Single nucleotide polymorphisms as prognostic and predictive biomarkers in renal cell carcinoma

Despite major advances in the knowledge of the molecular basis of renal cell carcinoma, prognosis is still defined using clinical and pathological parameters. Moreover, no valid predictive biomarkers exist to help us selecting the best treatment for each patient. With these premises, we aimed to analyse the expression and to determine the prognostic and predictive value of 64 key single nucleotide polymorphisms in 18 genes related with angiogenesis or metabolism of antiangiogenics in two cohorts of patients with localized and advanced renal cell cancer treated at our institution. The presence of the selected single nucleotide polymorphisms was correlated with clinical features, disease free survival, overall survival and response rate. In patients with localized renal cell cancer, 5 of these polymorphisms in 3 genes involved in angiogenesis predicted for worse disease free survival (VEGFR2: rs10013228; PDGFRA: rs2228230) or shorter overall survival (VEGFR2: rs10013228; VEGFR3: rs6877011, rs307826) (p < 0.05). Rs2071559 in VEGFR2 showed a protective effect (p = 0.01). In the advanced setting, 5 SNPs determined inferior overall survival (IL8: rs2227543, PRKAR1B: rs9800958, PDGFRB: rs2302273; p = 0.05) or worse response rate (VEGFA: rs699947, rs3025010 p ≤ 0.01)). Additionally 1 single nucleotide polymorphism in VEGFB predicted for better response rate rs594942 (p = 0.03). Genetic analysis of renal cell carcinoma patients might provide valuable prognostic/predictive information. A set of SNPs in genes critical to angiogenesis and metabolism of antiangiogenics drugs seem to determine post-surgical outcomes and treatment response in our series.


INTRODUCTION
Renal cell carcinoma (RCC) is the most common malignancy of the kidney with near 338.000 new diagnoses per year worldwide [1]. It is more frequent in men and 75% of the patients are diagnosed over 60 years of age. Incidence of RCC has increased steadily at 2% per year contributing to about 144.000 deaths in 2012 [2,3]. Diverse histological variants have been described including clear cell (75%), papillary (10%), chromophobe (5%) and others [4]. Approximately 25% of the patients present with advanced disease at diagnosis, and up to one third of those with localized disease that undergo surgery with a curative intention will recur requiring systemic treatment [5].
homonymous tumor suppressor gene, in which around 60% of the patients develop clear cell RCC (ccRCC). In normal conditions the VHL product (VHLp) creates a complex that targets hypoxia inducible factors 1 and 2 (HIF 1-2) for ubiquitin-mediated degradation. In the absence of VHLp by either mutation or methylation of VHL gene, HIF accumulates leading to exaggerated transcription of multiple genes involved in cell proliferation and angiogenesis such as the platelet-derived growth factor (PDGF), vascular endothelial growth factor (VEGF) and transforming growth factor [6][7][8][9]. The VEGF binds its receptor (VEGFR) and promotes proliferation and migration of endothelial cells, increased vascular permeability and revascularization during tumor development [10][11][12]. Similarly, PDGF and its receptors (PDGFRA, PDGFRB) play a critical role in regulating angiogenesis through controlling functions during the mesenchymal cell development. Signalling through PDGF also promotes cell migration, survival and proliferation and indirectly regulates angiogenesis by inducing transcription and secretion of VEGF [13]. These knowledge and the observation that around 90% of sporadic ccRCC have abnormal function of VHL has led to an intense drug development in RCC targeting VEGF, PDGF or their cognate receptors. Bevacizumab, a humanized monoclonal antibody against VEGF, was the first agent in this class to demonstrate activity in advanced RCC [14]. Thereafter multiple antiangiogenics such as the tyrosine kinase inhibitors sunitinib, sorafenib, pazopanib or axitinib and mTOR inhibitors such as temsirolimus or everolimus, have shown remarkable activity in advanced RCC becoming standard of treatment in different settings [15]. More recently other therapeutic strategies such as targeting the program-death 1 (PD-1) receptor or the hepatocyte growth factor receptor (MET) have also succeeded [16,17].
Although the availability of all these drugs has improved substantially the therapeutic results in RCC, approximately 40% of patients treated in first-line will not achieve an objective response and about 20-25% will present an early progression. Currently available prognostic systems fail to identify these patients and no adequate predictive factors of response have been validated in advanced RCC yet.
The variability in the genetic constitution of the individual in critical genes related to disease mechanisms or anti-cancer drug metabolism could explain this variable clinical course. Single nucleotides polymorphisms (SNPs) are the most common genetic variations in the DNA sequence, involve a single base and have a frequency of greater than 1% in at least one minor allele population [18]. Certain SNPs have already been identified as potential predictors of efficacy and/or toxicity in advanced RCC patients treated with tyrosine kinase inhibitors [19][20][21][22][23][24][25][26].
The present study aims to analyse the incidence of SNPs in genes related with angiogenesis or metabolism of antiangiogenics in patients with localized and advanced RCC and to test their potential as prognostic and/or predictive factors.

