Abstract

   Checkpoint kinase 1 (CHK1) is critical for cell survival under
   replication stress (RS). CHK1 inhibitors (CHK1i’s) in combination with
   chemotherapy have shown promising results in preclinical studies but
   have displayed minimal efficacy with substantial toxicity in clinical
   trials. To explore combinatorial strategies that can overcome these
   limitations, we perform an unbiased high-throughput screen in a
   non-small cell lung cancer (NSCLC) cell line and identify thioredoxin1
   (Trx1), a major component of the mammalian antioxidant-system, as a
   determinant of CHK1i sensitivity. We establish a role for redox
   recycling of RRM1, the larger subunit of ribonucleotide reductase
   (RNR), and a depletion of the deoxynucleotide pool in this
   Trx1-mediated CHK1i sensitivity. Further, the TrxR inhibitor auranofin,
   an approved anti-rheumatoid arthritis drug, shows a synergistic
   interaction with CHK1i via interruption of the deoxynucleotide pool.
   Together, we show a pharmacological combination to treat NSCLC that
   relies on a redox regulatory link between the Trx system and mammalian
   RNR activity.

   Subject terms: Lung cancer, Cancer therapeutic resistance,
   Non-small-cell lung cancer
     __________________________________________________________________

   The clinical application of inhibitors targeting checkpoint kinase 1
   (CHK1) is challenged by limited efficacy. Here, the authors identify
   that thioredoxin (Trx) system inhibition mediates sensitivity to CHK1
   inhibitor via regulating the activity of ribonucleotide reductase,
   demonstrating the synergistic effect of CHK1 inhibitor and inhibitors
   targeting Trx system in lung cancer models.

Introduction

   Lung cancer is one of the most frequently diagnosed cancers. Despite
   advances in targeted therapy and immunotherapy, lung cancer is still
   the leading cause of cancer-related deaths worldwide. Thus, the
   improved treatment approaches are urgently needed. The replication
   stress (RS) response protein ataxia telangiectasia and Rad3-related
   protein (ATR) and its main downstream factor, checkpoint kinase 1
   (CHK1), play important roles in cell survival under RS^[57]1. The
   inhibitors targeting CHK1 (CHK1i’s) have been shown to be a powerful
   strategy for treating solid tumors and hematological malignancies in
   preclinical studies^[58]2,[59]3. However, in most clinical trials,
   including those for treating non-small cell lung cancer (NSCLC), which
   accounts for 85% of all lung cancer cases, CHKi’s have failed to
   achieve their primary endpoints and have shown cumulative tissue
   toxicities in normal tissues^[60]4–[61]8. These findings significantly
   limit the clinical benefit of these agents. Thus, identifying
   combinatorial strategies that can enhance the sensitivity of tumor
   cells to CHKi’s, while limiting their toxicities, might be the key to
   improve the safety and efficacy of these compounds^[62]3.

   The side chain of a proteinogenic cysteine residue contains a terminal
   thiol (−SH) functional group. The sulfur atom at the core of the thiol
   is electron rich and its d-orbitals allow for multiple oxidation
   states^[63]9–[64]12. Reversible oxidation of thiols by forming the
   disulfide bond controls protein structure and biological functions.
   Thiol-based mammalian thioredoxin (Trx) and Grx-GSH systems are
   important antioxidant systems and maintain a proper balance of protein
   dithiol (reduced)–disulfide (oxidized) modifications through its
   disulfide reductase activity^[65]12,[66]13. The Trx system is comprised
   of reduced nicotinamide adenine dinucleotide phosphate (NADPH), Trx and
   thioredoxin reductase (TrxR). Human cells have two members of the Trx
   family genes; namely, cytosolic Trx1 (encoded by the TXN gene) and
   mitochondrial Trx2 (encoded by the TXN2 gene)^[67]14. The Trx system
   actively removes reactive oxygen species (ROS) and nitrogen species via
   redox regulation of peroxiredoxins. In addition, the Trx system is also
   critical for regulating transcription, apoptosis, cell growth and DNA
   synthesis via redox regulation of ribonucleotide reductase (RNR).
   However, the regulation of RNR by the Trx system is based on results
   from E. coli and/or yeast. The studies with mammalian cells are limited
   to the biochemical assays determining RNR activity regeneration by the
   Trx system, while RNR redox status in vivo has not been
   investigated^[68]12,[69]15.

   RNR promotes the reduction of ribonucleotides to their corresponding
   deoxyribonucleotides, providing a balanced supply of precursors for DNA
   synthesis and repair^[70]16. There are three known human RNR subunits,
   which are RRM1, RRM2 and p53R2. Based on the pathways of radical
   initiation and requirements of metal cofactors, the RNRs have been
   divided into three classes^[71]17, one of which is Class Ia RNR that is
   found in all types of eukaryotes (including human, mouse and yeast),
   several viruses, a few prokaryotes (including Escherichia coli) and
   some bacteriophages. Using E. coli Class Ia RNR as a prototype model,
   it has been proposed that the small subunit R2 (equivalent of human
   RRM2) contains a diiron-tyrosyl radical cofactor and operates as a
   radical chain initiator by generating a thiyl radical on cysteine
   Cys^439 in the active site of R1 (equivalent of human RRM1) via a
   long-range radical transfer pathway^[72]18–[73]20. Then, the thiyl
   radical of Cys^439 has been proposed to abstract the 3′ hydrogen atom
   from nucleoside diphosphate (NDP) and initiate its reduction^[74]21. As
   the active site cleft of the R1 subunit is not wide enough to permit
   the direct reduction by the external redox systems, the reduction of
   active site disulfide (Cys225 and Cys462) is carried out by a pair of
   shuttle cysteine residues in the C-terminal mobile tail of R1 subunit
   (Cys754 and Cys759) and then the resulting disulfide bond in the
   C-terminal tail of R1 is then reduced by external redox such as Trx or
   GSH system to continue the next catalytic cycle^[75]17,[76]22. A
   similar mechanism occurs in yeast but the Trx system is more important
   for Grx^[77]23. In vitro mutagenesis and kinetic studies support a
   critical role for the C-terminal cysteine pair of R1 in regeneration of
   the active site^[78]12,[79]17. Despite the extensive studies in
   bacteria and yeast, RRM1 redox regulation by the Trx system at the
   cellular level in human cells has not been reported. Although the role
   of the Trx system in fueling DNA synthesis during T-cell metabolic
   reprogramming was suggested via untargeted metabolomics analysis and
   pathway enrichment analysis under stimulated conditions, this study did
   not explore the regulatory role of the Trx system in RNR activity,
   especially with regards to RNR (RRM1)^[80]24.

   Active mammalian RNR consist of two homodimers of RRM1 and RRM2 each.
   In undisturbed cells, the RNR enzyme concentration and composition
   differ during the cell cycle. The level of RRM1 remains largely
   unchanged; however, RRM2 increases during the S phase and gets degraded
   while entering mitosis^[81]25–[82]27. A recent study identified
   acetylation of RRM2 as a regulatory mechanism governing RNR activity to
   promote cell growth^[83]28. In response to RS, RRM2 is
   transcriptionally increased by E2F1 via the ATR-CHK1 pathway^[84]28.
   Additionally, RRM1 and RRM2 relocate from the cytoplasm to the nucleus
   after genotoxic stress induced by UV irradiation^[85]26. Thus, RNR
   activity is a highly regulated process in the presence and absence of
   exogenous DNA damage.

   Here, we perform an unbiased high-throughput Decode Pooled shRNA
   library screen of the NSCLC cell line H1299 to identify the
   determinants of CHK1i sensitivity. Our screening identifies the TXN
   gene as one of the top hits. Further validation of the screening data
   show that the depletion of Trx1 or TrxR1 is synthetically lethal with
   CHK1i treatment due to the interruption of RNR activity in which both
   RRM1 and RRM2 are involved. Trx1 or TrxR1 depletion leads to an
   increase in RS because of an accumulation of oxidized RRM1 that results
   in defective dNTP biosynthesis. In parallel, activation of the
   E2F1-RRM2 pathway under states of increased RS is significantly
   restricted by CHK inhibition. Thus, limiting tyrosyl radical transfer
   activity of RRM2 interrupts long-range radical transfer in RRM1, which
   further halts dNTP production. Interestingly, we do not observe
   association of Trx system impairment-induced ROS with the RRM1 redox
   cycling or its effect on RS nor with the synthetic lethality between
   Trx system interruption and CHK1i. Auranofin (AUR), a small molecule
   inhibitor targeting TrxR that has been used clinically to treat
   rheumatoid arthritis upregulates RRM1 oxidation and shows synthetic
   lethality with CHK1i due to a depleted dNTP pool. Together, our results
   reveal that the Trx system serves as a regulator of RRM1 redox
   recycling in human cells and its inhibition increases RS due to a
   depleted dNTP pool that renders the cells sensitive to CHK1 inhibition.

Results

Trx1 is a determining factor of CHK1i sensitivity

   To identify factors that contribute to CHK1i sensitivity, we conducted
   a high-throughput screen using Decode Pooled Lentiviral shRNA library
   in the NSCLC cell line H1299 using LY2603618, a first generation
   specific CHK1i (Fig. [86]1a). Briefly, cells were transduced with an
   shRNA library consisting of 95,700 lentiviral shRNAs that targets
   18,205 unique human protein-coding genes. shRNA abundance was
   determined by next-generation sequencing and was compared with
   untreated cells using model-based analysis of genome-wide shRNA gene
   knockdown (KD). The effects of the shRNAs were quantified as
   standardized Z scores (Fig. [87]1b). Notably, the TXN gene, which
   encodes for Trx1, was identified as one of the top hits as five out of
   five shRNAs show sensitivity to CHK1i (Fig. [88]1b). We then focused on
   both Trx1 and its reductant TrxR1 because of its critical role in Trx1
   regulation. Scrutiny of mRNA expression from The Cancer Atlas Genome
   (TCGA) and ONCOMINE datasets confirmed significant elevation of Trx1
   and TrxR1 in NSCLC subsets lung adenocarcinoma (LUAD) and lung squamous
   cell carcinoma (LUSC) (Fig. [89]1c and Supplementary Fig. [90]S1a, b).
   Further data analysis revealed widespread overexpression of Trx1 and
   TrxR1 in LUAD (n = 560; Trx1 P < 0.00001; TrxR1 P < 0.00001) and LUSC
   (n = 546; Trx1 P < 0.00001; TrxR1 P < 0.00001). The expression of both
   Trx1 and TrxR1 in more than 50% of the patients with NSCLC are higher
   than the 75% quantile expression level in the normal samples
   (Supplementary Fig. [91]S1c). TCGA data sets also revealed that
   expression of Trx1 or TrxR1 significantly increases with advancement of
   the disease (Supplementary Fig. [92]S1d). Trx1 or TrxR1 protein
   expression was also examined by immunohistochemistry (IHC) in a cohort
   of tissue microarray (TMA) with their matched adjacent normal tissue
   (ANT). The median immunoreactive score (IRS) was significantly higher
   for the NSCLC than ANTs (Fig. [93]1d). It was found that 88% of the
   tumor samples were Trx1-positive compared with only 8% of the ANTs and
   80% of the sample were positive for TrxR1 positive, compared with only
   24% of the ANTs (Fig. [94]1e). Representative images of NSCLC and
   matched ANT are shown (Fig. [95]1f). Additionally, higher expression of
   Trx1 and TrxR1 in NSCLC subsets were positively associated with disease
   advancement, leading to poor overall survival of patients with LUAD
   (Supplementary Fig. [96]S1e). Collectively, our results and the data
   set analysis corroborate previous reports that annotated the
   association of elevated Trx1 and TrxR1 expression with poor prognosis
   and survival in different cancers^[97]29–[98]33.

Fig. 1. Genome-wide Decode Pooled shRNA library screening identifies Trx1 as
a determining factor of CHK1 inhibitor sensitivity.

