Introduction

Tumor growth and progression requires the formation of new blood vessels, a process called angiogenesis. Angiogenesis is a complex multifactorial process involving a variety of proangiogenic and proteolytic enzyme activators and inhibitors [1]. The most important regulator of angiogenesis is vascular endothelial growth factor (VEGF), which is overexpressed in several tumor tissues. VEGF is a disulfide-bonded dimeric glycoprotein, sharing close sequence homology with placenta growth factor, VEGF-B and VEGF-C, and lower sequence homology with platelet-derived growth factor (PDGF) [2]. VEGF plasma levels are highly predictive for tumor growth and survival rate of breast cancer patients [3, 4] and therapeutic strategies blocking VEGF action successfully inhibited tumor growth [4].

Several single nucleotide polymorphisms (SNPs) have been described in the VEGF gene, some of them have been associated with VEGF expression and/or clinical phenotypes. We have previously reported a significant association of one variant, VEGF 936C>T (rs3025039), with decreased risk for breast cancer [5]. This result was subsequently confirmed by one study [6], but refuted by two other studies [7, 8].

In order to replicate and expand previous data on the role of VEGF polymorphisms in breast cancer risk, we have determined seven SNPs and haplotypes of the VEGF gene in 804 incident breast cancer patients and 804 healthy age-matched population-based control subjects.

Materials and methods

Subjects

The Austrian “tumor of breast tissue: incidence, genetics, and environmental risk factors” (TIGER) study is an ongoing study investigating risk factors for breast cancer. TIGER consists of 804 consecutive female patients with histologically confirmed incident breast cancer without any other cancer diagnosis beside breast cancer. All patients were recruited between January 2000 and September 2004 from patients attending the Division of Oncology, Department of Internal Medicine, Medical University Graz, Austria. Patients were included in the aftercare measures program of the Division of Oncology Graz, providing follow-ups in regular intervals (3 months interval in years 1–3, 6 months interval in years 4–5, and 12 months interval in years 6–15 after diagnosis). Follow-up investigations included clinical check-up, laboratory (including CEA and CA15-3), radiological (bone scan, liver scan, chest X-ray, and mammograms), and gynecological analyses.

The majority of the participants of TIGER had not participated in any genetic association study before, a small fraction of TIGER participants (n = 18) were also included in a previous case–control study from the same Department [5].

For each patient of the TIGER study, one healthy female age-matched (±2 years) control subject was enrolled. Control subjects were recruited from local health screening studies, the presence of known current or previous malignant disease was excluded anamnestically.

The study was performed according to the Austrian Gene Technology Act and has been approved by the Ethical Committee of the Medical University Graz. Written informed consent was obtained from all participating subjects. All study participants (patients and controls) were Caucasians (Table 1).

Table 1 Primer and probe sequences for 5′-exonuclease assays
Full size table

Selection of VEGF polymorphisms

With the use of the public NCBI SNP database and available literature [9, 10–12], we selected VEGF candidate polymorphisms with a minor allele frequency of at least 0.10 and location in the promoter region, coding region or untranslated region of the VEGF gene. Using this approach, seven common VEGF polymorphisms were chosen for further analysis (Fig. 1).

Fig. 1
figure 1

Structure of the VEGF gene and position of candidate gene polymorphisms. Position of polymorphism are relative to the translation start, italic positions indicate alternative designations. Dashed lines indicate 13 kb region between upstream polymorphisms and downstream polymorphisms, containing the coding sequence (CDS) and seven introns

Full size image

DNA isolation and genotyping assays

Genomic DNA was isolated by standard procedures. VEGF genotypes were determined between November 2005 and August 2006 using 5′-nuclease assays (TaqMan). Reaction conditions were as described previously [13]. Primers and probe sets are summarized in Table 2. The laboratory staff responsible for genotyping were blinded for case/control status.

Table 2 Characteristics of breast cancer patients enrolled in the TIGER study
Full size table

Determination of VEGF plasma levels

The VEGF plasma levels were determined in 81 healthy volunteers (64 men and 17 women) using a commercially available enzyme immunoassay (human VEGF Quantikine, R&D Systems, Wiesbaden, Germany) as described previously [10]. All reactions were performed in duplicate. The assay was specific for VEGF165 and did not detect related molecules, e.g., PDGF or placental growth factor.

Construction of haplotypes and statistical analysis

Haplotypes and linkage disequilibrium were determined using the Haploview program (Version 2.05, http://www.broad.mit.edu/personal/jcbarret/haploview/). Assignment of individual haplotype pairs was performed by the PHASE Version 2.1 software [14]. Statistic analysis was done using SPSS 14.0 for Windows. Numeric values were analyzed by Student’s t-test, proportions of groups were compared by chi-squared test. Odds ratio (OR) and 95% confidence interval (95% CI) were calculated by logistic regression analysis. Threshold for significance was p < 0.05.

Due to the fact that precise frequencies of VEGF genotypes and haplotypes were not at the planning phase of the study, an a priori Power analysis was performed assuming a frequency of 0.1 for a hypothetical genetic risk marker. Using these condidtions, the present study had a Power of 0.99, 0.95 or 0.75 to detect or exclude an OR of 2.0, 1.7 or 1.5 for breast cancer. The statistical Power increased with higher frequencies and/or higher ORs, and decreased with lower frequencies and/or lower ORs of genetic markers.

