EGFR Mutations Compromise Hypoxia-Associated Radiation Resistance through Impaired Replication Fork–Associated DNA Damage Repair

Relative to well-oxygenated tumors, hypoxic tumors tend to be significantly more resistant to ionizing radiation (IR). Radiation-induced free radicals chemically react with DNA and cause DNA damage through the formation of a DNA radical. For DNA damage to persist, oxygen is required to oxidize the DNA radical and extend its half-life. Hypoxic cells are more radioresistant because, in conditions of low oxygen, oxidation of the DNA radical and radiation-induced DNA damage are limited (1). Oxygen enhances radiosensitivity by a radiation dose-modifying factor (DMF) called oxygen enhancement ratio (OER). OER is defined as the ratio of the radiation dose under anoxic or hypoxic conditions to the dose under conditions of a specific partial pressure of oxygen to produce the same biological effect. Alternatively, the hypoxia reduction factor (HRF) has been defined as the ratio of radiation dose at a specific partial pressure of oxygen to the dose under fully aerobic conditions (21% O₂) for the same biological effect (2).

Prolonged exposure to hypoxia itself can be cytotoxic, and tumor survival depends on an adaptive prosurvival response to hypoxia that could also augment radioresistance. Activation and downstream signaling of the EGFR in many cancers, including non–small cell lung cancer (NSCLC), has been linked to survival responses to hypoxia. These include inhibition of hypoxia-induced cell death (3), stimulation of pro-proliferative pathways (4), induction of epithelial–mesenchymal transition (5), and establishment of a positive feedback loop through induction of the hypoxia-inducible factor 1 (HIF-1α; ref. 6).

EGFR also has roles in the repair of radiation-induced DNA double-strand breaks (DSB). DSB repair in tumors is catalyzed by two signaling pathways, nonhomologous end joining (NHEJ) and homologous recombination (HR). Several studies have demonstrated a role for EGFR in NHEJ, which involves radiation-induced nuclear translocation of EGFR, interactions with the NHEJ enzyme, DNA-dependent protein kinase catalytic subunit (DNA-PKcs), and modulation of DNA-PKcs phosphorylation (7–11). Other studies have shown that EGFR-mediated DSB repair may also involve HR (12, 13). Recent studies have uncovered a less understood, kinase-independent role for EGFR in HR-mediated repair of cisplatin-induced DNA interstrand crosslinks, but not radiation-induced DSBs (14).

We previously demonstrated that NHEJ-mediated DSB repair is defective in NSCLCs harboring somatic, activating mutations in EGFR that were clinically linked to sensitivity to EGFR tyrosine kinase inhibitors (TKI; refs. 15–17). We showed that, despite elevated EGFR tyrosine phosphorylation and downstream signaling, NSCLCs harboring ΔE746-E750, L858R, or the TKI-resistant T790M mutant (MT) EGFR are profoundly radiosensitive (18). MT-EGFR–associated radiosensitivity manifests as pronounced delays in repair of radiation-induced DSBs, failure of EGFR nuclear import, abrogation of EGFR–DNA-PKcs interactions (8), and inhibition of DNA-PKcs phosphorylation at threonine 2609 (T2609; ref. 11), a pattern that is remarkably similar to EGFR blockade by the mAb, cetuximab (19). How expression of signaling-hyperactive, NHEJ-defective MT-EGFR affects hypoxia response is not known.

Evidence from a number of studies shows that hypoxia induces a unique DNA damage response (DDR), which is a cascade of signaling events that collectively orchestrate DNA damage recognition, DNA repair, cell-cycle arrest, or apoptosis (20). DNA DSBs do not usually occur in hypoxic cells (21). However, in response to hypoxia, decrease in ribonucleotide reductase (RNR) activity (22) causes nucleotide insufficiency, resulting in a rapid halt of DNA synthesis and transient cell-cycle arrest through replication fork–bound replication protein A (RPA) and activation of the ataxia telangiectasia–related checkpoint kinase (ATR) Cdc25 pathway (ATR/Chk1/Cdc25 pathway; ref. 23). Stalled replication forks are at a high risk for DNA DSBs, and repair of fork-associated DSBs is essential for survival. During conditions of transient hypoxia (<12 hours), HR but not NHEJ appears to be critical for survival (24). In contrast, during prolonged hypoxia (>12 hours), many components of the HR pathway are downregulated with a significant upregulation of several NHEJ components (25).

This study investigates the effect of activating L858R and ΔE746-E750 mutations in EGFR on radiation resistance in conditions of hypoxia. Our study has uncovered a novel relationship between MT-EGFR expression and replication stress in oxygen-replete and hypoxic conditions. We demonstrate that in the context of an altered DDR in hypoxic cells, expression of NHEJ-defective MT-EGFR, or an EGFR blockade by cetuximab, exerts a synthetic lethality effect and sensitizes hypoxic cells to radiation. This study has important clinical implications in the treatment of NSCLC patients, especially those at greatest risk of therapy failure due to tumor hypoxia.

