Quantitative dissection of radioresistance in cancer cells: Oxygen depletion-induced biological versus chemical mechanisms and baseline mechanisms governed by HIF-1β-NRF2/TXNRD1 Authors: Satoshi Takeuchi, Minoru Kobayashi, Peter Wai Tik Lee, Itsuki Takahashi, Toshitada Yoshihara, Hiroshi Harada Published: 2026 ABSTRACT The oxygen environment within malignant solid tumors is highly heterogeneous, with anoxic regions arising where oxygen supply is insufficient. It is known that oxygen depletion is one of the most influential factors that confer radioresistance on cancer cells. The resistance under low oxygen conditions has been attributed to both hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms resulting from reduced production of radiation-induced reactive oxygen species. However, the relative contributions of these mechanisms have remained unclear. Moreover, although hypoxia-inducible factors (HIFs) are suggested to play key roles in hypoxia-inducible genes-dependent biological mechanisms, it has remained uncertain whether and how HIFs affect the intrinsic radioresistance of cancer cells. Here, we developed an approach to separately analyze the mechanisms behind radioresistance under anoxia as either hypoxia-inducible genes-dependent biological or genes-independent chemical mechanisms and found that the relative contributions in HeLa cells were 64.4 ± 9.06% and 35.6 ± 4.32%, respectively. In addition, HIF-1β alone, rather than heterodimeric HIF-1, HIF-2, or HIF-1β/AhR, was found to enhance the baseline radioresistance of cancer cells, irrespective of oxygen availability. RNA-sequencing-based screening further identified NRF2 and TXNRD1 as downstream effectors of HIF-1β underlying this baseline radioresistance. Taken together, these findings delineate the quantitative contributions of distinct mechanisms and identify HIF-1β-mediated signaling as a key determinant of baseline radioresistance, providing a framework for understanding how oxygen availability shapes tumor radioresistance and, in turn, a basis for developing strategies to overcome it. Introduction Advances in radiation therapy technology have enabled the delivery of high radiation doses to tumor-specific regions while minimizing exposure to surrounding normal tissues. However, the presence of cancer cells that acquire radioresistance within tumor microenvironments remains a major obstacle to curative outcomes of radiation therapy, as these cells cause local recurrence and distant metastasis [[1], [2], [3], [4], [5]]. A growing body of evidence indicates that various endogenous and exogenous factors can induce radioresistance in cancer cells, with low oxygen conditions (including hypoxia and anoxia) being one of the most important [[6], [7], [8], [9]]. The oxygen distribution within malignant solid tumors is highly heterogeneous [2,10] (Supplementary Figure S1). Cancer cells located near tumor blood vessels receive sufficient oxygen, thereby forming normoxic regions [2]. As oxygen is consumed by these cells, the partial pressure of oxygen decreases with increasing distance from the vessels, generating hypoxic regions approximately 70-100 μm away from vessels [2]. At greater distances, oxygen gradients reach peri‑necrotic regions that are anoxic and adjacent to necrosis. In these chronically hypoxic to anoxic regions, expression of the so-called hypoxia-inducible genes is upregulated, reflecting cellular adaptation to oxygen deficiency. In addition to such chronically hypoxic and anoxic regions, acutely or transiently hypoxic or anoxic regions can also arise, even in close proximity to blood vessels, due to temporary vascular occlusion or fluctuations in blood flow [2]. In these regions, on the other hand, the duration of oxygen deprivation is too short to induce the expression of hypoxia-inducible genes. In normoxic regions, reactive oxygen species (ROS) produced by ionizing radiation are stabilized through reactions with oxygen, leading to the formation of long-lived radicals that can cause DNA damage over extended periods [11,12]. Furthermore, oxygen molecules react with DNA radicals at sites of damage generated by irradiation, converting them into chemically stable and irreparable lesions, thereby enhancing the cytotoxic effect of radiation [11,12]. In contrast, in the absence of oxygen, oxygen-mediated stabilization of ROS and fixation of DNA damage do not occur [13]. Consequently, low oxygen conditions reduce DNA damage mediated by these radiation-chemical reactions, leading to the radioresistance of cancer cells. In addition to these low oxygen-dependent chemical mechanisms, there are specific biological mechanisms that are activated under low oxygen conditions and contribute to radioresistance [4,14,15]. Among these, one well-characterized example is the hypoxia-inducible genes-dependent mechanisms mediated by hypoxia-inducible factor 1 (HIF-1), a master transcription factor for adaptation to low oxygen conditions [1]. HIF-1 belongs to the HIF family, which consists of HIF-1, HIF-2, and HIF-3. Each of them is a heterodimer composed of a unique regulatory subunit, HIF-1α, HIF-2α, and HIF-3α, respectively, and a common subunit shared by all three, HIF-1β (also known as the aryl hydrocarbon receptor nuclear translocator, ARNT) [10,16]. In vivo studies using tumor-bearing mouse models have shown that pharmacological or genetic inhibition of HIF-1 enhances the therapeutic effect of radiation and significantly delays tumor growth after radiation therapy [17,18]. In this line of research, our laboratory previously elucidated the underlying mechanism: activation of HIF-1 under low oxygen conditions induces VEGF expression in cancer cells, and the secreted VEGF protects vascular endothelial cells from radiation-induced damage, thereby attenuating the efficacy of radiotherapy [19]. However, whether and how HIF family proteins affect the radioresistance of cancer cells per se remain poorly understood. Beyond the HIF-1-dependent pathways, HIF-1-independent pathways have also been implicated in hypoxia-inducible genes-dependent biological mechanisms for radioresistance. For instance, a previous study using immortalized fibroblasts showed that exposure to hypoxia induces expression of cell cycle regulators independently of HIF-1, resulting in cell cycle arrest [20]. According to the Bergonié-Tribondeau law in radiation biology, cells with high proliferative capacity are generally more radiosensitive; thus, hypoxia-induced cell cycle arrest could potentially enhance radioresistance in cancer cells [21]. In addition, there exists another category of biological mechanisms that is independent of hypoxia-inducible genes and oxygen-availability. Cancer cells exhibit different baseline levels of radioresistance, determined by their genetic and epigenetic backgrounds, for example by the expression of genes involved in antioxidant systems (e.g., glutathione biosynthesis and the thioredoxin system) and DNA damage response and repair pathways. Importantly, this baseline resistance is independent of hypoxia-inducible pathways and therefore holds true regardless of oxygen availability [13]; however, its precise mechanisms also remain poorly understood. Overall, the available evidence suggests that each cancer cell exhibits its own baseline radioresistance depending on its genetic and epigenetic backgrounds, regardless of oxygen availability. Furthermore, specifically under low oxygen conditions, cancer radioresistance is augmented by both hypoxia-inducible genes-independent chemical mechanisms and hypoxia-inducible genes-dependent biological mechanisms (via HIF-1-dependent or -independent pathways), with the latter mechanisms operating only when hypoxia-inducible genes are sufficiently expressed under chronically low oxygen conditions (Supplementary Figure S1). However, the relative importance of these two mechanisms under chronically anoxic conditions is still largely