Valproic Acid Sensitizes Hepatocellular Carcinoma Cells to Proton Therapy by Suppressing NRF2 Activation

Although efficacy of combined histone deacetylase (HDAC) inhibitors and conventional photon radiotherapy is being tested in clinical trials, their combined effect with proton beam radiotherapy has yet to be determined. Here, we compared combined effect of valproic acid (VPA), a class I and II HDAC inhibitor and antiepileptic drug with proton and photon irradiation in hepatocellular carcinoma (HCC) cells in vitro and in vivo. We found that VPA sensitized more Hep3B cells to proton than to photon irradiation. VPA prolonged proton-induced DNA damage and augmented proton-induced apoptosis. In addition, VPA further increased proton-induced production of intracellular reactive oxygen species and suppressed expression of nuclear factor erythroid-2-related factor 2 (NRF2), a key transcription factor regulating antioxidant response. Downregulation of NRF2 by siRNA transfection increased proton-induced apoptotic cell death, supporting NRF2 as a target of VPA in radiosensitization. In Hep3B tumor xenograft models, VPA significantly enhanced proton-induced tumor growth delay with increased apoptosis and decreased NRF2 expression in vivo. Collectively, our study highlights a proton radiosensitizing effect of VPA in HCC cells. As NRF2 is an emerging prognostic marker contributing to radioresistance in HCC, targeting NRF2 pathway may impact clinical outcome of proton beam radiotherapy.

in progress. Pelvic radiation and vorinostat (PRAVO) study was the first clinical trial of combination of HDAC inhibitors with palliative RT confirming the safety and tolerability of the combination 18 . A retrospective study of valproic acid (VPA) treatment during RT for glioblastoma showed association of VPA and improved survival 19 . A phase 2 study of combination of RT, temozolomide and VPA for glioblastoma revealed combined VPA/RT/ temozolomide was well tolerated and expected to improve outcomes 20 .
Particle beams such as protons and carbon ions have different biological activities compared with the same physical dose of photon, so called relative biological effectiveness (RBE). Proton beam shows superior dose distribution compared with photon due to "Bragg peak" and has similar or slightly higher RBE as 1.1 than the photon 21 . However, the RBE of proton could be changed by several circumstances, like energy of proton, type of irradiated tissue, and dose/fraction. It is also reasonable to expect that RBE could be modified by the addition of chemotherapeutic agents 22 .
VPA is a well-known anti-epileptic drug that has been used for about 50 years to treat seizure and inhibits class I and IIa HDACs and its usage in cancer treatment has been recently described 5,23 . Synergistic or additive effect of VPA with conventional photon therapy has been extensively studied in various cancers [24][25][26][27][28] , but combination effect of VPA with proton therapy has yet to be determined. Using clinical radiation therapy machines, we report here for the first time that VPA acted as proton radiosensitizer in hepatocellular carcinoma (HCC) cells in vitro and in vivo. As proton beam therapy is a favorable option for HCC treatment, our findings suggest that VPA may have benefit in treating HCC with proton therapy.

VPA treatment inhibits deacetylation of histones and minimally reduces cell proliferation.
To test whether VPA, an antiepileptic drug, acts as HDAC inhibitor in human HCC cells, we determined acetylation of histone H4 after VPA treatment in Hep3B. Western blot analysis showed that acetylated histone H4 was low in untreated cells and increased at 6 h after 1 mM VPA treatment (Fig. 1a). VPA treatment also increased the acetylation of histone H4 in a concentration-dependent manner (Fig. 1b), confirming function of VPA as HDAC inhibitor in HCC cells. Next, the effect of VPA on HCC cell proliferation was assessed by using the colorimetric MTT metabolic activity assay. As VPA concentration increased, cell proliferation slightly decreased; 5 mM VPA reduced Huh7 and Hep3B cell proliferation by 11.0% and 25.6%, respectively, which suggests that Hep3B cells are more sensitive to VPA than Huh7 cells (p < 0.001) (Fig. 1c).