Results

Tumor characteristics of TIGER participants are presented in Table 2. VEGF genotypes did not deviate from the Hardy Weinberg equilibrium in patients or controls. Haplotype analysis showed two separate blocks of high-linkage disequilibrium, formed by five polymorphisms upstream of the coding sequence (promoter and 5′ untranslated region) and two polymorphisms downstream of the coding sequence, respectively (Fig. 2). None of the single polymorphisms or haplotypes was significantly associated with the presence of breast cancer (Table 3).

Fig. 2
figure 2

Linkage disequilibrium of VEGF polymorphisms. Values in squares are LD′ between single markers. Dark squares indicate high r 2 and bright squares indicate low r 2 values

Full size image
Table 3 Allele frequencies and association of VEGF single markers and haplotypes with breast cancer
Full size table

Tumor characteristics of breast cancer patients stratified by VEGF genotypes and haplotypes are summarized in Tables 4 and 5. Five associations were below the significance treshold of 0.5. Applying Bonferroni correction for multiple testing, only the association of the -634G>C polymorphism with tumor size remained statistically significant. VEGF genotypes or haplotypes were furthermore not associated with HER2neu overexpression, estrogen receptor status or progesteron receptor status (data not shown).

Table 4 Relationship between VEGF genotypes and breast cancer characteristics
Full size table
Table 5 Relationship between VEGF haplotypes and breast cancer characteristics
Full size table

The potential association of VEGF polymorphisms and haplotypes with VEGF plasma levels were determined in a multivariate linear regression analysis. Polymorphisms and haplotypes were entered assuming codominant effects (0 = polymorphism/haplotype not present; 1 = one copy present; 2 = two copies present). None of the VEGF polymorphism or haplotypes was a significant predictor of VEGF plasma levels. This did not change when sex and age were entered in the model as additional covariates (Table 6).

Table 6 Multivariate linear regression of VEGF plasma levels
Full size table

Discussion

Aim of the present study was to re-evaluate the association of VEGF gene polymorphisms and their haplotypes with breast cancer risk in a large case–control study including incident patients and population-based control subjects. No significant differences in allele, genotype, and haplotype distribution of the VEGF gene polymorphisms between breast cancer cases and controls were detected. Our study does not support the notion that VEGF polymorphisms do modify the risk of breast cancer.

Our data are in contrast to a case–control study we had performed previously [5]. In that study, including 500 Caucasian breast cancer cases and 500 controls, we observed a decreased risk for breast cancer in carriers of VEGF 936T allele. VEGF genotypes in that study were determined by a PCR-RFLP, which may be more error-prone than the TaqMan assay used in the present study. Furthermore, the present study included only incident breast cancer patients, whereas in the previous study incident and as well as prevalent patients had been included.

The fact that the significant association between the VEGF 936C>T polymorphism and breast cancer was not replicated in the present study remains puzzling. On the other hand, non-replication of significant primary genetic association results is a well-known phenomenon in the field of genetic epidemiology. Consequently, a number of methodical papers on genetic association studies have stressed the importance of studies confirming (or confuting) results of primary reports [15–17]. Replication of association studies is imperative to draw firm conclusions about the role of genetic risk factors.

Recently, Jin and co-workers investigated the association between four VEGF polymorphisms (-2578C>A, -1154G>A, -634G>C, and 936C>T) and breast cancer risk [8]. As main result of their study, no association between VEGF polymorphisms or haplotypes and the presence of breast cancer was observed. This is in line with our findings. However, Jin and co-workers reported an association of the VEGFR -634CC genotype and the -2578/-634 CC haplotype with high-tumor aggressiveness (large tumor and high-histologic grade). Interestingly, the same genotype was associated with smaller tumor size and had no effect on histologic grade in the present study. These opposing results underline again the utmost importance of replication of genetic studies.

In a study by Jacobs and co-workers, VEGF alleles -2578C and −1154G were associated with invasive, but not with in situ breast cancer. VEGF polymorphisms −634G>C and 936C>T were not related to breast cancer susceptibility [7]. Kataoka and co-workers reported that breast cancer risk was influenced by the VEGF 936C>T polymorphism, but not by the −1498T>C or the −634G>C polymorphism [6]. Taken together, on the basis of currently available data, a clear effect of VEGF genotypes on breast cancer risk is unlikely.

The hypothesis that polymorphisms of the VEGF might influence breast cancer risk has been built upon the notion that VEGF gene polymorphism are associated with altered VEGF gene expression. In the present study, we were unable to detect any clear effects of VEGF genotypes on VEGF plasma levels. This is in line with a recent publication from Berrahmoune and co-workers, who reported that plasma VEGF concentrations were under strong genetic control in healthy families, but not influenced by VEGF genotypes at positions −1498, −634 or 936 [18]. It is likely that substantial genetic determinants of vascular growth might be found in other candidates genes, such as those for hypoxia inducable factor (HIF1), VEGF receptor 1 (Kinase Insert Domain Receptor; KDR) or VEGF receptor 2 (FMS-Related Tyrosine Kinase 1; FLT [19]).