|Home | About | Journals | Submit | Contact Us | Français|
Epidemiological studies of underground miners suggested that occupational exposure to radon causes lung cancer with squamous cell carcinoma (SCC) as the predominant histological type. However, the genetic determinants for susceptibility of radon-induced SCC in miners are unclear. Double-strand breaks induced by radioactive radon daughters are repaired primarily by non-homologous end joining (NHEJ) that is accompanied by the dynamic changes in surrounding chromatin, including nucleosome repositioning and histone modifications. Thus, a molecular epidemiological study was conducted to assess whether genetic variation in 16 genes involved in NHEJ and related histone modification affected susceptibility for SCC in radon-exposed former miners (267 SCC cases and 383 controls) from the Colorado plateau. A global association between genetic variation in the haplotype block where SIRT1 resides and the risk for SCC in miners (P = 0.003) was identified. Haplotype alleles tagged by the A allele of SIRT1 rs7097008 were associated with increased risk for SCC (odds ratio = 1.69, P = 8.2×10−5) and greater survival in SCC cases (hazard ratio = 0.79, P = 0.03) in miners. Functional validation of rs7097008 demonstrated that the A allele was associated with reduced gene expression in bronchial epithelial cells and compromised DNA repair capacity in peripheral lymphocytes. Together, these findings substantiate genetic variation in SIRT1 as a risk modifier for developing SCC in miners and suggest that SIRT1 may also play a tumor suppressor role in radon-induced cancer in miners.
Radon is an inert gas released during the decay of radium-226, the fifth daughter of uranium-238. Radon is ubiquitous in indoor and outdoor air and contaminates many underground mines (1). Epidemiological studies of underground miners suggested that occupational exposure to radon causes lung cancer in smokers and never smokers (2,3). Moreover, exposure to radon and tobacco smoke through uranium mining is associated with an additive to multiplicative increase in lung cancer risk (4). Underground uranium miners have elevated risk for lung cancer of all histology types with squamous cell carcinoma (SCC) as the predominant histological type (4,5). The predominance of SCC in uranium miners is most likely due to the fact that radon daughters and tobacco carcinogens that attach to the silica dust or diesel emission particles deposit at the highest concentration in the bronchial epithelium, the site for SCC development (1,6,7).
The radon daughters generated from the decay of radon emit α particles whose high energy damages DNA to cause mainly double-strand breaks (DSBs) (1,8). Carcinogens and mutagens within tobacco smoke can also either directly or indirectly induce a wide spectrum of DNA damage that includes DSBs (9). DSBs represent one of the most detrimental forms of DNA damage and if not properly repaired, it can trigger cell death or malignant transformation (10). DSBs are repaired in mammalian cells primarily by non-homologous end joining (NHEJ), in which the two broken parts of the chromosomes are ligated (11).
Genomic DNA and histones form a highly condensed structure known as chromatin. The aforementioned DNA repair requires dynamic changes in surrounding chromatin, including changes in nucleosome repositioning and histone modifications (10,12). The best-characterized chromatin alteration in DNA repair is the phosphorylation of the histone variant H2AX (γ-H2AX) by DNA damage response protein kinases (13). This modification helps to stabilize the interaction of repair factors (e.g. MRE11A) within the break sites by affecting chromatin configuration. Histone acetylases and deacetylases also localize to sites of DNA damage to facilitate repair by increasing access of repair proteins to the break site, repressing transcription at sites of damage, restoring the local chromatin structure after repair is done and turning off the DNA damage response (12,14).
Suboptimal DNA repair capacity (DRC) for DSBs may be used as an inherent susceptibility factor for lung cancer risk assessment in smokers. Support for this supposition was provided through several studies from our group and others (15–18). Enzymatic activity of DNA-dependent protein kinase catalytic subunit (a key player in DNA repair, which is activated by DSBs and NHEJ phosphorylating proteins) in peripheral mononuclear cells and bronchial epithelial cells is inversely associated with risk for lung cancer (15). Promoter methylation of tumor suppressor genes detected in sputum from smokers provides an assessment of the extent of field cancerization that in turn predicts early lung cancer (16). A highly significant association was observed between DRC for DSBs measured in lymphocytes and the propensity for gene methylation detected in sputum from cancer-free smokers from the Lovelace Smokers Cohort (LSC) (17). A significant association between DSB DRC and risk for lung cancer was further observed in peripheral lymphocytes collected 0.3–6 years prior to the lung cancer diagnosis (18). The fact that uranium miners are exposed to a high level of radiation and tobacco carcinogens and DSB DRC is a risk modifier for lung cancer in smokers led us to conduct a molecular epidemiological study to assess whether genetic variants in genes involved in NHEJ and related histone modification affect susceptibility of radon-induced SCC in former miners from the Colorado plateau. The SNPs significantly associated with risk for SCC were also tested in relation to overall survival of SCC in miners. The functional validation of the SNPs with the strongest evidence for association with the risk for SCC was conducted by examining their effect on gene expression and DSB repair capacity.
