The KEAP1–NFE2L2/NRF2 Axis in Non-Small Cell Lung Cancer Radioresistance: Redox Homeostasis and Emerging DNA Damage Response Mechanisms Authors: Xinyu Jia, Eerdemutu, Yuan Ren, Shiduo Yang, Chang Yang, Qiaochu Ding, Yu Lin Published: 2026 ABSTRACT AbstractRadioresistance and local recurrence remain major barriers to effective radiotherapy in non-small cell lung cancer (NSCLC). Loss-of-function KEAP1 alterations or activating NFE2L2 alterations can stabilize NRF2, but do not alone establish sustained transcriptional activity or functional dependency. This focused narrative review evaluates clinical radiotherapy studies and mechanistically informative preclinical studies linking the KEAP1–NFE2L2/NRF2 axis to NSCLC radioresistance. We prioritized clinical studies reporting radiotherapy-specific outcomes and preclinical studies coupling NRF2-related molecular status or perturbation with radiation-response endpoints; contextual studies informed metabolic, DNA damage response (DDR), immune and normal-lung effects. Evidence most consistently supports NRF2-mediated redox protection through glutathione-dependent defense, cellular reducing capacity and antioxidant enzymes, limiting radiation-induced reactive oxygen species (ROS) accumulation and oxidative injury. Limited studies further suggest that NRF2 may affect DNA-damage signaling, checkpoint control and repair. The detailed RPA32–TOPBP1–ATR–CHK1 model is therefore considered proposed rather than established in NRF2-active NSCLC. Retrospective clinical studies associate pathogenic KEAP1/NFE2L2 alterations with impaired local control in some radiotherapy-treated cohorts, but do not justify treating genomic status, protein abundance, transcriptional activity and functional dependency as equivalent measures or demonstrate treatment-predictive value. NRF2-mediated normal-lung protection also constrains systemic inhibition. Prospective studies integrating molecular classification, radiation-response endpoints, local control and normal-tissue toxicity are required before biomarker-guided radiosensitization can be considered. Introduction Radiotherapy is used in distinct non-small cell lung cancer (NSCLC) settings. Medically inoperable, node-negative early-stage disease is generally treated with stereotactic body radiotherapy (SBRT), which delivers ablative doses in a few fractions without concurrent systemic therapy.1 Unresectable locally advanced disease is generally treated with conventionally fractionated concurrent chemoradiotherapy, followed by consolidation durvalumab in eligible patients without progression; chemoradiotherapy alone remains relevant when immunotherapy is not given.2 These settings differ in target volume, fractionation, systemic exposure and risks of locoregional and distant failure, so biomarker associations should be interpreted within treatment-specific cohorts. Radioresistance can impair local control, whereas normal-tissue toxicity limits treatment intensification. Relevant mechanisms include DNA repair, hypoxia, epithelial–mesenchymal transition, tumor-microenvironment remodeling and genomic alterations.3 Alterations in Kelch-like ECH-associated protein 1 (KEAP1) and nuclear factor erythroid 2 like 2 (NFE2L2), which encodes nuclear factor erythroid 2-related factor 2 (NRF2), are candidate determinants of NSCLC radioresponse. KEAP1 normally promotes NRF2 degradation; loss-of-function alterations in KEAP1 or activating alterations in NFE2L2 can sustain antioxidant response element (ARE)-regulated transcription.4 Histologic and co-mutational context is relevant: lung adenocarcinoma and squamous-cell carcinoma differ in alteration distribution, and co-alterations involving KRAS, STK11/LKB1 and TP53 may modify pathway activity, metabolic dependency and radiation response.4–9 In localized NSCLC, KEAP1 or NFE2L2 mutations have been associated with local recurrence after radiotherapy but not surgery; their independent predictive value remains unvalidated.10,11 As illustrated in Figure 1, the principal tumor-intrinsic effects of the KEAP1–NFE2L2/NRF2 axis on radiotherapy response involve two interconnected processes: maintenance of redox homeostasis and emerging NRF2-associated DDR mechanisms. Ionizing radiation induces both direct DNA damage and indirect injury mediated by reactive oxygen species (ROS), whereas sustained NRF2 activation can increase glutathione metabolism, NADPH-dependent reducing capacity and antioxidant enzyme activity, thereby limiting radiation-induced oxidative damage. Evidence that NRF2 directly regulates the DNA damage response (DDR) is more limited. One experimental study reported an NRF2–ATR interaction and NRF2-dependent ATR–CHK1 activation after DNA damage, and a subsequent study from the same investigative group proposed that NRF2 promotes RPA32 phosphorylation and TOPBP1 recruitment.12,13 This focused review examines clinical and mechanistic evidence linking aberrant KEAP1–NFE2L2/NRF2 signaling to NSCLC radioresistance, with particular emphasis on redox homeostasis and emerging DDR mechanisms. It also considers the immune context of radiotherapy–immunotherapy combinations, candidate indicators of reduced radiosensitivity and potential intervention strategies, while critically assessing the strength and translational limitations of the available evidence. Figure 1Evidence-qualified model of KEAP1–NFE2L2/NRF2-associated radioresistance in non-small cell lung cancer. Ionizing radiation produces reactive oxygen species (ROS)-mediated oxidative stress and direct DNA damage, whereas KEAP1 loss-of-function or activating NFE2L2 alterations can sustain NRF2 stabilization, nuclear translocation and antioxidant response element (ARE) target-gene transcription. Solid arrows denote established or directly supported relationships. NRF2-mediated redox protection, including increased glutathione (GSH)/thioredoxin (Trx) capacity, nicotinamide adenine dinucleotide phosphate (NADPH) regeneration and antioxidant enzyme activity, limits ROS-mediated injury. The RPA-coated single-stranded DNA (ssDNA)–TOPBP1–ATR–CHK1 sequence represents established DNA damage response (DDR) biology independently of NRF2. Dashed purple arrows denote the proposed direct involvement of NRF2 in RPA32 phosphorylation and TOPBP1 recruitment; the specific RPA32–TOPBP1 link was reported principally in one experimental study and has not been independently validated in molecularly defined NSCLC radiotherapy models or clinical cohorts. Redox protection and the proposed NRF2-associated DDR mechanism may support post-irradiation tumor-cell repair and survival and thereby contribute to NSCLC radioresistance. The gray arrow denotes radiotherapy selection pressure and enrichment of surviving resistant cells rather than a direct molecular feedback pathway. The red cross indicates loss of KEAP1 function; upward and downward arrows indicate increases and decreases, respectively.A diagram on NRF2-related radioresistance in NSCLC highlights Radiotherapy Exposure, Tumor Molecular Context and Sustained NRF2 Activity. Ionizing radiation induces ROS and DNA damage. KEAP1/NFE2L2 dysregulation stabilizes NRF2, leading to nuclear translocation and ARE gene transcription. NRF2-ARE antioxidant mechanisms enhance GSH/Trx capacity, NADPH regeneration and antioxidant enzymes, mitigating ROS damage. The DNA damage response involves DDR core and NRF2′s role in RPA32 phosphorylation and TOPBP1 recruitment, aiding checkpoint maintenance and repair, boosting post-irradiation tumor-cell survival and contributing to NSCLC radioresistance and local recurrence risk. Arrows depict established, proposed and selection/enrichment pathways, with symbols for KEAP1 loss-of-function and level changes.Mechanistic schematic showing radiotherapy-induced ROS and DNA damage in NSCLC. KEAP1 loss or NFE2L2 activation stabilizes NRF2, which enhances redox defenses and RPA32–TOPBP1–ATR–CHK1-associated DNA damage responses, supporting checkpoint maintenance, DNA repair and post-irradiation survival and thereby contributing to radioresistance and potentially impaired local control.Abbreviations: ARE, antioxidant response element; ATR, ataxia telangiectasia and Rad3-related; CHK1, checkpoint kinase 1; DDR, DNA damage response; GSH, glutathione; KEAP1, Kelch-like ECH-associated protein 1; NADPH, reduced nicotinamide adenine dinucleotide phosphate; NFE2L2, nuclear factor erythroid 2 like 2; NRF2, nuclear factor erythroid 2-related factor 2; NSCLC, non-small cell lung cancer; ROS, reactive oxygen species; RPA, replication protein A; RPA32, 32-kDa subunit of replication protein A; ssDNA, single-stranded DNA; TOPBP1, DNA topoisomerase II-binding protein 1; Trx, thioredoxin. Evidence-qualified model of KEAP1–NFE2L2/NRF2-associated radioresistance in non-small cell lung cancer. Ionizing radiation produces reactive oxygen species (ROS)-mediated oxidative stress and direct DNA damage, whereas KEAP1 loss-of-function or activating NFE2L2 alterations can sustain NRF2 stabilization, nuclear translocation and antioxidant response element (ARE) target-gene transcription. Solid arrows denote established or directly supported relationships. NRF2-mediated redox protection, including increased glutathione (GSH)/thioredoxin (Trx) capacity, nicotinamide adenine dinucleotide phosphate (NADPH) regeneration and antioxidant enzyme activity, limits ROS-mediated injury. The RPA-coated single-stranded DNA (ssDNA)–TOPBP1–ATR–CHK1 sequence represents established DNA damage response (DDR) biology independently of NRF2. Dashed purple arrows denote the proposed direct involvement of NRF2 in RPA32 phosphorylation and TOPBP1 recruitment; the specific RPA32–TOPBP1 link was reported principally in one experimental study and has not been independently validated in molecularly defined NSCLC radiotherapy models or clinical cohorts. Redox protection and the proposed NRF2-associated DDR mechanism may support post-irradiation tumor-cell repair and survival and thereby contribute to NSCLC radioresistance. The gray arrow denotes radiotherapy selection pressure and enrichment of surviving resistant cells rather than a direct molecular feedback pathway. The red cross indicates loss of KEAP1 function; upward and downward arrows indicate increases and decreases, respectively. Literature Search Strategy A focused literature search was conducted in PubMed/MEDLINE, Web of Science Core Collection and Scopus from database inception, with the final search performed on August 11, 2026. Combinations of terms related to the pathway (“KEAP1,” “NFE2L2” and “NRF2”), disease (“NSCLC,” “lung adenocarcinoma” and “lung squamous-cell carcinoma”) and radiotherapy or relevant mechanisms (“radiotherapy,” “radioresistance,” “radiosensitivity,” “redox,” “oxidative stress,” “DNA damage response” and “immunotherapy”) were used. Clinical studies were prioritized when they defined the biomarker and reported radiotherapy-specific outcomes. Preclinical studies were prioritized when molecular status or experimental perturbation was linked to radiation-response endpoints. Studies without irradiation or outside NSCLC were included only for directly relevant context concerning metabolism, DDR, immunity or normal-lung injury. Reference lists of relevant articles and reviews were screened. Studies unrelated to NSCLC radioresponse or these predefined contextual domains were excluded. No quantitative synthesis or formal risk-of-bias assessment was performed. KEAP1 and NFE2L2 Alterations in NSCLC Radioresistance Molecular Basis of KEAP1 and NFE2L2 Alterations in NSCLC KEAP1 and NFE2L2 have attracted considerable attention in NSCLC because alterations in either gene can disrupt the physiological regulation of NRF2. Under basal conditions, KEAP1 functions as an adaptor for the cullin 3 (CUL3)-containing E3 ubiquitin ligase complex, promoting NRF2 ubiquitination and proteasomal degradation. This process maintains NRF2-dependent transcription at a relatively low level. In response to oxidative or electrophilic stress, KEAP1-mediated repression is attenuated, allowing NRF2 to accumulate, translocate to the nucleus and activate transcriptional programs involved in redox homeostasis and cellular protection.14 Loss-of-function alterations in KEAP1 or activating alterations in NFE2L2 can uncouple this stress-responsive pathway from its normal regulation, leading to sustained NRF2 activation and prolonged expression of downstream defense genes. Constitutive pathway activation promotes oxidative-stress tolerance and an aggressive tumor phenotype.5 NRF2-dependent mechanistic target of rapamycin complex 1 (mTORC1) signaling has also been identified as a potential therapeutic vulnerability in lung squamous cell carcinoma.15 Individual variants are not functionally equivalent, and a genomic alteration alone does not establish sustained NRF2 activity or dependency. The relevance of this molecular dysregulation to radiotherapy arises from the central contribution of ROS and DNA damage to radiation-induced tumor cell killing. Persistent NRF2 activation may strengthen redox buffering and other cytoprotective responses, thereby reducing radiation-induced injury and increasing survival after irradiation.5,14 Accordingly, the relevance of a KEAP1 or NFE2L2 alteration to radioresistance depends on its pathogenicity, downstream NRF2 pathway activity and irradiation-specific functional consequences. KEAP1 and NFE2L2 Alterations and Local Recurrence After Radiotherapy Clinical radiotherapy studies have generally defined the biomarker at the genomic level rather than by direct measurement of NRF2 pathway activity. In a cohort of 232 patients with localized NSCLC, Binkley et al classified tumors according to pathogenic KEAP1 or NFE2L2 alterations and found that these alterations were associated with a higher risk of local recurrence after radiotherapy but not after surgery.11 This treatment-specific association is consistent with a mutation-associated radioresistant phenotype. However, it does not demonstrate that every altered tumor had sustained NRF2 transcriptional activity or was functionally dependent on NRF2. RNA sequencing was available for only 41 radiotherapy-treated tumors, and NFE2L2 target-gene expression did not predict local recurrence. The principal clinical finding was therefore based on tumor genotype rather than a validated transcriptional pathway-activity measure. Two retrospective studies evaluated genomic KEAP1/NFE2L2 status in patients with stage III NSCLC treated with concurrent chemoradiotherapy and durvalumab. In the study by Shaverdian et al, KEAP1/NFE2L2-altered tumors had a 12-month locoregional failure incidence of 62% after chemoradiotherapy alone and 25% after chemoradiotherapy followed by durvalumab (P=0.021), whereas alteration status did not predict locoregional failure in the durvalumab-treated cohort.16 In a subsequent cohort of 81 patients receiving chemoradiotherapy and durvalumab, the 24-month locoregional failure rate was 52% in altered tumors and 27% in wild-type tumors (P=0.05), but KEAP1/NFE2L2 status was not associated with locoregional failure in Cox analysis.17 These nonrandomized genomic analyses did not measure NRF2 transcriptional activity, immune-cell composition or functional dependency. They therefore suggest that immune consolidation may modify mutation-associated local failure, but do not establish a treatment-predictive interaction between NRF2 activation and durvalumab. These clinical studies therefore demonstrate associations between genomic alterations and outcomes after multimodality treatment. They do not establish that all genomically altered tumors are transcriptionally NRF2-active, that all NRF2-active tumors harbor canonical KEAP1 or NFE2L2 alterations, or that NRF2 dependency mediates each observed local failure. Treatment context, variant pathogenicity, histology and co-occurring alterations may all modify the association. KEAP1/NFE2L2 status should consequently remain a candidate genomic feature for radiotherapy research rather than a validated surrogate for functional NRF2 activation or a treatment-predictive biomarker. The principal clinical studies evaluating KEAP1/NFE2L2-related biomarkers in radiotherapy-treated NSCLC are summarized in Table 1. Table 1Clinical Evidence Linking KEAP1–NFE2L2/NRF2-Related Biomarkers to Radiotherapy Outcomes in NSCLCStudy (Ref).Design and PopulationHistology/StageRadiotherapy and Systemic TherapyBiomarker DefinitionEndpointPrincipal and Adjusted FindingsKey LimitationsSitthideatphaiboon et al, 20217Retrospective cohort; n=194 stage I–III NSCLC, with the principal RT analysis involving 164 patients with stage III disease.Stage I–III NSCLC; mixed histology.RT with or without chemotherapy; numerical RT prescriptions and systemic regimens were not uniform across the cohort.Clinical analysis of STK11/LKB1 mutation; KEAP1 was not assessed clinically. KEAP1/NRF2 dependency was examined in the linked preclinical component.LRR, DM, DFS and OSSTK11/LKB1 mutation was associated with increased LRR (P=0.0108), shorter DFS (HR, 2.530; P=0.0029) and shorter OS (HR, 2.198; P=0.0263).Indirect NRF2 context because KEAP1 was not assessed clinically; retrospective design; heterogeneous treatment; establishes association, not RT-specific prediction.Sitthideatphaiboon et al, 20248Retrospective cohort; n=178 stage III locally advanced NSCLC treated with CRT; IHC evaluable in 74 patients.76% non-squamous and 24% squamous in the full cohort; stage III.At least 60 Gy in 82% of patients; concurrent chemotherapy in 89%, with some neoadjuvant or consolidation therapy; no consolidation durvalumab.Semiquantitative LKB1 IHC using an outcome-derived H-score cutoff of 17.5; NRF2 and NQO1 IHC also assessed; no tumor genotyping.LRR, DM, DFS and OSTwo-year LRR was 68.8% with low LKB1 vs 31.3% with high LKB1 (HR, 3.90; 95% CI, 1.50–10.11; P=0.0001). The association was observed in non-squamous tumors (HR, 5.30; 95% CI, 1.97–14.27; P=0.0001), but not in squamous tumors. On multivariable analysis, low LKB1 was associated with shorter DFS (HR, 2.15; P=0.009) and OS (HR, 2.03; P=0.04).IHC subset of 74; outcome-derived cutoff; no genotyping; retrospective design; association does not establish predictive utility.Binkley et al, 202011Retrospective, single-institution cohort; n=232 localized NSCLC: CRT, n=47; SABR, n=50; surgery, n=135.87.1% adenocarcinoma, 10.3% squamous and 2.6% other; stage IA1–IIIC, with stage distribution differing by treatment group.Definitive conventionally fractionated CRT or high-dose SABR; surgery served as a non-RT comparator. Concurrent