Common TDP1 Polymorphisms in Relation to Survival among Small Cell Lung Cancer Patients: A Multicenter Study from the International Lung Cancer Consortium

Small-cell lung cancer (SCLC) is the most aggressive form of lung cancer, with a 5-year survival of only 7% (1). Despite rapid advances in cancer therapy, treatment and overall survival of SCLC patients has changed little over the past few decades (2, 3). Unlike non–small cell lung cancer (NSCLC), in which several prognostic and predictive biomarkers have been identified and targeted clinically (4), there are relatively few markers to predict survival or to guide treatment selection for SCLC patients (reviewed in refs. 2 and 5).

A combination of platinum chemotherapy and a DNA topoisomerase inhibitor is the first-line chemotherapy for treating SCLC patients (6). DNA topoisomerases (TOP1 and TOP2) are important players during DNA replication and transcription as they introduce transient DNA strand breaks (7). TOP1 inhibitors (e.g., irinotecan, topotecan) and TOP2 inhibitors (e.g., etoposide, teniposide) bind to DNA topoisomerases and generate drug-stabilized DNA cleavage complexes, which eventually result in tumor cell death (8, 9). Tyrosyl-DNA phosphodiesterase (TDP1) plays a role in repairing both TOP1- and TOP2-mediated DNA damage (10, 11) and it is believed to be responsible for drug resistance to DNA topoisomerase inhibitors (12, 13). A study in SCLC cell lines suggests that the TDP1/TOP1 ratio may be an indicator for the response of SCLC to topotecan (14); however, confirmation in SCLC tissue is lacking. Limited available tissue for such confirmation presents a challenge, as only a small portion of SCLC patients receive surgical resection.

Developing a blood-based marker to predict drug response would be useful to inform appropriate treatment for SCLC patients. As TDP1 plays a role in resistance to DNA topoisomerase inhibitors, it is plausible that patients carrying a TDP1 variant may respond differently to treatment, thus having different survival outcomes. There are very few studies on TDP1 single-nucleotide polymorphisms (SNP; refs. 15–17) and, to the best of our knowledge, none have examined TDP1 SNPs in relation to SCLC survival. In this study, we investigated whether common TDP1 SNPs are associated with overall survival among SCLC patients in a multicenter study from the International Lung Cancer Consortium (ILCCO, http://ilcco.iarc.fr).

This study consists of 898 SCLC patients from 10 ILCCO studies that have data on patient survival time and vital status (Table 1). Further details on the study population and source of data for each study are provided in the Supplementary Text. All participants provided written informed consent, and each study was approved by its local institutional review board. For the current study, SCLC includes small-cell carcinoma, combined small-cell cancer, and neuroendocrine carcinoma (ICD-O 8013, 8041, 8042, 8043, 8044, 8045, 8246).

Tag SNPs for the TDP1 gene region (±2.5 kb of the coding sequence) were identified using the Genome Variation Server (http://gvs.gs.washington.edu/GVS150/). SNPs were classified into bins with a pairwise linkage disequilibrium (LD) threshold of r² ≥ 0.8 using the IdSelect algorithm (18). The list of TDP1 tag SNPs based on the HapMap Phase I and II Centre d'Etude du Polymorphism Humain (CEU) population is shown in the Supplementary Table S1. One SNP per bin of the tag SNPs with an average minor allele frequency (MAF) ≥ 5% (a total of six SNPs) was selected by prioritizing on the SNP function class and predicted genotyping success based on Illumina assay design score. Six TDP1 tag SNPs (rs9488, rs942190, rs1286927, rs2401863, rs4143999, and rs12880397) were genotyped on 1,586 healthy controls and 793 lung cancer cases (including 137 SCLC) from the β-Carotene and Retinol Efficacy Trial (CARET) as part of a study on germline variation in DNA repair genes and lung cancer risk (19, 20). Four of the six SNPs had low MAF among SCLC patients (0.03–0.07) and were excluded from further investigation as a very large sample size would be needed to determine the effect of these SNPs. Thus, only two SNPs (rs942190 and rs2401863) were chosen for the current pooled analysis. These two SNPs are partially correlated, especially among individuals of European ancestry with r² of 0.63 [r²_{(East Asian)} = 0.26].