A nested case-control study was conducted in a cohort of approximately 17 000 radon-exposed workers who participated in sputum cytology screening for lung cancer detection between 1957 and 2002. These workers were former uranium miners who worked underground at the Colorado plateau. Approximately, 17 000 male workers contributed over 300 000 sputum specimens among which approximately 100 000 samples were longitudinal sputum specimens. All sputum samples were collected in Saccomanno fixative (alcohol [50%] and carbowax [2%]) for cytology screening to detect abnormal cells suggestive of carcinoma. The majority of the sputum cytology slides are archived at the St Mary’s Saccomanno Research Institute (SRI, Grand Junction, CO).
The St Mary’s Hospital Cancer Registry and SRI’s Cancer Research Database were used to identify uranium miners who participated in sputum cytology screening and were diagnosed with SCC. Life-time lung cancer-free miners with archived sputum slides collected during the period similar to that for the SCC cases were identified to comprise the control group. The miners selected as the controls have cumulative radon exposure at work expressed as working level month (WLM) and age at death similar to other life-time lung cancer-free uranium miners. All the study subjects were deceased. Year of birth, death and sputum collection, ethnicity, smoking history prior to the sputum collection (pack years and smoking status [current, former and never]) and WLM were obtained for most subjects. Because over 95% of miners were non-Hispanic white (NHW), the study was restricted to this ethnic group. Additional information including year of diagnosis and cause of death was obtained for SCC patients. WLM, a time-integrated measure, is the product of time in working months (1 month = 170h) and working levels. One working level equals any combination of radon progeny in 1 l of air that results in the ultimate emission of 130 000 MeV of energy from α particles. No personal identifiers accompanied the transfer of material from SRI to Lovelace Respiratory Research Institute. This study was conducted under an institutional review board-approved protocol.
Sputum cytology slides archived at the SRI from the patients identified above were selected for recovery of the genomic DNA. The sputum specimens proximal to cancer diagnosis or the last follow-up for controls were used for DNA isolations as described in the Supplemental Materials and methods, available at Carcinogenesis Online. DNA was successfully recovered from 267 cases and 383 controls.
Sixteen genes involved in NHEJ DNA repair (PRKDC, Ku70/XRCC6, Ku80/XRCC5, XRCC4, LIG4, Artemis/DCLRE1C and POLM), histone acetylation (CBP/CREBBP, p300/EP300, TIP60/HTATIP, GCN5/PCAF/KAT2B and MOF/MYST1) and histone deacetylation (HDAC1, HDAC2, SIRT1 and SIRT2) were selected (Supplemental Table I, available at Carcinogenesis Online). The linkage disequilibrium (LD) structure for common (minor allele frequency > 0.10) SNPs in the gene regions expanded by 20kb upstream and 10kb downstream was defined using the HapMap population with European ancestry by Haploview software (version 3 release R2). If the upstream or downstream boundaries were within a haplotype block, the corresponding boundary was further expanded to the boundary of the block. Tag SNPs were selected using the pairwise r 2 algorithm (r 2 > 0.8). For the bins with more than two SNPs, two SNPs were selected as the tags to minimize significant loss of the information due to genotyping failure and to serve as a quality assurance step by comparing their associations with risk for SCC. Non-synonymous SNPs were also included. All selected SNPs had design score >0.4 for the Illumina Golden Gate Genotyping platform. A 384-plex oligo pool assay (OPA) interrogating all the tag SNPs was designed and synthesized (Supplemental Table I, available at Carcinogenesis Online).