NSCLC cell lines with either WT-EGFR (NCI-A549), or ΔE746-E750-T790M (NCI-H820), ΔE746-E750 (NCI-HCC827) or L858R-T790M (NCI-H1975) mutated forms of EGFR, used in this study were from the ATCC and maintained as described previously (8). CDK4/hTERT immortalized human bronchial epithelial (HBEC) cells stably expressing the V5-tagged wild-type EGFR (WT), L858R, and ΔE746-E750 forms of EGFR were maintained as described in ref. 11. For induction of hypoxia, cells were cultured in poly-l-lysine (Sigma-Aldrich) or attachment factor (AF; Thermo Fisher Scientific) coated glass petri dishes to avoid the problem of oxygen (O₂) dissolved in plastic. Similarly, for imaging studies, cells were first seeded on glass coverslips, and the coverslips were submerged in glass petri dishes containing medium. Petri dishes containing cells were incubated in an open Ziploc bag at 5% CO₂, 0.1% O₂, and 37°C in a ProOx 110-controlled N₂/CO₂ gas environment chamber, i-Glove (BioSpherix). Dissolved oxygen in medium, measured using a ruthenium-based oxygen probe, reached desired levels (0.1%) within 30 to 45 minutes of incubation in hypoxia chamber. For irradiation under hypoxia, Ziploc bags were hermetically sealed while still under hypoxic conditions before being exposed to radiation in either XRAD320 (dose rate 117 cGy/minute, Precision X-ray) or Cs137 irradiator (dose rate 347 cGy/minute, J.L. Shephard and associates).

All cell lines were tested weekly for mycoplasma contamination using the MycoAlert Kit (Lonza) and authenticated twice annually using professional authentication services (Genetica-LabCorp Cell line testing).

Clonogenic survival was measured as described before (8, 11, 18). For survival response in hypoxic state, cells were exposed to 0.1% O₂ for 24 hours, irradiated as described above, allowed to recover for 8 hours under hypoxic conditions, and allowed to form colonies over 7 to 10 days in aerobic environment (21% O₂). Where inhibitors were used, cells were incubated in aerobic or hypoxic environments as above and then received vehicle, PARP1/2 inhibitor, ABT-888, (Cayman Chemical) for 2 hours, or 50 μg/mL (∼345 nmol/L) humanized mAb anti-EGFR cetuximab (Imclone/Bristol-Myers Squibb/Eli Lilly) for 6 hours prior to irradiation. Plating efficiency (PE) values were used to determine effective concentration for 50% survival (EC₅₀) for ABT-888. Surviving fraction (SF) values were normalized to PE and plotted as a function of radiation dose. The data were fit with the linear quadratic equation using SigmaPlot version 12.5 (Systat Software; www.systatsoftware.com).

HBEC cells stably expressing WT, L858R, or ΔE746-E750 forms of EGFR were left under aerobic conditions (21% O₂) or exposed for 24 hours to a hypoxic environment containing 0.1% O₂. Illumina Whole Genome HumanHT12 v4 Expression BeadChip was used in this study. Total RNA was extracted, amplified, transcribed into biotin-labeled cRNA, and hybridized with streptavidin-Cy3 (GE Healthcare) using standard Illumina protocols as described previously (26). Slides were scanned on an Illumina Beadstation. Summarized expression values for each probe sets were generated using BeadStudio 3.1 (Illumina Inc.). The data were background subtracted and quantile–quantile normalized across samples using MBCB algorithm (27). A two-sample t test was performed between WT EGFR and MT-EGFR (L858R and ΔE746-E750 combined) samples or, separately, between 21% O₂ and 0.1% O₂ samples within WT, L858R and ΔE746-E750 cohorts. Genes with P < 0.01 and fold change greater than 2 were considered as changed with statistical significance. The microarray data are deposited in NCBI's Gene Expression Omnibus (GEO) (28) and are accessible through GEO accession number GSE95564.

Approximately 10⁶ NSCLC or HBEC3KT cells were exposed to an aerobic (21% O₂) or hypoxic (0.1% O₂) environment for 24 hours. Whole-cell lysates were prepared as described previously (8, 11) Antibodies specific to RAD50 (# sc20155), RAD51 (# sc-8349), MRE11 (# sc5859), β-actin (# sc-69879), Santa Cruz Biotechnology, NBS1/Nibrin (# 05-616), RNR/RRM2 (# ABC106), Ku80 (# 05-393), EMD-Millipore (Upstate Biotechnology), and HIF1α (# GTX628480), GeneTex were used. Blots were imaged using the Chemidoc MP imaging system (Bio-Rad), and band densitometric analysis was performed using the ImageLab software version 4.1 (Bio-Rad). After background subtraction, density of a given band on a blot was normalized to density of β-actin band from the same sample on that particular blot to obtain relative density.

NSCLC cells were seeded in duplicate on glass bottom 96-well plates and then exposed for 24 hours to aerobic or hypoxic environments. For reoxygenation experiments, cells were returned to an aerobic environment and harvested at various time points for immune-detection of 53BP1 and γH2AX foci as described previously (8, 11, 18).