   [99]Fig. 1
   [100]Open in a new tab

   a Schematic diagram (Created with BioRender.com released under a
   Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International
   license) illustrating the screening workflow for genome-wide
   loss-of-function screening. gDNA was isolated from the reference and
   experimental populations of transduced H1299 cells. CHK1i, CHK1
   inhibitor (LY2603618). b A volcano plot showing the identified target
   genes from the screen in a. The colored dots are significant genes that
   have a fold change >1.5 and any point not gray is significant P < 0.05.
   TRX1 is indicated as one of the top candidates from the screen. c The
   analysis of TCGA data sets showing the transcript expression of Trx1
   and TrxR1 NSCLC subsets (LUSC & LUAD) compared to normal tissue. Top
   and bottom of the box indicates the 75th and 25th percentile,
   respectively. The whisker represents 1.5 times the interquartile range
   from the box; p-values determined by a two-tailed Mann–Whitney test. d
   The median Trx1 and TrxR1 IHC score (IRS) of the NSCLC tumors (n = 50)
   was significantly higher than that of matched adjacent normal tissue
   (ANT) (n = 50) Top and bottom of the box indicates the 75th and 25th
   percentile, respectively. The whisker represents 1.5 times the
   interquartile range from the box; p-values determined by a two-tailed
   Mann–Whitney test. IRS score was determined from the staining intensity
   (SI) and percentage of positive cells. e Bar graph of samples showing
   positive and negative staining for Trx1 and TrxR1 in tumors and ANT. f
   Representative IHC images of NSCLC tumors (n = 50) with matched ANT
   (n = 50); Scale bar, 100 μm. Tissue staining for Trx1 and TrxR1 was
   performed once on tissue micro array (TMA) containing tumors (n = 50
   individual tissue) with matched ANT (n = 50 individual tissue).
   [Statistical information: Box plots (c and d) are represented as median
   value ± SD. P value was determined by a two-tailed Mann–Whitney test.;
   ***p ≤ 0.0001; ****p < 0.0001].

   Interestingly, in contrast to a frequent upregulation of the Trx
   system, defects in Grx-GSH system components compromises GSH
   homeostasis in NSCLC^[101]34. Several findings indicate the redundancy
   between cytosolic Trx1 and the Grx-GSH system, which act as backup
   systems for each other^[102]34. The Grx-GSH system and its regulators
   comprises many components, and a defect in any of them could result in
   impaired system function. Typically, the sensitivity to Trx1/TrxR1
   inhibition is employed as an indicator of Grx-GSH deficiency^[103]34.
   Given that NSCLC H1299 cells carry a defective Grx-GSH system^[104]34,
   our study focuses on the impact of Trx1 and TrxR1 inhibition in NSCLC
   cells with a defective Grx-GSH system.

A synergistic interaction occurs between Trx1/TrxR1 KD and CHK1 inhibition

   To validate our screening results, we first depleted Trx1 or TrxR1 in
   two NSCLC cell lines (H1299 and Calu-6) (Fig. [105]2a), both of which
   are reported to have a deficient Grx-GSH system, to determine the
   effect of CHK1i treatment. By cellular toxicity results we found that
   the Trx1 or TrxR1 depletion increases CHK1i sensitivity, leading to a
   marked increase in toxicity and cell death in a dose-dependent manner
   (Fig. [106]2b). Similarly, treatment with CHK1i for 24 h significantly
   attenuated the proliferation of Trx1- or TrxR1-depleted cells, compared
   to scrambled control (shCON) (Fig. [107]2c, d). Of note, depletion of
   Trx1 or TrxR1 also magnified ATR inhibitor (ATRi) sensitivity,
   resulting in cessation of cell proliferation (Supplementary
   Fig. [108]S2a, b). Additionally, we performed the apoptosis assay in
   vitro using Annexin V and propidium iodide (PI) staining of NSCLC cells
   treated with a CHK1i followed by Trx1 or TrxR1KD and found a higher
   occurrence of apoptosis in the Trx1 or TrxR1 KD cells treated with the
   CHK1i compared to either condition alone (Supplementary Fig. [109]S3a,
   b, FACS gating strategies are presented in Supplementary
   Fig. [110]S17). Consistent with the increased apoptosis, a higher
   cleaved PARP1 in H1299 cells treated with the CHK1i followed by Trx1 or
   TrxR1 KD was also observed (Supplementary Fig. [111]S3c).

Fig. 2. Trx1 or TrxR1 depletion and CHK1i show a synergistic interaction in
vitro and in vivo.

   [112]Fig. 2
   [113]Open in a new tab

   a Western blots of Trx1 and TrxR1 in H1299 and Calu-6 NSCLC cells after
   shRNA-mediated depletion. b Cellular toxicity assays of the indicated
   cell lines with the indicated knockdowns to measure cell survival after
   treatment with CHK1i for 72 h (n = 4; biological repeats), error bars
   represent ±SD. c, d Proliferation assays of the indicated cells lines
   with the indicated knockdowns and after treatment with CHK1i for 24 h
   (n = 3 biological repeats), error bars represent ±SD. e Representative
   photographs of the excised tumors from the indicated groups. f The
   tumor weight from each indicated group at the endpoint of the
   experiment (n = 8; individual tumors in each group; error bars
   represent ±SD). g Tumor growth curves in the indicated groups (n = 10
   animals for each group), error bars represent ±SD. h Survival of the
   indicated groups as analyzed by Kaplan–Meier analysis. (number of
   animals in each group is indicated in figure in parentheses). i
   Representative TUNEL images of the tissue excised from the indicated
   groups treated with or without CHK1i. T1-T3 are the number of tumor
   tissue. Scale bar, 100 μm. j Bar graph showing relative TUNEL intensity
   in the indicated groups (n = 3 individual tumors from each group; error
   bars represent ±SD). [Statistical information: Data are represented as
   mean value ± SD. The p- values were calculated using one way ANOVA for
   multiple comparison; **p ≤ 0.005; ***p ≤ 0.0001; ****p < 0.0001].

   To verify our in vitro results, we conducted an in vivo assay using
   a xenograft model. We started CHK1i treatment after the tumor reached
   at least a size of 100 mm^3 (Supplementary Fig. [114]S4a). Animals
   bearing xenograft tumors derived from Trx1- or TrxR1-depleted cells
   showed a significant decrease in tumor size and tumor weight following
   CHK1i treatment (Fig. [115]2e, f), leading to significant inhibition of
   tumor growth (Fig. [116]2g). The significant decrease in tumor growth
   by CHK1i treatment allowed for increased overall survival, especially
   in animals bearing tumors form Trx1- or TrxR1-depleted cells
   (Fig. [117]2h). Of note, a similar body weight among the various groups
   indicated no apparent systematic toxicity following CHK1i
   administration during treatment (Supplementary Fig. [118]S4b). We then
   determined the effect of the different treatments on cancer cell death
   by conducting H&E staining (Supplementary Fig. [119]S4c) and the TUNEL
   assay in xenograft tumor tissue. A significantly higher degree of
   apoptosis as measured by TUNEL assays (Fig. [120]2I, j) in the group
   treated with both CHK1i and Trx1 KD or TrxR1 KD compared to the
   monotherapies. In summary, both in vitro and in vivo findings suggest a
   potent and significant synergistic interaction between Trx1 or TrxR1
   depletion and CHK1 inhibition.

Trx1/TrxR1 KD in combination with CHK1i leads to profound RS

   Cells with high RS depend on ATR/CHK1 for survival, which renders cells
   sensitive to ATR/CHK1 inhibitors^[121]35–[122]37. To test the
   hypothesis that Trx1 or TrxR1 depletion increases RS and activates
   ATR/CHK1, thereby leading to the increased sensitivity of the CHK1i, we
   first assessed by Western blot the phosphorylation of ATR/CHK1, RPA32
   (p-RPA32) and histone protein H2AX (γH2AX), markers of RS and/or DNA
   double-strand breaks (DSBs). We found that Trx1 or TrxR1 depletion
   distinctly increased p-CHK1(S345) and p-ATR (1989) and p-RPA32 and
   γH2AX compared with shCON, an indication of increased spontaneous RS
   (Supplementary Fig. [123]S5a). In addition, Trx1 or TrxR1 depletion
   also led to an increase in chromatin-associated RPA2, p-RPA32 and γH2AX
   (Supplementary Fig. [124]S5b). Furthermore, nuclear intensity
   quantification of RPA32, pRPA32 and γ-H2AX together with foci formation
   revealed significantly increased nuclear intensities and nuclear foci
   in Trx1- or TrxR1-depleted cells (Supplementary Fig. [125]S5c–h).
   Supporting these findings, Trx1 or TrxR1 depletion-induced RS resulted
   in the accumulation of cells in the S phase as detected by flow
   cytometry (Supplementary Fig. [126]S5i, j, FACS gating strategies are
   presented in supplementary file, Supplementary Fig. [127]S17). In
   contrast, Trx1 or TrxR1 inhibition has much less impact in RS
   untransformed lung cells (Supplementary Fig. [128]S5k). Thus, Trx1 or
   TrxR1 depletion results in spontaneous RS and DSBs. Of note, Trx1 or
   TrxR1 KD had no obvious impact on RS (Supplementary Fig. [129]S6a–d)
   and cytotoxicity (Supplementary Fig. [130]S6e) in the A549 cells that
   have an intact GSH system (Supplementary Fig. [131]S6).

   Contingent with the synergistic inhibition findings (Fig. [132]2), we
   sought to determine the effect of the combinatorial inhibition of the
   Trx system and CHK1 on RS. CHK1i exposure in Trx1 or TrxR1-depleted
   cells further intensified RS compared to shCON cells (Fig. [133]3a).
   Moreover, immunostaining showed further increases in nuclear
   intensities of p-RPA32 and γH2AX (Fig. [134]3b, c) and nuclear foci
   (Fig. [135]3d–f), following CHK1i treatment, particularly in Trx1 or
   TrxR1-depleted cells. Neutral comet assays also suggested a greater
   increase in DSBs in Trx1 or TrxR1-depleted cells following CHK1i
   treatment compared to shCON cells (Fig. [136]3g, h). Of note, Trx1 or
   TrxR1-depleted cells treated with the CHK1i exhibited a significantly
   higher proportion of cells with pan γH2AX staining, compared to CHK1
   inhibition and/or Trx1 or TrxR1 depletion alone (Fig. [137]3i). In
   support of the hypothesis that increased RS and DSBs lead to the cell
   death, we noticed a significant increases in higher nuclear
   fragmentation following CHK1i treatment, a typical cell death caused by
   mitotic catastrophe particularly in Trx1 or TrxR1-depleted cells
   (Fig. [138]3j, k). Lastly, cells with depleted Trx1 or TrxR1 showed
   significantly higher TUNEL staining compared to depletion or CHK1
   inhibition alone (Fig. [139]3l, m). Of note, the extent of RS and DNA
   damage in Trx1 or TrxR1 KD tumors treated with CHK1i that was described
   in Fig. [140]2 were also profound (Supplementary Fig. [141]S7).
   Collectively, inhibition of CHK1 coupled with Trx1 or TrxR1 depletion
   induces a greater increase in the level of RS and unrepaired DSBs
   compared to either inhibition alone, leading to cell death.

Fig. 3. Trx1 or TrxR1 depletion in combination with CHK1i leads to profound
RS, replication fork collapse and cell death.

   [142]Fig. 3
   [143]Open in a new tab

   a Western blot analysis of markers of RS and DNA damage in H1299 cells
   treated with 1 µM CHK1i for 1 h. b, c The nuclear intensities of
   individual cells of γH2AX and p-RPA32(S33) following CHK1i treatment
   (n ≥ 1500 cells from 3 biological repeats), error bars represent ±SD. d
   Representative immunofluorescence images of γH2AX- and
   p-RPA32(S33)-positive foci in H1299 cells. Scale bar, 50 μm. The
   percentage of H1299 cells with ≥5 foci of γH2AX (e) and p-RPA32(S33)
   (f) following CHK1i treatment for 1 h. (n ≥ 100 cells/repeat from 3
   biological repeats; error bars represent ±SD). Representative images
   from neutral comet assays to measure the degree of DNA double strand
   breaks in H1299 cells treated with 1 µM CHK1i for 2 h (g) and the
   quantitation of their tail lengths (h) (n ≥ 50 cells/repeat from 3
   biological repeats), error bars represent ±SD. i Percentage of cells
   with Pan-γH2AX staining after CHK1i treatment (n ≥ 100 cells/repeat
   from 3 biological repeats; error bars represent ±SD). Representative
   immunofluorescent images of nuclear fragmentation (DAPI staining) and
   F-actin staining (Phalloidin) (j) after CHK1i treatment for 24 h and
   the quantitation of the percentage of cells with nuclear fragmentation
   (k) (n = ≥100 cells/repeat from 3 biological repeats; error bars
   represent ±SD) Scale bar, 50 μm. TUNEL staining (l) and its
   quantitation (m) of H1299 cells treated with CHK1i (1 µM) for 6 h.
   (n ≥ 100 cells/repeat from 3 biological repeats; error bars represent
   ±SD) Scale bar, 50 μm. [Statistical information: Data are represented
   as mean value ± SD. The p-values were calculated using one way ANOVA
   for multiple comparison. Red line in dot plot (b, c and h) indicates
   mean; **p ≤ 0.005; ***p ≤ 0.0001; ****p < 0.0001; ns non-significant].