The performance of the OPA was initially tested in lymphocyte DNA samples (n = 182) from NHW LSC members who were previously genotyped using the Illumina Human Omni 2.5M chip (S Leng, unpublished data) . Genotype concordance for 206 shared SNPs on this OPA and the Illumina Human Omni 2.5M chip was 99.6%. Twelve SNPs that had disconcordant genotypes (>2%) between the platforms (n = 6), failed the Hardy-Weinberg Equilibrium test in the 182 DNA samples (P < 0.00013; n = 3) or had low SNP-wise call rate (<0.75; n = 3) were removed from further analysis (Supplemental Table I, available at Carcinogenesis Online). The DNA samples from 267 SCC cases and 383 controls were randomly ordered and loaded onto eight 96-well plates. Inter- and intra-plate duplicate DNA samples (n = 20 pairs) selected from the 182 LSR samples were also included for quality assurance. Samples were genotyped by the Illumina Golden Gate Genotyping assay on the BeadXpress platform. An additional 15 SNPs were removed from further analysis due to call rates <0.75 in miner DNA samples (Supplemental Table I, available at Carcinogenesis Online). The average agreement for genotypes for the 357 SNPs between duplicates was >99.4%.
Messenger RNA was isolated from primary normal human bronchial epithelial cell (NHBEC) cultures (n = 45), SCC and adenocarcinoma tumor normal pairs (n = 20 pairs each). These samples were collected from patients who are NHW. NHBEC cultures were established from airway biopsies obtained at bronchoscopy from current or former smokers. TaqMan real-time PCR was conducted to quantify gene expression in complementary DNA using the threshold cycle method with proliferating cell nuclear antigen (PCNA) and β-actin as the endogenous controls. PCNA and β-actin were highly correlated in the NHBEC samples (Pearson correlation coefficient = 0.86, P < 0.0001) (19).
A random block design was used to examine the association between the SIRT1 rs7097008 genotype and mutagen sensitivity in lymphocytes exposed to X-ray radiation. Cryopreserved lymphocytes (n = 34) were selected from NHW LSC members with the SIRT1 SNP (rs7097008) genotypes (A/A and C/C) that were matched 1:1 by age (every 10 years), gender, current smoking status and pack years. Phytohemagglutinin (M form) -stimulated lymphocytes were treated with 1.44 gray X-ray radiation to evaluate the generation of chromosome aberrations as an index of DSB DRC (17). The radiation dose was defined through dose-response studies using a lymphoblastoid cell line (GM00131). The dose selected caused obvious genotoxicity but minimal cytotoxicity. Metaphase spreading was made as described in ref. 17 and mutagen sensitivity was expressed as the mean number of chromatid breaks per 100 metaphases.
Quantile–quantile (Q–Q) plots were assessed for different thresholds to determine a cut-off for an acceptable sample-wise call rate (SCR). Deviation from the unit diagonal to the upper end of the plot indicates association with extreme P values that one would not expect under the null hypothesis accounting for multiple comparisons. The SNPs with extreme P values contain both real and false-positive associations. The extent of the latter is indicated by the reduction in departure from the null as the SCR increases.
Logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI) for the association between case-control status and each SNP using an additive inheritance model. Cumulative radon exposure (WLM) and smoking history are two major risk factors for SCC incidence in uranium miners. For the subjects evaluated in this study, smoking history included smoking status (current-, former- or never-smoker) and pack years at the time of sputum collection. WLM data were not reported for 31 SCC cases. Inclusion of the smoking history at sputum collection and cumulative WLM in the model did not affect the estimates for SIRT1 SNPs, thus, the results from models without adjustment for covariates were presented. False-positive report probability (FPRP) was calculated to address the robustness of the findings for individual SNPs that were most significantly (P < 0.01) associated with risk for SCC in miners (20). Prior probability was assigned to different SNPs based on their predicted functional potential assessed by searching the Functional SNP Prediction module of SNPinfo Web Server (21) and their association with disease phenotypes and/or functional readouts in the published work by others (22–28). Associations with FPRP ≤ 0.15 are considered noteworthy.
A competing risk model was used to derive the SNP-associated lung cancer-specific (HR) because some long-time SCC survivors died due to non-lung cancer causes (29). The P value and HR and its 95% CI were calculated using the Cox regression model with adjustment for age at lung cancer diagnosis. Tumor stage was not available for inclusion into the models. The Kaplan–Meier estimator was also used to estimate the survival function between genotypes. Because the survival functions showed departure from each other at early times, the Wilcoxon test that puts more weight on the earlier times rather than the later times were used in place of the log rank test to assess differences between genotypes.
