Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Sep 1.
Published in final edited form as: Eur J Cancer. 2020 Jul 10;136:84–94. doi: 10.1016/j.ejca.2020.04.049

Genetic variants in TKT and DERA in the NADPH pathway predict melanoma survival

Ning Gu 1,2,3,4,†, Wei Dai 3,4,5,†, Hongliang Liu 3,4, Jie Ge 3,4, Sheng Luo 6, Eunyoung Cho 7,8, Christopher I Amos 9, Jeffrey E Lee 10, Xin Li 11, Hongmei Nan 11, Hua Yuan 1,2,12,*, Qingyi Wei 3,4,13,*
PMCID: PMC7540967  NIHMSID: NIHMS1598882  PMID: 32659474

Abstract

Background:

Cutaneous melanoma (CM) is the most lethal type of skin cancers. Nicotinamide adenine dinucleotide phosphate (NADPH) plays an important role in anabolic reactions and tumorigenesis, but many genes are involved in the NADPH system.

Methods:

We used 10,912 SNPs (2,018 genotyped and 8,894 imputed) in 134 NADPH-related genes from a genome-wide association study (GWAS) of 858 patients from The University of Texas MD Anderson Cancer Center (MDACC) in a single-locus analysis to predict CM survival. We then replicated the results in another GWAS dataset of 409 patients from the Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (HPFS).

Rsults:

There were 95/858 (11.1%) and 48/409 (11.7%) patients who died of CM, respectively. In multivariable Cox regression analyses, we identified two independent SNPs (TKT rs9864057 G>A and DERA rs12297652 A>G) to be significantly associated with CM-specific survival [hazards ratio (HR) of 1.52, 95% confidence interval (CI)=1.18-1.96, P=1.06×10−3 and 1.51 (1.19-1.91, 5.89×10−4)] in the meta-analysis, respectively. Furthermore, an increasing number of risk genotypes of these two SNPs was associated with a higher risk of death in the MDACC, the NHS/HPFS, and their combined datasets (Ptrend<0.001, =0.004, and <0.001, respectively). In the expression quantitative trait loci (eQTL) analysis, TKT rs9864057 G>A and DERA rs12297652 A>G were also significantly associated with higher mRNA expression levels in sun-exposed lower-leg skin (P=0.043 and 0.006, respectively).

Conclusions:

These results suggest that these two potentially functional SNPs may be valuable prognostic biomarkers for CM survival, but larger studies are needed to validate these findings.

Keywords: cutaneous melanoma, nicotinamide adenine dinucleotide phosphate, single-nucleotide polymorphism, genome-wide association study, melanoma-specific survival

1. Introduction

Cutaneous melanoma (CM) is the most lethal type of skin cancers. Although CM is much less common than basal or squamous cell carcinomas, it accounts for the majority of skin cancer deaths, because of its tendency to invade and spread during the course of the disease [1,2]. It has been estimated that 96,480 new CM cases, about 5.5% of all of new cancer cases, will be diagnosed in the United States in 2019, and 7,230 patients will die of this disease [3]. Between 2008 and 2014 in the US, the overall five-years survival rate after diagnosis was 91.8% [4], while patients with a late or distant-stage CM generally have a much poorer prognosis with a five-year survival rate of only approximately 18% [2]; survival is particularly poor for those patients with visceral metastasis (a median survival of 6 months) [5]. To better identify high-risk subgroups with a poor survival, it is urgent to identify additional factors, such as genetic variants, that are involved in CM prognosis. Such investigations are likely to help identify clinically relevant mechanisms, allowing for more accurate selection of CM patients for the most effective management and treatment.

Nicotinamide adenine dinucleotide phosphate (NADPH) plays a key role in anabolic reactions in all cellular organisms as a reducing agent [6]. The NADPH system generates free radicals in biological immune cells to destroy pathogens through a process called the respiratory burst [7]. Through the NADPH system, cells produce reactive oxygen species (ROS) that cause cellular and tissue damage by oxidizing both DNA and membranes, thus contributing to cellular dysfunction, aging, neurodegeneration, cell death and cancer [8]. The NADPH system also promotes migration of inflammatory cells and modulates radiation-induced senescent cells by producing ROS [9]. In cancer cells, defects in the mitochondrial oxidative metabolism provide reduced equivalents through NADPH for metabolizing hydroperoxides to increase production of superoxide, hydrogen peroxide, and hydroperoxide [10]. The NADPH oxidases are an important site of ROS generation that is regulated in cancer cells and T cells in the context of antitumor immunity [11]. Several publications have highlighted the impact of NADPH on CM. For example, melanoma cells are characterized by altered redox signaling, especially in the form of higher levels of ROS, than those required for normal cell signals [12]. Two isoforms of NADPH (NOX1 and NOX4) are overexpressed, producing high ROS levels in melanoma cells, and the abundant ROS then stimulates the proliferation of tumor cells [13-15]. Some studies have shown that ultraviolet-induced chemical excitation of melanin fragments leads to DNA damage by initiating the ROSdominated NADPH [16-18].

Many genetic variants have been found to be associated with cancer risk or survival patients with CM [1, 19]. However, the biological relevance of most identified genetic variants, such as single nucleotide polymorphisms (SNPs), remains unknown. Understanding of genetic factors that contribute to the NADPH pathway offers opportunities for prediction and identification of high-risk CM patients for personalized managenment and treatment.

In the present study, therefore, we systematically examined associations between genetic variants in NADPH-related genes and CM survival, by using publicly available genotyping datasets and also evaluated correlations between significant SNPs and their gene expression levels to identify biological mechanisms that may underlie the observed associations.