unknown. How do cancer cells acquire radioresistance within the extremely heterogeneous oxygen environment of tumor tissue? Cancer cells in normoxic regions exhibit baseline radioresistance, which is primarily determined by their genetic and epigenetic backgrounds in a manner independent of hypoxia-inducible genes (Supplementary Figure S1). In chronically hypoxic or anoxic regions, on the other hand, cancer cells are thought to additionally acquire radioresistance through both hypoxia-inducible genes-independent chemical mechanisms and hypoxia-inducible genes-dependent biological mechanisms (Supplementary Figure S1) [2]. In contrast, in acutely hypoxic or anoxic regions, although oxygen tension is comparable to that observed in chronic regions, the exposure of cancer cells to low oxygen conditions is too brief to induce the expression of hypoxia-inducible genes [2]. Consequently, hypoxia-inducible genes-dependent biological mechanisms are unlikely to play a major role, whereas hypoxia-inducible genes-independent chemical mechanisms, together with baseline mechanisms, are presumed to be more important (Supplementary Figure S1). Nevertheless, the mechanisms underlying baseline radioresistance remain poorly understood. In the present study, we established a rational approach to distinguish hypoxia-inducible genes-dependent biological mechanisms from genes-independent chemical mechanisms, and quantitatively assessed their relative contributions to radioresistance under anoxia. Moreover, our mechanistic analyses revealed that HIF-1β alone, rather than heterodimeric HIF-1, HIF-2, and HIF-1β/AhR, enhances the intrinsic radioresistance of cancer cells. RNA sequencing-based screening experiments identified NRF2 and TXNRD1 as effectors involved in HIF-1β-mediated, but hypoxia-inducible genes-independent, baseline radioresistance of cancer cells. Results Establishment of in vitro models recapitulating acute and chronic tumor anoxia As a first step, we sought to optimize in vitro experimental conditions that could mimic two intratumoral oxygen microenvironments: 1) chronically anoxic regions, where chronic anoxia-induced radioresistance is thought to be augmented by both hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms; and 2) acutely anoxic regions, where acute anoxia-induced radioresistance is assumed to be augmented mainly by hypoxia-inducible genes-independent chemical mechanisms. In both regions, such anoxia-dependent radioresistance is superimposed on the baseline radioresistance determined by the intrinsic genetic and epigenetic backgrounds of individual cancer cells. We first confirmed that the expression of carbonic anhydrase 9 (CA9), a representative hypoxia-inducible gene, was markedly induced by anoxic culture at < 0.1% O2 for 16 h, at both the mRNA and protein levels (Fig. 1A, B). A clonogenic cell survival assay further demonstrated that radioresistance of cancer cells was significantly enhanced under this anoxic culture condition, compared with the normoxic condition (20% O₂) (Fig. 1C). Based on these results, we designated this culture condition as the in vitro chronic anoxia model in the subsequent experiments.Fig. 1Establishment of in vitro models recapitulating acute and chronic tumor anoxia.A. Schematic diagram illustrating three oxygen conditions (chronic anoxia, acute anoxia induced by N2 gas flushing, and normoxia) and the experimental timeline.B,C. qPCR (left) and Western blot (right) analyses of relative CA9 expression (B), and a clonogenic cell survival assay following γ-irradiation at the indicated doses using HeLa cells (C) after incubation of HeLa cells under normoxia or chronic anoxia as illustrated in A. n = 3.D. Schematic diagram illustrating the experimental timeline for monitoring intracellular oxygen partial pressure with phosphorescent probe, BTPDM1.E. Phosphorescence intensity in HeLa cells after treatment as illustrated in D. n = 4.F,G. qPCR (left) and Western blot (right) analyses of relative CA9 expression (F) and luciferase-based reporter assay of relative HIF-1 activity in HeLa cells (G) after incubation under normoxia, acute anoxia induced by N2 gas flushing for 30 min, or chronic anoxia as illustrated in A. n = 3.H,I. The same experiments as E and F were performed in H and I, respectively, using SCC7 cells. n = 4 (H) and 3 (I).Data are presented as mean ± SD. ns: not significant, ***p < 0.001, ****p < 0.0001 (Student’s two-tailed t-test (B) or One-way ANOVA with Dunnett’s (E,H) or Tukey’s test (F,G,I)).Fig 1 dummy alt text Establishment of in vitro models recapitulating acute and chronic tumor anoxia. A. Schematic diagram illustrating three oxygen conditions (chronic anoxia, acute anoxia induced by N2 gas flushing, and normoxia) and the experimental timeline. B,C. qPCR (left) and Western blot (right) analyses of relative CA9 expression (B), and a clonogenic cell survival assay following γ-irradiation at the indicated doses using HeLa cells (C) after incubation of HeLa cells under normoxia or chronic anoxia as illustrated in A. n = 3. D. Schematic diagram illustrating the experimental timeline for monitoring intracellular oxygen partial pressure with phosphorescent probe, BTPDM1. E. Phosphorescence intensity in HeLa cells after treatment as illustrated in D. n = 4. F,G. qPCR (left) and Western blot (right) analyses of relative CA9 expression (F) and luciferase-based reporter assay of relative HIF-1 activity in HeLa cells (G) after incubation under normoxia, acute anoxia induced by N2 gas flushing for 30 min, or chronic anoxia as illustrated in A. n = 3. H,I. The same experiments as E and F were performed in H and I, respectively, using SCC7 cells. n = 4 (H) and 3 (I). Data are presented as mean ± SD. ns: not significant, ***p < 0.001, ****p < 0.0001 (Student’s two-tailed t-test (B) or One-way ANOVA with Dunnett’s (E,H) or Tukey’s test (F,G,I)). In order to optimize the culture conditions that mimic acutely anoxic regions, we next aimed to determine the minimum duration of nitrogen (N2) gas flushing required to lower intracellular oxygen partial pressure to the equivalent levels observed in the chronic anoxia model (< 0.1% O2 for 16 h), without inducing hypoxia-inducible gene expression. To assess intracellular oxygen partial pressure, we employed an optical probe, BTPDM1, whose phosphorescence is quenched under oxygenated conditions but not under low oxygen conditions [22]. Intracellular oxygen levels in probe-treated cancer cells were quantified by measuring phosphorescence intensity after various durations of N2 gas flushing (Fig. 1D, E). Phosphorescence intensity increased in a time-dependent manner with longer flushing times and reached a plateau after approximately 20-30 min of flushing. Among the tested conditions, flushing for 30 min was found to yield the most reproducible results, with phosphorescence intensity reaching levels comparable to those observed in cells exposed to 16 h of chronic anoxia (< 0.1% O2) (Fig. 1E). Importantly, the detected phosphorescence was confirmed to originate from probes internalized into cells rather than from residual probes in the culture medium (Supplementary Figure S2). We next analyzed the induction of hypoxia-inducible gene expression by the 30- min N2 gas flushing. Quantitative real-time PCR and Western blotting analyses showed no increase in CA9 mRNA and protein levels, respectively (Fig. 1A, F). Furthermore, a luciferase assay employing the 5HRE-Luc reporter gene, which expresses luciferase in a HIF-1-dependent manner under low oxygen conditions [23], demonstrated that luciferase bioluminescence was not induced by the 30-min N2 gas flushing but was robustly induced by a 16-h exposure to chronic anoxia (Fig. 1A, G). Collectively, these results demonstrate that anoxia achieved by 30-min N2 gas flushing establishes an acute anoxia model, in which intracellular oxygen is sufficiently reduced to levels equivalent to those in the