Gene polymorphisms in the PI3K/AKT/mTOR signaling pathway contribute to prostate cancer susceptibility in Chinese men

In this hospital-based case-control study of 413 prostate cancer (PCa) cases and 807 cancer-free controls, we investigated the role of functional single nucleotide polymorphisms (SNPs) of pivotal genes in the PI3K/AKT/mTOR pathway. We genotyped 17 SNPs in mTOR, Raptor, AKT1, AKT2, PTEN, and K-ras and found that 4 were associated with PCa susceptibility. Among the variants, the homozygote variant CC genotype of mTOR rs17036508 C>T were associated with higher PCa risk than the wild TT genotypes (adjusted OR = 3.73 (95% CI = 1.75-7.94), P = 0.001). The GT genotype of mTOR rs2295080 G>T was more protective than the TT genotypes (adjusted OR=0.54 (95% CI=0.32-0.91), P=0.020). The distributions of Raptor rs1468033 A>G genotypes differed between cases and controls, especially in subgroups defined by age, BMI, smoking status, and ethnicity. The CT/CC genotypes of AKT2 rs7250897 C>T were associated with an increased risk of PCa, particularly in subgroups of age >71 and BMI >24 kg/m2. These findings suggest that SNPs in the PI3K/AKT/mTOR pathway may contribute to the risk of PCa in Chinese men.


INTRODUCTION
Prostate cancer (PCa) is the leading malignancy in developed nations and ranks second in cancer deaths worldwide [1,2]. Its incidence has increased rapidly in China [3,4], with an estimated incidence of 1 in 10000 in 2010 [5]. Although environmental and genetic factors are associated with PCa carcinogenesis, the actual causes are unknown. Published genome-wide association studies (GWAS) identified 40 single nucleotide polymorphisms (SNPs) that were associated with human PCa risk [6]. Lu et al. demonstrated that SNPs associated with PCa risk were enriched in the androgen receptor (AR)-binding sites [7,8]. These findings led to further functional studies to better understand PCa susceptibility.
The PI3K/AKT/mTOR signaling pathway is involved in many human malignancies, including PCa [9][10][11]. Nearly 25-70% of PCa cases show altered PI3K/AKT/ mTOR signaling with higher prevalence in metastatic tumors. AKT is a proto-oncogene that is phosphorylated by a variety of extracellular stimuli and regulates multiple cellular processes involved in cell survival, growth, differentiation and proliferation. Genomic mutations of PTEN are reported for nearly 50% of primary PCa, especially in advanced disease [12]. Although AR is not a known physiological substrate of AKT, two AKT phosphorylation consensus sequences were identified in AR [13,14]. It was postulated that some AKT substrate, eukaryotic initiation factor 4E-binding protein-1 (4EBP1) and ribosomal protein S6 kinase (P70) were associated with AR synthesis and resulted in PCa. Given these www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 37), pp: 61305-61317 Research Paper findings, the role of PI3K/AKT/mTOR pathway in PCa carcinogenesis needs to be established.
Previous pre-GWAS studies demonstrated that genetic variants in PI3K/AKT/mTOR pathway were associated with cancer risk, including PCa [15,16]. However, they needed to be experimentally verified [7,8]. In the present study, we analyzed if 17 SNPs in six pivotal genes (K-ras, PTEN, AKT1, AKT2, mTOR, and Raptor) of the mTOR pathway were associated with PCa susceptibility.

Characteristics of the subjects
Overall, there were no differences in distributions of age, smoking status, and BMI index between the 413 cases and 807 cancer-free controls (

Association of high-order interactions with PCa risk
We performed the multifactor dimensionality reduction (MDR) analysis by including the genotypes of four significant genetic factors (mTOR rs17036508 CC and rs2295080 TT, Raptor rs1468033 AG/AA, and AKT2 rs7250897 TT/CT) and four environmental risk factors (age at diagnosis, smoking status, race, and BMI). Among all 8 factors, the Raptor rs1468033 A>G variant was the best one-factor model. Likewise, interactions between mTOR rs17036508 CC, rs2295080 TT and Raptor rs1468033 AG/AA represented the best threefactor model. Age was the most promising environmental risk factor associated with the four genetic factors. The details are presented in Table 5. Subsequent hierarchical cluster analysis placed BMI and smoking status, race and rs17036508 C>T, rs2295080 G>T and rs1468033 A>G, age and rs7250897 C>T on the same branch. This suggested that interactions in this eight-locus model may modulate PCa risk (data not shown).
Finally, we analyzed the false-positive report probability (FPRP) values at variant prior probability levels for all positive findings (Table 6). Some higher statistical power (81.7%-89.6%) was observed under the assumption of prior probability of 0.25. However, some significant findings were still noteworthy at prior probability of 0.01 in spite of the lower statistical power (range from 5% to 45%). Some findings with greater FPRP values need further validation in larger studies.

