MDM2 SNP309 is associated with high grade node positive breast tumours and is in linkage disequilibrium with a novel MDM2 intron 1 polymorphism

MDM2, encoded by the human homologue of Murine Double Minute oncogene, is the principal negative regulator of p53, a transcription factor which plays key roles in cell division and response to DNA damage [1, 2]. p53 is frequently mutated in cancer resulting in defective functions, including apoptotic and cell cycle arrest programs [3]. MDM2 controls p53 levels and activity by a number of different mechanisms, including direct inhibition of the transcriptional activity of p53 [4]. In addition, MDM2 acts as an E3 ubiquitin ligase targeting p53 for nuclear export and proteosomal degradation [5]. Furthermore, as MDM2 is a transcriptional target for p53, through the P2 inducible promoter located in intron 1, a finely balanced negative feedback loop mechanism exists [6].

Development of cancer is often associated with defects in this p53-MDM2 regulatory circuit, and in cells with wild type p53 other alterations in the p53 pathway are often observed [3]. MDM2 is overexpressed in a number of different cancers and in breast cancer, where only 30% of tumours have mutated p53, some 40% display overexpression of MDM2, although amplifications are rare [7‐10].

A single nucleotide polymorphism (SNP) within intron 1 of MDM2, a T to G substitution (T>G) at position 309 (SNP309) (rs2279744), has been shown to lead to enhanced binding of the Sp1 transcription factor resulting in elevated levels of both MDM2 mRNA and protein, thereby attenuating the p53 response [11]. In Li-Fraumeni patients, individuals homozygous or heterozygous for SNP309 (G/G or G/T) were shown to develop cancer at an earlier age than wild type individuals; in patients with sporadic soft tissue sarcomas, the 309G/G genotype correlated with an average 12 year earlier age of diagnosis [11]. SNP309 was thus postulated as a potential modulator of cancer susceptibility [11].

Subsequent studies of the SNP309 polymorphism have demonstrated variable frequencies of 309G/G depending on race and ethnicity [12‐15]. Similarly, the association between SNP309 and development of cancer has produced conflicting data (reviewed in [16]). This study therefore sought to sequence the MDM2 intron 1 region around SNP309 in detail and determine SNP frequencies from a control cohort of Scottish Caucasians (n = 275) and a cohort of geographically matched Scottish Caucasian women with breast cancer (n = 299). The MDM2 SNP genotypes were examined to determine if they could be linked to an increased cancer susceptibility, age of cancer diagnosis, pathological variables and clinical outcome.

The intron 1 region of MDM2 was examined in a Scottish population of 299 breast cancer patients and 275 cancer free controls to establish any associations between MDM2 SNPs and breast cancer. In the breast cancer cohort MDM2 SNP309 was also analysed with respect to age at cancer diagnosis and pathological variables. The well characterised SNP309 and three additional SNPs: 344T>A, 285G>C and 443G>T, were identified in both the control and cancer populations. SNP 344T>A has been reported previously [11], but not studied in detail, and the latter two SNPs, to our knowledge, represent novel polymorphisms. SNP443G>T lies within the conserved 5'-RRRCWWGYYY [0,13]RRRCWWGYYY-3' p53 response element [6, 25], implying that this SNP may alter the ability of p53 to bind and regulate the MDM2 intron 1 promoter, thereby influencing cancer susceptibility. However, a larger cohort would be required to test the clinical significance of this observation.

Neither the 344T>A, nor the 443G>T SNP were observed with the 309G/G genotype, probably due to the very low frequencies of each allele. In contrast, SNP285 G>C, which also has a low prevalence was always observed with the SNP 309T>G, either in the heterozygous (309T/G) or homozygous (309G/G) states, but never observed with the commonest 309T/T allele. This implies that 309SNP and 285SNP are in linkage disequilibrium and was confirmed by haplotype analysis. SNP 285 and SNP309 thus may influence each other and together modulate the levels of MDM2 produced. Supporting this potential role of SNP285 enhancing the effects of 309G/G, we observed the double homozygous genotype (285C/C, 309G/G) only in the breast cancer cohort and the heterozygous/homozygous genotype (285G/C, 309G/G) was more prevalent in the breast cancer cohort. Moreover, SNP285 appeared to deviate from Hardy-Weinberg equilibrium only in the breast cancer cohort. However, although potentially interesting, the numbers in our study are exceptionally small and therefore to verify if the combination of 309G/G together with the 285C variant may be breast cancer associated, a much greater number of patients would be required.

The comparable frequencies of T/T, G/T and G/G SNP309 for the control and breast cancer cohorts suggests that the SNP309G/G genotype did not significantly influence breast cancer occurrence within this Scottish population, consistent with studies of both sporadic and familial breast cancer [15, 16, 21, 26‐34]. Similarly, no increased cancer risk, associated with the 309G/G genotype, has been observed for other cancer sites including: colorectal cancer [35‐37], uterine leiomyosarcomas [36], squamous cell carcinoma of the head and neck [36] and lung cancer [38]. In contrast, several studies, encompassing a diverse array of other tumour types but also including colorectal and lung cancer, have reported an increased risk of cancer occurrence associated with 309G/G [14, 39‐43]. While the frequencies observed in our studies are similar to those reported in other Caucasian populations, SNP309 genotypic frequencies do vary considerably between different ethnic groups [15, 31]. For example, approximately 3% of African Americans carry the double homozygous 309G/G locus compared to frequencies of approximately 30% in the Japanese and Korean populations [12, 13, 15, 31]. This may, in part, explain the discrepant associations between the 309G/G genotype and cancer. Additionally, from our work, other SNPs e.g. SNP285, which may also display variations in demographic frequencies, may influence the penetrance of SNP309. Furthermore, the higher frequency G allele in females confirms that appropriate controls should be used when interpreting results.