VPA increases sensitivity of HCC cells to proton irradiation.
To compare the potential of VPA in sensitizing HCC cells to photon and proton irradiation, we performed clonogenic survival assay. Considering that photon and proton beams have different dose distribution profiles, HCC cells were positioned at water-equivalent depth of 2 cm (for photon irradiation) or 17.2 cm (for proton irradiation, in the middle of the spread-out Bragg peak) using water-equivalent solid phantoms (Fig. 2a). The Hep3B cells were pre-treated with 1 mM VPA for 3 h and then were irradiated with the same physical doses of photon or proton beam. The cell survival curves were fit to a linear-quadratic model. Comparison of survival curves of proton and photon-irradiated cells showed that proton killed more Hep3B cells than photon (Fig. 2b). Proton RBE of Hep3B was 1.08 at survival fraction (SF) of 0.1, which is close to generic RBE of 1.1 that is being used in the clinic. VPA increased radiosensitivity with both proton and photon irradiation, but it killed more Hep3B cells when combined with proton. Value of SF at 6 Gy (SF6) of photon and VPA was much less than that of SF6 of proton and VPA (0.090 vs 0.024). VPA treatment increased proton RBE from 1.08 to 1.17 (Fig. 2c), suggesting a potential role of VPA in proton sensitization.
VPA enhances proton irradiation-induced cell cycle arrest. To understand how VPA potentiates proton radiosensitization in HCC cells, we first evaluated effect of VPA on cell cycle progression in Hep3B cells (Fig. 3). The majority of untreated cells stayed in G1 phase, which was minimally affected by 1 mM VPA alone. At 24 h post-irradiation, percentage of cells in the G2/M phase increased from 32.9% to 71.4% for photon and 70.1% for proton (p < 0.001) and then decreased to 45.8% and 49.2% at 72 h post-irradiation (p < 0.05). When combined with 1 mM VPA, percentage of G2/M arrested cells increased to 73.4% and 80.1% at 24 h (p < 0.001) and 59.9% and 58.6% at 72 h (p < 0.001), which was higher than radiation alone. It suggested that VPA may attenuate recovery from radiation-induced cell cycle arrest.
VPA attenuates repair of proton-induced DNA damages. To test whether VPA affects radiationinduced DNA damage repair, we monitored formation of γ-H2AX foci, a sensitive biomarker of DNA double strand breaks. Immunofluorescence analysis showed that γ-H2AX foci formation prominently increased with either photon or proton irradiation at 2 h and then decreased at 24 h ( Fig. 4a and b). VPA treatment alone did not have an impact on γ-H2AX foci formation in non-irradiated cells but it markedly attenuated resolution of radiation-induced γ-H2AX foci ( Fig. 4a and b), which was more evident in proton-irradiated cells (p < 0.01). Effect of VPA in DNA damage response signaling was determined by western blot analysis of γ-H2AX, phosphorylated ATM and ATR. γ-H2AX expression disappeared at 72 h of post-irradiation but it remained persistent in cells co-treated with VPA and proton beam irradiation (Fig. 4c). Phosphorylation of ATM and ATR was slightly higher in proton-irradiated cells and strongly enhanced by VPA co-treatment at 72 h of post-irradiation (Fig. 4d).
VPA enhances proton irradiation-induced apoptosis. Next, we determined effect of VPA on radiationinduced apoptosis using flow cytometry. Irradiation of photon and proton beam increased apoptotic cell population from 5.0% to 9.9% and 10.0%, respectively (Fig. 5a). The co-treatment with VPA further increased the proportion of apoptosis: combination of proton and 1 mM VPA (20.1%) induced more apoptotic cell death than that of photon and 1 mM VPA (12.7%).
The increment of apoptosis by combination treatment with proton irradiation and VPA was further confirmed by western blot analysis (Fig. 5b). The cleavage of PARP was induced by either photon or proton irradiation, which was further enhanced by VPA treatment. Cleaved caspase-3 level was the most prominent in the Hep3B cells co-treated with VPA and proton irradiation.

