Individual Differences in the Response of Human β-Lymphoblastoid Cells to the Cytotoxic, Mutagenic, and DNA-Damaging Effects of a DNA Methylating Agent, N-Methylnitrosourethane

Abstract

Metabolic activation of many carcinogens leads to formation of reactive intermediates that form DNA adducts. These adducts are cytotoxic when they interfere with cell division. They can also cause mutations by miscoding during DNA replication. Therefore, an individual’s risk of developing cancer will depend on the balance between these processes as well as their ability to repair the DNA damage. Our hypothesis is that variations of genes participating in DNA damage repair and response pathways play significant roles in an individual’s risk of developing tobacco-related cancers. To test this hypothesis, 61 human B-lymphocyte cell lines from the International HapMap project were phenotyped for their sensitivity to the cytotoxic and genotoxic properties of a model methylating agent, N-nitroso-N-methylurethane (NMUr). Cell viability was measured using a luciferase-based assay. Repair of the mutagenic and toxic DNA adduct, O⁶-methylguanine (O⁶-mG), was monitored by LC-MS/MS analysis. Genotoxic potential of NMUr was assessed employing a flow-cytometry based in vitro mutagenesis assay in the phosphatidylinositol-glycan biosynthesis class-A (PIG-A) gene. A wide distribution of responses to NMUr was observed with no correlation to gender or ethnicity. While the rate of O⁶-mG repair partially influenced the toxicity of NMUr, it did not appear to be the major factor affecting individual susceptibility to the mutagenic effects of NMUr. Genome-wide analysis identified several novel single nucleotide polymorphisms to be explored in future functional validation studies for a number of the toxicological end points.

Graphical Abstract

INTRODUCTION

Cancer risk is determined in part by gene-environment interactions.¹ Lung cancer is an important example of gene-environment interactions where approximately 90 percent of lung cancers are attributed to smoking; however, only 15% of smokers get lung cancer.² Literature reports indicate that genetic factors contribute significantly to a person’s susceptibility to tobacco-induced lung cancer.^3,4 A major factor affecting lung cancer risk is genetic variation in nicotine metabolism;^5–8 if an individual is a rapid nicotine metabolizer, they smoke more intensely, driving their exposure to tobacco smoke chemicals.^6,9 However, genetic differences in nicotine metabolism does not account for all the individual variation in cancer risk.

Tobacco products and smoke are rich in chemicals that can damage DNA. An individual’s cancer risk when exposed to a genotoxic carcinogen depends upon the balance of activation and detoxification metabolic pathways as well as the individual’s ability to repair the resulting DNA damage. Polymorphisms in DNA repair and cell cycle proteins have been associated with altered cancer risk in individuals exposed to tobacco smoke and/or products.^10–14 The functional consequences of these genetic changes are not well understood. The major DNA damage caused by tobacco smoke is not known, in part, because tobacco smoke is such a diverse mixture, containing more than 7000 chemicals where more than 70 are known carcinogens.¹⁵ It is likely that an individual’s genetics will determine which tobacco smoke chemicals they will be susceptible to.

Our long-term goal is to determine how genetic susceptibility impacts an individual’s response to DNA damage from tobacco chemicals. Our initial studies focused on methyl DNA damage that is generated from two tobacco smoke carcinogens, N-nitrosodimethylamine (DMN) and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK). We chose this DNA-damage pathway because much is known about how cells handle DNA methyl damage. To test the hypothesis that individual differences in DNA repair capacity influence an individual’s sensitivity the cytotoxic and genotoxic effects of tobacco smoke chemicals, Epstein–Barr virus (EBV)-transformed B-lymphocyte cell lines derived from European Americans (CEU) and Yoruban (African, YRI) populations in the International HapMap study were exposed to the model methylating agent, N-methylnitrosourethane (NMUr). NMUr has previously been used as a model for DMN and NNK in cell models that lack expression of cytochrome P450.¹⁶ It is activated by ester hydrolysis to form the same reactive methanediazohydroxide that is formed by enzymatically mediated α-carbon hydroxylation of DMN and NNK. This reactive intermediate generates methyl DNA adducts such as 7-methylguanine (7-mG) and O⁶-methylguanine (O⁶-mG).¹⁷

The DNA adduct primarily responsible for the cytotoxicity and mutagenic properties of NMUr is O⁶-mG.^16,18 There are several pathways by which it triggers its genotoxic effects (Scheme 1). The main pathway protecting against these effects is O⁶-alkylguanine DNA alkyltransferase (AGT)-mediated repair. AGT removes the methyl group from the O⁶-position of guanine and transfers it to an active site cysteinyl residue, resulting in its inactivation.^18–20 Therefore, baseline levels of AGT determine the initial repair capacity of a cell. Unrepaired O⁶-mG miscodes during replication, causing polymerases to preferentially insert a T opposite this modified base, rather than the correct C.²¹ Mismatch repair (MMR) proteins are able to recognize this mismatch and initiate repair. In the absence of AGT, MMR results in a futile repair cycle, which ultimately results in cell death.²² In the absence of MMR, the mismatch escapes repair, and the cell proceeds through cell division resulting in enhanced mutagenesis.^23,24 Homologous recombination (HR) is an alternate mechanism for the cell to repair this mismatch, using the unmodified sister chromosome as a template. Loss of HR sensitizes MMR proficient cells to the toxic effects of methylating agents.²⁵ Therefore, genetic variations in the proteins involved in these processes are likely to influence the toxic and mutagenic properties of tobacco carcinogens. Since mutagenic activity is enhanced when the damaged cell does not die, a measurement of interindividual variation in sensitivity to the cytotoxic effects of these model reactive metabolites relative to their sensitivity to their genotoxic effects is critical in understanding how genetic variations in DNA damage repair and DNA damage response proteins influence lung cancer risk.

Scheme 1.

Pathways by Which O⁶-mG Triggers Its Genotoxic Effects

Lymphocytes have been a useful surrogate for human sensitivity to mutagens and have been extensively employed to explore interindividual variation in DNA repair,^2,26 allowing focus on specific toxicological end points without complications from the many variables that occur in vivo. The value of the HapMap cell lines is that they are immortal lymphocyte cell lines from individuals of different ethnic backgrounds.²⁷ Therefore, it is possible to obtain sufficient amounts of sample required for the proposed end points. In addition, there are multiple sources for the genetic information on the individual cell lines: ~4 million single nucleotide polymorphisms (SNPs) were determined by the HapMap project;²⁷ genetic variation in DNA repair and cell cycle pathways were determined by the NIEHS environmental genome project;²⁸ and whole genome sequencing of many cell lines was performed through the 1000 Genomes Project.²⁹ All this information is publically available for download.

To determine if individuals differ in their ability to repair O⁶-mG and if these differences influence the cytotoxic and mutagenic effects of NMUr, a series of experiments were conducted in CEU and YRI cell lines as outlined in Scheme 2. Cells were treated with NMUr for 1 h. Cytotoxicity was assessed 48 h later with Cell-titer Glo. Mutation frequency in the PIG-A gene caused by NMUr (0, 10, or 20 μM) was determined 2 weeks later. The levels of 7-mG and O⁶-mG were measured 0 and 6 h after a 1 h exposure to 10 μM MNUr to assess the repair of O⁶-mG. We then determined if these toxicological end points were correlated with one another as well as with genetic variations reported in the HapMap Project.

