Introduction

The survival of colon cancer (CC) has improved in recent decades, however this disease remains the third cause of cancer death worldwide [1]. Patients with stages II and III CC undergo a complete surgical resection with curative intent. However, the risk of tumor recurrence is considerable, especially in patients with stage III and high-risk stage II disease. In these stages, adjuvant chemotherapy following surgery is recommended.

Several biomarkers have been correlated with prognosis of CC. Microsatellite instability, for example, implies a favorable outcome in stage II patients. Other putative biomarkers include BRAF, KRAS PIK3CA mutations and CDX2 expression, but few studies have described the role of molecules such as VEGF and interleukins in the likelihood of recurrence in stage II–III colon cancer [2,3,4,5].

The vascular endothelial growth factor (VEGF) pathway plays a key role in tumor-induced angiogenesis, which promotes the growth and progression of solid tumors [6, 7]. Activation of this pathway may confer a worse prognosis in stages II–III colon cancer as it can promote the switch from dormant tumor cells not removed by surgery to proliferative cells [8]. Multiple proteins are involved in the VEGF pathway, particularly VEGFA, a major mediator of angiogenesis whose synthesis is stimulated under hypoxic conditions by HIFα. Several receptors (VEGFR1/FLT1, VEGFR2/KDR, VEGFR3 and FGFR), neuropilins (NRP1) and integrins (ITGAV), among other proteins are involved in the process. The binding of circulating VEGFA to its main receptors, VEGFR1 and VEGFR2, promotes cell migration, survival and proliferation by triggering a protein cascade that involves PRKCE, GRB2, RAS and MAP kinases [9,10,11].

Heritable functional variation in genes involved in the angiogenesis process may impact on the angiogenic switch, tumor progression and, hence, on clinical cancer outcomes. In the present work, we analyzed 28 genetic variants in a panel of 12 VEGF-dependent genes (VEGFA, VEGFR1/FLT1, VEGFR2/KDR, GRB2, ITGAV, KISS1, KRAS, PRKCE, HIF1α, MAP2K4, MAP2K6, MAPK11). The goal of this study was to evaluate the association between the selected genetic variants and the clinical outcome in a cohort of 347 stages II–III CC patients.

Materials and methods

Patient population

Three-hundred and forty-seven patients with stage II–III CC who underwent radical surgery between 2009 and 2014 at Hospital de la Santa Creu i Sant Pau (HSCSP, Barcelona, Spain) were consecutively included in the study. Patients with rectal tumor were excluded. Patient data including tumor localization, histological tumor grade, lymph node sampling and vascular or perineural invasion were collected retrospectively through chart review. Patients were treated with fluoropyrimidine-based adjuvant chemotherapy at clinician discretion. Blood samples from 345 patients were available for the genetic analyses. The study was approved by the Institutional Ethics Committee at HSCSP and all the study participants gave informed consent for the analysis of molecular correlates.

Genetic studies

We analyzed 28 single-nucleotide polymorphisms (SNPs) in 12 candidate genes involved in the VEGF pathway. The selected SNPs were (i) SNPs associated with lymphoblastoid cell line (LCL) mRNA expression of VEGF pathway genes [12] and (ii) functional variants associated with cancer survival according to studies reported in the literature [5, 13,14,15,16,17,18,19,20,21,22]. The SNPs that were Tag SNPs had an r 2 greater than 0.8. All SNPs had a minor allele frequency (MAF) >5% in the Caucasian population. Table 1 shows detailed information about the SNPs.

Table 1 Selected polymorphisms in VEGF-related genes
Full size table

Genomic DNA was automatically extracted from peripheral whole-blood samples (Autopure, Qiagen, Hilden, Germany). SNP genotyping was performed by means of real-time PCR using TaqMan® SNP genotyping assays (Applied Biosystems, Foster City, CA, USA) and 48.48 dynamic arrays on the BioMarkTM system (Fluidigm, San Francisco, CA, USA). Patients’ characteristics and clinical outcome were unknown to the investigator conducting the genetic analyses. SNP allele frequencies were comparable to those reported in the 1000 Genomes project. Genotyping was successful in atleast 99% of cases for each SNP analyzed. The quantity and/or the quality of the extracted DNA were the most common causes of failure.

