 (≤ 0.001); DU145 (≤ 0.001); LNCaP (≤ 0.067)) (Fig. 1b). The addition of H 2 O 2 in serum-containing cells and serum-deprived cells displayed no significant change in ROS generation compared to their respective controls in PC3. However, this increase in ROS generation with the addition of H 2 O 2 was observed only in serum-containing cells in DU145, but in both serum-containing and serum-deprived cells of LNCaP. As expected, pre-treatment with ROS-scavenger, NAC, prior to stimulation with H 2 O 2 , inhibited significant ROS generation. We note the observation of depriving cells of serum only increases ROS generation, and additional ROS stimulation did not increase ROS production. It is believed that serum withdrawal causes cells to stop growing and initiates apoptosis; however, we observed a morphology that is inconsistent with apoptosis in serum-deprived samples exposed to ROS 42,43 (Fig. 1a). To determine whether serum deprivation primes cells for adaptive survival to increasing oxidative stress, we investigated the viability of cells grown under serum-deprived conditions. Our results indicate that PC3 and DU145 prostate cancer cells were viable when serum-deprived and Figure 1. (a) Serum deprivation prevents an apoptotic phenotype in prostate cancer cells. One hundred fifty thousand (1.5 × 10 5 ) PC3 (i), DU145 (ii), and LNCaP (iii) prostate cancer cell lines were plated in 6-well plates prior to serum deprivation and treatment with 5 mM n-acetyl-cysteine (NAC) for 1 h, and/or 250 μM H 2 O 2 at various time points (0, 4, 8 h). Phase contrast microscopy (20×) was used to capture images. Arrows highlight areas of cells with a cell death phenotype. Scale bar = 50 μm. (b) Serum deprivation generates ROS. Ten thousand (1 × 10 4 ) serum-containing and serum-deprived PC3 (i), DU145 (ii), and LNCaP (iii) prostate cancer cells were plated in black 96-well microplates. Cells were incubated for 1 h with 10 μM of 2′,7′-Dichlorofluorescin diacetate (DCFDA) followed by stimulation with 250 μM H 2 O 2 alone, pre-treatment with NAC (5 mM) followed by stimulation with 250 μM H 2 O 2 , or 2 μM Doxorubicin (Doxo; positive control to induce ROS) for 2 h via a microplate reader to detect ROS generation. The mean ± SEM of data were obtained from four independent replicate experiments. Statistical analysis (one-way ANOVA) was performed with GraphPad Prism (***p < 0.001; **p < 0.01; *p < 0.05). (c) Serum deprivation preserved viability during oxidative stress during. Cell viability was measured in PC3 (i), DU145 (ii), and LNCaP (iii) prostate cancer cells. Two hundred thousand (2 × 10 5 ) serum-containing and serum-deprived cells were either pre-treated with 5 mM n-acetyl-cysteine (NAC) and/or treated with 250 μM H 2 O 2 for 4 h. Cell viability was measured via a live/ dead cell assay (ThermoFisher) according to manufacturer's protocol. Green = live; Red = dead). Quiescence is a consequence of serum-deprivation and helps to manage oxidative stress. Serum deprivation in vitro induces proliferation arrest to protect cells from toxicities; consequently, a reversible cell cycle arrest (quiescence) is associated with this phenotype and cellular consequences are multidrug chemo-resistance and a propensity to evade apoptosis 18,45 . Therefore, we examined proliferation during oxidative stress via the 5-ethynyl-2′-deoxyuridine (EdU) assay. Serum-deprived PC3 (Fig. 3a) and DU145 (Fig. 3b) cells exhibited proliferation arrest, as expected, and continued after exposure to H 2 O 2 for 24 h. Moreover, cell cycle arrest was also observed when cells were grown with serum and stimulated with H 2 O 2 . These observations simply report that a stress event (via loss of serum or OS) halts proliferation. Doxorubicin, a potent inducer of apoptosis, served as a control.