chronic anoxia model, yet hypoxia-inducible gene expression is not upregulated. We further investigated whether this acute anoxia model is applicable to other cell types by conducting analogous experiments with the murine squamous cell carcinoma cell line SCC7 (also known as SCC VII). Consistent with the results in HeLa cells, 30- min N2 gas flushing increased phosphorescence intensity to levels comparable to those of cells exposed to 16 h of chronic anoxia, without inducing CA9 mRNA and protein expression (Fig. 1H, I). Relative contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms to chronically anoxia-induced radioresistance In chronically anoxic regions in tumors, both hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms are considered to contribute to anoxia-induced radioresistance. In contrast, in acutely anoxic regions, hypoxia-inducible genes-independent chemical mechanisms are considered to predominate. Cancer cells exhibit baseline radioresistance depending on their intrinsic backgrounds regardless of oxygen conditions. On this basis, the difference in radioresistance between normoxic and acutely anoxic regions is attributed to hypoxia-inducible genes-independent chemical mechanisms, while that between acutely and chronically anoxic regions is attributed mainly to hypoxia-inducible genes-dependent biological mechanisms. To quantify the relative contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms to cancer cell radioresistance in chronically anoxic regions, we next conducted a clonogenic cell survival assay. Cancer cells exhibited increased radioresistance under the two anoxic conditions compared with normoxic conditions (Fig. 2A). As anticipated, HeLa cells under chronic anoxia exhibited greater radioresistance than those under acute anoxia, likely reflecting, at least in part, the contribution of hypoxia-inducible genes-dependent biological mechanisms (Fig. 2A). According to the cell survival data after 8 Gy γ-irradiation presented in Fig. 2B, which are derived from the data in Fig. 2A, 64.4% of the radioresistance acquired by chronic anoxia treatment, which was calculated using the formula described in Materials and Methods as (0.5789 - 0.3476) × 100 / (0.5789 - 0.2199), was attributable to hypoxia-inducible genes-dependent biological mechanisms (Fig. 2B, C). In contrast, the remaining 35.6%, which was also calculated using the formula described in Materials and Methods as (0.3476 - 0.2199) × 100 / (0.5789 - 0.2199), was attributable to hypoxia-inducible genes-independent chemical mechanisms (Fig. 2B, C).Fig. 2Relative contributions of hypoxia inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms to chronically anoxia-induced radioresistance.A. Clonogenic cell survival assay following γ-irradiation at the indicated dose using HeLa cells incubated under normoxia, acute anoxia induced by N2 gas flushing for 30 min, or chronic anoxia as illustrated in Figure 1A. n = 3.B. Surviving fraction at 8 Gy irradiation in the experiment shown in A. n = 3.C. Percentage contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms calculated from the results of B. n = 3.D-F. The same analyses as in A-C were performed using SCC7 cells. n = 3.Data are presented as mean ± SD. ns: not significant, *p < 0.05, **p < 0.01, ****p < 0.0001 (One-way ANOVA with Tukey’s test (B,E)).Fig 2 dummy alt text Relative contributions of hypoxia inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms to chronically anoxia-induced radioresistance. A. Clonogenic cell survival assay following γ-irradiation at the indicated dose using HeLa cells incubated under normoxia, acute anoxia induced by N2 gas flushing for 30 min, or chronic anoxia as illustrated in Figure 1A. n = 3. B. Surviving fraction at 8 Gy irradiation in the experiment shown in A. n = 3. C. Percentage contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms calculated from the results of B. n = 3. D-F. The same analyses as in A-C were performed using SCC7 cells. n = 3. Data are presented as mean ± SD. ns: not significant, *p < 0.05, **p < 0.01, ****p < 0.0001 (One-way ANOVA with Tukey’s test (B,E)). We extended this analysis to another cancer cell line, SCC7, and found that although levels of surviving fraction under each oxygen condition differed from those in HeLa cells, the relative contributions of the two mechanisms showed a similar trend, 70.4% and 29.6%, respectively (Fig. 2D, E, F). To further assess the robustness of these estimates within the same experimental framework, we additionally calculated the relative contributions using survival data obtained after 4 Gy γ-irradiation, derived from the data in Fig. 2A and D (Supplementary Figure S3A, B, D, E). In addition, we performed an isoeffect comparison using D60 values derived from the survival curves, where D60 represents the radiation dose required to reduce the surviving fraction to 60% (Supplementary Figure S3C, F). Although the absolute percentage values varied depending on the radiation dose and calculation method, all analyses consistently demonstrated that hypoxia-inducible genes-dependent biological mechanisms accounted for the major fraction of chronically anoxia-induced radioresistance, whereas hypoxia-inducible genes-independent chemical mechanisms contributed to a lesser extent. Limited role of HIF-1 and HIF-2 in radioresistance of cancer cells per se under chronic anoxia HIF-1 is widely recognized as a master transcription factor mediating cellular adaptation to hypoxia and anoxia. To directly investigate the role of HIF-1 in hypoxia-inducible genes-dependent biological mechanisms under chronic anoxia, we performed loss-of-function studies using the HIF1A knockout (KO) clone [24] (Fig. 3A). Clonogenic cell survival assay at 4 Gy and 8 Gy γ-irradiation showed that HIF1A KO did not affect cell survival under any oxygen conditions and did not alter the relative contributions of hypoxia-inducible genes-dependent biological mechanisms (Fig. 3B, C, D). An isoeffect comparison using D60 values also demonstrated little effect of HIF1A KO on the relative contribution (Fig. 3E). Moreover, reintroduction of HIF-1α expression vector to the KO cells did not enhance radioresistance (Fig. 3F, G, H). In addition, forced expression of a constitutively active mutant of HIF-1α (HIF-1α 3A), which is known to induce the expression of HIF-1 downstream genes even under normoxic conditions, as demonstrated in Fig. 3I, did not enhance cancer cell radioresistance in vitro (Fig. 3J, K).Fig. 3Limited role of HIF-1 and HIF-2 in radioresistance of cancer cells per se under chronic anoxia.A-E. HeLa wild-type cells (WT) and the HIF1A KO clone (HIF1A KO) were cultured under the indicated oxygen conditions, and subjected to Western blotting for the indicated proteins (A) and to a clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (B,C), as illustrated in Figure 1A. Percentage contributions of the hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms at the indicated radiation doses were calculated based on the results shown in C (D) or on the D60 values obtained from the survival curves in B (E). D60 values under normoxia, acute anoxia, and chronic anoxia in WT and HIF1A KO cells are 2.863 Gy, 3.609 Gy, and 7.259 Gy (WT), and 2.879 Gy, 3.635 Gy, and 7.571 Gy (HIF1A KO) respectively. Contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms were calculated as (7.259 - 3.609) × 100 / (7.259 - 2.863) = 83.0% and (3.609- 2.863) × 100 / (7.259 - 2.863) = 17.0% for WT cells, and as (7.571- 3.635) × 100 / (7.571 - 2.879) = 83.9% and (3.635 - 2.879) × 100 / (7.571 - 2.879) = 16.1% for HIF1A KO cells, respectively. n = 3 (B-D).F-H. HeLa HIF1A KO clone (HIF1A KO) stably transfected with HIF-1α expression vector (HIF-1α) or its corresponding empty vector (EV) was cultured under the indicated oxygen conditions and subjected to Western blotting for the indicated proteins (F) and clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (G,H) as