The PHASE program was used to reconstruct the haplotypes and to calculate their estimated probabilities from the tag SNP data in a haplotype block (30). The probabilities of the common haplotypes for each individual were used as explanatory variables in the logistic or Cox regression models. Haplotypes with frequency <4% were combined into one group.
The association of mutagen sensitivity with the SIRT1 rs7097008 genotype was analyzed using a generalized linear mixed model to decompose the variation due to the genotyping of rs7097008, block factor (age, gender, current smoking status and pack years), freezer storage time, living cell counts and percent of viable cells. The values for the chromatid breaks were log transformed to satisfy the normality and homoscedasticity assumptions. The association between SIRT1 expression and number of C alleles for rs7097008 in primary NHBECs was analyzed using linear regression. Statistical analyses were conducted using SAS 9.2 and PLINK (31).
Approximately, half of the DNA samples generated from the 650 miners had SCRs > 0.96 (25th–75th percentile of 0.75–0.99) for the 357 SNPs that passed the quality check. The initial analysis for SNP association with risk for SCC was conducted using all miners (n = 650). Q–Q plots were generated to characterize the impact of sample quality as reflected by the SCR on the association signals. The departure from the diagonal line was reduced as the threshold of the SCR increased (Figure 1A). A SCR of 0.65 was selected as the threshold for including miners for the SNP association analysis because call rates greater than this have little additional effect on the Q–Q plot. Thus, the final analysis was conducted in 241 SCC cases and 280 controls. The average call rates for SCC cases and controls were 0.95 and 0.91, respectively, with 90% of samples with SCR > 0.80.
On average, cases have 1.5-fold higher cumulative radon progeny exposure than controls, consistent with radon progeny exposure as a risk factor for SCC in miners (Table I). Cases were older and had higher pack years than controls at the time of the latest sputum collection. Half of the cases died within 6 months of diagnosis. Cases were an average of 6 years younger at death than controls. All study subjects were male and NHWs.
Thirty seven of the 357 SNPs that passed the quality check were nominally associated with risk for SCC in miners (P < 0.05, Supplemental Table I, available at Carcinogenesis Online). Most strikingly, 6 of the 11 tag SNPs in SIRT1, including rs10997817, rs7097008, rs7895833, rs12414281, rs17533847 and rs747024, were associated with increased risk for SCC (Supplemental Table I, available at Carcinogenesis Online). Rs7097008 had the most significant association with the risk for SCC in SIRT1 with each A allele conveying 40% of increased risk (P = 0.003). Furthermore, consistent association with the risk for SCC was observed between the two tag SNPs selected from the same bins in SIRT1 (Supplemental Table I, available at Carcinogenesis Online). No interaction between radon exposure and rs7097008 genotype for the risk of SCC was identified (P = 0.25, data not shown) possibly due to the limited power for the detection of the gene–environment interaction. Moreover, the A allele of rs7097008 associated with increased risk for SCC was also associated with better prognosis (n = 217, HR = 0.81, 95% CI = 0.67–0.99, P = 0.04, data not shown). The Kaplan–Meier survival curves revealed significant difference between SCC cases with 0, 1 and 2 copies of the rs7097008 A allele (P = 0.04 for Wilcoxon test) with the largest difference in survival time observed when 80% of cases were deceased (Figure 1B).
FPRPs were calculated to address the robustness of the most significant (P < 0.01) associations with the risk for SCC in miners (Supplemental Table I, available at Carcinogenesis Online). Rs20551 is a non-synonymous SNP in EP300 and rs7097008 is in high LD with multiple promoter SNPs that may affect the binding of transcriptional factors (21). In addition, both SNPs were reported to be associated with multiple disease phenotypes and/or functional readout(s) (22–28). Thus, a higher prior probability (0.10) was assigned to these two SNPs. Lower probabilities (0.01 and 0.001) were assigned to the rest of the SNPs because of either lack of predicted functional potential or no disease association identified or both. Associations of rs20551 and rs7097008 with risk for SCC in miners have FPRP ≤ 0.15 and should be considered noteworthy.