ROS does not play a critical role in Trx1 KD-induced RS or the synergy
between Trx1 KD and CHK1 inhibition

   The Trx system protects cells by its antioxidant activity via removal
   of peroxides and ROS through electron donation to downstream
   thioredoxin peroxidases, such as peroxiredoxins (Prxs). Given that ROS
   can induce DNA damage, ROS might be responsible for the increased RS in
   Trx1 or TrxR1-depleted cells. Thus, we measured the levels of
   intracellular ROS using the fluorescent probe CM-H^2DCFDA that is
   commonly employed and react with several ROS including hydrogen
   peroxide, hydroxyl radicals and peroxinitrite^[144]38 and found that
   Trx1 or TrxR1 depletion increased the accumulation of intracellular ROS
   as indicated by a higher DCFDA signal (Fig. [145]4a). Simultaneously,
   we used two different ROS scavenger N-acetyl cysteine (NAC) and
   α-tocopherol (vitamin E) to scavenge intracellular ROS induced by Trx1
   or TrxR1 depletion to the basal level ROS as compared to shCON. NAC is
   frequently employed as a source of sulfhydryl groups to cells as an
   acetylated precursor of reduced GSH. NAC can also interact directly
   with ROS and nitrogen species because it is a scavenger of oxygen free
   radicals^[146]39. α-tocopherol can significantly scavenge both peroxyl
   and singlet oxygen species^[147]40,[148]41, both of which can cause DNA
   damage^[149]42,[150]43. Therefore, we used NAC and α-tocopherol as
   scavengers to remove the DNA damage-inducing ROS that are produced in
   cells upon inhibition of the Trx system. By titration, we found that
   ~1 mM NAC and ~10 μM of α-tocopherol was sufficient to bring the Trx1
   or TrxR1 depletion induced-ROS level to almost basal levels
   (Fig. [151]4a and Supplementary Fig [152]S8a). We, then measured the
   changes in lipid peroxidation as secondary verification for ROS
   scavenging in cells treated with NAC or α-tocopherol following Trx1 or
   TrxR1 KD. Knockdown-induced lipid peroxidation was brought down
   significantly close to shCON groups (Fig. [153]4b and Supplementary
   Fig. [154]S8b). Surprisingly, treatment with NAC or α-tocopherol failed
   to mitigate Trx1 or TrxR1 depletion-induced RS, as reflected by no
   profound changes in the RS markers (Fig. [155]4c, d). Additionally, no
   significant change was found in Trx1 or TrxR1 depletion-induced p-RPA32
   foci in the cells with or without NAC (Fig. [156]4e). Interestingly, we
   observed a decrease in γH2AX foci induced by Trx1 or TrxR1 depletion
   following NAC treatment, relative to untreated cells (Fig. [157]4f).
   This might be caused by the fact that apoptosis can also lead to
   increased γH2AX, and ROS scavenger treatment might have an impact on
   apoptosis. Consistent with a lack of impact on Trx1 or TrxR1
   depletion-induced RS, NAC intervention had no obvious impact on Trx1 or
   TrxR1 depletion-induced cell growth repression (Fig. [158]4g, h) and
   failed to decrease CHK1i sensitivity following Trx1 or TrxR1 depletion
   (Fig. [159]4i, j). Collectively, our results suggest that Trx1 or TrxR1
   depletion-induced ROS is not responsible for RS and ROS scavenging does
   not mitigate the enhanced sensitivity to CHK1i following Trx1 or TrxR1
   KD.

Fig. 4. ROS is not responsible for Trx1 or TrxR1 depletion-induced RS.

   [160]Fig. 4
   [161]Open in a new tab

   a Intracellular ROS levels in H1299 cells treated with 1 mM NAC and
   α-tocopherol (10 μM) as measured by DCFDA staining (n = 3 biological
   repeats; error bars represent ±SD). b Bar graph showing changes in
   lipid peroxidation in indicated groups treated with or without NAC and
   α-tocopherol (n ≥ 1000 cells/repeat from 2 biological repeats; error
   bars represent ±SD). c, d Representative western blots of the indicated
   RS markers in the experimental condition after NAC and α-tocopherol
   treatment of H1299 cells. Percent of H1299 cells with ≥5 foci of
   p-RAP32(S33) (e) and γH2AX (f) after NAC treatment (1 mM). (n ≥ 100
   cells/repeat from 3 biological repeats; error bars represent ±SD).
   Relative cell proliferation of H1299 cells depleted for Trx1 (g) or
   TrxR1 (h) after NAC treatment (1 mM) (n = 2 biological repeats; error
   bars represent ±SD). Relative cell proliferation of H1299 cells after
   Trx1 depletion (i) or TrxR1 depletion (j) and CHK1i treatment and NAC
   treatment (1 mM) (n = 2 biological repeats; error bars represent ±SD).
   [Statistical information: Data are represented as mean value ± SD. The
   p-values were calculated using one way ANOVA for multiple comparison;
   *p ≤ 0.05; **p ≤ 0.001; ns non-significant].

Trx1/TrxR1 KD impairs replication fork elongation due to a depleted dNTP pool

   To determine the mechanisms by which Trx1 or TrxR1 depletion increases
   RS, we first determined the rate of DNA synthesis during S phase
   progression via a BrdU incorporation assay. By doing so, we found an
   increased accumulation of cells in early S phase, which typically
   occurs when cells have a deficient dNTP pool (Supplementary
   Fig. [162]S9a, b, FACS gating strategies are presented in Supplementary
   Fig. [163]S17). Indeed, we observed a significant decrease in the dNTP
   pool following Trx1 or TrxR1 depletion (Fig. [164]5a–d). To
   characterize the fate of replication forks caused by a disbalanced dNTP
   pool, we analyzed the replication elongation, initiation and
   termination that are frequently affected during RS using the DNA fiber
   technique^[165]44. We found that Trx1 or TrxR1 depletion significantly
   impeded replication fork progression (Fig. [166]5e, f), which resulted
   in a higher rate of fork termination and simultaneous initiation of new
   origin firings, especially in Trx1 or TrxR1-depleted conditions
   (Fig. [167]5g). We validated the contribution of a reduced dNTP pool to
   replication fork stalling following Trx1 or TrxR1 depletion via
   reconstitution with extracellular dNTPs. This repletion significantly
   restored replication fork elongation in Trx1 or TrxR1- depleted cells
   (Fig. [168]5h). Similarly, exogenous addition of dNTPs remarkably
   mitigated Trx1 or TrxR1 depletion-induced RS, as indicated by decreased
   p-CHK1, p-RPA2 and γ-H2AX levels (Fig. [169]5i). Of note, considering
   the ROS-independent RS induction in Trx1 or TrxR1-depleted cells, the
   results of DNA fiber assays also suggested no mitigation of replication
   elongation or decreases in new origin firing upon NAC treatment in Trx1
   or TrxR1-depleted cells. (Fig. [170]5–l). Taken together, we conclude
   that Trx1 and TrxR1 depletion-induced RS is tightly linked with an
   imbalanced dNTP pool.

Fig. 5. Trx1 or TrxR1 depletion abrogates dNTP production and interrupts
replication fork elongation.

   [171]Fig. 5
   [172]Open in a new tab

   Levels of dATP (a), dGTP (b), dCTP (c) and dTTP (d) after Trx1 or TrxR1
   depletion in H1299 cells (n = 3 biological repeats; error bars
   represent ±SD). e Representative image of DNA fiber tracks in the
   indicated groups after 40 min pulses with IdU and CldU. f Replication
   fork progression as measured by relative CldU track lengths (CldU to
   IdU ratio) of the groups in e. A CIdU-to-IdU ratio closer to ~1 one
   indicates a normal speed of replication fork with a ratio <1 indicating
   a slower progression of the fork. 100–200 fibers (n = 3 biological
   repeats; ≥100–200 fibers/repeat) were counted per conditions/experiment
   and the ratio of each fiber is represented in dot plot. The median line
   is indicated in red. g The percentage of replication fork progression,
   elongation termination and new origin firing in the indicated groups.
   (n = 3 biological repeats; ≥100 fibers/repeat; error bars represent
   ±SD). h Effects on replication fork progression after exogenous dNTP
   treatment in the indicated groups (n = 3 biological repeats; ≥100–200
   fibers/repeat). i Representative western blots of the indicated RS
   markers after addition of exogenous dNTPs to H1299 cells that have been
   depleted for Trx1 or TrxR1. j Representative images of DNA fibers from
   the indicated cells treated with 1 mM NAC for 24 h. k The effect of NAC
   treatment on replication fork progression in the indicated cells.
   (n = 3 biological repeats; ≥100–200 fibers/repeat). l The effect of NAC
   treatment on replication fork progression, elongation termination and
   new origin firing in the indicated groups (n = 3 biological repeats;
   ≥100 fibers/repeat; error bars represent ±SD). [Statistical
   information: Data are presented as mean value ± SD. The p-values were
   calculated using one way ANOVA for multiple comparison. Red line in dot
   plot (f, h and k) indicates mean; *≤0.05; **** p < 0.0001; ns
   non-significant].

RRM1 cysteine oxidation is required for Trx1 KD-induced RS

   RNR, the rate limiting enzyme for dNTP synthesis, relies on both
   internal and external electron transfer flow to maintain redox cycling;
   i.e., enzymatic activity^[173]16. Reportedly, RRM1 is a substrate of
   the Trx system in E. coli and Saccharomyces
   cerevisiae^[174]15,[175]45,[176]46 although it has not yet been shown
   in mammalian cells. Given the impaired DNA replication fork progress
   and imbalanced dNTP pool in Trx1 or TrxR1-depleted cells
   (Fig. [177]5a–d), we hypothesized that Trx system inhibition-induced RS
   is associated with RRM1 enzymatic inactivity in human cells. To test
   our hypothesis, we first determined if Trx1 depletion alters the levels
   of oxidized RRM1 using iodoacetamide (IAA) to alkylate free thiols. IAA
   modification in protein thiols (reduced form) results in thiol
   alkylation and adds one negative charge per thiol modified.
   Consequently, reduced Trx1 with negative charges can be separated from
   the oxidized form due to higher negative charge during non-reducing SDS
   gel electrophoresis^[178]47. Scrambled shRNA (shCON)-treated cells
   predominantly exhibited a faster migrating form of RRM1, whereas
   depletion of Trx1 resulted in accumulation of slower migrating RRM1
   (Fig. [179]6a). Notably, protein extract reduction with DTT prior to
   gel loading led to the complete disappearance of the upper band,
   indicating that the slower migrating band represents the oxidized form
   of RRM1 while the lower band is the reduced form (Fig. [180]6a). This
   indicates that Trx system inhibition interrupted the electron flow
   between Trx1 and its substrate RRM1. Thus, Trx1 depletion-induced an
   alteration in electron flow, leading to the accumulation of oxidized
   RRM1. Of note, NAC treatment failed to exert any notable impact on RRM1
   oxidation in the Trx1 depletion group, again indicating a
   ROS-independent change in RRM1 redox recycling caused by the depletion
   (Fig. [181]6b).

Fig. 6. RRM1 cysteine oxidation is required for Trx1 KD-induced RS.