illustrated in Figure 1A. n = 3 (G,H).I-K. HeLa cells stably transfected with either a wild-type HIF-1α expression vector (WT), a constitutively active mutant HIF-1α 3A expression vector (3A), or its corresponding empty vector (EV) were cultured under the indicated oxygen conditions and subjected to Western blotting for the indicated proteins (I, left), qPCR for CA9 (I, right), and clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (J,K) as illustrated in Figure 1A. n = 3 (I right, J,K).L-Q. HeLa wild-type cells (WT), HeLa HIF2A (EPAS1) KO clones (KO #1 and #2), and HeLa HIF1A/HIF2A (EPAS1) double KO clones (double KO #1 and #2) were cultured under the indicated oxygen conditions, and subjected to Western blotting for the indicated proteins (L,O) and clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (M,N,P,Q) as illustrated in Figure 1A. n = 3 (M,N,P,Q).Data are represented as mean ± SD. ns: not significant, **p < 0.01, ****p < 0.0001 (One-way ANOVA with Tukey’s (B-D,G,H,J,K,M,N,P,Q) or Dunnett’s test (I)).Fig 3 dummy alt text Limited role of HIF-1 and HIF-2 in radioresistance of cancer cells per se under chronic anoxia. A-E. HeLa wild-type cells (WT) and the HIF1A KO clone (HIF1A KO) were cultured under the indicated oxygen conditions, and subjected to Western blotting for the indicated proteins (A) and to a clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (B,C), as illustrated in Figure 1A. Percentage contributions of the hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms at the indicated radiation doses were calculated based on the results shown in C (D) or on the D60 values obtained from the survival curves in B (E). D60 values under normoxia, acute anoxia, and chronic anoxia in WT and HIF1A KO cells are 2.863 Gy, 3.609 Gy, and 7.259 Gy (WT), and 2.879 Gy, 3.635 Gy, and 7.571 Gy (HIF1A KO) respectively. Contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms were calculated as (7.259 - 3.609) × 100 / (7.259 - 2.863) = 83.0% and (3.609- 2.863) × 100 / (7.259 - 2.863) = 17.0% for WT cells, and as (7.571- 3.635) × 100 / (7.571 - 2.879) = 83.9% and (3.635 - 2.879) × 100 / (7.571 - 2.879) = 16.1% for HIF1A KO cells, respectively. n = 3 (B-D). F-H. HeLa HIF1A KO clone (HIF1A KO) stably transfected with HIF-1α expression vector (HIF-1α) or its corresponding empty vector (EV) was cultured under the indicated oxygen conditions and subjected to Western blotting for the indicated proteins (F) and clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (G,H) as illustrated in Figure 1A. n = 3 (G,H). I-K. HeLa cells stably transfected with either a wild-type HIF-1α expression vector (WT), a constitutively active mutant HIF-1α 3A expression vector (3A), or its corresponding empty vector (EV) were cultured under the indicated oxygen conditions and subjected to Western blotting for the indicated proteins (I, left), qPCR for CA9 (I, right), and clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (J,K) as illustrated in Figure 1A. n = 3 (I right, J,K). L-Q. HeLa wild-type cells (WT), HeLa HIF2A (EPAS1) KO clones (KO #1 and #2), and HeLa HIF1A/HIF2A (EPAS1) double KO clones (double KO #1 and #2) were cultured under the indicated oxygen conditions, and subjected to Western blotting for the indicated proteins (L,O) and clonogenic cell survival assay to quantify the surviving fraction after the indicated dose of γ-irradiation (M,N,P,Q) as illustrated in Figure 1A. n = 3 (M,N,P,Q). Data are represented as mean ± SD. ns: not significant, **p < 0.01, ****p < 0.0001 (One-way ANOVA with Tukey’s (B-D,G,H,J,K,M,N,P,Q) or Dunnett’s test (I)). We hypothesized that HIF-2α, a factor structurally and functionally similar to HIF-1α, might compensate for the effects of HIF-1α loss [25]. To test this possibility, we next carried out clonogenic cell survival assays using HIF2A (EPAS1 herein) single KO and HIF1A/EPAS1 double KO clones (Fig. 3L-Q; Supplementary Figure S4, S5). Neither EPAS1 single KO nor HIF1A/EPAS1 double KO reduced radioresistance under both normoxic and chronically anoxic conditions (Fig. 3M, N, P, Q). These findings indicate that, under our in vitro conditions, HIF-1 and HIF-2 make only minor contributions to hypoxia-inducible genes-dependent radioresistance under chronically anoxic conditions. HIF-1β maintains baseline radioresistance through regulation of intracellular ROS in a manner independent of AhR HIF-1β is a class II member of the basic helix-loop-helix-PER-ARNT-SIM (bHLH-PAS) family. Class I and class II proteins form heterodimers that function as transcription factors [26]. Representative class I bHLH-PAS partners of HIF-1β include HIF-1α, HIF-2α, and AhR; thus, deletion of HIF-1β disrupts a broad array of HIF-1β-dependent pathways, including HIF-1- and HIF-2-dependent signaling [16,26,27]. To investigate the involvement of HIF-1β in radioresistance, we generated a HIF1B (ARNT) KO clone using CRISPR/Cas9 system (Figure 4A; Supplementary Figure S6). A clonogenic cell survival assay with 8 Gy γ-irradiation demonstrated that HIF1B (ARNT) KO significantly reduced radioresistance consistently across all three oxygen conditions tested, normoxia, acute anoxia, and chronic anoxia (Fig. 4B). Reintroduction of HIF-1β expression vector into the KO cells restored radioresistance under all oxygen conditions (Fig. 4A, B). The result that cell survival under normoxic conditions was decreased by HIF1B (ARNT) KO suggests that the baseline radioresistance of HeLa cells is dependent on HIF-1β.Fig. 4HIF-1β maintains baseline radioresistance through regulation of intracellular ROS in a manner independent of AhR.A-C. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone stably transfected with HIF-1β expression vector (HIF-1β) or its corresponding empty vector (EV) were cultured under the indicated oxygen conditions, and subjected to Western blotting for the indicated proteins (A), clonogenic cell survival assay to quantify surviving fraction after 8 Gy γ-irradiation as illustrated in Figure 1A (B), and DCFH-DA assay for intracellular ROS levels (C). n = 3 (B,C).D,E. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone (KO) were cultured under normoxic conditions in the presence or absence of 10 μM AhR inhibitor CH-223191 from 24 h post-seeding and irradiated with 8 Gy γ-ray 12 h later. The cells were continuously exposed to the drug until 24 h post-irradiation, with medium renewed every 12 h. To confirm sustained drug activity, CYP1A1 mRNA levels were analyzed by qPCR at the time of irradiation and 24 h later (D). Clonogenic survival was assessed by colony formation approx. 12 days post-irradiation (E).Data are represented as mean ± SD. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (One-way ANOVA with Tukey’s test (B,D,E) and two-tailed Student's t-test (C)).Fig 4 dummy alt text HIF-1β maintains baseline radioresistance through regulation of intracellular ROS in a manner independent of AhR. A-C. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone stably transfected with HIF-1β expression vector (HIF-1β) or its corresponding empty vector (EV) were cultured under the indicated oxygen conditions, and subjected to Western blotting for the indicated proteins (A), clonogenic cell survival assay to quantify surviving fraction after 8 Gy γ-irradiation as illustrated in Figure 1A (B), and DCFH-DA assay for intracellular ROS levels (C). n = 3 (B,C). D,E. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone (KO) were cultured under normoxic conditions in the presence or absence of 10 μM AhR inhibitor CH-223191 from 24 h post-seeding and irradiated with 8 Gy γ-ray 12 h later. The cells were continuously exposed to the drug until 24 h post-irradiation, with medium renewed every 12 h. To confirm sustained drug activity, CYP1A1 mRNA levels were analyzed by qPCR at the time of irradiation and 24 h later (D). Clonogenic survival was assessed by colony formation approx. 