Genetic variants in one-carbon metabolism-related genes contribute to NSCLC prognosis in a Chinese population


  • Guangfu Jin MD, PhD,

    1. Department of Epidemiology and Biostatistics, Cancer Center, Nanjing Medical University, Nanjing, China
    Search for more papers by this author
    • The first and second authors contributed equally to this work.

  • Jinlin Huang PhD,

    1. Berkeley Biotech Inc, Taizhou, Jiangsu, China
    Search for more papers by this author
    • The first and second authors contributed equally to this work.

  • Zhibin Hu MD, PhD,

    1. Department of Epidemiology and Biostatistics, Cancer Center, Nanjing Medical University, Nanjing, China
    Search for more papers by this author
  • Juncheng Dai MD,

    1. Department of Epidemiology and Biostatistics, Cancer Center, Nanjing Medical University, Nanjing, China
    Search for more papers by this author
  • Rong Tang PhD,

    1. Berkeley Biotech Inc, Taizhou, Jiangsu, China
    Search for more papers by this author
  • Yijiang Chen MD,

    1. Department of Thoracic Surgery, The First Affiliated Hospital of Nanjing Medical University, Nanjing, China
    Search for more papers by this author
  • Lin Xu MD,

    1. Department of Thoracic Surgery, Jiangsu Cancer Hospital, Nanjing, China
    Search for more papers by this author
  • Xinen Huang MD, PhD,

    1. Department of Oncology, Jiangsu Cancer Hospital, Nanjing, China
    Search for more papers by this author
  • Yongqian Shu MD,

    Corresponding author
    1. Department of Epidemiology and Biostatistics, Cancer Center, Nanjing Medical University, Nanjing, China
    2. Department of Oncology, Jiangsu Key Discipline of Medicine, The First Affiliated Hospital of Nanjing Medical University, Nanjing, China
    • Department of Oncology, Jiangsu Key Discipline of Medicine, The First Affiliated Hospital of Nanjing Medical University, Nanjing, 210029, China
    Search for more papers by this author
  • Hongbing Shen MD, PhD

    Corresponding author
    1. Department of Epidemiology and Biostatistics, Cancer Center, Nanjing Medical University, Nanjing, China
    2. Department of Oncology, Jiangsu Key Discipline of Medicine, The First Affiliated Hospital of Nanjing Medical University, Nanjing, China
    • Department of Epidemiology and Biostatistics, Cancer Center, Nanjing Medical University, 140 Hanzhong Rd., Nanjing 210029, China
    Search for more papers by this author
    • Fax: (011) 86-25-8652-7613



One-carbon metabolism plays a critical role in DNA methylation and DNA synthesis. Variants of genes involved in one-carbon metabolism may result in aberrant methylation and/or DNA synthesis inhibition, and ultimately modulate the initiation and progression of tumors. In this study, the authors hypothesized that polymorphisms in one-carbon metabolism-related genes may contribute to the prognosis of nonsmall cell lung cancer (NSCLC).


The authors screened 57 potentially functional single nucleotide polymorphisms (SNPs) from 11 candidate genes involved in one-carbon metabolism and genotyped them in a cohort of 568 NSCLC patients by using Illumina Golden Gate platform. The Kaplan-Meier method with log-rank test and Cox proportional hazards model were used for survival analyses.


Variant alleles were significantly associated with favorable survivals of NSCLC for MTR rs3768160 A>G (allelic hazards ratio [HR], 0.78; 95% confidence interval [CI], 0.62-0.98), MTRR rs2966952 G>A (allelic HR, 0.84; 95% CI, 0.71-0.99) and DHFR rs1650697 G>A (allelic HR, 0.83; 95% CI, 0.70-0.99) and with unfavorable prognosis for MTHFD1 rs1950902 G>A with borderline significance (allelic HR, 1.18; 95% CI, 0.99-1.40). In addition, the combined genotypes of these four SNPs showed a locus-dosage effect on NSCLC survival (Ptrend = 6.9 × 10−5). In the final multivariate Cox regression model, combined genotypes based on 3 categories may be an independent prognostic factor for NSCLC with adjusted trend HR of 0.78 (95% CI, 0.66-0.92).


Genetic variants in one-carbon metabolism pathway may be candidate biomarkers for NSCLC prognosis. Cancer 2010. © 2010 American Cancer Society.

