Association of the MDM2 SNP285 and SNP309 Genetic Variants with the Risk, Age at Onset and Prognosis of Breast Cancer in Central European Women: A Hospital-Based Case-Control Study

SNP309T>G (rs2279744) and SNP285G>C (rs117039649) in the MDM2 promoter are thought to have opposite effects on the binding of transcription factor SP1 (specificity protein 1), and consequently on MDM2 expression, p53 levels, cancer risk, age at onset, and prognosis. Here, we genotyped SNP309 and SNP285 in 406 Austrian breast cancer patients and 254 female controls. The SNP309GG genotype was associated with an increased breast cancer risk in p53 negative (OR, 1.82; 95% CI, 1.09–3.03; p = 0.02), but not p53 positive or unselected patients. In contrast, the SNP309TT genotype was associated with an earlier age at onset (TT, 57.0 ± 12.9; TG, 58.6 ± 13.9; GG, 59.7 ± 15.0 years; p = 0.048). 31% of SNP309TT, 26% of TG, and 13% of GG tumors were p53 positive (p = 0.034), indicating a lower selective pressure to mutate TP53 in the presence of the G-allele. Moreover, SNP309TT patients exhibited a shortened metastasis-free survival in multivariable analysis. Censoring carriers of the SNP285C-allele hardly altered the strength of these associations of SNP309, thus challenging the proposed antagonistic function of SNP285C towards SNP309G. The minor SNP285C-allele tended to be non-significantly associated with an increased breast cancer risk and a poor disease-free and metastasis-free survival, which may be bystander effects of its complete linkage disequilibrium with SNP309G. We conclude that the SNP309G-allele attenuates the p53-response and leads to a higher breast cancer risk, but also to a later onset of breast cancer and a trend towards a good prognosis.


Introduction
The tumor suppressor gene TP53 is the most commonly mutated gene in human cancer [1]. In breast cancer, mutation of TP53 occurs at a frequency of 25-35% and is often associated with an earlier age at onset, a poor prognosis and a triple-negative molecular subtype [2][3][4][5]. p53 is a transcription factor and maintains genomic integrity primarily by regulating downstream target genes to induce apoptosis, cell cycle arrest, senescence, or other stress responses to halt the progression of cancer [6,7]. In addition to mutation, decreasing the level of wildtype TP53 is also a major mechanism of functional inactivation of p53 [8]. MDM2 (Mouse Double Minute 2 homolog) is a key negative regulator of p53, being both a target gene of p53 and an E3 ubiquitin ligase that targets p53 for proteasomal degradation, which maintains low p53 protein levels in the absence of stress [8][9][10]. Mdm2 knockout in mice leads to early embryonic lethality, whereas mice with a deletion of both Mdm2 and

Distribution of MDM2 SNP285 and SNP309 Genotypes
The promoter SNPs rs117039649 (SNP285G>C) and rs2279744 (SNP309T>G) in intron 1 of the MDM2 gene (hereafter referred to as SNP285 and SNP309, respectively) were genotyped in an Austrian hospital-based case-control study of 406 breast cancer patients and 254 control subjects. Tables S1 and S2 show the clinical and histopathological characteristics as well as the frequency of the SNP285 and SNP309 genotypes within the study population. The control population was in Hardy-Weinberg equilibrium for both SNP285 (p = 0.34) and SNP309 (p = 0.39). Likewise, the patient population was in Hardy-Weinberg equilibrium for both SNPs (p = 0.76 for SNP285; p = 0.32 for SNP309). The frequency of the minor G-allele of SNP309 (MAF) was 0.365 in patients and 0.347 in controls, close to the 0.367 MAF reported by the 1000 Genomes Project population database [31]. SNP285 genotype GC was observed in 12/254 controls (4.7%; MAF = 0.024), and 23/406 patients (5.7%; MAF = 0.028), all others exhibited genotype GG (Table S1). This frequency of the GC genotype in Austria is somewhat lower than the >7% frequency observed in Northwestern European populations from Norway, UK, and the Netherlands, and closer to the frequencies reported for Southern and Eastern European populations [20,22,25,31]. The SNP285C-allele was observed only in carriers of the SNP309G-allele (genotype GG or TG), consistent with previous reports [20,22,25]. Accordingly, SNP285 and SNP309 were in complete linkage disequilibrium in our study population (D = 0.998; p = 1.1 × 10 −15 ), indicating that the C-allele of SNP285 is present exclusively in a distinct SNP285C/SNP309G haplotype. 9.6% (17/178) of SNP309TG patients (8.2% of controls), and 10.2% (6/59) of SNP309GG patients (7.4% of controls) exhibited the SNP285GC genotype.