RESULTS
One hundred and two patients were initially included in the study, 65% were male and the median age was 62 years (range 29-83 years). Three patients were excluded from the final analysis due to incomplete clinical information available. The median of follow-up was 62 months. Table 1 shows clinical characteristics for localized (a) and metastatic (b) patients and the association of these characteristics with disease/progression free survival (DFS and PFS) and overall survival (OS) (c).
One triallelic SNP (rs2032582) was excluded from the analysis due to inconsistent results with the array utilized. The minor allele frequencies (MAF) of the others 62 polymorphisms genotyped ( Table 2) were consistent with the data described elsewhere for European and Iberian population (1000 genomes, dbSNP database) and all SNPs were in Hardy-Weinberg equilibrium ( p > 0.05). Table 2 shows the characteristics for the 62 polymorphisms genotyped and frequency in our tumor samples in localized and metastatic patients.
Patients were classified in two cohorts for analysis purposes: localized and metastatic. A number of SNPs showed either a protective or adverse effect (Table 3A). Thus, in patients with localized tumors, one polymorphism, rs2071559 in VEGFR2 gene was associated with a protective effect: the mean of patients with this SNP presented a DFS of 49 month vs. 19 months when the SNP was absent. Another two, rs2228230 and rs10013228 in two genes (PDGFRA and VEGFR2) were significantly associated with worse DFS in the multivariate analysis. Accordingly, the absence of rs2228230 associated with an increased DFS (43 months) compared with 25 months in those patients harbouring the SNP. For rs10013228 the deleterious effect in DFS was even of a larger magnitude (62 months vs. 31 months). Additionally, rs10013228 was also significantly associated with a shorter OS (136 vs. 120 months). Other two SNPs (rs307826 and rs6877011) in VEGFR3 were also confirmed as predictors of shorter OS (127 vs. 96 months and 139 vs. 30 months, respectively).