Lung cancer is the most common malignancy and the world's leading cause of cancer deaths in both males and females.1 Nonsmall cell lung cancer (NSCLC) accounts for 80% of all types of lung cancer, and <15% of patients survive more than 5 years.1 In clinical practice, even in patients with comparable histology and tumor stage, there is great variation in prognosis and treatment outcome. This variation is due in part to genetic diversity caused by single nucleotide polymorphisms (SNPs) in genes involved in tumor progression or treatment response. Identifying stable prognosis biomarkers could help guide medical care and improve patients' survival through personalized treatment.

Folate and related nutrients are essential for DNA synthesis and methylation by a complex set of reactions known as one-carbon metabolism.2 Folate and methionine directly enter the pathway acting as methyl donors for DNA synthesis and methylation, whereas vitamins B6 and B12 serve as enzymatic cofactors (summarized in Fig. 1). Deficient supply of methyl can induce DNA global hypomethylation, resulting in chromosomal instability, transposon reactivation, loss of imprinting, and activation of proto-oncogenes. Conversely, it can also result in promoter hypermethylation, which is associated with the inactivation of tumor suppressor genes involved in detoxification, DNA repair, cell cycle control, apoptosis, etc.3 Methyl deficiency can induce deficient conversion of dUMP to dTMP leading to uracil misincorporation into DNA. The repair activity by uracil glycosylase can lead to DNA strand breaks, resulting in enhanced mutagenesis and apoptosis.4 Because their roles are soessential, the enzymes in one-carbon metabolism have been considered as the targets for antitumor therapies.5, 6

Figure 1.

Overview of one-carbon metabolism links to methylation reactions and nucleotide synthesis is depicted. The critical enzymes involved in this set of reactions include betaine-homocysteine methyltransferase (BHMT); cystathionine-beta-synthase (CBS); dihydrofolate reductase (DHFR); folate receptor (FOLR); 10-formyltetrahydrofolate dehydrogenase (FTHFD); 5,10-methylenetetrahydrofolate dehydrogenase/5,10-methenyltetrahydrofolate cyclohydrolase/10-formylotetrahydrofolate synthetase (MTHFD1); 5,10-methylenetetrahydrofolate reductase (MTHFR); methionine synthase (MTR); methionine synthase reductase (MTRR); serine hydroxymethyltransferase (SHMT1); reduced folate carrier (SLC19A1/RFC); and thymidylate synthetase (TS). The abbreviations for related nutrients, substrates, or products in the diagram are as follows: B6, vitamin B6; B12, vitamin B12; DHF, dihydrofolate; dTMP, deoxythymidine monophosphate; dUMP, deoxyuridine monophosphate; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; THF, tetrahydrofolate.

Genetic variants of the genes coding key enzymes in one-carbon metabolism may influence their function in one-carbon supply, resulting in aberrant methylation or DNA synthesis inhibition. These variants may also be implicated in the initiation and progression of tumors.3, 7 Molecular epidemiologic studies8-10 concentrating on the risk and outcome of multiple cancers, including lung cancer, have explored functional polymorphisms in key genes in one-carbon metabolism. Only one study on lung cancer prognosis has been carried out by Matakidou et al,8 who investigated the effects of 14 nonsynonymous SNPs mapping to 6 genes in one-carbon metabolism pathway on lung cancer survival with a cohort of 619 Caucasian patients. This study found associations between lung cancer survival and variations in MTHFR, MTHFS, and MTRR, indicating that functional polymorphisms in one-carbon metabolism pathway may modulate the prognosis of lung cancer.

In light of previous evidence, we hypothesized that potentially functional polymorphisms in key genes in one-carbon metabolism pathway may contribute to the prognosis of NSCLC patients. Considering that the effect of single variations may be small, we screened a total of 57 SNPs from 11 genes in the one-carbon metabolism pathway (BHMT, CBS, DHFR, FTHFD, MTHFD1, MTHFR, MTR, MTRR, SHMT1, SLC19A1, and TS) to evaluate their effects on lung cancer survival individually or jointly in a cohort of 568 NSCLC patients in a Chinese population.


