Inhibition of Thioredoxin/Thioredoxin Reductase Induces Synthetic Lethality in Lung Cancers with Compromised Glutathione Homeostasis

Glutathione (GSH)/GSH reductase (GSR) and thioredoxin (TXN)/thioredoxin reductase (TXNRD) are two parallel, compensating thiol-dependent antioxidant pathways that regulate and maintain cellular thiol redox homeostasis and protein dithiol/disulfide balance (1, 2). Both GSH and TXN catalyze the reduction of disulfide bonds in proteins to dithiols, which is critical for maintaining protein functions (Fig. 1A). Oxidized GSH (GSSG) and TXN generated during this catalytic process are converted to their reduced states by GSR and TXNRD, respectively, both using NADPH generated in the pentose phosphate cycle as the hydrogen donor. Functional deficiency in one of these systems will render cells dependent on compensation by the other system for survival. In GSR-deficient yeasts (3) and Drosophila (4), the TXN/TXNRD pathway is required for the reduction of GSSG to GSH. Similarly, GSH acts as a backup for human TXNRD1 to reduce TXN, thereby preventing cell death induced by gold compound–mediated TXNRD1 inhibition (5). Combined inhibition of the GSH and TXN pathways has led to synergistic cancer cell death (6, 7). Because synthetic lethality interactions occur significantly more frequently among genes involved in similar biological processes, such as those in parallel or compensating pathways (8), we hypothesized that loss and/or reduced expression of enzymes required for GSH homeostasis, including GSR and glutamate-cysteine ligase (the first rate-limiting enzyme of GSH synthesis), in a subgroup of cancers predisposes them to vulnerability to treatment with TXNRD inhibitors, providing a mechanism-based synthetic lethality therapy for these cancers.

The human GSR gene is located on chromosome 8p12, a region frequently lost in lung cancers (9, 10) and other types of cancers, including breast (11), head and neck (12), prostate (13), bladder (14), and gastric/colorectal (10) cancers and lymphoma (15). Accumulating evidence has shown that loss of chromosome 8p increases tumor invasiveness and metastasis potential (16–18) and confers resistance to chemotherapy and poor survival in patients with cancer (13, 18). Targeted deletion of chromosome 8p yielded altered lipid metabolism and increased autophagy, which trigger invasiveness and confer tumor growth under stress conditions (18). Unfortunately, therapeutic strategies for cancers with chromosome 8p loss have yet to be developed. Loss of the GSR gene may offer a unique opportunity for treatment of cancer in a subgroup of patients whose cancer has chromosome 8p aberrations.

To explore the concept of targeting cancers in which expression of enzymes required for GSH homeostasis is reduced or lost, we searched The Cancer Genome Atlas (TCGA) database for genetic alterations and found that deep GSR deletion was detected in 6% of lung adenocarcinomas, 3.5% of squamous cell lung cancers, and 4% to 9% of prostate, breast, liver, ovarian, and colon cancers, suggesting that loss of the GSR gene occurs in a subset of human cancers. We then tested whether loss of GSR in cancer cells will result in their vulnerability to auranofin, a TXNRD inhibitor (19). Our results show that reduced expression of enzymes required for GSH homeostasis sensitized cancer cells to TXNRD inhibitor auranofin.

Human non–small cell lung cancer (NSCLC) cell lines, including A549, H2023, H1993, and DFC1024, were obtained from Dr. John Minna's laboratory (University of Texas Southwestern Medical Center) and propagated in monolayer culture in DMEM supplemented with 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. All cells were maintained in a humidified atmosphere containing 5% CO₂ at 37°C. Cell lines were tested for mycoplasma with the LookOut Mycoplasma PCR Detection Kit (Sigma-Aldrich) and authenticated per routine at our institution's Characterized Cell Line Core using a short tandem repeat DNA fingerprint assay with a PowerPlex 16 HS kit (Promega).