chemotherapy was used in the CRT cohort.Targeted sequencing (130–198 genes) with functional classification of pathogenic KEAP1/NFE2L2 variants; RNA sequencing in 41 RT-treated tumors for an NRF2-target signature.LRIn the combined RT cohort, 2-year LR was 42.4% for KEAP1/NFE2L2-mutant tumors vs 12.5% for wild-type tumors (P=0.005). In the CRT cohort, mutation status remained associated with LR on multivariable analysis (HR, 5.17; 95% CI, 1.30–20.58; P=0.02). No association was observed after surgery (P=0.60). The RNA signature was not associated with LR (P=0.93).Retrospective design; small molecular subgroups; heterogeneous treatment cohorts; RNA-sequencing subset of 41; genotype–pathway discordance; no randomized treatment-interaction test.Shaverdian et al, 202116Retrospective cohort; n=120 stage III NSCLC: cCRT alone, n=54; cCRT followed by durvalumab, n=66.Stage III NSCLC; mixed histology.Definitive cCRT; consolidation durvalumab in 66 patients. Numerical RT prescriptions were not uniform across the reported cohort.Tumor genomic profiling for KEAP1/NFE2L2 mutations; 20 of 54 cCRT-only tumors and 18 of 66 durvalumab-treated tumors were altered.LRFWith cCRT alone, KEAP1/NFE2L2 mutation was associated with increased LRF on multivariable analysis (HR, 3.9; 95% CI, 1.6–9.8; P=0.003). In the durvalumab cohort, mutation status was not associated with LRF (P=0.545). Among altered tumors, 12-month LRF was 62% with cCRT alone vs 25% with durvalumab (P=0.021).Nonrandomized, retrospective comparison; small molecular subgroups; treatment-era differences; genomic biomarker only; no pathway-activity or immune measurements; no randomized predictive interaction.Lebow et al, 202317Retrospective cohort; n=81 unresectable stage III NSCLC treated with cCRT and consolidation durvalumab.69% adenocarcinoma, 22% squamous and 9% other; unresectable stage III.RT at 2 Gy/fraction to a median dose of 60 Gy (range, 56–66 Gy), concurrent platinum-doublet chemotherapy, then durvalumab 10 mg/kg every 2 weeks for up to 12 months (median, 6 months).Prospective comprehensive targeted profiling with OncoKB annotation; KEAP1/NFE2L2 alterations in 11 of 81 tumors; TMB also evaluated.24-month LRF and PFSLRF was 52% in KEAP1/NFE2L2-altered tumors vs 27% in unaltered tumors (P=0.05). On Cox analysis, only TMB was associated with LRF (HR, 0.17; 95% CI, 0.03–0.64; P=0.02) and PFS (HR, 0.45; 95% CI, 0.21–0.90; P=0.03); KEAP1/NFE2L2 status was not independently significant.Retrospective design; only 11 altered tumors; single treatment context; genotype without pathway-activity measurement; no treatment-interaction test.Abbreviations: cCRT, concurrent chemoradiotherapy; CI, confidence interval; CRT, chemoradiotherapy; DFS, disease-free survival; DM, distant metastasis; HR, hazard ratio; IHC, immunohistochemistry; LR, local recurrence; LRF, locoregional failure; LRR, locoregional recurrence; NSCLC, non-small cell lung cancer; OS, overall survival; PFS, progression-free survival; RT, radiotherapy; SABR, stereotactic ablative radiotherapy; TMB, tumor mutational burden. Table 1Clinical Evidence Linking KEAP1–NFE2L2/NRF2-Related Biomarkers to Radiotherapy Outcomes in NSCLCStudy (Ref).Design and PopulationHistology/StageRadiotherapy and Systemic TherapyBiomarker DefinitionEndpointPrincipal and Adjusted FindingsKey LimitationsSitthideatphaiboon et al, 20217Retrospective cohort; n=194 stage I–III NSCLC, with the principal RT analysis involving 164 patients with stage III disease.Stage I–III NSCLC; mixed histology.RT with or without chemotherapy; numerical RT prescriptions and systemic regimens were not uniform across the cohort.Clinical analysis of STK11/LKB1 mutation; KEAP1 was not assessed clinically. KEAP1/NRF2 dependency was examined in the linked preclinical component.LRR, DM, DFS and OSSTK11/LKB1 mutation was associated with increased LRR (P=0.0108), shorter DFS (HR, 2.530; P=0.0029) and shorter OS (HR, 2.198; P=0.0263).Indirect NRF2 context because KEAP1 was not assessed clinically; retrospective design; heterogeneous treatment; establishes association, not RT-specific prediction.Sitthideatphaiboon et al, 20248Retrospective cohort; n=178 stage III locally advanced NSCLC treated with CRT; IHC evaluable in 74 patients.76% non-squamous and 24% squamous in the full cohort; stage III.At least 60 Gy in 82% of patients; concurrent chemotherapy in 89%, with some neoadjuvant or consolidation therapy; no consolidation durvalumab.Semiquantitative LKB1 IHC using an outcome-derived H-score cutoff of 17.5; NRF2 and NQO1 IHC also assessed; no tumor genotyping.LRR, DM, DFS and OSTwo-year LRR was 68.8% with low LKB1 vs 31.3% with high LKB1 (HR, 3.90; 95% CI, 1.50–10.11; P=0.0001). The association was observed in non-squamous tumors (HR, 5.30; 95% CI, 1.97–14.27; P=0.0001), but not in squamous tumors. On multivariable analysis, low LKB1 was associated with shorter DFS (HR, 2.15; P=0.009) and OS (HR, 2.03; P=0.04).IHC subset of 74; outcome-derived cutoff; no genotyping; retrospective design; association does not establish predictive utility.Binkley et al, 202011Retrospective, single-institution cohort; n=232 localized NSCLC: CRT, n=47; SABR, n=50; surgery, n=135.87.1% adenocarcinoma, 10.3% squamous and 2.6% other; stage IA1–IIIC, with stage distribution differing by treatment group.Definitive conventionally fractionated CRT or high-dose SABR; surgery served as a non-RT comparator. Concurrent chemotherapy was used in the CRT cohort.Targeted sequencing (130–198 genes) with functional classification of pathogenic KEAP1/NFE2L2 variants; RNA sequencing in 41 RT-treated tumors for an NRF2-target signature.LRIn the combined RT cohort, 2-year LR was 42.4% for KEAP1/NFE2L2-mutant tumors vs 12.5% for wild-type tumors (P=0.005). In the CRT cohort, mutation status remained associated with LR on multivariable analysis (HR, 5.17; 95% CI, 1.30–20.58; P=0.02). No association was observed after surgery (P=0.60). The RNA signature was not associated with LR (P=0.93).Retrospective design; small molecular subgroups; heterogeneous treatment cohorts; RNA-sequencing subset of 41; genotype–pathway discordance; no randomized treatment-interaction test.Shaverdian et al, 202116Retrospective cohort; n=120 stage III NSCLC: cCRT alone, n=54; cCRT followed by durvalumab, n=66.Stage III NSCLC; mixed histology.Definitive cCRT; consolidation durvalumab in 66 patients. Numerical RT prescriptions were not uniform across the reported cohort.Tumor genomic profiling for KEAP1/NFE2L2 mutations; 20 of 54 cCRT-only tumors and 18 of 66 durvalumab-treated tumors were altered.LRFWith cCRT alone, KEAP1/NFE2L2 mutation was associated with increased LRF on multivariable analysis (HR, 3.9; 95% CI, 1.6–9.8; P=0.003). In the durvalumab cohort, mutation status was not associated with LRF (P=0.545). Among altered tumors, 12-month LRF was 62% with cCRT alone vs 25% with durvalumab (P=0.021).Nonrandomized, retrospective comparison; small molecular subgroups; treatment-era differences; genomic biomarker only; no pathway-activity or immune measurements; no randomized predictive interaction.Lebow et al, 202317Retrospective cohort; n=81 unresectable stage III NSCLC treated with cCRT and consolidation durvalumab.69% adenocarcinoma, 22% squamous and 9% other; unresectable stage III.RT at 2 Gy/fraction to a median dose of 60 Gy (range, 56–66 Gy), concurrent platinum-doublet chemotherapy, then durvalumab 10 mg/kg every 2 weeks for up to 12 months (median, 6 months).Prospective comprehensive targeted profiling with OncoKB annotation; KEAP1/NFE2L2 alterations in 11 of 81 tumors; TMB also evaluated.24-month LRF and PFSLRF was 52% in KEAP1/NFE2L2-altered tumors vs 27% in unaltered tumors (P=0.05). On Cox analysis, only TMB was associated with LRF (HR, 0.17; 95% CI, 0.03–0.64; P=0.02) and PFS (HR, 0.45; 95% CI, 0.21–0.90; P=0.03); KEAP1/NFE2L2 status was not independently significant.Retrospective design; only 11 altered tumors; single treatment context; genotype without pathway-activity measurement; no treatment-interaction test.Abbreviations: cCRT, concurrent chemoradiotherapy; CI, confidence interval; CRT, chemoradiotherapy; DFS, disease-free survival; DM, distant metastasis; HR, hazard ratio; IHC, immunohistochemistry; LR, local recurrence; LRF, locoregional failure; LRR, locoregional recurrence; NSCLC, non-small cell lung cancer; OS, overall survival; PFS, progression-free survival; RT, radiotherapy; SABR, stereotactic ablative radiotherapy; TMB, tumor mutational burden. Clinical Evidence Linking KEAP1–NFE2L2/NRF2-Related Biomarkers to Radiotherapy Outcomes in NSCLC Abbreviations: cCRT, concurrent chemoradiotherapy; CI, confidence interval; CRT, chemoradiotherapy; DFS, disease-free survival; DM, distant metastasis; HR, hazard ratio; IHC, immunohistochemistry; LR, local recurrence; LRF, locoregional failure; LRR, locoregional recurrence; NSCLC, non-small cell lung cancer; OS, overall survival; PFS, progression-free survival; RT, radiotherapy; SABR, stereotactic ablative radiotherapy; TMB, tumor mutational burden. Genomic Alterations, Pathway Activation and Functional Dependency Genomic alterations, protein abundance, transcriptional pathway activity and functional dependency represent distinct biological levels. Sequencing can identify KEAP1 or NFE2L2 alterations, but