A Functional Polymorphism (rs937283) in the MDM2 Promoter Region is Associated with Poor Prognosis of Retinoblastoma in Chinese Han Population

The effect of single nucleotide polymorphisms (SNPs) at MDM2 has been investigated in several cancer types. Three MDM2 SNPs(rs937283, rs2270744 and rs769412) have previously been suggested to be positively correlated with cancer. In this study, we aimed to explore the association of rs937283, rs2270744 and rs769412 polymorphisms with retinoblastoma (RB) risk, clinicopathological characteristics, and prognosis. Compared with wild-type genotype AA at rs937283, individuals carrying AG and GG genotype had a significantly increased risk for developing RB (OR = 1.86, 95% CI 1.13–3.08; OR = 2.48, 95% CI 1.10–5.62, respectively). RB patients with allele G at rs937283 were more susceptible to invasion and high tumor aggression (OR = 2.42, 95% CI 1.43–4.11; OR = 2.15, 95% CI 1.27–3.64, respectively). Kaplan-Meier curves and log-rank results revealed that RB patients harboring genotype GG and G allele at rs937283 had worse survival (P < 0.02 and P < 0.01, respectively). In addition, the A to G substitution at rs937283 significantly enhanced the transcription activity of the MDM2 gene in vitro. In vivo, we found that MDM2 mRNA and protein were overexpressed in individuals who carried the G allele at rs937283. This study suggested that the MDM2 rs937283 polymorphism is a novel functional SNP both in vitro and in vivo as well as a biomarker for poor prognosis in RB.

transcriptionally activate MDM2 (murine double-minute 2 homology), which in turn could lead to increased proteolytic degradation of p53, thus MDM2 functions as a negative regulator of p53 pathway 10 . Furthermore, MDM2 could interact with pRB and bind to the activation domain of E2F1 transcription factor which could inhibit pRB regulatory function 11 . Thus, MDM2 was identified as a modifier gene in RB 12 . Recent data suggested that MDM2 gene polymorphisms may contribute to increased MDM2 basal expression and increase cancer susceptibility. Specifically, MDM2 rs2279744 (SNP T309G) polymorphism has been shown to produce a higher-affinity DNA-binding site for Sp1, which could increase MDM2 mRNA and protein expression, and thus increase the degradation of p53 and hinder p53-induced apoptosis 13 . Increased MDM2 expression is related to elevated cancer risk in sporadic and hereditary malignancies and increased likelihood of distant metastases 14 . Previous studies have found a positive association between MDM2 rs2279744 polymorphism and RB development 9,15 . In addition to rs2279744, recent studies have suggested the potential association of rs769412 and rs937283 in the MDM2 loci with cancer development 16,17 . rs769412 polymorphism (also known as G354A) is a A to G substitution located at the 354 nucleotide in the exon 12 of MDM2 gene, while rs937283 (also known as G2164A) polymorphism could lead to an A to G base change at the 2164 nucleotide in the promoter region of MDM2 gene 18,19 . Previous studies have reported rs769412 and rs937283 polymorphism were associated with the development of several cancer types 16,18 . Until now, there has been no report investigating the association of these two polymorphisms with RB development. To determine the role of MDM2 polymorphism in RB, we analyzed the distribution of rs937283, rs2270744 and rs769412, and assessed the association of these polymorphisms with clinicopathological characteristics and prognosis in Chinese RB patients. Further, we conducted the molecular work to confirm whether the variants could alter MDM2 expression.

Results
Participant characteristics. A study on 137 RB patients and a control group of 150 non-carriers of RB was undertaken. All participants were Han ethnicity from the same region in Shandong of China and there was no statistical difference in gender between RB patients and normal controls (P = 0.47). The clinical characteristics of RB patients were presented in Table 1.
Association of rs937283 with RB risk. As shown in Table 2, significant differences in the distribution of rs937283 were observed between RB patients and normal controls. The analysis did not yield a significant deviation from HWE in control group (P = 0.196). Compared with WT genotype AA at rs937283, individuals carrying genotype AG and GG had significantly increased risk for developing RB (OR = 1.86, 95% CI 1.13-3.08; OR = 2.48, 95% CI 1.10-5.62, respectively). Significantly increased risk for RB was also observed in subjects with genotype AG or GG under a recessive model (OR = 1.98, 95% CI 1. 23-3.17). In addition, individuals with G allele at rs937283 had an almost 74% increased risk of RB development when compared with A allele (OR = 1.74, 95% CI 1.21-2.51).