Study Population

Since July 2003, patients with histologically confirmed lung cancer were recruited in the Nanjing Lung Cancer Study (NJLCS),11 originally designed as a case-control study, from the Cancer Hospital of Jiangsu Province and the First Affiliated Hospital of Nanjing Medical University, Nanjing, China. All patients were interviewed face-to-face by trained interviewers at the time of recruitment to obtain demographic data (eg, age and sex) and exposure information (eg, smoking status). Each patient donated 5 mL venous blood after signing informed consent. A total of 1099 NSCLC patients were recruited in the NJLCS until April 2008, of which 828 (75.3%) patients had complete, clinical information and responded in the first follow-up at the third month after enrollment.12 Patients who entered the case cohort were prospectively followed-up every 3 months through personal or family contact until death or the latest follow-up. The median follow-up time was 18.8 months up to the last follow-up date, July 2009. During the follow-up period, 131 (15.8%) patients were lost to follow-up; 558 were followed more than 36 months, 435 died of NSCLC, and 5 died from other causes. Those lost to follow-up or dead from other causes were considered as censored data in the analyses. This study was approved by the institutional review boards of Nanjing Medical University.

SNPs Selection

As shown in Figure 1, we focused on 11 key genes in the one-carbon metabolism pathway, including BHMT, CBS, DHFR, FTHFD, MTHFD1, MTHFR, MTR, MTRR, SHMT1, SLC19A1, and TS, and searched the common SNPs (minor allele frequency, MAF ≥ 0.05 in Chinese population) with potential functional significance — ie, those located at 5′ flaking regions, 5′ untranslated regions (UTRs), coding regions, or 3′ UTRs, according to NCBI dbSNPs (build 36, last search date February, 2008). As a result, 57 SNPs were selected for genotyping; the details are shown in Table 1.

Table 1. The Results of SNPs Selection, Genotyping, and Analyses with NSCLC Survival
GeneLocationSNPAllelesReference MAFaIllumina ScoreCall RateMAF in PatientsP for HWEbAnalysisAllelic PcDominant PcRecessive Pc
  • a

    Minor allele frequencies in Chinese population according to NCBI dbSNPs.

  • b

    Hardy-Weinberg equilibrium (HWE) was tested by goodness-of-fit χ2 test.

  • c

    Univariate Cox regression was performed in log-additive, dominant, and recessive models, respectively.