Since we did not observe apoptosis or proliferation in our serum-deprived cells, nor when further exposed to oxidative stress via H 2 O 2 , we explored whether epithelial tumor cells may temporarily transition to a quiescent phenotype for management and survival during serum deprivation and/or oxidative stress 46 . We examined the morphology of serum-deprived PC3 and DU145 cells alone or serum-deprived and stimulated with H 2 O 2 for 1 h (Fig. 4ai, bi). It is reported, cells which display a shrunken, round, and flat morphology is indicative of a quiescent phenotype 18,33 . Indeed, we observed PC3 and DU145 cells cultured without serum, as well serumdeprived cells with H 2 O 2 , displayed a rounder and flatter phenotype compared to cells in serum suggesting that www.nature.com/scientificreports/ cells submit to a quiescent phenotype for survival and adaption to stress that may be a consequence of nutrient deprivation or downstream stress.
To confirm the quiescent phenotype on a molecular level, serum-deprived PC3 and DU145 prostate cancer cells were examined for expression of quiescent markers: (1) tumor suppressor gene, retinoblastoma (RB) 23,47 ; and (2) CDK inhibitor, cyclin-dependent kinase inhibitor 1B (p27 Kip1 ) 24 . It is well known that RB is an enforcer of quiescence 47 , and as such, the diminishing expression of phosphorylated-RB (pRB) in serum-deprived cells and serum-deprived cells exposed to H 2 O 2 indicates the onset of quiescence under stress (Fig. 4aii,bii). Accordingly, an accumulation of p27 Kip1 in the nucleus is indicative of G0/G1 arrest and quiescence 48 , and an observation of strong p27 Kip1 protein expression in the same samples further confirms that quiescence is necessary for cells to survive oxidative stress (Fig. 4aii,bii). Furthermore, we detected more accumulation of p27 Kip1 in the nucleus of PC3 (Fig. 4aiii) and DU145 (Fig. 4biii) cells in samples that were starved and further exposed to H 2 O 2 supporting that quiescence provides adaptive protection for tumor cells under stress to ensure long-term survival.
RelA/p65 (NF-κB) translocates to the nucleus in response to oxidative stress adaptation. NF-κB is recognized as a redox-sensitive transcription factor and is a major player in the cellular response to oxidative stress 8,49 . To determine the involvement of NF-κB, we first investigated the localization of RelA/p65, the major NF-κB activating subunit, in our samples 50 . In serum-deprived DU145 and PC3 cells exposed to H 2 O 2 ( Fig. 5a,b), RelA/p65 was detected with higher nuclear expression via cellular fractionation compared to controls. Likewise, observations were mirrored via immunocytochemistry where RelA/p65 was primarily nuclear in serum-deprived DU145 and PC3 cells, and serum-deprived cell lines exposed to H 2 O 2 (Fig. 5c,d). TNFα served as a positive control for nuclear localization of NF-κB 51,52 . Although a more complex mechanism is likely involved, this implicates the NF-κB-RELA/p65 pathway in contributing to the adaptability of cancer cells to oxidative stress environments.
RelA/p65 (NF-κB) and quiescence cooperate during oxidative stress. NF-κB is commonly associated with growth and the inflammatory response; however, some studies have indicated a role in quiescence 26,53,54 .
Recently, literature has demonstrated NF-κB to be implicated in protecting cells from apoptosis [55][56][57] . To first assess whether NF-κB is required for oxidative stress adaptation, we transiently silenced RelA/p65 and examined the apoptosis profile upon exposure to ROS. Compared to prior results (Fig. 2b), the percentage of total dead DU145 cells significantly increased in serum-deprived cells exposed to H 2 O 2 with the knockdown of RelA/ p65 (Fig. 6). We did not observe this effect in PC3 cells where there was no change in apoptosis upon RelA/p65 knockdown. Initially, we observed a decrease in pRB protein expression and an accumulation of nuclear p27 Kip1 in samples treated with TNFα, an inducer of RelA/p65 (NF-κB), suggesting a cooperation between NF-κB signaling and the quiescence program. Therefore, we sought to examine whether inhibition of quiescence reduced NF-κB-mediated cell survival during oxidative stress. We used two different quiescence inhibitors targeting Mirk/Dyrk1B (AZ191 and NCGC00185981-05/ML195) proteins regulating quiescence by stabilizing p27 Kip1 phosphorylation and nuclear localization, and inducing the degradation of cyclin D 21,39 . We examined p27 Kip1 nuclear accumulation in the presence of each inhibitor in DU145 prostate cancer cells (Fig. 7), and via immunocytochemistry, AZ191 (3-10 µM) was a more potent inhibitor of p27 Kip1 nuclear expression (Fig. 7a) indicating that the cells were not quiescent. Likewise, we could not resolve a distinct nuclear localization of RelA/p65 with 5-10 µM AZ191 (Fig. 8), suggesting that a quiescent phenotype and NF-κB may synergistically protect tumor cells from oxidative stress.