VPA enhances proton-induced accumulation of reactive oxygen species (ROS). Since radiation-
induced DNA damage is mainly due to ROS generation following water radiolysis, we determined the effect of VPA on radiation-induced accumulation of ROS level in Hep3B cells. The ROS sensitive fluorescent dye, 2′, 7′-dichlorodihydrofluorescein diacetate (H2DCF-DA) was used for ROS detection. Flow cytometry analysis showed that irradiation increased intracellular ROS level and the percentage of H2DCF-DA positive cells was even higher upon proton than photon irradiation (Fig. 6a). The percentage of H2DCF-DA positive cells were further increased by the addition of 1 mM VPA, suggesting that VPA enhanced proton-induced ROS production.
Next, we tested effect of VPA on signaling of nuclear factor erythroid 2-related factor 2 (NRF2), a key regulator of cellular redox homeostasis. Western blot analysis of NRF2 showed irradiation increased NRF2 expression as seen in previous reports (Fig. 6b). Interestingly, NRF2 expression was more induced by proton irradiation than photon irradiation, which would be due to more ROS production upon proton irradiation than photon irradiation. Expression of heme oxygenase 1, HO-1, a target gene of NRF2, was dramatically increased by proton irradiation. However, VPA treatment abrogated the proton-induced expressions of both NRF2 and HO-1 (Fig. 6b). To test whether NRF2 level is related to sensitivity to proton irradiation, we depleted NRF2 in Hep3B by shRNA transfection (Fig. 6c). The increment of apoptotic cell population in NRF2-depleted cells was seen upon proton irradiation compared to photon irradiation (p < 0.01) and control shRNA cells (p < 0.01), suggesting that NRF2 may protect cells from proton-induced cell killing (Fig. 6d).

VPA enhances proton-mediated suppression of Hep3B xenograft tumor growth.
We evaluated the in vivo efficacy of VPA on proton radiosensitization using Hep3B mouse xenograft model. Nude mice implanted with Hep3B cells were treated with PBS (control), VPA (300 mg/kg, every three days), photon irradiation alone (three daily fractions of 3 Gy), photon irradiation and VPA, proton irradiation alone (three daily fractions of 3 Gy) or proton irradiation and VPA (Fig. 7a).
The Hep3B tumor growth of each treatment group was compared on day 15. Photon (256.1 ± 161.5, p < 0.05) and proton irradiation (175.5 ± 93.1, p < 0.01) significantly inhibited Hep3B tumor growth, compared to control group (1,038.1 ± 555.4 mm 3 ) (Fig. 7b). VPA alone did not inhibit tumor growth (852.2 ± 555.4 mm 3 , p = 0.619). No significant difference between proton irradiation and photon irradiation group was seen (p = 0.326). To clarify effect of VPA on tumor growth, we calculated tumor growth delay (TGD) that is defined as a difference in the time to reach tumor volume of 500 mm 3 between control group and experimental group (Fig. 7c). Either photon or proton irradiation significantly delayed tumor growth (p < 0.001); the TGDs of photon irradiation and proton irradiation groups relative to control group were 18.0 days and 25.2 days, respectively and proton was more In accordance with in vitro data, TUNEL assay on tissue sections from the transplanted tumours revealed that both irradiations increased apoptotic cells in vivo, which were further enhanced by VPA treatment (Fig. 7d). Quantification data confirmed that VPA alone did not affect apoptosis on the tissues but its combined treatment with either photons or protons led to a significant enhancement of apoptosis in the tumour tissues (p < 0.001, Fig. 7e). In addition, immunohistochemical analysis on the tumour tissues showed that NRF2 expression was greatly induced by both irradiations (p < 0.001, Fig. 7f), which is well consistent with in vitro results (Fig. 6b). VPA treatment suppressed radiation-induced NRF2 expression (p < 0.001). Collectively, our data suggest that VPA exerted in vivo radiosensitizing effect via induction of apoptosis and suppression of NRF2.