MTHFRExon 12_3UTRrs4846048A:G0.102Pass99.80.0911.000Include.438.338.520
MTHFRExon 12_3UTRrs3737967G:A0.102Pass1000.0821.000Include.120.128.532
MTHFRExon 12_3UTRrs1537516G:A0.100Pass1000.0601.000Include.088.086.676
MTHFRExon 12_3UTRrs4846049C:A0.200Pass1000.172.541Include.603.961.096
MTHFRExon 12_R594Qrs2274976G:A0.100Pass98.40.1063.7×10−21Include.207.173.413
MTHFRExon 8_E429Ars1801131A:C0.200Pass1000.169.441Include.499.862.115
MTHFRExon 5_A222Vrs1801133G:A0.511Pass99.80.443.388Include.556.827.198
MTHFRExon 2_P39Prs2066470C:T0.100Fail--------------
MTHFR5′ flankingrs3737965G:A0.100Pass1000.0841.000Include.181.197.532
MTR5′ flankingrs16834388C:A0.396Pass1000.450.471Include.646.541.911
MTRExon 26_D919Grs1805087A:G0.089Pass1000.101.397Include.155.199.323
MTRExon 33_3UTRrs2853522C:A0.467Pass1000.436.775Include.235.351.316
MTRExon 33_3UTRrs2853523C:A0.227Pass99.80.192.067Include.814.425.289
MTRExon 33_3UTRrs1050993G:A0.222Pass1000.191.066Include.813.424.289
MTRExon 33_3UTRrs3768159A:G0.200Pass99.80.2021.000Include.172.211.363
MTRExon 33_3UTRrs3768160A:G0.209Pass99.10.151.121Include.031.049.128
MTRExon 33_3UTRrs1804742G:A0.067Pass99.80.100.373Include.146.185.328
MTRExon 33_3UTRrs1050996G:C0.227Pass1000.191.058Include.752.370.289
MTRExon 33_3UTRrs10925263A:G0.221Pass1000.247.844Include.090.174.138
FTHFDExon 21_D793Grs1127717A:G0.100Pass1000.121.207Include.790.794.911
FTHFDExon 12_S481Grs2276724A:G0.211Pass81.70.1871.000Exclude------
MTRR5′ flankingrs2966952G:A0.333Pass99.80.342.476Include.044.030.371
MTRRExon 2_I22Mrs1801394A:G0.250Pass1000.255.977Include.299.318.580
MTRRExon 5_S175Lrs1532268G:A0.122Pass65.60.4323.3×10−14Exclude------
MTRRExon 5_L179Lrs161870A:G0.065Pass1000.1921.000Include.141.055.767
MTRRExon 7_K350Rrs162036A:G0.233Pass99.10.1921.000Include.157.063.756
MTRRExon 9_L385Lrs2287779G:A0.122Pass1000.168.589Include.889.996.641
MTRRExon 9_R415Crs2287780G:A0.122Pass1000.168.589Include.889.996.641
MTRRExon 10_P450Rrs16879334C:G0.122Pass1000.168.589Include.889.996.641
MTRRExon 14_H595Yrs10380G:A0.200Pass99.80.1621.000Include.353.171.487
MTRRExon 14_A637Ars1802059G:A0.122Pass1000.173.333Include.923.773.587
MTRRExon 15_3UTRrs9332G:A0.188Pass1000.1621.000Include.391.195.475
MTRRExon 15_3UTRrs8659T:A0.311Pass98.60.3091.000Include.149.087.781
MTRRExon 15_3UTRrs10520873A:G0.182Pass1000.137.495Include.062.077.343
BHMT5′ flankingrs3756548A:G0.089Pass99.80.0671.000Include.116.158.971
BHMT5′ flankingrs562487T:C0.400Fail--------------
BHMTExon 6_R239Qrs3733890G:A0.289Pass98.20.343.772Include.361.347.639
BHMTExon 8_3UTRrs585800A:T0.068Pass1000.0861.000Include.826.650.252
DHFRExon 1_5UTRrs1650697G:A0.422Pass99.50.341.203Include.040.147.044
DHFR5′ flankingrs408626G:A0.489Pass99.80.371.593Include.090.155.185
MTHFD15′ flankingrs1076991A:G0.300Pass79.10.222.712Exclude------
MTHFD1Exon 6_K134Rrs1950902G:A0.222Pass99.80.2961.000Include.058.189.043
MTHFD1Exon 22_R653Qrs2236225G:A0.222Pass99.80.234.942Include.632.308.400
SHMT1Exon 12_3UTRrs12952556A:G0.091Pass1000.059.249Include.497.671.170
SHMT1Exon 12_3UTRrs1979276G:A0.089Pass99.80.060.272Include.400.551.168
SHMT15′ flankingrs638416G:C0.476Fail--------------
SHMT15′ flankingrs643333C:A0.083Pass1000.057.525Include.573.648.501
TYMS5′ flankingrs2853741A:G0.411Pass1000.462.906Include.898.964.864
TYMSExon 7_3UTRrs699517A:G0.300Pass99.80.329.726Include.824.879.813
TYMSExon 7_3UTRrs2790A:G0.354Pass1000.392.864Include.847.440.486
CBSExon 17_3UTRrs706209A:G0.329Pass99.80.375.781Include.376.170.864
CBSExon 17_3UTRrs706208G:A0.354Pass98.00.2501.2×10−20Include.270.270--
CBSExon 12_A360Ars1801181T:C0.375Fail--------------
SLC19A1Exon 6_3UTRrs1051298A:G0.466Pass1000.468.606Include.686.512.989
SLC19A1Exon 6_3UTRrs1051296C:A0.466Pass99.80.466.834Include.693.468.911
SLC19A1Exon 2_H27Rrs1051266A:G0.500Pass99.80.497.817Include.480.244.994
SLC19A1Exon 2_5UTRrs3177999G:A0.500Pass99.80.480.215Include.338.272.661


Genotyping was performed by using Illumina Golden Gate platform (Berkeley Biotech, Taizhou, China). All selected SNPs were first evaluated for chip design, and 4 SNPs with scores <.50 were excluded (Table 1). Before genotyping, DNA quantity and quality were assessed using both fluorometer and agarose gel electrophoresis. Finally, 568 samples reaching the criteria were selected for genotyping. There were no significant differences for age, sex, histology, or chemotherapy or radiotherapy between the patients included (n = 568) and those excluded (n = 260; P > .05), but the proportion of stage and surgical status was different between the 2 groups (P < .05). Information on genotyping assay is available upon request. Quality control was followed according to standard operation criteria (ie, 1 blank well and 3 repeated samples) described in our previous study.13 As shown in Table 1, genotyping results were further evaluated for quality control. Subjects with genotyping call rate <90% (n = 4 samples) were excluded from further analysis. Those SNPs with genotyping call rate <90% (n = 3 SNPs) were also removed from final analysis. Ultimately, 564 NSCLC patients and 50 SNPs (from 11 genes in the one-carbon metabolism pathway)were included in this study.