Scheme 2.

Overall Experimental Design

MATERIALS AND METHODS

N-Methylnitrosourethane (NMUr).

N-nitroso-N-methylurethane (NMUr) was purchased from MRI Global Chemical Carcinogen Repository (Kansas City, MO).

Cells Lines.

The experiments were performed on Epstein–Barr-transformed B-lymphoblastoid cell lines derived from individuals from Utah (Hapmappt01, CEU) and Yoruba in Ibadan, Nigeria (Hapmappt03, YRI), purchased from the Coriell Institute for Medical Research (Camden, NJ). In total, 61 cell-lines (CEU, 16 male and 15 female; YRI, 15 male and 15 female) were employed in these studies (Table S1). The cell lines were randomly selected from the parental cell lines. They were maintained in RPMI-1640 media (Life Technologies, Carlsbad, CA) supplemented with 15% of heat-inactivated fetal bovine serum (FBS, Life Technologies, Carlsbad, CA) at 37 °C in a humidified 5% CO₂ atmosphere. Individual cell lines were grown for at least 2 passages prior to plating for the three phenotyping experiments. For each cell line, all three experiments were performed at the same time. To minimize variation between experiments with the same cell line, all the experiments used cells that had been in culture for no more than 2 weeks. A cell line derived from Mantle cell lymphoma was kindly provided by Dr. David Araten, New York University.

Cytotoxicity Assay.

Cells were seeded 100 000 cells/well in 6-well plates (total volume: 2 mL) 24 h before they were exposed to NMUr for 1 h (0, 5, 10, 20, 40, 50, 75, 100, or 200 μM). At 1 h, porcine liver esterase (1 unit/well, Sigma-Aldrich, St. Louis, MO) was added to inactivate any extracellular NMUr. This step removed excess NMUr in order to reduce variation between experiment replicates. Each treatment was performed in duplicate. After 48 h, three 50-μL aliquots were removed from each well and combined with 50 μL of CellTiter-Glo Luminescent Cell Viability reagent (Promega Corporation, Madison, WI) in 96-well plates and incubated for 20 min on a shaker. The plates were analyzed using a Synergy H1Multi-Mode Reader (BioTek, Winooski, VT) operating in the luminescent mode with an acquisition gain of 134 according to manufacturer’s instructions. Cell survival was normalized to the luminescence of the cells treated with 0 μM NMUr. IC₂₀, IC₅₀, and IC₈₀ values were determined using the Gen5Microplate Reader and Imager Software. Seventeen cell lines were repeated at a different time to determine the technical variation of the assay. The IC₂₀ values varied by 13% ± 13%, and the IC₅₀ values varied by 10% ± 10%.

PIG-A Mutagenesis Assay.

Cell Treatment.

Cells were seeded in T-25 flasks (500 000 cells/flask, 10 mL total volume) 24 h before they were treated with 0, 10, or 20 μM NMUr for 1 h. At 1 h, porcine liver esterase (2 units/flask) was added to inactivate any extracellular NMUr. The cells were then allowed to recover for 2 weeks, during which time, the cultures were split to 2.5 × 10⁶ cells/10 mL whenever the concentration exceeded 1 × 10⁶ cells/mL.

Flow Cytometry.

Five million live cells were collected and washed with fluorescence-activated cell sorting (FACS) buffer (colorless RPMI-1640 supplemented with 15% FBS; 500 μL). The cells were then incubated with a mixture of primary antibodies against three glycophosphatidylinositol (GPI)-linked proteins diluted in FACS buffer: anti-CD48 (1:200 dilution, Serotec, Atlanta, GA), anti-CD55 (1:10 dilution, Serotec Atlanta, GA), and anti-CD59 (1:20 dilution, Serotec, Atlanta, GA) for 30 min on ice (total volume: 50 μL). The cells were then washed twice with FACS buffer (500 μL) and incubated with R-phycoerythrin-conjugated (PE) goat antimouse antibodies (1:5 dilution, BioLegend, San Diego, CA) for 30 min on ice in dark (total volume: 50 μL). Cells were washed twice with FACS buffer (500 μL) and incubated in FACS buffer containing allophycocyanin-conjugated (APC) antibody specific for CD19 (1:10 dilution, BioLegend, San Diego, CA) for 30 min on ice in the dark (total volume: 50 μL). This protein is a non GPI-linked protein used for selection of lymphocytes. After the final staining, the cells were washed twice, resuspended in FACS buffer (500 μL), and kept on ice until the analysis. In some cases, the cells were resuspended in 4% paraformaldehyde (500 uL) and kept at 4 °C until analysis.

Flow Cytometry Analysis.

Cells were analyzed on a BD LSRII/ Fortessa H4760 instrument (BD Bioscience, San Jose, CA) using FACSDiva and Flowjo softwares in the University of Minnesota Masonic Cancer Center’s Flow-Cytometry Shared Resource. Five minutes before the analysis of each tube, 0.5 μL of Sytox Blue (Life Technologies, Carlsbad, CA) was added. Cells were then passed through a 35 μm strainer (Corning, NY). Cells were analyzed at a rate of 3000–4000 cells per second. Figure S1 shows gating strategy performed for quantification of the mutant fraction. Forward and side scatters (gain values set at BD LSRII/Fortessa H4760 default levels) were used to select for a cell population. Live cells were selected on the basis of the exclusion of SytoxBlue fluorescence. In addition, live cells were also positively identified by the expression of CD19, as indicated by APC fluorescence (y-axis). The geometric mean of APC fluorescence of SytoxBlue-negative population was defined, and the dead cells were excluded as having mean fluorescence <10% of the total population. The PIG-A mutants were identified by the negative expression of CD48, CD55, and CD59 proteins as indicated by PE fluorescence (x-axis). The geometric mean of PE fluorescence of live cells was defined, and the gates were chosen on the basis of a 2% cutoff of fluorescence level. The number of mutants was normalized per 10⁶ cells counted.

DNA Repair Assay.

Cell Treatment.

Cells (5 × 10⁶ cells) were seeded in T25 flasks (total volume: 10 mL) 24 h before treatment with 0 (n = 1) or 10 μM (n = 2) NMUr. After 1 h at 37 °C, esterase (5 U/flask) was added to hydrolyze any residual NMUr. Aliquots of 4 mL were removed at 0 or 6 h after the addition of esterase. The cells were pelleted by centrifuge (1200 rpm, 5 min). After the supernatant was removed, the cells were repelleted by centrifugation (1200 rpm, 1 min). DNA was isolated immediately from the cell pellet using the EZNA Blood DNA Mini-Kit (Omega Biotek, Norcross, Georgia) with using the manufactor’s protocol. Purity was assessed by measuring absorbance at 230, 260, and 280 nm using NanoDrop instrument (Thermo Scientific, Wilmington, DE), and samples were stored at −20 °C until analysis.

Preliminary studies were performed with cell lines NA12864 and NA19172 to determine the time points used with the larger study. Cells (15 × 10⁶) were seeded in T75 flasks (total volume: 30 mL). Triplicate incubations were conducted with 0 or 10 μM NMUr where esterase was added after 1 h at 37 °C. Aliquots (4 mL) were collected 0, 4, 6, and 24 h after the addition of esterase. DNA was isolated as described above.

DNA Adduct Levels.