DISCUSSION
Despite major advances in the knowledge of the molecular basis and therapeutics of RCC, prognostic and predictive estimation remains largely based on clinical and blood test parameters. This is an exploratory pharmacogenetic study designed to identify SNPs that could contribute to select patients with better prognosis and /or higher chances of benefiting from systemic treatment. We studied 62 polymorphisms from 18 genes in 99 patients on the basis of allele frequency and functionality evidence. Our study showed that the presence of certain SPNs was statistically associated with the progression of the disease, the response to treatment and the overall survival in this RCC patient population.
In patients with localized disease, the SNPs that had clinical significance were those positioned in receptors of VEGF and PDGF such as VEGFR2, VEGFR3 or PDGFR. SNPs located in these genes could potentially influence the activation of their cognate signaling pathways, which is a well-established mechanism of RCC tumorigenesis. We found that patients wild type for rs10013228 have a better DFS and OS. No studies in European populations or in RCC patients have been found in this regard. To our knowledge, the only reference in the literature of this SNP comes from a Chinese cohort of localized colorectal cancer patients where it had shown a protective effect [27]. Rs2071559 is a promoter SNP associated with VEGFR2 transcription activity [28]. In our study the AA genotype was associated with a protective effect increasing the DFS. These results are in concordance with data from other reported studies. In a recent metastatic RCC analysis [28], this polymorphism was shown to predict for sorafenib (an anti-VEGFR) efficacy. Promising results have been also described in metastatic colorectal cancer where this VEGFR2 polymorphism was significantly associated with increased PFS and OS in multivariate analysis in metastatic patients treated with first-line oxaliplatin-based chemotherapy regardless the KRAS mutational status [29]. Likewise a study in patients with localized colorectal cancer suggested a protective role for rs2071559, especially in patients that had received chemotherapy [27]. Data from other tumor types also pointed in a similar direction. An analysis in hepatocellular carcinoma patients treated with sorafenib showed that the presence of rs2071559 was a predictor of better outcomes [30]. SNPs in VEGFR3 were also associated with treatment outcomes. Thus, the absence of the SNPs rs307826 and rs6877011 were predictors of better outcome. This is consistent with other reports in RCC patients treated with the anti-VEGFR sunitinib where the presence of the genetic variant rs307826 or rs6877011 was associated with a shorter DFS [19] and OS [31].
Our study also found that SNPs in the PDGFRA gene such as rs2228230 significantly associated with worse prognosis. No previous reports exist about this SNP in RCC. Its presence has been reported in rare cancers such as extraintestinal stromal tumors and cervical adeno-squamous carcinoma, nevertheless its prognostic or predictive role remains largely unexplored [32,33]. In our series we could not confirm a variation in response to different PDGFR-inhibitors such as sunitinib or sorafenib based on the presence of this SNP. The limited sample size when stratifying by treatment arms could explain these results.
Three SNPs were found relevant at predicting survival in advanced RCC patients. One of them in the Interleukin 8 (IL8) gene (rs2227543) is a 3 prime UTR variant, and therefore variations in these regions could significantly impact in the metabolism of the protein. IL-8 is a pro-inflammatory chemokine that execute an angiogenic function, thus, variations on this gene could influence tumor cell growth and angiogenesis. Only one report has associated this SNP with cancer, suggesting  a potential role of genetic variations in IL genes as predictors of shorter DFS and OS in colorectal tumors. [34]. Likewise in our series, the presence of this genetic variant was associated with shorter OS. The other two SNPs relevant in the advanced cohort (rs9800958 and rs2302273) are located in PRKAR1B, an oncogene related with cell growth and PDGFRB respectively. Both demonstrated a protective effect in our series with longer OS for the patients that harbour these variants. Rs9800958 is an intron variant of PRKAR1B and rs2302273 is located in the 5 prime UTR variant of PDGFRB and therefore could affect the gene product by altering the binding of the transcription factor [35]. However, no data about the precise role of these SNPs in cancer has been communicated yet.
When looking at prediction of response SNPs in the VEGFA gene resulted of interest. The polymorphism rs699947 predicted worse prognosis in our analysis. This variant has been evaluated in metastatic RCC by other groups with contradictory results. In some series appears as a positive prognostic factor [28,36,37] while others deny its prognostic or predictive value [19,38]. In the same gene, the presence of rs3025010 in our series was associated to worse prognosis. There are only two oncology reports about this SNP, one in non-small cell lung cancer [39] and other in hepatocellular carcinoma [40] but neither of them established any correlation between the SNP and the response rate.
On the other hand, the presence of rs594942 in VEGFB has been associated with better response in our series. We have found only one citation of this polymorphism in metastatic colorectal cancer but without significance in the study [41].
All these results show the variability on the interpretation of polymorphisms depending on the type of cancer or the populations where they are evaluated. Nevertheless, the present exploratory study identified a set of SNPs that could improve prognostic and predictive estimation in RCC patients. Yet, the study might have a number of limitations that need to be taken into account. First the treatment varied across patients, although the majority (86%) received tyrosine kinase inhibitors targeting VEGFR/PDGR. This fact could compromise the real predictive value of these genetic variants. Another limitation of the study is the multiple testing. In a relatively small cohort of patients, multiple SNPs (variables) are evaluated. Therefore, these results need to be cautiously interpreted and require further validation in larger series. Yet, the data here presented are hypothesis generating and could eventually help in optimizing patient selection in cancer therapeutics and improve prognostic estimation through genetic characterization. Upstream gene: the sequence variant is located in the 5′ position of the gene. Downstream gene: the sequence variant is located in the 3′ position of the gene. www.impactjournals.com/oncotarget