A549–GSR-knockout cells were generated by using the CRISPR/Cas9 genome editing system. Briefly, GSR-specific guide RNA (gRNA) expression vectors were generated as described (20). LentiCas9-Blast (plasmid 52962) and gRNA cloning vector lentiGuide-Puro (plasmid 52963) were obtained from Addgene. The CRISPR guide sequences targeting GSR were GATCATGCATGAATTCAGAG (gRNA1) and GTGAGGGTAAATTCAATTGG (gRNA2). A549 cells were cotransfected with lentiCas9-Blast and gRNA expression vectors. The transfected cells were selected using puromycin (2 μg/mL) for 48 hours. Isolated single colonies were analyzed by Western blot. The colonies without GSR expression were subjected to detection of genomic mutations by DNA sequencing.

For siRNA-mediated gene knockdown, scramble control siRNA or siRNA targeting the GSR or the glutamate-cysteine ligase catalytic subunit (GCLC) gene (all, Sigma-Aldrich) was transfected into cells with oligofectamine (Sigma-Aldrich) according to the manufacturer's instructions. Plasmids for lentiviral vectors expressing cDNA or short-hairpin RNA (shRNA) used in this study were obtained from Open Biosystems through our institution's shRNA and ORFeome Core facility. Lentiviral vectors were packaged in 293 cells after cotransfection with lentiviral packaging plasmids by using Fugene 6 as directed by the manufacturer (Promega). The medium from the transfected 293 cells was filtered through a sterile 0.22-μm filter and used to infect target cells in the presence of 8 μg/mL polybrene. After selection with puromycin (1–2 μg/mL) or blasticidin (2–10 μg/mL), the cells were pooled together for subsequent studies. The proportion of stably transfected cells (expressing GFP or RFP in the lentiviral backbone) was usually >85% after selection.

The cell viability was determined using the sulforhodamine B assay as previously described (21) or using a CellTiter-Glo assay kit (Promega) according to the manufacturer's instructions. Cell viability following siRNA transfection was assayed as follows: cells were seeded in 96-well plates at 2,000 cells/per well for 24 hours. siRNA transfection was performed with 50 μL/well siRNA transfection mix in OptiMEM containing 0.2 μL of oligofectamine and 200 nmol/L of siRNA; after 4-hour incubation at 37°C, another 50 μL of regular culture medium containing 20% FBS was added to each well. Auranofin was added 12 hours after siRNA transfection, and cell viability was assayed 72 hours later. The percentage of viable cells was determined relative to the viability of the controls, which was set at 100%. Each experiment was performed in triplicate and repeated 3 times. Number of apoptotic cells (sub-G₁) were measured using fluorescence-activated cell sorting after Annexin V/propidium iodide staining.

Antibodies for detecting human GSR, GCLC, glutamate-cysteine ligase modifier subunit (GCLM), TNXRD1, NAD(P)H quinone dehydrogenase 1 (NQO1), and beta-actin were obtained from Santa Cruz Biotechnology or Abcam. Cells were harvested after stable selection or at 48 hours after siRNA transfection. In brief, cells were washed in PBS solution, collected, and then subjected to lysis in RIPA buffer containing proteinase inhibitor cocktail and phosphatase inhibitor cocktail (Roche). The lysates were subjected to centrifugation at 10,000 × g at 4°C for 10 minutes and the supernatants removed. The total protein (30 μg) was fractioned by 10% SDS-PAGE and electrophoretically transferred to Immobilon-FL PVDF membranes. The membrane was blocked with blocking buffer (Li-Cor) at room temperature for 1 hour, and then incubated with goat anti-human P53 IgG (1:1,000) for 2 hours or with the primary antibody at 4°C overnight. After washing with PBS containing 0.05% Tween 2 (PBST), the membrane was incubated with IRDye infrared secondary antibody (Li-Cor) for 1 hour at room temperature. The membrane was washed with PBST again, and protein expression was detected with the Odyssey Infrared Imaging System.

