Purpose: Colorectal cancer is routinely treated with a 5-fluorouracil (5-FU)–based chemotherapy. 5-FU incorporates into DNA, and the base excision repair (BER) pathway specifically recognizes such damage. We investigated the association of single-nucleotide polymorphisms (SNP) in the 3′-untranslated regions (UTR) of BER genes, and potentially affecting the microRNA (miRNA) binding, on the risk of colorectal cancer, its progression, and prognosis. SNPs in miRNA-binding sites may modulate the posttranscriptional regulation of gene expression operated by miRNAs and explain interindividual variability in BER capacity and response to 5-FU.

Experimental Design: We tested 12 SNPs in the 3′-UTRs of five BER genes for colorectal cancer susceptibility in a case–control study (1,098 cases and 1,459 healthy controls). Subsequently, we analyzed the role of these SNPs on clinical outcomes of patients (866 in the Training set and 232 in the Replication set).

Results: SNPs in the SMUG1 and NEIL2 genes were associated with overall survival. In particular, SMUG1 rs2233921 TT carriers showed increased survival compared with those with GT/GG genotypes [HR, 0.54; 95% confidence interval (CI), 0.36–0.81; P = 0.003] in the Training set and after pooling results from the Replication set. The association was more significant following stratification for 5-FU–based chemotherapy (P = 5.6 × 10−5). A reduced expression of the reporter gene for the T allele of rs2233921 was observed when compared with the common G allele by in vitro assay. None of the genotyped BER polymorphisms were associated with colorectal cancer risk.

Conclusions: We provide the first evidence that variations in miRNA-binding sites in BER genes 3′-UTR may modulate colorectal cancer prognosis and therapy response. Clin Cancer Res; 19(21); 6044–56. ©2013 AACR.

This article is featured in Highlights of This Issue, p. 5809

Translational Relevance

Base excision repair (BER) is the main pathway involved in resolving DNA lesions, including misincorporated uracil (or its analogs), as a result of chemotherapy with 5-fluorouracil (5-FU). A fine modulation of BER genes expression, such as the posttranscriptional regulation exerted by microRNAs (miRNA), could affect the efficiency of this repair system. The presence of single-nucleotide polymorphisms (SNP) within 3′-untranslated regions (UTR) of target genes could alter the binding with specific miRNAs, modulating gene expression and ultimately affecting cancer susceptibility, prognosis, and therapy outcomes. This study comprehensively evaluated the associations between SNPs in miRNA target–binding sites of BER genes and clinical outcomes in patients with colorectal cancer. Variations in the SMUG1 gene provided biologically plausible predictors of survival, with an additional effect of 5-FU chemotherapy. The validation of those SNPs in interaction with the clinical variables may allow the identification of patients with colorectal cancer with a specific prognosis, which can aid the design of personalized cancer therapy.

Colorectal cancer is the second most common cancer in Western countries (1). Although it is a preventable and potentially curable malignancy, it still represents the second leading cause of cancer-related death among men and women (2).

Impaired DNA repair capacity and DNA damage accumulation may lead to cancer development. In particular, reduced DNA repair activity due to differential genetic background in DNA repair genes may modulate individual cancer risk (3). Base excision repair (BER) is the main pathway involved in resolving spontaneous, alkylating, and oxidative DNA lesions (4). Defects in BER have been linked to several diseases and the disruption of BER coordination or efficacy can potentially foster tumorigenesis. Subtle variations in BER genes may be related to disease proneness or susceptibility, and this is particularly important when considering relevant exposures or during the cumulative life span of an individual (5). In our previous study, subjects homozygous for the variant alleles of two exon-localized polymorphisms of BER genes, APE1 and hOGG1, were at an increased risk of colorectal cancer (6). Such variant alleles were also associated with significantly lower BER rates in healthy subjects, suggesting the importance of DNA repair in cancer susceptibility, as well as gene–gene and gene–environment interactions (7).

BER also removes uracil (or its analogs) misincorporated into the DNA as a result of chemotherapy with 5-fluorouracil (5-FU; ref. 8). 5-FU is widely used in the treatment of a variety of solid tumors, including colorectal cancer (9). Intracellular metabolites of 5-FU inhibit thymidylate synthase exerting a cytotoxic effect or they may incorporate into RNA and DNA and ultimately activate apoptosis. In this sense, DNA repair pathways, particularly BER, deserve more attention as they represent important factors modulating the cell response to 5-FU and its resistance (8).

Notably, a fine regulation of BER genes, such as posttranscriptional and posttranslational regulation could affect the efficiency of this repair system. In the last decade, there has been increasing interest in the role of posttranscriptional regulation of gene expression by microRNAs (miRNA) and their possible effect on cancer risk and clinical outcomes. miRNAs are a class of short noncoding RNAs of approximately 20 to 25 nucleotides that modulate gene expression through canonical base pairing between the seed sequence and its complementary seed match sequence, mainly located in the 3′-untranslated region (UTR) of target mRNAs (10). Approximately, one third of the protein-coding genes are controlled by miRNAs, thus almost all cellular pathways are directly or indirectly influenced by these molecules (11, 12). Aberrant miRNA expression and/or function is frequently observed in many malignancies, including colorectal cancer (13). Furthermore, sequence variations, such as single-nucleotide polymorphisms (SNP), in the seed region or in a target gene site could alter the expression of miRNA targets with consequences on protein translation (14). When SNPs occur in 3′-UTRs, they may interfere with mRNA stability and translation by altering polyadenylation, protein::mRNA and miRNA::mRNA regulatory interactions, which may lead to altered cancer susceptibility (15–17). Polymorphic target sites for miRNAs may play a role in modulating important pathways such as DNA repair (18), thereby potentially affecting the individual risk of colorectal cancer (19, 20).

The aim of the present study was to investigate the role of polymorphisms in potential miRNA-binding sites within the 3′-UTRs of BER genes in modulating the risk of colorectal cancer, its progression, and prognosis. From all identified polymorphisms within the 3′-UTRs, using a series of a priori hypotheses, 12 SNPs in five BER genes were selected and analyzed in a case–control study on 1,098 colorectal cancer cases and 1,459 controls from the Czech Republic. In this country, colorectal cancer constitutes a serious health problem as is among the countries with the highest rate of incidence and mortality worldwide (21). The same set of BER SNPs were additionally investigated for their relationship to the patients' prognosis and for the effect of 5-FU–based therapy on 1,098 patients (866 in the Training set and 232 in the Replication set) with a detailed follow-up. In addition, the effect of an identified relevant SNP on the regulation of gene expression was further tested by an in vitro functional assay.

Study population and data collection

Blood samples from 1,098 subjects were collected among patients with histologically confirmed colorectal cancer, recruited between September 2003 and October 2010 from several oncological departments in the Czech Republic [Prague (three), Benesov, Brno, Liberec, Ples, Pribram, Usti nad Labem, and Zlin]. Two control groups, whose samples were collected at the same time of cases recruitment, were included in the study. The first group consisted of 688 hospital-based individuals admitted to five of the above-mentioned gastroenterologic departments that had negative colonoscopy results for malignancy or idiopathic bowel diseases (CFCC, cancer-free colonoscopy inspected controls). The reasons for undergoing the colonoscopy were: (i) positive fecal occult blood test, (ii) hemorrhoids, (iii) abdominal pain of unknown origin, and (iv) macroscopic bleeding. The second group of controls consisted of 781 healthy blood donor volunteers (HBDV) collected from a blood donor center in Prague. All individuals were subjected to standard examinations to verify the health status for blood donation and were cancer-free at the time of the sampling. A detailed description of colorectal cancer cases and controls is reported elsewhere (22, 23).

All subjects were informed and provided written consent to participate in the study and to approve the use of their biologic samples for genetic analyses, according to the Helsinki declaration. The design of the study was approved by the local Ethics Committee. Study subjects provided information on their lifestyle habits, body mass index (BMI), diabetes, and family/personal history of cancer, using a structured questionnaire to determine demographic characteristics and potential risk factors for colorectal cancer.