Association of rs2279744 with RB risk. The distribution and statistical analyses of rs2279744 genotype in RB patients and normal controls are also described in Table 2. The value for the χ 2 tests of HWE was 0.777 for controls. Similar to rs2279744 polymorphism, we did not observe a positive association between rs2279744 Linkage Disequilibrium(LD) analysis. Linkage disequilibrium of rs937283, rs2279744 and rs769412 were analyzed using Haploview 4.2 software (Fig. 1). The result revealed that these three SNPs were not in LD (r 2 = 0.20, r 2 = 0.11, r 2 = 0.06; respectively).
Association of rs937283, rs2279744 and rs769412 with clinicopathological characteristics in RB patients. Association of rs937283, rs2279744 and rs769412 with RB prognosis. Kaplan-Meier curves were constructed to evaluate the association of survival rate with rs937283, rs2279744 and rs769412 SNPs. Significant difference in survival rate was detected among patients with different genotypes at rs937283, but not rs769412 or rs2279744 (Fig. 2). Kaplan-Meier curves and log-rank results revealed that RB patients carrying genotype GG wat rs937283 had shorter survival time than those with genotype AA and AG alone ( Fig. 2a, P < 0.05, P < 0.02, respectively). Consistently, RB patients carrying G allele (AG + GG) at rs937283 also had worse survival (Fig. 2b).
In addition, the results of the multivariate analysis of survival time using the Cox proportional hazards model are presented in Table 4. Tumor invasion and lag-time were two independent risk factors for poor overall survival in RB patients (HR = 3.17, 95% CI 1.28-4.66; HR = 3.87, 95% CI 1.72-5.52). However, rs937283, but not rs769412 or rs2279744, was independently associated with overall survival. Genotype AG + GG carriers of rs937283 exhibited a significantly increased risk of poor overall survival versus GG genotype (HR = 2.01, 95% CI 1.04-3.95), suggesting that AG + GG genotype was an independent risk factor for overall survival in RB patients.
Effects of rs937283 polymorphism on transcriptional activity. To evaluate whether the promoter activity could be affected by rs937283 polymorphism, we constructed luciferase reporter vectors (pGL3) with either A or G allele and used them for transient transfections with Y79, WERI-Rb1 and HeLa cells as mentioned in method section. Compared to those with A allele, the vectors with rs937283 G allele yielded a 90% to 150% increase in the relative luciferase activities in all three types of cell lines (P < 0.05 for all, Fig. 3). These results suggested that rs937283 G allele indeed could increase transcriptional activity of the MDM2 gene in vitro. Association of rs937283 polymorphism with mRNA and protein expression levels of MDM2.
To assess the effect of rs937283 polymorphism on MDM2 expression, we further analyzed mRNA and protein expression levels of MDM2 in RB patients. The effect of rs937283 polymorphism on mRNA expression was evaluated by quantitative real time PCR (qRT-PCR). Similar to transcriptional activity, the results revealed that MDM2 mRNA levels were significantly higher in individuals who carrying AG and GG genotype than those with AA genotype(P < 0.05 for both, Fig. 4a). The differences in the mRNA level between individuals with the AG and GG genotype were also statistically significant(P < 0.05, Fig. 4a). In consistent with the mRNA expression level, the protein expression levels of RB patients carrying the AG or GG genotype were also significantly higher than that of those with AA genotype (Fig. 4b). Conversely, p53 protein expression was significantly lower in RB patients with AA genotype when compared with AG or GG genotype (Fig. 4b). Taken together, the observation of higher MDM2 expression and lower p53 expression in RB patients with GG genotype than those with other genotypes suggested GG genotype at rs937283 as a risk factor for RB.  