Animals were maintained and experiments carried out in accordance with the Guidelines for the Care and Use of Laboratory Animals (National Institutes of Health Publication 85-23) and the institutional guidelines of MD Anderson Cancer Center. Patient-derived xenografts (PDX) were established from clinical surgically resected specimens or from pleural fluid as we previously reported (22), except that NOD-SCID mice with null mutations of the gene encoding the IL2 receptor γ (NSG) obtained from The Jackson Laboratory were used instead of regular NOD-SCID for generation and passaging of PDXs. This study is approved by Institutional Animal Care and Use Committee and Institutional Review Board at The University of Texas MD Anderson Cancer Center. All clinical samples were collected with informed consent from the patients. Each PDX was subcutaneously inoculated into the dorsal flank of nude mice. When the tumors had grown to 3 to 5 mm in diameter, the mice were assigned randomly into two groups (n = 5/group) and treated with intraperitoneal administration of auranofin (10 mg/kg) in a solvent comprising 2% dimethylsulfoxide, 10% ethanol, and 5% polyethylene glycol 400 or the solvent without auranofin as controls. Treatment was administered daily for the first week, then twice a week for the duration of the experiment. Tumor growth was monitored every 2 to 3 days and was normalized to tumor volume at the beginning of treatment, which was set to 1. Subcutaneous tumors were measured with calipers, and tumor volume was calculated according to the formula V = ab²/2, where a is the largest diameter and b is the smallest. The experiment was ended, and the mice were euthanized when the largest tumors reached 15 mm in diameter. This in vivo experiment was performed 3 times.

The half-maximal effective dose (IC₅₀) values were calculated with the best-fit dose-response model selected by the residual standard error using the drexplorer package (23). The mRNA levels determined by microarray assay were available for 76 of 129 NSCLC cell lines tested for auranofin sensitivity. Association between gene expression and IC₅₀ value was assessed using the Spearman rank correlation. All statistical analyses were performed using the R software. P values were obtained using computed t statistics. The significance of the in vivo animal study data was determined by using the Mann–Whitney U test for two treatment groups (two tailed). A P < 0.05 was considered statistically significant.

To test the concept of targeting cancers with compromised GSH homeostasis, we searched the TCGA database for alterations of the GSR, GCLC, and GCLM genes using the cBioPortal website (http://cBioPortal.org). We found that deep GSR deletion was detected in 6% of lung adenocarcinomas, 3.5% of squamous cell lung cancers, and 4% to 9% of prostate, breast, liver, ovarian, and colon cancers (Fig. 1B and C). In both lung adenocarcinoma and squamous cell cancer, most deletions of GSR are concurrent with deletions of TUSC3, a gene located in chromosome 8p22, indicating coexistence of GSR deletion and chromosome 8p loss. Deletions of the GCLC and GCLM genes, which encode the two protein subunits of glutamate-cysteine ligase, were detected occasionally in lung adenocarcinoma or squamous cell cancer (about 1%). Deletion of GCLM was relatively more frequent (about 2%–3%) in prostate and pancreatic cancers. This result indicates that there is a subset of human cancers that are likely compromised in GSH homeostasis. Deletion of the GSR gene in a lung adenocarcinoma did not overlap significantly with EGFR mutation or amplification but did overlap TP53 and KRAS gene mutations, suggesting that available targeted agents for cancer are not effective in cancers with GSR, GCLC, or GCLM mutation.

To test whether loss of enzymes required for GSH homeostasis sensitizes cancer cells to TXNRD inhibition, we knocked out the GSR gene in auranofin-resistant A549 lung cancer cells using CRISPR/Cas9 technology with two GSR-specific gRNAs (Fig. 2A). We identified five A549 clones with loss of GSR expression by Western blot analysis and verified genomic deletions in both alleles of the GSR gene using PCR-mediated genomic sequencing of the target sequence in one of the clones (Fig. 2B). A dose-dependent cell viability assay on two of the five clones demonstrated that knockout of the GSR gene led to a dramatic increase in the sensitivity of A549 cells to auranofin (Fig. 2C). According to Annexin V/propidium iodide staining, the rate of auranofin-induced apoptosis was markedly higher in A549/GSR^−/− cells (16.4%) than in A549 cells wild-type for GSR (3.8%; Fig. 2D–F). A similar dose-dependent cytotoxic effect was observed for another TXNRD inhibitor aurothioglucose (5) in A549/ GSR^−/− cells but not in the parental A549 cells (Supplementary Fig. S1).