Discussion
In this study, the radiosensitizing effect of a HDAC inhibitor, VPA on human HCC cells was evaluated with two different types of radiation, photon and proton using in vitro and in vivo models. To the best of our knowledge, this is the first study to directly compare the combined effect of a HDAC inhibitor on photon and proton irradiations. It is intriguing that VPA exerted a stronger sensitizing effect when combined with proton irradiation, compared to photon irradiation. Enhanced DNA damages and accumulated ROS production were seen when VPA and proton irradiation was co-treated. Consistent with in vitro data, VPA enhanced proton-mediated suppression of xenograft tumor growth in vivo, suggesting that VPA could be a valuable proton sensitizer in HCC treatments.
The RBE of proton irradiation is not an issue because it is well recognized, studied, and actively adapted in oncologic fields 21 . It is demonstrated that proton RBE can be affected by initial proton beam energy, linear energy transfer, dose per fraction, cell types, and proton therapy machines. Regardless of the complexity and/or uncertainty, "generic RBE" which means a single all-round value of RBE 1.1 is widely used in most clinical proton therapy centers. Radiosensitizing effect of chemotherapy drugs like cisplatin, 5-fluorouracil and capecitabine on conventional photon therapy has been well studied and is simply being applied to proton treatment in a clinical setting [29][30][31] . The use of proton beam therapy and capecitabine for patients with resectable pancreatic ductal adenocarcinoma is well tolerated with favorable local control 29 . In an esophageal cancer study, proton beam therapy combined with cisplatin and 5-fluorouracil seemed to be feasible with regard to late cardio-pulmonary toxicity 30 . In addition, high dose proton beam therapy with cisplatin and vinorelbine is safe to use for treatment of unresectable stage III non-small cell lung cancer 31 . Considering differences in biological effects between proton and photon irradiations, there is, however, no doubt that proton RBE could be affected by other chemotherapeutic agents. This prompted us to investigate an effect of HDAC inhibitors that are recently highlighted as a radiosensitizer, on proton RBE.
In addition to their potential as anticancer therapeutics 10 , HDAC inhibitors have been tested as radiosensitizer. For example, suberoylanilide hydroxamic acid (SAHA, vorinostat) enhanced photon irradiation-mediated cell killing by inhibiting DNA double strand break repair 17 . SAHA inhibited radiation-induced upregulation of RAD50 and RAD51 in diverse types of cancer cells, suggesting a possible role of impaired homologous recombination (HR) repair pathway in SAHA-induced radiosensitization 14,16,17 . SAHA-mediated downregulation of Ku70 and Ku80 protein was seen in colon cancer cells, suggesting implication of non-homologous end-joining (NHEJ) DNA repair pathway in the radiosensitization 15 . As with SAHA, increased efficacy by combination treatment with VPA and photon irradiation has been reported in preclinical and/or clinical setting 25,26,32 . In consistent with the previous results, enhanced radiation sensitivity by VPA was also seen in our study with HCC cells. The sensitizing effect of VPA was higher in proton-irradiated cells than photon-irradiated cells; VPA enhanced proton-induced apoptosis both in vitro and in vivo. Our findings indicate that stronger sensitizing effect of VPA in response to proton over photon irradiation increased proton RBE, which may allow to de-escalate prescription doses and protect normal tissues in a clinical setting.