MDM2 SNP309, SNP285, and Breast Cancer Risk
To assess the breast cancer risk associated with the different SNP309 genotypes and alleles, odds ratios (OR), 95% confidence intervals (CI), and p-values were determined. This analysis revealed moderately, but non-significantly increased odds ratios associated with the GG genotype, but not the TG genotype, indicating a recessive inheritance mode (Table 1). Censoring individuals with the SNP285GC genotype, or adjusting for SNP285, did not appreciably alter this association, consistent with previous studies [20,25,32] (Table 1). We next analyzed the association of SNP285 with breast cancer risk. SNP285CC patients or controls were not observed in our study population, so we compared genotype GC vs. GG, which revealed no significant association (Table 2). Excluding individuals with the SNP309TT genotype from the analysis (which exclusively exhibited the SNP285GG genotype), or adjusting for SNP309, did not appreciably affect the observed odds ratios. Moreover, no significant association of SNP285 with breast cancer risk was observed in SNP309TG or SNP309GG individuals (Table 2).
Since associations of SNP309 with the age at onset, estrogen receptor (ER) status and p53 status have been reported [14,17,[33][34][35], odds ratios in relevant breast cancer subpopulations were determined. In this exploratory analysis, no significant associations of SNP309 with breast cancer risk were observed in any subpopulation under a dominant inheritance model (Table S3. In contrast, a significantly elevated odds ratio associated with the GG genotype was observed in p53 negative patients in a recessive inheritance model (OR, 1.82; 95% CI, 1.09-3.03; p = 0.02; Table 3). A significant association was also observed in patients with >10% Ki67-positive tumor cells (OR, 1.90; 95% CI, 1.04-3.47; p = 0.04). A trend at borderline significance for an elevated odds ratio in the recessive model was also observed in patients older than 55 years, in patients with large tumors (pT2-4; >2 cm), and in stage II-IV patients ( Table 3). Analyses of SNP285GC vs. GG did not reveal any association in these subgroups ( Table 3). The associations of SNP285GC vs. GG predominantly exhibited the same pattern as that of SNP309GG vs. TT + TG. Accordingly, higher odds ratios for both SNPs were observed in p53 negative than in p53 positive patients, as well as in ER negative, PR positive, Ki67 high, high stage, low grade, and pT2-4 patients, which may be a consequence of their complete linkage disequilibrium (Table 3).

Association of MDM2 SNP309 with Age at Breast Cancer Onset
Associations of SNP309 with the onset age of hereditary breast cancer have been reported in several studies [14,17,34,[36][37][38][39][40]. In our study population, we found the mean age of breast cancer onset for patients with the TT, TG, and GG genotype to be 57.0 ± 12.9 years (median, 56.8), 58.6 ± 13.9 years (median, 59.4), and 59.7 ± 15.0 years (median, 60.4), respectively (p < 0.05, log-rank test; Figure 1a). When carriers of the SNP285GC genotype were excluded from this analysis, somewhat more pronounced differences in the age at onset were observed: TT, 57.0 ± 12.9 years (median, 56.8; unaltered since none of them was SNP285GC); TG, 58.5 ± 13.7 years (median, 59.3); GG, 60.4 ± 14.9 years (median, 61.2; p < 0.04; Figure 1b). A similar trend to an approximately 3-4 years younger mean and median age at onset associated with the TT genotype was also observed in ER positive patients, but not in p53 negative patients ( Figure S1. No significant differences in the age at onset were observed between patients with the SNP285GG vs. GC genotypes ( Figure S1). onset for patients with the TT, TG, and GG genotype to be 57.0 ± 12.9 years (median, 56.8), 58.6 ± 13.9 years (median, 59.4), and 59.7 ± 15.0 years (median, 60.4), respectively (p < 0.05, log-rank test; Figure  1a). When carriers of the SNP285GC genotype were excluded from this analysis, somewhat more pronounced differences in the age at onset were observed: TT, 57.0 ± 12.9 years (median, 56.8; unaltered since none of them was SNP285GC); TG, 58.5 ± 13.7 years (median, 59.3); GG, 60.4 ± 14.9 years (median, 61.2; p < 0.04; Figure 1b). A similar trend to an approximately 3-4 years younger mean and median age at onset associated with the TT genotype was also observed in ER positive patients, but not in p53 negative patients ( Figure S1. No significant differences in the age at onset were observed between patients with the SNP285GG vs. GC genotypes ( Figure S1). Figure 1. Association of SNP309 genotypes with the age at breast cancer onset. Violin plots (left) and curves of the cumulative breast cancer incidence (right) of the indicated age at onset are shown for (a) all patients of the study population, and (b) patients with the SNP285GG genotype. SNP309 genotypes (TT, TG, GG) and numbers of patients (n) are indicated. Numbers in left panels indicate the median age at breast cancer onset of each genotype (indicated by white dots).

