JmjC-KDMs KDM3A and KDM6B modulate radioresistance under hypoxic conditions in esophageal squamous cell carcinoma

Esophageal squamous cell carcinoma (ESCC), the most frequent esophageal cancer (EC) subtype, entails dismal prognosis. Hypoxia, a common feature of advanced ESCC, is involved in resistance to radiotherapy (RT). RT response in hypoxia might be modulated through epigenetic mechanisms, constituting novel targets to improve patient outcome. Post-translational methylation in histone can be partially modulated by histone lysine demethylases (KDMs), which specifically removes methyl groups in certain lysine residues. KDMs deregulation was associated with tumor aggressiveness and therapy failure. Thus, we sought to unveil the role of Jumonji C domain histone lysine demethylases (JmjC-KDMs) in ESCC radioresistance acquisition. The effectiveness of RT upon ESCC cells under hypoxic conditions was assessed by colony formation assay. KDM3A/KDM6B expression, and respective H3K9me2 and H3K27me3 target marks, were evaluated by RT-qPCR, Western blot, and immunofluorescence. Effect of JmjC-KDM inhibitor IOX1, as well as KDM3A knockdown, in in vitro functional cell behavior and RT response was assessed in ESCC under hypoxic conditions. In vivo effect of combined IOX1 and ionizing radiation treatment was evaluated in ESCC cells using CAM assay. KDM3A, KDM6B, HIF-1α, and CAIX immunoexpression was assessed in primary ESCC and normal esophagus. Herein, we found that hypoxia promoted ESCC radioresistance through increased KDM3A/KDM6B expression, enhancing cell survival and migration and decreasing DNA damage and apoptosis, in vitro. Exposure to IOX1 reverted these features, increasing ESCC radiosensitivity and decreasing ESCC microtumors size, in vivo. KDM3A was upregulated in ESCC tissues compared to the normal esophagus, associating and colocalizing with hypoxic markers (HIF-1α and CAIX). Therefore, KDM3A upregulation in ESCC cell lines and primary tumors associated with hypoxia, playing a critical role in EC aggressiveness and radioresistance. KDM3A targeting, concomitant with conventional RT, constitutes a promising strategy to improve ESCC patients’ survival.


Introduction
Esophageal cancer (EC) is the eighth most common cancer worldwide and the sixth most common cause of death from cancer 1,2 . Esophageal squamous cell carcinoma (ESCC) is the most common histological subtype [3][4][5] . Although most patients are diagnosed with a loco-regional disease, surgery remains the cornerstone of curative-intent treatment, despite the high morbidity and mortality rates 6,7 . Indeed, overall 5-year survival rates do not exceed 15-20% 8 . In addition to surgery, radiotherapy (RT) is often used as the first-line treatment of EC, both as the main therapeutic strategy or in neoadjuvant context, combined with chemotherapy, entailing similar survival rates in advanced ESSC 7,9 .

Hypoxia decreases RT response in ESCC
In both in vitro hypoxic conditions, HIF-1α chemical induction with 50 µM CoCl 2 or 0.5-1% of O 2 levels, nuclear HIF-1α and cell membrane CAIX expression were increased in all cell lines, although a more impressive effect was observed in Kyse-30 and OE21 cells [which did not express these proteins in normoxia (21% O 2 levels)] ( Supplementary Fig. S1A, B).

