Uracil misincorporation into DNA is a consequence of pemetrexed inhibition of thymidylate synthase. The base excision repair (BER) enzyme uracil–DNA glycosylase (UNG) is the major glycosylase responsible for removal of misincorporated uracil. We previously illustrated hypersensitivity to pemetrexed in UNG−/− human colon cancer cells. Here, we examined the relationship between UNG expression and pemetrexed sensitivity in human lung cancer. We observed a spectrum of UNG expression in human lung cancer cells. Higher levels of UNG are associated with pemetrexed resistance and are present in cell lines derived from pemetrexed-resistant histologic subtypes (small cell and squamous cell carcinoma). Acute pemetrexed exposure induces UNG protein and mRNA, consistent with upregulation of uracil–DNA repair machinery. Chronic exposure of H1299 adenocarcinoma cells to increasing pemetrexed concentrations established drug-resistant sublines. Significant induction of UNG protein confirmed upregulation of BER as a feature of acquired pemetrexed resistance. Cotreatment with the BER inhibitor methoxyamine overrides pemetrexed resistance in chronically exposed cells, underscoring the use of BER-directed therapeutics to offset acquired drug resistance. Expression of UNG-directed siRNA and shRNA enhanced sensitivity in A549 and H1975 cells, and in drug-resistant sublines, confirming that UNG upregulation is protective. In human lung cancer, UNG deficiency is associated with pemetrexed-induced retention of uracil in DNA that destabilizes DNA replication forks resulting in DNA double-strand breaks and cell death. Thus, in experimental models, UNG is a critical mediator of pemetrexed sensitivity that warrants evaluation to determine clinical value. Mol Cancer Ther; 12(10); 2248–60. ©2013 AACR.
Lung cancer remains a leading cause of mortality worldwide. More than 80% of lung cancer cases are non–small cell lung cancer (NSCLC) and of newly diagnosed patients, 40% have advanced disease (1). Chemotherapy is a cornerstone therapy for patients with advanced and recurrent lung malignancy (2).
Pemetrexed, a multitarget antifolate, has proven clinical activity against nonsquamous NSCLC. It is currently approved as single agent second-line therapy for patients with advanced NSCLC (3) and as first-line therapy in combination with cisplatin for patients with advanced NSCLC (4). Pemetrexed maintenance therapy results in statistically significant progression-free and overall survival compared with placebo (5). Still, despite initial responses, patients receiving pemetrexed-based chemotherapy ultimately progress highlighting a need for novel strategies to enhance pemetrexed efficacy.
Pemetrexed inhibits several key enzymes in folate-dependent metabolism including thymidylate synthase (TYMS), dihydrofolate reductase (DHFR), and glycinamide ribonucleotide formyl transferase (GARFT) causing decreased nucleotide synthesis (6). Importantly, pemetrexed metabolites are most active against TYMS, the primary target for pemetrexed cytotoxicity (7). Inhibition of TYMS limits reductive methylation of deoxuridine nucleotides (dUMP) to form deoxythymidine nucleotides (dTMP). As a result, aberrantly large pools of dUMP accumulate, are phosphorylated to dUTP and are used in DNA synthesis in place of dTTP (8). The misincorporated genomic uracil that results is a substrate for uracil–DNA glycosylase (UNG)-initiated base excision repair (BER; ref. 3). Although uracil accumulation is a well documented sequelae of cellular exposure to pemetrexed and other TYMS inhibitors (9), the impact of uracil misincorporation on genomic stability and cellular survival remains undefined.
Pemetrexed resistance has been evaluated in lung and other cancer cell types (10–13). In NSCLC cell lines, elevated expression of TYMS, DHFR, and GARFT corresponds to pemetrexed resistance (13–15). UNG activity, however, is seldom cited as a mechanism of antifolate resistance, despite the apparent toxicity of uracil–DNA in some human (16, 17) and nonmammalian model systems (18, 19).
Recently, we reported that DLD1 human colon cancer cells lacking UNG are hypersensitive to pemetrexed-induced uracil accumulation resulting in cell-cycle arrest, DNA double-strand break (DSB) formation, and apoptosis (20). As pemetrexed is primarily used in the treatment of lung cancer and is limited by a response rate of 30% to 40% with no long-term sustained responses, we evaluated the relationship between UNG expression and pemetrexed response in human lung cancer cell lines. Gene expression data in cell lines and primary human lung tumor tissue samples suggest a spectrum of UNG expression in lung cancer specimen that is significantly correlated with pemetrexed response. On the basis of the evidence of DNA replication fork instability in the context of deficient uracil excision, we propose a novel role for misincorporated uracil as a genotoxic lesion that contributes to antifolate-induced DSB formation and cell death. Induction of UNG in response to acute and chronic pemetrexed exposure also suggests UNG activity limits pemetrexed cytotoxicity. Differential UNG expression among lung cancer histologic subtypes motivates the investigation of UNG as a clinical predictive marker for pemetrexed response. The correlation between UNG expression and pemetrexed sensitivity in experimental models justifies targeting UNG to enhance pemetrexed anticancer activity NSCLC.