   [182]Fig. 6
   [183]Open in a new tab

   a Representative western blots of oxidized and reduced RRM1 from H1299
   cells under non-denaturing (−DTT) or denaturing (+DTT) conditions
   following iodoacetamide carboxyamidomethylation and dephosphorylation
   of the extract. b Representative western blots of oxidized and reduced
   RRM1 from the indicated H1299 cells and after treatment with 1 or 2 mM
   NAC for 24 h. c Representative western blots of oxidized and reduced
   RRM1 from Trx1-depleted H1299 cells transfected with Trx1-WT or Trx1
   redox mutants. d Representative western blots of RS markers in
   Trx1-depleted H1299 cells transfected with Trx1-WT or Trx1 redox
   mutants. e The levels of dATP in the indicated group of cells (n = 3
   biological repeats; error bars represent ±SD). f The percentage of
   cells in S phase among the indicated group of cells (n = 3 biological
   repeats; error bars represent ±SD). g The effect on replication fork
   progression in the indicated group of H1299 cells (n = 3 biological
   repeats; ≥100 fibers/repeat). h Violin plots of relative γH2AX
   intensity among the indicated groups of cells (n = 2 biological
   repeats; ≥500 cells/repeat). i A schematic diagram (Created with
   BioRender.com released under a Creative Commons
   Attribution-NonCommercial-NoDerivs 4.0 International license)
   illustrating the electron transfer flow between RRM2-RRM1 and RRM1 and
   the Trx system. j Representative western blots of oxidized and reduced
   mutant RRM1 expression in non-reducing (−DTT) and reducing (+DTT)
   conditions. k Representative western blots of oxidized and reduced
   mutant RRM1 in the indicated cells with or without Trx1 depletion. l
   Representative western blots of the indicated proteins in Trx1-depleted
   cells expressing various mutant forms of RRM1. [Statistical
   information: Data are presented as mean value ± SD. The p-values were
   calculated using one way ANOVA for multiple comparison. Red line in dot
   plot (g) and violine plots (h) indicates mean; *≤0.05; **p ≤ 0.005;
   ****p < 0.0001; ns non-significant].

   To determine if RRM1 oxidation contributes to Trx system
   inhibition-induced RS, we next determined the linkage between RRM1
   oxidation with RS using wild-type or redox-inactive Trx1 mutants. The
   functional activity of Trx1 is critically associated with the conserved
   disulfide motif (Cys^32-Gly-Pro-Cys^[184]35 of hTrx1) in the active
   site, which is indispensable for Trx1 redox activity. We exogenously
   expressed wild-type Trx1 (Flag-Trx1-WT) and redox mutant Trx1
   (Flag-Trx1(C32S) and Flag-Trx1(C35S)) in cells expressing 3’UTR shRNA
   of Trx1. Trx1-WT re-expression in Trx1-depleted cells rescued the
   effect on oxidized RRM1 accumulation in response to Trx1 depletion,
   whereas redox inactive Trx1 expression failed to rescue this effect
   (Fig. [185]6c). Correspondingly, Trx1-WT abolished the increase in
   phosphorylation of RS markers triggered by Trx1 depletion, whereas
   redox mutant Trx1 (Trx1 C32S, C35S) showed no significant impact on
   this readout (Fig. [186]6d). Additionally, unlike the redox mutants,
   Trx1-WT expression was associated with a significant repletion of the
   dNTP pool (Fig. [187]6e), and it facilitated S-phase progression
   (Fig. [188]6f). FACS gating strategies are presented in supplementary
   Fig. [189]S17. Scrutiny of DNA fibers revealed that WT-Trx1
   significantly restored the Trx1 depletion-mediated replication
   impediment; however, redox mutants Trx1 (Trx1 C32S, C35S) failed to
   mitigate the slower replication rate (Fig. [190]6g). Reinstatement of
   DNA replication and dNTP pool replenishment significantly diminished
   RS-associated DNA damage (Fig. [191]6d, h). Finally, we determined the
   CHK1i sensitivity in Trx1-depleted cells following WT or redox mutant
   Trx1 (C32S, C35S) expression. We noted a significant recovery in cell
   proliferation in cells re-expressing WT Trx1, whereas expression of the
   redox mutants failed to impact cell proliferation (Supplementary
   Fig. [192]S10a). Similarly, Trx1-WT-expression decreased CHK1i
   sensitivity, whereas expression of the redox mutants failed to curtail
   CHK1i sensitivity in Trx1-depleted cells (Supplementary
   Fig. [193]S10b). Collectively, results suggests that RRM1 oxidation is
   responsible for Trx1 depletion-mediated RS that leads to enhanced
   sensitivity to CHK1i.

   Reportedly, in the R1 (equivalent to human RRM1) subunit of E. coli
   RNR, a narrow entrance to the catalytic site sterically prohibits a
   direct reduction of the active-site disulfide by Trx or Grx. Five
   residues in R1 are essential for its activity at the active sites;
   namely, Cys225, Asn437, Cys439, Glu441 and Cys462 in E. coli^[194]18.
   R2 subunit-initiated transfer of tyrosyl radical [Tyr122 (Y•)] to
   Cys462 in R1 facilitates the formation of first disulfide bond
   formation within the R1 active site between Cys225 and Cys462 and is
   subsequently transposed to a C-terminal disulfide bond between C754 and
   C759 via a long-range electron transfer pathway. Finally, disulfide
   bond formation between two C-terminal cysteines (Cys754 and Cys759 in
   E. coli, Cys787 and Cys790 in mammals) then acts as substrate for the
   external Trx system to be reduced to facilitate the conversion
   (Fig. [195]6i). It is well established that most eukaryotes and
   prokaryotes exhibit two highly conserved cysteine residues in the
   C-terminal regions at Cys787 and Cys790 of RRM1, which are required for
   the reduction of the catalytic site. Thus, we use RRM1 redox mutants
   (Myc-RRM1 C787S and C790S) that are devoid of both internal and
   external electron exchange capabilities due to serine substitution to
   investigate its impact on Trx1 depletion-induced RS. Additionally, we
   also mutated another cysteine (Cys779) proximal to the C-terminal tail
   that is closer to the Trx1 substrate Cys787-X-X-Cys790. The site Cys779
   might be involved in regulation of RNR activity by other mechanism, but
   it has no direct role in exchange of electrons through the external Trx
   system.

   We started with assessment of RRM1 reduction-oxidation following
   Cys→Ser substitution in the previously indicated cystine residues. We
   found that single mutations at Cys787S and Cys790S and double mutations
   (Cys787S/Cys790S) increased the accumulation of RRM1 oxidation
   (Fig. [196]6j). However, RRM1-WT-and a single mutation at C779S did not
   show detectable amount of oxidized form accumulation. Interestingly, a
   single Cys substitution at C779S does not result in accumulation of
   oxidized RRM1; however, the C779S mutation in combination with Trx1
   substrate sites Cys787 and/or Cys790 (C779S/C787S; C779S/C790S and
   C779S/C787S/C790S) leads to accumulation of oxidized RRM1. This
   indicates the redundancy of Cys779 in redox recycling of RRM1 and no
   participation in electron exchange via external Trx system
   (Fig. [197]6j). Reportedly, overexpression of wild-type RRM1 in
   Trx1-deficient S. cerevisiae rescued Trx deficiency-linked phenotypes
   and decreased the oxidized RRM1 due to a simultaneous increase in RRM1
   reduction status^[198]23. Similarly in human cells, we observed that
   exogenous Myc RRM1-WT and C779S does not accumulate in their oxidized
   form under the experimental condition, suggesting their efficient redox
   recycling (Fig. [199]6k). Consistent with these results, we found
   expression of exogenous RRM1-WT reduced Trx1-depletion-induced RS
   whereas a RRM1-redox mutation failed to do so, which is consistent with
   the previous observation in yeast showing that RRM1-WT, but not a redox
   mutant, rescues the RS in Trx1-depleted cells (Fig. [200]6l). The C779S
   expression was found to have no impact on the oxidation level of RRM1
   and the Trx1 depletion-induced RS, like the effect of RRM1 WT. In
   summary, our study suggests that Trx1 KD leads to RS and slows
   replication forks velocity by increasing RRM1 oxidation and thus
   decreasing the dNTP pool in mammalian cells. Therefore, our finding
   establishes that RRM1 serves as one of the direct substrates of Trx1,
   which undergoes redox cycling via the Trx system, and C787 and/or C790
   oxidation of RRM1 are critical for Trx1 KD-induced RS.

CHK1 inhibition disrupts Trx1/TrxR1 depletion-mediated E2F1-RRM2 pathway

   As previously reported, the CHK1-E2F1-RRM2 pathway is activated under
   conditions of RS^[201]28. We found that Trx system interruption
   triggered ATR/CHK1 activation (Supplementary Fig. [202]S5a–[203]j) and
   increased the expression of RRM2, which is accompanied by
   overexpression of the transcription factor E2F1 (Fig. [204]7a). CHK1
   inhibition abrogated Trx1 or TrxR1 depletion-induced E2F1-RRM2
   induction (Fig. [205]7b). Mechanistically, functional RNR depends on
   both RRM1 and RRM2, and RNR activity require long-range electron
   transfer from RRM2 to RRM1 to produce dNTPs (Figs. [206]6i and
   [207]7c). In support of the result that CHK1i limits RRM2 expression in
   Trx1 or TrxR1-depleted cells, CHKi leads to a further decline in dNTP
   production in cells depleted of Trx1 or TrxR1 alone (Fig. [208]7d and
   Supplementary Fig. [209]S11). Furthermore, CHK1i significantly
   decreased DNA replication fork velocity in Trx1 or TrxR1-depleted cells
   (Fig. [210]7e, f). Therefore, RRM2 inhibition by CHKi further impaired
   RNR activity and dNTP biosynthesis, leading to a slow replication fork
   progression in cells with Trx system disruption.

Fig. 7. Trx1 or TrxR1 depletion triggers the CHK1-E2F1-RRM2 pathway, which
can be inhibited by CHK1i.

   [211]Fig. 7
   [212]Open in a new tab

   a Representative western blots of the indicated proteins of the
   E2F1-RRM1 pathway after Trx1 or TrxR1 depletion. b Representative
   western blots of E2F1 and RRM2 upon Trx1 or TrxR1 depletion and CHK1i
   treatment at the indicated doses. c A schematic diagram (Created with
   BioRender.com released under a Creative Commons
   Attribution-NonCommercial-NoDerivs 4.0 International license)
   illustrating the expected impact of CHK1i on Trx1 or TrxR1
   depletion-mediated E2F1-RRM2 axis alterations and subsequent RRM1
   activity. d The levels of dATP in cells with Trx1 or TrxR1 depletion
   with or without CHK1i treatment (n = 3 biological repeats; error bars
   represent ±SD). Effects on replication fork progression in cells with
   Trx1 or TrxR1 depletion with or without CHK1i treatment as visualized
   by track lengths (e) and their relative quantification (f) (n = 3
   biological repeats; ≥100 fibers/repeat). Representative western blots
   of oxidized and reduced RRM1 in Trx1-depleted cells with or without HU
   or CHK1i treatment (g) and their quantification below each set of
   western blots (h) (n = 3 biological repeats; error bars represent ±SD).
   Representative western blots of oxidized and reduced RRM1 in cells
   expressing various redox mutants of RRM1 treated with CHK1i or HU (i)
   and their quantification (j) (n = 2 biological repeats). [Statistical
   information: Data are presented as mean value ± SD. The p-values were
   calculated using one way ANOVA for multiple comparison. Red line in dot
   plot (f) indicates mean; *≤0.05; **p ≤ 0.005; ***p ≤ 0.001;
   ****p < 0.0001; ns non-significant].

   To validate our hypothesis that CHK1i-induced inhibition of RRM2
   expression indeed impairs RNR activity, we determined the impact of
   CHK1i on RRM1 oxidation in the cells with or without Trx1 depletion.
   The cells were treated with the CHK1i or hydroxyurea (HU), an
   established RNR inhibitor that inhibits RRM2 via scavenging the
   iron-tyrosyl free radical^[213]48. HU significantly decreased the
   accumulation of oxidized RRM1 triggered by Trx1 depletion
   (Fig. [214]7g, h, left panel). Similarly, CHK1i also abrogated Trx1
   depletion-induced RRM1 oxidation (Fig. [215]7g, h, right panel).
   Mechanistically, a long-range intermolecular electron transfer into
   RRM1 (R1 subunit) is initiated at Tyr122 cofactor (Y•) at a di-nuclear
   iron cluster (Fe^IIIFe^III) in RRM2 (R2 subunit) (Fig. [216]6i). Thus,
   a RRM2-localized radical initiates the formation of the first disulfide
   bond in RRM1. When RRM2 is inhibited by CHK1i or HU, the electron
   transfer flow was interrupted at the initial step; thus, leading to the
   decrease in the oxidized RRM1.