SIRT1 resides in a 279kb haplotype block (chr 10: 69232957–69513313; NCBI B36) that also contains two other genes (DNAJC12 and HERC4) (Supplemental Figure I, available at Carcinogenesis Online). Eight common (>4%) haplotype alleles were constructed using Haploview software in the HapMap population with European ancestry (version 3 release R2, Figure 2A) and accounted for 90% of the variation in the SIRT1 block (Figure 2A). Interestingly, the composite gray white pattern of the 59 common SNPs with gray representing variant alleles that represents the linear combination of the SNPs along the chromosome in the haplotype block can be grouped into two categories (Figure 2A and Table II). Category A contains four haplotype alleles that carry the minimal number of variant alleles with a total frequency of 0.63. Category B contains the remaining four haplotype alleles with majority of SNPs carrying the variant alleles with a total frequency of 0.28. Haplotype alleles across the two categories can be fully differentiated by two tag SNPs (rs7097008 and rs10823120). In addition, the minor alleles of six SNPs that were associated with SCC (P ≤ 0.06) tagged at least one of the four haplotype alleles in category B (Figure 2A).
A haplotype-based analysis was conducted to integrate the genetic variation in each individual genomic locus because multiple SIRT1 tag SNPs were associated with the risk for SCC in miners. This analysis was conducted in a subset of miners (n = 481) with ≥8 out of 11 SIRT1 tag SNPs successfully genotyped to ensure the accuracy of the haplotype reconstruction. Eight tagging SNPs (rs10997817, rs7895833, rs17454621, rs10823120, rs1812671, rs497849, rs10997908 and rs747024) allowed the construction of same eight common haplotype alleles (>4%) in the miners with a cumulative frequency of 71% (Table II). The likelihood ratio test indicated a global association between genetic variation in the haplotype block where SIRT1 resides and the risk for SCC in miners (P = 0.003). A logistic regression model developed by selectively removing haplotypes identified the four common haplotype alleles in category A that were associated with increased risk for SCC in miners (Table II). A risk haplotype allele score was generated by summing the probabilities of these four common haplotype alleles for each individual. A highly significant increased risk for SCCs was observed for miners carrying the risk haplotype alleles (OR = 1.69 per risk haplotype allele, P = 8.2×10−5). Furthermore, this risk haplotype allele score was also associated with better prognosis (HR = 0.79, 95% CI = 0.64–0.98, P = 0.03).
Characterization of the LD for common (>4%) SNPs (n = 377) and insertion/deletion polymorphisms (n = 42) located in the 279kb haplotype block where SIRT1 resides using the 1000 Genome Pilot1 CEU population identified 168 variants that have high LD (r 2 > 0.7) with rs7097008 (Supplemental Figure II, available at Carcinogenesis Online). Further expanding this region to 500kb upstream and downstream of rs7097008 did not identify any other variants in high LD with rs7097008 (r 2 < 0.25) (Supplemental Figure II, available at Carcinogenesis Online). Thus, it is unlikely that the association seen for rs7097008 is due to long-range LD stemming from functional SNPs outside these candidate regions. The functional potential for these 168 SNPs was assessed with no non-synonymous SNPs identified in SIRT1. However, three SNPs (rs932658, rs2394443 and rs3758391) in the SIRT1 promoter were predicted to affect the binding of transcription factors and one deletion/insertion polymorphism (rs35461348) within the 3' untranslated region was predicted to affect the binding of microRNAs (data not shown). Thus, the association of the rs7097008 genotype with SIRT1 expression was assessed in NHBECs and normal lung tissues from NHW subjects. The number of rs7097008 C alleles was correlated with SIRT1 expression in a dose-response manner as seen by a 39% and 65% increase in gene expression per allele in NHBECs using either β-actin or PCNA as the endogenous control (P = 0.02 for both; Figure 2B). This association was further validated in 40 normal lung tissues (P = 0.03, data not shown). Rs35461348, a deletion/insertion polymorphism (−/AAAG) is located within the 3' untranslated region of SIRT1 and its AAAG allele is in high LD (r 2 > 0.83) with the rs7097008 C allele in NHWs. A TaqMan allelic discrimination assay was designed to obviate the need for an endogenous control gene for relative quantification, thus allowing for a more precise assessment of the relative abundance between the two alleles in complementary DNA generated from heterozygous NHBECs. Expression of the AAAG allele was 30% greater compared with the deletion allele (P = 0.0008; Figure 2C).