IOX1 and KDM3A knockdown increases ESCC RT response
ESCC cells treatment with 50 µM of IOX1 increased H3K9me2 and slightly alter H3K27me3 levels (Fig. 3a, b), whereas decreased protein expression of KDM3A and KDM6B was also apparent in both in vitro hypoxic conditions (Fig. 3b).
Remarkably, for all ESCC cells, IOX1 treatment combined with 2 Gy irradiation significantly increased the % of global DNA fragmentation and cell apoptosis for most of the time points, compared to 50 µM CoCl 2 and 0.5-1% O 2 conditions (Fig. 3d, e and Supplementary Fig. S2B). Furthermore, cell migration capability was significantly decreased in IOX1 treated cells (Fig. 3f). Of note, 50 µM IOX1 effect was significantly lower in normal esophageal Het-1A cell line than in ESCC cell lines ( Supplementary  Fig. S3), both for cell viability and apoptosis (Supplementary Fig. S3).
To unveil whether KDM3A is implicated in ESCC radioresponse, KDM3A knockdown (KDM3A-KD) was performed in the Kyse-410 cell line (Fig. 4). KDM3A and (see figure on previous page) Fig. 1 Effect of hypoxia on RT response, DNA damage, cell migration, and apoptosis. a Cell surviving fraction in three ESCC cell lines irradiated with [0-8] Gy range concentration under normoxia, 50 µM CoCl 2 and hypoxia through SHMT model analysis. Results are presented as mean±SD of at least 3 independent experiments. b DNA damage of 2 Gy irradiated ESCC cells between 0 and 24h, under normoxia, 50 µM CoCl 2 and hypoxia by comet assay. The results are the mean of at least 50 comets per condition. All values of DNA fragmentation (tail moment) were normalized to control (0Gy). Further, hypoxic conditions (50µM CoCl 2 or 0.5-1% O 2 ) were compared to normoxia. c Representative pictures of nuclear γ-H2AX staining of 2 Gy irradiated ESCC cells under normoxia, 50 µM CoCl 2 and hypoxia conditions. All pictures were taken from Olympus IX51 microscope at ×400 magnification (scale bar 20 μm). IF quantification was done using ImageJ software (version 1.6.1, from National Institutes of Health) and represented as a fold change between 2 Gy irradiated cells and non-irradiated control. IF, fluorescence intensity. d ESCC cell apoptosis under normoxia, 50 µM CoCl 2 and hypoxia conditions with 2 Gy of IR. Results are presented as mean±SD of at least three independent experiments. e ESCC cell migration through wound-healing assay, after 24 h of 2 Gy treatment normalized 0 h. Results are the mean±SD of at least three independent experiments; Irradiated cells are compared to non-irradiated cells in each condition. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

KDMs activity inhibition impairs DSB DNA damage repair after ionizing radiation exposure
To further understand whether KDMs activity inhibition with IOX1 treatment affects DNA damage repair (DDR), enriched γ-H2AX foci were found after 0.5 h 2 Gy IR until 24 h in all ESCC cells treated with IOX1, under hypoxic conditions (50 µM CoCl 2 and 0.5-1% O 2 ), indicating increased DSBs and less DDR (Fig. 5a). Accordingly, reduced DDR effectors relative protein levels were observed in hypoxic-IOX1 ESCC treated cells after 2 Gy IR (Fig. 5b, c). Of note, both homologous recombination (HR) and non-homologous end-joining (NHEJ) repair pathways were disturbed (Fig. 5c). Phospho-ATM (γ-ATM) and DNA-PKcs are common DNA damage kinases activated in response to DSBs formation in each repair pathway, HR and NHEJ, respectively 32 . Herein we found reduced γ-ATM and DNA-PKcs protein levels after 24 h of 2 Gy IR exposure in hypoxic-IOX1 treated ESCC cells compared with respective hypoxic controls, suggesting a DDR network deficiency (Fig. 5b, c). Although DNA-PKcs activation was maintained after 24 h of 2 Gy IR it was less pronounced in IOX1 treated cells (Fig. 5b, c). Because DNA-PKcs is known to critically interact with Ku70/80 heterodimer to signalize DDR kinase activity during classic NHEJ repair pathway we tested that in our cells 33 . Overall, hypoxic-IOX1 treated cells displayed less DNA-PKcs activation, although subtle differences were found for Ku80 protein expression (except for Kyse-30 cells with 0.5-1% O2 + 50 µM IOX1), suggesting that IOX1 did not consistently influence Ku80 expression (Fig. 5b). Conversely, NHEJ factor 1 or Cernunnos, also known as XRCC4-like factor (XLF), which is another critical core component of NHEJ repair pathway to endure gapfilling 34 , exhibited reduced activation in hypoxic-IOX1 treated cells compared with respective hypoxic controls, after 24 h of 2 Gy IR exposure, with the exception for kyse-410 with 50 µM (CoCl 2 + IOX1) (Fig. 5b, c).
Moreover, IOX1 reduced the relative protein expression of all studied HR components (Fig. 5b, c), in accordance with NHEJ network variations. Additionally, defects on the major DDR mechanisms were more evident with IOX1 addition in hypoxic-ESCC cells (Fig. 5b, c).
Finally, IOX1 strongly reduced protein levels of both Mre11 and p95/NBS1 after 24 h of IR, suggesting a compromised DDR complex activity and further supporting previous results (Fig. 5b, c). Taken together, these findings reveal that DDR proteins hypoxic-dependent modulation after IR exposure was mainly abrogated with JmjC-KDMs activity inhibition by IOX1, in all ESCC cell lines.
Additionally, an overall decrease of Ki-67 and γ-p53 expression was displayed by CAM microtumors treated with IOX1 alone or combined with radiation, compared with the respective control ( Fig. 6e and Supplementary  Fig. S4). Conversely, increased cleaved caspase 3 and γ-H2AX expression was depicted by treated CAM microtumors, although no differences were apparent between tumors exposed to combined treatment or irradiated only ( Fig. 6e and Supplementary Fig. S4).
Finally, KDM3A and KDM6B decreased expression was accomplished in CAM tumors treated with IOX1 inhibitor ( Fig. 6e and Supplementary Fig. S4), although only H3K9me2 was significantly increased in target marks ( Fig. 6e and Supplementary Fig. S4), supporting in vitro findings, and thus, indicating KDM3A as a key targetable molecule to radiosensitize hypoxic ESCC.