Materials and Methods
Cell lines and reagents
Pemetrexed was purchased from LC Laboratories. Thymidine, 5-fluorouracil, cisplatin, methoxyamine-HCL, and raltitrexed were purchased from Sigma Aldrich. Temozolomide was purchased from O-Chem, Inc. All cell lines were obtained from American Type Culture Collection (ATCC) and expanded upon delivery into numerous vials of low-passage cells for cryopreservation. Cells were passaged for up to 3 months. Cell line characterization by ATCC is conducted through short tandem repeat typing. Reauthentication was not conducted. Adherent cells were maintained in complete Dulbecco's modified Eagle medium (10% FBS, 2 mmol/L l-glutamine) and suspension cells were maintained in complete RPMI-1640 (10% FBS, 2 mmol/L l-glutamine) at 37°C in a 5% CO2 incubator.
Propidium iodine (PI) staining of methanol fixed cells was for cell-cycle determinations. Where indicated, fluorescein isothiocynate (FITC)-labeled proliferating cell nuclear antigen (PCNA) antibody (PCNA–FITC, Abcam) was added for PCNA detection. Uni-parameter (PI-only) and dual-parameter (PI + PCNA–FITC) analysis was conducted on a Coulter flow cytometer (EPICS-XL-MCL). Cell-cycle histograms (PI) and PCNA dot plots (PCNA–FITC) were deconvoluted from ≥20,000 events using FlowJo software.
Protein extracts (25 μg) were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane (Millipore). Nonspecific binding sites were blocked in 5% milk in PBST (1X PBS + 0.1% Tween-20). Incubation with primary antibody at 4°C in 5% bovine serum albumin/PBS was followed by incubation with horseradish peroxidase-conjugated secondary antibody in 2.5% milk in PBST. Proteins were visualized with enhanced chemiluminescence reagent (Amersham Corp.). Chromatin-bound proteins were extracted from formaldehyde (1%) cross-linked cells using Pierce Chromatin Prep Module (Thermo Pierce). Antibody sources: UNG-23936 (39 kDa band, nuclear UNG) and PCNA (Abcam); Tubulin (Calbiochem); γH2AX and Histone-H3 (Millipore); Cleaved PARP (Cell Signaling); and p-chk1, chk1, cdc2, and p-cdc2 (Santa Cruz).
UNG activity assay
UNG activity was measured using a 40-mer oligodeoxynucleotide duplex:
5′[HEX] GTAAAACGACGGCCAGTGUCTTCGAGCTCGGTACCCGGGG (top)
In fluorescent images, the top and bottom strands seem green and red, respectively. Oligonucleotide duplexes were incubated with either purified enzymes (1 unit) or whole-cell extract (2.5 μg) at 37°C for 30 minutes. The reaction was heat-killed at 95°C, reaction products were resolved by electrophoresis on denaturing 20% polyacrylamide gels and visualized with a Typhoon 9200 fluorescence imager (Amersham Bioscience). UNG activity (percentage of cutting) was defined as the fluorescence density of the cut band (20-mer) relative to the sum of the fluorescence intensity of the cut (20-mer) and uncut (40-mer) bands using ImageQuant software (Amersham BioScience).
Abasic (AP) site detection
Following drug treatment cellular DNA extracts were labeled with a biotinylated aldehyde reactive probe (ARP) for chemiluminescent AP site detection as previously described (21). For UNG-deficient cells, an additional incubation at 37°C with recombinant UNG (1U UNG/100 μg DNA) liberated genomic uracil before ARP labeling. Quantitative densitometry was conducted using Image J software.
Neutral comet assay
Treated cells were processed for comet tail formation under neutral comet assay conditions according to the manufacturers instructions for cell lysis and single-cell electrophoresis (Trevigen). Tail lengths were recorded for at least 50 comets on 2 separate slides (∼100 cells per treatment) using ImageJ software.