variant detection does not demonstrate increased NRF2 protein abundance, nuclear localization or downstream transcription. NRF2 or NQO1 protein expression provides information at the protein level but does not establish the full transcriptional output of the pathway. Transcriptional signatures estimate downstream activity, whereas functional dependency requires causal evidence that perturbing NRF2 or a specific downstream process changes the relevant phenotype. None of these measurements can substitute for direct assessment of radiation response. The 46-gene K1N2 score was developed from transcriptional features associated with KEAP1/NFE2L2 pathway dysregulation and identified tumors with canonical mutations as well as potentially activating alterations in other genes.18 It also outperformed mutation status in analyses of patient survival and tumor hypoxia. The K1N2 score should therefore be interpreted as a transcriptional surrogate of pathway activity rather than a direct measure of NRF2 dependency. It has not been validated against post-irradiation ROS dynamics, clonogenic survival, tumor control or local recurrence in a prospectively treated radiotherapy cohort. A subsequent study benchmarked NQO1 immunohistochemistry and single-gene RNA measurements against mutation status and K1N2-defined pathway activity.19 The performance of NQO1 immunohistochemistry varied by histology, and single-gene RNA measurements remained surrogate tests rather than functional radiation-response assays. Functional dependency is most directly supported when genetic perturbation and rescue alter radiation-related endpoints. RNA interference-mediated NRF2 depletion in A549 and H460 cells, together with Nrf2 gain- and loss-of-function models, linked NRF2 function to oxidative injury and radiation response in those experimental systems.20 These findings demonstrate a causal contribution in the tested models but do not establish uniform dependency across heterogeneous NSCLC. Pharmacological studies showing pathway suppression and radiosensitization provide complementary evidence but should be distinguished from genetic dependency experiments.21,22 Similarly, A549R, H460R and H1299R cells generated through repeated irradiation represent experimentally selected radioresistant populations rather than a genomic subgroup defined by KEAP1 or NFE2L2 status.12 Accordingly, KEAP1/NFE2L2-altered should denote genomic status; NRF2-high should denote a protein-based measurement; transcriptionally NRF2-active should be reserved for evidence based on target-gene expression or a pathway signature; and NRF2-dependent should be used only when causal perturbation changes the relevant biological phenotype. A radioresistant phenotype additionally requires radiation-specific evidence, such as clonogenic survival, tumor response or local-control outcomes. Prospective integration of these measurements will be necessary to determine which level, or combination of levels, has clinically relevant value. Molecular Context and Co-Occurring Alterations NRF2-associated phenotypes are not uniform across NSCLC. Oncogenic KRAS can increase NRF2 transcription in experimental models, but KRAS mutation alone does not establish sustained NRF2 pathway activation.23 In KRAS-mutant lung adenocarcinoma, tumors with alterations in serine/threonine kinase 11 (STK11), which encodes liver kinase B1 (LKB1), are enriched for concurrent KEAP1 alterations. In models defined by concurrent KRAS mutation, LKB1 loss and KEAP1/NRF2 pathway activation, LKB1 deficiency increased energetic and redox stress, whereas KEAP1 loss and NRF2-dependent metabolic adaptation produced greater glutamine dependence and sensitivity to glutaminase inhibition.24 Because this study examined metabolic dependency rather than irradiation, it identifies a context-specific vulnerability but does not by itself demonstrate radiosensitization. Radiotherapy-specific evidence is available for the LKB1-deficient context but remains limited. In a retrospective analysis of 164 patients with stage III NSCLC treated with radiotherapy, STK11 mutations were associated with a higher rate of locoregional recurrence. Isogenic cell-line and xenograft experiments in the same study showed that LKB1 loss reduced radiation-induced oxidative stress and increased radioresistance, whereas KEAP1 re-expression or glutaminase 1 (GLS1) inhibition increased radiosensitivity.7 These results support a functional interaction between LKB1 loss and KEAP1/NRF2-dependent redox buffering. However, KEAP1 alterations were not assessed in the clinical cohort, and the retrospective association does not establish STK11 as a treatment-predictive biomarker. A separate retrospective cohort using immunohistochemistry found that low LKB1 expression was associated with increased locoregional recurrence and inversely correlated with NRF2 and NAD(P)H quinone oxidoreductase 1 (NQO1) expression in non-squamous, but not squamous, NSCLC.8 This subgroup analysis was based on 74 evaluable tumors, used an outcome-derived LKB1 cutoff and lacked tumor genotyping, so the apparent histology dependence requires validation. In squamous NSCLC, loss of tumor protein p53 (TP53) provides a distinct experimental context. In a murine model initiated by combined Trp53 and Keap1 deletion in airway basal stem cells, Keap1 loss promoted NRF2 activity, aggressive tumor behavior and resistance to radiotherapy.6 These findings demonstrate NRF2-associated radioresistance in a TP53-deficient, Keap1-deficient squamous model, but do not establish TP53 co-mutation as an independent modifier of radiation response in patients. Collectively, the available evidence supports context-dependent interactions rather than a single co-mutation pattern applicable to all NRF2-activated NSCLC. Histology, co-alterations and direct measures of pathway activity should therefore be incorporated into future radiotherapy studies. Immune Microenvironment and Radiotherapy–Immunotherapy Combinations KEAP1–NFE2L2/NRF2 signaling may influence treatment response through immune remodeling. In antigenic murine lung adenocarcinoma, Keap1 loss reduced conventional type 1 dendritic cells and impaired CD8+ T-cell responses; genetic experiments implicated NRF2 hyperactivation, and Keap1-mutant tumors were less responsive to immune checkpoint blockade.25 Similar immune depletion was observed in human tumors. A redox-high lung adenocarcinoma phenotype associated with KEAP1, STK11, NFE2L2 and SMARCA4 alterations was linked to fewer tissue-resident memory CD8+ T cells and inferior immunotherapy outcomes.26 These data are largely confined to lung adenocarcinoma; the composite phenotype cannot isolate an NRF2-specific effect, and neither study evaluated irradiation. A recent review of acyl-CoA oxidase 2 (ACOX2) illustrates a broader metabolic–DDR–immune interface. In clear-cell renal cell carcinoma, ACOX2 was reported to destabilize the MRE11–RAD50–NBS1 complex, restrict homologous recombination repair, increase cytosolic DNA and activate cyclic GMP–AMP synthase–stimulator of interferon genes (cGAS–STING) signaling.27 Because these findings did not involve irradiation, NSCLC or KEAP1–NRF2 manipulation, they provide cross-cancer mechanistic context rather than direct evidence for NRF2-dependent radiotherapy–immunotherapy response. Radiotherapy can promote antigen release and immune activation but also lymphocyte depletion and myeloid immunosuppression.28 Impaired dendritic-cell priming or CD8+ T-cell infiltration could limit radiotherapy–immunotherapy cooperation, but this remains untested in NRF2-activated NSCLC. Retrospective stage III data suggest that durvalumab may mitigate mutation-associated locoregional failure, but nonrandomized designs and small molecular subgroups preclude a treatment-predictive interpretation.16,17 Prospective cohorts should integrate pathway activity, co-alterations, immune composition and radiotherapy-specific outcomes. NRF2-Mediated Maintenance of Redox Homeostasis Radiation-Induced Oxidative Stress Ionizing radiation damages tumor cells through both direct and indirect mechanisms. Direct ionization affects DNA and other biological macromolecules, whereas radiolysis of water generates reactive intermediates that contribute to oxidative injury.29 These effects extend beyond DNA to proteins, lipids and metabolites, exposing irradiated cells to an acute oxidative burden.30 In NSCLC, survival after irradiation therefore depends not only on the extent of the initial DNA damage, but also on the capacity of tumor cells to limit ROS accumulation and restore redox balance. Radiation-induced ROS contribute to DNA damage and cell death by promoting the accumulation of oxidative lesions.28 Persistent ROS production can amplify oxidative injury and increase the likelihood of irreversible cellular damage. By contrast, rapid mobilization of cellular reducing capacity and antioxidant defenses may limit the magnitude and duration of this injury. In NSCLC with sustained NRF2 activation, including subsets harboring functionally relevant KEAP1 or NFE2L2 alterations, enhanced redox buffering may attenuate early oxidative damage after irradiation. Constitutive NRF2 activation reduces radiation cytotoxicity in selected NSCLC cells,20 supporting redox homeostasis