Gender   Female  37  31  10  105  51  26  32  20  84  72  56  18  4  130  26   Male  23  28  8  74  44  15  27  17  57  61  42  12  5  96  22 OR (95%) 1.0 (− )  Discussion RB has been the archetypal cancer model that follows the two-hit hypothesis for the initiation of tumor 8 . Although inactivation of the RB1 tumor suppressor gene seems sufficient for the onset of this tumor, the development of RB is potentially modified by the presence of numerous additional genetic mutations in RB patients 8,20 . In the present study, we evaluated the association between rs769412, rs937283 and rs2279744 polymorphisms at MDM2 on RB risk and prognosis. Allele G and genotype GG at rs937283 increased the risk of RB development and were associated with invasion, high tumor aggression. AG + GG at rs937283 was associated with poor prognosis, and  was also identified as an independent risk factor for RB prognosis. In addition, our results also showed that the A to G substitution at rs937283 significantly enhanced the transcription activity of the MDM2 gene in vitro. Furthermore, we found that MDM2 mRNA and protein were overexpressed in vivo in individuals who carried the G allele, suggesting that the MDM2 rs937283 polymorphism is indeed a functional SNP both in vitro and in vivo as well as a biomarker for risk and prognosis of RB. During DNA damage response, TP53 prevents cell proliferation through several mechanisms such as cell cycle arrest and apoptosis. Abnormality in the p53 pathway can yield negative effects on the homeostasis of normal cells; thus, MDM2 can attenuate excess p53 effect as a principal regulator of p53. Specifically, MDM2 can exert its function as an E3 ligase and interact with the terminal transaction domain of TP53 21 . The production of MDM2 was activated by the cytoplasmic presence of p53 through a negative feedback loop, and as such, both proteins were intricately dependent on each other for proliferation and growth in normal conditions 22 . However, previous in vitro studies reported that MDM2 showed p53-independent oncogenic properties that could regulate proliferation, apoptosis, tumor invasion, and metastasis [23][24][25] . MDM2 protein levels and function were precisely controlled at the transcriptional, translational, and post-translational levels 26,27 . Therefore, various SNPs occurring in the MDM2 gene could potentially dysregulate its production.

A/A A/G G/G A G G/G G/T T/T G T A/A A/G G/G A G
Until now, the most widely studied MDM2 polymorphism was rs2279744, which is a T > G nucleotide change located in the promoter region of MDM2. rs2279744 can increase the binding affinity of the transcriptional activator SP1 for the MDM2 promoter 13 . This higher affinity potentially leads to increased production of MDM2 mRNA and protein. The increased levels of MDM2, coupled with normal production of p53, may result in lower-than-normal cytoplasmic levels of p53. This in turn can augment DNA damage and hasten cellular transformation 28 . Due to the potential effect of rs2279744 polymorphism on MDM2 production, previous studies have described the role of rs2279744 polymorphism in different neoplasms including lung cancer, cervical cancer, melanoma, and breast cancer [29][30][31] . However, these studies yielded inconsistent results. Ryan B et al. reported that rs2279744 was associated with an increased risk of lung cancer, however, Roszak A et al. did not observe any association of rs2279744 with cervical cancer development and clinicopathological features 29,30 . Recently, one study conducted by Chen et al. also did not observe the significant association of rs2279744 polymorphism with RB risk in a Chinese population 9 . In consistent to Chen's result, we also could not identify a positive association  in Chinese RB patients in the present study. One possibility for this discrepancy was ethnicity. For example, MDM2 rs2279744 polymorphism was not a risk factor for cervical cancer in northeastern Brazilian, Caucasian, or African-American ethnicities, but a stratification-based ethnicity study revealed that the rs2279744 was a significant risk factor for cervical cancer in an Asian population [32][33][34] . Admittedly, different cancer types may contribute to the difference in association of rs2279744 and tumor risk. Another highly studied MDM2 polymorphism is rs117039649, a G > C nucleotide mutation, which has been reported as a potential antagonist to rs2279744. rs117039649 may override the effect of rs2279744 on SP1-mediated transcription and lead to an overall decrease in MDM2 protein production in the presence of both variant alleles 35 . However, a previous study reported that the rs117039649 SNP in the MDM2 promoter region was not present in Asian populations but was identified in a Caucasian population with an allele frequency of approximately 8% 13 . Herein, we also focused on two additional polymorphisms (rs769412 and rs937283) at MDM2 gene and investigated their association with RB risk.