The majority of genotype data for our pooled analysis were obtained from the OncoArray, a custom array manufactured by Illumina which contains approximately 500K SNPs that provide genome-wide coverage of most common genetic variants along with markers of interest for common cancers (21). Genotype data from the Mayo Clinic and part of the genetic data from the Lunenfeld-Tanenbaum Research Institute were from existing genome-wide association studies (GWAS). Samples from CARET participants were genotyped using a custom-designed 384-plex GoldenGate assay (Illumina). Samples from Japan were genotyped using a pre-design (for rs942190) and a custom-design (for rs2401863) TaqMan assay (Applied Biosystems). Race-specific genotype frequencies for both SNPs were in agreement with Hardy–Weinberg equilibrium (χ²P for rs942190 among Whites, rs942190 among Asians, rs2401863 among Whites, and rs2401863 among Asians were 0.41, 0.14, 0.54, and 0.20, respectively).

Clinical and genotype data were harmonized across studies. Characteristics of all 898 patients by study site are summarized in Supplementary Table S2. Race was imputed as White for the 96 patients of unknown race since their genotype distributions for both SNPs were similar to those of White patients (Supplementary Table S3). Tumor stage was classified as limited stage (LS or stage I–III) and extensive stage (ES or stage IV). Chemotherapy drug use was classified as “TOP1 inhibitor” (received any TOP1 inhibitor along the courses of chemotherapy), “TOP2 inhibitor” (received any TOP2 inhibitor along the courses of chemotherapy), and “Other/Unknown” (i.e., not known to have received any TOP1 or TOP2 inhibitor). The “TOP1 inhibitor” group and the “TOP2 inhibitor” group also contained patients who received both TOP1 and TOP2 inhibitors, either at the same time (n = 3) or switching from one to the other during the course of chemotherapy (n = 40). The primary outcome was overall mortality as of 36 months postdiagnosis (when deaths are commonly attributed to lung cancer), measured from the date of lung cancer diagnosis until the date of death, last contact, or censoring at 36 months follow-up, whichever occurred first. Disease-specific survival was not examined, as cause of death was missing for 43% of the patients.

Survival analyses were performed using Kaplan–Meier survival plots and Cox proportional hazard regression models with a robust estimator of variance adjusting for age, sex, race (White vs. Asian), and tumor stage. Analyses were conducted to evaluate genotype and haplotype associations with overall mortality at 36 months postdiagnosis. Six patients with no follow-up data and two patients with no genotype data for both SNPs were excluded from survival analyses. Of the remaining 890 patients, 6 and 2 did not have genotype data for rs942190 and rs2401863, respectively. As the SNP genotype frequencies were quite different between Whites and Asians, we also conducted a subgroup analysis by race for each SNP. Analyses were performed using STATA 14 (StataCorp). Haplotype analysis was performed using the THESIAS (Testing Haplotype Effects In Association Studies) software version 3.1 (22), which is based on the Stochastic expectation maximization algorithm (23). HR and 95% confidence intervals (CI) adjusting for age, sex, and tumor stage were calculated using the most common haplotype as the reference. Haplotype analysis was performed on White patients only as the two SNPs were correlated among Whites and sample sizes for Asian and other races were limited.