For cell-cycle distribution of γH2AX foci, while still under aerobic or hypoxic states, cells were pulsed for 10 minutes with 10 μmol/L of the nucleoside analogue, 5-ethynyl-2′-deoxyuridine (EdU) before being processed for immunocytochemistry using the Alexa Fluor-488 Click-iT EdU Imaging Kit (Invitrogen), mouse anti-γH2AX (pSer 139; # 05-636), rabbit anti-phospho histone H3 (pSer 10; # 04-817) from EMD-Millipore (Upstate Biotechnology), and rabbit anti-cyclin B1 antibody (# 4138), Cell Signaling Technology. Fluorescent images were acquired under the IN Cell Analyzer 2000 high content automated imaging system (GE Healthcare) using a 40× objective and analyzed using the IN Cell Investigator software (GE Healthcare). γH2AX foci in EdU-positive S-phase, Cyclin B1–positive G₂ phase, and histone H3–positive M-phase cells were enumerated. The G₁ subpopulation was estimated by subtracting the sum of S, G₂, and M phase cells from the total number of nuclei. In each phase, the fraction of nuclei harboring >5 foci was calculated.

In a separate experiment, to evaluate extent of radiation-induced DNA damage, NSCLC cells were processed for immunocytochemistry at 45 minutes following 0 or 1 Gy radiation and 15 minutes after an EdU pulse. Images were acquired in multiple focal z-planes using a Nikon A1rsi scanning confocal microscope with an oil immersion 100× (NA1.49) objective. Images within a Z-stack were collapsed to generate a single composite image and total number of nuclei, number of γH2AX foci per nucleus, and EdU⁺ S-phase nuclei were enumerated using the Cell Profiler image analysis software.

NSCLC and HBEC3KT cells were seeded in 96-well plates and pulse labeled as described above. To obtain clear RPA foci, cells were subjected to in situ fractionation as described in ref. 29, labeled with Alexa 488 Click-it reagent and immunostained with rabbit anti-γH2AX and mouse anti-RPA (# MAB286, EMD Millipore), followed by incubation with secondary antibodies and DAPI. Fluorescent images were acquired under the IN Cell Analyzer 2000 system using a 40× objective. γH2AX foci, RPA foci and EdU-positive (S-phase) nuclei were enumerated and number of S-phase nuclei with merged RPA and γH2AX foci was obtained.

Replication fork dynamics were evaluated by DNA fiber assay as described in ref. 30. Briefly, NSCLC cells were seeded in AF-coated glass 60-mm dishes and exposed for 24 hours to aerobic or hypoxic environments. Cells were pulse labeled with 25 μmol/L 5′ iodo 2′deoxyuridine (IdU) for 20 minutes followed by 10-fold excess (250 μmol/L) of 5′ chloro 2′ deoxy uridine (CldU) for a further 20 minutes. Cells were detached and diluted to a density of 7 × 10⁵ cells/mL. Nuclear suspension (2 μL) was spotted on glass slides, dried, and fixed. Samples were blocked with 5% BSA in PBS and incubated with a 1:25 dilution of IdU-specific mouse anti-BrdUrd (BD Biosciences) and 1:400 dilution of CldU-specific rat anti-BrdUrd (Abcam). Samples were incubated with a 1:500 dilution of Cy3-conjugated sheep anti-mouse and 1:400 dilution of Alexa-488–conjugated goat anti-rat secondary antibodies and mounted. Images were acquired with the Nikon A1rsi scanning confocal microscope. A minimum of 200 well-separated fibers were counted per sample. The number and juxtaposition of IdU- and CldU-stained fibers were manually measured using NIS Elements software (v 4.0). Replication structures such as stalled forks (IdU⁺/CldU⁻), active forks (IdU⁺/CldU⁺), new origins (IdU⁻/CldU⁺), elongating forks (CldU-IdU-CldU), and terminating forks (IdU-CldU-IdU) were enumerated. Length of DNA synthesized in 20-minute nucleotide pulses was measured using a DNA extension factor of 2.59 kbp/μm and relative fork velocity in aerobic and hypoxic conditions was calculated as a ratio of CldU/IdU lengths. A CldU/IdU ratio approximately 1.0 indicated unperturbed replication and CldU/IdU ratio <1 was considered slower replication.

We compared radiation response in four different NSCLC cell lines following a 24-hour exposure to either an aerobic (21% oxygen) or a hypoxic (0.1% oxygen) environment. In all three MT-EGFR NSCLC cell lines, exposure to hypoxia reduced PE across a wide range of oxygen concentrations, while WT-EGFR expressing A549 cells were relatively unaffected (Supplementary Fig. S1A–S1C). When normalized for PE, hypoxic WT-EGFR–expressing A549 cells were significantly more radioresistant relative to aerobic A549 cells (Fig. 1A). In contrast, hypoxia-associated radioresistance in MT-EGFR–expressing NSCLCs was marked reduced (Fig. 1B–D).

To quantify hypoxia-associated radiation resistance, radiation DMFs, such as OER or HRF, are frequently used. Here, we use the term hypoxia reduction factor (signaling; HRF_S), which encompasses not only physiochemical modification of DNA but also the biological or enzymatic processes that can influence survival under hypoxic conditions. We define HRF_S as the ratio of the radiation dose at a specific oxygen concentration (0.1%) to the radiation dose under fully aerobic conditions (21% oxygen) for the same reduction in clonogenic survival. Data in Fig. 1E reveal that HRF_S across a range of SFs was consistently and significantly higher in WT-EGFR–expressing A549 cells (mean HRF_S, 2.28) compared with mutant EGFR–expressing cells (mean HRF_S, 1.12–1.58). Our panel consisted of NSCLCs with low (H820), medium (H1975), and high (HCC827) expression of MT-EGFR (Fig. 1F). Despite these wide differences in expression levels, all three cell lines exhibited marked reduction in HRF_S relative to wild EGFR-expressing NSCLC, A549, indicating that expression level was not a factor in HRF_S reduction.