Colony survival assay
For cells in suspension, colony survival was determined by crystal violet staining of colonies formed after 10-day exposure of 5 × 103 cells to pemetrexed in soft agar. For adherent cells, colony survival was determined by methylene blue staining of colonies formed after 10-day exposure of 100 cells to pemetrexed in 6-well culture dishes. Only colonies containing ≥50 cells were counted. Data points represent percentage of colonies relative to untreated control averaged over 3 experiments.
Lung cancer cDNA microarray was purchased from Origene. For cell lines, total RNA was extracted from cells using RNAqueous-4PCR Kit (Ambion). Random hexamers (Invitrogen) were used to PCR amplify cDNA from 1 μg of RNA extract. TaqMan MGB probes (FAMTM dye labeled, Applied Biosystems) for nuclear UNG (UNG2)-, SMUG1-, MBD4-, TDG-, TYMS-, and βPol-amplified cDNA using 40 cycles of PCR in an ABI 7500 Fast Real-Time PCR System (Applied Biosystems). Target quantification was achieved after normalization to β-actin amplification as an endogenous control and is presented as relative quantification values or log-mean relative quantification values.
siRNA and shRNA transfection
UNG-directed short hairpin RNA (shRNA) and siRNA plasmids were purchased from Origene. Transfection was carried out according to manufacturer specifications. Stably transfected clones (shRNA) were selected with puromycin with subsequent expansion of a well isolated colony of cells.
Pemetrexed-resistant cell lines
H1299 cells were exposed to step-wise increasing concentrations of pemetrexed over a period of 4 months. UNG expression was monitored in the bulk population at 2, 8, 12, and 16 weeks before administration of the induction dosage and at 24 weeks in a population of bulk chronically exposed cells that had been without pemetrexed for 8 weeks. H1299 cells capable of growth in 50 nmol/L pemetrexed selection pressure were subcloned by limiting dilution in 96-well plates. Independent sublines were designated H1299/PR-1 and H1299/PR-2.
Xenografts in NOD/SCID mice
Tumor cells in early passage (5 × 106) were injected into bilateral flanks of female nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice (6 weeks old). When tumor volumes reached 100 mm3, mice were divided into control (n = 4) and treatment (n = 6) groups. Mice-bearing tumors were treated with pemetrexed (150 mg/kg) or 100 μL sterile PBS (control) by daily intraperitoneal injection for 5 consecutive days. Tumor measurements were taken every 2 days and response was quantified by tumor volume.
Pooled lung cancer microarray dataset analysis
Oncomine (Compendia Bioscience) was to analyze data from published microarray datasets (22–24). Data were normalized to set log-median expression of nonmalignant samples to 1.
Results are presented as the mean ± SEM. Significance, assigned for P values < 0.05, was determined by unpaired 2-tailed Student t test with standard software (GraphPad Prism). Correlations of gene expression with pemetrexed IC50 were estimated using Pearson correlation coefficient. Expression of expression UNG and other genes of interest was compared using analysis of variance (ANOVA) followed by Tukey pair-wise comparison procedure. The effects of multiple genes on drug IC50 were estimated using multivariable regression models, i.e., IC50 = intercept + coefficient1(gene1) + coeficient2(gene2) + ϵ.
A spectrum of UNG expression exists in human lung cancer
Our previous observation of profound pemetrexed sensitivity in UNG−/− colon cancer cells (20) prompted us to investigate the value of UNG as a mechanistic and predictive marker for pemetrexed response in human lung cancer. To do this, we evaluated UNG expression and pemetrexed sensitivity (IC50) in a panel of 8 lung cancer cell lines and 2 nonmalignant lung cell lines. Details of the cell lines used in this study are summarized in Supplementary Table S1. UNG protein (Fig. 1A) and transcript (Fig. 1B) levels were significantly higher in lung cancer cell lines compared with nonmalignant lung epithelial cells (WI38 and IMR90). In addition, lung cancer cell lines derived from small cell (H69) and squamous cell (Calu-1) carcinoma, known to be clinically unresponsive to pemetrexed (25–27), had higher levels of UNG compared with adenocarcinoma and large cell carcinoma cell lines (Fig. 1A and B). Pemetrexed IC50 was determined from colony survival experiments (Fig. 1C). Plotting UNG levels (protein band density; Fig. 1D) or UNG mRNA levels (relative quantification value; Fig. 1E) against pemetrexed IC50 for each cell line indicated that UNG expression was well correlated with pemetrexed IC50 (protein–Pearson r = 0.79, P = 0.021; mRNA–Pearson r = 0.71, P = 0.047). Pemetrexed IC50 was not significantly correlated with mRNA expression of other BER genes, including other glycosylases with uracil excision activity (SMUG1, TDG, and MBD4), DNA polymerase β (Polβ) or the pemetrexed target gene, TYMS (Supplementary Table S2). UNG mRNA levels were still marginally predictive of the pemetrexed IC50 (P < 0.1) when paired with other pathway-specific genes in multivariable regression analysis (Supplementary Table S3). Overall, these data illustrate a spectrum of UNG expression in human lung cancer cell lines that is positively correlated with pemetrexed response.