Follow-up of the patients

Eight hundred sixty-six of the colorectal cancer cases were monitored with follow-ups until August 31, 2011. They represented the initial Training set recruited in nine oncological departments in Prague and throughout the Czech Republic. For these subjects, clinical data at the time of diagnosis, including location of the tumor, UICC (Union for International Cancer Control) tumor–node–metastasis (TNM) stage system, grade, and adjuvant chemotherapy treatment were collected, as well as information about distant metastasis, relapse, and date of death. Because 223 patients had incomplete clinical information, they were excluded from the analyses. Three hundred and nineteen colorectal cancer cases received a 5-FU–based adjuvant regimen as first-line postoperative therapy. The therapy consisted of either a Mayo regimen, delivered as a bolus infusion of 5-FU (425 mg/m2) and leucovorin (10 mg/m2) for 5 days every 4 weeks six times or a simplified DeGramond regimen that consisted of a 2-hour intravenous infusion of leucovorin (200 mg/m2), then a 5-FU intravenous bolus (400 mg/m2) followed by a 46-hour 5-FU continuous intravenous infusion (2,400–3,000 mg/m2). Three hundred twenty-four subjects did not receive any adjuvant chemotherapy after surgery. In this study, the outcome variables measured were 5-FU–based chemotherapy, overall survival (OS; time from diagnosis until death or censorship), and event-free survival (EFS; time of surgery or end of chemotherapy until date of relapse, death, or censorship).

A second group consisting of 232 patients with colorectal cancer (Replication set) was recruited between 2008 and 2010 in the surgical department of Thomayer Hospital (Prague, Czech Republic). Similar to the Training set, detailed information were available on clinical data at the time of diagnosis, including location of the tumor, UICC TNM stage system, grade, and adjuvant chemotherapy treatment (5-FU chemotherapy regimens were similar to those used for the Training set). Follow-up for this second group of patients was performed until March 31, 2013.

Selection of candidate genes

From the complete list of DNA repair genes available online (http://sciencepark.mdanderson.org/labs/wood/DNA_Repair_Genes.html#BER; March 2012 version), 16 genes were retrieved in the BER pathway.

Selection of the SNPs in miRNA target–binding sites

For each BER gene, SNPs within target-binding sites for miRNAs were identified by using the freely available software MicroSNiper [http://cbdb.nimh.nih.gov/microsniper (24); updated at March 2012]. The 50 detected SNPs were tested for minor allele frequency (MAF; >5% in Caucasian populations) in the SNP database (dbSNP; http://www.ncbi.nlm.nih.gov/SNP/) to reach an appropriate representation of all genotypes in our set of cases and controls. The selection of SNPs was primarily done on the basis of HAPMAP CEU population; whenever this was not possible, other reference populations were checked (i.e., 1,000 genomes, phase I, CEU population). SNPs with the required MAF were further tested for the possibility to be in linkage disequilibrium using HaploView (v. 4.2) with the data from HapMap v. 3, release R2 in the CEU population. The workflow for the selection of the SNPs is depicted in Fig. 1.

Figure 1.

Workflow strategy for selection and analyses of SNPs residing in potential miRNA-binding sites in 3′-UTR of BER genes.

Figure 1.

Workflow strategy for selection and analyses of SNPs residing in potential miRNA-binding sites in 3′-UTR of BER genes.

Close modal

SNP genotyping

Genomic DNA was isolated from peripheral blood lymphocytes using standard procedures. The DNA from cases and controls was randomly placed on plates where an equal number of samples could be run simultaneously. The 12 selected SNPs were genotyped using the KASP genotyping assay, a competitive allele-specific PCR SNP genotyping system (LGC Genomics). For quality control purposes, duplicate samples (5% of the total numbers of samples) were repeated for each SNP, non-template controls were included in each plate, and Hardy–Weinberg equilibrium test was performed.

DNA cloning and in vitro assay

A dual-luciferase reporter assay was used to investigate whether for SMUG1 rs2233921 alleles were associated with differential gene expression. Initially, a 640 bp fragment of the 3′-UTR region of SMUG1 containing the G-allele of rs2233921 was PCR-amplified. The PCR primers were both 41 bp long. The twenty bases at the primers 3′ ends were specific to the region to be amplified, whereas the 15 bases at the 5′ ends were homologous to either side of the XhoI restriction site within the multiple cloning site of the pmirGLO vector (Promega). Between the two sequences, each primer was also designed to include a XhoI restriction site sequence (C⁁TCGAG). The complete sequences were: sense primer, AACGAGCTCGCTAGCCTCGAGGTGCCCTTGGGGCCTTGCAT; and antisense primer, CAGGTCGACTCTAGACTCGAGTACAGATGAGGAGCCTGAGG. The vector was linearized with XhoI (NEB Inc.) and the PCR product was cloned downstream from the firefly luciferase (Photinus pyralis) reporter gene, using the Clone EZ PCR Cloning Kit (GenScript). As suggested by the manufacturers, competent cells DH5α (NZYTech) were used for transformation after the cloning reaction. To obtain a vector with a SMUG1 3′-UTR bearing the T allele of rs2233921, the construct underwent site-specific mutagenesis using the QuickChange Lightning Site-Directed Mutagenesis Kit (Agilent). The sequences of the mutagenic primers were: sense, TGACAACACATCTCCTGTACTGGAGCAAAAGGTCC; and antisense, GGACCTTTTGCTCCAGTACAGGAGATGTGTTGTCA. Following the digestion of the parental (methylated) supercoiled dsDNA with Dpn I, XL10-Gold ultra-competent cells (Agilent) were used for transformation.

For the functional assay, HCT-116 cells were plated at a density of approximately 5 × 104 cells per well in 24-well plates and incubated overnight at 5% CO2, 37°C in a humidified incubator. Cells were transiently transfected at about 80% confluence using 3 μL of Polyfect transfection reagent (Qiagen), and 0.4 μg of the chimeric construct carrying the T or the G allele. The pmirGLO vector contains the luciferase gene from Renilla reniformis (hRluc-neo), acting as a control reporter to normalize transfection efficiency. Three independent experiments were carried out. Six replicates of each experimental point were performed in each experiment. Twenty-four hours after transfection, cells were washed with a PBS solution and lysed with 100 μL of Passive Lysis Buffer (PLB; Promega) for optimal stability of the firefly and Renilla luciferase reporter enzymes. The culture vessel was shaken for 15 minutes at room temperature. The lysates were then transferred to 1.5-mL microcentrifuge tubes and centrifuged at 13,000 rpm for 1 minute to pellet the debris. Supernatants were transferred to clean tubes and used for measuring the activity of firefly and Renilla luciferases. The assays were carried out using the Dual-Luciferase Reporter Assay Kit (Promega). Because the same vector carries both the luciferase and the Renilla genes, the experimental variability was greatly reduced allowing a more precise evaluation of small biologic differences such as those reflecting the effect of the different tested alleles.

For each transfection, luminescence intensity was evaluated by a luminometer (Optima FluoStar, BMG), and luciferase activities were averaged from four measurements. The luminescence intensities of firefly and Renilla luciferase of the nontransfected cells (background) were subtracted from the values obtained for the transfected cells with the pmiRGLO vector containing the 3′-UTR. The luminescence of the Renilla luciferase was used as the reference value to calculate the value of firefly luciferase (Luc/Ren ratio of luminescence).

Statistical analyses

Chi-square test (1 degree of freedom), with a type-I error threshold set at α = 0.05, was used to verify whether the genotypes were in Hardy–Weinberg equilibrium in controls. The multivariate logistic regression (MLR) analysis was used to test the association between genotypes and risk of colorectal cancer. The covariates analyzed for inclusion in the multivariate model were sex, age, smoking habit (non-smokers vs. smokers and ex-smokers), BMI, familial history of colorectal cancer, education level (high, intermediate, and low), and living area (country, suburbs, and town). The associations between SNPs and colorectal cancer risk were calculated by estimating the ORs and their 95% confidence intervals (CI), adjusted for both continuous and discrete covariates. For all the genotypes, regression coefficients for additive model were estimated. For all SNPs with significant P values per genotype, the dominant or recessive models were also calculated. The model with the highest likelihood was additionally checked for the significance of possible interaction terms in the MLR analysis. Statistical analyses were performed using R (http://www.rproject.org).

OS in patients with colorectal cancer was evaluated using the date of death or the date of follow-up termination as the endpoint. For the EFS, in patients who did not have distant metastasis at the time of diagnosis, date of relapse, death, or end of the study was used as the endpoint of follow-up. EFS was defined as the time from surgery/end of therapy to the occurrence of distant metastasis, recurrence, or death, whichever came first. The survival curves for OS and EFS were derived by the Kaplan–Meier method (R version 2.14-2, Survival package). The relative risk of death was estimated as HR using Cox regression (R version 2.14-2, Survival package). Multivariate survival analyses were adjusted for age, gender, TNM, and chemotherapy.