Recent preclinical studies have shown that HR repair pathway is more necessary for repair of proton-induced DNA damages than NHEJ [33][34][35] . Defects in HR pathway increased proton RBE in lung cancer cells 34 and SAHA-mediated RAD51 downregulation enhanced proton sensitizing effect in lung adenocarcinoma A549 cells 33 . In our study with HCC cells, altered expression of HR pathway proteins was not seen in VPA-treated cells. Once tumors were palpable, they were irradiated with 3 Gy for 3 consecutive days for a total 9 Gy. Mice were treated with intraperitoneal injections of VPA (300 mg/kg/day) every 3 days. Shown are mean tumor volumes and standard deviation per group (n = 4). (c) Tumour growth delay was determined by calculating days each tumour taken to reach 500 mm 3 . n.s. not significant; *p < 0.05; ***p < 0.001. (d) TUNEL assay detected more apoptotic cells in tissues cotreated with proton and VPA compared to proton alone or combined photon and VPA. Little was seen in tissues treated with VPA alone. Scale bar, 400 µm. (e) Quantification of the TUNEL positive cell density in tumor tissue sections. n.s. not significant; **p < 0.01; ***p < 0.001. (f) Proton and photon irradiations increased NRF2 expression on tumour tissues, which was suppressed by co-treatment with VPA. NRF2 expression was assessed by immunohistochemistry. Scale bar, 400 µm. (g) Quantification of NRF2 expression in tissue samples. *p < 0.05; ***p < 0.001.
Instead, co-treatment with VPA increased proton-induced G2/M arrest after 72 h and attenuated proton-induced DNA damage repair, which was judged by persistence of γH2AX foci and prolonged activation of ATM/ATR. Thus, this suggested implication of a different mechanism rather than suppression of expression of DNA damage repair genes in VPA-mediated proton radiosensitization.
The DNA damages induced by sparsely ionizing radiations, like photon and proton beam is mainly caused by ROS produced during water radiolysis. In our study, more accumulation of ROS was seen in the Hep3B cells treated with proton than photon, which is consistent with previous studies 36,37 . Level of NRF2, a key transcription factor protecting from oxidative stress, increased concomitantly with the increase in ROS level after irradiations, which was prominent in proton-irradiated cells. VPA alone also slightly increased NRF2 level, which might be related to increased ROS level. It has been shown that VPA 38,39 and other HDAC inhibitors [40][41][42][43] increased intracellular ROS levels through NADPH oxidase activation. Even though ROS levels were further increased by co-treatment with radiation and VPA, proton-induced NRF2 expression was suppressed by VPA treatment. Moreover, VPA-mediated suppression of expression of HO-1, a target gene of NRF2, suggested that VPA inhibited radiation-induced activation of NRF2 signaling. Suppression of NRF2 expression by VPA [44][45][46] and other HDAC inhibitors 47,48 were also seen in cancer and other disease and is proposed to be a plausible approach to enhance antitumor activity of TRAIL or EGFR inhibitors. Taken together, our data suggest that VPA suppressed induction of NRF2 in response to proton-mediated ROS overproduction and downregulated expression of anti-oxidative genes such as HO-1, thereby resulting in enhancement of RT-induced apoptosis and reduction in clonogenic survival.
NRF2 is frequently activated in HCC (5-15%) and is considered as one of oncogenic drivers including TERT and β-catenin 49,50 . Activated NRF2 confers radioresistance and chemoresistance by protecting cells from radiation-induced oxidative stress and detoxifying broad spectrum of anticancer drugs. In a non-small lung cancer, genetic modulation of NRF2 levels affects radiation sensitivity and NRF2 knockdown potentiates radiation-induced apoptosis 51,52 , which is well consistent with our results. Similar findings were also seen in various cancers including prostate 53,54 , squamous cell lung cancer 55 , hepatocellular carcinoma 56 and glioblastoma cells 57 . A recent study on squamous cell lung cancer patients revealed that NRF2 status is a predictive marker of local recurrence after radiation therapy 58 . As our data clearly showed that depletion of NRF2 by VPA treatment or shRNA transfection made cancer cells much more sensitive to proton irradiation, targeting NRF2 pathway may impact outcome of proton radiation therapy.