To rule out effects from genetic background, we stably overexpressed WT- and MT-EGFR forms in HBEC cells. Again, hypoxia alone significantly inhibited survival in HBEC cells overexpressing L858R or ΔE746-E750 EGFR but had no effect on cells overexpressing the WT-EGFR (Supplementary Fig. S1B). Cells expressing L858R or ΔE746-E750 EGFR-expressing cells (Fig. 2B and C) exhibited significantly reduced HRF_S (means, 1.66 and 1.76) compared with WT-EGFR (mean, 2.23; Fig. 2A). Figure 2E shows that WT-EGFR and MT-EGFR forms were overexpressed to similar levels in HBEC cells and were nearly three orders of magnitude higher relative to mock-transfected controls. These levels were comparable with levels of expression found in most MT-EGFR–expressing NSCLCs, such as H1975 and HCC827 (Fig. 1E). Again, overexpression alone was not a factor in HRF_S reduction because similar results were obtained when L858R or ΔE746-E750 EGFR were ectopically expressed at lower levels in A549 NSCLCs (Supplementary Fig. S5). Even though MT-EGFR expression was only 5.5- to 6.7-fold over endogenous WT EGFR (Supplementary Fig. S5A), this level of MT-EGFR expression was sufficient to exert a strong dominant negative effect in A549 cells by significantly reducing PE (Supplementary Fig. S5B, left), radioresistance (Supplementary Fig. S5B and S5C), and HRF_S (Supplementary Fig. S5D).

To examine whether mutation status of EGFR influenced hypoxia-induced changes in gene expression and whether these effects could account for reduced HRF, we compared gene expression patterns associated with WT, L858R, or ΔE746-E750 EGFR stably expressed in the isogenic background of HBEC cells. In aerobic conditions, relative to WT-EGFR–expressing cells, 226 genes were differentially expressed 2-fold or more (P < 0.01) in MT-EGFR HBECs. In contrast, in hypoxic conditions, the number was nearly 8 times larger, with 1,793 genes differentially expressed between the two cohorts. Unsupervised clustering revealed significant differences between WT- and MT-EGFR in both aerobic and hypoxic conditions (Fig. 3A) and gene expression changes in L858R and ΔE746-E750–expressing cells clustered together. In aerobic conditions (Fig. 3B), relative to WT-EGFR, MT-EGFR expression was associated with upregulation of cell-cycle activators, including CCNE1 and CCNE2, and downregulation of cell-cycle inhibitors, such as CDKN2A, CDKN2D, and CDKN1A. This is consistent with the elevated pro-proliferative signaling associated with the MT-EGFR. In hypoxic conditions, a very different pattern emerged. Although expression of NHEJ genes, PRKDC (DNA-PKcs), XRCC6 (KU70), XRCC5 (KU80), or LIG4 (ligase IV) was unchanged, the hypoxic state was associated with downregulation of many HR DNA repair components, including RAD50, RAD51, MRE11, and NBN (NBS1). Of these, mRNA levels of RAD50 and MRE11 appeared significantly lower in MT-EGFR–expressing cells relative to WT-EGFR cells. Data from Western blot and densitometric analysis (Fig. 3C and D) largely validated the microarray profile. For example, at both mRNA and protein levels, RAD51 levels reduced to the same extent in hypoxic WT- and MT-EGFR–expressing cells. Likewise, at both transcript and protein levels, RAD50 remained largely unaffected by hypoxia in WT-EGFR cells but was significantly reduced in L858R or ΔE746-E750 EGFR-expressing HBEC cells. However, MRE11 and NBS1 appeared to be differentially downregulated at the mRNA level but not at the protein level. Remarkably similar trends were observed in WT- and MT-EGFR–expressing NSCLC cells. The data in Fig. 3 not only confirm hypoxia-associated HR downregulation but also suggest that MT-EGFR expression may exert an inhibitory effect on additional HR components, such as RAD50.

We next examined the effect of hypoxia and reoxygenation on two well-established surrogates of DNA damage, 53BP1 and γH2AX, which localize to sites proximal to DSBs and form distinct foci (Fig. 4). Relative to WT-EGFR–expressing A549 cells (0.5 foci per nucleus), MT-EGFR–expressing H820, HCC827, and H1975 cells harbored dramatically elevated levels of basal 53BP1 (Fig. 4A) and γH2AX foci (2.5–7.5 foci per nucleus; Fig. 4B). A 24-hour exposure to 0.1% O₂ resulted in a modest 5-fold increase in γH2AX foci in A549 cells, which returned to near baseline levels, 2 hours following reoxygenation. Interestingly, 53BP1 foci were undetectable in hypoxic A549 cells, even after reexposure to 21% O₂. In striking contrast, all three MT-EGFR–expressing NSCLCs tested showed high basal levels of 53BP1 and γH2AX foci. Exposure to a 0.1% O₂ environment resulted in 2.5- to 7-fold decrease in 53BP1. However, reexposure to an aerobic environment not only led to a further increase in γH2AX foci but also a dramatic resurgence of 53BP1 foci to levels that were nearly 2-fold higher compared with aerobic conditions. Highly similar results were observed when WT- and MT- EGFR were ectopically expressed in HBEC cells (Fig. 4C). Relative to WT-EGFR–expressing cells, cells expressing ΔE746-E750 and L858R exhibited significantly higher levels of γH2AX foci at 21% O₂ that further increased upon exposure to a hypoxic environment.