Selected characteristics of patients included in the survival analyses are presented in Table 2. The majority of patients were male, non-Hispanic White, and either current or former smokers. There was a slightly higher proportion of patients with limited stage than extensive stage SCLC. Treatment was unknown for approximately 25% of the patients. Almost 90% of patients with known treatment received chemotherapy, among whom most received a TOP2 inhibitor. Approximately 90% of patients had died by the time of last follow-up and 87.5% of deceased patients died within 36 months after diagnosis of SCLC. The median follow-up time for patients who were alive at last follow-up was 73 months (ranged from 3 to 234 months). The allele frequencies of the two SNPs differed between persons of European and East Asian ancestry. The MAFs of rs942190 (G allele) for White and Asian patients in this study were 0.49 and 0.23, respectively, and for rs2401863 (C allele) were 0.38 and 0.52, respectively. Mean age at diagnosis was similar for patients in each genotype group. There was a slightly lower proportion of female and tumors of limited stage among patients with the rs942190 AA genotype compared with patients with the other two rs942190 genotypes. The proportions of tumor stage were comparable by rs2401863 genotype.

Kaplan–Meier analyses for all patients with known vital status and genotype demonstrated poorer survival for patients homozygous for the minor allele (GG) of rs942190 compared with those carrying the other two genotypes (Fig. 1A). For rs2401863, better survival was associated with carrying both minor alleles (CC); however, the association was not statistically significant (Fig. 1B).

The results from multivariable Cox regression analyses (Table 3) were consistent with the results from Kaplan–Meier analyses. Patients carrying GG of rs942190 had poorer survival compared with those with the AA genotype, with a HR of 1.36 (95% CI, 1.08–1.72; P = 0.01), but no association with survival was observed for patients with the heterozygous (AG) genotype (HR = 1.04, 95% CI, 0.84–1.29; P = 0.72). The HRs associated with the presence of the GG genotype were in the same direction for Whites and Asians. For rs2401863, patients carrying two minor alleles (CC genotype) tended to have better survival than patients carrying the AA genotype (HR = 0.79, 95% CI, 0.61–1.02; P = 0.07); however, this inverse association with survival was observed only in White patients (HR = 0.71, 95% CI, 0.54–0.94; P = 0.02). The association was, if anything, in the opposite direction among Asian patients (HR = 2.11; 95% CI, 0.90–4.95, P = 0.09). The most common haplotype among White patients was the haplotype containing the G allele of rs942190 and the A allele of rs2401863 (the risk haplotype). The haplotype containing the rs942190 A allele and the rs2401863 C allele was associated with better survival, compared with the most common haplotype (HR = 0.84; 95% CI, 0.73–0.95; P = 0.008).

We also examined potential functional consequences of the two SNPs using a single tissue expression quantitative trait loci (eQTL) analysis from the Genotype-Tissue Expression (GTEx) Project (www.gtexportal.org). The GTEx Project, funded by The National Institutes of Health Common Fund, has collected and analyzed genomic variation from blood and gene expression in multiple tissues of the nondiseased donor to determine how genetic variation affects gene expression in human tissues (24). On the basis of the analysis available from the GTEx website, TDP1 gene expression was higher in lung tissues of people with the GG genotype of rs942190 than of people with AG or AA genotypes (P = 0.0008; Fig. 2). In contrast, there was minimal difference of TDP1 gene expression in lung tissue across rs2401863 genotypes (P = 0.12).

On the basis of SNP functional prediction (http://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html; ref. 25), rs942190 may affect TDP1 expression by residing in a transcription factor binding site (TFBS). We further investigated which transcription factors (TF) bind to this region using data from the Encyclopedia of DNA Elements (ENCODE). The ENCODE project has performed a large number of ChIP-seq experiments on multiple cell lines to identify TFBSs across the human genome (26–28). Table 4 shows the list of 20 TFs identified by ENCODE that could bind to the TFBS in the location that rs942190 resides. In addition, we used a freely accessible web-based application called ePOSSUM (http://mutationtaster.charite.de/ePOSSUM/) by Hombach and colleagues (29) to assess the impact of the T(A) and C(G) alleles of rs942190 on TF binding. ePOSSUM allows user to enter either the genomic position of the SNP (based on human genome assembly GRCh37) or the wild-type and variant sequences. The output shows predicted TF binding scores of 81 TFs from three sources (JASPAR, HT-SELEX, and hPDI) as well as a summary of prediction whether the genetic alteration leads to the gain or loss of TF binding. Of the 20 TFs identified by ENCODE, nine have data available on the ePOSSUM website. Using rs942190 location (Chr14:90422414T>C) as an input on the website, six of the nine TFs were predicted to have a different binding affinity to the T(A) and C(G) allele. These included CTCF, MAX, MYBL2, RBBP5, TFAP2A, and TFAP2C (Table 4).