Statistical Analysis

The chi-square goodness-of-fit test was used to assess Hardy-Weinberg equilibrium by comparing observed genotype frequency to expected frequency. Median survival time (MST) was calculated from the date of NSCLC diagnosis to date of patient's death or last follow-up. The Kaplan-Meier method and log-rank test were used to compare survival time in different subgroups categorized by patient characteristics, clinical features, and genotypes. Univariate and multivariate Cox regression analyses were performed to estimate crude and adjusted hazard ratios (HR) and their 95% confidence intervals (CIs). Cox stepwise regression was also conducted to determine predictive factors for NSCLC prognosis, with a significance level of 0.050 for entering and 0.051 for removing the respective variables. Multiplicative interactions were assessed by Cox regression. Heterogeneity between subgroups was assessed with the chi-square–based Q-test and was considered significant when P < 0.10. All statistical analyses were carried out by Statistical Analysis System software, version 9.1.3, (SAS Institute, Cary, NC).


Patient Characteristics and Clinical Features

As shown in Table 2, 564 patients were included in the final analyses, of which 308 deaths were observed during the whole follow-up period. The overall median survival time was 25.0 months. In the patient cohort, the median age was 60 years (range, 25-83 years) and 76.6% (n = 432) were males. For histological type, there were 351 (62.2%) adenocarcinomas, 182 (32.3%) squamous cell carcinomas, and 31 (5.5%) others, including large cell, undifferentiated, and mixed-cell carcinomas. Survival was not statistically different among strata by age (P = .762), sex (P = .804), or histology (P = .495) (Table 2). As expected, survival time was significantly shorter as clinical stage increased (log-rank P = 8.4 × 10−27; Ptrend = 3.4 × 10−24). Tumor resection could improve prognosis for NSCLC patients (P = 1.1 × 10−23); however, it seemed that treatments with chemotherapy or radiotherapy did not (P = .110). Multivariate Cox regression analyses revealed that clinical stage remained the pivotal predictor for NSCLCL survival (HRtrend, 1.70; 95% CI, 1.45-1.97), whereas combining surgical resection with chemotherapy or radiotherapy could benefit NSCLC patients (HR, 0.57; 95% CI, 0.43-0.75 and HR, 0.63; 95% CI, 0.45-0.89, respectively).

Table 2. Patient Characteristics and Clinical Features
VariablePatients n=564Deaths n=308MSTLog-rank P
No.%No.No. of Months
  • a

    Other carcinomas include large cell, undifferentiated, and mixed-cell carcinomas.

Age    .762
Sex    .804
Histology    .495
 Squamous cell18232.39724.2 
Stage    8.4×1027
Surgical operation    1.1×1023
Chemotherapy or radiotherapy    .110

Genotyping Results and Their Associations With NSCLC Survival

Details of the 50 SNPs included in the final analyses are summarized in Table 1. Univariate Cox regression analysis was used to test the association of each SNP with NSCLC survival in log-additive, dominant, and recessive models. Four SNPs located at different genes were identified (allelic P = .031,.044,.040, and .058 for MTR rs3768160 A>G, MTRR rs2966952 G>A, DHFR rs1650697 G>A, and MTHFD1 rs1950902 G>A, respectively). As suggested in Table 3, variant alleles were significantly associated with favorable survivals of NSCLC for MTR rs3768160 (allelic HR, 0.78; 95% CI, 0.62-0.98), MTRR rs2966952 (allelic HR, 0.84; 95% CI, 0.71-0.99), and DHFR rs1650697 (allelic HR, 0.83; 95% CI, 0.70-0.99). Prognosis was unfavorable with borderline significance for MTHFD1 rs1950902 variant allele (allelic HR, 1.18; 95% CI, 0.99-1.40). When we used dominant and recessive models (Table 3), it seemed that rs3768160 (log-rank P = .048) and rs2966952 (log-rank P = .030) were effective in the dominant model, whereas rs1650697 (log-rank P = .042) and rs1950902 (log-rank P = .042) were significant in the recessive model. However, when adjusted for age, gender, histology, stage, and chemo-or radiotherapy, these associations were not statistically significant.