Measurement of UNG expression in primary human lung cancer tissue cDNA microarrays evinced significant variation in UNG expression among histologic subtypes (Fig. 1F). Similar to cell lines, UNG was elevated in lung cancer compared with nonmalignant tissue cDNAs. In addition, cDNA from small cell and squamous cell carcinoma had significantly higher UNG expression compared with adenocarcinoma (P < 0.0001; Tukey procedure). Pooled analysis of published microarray data (22–24, 28, 29) corroborated our findings (Fig. 1G). These data also showed that UNG expression was correlated with higher grade adenocarcinoma (28) and reduced 1 year survival (24). Higher UNG levels in lung cancer histologic subtypes that are typically resistant to pemetrexed (25–27) suggests there may be a clinically relevant correlation between UNG expression and pemetrexed response. TYMS, which has been associated with pemetrexed resistance, was also differentially expressed among the various lung cancer subtypes in the datasets analyzed (22–24, 28, 29). Squamous but not small cell carcinoma TYMS expression was significantly higher than adenocarcinoma (Fig. 1H). We were unable to locate tissue sets from recent pemetrexed clinical trials in lung cancer to interrogate for UNG expression. We have initiated prospective collection of lung cancer tissue samples from patients receiving pemetrexed to directly evaluate the relationship between UNG expression and pemetrexed response.
Loss of UNG expression increases lung cancer sensitivity to pemetrexed
We validated the correlation between UNG expression and pemetrexed response through direct targeting of UNG expression with siRNA and shRNA. For this analysis we used adenocarcinoma cell lines, A549 and H1975, which have moderate and high baseline UNG expression, respectively. Targeting of A549 cells with UNG-directed siRNA resulted in 70% reduction of UNG protein expression (Fig. 2A and B). A549 siRNA cells were 7-fold more sensitive to pemetrexed compared with parental cells (A549 siUNG IC50 = 60.54 nmol/L; A549 parental IC50 = 416.1 nmol/L, P < 0.0001; Fig. 2C). Similarly, stable transfection of H1975 cells with UNG-directed shRNA resulted in 50% to 60% knockdown of UNG protein expression (Fig. 2D and E). In colony survival assays (Fig. 2F), shRNA-targeted H1975 clones were more than 5-fold more sensitive to pemetrexed than parental cells (IC50 values: H1975 = 776.6 nmol/L; H1975 66 shUNG = 130.9 nmol/L; H1975 67 shUNG = 146.2 nmol/L, P < 0.0001). UNG-deficient H1975 cells displayed some cross sensitivity to raltitrexed and 5-fluorouracil but not temozolomide or cisplatin suggesting the effects of UNG loss are specific to TYMS inhibitors (Supplementary Table S4).
To determine the impact of UNG expression on pemetrexed sensitivity in vivo, H1975 and H1975 66 shUNG cells were xenografted subcutaneously into NOD/SCID mice. Treatment with 5 daily consecutive intraperitoneal injections of pemetrexed (150 mg/kg) evinced a tumor quadrupling time of 9.68 ± 0.31 days in H1975 wild-type tumors compared with 13.67 ± 0.28 days for H1975 shUNG tumors. At day 20, pemetrexed-treated H1975 66 shUNG tumors were 55.6% less than untreated controls, whereas H1975 wild-type tumors were only 37.8% less than untreated controls, P < 0.001 (Fig. 2G). These data indicate a significant increase in the antitumor effect of pemetrexed on tumors with lower levels of UNG expression, in vivo.