To our knowledge, this study is the first to investigate germline variation of TDP1 in relation to survival among SCLC patients. Leveraging data from ten ILCCO studies, we analyzed a fairly large cohort of SCLC patients with near complete follow-up at 36 months. Of the two SNPs examined, we found the rs942190 GG genotype to be associated with poorer overall survival compared with AA genotype.

Several lines of evidence support the potential function of rs942190 including the results from GTEx, ENCODE, and ePOSSUM. It has been shown that overexpression of TDP1 in cell lines could counteract the effect of DNA topoisomerase inhibitors (30); therefore, one would expect that patients with higher TDP1 in lung tissue may have more resistance to treatment with DNA topoisomerase inhibitors. The observed higher TDP1 expression in lung tissue of individuals with the rs942190 GG genotype from the GTEx analysis is in line with our finding that patients with the GG genotype had poorer survival than patients with the other two genotypes. However, this observation is based on healthy tissue; the effect of rs942190 GG genotype may be different in tumor tissue. No known studies have compared TDP1 expression in tumor versus adjacent nontumor tissue from SCLC patients, although increased TDP1 expression has been found in tumor tissue relative to adjacent nontumor tissue from NSCLC patients (31, 32). Conversely, we did not find a clear association between rs2401863 genotype and survival, which is consistent with the lack of association between the rs2401863 genotype and TDP1 expression in lung tissue. Our observed association of the rs2401863 genotype with survival among White patients only may be due to the linkage with rs942190 SNP.

The difference in TDP1 expression by rs942190 genotypes may be the result of differences in TF binding affinity. This SNP is located in the TFBS where several TFs bind (confirmed by cell line experiments from the ENCODE project). At least six TF were predicted (based on in silico analysis) to have different binding affinity between T(A) and C(G) alleles of rs942190. Although these six TFs are mostly well known, the effect of these TFs specifically on TDP1 gene expression has not been reported. CTCF could function as an enhancer or repressor (33); thus, the attenuation of binding may result in either increasing or decreasing transcription. MAX could also be either an enhancer (forming heterodimers with MYC, MYC-MAX) or a repressor (forming heterodimers with MAD, MAD-MAX or homodimers, MAX-MAX; refs. 34–36). Overexpression of MYBL2 has been found in several cancer types and associated with poor patient outcomes (reviewed in ref. 37). Moreover, studies in cell lines suggest that overexpression of MYBL2 is associated with resistance to chemotherapeutic agents (including etoposide) and radiation (38–40). It is plausible that one mechanism of resistance to topoisomerase inhibitor or radiation is through activation of TDP1 expression by MYBL2. Patients with the rs942190 GG genotype could have higher MYBL2 binding affinity, thus having higher TDP1 expression that causes their tumors to be relatively more resistant to the treatment. Further study is needed to investigate this possibility. The protein encoded by TFAP2A and TFAP2C (AP-2α and AP-2γ) could activate or repress transcription of their target genes (41, 42). One study has found that decreasing AP-2α and AP-2γ function in breast cancer cell lines leads to an increase in sensitivity to a topoisomerase inhibitor and radiation (43). However, further study is needed to determine if there is a link between AP-2α and AP-2γ and TDP1 expression.