DISCUSSION
In the current single institution based case-control study, functional SNPs of six pivotal genes of PI3K/AKT/ mTOR pathway were associated with PCa risk. Briefly, the mTOR rs17036508 CC, mTOR rs2295080 GG/GT, Raptor rs1468033 AG/AA, and AKT rs7250897 TT/CT genotypes were associated with PCa risk, especially in age, BMI, smoker-status, and ethnic subgroups. To the best of our knowledge, this is the first post-GWAS study analyzing associations of these six pivotal genes of PI3K/ AKT/mTOR pathway with PCa risk.
The mTOR gene is a critical cellular protein with more than 2651 SNPs reported across the whole region [17]. However, only few SNPs have been associated with PCa susceptibility. Two studies showed that mTOR rs2295080 GT/GG genotypes protected against PCa risk in Han Chinese populations [18,19]. These findings were consistent with mixed Chinese populations in the present study, but not in separate ethnic subgroups (Han or Uygur). The effects of the mTOR rs2295080 GT/GG genotypes were also observed in esophageal squamous cell carcinoma, gastric carcinoma, and renal cell carcinoma [20][21][22]. In vitro and in vivo studies suggested that mTOR rs2295080 T allele probably increased the affinity of special transcription factors to this promoter region and contributed to enhanced mTOR activity [20]. Variant mTOR rs2295080 was linked to eight potential functional variants of mTOR with higher linkage disequilibrium Two-sided chi-square tests were used to calculate differences in the frequency distribution of genotypes between cases and controls.  [24], and esophageal squamous cell carcinoma [25]. Similarly, as a probable transcription factor binding site of mTOR, variant rs1883965 was associated with esophageal carcinoma, gastric cancer, and hepatocellular carcinoma [26]. Additionally, it was predicted that variant rs17036508 of mTOR was located within a miRNA binding site and an exonic splicing enhancer or silencer motif, thereby affecting pre-RNA splicing. The association between rs17036508 and PCa was also observed previously [19]. In the present study, we found that homozygote variant carriers of rs17036508 were more likely to develop cancer compared to homozygote wild carriers. In theory, polymorphism rs17036508 located in 3'-UTR of angiopoietin-like 7 gene (ANGPTL7), resulting in upregulation of ANGPTL7 by hypoxia in   cancer cells, which exerts a pro-angiogenesis effect [27]. Taken together, it is biologically plausible that these two polymorphisms might mediate tumor formation by regulating the expression of mTOR and ANGPTL7 simultaneously.
The Raptor gene, located in 17q25.3 with 34 exons, regulates responses to nutrient and insulin levels [28]. The Raptor protein forms a stoichiometric complex with the mTOR kinase, and is associated with 4EBP1and P70. Previous in vitro and in vivo studies have shown that Raptor acts as a scaffold protein that regulates mTORdependent signaling [29]. One of the mechanisms involves changes in phosphorylation status of Raptor. However, the mechanism through which Raptor regulates these processes is only beginning to be established. In a recent study, Raptor gene polymorphisms were associated with increased risk for bladder cancer; physical activity, energy balance and genetic variants in the mTOR pathway cocoordinately influenced bladder cancer risk [30]. In this study, variant Raptor rs1468033 was associated with PCa risk, particularly in subgroups of age, BMI, and smoking status, similar to previous findings. Silico analysis indicated that rs1468033 was found with several potential functional polymorphisms of Raptor gene, including rs2292639, rs499609, rs6565500, and rs9899178. Among these, rs2292639, rs4999609, and rs6565500 were predicted as transfactor binding sites, whereas rs9899178 was predicted as splicing site. It is plausible that these variants modulate the expression of Raptor gene, and affect mTOR activity.
Although mTOR activity was previously implicated in promoting PCa cell invasion, the role of AKT2 was not known. The different AKT isoforms (AKT1, AKT2, and AKT3) play contradictory regulatory roles; for example, AKT1 activation inhibits cell migration whereas AKT2 promotes it [31,32]. Many studies have indicated that AKT2 is a more promising therapeutic target than PI3K due to its involvement in normal cellular processes. AKT2 downregulated GSK3b, which modulates cell migration [33]. Additionally, AKT negatively regulates Mdm2 expression during abnormal stress and promotes tumorigenesis [34]. Recently, Chen et al. reported that AKT2 rs7254617 increased prostate cancer risk [18]. Also, a Caucasian study showed that AKT2 variant rs3730050 was associated with poor prognosis of bladder cancer patients. Therefore, AKT isoforms are cancer susceptibility genes. In this study, AKT2 rs7250897 was associated with increased PCa risk by a dominant genetic model in subgroups of age >71and BMI >24kg/m 2 , indicating that rs7250897 altered AKT2 expression, which subsequently affected synthesis of adipose-related proteins. Since the three variants were far from each other, we postulated that they were associated with carcinogenesis by different mechanisms. These findings need to be further validated.
There were few limitations in the present study that need to be addressed. First, the limited sample size in our study may have decreased detection of weaker genetic effects in carcinogenesis. Second, information regarding predisposition to PCa was not collected for the analysis and might confuse the stratified positive associations. Third, the case-control study may have inherent selection and information biases. Because of insufficient medical records, we could not perform correlative analyses between stages of PCa and variants of AKT pathway that may have provided more information in PCa carcinogenesis. Moreover, further in vitro and in vivo experiments are necessary to unravel molecular mechanisms for the genetic associations that we have postulated in this study.  In summary, we showed that variants of PI3K/AKT/ mTOR signal pathway genes may associate with PCa risk. The combined genotypes of these variants enhanced PCa risk with increasing numbers of putative high-risk genotypes. Moreover, the three-factor model (rs17036508, rs2295080, rs1468033) was the best model to predict PCa risk. In conclusion, our study postulated that the genetic variants may alter the expression and activity of mTOR leading to PCa susceptibility.