as a central component of NRF2-associated radioresistance. This evidence does not establish the same dependence in every NRF2-active tumor. NRF2 Target Genes and Redox Defense Sustained NRF2 activation shapes the cellular response to irradiation through a coordinated network of redox-defense mechanisms rather than through a single downstream target. NRF2 regulates glutathione (GSH)- and thioredoxin (Trx)-dependent antioxidant systems and also supports metabolite detoxification, reduced nicotinamide adenine dinucleotide phosphate (NADPH) regeneration and heme metabolism.31 Together, these processes increase cellular reducing capacity, limit ROS accumulation after irradiation and restrict the accumulation of oxidative damage. In lung cancer cells, constitutive NRF2 activation upregulates components of the GSH and Trx systems, together with pathways involved in detoxification and drug efflux. Conversely, NRF2 depletion increases basal intracellular ROS levels, indicating that NRF2 actively regulates cellular redox status rather than merely serving as a marker of cellular stress.32 NRF2 can also redirect glucose and glutamine metabolism to support antioxidant defense and cell proliferation.33 Under irradiation, these metabolic adaptations may enhance ROS neutralization and attenuate early oxidative injury. In A549 and H460 lung cancer cells, glutamine contributed to GSH synthesis and was linked to radiosensitivity, supporting GSH maintenance as an important metabolic component of the response to irradiation.34 These protective programs may be persistently maintained in NSCLC with sustained NRF2 activation, including subsets harboring functionally relevant KEAP1 or NFE2L2 alterations. Coordinated enhancement of antioxidant systems, NADPH-generating metabolism, detoxification and transport may reduce the extent to which radiation-induced ROS are converted into lethal cellular damage.23,31–33 The effects of NRF2 target genes therefore extend beyond direct ROS scavenging and may support survival after irradiation by maintaining a broader redox-protective state. The relative contribution of individual downstream programs may nevertheless vary according to tumor genotype, metabolic context and the degree of NRF2 pathway activation. Redox Defense and Radioresistance Ionizing radiation can itself activate NRF2-dependent antioxidant responses.35 In tumors with sustained NRF2 activity, this inducible response may reinforce constitutive redox buffering and increase post-irradiation survival. Experimental studies support a relationship between NRF2-regulated redox defense and radiation response, but they use different definitions of pathway involvement. RNA interference-mediated NRF2 depletion in A549 and H460 cells increased endogenous ROS and radiation-associated oxidative injury, while Nrf2 gain- and loss-of-function models demonstrated corresponding differences in radiation sensitivity.20 These experiments directly tested NRF2 function in selected cellular systems. By contrast, the adaptive response induced by low-dose irradiation in A549 cells was associated with ROS, autophagy and NRF2-related signaling but represented an induced phenotype in a single cell line rather than a genomically defined NRF2-active subgroup.36 Pharmacological studies provide a separate level of evidence. IM3829 reduced NRF2-binding activity and target-gene expression and enhanced the radiation response of H1299, A549 and H460 cells and lung cancer xenografts.21 Brusatol treatment in A549 cells was also accompanied by reduced NRF2-associated signaling, increased ROS and DNA damage and greater radiosensitivity.22 These findings support radiosensitization associated with pharmacological suppression of NRF2-regulated programs, but compound-based results alone do not establish NRF2-specific functional dependency. Collectively, the available experiments show that NRF2-mediated redox defense can modulate radiation response in particular models; they do not demonstrate that every KEAP1/NFE2L2-altered, NRF2-high or transcriptionally NRF2-active tumor is equally dependent on this pathway. NRF2 and the Tumor Immune Microenvironment Radiotherapy can engage antitumor immunity through cytosolic DNA sensing. In immunogenic tumor models, cyclic GMP–AMP synthase–stimulator of interferon genes (cGAS–STING) signaling in dendritic cells was required for radiation-induced type I interferon (IFN) production and CD8+ T-cell cross-priming.37 Whether NRF2-dependent redox buffering modifies damage-derived immune signaling after irradiation remains untested in molecularly defined NSCLC. More direct evidence links tumor-cell-intrinsic NRF2 activation to an immunosuppressive microenvironment outside the radiotherapy setting. In an antigenic KRAS-driven lung adenocarcinoma model, Keap1 mutation reduced CD103 dendritic-cell accumulation and impaired CD8 T-cell infiltration, activation and effector cytokine production, with weaker responses to anti-programmed cell death protein 1 (PD-1) therapy.25 Human KEAP1-mutant lung adenocarcinomas also showed lower CD3, CD8 and PD-1-positive CD8 T-cell infiltration. CRISPR-mediated Nrf2 deletion in Keap1-mutant tumor cells restored CD103 dendritic-cell accumulation and T-cell cytokine production, supporting a causal contribution of tumor-cell-intrinsic NRF2 activation. However, the mouse model relied on H-Y-driven antigenicity and did not include irradiation. Evidence combining irradiation, immunotherapy and NRF2 measurement remains preliminary. In Lewis lung carcinoma, 1 Gy × 5 plus anti-PD-1 increased CD8+ T-cell infiltration and IFN-γ and was accompanied by lower NRF2, heme oxygenase 1 (HO-1) and glutathione peroxidase 4 (GPX4) expression and increased ferroptosis-associated readouts.38 A Phase I component treated six patients with chemoimmunotherapy-resistant NSCLC using a total dose of 6 Gy delivered in five 1.2-Gy fractions plus anti-PD-1 and reported an objective response rate of 33.3%. The preclinical experiments lacked NRF2 perturbation or rescue, and the clinical component did not assess KEAP1/NFE2L2 status or functional activity. Clinical observations after concurrent chemoradiotherapy and durvalumab likewise do not establish a treatment-predictive interaction because the studies were retrospective, used genomic classification and yielded nonuniform adjusted results.16,17 Validation requires immunocompetent isogenic NSCLC models with defined pathway status, clinically relevant fractionation, immune profiling and tumor-control endpoints, followed by prospective molecularly annotated cohorts. Emerging NRF2-Associated Mechanisms in the DNA Damage Response DNA Damage Response Following Ionizing Radiation Among the lesions induced by ionizing radiation, DNA double-strand breaks (DSBs) are particularly consequential. Unrepaired or incorrectly repaired DSBs can lead to chromosomal instability, sustained cell cycle arrest and cell death.39 The fate of an irradiated cell therefore depends not only on the initial extent of DNA damage, but also on its capacity to recognize the damage, activate cell cycle checkpoints and complete repair. These processes determine whether the lesions are resolved or persist sufficiently to compromise long-term proliferative capacity. The DDR coordinates damage recognition, signal transduction, checkpoint activation and DNA repair. Ataxia-telangiectasia mutated (ATM), ataxia telangiectasia and Rad3-related (ATR), and DNA-dependent protein kinase (DNA-PK) are major upstream kinases that respond to distinct but overlapping forms of genomic stress.40 ATM and DNA-PK are prominently involved in responses to DSBs, whereas ATR is recruited primarily to replication protein A-coated single-stranded DNA generated by replication stress or DNA-end resection. Through their downstream signaling networks, these kinases regulate checkpoint activation and repair, thereby influencing whether irradiated tumor cells recover, undergo prolonged arrest or lose their capacity for continued proliferation. Although sustained NRF2 activation can attenuate radiation-induced oxidative stress, irradiation still generates DSBs and replication-associated lesions that require checkpoint control and DNA repair.12,13,39–42 Evidence assigning NRF2 a direct role in DDR signaling is less mature than evidence supporting its redox functions. The relevant mechanistic observations are derived primarily from two related experimental studies from the same investigative group and require interpretation according to the cell systems, irradiation schedules and functional endpoints used.12,13 These findings generate a plausible mechanism linking NRF2 to damage processing and checkpoint regulation, but do not establish DDR remodeling as a universal feature of NRF2-activated NSCLC. Proposed NRF2 Involvement in the RPA32–TOPBP1–ATR–CHK1 Pathway The general organization of ATR signaling is established independently of NRF2. During replication stress or DNA-end resection, exposed single-stranded DNA is bound by replication protein A (RPA), forming a platform for recruitment of the ATR–ATRIP complex. ATR-interacting protein (ATRIP) mediates ATR localization to RPA-coated single-stranded DNA, whereas DNA topoisomerase II-binding protein 1 (TOPBP1) promotes ATR activation and downstream signaling