This preclinical study clearly showed a reliable combined effect of VPA with proton irradiation in HCC in vitro and in vivo, but there remain several obstacles for clinical use of VPA upon proton therapy for HCC treatment. Whereas the optimum dose of VPA as antiepileptic drug is well documented, its optimal dose for radiosensitization is still under investigation 20 . Based on the previous preclinical studies 25,27,59 , we used 300 mg/kg/ day for our in vivo study, but it is higher than the maximum daily recommended dose (60 mg/kg/day) used for epilepsy. Furthermore, hepatotoxicity would be an issue of VPA treatment in HCC patients 60 . Saha et al. showed that valproic acid dose-dependently inhibited viability of human HCC cells such as HepG2 and SNU475 but not normal hepatocyte MIHA cells 61 . Another study showed that VPA radiosensitized glioblastoma cells but not normal hippocampal neurons, suggesting the effect of VPA may be limited to malignant cells 25 . Maintenance of liver function is important for treatment of HCC because HCC is mostly developed in liver cirrhosis background. Thus, more comprehensive dose optimization of VPA for radiosensitization will be needed before clinical use for HCC patients.

Conclusion
Our study highlights a radiosensitizing effect of VPA, particularly with proton radiation in HCC treatment in vitro and in vivo. VPA potentiated the effect of photon or proton irradiation in HCC cells, resulting in an increase in proton RBE. As proton is only 10% more effective than photon, it should be interesting to address effect of VPA on carbon beam irradiation which has higher RBE. NRF2 is an emerging prognostic marker of HCC and other cancer types and contributes to radioresistance. Thus, targeting NRF2 with VPA may be a promising strategy for cancer patients who receive proton radiation therapy.

Materials and Methods
Cell culture. The HCC cell lines, Hep3B and Huh7 were purchased from Korean Cell Line Bank and were cultured in DMEM medium supplemented with 10% fetal bovine serum and 1x Antibiotic-Antimycotic (Gibco, Carlsbed, CA, USA). All cultures were maintained at a humidified 37 °C incubator with 5% CO 2 atmosphere and were routinely passaged every 2-3 days.
Irradiation of photon and proton. Cells were seeded into a 6-well plate or a 10-cm culture plate and then were irradiated with either photon or proton beam next day. Photon beam irradiation was performed with a linear accelerator Varian Clinac 6EX machine (Varian Medical Systems, Palo Alto, CA, USA) at Samsung Medical Center. The cell dishes were placed under 2 cm thickness solid water phantom with source surface distance of 100 cm and field size of 30 × 30 cm and were irradiated with 6-MV photons at a dose rate of 3.96 Gy per min. The photon absolute dose was calibrated according to TG-51 and verified with Gafchromic film to 1% accuracy.
Proton beam irradiation was performed with a proton therapy machine of Samsung Proton Therapy Center in Seoul, South Korea 11 . Cells were irradiated with 230 MeV proton beam with wobbling method 62 and the field was collimated by 18 × 12 cm Brass block. The range of proton beam was 22.8 cm (distal 90%) and the spread-out Bragg peak (SOBP) width was 11.2 cm (distal 90% to proximal 95%). The irradiation point of cell dishes was the middle of SOBP (17.2 cm) and a dose rate of 2.14 Gy per min. To minimize the dose uncertainty, the absolute dose was verified according to TRS-398 for proton therapy to 1% accuracy. The graded doses of 2, 4, 6 and 10 Gy were delivered by application of anterior beams of photon or proton irradiation.
Cell proliferation assay. Colorimetric MTT (thiazolyl blue tetrazolium bromide) metabolic activity assay was used to determine proliferation of cells treated with VPA. Cells were seeded at 1 × 10 3 cells/well into a 96-well plate and on the next day the media were replaced with fresh media containing 0, 0.25, 0.5, 1, 2 and 5 mM VPA. After 48 h incubation, MTT solution (0.5 mg/ml) was added and population of metabolically active cells was calculated based on formazan formation, which was monitored by measuring the absorbance at 540 nm using SpectraMax i3 microplate reader (Molecular Devices, Sunnyvale, CA, USA). The cell proliferation was calculated as percentage of untreated control.