Patients and controls
We recruited 413 newly diagnosed PCa cases and 807 frequency-matched cancer-free controls from genetically unrelated Chinese Han and Uygur participants in Xinjiang province between January 2003 and January 2015. The cases were histopathologically confirmed as primary prostate adenocarcinoma at the First Affiliated Hospital of Xinjiang medical University (XJMU). Pathological grades of the PCa were determined by Gleason scores from the radical prostatectomy specimens according to the latest WHO criteria [35]. The recruited healthy controls were extracted from males who had health check-up in the First Affiliated Hospital of XJMU during the same period. Individuals with serum PSA >4ng/ ml were excluded from the control group.
All subjects were interviewed with a written informed consent. Response rate was 92% and 90% for cases and controls, respectively. The experimental and research protocols were approved by the Institutional Review Board of XJMU.

Selection of single nucleotide polymorphisms
For the six pivotal genes in PI3K/AKT/mTOR pathway, selection strategy for the potentially functional SNPs was based on the NCBI dbSNP database (http:// www.ncbi.nlm.nih.gov/projects/SNP) and SNPinfo web server (http://snpinfo.niehs.nih.gov/snpfunc.htm). The criteria were: 1) minor allele frequency (MAF) was at least 5% in Chinese populations; 2) SNPs were potentially functional according to SNPinfo prediction platform. Ultimately, 17 potentially functional variants were selected involving mTOR (rs1034528 C>G, rs1703658 C>T, rs12122605 T>C, and rs2295080 G>T), Raptor (rs2271610 C>G, rs2271612 T>C, rs2292639 A>C, rs3751932 T>C, rs3751934 A>C), AKT1 (rs2494750 C>G, rs2494752 G>A), AKT2 (rs2304186 T>G, rs7250897 C>T, rs7254617 A>G), PTEN (rs701848 C>T), and K-ras (rs7312175 A>G) based on the bioinformatics analysis performed with HaploView software 4.2. All these SNPs were genotyped by the TaqMan real-time PCR method as described previously [36]. In this study, the quality control strategy was established as follows: (1) the discrepancy rate in all positive controls was less than 0.1%; (2) the results with >95% call rates and 100% concordance for duplicated specimens were favorable for further analysis.

Statistical analysis
We performed the Pearson's χ 2 -test for the differences in selected variables between cases and controls. Crude and adjusted ORs and their 95% CIs were computed from both univariate and multivariate unconditional logistic regression models. We further evaluated the stratified associations based on the significant genetic models accompanied by the homogeneity Q-tests for the strata. For all the significant findings, we calculated FPRP with the assumption of different prior probabilities to detect any possible false positive associations. Only FPRP values<0.2 were considered noteworthy [37]. All statistical analyses were performed with SAS 9.1 statistical software (SAS, Cary, NC, USA) unless stated otherwise. All P values were two-sided with a significance level of P<0.05.
The multifactor dimensionality reduction (MDR) analysis was conducted by the MDR V2.0 beta 8.2 software (http://www.multifactordimensionalityreduction. org/) to identify the best n-factor interaction model [38]. We further performed the interaction dendrograms and graphs for risk loci of this study, and the color of branches and lines referred to the type of interaction, with green-toyellow-to-red indicating weak-to-strong interactions.

Author contributions
QL conceived and designed the experiments; TL, AG and FW performed the experiments; HA and XS analyzed the data; WC contributed reagents, materials and analysis tools; TL and QL wrote the paper.