DNA (~20 μg of DNA) in 6 mM sodium phosphate, pH 7.0 was spiked with [¹³C²H₃]-7-mG (1 pmol) (final volume: 500 μL) and heated at 80 °C for 30 min on a shaking heating block (Eppendorf, Hauppauge, NY; 1000 rpm). The samples were cooled on ice for 5 min before 100 μL of the hydrolysis solution was transferred to a Nanosep 10k spin-filter (Pall Corporation, Port Washington, NY) and centrifuged at 6000 rpm for 10 min. The eluant was stored at −20 °C for LC-MS/MS analysis of 7-mG. The remaining hydrolysis solution (400 μL) was acidified with 50 μL of 1 N HCl, spiked with 60 fmol of [²H₃]-O⁶-mG, and heated at 80 °C for 30 min on a shaking heating block (1000 rpm). The samples were cooled on ice for 10 minutes, and a portion (75 μL) was removed for guanine analysis. The remaining hydrolysate was applied to a polymeric StrataX cartridges (Phenomenex, Torrance, CA; 30 mg, 1 mL) which had been preconditioned with 1 mL of methanol and 1 mL of ultrapure water. The cartridge was subsequently washed with 1 mL of ultrapure water followed by 1 mL of 10% (v/v) methanol:-water. The analyte, O⁶-mG, was eluted with 1 mL of methanol, which was collected in glass vials. Samples were immediately dried by speed-vacuum and stored at −20 °C until analysis. These samples were reconstituted in 50 μL of 20 mM ammonium acetate containing 5% 75:25 (v/v) methanol:acetonitrile for LC-MS/MS analysis.

Adduct (7-mG and O⁶-mG) quantification was performed using a TSQ Vantage mass spectrometer (Thermo Scientific, Waltham, MA) attached to an Eksigent (Dublin, CA) nanoLC-ultra 2D pump and as-C autosampler system. Mixtures (8 μL) were separated on a Thermo Aquasil C18 column (150 × 0.5 mm, 5 μm). Analytes were eluted from the column with 25 mM ammonium acetate (solvent A) and 75:25 (v/v) methanol:acetonitrile (solvent B) using the following gradient at a flow rate of 10 μL/min: The gradient was linearly increased from 95% solvent A/5% solvent B to 80% solvent A/20% B over 13 min and then increased to 25% solvent A/75% B over 3 min. The column was returned to initial conditions in 10 min and equilibrated for 14 min before the next injection. The source spray voltage was 3200 V. Nitrogen sheath gas pressure was 32 mTorr. Q2 argon gas pressure was 1.3 mTorr. The declustering voltage was 6 V. The scan width was 0.1 m/z, and the scan time was 0.3 s. Q1 and Q3 resolution was 0.70 fwhm. The electrospray ionization source was set to operate in negative ion mode with selective reaction monitoring. Mass transitions for both 7-mG and O⁶-mG were m/z 166 → m/z 149. Secondary transitions for 7-mG (m/z 166 → m/z 124) and O⁶-mG (m/z 166 → m/z 121) were also monitored to determine any possible coelutors. [¹³C²H₃]-7-mG was monitored using m/z 170 → m/z 153, and 170 → m/z 128, and [²H₃]-O⁶-mG was monitored using m/z 169 m/z 152 and m/z 169 m/z 124. The retention times for 7-mG and O⁶-mG were 9.6 and 15.0 min, respectively. The amount of 7-mG and O⁶-mG in each sample was determined from the product of the area ratio (unlabeled DNA adduct/labeled standard) and the amount of internal standard added. Calibration curves were constructed from known amounts of authentic standards spiked with known amounts of internal standard. Guanine concentrations were determined as previously described.³⁰ The amount of DNA adducts was normalized to the amount of the guanine in each sample.

Statistical Analysis.

Linear regression models were implemented to compare cytotoxicity, mutagenesis, and DNA-repair end points between CEU and YRI. Some of the cell lines were assessed more than once for all the end points. In this case, the values were averaged prior to statistical analysis. The p values are reported with and without cell turnover rate adjustment. The effect of gender for each of the outcomes was assessed using a simple linear regression model, and no significant effect was observed. Correlations between the toxicological outcomes in a pooled data set (CEU + YRI) were determined with Spearman correlation. Correlations between toxicological outcomes were conducted separately for each CEU and YRI data sets using Pearson’s correlation. Mutation counts as well as the IC values were log transformed prior to calculating correlations.

Genome-wide association studies (GWAS) were performed with the three end points of interest for this study: cytotoxicity, DNA methylation, and mutagenesis. The end points for assessing cytotoxicity were IC₂₀, IC₅₀, and IC₈₀ values. For the evaluation of genetic factors affecting levels of DNA methylation and repair, we used levels of 7-mG and O⁶-mG at both t = 0 h and t = 6 h as well as the ratio of 7-mG to O⁶-mG at 0 or 6 h. For mutagenesis, the absolute mutation frequency (mutants per 10⁻⁶ cells) caused by 0, 10, or 20 μM NMUr were assessed for genome-wide significance as well as the difference in mutation counts between cells treated with 10 μM NMUr and those treated with 0 μM NMUr and the difference in mutation counts between cells treated with 20 μM NMUr and those treated with 0 μM NMUr. End point values were averaged across replicates to reduce variability in the measurements. Mutation counts as well as the IC values were log transformed prior to analyses to better adhere to the Normality assumption.

Genotype data on the 61 cell-lines (31 CEU and 30 YRI cell-lines) were obtained from NCBI’s most recent phaseII+III hapmap database.³¹ These genotypes were converted to minor allele frequencies using R statistical software version 3.6.0 to prepare for analysis. Data on 471 SNP were excluded prior to analysis due to inconsistencies specified by NCBI.³¹

Separate GWAS analyses were conducted for data collected on the CEU and YRI cell lines, investigating the relationship between each cytotoxicity, mutagenicity, and DNA adduct levels’ end points and each SNP for both groups using R statistical software version 3.6.0.³² Specifically, we separately fit linear regression models for these end points using each SNP as a covariate while controlling for gender. Statistical significance of the associations between each SNP and toxicological end point was determined by comparing two-tailed p-values to the traditional GWAS threshold of 5 × 10⁻⁸ to account for multiple comparisons.

RESULTS

The Cytotoxic Effects of NMUr Vary with Individual Lymphoblastoid Cell Lines.

NMUr cytotoxicity was determined in lymphoblastoid cell lines from 61 different individuals (CEU, 16 male and 15 female; YRI, 15 male and 15 female). The cells were treated with increasing concentrations of NMUr (0–200 μM) for 1 h, and cell viability was assessed with CellTiter Glo 48 h later (Table S2). There were significant interindividual differences in response to the cytotoxic effects of the methylating agent with IC₅₀ values ranged from 3.6 to 103 μM overall (Figure 1 and Table S3). The IC₅₀ values ranged from 3.6–95 μM in the CEU cell lines and from 36–103 μM in the YRI cell lines. There was a small difference in the average IC₅₀ value between the two populations with the YRI population slightly less sensitive to the toxic effects of NMUr: CEU, 54 ± 25 versus YRI, 67 ± 17 (p value = 0.03) (Table 1). This difference remained significant after adjusting for cellular turnover rate (Table 1). IC₂₀ values were not significantly different between the two ethnic groups (Table 1).

Figure 1.