Limited uracil removal is associated with increased DNA damage in UNG-deficient cells
UNG is the major glycosylase responsible for removing uracil that is misincorporated during DNA replication (30). To relate pemetrexed sensitivity in UNG-deficient cells to the retention of uracil in DNA, we first evaluated uracil excision capacity of UNG-knockdown cells. UNG knockdown did not alter expression of other glycosylases capable of uracil excision (Fig. 3A). Protein extracts from untreated H1975 66 shUNG cells had diminished capacity to excise uracil from a synthetic oligonucleotide duplex containing a single-uracil residue (Fig. 3B and C). In pemetrexed-exposed cells, AP site detection was used as a surrogate measure of accumulated uracil. H1975 parental and H1975 66shUNG cells were treated with pemetrexed for 0 to 48 hours. Following treatment, DNA extracts from treated cells were labeled with chemiluminescent ARP that binds glycosylase-generated AP sites. Compared with control cells, H1975 shUNG cells had decreased AP site detection, P < 0.05 (Fig. 3D). Lack of AP site formation following pemetrexed exposure suggests decreased uracil excision and accumulation of uracil bases in DNA (20). To verify that uracil was retained in the DNA of cells with low UNG expression, we incubated DNA extract of pemetrexed-treated H1975 shUNG cells with recombinant UNG enzyme before labeling with ARP to detect AP sites. This in vitro uracil excision reaction resulted in the chemiluminescent detection of AP sites thereby confirming the persistence of uracil in H1975 shUNG DNA (Fig. 3E).
To determine the mechanisms responsible for enhanced pemetrexed sensitivity in UNG-deficient cells, we compared cell-cycle progression and expression of DNA damage response proteins in pemetrexed-treated UNG competent and UNG-knockdown cells. Stable knockdown of UNG (H1975 66 shUNG cells) conferred increased sensitivity to pemetrexed-mediated accumulation of early S-phase and sub-G1 cells (Fig. 4A), despite similar doubling times (Supplementary Table S1). S-phase accumulation was accompanied by induction of phospho-cdc2, phospho-chk1, and cyclin B1 (Fig. 4B). H1975 66 shUNG cells also had increased DNA DSB formation, as indicated by increased levels of γ-H2AX (Fig. 4B) and significantly increased comet tail lengths (P < 0.0001) in neutral assay conditions (Fig. 4C). H1975 66 shUNG cells were also more sensitive to pemetrexed-induced apoptosis as indicated by increased levels of cleaved poly-ADP ribose polymerase (PARP; Fig. 4B). These data are consistent with the DNA damage response observed in DLD1 UNG−/− cells treated with pemetrexed (20) and support our prior conclusion that cells lacking UNG are more sensitive to pemetrexed induced cell-cycle arrest, DNA DSB formation, and apoptosis.
Prolonged S-phase arrest is known to result in DNA DSB formation due to the collapse of stalled replication forks (31, 32). We evaluated the stability of the DNA replication fork in pemetrexed-treated lung cancer cells with native and reduced levels of UNG protein. Nucleotide incorporation experiments using chlorodeoxyuridine (CldU) and iododeoxyuridine (IdU) indicate decreased post—pemetrexed treatment nucleotide incorporation (data not shown). However, because CldU and IdU compete with deoxuridine (dU) for incorporation sites, these data are difficult to interpret prompting us to use an alternative measure of replication stability. Using biparametric flow cytometry we measured the dissociation of the replication fork processivity factor, PCNA at the single-cell level in H1975 cells expressing UNG-directed shRNA. Similar assays have been used to detect replication fork disassembly following etoposide and hydroxyurea exposure (33, 34). PCNA dissociation was determined by the percentage of cells in S-phase having low PCNA (red box; Fig. 4D). In wild-type cells, less than 1% of cells in treated and untreated samples had low PCNA staining in S-phase compared with 3.88 ± 0.93% and 7.81 ± 1.19%, H1975 shUNG cells treated with 25 nmol/L pemetrexed for 6 and 24 hours, respectively, P < 0.001 (Fig. 4D). As a complimentary experiment, we examined the expression of chromatin-bound PCNA after pemetrexed treatment by Western blot of chromatin cellular extracts. Pemetrexed treatment resulted in reduced expression of chromatin-bound PCNA in shUNG cells (Fig. 4E). Such dispersal of PCNA and other replication fork components from chromatin is indicative of collapsing replication forks (34) and our data implicate pemetrexed-mediated replication fork instability and subsequent fork collapse in the mechanism of DSB formation and cell death observed in pemetrexed-treated UNG-deficient cells.
Lastly, to link the accumulation and retention of genomic uracil with pemetrexed cytotoxicity, we used supplemental thymidine to promote salvage pathway production of dTMPs (Fig. 4F). In H1975 cells treated with varying concentrations of pemetrexed alone or in the presence of 10 μmol/L supplemental thymidine, the addition of thymidine rescued pemetrexed sensitivity in H1975 shUNG cells (Fig. 4G). Supplemental thymidine also significantly decreased pemetrexed-mediated induction of γ-H2AX (Fig. 4H). Importantly, AP site detection in H1975 parental cells and in H1975 shUNG DNA extracts incubated in vitro with purified UNG was also limited by the addition of thymidine (Supplementary Fig. S1). Reduced UNG excision (fewer AP sites) suggests supplemental thymidine dampens genomic uracil misincorporation. These data support the hypothesis that pemetrexed-induced cell-cycle arrest and DSB formation are consequences of uracil misincorporation.