2. Patients and methods

2.1. Study populations

In the present study, we used two publically available genome-wide association study (GWAS) genotyping datasets: a discovery dataset from The University of Texas MD Anderson Cancer Center (MDACC) and a replication dataset by from the Nurses’ Health Study (NHS) and the Health Professionals Follow-up Study (HPFS). There were 95/858 (11.1%) and 48/409 (11.7%) patients who died of CM, respectively. All patients from the discovery dataset gave written informed consent, and the protocol was approved by the MDACC Institutional Review Board. For the replication dataset, a written informed consent was also obtained from each subject, and the study protocol was approved by the institutional review boards of Brigham and Women’s Hospital and the Harvard T.H. Chan School of Public Health, and those of participating registries as required.

2.2. Discovery Dataset

The discovery dataset included 858 non-Hispanic white patients with CM who were recruited between March 1993 and August 2008 to be participants in a hospital-based case-control study [20]. All CM patients were classified according to the American Joint Committee on Cancer (AJCC) staging system [21] and followed using standardized methods and guidelines [22]. Demographic and clinical variables such as age, sex, Breslow thickness, tumor stage, ulceration, and mitotic rate were available in the datasets obtained from the dbGaP database (accession: phs000187.v1.p1) [23] (Supplementary Table 1).The details of genotyping information and data quality control have previously been reported [20]. Using the MACH software program and the 1000 Genomes Project CEU population (March 2010 release) as the reference, data from the MDACC study have been imputed (imputation quality r2 ≥ 0.8) [24].

2.3. Replication Dataset

The replication dataset from the NHS/HPFS was produced by merging two subdatasets from two other studies: one having 317 female cases from the NHS and the other having 177 male cases from the HPFS [25]. There were 409 non-Hispanic white patients in the NHS/HPFS GWAS dataset. Participants were enrolled in the NHS in 1976 and the HPFS in 1986, and were diagnosed after the enrollment up to the 2008 follow-up cycle for both cohorts. Clinical information on age, sex, survival outcome, and genotype data were available. Genotype data were generated using the Illumina HumanHap610 array and imputed using the MACH software program by using the 1000 Genomes Project CEU population (Northern Europeans from Utah) database (phase I v3, March 2012, with imputation quality r2 ≥ 0.8) as reference panel [26, 27]. The study protocol was approved by the institutional review boards of the Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health, and those of participating registries as required.

2.4. Gene and SNP extraction

Because females carry two copies of the X chromosome but males are heterozygous, and there is no standard statistical data based on sex-specific analysis, five genes in the X chromosome (but no seudogenes and none on Y chromosome) were excluded.

Pseudogenes were also excluded because they have no biological function. A total of 134 NADPH pathway genes located on autosomes were extracted from the online Molecular Signatures Database, which includes gene sets extracted from original research publications as well as and the entire collections from the online resources, such as GO and KEGG (http://software.broadinstitute.org/gsea/msigdb/index.jsp) (Supplementary Table 1). All selected genes were expanded with ±2-kb flanking regions and mapped to all the SNPs identified in the MDACC GWAS dataset following standard quality-control criteria, including minor allele frequency (MAF) ≥ 0.05, genotyping rate ⩾ 95%, and Hardy-Weinberg equilibrium (HWE) P value ⩾ 1×10−5. Consequently, 10,912 [with 2,018 (18.5%) genotyped and 8,894 (81.5%) imputed] common SNPs in the NADPH pathway genes were extracted from the MDACC GWAS dataset. A detailed flowchart is shown in Figure 1.

Figure 1:

Figure 1:

Open in a new tab

Research flowchart. Abbreviations: NADPH: nicotinamide adenine dinucleotide phosphate; SNP: single-nucleotide polymorphism; MDACC: The University of Texas M.D. Anderson Cancer Center; FPRP: false-positive report probability; TKT: transketolase; DERA: deoxyribose-phosphate aldolase; AUC: area under the curve; ROC: receiver operating characteristic; GWAS: genome-wide association study; MAF: minor allele frequency; HWE: Hardy-Weinberg equilibrium; NHS: the Nurses’ Health Study; HPFS: the Health Professionals Follow-up Study.

2.5. Statistical analysis

CM-specific survival (CMSS) was the outcome of interest in the present study, calculated based on the date of CM diagnosis to the last follow-up or CM-related death date. We used multivariate Cox proportional hazards regression models and the GenABEL package of R software to compute the hazards ratio (HR) and 95% confidence interval (CI) for associations between SNPs and CMSS in an additive genetic model. Multivariate analysis of the MDACC dataset was adjusted by age, sex, Breslow thickness, tumor stage, ulceration, and mitotic rate. Multivariate analysis of the NHS/HPFS dataset as well as the combination of the two datasets were adjusted only for age and sex, because clinical factors were not available in the NHS/HPFS dataset.

Because most of the SNPs in the present study were imputed based on linkage disequilibrium (LD) among them, false-positive report probability (FPRP) has been suggested as a better probability threshold for these types of datasets for multiple testing correction than the false discovery rate (FDR) [28]. In the FPRP calculation, a prior probability of 0.10 was selected to detect an HR of 2.0 for an association with variant genotypes or minor alleles of each SNP. SNPs with FPRP < 0.2 were selected for replication in the NHS/HPFS dataset. We also used a meta-analysis to combine the results from the discovery and replication sets. When there was no heterogeneity between the MDACC and the NHS/HPFS datasets (Cochran’s Q test P-value > 0.100 and the heterogeneity statistic (I2) < 50%), we used a fixed-effects model; otherwise, a random-effects model was used. To identify independent predictive SNPs for CMSS, the validated SNPs together with clinical prognostic variables were included in a multivariate stepwise Cox model only using the MDACC dataset that had more covariates available for further adjustment. We used Kaplan-Meier survival curves and log-rank tests to evaluate the effect of selected SNPs on the cumulative probability of CMSS. We also used the Chi-square-based Q-test with P < 0.05 to evaluate effect difference in the stratified analyses. We used the receiver operating characteristic (ROC) curve, which relies on the value of area under the curve (AUC), to illustrate the CMSS prediction with the sensitivity and specificity. Time-dependent AUC and ROC analyses were performed using the two R packages of survival and timeROC [29]. The expression quantitative trait loci (eQTL) analysis for associations with the alleles and genotypes of the significant SNPs was assessed by linear regression analysis using the R software. All analyses were performed with SAS (version 9.3.3; SAS Institute, Cary, NC, USA) unless otherwise specified. All reported P values were two-sided.