   To further investigate the CHK1i-mediated RRM2 inhibition and its
   effect on the long-range intermolecular electron transfer to RRM1, we
   next determined the RRM1 oxidation status in CHK1i- or HU-treated cells
   expressing Myc-RRM1 WT and its redox mutants. As with HU, CHK1i
   significantly abrogated the accumulation of the oxidized form of the
   redox mutants of RRM1, which supports the critical role of RRM2 in the
   initiation of electron transfer within RRM1 (Fig. [217]7i, j).
   Therefore, simultaneous inhibition of RRM2 by CHKi in the absence of
   the external electron donor Trx1 renders RNR inactive to a greater
   extent than CHK1i treatment alone, leading to a drastic loss of the
   dNTP pool that further sensitizes the anti-tumor effects of the drug.

The TrxR1 inhibitor AUR exerts an antitumor effect via synergistic
interaction with CHK1i in NSCLC cells

   Inhibitors of the Trx system have been developed, but most of them have
   failed to translate to the clinic due to high toxicity and insufficient
   targeting efficacy. However, AUR, a gold-containing triethyl phosphine
   compound that was originally approved to treat rheumatoid arthritis, is
   a well-known inhibitor of TrxR. Therefore, we next determined if AUR
   has a similar impact on RS and CHK1i sensitivity as Trx1 or TrxR1
   depletion. We noted a significant dose-dependent attenuation in cell
   growth in several NSCLC cells following AUR treatment (Supplementary
   Fig. [218]S12a). Additionally, AUR exacerbated CHK1-mediated cell
   killing in a panel of different NSCLC cell lines and severely limited
   cell viability (Fig. [219]8a). Concordantly, combined AUR and CHK1i
   treatment significantly decreases cell proliferation compared to either
   treatment alone (Supplementary Fig. [220]S12b). An ATRi also shows a
   similar synergistic interaction with AUR (Supplementary
   Fig. [221]S12c). To investigate whether the combination of AUR and
   CHK1i is synergistic or additive in nature, we used SynergyFinder 2.0
   to generate an inhibitory dose response matrix of AUR and CHK1i and
   found a dose-dependent increase in cell toxicity for the combinatorial
   treatment (Supplementary Fig. [222]S12d). The drug interactive
   algorithm utilizes four different synergy models (HAS, Bliss, Loewe and
   ZIP) as a reference and uses their extensions to calculate the synergy
   score for combinatorial data of multiple drugs. A synergy scores lower
   than −10 indicates an antagonistic response and a score between −10 to
   +10 indicates an additive effect. A synergy score >+10 indicates a
   synergistic interaction between the given drugs. The synergy score
   anticipated by all four different synergy models, ZIP (score: 16.54)
   (Fig. [223]8b), Loewe (score: 20.94) (Fig. [224]8c), Bliss (score:
   16.42) (Fig. [225]8d) and HSA (score: 25.9) (Fig. [226]8e), generated a
   score greater than +10; indicating a synergistic interaction between
   AUR and CHK1i. AUR exerted a massive synergistic potency shift of
   34752.01-fold against CHK1i and CHK1i displayed synergistic potency
   shift of 7.025-fold against AUR. However, AUR and CHK1i displayed only
   a 0.15-fold negative cooperativity in combination (Supplementary
   Fig. [227]S12e).

Fig. 8. AUR and CHK1i have a synergistic interaction with regard to their
antitumor activity and RS.

   [228]Fig. 8
   [229]Open in a new tab

   a Relative survival of different NSCLC cell lines treated with AUR or
   AUR + CHK1i. (n = 3 biological repeats in triplicates; error bars
   represent ±SD). Different synergy scores, such as ZIP (b), Loewe (c),
   Bliss (d) and HSA (e) indicate the true synergy between CHK1i and AUR.
   A synergy score ≥10 is considered to be synergistic. f Growth rate of
   the tumors in the indicated groups over the indicated course of time
   (n = 10 animals in each group; error bars represent ±SD). g and the
   survival rates by Kaplan–Meier analysis (Control n = 8; AUR n = 9;
   CHK1i n = 8; AUR+CHK1i n = 10). Representative images of tumors excised
   from indicated groups (h), the average weight of the excised tumors
   (n = 6 tumors form each group; error bars represent ±SD) at the
   endpoint of treatment (i), of a mouse xenograft H1299 model of NSCLC
   treated with CHK1i (25 mg/Kg) and AUR (10 mg/Kg). j–m The synergy
   between AUR and CHKi, as detected by a PDX model (Control n = 8; AUR
   n = 8; CHK1i n = 10; AUR + CHK1i n = 8; error bars represent ±SD). The
   growth rate of PDX-72 tumors in indicated treatment groups, over the
   indicated course of time and the survival rates by Kaplan–Meier
   analysis (n = 5 animal/group) (k). Representative images of PDX-72
   tumors excised from indicated groups (l), the average weight of the
   excised tumors at the endpoint of treatment (Control n = 7; AUR n = 7;
   CHK1i n = 8; AUR + CHK1i n = 9 individual tumor from the indicated
   groups; error bars represent ±SD) (m). n Tumor organoid generated from
   PDX-72 tumors (PDO-72) were stained with Hoechst, Calacein AM and PI to
   stain nucleus, live cells and dead cells respectively. Representative
   organoids in indicated treatment groups. o The PI intensity, an
   indicator of cell death of organoids, in each treatment groups (n ≥ 50
   organoids/groups from 2 biological repeats; error bars represent ±SD).
   p Representative images of 10-day-old organoids treated with the
   indicated inhibitors for 72 h. q The growth and size of the organoids
   from each treatment group (n ≥ 50 organoids/groups from 2 biological
   repeats; error bars represent ±SD). [Statistical information: Data are
   presented as mean value ± SD. The p- values were calculated using one
   way ANOVA for multiple comparison; *p ≤ 0.05; **p ≤ 0.005;
   ***p ≤ 0.001].

   Furthermore, we validated our in vitro combinatorial treatment
   strategies using an in vivo xenograft model. We treated tumor-bearing
   animals with AUR (10 mg/kg) and CHK1i (25 mg/kg) (Supplementary
   Fig. [230]S13a). Animals treated with the combined AUR and CHK1i showed
   significant reduction in tumor growth (Fig. [231]8f), leading to
   increased overall survival (Fig. [232]8g). Overall tumor burden which
   was indicated by tumor size and weight was significantly reduced in
   combinatorial treatment group (Fig. [233]8h, i). Of note, no
   significant loss in overall body weight indicates that the mice
   tolerated the combination therapy with no apparent toxicity
   (Supplementary Fig. [234]S13b). Additionally, we also assessed the
   changes in RS, DNA damage and apoptosis across the treatment groups and
   found that tumors treated with the combined inhibitors had greater RS,
   DNA damage and apoptosis (Supplementary Fig. [235]S13c–f).

   Because of the fidelity of patient-derived xenograft (PDX) models to
   clinical scenarios, these models have been widely employed in exploring
   therapeutic effects. To further validate our results from the xenograft
   model, we utilized the PDX (PDX-72) of NSCLC^[236]49 to determine the
   synergy of the combinatorial treatment approach. A combination of AUR
   and CHKi significantly slowed the tumor growth rate compared to
   monotherapy (Fig. [237]8j), resulting in an increased survival rate
   (Fig. [238]8k) and an overall smaller tumor burden as assessed by
   significantly less tumor size and weight (Fig. [239]8l, m). The
   treatment outcome of PDX-72 suggests a high sensitivity to the
   combination. Of note, the sensitivity of PDX-72 to AUR alone suggests a
   deficiency of Grx-GSH system. Additionally, we also established that
   the inhibition of PDX-72 growth by the combinatorial treatment is due
   to greater apoptosis as measured by a TUNEL assay of tumor tissues
   (Supplementary Fig. [240]S14a–[241]c). Further, the significant impact
   of the AUR and CHKi combination was further validated by generating PDX
   model-derived organoids (PDOs). We dissociated the PDX tumor and
   isolated the human cancer cells via a human cancer cell isolation kit
   and validated the human cancer cells by human-specific EpCAM (CD326)
   antibody-based flow cytometry (Supplementary Fig. [242]S14d, FACS
   gating strategies are presented in Supplementary File, Supplementary
   Fig. [243]S17). Combinatorial treatment led to a significant increase
   in cell death as measured by PI uptake in dead cells compared to the
   monotherapy (Fig. [244]8n, o). Furthermore, we detected a higher level
   of cleaved PARP1 in PDOs treated with the combination therapy as
   compared to untreated PDOs (Supplementary Fig. [245]S14e). To exclude
   the possibility of off-target effects of AUR, we further validated the
   results by utilizing the Trx-specific inhibitor PX-12 and found that
   PX-12 also significantly decreased the organoid formation rate,
   resulting in smaller organoid formation as compared to untreated or
   mono treatment (Fig. [246]8p, q).

   Taken together, our results suggest a strong synthetic lethality occurs
   between CHK1 inhibition and Trx system inhibition in NSCLC cells in
   vitro and in tumors in vivo.

AUR triggers RRM1 oxidation and dysregulates the dNTP pool and has a synergy
with CHK1i

   Finally, we investigated whether the pharmacological inhibition of the
   Trx system mimics the Trx system depletion-mediated RS-associated
   phenotypes. We first determined the impact of AUR on RRM1 redox
   recycling and found accumulation of oxidized RRM1 accompanied with
   increased oxidized aggregation of its electron doners Trx1 and TrxR1
   following AUR treatment (Fig. [247]9a). Inactivation of RRM1 redox
   cycling by AUR due to disruption of electron flow abrogates the RNR
   enzymatic activity, resulting in defective dNTP biogenesis
   (Fig. [248]9b). Additionally, AUR treatment increased the production of
   intracellular ROS (Supplementary Fig. [249]S15a) and resulted in
   S-phase accumulation of the cells, but NAC co-treatment failed to
   rescue AUR-induced cell cycle arrest, suggesting a ROS-independent cell
   cycle arrest (Fig. [250]9c, d; FACS gating strategies are presented in
   Supplementary Fig. [251]S17). Additionally, exogenous reconstitution of
   the dNTPs significantly mitigated AUR-induced impediment of the
   replication fork (Fig. [252]9e, f). Consistent with the Trx1 or TrxR1
   depletion data, AUR treatment alone increases RS in NSCLC cells
   (Supplementary Fig. [253]S15b). CHKi1 further intensified the
   AUR-induced RS and produced massive DNA damage, leading to the
   accumulation of unpaired DSBs (Fig. [254]9g–j). By DNA fiber assays we
   found that combinatorial treatment with CHK1i and AUR exerted a more
   profound defect in the progression of the replication tracks and
   resulted in increased replication termination (Fig. [255]9k, l).
   Additionally, we assessed the apoptotic efficacy of AUR and CHKi in
   NSCLC cells via annexin staining and found a significantly higher
   percentage of apoptotic cells among the combination treatment group
   compared to single treatment (Supplementary Figs. [256]S15c, d, FACS
   gating strategies are presented in Supplementary Fig. [257]S17).

Fig. 9. AUR treatment increases RRM1 oxidation and impairs replication fork
elongation due to depletion of the dNTP pool.

   [258]Fig. 9
   [259]Open in a new tab

   a Representative western blots of oxidized and reduced RRM1 in H1299
   cells treated with AUR. b Levels of dATP in the AUR-treated cells
   (n = 4 biological repeats; error bars represent ±SD). Effects of NAC on
   the accumulation of S phase cells caused by AUR treatment (c) and its
   quantification (n = 2 biological repeats; error bars represent ±SD)
   (d). FACS gating strategies are presented in Supplementary
   Fig. [260]S17. Effects of AUR treatment on replication fork progression
   with or without dNTP supplementation as visualized by DNA fiber tracks
   (e) and their relative quantification (f) (n = 3 biological repeats;
   ≥100 fibers/repeat). g Representative western blots of RS markers in
   cells treated with AUR and an CHK1i. Percent of cells with
   p-RPA32(S33)-positive foci (h), those with ≥5 foci of γH2AX (i) and
   with pan-γH2AX nuclear staining (j) in those treated with AUR, a CHK1i
   or both (n ≥ 100 cells/repeat from 3 biological repeats; error bars
   represent ±SD). k The effect of combined AUR and CHK1 inhibition or
   monotherapy on the percentage of replication fork progression,
   elongation termination and new origin firing (n = 3 biological repeats;
   ≥100 fibers/repeat; error bars represent ±SD). l The effects of AUR and
   CHK1 inhibition or monotherapy on replication fork progression. The dot
   plots show the relative CIdU track length (n = 3 biological repeats;
   ≥100 fibers/repeat). m A schematic diagram (Created with BioRender.com
   released under a Creative Commons Attribution-NonCommercial-NoDerivs
   4.0 International license) illustrating our working model of how
   Trx1/TrxR1 depletion or inhibition synergistically interacts with CHK1i
   to promote more anti-tumor effects than either therapy alone. In this
   model, both RRM1 and RRM2 are involved in the synergistic interaction
   between Trx system inhibition and CHK1i. [Statistical information: Data
   are presented as mean value ± SD. The p-values were calculated using
   one way ANOVA for multiple comparison. Red line in dot plot (f and l)
   indicates mean; *p ≤ 0.05; ** p ≤ 0.005; *** p ≤ 0.001; ****
   p ≤ 0.0001; ns: non-significant].