UNG is induced in response to acute and chronic pemetrexed exposure
Resistance to anticancer agents that induce DNA damage has long been associated with upregulation of DNA repair genes. Pemetrexed resistance in cells with high UNG expression led us to hypothesize that UNG, by limiting uracil–DNA, promotes survival of pemetrexed-exposed cells. We assessed the impact of acute and chronic pemetrexed exposure on lung cancer cell expression of UNG. Acute pemetrexed exposure in the pemetrexed-sensitive adenocarcinoma cell line, H460 revealed time- and dose-dependent induction of UNG protein and transcript (Fig. 5A–D). Supplemental thymidine dampened the UNG induction response, linking the observation of UNG induction to TYMS inhibition and consequent uracil accumulation (Fig. 5E).
To investigate whether chronic pemetrexed exposure would select for cells with elevated UNG expression, we established pemetrexed-resistant sublines. We chose to induce resistance in H1299 cells (adenocarcinoma), the most pemetrexed sensitive cell line in our panel, which also expressed low levels of UNG. Sequential exposure of H1299 parental cells to increasing concentrations of pemetrexed over a 16-week period resulted in chronically elevated UNG protein expression that persisted for 8 weeks when pemetrexed was withdrawn (Fig. 6A). Clonogenic sublines, PR-1 and PR-2, were established and colony survival revealed 25-fold and 71-fold relative resistance to pemetrexed compared with parental cells, P < 0.0001 (Fig. 6B). Western blot analysis confirmed induction of UNG in PR1 and PR2 (Fig. 6C). UNG activity was also enhanced, as indicated by UNG cutting assay (Fig. 6D and E) and increased AP site detection in DNA extracts from pemetrexed treated pemetrexed-resistant sublines (Fig. 6F). PR-1 and PR-2 cells displayed cross-resistance to the TYMS inhibitors raltitrexed and 5-fluorouracil but were not resistant to other DNA damaging chemotherapeutics such as cisplatin or temozolomide that are not known to induce DNA repair through UNG-initiated BER (Supplementary Fig. S2). Transfection of UNG-directed siRNA into PR-1 cells restored pemetrexed sensitivity (Fig. 6G), indicating that UNG expression contributes significantly to the development of acquired pemetrexed resistance. The acute and chronic induction of UNG protein in lung cancer cells exposed to pemetrexed suggests that UNG activity is a prosurvival response to pemetrexed-induced uracil incorporation into DNA and the resulting DNA damage.
Increased BER gene expression and activity prompted us to investigate BER inhibition as a means to resensitize chronically exposed cells to pemetrexed. The BER inhibitor methoxyamine covalently binds the aldehyde of glycosylase-formed AP sites and blocks downstream BER. This agent is now in phase I clinical trials. Recent data has documented inhuman tolerance and potential efficacy when combined with pemetrexed (35). Because methoxyamine is well tolerated at 3 mmol/L in cell culture and does not impact cellular sensitivity to non-AP site producing chemotherapeutics, we have surmised that the potentiation of cytotoxicity is due primarily to methoxyamine interaction with AP sites versus nonspecific interactions with other intracellular aldehydes. Methoxyamine-bound AP sites are substrates for topoisomerase IIα (TOPOIIα) cleavage and DNA DSB formation. Cells with elevated TOPOIIα are more sensitive to methoxyamine potentiation of DNA damaging agent cytotoxicity (36). Interestingly, TOPOIIα expression is elevated in pemetrexed-resistant histologic subtypes of lung cancer primary tissues compared with adenocarcinoma (22–24) and we observed upregulation of TOPOIIα in both pemetrexed-resistant H1299 subclones compared with parental cells (Supplementary Fig. S3). Potentiation of pemetrexed cytotoxicity by methoxyamine is attenuated in UNG-deficient cells (ref. 20 and Supplementary Table S4) suggesting that methoxyamine effects are critically linked to UNG expression and activity. In colony survival experiments, 3 mmol/L methoxyamine cotreatment restored cellular sensitivity to pemetrexed in PR-1 cells (Fig. 6H) further suggesting that pemetrexed response is critically linked to uracil excision by UNG and in doing so highlighting the use of BER blockade to override acquired pemetrexed resistance.