3. Results

3.1. Characteristics of study populations

The analyses included 858 patients from the MDACC dataset and 409 patients from the NHS/HPFS dataset. The age range at diagnosis was between 17 and 94 years with a mean age of 52.4 years (52.4±14.4) in the MDACC dataset, compared with 34-87 years and 61.1 years (61.1 ± 10.8 years) in the NHS/HPFS dataset. The male/female ratio was 57.8% (496)/42.2% (362) in the MDACC dataset, compared with 33.7% (138)/66.3% (271) in the NHS/HPFS dataset. Univariate analysis revealed that age, sex, tumor stage, Breslow thickness, ulceration, and mitotic rate were all significantly associated with CMSS in the MDACC dataset, while age but not sex was statistically significantly associated with CMSS in the NHS/HPFS dataset. The median follow-up time for patients was 81.1 months in the MDACC dataset and 179.0 months in the NHS/HPFS dataset. Death rates in the MDACC dataset (95/858, 11.1%) and the NHS/HPFS dataset (48/409, 11.7%) were similar (Supplementary Table 2).

3.2. Gene and SNP extraction

After the exclusion of five genes in the X chromosome, we included 134 NADPH-related genes from the online database (MSigDB) (Supplementary Table 1). Initially, we extracted 10,912 common SNPs from the MDACC GWAS dataset (2,018 genotyped SNPs and 8,894 imputed SNPs). The associations between all these SNPs and CMSS were presented in a Manhattan plot (Supplementary Figure 1). Using P < 0.05 as a threshold, we identified 834 SNPs as significantly associated with CMSS in the discovery phase. After further screening for false-positive findings with a FPRP < 0.2, 341 SNPs were selected for further replication (Figure 1).

In the replication analyses, we continued to use the Cox regression analysis (including only the two available variables of age and sex) to verify the 341 SNPs in the NHS/HPFS dataset. Five SNPs in two genes [rs9864057, rs17234092, rs17306163, rs62255994 in TKT (transketolase) and rs12297652 in DERA (deoxyribose-phosphate aldolase)] remained statistically significant (P < 0.05) for CMSS (Figure 1). Subsequently, LD analysis showed a high LD (i.e., the same LD plots) among the four SNPs in TKT gene. After meta-analysis of these five SNPs, the same associations remained statistically significant, and there was no statistically significant heterogeneity in the results for these five SNPs between the two datasets (Table 1). However, in the univariate analysis as shown in Supplementary Table 2, sex was a risk factor for the MADCC datasets but not for the NHS/HPFS dataset, but in the multivariable medling, sex was no longer a risk factor for the MDACC dataset (Table 2).

Table 1.

Meta-analysis of five validated SNPs using two published melanoma GWAS datasets of the MDACC and NHS/HPFS datasets

MDACC (n=858)
 NHS/HPFS (n=409)
 Meta-analysis
SNP Allele1 Gene Chr EAF HR (95% CI) 2 P2 EAF HR (95% CI) 3 P3 Phet I2 HR (95% CI)4 P4
rs12297652 A/G DERA 12p12.3 0.38 1.51 (1.13-2.01) 0.005 0.40 1.51 (1.00-2.27) 0.048 1.000 0.000 1.51 (1.19-1.91) 5.89×10−4
rs9864057 G/A TKT 3 p21.1 0.23 1.51 (1.10-2.06) 0.003 0.23 1.55 (1.00-2.40) 0.048 0.924 0.000 1.52 (1.18-1.96) 1.06×10−3
rs17234092 C/T TKT 3p21.1 0.24 1.53 (1.13-2.09) 0.006 0.24 1.65 (1.08-2.52) 0.021 0.778 0.000 1.57 (1.22-2.01) 3.84×10−4
rs17306163 T/A TKT 3p21.1 0.24 1.53 (1.13-2.09) 0.006 0.24 1.65 (1.08-2.52) 0.021 0.778 0.000 1.57 (1.22-2.14) 3.84×10−4
rs62255994 C/T TKT 3p21.1 0.23 1.60 (1.18-2.18) 0.010 0.23 1.58 (1.03-2.43) 0.037 0.963 0.000 1.59 (1.24-2.05) 2.90×10−4
Open in a new tab

Abbreviations: SNP: single nucleotide polymorphism; EAF: effect allele frequency; HR: hazards ratio; CI: confidence interval; Phet: P value for heterogeneity by Cochrane’s Q test; GWAS: genome-wide association study; MDACC: The University of Texas M.D. Anderson Cancer Center; NHS: the Nurses’ Health Study; HPFS: the Health Professionals Follow-up Study.

1

Reference allele/effect allele.

2

Adjusted for age, sex, Breslow thickness, distant/regional metastasis, ulceration and mitotic rate in the MDACC dataset.

3

Adjusted for age and sex in the NHS/HPFS dataset.

4

Meta-analysis in a fix-effects model, when the heterogeneity statistic (I2) < 50% or 0.5.

Table 2.