Variation in NMUr IC₅₀ values determined in CEU and YRI lymphoblastoid cell lines. Cells were exposed to NMUr (0, 5, 10, 20, 40, 50, 75, 100, or 200 μM) for 1 h. Excess NMUr was removed with the addition of esterase. Each treatment was performed in duplicate. At 48 h after exposure, three 50-μL aliquots were removed from each well and combined with CellTiter-Glo Luminescent Cell Viability reagent to determine cellular ATP levels as a measure of cytotoxicity. Cell survival was normalized to the luminescence of the cells treated with 0 μM NMUr. IC₅₀ values were determined using the Gen5Microplate Reader and Imager Software. The dotted line indicates the population average.

Table 1.

Comparison of Toxicological End Points between the CEU and YRI Cell Lines^c

	CEU	YRI	p	pb
n	31	30
gender = male (%)	15 (48.4)	15 (50.0)	1.000
cytotoxicity end pointsc	mean ± SD	mean ± SD
IC20 (μM)	26.4 ± 15.1	30.8 ± 9.7	0.185	0.192
IC50 (μM)	54.4 ± 24.6	66.5 ± 17.3	0.030a	0.036a
IC80 (μM)	134 ± 38	153 ± 29	0.046a	0.050a
% survival at 10 μM	87.9 ± 20.1	96.2 ± 6.2	0.036a	0.042a
mutagenesis end pointsd
mutants at 0 μM NMUr (per 106 cells)	88.3 ± 153	61.6 ± 102	0.427	0.387
mutants at 10 μM NMUr (per 106 cells)	114 ± 221	72.2 ± 140	0.381	0.362
mutants at 20 μM NMUr (per 106 cells)	140 ± 251	82.5 ± 138	0.273	0.249
difference in number of mutants at 10 and 0 μM NMUr	32.4 ± 76.5	17.2 ± 40.3	0.339	0.359
difference in number of mutants at 20 and 0 μM NMUr	54.5 ± 118	24.6 ± 41.4	0.193	0.191
difference in number of mutants at 20 and 10 μM NMUr	42.8 ± 84.1	15.5 ± 27.8	0.096	0.090
DNA repair end pointse
7-mG at 0 h (pmol/μmol guanine)	181 ± 21.5	165 ± 23	0.005a	0.006a
7-mG at 6 h (pmol/μmol guanine)	139 ± 15	129 ± 19	0.020a	0.029a
O6-mG at 0 h (pmol/μmol guanine)	11.2 ± 5.2	8.7 ± 2.6	0.023a	0.027a
O6-mG at 6 h (pmol/μmol guanine)	6.7 ± 5.0	3.9 ± 2.2	0.006a	0.008a
7-mG/O6-mG at 0 h (mean ± sd)	19.2 ± 7.5	20.0 ± 4.3	0.625	0.658
7-mG/O6-mG at 6 h (mean ± sd)	33.8 ± 22.8	41.3 ± 16.4	0.146	0.152
turnover rate (in days)	1.74 ± 0.51	1.87 ± 0.64	0.356

^aSignificant difference between groups.

^bp value after adjusting for turnover rate.

^cCells were exposed to NMUr (0, 5, 10, 20, 40, 50, 75, 100, or 200 μM) for 1 h. Cytotoxicity was measured 48 h after exposure with CellTiter-Glo Luminescent Cell Viability reagent. Cell survival was normalized to the luminescence of the cells treated with 0 μM NMUr. IC₂₀, IC₅₀, and IC₈₀ values were determined using the Gen5Microplate Reader and Imager Software.

^dCells were treated with 0, 10, or 20 μM NMUr for 1 h. Mutagenesis in the PIGA gene was determined 2 weeks later as described in the Materials and Methods section.

^eCells were treated with 0 or 10 μM NMUr for 1 h. DNA was isolated from cells at 0 or 6 h after this exposure. DNA adduct levels were measured by the LC-MS/MS assay described in the Materials and Methods section.

The Mutagenic Activity of NMUr Varies with Individual Lymphoblastoid Cell Lines.

The mutagenic activity of NMUr in the lymphoblastoid cell lines was determined using an in vitro mutagenesis assay, which screens for mutations in the PIG-A gene following exposure to NMUr (Figure 2). A similar approach has been reported in TK6 cells.³³ We chose this assay over the more common hprt mutagenesis assay since mismatch repair variants impact the response to both 6-thioguanine and methylating agents.^23,24,34 In addition, the hprt mutagenesis assay is very labor intensive.²⁶ Because the PIG-A gene codes for a critical enzyme in the biosynthesis of GPI anchor proteins, mutations in this gene results in the loss of GPI-anchored cell surface antigens.³⁵ A flow cytometry assay has been developed to detect these mutations,³⁶ and it has been used to determine human somatic mutation rates.³⁷ Therefore, this mutagenesis assay is higher throughput than the traditional hprt mutagenesis assay. To overcome the challenge of detecting very rare GPI⁻ cells in the presence of >1 × 10⁶ GPI⁺ cells, antibodies to three independent GPI-anchored proteins CD48, CD55, and CD59 were used. These antibodies were stained with R-phycoerythrin (PE)-conjugated primary antibodies. All live lymphocytes were selected by their ability to exclude SytoxBlue fluorescence. Further selection of live cells was achieved with APC-conjugated antibodies specific for CD19, a non-GPI-linked protein. This reduced the background noise in the quadrant containing the GPI⁻ cells. The gating strategy is shown in Figure S1. Preliminary studies demonstrated that the PIG-A mutant frequency stabilized 2 weeks following chemical exposure (data not shown); this observation was similar to that reported for TK6 cells.³³ To determine the reproducibility of the assay, it was performed in several cell lines using different aliquots of that cell line (Figure 3). The number of mutants induced was not associated with the cytotoxic effects of NMUr in that particular cell line at the concentration tested (Figure 3).

Figure 2.

(Top) Schematic representation of molecular mechanism of PIG-A assay. Intact PIG-A gene leads to the expression of surface markers CD48, CD55, and CD59. This expression is abolished once PIG-A gene is mutated. The expression of CD19 is not mediated by PIG-A gene, and so this marker can be used as a cellular integrity control. (Bottom) Representative fluorescent plots increased mutant count in a cell-line treated with 20 μM NMUr. Quadrant Q3 represents the mutant cells, as these have the reduced expression of CD48, CD55, and CD59 markers judged by the fluorescent intensity of PE.

Figure 3.

Mutation frequency in the PIG-A gene as compared to the cytotoxicity induced by a 1 h exposure to 0, 10, or 20 μM NMUr in three different lymphoblastoid cell lines. Excess NMUr was removed with the addition of esterase, and the cells were grown for an additional 2 weeks prior to performing the PIG-A flow cytometry assay described in the Material and Methods section. The red line is mutagenesis experiment 1, and the blue line is mutagenesis experiment 2. The data for each experiment is an average of duplicate measurements at 0, 10, and 20 μM NMUr. The cytotoxicity data for 0, 10, and 20 μM NMUr was obtained from cells treated at the same time as the mutagenesis experiments. Cell viability (green bars) was determined with Cell-titer Glo, and the data are an average of the two experiments (n = 6 at each concentration for each experiment). The error bars represent standard deviation.