The present study reveals a significant role for UNG-directed BER as a determinant of pemetrexed sensitivity in lung cancer. We observe a spectrum of UNG expression in human lung cancer that is well correlated with pemetrexed IC50 in cell lines and trends higher in pemetrexed-resistant small and squamous cell lung cancer subtypes. UNG has been identified as a prognostic marker in NSCLC (37). Among lung adenocarcinoma tissues, relationships between elevated high UNG expression and both decreased survival (24, 29,) and advanced disease stage (28) were noted. While it would seem obvious to also review prior studies to evaluate UNG as a predictive clinical predictive marker for pemetrexed response, we have been unable to find samples that allow this analysis. We are now collecting prospective samples to make this determination.
Other DNA glycosylases were not significantly associated with pemetrexed sensitivity (Fig. 1F) and did not compensate for UNG loss (Fig. 3A–C) suggesting UNG is the major glycosylase for uracil removal in pemetrexed-treated cells. The correlation between UNG expression and pemetrexed IC50 remained marginally significant in multivariable regression models controlling for the expression of other BER genes and TYMS, with slight improvements in coefficients of determination. This finding introduces a level of genetic complexity to our discussion of pemetrexed response in lung cancer by showing that, whereas not significantly predictive of pemetrexed IC50 alone, other BER and pathway-specific genes contribute to pemetrexed response. To further clarify the combinatorial role of UNG and other genes, we propose future studies to determine pathway-specific predictive gene signatures for pemetrexed response in lung cancer tissue of patients receiving pemetrexed.
Prolonged cellular exposure to TYMS inhibitors results in growth arrest or resistance (38) and patients receiving pemetrexed ultimately progress. Continued clinical success of pemetrexed and other TYMS inhibitor chemotherapeutics therefore depends upon biomarker-based patient selection. TYMS levels have been studied for predictive value in pemetrexed sensitivity. Increased intratumor levels of TYMS, observed in highly proliferating tumors, limit dTTP pool depletion and contribute to TYMS inhibitor resistance (14, 39). In addition, in lung cancer models of acquired pemetrexed resistance, TYMS is consistently elevated (13). Significant correlation between expression of TYMS and pemetrexed IC50 in NSCLC cell lines (14, 15) and a modest survival advantage in patients (15) have also been reported. High TYMS expression has been reported in high-grade small cell lung carcinoma (SCLC; ref. 40). However, TYMS has failed to predict SCLC response to pemetrexed combination therapy (41) and recent data suggest that TYMS has less predictive value beyond second-line therapy (42). In our analysis, TYMS was significantly elevated in cell lines (PR1 and PR2) with acquired pemetrexed resistance. We did not, however, observe significant correlation between pemetrexed IC50 and TYMS expression in our panel of cell lines.
Like TYMS, UNG expression is correlated with cellular proliferation. UNG-initiated BER has been observed at replication foci, illustrating coordination of DNA replication and repair of uracil–DNA (43, 44). Despite observations of high UNG expression in rapidly proliferating cells, neither loss of UNG nor chronic pemetrexed exposure altered cellular doubling time (Supplementary Table S1). Thus the predictive value of UNG for pemetrexed response extends beyond the association of UNG with replication.
Fluctuations in UNG expression significantly impact pemetrexed sensitivity, consistent with the prior observation that glycosylase activity is a major rate-determining step of BER (45). Recently, we reported that UNG−/− DLD1 cells accumulated uracil and were hypersensitive to pemetrexed (20). Here, through siRNA and shRNA knockdown of UNG we show the consistency of this phenotype in lung cancer cell lines, a clinically relevant model system. Decreased AP site formation in pemetrexed-treated UNG knockdown lung cancer cells suggests reduced uracil removal. Previous publications have suggested that substituting uracil for thymine reduces background DNA methylation, alters DNA structure and interferes with high-affinity protein–DNA interactions (46). Heavily uracilated DNA may therefore impede access and/or activity of transcription factors and replication fork proteins. In our study, we have observed compromised replication fork stability in the absence of uracil excision by UNG. Sensitivity to pemetrexed in UNG-deficient cells is rescued by supplemental thymidine, which attenuates UNG induction response and AP site formation in parental cells. On the basis of these data, we propose a novel hypothesis for thymineless death in UNG-deficient cells wherein lack of repair of misincorporated uracil leads to the collapse of DNA replication forks and triggers apoptosis.