through checkpoint kinase 1 (CHK1).41,43–45 During DSB processing, DNA-end resection can therefore generate the substrate required for ATR-dependent checkpoint signaling, complementing the predominantly ATM-dependent early response.46 These relationships represent established DDR biology; the unresolved question is whether NRF2 directly and reproducibly modifies this signaling architecture in NSCLC after irradiation. Two related studies from the same research group provide the principal evidence linking NRF2 directly to ATR-dependent damage signaling. The earlier study reported that NRF2 depletion in A549, H460 and H1299 cells attenuated radiation-induced ATR–CHK1 activation and G2 arrest, whereas NRF2 re-expression rescued these effects in A549 cells; homologous recombination and xenograft radiation responses were also evaluated.13 A subsequent study using parental and experimentally selected radioresistant A549, H460 and H1299 cells reported NRF2-dependent RPA32 phosphorylation, TOPBP1 recruitment and ATR–CHK1 activation after DNA damage.12 The detailed RPA32–TOPBP1–ATR–CHK1 model remains proposed. Some experiments used camptothecin, laser microirradiation or targeted DSBs rather than therapeutic irradiation; patient analyses addressed prognosis rather than radiotherapy-specific outcomes. The complete pathway has not been independently reproduced in molecularly defined KEAP1/NFE2L2 NSCLC systems, and its prevalence, causal importance and predictive value remain unknown. Independent studies should connect genetically validated damage-site interactions to repair kinetics and clonogenic survival under clinically relevant fractionation. DNA Repair and Post-Irradiation Survival Whether the DDR contributes to radioresistance ultimately depends on the ability of tumor cells to resolve radiation-induced damage and retain long-term proliferative capacity. Histone H2AX phosphorylated at Ser139 (γH2AX) is widely used to monitor DNA damage and repair kinetics after irradiation. Early γH2AX accumulation primarily reflects damage induction, whereas persistent γH2AX foci are commonly interpreted as evidence of residual or difficult-to-repair lesions.47 Residual γH2AX foci have also been associated with clonogenic survival, linking the persistence of DNA damage to long-term proliferative capacity after irradiation.48 However, fewer residual γH2AX foci do not necessarily indicate enhanced repair, as they may also result from a lower initial damage burden. Assessment of γH2AX kinetics, rather than a single post-irradiation measurement, is therefore required to distinguish reduced damage induction from accelerated damage resolution. Cell cycle checkpoints can provide irradiated cells with time to process DNA damage before mitotic entry. The G2 checkpoint is particularly relevant in this setting, although the amount of damage required for its activation and the duration of checkpoint arrest vary among cellular backgrounds.49 In patient-derived xenograft models of NSCLC brain metastases, the ATR inhibitor M6620 enhanced the response to radiotherapy and was accompanied by reduced clonogenic survival, impaired DSB repair, attenuation of G2 checkpoint control and increased mitotic cell death.50 Similarly, combining AZD6738 with irradiation increased radiosensitivity in NSCLC models and produced greater tumor growth delay in H460 xenografts.51 These studies support the contribution of ATR-dependent checkpoint signaling and DNA repair to post-irradiation survival. They do not, however, demonstrate that tumors with sustained NRF2 activation are uniformly dependent on ATR signaling. Evaluation of this proposed NRF2-associated DDR mechanism should not rely on a single marker. γH2AX immunofluorescence or immunoblotting measures a damage-associated signal but is not specific to one lesion or repair process, whereas comet assays estimate strand-break burden without identifying the operative repair pathway. Checkpoint assays assess signaling, and clonogenic assays determine long-term proliferative survival. Recent methodological reviews therefore emphasize combining complementary readouts and accounting for assay-specific limitations; single-cell and advanced imaging approaches may improve resolution but do not replace irradiation-specific functional endpoints.52 In molecularly defined NSCLC, comparable initial damage followed by more rapid damage resolution, sustained checkpoint control and greater clonogenic survival after irradiation would provide stronger support for an NRF2-associated repair phenotype. Genetic NRF2 perturbation and rescue would be required to establish causal dependency. This framework has not been validated consistently across NRF2-activated NSCLC models. Key preclinical studies linking NRF2-related molecular context or perturbation to radiation-response endpoints are summarized in Table 2. Table 2Preclinical Evidence Linking NRF2-Related Molecular Context or Perturbation to Radiation ResponseStudy (Ref).Model and Molecular ContextRadiation SchedulePerturbation/InterventionFunctional EndpointsMain FindingsKey LimitationsJeong et al, 20176Murine TP53-null LSCC (P-LSCC) and TP53/Keap1 double-deletion LSCC (K/P-LSCC); tumorspheres and xenografts.Clonogenic/tumorsphere experiments, 0–10 Gy; 5 Gy for mechanistic assays; tumors received a single 6-Gy fraction.Genetic Keap1 deletion; xCT inhibition with sulfasalazine; NAC rescue.ROS, DNA damage, clonogenic and tumorsphere survival, tumor growth and survival.Keap1 loss activated NRF2 and increased radioresistance; xCT inhibition partially restored radiosensitivity.TP53-null murine squamous context; single-fraction tumor irradiation; no normal-tissue assessment.Sitthideatphaiboon et al, 20217A549, H460 and Calu-6 LKB1-isogenic models; A549 KEAP1-isogenic models; A549 xenografts.Clonogenic dose series; 8 Gy for mechanistic experiments; xenografts received a single 5-Gy fraction.LKB1 gain/loss; KEAP1 re-expression; CB-839 (10 nM in clonogenic assays; 1 µM for 4 h before 8 Gy in mechanistic assays).ROS, apoptosis, clonogenic survival, tumor growth and time to doubling.LKB1 loss increased radioresistance through a KEAP1/NRF2-dependent redox state; KEAP1 restoration or GLS1 inhibition increased radiosensitivity.Few isogenic models; single-fraction irradiation; clinical cohort did not assess KEAP1; no normal-lung assessment.Binkley et al, 202011H1299 KEAP1-null cells vs parental cells; H1437 cells with KEAP1 knockdown; KEAP1-mutant A549 cells with or without wild-type KEAP1 re-expression.Single-dose 2–5 Gy; CB-839 pretreatment for 24 h (0.1–500 nM); 2 or 4 Gy in mechanistic assays.Genetic KEAP1 loss or re-expression; GLS1 inhibitor CB-839 (telaglenastat); NAC rescue.Clonogenic survival, ROS, GSH/GSSG and γH2AX.CB-839 preferentially radiosensitized KEAP1-deficient models; NAC rescue and genetic reversal supported a redox/metabolic dependency.In vitro only; selected cell lines; acute single-dose irradiation; no tumor or normal-lung therapeutic-ratio assessment.Sun et al, 202412Parental A549, H460 and H1299 cells and derived A549R, H460R and H1299R radioresistant cells; A549 NRF2 knockout/rescue; A549R lung-colonization tumors.Radioresistant cells were selected with repeated 2-Gy exposures every 6–9 days to a cumulative 20 Gy; 6 Gy for survival challenge and 4–8 Gy for mechanistic assays; lung tumors received 2+4+2 Gy weekly (total, 8 Gy).NRF2 knockdown/knockout and rescue; RPA32 or TOPBP1 knockdown/overexpression.RPA32 phosphorylation, TOPBP1 recruitment, ATR–CHK1 signaling, DNA-damage foci, clonogenic survival and tumor burden.The study proposed an NRF2–RPA32–TOPBP1–ATR mechanism associated with the acquired radioresistant phenotype.Selected radioresistant lines may contain unrelated adaptations; molecular KEAP1/NFE2L2 context was not fully defined; same investigative group as Ref. 13; the complete pathway was not genetically disrupted and rescued in vivo; lung-colonization rather than orthotopic primary-lung model.Sun et al, 202013A549, H460 and H1299 cells; A549 NRF2 knockout and rescue; DR-GFP U2OS HR reporter; A549 xenografts.Clonogenic dose series, 0–6 Gy; 8-Gy 137Cs γ-irradiation for principal mechanistic experiments; xenografts received a single 8-Gy fraction.NRF2 knockdown/knockout and rescue; ATR inhibitor VE-821; NAC; brusatol 2 mg/kg in xenografts.ATR–CHK1 signaling, G2 arrest, HR, RAD51/BRCA1 foci, clonogenic survival and tumor growth.NRF2 supported ATR-dependent G2 checkpoint signaling, HR-associated readouts and survival; brusatol increased xenograft radiation response.HR reporter used non-NSCLC U2OS cells; several assays used camptothecin rather than irradiation; brusatol specificity limitations; NRF2-knockout A549 cells did not form tumors, precluding in vivo genetic rescue; same investigative group as Ref. 12.Singh et al, 201020Human A549 and H460 NSCLC cells with constitutive NRF2 activity; Keap1−/− and Nrf2−/− MEFs.Single-dose γ-irradiation; clonogenic dose series up to 6 Gy; mechanistic experiments commonly used 6 Gy in A549 and 4 Gy in H460 cells.Stable NRF2 shRNA; genetic gain/loss in MEFs; antioxidant rescue with NAC, GSH or Trolox.ROS and protein oxidation, apoptosis and clonogenic survival.NRF2 depletion increased radiation-associated oxidative stress and radiosensitivity; antioxidant rescue and genetic