Discussion
ESCC is highly incident worldwide, entailing poor prognosis and low overall survival rates 1,35 , despite some therapeutic advances over the last years 7,36 . Thus, the identification of new molecular targets that might improve therapeutic efficacy for advanced ESCC is urgently needed. Hypoxia has been associated with poor prognosis in EC, namely due to resistance to RT, the gold standard therapy for advanced stages 12,37 . Indeed, EC patients with nonhypoxic tumors, displaying HIF-1α downregulation, endure complete chemoradiotherapy response, contrary to patients  Recently, hypoxia has been associated with epigenetic deregulation 21 . Indeed, histone lysine demethylases, namely JmjC-KDMs superfamily was found to be regulated by oxygen levels and/or HIF-1α transcription factor 38 . Specifically, both low O 2 levels and HIF-1α expression were reported to induce KDM3A and KDM6B expression 23 . Still, HIFs might be regulated by 2oxoglutarate-dependent members, which involves JmjC-KDMs family, as well as, prolyl hydroxylases (PHD) 23 . Accordingly, in our hands reduced nuclear HIF-1α expression levels were found in KDM3A-KD cells.
Furthermore, nuclear HIF-1α expression levels were not significantly altered in IOX1 CAM-associated microtumors, whereas a significant reduction was found after 2 Gy IR and after combined treatment. This was followed by decreased tumor volume and consequently, decreased hypoxic foci. Additionally, IOX1 was also shown to have a higher selectivity for Pan-JmjC-KDMs activity inhibition than PHDs 39 . Indeed, a similar HIF-1α expression trend in IOX1 Kyse-410 microtumors may be partially explained by intrinsic expression levels in the control condition and the apparent hypoxic foci in 3D tumors.
Additionally, findings from in vitro assays using different cell lines suggested that under hypoxic conditions, HIF-1α recruits KDM3A, promoting H3K9me2 demethylation, and increasing gene transcription 40 . Remarkably, we showed that in ESCC cells, both KDM3A and KDM6B expression was upregulated in parallel with low oxygen levels and /or HIF-1α overexpression, with the latter specifically bound to KDM3A and KDM6B promoter region under hypoxic conditions. Thus, our findings both confirm and extend previously published observations.
Additionally, several studies suggested that both KDM3A and KDM6B were putative therapeutic targets in different cancer models, not including ESCC. Indeed, KDM3A inhibition decreased estrogen receptor positive breast cancer cells' proliferation 41 whereas it was implicated in stemness and chemoresistance in ovarian cancer 42 . Furthermore, KDM3A targeting increased response to anti-angiogenic therapies, disclosing a role in tumor angiogenesis 43 Interestingly, in vitro studies in lung and breast cancer cells demonstrated that KDM6A inhibition decreased cell survival and improved RT response, through H3K27me3 enhancement 44 . Furthermore, two other KDMs, KDM4C, and PHF8 were associated with ESCC malignant features. Indeed, KDM4C targeting decreased ESCC stemness properties 45 , whereas PHF8's inhibition promoted apoptosis and decreased ESCC cell proliferation and invasion 46 . Interestingly, our in vitro data we demonstrated that radioresistant phenotype observed under hypoxic conditions was abrogated with both JmjC-KDMs activity inhibition and KDM3A-KD, promoting radiosensitization in ESCC cells, in line with the results obtained with IOX1 inhibitor. Of note, hypoxic-dependent KDM3A seems to play a critical role in RT response modulation in in vitro and in vivo experiments. Furthermore, radiosensitized hypoxic-IOX1 treated ESCC cells impaired DDR network, with decreased relative protein levels of the major DDR effectors. In the same vein and as a consequence of DNA damage repair deficiency, γ-H2AX was independently maintained overtime after cell replication in hypoxic-induced IOX1 cells, suggesting the persistence unrepaired DNA DSBs. Remarkably, a similar function of a KDM5B inhibitor, JIB-04, was reported to radiosensitize lung cancer 47 . Defects in DNA repair dynamics prevents DDR resolution, due to endless γ-H2AX activation and impaired recruitment of the major HR and NHEJ repair effectors 47 .
Those results were further supported by in vivo experiments using the CAM assay. Additionally, the reduction on microtumor aggressiveness features was