To determine whether hypoxia-associated increase in γH2AX foci affected all cells or only cells in specific phases of cell-cycle progression, cell-cycle distribution of basal and hypoxia-induced γH2AX foci in WT- and MT-EGFR NSCLCs was assessed (Fig. 4D). In aerobic A549 cells, the fraction of foci-bearing cells (with >5 γH2AX) was extremely low (0.2%–0.8%) in all phases of the cell cycle. but increased approximately 23-fold (up to 14.5%) following a 24-hour exposure to 0.1% O₂. The increase in foci-bearing fractions was observed almost exclusively in the S-phase subpopulation, while those in G₁, G₂, and M subpopulations remained essentially unchanged. In striking contrast, foci-bearing fractions in aerobic MT-EGFR–expressing NSCLCs were significantly elevated in all phases of the cell cycle, although the most dramatic difference was in the S-phase subpopulation (24- to 31-fold higher relative to A549). At 24 hours of exposure to 0.1% O₂, foci-bearing fractions in all three MT-EGFR NSCLC cell lines were significantly higher (2- to 3.5-fold) compared with aerobic state but, again, these increases were most pronounced in the G₁ and S subpopulations. The data suggest that NSCLCs with activating ΔE746-E750 and L858R mutations differ significantly from WT-EGFR cells in basal and hypoxia-associated DDR.

We next used confocal microscopy to examine the extent and magnitude of radiation-induced DNA damage in aerobic and hypoxic WT- and MT-EGFR–expressing NSCLC cells 45 minutes after exposure to 1 Gy radiation. Representative images of confocal scans are shown in Fig. 5A. Regardless of EGFR status, in aerobic conditions, all three NSCLCs, regardless of EGFR status, registered essentially similar levels of γH2AX foci per nucleus (39.5–42.5 ± 3.5) in response to 1 Gy (Fig. 5B). Moreover, the fraction of foci-bearing nuclei (with >10 foci) in aerobic state after 1 Gy was also essentially similar; 80% to 85% (±10.11) of nuclei had >10 γH2AX foci in all three cell lines. As expected, exposure to 1 Gy in an oxygen-deficient environment reduced the number of γH2AX foci but to differing extents depending on EGFR status. A549 cells irradiated under hypoxic conditions showed an approximately 40% decrease in foci relative to aerobic irradiated cells but only 25% and 2% reductions in H1975 and HCC827, respectively (Fig. 5B). When foci-bearing EdU⁺, S-phase subpopulation was considered, a different pattern emerged (Fig. 5C). First, in all three NSCLCs, relative to aerobic cells, the mean number of radiation-induced γH2AX foci in EdU⁺ nuclei remained unchanged with hypoxia. Moreover, S-phase foci-bearing fractions (>10 foci) in all three cell lines was either the same (HCC827) or slightly higher (A549 and H1975) with hypoxia. The data in Figs. 4D and 5C suggested that hypoxia-associated increases in the S-phase γH2AX foci likely involved replication events in the S-phase.

To verify whether this is indeed the case, we enumerated fractions of S-phase nuclei harboring >20 γH2AX foci that also contained >20 foci of RPA. In NSCLCs and HBEC cells, WT-EGFR expression was associated with a 5-fold (HBEC) to 9-fold (A549) hypoxia-associated increase in the RPA⁺/H2AX⁺ S-phase fraction (Fig. 5D and E). In aerobic and hypoxic cells expressing MT-EGFRs, this fraction was significantly higher compared with WT-EGFR cells. Camptothecin, a selective inhibitor of topoisomerase I (Top1), which stabilizes Top1-linked single-stranded nicks, also caused a dramatic increase in the RPA⁺/H2AX⁺ S-phase fraction (Supplementary Fig. S3A), but this increase was evident in both WT- and MT-EGFR cells. In contrast, the effects of hydroxyurea (Supplementary Fig. S3B), which blocks initiation and elongation phases of replication, resembled those of hypoxia, with modest HU-induced increase in RPA⁺/H2AX⁺ fractions in WT-EGFR NSCLCs and HBECs but dramatically high basal and HU-induced fractions of these events in MT-EGFR cells. The data indicate that in both aerobic and hypoxic states, replicating MT-EGFR–expressing cells harbor a significantly higher burden of γH2AX foci compared with WT-EGFR cells.