Discussion
Oxidative stress is considered to be one of the mechanisms that trigger early stages of prostate disease lesions, particularly prostatitic hyperplasia, benign prostatitic hyperplasia 58-60 , proliferative inflammatory atrophy (PIA) 61,62 , and others. Overall, if not subsided, the consequence(s) of oxidative stress result in a significant decrease in the antioxidant systems leading to lipid, protein, and DNA damage. However, at levels that are still under investigation, the resulting ROS during a stressful event can prime biochemical molecules to allow prostate cancer to develop and progress, such as deactivating tumor suppressors 35,63 or increasing expression of pro-migratory signaling axes 35 . These dichotomic roles for ROS make it difficult to assess its clinical efficacy. For instance, questions that may arise during a clinical observation are: (1) is the observance of oxidative stress in a clinical BPH simply the result of aging tissue and vascular deterioration; (2) or is it the onset of tumor development; or (3) is the signaling potential of ROS collateral damage in cancer chemotherapy with the eventual outcome of a migratory tumor cell 34,[64][65][66][67] . The preventative role of oxidative stress regulators is thought to protect the prostate from tumor development; however, chronic stress over time induces somatic mutations in DNA, lipids and proteins resulting in neoplastic transformation due to alterations in metabolic checkpoints. Additionally, the byproducts of ROS-based therapy are now being acknowledged to help propagate, amplify, and create a mutagenic and oncogenic microenvironment that is beneficial to a transforming metastatic tumor cell 68 . Thus, the metabolic relationships that are regulated by oxidative stress and the onset of prostate tumorigenesis remain an enigma.
In general, an advanced tumor is conditioned to survive in the poorest conditions i.e. when oxygen, glucose 69 and amino acids 70 are not accessible for metabolism during intravasation, extravasation and migration. Hypoxia and poor nutrition are common in a tumor mass due to insufficient vascularization of a heterogeneous and/or mobile tumor 71 . Angiogenesis is one of the hallmarks of survivability because neovascularization provides the nutrients and oxygen necessary for tumor sustainability. However, considering that tumors survive very well during the metastasis process and in a heterogeneous tumor mass with a limited vascular supply, there must be additional markers to access tumor survivability. Izuishi et al. posited two theoretical ways of adaptations to an insufficient oxygen and nutrient supply 72 . In brief, one way is by increasing the supply through angiogenesis, and www.nature.com/scientificreports/ the other way is by developing tolerance and alternate coping mechanisms to survive. Eventually, tumors grow beyond its ability to coordinate a vascular supply, and possibly, only the tumors cells that have learned to cope or adapt in a deprived environment might be the phenotype that advances to malignancy. Much therapeutic attention is paid to angiogenesis since this is a crucial event for survival and migration; however, therapeutic focus should expand to tumor populations that outgrow angiogenesis-dependent survival and can tolerate nutrient deprivation and oxidative stress. We observed that starvation primed prostate cancer cells for further insult to oxidative stress. Although not investigated in this study, we have previously observed concomitant expression of HIF1-α and the chemotactic receptor, CXCR4, in DU145 prostate cancer cells under H 2 O 2 oxidative stress 35,36 . Herein, our data support the notion that nutrient deprivation primes tumors for adaptation to stress and may serve as a determinant for tumor survival during stress and tumor progression 73 . Particularly, prostate cancer cells treated with both serum and H 2 O 2 were apoptotic, a phenotype that was not observed in cells that were initially starved. Quiescent states allow diverse microorganisms to survive for long terms without nutrients in contrast to a proliferative "active" state that maintains itself by filtering out damaged molecules. During a state of quiescence, there must be a specialized