For the in vitro assays, the ratios (Luc/Ren) of the measurements of luminescence, each subtracted of its respective background, were compared between genotypes using the multifactor ANOVA with interactions (MANOVA), where “experiment” and “genotype” were entered as independent factors in the model. The statistical tests were two-tailed and carried out using Statgraphics Centurion software (StatPoint Technologies).

SNP selection and prediction of their effects on target sites

Out of 16 genes involved in the BER pathway, two did not present any SNPs within their 3′-UTRs (MUTYH and MPG), whereas 50 SNPs were identified in the remaining 14 genes. A SNP in OGG1 (rs7373786) was not located within a known miRNA-binding site. All 49 remaining SNPs were tested for MAF (>5% in Caucasian populations) in dbSNP. Thirty-three SNPs did not reach the required MAF, and therefore were excluded from the study. An additional four SNPs were found to be in the same haplotype block (D′, 1; r2, 1) of other identified SNPs (NEIL2 rs804292 with rs1534862; NEIL3 rs1055677 and rs34652756 with rs1055678; and rs2740438 with rs2740439). Therefore, only one SNP for each group was considered (rs1534862, rs1055678, and rs2740439, respectively). The final set of 12 selected SNPs within putative miRNA-binding sites in the 3′-UTR of five BER genes is reported in Table 1.

Table 1.

Selected SNPs in the 3′-UTR of BER genes, with MAF >0.05 (in Caucasians) and within miRNAs target

Gene IDdbSNP IDPositionVariationMAF (population studied)
SMUG1 rs2233921 80 G/T T = 0.482 (HAPMAPCEU
 rs971 422 G/A A = 0.320 (HAPMAPCEU
MBD4 rs2307285 383 A/G G = 0.197 (PDR90
 rs2307294 447 G/C C = 0.100 (NIH-PDR90
NEIL2 rs1534862 21 C/T T = 0.219 (HAPMAPCEU
 rs804292a 133 A/G G = 0.292 (HAPMAPCEU
 rs6997097 190 T/C C = 0.055 (PDR90) 
 rs8191670 356 T/C C = 0.25 (PDR90
 rs2740439 866 G/C C = 0.08 (HAPMAPCEU
 rs4639 969 A/G G = 0.401 (HAPMAPCEU
 rs1043180 1,039 C/T T = 0.111 (HAPMAPCEU
NEIL3 rs1055677b 173 C/G G = 0.068 (HAPMAPCEU
 rs34652756b 132 T/C C = 0.067 (HAPMAPCEU
 rs1055678 342 C/T T = 0.068 (HAPMAPCEU
LIG3 rs1052536 50 C/T T = 0.500 (HAPMAPCEU
Gene IDdbSNP IDPositionVariationMAF (population studied)
SMUG1 rs2233921 80 G/T T = 0.482 (HAPMAPCEU
 rs971 422 G/A A = 0.320 (HAPMAPCEU
MBD4 rs2307285 383 A/G G = 0.197 (PDR90
 rs2307294 447 G/C C = 0.100 (NIH-PDR90
NEIL2 rs1534862 21 C/T T = 0.219 (HAPMAPCEU
 rs804292a 133 A/G G = 0.292 (HAPMAPCEU
 rs6997097 190 T/C C = 0.055 (PDR90) 
 rs8191670 356 T/C C = 0.25 (PDR90
 rs2740439 866 G/C C = 0.08 (HAPMAPCEU
 rs4639 969 A/G G = 0.401 (HAPMAPCEU
 rs1043180 1,039 C/T T = 0.111 (HAPMAPCEU
NEIL3 rs1055677b 173 C/G G = 0.068 (HAPMAPCEU
 rs34652756b 132 T/C C = 0.067 (HAPMAPCEU
 rs1055678 342 C/T T = 0.068 (HAPMAPCEU
LIG3 rs1052536 50 C/T T = 0.500 (HAPMAPCEU

aIn linkage disequilibrium (LD) with rs1534862 (Haploview 4.2). Not genotyped.

bIn LD with rs1055678 (Haploview 4.2). Not genotyped.

Case–control study

The total number of patients with colorectal cancer included in the study was 1,095, among which approximately two third were diagnosed with colon and the rest with rectal cancer (3 of them were excluded because of missing tumor site information). Out of the 1,469 controls, 688 were CFCC and 781 were HBDV. Compared with subjects of both control groups, colorectal cancer cases were more likely to be older and have a slightly higher BMI. Compared with the HBDV group, they were more likely to have a positive family history of colorectal cancer and lower formal education (Table 2).

Table 2.

Characteristics of the study population

Colorectal cancer casesControl group I (CFCC)Control group II (HBDV)All controlsOR (95% CI)P
All subjects 1,098 688 781 1,469   
Diagnosis Colon cancer 725 — — —   
 Rectal cancer 370 — — —   
Age, y 47≤ 94 164 427 591 Ref  
 48–55 208 145 277 422 3.10 (2.36–4.09) <0.01 
 56–65 370 209 77 286 8.13 (6.25–10.66) <0.01 
 >65 423 170 170 15.37 (11.66–20.44) <0.01 
Sex Females 435 317 343 660 Ref  
 Males 660 371 438 809 1.23 (1.05–1.45) 0.01 
BMI 23.7≤ 184 154 215 369 Ref  
 23.7–26.2 192 147 213 360 1.07 (0.83–1.37) 0.61 
 26.3–28.9 226 139 184 323 1.40 (1.10–1.79) 0.01 
 >28.9 222 172 157 329 1.35 (1.06–1.73) 0.02 
Smoking No 536 364 451 815 Ref  
 Yesa 501 254 327 581 1.31 (1.12–1.54) <0.01 
Family history of colorectal cancer No 726 486 718 1,204 Ref  
 Yes 144 90 52 142 1.68 (1.31–2.16) <0.01 
Living area City 511 338 614 952 Ref  
 Suburbs 128 118 53 171 1.39 (1.08–1.79) 0.01 
 Countryside 242 157 112 270 1.67 (1.36–2.05) <0.01 
Education Basic 266 171 53 224 Ref  
 Medium 469 327 492 820 0.48 (0.39–0.59) <0.01 
 High 138 114 231 345 0.34 (0.26–0.44) <0.01 
Colorectal cancer casesControl group I (CFCC)Control group II (HBDV)All controlsOR (95% CI)P
All subjects 1,098 688 781 1,469   
Diagnosis Colon cancer 725 — — —   
 Rectal cancer 370 — — —   
Age, y 47≤ 94 164 427 591 Ref  
 48–55 208 145 277 422 3.10 (2.36–4.09) <0.01 
 56–65 370 209 77 286 8.13 (6.25–10.66) <0.01 
 >65 423 170 170 15.37 (11.66–20.44) <0.01 
Sex Females 435 317 343 660 Ref  
 Males 660 371 438 809 1.23 (1.05–1.45) 0.01 
BMI 23.7≤ 184 154 215 369 Ref  
 23.7–26.2 192 147 213 360 1.07 (0.83–1.37) 0.61 
 26.3–28.9 226 139 184 323 1.40 (1.10–1.79) 0.01 
 >28.9 222 172 157 329 1.35 (1.06–1.73) 0.02 
Smoking No 536 364 451 815 Ref  
 Yesa 501 254 327 581 1.31 (1.12–1.54) <0.01 
Family history of colorectal cancer No 726 486 718 1,204 Ref  
 Yes 144 90 52 142 1.68 (1.31–2.16) <0.01 
Living area City 511 338 614 952 Ref  
 Suburbs 128 118 53 171 1.39 (1.08–1.79) 0.01 
 Countryside 242 157 112 270 1.67 (1.36–2.05) <0.01 
Education Basic 266 171 53 224 Ref  
 Medium 469 327 492 820 0.48 (0.39–0.59) <0.01 
 High 138 114 231 345 0.34 (0.26–0.44) <0.01 

NOTE: Significant results are given in bold.

Numbers may not add up to 100% of available subjects because of missing data.

aEx-smokers are included into this group.

The genotype and allele frequencies for the investigated SNPs and their association with colorectal cancer risk after adjustment for all listed confounders are reported in Supplementary Table S1, together with the calculation of the Hardy–Weinberg equilibrium in controls. For all analyzed SNPs, the genotype calls had 99% concordance with duplicate quality controls. Only one SNP, rs2307294, was not in Hardy–Weinberg equilibrium in the control group (χ2 = 9.01 and P = 0.01), therefore it was excluded from further analyses.

