The rs11515 Polymorphism Is More Frequent and Associated With Aggressive Breast Tumors with Increased ANRIL and Decreased p16INK4a Expression

Chromosome position 9p21 encodes three-tumor suppressors p16INK4a, p14ARF, and p15INK4b and the long non-coding RNA ANRIL (antisense non-coding RNA in the INK4 locus). The rs11515 single-nucleotide polymorphism in the p16INK4a/p14ARF 3′-untranslated region is associated with glioblastoma, melanoma, and other cancers. This study investigated the frequency and effect of rs11515 genotypes in breast cancer. Genomic DNA samples from 400 women (200 with and 200 without a diagnosis of breast cancer) were genotyped for the rs11515 major (C) and minor (G) alleles. The rs11515 polymorphism was also investigated in 108 heart tissues to test for tissue-specific effects. Four 9p21 transcripts, p16INK4a, p14ARF, p15INK4b, and ANRIL were measured in breast tumors and myocardium using quantitative PCR. Heterozygotes (CG genotype) were more frequent in women with breast cancer compared to the control population (P = 0.0039). In those with breast cancer, the CG genotype was associated with an older age (P = 0.016) and increased lymph node involvement (P = 0.007) compared to homozygotes for the major allele (CC genotype). In breast tumors, the CG genotype had higher ANRIL (P = 0.031) and lower p16INK4a (P = 0.006) expression compared to the CC genotype. The CG genotype was not associated with altered 9p21 transcripts in heart tissue. In breast cancer, the rs11515 CG genotype is more frequent and associated with a more aggressive tumor that could be due to increased ANRIL and reduced p16INK4a expression. The absence of association between rs11515 genotypes and 9p21 transcripts in heart tissue suggests this polymorphism has tissue- or disease-specific functions.

Given the association of the CDKN2A rs11515 and the functions of proteins encoded at 9p21 in cancer, the current study investigated rs11515 genotypes in breast cancer.

Patients
For the breast cancer analyses, the patient cohort consisted of 200 women diagnosed with primary breast cancer who underwent surgery at Christchurch hospital, New Zealand. The mean age for the breast cancer cohort was 59.7 years (95% CI 57. 6-61.8). Ethnicity data were available for 152 women (87% identified as European, 3% Maori, 3% Maori and European, and 7% identified with other ethnic groups). The control cohort consisted of 200 healthy women from the New Zealand population. The mean age of the control cohort was 56 years (95 CI percentile 54-58.5 years). Ethnicity data were available for all in the control cohort. Eightyfive percent identified as European, 3% Maori, 5% Maori and European, and 7% identified with other ethnic groups. There was no significant difference in age between the patient and control cohorts. Frozen breast tumors were available for 25 women (12 women with the CG and 13 with the CC genotype). Tumor tissue was limited for four individuals with the CG genotype; therefore, some analyses were performed on 8 instead of 12 tumors. The breast tumors were selected so that four variables were matched between those with the CG and CC genotypes (tumor grade, presence of lymph/vascular invasion, ≥2 lymph nodes affected, and an estrogen-receptor positive tumor). All breast tumors were from women who did not have previous treatment with radiotherapy or chemotherapy before surgery.
For the analysis of heart tissue, tissue from the left ventricular free wall of the myocardium of organ donors (n = 108) was collected by the Cleveland Clinic Kaufman Center for the Heart Failure Human Heart Tissue Bank between 1993 and 2006, as previously described (27).
For the breast tumor study, ethical approval was obtained in New Zealand, and all women gave written informed consent for inclusion in the study (Ethics Committee Approvals 02.06.98/5.11.09 and LRS/10/09/035). Standard procedures were followed, which included culturally appropriate tissue handling and disposal protocols (28). For the heart tissue study, ethical approval was obtained from the Cleveland Clinic Internal Review Board (IRB 2378), and the study adhered to the principles outlined in the Declaration of Helsinki. All procedures followed were in accordance with institutional guidelines and all families and/or patients provided informed consent.

rs11515 genotyping
The C and G rs11515 alleles were genotyped in the breast tumor and control cohorts using a PCR and restriction enzyme digestion method. Genomic DNA was extracted from blood leukocytes using the Dneasy Blood & Tissue Kit (Qiagen, Veno, Netherlands) according to the manufacturer's instructions. A 319 bp region was amplified using the following primers: forward, 5′-TGCCACACATCTTTGACCTC-3′; reverse, 5′-GCAGAAGCGGTGTTTTTCTT-3′. Following conformation of the correct PCR product, amplified DNA was digested with the MspI restriction enzyme. Restriction enzyme-digested PCR products were separated by agarose gel electrophoresis to identify the rs11515 genotype (C allele, digested band; and G allele, undigested band). DNA sequencing was performed on 20 PCR products to confirm genotypes as described previously (17).
The C500 and G500 rs11515 alleles were genotyped in heart tissue using a Taqman SNP assay (C_12096259_10, Applied Biosystems, Foster City, CA, USA). Genomic DNA was extracted from frozen tissue as previously described (27) and from samples genotyped in duplicate using the Lightcycler 480 platform and Lightcycler 480 software, version 1.5.0 (Roche, Indianapolis, IN, USA). Reactions were optimized for 5 μL volumes with 0.5× the recommended probe concentration.

rs3088440 genotyping
For the breast tumor study, rs3088440 alleles were genotyped in the patient and control cohorts using a PCR and restriction enzyme digestion method. A 180 bp region was amplified from blood leukocyte genomic DNA using the following primers: forward, 5′-TAGATCATCAGTCACCGAAGG-3′; reverse, 5′-CATTTACGGTAGTGGGGGAAG-3′. Following conformation of the correct PCR product, amplified DNA was digested with the HaeIII restriction enzyme. Restriction enzyme-digested PCR products were separated by agarose gel electrophoresis to identify the major allele (digested band) and the minor allele (undigested band). DNA sequencing was performed on 20 PCR products to confirm genotypes as described previously (17). The rs11515 Polymorphism in Breast Cancer Frontiers in Oncology | www.frontiersin.org