Irradiation
IR was performed at room temperature (R/T) with normal oxygen levels, using TrueBeam linear accelerator as irradiation source within a field of 25×25 cm 2 , a photon energy of 6MV and a dose rate of 600 MU/min. For hypoxic experiments (0.5-1% O 2 ) cells were maintained under oxygen deprivation during all experimental timeline, before and after IR radiation exposure for all in vitro assays. Nonetheless, In vivo chicken CAM irradiation was carried out in a microSelectronv3 Iridio-192 brachytherapy (192-Ir-mHDR-v2r) at 2 Gy per egg/pulse for chicken embryo's protection (Image planning represented on Supplementary  Fig. S5) since only a low dose rate brachytherapy was attained by the animal.

RNA extraction, quantification, cDNA synthesis and RT-qPCR
Tumor cell RNA was extracted by a ribozol reagent method. Revert Aid RT Kit (ThermoScientific Inc.) was used for cDNA synthesis, according to the manufacturer's instructions. RT-qPCR was performed in Light-Cycler480II (Roche) using Xpert Fast SYBER Mastermix Blue (GE22.2501, Grisp) with specific designed primers (Supplementary Table S3). GUSβ was used as endogenous control.

Total protein extraction, quantification, and SDS-PAGE western blot
For next experiments all antibodies details are described in Supplementary Table S4.
Briefly, cells were scraped in lysis Buffer (Kinexus Bioinformatics Corporation, Vancouver, British Columbia, Canada) on ice. Protein quantification was performed using the Pierce BCA Protein Kit (Thermo Scientific Inc.), according to the manufacturer's instructions. Western blot (WB) was performed as previously described 53 using specific primary antibodies. After primary antibody incubation overnight at 4°C, specific conjugated horseradish peroxidase secondary antibodies (Bio-Rad, USA) were incubated 1 h R/T. Chemiluminescence was detected with Clarity WB ECL substrate (Bio-Rad, USA) and evaluated using ImageJ software (version 1.6.1, from National Institutes of Health). β-actin served as control of the total loaded protein.

Immunofluorescence and immunocytochemistry
ESCC cells were seeded in cover slips into culture plates and fixed with 4% paraformaldehyde. For nuclear proteins, cells were permeabilized with 0.25% Triton X-100 solution in 1x phosphate-buffer saline (PBS 1x).
Phenotypic assays IOX1 effects on cell viability were assessed by MTT assay, following previously reported procedures 54 . Apoptosis was evaluated after 24 h of 2 Gy IR and after 48 h and 72 h of IOX1 induction and hypoxic stimulation, respectively, using APOPercentage assay kit (Biocolor Ltd., Belfast, Northern Ireland, UK), according to manufactured instructions. Concerning wound-healing assay, wild-type (WT) ESCC cells or Kyse-410 KDM3A-KD and scramble were seeded and exposed to 50 µM CoCl 2 or hypoxia. Subsequently, when applicable, cells were treated with 50 µM IOX1 24 h before 2 Gy IR. Then, cells growth at 95% of confluence and two parallel "wounds" in each well (initial slope) were done. Then, relative migration distance was analyzed by beWound -Cell Migration Tool (version 1.5) calculating % cell migration = (A/B)/C*100 (A, width of cell wound at initial slope; B, width of cell wound at several time points; C, width mean of cell wound at initial slope)]. KDM3A-KD cell proliferation assay was assessed after 24 h of 2 Gy IR and 48 h of 50 µM CoCl 2 and hypoxia induction, using Cell proliferation ELISA BrdU (5-bromo-2'deoxyuridine) assay kit (Roche Applied Sciences, Penzberg, Germany), according to manufactured instructions.

