Genetic variation and antioxidant response gene expression in the bronchial airway epithelium of smokers at risk for lung cancer

Abstract

Prior microarray studies of smokers at high risk for lung cancer have demonstrated that heterogeneity in bronchial airway epithelial cell gene expression response to smoking can serve as an early diagnostic biomarker for lung cancer. As a first step in applying functional genomic analysis to population studies, we have examined the relationship between gene expression variation and genetic variation in a central molecular pathway (NRF2-mediated antioxidant response) associated with smoking exposure and lung cancer. We assessed global gene expression in histologically normal airway epithelial cells obtained at bronchoscopy from smokers who developed lung cancer (SC, n = 20), smokers without lung cancer (SNC, n = 24), and never smokers (NS, n = 8). Functional enrichment analysis showed that the NRF2-mediated, antioxidant response element (ARE)-regulated genes, were significantly lower in SC, when compared with expression levels in SNC. Importantly, we found that the expression of MAFG (a binding partner of NRF2) was correlated with the expression of ARE genes, suggesting MAFG levels may limit target gene induction. Bioinformatically we identified single nucleotide polymorphisms (SNPs) in putative ARE genes and to test the impact of genetic variation, we genotyped these putative regulatory SNPs and other tag SNPs in selected NRF2 pathway genes. Sequencing MAFG locus, we identified 30 novel SNPs and two were associated with either gene expression or lung cancer status among smokers. This work demonstrates an analysis approach that integrates bioinformatics pathway and transcription factor binding site analysis with genotype, gene expression and disease status to identify SNPs that may be associated with individual differences in gene expression and/or cancer status in smokers. These polymorphisms might ultimately contribute to lung cancer risk via their effect on the airway gene expression response to tobacco-smoke exposure.

Figure 1. The workflow for examining the relationship between gene expression variation and genetic variation in the NRF2-mediated antioxidant response pathway associated with smoking exposure and lung cancer.

(A) We assessed microarray gene expression profiles of histologically normal airway epithelial cells obtained by bronchoscopy from smokers with suspicion of lung cancer and from a control group of never smokers. (B) We identified that the antioxidant response pathway regulated by the transcription factor NRF2 differed among these groups of subjects. We found that the expression of MAFG (a binding partner of NRF2) was correlated with the expression of NRF2 pathway genes. (C) Bioinformatics strategies were used to identify putative regulatory SNPs in NRF2 binding sites and to select tagging SNPs for NRF2-mediated genes. (D) The MAFG locus was sequenced in our study subjects. We identified SNPs that were associated with individual differences in: (E) gene expression and/or (F) cancer status by integrating gene expression, genotype and lung cancer status.

Figure 2. Pathway analysis and NRF2 pathway gene expression.

(A) Ingenuity Pathway Analysis revealed differential effects of cigarette smoking in smokers without cancer (SNC) and smokers with cancer (SC) relative to nonsmokers (NS). Six canonical pathways were significant enriched in the 210 over-expressed probe-sets when comparing SNC with NS (blue bars). Three pathways were significantly enriched in the 263 over-expressed probe-sets when comparing SC versus NS (red bars). We observed a very pronounced difference in the significance level of “NRF2-mediated oxidative stress pathway” between SNC and SC. (B) The mRNA levels of the interacting, regulatory components of the NRF2 pathway displayed as boxplots. A boxplot depicts a dataset through five-number summaries: the smallest observation, lower quartile, median, upper quartile, and largest observation. Compared with NS, the master regulator NRF2 showed no difference between SNC and SC. However, the binding partner MAFG was significantly lower in SC, and the competitor NRF1 was significantly higher in SC. NRF3, KEAP1, and BACH1 mRNA showed no significant changes. (C) We observed that 22 genes with known NRF2 binding sites showed significant differences among the groups at FDR 0.1 level. Consistently, the expression of these genes in SNC was higher than that in NS; and most of these genes have lower expression in SC than that in SNC. This pattern was similar to MAFG expression pattern.

Figure 3. MAFG silencing attenuates downstream antioxidant and Phase II gene expression.

Following transient transfection with MAFG siRNA in the A549 airway cell line, gene expression was measured using real-time qPCR. Transfection with scrambled control siRNA produced a general increase in NRF2 pathway genes (black bars) relative to nontransfected cells (set at 100%). (A) MAFG gene expression was significantly reduced (55%) compared to non-specific siRNA control. (B) GCLC, NQO1, SLC7A11, and TXNRD1 gene expression was significantly reduced with MAFG silencing compared to non-specific siRNA controls. * (p≤0.05, t-test). All data presented as mean ± SEM (n = 3).

Figure 4. SNPs associated with the expression of 3 known ARE genes, AKR1C1, AKR1C2, and AKR1C3 among smokers.

We genotyped 25 SNPs in the genomic region containing AKR1C1, AKR1C2, and AKR1C3. In each plot of expression verse genetype, circles were log2 expression, and lines were linear regression trend lines. The genotypes of SNP rs1241488 associated with the expression levels of all 3 genes. The genotypes of SNP rs1090439 associated with the expression levels of AKR1C1 and AKR1C2. Note: The SNP rs17134158 only had 2 genotypes AG and GG in smokers, but did have AA genotype in NS and its overall minor allele frequency >0.05. Linkage disequilibrium analyses indicated these 3 SNPs were in linkage (r²>0.88).

Figure 5. Plots of association between SNP genotype and gene expression for selected SNPs.

In each plot, SC, SNC and NS are colored with red, blue, and black, respectively. (A) The association of a putative ARE SNP rs3753660 in the promoter of EPHX1 gene displays distinct trends in SC and SNC; (B) DUSP1 putative ARE SNP rs17658295 associated with its expression. The minor allele was significantly associated with higher expression and the significance level was more pronounced in the cancer group; C) GCLC intronic SNP rs670548 minor allele associated with lower expression among all subjects; (D) GCLC 3′ downstream SNP rs2397146 minor allele associated with higher expression among all subjects.

Figure 6. An example of how our three-part approach for association analysis of expression, genotype, and phenotype data may reveal biologically plausible SNPs.

The MAFG 3′UTR SNP at chr17:77469864 may potentially contribute to phenotype (lung cancer status) via gene expression, based on: (A) Expression of MAFG was higher in smokers without cancer than that in smokers with cancer; (B) Genotype GG displays a trend toward higher expression levels of MAFG (#, black line); slopes of genotype by expression plots differ between groups (* red dotted line vs blue dotted line); (C) Genotype GG, associated with higher expression, was more common in smokers without cancer; (D) Hypothesis. Individuals with genotype GG display higher MAFG expression in bronchial epithelial cells; MAFG expression higher in smokers without cancer suggesting it is protective against lung cancer; GG genotype is less frequent among cancer group, consistent with a protective effect.