The assay was then performed concurrently with the cytotoxicity assay, employing 61 lymphoblastoid cell lines (CEU, 16 male and 15 female; YRI, 15 male and 15 female). The mutation frequency in the PIG-A gene was determined 2 weeks following a 1 h treatment with 0, 10, or 20 μM NMUr. These concentrations were selected to ensure that the impact of toxicity did not interfere with our ability to detect mutations; the average survival at 10 μM NMUr was 88 ± 20% in the CEU cell lines and 96 ± 6% in YRI cell lines (Table 1). Because of the wide range in background mutations (median: 36 mutants per 10⁶ cells where the range was 1.9–650 mutants per 10⁶ cells), we compared cell lines based on the increase in mutations caused by 10 or 20 μM NMUr (Figure 4 and Table S4). Overall, a variety of genotoxic dose response profiles were observed. Roughly half of the cell lines were insensitive to the NMUr concentrations tested. The number of mutants per 10⁶ cells declined with NMUr exposure in four of the nonresponsive cell lines; two of these cell lines were very sensitive to the toxic effects of NMUr (IC₅₀ < 5 μM), whereas two cell lines were not (IC₅₀ > 45 μM). The remaining cell lines exhibited either concentration-dependent increase in PIG-A mutants or were more mutagenic either at 10 μM NMUr or 20 μM NMUr. While there was no significant difference in the increase of mutation frequency caused by 10 or 20 μM NMUr between the two populations (Table 1), there were more CEU cell lines that responded to the mutagenic effects of NMUr than YRI cell lines (Figure 4).

Figure 4.

Increase in mutation frequency in the PIG-A gene in (A) CEU and (B) YRI cell lines treated with 0, 10, or 20 μM NMUr for 1 h. Excess NMUr was removed with the addition of esterase, and the cells were grown for an additional 2 weeks prior to performing the PIG-A flow cytometry assay described in the Material and Methods section. The red bars are the increase caused by 10 μM NMUr, and the blue bars are the increase caused by 20 μM NMUr.

The Repair of O⁶-mG Varies with Lymphoblastoid Cell Line.

The levels of DNA damage and repair were measured in cell lines treated with 10 μM NMUr. The average survival of the cell lines at this concentration was 88 ± 20% in the CEU cell lines and 96 ± 6% in YRI cell lines; this concentration was selected as it was expected that repair would not be saturated under these treatment conditions. Levels of 7-mG and O⁶-mG were determined by an established LC-MS/ MS assay;¹⁶ active repair of 7-mG in these cell lines was minimal in comparison to O⁶-mG. Some loss of 7-mG could occur from spontaneous depurination. Preliminary studies demonstrated that the ratio of 7-mG/O⁶-mG plateaued 6 h after a 1 h treatment with 10 μM NMUr (Figure 5), so we chose to measure DNA adduct levels at 0 and 6 h after the 1 h treatment with NMUr for the DNA repair studies.

Figure 5.

Time dependent change of the levels of (A) 7-mG, (B) O⁶-mG, and (C) ratio of 7-mG/O⁶-mG in two different cell lines. Cells were collected for DNA isolation 0, 4, 6, and 24 h after a 1 h treatment with 10 μM NMUr. The 7-mG and O⁶-mG levels were measured by LC-MS/MS and normalized to guanine. Data is shown as average ratio ± SD (n = 3).

Treatments for DNA adduct levels were performed concurrently with the cytotoxicity and mutagenicity levels. The levels of 7-mG and O⁶-mG levels varied significantly from cell line to cell line (Table S5), but some trends became apparent when separating the two ethnic groups. The CEU cell lines had higher levels of 7-mG at both time points as compared with the YRI cell lines (Table 1). Similarly, O⁶-mG levels in the CEU cell lines were also higher at both time points as compared with the YRI cell lines (Table 1). However, the ratio of 7-mG to O⁶-mG was not significantly different between the two groups at either time point (Figure 6 and Table 1). This indicates that the CEU cell lines generated greater levels of NMUr-derived DNA methylation than the YRI cell lines, but the average rate of repair was similar between the two groups.

Figure 6.

Ratio of 7-mG/O⁶-mG at 6 h following a 1 h exposure of lymphoblastoid cell lines to NMUr as a measure of O⁶-mG repair. Cells were exposed to NMUr for 1 h. Excess NMUr was removed by the addition of esterase, and the cells were collected for DNA isolation 6 h later. Levels of 7-mG and O⁶-mG were measured by the LC/MS/MS assay described in the Materials and Methods section. The dotted line indicates the population average.

The Repair of O⁶-mG Is Correlated to NMUr Cytotoxicity in CEU but Not YRI Cell Lines.

The data sets from the two populations were initially pooled to investigate the correlation among the various experimental end points. IC₂₀ values were correlated with the levels of O⁶-mG at both 0 and 6 h (−0.66 and −0.74, respectively) as well as the ratio of 7-mG/O⁶-mG at 0 and 6 h (0.67 and 0.75, respectively). Correlations were also observed between the IC₅₀ values and the levels of O⁶-mG at both 0 and 6 h (−0.53 and −0.63, respectively) as well as the ratio of 7-mG/O⁶-mG at 0 and 6 h (0.55 and 0.63, respectively). There was also a correlation between the percent survival at 10 μM NMUr and the ratio of 7-mG/O⁶-mG at 6 h (0.56). There was no correlation of any of the cytotoxicity or DNA adduct/repair end points with mutation frequency.

When a second analysis was performed separately for each population, significant population differences were observed (Tables S6 and S7). In the CEU cell lines, the correlation between IC₂₀ values and the levels of O⁶-mG at 0 and 6 h were −0.78 and −0.83, respectively (p value = 3.1 × 10⁻⁷ and 6.5 × 10⁻⁹, respectively); these correlations were −0.26 and −0.45, respectively (p value = 0.066 and 0.0075, respectively) in the YRI cell lines. There was also a strong correlation between IC₂₀ values and the ratio of 7-mG/O⁶-mG at 0 and 6 h in the CEU cell lines (0.71 and 0.82, respectively; p values = 7.7 × 10⁻⁶ and 2.4 × 10⁻⁸, respectively). This correlation was not as significant in the YRI cell lines (0.49 and 0.46, respectively; p values <0.01). Similarly, there were population differences in the correlations observed between IC₅₀ and O⁶-mG levels or the ratio of 7-mG/O⁶-mG at 0 and 6 h. The levels of O⁶-mG at 0 and 6 h were strongly correlated to NMUr IC₅₀ values in the CEU cell lines (−0.75 and −0.85, respectively; p values <1 × 10⁻⁷). The correlations between the adduct ratio at 0 and 6 h and IC₅₀ values were 0.66 and 0.77, respectively, in the CEU cell lines (p value <1 × 10⁻⁵). In contrast, correlations between O⁶-mG levels or adduct ratios at either time point were not significantly correlated to NMUr IC₅₀ values in the YRI cell lines (Table S7).

There was no correlation of any of the cytotoxicity or DNA adduct/repair end points with mutation frequency in either cell line.

Several Genome-Wide Associations Were Observed for the Toxicological End Points.