To conclusively determine whether MT-EGFR expression has any effect on kinetics of replication fork progression, we examined kinetics of fork initiation, fork extension, and fork progression in NSCLCs A549, H1975, HCC827 after a 24-hour exposure to 21% or 0.1% O₂ using the DNA fiber assay (30). This assay involves sequential 20-minute pulses of two nucleotide analogues, IdU and CldU, and detection of their incorporation in replicating DNA by differently labeled fluorescent antibodies. Stalled forks (red only fibers) active forks (red–green fibers), terminating forks (red–green–red), and new origins (green only) fibers (Fig. 6A and B) were enumerated. Ratio of CldU:IdU lengths was used to determine relative rates of fork progression, with lower ratios representing replication stalling or slowing and higher ratios representing unperturbed replication (Fig. 6C). Data in Fig. 6B reveal that, in aerobic WT-EGFR–expressing A549 NSCLCs, a small fraction (7.5%) of all replicons occurred as stalled replication forks, while the majority of structures were active forks (70%) or successful terminating replicons (13.5%). Exposure to 0.1% for 24 hours resulted in a 7-fold increase in the fraction of stalled replication forks (52%), with a 50% reduction in the fraction of active forks (35%) and terminating replicons (6%). In striking contrast, even in the aerobic state, almost 20% to 26% of replicons in MT-EGFR–expressing H1975 and HCC827 NSCLCs appeared as stalled forks, an approximately 3 to 4 higher baseline relative to WT-EGFR–expressing A549 cells. In hypoxic conditions, this fraction of stalled forks formed the majority of replicons (∼63%) in HCC827 and equaled active forks (45%) in H1975. Active and terminating forks constituted a combined 83% of all replicons in aerobic A549 cells, but formed a significantly lesser proportion (65%–70%) of all replicons in H1975 and HCC827. In hypoxia, there was a further reduction in the combined fraction of active and terminating forks (31% and 45% in HCC827 and H1975, respectively). The high proportion of stalled replication forks clearly seems to have affected fork velocity. Even in aerobic conditions, relative fork velocity measured as CldU/IdU ratio was significantly reduced in H1975 (0.65 ± 0.04) and HCC827 (0.66 ± 0.03), compared with A549 cells (0.89 ± 0.04; Fig. 5C). In hypoxic conditions, whereas WT-EGFR NSCLC registered a ratio of 0.60 ± 0.035, H1975 and HCC827 exhibited ratios of 0.49 and 0.39, respectively.

Many studies have shown that, during prolonged hypoxia, levels of a key enzyme in nucleotide metabolism, RNR are drastically reduced (22) and nucleotide depletion inhibits or delays replication initiation, leading to stalled replication forks and reduced fork velocities (31, 32). We tested whether replication stress encountered in aerobic MT-EGFR NSCLCs was due to RNR deficiency. Consistent with other reports (22, 33), prolonged 24-hour exposure to 0.1% O₂ almost completely abrogated RNR levels in all cell lines, regardless of EGFR status (Fig. 6D and E). The data indicate that, despite adequate levels of RNR, replication fork progression in MT-EGFR NSCLCs is significantly compromised in aerobic conditions, and RNR-depleted hypoxic conditions may exacerbate replication stress.

Synthetic lethality occurs when simultaneous inactivation of two complementary genes or pathways causes loss of cell viability or reproductive capacity, whereas inactivation of only one gene/pathway does not. Synthetic lethality has been demonstrated by inhibition of the PARP, an abundant nuclear protein with an important function in an alternate NHEJ DSB repair pathway involving XRCC1 and ligase III (34). Multiple studies have shown that inhibition of PARP-mediated alternate NHEJ has a profound synthetic lethality effect in the context of HR deficiency in BRCA-deficient breast cancers (35) as well as HR-downregulated hypoxic tumors (36–39). As MT-EGFR expression (8, 11, 18) or EGFR blockade by the anti-EGFR mAb, cetuximab (7, 19) also compromises NHEJ, we reasoned that cetuximab might have a similar synthetic lethality effect on HR-downregulated hypoxic tumors. As a positive control for synthetic lethality, we first confirmed the effects of the PARP inhibitor, ABT-888, on PE in NSCLCs in aerobic and hypoxic conditions (Fig. 7A). Treatment with ABT-888 had no effect on survival in aerobic A549 cells but significantly reduced of A549 cells in the hypoxic state with an effective concentration (EC₅₀) of 0.5 μmol/L. Consistent with the results in Fig. 1E, hypoxia alone was sufficient to reduced PE in MT-EGFR–expressing NSCLCs. Surprisingly, ABT-888 had no effect on PE of any of the MT-EGFR NSCLCs cells even in hypoxic conditions. ABT-888 at a concentration of 1 μmol/L had no effect on radiosensitivity of aerobic A549 NSCLC cells but significantly increased A549 radiosensitivity in hypoxic conditions (Fig. 7B). D₃₇ is the radiation dose required to reduce the SF down to 37% (SF = 0.37). In aerobic conditions, the difference in D₃₇ values between untreated and ABT-888–treated A549 cells was only 0.88 Gy but it was approximately 3.5-fold higher in hypoxic conditions. EGFR blockade by cetuximab had a surprisingly similar effect (Fig. 7C). The difference in D₃₇ values between untreated and cetuximab-treated was 1.02 Gy in aerobic and 3.05 Gy in hypoxic conditions, again 3-fold higher. Both ABT-888 and cetuximab reduced HRF to similar extents (Fig. 7D). Moreover, both ABT-888 and cetuximab were significantly more effective in hypoxic conditions (Fig. 7E). The DMF is defined as the ratio of the radiation dose without agent to the radiation dose with agent to decrease SF to the same extent. In aerobic conditions, compared with ABT-888, cetuximab was slightly more effective in augmenting radiation in aerobic A549 cells (Fig. 7E). However, mean DMF for both ABT-888 and cetuximab was significantly higher in hypoxic conditions (2.0 and 1.78 Gy, respectively) compared with aerobic conditions (1.12 and 1.29 Gy, respectively). The data indicate that, like ABT-888, EGFR blockade by cetuximab was significantly more effective in radiosensitizing A549 cells in the hypoxic state compared with cells in the aerobic state.