The majority of the data used in this analysis came from etiologic studies of lung cancer, and so the data on treatments received by patients often were limited. The treatment method and the name of chemotherapeutic agents used were unknown for approximately 30% of patients in this study. The majority of patients with unavailable treatment data were from the Harvard cohort. However, the chemotherapy regimen most commonly used in initial treatment of SCLC at Harvard is etoposide (TOP2 inhibitor) plus cisplatin or carboplatin, a regimen similar to that of other institutions. Thus, we would expect that the majority of patients with unknown treatment would have received similar treatment to the rest of patients. On the basis of the available data, we explored whether the association of rs942190 with survival differed between patients who received TOP1 and TOP2 inhibitors. We found a stronger association among patients who received TOP1 inhibitor (HR comparing GG vs. AA adjusting for age, sex, race, and tumor stage = 1.58; 95% CI, 0.87–2.87) compared with those receiving TOP2 inhibitor (aHR = 0.99; 95% CI, 0.73–1.34). When we excluded patients who received both TOP1 and TOP2 inhibitors, the magnitude of association was stronger among patients receiving a TOP1 inhibitor (n = 47, aHR = 1.92; 95% CI, 0.73–5.06). The adjusted HR for those receiving a TOP2 inhibitor without a TOP1 inhibitor (n = 354) was 0.96 (95% CI, 0.69–1.33). However, as the sample size for patients who received a TOP1 inhibitor is quite small, and important data such as chemotherapy completion and response to treatment were unavailable, we are not able to conclude that the association of rs942190 with survival differs among patients receiving different type of topoisomerase inhibitors.

In addition to repairing DNA damage produced by TOP1 and TOP2 inhibitors, an effect of TDP1 on DNA repair caused by radiation has been reported (11, 44). Thus, we further explored the association of rs942190 genotype with overall survival among patients known to have received radiation (n = 290), and found that patients with the GG genotype tended to have poorer survival compared with patients with the AA genotype (aHR = 1.36; 95% CI, 0.95–1.97). The association was stronger among patients who received both a TOP1 inhibitor and radiotherapy (n = 36, aHR = 4.31; 95% CI, 1.1–16.80) but not for those who received a TOP2 inhibitor and radiation (n = 221, aHR = 1.18; 95% CI, 0.78–1.81). We did not observe an association with survival among those who did not receive radiotherapy (n = 287, aHR comparing rs942190 GG to AA genotype = 1.09; 95% CI, 0.76–1.55).

In conclusion, our study suggests an association between rs942190 genotype and overall survival at 36 months after SCLC diagnosis. The association may be different among patients who received different treatment regimens, with respect to both chemotherapy and radiation. Further assessment of the genotype–survival association in a larger study with more detailed and complete treatment data is needed to confirm our findings.

J.K. Field is a consultant/advisory board member for Epigenomics and Vision Gate. G. Liu reports receiving other commercial research support from AstraZeneca, speakers bureau honoraria from Takeda, and is a consultant/advisory board member for AstraZeneca, Novartis, Pfizer, and Takeda. No potential conflicts of interest were disclosed by the other authors.

The views and opinions of and endorsements by the author(s) do not reflect those of the US Army or the Department of Defense.

Conception and design: P. Lohavanichbutr, M.W. Marcus, P. Yang, C. Chen

Development of methodology: J.K. Field, P. Yang, C. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.I. Amos, S.M. Arnold, D.C. Christiani, M.P.A. Davies, J.K. Field, E.B. Haura, R.J. Hung, T. Kohno, M.T. Landi, G. Liu, M.W. Marcus, G.M. O'Kane, M.B. Schabath, K. Shiraishi, A. Tardón, P. Yang, K. Yoshida, X. Zong, G.G. Goodman, C. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Lohavanichbutr, L.C. Sakoda, C.I. Amos, S.M. Arnold, D.C. Christiani, R.J. Hung, Y. Liu, K. Shiraishi, P. Yang, R. Zhang, G.G. Goodman, C. Chen