SIRT1 plays an important role in DNA damage response and genome integrity by maintaining chromatin structure and DNA damage repair foci formation (32). The mutagen sensitivity assay was conducted to assess whether the risk allele (A) of rs7097008 was associated with reduced DRC in lymphocytes challenged by 1.44 Gy X-ray radiation. Up to 40% increase in the number of chromatid breaks was observed in A allele homozygotes compared with C allele homozygotes (P = 0.004; Figure 2D), suggesting that risk allele carriers have compromised DRC toward DSBs.
SIRT1 alleles associated with increased risk for SCC are associated with reduced gene expression and compromised DRC toward DSBs, suggesting that SIRT1 may be a tumor suppressor in lung SCC development in uranium miners. To extend these findings, SIRT1 expression was evaluated independent of genotype in lung tumor normal pairs, 20 pairs each for lung SCCs and adenocarcinomas collected from non-miner smokers. Intriguingly, SIRT1 expression was reduced to an average of 48% and 67% in tumors versus normal lung tissues for lung adenocarcinomas (P = 2.4×10−8) and SCCs (P = 1.8×10−13), respectively. In addition, the reduction of SIRT1 expression in SCCs compared with lung adenocarcinomas was also validated using The Cancer Genome Atlas lung cancer data (33). The log R ratio for the SIRT1 expression in tumors relative to a Stratagene Universal Reference was −0.28±0.05 for SCCs (n = 150) and 0.38±0.09 for lung adenocarcinomas (n = 32) (P = 8×10−10).
This is the first study to comprehensively assess the association between genetic variants in NHEJ and related histone modification genes and susceptibility and prognosis of lung SCC in a unique occupational cohort of uranium miners with a high cumulative radon exposure. Our findings suggest a role of genetic variation in SIRT1 as a risk modifier for developing SCC in this population. Furthermore, the alleles associated with increased risk for SCC confer a favorable prognosis in cases. Finally, functional validation provides corroborating evidence for the population-based associations by demonstrating that the SIRT1 alleles are associated with reduced gene expression and compromised DRC toward DSBs.
SIRT1 is a mammalian nicotinamide adenine dinucleotide-dependent histone/protein deacetylase and a homolog of yeast silent information regulator 2 that is required for replicative lifespan extension upon calorie restriction (34). SIRT1 has direct or indirect roles in epigenomic regulation (heterochromatin formation) by deacetylating histones and chromatin modifiers such as Suv39h1 (35–37). Detailed studies have provided compelling evidence that SIRT1 is involved in epigenetic modifications of both local chromatin structure and DNA repair machineries to facilitate DNA damage repair. In response to DNA damage, SIRT1 is recruited to DNA DSB sites to facilitate remodeling of local chromatin structures presumably to support repair (32,38). Multiple DNA damage repair factors are modified by SIRT1 through deacetylation, including xeroderma pigmentosum c protein, Werner syndrome protein, Nijmegen breakage syndrome protein and Ku70 (38–41). Loss of SIRT1 results in impaired DRC toward several major types of DNA damage including DSBs and causes chromosomal abnormality and translocation in mouse embryonic cells (38,41,42). Thus, these studies strongly support our findings that SIRT1 alleles associated with reduced gene expression are associated with compromised DRC toward DSBs and increased risk for SCC in miners. Furthermore, the reduction of the SIRT1 expression in lung adenocarcinomas and SCCs from non-miner smokers suggests that SIRT1’s tumor suppressor role may extend to smoking-induced lung cancer in the general population. A strong support for this premise is provided by a most recent study that identified SIRT1 pathway as a key pathway being activated in response to tobacco smoke exposure in primary human bronchial epithelial cells (43). Furthermore, SIRT1 activity was downregulated in non-small cell lung cancers from smokers, suggesting that SIRT1 could play a tumor suppressor role in smoking-induced lung cancer in general population. Reduced expression of SIRT1 has also been demonstrated in human tumors derived from skin, breast, liver, ovarian, prostate, bladder and brain (41,42).