12 days post-irradiation (E). Data are represented as mean ± SD. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (One-way ANOVA with Tukey’s test (B,D,E) and two-tailed Student's t-test (C)). Because previous studies have reported that HIF-1β is associated with antioxidant capacity in cancer cells, a property that plays a critical role in their radioresistance [28], we next examined the possibility that HIF-1β influences intracellular ROS levels. Quantification of intracellular ROS levels using a DCFH-DA-based assay showed that HIF1B (ARNT) KO increased intracellular ROS levels particularly after irradiation, and that reintroduction of the HIF-1β expression vector effectively reduced them (Fig. 4C). These findings collectively indicate that HIF-1β is crucial for maintaining low intracellular ROS levels, thereby potentially enhancing hypoxia-inducible genes-independent baseline radioresistance. Previous studies have reported that HIF-1β interacts with AhR to induce the expression of proteins that contribute to the regulation of cellular redox balance [27]. We, therefore, hypothesized that HIF-1β promotes radioresistance by inducing antioxidant gene expression in cooperation with AhR. Indeed, the expression of a representative AhR-downstream gene CYP1A1, which plays an important role in the regulation of redox homeostasis [29], was significantly suppressed by CH-223191, an inhibitor of the nuclear translocation of HIF-1β-AhR complex, and by HIF1B (ARNT) KO (Fig. 4D). However, the clonogenic cell survival assay clearly demonstrated that CH-223191 influenced neither the radioresistance of the HeLa—HIF1B (ARNT) KO clone nor that of the parental HeLa cells, suggesting that AhR is unlikely to contribute to radioresistance of HeLa cells in our experimental setting (Fig. 4E). Contribution of NRF2 and TXNRD1 to HIF-1β-mediated radioresistance in cancer cells To identify HIF-1β downstream effectors responsible for baseline radioresistance, we next performed an RNA sequencing-based screening. We searched for genes whose expression decreased upon HIF1B (ARNT) knockout and was restored by reintroduction of the HIF-1β expression vector under normoxic conditions. Among a total of 57,773 annotated transcripts (including both protein-coding genes and non-coding RNA such as lncRNA and small RNAs), 8,060 transcripts exhibited this expected expression pattern (Fig. 5A). To identify genes related to cellular antioxidant capacity, we performed Gene Ontology enrichment analysis and extracted 6 genes annotated to both “cell redox homeostasis (maintenance of the balance between ROS production and elimination)” and “response to oxidative stress (stimulus-dependent antioxidant responses)” (Fig. 5B, C). Based on the result that HIF-1β exhibited no influence on intracellular levels of reduced glutathione (GSH), one of the major intracellular antioxidants (Fig. 5D), we decided to focus particularly on 2 genes that are not directly involved in GSH biosynthesis; thioredoxin reductase 1 (TXNRD1), a key enzyme that maintains redox homeostasis by reducing thioredoxin, and nuclear factor erythroid 2-related factor 2 (NRF2), a master transcription factor regulating the expression of a wide variety of antioxidant genes (Fig. 5C). Clonogenic cell survival assays showed that the reduction of radioresistance by HIF-1B (ARNT) KO was abolished under essentially all oxygen conditions when NRF2 and TXNRD1 were inhibited by their specific inhibitors, ML385 and auranofin, respectively (Fig. 5E, F). These findings suggest that NRF2 and TXNRD1 both function within the same pathway as HIF-1β in maintaining baseline radioresistance in cancer cells. We also examined the effects of the inhibitors alone and found that treatment of WT cells with ML385 or auranofin significantly reduced the surviving fraction under normoxic and acutely anoxic conditions, whereas no significant reduction was observed under chronic anoxia (Fig. 5E, F).Fig. 5Contribution of NRF2 and TXNRD1 to HIF-1β-mediated radioresistance in cancer cells.A. RNA sequencing-based screening under normoxia identified 12,237 transcripts whose expression decreased upon HIF1B (ARNT) KO, of which 8,060 were restored by reintroduction of HIF-1β, out of a total of 57,773 annotated transcripts.B,C. Gene Ontology enrichment analysis of the transcripts identified in A revealed 6 genes annotated to both “cell redox homeostasis” and “response to oxidative stress”. Venn diagram showing the overlap between the 2 indicated GO terms (B). A list of the 6 overlapping genes (C).D. Colorimetric GSH assay quantifying intracellular GSH levels in HeLa HIF1B (ARNT) KO clone stably transfected with either a HIF-1β expression vector (HIF-1β) or its corresponding empty vector (EV) after incubation under normoxic conditions. n = 3.E,F. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone (KO) were cultured under the indicated oxygen conditions in the presence or absence of 10 μM NRF2 inhibitor ML385 (12 h immediately prior to irradiation and 24 h post-irradiation) (E), or 2 μM TXNRD1 inhibitor auranofin for (12 h immediately prior to irradiation) (F), and then subjected to clonogenic cell survival assays to quantify surviving fraction after 8 Gy γ-irradiation as illustrated in Figure 1A. n = 3.G,H. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone (KO) were cultured under the indicated oxygen conditions, treated with or without 10 μM ML385 for 36 h (G) or 2 μM auranofin for 12 h (H), irradiated with the indicated dose of γ-rays, and then subjected to the DCFH-DA assay to quantify intracellular ROS levels. n = 3.Data are presented as mean ± SD. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (One-way ANOVA with Tukey’s test (E) and two-tailed Student's t-test (D,F-H)).Fig 5 dummy alt text Contribution of NRF2 and TXNRD1 to HIF-1β-mediated radioresistance in cancer cells. A. RNA sequencing-based screening under normoxia identified 12,237 transcripts whose expression decreased upon HIF1B (ARNT) KO, of which 8,060 were restored by reintroduction of HIF-1β, out of a total of 57,773 annotated transcripts. B,C. Gene Ontology enrichment analysis of the transcripts identified in A revealed 6 genes annotated to both “cell redox homeostasis” and “response to oxidative stress”. Venn diagram showing the overlap between the 2 indicated GO terms (B). A list of the 6 overlapping genes (C). D. Colorimetric GSH assay quantifying intracellular GSH levels in HeLa HIF1B (ARNT) KO clone stably transfected with either a HIF-1β expression vector (HIF-1β) or its corresponding empty vector (EV) after incubation under normoxic conditions. n = 3. E,F. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone (KO) were cultured under the indicated oxygen conditions in the presence or absence of 10 μM NRF2 inhibitor ML385 (12 h immediately prior to irradiation and 24 h post-irradiation) (E), or 2 μM TXNRD1 inhibitor auranofin for (12 h immediately prior to irradiation) (F), and then subjected to clonogenic cell survival assays to quantify surviving fraction after 8 Gy γ-irradiation as illustrated in Figure 1A. n = 3. G,H. HeLa wild-type cells (WT) and HeLa HIF1B (ARNT) KO clone (KO) were cultured under the indicated oxygen conditions, treated with or without 10 μM ML385 for 36 h (G) or 2 μM auranofin for 12 h (H), irradiated with the indicated dose of γ-rays, and then subjected to the DCFH-DA assay to quantify intracellular ROS levels. n = 3. Data are presented as mean ± SD. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (One-way ANOVA with Tukey’s test (E) and two-tailed Student's t-test (D,F-H)). To examine whether the HIF-1β-NRF2 and the HIF-1β-TXNRD1 axes contribute to ROS reduction, we next conducted DCFH-DA-based assays (Fig. 5G, H). HIF1B (ARNT) KO significantly increased intracellular ROS levels compared with WT cells, and this increase was suppressed by reintroduction of HIF-1β, indicating that HIF-1β negatively regulates intracellular ROS. In addition, inhibition of NRF2 with ML385 abolished the increase in ROS levels induced by HIF1B (ARNT) KO, suggesting that NRF2 functions in the same pathway as HIF-1β in regulating intracellular ROS (Fig. 5G). In contrast, the significant increase in intracellular ROS levels induced by HIF1B (ARNT) KO was still observed in the presence of the TXNRD1 inhibitor auranofin (Fig. 5H), suggesting that TXNRD1 may contribute to radioresistance through mechanisms that are not fully captured by the DCFH-DA-based assay. Discussion In this study, we first disentangled hypoxia-inducible genes-dependent biological mechanisms from hypoxia-inducible genes-independent chemical mechanisms and quantified their respective contributions to cancer cell radioresistance under chronic anoxia. We then conducted mechanistic analyses and found that, although HIF-1 and HIF-2 were dispensable for the hypoxia-inducible genes-dependent component in our in vitro system, HIF-1β promoted baseline radioresistance by enhancing antioxidant capacity in an AhR-independent yet NRF2- and TXNRD1-dependent manner. In our in vitro system under chronic anoxia, hypoxia-inducible genes-independent chemical mechanisms accounted for only 29.6% of radioresistance in SCC7 cells, whereas hypoxia-inducible genes-dependent biological mechanisms accounted for 70.4%, implying a dominant role for the hypoxia-inducible genes-dependent component. To our knowledge, this is the first study to quantitatively evaluate the relative contributions of these two mechanisms, and the results may provide a critical basis for developing strategies to overcome the radioresistance of chronically anoxic regions, which are located in peri‑necrotic areas. Nevertheless, it is essential to experimentally verify whether the insights obtained from these in vitro studies are truly operative within tumor tissues. CA9 expression, used as a representative hypoxia-inducible gene marker, did not increase during 30 min of acute anoxia relative to normoxia, suggesting that hypoxia-inducible genes-dependent biological mechanisms are unlikely to be activated within this timeframe. Consistent with this transcriptional observation, we did not observe detectable accumulation of HIF-1α protein after 30 min of acute anoxia across multiple cancer cell lines. Given that oxygen tension is sufficiently reduced under these conditions (Fig. 1E), this finding suggests that short-term anoxia is not sufficient to induce HIF-1α accumulation to levels detectable by Western blotting. Consistent with this interpretation, HIF-1α protein was readily detected under chronic anoxia conditions (Fig. 1F and I), indicating that sustained oxygen deprivation is required for robust protein stabilization. This observation is in line with our previous report showing that even 2 h of anoxic treatment did not result in detectable HIF-1α accumulation under certain experimental conditions [3]. Together, these findings suggest that the kinetics and magnitude of HIF-1α stabilization are highly context-dependent and may require prolonged hypoxic exposure for accumulation to detectable levels. Nevertheless, post-translational modifications often occur far more rapidly than transcriptional or translational responses. For example, upon reoxygenation, HIF-1α is ubiquitinated and degraded by the proteasome with a half-life of ∼ 5 min [16], and phosphorylation of H2AX at serine 139, an early step in double-strand break repair, occurs within seconds after irradiation [30]. These findings suggest that while a 30-min acute anoxia treatment may not trigger transcript-level responses, oxygen-sensitive post-translational modifications might be initiated during this period. Further optimization of the N2-gas flushing regimen to establish acutely anoxic conditions that reduce oxygen partial pressure to the same extent as under chronic anoxia, yet do not initiate post-translational modifications, may allow a more precise dissection of radioresistance mechanisms and bring us closer to their core determinants. In the present study, we adopted a 30-min anoxia treatment, which we considered sufficient to lower intracellular oxygen partial pressure without yet inducing hypoxia-inducible gene expression. Previous pulsed electron paramagnetic resonance imaging (EPRI) studies have shown that oxygen partial pressure in tumors fluctuates on the order of minutes to tens of minutes [31], consistent with the report by Chaplin et al. that acute hypoxic or anoxic regions around tumor vasculature fluctuate approximately every 20 min [32]. Thus, our 30-min anoxia model reasonably reflects the temporal dynamics of acute hypoxia in vivo. However, when acute anoxia results from transient vascular occlusion, not only oxygen but also nutrients such as glucose become limited. Because glucose levels have also been shown to modulate the radioresistance of anoxic cancer cells [33], future in vitro models of acute anoxia should account for multiple factors, including nutrient availability. Our clonogenic cell survival assays suggested that HIF-1α makes only a limited contribution to radioresistance in vitro, in contrast to findings in HeLa tumor-bearing mouse models [17]. In solid tumors, microenvironmental factors, particularly the vasculature, can be critical. In HeLa tumors, for instance, HIF-1-mediated radioresistance has been attributed to protection of tumor vasculature via VEGF secreted from surviving HIF-1-active cancer cells, thereby sustaining oxygen and nutrient supplies [19]. Such microenvironment-dependent mechanisms are not captured in vitro, potentially explaining the discrepancy. Moreover, the effects of HIF-1α appear to be cell type-dependent: in U87 cells, siRNA-mediated inhibition reduced radioresistance by suppressing anti-apoptotic activity [34], whereas in HCT116 cells, in which HIF-1α induced apoptosis through the p53 pathway, inhibition of HIF-1α increased radioresistance, consistent with reduced apoptosis [35]. These results highlight the cell context dependence of HIF-1α’s role in radioresistance. Similar context dependence has been reported for HIF-2α. Its expression correlates with poor prognosis and reduced response to radiotherapy across multiple cancers [[36], [37], [38]]. In A498 cells, HIF-2α knockdown reduced radioresistance by promoting p53-dependent apoptosis [39]. AhR likewise appears to play cell context-dependent roles in radioresistance. For example, in our in vitro HeLa model, pharmacological inhibition of AhR with CH-223191, although it significantly suppressed the expression of the AhR target gene CYP1A1, did not alter clonogenic survival under normoxia, nor did it modify the radiosensitization conferred by HIF1B (ARNT) KO, indicating that AhR is unlikely to contribute to radioresistance in our setting. In PC9 tumors, it promoted immune evasion following radiotherapy [40], whereas, in A549 cells, AhR enhanced radioresistance by activating DNA repair pathways [41]. Our RNA-Seq analysis indicated a potential regulation of NRF2 mRNA levels by HIF-1β. Consistent with this, the decrease in radioresistance caused by HIF1B (ARNT) KO was abolished in the presence of the pharmacological NRF2 inhibitor ML385 in clonogenic cell survival assay, suggesting that NRF2 acts downstream of HIF-1β. These observations are consistent with the report by Miao et al. that HIF-1β binds the NRF2 promoter and upregulates NRF2 expression [42]. Nevertheless, it should be noted that the study by Miao et al. reported that HIF-1β functions in cooperation with AhR, which differs from our observation of an AhR-independent role of HIF-1β. As mentioned above, this discrepancy may be attributable to cell-context-dependent differences. Here, we identified NRF2 and TXNRD1 as downstream effectors of HIF-1β. Our DCFH-DA assays demonstrated that the increase in ROS induced by HIF1B (ARNT) KO was abolished by NRF2 inhibition (ML385) but persisted with TXNRD1 inhibition (auranofin). Since DCFH-DA