None of the polymorphisms genotyped in this study were significantly associated with colorectal cancer risk. To overcome issues due to the age effect (mean age in cases = 61.7 years; mean age in controls = 50.6 years; t test = 24.4; P ≤ 0.0001), cases and controls were matched by age quartiles through bootstrap sampling (10 repetitions). For each subset, the association analyses were repeated and the results averaged. With this approach, we could ascertain that no changes were observed in the 10 different resamplings. In addition, the analysis was repeated in a subgroup of 800 cases and 800 controls matched for sex and age (after matching, mean age in cases = 59.0 years, mean age in controls = 57.7 years) and similar results were observed as in the whole group of study; none of the SNPs were associated with cancer risk (data not shown).

Survival analysis

Training set.

The average (median) OS and EFS for patients was 87.8 (72.0) and 83.0 (68.1) months, respectively. In the preliminary univariate assessment of covariates, gender, age, smoking habit, TNM status, and chemotherapy treatment were associated with OS (Table 3). Advanced age, male gender, and current smoking status were related to a shortened OS. Men in particular also showed a higher risk of relapse or metastasis (OS: HR, 1.54; 95% CI, 1.22–1.94; P < 0.01; EFS: HR, 1.43; 95% CI, 1.14–1.80; P < 0.01). Four established prognostic factors (TNM status and chemotherapy treatment) were associated with patient survival. Moreover, TNM status was also associated with an increased risk of recurrence.

Table 3.

Clinical and anamnestic characteristics significantly affecting OS and EFS of the colorectal cancer patients with complete follow-up (Cox regression) in the Training set and the Replication set

Training setReplication setPooled
OSEFSOSEFSOSEFS
NaHR (95% CI)PHR (95% CI)PNaHR (95% CI)PHR (95% CI)PNaHR (95% CI)PHR (95% CI)P
Sex Females 357 Ref.  Ref.  78 Ref.  Ref.  435 Ref.  Ref.  
 Males 509 1.54 (1.22–1.94) <0.01 1.43 (1.14–1.80) <0.01 154 2.18 (1.05–4.52) 0.04 1.23 (0.65–2.32) 0.51 663 1.54 (1.24–1.92) <0.01 1.36 (1.09–1.68) <0.01 
Age, y 55≤ 235 Ref.  Ref.  61 Ref.  Ref.  296 Ref.  Ref.  
 56–62 203 1.40 (1.00–1.94) 0.05 1.32 (0.97–1.78) 0.07 59 1.12 (0.47–2.70) 0.80 2.01 (0.88–4.59) 0.10 262 1.42 (1.04–1.93) 0.03 1.41 (1.06–1.87) 0.02 
 63–69 221 1.33 (0.96–1.84) 0.08 1.07 (0.79–1.46) 0.66 54 1.40 (0.60–3.31) 0.44 2.15 (0.93–4.98) 0.07 537 1.39 (1.03–1.86) 0.03 1.20 (0.91–1.59) 0.20 
 >69 207 1.87 (1.37–2.57) <0.01 1.05 (0.76–1.45) 0.77 58 2.43 (1.08–5.50) 0.03 1.01 (0.38–2.73) 0.98 265 2.02 (1.50–2.71) <0.01 1.05 (0.77–1.43) 0.75 
Smoking habit No 428 Ref.  Ref.  180 Ref.  Ref.  537 Ref.  Ref.  
 Yesb 116 1.54 (1.13–2.09) 0.01 1.26 (0.91–1.75) 0.16 42 2.60 (1.33–5.08) 0.005 0.97 (0.53–1.78) 0.92 157 1.25 (1.02–1.54) 0.04 1.14 (0.93–1.41) 0.21 
pT 32 Ref  Ref.  19 Ref.  Ref.  51 Ref.  Ref.  
 130 2.66 (0.82–8.67) 0.10 2.18 (0.77–6.16) 0.14 39 2.16 (0.25–18.53) 0.48 0.98 (0.09–10.80) 0.99 169 2.72 (0.97–7.62) 0.06 2.18 (0.86–5.57) 0.10 
 420 5.84 (1.87–18.28) <0.01 5.26 (1.95–14.16) <0.01 122 4.76 (0.65–35.90) 0.13 6.06 (0.83–44.44) 0.08 542 6.05 (2.25–16.26) <0.001 5.74 (2.36–13.92) <0.001 
 90 12.30 (3.85–39.23) <0.01 8.09 (2.93–22.33) <0.01 46 4.43 (0.57–34.58) 0.16 4.94 (0.64–38.31) 0.13 136 9.45 (3.45–25.89) <0.0001 7.10 (2.86–17.64) <0.001 
pN 359 Ref  Ref.  149 Ref.  Ref.  508 Ref.  Ref.  
 213 2.04 (1.56–2.67) <0.01 1.76 (1.35–2.30) <0.01 48 2.04 (1.03–4.07) 0.04 1.71 (0.81–3.61) 0.16 261 2.14 (1.67–2.74) <0.01 1.89 (1.47–2.42) <0.01 
 51 4.69 (3.17–6.94) <0.01 3.51 (2.42–5.07) <0.01 17 1.47 (0.44–4.95) 0.53 3.91 (1.73–8.82) <0.01 68 3.43 (2.38–4.95) <0.01 3.50 (2.51–4.90) <0.01 
pM 545 Ref  Ref.  189 Ref.  Ref.  734 Ref.  Ref.  
 150 4.67 (3.68–5.93) <0.01 4.11 (3.28–5.15) <0.01 29 6.07 (3.06–12.04) <0.01 6.41 (3.32–12.39) <0.01 179 4.82 (3.86–6.04) <0.01 4.59 (3.71–5.68) <0.01 
5-FU–based chemotherapy Yes 319 Ref    102 Ref  Ref  421 Ref  Ref  
 No 324 1.66 (1.29–2.12) <0.01 1.14 (0.90–1.45) 0.27 121 0.71 (0.39–1.30) 0.27 0.13 (0.06–0.29) <0.01 445 1.43 (1.14–1.80) <0.01 0.85 (0.68–1.06) 0.14 
Training setReplication setPooled
OSEFSOSEFSOSEFS
NaHR (95% CI)PHR (95% CI)PNaHR (95% CI)PHR (95% CI)PNaHR (95% CI)PHR (95% CI)P
Sex Females 357 Ref.  Ref.  78 Ref.  Ref.  435 Ref.  Ref.  
 Males 509 1.54 (1.22–1.94) <0.01 1.43 (1.14–1.80) <0.01 154 2.18 (1.05–4.52) 0.04 1.23 (0.65–2.32) 0.51 663 1.54 (1.24–1.92) <0.01 1.36 (1.09–1.68) <0.01 
Age, y 55≤ 235 Ref.  Ref.  61 Ref.  Ref.  296 Ref.  Ref.  
 56–62 203 1.40 (1.00–1.94) 0.05 1.32 (0.97–1.78) 0.07 59 1.12 (0.47–2.70) 0.80 2.01 (0.88–4.59) 0.10 262 1.42 (1.04–1.93) 0.03 1.41 (1.06–1.87) 0.02 
 63–69 221 1.33 (0.96–1.84) 0.08 1.07 (0.79–1.46) 0.66 54 1.40 (0.60–3.31) 0.44 2.15 (0.93–4.98) 0.07 537 1.39 (1.03–1.86) 0.03 1.20 (0.91–1.59) 0.20 
 >69 207 1.87 (1.37–2.57) <0.01 1.05 (0.76–1.45) 0.77 58 2.43 (1.08–5.50) 0.03 1.01 (0.38–2.73) 0.98 265 2.02 (1.50–2.71) <0.01 1.05 (0.77–1.43) 0.75 
Smoking habit No 428 Ref.  Ref.  180 Ref.  Ref.  537 Ref.  Ref.  
 Yesb 116 1.54 (1.13–2.09) 0.01 1.26 (0.91–1.75) 0.16 42 2.60 (1.33–5.08) 0.005 0.97 (0.53–1.78) 0.92 157 1.25 (1.02–1.54) 0.04 1.14 (0.93–1.41) 0.21 
pT 32 Ref  Ref.  19 Ref.  Ref.  51 Ref.  Ref.  
 130 2.66 (0.82–8.67) 0.10 2.18 (0.77–6.16) 0.14 39 2.16 (0.25–18.53) 0.48 0.98 (0.09–10.80) 0.99 169 2.72 (0.97–7.62) 0.06 2.18 (0.86–5.57) 0.10 
 420 5.84 (1.87–18.28) <0.01 5.26 (1.95–14.16) <0.01 122 4.76 (0.65–35.90) 0.13 6.06 (0.83–44.44) 0.08 542 6.05 (2.25–16.26) <0.001 5.74 (2.36–13.92) <0.001 
 90 12.30 (3.85–39.23) <0.01 8.09 (2.93–22.33) <0.01 46 4.43 (0.57–34.58) 0.16 4.94 (0.64–38.31) 0.13 136 9.45 (3.45–25.89) <0.0001 7.10 (2.86–17.64) <0.001 
pN 359 Ref  Ref.  149 Ref.  Ref.  508 Ref.  Ref.  
 213 2.04 (1.56–2.67) <0.01 1.76 (1.35–2.30) <0.01 48 2.04 (1.03–4.07) 0.04 1.71 (0.81–3.61) 0.16 261 2.14 (1.67–2.74) <0.01 1.89 (1.47–2.42) <0.01 
 51 4.69 (3.17–6.94) <0.01 3.51 (2.42–5.07) <0.01 17 1.47 (0.44–4.95) 0.53 3.91 (1.73–8.82) <0.01 68 3.43 (2.38–4.95) <0.01 3.50 (2.51–4.90) <0.01 
pM 545 Ref  Ref.  189 Ref.  Ref.  734 Ref.  Ref.  
 150 4.67 (3.68–5.93) <0.01 4.11 (3.28–5.15) <0.01 29 6.07 (3.06–12.04) <0.01 6.41 (3.32–12.39) <0.01 179 4.82 (3.86–6.04) <0.01 4.59 (3.71–5.68) <0.01 
5-FU–based chemotherapy Yes 319 Ref    102 Ref  Ref  421 Ref  Ref  
 No 324 1.66 (1.29–2.12) <0.01 1.14 (0.90–1.45) 0.27 121 0.71 (0.39–1.30) 0.27 0.13 (0.06–0.29) <0.01 445 1.43 (1.14–1.80) <0.01 0.85 (0.68–1.06) 0.14 