   We then tested PX-12 and found that it effectively inhibited the
   proliferation of NSCLC cell lines in a synthetic lethal manner when
   combined with the CHK1i (Supplementary Fig. [261]S16a). The combined
   treatment of PX-12 and CHK1i resulted in an increased level of cleaved
   PARP1 like AUR combination with CHKi (Supplementary Fig. [262]S16b).
   Additionally, PX-12 treatment led to an increase in the oxidation of
   RRM1 (Supplementary Fig. [263]S16c). Thus, our data, obtained with
   PX-12 demonstrates a synthetic interaction with CHK1i in the treatment
   of NSCLC, which is like what we found in the cells treated with AUR and
   CHKi. Therefore, there is a synergistic interaction between inhibition
   of different components of the Trx system and CHK1 in the treatment of
   NSCLC.

Discussion

   A steady and sufficient supply of dNTPs is required for error-free
   replication of DNA in proliferating cells. The maintenance of a
   balanced pool of dNTPs is tightly regulated by RNR and interruption of
   its enzymatic activity results in insufficient dNTPs and its associated
   phenotypes, such as RS and S-phase cell accumulation^[264]50,[265]51.
   Our study reveals three main findings (Fig. [266]9m). Firstly, we found
   that Trx system inhibition-triggered RS relies on Trx system-mediated
   redox recycling of RRM1 activity in mammalian cells that is critically
   dependent on the C-terminal cystines of RRM1, but that is independent
   of ROS. Secondly, inhibition of Trx system-induced elevation in RRM2 by
   CHK1i terminates its tyrosyl radical production activity and shuts off
   long-range electron transfer to RRM1. And thirdly, combination of Trx
   system inhibition and CHK1i can serve as a potent combinational therapy
   to strikingly reduce tumor growth and prolong survival. And notably,
   these findings suggest that AUR could be repurposed to treat NSCLC when
   combined with a CHKi.

   The side chain of cysteine (Cys) undergoes both reversible and
   irreversible redox post-translational modifications (redox-PTMs) and is
   tightly linked with structural and functional integrity of the
   proteins. Here we demonstrate that the Trx system acts as an electron
   donor for RRM1 in mammalian cells and interruption of this system
   increases RS due to dNTP pool depletion due to inhibition of redox
   cycling of RRM1. Further, we demonstrate that the depletion or
   inhibition of the Trx system increases RRM1 oxidation, resulting in
   slow replication fork progress due to an imbalanced dNTP pool. The
   exogenous addition of dNTPs rescues the RS induced by interruption of
   the Trx system, which supports the role of RRM1 redox recycling in DNA
   precursor synthesis. Consistent with previous reports showing that
   replication initiation is important for the rescue of collapsed
   replication forks that are induced by HU-mediated RNR
   inhibition^[267]44 and that the firing of multiple origins in cells
   with lower dNTPs increases consumption of dNTPs and fork
   termination^[268]52,[269]53, we observed an increase in new origin
   firing and fork termination rate in response to Trx system
   interruption. Also, the RRM1 oxidation status is closely associated
   with Trx system interruption-induced RS, which is demonstrated by using
   a Trx1 -WT and Trx1- redox mutant expression system. Trx1 contains a
   thiol-exchange catalytic site at Cys32-X-X-Cys35 and alterations in
   Cys32 and/or Cys35 can interrupt the flow of electrons due to increased
   RRM1 oxidation. Evidently, catalytically active Trx1-WT facilitates the
   recycling of oxidized RRM1 in response to Trx1 loss, resulting in
   striking mitigation of Trx1 depletion-induced RS and dNTP loss, but
   redox mutants Trx1(C32S and C35S) were unable to recycle the RRM1
   oxidation and/or failed to restore Trx1 depletion-linked phenotypes,
   such as increased RS and early S phase accumulation, as well as dNTP
   pool scarcity. The observation that the increased dNTPs accompanied by
   restoration of replication velocity occurred in Trx1-depleted cells
   that express catalytically active Trx-WT, but not a redox mutant,
   suggests that thiol-exchange of Trx1 is critical in regulating
   enzymatic activity of RNR in mammals. Finally, consistent with the
   observation in yeast^[270]23, we demonstrated that expression of
   RRM1-WT restored the Trx1 depletion-induced RS, whereas expression of a
   RRM1 redox mutant failed to do so. Therefore, our data strongly suggest
   that RRM1 oxidation contributes to the Trx system interruption-induced
   RS by causing dNTP deficiency.

   The formation of individual cysteine redox-PTMs depends on many
   factors, including the reactivity of the individual cysteine residue,
   its surrounding environment and the composition of the local redox
   environment. A total of 16 cysteine residues are distributed between
   the N-terminus and the C-terminus of the 792 amino acid-long mammalian
   RRM1; however, the E. coli and S. cerevisiae isoforms contain 13
   cysteine residues along their 811 amino acid-long chains. As previously
   reported, two cysteines in the C-terminus of R1/RRM1 (Cys754 and Cys759
   in E. coli R1 and Cys787 and Cys790 in mammalian RRM1) are critical for
   the last step of internal electron transfer. The disulfide bond between
   these two cysteines needs to be reduced by the external redox system to
   complete the redox and recycling of R1/RRM1. Consistent with the
   results from bacteria and yeast, we found that the two cysteine (Cys787
   and Cys790) are critical for the recycling of RRM1 oxidation in
   mammalian cells. Our findings using C-terminal cysteine (Cys787 and
   Cys790) mutants, which fail to undergo thiol-exchange due to Cys→Ser
   mutation, suggest that thiol exchange at the disulfide bond between
   Cys787 and Cys790 site is critical for RRM1 oxidation recycling.
   Indeed, mutations in these two cysteines block the transfer of
   electrons, as we found an increased oxidized form of RRM1 in cells
   expressing C787A and/or C790A. In addition, a single mutation at Cys787
   or at Cys790 significantly abolishes the RRM1 recycling, indicating
   that Cys787 and Cys790 are involved in Trx1-mediated RRM1 recycling in
   mammalian cells. In contrast, a single mutation in another cystine
   (C779S) proximal to C-terminal Cys787 and Cys790 has no effect on RRM1
   recycling. Although it has been suggested that Cys779 in RRM1 is
   essential for regeneration of RNR activity in vitro in mammalian
   cells^[271]54, in our study it does not participate in thiol exchange
   with Trx1 in mammalian cells. It would be very interesting to determine
   the role of other cysteines in playing specific roles in determining
   the activity of human RRM1.

   Of note, Trx1 or TrxR1 depletion increases ROS, which could cause the
   oxidized base damage. During the oxidized base repair by base excision
   repair (BER) pathway, an apurinic site (AP site) is generated and then
   processed further into DNA single-strand breaks via backbone incision
   of AP-endonuclease 1. Theoretically, the AP site is processed to DNA
   single-strand breaks or DSBs when encountered within the replication
   forks^[272]55. Thus, we originally speculated that increased ROS
   contributes to Trx1 or TrxR1 inhibition-induced RS. Surprisingly, ROS
   is not responsible for the Trx1 or TrxR1 inhibition-induced RS because
   the classic ROS scavenger NAC has no significant impact on RS, RRM1
   oxidation and antitumor activity as a monotherapy or in combination
   therapy. Thus, the ROS-induced DNA damage in the condition of Trx1 or
   TrxR1 interruption are likely rapidly repaired given the fast nature of
   repair of ROS-induced DNA damage^[273]56,[274]57. It would be very
   interesting to determine if Trx1 or TrxR1 inhibition-induced ROS is
   important for the RS and DNA damage and tumor cell growth suppression
   in BER-deficient cells

   RRM2 has been identified as the regulatory subunit of RNR due to its
   tyrosyl radical-producing activity, which is important for overall RNR
   activity. RRM2 transcriptional expression and relocation is regulated
   by ATR/CHK1 pathway during RS^[275]28,[276]58. We found that
   CHK1-mediated E2F1-RRM2 induction was triggered during Trx1 or TrxR1
   depletion-induced RS, suggesting a RRM2 expression-associated mechanism
   may compensate for the dNTP deficiency due to RRM1 oxidation. This
   compensatory mechanism could funnel the tyrosyl radical to RRM1 to
   generate sufficient dNTPs to rescue cells stalled in S-phase. In
   support of our hypothesis, CHK1 inhibition abrogates the Trx1 or TrxR1
   depletion-induced RRM2 expression and leads to deceased RRM1 oxidation
   that depends on the RRM2-mediated tyrosyl radical transfer and more
   profound dNTP pool scarcity. The additional evidence supporting the
   loss of tyrosyl radical production capacity of RRM2 following CHK1
   inhibition is from the direct measurement of RRM1 oxidation. CHK1
   inhibition in Trx1-depleted cells significantly decreases the
   accumulation of oxidized RRM1, probably because CHK1i-mediated
   abrogation of RRM2 expression resulted in lower RRM1 oxidation due to
   decreased tyrosyl activity in Trx1-depleted cells. Further, the
   accumulation of oxidized RRM1 due to mutation in C-terminal cystines
   (Cys787 and Cys790) was also abrogated following CHK1i or HU treatment
   as both mediated the loss of RRM2 activity, leading to the inhibition
   of long-range electron transfer to RRM1, subsequently leading to
   abrogating RRM1 oxidation because the RRM2-mediated initial tyrosyl
   radical transfer is blocked. Thus, our hypothesis is that the
   disruption of the Trx system induces oxidation of RRM1, a crucial step
   for the Trx system interruption-induced RS. This occurs by impairing
   dNTP production due to decreased RNR activity. When CHK1 activity is
   inhibited in these cells, CHK1 inhibition can further hinder RNR
   activity by affecting RRM2 through an E2F1-mediated mechanism
   (Fig. [277]9m). Therefore, there is a synergistic interaction between
   Trx1 or TrxR1 inhibition and CHK1i.

   The ATR/CHK1 is important for cell survival under conditions of RS via
   promotion of cell cycle checkpoint, DNA synthesis, replication fork
   stability and regulation of replication initiation^[278]28. Thus, we
   cannot exclude the impact of CHKi-induced cell cycle checkpoint
   interruption is also important for the synergy because inhibition of
   Trx system and CHK1 leads to increased catastrophe, a situation where
   cells try to move to a mitotic stage with the damaged DNA because of
   the impaired cell cycle checkpoint. Thus, in addition to the DNA
   synthesis deficiency, abrogation of cell cycle checkpoints might also
   contribute to CHK1i-induced antitumor activity, especially with Trx
   system interruption.