We observed similar results when we knocked down the GSR and/or GCLC gene in H2023 and DFC1024 NSCLC cells via transfection of either siRNA or lentivirus-mediated shRNA. Knockdown of either GSR or GCLC sensitized these cells to auranofin (Fig. 3A–H). In contrast, enforced expression of the GSR gene in H1993 cells increased viable cells after treatment with auranofin at the doses of 0.16 to 0.62 μmol/L (Supplementary Fig. S2). These results provided proof-of concept evidence that cancer cells with loss of or reduced expression of enzymes required for GSH homeostasis can be selectively eliminated by treatment with a TXNRD antagonist.

Our previous study showed that auranofin's anticancer activity was inversely correlated with TXNRD1 expression and enzymatic activity in 10 NSCLC cell lines (24). To further investigate potential predictive biomarkers of NSCLC sensitivity to auranofin, dose responses to auranofin were examined in 129 NSCLC cell lines. The cells were treated with different concentrations of auranofin ranging from 78 nmol/L to 5 μmol/L. Dose-dependent cell viability was determined using a CellTiter-Glo assay kit. These cells had varying sensitivity to auranofin. Twenty-eight of the 129 cell lines had an IC₅₀ less than 1 μmol/L (Fig. 4A; Supplementary Table S1). Illumina mRNA expression profiling data on 11,544 unique genes available for 76 of these 129 NSCLC cell lines were analyzed by the Spearman rank correlation test to determine the correlation between auranofin response and mRNA levels in these cell lines. The result showed that the levels of GSR (r = 0.43, P = 7 × 10⁻⁵), GCLC (r = 0.44, P = 3.7 × 10⁻⁵), TXNRD1 (r = 0.44, P = 4 × 10⁻⁵), NQO1 (r = 0.52, P = 8.5 × 10⁻⁷), and glucose-6-phosphate dehydrogenase (G6PD; r = 0.44, P = 4.1 × 10⁻⁵) gene expression were all significantly correlated with auranofin activity in these cell lines (Supplementary Table S2). In addition, Western blot analysis of the expression of GSR, GCLC, GCLM, NQO1, and TXNRD1 in 22 of these NSCLC cell lines (12 auranofin-sensitive and 10 auranofin-resistant) demonstrated dramatic differences in the expression of GSR, GCLC, and NQO1 proteins in these 22 cell lines (Fig. 4B and C). The auranofin-sensitive cancer cell lines expressed low levels of these three genes, whereas the auranofin-resistant cell lines expressed high levels of these genes, suggesting that the levels of these three genes can be used as a signature to distinguish auranofin-sensitive and -resistant NSCLCs.

Notably, expressions of the GSR and GCLC genes are concurrently low in the auranofin-sensitive NSCLC cell lines. We searched the TCGA database to see if the expression of GSR and GCLC genes is correlated in human primary lung cancers. We found that the GSR gene was coexpressed with the SRXN1, TXNRD1, GCLM, G6PD, NQO1, and GCLC genes in both lung adenocarcinomas and squamous cell lung cancers, with Pearson r values ranging from 0.76 to 0.50 (Supplementary Table S3), indicating that expression of these genes occurs concurrently in primary lung cancers.

Some recent studies demonstrated that PDXs have response rates similar to those of primary tumors for several anticancer drugs (25, 26). We established a panel of PDXs from surgical or pleural fluid NSCLC specimens. Our previous study, which carried out targeted sequencing of 200 cancer-related genes in 23 NSCLC PDXs, revealed that 93% of the mutations detected in primary tumors also occurred in their corresponding PDXs, indicating that PDXs can recapitulate the mutations in primary tumors (22).