gain/loss supported an NRF2-dependent redox-protective phenotype.Selected cell lines and MEFs; no in vivo tumor-control experiment; not a clinically defined genotype cohort.Lee et al, 201221H1299, A549 and H460 cells; H1299 and A549 xenografts.Single-dose irradiation in vitro; xenografts received local 8-Gy 60Co γ-irradiation.NRF2 inhibitor IM3829; 5 mg/kg/day in vivo.ARE binding and target expression, ROS, apoptosis, clonogenic survival and tumor growth.IM3829 increased radiation response in high-NRF2 cells and xenografts.Pharmacologic specificity and off-target effects were not fully established; acute single-dose irradiation; no genetic rescue.Sun et al, 201622A549 cells, selected after screening H1299, A549 and H460 cells.Clonogenic dose series, 0–6 Gy; 80 nM brusatol plus 6 Gy for principal mechanistic experiments; 10 Gy in selected viability assays.Brusatol.ROS, viability, clonogenic survival, comet assay, γH2AX, apoptosis and cell cycle.Brusatol reduced NRF2-associated signaling and increased radiation-associated damage and radiosensitivity.Predominantly one cell line; brusatol has limited NRF2 specificity and affects global protein synthesis; no in vivo experiment.Chen et al, 201536A549 cells.A 0.05-Gy α-particle priming dose followed by a 0.75-Gy challenge dose.shNRF2; HO-1 inhibition with ZnPP; autophagy inhibition and ROS modulation.Cell survival, autophagy, NRF2/HO-1 signaling and ROS.A low-dose adaptive response was linked to ROS–autophagy–NRF2–HO-1 signaling.Single cell line; α-particle exposure and nonclinical schedule; induced signaling rather than a genomically defined NRF2 state.Luo et al, 202538LLC and LLC-OVA cells; syngeneic subcutaneous LLC tumors; murine CD8+ T-cell co-culture.Tumors received 1 Gy/day on days 8–13 for five fractions (total, 5 Gy); cells received 0.5 Gy once.Anti-PD-1 in mice (200 µg/day) or anti-PD-L1 in co-culture; ML385 was used as a positive-control NRF2 inhibitor.Tumor growth, CD8+ infiltration and IFN-γ, ROS, GSH, MDA, Fe2+, lipid peroxidation and NRF2/HO-1/GPX4 signaling.Low-dose RT plus immune-checkpoint blockade increased antitumor immunity and ferroptosis-associated readouts, accompanied by reduced NRF2/HO-1/GPX4 signaling.No NRF2 genetic perturbation/rescue; LLC was not molecularly defined by KEAP1/NFE2L2 status; small groups; pathway change was correlative rather than causal; low-dose schedule.Note: ARE, antioxidant response element; ATR, ataxia telangiectasia and Rad3-related protein; BRCA1, breast cancer gene 1; CHK1, checkpoint kinase 1; DR-GFP, direct-repeat green fluorescent protein; Fe2+, ferrous iron; GPX4, glutathione peroxidase 4; GSH, reduced glutathione; GSSG, oxidized glutathione; HO-1, heme oxygenase 1; HR, homologous recombination; ICI, immune checkpoint inhibitor; IFN-γ, interferon-γ; IR, ionizing radiation; KEAP1, Kelch-like ECH-associated protein 1; LKB1, liver kinase B1; LLC, Lewis lung carcinoma; LSCC, lung squamous cell carcinoma; MDA, malondialdehyde; MEF, mouse embryonic fibroblast; NAC, N-acetylcysteine; NRF2, nuclear factor erythroid 2-related factor 2; NSCLC, non-small cell lung cancer; OVA, ovalbumin; RAD51, RAD51 recombinase; ROS, reactive oxygen species; RPA32, 32-kDa subunit of replication protein A; RT, radiotherapy; shRNA, short hairpin RNA; TOPBP1, DNA topoisomerase II-binding protein 1; TP53, tumor protein p53; xCT, cystine/glutamate antiporter; γH2AX, histone H2AX phosphorylated at serine 139; ZnPP, zinc protoporphyrin. Table 2Preclinical Evidence Linking NRF2-Related Molecular Context or Perturbation to Radiation ResponseStudy (Ref).Model and Molecular ContextRadiation SchedulePerturbation/InterventionFunctional EndpointsMain FindingsKey LimitationsJeong et al, 20176Murine TP53-null LSCC (P-LSCC) and TP53/Keap1 double-deletion LSCC (K/P-LSCC); tumorspheres and xenografts.Clonogenic/tumorsphere experiments, 0–10 Gy; 5 Gy for mechanistic assays; tumors received a single 6-Gy fraction.Genetic Keap1 deletion; xCT inhibition with sulfasalazine; NAC rescue.ROS, DNA damage, clonogenic and tumorsphere survival, tumor growth and survival.Keap1 loss activated NRF2 and increased radioresistance; xCT inhibition partially restored radiosensitivity.TP53-null murine squamous context; single-fraction tumor irradiation; no normal-tissue assessment.Sitthideatphaiboon et al, 20217A549, H460 and Calu-6 LKB1-isogenic models; A549 KEAP1-isogenic models; A549 xenografts.Clonogenic dose series; 8 Gy for mechanistic experiments; xenografts received a single 5-Gy fraction.LKB1 gain/loss; KEAP1 re-expression; CB-839 (10 nM in clonogenic assays; 1 µM for 4 h before 8 Gy in mechanistic assays).ROS, apoptosis, clonogenic survival, tumor growth and time to doubling.LKB1 loss increased radioresistance through a KEAP1/NRF2-dependent redox state; KEAP1 restoration or GLS1 inhibition increased radiosensitivity.Few isogenic models; single-fraction irradiation; clinical cohort did not assess KEAP1; no normal-lung assessment.Binkley et al, 202011H1299 KEAP1-null cells vs parental cells; H1437 cells with KEAP1 knockdown; KEAP1-mutant A549 cells with or without wild-type KEAP1 re-expression.Single-dose 2–5 Gy; CB-839 pretreatment for 24 h (0.1–500 nM); 2 or 4 Gy in mechanistic assays.Genetic KEAP1 loss or re-expression; GLS1 inhibitor CB-839 (telaglenastat); NAC rescue.Clonogenic survival, ROS, GSH/GSSG and γH2AX.CB-839 preferentially radiosensitized KEAP1-deficient models; NAC rescue and genetic reversal supported a redox/metabolic dependency.In vitro only; selected cell lines; acute single-dose irradiation; no tumor or normal-lung therapeutic-ratio assessment.Sun et al, 202412Parental A549, H460 and H1299 cells and derived A549R, H460R and H1299R radioresistant cells; A549 NRF2 knockout/rescue; A549R lung-colonization tumors.Radioresistant cells were selected with repeated 2-Gy exposures every 6–9 days to a cumulative 20 Gy; 6 Gy for survival challenge and 4–8 Gy for mechanistic assays; lung tumors received 2+4+2 Gy weekly (total, 8 Gy).NRF2 knockdown/knockout and rescue; RPA32 or TOPBP1 knockdown/overexpression.RPA32 phosphorylation, TOPBP1 recruitment, ATR–CHK1 signaling, DNA-damage foci, clonogenic survival and tumor burden.The study proposed an NRF2–RPA32–TOPBP1–ATR mechanism associated with the acquired radioresistant phenotype.Selected radioresistant lines may contain unrelated adaptations; molecular KEAP1/NFE2L2 context was not fully defined; same investigative group as Ref. 13; the complete pathway was not genetically disrupted and rescued in vivo; lung-colonization rather than orthotopic primary-lung model.Sun et al, 202013A549, H460 and H1299 cells; A549 NRF2 knockout and rescue; DR-GFP U2OS HR reporter; A549 xenografts.Clonogenic dose series, 0–6 Gy; 8-Gy 137Cs γ-irradiation for principal mechanistic experiments; xenografts received a single 8-Gy fraction.NRF2 knockdown/knockout and rescue; ATR inhibitor VE-821; NAC; brusatol 2 mg/kg in xenografts.ATR–CHK1 signaling, G2 arrest, HR, RAD51/BRCA1 foci, clonogenic survival and tumor growth.NRF2 supported ATR-dependent G2 checkpoint signaling, HR-associated readouts and survival; brusatol increased xenograft radiation response.HR reporter used non-NSCLC U2OS cells; several assays used camptothecin rather than irradiation; brusatol specificity limitations; NRF2-knockout A549 cells did not form tumors, precluding in vivo genetic rescue; same investigative group as Ref. 12.Singh et al, 201020Human A549 and H460 NSCLC cells with constitutive NRF2 activity; Keap1−/− and Nrf2−/− MEFs.Single-dose γ-irradiation; clonogenic dose series up to 6 Gy; mechanistic experiments commonly used 6 Gy in A549 and 4 Gy in H460 cells.Stable NRF2 shRNA; genetic gain/loss in MEFs; antioxidant rescue with NAC, GSH or Trolox.ROS and protein oxidation, apoptosis and clonogenic survival.NRF2 depletion increased radiation-associated oxidative stress and radiosensitivity; antioxidant rescue and genetic gain/loss supported an NRF2-dependent redox-protective phenotype.Selected cell lines and MEFs; no in vivo tumor-control experiment; not a clinically defined genotype cohort.Lee et al, 201221H1299, A549 and H460 cells; H1299 and A549 xenografts.Single-dose irradiation in vitro; xenografts received local 8-Gy 60Co γ-irradiation.NRF2 inhibitor IM3829; 5 mg/kg/day in vivo.ARE binding and target expression, ROS, apoptosis, clonogenic survival and tumor growth.IM3829 increased radiation response in high-NRF2 cells and xenografts.Pharmacologic specificity and off-target effects were not fully established; acute single-dose irradiation; no genetic rescue.Sun et al, 201622A549 cells, selected after screening H1299, A549 and H460 cells.Clonogenic dose series, 0–6 Gy; 80 nM brusatol plus 6 Gy for principal mechanistic experiments; 10 Gy in selected viability assays.Brusatol.ROS, viability, clonogenic survival, comet assay, γH2AX, apoptosis and cell cycle.Brusatol reduced NRF2-associated signaling and increased radiation-associated damage and radiosensitivity.Predominantly one cell line; brusatol has limited NRF2 specificity and affects global protein synthesis; no in vivo experiment.Chen et al, 201536A549 cells.A 0.05-Gy α-particle priming dose followed by a 0.75-Gy challenge dose.shNRF2; HO-1 inhibition with ZnPP; autophagy inhibition and ROS modulation.Cell survival, autophagy, NRF2/HO-1 