Table 3. Polymorphisms of Key Genes in One-carbon Metabolism Pathway and NSCLC Survival
GenotypePatientsDeathsMST (Months)Crude HR (95% CI)Adjusted HR (95% CI)a
  • a

    Adjusted for age, sex, histology, stage, surgical operation and chemotherapy or radiotherapy.

MTR rs3768160n=559n=303   
 AG1336131.30.80 (0.61-1.07)0.88 (0.66-1.17)
 GG18836.80.55 (0.27-1.12)0.65 (0.32-1.33)
 G allele   0.78 (0.62-0.98)0.85 (0.68-1.07)
 AG + GG1516932.90.76 (0.58-0.99)0.85 (0.64-1.11)
MTRR rs2966952n=563n=307   
 AG24512428.20.79 (0.62-1.00)0.83 (0.65-1.06)
 AA703427.70.76 (0.52-1.10)0.87 (0.59-1.26)
 A allele   0.84 (0.71-0.99)0.89 (0.75-1.06)
AG + AA31515828.00.78 (0.62-0.98)0.84 (0.67-1.05)
DHFR rs1650697n = 561n = 305   
 AG26714225.00.90 (0.71-1.14)0.84 (0.66-1.06)
 AA582452.90.62 (0.40-0.95)0.62 (0.40-0.96)
 A allele   0.83 (0.70-0.99)0.81 (0.67-0.97)
 GG + AG50328124.21.001.00
 AA582452.90.65 (0.43-0.99)0.68 (0.45-1.04)
MTHFD1 rs1950902n=563n=308   
 AG23512823.51.10 (0.87-1.39)1.10 (0.86-1.39)
 AA493318.71.51 (1.04-2.21)1.41 (0.97-2.07)
 A allele   1.18 (0.99-1.40)1.16 (0.97-1.37)
 GG + AG51427525.91.001.00
 AA493318.71.45 (1.01-2.09)1.37 (0.94-1.95)

Combined Effects and Stratified Analyses

In view of the modest effect of a single variant, we combined all 4 SNPs to assess effect on NSCLC survival. As shown in Figure 2 and Table 4, we found that the more favorable genotypes the patients carried, the longer they survived, suggesting a locus-dosage effect between combined genotypes and survival (Ptrend = 6.9 × 10−5). Because the number of patients carrying 0 or 4 favorable genotypes was very limited (n = 15 and 9, respectively), we trichotomized the patients by combining those with 0 or 1 and those with 3 or 4, respectively. We found that patients with 0-1 had a MST of 19.4 months, whereas those with 2 or 3-4 favorable genotypes had MSTs of 27.7 and 59.3 months, respectively (log-rank P = 1.2 × 104). After adjusting for age, sex, histology, stage, surgical, and chemo-or radiotherapy, risk of death significantly decreased by 22%, or 42% for those with 2 or 3-4 favorable genotypes, with a significantly ranked trend and adjusted HR of 0.77 (95% CI, 0.65-0.91; P = .002).

Figure 2.

Kaplan-Meier plots of survival are arranged by combined genotypes of one-carbon metabolism-related genes for NSCLC: Chart 2a, quinquefid group (log-rank P = 1.1 × 10−3), Chart 2b, trichotomized group (log-rank P = 1.2 × 10−4); the numbers “0-4” denote the amount of favorable genotypes.

Table 4. Combined Effects of Polymorphisms in One-Carbon Metabolism-Related Genes on NSCLC Prognosis
Combined Genotypesa (Favorable Genotypes Carried)PatientsbDeathsMST (Mo)Crude HR (95% CI)Adjusted HR (95% CI)c
  • a

    rs3768160AG/GG, rs2966952AG/AA, rs1650697AA and rs1950902GG/AG are presumed to be favorable genotypes.

  • b

    Total 554 patients were successfully genotyped for the above 4 SNPs.

  • c

    Adjusted for age, sex, histology, stage, surgical operation and chemotherapy or radiotherapy.

  • d

    NA (not available) means that median survival time (MST) could not be calculated.