Association of MDM2 SNP309 Genotype with p53 Status
The p53 status was retrieved from clinical records, which had been determined by immune histochemistry (IHC) as part of the routine clinical breast cancer diagnostics in Austria. p53 IHC positivity indicates an inactivating mutation of p53, since it leads to a defective p53-MDM2 negative feedback loop and hence an accumulation of stabilized p53 protein [5,[41][42][43][44][45]. The presence of ≥10% p53-positive tumor cell nuclei was scored as p53 positive. p53 status was then analyzed as a function of SNP309 genotype ( Table 4). The fraction of p53-positive tumors was 31.1% (47/151) in SNP309TT patients, 25.9% (38/147) in TG patients, but only 13.2% (7/53) in GG patients (p = 0.03; Fisher's exact test), indicating a higher rate of p53 mutation with increasing number of SNP309 T-alleles (Table 4).

Association of MDM2 SNP309 Genotype with Breast Cancer Prognosis
The overall survival (OS), disease-free survival (DFS) and metastasis-free survival (MFS) of patients with the three SNP309 genotypes were compared in Kaplan-Meier analyses of unselected patients, SNP285GG patients, ER-positive patients, and patients with wildtype TP53. Detailed follow-up records for this survival analyses were available for 132 of the genotyped patients. For 115 of these 132 patients, TP53 status was determined by DNA sequencing in addition to IHC [46], and these sequencing data were used in the survival analyses. No significant differences in any of these analyses were observed, although the TT-genotype tended to be associated with a slightly poorer metastasis-free survival ( Figure 2). Accordingly, multivariable Cox proportional hazards regression analyses of the disease-free and metastasis-free survival of the variables SNP309 genotype, SNP285 genotype, ER-status, p53-status, and lymph node status were performed. Each variable was analyzed individually in parallel (Tables 5 and 6). Lymph node status, i.e., the presence vs. absence of lymph node metastases was the strongest prognostic marker in these analyses. ER-status considerably gained prognostic power in the multivariable compared to the univariable analyses, whereas p53-status showed the opposite behavior. The SNP309T-allele was significantly associated with a poorer survival in the multivariable analysis of the metastasis-free survival (p < 0.04; Table 6). Conversely, the SNP285GC-genotype was non-significantly associated with a poorer survival in both analyses (Tables 5  and 6).

Discussion
The discovery of SNP285 in MDM2 promoter P2 has led to an elegant hypothesis potentially solving a long-standing conundrum: Why is the SNP309G-allele associated with an increased cancer risk in Asian, but not in Caucasian populations [16][17][18][19][20][21]27,29]? According to this hypothesis, SNP309 likely is associated with cancer risk, consistent with the finding that genetically engineered