Predictors of CMSS selected by stepwise Cox regression analysis in the MDACC study

Parameter1 Category2 Frequency3 HR (95% CI) P
Age Continuous 836 1.02 (1.00-1.03) 0.062
Sex Female/Male 353/483 1.36 (0.85-2.19) 0.198
Regional/distant metastasis No/Yes 693/143 3.98 (2.59-6.10) <.001
Breslow thickness(mm) Continuous 836 1.16 (1.10-1.22) <.001
Ulceration No/Yes 681/155 2.51 (1.63-3.88) <.001
Mitotic rate (mm2) ≤1/>1 271/565 2.83 (1.37-5.87) 0.005
TKT rs9864057 G>A GG/GA/AA 491/303/42 1.54 (1.12-2.11) 0.008
DERA rs12297652 A>G AA/AG/GG 316/394/126 1.52 (1.14-2.02) 0.004
Open in a new tab

Abbreviations: CMSS: cutaneous melanoma-specific survival; MDACC: MD Anderson cancer Center; HR: hazards ratio; CI: confidence interval;

1

Stepwise analysis included age, sex, regional/distant metastasis, Breslow thickness, ulceration, mitotic rate and two SNPs in two genes (rs9864057 in TKT; rs12297652 in DETA);

2

The leftmost was used as the reference.

3

Twenty-two missing observations of ulceration were excluded; 836 patients remained for the stepwise analysis.

3.3. Genetic variants in NADPH-related genes as independent survival predictors

The four SNPs in the same LD plot of TKT were subsequently annotated and filtered using SNPinfo (http://snpinfo.niehs.nih.gov/snpinfo/snpfunchtml), RegulomeDB (http://www.regulomedb.org/) and F-SNP (http://compbio.cs.queensu.ca/F-SNP) for in silico functional prediction. As shown in Supplementary Table 3, out of the four LD SNPs on TKT, TKT rs9864057 is located at the potential enhancer regions of 16 tissues and the DNAse I sensitive site, while DERA rs12297652 may modify the biding activity of LXR. Therefore, we selected these two SNPs as the representative SNPs for further analysis. In stepwise Cox regression analysis, including the two tagging SNPs, with adjustment for other clinical covariates in the MDACC dataset, these two SNPs (TKT rs9864057 G>A and DERA rs12297652 A>G) remained significant and independent predictors of CMSS (Table 2). All SNPs, both genotyped and imputed, in TKT and DERA are shown in regional association plots containing 200-kb up and downstream of rs9864057 and rs12297652, respectively (Supplementary Figure 2).

3.4. Two independent SNPs as CM survival predictors in the MDACC dataset, the NHS/HPFS dataset, and the combination of both datasets

We further examined associations between the identified independent SNPs and CM survival in the MDACC dataset, the NHS/HPFS dataset, and their combination in multivariate analysis. The risk alleles of these two independent SNPs were rs9864057 A in TKT and rs12297652 G in DERA (Table 3).

Table 3.

Associations between two independent SNPs and CMSS in the MDACC dataset, the NHS/HPFS dataset, and the combined dataset

Genotype MDACC (n=836)
 NHS/HPFS (n=409)
 MDACC + NHS/HPFS (n=1,267)
Frequency Multivariate analysis Frequency Multivariate analysis Frequency Multivariate analysis
All Death (%) HR (95% CI)1 P1 All Death (%) HR (95% CI)2 P2 All Death (%) HR (95% CI)3 P3
TKT rs9864057 G>A
  GG 491 44 (8.96) 1.00 239 22 (9.21) 1.00 746 69 (9.25) 1.00
  GA 303 39 (12.87) 1.84 (1.19-2.86) 0.006 148 22 (14.86) 1.71 (0.94-3.09) 0.077 457 62 (13.57) 1.58 (1.12-2.22) 0.009
  AA 42 8 (19.05) 1.74 (0.78-3.88) 0.177 22 4 (18.18) 2.11 (0.72-6.19) 0.176 64 12 (18.75) 2.23 (1.21-4.13) 0.011
  Trend test 0.010 0.049 0.001
  GA+AA 345 47 (13.62) 1.82 (1.20-2.77) 0.005 170 26 (15.29) 1.76 (0.99-3.11) 0.053 521 74 (14.20) 1.66 (1.19-2.30) 0.003
DERA rs12297652 A>G
  AA 316 21 (6.65) 1.00 145 10 (6.90) 1.00 471 34 (7.22) 1.00
  AG 394 50 (12.69) 2.03 (1.22-3.40) 0.007 200 29 (14.50) 2.11 (1.03-4.34) 0.042 604 80 (13.25) 1.86 (1.24-2.78) 0.002
  GG 126 20 (15.87) 2.24 (1.20-4.15) 0.011 64 9 (14.06) 2.23 (0.91-5.50) 0.081 192 29 (15.10) 2.18 (1.33-3.57) 0.002
  Trend test 0.005 0.049 0.001
  AG+GG 520 70 (13.46) 2.09 (1.28-3.41) 0.003 264 38 (14.39) 2.14 (1.07-4.30) 0.033 796 109 (13.69) 1.93 (1.32-2.84) 0.001
Number of risk genotypes4
  0 184 9 (4.89) 1.00 89 6 (6.74) 1.00 281 17 (6.05) 1.00
  1 439 47 (10.71) 2.46 (1.20-5.05) 0.014 206 20 (9.71) 1.42 (0.57-3.55) 0.458 655 69 (10.53) 1.78 (1.05-3.03) 0.033
  2 213 35 (16.43) 4.24 (2.02-8.89) <.001 114 22 (19.30) 3.08 (1.24-7.62) 0.015 331 57 (17.22) 3.12 (1.81-5.36) <.001
  Trend test <.001 0.004 <.001
  0-1 623 56 (8.99) 1.00 295 26 (8.81) 1.00 936 86 (9.19) 1.00
  2 213 35 (16.43) 2.12 (1.38-3.26) <.001 114 22 (19.30) 2.38 (1.35-4.20) 0.003 331 57 (17.22) 2.02 (1.45-2.83) <.001
Open in a new tab

Abbreviations: SNP: single-nucleotide polymorphisms; CMSS: cutaneous melanoma-specific survival; MDACC: The University of Texas M.D. Anderson Cancer Center; NHS: the Nurses’ Health Study; HPFS: the Health Professionals Follow-up Study; HR: hazards ratio; CI: confidence interval.