EGFR activation and downstream signaling has well-documented roles in prosurvival mechanisms that contribute to radiation resistance not only in aerobic but also in hypoxic environments. This study proffers evidence that activating ΔE746-E750 and L858R mutations in EGFR have the opposite effect on hypoxia-associated radiation resistance and survival. Compared with WT-EGFR, MT-EGFR expression in NSCLCs or HBEC significantly compromised survival (Supplementary Fig. S1A and S1B) and diminished HRF (Figs. 1 and 2). Deficiency of the NHEJ enzyme, DNA-PKcs, similarly reduced survival in hypoxic but not aerobic conditions (Supplementary Fig. S1C). Hypoxia-induced PARP cleavage, a reliable indicator of apoptosis, was observed in all NSCLCs, regardless of EGFR status, but not in HBEC cells, (Supplementary Fig. S1D). Thus, the difference in survival between WT- and MT-EGFR–expressing cells is unlikely due to apoptosis.

We used HRF_S as a measure of hypoxia-associated radiation resistance from physicochemical modification of DNA as well as biochemical signaling. Although oxygen concentration can significantly influence HRF_S, variations in cellular reducing environment can also impact HRF_S. Reduction of the DNA radical by thiol (–SH) containing compounds restores DNA to its original (reduced) form, thereby limiting radiation-induced DNA damage. Thus, cells with high thiol content are generally less sensitive to radiation, and the effect of oxygen (or hypoxia) on HRF_S in these cases is less apparent (40). Thiol content was not a factor in the decreased HRF_S because thiol levels were similar between WT- and MT-EGFR NSCLCs and ectopic MT-EGFR expression in HBEC cells had no impact on thiol levels (Supplementary Fig. S2). Moreover, microarray analysis of WT- and MT-EGFR–expressing HBEC cells (Fig. 3, GEO accession number GSE95564) or Oncomine database analysis of a large dataset with 99 WT- and 127 MT-EGFR tumor samples from NSCLC patients (GEO accession number GSE31210) showed no significant differences in antioxidant metabolic pathway between WT- and MT-EGFR NSCLCs. It appears unlikely that the reduced HRF_S in MT-EGFR–expressing cells was due to physicochemical modulation of the DNA radical by oxygen or thiols.

A more likely mechanism could be defective DDR. We previously demonstrated that NSCLCs and HBEC expressing MT-EGFR are defective in NHEJ-mediated repair of radiation induced DNA damage. Chronically hypoxic tumors exhibit a markedly altered DDR relative to aerobic tumors. A number of studies have demonstrated that, due to a selective downregulation of HR, hypoxic cells are overreliant on NHEJ as the sole DSB repair pathway (25, 41). In agreement with these reports, our data confirm that key proteins in the HR DNA repair pathway, including RAD50 and RAD51, are downregulated during prolonged hypoxia. Interestingly, ectopic expression of ΔE746-E750 and L858R-mutant EGFR not only reduced RAD51 levels, as did WT-EGFR, but also significantly downregulated levels of RAD50, which were relatively unaffected by WT-EGFR expression. RAD51 exclusively participates in HR (42). In contrast, RAD50, a crucial component of the MRE11-RAD50-NBS1 (MRN) complex, is required for both NHEJ and HR (43). Thus, in addition to their reported defects in NHEJ (8, 10, 11, 18), MT-EGFRs may exert an additional level of NHEJ and HR suppression in hypoxic conditions.

In hypoxic state, replication stress appears to be independent of EGFR mutation status and could be, in part, due to RNR downregulation (Fig. 5D), which is known to cause nucleotide depletion and replication arrest (33). However, in aerobic conditions, where RNR levels were similar, MT-EGFR NSCLCs showed elevated replication stress compared with WT-EGFR NSCLC. One possibility is that MT-EGFRs are defective in NHEJ (8, 10, 11, 18) and an accumulation of unresolved DSBs in the S-phase may arrest or slow down fork progression (Fig. 4). An alternative mechanism may involve deregulation of the S-phase checkpoint control, resulting in fork-associated DNA damage. A recent study found that, in addition to NHEJ-mediated DSB repair, DNA-PKcs facilitates S-phase checkpoint control (44), which involves Ataxia telangiectasia–related (ATR) protein, checkpoint kinase 1, Chk1, and Cdc25a (45). DNA-PKcs is phosphorylated at T2609 by ATR in response to replication stress (46). We previously demonstrated that MT-EGFR expression in NSCLCs and HBEC cells completely abrogates DNA-PKcs T2609 phosphorylation (11). However, unlike MT-EGFR expression (Fig. 5C), DNA-PKcs deficiency had no effect on CldU/IdU ratio in aerobic conditions (44). Thus, although S-phase checkpoint impairment may elevate DSBs, the replication stress in aerobic MT-EGFR NSCLCs may involve additional factors.