Clonogenic assay. Cells were seeded into a 6-well plate in triplicate (300 cells/well for 0 and 2 Gy, 600 cells/ well for 4 Gy and 1000 cells/well for 6 Gy) and then incubated overnight. The cells were pre-treated with 1 mM VPA for 3 h and were irradiated with either photons or protons of doses of 0, 2, 4 and 6 Gy. After incubating for 14 days, cells were fixed and stained with methanol containing 1% crystal violet and colonies containing 50 or more cells were counted. Plating efficiency was calculated as % of colonies from seeded cells and cell survival at each irradiation dose was determined by dividing the plating efficiency of the irradiated cells by that of the mock-irradiated control. The survival curves were fitted and analysed using linear-quadratic model [SF = exp(− αD − βD 2 )] using GraphPad Prism 7.02 (GraphPad Software, La Jolla, CA, USA); SF is the survival fraction and D is the absorbed dose. Proton RBE was calculated as the ratio of the physical doses of photon radiation and protons that correspond to the survival fraction of 0.1.
Western blot analysis. Cells irradiated with photon or proton were harvested in a lysis buffer and an equal amount of proteins was subjected to SDS-PAGE. Proteins were transferred onto nitrocellulose membrane and blots were probed with antibodies specific for the indicated proteins. The protein bands were visualized by Amersham enhanced chemiluminescence reaction kit (GE healthcare, Piscataway, NJ, USA) and exposure to X-ray film.
Immunofluorescence staining. 2 × 10 4 Hep3B cells were seeded on a cover glass (Marinfild Inc., Rochester, NY, USA) a day before irradiation. The cells were pre-treated with VPA for 3 h and then exposed to photon or proton and then fixed with 4% formaldehyde at the indicated times. After permeabilized using 0.01% Triton X-100 and blocked with 2% bovine serum albumin for 30 min, cells were incubated with phospho-S193 H2AX antibody for 2 h, followed by incubating with Alexa Fluor 488-conjugated secondary antibody (Life Technologies, Paisley, UK) and 4′,6-diamidino-2-phenylindole (DAPI) for 30 min at room temperature. After mounting cover glass with SlowFade anti-fade reagent (Molecular Probes, Eugene, OR, USA), immunofluorescent images were acquired using fluorescent microscope (Zeiss Observer D1, Carl Zeiss Co., Ltd., Jena, Germany).
Reactive oxygen species (ROS) measurement. Cells were pre-treated with 1 mM VPA for 3 h and then incubated with 20 μM Carboxy-H2DCFDA for 30 min, followed by exposure to 6 Gy of photon or proton irradiation. After 72 h post-irradiation, the cells were harvested and ROS level was analysed by flow cytometry (FACSVerse, Becton-Dickinson, CA, USA).
Cell cycle analysis. Cell cycle analysis was performed by flow cytometry. 2 × 10 5 cells were plated in 6-well plates and allowed to attach overnight. Then, cells were pre-treated with VPA for 3 h and then exposed to photon or proton. At 48 h of post-irradiation, cells were collected, fixed with pre-chilled 70% ethanol and washed with phosphate-buffered saline (PBS). The cells were resuspended in 1 ml of PBS containing 1 mg/ml RNase and 50 μg/ ml propidium iodide, incubated in the dark for 30 min at 37 °C, and analysed by flow cytometry.
shRNA transfection. Transient transfection with short hairpin RNAs was performed to knock down NRF2 expression. The shRNA plasmid for NRF2 depletion was purchased from Sigma (TRCN0000273494,) and its sequence is: 5′-CCGGAGTTTGGGAGGAGCTATTATCCTCGAGGATAATAGCTCCTCCCAAACTTTTTTG-3′. For transfection, Hep3B cells (1 × 10 5 cells) were seeded on 6 well plates and incubated with a mixture of 2 µg of NRF2 shRNAs and 10 µl of Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) for 4 h and then the media was replaced with fresh one. Depletion of NRF2 in Hep3B was confirmed by western blot analysis using anti-NRF2 antibody.