NOTE: Significant results are given in bold.

aNumbers may not add up to 100% of available subjects because of missing information.

bEx-smokers excluded.

The data analysis showed that rs2233921 in SMUG1 gene was significantly associated with OS (HR, 0.62; 95% CI, 0.39–1.00; P = 0.05) but not with EFS (Table 4; Supplementary Table S2). In the dominant model, the association of the T allele with a better survival was even more pronounced (HR, 0.54; 95% CI, 0.36–0.89; P = 0.003). The univariate Kaplan–Meier plots for OS also reflected these results (log-rank test for the dominant model P = 0.008; Fig. 2A). In contrast, carriers of the GA and TC genotypes of SMUG1 rs971 and NEIL2 rs6997097 polymorphisms were associated with a shorter survival (HR, 1.38; 95% CI, 1.02–1.86; P = 0.04; HR, 1.77; 95% CI, 1.23–2.53; P = 0.002, respectively). However, in the dominant model, only C allele carriers of rs6997097 in NEIL2 showed a shorter survival (HR, 1.73; 95% CI, 1.21–2.47; P = 0.003; Table 4; Supplementary Table S2).

Figure 2.

Kaplan–Meier OS curves of patients with colorectal cancer stratified for SMUG1 rs2233921 genotypes. A and B, dominant model and stratification according to 5-FU–based chemotherapy for the Training set. C and D, for the pooled analysis (Training set + Replication set).

Figure 2.

Kaplan–Meier OS curves of patients with colorectal cancer stratified for SMUG1 rs2233921 genotypes. A and B, dominant model and stratification according to 5-FU–based chemotherapy for the Training set. C and D, for the pooled analysis (Training set + Replication set).

Close modal
Table 4.

SNPs associated with OS of patients in each set and in the pooled population (Cox regression for adjusted estimates)

Training setReplication setPooled
Gene dbSNP IDGenotypeNaEventsExpectedHR (95% CI)bPNaEventsExpectedHR (95% CI)bPNaEventsExpectedHR (95% CI)bP
SMUG1 rs2233921 GG 205 94 85.3 Ref  76 15 15.3 Ref  281 109 104.0 Ref  
 GT 392 181 171.9 1.24 (0.90–1.71) 0.19 106 20 20.1 0.99 (0.41–2.41) 0.98 498 201 188.0 1.19 (0.89–1.61) 0.25 
 TT 121 38 55.8 0.62 (0.39–1.00) 0.05 47 8.6 0.35 (0.08–1.49) 0.15 168 47 65.0 0.61 (0.39–0.95) 0.03 
 GG+GT 597 275 257.2 Ref  182 35 35.42 Ref  779 310 292.0 Ref  
 TT 121 38 55.8 0.54 (0.36–0.81) 0.003 47 8.58 0.35 (0.09–1.33) 0.12 168 47 65.0 0.55 (0.37–0.80) 0.002 
SMUG1 rs971 GG 262 102 117.2 Ref  87 17 15.6 Ref  349 119 131.6 Ref  
 GA 368 173 153.5 1.38 (1.02–1.86) 0.04 111 21 23.4 1.82 (0.71–4.68) 0.21 479 194 178.0 1.41 (1.06–1.87) 0.02 
 AA 92 35 39.3 0.92 (0.57–1.50) 0.75 31 6.0 2.13 (0.63–7.27) 0.23 123 42 45.4 1.12 (0.72–1.74) 0.63 
 GG 262 102 117.2 Ref  87 17 15.6 Ref  349 119 132.0 Ref  
 GA+AA 460 208 192.8 1.27 (0.95–1.70) 0.10 142 28 29.4 1.89 (0.76–4.70) 0.17 602 236 223.0 1.35 (1.03–1.77) 0.03 
NEIL2 rs6997097 TT 618 258 266.8 Ref  191 37 36.8 Ref  809 295 304.1 Ref  
 TC 106 54 43.3 1.77 (1.23–2.53) <0.01 34 6.25 0.68 (0.19–2.40) 0.54 140 60 49.3 1.68 (1.19–2.38) 0.003 
 CC 2.9 0.87 (0.12–6.37) 0.89 2.6 1.11 (0.15–8.06) 0.92 
 TT 618 258 266.8 Ref  191 37 36.8 Ref  809 295 304.1 Ref  
 TC+CC 111 55 46.2 1.73 (1.21–2.47) 0.003 34 6.25 0.68 (0.19–2.40) 0.54 145 61 51.9 1.66 (1.18–2.34) 0.003 
Training setReplication setPooled
Gene dbSNP IDGenotypeNaEventsExpectedHR (95% CI)bPNaEventsExpectedHR (95% CI)bPNaEventsExpectedHR (95% CI)bP
SMUG1 rs2233921 GG 205 94 85.3 Ref  76 15 15.3 Ref  281 109 104.0 Ref  
 GT 392 181 171.9 1.24 (0.90–1.71) 0.19 106 20 20.1 0.99 (0.41–2.41) 0.98 498 201 188.0 1.19 (0.89–1.61) 0.25 
 TT 121 38 55.8 0.62 (0.39–1.00) 0.05 47 8.6 0.35 (0.08–1.49) 0.15 168 47 65.0 0.61 (0.39–0.95) 0.03 
 GG+GT 597 275 257.2 Ref  182 35 35.42 Ref  779 310 292.0 Ref  
 TT 121 38 55.8 0.54 (0.36–0.81) 0.003 47 8.58 0.35 (0.09–1.33) 0.12 168 47 65.0 0.55 (0.37–0.80) 0.002 
SMUG1 rs971 GG 262 102 117.2 Ref  87 17 15.6 Ref  349 119 131.6 Ref  
 GA 368 173 153.5 1.38 (1.02–1.86) 0.04 111 21 23.4 1.82 (0.71–4.68) 0.21 479 194 178.0 1.41 (1.06–1.87) 0.02 
 AA 92 35 39.3 0.92 (0.57–1.50) 0.75 31 6.0 2.13 (0.63–7.27) 0.23 123 42 45.4 1.12 (0.72–1.74) 0.63 
 GG 262 102 117.2 Ref  87 17 15.6 Ref  349 119 132.0 Ref  
 GA+AA 460 208 192.8 1.27 (0.95–1.70) 0.10 142 28 29.4 1.89 (0.76–4.70) 0.17 602 236 223.0 1.35 (1.03–1.77) 0.03 
NEIL2 rs6997097 TT 618 258 266.8 Ref  191 37 36.8 Ref  809 295 304.1 Ref  
 TC 106 54 43.3 1.77 (1.23–2.53) <0.01 34 6.25 0.68 (0.19–2.40) 0.54 140 60 49.3 1.68 (1.19–2.38) 0.003 
 CC 2.9 0.87 (0.12–6.37) 0.89 2.6 1.11 (0.15–8.06) 0.92 
 TT 618 258 266.8 Ref  191 37 36.8 Ref  809 295 304.1 Ref  
 TC+CC 111 55 46.2 1.73 (1.21–2.47) 0.003 34 6.25 0.68 (0.19–2.40) 0.54 145 61 51.9 1.66 (1.18–2.34) 0.003 

NOTE: Significant results are given in bold.

aNumbers may not add up to 100% of available subjects because of missing information.

bAdjusted for sex, age, TNM, and chemotherapy.