protective biochemistry to counter damage for a cell to choose it as a first responding "safe house" to stress. For instance, Longo et al. reported that quiescent yeast cells retained a capacity to detect and respond to oxidative damage, making it a preferred phenotype during stress 72,74 . Additionally, Hill et al. demonstrated cellular stress induced a protective sleep-like state (quiescence) in C. elegans, and quiescence-defective organisms showed elevated expression of stress reporter genes and were impaired for survival, and suggest a deeply conserved function of quiescence 75 . Furthermore, quiescent and self-renewing stem cells easily reside in quiescence during hypoxia and oxidative stress 76,77 . Perhaps, this is what we are observing in cancer. Autophagy is another consideration for phenotypic adaptation to stress 78 . Izuishi et al. demonstrated that some tumor cells acquired strong tolerance for nutrient deprivation. Pancreatic cancer cell lines, which are notoriously hypo-vascular as malignancy increases, survived for considerably longer periods under extremely low nutrient conditions than cell lines of liver cancer suggesting that tumor cells that have acquired the ability to survive an unfavorable microenvironment might be the most aggressive malignancy, and correlates with poor differentiation of tumors. In comparison, we observed that nutrient deprivation primed prostate cancer cells for tolerance of oxidative stress and adaptation for longer periods compared to cells with serum.
Adaptation and survival depended on a quiescent phenotype and transcription factor NF-κB which are usually responsive to low nutrient conditions and stress. Literature also suggests that NF-κB has alternate responses to oxidative stress depending on the context of the environment. While most studies focus on pro-apoptosis during times of stress 8 , less attention is paid to the anti-apoptotic functions of Rel/NF-κB complexes. In some situations, NF-κB promotes apoptosis in response to a specific cell type and stimulus [79][80][81][82] . For instance, Babaei et al. demonstrated that PC3 cells treated with biseugenol B, a cytotoxic agent, repressed the apoptosis-inhibitor activity of NF-κB thereby preventing NF-κB nuclear translocation 83 . Additionally, some Bcl-2 family member genes are up-regulated by NF-κB following the onset of stressful stimuli such as γ-irradiation and UV-radiation. Conversely, Bcl-2 has a reciprocal relationship with NF-κB whereby Bcl-2 has an inhibitory effect on NF-κB activity through stabilizing IkBa, inhibiting RelA transactivation, and interfering with the nuclear translocation of Rel family members 84,85 . . Images were acquired via flow cytometry (Accuri C6 Cytometer; BD Biosciences); data was analyzed using FlowJo (v10). The mean ± SEM of data were obtained from three independent replicate experiments. Statistical analysis (one-way ANOVA) was done with GraphPad Prism (***p < 0.001; **p < 0.01; *p < 0.05). www.nature.com/scientificreports/ Early in our study, LNCaP demonstrated poor survivability during oxidative stress even when initially starved. We suspect that a p53 functional status determines the ability of ROS to induce different responses to death in LNCaP cells versus DU145 and PC3 cells. The LNCaP cell line has a wild-type, functional p53 while DU145 bears a mutant p53, and PC3 bears a frameshift producing a stop codon and an allele deletion 86 . The p53 transcription factor is a critical element in the cell's ability to regulate the cell cycle and its response to DNA damage 86 . One of the most important unknowns in investigating p53 is how it determines a cellular outcome (cell cycle arrest vs. senescence vs. apoptosis) via regulation of outcome-specific target genes 87 . ROS act as both an up-stream signal that triggers p53 activation and as a downstream factor that mediates apoptosis 87 . Death is not the only outcome of p53 signaling during oxidative stress; however, Zhao et al. described that once in the mitochondria, p53 inhibits mitochondrial superoxide dismutase (MnSOD), playing a direct role in promoting apoptosis 88 . In addition, basal levels of p53 also has an antioxidant role, and the outcome depends on the context of the cell 88 .