The publically available SNP HapMap data were used to determine if there were SNPs that were significantly associated with any of the measured end points (p value < 5 x 10⁻⁸). This analysis was performed separately on the two populations. For the CEU cell lines, a number of SNPs achieved genome-wide significance for a variety of end points (Table 2, Figure S2 and S3). The cytotoxicity end points IC₂₀ and IC₅₀ had 5 and 51 significant SNPs, respectively. Four of these SNPs were overlapping; rs11258248 and rs3781080 mapped to the MCM10 gene on chromosome 10 and rs6848554 and rs6823445 mapped to LOC107986236 on chromosome 4 (Figure 7). MCM10 (Minichromosome Maintenance 10 Replication Initiation Factor) is an highly conserved protein involved in the initiation of genome replication.³⁸ LOC107986236 is an uncharacterized noncoding RNA gene. There were 17 SNPs with the same region of chromosome 4 assigned to LINC02355 that were associated with IC₅₀; this region had the most significant p-value and codes for a long intergenic nonprotein coding RNA.

Table 2.

SNPs Identified in the GWAS Analysis of the CEU Lymphoblastoid Cell Lines^a

outcome	SNP	CHR	chromosome base position	N	t test statistic	P-value	gene	MAF CEU	MAF YRI
IC20	rs11258248	10	13200379	31	7.7825	2.3 × 10−08	MCM10	0.417	0.466
IC20	rs3781080	10	13203753	31	−7.7825	2.3 × 10−08	MCM10	0.417	0.466
IC20	rs6848554	4	132040717	31	7.6367	4.2 × 10−08	LOC107986236	0.431	0.500
IC20	rs6823445	4	132041142	31	−7.9072	1.3 × 10−08	LOC107986236	0.435	0.467
IC20	rs11982165	7	95650656	31	7.8835	1.8 × 10−08	LOC107986746	0.417	0.362
IC50	rs11258248	10	13200379	31	7.5359	4.2 × 10−08	MCM10	0.417	0.466
IC50	rs3781080	10	13203753	31	−7.5359	4.2 × 10−08	MCM10	0.417	0.466
IC50	rs1251410	10	31953101	31	−8.9678	1.0 × 10−09	LOC107984219	0.452	0.500
IC50	rs2039617	10	95999446	31	8.8902	1.7 × 10−09	CC2D2B	0.450	0.310
IC50	rs2429169	12	1920571	31	8.4654	8.3 × 10−09	CACNA2D4	0.464	0.500
IC50	rs6489697	12	6159183	31	7.5692	3.8 × 10−08	#N/A	0.450	0.500
IC50	rs6575816	14	101074680	31	8.5268	1.4 × 10−08	LOC105370670	0.462	0.241
IC50	rs1423947	16	60105993	31	8.3438	4.5 × 10−09	#N/A	0.468	0.433
IC50	rs12462702	19	16138163	31	−10.2135	6.4 × 10−09	HSH2D	0.476	0.483
IC50	rs6001605	22	39490212	31	−8.6996	1.9 × 10−09	MGAT3	0.435	0.433
IC50	rs12157593	22	39492856	31	−8.6488	2.9 × 10−09	#N/A	0.433	0.431
IC50	rs6001609	22	39496854	31	−8.6488	2.9 × 10−09	#N/A	0.433	0.345
IC50	rs1435603	2	141911113	31	−8.6090	6.0 × 10−09	LRP1B	0.464	#N/A
IC50	rs10180758	2	220735307	31	9.1207	9.5 × 10−09	#N/A	0.458	0.440
IC50	rs2241962	3	123102631	31	−8.3603	2.0 × 10−08	PDIA5	0.442	0.283
IC50	rs6848554	4	132040717	31	7.6045	4.5 × 10−08	LOC107986236	0.431	0.500
IC50	rs6823445	4	132041142	31	−7.9134	1.3 × 10−08	LOC107986236	0.435	0.467
IC50	rs10027129	4	149192371	31	−9.1272	1.4 × 10−09	LINC02355	0.448	0.397
IC50	rs10027206	4	149192466	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs10027376	4	149192626	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs17025850	4	149193527	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.379
IC50	rs10032896	4	149193908	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs17025856	4	149198128	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.375
IC50	rs17025858	4	149200362	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs9993655	4	149200870	31	8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs17025897	4	149208021	31	8.9678	1.0 × 10−09	LINC02355	0.452	0.448
IC50	rs1154734	4	149210753	31	8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs10015540	4	149213378	31	8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs17025934	4	149214172	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs17025936	4	149214505	31	−8.8902	1.7 × 10−09	LINC02355	0.450	0.431
IC50	rs10021160	4	149214871	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.397
IC50	rs10031174	4	149218498	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.379
IC50	rs9996888	4	149233761	31	8.9678	1.0 × 10−09	LINC02355	0.452	0.379
IC50	rs10212691	4	149236287	31	−8.9678	1.0 × 10−09	LINC02355	0.452	0.383
IC50	rs13117940	4	179076089	31	8.9678	1.0 × 10−09	#N/A	0.452	0.500
IC50	rs2624445	5	117412262	31	−9.0697	1.1 × 10−09	#N/A	0.450	0.367
IC50	rs2560578	5	117412288	31	−9.0697	1.1 × 10−09	#N/A	0.450	0.367
IC50	rs2624444	5	117412592	31	−10.8119	6.5 × 10−11	#N/A	0.446	#N/A
IC50	rs2624538	5	117422976	31	−7.7139	3.5 × 10−08	LINC00992	0.431	0.367
IC50	rs2560533	5	117426811	31	7.5046	3.6 × 10−08	LINC00992	0.452	0.367
IC50	rs2624508	5	117462598	31	−9.0697	1.1 × 10−09	LINC00992	0.450	0.433
IC50	rs17576261	6	15382549	31	8.8902	1.7 × 10−09	JARID2	0.450	0.500
IC50	rs9367745	6	60351710	31	8.4731	5.9 × 10−09	#N/A	0.466	0.328
IC50	rs1116886	6	58064487	31	−8.5905	3.3 × 10−09	LOC101927293	0.467	0.426
IC50	rs9404733	6	58118888	31	9.0265	2.4 × 10−09	LOC101927293	0.464	0.467
IC50	rs9404736	6	58153188	31	−8.3137	1.6 × 10−08	LOC101927293	0.463	0.463
IC50	rs10244385	7	13226870	31	−7.5666	3.1 × 10−08	#N/A	0.452	0.483
IC50	rs9769512	7	72407765	31	−8.4453	6.3 × 10−09	CALN1	0.466	0.286
IC50	rs13264649	8	41409497	31	−8.4430	6.3 × 10−09	#N/A	0.466	0.467
IC50	rs1414141	9	115290467	31	−8.8902	1.7 × 10−09	DEC1	0.450	0.286
IC50	rs4978648	9	115712500	31	−7.5046	3.6 × 10−08	#N/A	0.452	0.467
IC80	rs6504860	17	53800974	31	−8.0909	2.6 × 10−08	#N/A	0.389	#N/A
IC80	rs6504863	17	53801314	31	7.8304	2.0 × 10−08	#N/A	0.383	0.259
IC80	rs2331866	17	53801810	31	7.8304	2.0 × 10−08	#N/A	0.383	0.204
log MF at 20	rs11904059	2	8233267	31	7.8728	1.8 × 10−08	LINC00299	0.242	0.283
MF20v0	rs4956349	4	141147319	31	7.8466	1.9 × 10−08	RNF150	0.267	0.357
MF20v0	rs11755314	6	76630431	31	−9.8426	2.0 × 10−10	#N/A	0.383	0.431
MF20v10	rs7649990	3	54187304	31	−8.0610	1.2 × 10−08	CACNA2D3	0.333	0.414
MF20v10	rs6445632	3	54188432	31	−8.0610	1.2 × 10−08	CACNA2D3	0.333	0.414
MF20v10	rs4076708	3	54190793	31	−8.1967	6.4 × 10−09	CACNA2D3	0.339	0.450
MF20v10	rs11717730	3	54191357	31	8.1967	6.4 × 10−09	CACNA2D3	0.339	0.450
O6-mG 0 h	rs11982165	7	95650656	31	−7.9845	1.4 × 10−08	LOC107986746	0.417	0.362
O6-mG 6 h	rs11982165	7	95650656	31	−8.1687	9.0 × 10−09	LOC107986746	0.417	0.362

^aMF = Mutants per 10⁶ cells. MF10v0 = difference between MF 10 μM and 0 μM NMUr. MF20v0 = difference between MF 20 μM and 0 μM MNUr. MF20v10 = difference between MF 20 μM and 10 μM NMUr.