Because SIRT1 is an important regulator of energy metabolism through its impact on glucose and lipid metabolism, extensive studies were conducted to assess whether genetic variation in SIRT1 was associated with metabolism syndromes, such as obesity and type II diabetes (22–26). The major alleles in NHWs tagging haplotype alleles in category A defined in this study or in high LD with rs7097008 allele A all have consistent association with increased risk for adult obesity across several independent large cohorts (22–24). In addition, these SNPs also interact with non-genetic factors such as prenatal exposure to famine and sex to increase the risk for type II diabetes in NHWs and Pima Indians (25,26). Thus, although the endpoints in these studies are not lung cancer, their consistent association with SIRT1 SNPs across studies provide strong support for our findings and suggest there must be one or multiple causative SNPs among the 168 genetic variants (SNPs and insertion/deletion polymorphisms) in high LD with rs7097008 that drive the associations. Indeed, this premise is supported by a study that identified a novel TP53 binding site in the distal human SIRT1 promoter that mediates the calorie restriction-induced SIRT1 over expression (34). More importantly, a C/T polymorphism (rs3758391) that is in high LD with rs7097008 (r 2 > 0.8) in NHWs lies in this novel TP53 binding site with the common allele C destroying the binding of TP53 and impairing the calorie restriction-induced over expression of SIRT1 and SIRT1 target genes AMPKα2 and PGC-1β in human skeletal muscles (34). Finally, a meta-analysis was conducted to summarize five lung cancer genome-wide association studies (GWAS) that contained 4750 cases and 16 040 controls who were Caucasian smokers (44). These five GWAS include MDACC, Toronto, Germany, DeCODE and Institute of Cancer Research. A suggestive association was identified between SIRT1 rs7097008 and risk for non-small cell lung cancer (P = 0.09; S Leng et al. unpublished data). Thus, the broader association with multiple metabolism syndromes, the identification of a functional promoter SNP that is in high LD with rs7097008 and a suggestive association with the risk for smoking-induced lung cancer provide additional support for the population-based association observed in this study.
SIRT1 alleles associated with reduced gene expression were associated with favorable lung cancer-specific overall survival in SCC cases in miners, although information for tumor stage and therapy was not adjusted in the models due to its lack of availability. Lower SIRT1 expression in lung SCC tumors was also associated with greater overall survival with adjustment for age at diagnosis and tumor stage in The Cancer Genome Atlas patients who received chemotherapy and/or radiation therapy (n = 26, HR = 0.10 per 2-fold change of SIRT1 expression, 95%CI = 0.01–0.89, P = 0.03), but not in patients without chemotherapy and/or radiotherapy (n = 126, HR = 1.34, 95% CI = 0.77–2.33, P = 0.29). In addition, reduced SIRT1 expression was reported to be associated with better prognosis for breast carcinoma and diffuse large B-cell lymphoma in patients who received chemotherapy and/or radiotherapy (45–47). This is consistent with the notion that patients with a lower SIRT1 expression should respond better to chemotherapy and/or radiation therapy due to compromised DNA repair. Moreover, a most recent study showed that SIRT1 over expression promoted acquisition of both BCR-ABL mutations in chronic myelogenous leukemia (CML) cells treated with tyrosine kinase inhibitors and mutations of hypoxanthine phosphoribosyl transferase in CML and other tumor cells treated with the chemotherapeutic agent camptothecin (48). Mechanistic studies found that although SIRT1 can enhance cellular DNA damage response in CML and non-CML tumor cells, it can also alter the function of DNA repair machinery by stimulating the activity of error-prone DNA damage repair process (48). Our new findings provide additional insight into how sequence variants within SIRT1 can impact the function of this emerging tumor suppressor gene.
NIEHS National Institute of Environmental Health Sciences (R01 ES015262), US Department of Energy (DE-FG02-09ER64783) and NIH/NCI National Institute of Health/National Cancer Institute (P30 CA118100). Support for lung cancer GWAS meta-analysis was through NIH National Institute of Health (U19CA148127).
We gratefully acknowledge all the principal investigators of the Transdisciplinary Research in Cancer of the Lung for the lung cancer GWAS results for the meta-analysis of SIRT1 SNPs. The meta-analysis was conducted using the five studies led by Drs Christopher I.Amos at Dartmouth College in USA, Richard Houlston at Institute of Cancer Research in UK, Rayjean J.Hung at Samuel Lunenfeld Research Institute of Mount Sinai Hospital in Canada, Kari Stefansson at DeCODE Genetics in Iceland and H.-Erich Wichmann at Institute of Epidemiology I and Ludwig-Maximilians-Universität Munich in Germany. We also thank Mr Thomas J.Gagliano at Lovelace Respiratory Research Institute for the scientific editing of all figures.
Conflict of Interest Statement: None declared.