SNPs rs769412 and rs937283 are two novel polymorphisms at MDM2 gene that have been identified in Chinese populations. Recently, Wang et al. reported the frequency of MDM2 rs937283 G variant allele was almost 5.8% in a control population, while the GG genotype and G allele were found to have positive correlations with the risk of larynx carcinoma in Chinese Han population 16 . Consistent with that study, our present results revealed the GG genotype and G allele could also increase the risk of RB development with the minor frequency of approximately 7.3%. In Wang's study, the authors did not discuss the association of this polymorphism with tumor features or tumor prognosis, which we address in our present study. Our result showed positive correlations between rs937283 G allele and tumor invasion and tumor aggression. From the survival analysis, we observed evidence for an independent association between rs937283 G allele with survival, wherein the G allele was associated with an increased risk of poor survival. In tumor, overexpression of MDM2 could substitute for inactivation of p53 in the absence of p53 mutations 36 . Accumulated evidence supported that elevated expression of MDM2 could cause tumorigenesis and was often associated with adverse clinical behaviors of tumors such as invasion, poor outcome, metastasis 37 . Our results proposed that the G > A mutation on rs937283 have ability of increasing MDM2 production and decreasing p53 level. G allele at rs937283 could significantly enhance the transcription activity of MDM2 gene in vitro. In vivo, both MDM2 mRNA and protein levels were elevated in RB patients who carried G allele.
It remains controversial concerning the role of rs769412 in tumor development. In Wang's study, the authors also investigated the association of rs769412 with larynx carcinoma risk 16 . Although no significant association was observed, their results specifically revealed that rs769412 was associated with the significantly increased risk of larynx carcinoma among drinkers. Pine S et al. reported no association between rs769412 polymorphism and lung cancer risk 38 , but Rajaraman P et al. found rs769412 polymorphism was associated with significantly reduced risk of glioma 39 . In our study, we also did not find a positive association between rs769412 and RB risk. Although we could observe a positive association of rs769412 polymorphism with RB invasion, a small number of individual carrying homozygous GG may limit the accuracy and reliability of this association.
Apparently, there were several limitations in the present study. First, the sample size of the present study was relatively small, which could limit the accuracy and reliability of results. Especially, the specific role of rs769412 in RB required to be further confirmed in a large cohort study due to the low minor allele frequency of rs769412 in Chinese population. Secondly, although we found rs937283 mutation could enhance transcriptional activity of MDM2 gene, the specific mechanism was not clear. One possibility we speculate may be that the A to G substitution of this polymorphism could affect the binding affinity of some transcriptional activator, which will be addressed in future functional studies. Thirdly, we did not evaluated the relationship of these two polymorphisms with other functional MDM2 SNPs such as rs2279744, although previous studies did not observe the differential frequency of rs2279744 polymorphism in RB patients 9 . Forth, the function or power of these two gene variants as a major predictor factor of RB development or RB prognosis in clinical practice was relatively low. Besides, we only conducted this study in Chinese Han population, the frequency of these two polymorphisms in other ethnic groups required to be confirmed. Thus, this study should be replicated in larger independent cohorts of different ethnicities.
In conclusion, our genetic assessment is the first study to evaluate the association of rs769412, rs937283 and rs2279744 with RB. The results showed that rs937283 were associated with the increased risk of RB development. The G allele at rs937283 could increase MDM2 production through enhancing transcriptional activity. In addition, G allele at rs937283 could reflect poor prognosis of RB patients, and may be an independent factor for predicting poor prognosis for RB. MDM2 polymorphisms analysis. Genomic DNA was extracted from blood samples using a Genomic DNA kit (Axygen, CA, USA) according to the manufacturer's protocol and genomic DNA samples were stored at − 20 °C until use. Three validated SNPs rs937283, rs2279744 and rs769412 in the MDM2 gene that were previously found to be associated with cancer development were analyzed. Genotyping for these three SNPs was performed using custom TaqMan ® SNP Genotyping Assays. Genotyping or allele analysis was carried out with the ABI Prism 7900HT genetic detection system with Taqman ® Genotyping Master Mix (ThermoFisher, OK, USA).