How do cells recognize nutrient starvation? Izuishi et al. suggest that cells seem to recognize the amount of AMP and AMP-activated protein kinase in addition to stress signals for hypoxia, nutrient starvation, and physical stresses. Therefore, it is also probable that as cells increase in malignancy, they have already acquired constitutive tolerance for nutrient and oxygen starvation through multiple carcinogenesis steps 73 . Other considerations are a crosstalk between the mTORC1 and eIF2α pathways 89 , and AKT-mediated activation of NF-κB 90 . Early works attribute much of cellular survivability during oxidative stress to AKT 90 where Song et al. described that inhibition of AKT phosphorylation induced decreases in sequential NF-κB signaling after 30 min of transient focal cerebral ischemia along with decreases in downstream survival signals of the AKT pathway. We also know that NF-κB activation and nuclear translocation can be blocked by PI3K/Akt inhibitors 91 . In addition to the NF-κB/AKT One million (1.0 × 10 6 ) serum-containing and serum-deprived PC3 (a) and DU145 (b) prostate cancer cells were treated with 250 μM H 2 O 2 , or co-treated with 10 mM n-acetyl-cysteine (NAC) or 2 μM Doxorubicin (Doxo) for 24 h. Cells were incubated with 20 μM EdU label per manufacturer's instructions; control cells were cultured in 10% FBS without EdU. Images were acquired via flow cytometry (Accuri C6, BD Biosciences); data was analyzed using FlowJo (v10). The percentage of gated cells (EdU + ) is highlighted. (c) A graphical representation of EdU + PC3 cells. (d) A graphical representation of EdU + DU145 cells. The mean ± SEM of data were obtained from three independent replicate experiments. Statistical analysis (one-way ANOVA) was done with GraphPad Prism (***p < 0.001; **p < 0.01; *p < 0.05).
Scientific RepoRtS | (2020) 10:12505 | https://doi.org/10.1038/s41598-020-68668-x www.nature.com/scientificreports/ relationship, MAPK is also involved in NF-κB signaling in a context dependent manner [92][93][94] . Moreover, in our own work, we demonstrated that ROS accumulation permitted AKT and CXCR4-mediated functions through PTEN catalytic inactivation. In this case, a disulfide bridge formed within the catalytic cleft of PTEN, inhibiting its suppressive functions, which allowed ROS to freely orchestrate signaling. We observed increased phosphorylated AKT (p-AKT) and CXCR4 expression, independent of any ligands, which were abrogated by a ROS scavenger in prostate cancer cells. ROS-mediated catalytic inactivation of PTEN did not affect its expression, yet enhanced cell migration and invasion in a CXCR4-dependent manner 35 . We also observed a contrary relationship between ROS and AKT where ROS facilitated cell death through activation of AKT. We initially observed that ROS increased expression of p-AKT in 22Rv1 human prostate cancer cells. The tumor suppressor, PTEN, a negative regulator of AKT signaling, was rendered catalytically inactive through oxidation by ROS, although the expression levels remained consistent. Despite these events, cells still underwent apoptosis. Further investigation into apoptosis revealed that expression of the tumor suppressor pVHL increased and contains a target site for p-AKT phosphorylation. pVHL and p-AKT associated in vitro, and knockdown of pVHL rescued HIF1α expression and the cells from apoptosis 36 . With all of this literature describing the relationship between ROS, AKT and cell survival, we believe our data is novel because there are no studies that describe the physical state of surviving epithelial tumor cells during an oxidative stress event. In literature, the phenom is reserved in cancer stem cells.
With regards to therapy, trans-arterial chemoembolization, damaging the blood supply of a tumor to prevent delivery of oxygen, growth factors, nutrients and others, has shown great success in providing significant improvement in overall survival, disease-free survival, and recurrence rates 95 . Although not considered curative, this intervention is dominant in liver cancer. The procedure is being considered for prostate cancer with hopes of causing irreversible necrosis of prostate tissue and causing the gland to shrink and soften 96,97 . This type of therapy would be the closest to nutrient deprivation therapy available for prostate lesions but would likely only be effective on a premalignant primary tumor mass versus a metastatic cells and/or cells that have adapted to nutrient deficiency; the challenge still remains in identifying tumors cells that are fully adapted to hostile conditions. Nevertheless, our results bring to the forefront a tumor cell phenotype that is often underappreciated, yet critical to metastasis, relapse and likely death.