Statistical analyses

The endpoints of the study were recurrence-free survival (RFS) and overall survival (OS). RFS was calculated from the date of surgery until the date of first recurrence. RFS was censored at the last follow-up or the time of death if the patient remained tumor recurrence-free at that time. OS was defined as the time from the surgery date until death from any cause or last follow-up. We also analyzed survival after relapse (SaR), which was defined as the time from the relapse date until death from any cause or last follow-up. All patients were included in the colon cancer surveillance program of HSCSP, providing history and undergoing physical examination and CEA determination every 3 months for 2 years and every 6 months at years 3–5 after surgery, colonoscopy at year 1 and thereafter every 3–5 years, and computed tomographic scans of chest and abdomen every year. The associations between polymorphisms and clinicopathological features were evaluated using Fisher’s exact test. The associations of polymorphisms with RFS and OS were analyzed using Kaplan–Meier curves and a log-rank test. Three different models of inheritance were considered to evaluate associations with outcome variables—additive, dominant and recessive. To identify markers associated with the outcomes independently from clinical data, each SNP was fitted to a Cox model in which the statistically significant clinicopathological variables in the univariate analyses were included as covariates. For each SNP, Hardy–Weinberg equilibrium was assessed using an exact test. All statistical tests were performed at 95% significance. All analyses were performed using SPSS (version 19.0, IBM).

Results

We studied a total of 347 patients. Two-hundred and five were diagnosed as stage II and 142 were classified as stage III CC. Following total surgical resection of the tumor, 47 patients received adjuvant fluorouracil plus leucovorin (FL) or oral fluoropyrimidine capecitabine, and 162 patients were treated with oxaliplatin added to FL (FOLFOX) or capecitabine (XELOX). The remaining 138 patients did not receive adjuvant therapy.

Table 2 shows the baseline clinical characteristics. The median follow-up was 46.3 months (range, 9.3–91.5 months). Seventy-one patients (20.5%) relapsed during the follow-up. Thirty-four of the 71 corresponded to 16.6% of the stage II patients, while the remaining 37 patients represented 26.1% of the stage III group.

Table 2 Baseline patient characteristics
Full size table

The stage II and stage III probability of 3-year recurrence-free survival was 0.876 ± 0.023 and 0.800 ± 0.034, respectively. The stage II and stage III probability of 5-year overall survival was 0.860 ± 0.025 and 0.766 ± 0.036, respectively.

Stage classification was significantly associated with RFS (p = 0.028) and OS (p = 0.001). Tumor grade was significantly associated with OS (p = 0.032). Localization of the tumor and administration of adjuvant therapy were associated with OS (p = 0.05). Perineural invasion and the number of resected lymph nodes were also significantly associated with RFS (p = 0.004 and p = 0.05, respectively). Thus, these covariates were included in the multivariate model.

Genetic determinants and RFS

In the univariate analysis of the stage II and III patients, two genetic variants were significantly associated with RFS: rs61764370 in the KRAS gene (p = 0.041) and rs699947 in the VEGFA gene (p = 0.035). These results are detailed in Table 3. Non-significant genetic variants are presented in the Supplementary Table. Stage II patients showed a significant association between RFS and four genetic variants: rs35251833 in the ITGAV gene (p = 0.018), rs71745629 in the KISS1 gene (p = 0.018), rs61764370 in the KRAS gene (p = 0.046) and rs4953299 in the PRKCE gene (p = 0.033) (Table 4). No association was found between any of the SNPs evaluated and RFS in stage III colon cancer patients. In the multivariate analysis, none of the variants remained significantly associated with RFS.

Table 3 Germline polymorphisms in the VEGF pathway and univariate analysis for time to recurrence and overall survival in patients with stage II or III colon cancer
Full size table
Table 4 Germline polymorphisms in the VEGF pathway associated with RFS/OS by tumor stage
Full size table

Genetic determinants and OS

In the univariate analysis of the stage II and III patients, three genetic variants were significantly associated with OS: rs11656130 (MAP2K6) (p = 0.017), rs9513070 (VEGFR1) (p = 0.041) and rs1551641 (VEGFR2) (p = 0.049). Table 3 shows these results. Polymorphisms not significantly associated with OS are detailed in the Supplementary Table. Stage II patients showed a significant association between OS and six genetic variants corresponding to four different genes: rs11549465 (HIF1α) (p = 0.003); rs1137282 (KRAS) (p = 0.001); rs833061, rs1570360 and rs699947 (VEGFA) (p = 0.038, p = 0.009 and p = 0.035, respectively) and rs9513070 (VEGFR1) (p = 0.025) (Table 4). No association was found between any of the SNPs evaluated and OS in stage III colon cancer patients.