Discussion
The discovery of SNP285 in MDM2 promoter P2 has led to an elegant hypothesis potentially solving a long-standing conundrum: Why is the SNP309G-allele associated with an increased cancer risk in Asian, but not in Caucasian populations [16][17][18][19][20][21]27,29]? According to this hypothesis, SNP309 likely is associated with cancer risk, consistent with the finding that genetically engineered SNP309GG mice are more tumor-prone than SNP309TT mice [47]. However, this association is apparently confounded in Europeans by the adjacent SNP285, which also affects SP1 binding to another site, antagonizes the effects of SNP309, and is present at a low frequency in Europeans, but absent in Asians [20,22]. On the other hand, the present study, together with previous ones, found no evidence to support this hypothesis. Like many previous studies in European populations, we found that SNP309 is not significantly associated with breast cancer risk, although a trend for an increased risk associated with the GG genotype in the recessive inheritance model was observed [16,18,19]. However, when we corrected for the potential confounding effect of SNP285, we still did not observe any significant associations. In fact, the observed odds ratios associated with SNP309 genotypes hardly changed upon exclusion of carriers of the SNP285C allele or adjustment for SNP285, consistent with previous reports [25,32]. We conclude that the discordant results between European and Asian populations with respect to the association of SNP309 with cancer risk must be due to differences other than the presence vs. absence of the variant C-allele of SNP285.
In the initial study of SNP309, the GG genotype was found to be associated with a significantly younger age at cancer onset in Li-Fraumeni patients [14]. This finding was confirmed in most subsequent studies in Li-Fraumeni patients [36][37][38][39][40]. Interestingly, an association of the SNP309GG genotype with an earlier onset was also observed in several studies of familial breast cancer, particularly in BRCA1 mutation carriers [17,27,48,49]. However, most studies of sporadic breast cancer did not find such a premature onset associated with the SNP309G-allele [18,21,[50][51][52][53]. Here we report an effect opposite to that in familial cancer syndromes: In our study population, the SNP309TT genotype is associated with a younger age at breast cancer onset, consistent with a previous report [17]. Like our study, Lum et al. reported an association of the SNP309GG genotype with an increased breast cancer risk, but a delayed onset [17]. This is unexpected given the proposed molecular function of the SNP309G-allele to increase the expression of MDM2 and hence to attenuate the p53-pathway. We hypothesize that this seemingly paradoxical association may be due to the differential rate of p53 positivity found by us and others, which is low in SNP309GG, but high in TT-tumors (see discussion below). It has been convincingly established that TP53 mutant patients have a much younger age at onset [2][3][4]. Accordingly, we propose that the SNP309TT genotype leads to a higher selective pressure to mutate TP53 and hence to a much higher rate of these mutations, which then causes the earlier mean age at onset associated with that genotype. This effect of differential somatic TP53 mutation rate obviously does not exist in Li-Fraumeni patients who have a germline mutation in TP53, which may explain why they differ from sporadic cancer cases with respect to the association of SNP309 with the age at cancer onset. Consistently, we found an association of the age at breast cancer onset with MDM2 mRNA expression. Significantly elevated mean MDM2 mRNA levels were observed in patients with an age ≥55 years, as well as in post-menopausal patients (H. Miedl and M. Schreiber, unpublished data).
We found that the SNP309GG genotype is significantly associated with an increased breast cancer risk in p53 negative patients, who harbour a functional p53 protein (OR, 1.82; 95% CI; 1.09-3.03; and p = 0.02), but not in p53 positive or unselected patients (Tables 1 and 3), consistent with a previous report [50]. Thus, the impact of SNP309 on breast cancer risk appears to be limited to tumors in which p53 is functional, which is biologically plausible since the key function of MDM2 to negatively regulate p53 becomes irrelevant if p53 is inactivated by mutation. Moreover, we found evidence for an inverse correlation of TP53 mutation with the presence of the variant SNP309G-allele. TP53 was positive in 31% of SNP309TT patients, 26% of TG patients, but only 13% of GG patients (p < 0.04; Table 4). Since the SN309G allele leads to a higher expression of MDM2 and consequently lower levels of p53, presence of the SNP309TG and particularly the GG genotype may, at least in part, functionally substitute the inactivation of TP53 by mutation. Thus, if p53 tumor suppressor activity is already attenuated by overexpression of MDM2, the selective pressure to additionally mutate TP53 may be reduced in the presence of the SNP309G allele, and this may be reflected in the lower rates of p53 positivity observed in carriers of the SNP309GG and TG genotypes. Interestingly, amplification of MDM2, another clinically relevant mechanism in carcinogenesis to increase MDM2 and decrease p53 expression, was also found to be associated with a reduced rate of TP53 mutation [10,13,54]. Like in the present study, a reduced rate of TP53 mutations in carriers of the SNP309G allele has been reported for non-small cell lung cancer [55], hepatocellular carcinoma [35], and bladder cancer [56]. In contrast, no significant association was found in oral cancer, pancreatic cancer, and two studies of colorectal cancer [57][58][59][60]. A lower frequency of p53 positivity was also found in breast cancer patients with the SNP309GG genotype compared to those with the SNP309TT genotype [50]. An analogous association of a germline polymorphism with a differential mutation rate of TP53 was also found for the R72P coding SNP (rs1042522) in TP53 itself [46,61].
We found no significant association of SNP309 genotype with survival, consistent with most reports in breast cancer and other tumor types [50,[62][63][64][65]. However, the TT genotype tended to be associated with a slightly poorer metastasis-free survival, and this association was significant in multivariable analysis (Table 6). Since the SNP309T allele is thought to lead to lower MDM2 and hence higher p53 levels [14], we expected it to rather be associated with an improved survival. However, the TT genotype has been found associated with a poor prognosis previously, including in a breast cancer study in BRCA1/2 mutation carriers [66][67][68]. We hypothesize that this counter-intuitive association is due to the much higher rate of TP53 mutation in SNP309TT patients. To the best of our knowledge, we here report the first analysis of the association of SNP285 with breast cancer survival. We found the SNP285GC genotype non-significantly associated with a poorer disease-free and metastasis-free survival in univariable and multivariable analyses (Tables 5 and 6). However, this result is based on the analysis of only four patients with the GC genotype and 128 with the SNP285GG genotype and should be interpreted with caution. The association of SNP285 with survival has previously been analyzed in two lung cancer studies. One study found no association, whereas the other study found an association of the SNP285GC genotype with a worse progression-free survival compared to the GG genotype, but no association with the overall survival [32,55].