We examined GSR and GCLC expression in 19 NSCLC PDXs (F1–F3 generations) using Western blot analysis. The results demonstrated that six of the PDXs had low GSR and GCLC expression (Fig. 5A). In two of these PDXs, Western blot analysis revealed that the levels of GSR and GCLC expression were consistent in the F1 and F3 generations. We then performed an experiment to determine the in vivo activity of auranofin in two PDXs (PDX-2 and PDX-12) with relative low levels of GSR and GCLC expression, and in one PDX model with relatively high GSR and GCLC expression (PDX-1). The results demonstrated that treatment with auranofin resulted in significant suppression of the growth of both GSR/GCLC-low PDX models in vivo (P < 0.05 at 2 weeks after the treatment began), but not in the GSR/GCLC-high model (Fig. 5B–D). Treatment with auranofin led to 65% and 68% inhibition of PDX-2 and PDX-12 tumor growth, respectively, compared with that in control mice. Weight loss was not detectable in PDX-12 and PDX-1 mice and not significant in PDX-2 mice. Together, these findings indicate that treatment of NSCLC PDX-bearing mice with auranofin was effective in GSR/GCLC-low tumors and well tolerated.

Our results support the hypothesis that cancers with loss and/or reduced expression of enzymes required for GSH homeostasis, including GSR and GCLC, can be targeted using a synthetic lethality strategy with TXNRD inhibitors. These results were in consistence with reports by others that combined inhibition of GSH and TXN antioxidant pathways leads to a synergistic cancer cell death in vitro and in vivo (6, 27). Synthetic lethality therapy has been exploited for treatment of cancers with mutations in the KRAS (28, 29), BRCA (30, 31), and TP53 (32) genes. The success of using PARP inhibitors in treatment of BRCA-mutant cancers (33, 34) has demonstrated the clinical feasibility of targeting cancers via synthetic lethality. Because the GSR gene is located in chromosome 8p, a region that is frequently deleted in several types of human cancers, this strategy may be applied to cancers whose chromosome 8p deletion involves the GSR gene.

Chromosome 8p deletions detected by analysis for LOH have been frequently reported in cancers of lung, colon, breast, prostate, or liver. Chromosome fragile sites located at 8p22 (FRA8B) and at 8p12 (within the NRG1 gene) have been proposed to account for the frequent 8p aberrations observed in cancers (35). The GSR gene is located between these sites, suggesting that its loss might be common in cancers with chromosome 8p deletions. Efforts have been made to identify possible tumor-suppressor genes within the 8p region. A study that used shRNA-mediated knockdown of the mouse orthologs of the 21 genes frequently deleted on human chromosome 8p as detected in human cancers found that multiple genes on chromosome 8p can cooperatively inhibit tumorigenesis in a murine hepatocellular carcinoma model, and that their cosuppression can synergistically promote tumor growth (36). A recent study of 8p LOH established in human mammary epithelial cell line MCF10A by using TALEN-based genomic engineering revealed that 8p LOH triggers invasiveness and resistance to chemotherapeutic drugs and confers tumor growth under stress conditions due to increased autophagy (18), providing molecular insights into the poorer clinical outcomes of patients with cancer with a chromosome 8p deletion.