Parallels can be drawn between MT-EGFR and MYC oncogene–associated replication stress. MYC influences replication at multiple levels, including direct interaction with prereplication complex proteins, regulation of replication proteins, CDC6, CDK and minichromosome maintenance (MCM) proteins, such as MCM 2-7 and 10 at origin sites (47), and upregulation of cyclin E. Microarray gene expression analysis showed no differences in MYC mRNA between WT- and MT-EGFR–expressing HBECs. However, many of MYC upregulated proteins, including Cyclin E, MCM 2-7, and 10, were significantly overexpressed in MT-EGFR–expressing cells relative to WT-EGFR cells (Supplementary Fig. S4). Like MT-EGFR–expressing cells (Fig. 4), MYC overexpression also results in fork-associated DSBs (48) and slowing of replication due to premature origin firing and increased origin density (47). It is therefore conceivable that overabundance of replication factors, together with preexisting DSBs due to NHEJ defects may, in part, be responsible for replication stress in aerobic MT-EGFR NSCLCs.

Our study finds interesting parallels between cetuximab-mediated EGFR blockade and PARP inhibition. Similar to PARP inhibition, expression of MT-EGFR exerted lethality, possibly through inhibition of NHEJ in an altered, HR-deficient DDR. First, MT-EGFR (Supplementary Fig. S1A and S1B), or DNA-PKcs deficiency (Supplementary Fig. S1C) similarly reduced survival of hypoxic cells but did not affect survival of aerobic cells. Second, MT-EGFR expression resulted in markedly reduced HRF_S in both NSCLCs and HBEC (Figs. 1 and 2) to the same extent as PARP inhibition (Fig. 7). Third, like PARP inhibition (Fig. 7) and (39) EGFR blockade by cetuximab was significantly more effective in increasing radiosensitivity of A549 cells in the hypoxic state (DMF: 1.78) compared with aerobic state (DMF: 1.29).

Interestingly, PARP inhibition did not affect survival of NHEJ-compromised MT-EGFR NSCLCs in either aerobic or hypoxic conditions (Fig. 7A), indicating that EGFR and PARP may be epistatic, with interdependent activities and mutually supportive roles in NHEJ-mediated DSB repair (49). Compared with the transient 2-hour ABT-888 treatment in our study, a 72-hour PARP inhibition by olaparib did show MT-EGFR NSCLCs to be slightly more sensitive than WT-EGFR NSCLCs (14). Alternatively, EGFR and PARP may share a synthetic lethal interaction. EGFR blockade by lapatinib in triple-negative breast cancers (50) and cetuximab in head and neck cancers (13) augmented ABT-888 cytotoxicity through possible suppression of NHEJ, HR, or both (12).

Our data support a model to explain MT-EGFR–associated radiosensitivity in aerobic and hypoxic conditions. In aerobic conditions, expression of NHEJ-defective MT-EGFR compromises repair of radiation-induced as well as fork-associated DSBs, resulting in increased radiosensitivity. In the hypoxic state, in the context of an altered HR downregulated DDR, expression of NHEJ-defective MT-EGFR or blockade of EGFR-mediated NHEJ has a catastrophic effect on DSB repair and causes synthetic lethality.

Understanding how EGFR mediates repair of fork-associated DSBs could elucidate novel therapeutic targets. Our finding that MT-EGFR expression or EGFR blockade has a synthetic lethal effect in hypoxic conditions suggests that anti-EGFR therapy in combination with radiotherapy could potentially be effective in treating not only hypoxic NSCLCs but also NSCLCs with mutations in HR DSB repair genes.

No potential conflicts of interest were disclosed.

Conception and design: M. Saki, H. Makino, J. Andrews, D. Saha, S. Burma, C.S. Nirodi

Development of methodology: M. Saki, H. Makino, L.-H. Ding, J. Andrews, S. Burma, C.S. Nirodi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Makino, N. Tomimatsu, L.-H. Ding, J.E. Clark, E. Gavin, K. Takeda, J. Andrews, M.D. Story, C.S. Nirodi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Saki, H. Makino, L.-H. Ding, J. Andrews, D. Saha, M.D. Story, S. Burma, C.S. Nirodi

Writing, review, and/or revision of the manuscript: M. Saki, L.-H. Ding, J.E. Clark, J. Andrews, D. Saha, M.D. Story, C.S. Nirodi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.S. Nirodi

Study supervision: C.S. Nirodi

Other (initial experiments while in the laboratory): P. Javvadi

The authors gratefully acknowledge Dr. David Boothman (University of Texas Southwestern Medical Center), Dr. Robert W. Sobol (University of South Alabama, Mitchell Cancer Institute), and Dr. Conchita Vens (Netherlands Cancer Institute) for their critical inputs during the course of this study.

This work was supported by the NIH (R01CA129364 to C.S. Nirodi, RO1CA149461, RO1CA197796, R21CA202403 to S. Burma, R21CA175879 to D. Saha), the National Aeronautics and Space Administration (NNX16AD78G to S. Burma, NNJ05HD36G to M.D. Story), and startup funds from the University of South Alabama Mitchell Cancer Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.