   Although the most dramatic sensitization has been observed in
   preclinical studies when CHK1 inhibitors are combined with HU or
   gemcitabine, two antimetabolites that inhibit RNR^[279]2,[280]59, only
   limited efficacy and toxicity has been noted in clinical
   trials^[281]5,[282]8,[283]60. What made us believe that combination of
   CHK1 and Trx1 or TrxR1 inhibition might be better than the combination
   with RNR targeting chemotherapy is because of the following three
   reasons. Firstly, elevated Trx1 or TrxR1 expression in tumor cells
   cause them to rely on the Trx system to maintain redox
   homeostasis^[284]61. In support of the hypothesis that cancer cells are
   highly dependent on Trx1 or TrxR1 for survival due to their high
   metabolism and proliferation, loss of the Trx system showed no effect
   on normal replicative potential. For instance, Txnrd1-deficient mouse
   hepatocytes have equivalent developmental and regenerative
   proliferative potentials^[285]62. Indeed, we found that Trx1 or TrxR1
   inhibition has much less impact in RS in the untransformed lung cells.
   Secondly, a very broad deficiency of the Grx-GSH system was observed in
   human tumors, including NSCLC, which provides a unique opportunity to
   target a subset of NSCLC by Trx1 or TrxR1 inhibition. It is quite
   common that loss of the Grx-GSH system due to deficiency in its
   components and other upstream factors render cancer cells highly
   reliant on the Trx system and show synthetic lethality with Trx system
   inhibition^[286]34,[287]63–[288]66. Lastly, CHK1i, by regulating the
   RRM2 compensation pathway, provides the foundation for synthetic
   lethality. Of note, our study also suggests the potential for combining
   Trx1 or TrxR1 inhibition with homologous recombination (HR) inhibitors
   as the stalled replication forks triggered by inhibition of RNR
   activity highly depends on HR-associated repair^[289]44. It would be
   quite interesting to determine if Trx1 or TrxR1 inhibition can be
   combined with HR inhibitors to suppress tumor growth in NSCLC and other
   human cancers.

   We anticipate that the status of glutathione-glutaredoxin (Grx)-GSH
   system (NADPH, glutathione reductase, GSH and Grx) might impact
   combinatorial strategies via compensation of Trx system loss^[290]34.
   Nevertheless, it remains largely uncertain, particularly in mammals, if
   these two systems have an overlapping role in promoting RNR redox
   recycling and dNTP synthesis. Although it has been reported that Trx
   and Grx-GSH are dithiol electron donors of E. coli RR and Grx1 is the
   most efficient electron donor for the enzyme activity^[291]17, the Trx1
   and Grx1 systems show similar catalytic efficiencies as recombinant
   mouse RR complex^[292]17,[293]67. Interestingly, a recent study
   suggested that for human cells, hTrx1 is a much more efficient
   reductase for RRM1 regeneration than hGrx1^[294]68. Our study supports
   the concept that both systems might be important in terms of its
   regulation of RNR as Trx1 or TrxR1 inhibition has a better
   sensitization activity in cells with a deficient Grx-GSH system. Thus,
   investigating a potential a compensatory role for the Trx and Grx-GSH
   systems in the regulation of RNR activity in mammalian cells is
   certainly warranted in future studies, including the efficacy of Trx1
   or TrxR1 and CHK1 combinatorial inhibition in treating cancer cells
   with a Grx-GSH deficiency. It is worth noting that other factors may
   also be important, such as the p53 status as it has been shown that it
   affects the sensitivity of CHK1 inhibitors. p53 is important for the
   G1/S checkpoint. Disruption of G1/S checkpoint control due to loss of
   p53 leaves cells reliant on G2/M arrest for DNA repair when the cells
   are challenged with DNA damaging agents^[295]2,[296]69,[297]70. CHK1
   phosphorylates and inhibits its substrates, the phosphatases CDC25C and
   CDC25A, leading to arrest at the G2/M checkpoint^[298]2. Therefore,
   p53-deficient cells are normally more sensitive to CHK1
   inhibitor-associated cancer therapy. As most of the cell lines in our
   study harbored TP53 mutations, it would be interesting to investigate
   the impact of the p53 status on the efficacy of this combination as
   well.

   The Trx system is suggested to be a promising target for cancer
   therapy^[299]71. The anti-rheumatoid arthritis drug AUR is now
   recognized as a potent TrxR inhibitor, and attempts have been made to
   repurpose AUR for the treatment of infection and
   cancers^[300]72–[301]75. However, the molecular mechanisms underlying
   the anticancer effects of AUR are not yet completely understood, which
   limits its choice for a combination therapy. Currently, most studies
   have focused on ROS and mitochondrial function, including proteasomal
   inhibition-linked mechanisms, in the mechanisms underlying the
   anti-tumor effects of AUR^[302]65,[303]76–[304]78. In human lung
   cancer, AUR induces cytotoxicity via ROS production in stem cell-like
   cancer cells^[305]75. The primary molecular targets of AUR have also
   been described to be mitochondrial and (to a lesser extent) cytoplasmic
   TrxR (TrxR1)^[306]76. The anticancer activity and anti-rheumatoid
   impact of AUR have been found to be linked to inducible ROS and/or
   glycolysis^[307]75,[308]77,[309]79–[310]82. In our study, AUR has a
   synergistic interaction with CHKi and AUR treatment increased RS due to
   poor dNTP production, resulting in defective replication fork
   elongation following accumulation of oxidized RRM1. Here, we have shown
   that AUR treatment leads to RS and DNA damage and identify an
   unexplored mechanism of redox inhibition of RRM1 consequently causing
   RS. Thus, Trx system inhibition-induced deficiency most likely
   contributes to RS and antitumor growth activity. In the future, the
   challenge will be to identify the determining factors for the antitumor
   activity of AUR and any different mechanisms that explain these
   effects, as well as identifying the patient population that would most
   benefit from its treatment.

   Our study provides mechanisms by which AUR in combination with CHK1i
   inhibits cancer cell growth via regulation of cysteine oxidation of
   RRM1. Thus, repurposing of this drug in combination with CHK1
   inhibitors for NSCLC treatment might be an area to focus on. Although
   AUR is considered safe for human use in treating rheumatoid arthritis
   and the drug has a well-known toxicity profile^[311]83, the dose used
   for its antitumor activity in current preclinical studies, including
   here, is higher compared to those used in clinical settings for the
   treatment of rheumatoid arthritis. The next clinical steps could be
   validation of AUR-induced CHK1i sensitization in NSCLC or other tumor
   models in clinical settings to study the antitumor activity in an
   immune component host. In addition, more specific and potent inhibitors
   targeting Trx1 or TrxR1 needs to be developed.

   In summary, the interruption of Trx1 or TrxR1 increases RS due to
   depletion of the dNTP pool by inhibition of RNR redox recycling. We
   propose that increased RRM1 oxidation following Trx system inhibition
   and the abrogation of RRM2 pathway by CHKi contributes to the
   synergistic interaction by the severe loss of RNR function. Our study
   reveals an underappreciated mechanism responsible for the Trx or TrxR1
   inhibition-induced RS, and we propose a potential combinational
   approach to treat a subset of NSCLC. We foresee great potential for
   future efforts towards developing inhibitors for targeting the Trx
   system as a monotherapy or in combination with other targeted
   therapies, from a redox perspective involving the Trx system.

Methods

Cell line, cell culture reagents and inhibitors

   All experiments complied with protocols and conditions approved by the
   Office of Research and Institutional Animal Care and Use Committee of
   The Ohio State University (Columbus, OH). The NSCLC cell lines H1299,
   A549, Calu-6, SK-MES-1, H2006, H460 and H1437 were purchased from ATCC
   and maintained in DMEM (H1299 and A549); MEM (Calu-6 and SK-MES-1) or
   RPMI1640 (H2006, H460 and H1437) medium supplemented with 10% FBS and
   1% penicillin/streptomycin. Cell lines were routinely screened for
   mycoplasma contamination using LookOut® Mycoplasma PCR detection kit
   from Sigma (Cat. MP0035). All cell lines were authenticated via STR
   profiling by Genomics Shared Resources at the Ohio State University.
   Cells were cultured at 37 °C in a humidified incubator at 95%
   humidified oxygen and 5% CO2. The CHK1 inhibitor LY2603618 was
   purchased from APExBIO (cat. #A8638) and the ATR inhibitor VE-821 (cat.
   #S8007) was purchased from Selleckchem. Auranofin (Cat. #A6733) and
   N-Acetyl cysteine (cat. #A9165) and Iodoacetamide (IAA, cat. #A3221)
   were purchased from Sigma Aldrich.

Genome-wide lentiviral shRNA screening

   A human whole-genome library (RHS6083, GE Dharmacon) was transfected
   into the NSCLC cell line H1299 to conduct the genome-wide lentiviral
   shRNA screening^[312]84. The screening procedure was conducted
   according to the manufacturer’s instructions. 1 μmol/L of LY2603618 was
   used for CHK1 inhibition. Genomic DNA extraction, PCR amplification of
   shRNA and PCR product purification were performed according to the
   protocol provide by RHS6083. Purified amplicons from different samples
   were pooled and analyzed with a Bioanalyzer (Agilent) and qPCR and then
   sequenced on an Illumina HiSeq 2500 platform on high output mode.
   TrimGalore was used for trimming and quality evaluation of the
   sequences from the shRNA samples. Sequences that passed the quality
   filters were aligned to the shRNA library sequences provided by
   Dharmacon (GE). The R package DESeq was used to determine whether there
   was differential expression. An initial significance cutoff was applied
   to the DESeq output, and only shRNAs that had an adjusted P < 0.05
   false discovery rate (FDR) were considered.

Immunohistochemistry (IHC)

   A tissue microarray (TMA) with 50 NSCLC samples and their matched
   adjacent normal tissue (ANT) were purchased from TissueArray (slides
   ID: LC10013c #199 and LC10019c #200) and slides were subjected to
   graded alcohol following deparaffinization. Monoclonal Trx1 (1:100;
   cell signaling #2429 clone-C63C6) and TrxR1 (1:50; #15140 clone-D1T3D)
   antibodies were used for IHC. The median immunoreactivity score (IRS)
   was determined from the staining intensity (SI) and percentage of
   positive cells (PP): IRS = SI × PP. An IRS ≥ 1–3 score was classified
   as positive.

Plasmids and lentiviral preparation

   Plasmids pCMV ΔR8.2 (Addgene ID-12263) and pCMV-VSVG (Addgene ID-8454)
   for lentiviral packaging were procured from Addgene. Flag-Trx1 WT
   (Addgene ID-21283); Flag-Trx1 C32S (Addgene ID-21284); Flag-Trx1 C35S
   (Addgene ID-21285) were procured from Addgene. RRM1 expressing plasmids
   (Myc-RRM1 WT; Myc-RRM1 C779S; Myc-RRM1 C787S; Myc-RRM1 C790S; Myc-RRM1
   C779S/C787S; Myc-RRM1 C779S/C790S and Myc-RRM1C779S/C787S/C790S) were
   described previously^[313]68. The validated shRNAs construct of Trx1
   (TRCN0000064278, TRCN0000064279) and TrxR1 (TRCN0000046533,
   TRCN0000046534) were purchased from Sigma-Aldrich and the lentivirus
   were produced using CaCl[2] mediated transfection into 293T cells.
   3’UTR targeting Trx1 shRNAs (3’UTR shTrx1-1; clone ID-V2LHS_275263 and
   shTrx1-2; clone ID- V3LHS_412949 used in Fig. [314]6C–H) were purchased
   from Dharmacon.

Western blot

   Immunoblotting was conducted as previously described^[315]85–[316]87.
   Briefly, total protein was extracted using protein extraction buffer
   (62 mM Tris-HCl pH 6.8; 2% SDS; 10% Glycerol and 10 mM DTT)
   supplemented with protease inhibitor cocktail (Roche #11836153001) and
   phosphatase inhibitors (Sigma-Aldrich #P2850) 1 mM PMSF (Sigma-Aldrich
   #93482). A total of 50 μg protein was separated on denaturing SDS-PAGE
   and transferred on to PVDG membrane. Non-specific biding was blocked
   using 5%BSA in TBST for 2 hrs at RT. Proteins of interest were detected
   using appropriate primary antibodies. Antibodies [(Trx1; #2429
   clone-C63C6; TrxR1; #15140 clone-D1T3D; p-CHK1 (S345); #12302
   clone-D12H3; p-CHK1 (S317); #8191 clone-D7H2; CHK1; #2360 clone- 2G1D5;
   p-ATR(T1989); #30632 clone-D5K8W; ATR; #13934 clone-E1S3S; Histone H3;
   #4499 clone-D1H2; Histone H2AX; #7631 clone-D17A3; RPA32; #2208
   clone-4E4 and RRM1; #8637 clone-D12F12; RRM2; #65939 clone-E7Y9J; E2F1;
   #3742, cleaved PARP1 #5625 clone-D64E10, Cleaved caspase 3; #9661
   clone-5A1E); antibodies were used at 1:1000 dilution] were purchased
   from Cell Signaling Technologies. The anti-p53R2 antibody was purchased
   from Santa Cruze Biotechnology (#sc-137174 clone A-5; 1:1500). The
   anti-β-actin antibody was purchased form Sigma Aldrich cat#A5441.
   Antibodies to detect p-RPA32(S4/8) (cat# A300-245A) and p-RPA32(S33)
   (cat# A300-246A) were purchased from Bethyl Laboratories. The
   anti-γH2AX antibody (cat#05-636; clone-JBW301; 1:2000) was procured
   from EMD Millipore. Anti-p-RPA32 (S33 and S4/8) antibodies were used at
   1:1500 and anti-β-actin was used at 1:5000 dilution. All other primary
   antibodies were use at 1:1000 dilution in 5% BSA in TBST at 4 °C
   overnight with gentle shaking. After incubation membranes were washed 3
   time with 1X TBST for 5 min. The membrane was incubated with relevant
   HRP-tagged secondary antibodies (Anti-rabbit IgG (#7074); Anti-mouse
   IgG (#7076) and Anti-rat IgG (#7077) from Cell signaling technology) at
   1:5000 dilution in 5% BSA at RT for 2 h with gentle shaking. After
   washing with TBST, the signal was detected using SuperSignal^TM West
   Pico PLUS Luminol reagent (Thermo Scientific #34578) on X-ray films.
   Full scan of blots is present in source data and supplementary data
   file.

