The 613 patients included in the study belonged to the population-based PopGen project in Schleswig-Holstein (Germany) 11. All individuals were of German origin, i.e. with both parents born in Germany. Patients fulfilling either the clinical Amsterdam or Bethesda criteria for hereditary non-polyposis colorectal cancer (HNPCC) were excluded from the study, as were patients with a history of malignant disease or inflammatory bowel disease. For a full description of the characteristics of the 613 patients see Table 1 and Castro et al. 12. The work has been approved by the corresponding ethical committees and subjects gave informed consent to the work.

a Comprises radiotherapy (n = 13), chemotherapy (n = 86), radio-chemotherapy (n = 59), immunotherapy (n = 3), chemotherapy plus 5FU/FA (n = 68), radio-chemotherapy plus 5FU/FA (n = 18), and other (n =3).

b Comprises radiotherapy (n = 5) and radio-chemotherapy (n = 63).

Candidate genes in the OXPHOS pathway (map00190 in the KEGG database 13) were selected based on a recently published meta-analysis of GEP studies on CRC prognosis 3. Three out of the seven genes reported in two GEP studies had SNPs in the putative regulatory regions (ATP5C1 from complex V, COX5B and COX6B1 from complex IV). Two additional genes with SNPs in putative regulatory regions, each reported in one GEP study, were also included, namely NDUFS2 from complex I and UQCRB from complex III. With these selection criteria, the four complexes from the OXPHOS chain which allow flowing of protons between the mitochondrial matrix and the intermembrane space, resulting into energy generation in form of ATP, were included in the study. Finally, SNPs in the genes GAPDH and HSPD1 were also selected, since both genes might define the cellular bioenergetic activity of the mitochondria 6.

SNPs within the set of candidate genes (Table 2 and Additional file 1: Table S1) were selected according to the following criteria: minor allele frequency (MAF) > 0.05 in HapMap CEPH population (Utah residents with ancestry from northern and western Europe - CEU), and location in potentially regulatory gene regions (5′ and 3′-UTR, promoter or non-synonymous coding SNPs). In regions with several SNPs in high linkage disequilibrium (LD), tagSNPs were selected for genotyping. LD between the SNPs was calculated using Haploview 14. The promoter SNP rs2802460 in ATP5C1 was reported to regulate expression of the gene in lymphoblastoid cell lines (eSNPs) on a genome-wide association study of global gene expression 15. Predictions of the possible impact of amino acid substitutions on the structure and function of the protein were performed with PolyPhen2 16.

Two additional SNPs were genotyped, but showed only one allele in the 613 samples: rs2071038 (COX5B) and rs955 (HSPD1).

MAF, minor allele frequency in the 613 CRC patients investigated; NDUFS2, NADH dehydrogenase (ubiquinone) Fe-S protein 2, 49kDa (NADH-coenzyme Q reductase); UQCRB, ubiquinol-cytochrome c reductase binding protein; COX5B, cytochrome c oxidase subunit Vb; COX6B1, cytochrome c oxidase subunit VIb, polypeptide 1; ATP5C1, ATP synthase, H + transporting, mitochondrial F1 complex, gamma polypeptide 1; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; HSPD1, heat shock 60kDa protein 1 (chaperonin).

The 613 DNAs used in the study were extracted from peripheral venous blood using the FlexiGene chemistry (Qiagen, Hilden, Germany) according to the manufacturer’s protocols. From them, 272 (44.4%) samples were subjected to genome-wide amplification using the illustra GenomiPhi V2 DNA Amplification Kit (GE Healthcare, Freiburg, Germany) according to manufacturer’s instructions. Genotyping was performed using 5 ng of either genomic or whole-genome amplified DNA in 384-well plate format, using KASPar assays on demand (KBiosciences, Hoddesdon, UK) and following the manufacturer’s protocols. Average genotyping call rate was 98.8%. All polymorphisms were in Hardy-Weinberg equilibrium.

SAS software version 9.2 (SAS Institute) was used in all analysis. Unadjusted association between the genotyped markers and patient characteristics were evaluated by chi-square test. Odds ratios (ORs) with 95% confidence intervals (CIs) were estimated using logistic regression. Effect of the different genotypes on CRC survival was evaluated using Kaplan-Meier method and was compared using log-rank testing. Analysis of different parameters for prognostic significance was done by univariate and multivariate Cox proportional hazard models. P values < 0.05 were considered statistically significant. Follow-up time was calculated from the date of CRC diagnosis to the CRC specific death or death by any cause or to the end of follow-up (date of last contact with the treating physician). Taking into account the possible biological effects of the selected genes/variants on cancer progression and a possible correlation between the variants, correction for multiple comparisons was excluded.

A total of 29 SNPs were identified in the regions of interest (5′- and 3′-UTR, promoter and non-synonymous coding SNPs) of the seven genes NDUFS2, UQCRB, COX5B, COX6B1, ATP5C1, GAPDH and HSPD1 ( Additional file 1: Table S1). Out of them, 18 SNPs were selected for the association studies, covering all gene regions with a putative regulatory function.

The two polymorphisms in the 3′-UTR region of UQCRB (complex III) showed association with CRC survival (Table 3 and Additional file 1: Table S2). The marker rs7836698 was statistically significantly associated with a decreased risk of death due to CRC (HR = 0.53, 95% CI 0.31–0.91 for TT), or due to any cause (HR = 0.52, 95% CI 0.32–0.85). For rs10504961, only the association with overall survival was statistically significant (HR = 0.64, 95% CI 0.42–0.99 for TT). Interestingly, for both UQCRB polymorphisms, the TT carriers were more likely to develop tumours in the colon than in the rectum, with ORs of 2.02 (95% CI 1.24–3.28) for rs7836698 (TT vs. CC) and 1.74 (95% CI 1.08–2.78) for rs10504961 (TT vs. CC). When survival was analysed according to the cancer site, both SNPs showed association exclusively in the group of patients with cancer located in the colon (HR for overall survival 0.42, 95% CI 0.22–0.82 for rs7836698 TT carriers and 0.46, 95% CI 0.25–0.83 for rs10504961 TT carriers), but not in the rectum (Table 3). The Kaplan-Meier survival curves representing the survival rates of the patients according to their genotypes, as well as the P values for the log-rank test are presented in Figure 1.

HR, hazard ratio; CI, confidence interval.

We examined also whether inclusion of other variables associated with CRC survival affected the parameter estimates for the UQCRB polymorphisms. The risk of death for carriers of the TT genotype at rs7836698 remained statistically significantly decreased after adjustment for age at diagnosis and TNM stage. Among patients with tumour located in the colon, the hazard ratios were similar both in the uni- and multivariate analyses for rs7836698 and rs10504961, although the statistical significance was sometimes lost.

Three of the 16 SNPs investigated showed an association with one of the different components that form the UICC stage (T, N and M) in a dominant model. Two SNPs in COX6B1 (complex IV) showed an association with the presence of affected lymph nodes (rs6510502, OR = 1.75, 95% CI 1.20–2.57; rs10420252, OR = 1.68, 95% CI 1.11–2.53). The polymorphism rs6510502 was also associated with distant metastasis (OR = 1.67, 95% CI 1.09–2.56). The marker rs7971637 in GAPDH was associated with tumour stage T3/T4 (OR = 1.60, 95% CI 1.11–2.30).