 116711119.41.19 (0.60-2.34)0.98 (0.49-1.94)
 227014027.70.85 (0.43-1.67)0.76 (0.39-1.51)
 3933659.30.55 (0.27-1.15)0.57 (0.28-1.20)
 493NAd0.59 (0.16-2.20)0.54 (0.14-2.04)
Locus trend   0.75 (0.65-0.86)0.80 (0.68-0.93)
Trichotomized group (favorable genotypes carried)
 227014027.70.73 (0.57-0.93)0.78 (0.61-1.00)
 3-41023959.30.48 (0.33-0.68)0.58 (0.41-0.84)
Locus trend   0.70 (0.59-0.83)0.77 (0.65-0.91)

We then performed stratification analyses by age, sex, histology, stage, surgical, and radio-or chemotherapy status, and found the protective effect of combined genotypes was more pronounced in patients who had undergone surgery (adjusted HR, 0.65; 95% CI, 0.51-0.82) than those who did not (adjusted HR, 0.95; 95% CI, 0.75-1.22; P = .029 for heterogeneity test). However, no significant differences between associations were detected in other subgroups (data not shown).

Multivariate Cox Regression Model for NSCLC Survival

The variables with a P value of <.10 in univariate analysis were subjected to stepwise Cox regression analysis. Four variables (P = 8.5, 10−12, 2.5 = 10−4, 3.4 × 10−3, and .019 for clinical stage, surgery, combined genotypes, and chemotherapy or radiotherapy treatment, respectively) proved significant predictors in our patient cohort (data not shown). When age and sex were enforced into the model as fixed factors, these 4 variables remained statistically significant (Table 5). Therefore, the combined genotypes for the 4 SNPs in one-carbon metabolism under discussion may compose an independent predictor for NSCLC survival with adjusted HRtrend of 0.78 (95% CI, 0.66-0.92).

Table 5. Results of Multivariate Cox Regression Analyses for NSCLC Survival
VariablesaβSEHR95% CIP
  • a

    In the multivariate regression model, age was entered as continuous variable, whereas stage was treated as ranked variable.

  • b

    ”0-4” Means the number of favorable genotypes carried and rs3768160AG/GG, rs2966952AG/AA, rs1650697AA, and rs1950902GG/AG are presumed as favorable genotypes.

Sex (female vs male)-.140.140.870.66-1.15.322
Stage (IV vs III vs II vs I).520.081.681.45-1.955.4×10−12
Surgical operation (yes vs no)-.520.140.600.45-0.792.4×10−4
Chemotherapy or radiotherapy (yes vs no)-.410.180.660.47-0.94.020
Combined genotypes (3-4 vs 2 vs 0-1)b-.250.090.780.66-0.923.7×10−3


In this lung cancer follow-up study, we screened 57 potentially functional SNPs from 11 key genes in one-carbon metabolism and investigated the effects of these variants on NSCLC prognosis in a Chinese population. We found that 4 SNPs— MTR rs3768160 A>G, MTRR rs2966952 G>A, DHFR rs1650697 G>A, and MTHFD1 rs1950902 G>A— were associated with NSCLC prognosis. The combined genotypes of these 4 SNPs showed a locus-dosage effect on NSCLC survival and were identified as an independent prognostic factor for NSCLC in the final multivariate Cox regression model.

As illustrated in Figure1, MTR, a vitamin B12 dependent enzyme essential for maintaining adequate intracellular folate pools, catalyzes the remethylation of homocysteine to methionine.14MTRR maintains the activity of MTR by reduction reactions and transfers the methyl group of methyltetrahydrofolate to homocysteine via methionine synthase-methylcobalamin as an intermediate methyl carrier.15DHFR catalyzes the reduction reaction of dihydrofolate into tetrahydrofolate, which is essential for conversion of ingested folic acid to reduced physiological folates before being used in cell metabolism.16DHFR is the target of methotrexate (MTX) and pemetrexed,17 two important antifolate chemotherapeutic agents widely used in the treatment of multiple malignancies, including lung cancer.18MTHFD1, a trifunctional cytosolic enzyme acting as 5,10-methylenetetrahydrofolate dehydrogenase, 5,10-methenyltetrahydrofolate cyclohydrolase, and 10-formylotetrahydrofolate synthetase, catalyzes 3 sequential reactions in the interconversion of one-carbon derivatives of tetrahydrofolate that are substrates for biosynthesis of thymidylate, purine nucleotides, and methionine.19