Figure 7.

Representative box plots of SNPs significantly associated with (A) IC₂₀ and (B) IC₅₀ in CEU cell lines.

Other genes with significant SNPs associated with the IC₅₀ end point in the CEU cell lines include DEC1 (rs1414141), JARID2 (rs17576261), CC2DSB (rs2039617s), MGAT3 (rs6001605), LRP1B (rs1435603), CALN1 (rs9769512), HSH2D (rs12462702), CACNA2D4 (rs2429169), and PDIA5 (rs2241962). JARID2 (Jumonji And AT-Rich Interaction Domain Containing 2) plays a role in regulating histone methylation and contributes to the negative regulation of cell proliferation signaling.³⁹ The function of DEC1 (Deleted in Esophageal Cancer 1) is unknown; however, it is thought to be a tumor suppressor gene.⁴⁰ LRP1B (LDL Receptor Related Protein 1B) is a low density lipoprotein receptor with possible tumor suppressor activity;⁴¹ its deletion is associated with chemotherapy resistance of ovarian cancers.⁴² CALN1 (Calneuron 1) is a calcium-binding protein.⁴³ HSH2D (Hematopoietic SH2 Domain Containing) is a target of the signaling pathways in T-cell activation and may influence apoptosis.^44,45 CACNA2D4 (Calcium Voltage-Gated Channel Auxiliary Subunit Alpha2delta 4) is part of the voltage-dependent calcium channel complex and regulates the kinetics of calcium channel activation/inactivation.⁴⁶ PDIA5 (Protein Disulfide Isomerase Family A Member 5) facilitates protein folding and thiol-sulfide interchange reactions and participates in the unfolded protein response.^47,48

Several SNPs were associated with a couple of the NMUr derived mutagenesis end points in the CEU cell lines. Two SNPs were associated with the difference in mutations caused by 20 μM NMUr relative to untreated controls; one SNP (rs4956349) is an intronic SNP for RNF150 (Ring Finger Protein 150) and the other is in an unassigned region of DNA. The difference between mutations at 10 and 20 μM NMUr was associated with 4 SNPs in the CACNA2D3 (Calcium Voltage-Gated Channel Auxiliary Subunit Alpha2delta 3) gene on chromosome 3. Like CACNA2D4, it is part of the voltage-dependent calcium channel complex and regulates the kinetics of calcium channel activation/inactivation.⁴⁹

One SNP was associated with the levels of O⁶-mG at both 0 and 6 h after a 1 h exposure to 10 μM NMUr. This SNP is an intronic variant of LOC107986746, a gene on chromosome 7 that codes for an uncharacterized noncoding RNA.

Fewer SNPs reached the level of genome-wide significance in the YRI cell lines (Table 3 and Figures S4 and S5). The SNP rs10941686, which is an intron variant for CDH18, was significantly associated with the IC₂₀ end point (Figure 8a); cadherin 18 is a member of the cadherin superfamily of integral membrane proteins that facilitates calcium-dependent cell–cell adhesion.⁵⁰

Table 3.

SNPs Identified in the GWAS Analysis of the YRI Lymphoblastoid Cell Lines

outcome	SNP	CHR	chromosome base position	N	t test statistic	P-value	gene	MAF CEU	MAF YRI
IC20	rs10941686	5	20284829	30	−7.8518	2.5 × 10−08	CDH18	0.250	0.293
IC80	rs10267660	7	24702895	30	−8.2682	3.4 × 10−08	GSDME	0.483	0.352
O6-mG 6h	rs1762436	10	129469457	30	−8.3437	8.0 × 10−09	MGMT	0.500	0.414
O6-mG 6h	rs7136711	12	89821844	30	−8.1145	1.8 × 10−08	#N/A	0.242	0.393
O6-mG 6h	rs11876394	18	7650692	30	8.1428	9.6 × 10−09	PTPRM	0.468	0.317
O6-mG 6h	rs10207395	2	30956931	30	−7.5009	4.5 × 10−08	GALNT14	0.483	0.450
O6-mG 6h	rs2694045	8	115811975	30	7.6596	3.1 × 10−08	LOC107986967	0.242	0.400

Figure 8.

Representative box plots of SNPs significantly associated with (A) IC₂₀ and (B) O⁶-mG levels at 6 h after a 1 h exposure to NMUr in YRI cell lines.

There were 5 SNPs associated with the levels of O⁶-mG at 6 h after NMUr treatment in the YRI cell lines. One of these SNPs is in the MGMT gene, which codes for the protein that repairs O⁶-mG (Figure 8b). Other genes with associated SNPs are PTPRM and GALNT14 as well as uncharacterized regions of DNA on chromosomes 12 and 8. PTPRM (Protein Tyrosine Phosphatase, Receptor Type M) plays a role in cell–cell adhesion.⁵¹ GALNT14 (Polypeptide N-Acetylgalac-tosaminyltransferase 14) catalyzes the transfer of N-acetyl-d-galactosamine to serine or threonine residues on mucins; changes in this protein can affect response to therapy or play role in cancer progression.⁵²

We then interrogated the GWAS data to determine the association of the GWAS significant SNPs with the other outcomes (Tables S8 and S9). In the CEU cell lines, many of the SNPs that were highly associated with the IC₂₀ or IC₅₀ end points were also associated with p < 1 × 10⁻⁵ with the IC₅₀ or IC₂₀ end point, respectively. The direction of the association was the same (if positive, positive; if negative, negative). The highly significant SNPs for these two outcomes were also tended to be associated with the levels of O⁶-mG at 0 and 6 h after a 1 h exposure to 10 μM NMUr, with the more significant association with the data at 6 h (p < 1 × 10⁻⁵). In this case, the association is opposite: when the toxicity data is positively associated with the genetic variation, the O⁶-mG adduct levels are negatively associated with the same genetic variation. This result is expected because higher levels of O⁶-mG are associated with increased toxicity (lower IC₂₀ or IC₅₀ concentrations).