Cell proliferation assay. To measure cell proliferation, an EdU (5-ethynyl-2′-deoxyuridine) Proliferation
Kit (Abcam) was employed according to the manufacturer's protocol. One million (1.0 × 10 6 ) serum-containing and serum-deprived PC3 and DU145 were stimulated as described above prior to the addition of 20 μM EdU for 3 h. Alternatively, control cells were incubated in 10% FBS without EdU label at 37 °C for 24 h. Cells were fixed with 3.7% formaldehyde in 1 × PBS at RT for 15 min followed by permeabilization with 0.5% Triton X-100 in 1 × PBS for 20 min at RT. Cells were incubated in reaction cocktail for 30 min at RT then washed twice with 1 × PBS. Cells were analyzed by flow cytometry (Accuri C6, BD Biosciences) for the detection of EdU-positive (EdU + ) cells. Data was analyzed using FlowJo (v10). Experiments were performed at least thrice, and statistical analysis was performed with GraphPad Prism (***p < 0.001; **p < 0.01; *p < 0.05).
Subcellular fractionation. Subcellular fractionation technique was performed as we've previously described 101 . Two million (2.0 × 10 6 ) serum-containing and serum-deprived PC3 and DU145 cells were stimulated with 250 μM H 2 O 2 or TNFα (0.1 ng/mL) for 30 min. Alternatively, select samples were pre-treated with 5 mM NAC for 1 h prior to 250 μM H 2 O 2 for 1 h. Cells were harvested for subcellular fractionation according to the manufacture's protocol (NE-PER Nuclear and Cytoplasmic Extraction Kit, ThermoFisher). Briefly, cells were lysed in a series of buffers, centrifuged to obtain a non-nuclear fraction and an intact nuclear pellet, and further lysed to isolate the nuclear fraction. Forty micrograms (40 μg) of total cell lysate were resolved by SDS-PAGE to detect RelA/p65 (1:1,000, Cell Signaling Technology). Topoisomerase I (1:1,000, Santa Cruz Biotechnology) and β-actin (1:1,000, Cell Signaling Technology) were used as loading controls.
Immunocytochemistry. Immunocytochemistry technique was performed as we've previously described 101 .
One hundred thousand (1 × 10 5 ) serum-containing and serum-deprived PC3 and DU145 cells were plated on glass coverslips (Fisher). Cells were stimulated as previously described above. Cells were fixed with 4% paraformaldehyde for 40 min at RT and washed with 1 × PBS and 0.1% Tris-glycine. Non-specific proteins were blocked in blocking solution (5% normal donkey serum/1% BSA/0.3% Triton X-100 in 1 × PBS) for 30 min at RT, prior to incubating with RelA/p65 ( Immunoblotting. Immunoblotting technique was performed as we've previously described 102 . Briefly, one million (1.0 × 10 6 ) serum-containing or serum-deprived cells were stimulated with 250 μM H 2 O 2 for 1 h or TNFα (0.1 ng/mL) for 30 min. Alternatively, cells were pre-treated with 5 mM NAC for 1 h followed by 250 μM H 2 O 2 for 1 h. Cells were lysed and sonicated in 1 × Cell Signaling Technology lysis buffer prior to incubation on ice for 30 min. Lysates were centrifuged at max speed for 10 min at 4 °C, and then equal amounts of protein per sample were separated by SDS-PAGE and transferred to PVDF membrane. Protein bound membranes were blocked in 5% BSA/1XTBST and subsequently incubated with primary antibodies p27 Kip1 (1:1,000, Cell Signaling Technology) or pRB (1:1,000, Cell Signaling Technology) overnight at 4 °C in 5% BSA/TBST. Beta actin (β-actin; 1:1,000; Santa Cruz Biotechnology) served as a loading control. Primary antibodies were detected by HRP-conjugated secondary antibodies (1:10,000, Jackson ImmunoResearch) diluted in 5% BSA/1 × TBST. Protein expression was detected with chemiluminescence (Luminata Western HRP Chemiluminescence Substrates; Millipore Sigma) on ChemiDoc MP Imaging System (Bio-Rad, USA).

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.