1

Adjusted for age, sex, Breslow thickness, distant/regional metastasis, ulceration and mitotic rate in Cox models of SNPs and CMSS in the MDACC dataset;

2

Adjusted for age and sex in the NHS/HPFS dataset;

3

Adjusted for age and sex in the combined dataset of MDACC and NHS/HPFS;

4

Risk genotypes were rs9864057 GA+AA, rs12297652 AG+GG;

5

Twenty-two missing observations of ulceration were excluded in the MDACC dataset.

Under an additive genetic model, the TKT rs9864057 A allele and the DERA rs12297652 G allele were significantly associated with CMSS in the MDACC dataset (trend test: P = 0.010 and 0.005, respectively). Under a dominant genetic model, the TKT rs9864057 A and DERA rs12297652 G genotypes were both associated with elevated risk of CMSS (HR=1.82, 95% CI =1.20-2.77 and P =0.005 for rs9864057 GA+AA vs. GG and HR=2.09, 95% CI =1.28-3.41 and P =0.003 for rs12297652 AG+GG vs. AA). Similar results were obtained in the NHS/HPFS dataset (trend test: P = 0.05 and P =0.05, respectively, and HR=1.76, 95% CI =0.99-3.11, P =0.053 and HR=2.14, 95% CI =1.07-4.30, P =0.03, respectively). Furthermore, when these two datasets were combined, the associations persisted (trend test: rs9864057, P = 0.001 and rs12297652, P =0.001; rs9864057 GA+AA vs. GG: HR=1.66, 95% CI = 1.19-2.30, P =0.003; rs12297652 AG+GG vs. AA: HR=1.93, 95% CI =1.32-2.84, P =0.001) (Table 3).

3.5. Analysis of combined genotypes of the two independent SNPs

To assess the combined effect of the two SNPs, we combined the risk genotypes of rs9865057 GA+AA and rs12297652 AG+GG into one variable as the number of risk genotypes. The trend test consistently showed that the risk increased with the number of risk genotypes in the MDACC dataset (P < 0.001), the NHS/HPFS dataset (P = 0.004), and the combined dataset (P < 0.001). Furthermore, when we dichotomized all patients into 0-1 risk genotypes and 2 risk genotypes, patients with 2-risk genotypes had a higher risk of death than those with 0-1 risk genotypes in the MDACC dataset (HR=2.12, 95% CI=1.38-3.26 and P < 0.001), the NHS/HPFS dataset (HR=2.38, 95% CI=1.35-4.20 and P = 0.003), and the combined dataset (HR=2.02, 95% CI=1.45-2.83, P <0.001). These results were further evaluated by Kaplan-Meier plot to visualize the associations between the number of risk genotypes and CMSS (Figure 2a-2f).

Figure 2:

Figure 2:

Open in a new tab

The independent SNPs and CMSS. (a-f) Kaplan-Meier survival curves of the exact numbers of risk genotypes (a) in the MDACC dataset, (c) in the NHS/HPFS dataset and (e) in these two combined dataset; dichotomiazed groups of risk genotypes (b) in the MDACC dataset, (d) in the NHS/HPFS dataset and (f) in the combined dataset. Abbreviations: SNP: single-nucleotide polymorphism; CMSS: cutaneous melanoma-specific survival; MDACC: The University of Texas M.D. Anderson Cancer Center; NHS: the Nurses’ Health Study; HPFS: the Health Professionals Follow-up Study.

3.6. Stratified analyses for the combined efficacy of risk genotypes on CMSS

We then performed stratified analyses by age and sex for the associations between CMSS and the number of risk genotypes in the MDACC dataset and the NHS/HPFS dataset using multivariate Cox regression analysis with adjustment where appropriate, looking for possible interaction between these covariates and SNPs. The risk of CM death associated with the 2-risk genotype was statistically higher for males ≤ 50 years in the MDACC dataset and for females > 50 years in the NHS/HPFS dataset. We found no evidence for any interactions between the subgroups (Supplementary Table 4).

3.7. ROC curve and time-dependent AUC to estimate CMSS survival

We also used inverse probability of censoring weighting estimators for the ROC curve and time-dependent AUC to estimate the improvement in prediction by adding risk genotypes to the model. In the combined dataset of both MDACC and NHS/HPFS, we found five-year risk of CMSS with clinical variables (age and sex) had an AUC=63.58%, while the addition of risk genotypes to the model significantly increased the AUC to 69.72% (P = 0.005) (Figure 3a-3b). However, such prediction was not statistically significant for both 5-year and 10-year survival prediction in the MDACC dataset with adjustment for age, sex, Breslow thickness, distant/regional metastasis and ulceration (Supplementary Figure 3a-3b) and in the NHS/HPFS dataset with adjustment for age and sex (Supplementary Figure 3c-3d). These results suggest that the sample sizes of the datasets are not large enough or more SNPs in different pathway genes need to be identified.

Figure 3:

Figure 3:

Open in a new tab

ROC and time-dependent AUC estimation for prediction of CM survival in the combined dataset of MDACC and NHS/HPFS as estimated by inverse probability of censoring weighting approach. (a) Five-year CM survival prediction by ROC curve, (b) Time-dependent AUC estimation: based on age, sex and the combined risk genotypes of the two genes. Abbreviations: ROC: receiver operating characteristic; AUC: area under the curve; MDACC: The University of Texas M.D. Anderson Cancer Center; NHS: the Nurses’ Health Study; HPFS: the Health Professionals Follow-up Study; CM: cutaneous melanoma.