The genetic interaction between GSH/GSR and TXN/TXNRD pathways may provide a therapeutic opportunity for cancers with compromised GSH homeostasis or with a chromosome 8p deletion that includes the GSR gene. Intriguingly, gene expression data from our lung cancer cell lines and from TCGA database indicate that the expression of the GSR, GCLC, and TXNRD1 genes is highly correlated in both lung adenocarcinomas and squamous cell lung cancers. The underlying mechanism is not yet clear. Nevertheless, the TCGA data showed that overexpression of the GSR, GCLC, and TXNRD1 were highly enriched in KEAP1-mutant lung adenocarcinoma (P = 1.2 × 10⁻⁸, 1.0 × 10⁻⁶, and 1.4 × 10⁻⁹, respectively) and in NFE2L2-mutant lung squamous cell cancer (P = 1.5 × 10⁻³, 4.7 × 10⁻⁴, and 6.1 × 10⁻⁹, respectively), supporting the roles of the KEAP1/NRF2 pathway in regulation of the expression of these oxidative stress–associated genes (37). It is possible that, unlike mutations in KEAP1/NRF2 pathways, low expression of GSR alone is not sufficient to trigger TXNRD1 overexpression.

In conclusion, we report that cancer cells with reduced or lost expression of the GSR gene have increased sensitivity to the TXNRD1 inhibitor auranofin, a gold complex that has been used to treat rheumatoid arthritis since the 1980s (38). The mechanism of auranofin-mediated immunomodulation in patients with rheumatoid arthritis is not entirely clear, although the primary molecular target of auranofin and other gold-containing drugs is documented to be TXNRD (19), the only enzyme that catalyzes the reduction of TXN with electrons from NADPH (39). The median inhibitory concentration (IC₅₀) of auranofin for inhibiting TXNRD is about 9 nmol/L, but it is 15 μmol/L for the closely related enzyme GSR (40). Long-term use of auranofin is well tolerated in both juvenile and elderly patients (41, 42). The discovery of auranofin's activity against Entamoeba histolytica infection also led to quick regulatory approval of auranofin in the United States for the treatment of amebiasis (43). Our previous study demonstrated that auranofin had single-agent activity in vitro and in vivo in a subset of lung cancer cells via inhibition of multiple key nodes in the PI3K/AKT/mTOR axis (24). It is not yet clear whether the similar mechanism occurs in vivo. Studies by others have shown that auranofin effectively induces apoptosis in chemoresistant cancer cells (44, 45). Auranofin and sirolimus combination therapy is currently under a phase II trial for treatment of ovarian cancer (trial registration NCT03456700). The correlations between auranofin sensitivity and expression levels of the GSR, GCLC, and NQO1 genes suggest that these three genes might be used as a predictive signature for identifying the patient subgroups whose tumors may respond to the treatment with a TXNRD inhibitor.

No potential conflicts of interest were disclosed.

Conception and design: X. Zhang, J.D. Minna, B. Fang

Development of methodology: X. Yan, X. Zhang, L. Wang, S. Wu, L. Li

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Yan, X. Zhang, L. Wang, R. Zhang, S. Wu, Q.H. Meng, V.B. Jensen, J.D. Minna, J.A. Roth, S.G. Swisher, J.V. Heymach

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Yan, X. Zhang, L. Wang, P. Tong, J. Wang, Q.H. Meng, L. Girard, B. Fang

Writing, review, and/or revision of the manuscript: X. Yan, X. Zhang, L. Li, Q.H. Meng, J.D. Minna, J.A. Roth, S.G. Swisher, J.V. Heymach, B. Fang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Yan, X. Zhang, R. Zhang, Q.H. Meng, J.A. Roth, B. Fang

Study supervision: B. Fang

We would like to thank Uma Giri for technical assistance and Kathryn Hale and the Department of Scientific Publications at The University of Texas MD Anderson Cancer Center for editorial review of the article.

This work was supported in part by the NIH through grants R01CA190628, Specialized Program of Research Excellence (SPORE) grant CA070907, The University of Texas PDX Development and Trial Center grant U54CA224065, and The University of Texas MD Anderson Cancer Center support grant P30CA016672 (used the Flow Cytometry and Cellular Imaging Facility, Gene Editing/Cellular Model Core Facility, shRNA and ORFeome facility, and Research Animal Support Facility), and by endowed funds to The University of Texas MD Anderson Cancer Center, including the Lung Cancer Moon Shot Program and the Sister Institution Network Fund.

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