After stratification of patients according to treatment (treated with 5-FU or not), the previously observed associations for OS analysis were confirmed only for SMUG1 rs2233921 in both groups (data not shown). The univariate Kaplan–Meier plot including the stratification for 5-FU–based chemotherapy provided a strong association with OS (log-rank test P = 5.6 × 10−5; Fig. 2B). In particular, a better survival was observed for patients carrying the TT genotype and undergoing 5-FU–based chemotherapy. On the other hand, SMUG1 rs971 and NEIL2 rs6997097 associated with a shorter survival only in those subjects who did not undergo 5-FU chemotherapy.

For EFS analysis, a SNP in NEIL2 (rs1534862) was associated with a lower risk of relapse or metastasis in both codominant and dominant models (for CT genotype HR, 0.73; 95% CI, 0.54–0.98; P = 0.04; for T allele HR, 0.73; 95% CI, 0.55–0.98; P = 0.03). Regarding rs1055678 in NEIL3, the TT genotype was not only associated with an increased risk of recurrence in the codominant model (HR, 7.98; 95% CI, 1.04–61.34; P = 0.05), but also in the dominant model (HR, 7.76; 95% CI, 1.01–59.52; P = 0.05; Supplementary Tables S2 and S3). Given the low numbers of cases with variant genotypes, the latter result should be considered with caution. Considering EFS, no association was observed after stratification according to chemotherapy treatment. The only exception was for rs1055678 in NEIL3; however, again, in this case only a reduced number of observations were available (data not shown).

Replication set and pooled analysis.

The average (median) OS and EFS for the Replication set were 82.5 (69.3) and 74.2 (58.7) months, respectively. In the univariate assessment of covariates, gender, age, smoking habit, TNM status, and chemotherapy treatment were found to be associated with either OS or EFS, similarly as in the Training set (see Table 3). In this second group, although not reaching statistical significance, all the SNPs previously found associated with either OS and EFS (and also after stratification for 5-FU–based chemotherapy) showed a similar trend of association with the exception of NEIL2 rs6997097 and rs1534862 (Table 4; Supplementary Tables S2 and S3). The pooled analysis of the two sets confirmed the previously observed associations with OS for SMUG1 rs2233921 (for TT genotype HR, 0.61; 95% CI, 0.39–0.95; P = 0.03; for T allele HR, 0.55; 95% CI, 0.37–0.80; P = 0.002) and for SMUG1 rs971 (for GA genotype HR, 1.41; 95% CI, 1.06–1.87; P = 0.02; for A allele HR, 1.35; 95% CI, 1.03–1.77; P = 0.03), as well as for NEIL2 rs6997097 (for TC genotype HR, 1.68; 95% CI, 1.19–2.38; P = 0.003; for C allele HR, 1.66; 95% CI, 1.18–2.34; P = 0.003). For rs2233921, these results reflected also the univariate Kaplan–Meier plots for OS (log-rank test for the dominant model P = 0.01; Fig. 2C), even after stratification for 5-FU–based chemotherapy (log-rank test P = 0.003; Fig. 2D). For EFS analysis, none of the previously associated SNPs were confirmed in the pooled analysis.

In vitro assay

For SMUG1 rs2233921, an in vitro assay was used to investigate whether its alleles were associated with differential gene expression. The results in the HCT-116 cell line derived from colorectal cancer are reported in Fig. 3. The G-to-T point mutation of rs2233921 (SMUG1) resulted in a reduction of the expression of luciferase in all the experiments performed. Combining results from all the experiments, when considering the luciferase expression with the G allele as the reference (100%), the average luciferase expression with the T allele of SMUG1 was reduced to 63.1% and this difference was statistically significant (P < 10−4).

Figure 3.

In vitro assay to test the luciferase activity according to SMUG1 rs2233921 different alleles (G or T). Six independent replicates in three independent experiments are summarized as mean values of relative intensity of luciferase/Renilla luminescence ratio. When considering the luciferase expression with the G allele as the reference (100%), the luciferase expression with the T allele of SMUG1 was 71.6%, 52.6%, and 65.2% respectively in the three experiments.

Figure 3.

In vitro assay to test the luciferase activity according to SMUG1 rs2233921 different alleles (G or T). Six independent replicates in three independent experiments are summarized as mean values of relative intensity of luciferase/Renilla luminescence ratio. When considering the luciferase expression with the G allele as the reference (100%), the luciferase expression with the T allele of SMUG1 was 71.6%, 52.6%, and 65.2% respectively in the three experiments.

Close modal

miRNAs are increasingly recognized as central players in diverse biologic processes, including DNA repair and DNA damage response (25). An increasing body of evidence has indicated miRNAs as cancer diagnostic, prognostic, and predictive clinical biomarkers (26, 27). The presence of SNPs within the 3′-UTRs of target DNA repair genes that alter the binding with specific miRNAs and modulate gene expression, may ultimately affect cancer susceptibility (18). Less is known about the effect of variations in such genes on patient survival and on the effect of therapy via miRNA modulation (28).

In the present study, we selected 12 SNPs within the 3′-UTRs of five genes from the BER pathway, which potentially alter miRNA::mRNA–binding affinity, and we investigated their role in colorectal cancer susceptibility in a case–control study. Subsequently, we tested those SNPs as prognostic factors in patients with colorectal cancer with detailed follow-up information.

The novel finding was that a SNP in the BER gene SMUG1 (rs2233921) was associated with patient survival in a set of 718 patients with colorectal cancer. In particular, carriers of the TT genotype showed a significantly increased OS compared with those with the GG genotype, and the results were even more pronounced in relation to the 5-FU–based chemotherapy. When another set of 229 patients was added, the above results were replicated, though reaching a more robust statistical significance only in the pooled analysis. The covariate analysis indicated that the two populations (Training and Replication sets), although of considerable different size, displayed similar clinical and anamnestic characteristics to justify a pooled analysis. However, we cannot rule out certain differences, for example due to a different time interval of follow-up and the collection of clinical information. Importantly, the observed associations remained significant even if we would apply the stringent Bonferroni's correction for multiple comparisons (Pthreshold = 0.0045, considering α = 0.05 and 11 independent tests).

The association of SMUG1 rs2233921 with increased OS is of particular interest, especially in relation to the widely used antimetabolite 5-FU–based chemotherapy. Variations in the DNA repair capacity may play a key role in the response to 5-FU exposure (8, 29). In fact, BER glycosylases are the principal enzymes involved in the cellular response to 5-FU–induced damage (8, 30, 31), with SMUG1 recognized as a main player (32). The overall picture suggests that SMUG1 excision activity is protective against toxicity caused by 5-FU (8). Other findings also imply that SMUG1 upregulation might mediate the development of tumor resistance to 5-FU treatment (33). Because no genetic variants have been found responsible for a differential functionality of SMUG1 enzyme (31), variations in posttranslational regulation, such as the one exerted by miRNAs, could be responsible for a different response to 5-FU treatment and survival. An in vitro assay, carried out to test functional differences between the different alleles of SMUG1 rs2233921, confirmed that the T allele was associated with a reduced expression of the reporter gene when compared with the common G allele. This finding supports the hypothesis that the presence of an “alternative” 3′-UTR placed in a “normal” cellular context (and exploiting the “normal” cellular machineries) may modulate the levels of expression of SMUG1. The assay was performed on colorectal cell lines, and it is conceivable that these cells express a miRNA set typical of colorectal cells. At present, we cannot state whether miRNAs, mRNA stability, or other mechanisms are involved in the observed genotype-dependent differential expression of SMUG1. However, this represents an interesting issue for further investigation. In fact, our outcomes support the relevant role of regulation of BER genes in response to DNA-damaging chemotherapy in the treatment of colorectal malignancy, which may be affected by interindividual variability.