Writing, review, and/or revision of the manuscript: P. Lohavanichbutr, L.C. Sakoda, C.I. Amos, S.M. Arnold, M.P.A. Davies, J.K. Field, E.B. Haura, R.J. Hung, M.T. Landi, G. Liu, M.W. Marcus, G.M. O'Kane, M.B. Schabath, A. Tardón, P. Yang, R. Zhang, G.G. Goodman, N.S. Weiss, C. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.I. Amos, D.C. Christiani, J.K. Field, G. Liu, M.W. Marcus, M.B. Schabath, S.A. Slone, R. Zhang, G.G. Goodman, C. Chen

Study supervision: G.G. Goodman, C. Chen

Other (principal investigator who brought in funding for the initiation of this study at the Fred Hutchinson Cancer Research Center): C. Chen

We would like to thank the study participants for their involvement.

ILCCO Data Repository was supported by Cancer Care Ontario Research Chair of Population Health, NIH (U19 CA148127), and Lunenfeld-Tanenbuaum Research Institute and Sinai Health System. The Cancer de Pulmon en Asturias (CAPUA) study (PI: Adonina Tardón) was supported by Fondo de Investigación Aanitaria, Instituto de Salud Carlos III, Consorcio de Investigación Biomédica en Red (CIBER) del Área de Epidemiología y Salud Pública, and University of Oviedo. The Environment And Genetics in Lung cancer Etiology (EAGLE) study (PI: Maria Teresa Landi) was supported by the Intramural Research Program of NIH, NCI, Division of Cancer Epidemiology and Genetics. The Epidemiology & Genetics of Lung cancer (EGLC) study (PI: Ping Yang) was supported by the National Cancer Institute and NIH (R03-CA77118, R01s-CA80127, CA84354, and HL107612). The Carotene and Retinol Efficacy Trial (CARET; PIs: Gary E. Goodman, Chu Chen) was supported by the National Cancer Institute and NIH (5-UM1-CA-167462, U01-CA63673, and R01-CA111703). The Harvard Lung Cancer Study (LCS; PI: David C. Christiani) was supported by NIH grants (R01CA092824, R01CA074386, and P30 ES000002). The Japan lung cancer study (PI: Kouya Shiraishi) was supported by the National Cancer Center Research and Development Fund (NCC Biobank). Kentucky Lung Cancer Research Initiative (LCRI; PI: Susanne M. Arnold) was supported by the Department of Defense (Congressionally Directed Medical Research Program, U.S. Army Medical Research and Materiel Command Program) under award number: 10153006 (W81XWH-11-1-0781). This research was also supported by unrestricted infrastructure funds from the UK Center for Clinical and Translational Science, NIH grant UL1TR000117, and Markey Cancer Center NCI Cancer Center Support Grant (P30 CA177558) Shared Resource Facilities: Cancer Research Informatics, Biospecimen and Tissue Procurement, and Biostatistics and Bioinformatics. The Liverpool Lung Project (LLP; PI: John K. Field) is supported by the Roy Castle Lung Cancer Foundation, UK. The work performed for the Toronto lung cancer study (PIs: Rayjean J. Hung, Geoffrey Liu) was supported by Ontario Institute for Cancer Research, the Canadian Cancer Society Research Institute (020214), Ontario Institute of Cancer and Cancer Care Ontario Chair Award (to R.J. Hung and G. Liu) and the Alan Brown Chair and Lusi Wong Programs at the Princess Margaret Hospital Foundation. The Total Lung Cancer (TLC) study (PI: Matthew B. Schabath) was supported by the following funding sources: James & Esther King Biomedical Research Program Grant (09KN-15), NIH Specialized Programs of Research Excellence (SPORE) grant (P50 CA119997), and an American Cancer Society Institutional Research Grant (93-032-13).

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