In the multivariate analysis, two SNPs retained their significance: rs1137282 in the KRAS gene and rs9513070 in the VEGFR1 gene. Stage II patients harboring a homozygous GG genotype in the KRAS SNP were significantly associated with decreased OS (HR = 10.82; 95% CI: 2.1–56.0; adjusted p = 0.005). In the same cohort of stage II CC patients, patients carrying at least one G allele of the SNP in the VEGFR1 gene were significantly associated with an increased OS (HR = 0.268; 95% CI: 0.07–0.95; adjusted p = 0.043).

Considering the clinical parameter survival after relapse, only the rs11656130 genetic variant in the MAP2K6 gene showed a significant association (p = 0.027). However, in the multivariate analysis this association did not remain statistically significant.

Genetic determinants and survival by localization of the tumor

In the univariate analysis, three genetic variants showed an association with RFS in patients with left colon cancer tumors: rs35251833 (ITGAV) (p = 0.032), rs833061 (VEGFA) (p = 0.036) and rs9513070 (VEGFR1) (p = 0.033) (Table 5). In the multivariate analysis, only two genetic variants retained their significance: rs9513070 of VEGFR1 gene and rs35251833 of ITGAV gene. Left colon cancer patients carrying at least one G allele in the VEGFR1 SNP showed a longer time to recurrence (HR = 0.345; 95% CI: 0.2–0.7; adjusted p = 0.005). The same cohort of left colon cancer patients harboring at least one G allele in the ITGAV variant showed an increased RFS (HR = 0.242; 95% CI: 0.1–0.6; adjusted p = 0.005).

Table 5 Germline polymorphisms in the VEGF pathway associated with RFS/OS by localization of the tumor
Full size table

Considering OS and tumor location, evaluation of the genetic variants included in the study showed a significant association with the rs9513070 variant of VEGFR1 gene (p = 0.031) in left-side tumors, and with two variants in the MAP2K6 gene: rs11656130 in right-sided tumors (p = 0.026) and rs2716191 in left-sided tumors (p = 0.029). Table 5 shows these results and the mode of inheritance. In the multivariate analysis, only the VEGFR1 variant retained its significance. Left colon cancer patients carrying atleast one G allele in the VEGFR1 SNP showed a longer overall survival (HR = 0.061; 95% CI: 0.01–0.5); adjusted p = 0.011).

Discussion

We investigated the role of germline polymorphisms as prognostic factors in stages II and III colon cancer patients in several genes involved in the angiogenesis pathway. We found that several genetic variants in the VEGFA and its main receptors, VEGFR1 and VEGFR2, were associated with RFS and/or OS and that these associations were influenced by tumor stage and location. Several polymorphisms in additional genes of the VEGF pathway such as MAP2K6, KRAS, PRKCE and ITGAV, as well as in HIFα, a protein that stimulates the synthesis of VEGFA, and in the metastasis suppressor KISS1, were also associated with survival. However, only rs9513070 in the VEGFR1 gene, rs1137282 in the KRAS gene and rs35251833 in the ITGAV gene retained their significance after multivariate adjustments.

In the present work, we discuss only those published studies that have identified polymorphisms in VEGF-related genes as predictive biomarkers for tumor recurrence in stages II or III colon cancer patients. We do not discuss data from case-control studies.