That unrepaired uracil–DNA elicits a profound cytotoxic response in pemetrexed-treated human cancer cells was unexpected given the normal development of young UNG−/− mice (30, 47) and comparable sensitivity of UNG+/+ and UNG−/− MEFs to fluoropyrimidine TYMS inhibition (48). Our data are consistent with earlier reports of a direct role for uracil misincorporation in pemetrexed cytotoxicity (49). Indeed, RNA interference-mediated silencing of dUTPase, an enzyme responsible for maintaining low dUTP levels, significantly enhances pemetrexed cytotoxicity presumably due to increased dUTP incorporation (49). At baseline, otherwise isogenic UNG-proficient and -deficient cells have comparable levels of DNA damage markers despite reduced capacity for uracil excision suggesting UNG loss is well tolerated in the absence of TS-inhibitor challenge. When exposed to pemetrexed, however, UNG−/− human cancer cells accumulate up to 40-fold more uracil compared with UNG+/+ controls (20). In contrast, only 1.5-fold and 8-fold increases in uracil were reported in 5-fluorouracil (5-FU) and 5-fluoro-2-deoxyuridine (FdUrd)-treated UNG−/− MEFs (48) and raltitrexed-treated 293 T cells expressing the bacteriophage UNG inhibitor Ugi (50), respectively. These studies concluded that UNG activity did not impact TYMS sensitivity (48, 50). We believe the differential sensitivity to various TYMS inhibitors with UNG loss points to an as yet undetermined threshold of genomic uracil tolerance in mammalian cells. Such a threshold has been suggested by Luo and colleagues (50) and is presumed to depend upon both the TYMS inhibitor used and the uracil excision capacity of the cells studied. We speculate previous observations of little correlation of UNG with TYMS sensitivity (48, 50) are due to uracil accumulation within tolerance levels in those model systems.
A clear advantage to the identification of UNG as a predictive marker for pemetrexed resistance is the ability to potentiate pemetrexed efficacy via BER inhibition. We show that UNG is induced by acute and chronic pemetrexed exposure in lung cancer cell lines and methoxyamine inhibition of BER restores pemetrexed sensitivity in chronically exposed cells. methoxyamine-bound AP sites are clastogenic, trapping TopoIIα in a cleavable complex resulting in DNA DSBs (36). Cells with high TopoIIα expression are particularly sensitive to DNA damaging agents when combined with methoxyamine. Like UNG, TopoIIα levels are elevated in human lung cancer and are highest in small- and squamous cell carcinoma (22, 24). Therefore, the pemetrexed/methoxyamine combination is a rational strategy to overcome pemetrexed insensitivity in certain lung cancer subtypes and to restore sensitivity in cells that acquire resistance due to chronic pemetrexed exposure. Pemetrexed/methoxyamine combination therapy has been pursued in phase I clinical trials involving solid tumors (35) resulting in a partial response in 56% of patients enrolled. Among the responders were 3 patients with squamous-cell lung carcinoma and 1 patient with squamous oropharyngeal carcinoma that notably had high TYMS levels (35). On the basis of these data, phase II and randomized controlled trials involving pemetrexed and methoxyamine are planned.
Tailoring chemotherapy based on histologic subtype and biomarker expression is a favorable strategy for aggressive, treatment-refractory malignancies such as lung cancer. Our observations that UNG expression is elevated in experimental models of pemetrexed-resistant lung cancer and correlates with pemetrexed efficacy prompt us to propose investigation of UNG as a novel predictive marker for pemetrexed in human lung cancer. Moreover, because UNG loss and BER inhibition with methoxyamine potently restore pemetrexed sensitivity in resistant cells, UNG-directed BER may be a novel therapeutic target, distinct from the folate metabolism pathway, for overcoming pemetrexed resistance in human lung cancer.
Disclosure of Potential Conflicts of Interest
S.L. Gerson has ownership interest (including patents) and is consultant/advisory board member of Tracon Pharma. No potential conflicts of interest were disclosed by the other authors.
Conception and design: L.D. Weeks, S.L. Gerson
Development of methodology: L.D. Weeks, S.L. Gerson
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.D. Weeks, S.L. Gerson
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.D. Weeks, P. Fu, S.L. Gerson
Writing, review, and/or revision of the manuscript: L.D. Weeks, P. Fu, S.L. Gerson
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.L. Gerson
Study supervision: S.L. Gerson
The authors thank the important contributions of Dr. Lili Liu. The authors also thank Alex Almasan, Clive Hamlin, Ruth Keri, Shigemi Matsuyama, and George Stark for their careful review of this manuscript.
This study was supported by grants from the NIH: CA86357 and CA 430703 (to S.L. Gerson). L.D. Weeks is the recipient of an F31 NRSA training grant from the National Cancer Institute (CA159614) and was also supported by an NIGHMS MSTP training grant T32-GM007250.
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.