Other interesting associations were found for SMUG1 rs971 or NEIL2 rs6997097 polymorphisms with a shorter survival in both the Training set and the pooled analysis. For an additional two SNPs, one in NEIL2 (rs1534862) and another in NEIL3 (rs1055678), the associations with risk of recurrence were found only in the Training set. Further investigations on a larger scale are needed to confirm the role of these SNPs. The modulation of NEIL2 and NEIL3 could also be of interest as both genes belong to a family of glycosylases specialized in the recognition of oxidized pyrimidines (31). NEIL2, in particular, recognizes lesions on single-stranded DNA and was also associated with the transcription process (34). Missense polymorphic variants in NEIL2 were associated with colorectal cancer risk, whereas variants in NEIL3 were associated with colorectal adenoma (reviewed in ref. 31). However, to the best of our knowledge, none of their polymorphic variants were evaluated for their functional effects or for their possible modulation in relationship with 5-FU therapy.

Unlike our previous studies (18, 19, 35), we did not observe any association of these 3′-UTR–localized SNPs with risk of colorectal cancer. With a study group of 1,095 cases and 1,469 controls, we had more than 80% power to detect an association for a recessive/dominant model with OR > 1.30 at α = 0.05 (with MAF = 0.30; power calculation by NAME software). We are aware of certain limitations in this case–control study, such as the different age distribution among the cases and controls. We attempted to control age effect by two different approaches: (i) matching cases and controls by age quartiles through bootstrap sampling (10 times), and (ii) repeating the analysis on a subgroup of 800 cases and 800 controls matched for sex and age. Nevertheless, none of the SNPs resulted in association with cancer risk.

The rationale for selecting the genes/polymorphisms in this study has been provided by MicroSNiPer, a Web-based tool that predicts the SNPs on putative miRNA targets (24). We did not rank and test SNPs according to allele-specific Gibbs-free binding energy, as did previous studies (18, 20), but we have analyzed systematically all polymorphisms retrieved in the 3′-UTR of BER genes with an adequate MAF in the population. An additional software available online (Mirnsnpscore; http://www.bigr.medisin.ntnu.no/mirsnpscore/), based on in silico predictions of SNP effects on miRNA-based gene regulation, was also queried to verify the regulatory effect of the selected SNPs. Mirnsnpscore sorts SNPs and calculates a DeltaS score that summarizes the predicted effect of a SNP on miRNA regulation. The above described five SNPs associated with clinical outcomes were again reported as having an effect on miRNA binding. In particular, rs2233921 had the highest DeltaS score among the SNPs predicted in the 3′-UTR of SMUG1 (DeltaS = 0.95 and 0.94 for miR770-5p and miR665, respectively; Supplementary Table S4).

Several miRNAs predicted to bind to the above-mentioned BER genes have been reported in the colorectal miRNAome (36). Interestingly, new findings suggest that miRNAs also affect drug resistance in colorectal cancer (37). For instance, Mir455-3p, predicted to bind in SMUG1 3′-UTR where rs2233921 resides, may target molecular pathways involved in cell survival after 5-FU or cisplatin chemotherapy (38). miRNA pharmacogenomics is a novel and promising field that investigates the role of miRNAs and polymorphisms affecting miRNAs and their target functions for the prediction of drug response or the improvement of drug efficiency (39). Recently, polymorphisms in miRNA-related genes were described as predictors of clinical outcomes in patients with stage III colorectal cancer treated with 5-FU–based chemotherapy (40). A proposed hypothesis was that the generation of a target gene expression control modification due to the SNPs in the miRNA precursor and in the mature molecules could have an effect on the response to 5-FU. We may also extend this hypothesis to the SNPs in miRNA-binding sites of our target BER genes. Accordingly, SNPs residing in the target sites of AP-2α gene were found to alter the miRNA–mRNA interaction resulting in differential target gene expression, which associated with cisplatin chemosensitivity (41). A different posttranslational regulation of an enzyme due to different miRNA sets binding according to specific alleles has been reported and elucidated in ref. (42). Thus, characterization of the polymorphisms in miRNA-related genes/target sites and the relative functional impact may provide a solid basis for miRNA-based therapeutic approaches.

In conclusion, we have identified SNPs potentially affecting miRNA binding to the 3′-UTR of BER genes among which SMUG1 rs2233921 was strongly associated with clinical outcome of patients with colorectal cancer treated with 5-FU–based chemotherapy. By in vitro assay, we have demonstrated that this polymorphism shows a differential expression of the reporter gene according to its different alleles, supporting a possible regulatory role on the expression of SMUG1. Further studies are needed to replicate these SNPs as predictive biomarkers in independent and larger populations, to more specifically characterize functional relevance of the identified genetic variants and to find the biologic mechanisms underlying the observed associations.

No potential conflicts of interest were disclosed.

Conception and design: B. Pardini, S. Landi, P. Vodicka, A. Naccarati

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Pardini, E. Barone, J. Novotny, M. Levy, S. Garritano, L. Vodickova, T. Buchler, P. Vodicka, A. Naccarati

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Pardini, F. Rosa, C. Di Gaetano, T. Buchler, S. Landi, A. Naccarati

Writing, review, and/or revision of the manuscript: B. Pardini, F. Rosa, E. Barone, J. Slyskova, J. Novotny, S. Landi, P. Vodicka, A. Naccarati

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Pardini, F. Rosa, J. Slyskova, M. Levy, S. Garritano, L. Vodickova, A. Naccarati

Study supervision: B. Pardini, F. Gemignani, P. Vodicka, A. Naccarati

Authors would like to thank T. O'Hearn II, B. O'Brien, and E. Van Emburgh for their excellent technical support.

This work was supported by the Grant Agency of the Czech Republic (grants CZ:GACR:GAP304/10/1286, CZ:GACR:GAP304/12/1585).

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.