We did not observe an association between 309SNP and earlier age of cancer diagnosis unlike previous investigations [11, 23, 33], although the majority of studies in breast cancer are consistent with our findings [16, 26‐28, 31, 32, 34]. Further studies by Bond et al., 2006, showed that accelerated age of onset in 309G/G patients was observed only in females and, in diffuse large B-cell lymphomas and soft tissue sarcomas the 309G/G genotype had a higher incidence in pre-menopausal compared to post-menopausal patients. In our breast cancer cohort, there was no difference in SNP309 frequencies in pre- and post-menopausal patients. In the same study Bond et al. also examined invasive breast cancer stratified by ER status in addition to menopausal status and demonstrated both an increased frequency of 309G/G and an earlier age of onset associated with ER positive tumours in pre-menopausal women. In the present series, a higher frequency of the 309G/G genotype was observed in the pre-menopausal ER positive patients (ER ≥ 4) compared to the post-menopausal ER positive patients which was further enhanced by selecting a highly ER positive group (ER ≥ 12). Nonetheless, within any sub-classification no differences in the mean age of breast cancer diagnosis and genotype were seen.

These findings, demonstrating a higher proportion of 309G/G patients in the ER highly positive pre-menopausal group, are in agreement with Bond et al., 2006 [23]. However, the proportions of pre- and post-menopausal highly ER positive tumours in the present study are very different to those reported previously: the highly ER positive (ER ≥ 12) pre-menopausal group had a frequency of 13.7% and the post-menopausal group had a frequency of 86.3%. In the two breast cancer cohorts studied by Bond et al., the Caucasian Ashkenazi-Jewish group had values of 35.4% and 64.6% for the pre- and post-menopausal highly ER positive groups respectively, and values of 50.4% (Pre) and 49.6% (Post) in the second Caucasian population. This suggests that the demographics of the Scottish Caucasian population are very different to the American Caucasian population, which also appear to differ based on ethnicity. This may also explain, in part, why we did not observe an association with the mean age of cancer diagnosis. It also suggests that estrogen may not be the only variable that impacts on the penetrance of SNP309.

Compared with clinical and pathological features, the 309G/G genotype appeared to be associated with higher tumour grade and node positive cancers. Combining these parameters together with tumour size, to calculate NPI values, it was therefore not unexpected that 309G/G was linked with an elevated NPI score. From this preliminary study, the 309G/G genotype in the Scottish breast cancer population therefore appears to correlate with poor prognostic indicators, although a much larger cohort would be required to verify these findings. This association of 309G/G with higher grade tumours and increased nodal involvement is supported by a recent study in nasopharyngeal carcinoma where the 309G/G genotype was associated with advanced lymph node metastasis [39]. Other corroborating evidence comes from the observation that over-expression of MDM2 in cancer has been linked with increased levels of metastasis, poorer response to therapy and bad prognosis [44]. In vitro studies, utilising cell lines, have also shown that cells gain enhanced metastatic potential when MDM2 is up-regulated during hypoxic conditions [45].

Our data, taken together with that of the literature, suggests that the 309 genotype may be influencing the grade and metastatic potential of breast tumours. However, the presence of the G allele does not appear to be adding to this alone as the 309G/T genotype was found to be associated with low grade tumours. One potential explanation for this is that each of the different alleles could be binding different transcription factors. Indeed, the P2 promoter of MDM2 is known to be regulated by a myriad of different transcription factors including Sp1, p53, Ap1-ETS and MYCN [6, 46‐48]. Levels of these regulatory proteins and whether they are mutant or wild type, modified or unmodified, could also impact on MDM2 transcription levels. One key regulator influencing the penetrance of SNP309 may be p53, particularly as it is known to be haploinsufficient [49]. p53 has been examined in terms of the p53 polymorphism Arg72Pro [13, 29, 32, 37, 50] and mutational status of p53 [26, 41, 51‐54] with conflicting results. Similarly, SNP285G>C that is in linkage disequilibrium with SNP309, identified in this study, may play a modulatory role influencing SNP309 penetrance. Thus, although all breast cancers can be classified on a histological and pathological level, this may in fact be determined by underlying SNP differences, some of which have greater influence than others depending on the population and the additional parameters examined.

In conclusion, within this Scottish population the MDM2 309 polymorphism was found to associate with high grade cancers with greater nodal involvement and by implication poorer prognosis, but had no impact on the age of diagnosis of breast cancer. Furthermore, SNP309 is in linkage disequilibrium with SNP285 which may further modulate the penetrance of polymorphism effects.

^§National Translational Cancer Research Network, ^‡Breast Cancer Research Scotland, ^#University of Dundee, ^†National Health Service Tayside, *Cancer Research UK.