In 2008, Lurje et al. studied 10 polymorphisms in eight genes involved in the angiogenesis pathway in a cohort of 125 stage III CC patients [5]. They demonstrated that patients with the T allele in rs3025039 ( + 936 C/T), located in the VEGFA gene, showed a significantly longer time to recurrence (TTR) (p = 0.003). However, they found no correlation between rs2010963 (−634 G/C), also located in the VEGFA gene, and TTR. Some time later, these same authors also analyzed both polymorphisms in a cohort of 109 stage II colon cancer patients [23]. Surprisingly, they found a contradictory result: rs3025039 was not associated with TTR, whereas rs2010963 was significantly associated with this outcome parameter (p = 0.028). They also analyzed rs1870377 of VEGFR2, finding no association between this polymorphism and TTR. More recently, Kjaer-Frifeldt et al. studied 3 polymorphisms in the VEGFA gene in a cohort of 698 stage II colon cancer patients. They found that rs699947 and rs833061 were significantly associated with TTR (p = 0.02 for both SNPs), whereas rs2010963 was not [24]. It should be emphasized that these studies evaluated a reduced number of SNPs in VEGF-related genes in a cohort of stages II and III colon cancer patients and that their results are contradictory. For this reason we designed a study that included functional SNPs of several VEGF-related genes, focusing especially on the ligands (VEGFA), receptors (VEGFR1 and VEGFR2), and other relevant genes in the VEGF signal cascade.

One of the results of the present study concerns rs9513070, a SNP located in the VEGFR1 gene. To the best of our knowledge, this is the first study that shows that the variant rs9513070 is associated with survival in both stage II CC and in left-sided primary tumor location. Tumor sidedness has a great impact on prognosis, as right and left colon cancer tumors harbor different biological features. Whereas right-sided tumors are more likely to present microsatellite instability, mutations in KRAS or BRAF or MLH1 methylation, left-sided tumors are associated with CIN, p53 or miRNAs [25,26,27,28,29]. Additionally, right colon cancers present a higher expression of pro-angiogenic factors, which justifies the better response to bevacizumab of these tumors [30]. Interestingly, rs9513070 variant had been associated with survival in a previous study conducted in advanced CC patients [21]. In that study, the median PFS and OS were superior in patients with the VEGFR1 rs9513070 A/A genotype (8.7 vs. 6.6 months; p = 0.001 and 26.4 vs. 16.1 months; p = 0.038, respectively). Expression studies conducted by Ning et al. showed that VEGF and VEGFR1 gene expression levels were significantly associated with TTR. One of their studies included 140 stage II–III CC patients treated with adjuvant chemotherapy [31]. They observed that patients with lower VEGF or VEGFR1 gene expression levels had longer time to recurrence than those with higher expression levels of these genes (p < 0.05, log-rank test). All these data, along with the results of the in silico analysis of this intron variant (http://compbio.cs.queensu.ca/F-SNP), strengthen the hypothesis that VEGFR1 could play a role in CC outcomes. However, until the mechanisms underlying all these findings are not fully elucidated an independent validation study that confirms our results is needed to highlight the clinical relevance of VEGFR1 polymorphisms.

Regarding the association between rs1137282 in the KRAS gene and OS, we observed that stage II patients harboring a homozygous GG genotype were significantly associated with a decreased OS. Our group recently conducted a study in 131 patients who underwent surgery for stage I-III non-small cell lung cancer. We found that the GG genotype was also associated with a poor prognosis [17]. In vitro experiments have demonstrated that the minor allele (G) increase luciferase activity significantly and in silico analyses predict that this SNP introduces an exon splicing enhancer motif and disrupts an exon splicing silencer motif [12]. These findings reinforce the role of this SNP as a prognostic biomarker in cancer.

Finally, the present study shows that rs35251833 SNP in the ITGAV gene correlated with RFS in left CC patients: G/G and G/A genotypes were significantly associated with an increased RFS. This variant is in high linkage-disequilibrium with rs1839123 that showed similar results in a previous study conducted by our group in metastatic colorectal cancer patients [14]. Bohanes et al. studied the role of genetic variants in several integrin genes and although they reported some associations in stages II–III CC, the rs35251833 was not included in their study [32]. In silico analysis of this intronic SNP suggests a potential role in transcriptional regulation (http://compbio.cs.queensu.ca/F-SNP).

In conclusion, this study provides evidence that germline polymorphisms in VEGFR1, KRAS and ITGAV genes are associated with prognosis in stages II–III CC patients. As stage and tumor location have also been correlated with prognosis, future genetic studies should stratify CC patients according to these parameters.