3.8. Expression quantitative trait loci (eQTL) analysis

To study the correlation between the two independent SNPs and mRNA expression levels of their corresponding genes, we used data on mRNA expression levels of these two SNPs for further statistical analysis using three in silico eQTL databases: 373 European descendants from the 1000 Genomes Project, The Cancer Genome Atlas (TCGA) database, and the Genotype-Tissue Expression (GTEx) project.

Through the GTEx project, we found that both TKT rs9864057 and DERA rs12297652 were significantly correlated with elevated mRNA expression levels in the skin of sun-exposed lower leg (P = 0.043 and 0.006, respectively), but not significantly associated with mRNA expression levels in the whole blood or in non-sun-exposed skin (Supplementary Table 5). The lymphoblastoid cell-line data from the 373 European descendants and the TCGA database did not demonstrate any correlation between these two SNPs and mRNA expression levels (data not shown).

4. Discussion

NADPH is a key factor in controlling redox reactions in human tissue, including tumor tissue, and NADPH is closely related to cancer cell growth, metabolism, migration, invasion, and metastasis [10]. However, few studies have investigated the roles of genetic variants in the NADPH-related genes in predicting outcomes of cancers, including melanoma. In the present study, we found that two-independent SNPs (TKT rs9864057 G>A and DERA rs12297652 A>G) were significantly associated with CMSS. Additional analyses suggested that their variant alleles were correlated with elevated mRNA expression levels in sun-exposed skin, a possible biological mechanism underlying the associations between the variants of NADPH-related genes and CM progression that led to poor survival.

TKT is located on chromosome 3p21.1, and this gene encodes a thiamine-dependent enzyme that plays a role in the pentose phosphate pathway. In mammals, transketolase joins the pentose phosphate pathway and glycolysis, and then transfers sugar phosphates into the main carbohydrate metabolism, an important factor in the production of NADPH for biosynthesis [30]. Because alteration in cellular metabolism is one of the hallmarks of cancers [31], several studies have reported the clinical significance of TKT in energy regulatory mechanism, especially in the survival of patients with cancers. For example, it was reported that TKT expression was correlated with tumor size in breast cancer; in addition, TKT expression was higher in lymph node metastases than in primary tumor or normal tissues of patients in whom high TKT levels were associated with poor survival after breast cancer [32]. TKT has also been associated with metastasis of ovarian and esophageal cancers, predicting poor survival [33, 34], but there are no such reports about melanoma survival. Other studies have reported significant associations between SNPs of the TKT gene and the morbidity/mortalities of some non-cancer diseases, particularly diabetic nephropathy [35-37]. Because melanoma is a cancer derived from neuroectoderm [38], a link between TKT genetic variants and nerve function has been reported [36]. Taken together, the findings in the present study support a significant role for TKT genetic variants in predicting CMSS.

DERA is located on chromosome 12p12.3 and encodes the human deoxyribose phosphate aldolase, which is involved in the glycosaminoglycan metabolism and the innate immune system. Deoxyribose phosphate aldolase activity levels have been correlated with DERA expression in most cell lines tested, and cells with high DERA activity can use deoxy nucleotide as a source of energy [39]. Through DERA, thymidine-derived 2-deoxy-D-ribose 5-phosphate enters the glycolytic pathway, affecting cancer cell growth, invasion, and metastasis [40]. Thymidine, via the glycolytic pathway, relies on DERA in vitro and in vivo to convert it into multiple thymidine-derived intermediate metabolites required for the survival of cells under low-glucose conditions; therefore, the number of viable cancer cells is thereby reduced, while DERA is also reduced in a microenvironment in which glucose was insufficient [41]. Inhibition of thymidine phosphorylase activity suppresses tumor growth by increasing the proportion of apoptotic cells and probably inhibiting angiogenesis in human epidermoid carcinoma cells [42]. These findings suggest that DERA plays a central role in the survival and growth of cancer cells, particularly under starvation conditions. Furthermore, enhanced DERA-dependent thymidine catabolism has been observed in human gastric cancer [40, 43]. DERA has not been previously reported to be associated with CM progression and prognosis.

While we found consistent evidence that two novel SNPs (TKT rs9864057 G>A and DERA rs12297652 A>G) in two NADPH-related gene predicted CMSS, the present study has some limitations. First, the available clinical variables in the two datasets were different, and the fact there were only two variables (age and sex) available in the NHS/HPFS dataset limited adjustment for other potentially relevant clinical covariates. The second limitation is that the predictive model was built on a non-Hispanic white population in the United States, which may limit generalization of our findings to other populations. Third, SNPs in other relevant genes may have been omitted due to limited knowledge regarding identification of NADPH-related genes. Finally, while differential expression is suggested by the in silico analyses, the exact molecular mechanisms of these two SNPs underlying the observed associations remain to be determined. Additional functional studies are needed to explore these newly identified SNPs in order to confirm their potential utility as biomarkers for CM prognosis.

Supplementary Material

1
2

Highlights.

  • In the pathway gene analysis, two novel and independent SNPs (TKT rs9864057 G>A and DERA rs12297652 A>G) were significantly associated with cutaneous melanoma-specific survival (HR=1.52, 95% CI=1.18-1.96, P=1.06×10-3] and 1.51 (1.19-1.91, 5.89×10-4)

  • The associations were in an allele dose-response manner (Ptrend<0.001 for rs9864057 variant genotypes, 0.004 for rs12297652 variant genotypes, and <0.001 for the number of their combined genotypes, respectively).