1.
Jemal
A
,
Siegel
R
,
Xu
J
,
Ward
E
. 
Cancer statistics, 2010
.
CA Cancer J Clin
2010
;
60
:
277
300
.
2.
Theodoratou
E
,
Montazeri
Z
,
Hawken
S
,
Allum
GC
,
Gong
J
,
Tait
V
, et al
Systematic meta-analyses and field synopsis of genetic association studies in colorectal cancer
.
J Natl Cancer Inst
2012
;
104
:
1433
57
.
3.
Berwick
M
,
Vineis
P
. 
Markers of DNA repair and susceptibility to cancer in humans: an epidemiologic review
.
J Natl Cancer Inst
2000
;
92
:
874
97
.
4.
Wilson
DM
 III
,
Kim
D
,
Berquist
BR
,
Sigurdson
AJ
. 
Variation in base excision repair capacity
.
Mutat Res
2011
;
711
:
100
12
.
5.
Wilson
DM
 III
,
Bohr
VA
. 
The mechanics of base excision repair, and its relationship to aging and disease
.
DNA Repair
2007
;
6
:
544
59
.
6.
Pardini
B
,
Naccarati
A
,
Novotny
J
,
Smerhovsky
Z
,
Vodickova
L
,
Polakova
V
, et al
DNA repair genetic polymorphisms and risk of colorectal cancer in the Czech Republic
.
Mutat Res
2008
;
638
:
146
53
.
7.
Vodicka
P
,
Stetina
R
,
Polakova
V
,
Tulupova
E
,
Naccarati
A
,
Vodickova
L
, et al
Association of DNA repair polymorphisms with DNA repair functional outcomes in healthy human subjects
.
Carcinogenesis
2007
;
28
:
657
64
.
8.
Wyatt
MD
,
Wilson
DM
 III
. 
Participation of DNA repair in the response to 5-fluorouracil
.
Cell Mol Life Sci
2009
;
66
:
788
99
.
9.
Longley
DB
,
Harkin
DP
,
Johnston
PG
. 
5-Fluorouracil: mechanisms of action and clinical strategies
.
Nat Rev Cancer
2003
;
3
:
330
8
.
10.
Bartel
DP
. 
MicroRNAs: target recognition and regulatory functions
.
Cell
2009
;
136
:
215
33
.
11.
Chen
CZ
,
Li
L
,
Lodish
HF
,
Bartel
DP
. 
MicroRNAs modulate hematopoietic lineage differentiation
.
Science
2004
;
303
:
83
6
.
12.
Xu
P
,
Guo
M
,
Hay
BA
. 
MicroRNAs and the regulation of cell death
.
Trends Genet
2004
;
20
:
617
24
.
13.
Slattery
ML
,
Wolff
E
,
Hoffman
MD
,
Pellatt
DF
,
Milash
B
,
Wolff
RK
. 
MicroRNAs and colon and rectal cancer: differential expression by tumor location and subtype
.
Genes Chromosomes Cancer
2011
;
50
:
196
206
.
14.
Duan
S
,
Mi
S
,
Zhang
W
,
Dolan
ME
. 
Comprehensive analysis of the impact of SNPs and CNVs on human microRNAs and their regulatory genes
.
RNA Biol
2009
;
6
:
412
25
.
15.
Mishra
PJ
,
Humeniuk
R
,
Longo-Sorbello
GS
,
Banerjee
D
,
Bertino
JR
. 
A miR-24 microRNA binding-site polymorphism in dihydrofolate reductase gene leads to methotrexate resistance
.
Proc Natl Acad Sci U S A
2007
;
104
:
13513
8
.
16.
Adams
BD
,
Furneaux
H
,
White
BA
. 
The micro-ribonucleic acid (miRNA) miR-206 targets the human estrogen receptor-alpha (ERalpha) and represses ERalpha messenger RNA and protein expression in breast cancer cell lines
.
Mol Endocrinol
2007
;
21
:
1132
47
.
17.
Jazdzewski
K
,
Murray
EL
,
Franssila
K
,
Jarzab
B
,
Schoenberg
DR
,
de la Chapelle
A
. 
Common SNP in pre-miR-146a decreases mature miR expression and predisposes to papillary thyroid carcinoma
.
Proc Natl Acad Sci U S A
2008
;
105
:
7269
74
.
18.
Naccarati
A
,
Pardini
B
,
Landi
S
,
Landi
D
,
Slyskova
J
,
Novotny
J
, et al
Polymorphisms in miRNA binding sites of nucleotide excision repair genes and colorectal cancer risk
.
Carcinogenesis
2012
;
33
:
1346
51
.
19.
Landi
D
,
Gemignani
F
,
Naccarati
A
,
Pardini
B
,
Vodicka
P
,
Vodickova
L
, et al
Polymorphisms within micro-RNA-binding sites and risk of sporadic colorectal cancer
.
Carcinogenesis
2008
;
29
:
579
84
.
20.
Landi
D
,
Moreno
V
,
Guino
E
,
Vodicka
P
,
Pardini
B
,
Naccarati
A
, et al
Polymorphisms affecting micro-RNA regulation and associated with the risk of dietary-related cancers: a review from the literature and new evidence for a functional role of rs17281995 (CD86) and rs1051690 (INSR), previously associated with colorectal cancer
.
Mutat Res
2011
;
717
:
109
15
.
21.
Center
MM
,
Jemal
A
,
Ward
E
. 
International trends in colorectal cancer incidence rates
.
Cancer Epidemiol Biomarkers Prev
2009
;
18
:
1688
94
.
22.
Pardini
B
,
Kumar
R
,
Naccarati
A
,
Prasad
RB
,
Forsti
A
,
Polakova
V
, et al
MTHFR and MTRR genotype and haplotype analysis and colorectal cancer susceptibility in a case-control study from the Czech Republic
.
Mutat Res
2011
;
721
:
74
80
.
23.
Polakova
V
,
Pardini
B
,
Naccarati
A
,
Landi
S
,
Slyskova
J
,
Novotny
J
, et al
Genotype and haplotype analysis of cell cycle genes in sporadic colorectal cancer in the Czech Republic
.
Hum Mutat
2009
;
30
:
661
8
.
24.
Barenboim
M
,
Zoltick
BJ
,
Guo
Y
,
Weinberger
DR
. 
MicroSNiPer: a web tool for prediction of SNP effects on putative microRNA targets
.
Hum Mutat
2010
;
31
:
1223
32
.
25.
Sharma
V
,
Misteli
T
. 
Non-coding RNAs in DNA damage and repair
.
FEBS Lett
2013
;
587
:
1832
9
.
26.
Iorio
MV
,
Croce
CM
. 
microRNA involvement in human cancer
.
Carcinogenesis
2012
;
33
:
1126
33
.
27.
Slaby
O
,
Bienertova-Vasku
J
,
Svoboda
M
,
Vyzula
R
. 
Genetic polymorphisms and microRNAs: new direction in molecular epidemiology of solid cancer
.
J Cell Mol Med
2012
;
16
:
8
21
.
28.
Teo
MT
,
Landi
D
,
Taylor
CF
,
Elliott
F
,
Vaslin
L
,
Cox
DG
, et al
The role of microRNA-binding site polymorphisms in DNA repair genes as risk factors for bladder cancer and breast cancer and their impact on radiotherapy outcomes
.
Carcinogenesis
2012
;
33
:
581
6
.
29.
Pettersen
HS
,
Visnes
T
,
Vagbo
CB
,
Svaasand
EK
,
Doseth
B
,
Slupphaug
G
, et al
UNG-initiated base excision repair is the major repair route for 5-fluorouracil in DNA, but 5-fluorouracil cytotoxicity depends mainly on RNA incorporation
.
Nucleic Acids Res
2011
;
39
:
8430
44
.
30.
Ingraham
HA
,
Tseng
BY
,
Goulian
M
. 
Mechanism for exclusion of 5-fluorouracil from DNA
.
Cancer Res
1980
;
40
:
998
1001
.
31.
Wallace
SS
,
Murphy
DL
,
Sweasy
JB
. 
Base excision repair and cancer
.
Cancer Lett
2012
;
327
:
73
89
.
32.
Kavli
B
,
Sundheim
O
,
Akbari
M
,
Otterlei
M
,
Nilsen
H
,
Skorpen
F
, et al
hUNG2 is the major repair enzyme for removal of uracil from U:A matches, U:G mismatches, and U in single-stranded DNA, with hSMUG1 as a broad specificity backup
.
J Biol Chem
2002
;
277
:
39926
36
.
33.
An
Q
,
Robins
P
,
Lindahl
T
,
Barnes
DE
. 
5-Fluorouracil incorporated into DNA is excised by the Smug1 DNA glycosylase to reduce drug cytotoxicity
.
Cancer Res
2007
;
67
:
940
5
.
34.
Banerjee
D
,
Mandal
SM
,
Das
A
,
Hegde
ML
,
Das
S
,
Bhakat
KK
, et al
Preferential repair of oxidized base damage in the transcribed genes of mammalian cells
.
J Biol Chem
2011
;
286
:
6006
16
.
35.
Landi
D
,
Gemignani
F
,
Pardini
B
,
Naccarati
A
,
Garritano
S
,
Vodicka
P
, et al
Identification of candidate genes carrying polymorphisms associated with the risk of colorectal cancer by analyzing the colorectal mutome and microRNAome
.
Cancer
2012
;
118
:
4670
80
.
36.
Cummins
JM
,
He
Y
,
Leary
RJ
,
Pagliarini
R
,
Diaz
LA
 Jr
,
Sjoblom
T
, et al
The colorectal microRNAome
.
Proc Natl Acad Sci U S A
2006
;
103
:
3687
92
.
37.
Boni
V
,
Zarate
R
,
Villa
JC
,
Bandres
E
,
Gomez
MA
,
Maiello
E
, et al
Role of primary miRNA polymorphic variants in metastatic colon cancer patients treated with 5-fluorouracil and irinotecan
.
Pharmacogenomics J
2011
;
11
:
429
36
.
38.
Hummel
R
,
Wang
T
,
Watson
DI
,
Michael
MZ
,
Van der Hoek
M
,
Haier
J
, et al
Chemotherapy-induced modification of microRNA expression in esophageal cancer
.
Oncol Rep
2011
;
26
:
1011
7
.
39.
Mishra
PJ
,
Bertino
JR
. 
MicroRNA polymorphisms: the future of pharmacogenomics, molecular epidemiology and individualized medicine
.
Pharmacogenomics
2009
;
10
:
399
416
.
40.
Lin
M
,
Gu
J
,
Eng
C
,
Ellis
LM
,
Hildebrandt
MA
,
Lin
J
, et al
Genetic polymorphisms in MicroRNA-related genes as predictors of clinical outcomes in colorectal adenocarcinoma patients
.
Clin Cancer Res
2012
;
18
:
3982
91
.
41.
Wu
Y
,
Xiao
Y
,
Ding
X
,
Zhuo
Y
,
Ren
P
,
Zhou
C
, et al
A miR-200b/200c/429-binding site polymorphism in the 3′ untranslated region of the AP-2alpha gene is associated with cisplatin resistance
.
PLoS ONE
2011
;
6
:
e29043
.
42.
Haas
U
,
Sczakiel
G
,
Laufer
SD
. 
MicroRNA-mediated regulation of gene expression is affected by disease-associated SNPs within the 3′-UTR via altered RNA structure
.
RNA Biol
2012
;
9
:
924
37
.