primarily detects hydroxyl radicals and peroxynitrite, but not superoxide or singlet oxygen [43,44], these results suggest that NRF2 suppresses ROS readily detected by this assay, thereby contributing to radioresistance. Notably, under normoxic conditions, differences in intracellular ROS levels among WT, HIF1B KO, and HIF-1β-rescued cells were inversely correlated with their radioresistance, supporting a role for HIF-1β-mediated ROS suppression in baseline radioresistance. In contrast, across different oxygen conditions, ROS levels did not consistently correlate with radioresistance, suggesting that additional mechanisms beyond ROS regulation contribute to radioresistance under anoxic conditions. Regarding TXNRD1, it may suppress other types of ROS that are not efficiently detected by DCFH-DA, or may contribute to radioresistance through mechanisms that are independent of classical antioxidant capacity. To further examine the context-dependent roles of NRF2 and TXNRD1, WT cells were treated with ML385 or auranofin prior to irradiation under normoxic, acute anoxic, and chronic anoxic conditions and subjected to clonogenic survival assays. Treatment with ML385 or auranofin significantly reduced the surviving fraction of WT cells under normoxic and acutely anoxic conditions, whereas no significant reduction was observed under chronic anoxia (Fig. 5E, F). These findings indicate that inhibition of NRF2 or TXNRD1 alone can reduce baseline radioresistance under normoxic and acutely anoxic conditions. This is consistent with NRF2 and TXNRD1 acting downstream of HIF-1β. In contrast, their effects appear limited under chronic anoxia, suggesting that additional radioresistance mechanisms induced under chronic anoxia may predominate and mask the contribution of these pathways. Furthermore, under chronic anoxia, inhibition of NRF2 did not fully abolish the difference in intracellular ROS levels between WT and HIF1B KO cells, suggesting that HIF-1β may regulate ROS through NRF2-independent pathways under prolonged anoxic conditions. Collectively, our data indicate that the HIF-1β-NRF2 and HIF-1β-TXNRD1 axes bolster baseline radioresistance of cancer cells irrespective of oxygen availability. These findings suggest that targeting these axes could improve the efficacy of radiotherapy. However, systemic inhibition may cause deleterious effects, given the essential physiological roles of these molecules and the fact that HIF-1β (ARNT) participates in multiple transcriptional partnerships beyond HIF-1. In hematopoietic cells, for example, NRF2 inhibition activates myeloid-derived suppressor cells, thereby enhancing metastasis [45]. Moreover, TXNRD1 inhibition can activate NRF2 and increase antioxidant capacity in cancer cells [46,47]. Therefore, to develop therapeutic strategies, it will be necessary to elucidate the detailed molecular mechanisms by which NRF2 and TXNRD1 function downstream of HIF-1β, confirm their generality across multiple cell lines in addition to HeLa, and further investigate their roles in tumors using tumor-bearing mouse models. Material & methods Cell culture A cervical cancer cell line, HeLa, and an immortalized human embryonic kidney cell line, HEK293TN, were purchased from the American Type Culture Collection and from System Biosciences, respectively. A murine squamous cell carcinoma cell line derived from a spontaneous primary tumor of the abdominal wall in C3H mice, SCC7, was kindly provided by Prof. Masahiro Hiraoka (Kyoto University, Japan). Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Nacalai Tesque) supplemented with 10% heat-inactivated fetal bovine serum (FBS; EQUITECH-BIO) and 1% (v/v) penicillin-streptomycin (Nacalai Tesque) at 37°C in a humidified incubator with 5% CO2 and 95% air. Cells were cultured at < 0.1% O2, 5% CO2 for chronically anoxic conditions, and at 1% O2, 5% CO2 for chronically hypoxic conditions, respectively, for the indicated time in a RUSKINN INVIVO2 400 Hypoxic Workstation (Ruskinn Technology Limited). To mimic oxygen microenvironments in acutely anoxic regions in vitro, cells were transferred to a custom-designed chamber, whereby oxygen in the culture medium was immediately excluded and the cells were exposed to arbitrary pO₂ conditions. The chamber was flushed with N2 gas at 3,480 mL/min, which had passed through a bubbling bottle to maintain high humidity. After N2 gas flushing for the indicated period, the inlet and outlet valves were closed to block gas flow, allowing the oxygen concentration inside the chamber to be stably maintained. This was confirmed by measuring phosphorescence intensity (Supplementary Figure S7). Reagents An AhR inhibitor, CH-223191 (Sigma-Aldrich, Cat#C8124), an NRF2 inhibitor, ML385 (Sigma-Aldrich, Cat#SML1833), a TXNRD1 inhibitor, auranofin (Sigma-Aldrich, Cat#A6733) were used. Plasmid constructs pCDH/HIF-1α was constructed as described previously [24]. To construct pCDH/HIF-1α 3A, which expresses the constitutively active mutant of HIF-1α harboring the P402A, P564A, and N803A mutations, a DNA fragment encoding human HIF-1α 3A was amplified from pcDNA4/HIF-1α 3A-myc-His [48] using the following primers: 5′-TAATACGACTCACTATAGGG-3′ and 5′-TATGATATCTCAGTTAACTTGATCCAAAGCTCTGA-3′, and inserted between the BamHI site and the blunted NotI site of pCDH-EF1α-MCS-IRES-Puro (System Biosciences, CD532A-2). To construct pCDH/HIF-1β-myc-His, pcDNA4/HIF-1β-myc-His [49] was digested with BamHI and PmeI to obtain the DNA fragment encoding HIF-1β fused with myc-tag, and the resulting fragment was inserted between the BamHI site and the blunted NotI site of pCDH-EF1α-MCS-IRES-Puro (System Biosciences, CD532A-2). To construct pCDH/HIF-1β for the preparation of lentivirus expressing wildtype HIF-1β, a DNA fragment encoding the C-terminal region of HIF-1β (294-789 a.a. region of HIF-1β) without the myc-tag sequence was then amplified from pCDH/HIF-1β-myc-His using the following primers: 5′-TGGCTCTGTAAAGGATGGGG-3′ and 5′-TATCTCGAGCTATTCTGAAAAGGGGGGAAACATAG-3′, digested with SpeI and XhoI, and inserted between the SpeI site (located within the HIF-1β CDS) and the XhoI site of pCDH/HIF-1β-myc-His. p5HRE-Luc was constructed as described previously [23]. pRL-TK was purchased from Promega. An oligonucleotide fragment encoding a gRNA targeting the gene of interest was inserted to the BbsI site of pSpCas9(BB)-2A-BSD [24], which constitutively expresses Cas9 and confers blasticidin resistance. Packaging vectors (pPACKH1 HIV Lentivector Packaging Kit, Cat#LV500A-1) were purchased from System Biosciences. Stable transfectants HEK293TN cells were co-transfected with pCDH-EF1α-MCS-IRES-Puro empty vector, pCDH/HIF-1α, and pCDH/HIF-1β together with packaging vectors (pPACKH1 HIV Lentivector Packaging Kit, System Biosciences, Cat#LV500A-1) and cultured for 3 days to produce control lentivirus (empty vector) and lentiviruses expressing HIF-1α, and HIF-1β, respectively, according to the manufacturer’s instructions. The virus-containing medium was collected and applied to the recipient cells pre-incubated with 8 μg/mL polybrene (Sigma Aldrich, Cat#H9268). Infected cells were selected with puromycin (Nacalai Tesque, Cat#19752-22). CRISPR/Cas9-mediated establishment of KO cells Plasmids expressing both gRNA and Cas9 were transiently transfected into cells, and to facilitate straightforward selection of transfected cells, cultures were supplemented with blasticidin at a final concentration of 10 μg/mL for 72 hours, starting 48 hours after transfection. The cells were then re-seeded under sparse conditions to allow colony formation, and single colonies were subsequently isolated. All KO clones were confirmed to carry the intended knockout by sequencing genomic DNA, as shown in Supplementary Figures S4–S6. HeLa HIF1A KO clone was established as described previously [24]. HeLa HIF2A (EPAS1) KO clones and HeLa HIF1B (ARNT) KO clone were established using sgRNA oligonucleotides targeting