  • TKT rs9864057 G>A and DERA rs12297652 A>G were significantly associated with higher mRNA expression levels in sun-exposed lower-leg skin (P=0.043 and 0.006, respectively).

Acknowledgments

We wish to thank all of the investigators and funding agencies that enabled the deposition of data in dbGaP that we used in this study. We would like to thank the Channing Division of Network Medicine at Brigham and Women’s Hospital for their support. We also thank the John Hopkins University Center for Inherited Disease Research for conducting high-throughput genotyping for this study. We would like to thank the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors assume full responsibility for analyses and interpretation of these data. We also thank all the individuals who participated in this project. The results published here are in whole or part based upon data generated by The Cancer Genome Atlas pilot project established by the NCI and NHGRI. Information about TCGA and the investigators and institutions that constitute the TCGA research network can be found at “http://cancergenome.nih.gov". The authors assume full responsibility for analyses and interpretation of these data.

Funding

The MD Anderson Study was supported by NIH/NCI R01 CA100264, 2P50CA093459 and R01CA133996 as well as by The University of Texas MD Anderson Cancer Center Various Donors Melanoma and Skin Cancers Priority Program Fund; the Miriam and Jim Mulva Research Fund; the McCarthy Skin Cancer Research Fund and the Marit Peterson Fund for Melanoma Research. The NHS and the HPFS were in part supported by NIH/NCI R01 CA49449, P01 CA87969, UM1 CA186107 and UM1 CA167552. Qingyi Wei was supported by start-up funds from Duke Cancer Institute, Duke University Medical Center and in part supported by the Duke Cancer Institute as part of the P30 Cancer Center Support Grant (Grant ID: NIH CA014236).

Abbreviations:

CM

cutaneous melanoma

SNP

single-nucleotide polymorphisms

GWAS

genome-wide association studies

MDACC

The University of Texas MD Anderson Cancer Center

NHS

the Nurses’ Health Study

HPFS

the Health Professionals Follow-up Study

CMSS

cutaneous melanoma-specific survival

HR

hazards ratio

CI

confidence interval

eQTL

expression quantitative trait loci

FPRP

false-positive report probability

LD

linkage disequilibrium

NADPH

nicotinamide adenine dinucleotide phosphate

ROC

receiver operating characteristic

AUC

area under the curve

TKT

transketolase

DERA

deoxyribose-phosphate aldolase

ROS

reactive oxygen species

Footnotes

Conflict of interest statement

The authors declare no conflict of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.Law MH, et a I, Genome-wide meta-analysis identifies five new susceptibility loci for cutaneous malignant melanoma. Nat Genet, 2015. 47(9): p. 987–995. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Guy GP Jr., et al. , Vital signs: melanoma incidence and mortality trends and projections - United States, 1982-2030. MMWR Morb Mortal Wkly Rep, 2015. 64(21): p. 591–6. [PMC free article] [PubMed] [Google Scholar]
  • 3.Siegel RL, Miller KD, and Jemal A, Cancer statistics, 2019. 2019. 69(1): p. 7–34. [DOI] [PubMed] [Google Scholar]
  • 4.Siegel RL, Miller KD, and Jemal A, Cancer statistics, 2018. CA Cancer J Clin, 2018. 68(1): p. 7–30. [DOI] [PubMed] [Google Scholar]
  • 5.Demierre MF, Epidemiology and prevention of cutaneous melanoma. Curr Treat Options Oncol, 2006. 7(3): p. 181–6. [DOI] [PubMed] [Google Scholar]
  • 6.Hanukoglu I and Rapoport R, Routes and regulation of NADPH production in steroidogenic mitochondria. Endocr Res, 1995. 21(1-2): p. 231–41. [DOI] [PubMed] [Google Scholar]
  • 7.Ogawa K, et al. , The association of elevated reactive oxygen species levels from neutrophils with low-grade inflammation in the elderly. Immun Ageing, 2008. 5: p. 13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Hervera A, et al. , Reactive oxygen species regulate axonal regeneration through the release of exosomal NADPH oxidase 2 complexes into injured axons. Nature Cell Biology, 2018. 20(3): p. 307–319. [DOI] [PubMed] [Google Scholar]
  • 9.Sakai Y, et al. , NADPH oxidase 4 mediates ROS production in radiation-induced senescent cells and promotes migration of inflammatory cells. Free Radic Res, 2018. 52(1): p. 92–102. [DOI] [PubMed] [Google Scholar]
  • 10.Hegedus C, et al. , Redox control of cancer cell destruction. Redox Biol, 2018. 16: p. 59–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Banerjee R, Introduction to the Thematic Minireview Series: Redox metabolism and signaling. J Biol Chem, 2018. 293(20): p. 7488–7489. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Meierjohann S, Oxidative stress in melanocyte senescence and melanoma transformation. Eur J Cell Biol, 2014. 93(1-2): p. 36–41. [DOI] [PubMed] [Google Scholar]
  • 13.Rouaud F, et al. , Mechanism of melanoma cells selective apoptosis induced by a photoactive NADPH analogue. Oncotarget, 2016. 7(50): p. 82804–82819. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Liu F, Gomez Garcia AM, and Meyskens FL Jr., NADPH oxidase 1 overexpression enhances invasion via matrix metalloproteinase-2 and epithelial-mesenchymal transition in melanoma cells. J Invest Dermatol, 2012. 132(8): p. 2033–41. [DOI] [PubMed] [Google Scholar]
  • 15.Yamaura M, et al. , NADPH oxidase 4 contributes to transformation phenotype of melanoma cells by regulating G2-M cell cycle progression. Cancer Res, 2009. 69(6): p. 2647–54. [DOI] [PubMed] [Google Scholar]
  • 16.Yuan TA, et al. , A Case-Control Study of the Genetic Variability in Reactive Oxygen Species-Metabolizing Enzymes in Melanoma Risk. Int J Mol Sci, 2018. 19(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Raad H, et al. , NADPH Oxidase-1 Plays a Key Role in Keratinocyte Responses to UV Radiation and UVB-Induced Skin Carcinogenesis. J Invest Dermatol, 2017. 137(6): p. 1311–1321. [DOI] [PubMed] [Google Scholar]
  • 18.Premi S, et al. , Photochemistry. Chemiexcitation of melanin derivatives induces DNA photoproducts long after UV exposure. Science, 2015. 347(6224): p. 842–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Manolio TA, Genomewide association studies and assessment of the risk of disease. N Engl J Med, 2010. 363(2): p. 166–76. [DOI] [PubMed] [Google Scholar]
  • 20.Amos CI, et al. , Genome-wide association study identifies novel loci predisposing to cutaneous melanoma. Hum Mol Genet, 2011. 20(24): p. 5012–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Balch CM, et al. , Final version of 2009 AJCC melanoma staging and classification. J Clin Oncol, 2009. 27(36): p. 6199–206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Gershenwald JE and Ross MI, Sentinel-lymph-node biopsy for cutaneous melanoma. N Engl J Med, 2011. 364(18): p. 1738–45. [DOI] [PubMed] [Google Scholar]
  • 23.Mailman MD, et al. , The NCBI dbGaP database of genotypes and phenotypes. Nat Genet, 2007. 39(10): p. 1181–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Li Y, et al. , MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol, 2010. 34(8): p. 816–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Song F, et al. , Exonuclease 1 (EXO1) gene variation and melanoma risk. DNA Repair (Amst), 2012. 11(3): p. 304–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Riedel KG, et al. , Ocular abnormalities in mucolipidosis IV. Am J Ophthalmol, 1985. 99(2): p. 125–36. [DOI] [PubMed] [Google Scholar]
  • 27.Biernacka JM, et al. , Assessment of genotype imputation methods. BMC Proc, 2009. 3 Suppl 7: p. S5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Wacholder S, et al. , Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J Natl Cancer Inst, 2004. 96(6): p. 434–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Chambless LE and Diao G, Estimation of time-dependent area under the ROC curve for long-term risk prediction. Stat Med, 2006. 25(20): p. 3474–86. [DOI] [PubMed] [Google Scholar]
  • 30.Sax CM, et al. , Transketolase gene expression in the cornea is influenced by environmental factors and developmentally controlled events. Cornea, 2000. 19(6): p. 833–41. [DOI] [PubMed] [Google Scholar]
  • 31.Vander Heiden MG, Cantley LC, and Thompson CB, Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science, 2009. 324(5930): p. 1029–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Tseng CW, et al. , Transketolase Regulates the Metabolic Switch to Control Breast Cancer Cell Metastasis via the alpha-Ketoglutarate Signaling Pathway. Cancer Res, 2018. 78(11): p.2799–2812. [DOI] [PubMed] [Google Scholar]
  • 33.Chao YK, et al. , Transketolase Serves a Poor Prognosticator in Esophageal Cancer by Promoting Cell Invasion via Epithelial-Mesenchymal Transition. J Cancer, 2016. 7(13): p. 1804–1811. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Ricciardelli C, et al. , Transketolase is upregulated in metastatic peritoneal implants and promotes ovarian cancer cell proliferation. Clin Exp Metastasis, 2015. 32(5): p. 441–55. [DOI] [PubMed] [Google Scholar]
  • 35.Tanhauserova V, et al. , Genetic variability in enzymes of metabolic pathways conferring protection against non-enzymatic glycation versus diabetes-related morbidity and mortality. Clin Chem Lab Med, 2014. 52(1): p. 77–83. [DOI] [PubMed] [Google Scholar]
  • 36.Spallone V, Might genetics play a role in understanding and treating diabetic polyneuropathy? Diabetes Metab Res Rev, 2017. 33(4). [DOI] [PubMed] [Google Scholar]
  • 37.Pacal L, et al. , Role of thiamine status and genetic variability in transketolase and other pentose phosphate cycle enzymes in the progression of diabetic nephropathy. Nephrol Dial Transplant, 2011. 26(4): p. 1229–36. [DOI] [PubMed] [Google Scholar]
  • 38.Little EG and Eide MJ, Update on the current state of melanoma incidence. Dermatol Clin, 2012. 30(3): p. 355–61. [DOI] [PubMed] [Google Scholar]
  • 39.Salleron L, et al. , DERA is the human deoxyribose phosphate aldolase and is involved in stress response. Biochim Biophys Acta, 2014. 1843(12): p. 2913–25. [DOI] [PubMed] [Google Scholar]
  • 40.Tabata S, et al. , Thymidine Catabolism as a Metabolic Strategy for Cancer Survival. Cell Rep, 2017. 19(7): p. 1313–1321. [DOI] [PubMed] [Google Scholar]
  • 41.Furukawa T, et al. , Thymidine phosphorylase in cancer aggressiveness and chemoresistance. Pharmacol Res, 2018. 132: p. 15–20. [DOI] [PubMed] [Google Scholar]
  • 42.Matsushita S, et al. , The effect of a thymidine phosphorylase inhibitor on angiogenesis and apoptosis in tumors. Cancer Res, 1999. 59(8): p. 1911–6. [PubMed] [Google Scholar]
  • 43.Hirayama A, et al. , Quantitative metabolome profiling of colon and stomach cancer microenvironment by capillary electrophoresis time-of-flight mass spectrometry. Cancer Res, 2009. 69(11): p. 4918–25. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1598882-supplement-1.docx (297.4KB, docx)
2
NIHMS1598882-supplement-2.pdf (354.1KB, pdf)

RESOURCES