Because MTR, MTRR, DHFR, and MTHFD1 are key players in one-carbon metabolism, the variants in these genes may modulate their respective enzyme activities via transcriptional or post-transcriptional regulation mechanisms. These genetic variants may also modify the initiation and progression of tumors, a possibility which has been extensively investigated in different kinds of tumors, including lung.8-10MTR A2756G (rs1805087) and MTRR A66G (rs1801394), both related to concentration of plasma homocysteine,20, 21 were reported to be involved in the development of lung cancer,9 whereas a Ser to Leu substitution at condon 175 in MTRR gene (rs1532268) was identified as a predictor of NSCLC prognosis in Caucasian women.8 Mishra et al22 reported a SNP 829C>T (rs1677669) near the miR-24 binding site in the 3′UTR of DHFR gene, which affects DHFR expression by interfering with miR-24 function. This polymorphic activity resulted in DHFR overexpression23 and methotrexate resistance. However, we did not observe a significant association of MTR rs1805087 and MTRR rs1801394 with NSCLC survival in our patient cohort. In addition, the minor allele frequencies of rs1532268 and rs1677669 are very low or nil in the Chinese population and, therefore, were not included in this study.

In the present study, we found that the variant alleles of MTR rs3768160 A>G, MTRR rs2966952 G>A, and DHFR rs1650697 G>A were possible, favorable biomarkers for NSCLC survival, whereas that of MTHFD1 rs1950902 G>A was unfavorable. The SNP rs3768160 A>G is located at 3′UTR of MTR with multiple miRNAs binding sites (predicted by This SNP in the vicinity of miRNAs binding site may interfere with miRNA function leading to differential expression of MTR,24 which may be the biological function for rs3768160 A>G involved in NSCLC prognosis. The SNP rs2966952 G>A, residing at the 5′ flanking region of MTRR gene (1187 bp apart from transcriptional start position), is predicted to destroy the binding site of transcriptional factor C/EBPb (according to TFSEARCH,, which may regulate the expression level of MTRR transcript and subsequently influence its enzyme activity. SNPs rs1650697 G>A and rs1950902 G>A are localized to exon 1 of DHFR and exon 6 of MTHFD1, respectively. The former is in the 5′ UTR and the latter results in an amino acid substitution from Arg to Lys at codon 134. However, the biological function of these SNPs remains unclear and needs to be clarified in future studies.

Although each of the above 4 SNPs (MTR rs3768160 A>G, MTRR rs2966952 G>A, DHFR rs1650697 G>A, MTHFD1 rs1950902 G>A) showed some potential for predicting NSCLC prognosis, they all seemed unstable, and none of them remained significant after adjustment. One possible explanation is that sample size in the current study was not large enough, which inhibited statistical analysis. Another reason is that the effect of a single variant is too small for both tumor initiation and progression, which was also shown in this study. Considering the limited effect of individual genetic variants, we tested them jointly using a pathway approach. Interestingly, in our patient cohort, we observed a significant, combined effect from the 4 SNPs on NSCLC prognosis in a locus-dosage manner; the combined genotypes turned out to be an independent prognosis factor. Specifically, the combined effect seemed more evident in patients with surgical resection, most of whom (308/360, 85.6%) were early stage patients. However, given the current sample size, these results were informative rather than conclusive, and need to be clarified by other studies (eg, biological interpretation and population confirmation).

Several potential limitations of the present study warrant consideration. First, although we comprehensively investigated a large number of SNPs in a one-carbon metabolism pathway, other potentially functional genetic variants were not included (eg, insert/deletion, tandem repeat polymorphisms) or failed to be genotyped (eg, SNPs with design score <0.50) by restriction with a genotyping platform. Second, 50 SNPs were included in the final analyses, which might have resulted in false-positive findings due to multiple comparisons. After Bonferroni corrections, none of the associations for single SNPs were still significant with the significance level of .001. False-positive report probability analysis (FPRP)25 revealed that all positive findings remained significant, assigning prior probability above. 1. However the associations may be less convincing with the prior probability of .01, when considering .5 as the FPRP level criterion derived from the potential, functional significance of the selected SNPs. Cautious interpretation of our results is needed before they are replicated by other studies. Third, 264 patients with clinical and follow-up information were excluded from the genotyping analysis because of low quantity DNA, which could potentially influence final estimation due to selection bias.

In conclusion, this study provides population evidence that genetic variants in one-carbon metabolism may contribute to NSCLC prognosis, especially through combining multiple polymorphisms. However, larger, well designed studies with different populations and functional evaluations are warranted to confirm these findings.


This work was supported in part by Key Grant of National Natural Science Foundation of China (30730080); National Outstanding Youth Science Foundation of China (30425001), National Natural Science Foundation of China (30972541, 30901233), Natural Science Foundation of Jiangsu Province (BS2006005) and Jiangsu Key Discipline of Medicine (XK200718).