Selection and characteristics of patients
Patients with localized and advanced RCC treated in the University Hospital "Virgen del Rocío" in the period 2000-2013 were included in the study. Paraffin embedded tumor samples were collected and patients were divided in two cohorts: those with localized disease and those with advanced RCC. The study protocol was approved by the Ethic Committee of Biomedical Investigation of Andalucía and conducted according to the principles of the Declaration of Helsinki.
The following inclusion criteria were considered: histologically confirmed diagnosis of primary RCC, complete clinical information and adequate tissue available (60%-75%). As clinical data the following were included: sex, age, date of diagnosis, TNM stage, histological type, tumor differentiation (Furhman grade), surgical treatment (partial or complete nephrectomy), systemic treatment (tyrosine kinase inhibitor (TKI) or mammalian target of rapamycin (mTOR) inhibitor, grade 3 or 4 toxicities, date of last visit or death and cause of death. All patients were treated following clinical guidelines and scientific evidence. Objective response was classified according to RECIST 1.1 as complete response (CR), partial response (PR), stable disease (SD), or progression of disease (PD).
The selection of the SNPs to be analyzed was not systematic. Given the particular biology of RCC and the drugs utilized for treatment of this cancer we first selected genes involved in angiogenesis and also those related to the mechanism of action of tyrosine kinase inhibitors targeting VEGFR/PDGFR. Additionally we took into consideration previous studies, allele frequency in European and Iberian population (reference 100 Genomes Project), Hardy-Weinberg Equilibrium (Genotype frequencies are determined by allele frequencies at that locus) and linkage disequilibrium between SNPs determined by Haploview v4.2 software. This can be perceived as a limitation of the study [42]. Indeterminate results were coded as missing values for statistical analysis.

DNA isolation and quantification
Paraffin embedded samples from patients with RCC were obtained from surgical specimens from nephrectomy. For each sample of 10 µm, paraffin was removed and DNA was isolated with DNA kit QiAGEN protocol. DNA concentration was determined by Nanodrop (Thermo Scientific, DE, USA).

Statistical analysis
The primary objective in the localized tumors cohort was to correlate the presence of SNPs with a worse DFS and OS. DFS was defined as the time between the diagnosis and the date of a radiological progression or death and OS as the time between the diagnosis and the date of death or last date of follow-up.
In the metastatic patients cohort overall RR, PFS and OS were analyzed and correlated with the presence SNPs. We considered overall RR as the percentage of CR and PR. The PFS was defined as the interval between the first day on systemic treatment and the date of radiological PD or death. Overall survival was defined as the time between the first day on treatment and the date of death or last date of follow-up.
Descriptive statistics were used to define the most relevant clinical features. The chi-squared test or Fisher's exact test were used in order to know the most relevant clinical variables to be included in the multivariate analysis. For this purpose, the DFS, PFS and OS parameters and RR variables with p < 0.25 or those considered clinically relevant based on the previous literature on RCC were selected. These characteristics were: for patients with localized disease: type of nephrectomy (partial/complete), Furhman Grade (3-4), TNM stage and for patients with metastatic disease: Furhman Grade (3)(4), TNM stage prognosis group (favorable vs intermediate/poor), metastasis lung and/or bones, Karnofsky, hemoglobin, time between nephrectomy and systemic treatment. ECOG was not included because the low number of patients in localized disease. All SNPs were tested in a univariant analysis for association with DFS, PFS and OS using Kaplan-Meier statistics and in a multivariate analysis using Cox proportional hazards to know the association between the presence of each SNPs and survival adjusting for potential confounding factors. Patients who had not progressed at database closure were censored at last follow-up. Also a chi-squared and a logistic regression were used to compare the presence of the SNPs and worse RR and the association of grade 3-4 toxicity with the presence of certain SNPs. P < 0.05 was considered significant. All analyses were performed using the Statistical Package for the Social Sciences software (SPSS 20.0 for Windows; SPSS Inc, Chicago, IL).