signaling and ROS.A low-dose adaptive response was linked to ROS–autophagy–NRF2–HO-1 signaling.Single cell line; α-particle exposure and nonclinical schedule; induced signaling rather than a genomically defined NRF2 state.Luo et al, 202538LLC and LLC-OVA cells; syngeneic subcutaneous LLC tumors; murine CD8+ T-cell co-culture.Tumors received 1 Gy/day on days 8–13 for five fractions (total, 5 Gy); cells received 0.5 Gy once.Anti-PD-1 in mice (200 µg/day) or anti-PD-L1 in co-culture; ML385 was used as a positive-control NRF2 inhibitor.Tumor growth, CD8+ infiltration and IFN-γ, ROS, GSH, MDA, Fe2+, lipid peroxidation and NRF2/HO-1/GPX4 signaling.Low-dose RT plus immune-checkpoint blockade increased antitumor immunity and ferroptosis-associated readouts, accompanied by reduced NRF2/HO-1/GPX4 signaling.No NRF2 genetic perturbation/rescue; LLC was not molecularly defined by KEAP1/NFE2L2 status; small groups; pathway change was correlative rather than causal; low-dose schedule.Note: ARE, antioxidant response element; ATR, ataxia telangiectasia and Rad3-related protein; BRCA1, breast cancer gene 1; CHK1, checkpoint kinase 1; DR-GFP, direct-repeat green fluorescent protein; Fe2+, ferrous iron; GPX4, glutathione peroxidase 4; GSH, reduced glutathione; GSSG, oxidized glutathione; HO-1, heme oxygenase 1; HR, homologous recombination; ICI, immune checkpoint inhibitor; IFN-γ, interferon-γ; IR, ionizing radiation; KEAP1, Kelch-like ECH-associated protein 1; LKB1, liver kinase B1; LLC, Lewis lung carcinoma; LSCC, lung squamous cell carcinoma; MDA, malondialdehyde; MEF, mouse embryonic fibroblast; NAC, N-acetylcysteine; NRF2, nuclear factor erythroid 2-related factor 2; NSCLC, non-small cell lung cancer; OVA, ovalbumin; RAD51, RAD51 recombinase; ROS, reactive oxygen species; RPA32, 32-kDa subunit of replication protein A; RT, radiotherapy; shRNA, short hairpin RNA; TOPBP1, DNA topoisomerase II-binding protein 1; TP53, tumor protein p53; xCT, cystine/glutamate antiporter; γH2AX, histone H2AX phosphorylated at serine 139; ZnPP, zinc protoporphyrin. Preclinical Evidence Linking NRF2-Related Molecular Context or Perturbation to Radiation Response Note: ARE, antioxidant response element; ATR, ataxia telangiectasia and Rad3-related protein; BRCA1, breast cancer gene 1; CHK1, checkpoint kinase 1; DR-GFP, direct-repeat green fluorescent protein; Fe2+, ferrous iron; GPX4, glutathione peroxidase 4; GSH, reduced glutathione; GSSG, oxidized glutathione; HO-1, heme oxygenase 1; HR, homologous recombination; ICI, immune checkpoint inhibitor; IFN-γ, interferon-γ; IR, ionizing radiation; KEAP1, Kelch-like ECH-associated protein 1; LKB1, liver kinase B1; LLC, Lewis lung carcinoma; LSCC, lung squamous cell carcinoma; MDA, malondialdehyde; MEF, mouse embryonic fibroblast; NAC, N-acetylcysteine; NRF2, nuclear factor erythroid 2-related factor 2; NSCLC, non-small cell lung cancer; OVA, ovalbumin; RAD51, RAD51 recombinase; ROS, reactive oxygen species; RPA32, 32-kDa subunit of replication protein A; RT, radiotherapy; shRNA, short hairpin RNA; TOPBP1, DNA topoisomerase II-binding protein 1; TP53, tumor protein p53; xCT, cystine/glutamate antiporter; γH2AX, histone H2AX phosphorylated at serine 139; ZnPP, zinc protoporphyrin. Identification of Radioresistance and Potential Intervention Strategies Candidate identification should integrate rather than conflate genomic, transcriptional and functional evidence. KEAP1/NFE2L2 sequencing identifies alterations, NRF2 target-gene expression and the K1N2 score estimate transcriptional activity, and radiation assays measure cellular phenotype.18,19 None is interchangeable or currently validated as a clinical radiation biomarker. Future molecular enrichment should combine variant pathogenicity with sustained pathway activity and, when an intervention is proposed, direct dependence on the targeted process. Potential interventions need not be directed against NRF2 itself. Some tumors with sustained NRF2 activation may develop increased dependence on glutamine metabolism, glutathione maintenance and redox buffering. In experimental KEAP1-mutant models, glutaminase inhibition increased radiosensitivity and was accompanied by glutathione depletion and increased radiation-associated DNA damage.11 LKB1-deficient models provide complementary evidence that KEAP1/NRF2-dependent redox buffering can support radioresistance and that GLS1 inhibition can increase radiosensitivity.7 These observations suggest that downstream metabolic dependencies may offer more readily testable intervention points than direct NRF2 inhibition, particularly when sustained pathway activation and the relevant metabolic dependence can both be demonstrated. The mechanistic relationship between NRF2 and ATR–CHK1 signaling also raises the possibility of a checkpoint-related vulnerability, but this remains inferential and requires direct validation in molecularly defined NSCLC radiotherapy models. Normal lung responses impose a major constraint on NRF2-related radiosensitization. NRF2 deficiency shortened survival and aggravated late tissue injury after thoracic irradiation in mice,53 whereas NRF2 overexpression attenuated acute histopathological injury, inflammatory-cell infiltration, cytokine alterations and oxidative damage after thoracic irradiation.54 Preferential activity against an NRF2-dependent tumor therefore cannot be equated with protection of the irradiated normal lung. The small-molecule NRF2 inhibitor ML385 preferentially impaired clonogenic growth in KEAP1-mutant H460 cells compared with isogenic cells in which wild-type KEAP1 was restored and did not inhibit the growth of non-tumorigenic BEAS-2B lung epithelial cells. ML385 also enhanced the antitumor activity of carboplatin in subcutaneous and orthotopic NSCLC models.9 However, ML385 was not evaluated in combination with thoracic radiotherapy, and these experiments did not assess radiation-induced pneumonitis or fibrosis. These findings provide preliminary evidence of pathway-associated differential activity, but do not establish a safe therapeutic window for NRF2 inhibition during lung irradiation. A more defensible development strategy would combine molecular enrichment with restricted target exposure. Candidate tumors should demonstrate not only a KEAP1 or NFE2L2 alteration, but also sustained NRF2 transcriptional activity and functional dependence on the metabolic or redox process being targeted. Targeting a validated downstream dependency, such as GLS1 dependence in selected KEAP1/NRF2-active models, may offer a means of avoiding systemic suppression of the broader NRF2 protective program; however, the available radiosensitization studies did not assess normal-lung injury.7,11 Tumor-targeted delivery and short-duration dosing around irradiation represent additional research strategies for limiting normal-tissue exposure, but neither approach has established a therapeutic window for NRF2-directed radiosensitization in thoracic radiotherapy. Before clinical translation, molecularly defined orthotopic or immunocompetent models using fractionated thoracic irradiation should assess tumor control together with acute lung inflammation and late pulmonary fibrosis. Further development would require improved tumor response without a material increase in normal-lung injury. Conclusions and Perspectives Aberrant activation of the KEAP1–NFE2L2/NRF2 axis may mark a molecular state associated with reduced radiosensitivity in NSCLC. Clinical and experimental evidence is strongest for NRF2-dependent redox protection. A smaller body of experimental evidence, derived mainly from two related reports from the same investigative group, suggests that NRF2 may also participate in damage-site signaling and ATR–CHK1 checkpoint control. The proposed DDR component, including RPA32 phosphorylation and TOPBP1 recruitment, has not been independently reproduced in molecularly defined NSCLC models or validated in prospective clinical radiotherapy cohorts. NRF2-associated radioresistance should therefore be viewed as a well-supported redox phenotype with an emerging DDR component. However, KEAP1 or NFE2L2 alterations, NRF2 protein abundance and transcriptional pathway scores have not been validated as interchangeable biomarkers for radiotherapy stratification, and no NRF2-based radiosensitizing strategy has been established for clinical use. Future studies should determine whether sustained pathway activation is accompanied by reproducible functional dependencies that can be safely targeted, rather than inferring therapeutic vulnerability from genomic status or protein expression alone. Prospective studies of patients receiving standardized radiotherapy should jointly assess variant pathogenicity, NRF2 protein abundance and localization, transcriptional activity, radiation-related functional readouts, immune-cell composition and, where applicable, immunotherapy exposure, together with local-control outcomes and normal-lung injury. Such evidence, together with preclinical demonstration of improved tumor control without a material increase in acute or late normal-lung injury, is required before biomarker-guided stratification or combination strategies can be evaluated clinically; these remain future research objectives rather than imminent applications.