Cell proliferation and cellular toxicity assay

   NSCLC cell lines were seeded in 96 well plates and treated with the
   desired concentration of inhibitors at the indicated durations. After
   completion of drug treatments, cells were incubated for an additional
   3 h in 6 mg/ml MTT solution in growth medium. Cell viability was
   assessed by absorption measurement of MTT crystals after solubilizing
   into MTT solvent.

Annexin V/PI apoptosis assay

   Apoptosis in cultured cells was measured using an eBioscience™ Annexin
   V Apoptosis Detection Kit (Invitrogen; #Cat- 88-8102-72) according to
   the manufacturer’s protocol.

Xenograft studies and the PDX model

   Athymic male and female mice (Strain code: 553, NCI Frederick) at
   4–6-week-old were used in this study. Animals were bred at The Ohio
   State University (Columbus, OH). All mice were maintained under barrier
   conditions, and the experiments were conducted using protocols and
   conditions approved by the Institutional Animal Care and Use Committee
   of The Ohio State University (Columbus, OH). Xenografts were
   established by injections of H1299 cells (0.5 × 10^6 cells)
   subcutaneously into flanks of the animals. Tumor diameters were
   measured with digital calipers, and the tumor volume in mm^3 was
   calculated using the formula: Volume = (width)^2 × length/2. Once tumor
   volume reached 100–150 mm^3, the mice were treated with vehicle control
   or a CHK1 inhibitor (25 mg/kg of LY2603618) via intraperitoneal
   injection (IP) twice a day for 3 days, followed by 4 days of rest over
   the course of 3 weeks. In case of combined CHK1i and AUR, AUR was given
   via IP at 10 mg/kg for 5 days every week accompanied by the same CHK1i
   regimen. PDX-72 was used to measure the effect of the efficacy of drug
   combination^[317]49. PDX72 was established by transplanting freshly
   removed NSCLC tumor tissue (KRAS G12C mutation) into NSG mice
   subcutaneously^[318]49. Anesthetized NSG (NOD scid gamma (NSG™)) mice
   were inoculated with small pieces of PDX-72 tumors. Animals were
   treated with buprenorphine at 0.1 mg/kg for 3 days for pain management.
   All mice were maintained under barrier conditions, and the experiments
   were conducted using protocols and conditions approved by the
   Institutional Animal Care and Use Committee of The Ohio State
   University (Columbus, OH). Tumor size ≈2000 mm^3 considered as the end
   point for all experiments and no tumor exceeded the size in this study.
   At the end point, animals were sacrificed, tumors were excised, weighed
   and photographed.

Immunofluorescence staining and intensity profiling of individual cells

   Immunofluorescence assays were performed as described
   previously^[319]85,[320]87,[321]88. Images were captured using confocal
   microscope at 20× and 63×. Higher magnification images were used to
   count the number of foci/cells. 20× were used to measure the intensity
   of staining in each cell using ImageJ. Briefly, images were first
   converted into 8-bit grayscale and the threshold was set to visualize
   the DAPI area and nuclear intensity of RPA32, p-RPA32 (S33) and γH2A
   was measured. Intensity of each nucleus (at least 1500–2000
   cells/group) was recorded, and the data was transferred into GraphPad
   prism for statistical analysis.

Neutral comet assay

   The Neutral Comet Assay was performed using the Comet Assay kit
   (4250-050-K) from Trevigen, following the manufacturer’s instructions.
   The length of comet tail was measured using Image J.

Measurement of ROS by DCFDA

   Intracellular ROS was measured by staining of cells with DCFDA under
   the desired treatment conditions. Briefly, cells were grown in
   multi-well plates and treated as indicated and harvested through
   trypsinization. Cells were washed in cold 1× PBS and stained with 10 µM
   DCFDA in the presence or absence of NAC for 30 min at 37 °C. DCFDA was
   then washed out and cells were resuspended in 1× PBS before detection
   of fluorescent signal. ROS assay flow cytometric histograms were
   generated and analyzed on an BD Fortessa instrument using FACSDIVA™
   software (BD Biosciences). DCFDA measurement was also assessed via
   microplate reader by measuring the DCFDA fluorescence at Ex/Em:
   ∼492–495/517–527 nm.

Cell cycle analysis

   Cells were grown for 2–3 generations in multiwall plates and
   logarithmically growing cells were harvested by trypsinization and
   fixed in 70% methanol at −20 °C overnight. After washing in 1× PSB
   twice cells were resuspended in 1× PSB containing 50 µg/ml PI &
   100 µg/ml RNase A and incubated at 37 °C for 1 h before analyzing on an
   BD Fortessa instrument using FACSDIVA™ software (BD Biosciences).

BrdU incorporation assay

   BrdU incorporation assay was performed using PhaseFlow^TM FITC BrdU kit
   according to the manufacturer’s instructions. Briefly, cells were
   pulsed with 10 μM BrdU or 30 min and stained with anti-BrdU antibody
   following DNA digestion. Cells were then stained with nuclear stain
   7-AAD, and data was recorded using an BD Fortessa instrument.

DNA fiber assay

   DNA fiber assays were performed as previously
   described^[322]84,[323]85,[324]88. Briefly, cells were pulsed with
   50 µM IdU for the indicated durations. IdU was washed out with
   pre-warmed 1× PBS and 200 µM CIdU was then added for desired duration
   with or without inhibitor or dNTPs. Cells were then harvested and lysed
   on a slide and DNA fibers were stretched via a drop rolling method.
   Fibers were dried and fixed in 1:3 acetomethenol solution for 10 mis.
   Fibers were denatured in 2.5 mM HCl and rehydrated in 70% ethanol
   followed by nonspecific blocking in 2% BSA + 1% goat serum for 30 min
   at RT. BrdU antibody (#347580, 1:20, BD Biosciences) and CldU antibody
   (ab6326, 1:200, Abcam) were incubated overnight at 4 °C. Next day,
   fibers were washed and appropriate 2° antibodies (Goat anti mouse Alexa
   fluor 594; Cat# A-11006 and Goat anti Rat Alexa Fluor 488; Cat# A-11006
   at 1:400 dilution) were incubated for 1 h at RT in dark. Fibers were
   mounted into antifade mounting medium. Fibers were imaged at 40× and
   fibers length was measured using Image J. Aa total of 100–200 fibers
   per condition was counted and experiment was repeated 3 time
   independently in double blind fashion.

BrdU incorporation assay

   BrdU incorporation assay was performed using PhaseFlow™ BrdU kit from
   Bio Legend (cat. 370704) according to manufacturer’s protocol.

dNTPs measurement

   The dNTP measurement was conducted as previously described^[325]89.
   Briefly, cell pellets (5 × 10^4 ∼ 2 × 10^6 cells) were washed twice
   with 1× DPBS (Mediatech, VA), and resuspended in 100 μl of ice-cold 60%
   methanol. Samples were vortexed vigorously to lyse the cells and then
   heated at 95 °C for 3 min, prior to centrifugation at 12,000 × g for
   30 s. The supernatants were collected and completely dried under
   vacuum, using a SpeedVac (Savant, NY) with medium heat. The dried
   pellets were subsequently resuspended in dNTP buffer (50 mM Tris-HCl,
   pH 8.0 and 10 mM MgCl2; 100 μl for 1 × 10^6 HeLa cells, and 10 μl for
   1 × 10^6 primary cells) and usually 1~2 μl of the extracted dNTP
   samples were used for each 20 μl single nucleotide incorporation
   reaction. The proper dilutions of the dNTP samples were prepared for
   the assay to make the primer extension values lie within the linear
   ranges of the dNTP incorporation (2~32% primer extension). The
   extracted dNTP samples were stored at −70 until used. Several different
   volumes of the extracted dNTP samples were also used to confirm the
   linearity of the primer extension. In addition, the dNTP samples were
   prepared from different cell numbers, depending on the recovery
   efficiency of the primary cells from each blood sample (see below).
   However, the dNTP content of each cell type was normalized by
   pmole/1 × 10^6.

Redox western blot

   Redox western blots were  performed as previously described^[326]47.
   Briefly, cells were harvested and pelleted by addition of 100% TCA to
   final concentration of 10% and pellet was washed twice in acetone.
   Alkylation of free thiols was achieved by resuspending the TCA
   extracted pellet into alkylation solution containing 75 mM
   iodoacetamide, 1% SDS, 100 mM Tris·HCl (pH 8), 1 mM EDTA following
   incubation at 25 °C for 15 min. Further, alkylated sample was
   dephosphorylated using calf intestinal phosphatase (CIP) for 1 h at
   37 °C. Dephosphorylated-alkylated protein samples were subjected to
   non-denaturing SDS-PAGE (-DTT; no boiling) or denaturing (+DTT and
   boiling) as indicated.

Synergy score

   Synergy scores were generated using online SynergyFinder 2.0 web
   application.

Organoid culture

   PDX-72 untreated tumors were dissociated using a human tumor
   dissociation kit (Miltenyi Biotech; Cat# 130-095-929) following the
   manufacturer’s protocol. Dissociated tissue was filtered via a 70 μm
   filter and subjected to isolation of human tumor cells using a human
   cancer cell isolation kit (Miltenyi Biotech; Cat# 130-108-339).
   Isolated tumor cells were then subjected to depletion of the mouse cell
   contamination using a Mouse cell depletion kit (Miltenyi Biotech; Cat#
   130-104-694). The purity of the human cancer cells was verified using
   anti-human-EpCAM (CD326) from Miltenyi Biotech (Cat3130-111-000; clone
   REA764; 1:100 dilution). These tumor cells were then imbedded in
   matrigel (Corning; Cat#356231) and cultured in organoid medium
   according to a previously published protocol^[327]90. The diameter of
   the organoids was measured with Image J. To detect the degree of cell
   death in organoids, organoid medium was supplemented with Hoechst
   (10 μg/ml) (Invitrogen, Cat#[328]H21486), Calcein AM (5 μM)
   (Invitrogen; cat#C3100MP) and PI (10 μg/ml) (Invitrogen; Cat#P1304MP)
   and incubated for 1 h under the growth condition. Organoids were then
   imaged on an Echo Revolve fluorescence microscope. The PI intensity of
   each organoid was measured with Image J.

Statistical analysis and reproducibility

   Results are depicted as mean ± SD unless indicated otherwise. GraphPad
   Prism 9.0 software (La Jolla, CA) was used for statistical analysis as
   described within figure legends. P-value ≤ 0.05 was considered
   statistically significant. DNA fiber assay and dNTP assays were
   performed in double blinded fashion. Animals were randomly divided in
   different groups before the initiation of the drug injections. Animals
   with visible occurrence of ulcer were excluded from the studies. All
   experiments were repeated at least 3 times independently with similar
   results unless indicated otherwise. One way ANOVA was used to compare
   multiple groups without any adjustment. Statistical analysis details
   are provided in the end of each figure legends to save the space; p
   values are indicated with * (*p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001;
   ****p ≤ 0.0001; ns: non-significant].

Reporting summary

   Further information on research design is available in the [329]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [330]Supplementary Information^ (6.1MB, pdf)
   [331]Reporting Summary^ (194.3KB, pdf)
   [332]Peer Review File^ (414.6KB, pdf)

Source data

   [333]Source Data^ (10.6MB, xlsx)

Acknowledgements