Colony formation
ESCC cells were seeded in 6-well culture plates at specific concentrations for each experimental group, as detailed in Supplementary Table S1. Then, after 48 h of hypoxia exposure or CoCl 2 addition, cells were exposed to IR and incubated at 37°C for 7 days. Experiments were carried out in all ESCC cell lines, whereas Kyse-30 cell line was not able to form colonies after CoCl 2 chemical induction. Also, 24 h after hypoxia stimulation and before IR, cells were treated with 20 µM IOX1. Colonies were stained with 25% (w/v) Giemsa. Colonies depicting more than 50 single cells were counted and analyzed using RAD ADAPT software (Biomedical Simulation Resource, USC, California, USA) 55 . Exponential single hit multi-target model (SHMT), S(D) = PE * [1-(1-exp (-D/D 0 )) n ) was used. Concerning statistics, D 0 represents the induction of one lethal event per cell becoming at 37% of viability, through the measurement of the ending slope resulting from a multiple event killing. Furthermore, D q , quasi-threshold dose represents the width of the curve shoulder. Additionally, sensitized enhancement ratio (SER) was evaluated according to D 0 (without sensitizer) / D 0 (with sensitizer). The sensitizer is IOX1 inhibitor or Kyse-410 KDM3A-KD cells.
Alkaline comet assay ESCC cells were treated with 2 Gy IR at 0 h, 0.5 h, 2 h, and 24 h. Briefly, cells were re-suspended in 0.5% low melting agarose (w/v) and immediately placed on a sheet previously covered with 1% normal melting agarose (w/v). Then, a cell lysis buffer (2.5 M NaCl, 100 mM Na 2 EDTA, 10 mM Tris Base, 1% Triton X-100), pH 10 was added. Electrophoresis was performed for 30 min at 21 V, 300 mA, 4°C. Lastly, the sheets were incubated in a neutralization buffer (0.4 M Tris-Base, pH 7.5,), followed staining with SybrGreen. Comet analysis was done using OpenComet v.1.3.1 56 . Global DNA damage (SSB and DSB) 57 evaluation was determined by measuring tail moment (tail % DNA x means of head x tail distance) and representative pictures taken with Olympus IX51 microscope at ×200 magnification. A sampling of at least 50 comets was included in the analysis.

CAM assay
Fresh fertilized eggs (PintoBar, Lda, Portugal) were incubated at 37°C in a humid environment. After 6 days of embryonic development, a window was opened into the eggshell under aseptic conditions. On day 10, Kyse-410 cells suspension in growth factor-reduced Matrigel (BD Biosciences) were seeded on CAM. Then, on day 13, a treated group, randomly selected, received IOX1 50 μM whereas a control group received only 1% DMSO in complete RPMI-1640. After 24 h, CAM was irradiated with 2 Gy. Lastly, on day 17, tumors were dissected and included in a paraffin block. Microtumor images were obtained on day 13 (0 h of treatments) and at day 17 (72 h of treatment). Relative perimeter in in ovo was assessed using CellSens software (version V0116, Olympus). Ex ovo pictures were obtained for blood vessels counting using Image J software.
Immunostaining of microtumors' sections was evaluated through a quantitative method using GenASIS software (Applied Spectral Imaging, ASI). Staining's evaluation was performed as described in the tissue immunoexpression subsection.

Statistics
Non-parametric tests (Kruskal-Wallis or Mann-Whitney U test) among groups with Bonferroni's correction were used to compare different conditions in in vitro assays through GraphPad Prism version 6.0. IHC results were analyzed by Pearson's chi-square or Fisher's exact test, using the SPSS 25.0 software. All results are shown as the mean ± SD for each group. For each analysis, p values were considered significant when inferior to 0.05 (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).