Methods
The final volume of the PCR system was 25 μ L including 12.50 ul of Master Mix, 1.25 ul of Assay Mix, 11.25 ul of ddH20, the PCR conditions were as follows: 95 °C for 15 s, 60 °C for 1 min, for 40 cycles. The allelic discrimination plots by automatic allele analysis were shown in Supplementary Fig. S1.
Construction of reporter plasmids. rs937283 polymorphism was located in the promoter of MDM2 gene, we then determined whether this polymorphism had an effect on gene expression in vitro. The MDM2 promoter-luciferase reporter plasmids containing either rs937283 A or G sequence were prepared by amplifying the 300-bp promoter region (from 2010 to 2309) by using primers The primers were 5′ -CGGGGTACCATGCTAGTACTGCTACCAG-3′ and 5′ -CCGCTCGAGATTCAGTAGCTGTCCTGAC-3′ . Two restriction sites including KpnI and XhoI were at 5′ -end of forward and reverse primers, respectively. To confirm the matched nucleotides and the plasmid containing either rs937283 A or G allele, the amplified fragments were sequenced. Both the amplified fragments and pGL3-basic vector (Promega, CA, USA) were digested by using the KpnI and XhoI enzymes (NEB, BioLabs lnc, USA). After digestion, the amplified fragments were cloned into pGL3-basic vector and the vectors were sequenced to confirm that there were no errors in the orientation and integrity of each construct.
pGL3 luciferase reporter containing rs937283 A or G allele was transfected into cells with the PolyJet DNA In Vitro Tranfection Reagent (Signagen Laboratories, MD, USA) according to manufacturer's instructions. The pGL3-basic vector without an insert was used as a negative control. Renilla plasmid (pRL-SV40) was co-transfected as an internal control. Luciferase assay was performed by using the Dual Glo Luciferase System (Promega, CA, USA). All luciferase activity was normalized against the activity of Renilla luciferase gene. Independent triplicate wells were done for each plasmid.
Protein isolation and western blotting. All 137 RB patients could divided into three groups according to the genotype at rs937283. We randomly selected 8 RB tumor tissues from each groups and then isolated protein. Frozen tumor tissues from RB patients were homogenized on ice and lysed in RIPA buffer containing protease inhibitor (cOmplete, Sigma, CA, USA) and PMSF. The homogenate were sonicated and centrifuged at 12,000 rpm at 4 °C for 5 min to remove cell debris. A BCA assay kit (Beoytime Biotech, Shanghai, China) was used to determine protein concentration and extracted proteins were separated by 8% SDS-PAGE, then separated proteins were transferred onto polyvinylidene difluoride membranes (Millipore, MA, USA). The membrane was blocked with 5% skim milk in TBST (TBS buffer with 0.1% Tween-20), and then incubated with human MDM2 antibody (1:1000 dilution, sc-5304, Mouse Monoclonal IgG, Santa Cruz, USA), p53 antibody (1:1000 dilution, ab28, Mouse Monoclonal IgG, Abcam USA) and GAPDH antibody (1:1000 dilution, sc-32233; Mouse Monoclonal IgG, Santa Cruz, USA) at room temperature for 2 h. Subsequently, horseradish peroxidase(HRP) -conjugated anti-mouse IgG were used as secondary antibody (1:5000 Dilution, Beoytime Biotech, Shanghai, China). Signals were captured by a CCD camera image system (Bio-Rad, CA, USA) with an HRP Chemiluminescent kit (Beoytime Biotech, Shanghai, China).

Statistical analysis.
Continuous data were compared between RB patients and normal controls using Student's t test, while categorical data were compared using Chi-square tests. Association between the genotypes and RB risk were estimated by odds ratios (ORs) using an unconditional logistic regression model. Survival probabilities were estimated by using Kaplan-Meier analysis, and significant differences were analyzed by using the log-rank test. Cox proportional hazards models were used to analyze the associations between genotypes with RB survival. Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using multivariable models. The deviation from Hardy-Weinberg equilibrium was assessed by using χ 2 tests. Differences in the expression levels of MDM2 among different genotypes were determined by Student's t test. All statistical analyses were performed in SPSS 22.0 (SPSS Inc., Chicago, IL) and GraphPad Prism 6.0 (CA, USA). P for the two-tailed test less than 0.05 was considered statistically significant.