Animal experiments. Six to seven-week-old female Balb/c nude mice were purchased from Orient Bio (Gapyeong, South Korea). 5 × 10 6 Hep3B cells were injected subcutaneously into the right hind leg. When the mean tumor volume reached 50~100 mm 3 , mice were randomized to six groups (n = 6) to receive the following treatments: (a) PBS, 0.1 mL, (b) VPA, (c) photon irradiation, (d) VPA + photon irradiation, (e) proton irradiation and (f) VPA + proton irradiation. Mice were pre-treated with intraperitoneal injections of VPA (300 mg/ kg/day) for 4 days. 300 mg/kg/day of VPA is within an acceptable range to administrate in mice 25,59 and no body weight change was seen in mice. From the 2nd day of drug treatment, the tumor-bearing right hind legs were Scientific REPORTS | 7: 14986 | DOI:10.1038/s41598-017-15165-3 irradiated with photons or protons at a dose of 3 Gy per fraction to a total dose of 9 Gy for 3 consecutive days. After irradiation, injection of VPA continued every 3 days until the day before sacrifice to keep the effect of VPA active in the mice. Tumor volumes were measured every 3 days with calipers and calculated according to following formula: volume = L × (W) 2 × 1/2 (L, length in mm; W, width in mm) 27,63,64 . The animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Samsung Biomedical Research Institute, which is accredited by an Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC, protocol number H-A9-003). All experiments were performed in accordance with relevant guidelines and regulations. TUNEL assays. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed to detect apoptosis in formalin-fixed paraffin-embedded (FFPE) tissues. Irradiated tumor tissues were fixed with 10% neutral buffered formalin (NBF) for 4 h and embedded in paraffin. After deparaffinization, TUNEL staining was performed using In Situ Cell Death Detection Kit (Roche Diagnostics, Mannheim, Germany). Images were captured using an Aperio ScanScope AT slide scanner (Leica Biosystems Inc. Buffalo Grove, Illinois, USA) and analysed using ImageScope software (Leica Biosystems).

Immunohistochemistry.
To evaluate expression of NRF2 in tumour tissues, immunohistochemistry (IHC) was performed. The sections sliced into 4 µm were deparaffinized in xylene, rehydrated in graded alcohol, and transferred to 0.01 M PBS, pH 7.4. After heat induced epitope retrieval (HIER) with citrate buffer (pH 6.0; Dako, Carpinteria, CA) for 3 min at 121 °C to reveal hidden antigen epitopes, endogenous peroxidase was blocked with 3% hydrogen peroxide in PBS for 10 min at room temperature. After washing in PBS buffer, sections were treated with serum free blocking solution (Dako) for 20 minutes at room temperature to block nonspecific binding. Subsequently, sections were incubated with anti-Nrf2 rabbit polyclonal antibody (1/100; Abcam, Cambridge, UK) overnight at 4 °C. After washing in PBS, the sections were incubated for 30 minutes at room temperature with HRP-labelled polymer conjugated secondary antibodies against mouse IgG (Dako) or rabbit IgG (Dako). The colour reaction was developed using the ready-to-use DAB (3,3′-diaminobenzidine) substrate-chromogen solution (Dako) for 5 minutes and then washed with distilled water. Finally, sections were lightly counterstained with Mayer's haematoxylin for 30 seconds before dehydration and mounting. Slides were scanned with Aperio ScanScope AT slide scanner (Leica Biosystems Inc. Buffalo Grove, Illinois, USA) at 20× magnification and analysed using ImageScope software (Leica Biosystems). Pixel counts were gated to strongly positive pixel counts using the Positive Pixel Count v9 (PPCv9) algorithm embedded in the program. Statistical analysis. All data was expressed as the mean ± SD from at least three independent experiments. Statistical analysis was performed using GraphPad Prism 7.02. Statistical significance was determined by unpaired, two-tailed Student's t-test and p < 0.05 was considered statistically significant.