upstream of exon 1 and within intron 1 of each target gene, according to the previously described procedures (Supplementary Figure S4-S6) [49]. For HeLa HIF2A (EPAS1) KO clone #1, the sgRNAs used were 5′-CTGCGGCCGCGCGTCCCCAT-3′ and 5′-GGGAGAGCATGTGCCCGGTT-3′, and clone #2, 5′-TCCCACTCGCTACCGGCGAT-3′ and 5′-AGTACGTGTAGCCCCCCCGG-3′. For the HeLa HIF1B (ARNT) KO clone, the sgRNAs were 5′-ACGCAAGGAGGGCGAGATTA-3′ and 5′-ACTGGCTGCCGCTCCAACGG-3′. HeLa HIF1A/HIF2A (EPAS1) double KO clones were established from the HeLa HIF1A KO clone [24] using sgRNA oligonucleotides targeting upstream of exon 1 and within intron 1 of HIF2A (EPAS1), according to the previously described procedures (Supplementary Figure S5) [49]. For double KO clone #1, the sgRNAs were 5′-CTGCGGCCGCGCGTCCCCAT-3′ and 5′-AGAAAAGGTAAGCGGGCGTC-3′. and clone #2, 5′-CTGCGGCCGCGCGTCCCCAT-3′ and 5′-GGGAGAGCATGTGCCCGGTT-3′. Reverse transcription and quantitative PCR (qPCR) Cells were lysed in Sepasol-RNA I Super G (Nacalai Tesque, Cat#09379-84) under the same oxygen conditions as during cell culture and 500 ng of total RNA was subjected to reverse transcription using the PrimeScript RT Reagent Kit (Takara Bio, Cat#RR037), according to the manufacturer’s instructions. Quantitative PCR was performed using TB Green Premix Ex Taq II (Takara Bio, Cat#RR820) and a Thermal Cycler Dice Real Time System Single (Takara Bio) with the following gene-specific primers; 5′-TGGCACCCAGCACAATGAA-3′ and 5′-CTAAGTCATAGTCCGCCTAGAAGC-3′ for human β-actin, 5′-ACCAGACAGTGATGCTGAGTGCTAA-3′ and 5′-TCAGCTGTAGCCGAGAGTCACC-3′ for human CA9, 5′-TCGGCCACGGAGTTTCTTC-3′ and 5′-GGTCAGCATGTGCCCAATCA-3′ for human CYP1A1, 5′-CATCCGTAAAGACCTCTATG-3′ and 5′-ATGGAGCCACCGATCCACA-3′ for mouse β-actin, and 5′-TGCTCCAAGTGTCTGCTCAG-3′ and 5′-CAGGTGCATCCTCTTCACTGG-3′ for mouse CA9. Western blotting Cell lysates were harvested with CelLytic M (Sigma-Aldrich, Cat#C2978) under the same oxygen conditions as during cell culture, sonicated with Bioruptor UCW-210 (Cosmo Bio) for 5 min (30 s on/off, 5 cycles), and centrifuged at 15,000 rpm for 15 min at 4°C. The resulting supernatant was subjected to Western blotting using anti-β-actin (Santa Cruz, Cat#sc-69879), anti-human HIF-1α (BD, Cat#610595), anti-mouse HIF-1α (Novus, Cat#NB100-479), anti-HIF-2α (Santa Cruz, Cat#sc-13596), anti-HIF-1β (Novus, Cat#NB100-124), or anti-CA9 (Novus, Cat#NB100-417) antibody as primary antibodies. Anti-mouse IgG HRP-linked Whole Ab (Cytiva, Cat#NA931) or anti-rabbit IgG HRP-linked Whole Ab (Cytiva, Cat#NA934) was used as secondary antibodies. Detection was performed using ECL Prime Western Blotting Detection Reagents (Cytiva, Cat#RPN2232). Assessment of intracellular partial pressure of oxygen using phosphorescent probe BTPDM1 Cells were seeded into 96-well plate (2 × 104 cells per well) and cultured for 16 h in phenol red-free medium (Gibco, Cat#21063-029) containing 500 nM BTPDM1. The culture medium was then replaced with fresh medium to remove extracellular BTPDM1, and the cells were cultured under the indicated oxygen conditions for the indicated period. Phosphorescence signals were detected using Amersham Imager 680 (Cytiva) equipped with custom filter settings (excitation: 460 nm, emission: Cy3 (585 nm −625 nm)) under the same oxygen conditions as the pre-culture. Phosphorescence intensity was quantified using ImageJ. Irradiation and clonogenic cell survival assay Cells were irradiated with the indicated dose of γ-rays using the Gammacell 40 Exactor (MDS Nordion International) under the same oxygen conditions as the pre-culture as illustrated in Fig. 1A, and cultured for 8 (for SCC7 cells) or 12 (for HeLa cells) additional days. Surviving colonies were fixed with 70% ethanol and stained with 0.5% crystal violet (w/v) in 20% methanol. Colonies with more than 50 cells were counted as surviving colonies. The surviving fraction at each radiation dose was calculated as the ratio of the plating efficiency at that dose to the plating efficiency at 0 Gy. Percentage contributions of hypoxia-inducible genes-dependent biological mechanisms and genes-independent chemical mechanisms were calculated as follows: hypoxia-inducible genes-dependent biological mechanisms = {(surviving fraction under chronic anoxia - surviving fraction under acute anoxia) / (surviving fraction under chronic anoxia - surviving fraction under normoxia)} × 100%; and hypoxia-inducible genes-independent chemical mechanisms = {(surviving fraction under acute anoxia - surviving fraction under normoxia) / (surviving fraction under chronic anoxia - surviving fraction under normoxia)} × 100%. RNA-Seq analysis After cells were cultured under normoxic or chronically anoxic (< 0.1% O2) conditions for 24 h, total RNA was extracted with TRIzol reagent (Ambion, Cat#15596026), and its quality was analysed using Agilent Bioanalyzer and TapeStation (Agilent Technologies). The samples were then subjected to rRNA depletion with the NEBNext rRNA Depletion Kit (NEB, Cat#E6310) and subsequently to directional library synthesis using the NEBNext Ultr Directional RNA Library Prep Kit for Illumina (NEB, Cat#E7420), according to the manufacturer’s instruction. After the libraries were quantified using the DNA High-sensitivity kit (Agilent, Cat#5067-4626), RNA sequencing was performed using Illumina NextSeq500 (Illumina) at Tsukuba i-Laboratory, University of Tsukuba. Gene Ontology analysis was performed using DAVID (https://david.ncifcrf. gov/) [50,51]. Luciferase assay Cells were seeded into 24-well plate (1 × 104 cells per well), cultured for 16 h, and transiently co-transfected with p5HRE-Luc and pRL-TK as an internal control to calculate relative luciferase activity. After 32 h, the cells were cultured under the indicated oxygen conditions for the indicated period, and cell lysates were harvested with 100 μL Passive Lysis Buffer (Promega). The luciferase assay was performed using the Dual Luciferase Assay kit (Promega), according to the manufacturer’s instructions. Colorimetric GSH assay Cells were seeded into 100-mm dish (1 × 106 cells per dish), cultured for 40 h, and harvested by trypsin digestion. Intracellular reduced glutathione (GSH) levels were quantified using the Glutathione Colorimetric Assay Kit (Abcam, Cat#ab239709), according to the manufacturer’s instructions. DCFH-DA assay Cells were seeded into 24-well plate (2.5 × 104 cells per well), cultured for 24 h, and treated with the indicated inhibitors for the indicated period. Intracellular ROS levels were quantified using the ROS Assay Kit -Highly Sensitive DCFH-DA- (Dojindo Laboratories, Cat#340-09811), according to the manufacturer’s instructions. Briefly, cells were stained with DCFH-DA dye at 37°C for 30 min and the DCF fluorescence intensity was analyzed using BD FACSCanto II Flow Cytometer (BD Biosciences) and DIVA software (BD Biosciences). Statistical analysis The results are presented as the mean ± SD. All experiments were independently repeated at least three times, and representative results are shown. Statistical significance of differences between two groups was determined using a two-tailed Student’s t-test, and among multiple groups using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc test (ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Approval of the submitted version of the manuscript Please check this box to confirm that all co-authors have read and approved the version of the manuscript that is submitted. Signatures are not required. CRediT authorship contribution statement Satoshi Takeuchi: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation. Minoru Kobayashi: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Conceptualization. Peter Wai Tik Lee: Writing – review & editing, Supervision, Resources. Itsuki Takahashi: Writing – review & editing, Investigation. Toshitada Yoshihara: Writing – review & editing, Resources. Hiroshi Harada: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.