There was no overlap between the significant SNPs or genes identified in the GWAS analysis between the two populations. For some of the SNPS, the cause is likely genetic since all the cell lines employed in the CEU or YRI populations had the exact same genotype for the SNP (i.e., rs1762436 in the CEU LCLs and rs1251410 in the YRI LCLs). However, for the majority of the SNPs, the significant association between SNP and outcome were only observed in one of the populations. Similar observations have been reported for other toxic chemicals.⁵³ Only one of the highly significant SNPs identified for the IC₅₀ end point in the CEU cell lines (rs2039617) was associated with the same end point in YRI cells (Table S8); the p value of 0.0017 was not significant when adjusted for multiple comparisons. Similarly, only one SNP, rs10941686, significantly associated with IC₅₀ in the YRI cell lines was also associated with IC₅₀ levels in the CEU cell lines (Table S9); the p value of this association (0.0037) was not significant when adjusted for multiple comparisons.

DISCUSSION

Sixty one EBV-transformed lymphoblastoid cell lines (31 CEU and 30 YRI) from the HapMap resource were screened for their sensitivity to the cytotoxic and mutagenic effects of a model methylating agent MNUr as well as for their ability to repair O⁶-mG. These cell lines showed a variation in the response to these toxicological outcomes with the CEU cell lines showing a much wider range than the YRI cell lines.

According to the pathways outlined in Scheme 1, we expected to see a correlation between the levels of O⁶-mG or the ratio of 7-mG/O⁶-mG and the cytotoxic end points. The levels of O⁶-mG or the ratio of 7-mG/O⁶-mG at both 0 and 6 h following a 1 h exposure to 10 μM MNUr were strongly correlated to the IC₂₀ values in both populations with a stronger correlation observed in the CEU cell lines. A similar strong correlation was observed between O⁶-mG or the ratio of 7-mG/O⁶-mG and IC₅₀ in the CEU cell lines but not the YRI cell lines. This confirms that O⁶-mG is the most cytotoxic DNA adduct formed by the reactive metabolite of MNUr.^16,22 The lower degree of correlation in the YRI cell lines suggests that other factors are contributing to the cytotoxic effects of NMUr in these cell lines. One possibility is that the YRI cell lines are more sensitive to another methyl DNA adducts such as 3-methyladenine. This adduct is a block to DNA replication and is cytotoxic.^22,54 Future studies will explore if it is better correlated to the cytotoxic properties of NMUr than O⁶-mG in the YRI cell lines.

Scheme 1 also predicts an association between levels of O⁶-mG or DNA repair ratios caused by NMUr and NMUr-derived mutation frequency. This relationship was not observed. Nor were there correlations between any of the cytotoxicity outcomes and mutation frequency as anticipated by the mechanism presented in Scheme 1. A likely reason for this is that the concentrations used in the mutation assay (0, 10, or 20 μM NMUr) did not increase the number of mutations in more than half of the cell lines. Future studies could determine if the use of higher concentrations of MNUr would allow for an association to be detected.

The variations between cell lines can be caused by genetic and nongenetic factors. The GWAS analysis identified several SNPs and genes associated with the different toxicological end points, demonstrating that genetic factors contribute to the variation we observed. Because of the small sample size, we were only able to detect significant effects for SNPs that had a high minor allele frequency.

In the CEU cell lines, a number of SNPs were associated with genes coding for proteins that may play important roles in the cellular response to the toxic effects of NMUr. The SNPs, rs11258248 and rs3781080, associated with both IC₂₀ and IC₅₀ are in intronic regions of the MCM10 gene. This gene codes for a protein that is essential in chromosomal replication, checkpoint signaling and DNA damage response.^38,55 Since replication activity is important for the toxic effects of O⁶-mG,⁵⁶ the involvement of MCM10 in the toxicity of methylating agents requires further exploration. Other interesting genes associated with the IC₅₀ end point in the CEU cell lines include JARID2, DEC1, LRP1B, CALN1, CACNA2D4, and PDIA5. Most of these proteins participate in cellular activities that could influence the cell’s response to a cytotoxic agent. Their participation in the toxicity of NMUr and other methylating agents should be targeted for further investigation.

A SNP connected with the MGMT gene was associated with the levels of O⁶-mG at 6 h in the YRI cell lines. This SNP was not detected in the CEU cell lines since every cell line tested was genetically identical at this location. A previous study in lymphoblastoid cell lines with temozolamide, a chemotherapy drug that methylates DNA in a manner similar to NMUr, identified a range of SNPs that were associated with temozolamide toxicity.⁵⁷ The most significant one, rs531572, did not reach genome-wide significance in our study, most likely because of our small sample size (MAF: CEU, 0.30; YRI, 0.31).

The SNPs identified in our study were unique to each population. This observation is not surprising since there are significant differences in gene expression between the two populations of cell lines for both genetic and nongenetic reasons.⁵⁸‘⁵⁹ One nongenetic factor is the age of the cell lines. The CEU cell lines were collected and transformed more than 20 years earlier than the YRI cell lines so differences in the EBV strains used for transformation as well as the number of freeze/thaw cycles could contribute to nongenetic differences between the two populations.⁶⁰ Analysis of genome-wide RNA sequencing data demonstrated that cell line age significantly impacts gene expression, indicating that some cell line differences are not due to intrinsic population characteristics.⁶¹ Furthermore, it has long been recognized that EBV-transformation of lymphocytes can affect gene expression;⁶² MGMT expression can be affected by this process.⁶³

Our study, despite its small sample size, demonstrates that there is substantial variability to the toxic and genotoxic properties of a DNA methylating agent and that there are some genetic variations that drive this variability. Our experimental approach, while labor intensive, provides information as to how an individual responds to an alkylating agent that simultaneously produces a number of cytotoxic or mutagenic DNA adducts. For example, O⁶-mG appeared to be the main driver of NMUr-derived cytotoxicity in CEU cell lines, whereas it was less important to the toxic effects observed in YRI cell lines. GWAS analysis identified several novel genes to be explored in future functional validation studies. This study represents the first step toward our goal to identify how genetic variation in DNA damage repair and response genes influence cancer risk associated with tobacco smoke exposure. It revealed that the individual biological response to a single DNA-damaging agent was substantially more diverse than predicted.

ABBREVIATIONS

7-mG

7-methylguanine

APC

allophycocyanin

CACNA2D3

calcium voltage-gated channel auxiliary subunit alpha2delta 3

CACNA2D4

calcium voltage-gated channel auxiliary subunit alpha2delta 4

CALN1

calneuron 1

CDH18

cadherin 18

CEU

Utah residents with Northern and Western European ancestry

DEC1

deleted in esophageal cancer 1

DMN

N-nitrosodimethylamine

EBV

Epstein–Barr virus

FACS

fluorescence-activated cell sorting

GALNT14

polypeptide N-acetylgalactosaminyltransferase 14

GPI

glycophosphatidylinositol

GWAS

genome-wide association studies

HR

homologous recombination

HSH2D

hematopoietic SH2 domain containing

JARID2

Jumonji and AT-rich interaction domain containing 2

LRP1B

low density lipoprotein receptor related protein 1B

MCM10

minichromosome maintenance 10 replication initiation factor

MGMT

O⁶-methyl DNA methyltransferase

NMUr

N-nitroso-N-methylurethane

NNK

4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone

O⁶-mG

O⁶-methylguanine

PDIA5

(Protein Disulfide Isomerase Family A Member 5)

PE

phycoerythrin

PIG-A

phosphatidylinositol-glycan biosynthesis class-A

PTPRM

Protein Tyrosine Phosphatase, Receptor Type M

RNF150

Ring Finger Protein 150

SNP

single nucleotide polymorphism

YRI

Yoruban