Study Population
This study was approved and is annually reviewed by the Institutional Review Board Only women of Central European descent from the same geographical area were included as patients or controls. The response rates for both patients and controls were >80%. Two patients were excluded from further analyses due to failure to genotype SNP309, one additional patient and one control due to failure to genotype SNP285, and one additional patient since genotyping of both SNPs was unsuccessful. Accordingly, all analyses reported here are based on the 406 breast cancer patients and 254 controls for whom both SNP309 and SNP285 were successfully genotyped.

Statistical Analyses
Statistical analyses were performed with R 3.3.2, an open-source language and environment for statistical computing [70]. Chi-square tests with Yates' continuity correction were used to evaluate whether the study population was in Hardy-Weinberg equilibrium. Confidence intervals and p-values associated with odds ratios were calculated by the mid-P exact method. As advocated [71], we did not adjust our subgroup analyses (Table 3) for multiple testing, since they should be considered exploratory. Differences between two groups with respect to age at onset were analyzed by two-sided t-tests, and between three or more groups by ANOVA. Statistical methods of survival analyses, as well as follow-up details of our study population have been described in References [46,72]. p-Values to Kaplan-Meier analyses and analyses of the cumulative breast cancer incidence were calculated by log-rank tests, as described in Reference [73]. The disease-free and metastasis-free survival were also analyzed by using a Cox proportional hazard model, unadjusted or adjusted for ER status, TP53 status, and lymph node status. The subcategories of these variables were coded as follows: SNP309 genotype, TT = 0, TG = 1, GG = 2; SNP285 genotype, GG = 0, GC = 1; ER status, positive = 0, negative = 1; TP53 status, wildtype = 0, mutated = 1 (determined by sequencing [46]); lymph node status, absence of lymph node metastases (pN0) = 0, presence of lymph node metastases (pN+) = 1. All p-values shown are two-sided. Associations with p-values < 0.05 were considered statistically significant.

Conclusions
We found that the SNP309GG genotype was associated with an older age at breast cancer onset, a good prognosis in multivariable analysis of the metastasis-free survival and a trend in the same direction in related survival analyses, an increased breast cancer risk in specific subgroups, and a rate of p53 positivity less than half compared to the SNP309TT genotype. These findings are consistent with a model in which the p53-response is attenuated in the presence of the SNP309GG genotype. A direct consequence of this attenuation may be that the risk of breast cancer and other cancers is increased, but that the selective pressure to mutate TP53 is reduced. An attenuated p53-response is presumably still functional to some extent as opposed to one that is completely inactivated by mutation, and consequently leads to a later onset of breast cancer and a trend towards a good prognosis. None of the observed associations of SNP285 were significant in our study, which may in part be due to the very low frequency of the variant C-allele. However, the observed trends towards an increased breast cancer risk and a poor survival are at odds with the proposed molecular function of the SNP285C-allele to reduce binding of SP1 and expression of MDM2. Accordingly, we conclude that these associations of SNP285C are bystander effects of its complete linkage disequilibrium with SNP309G. The odds ratios in subgroups according to SNP309 genotype (Table 2), and in clinical subgroups (Table 3) support this conclusion. For example, the odds ratios associated with both the SNP309GG and the SNP285GC genotype are higher in the Ki67-high than in the Ki67-low subgroup (Table 3). This is consistent with the molecular function of SNP309G to attenuate the anti-proliferative p53-response, but not with that of SNP285C to enhance it. These bystander effects may be more pronounced in our study population compared to previous ones due to the lower frequency of the SNP285GC genotype and the complete absence of SNP285CC subjects. Finally, the fact that SNP309 is associated with an increased cancer risk in most Asian but not European studies has been suggested to be due to the antagonistic molecular function and consequential confounding effect of SNP285, which is present in Europeans, but absent in Asians. However, the results of the present study do not support this hypothesis. Specifically, when we excluded SNP285GC individuals from our analyses, thus mimicking the situation in Asian populations, the observed associations (ORs) were still not significant, and remained essentially unaltered (Table 1).
In conclusion, we report here the first analysis of SNP285 in a Central European study population, the first analysis of the association of SNP285 with breast cancer survival, the first analysis of SNP285 in clinically relevant subgroups of breast cancer, and the first analysis of potential confounding effects of SNP285 on the association of SNP309 with the age at onset and survival.