CDKN2A gene Dosage
The gene dosage of CDKN2A, exon 1α (p16 INK4a ), and exon 1β (p14 ARF ) was estimated in 13 breast tumors with the CC and 12 tumors with the CG rs11515 genotype using a multiplex PCR assay, as previously described (29). Tumor DNA was extracted using the Qiagen Dneasy Blood & Tissue Kit according to the manufacturer's instructions (Qiagen, Limburg, The Netherlands). The intensity of the CDKN2A and β-globin PCR products was compared using Bio-Rad Quantity One software (Bio-Rad Laboratories, CA, USA) and Syngene GeneTools image software (Syngene, Cambridge, UK) as previously described (17).
For the study of heart donors, expression data for CDKN2A/ p16 INK4a and CDKN2A/p14 ARF was obtained in the current study, and that for CDKN2B and ANRIL generated as part of a previous study (27) and included in the current analysis. RT-qPCR was performed using Taqman Gene Expression assays for CDKN2A/ p16 INK4a (assay id: Hs02902543_mH), CDKN2A/p14 ARF (assay id: Hs99999189_m1), CDKN2B (assay id: Hs00793225_m1), and ANRIL (assay id: Hs01390879_m1) as described above. Expression levels were normalized to TPT1 (assay id Hs02621289_g1), EEF1A1 (assay id Hs00265885_g1) and signal recognition particle 14 kDa (SRP14, assay id Hs03055045_g1), three reference genes previously validated for use in human myocardium (32).

statistical analyses
The genotype frequencies were compared between cases and controls using the Chi-square test and SHEsis software (33). For breast tumors and heart samples, clinicopathologic measures were compared between genotypes using Chi-square and one-way ANOVA followed by pairwise comparisons tests. All gene expression levels displayed skewed distributions and were loge-transformed prior to analysis. Associations between gene expression and rs11515 genotypes were performed with analysis of variance (genotypic model), with adjustment for age, as appropriate. Correlations in expression levels between genes were tested with Pearson correlation. Statistical analyses were performed with SPSS version 22 software and P < 0.05 was taken as a significant difference.

resUlTs
The rs11515 Minor allele Was More Frequent in Women with Breast cancer The rs11515 alleles were genotyped in woman with breast cancer and healthy women. In women with breast cancer, 61% (n = 121) were homozygote for the major allele (CC, genotype), 38% (n = 76) were heterozygote (CG, genotype), and 1.5% (n = 3) were homozygote for the minor allele (GG, genotype). The genotypes of 20 samples were confirmed using DNA sequencing (7 typed as CC, 10 typed as CG, and 3 typed as GG). In the control population, 76% (n = 152) had the CC, 23% had the CG (n = 46), and 1% of women had the GG genotype. The genotype frequencies in controls were in Hardy-Weinberg equilibrium (P = 0.47). The CG genotype was more frequent in women with breast cancer compared to the control population (P = 0.0039).
Clinicopathologic variables (patient age at presentation, tumor size, the number of nodes involved, venous and lymphatic invasion, estrogen receptor, progesterone receptor, and human epidermal growth factor 2 receptor statuses) among rs11515 genotypes are shown in Table 1 and were compared between genotypes in the breast cancer cohort. The mean age for those with the CG allele was higher (63.3 years, 95% CI 60-66.5 years) compared to those with the CC genotype (58.2 years, 95% CI, 55.6-60.7 years, P = 0.016). Few women had the GG genotype, but all three women with this genotype were young (aged 23, 33, and 35 years).
The CG genotype was associated with a more aggressive tumor with greater lymph node involvement. Forty-five percent of women with the CG genotype had two or more nodes with malignant cells, and 38% had no nodal involvement. In those with the CC genotype, 21% had two or more nodes with malignant cells and 60% had no nodal involvement (P = 0.007). The tumor size and hormone receptor statuses did not differ between the rs11515 genotypes.
The rs11515 polymorphism is located in close proximity to another polymorphism associated with breast cancer (rs3088440) (34). To determine if rs3088440 genotypes were increased in the breast cancer cohort, the rs3088440 polymorphism was genotyped. In the breast cancer, cohort 78% (n = 155) were homozygote for the major allele, 21% were heterozygote (n = 42), and 1.5% (n = 3) were homozygote for the minor allele. In the control cohort, 82% (n = 164) were homozygote for the major allele, 17% were heterozygote (n = 34), and 1% (n = 2) were homozygote for the minor allele. The rs3088440 genotype was not significantly different in women with breast cancer compared to the control population. Nine individuals were heterozygote for both the rs11515 and rs3088440 SNPs.

The cg genotype Was associated With increased ANRIL expression in Breast Tumors
In glioblastoma patients, the CG genotype was associated with increased loss of p16 INK4a and p14 ARF due to gene deletions at 9p21 (17). To determine if this was the case in breast cancer, DNA was isolated from 25 breast tumors (12 from women with the CG genotype and 13 with the CC genotype), and multiplex PCR used to estimate the gene dosage of CDKN2A exon 1a (p16 INK4a ) and exon 1b (p14 ARF ) relative to an internal beta-globin gene fragment (29). All breast tumors had retained exon 1a and exon 1b (data not shown), suggesting the CG genotype was not associated with loss of p16 INK4a and p14 ARF gene dosage. Polymorphisms in CDKN2A/2B/ANRIL may regulate gene expression within the cluster. To test this hypothesis, we investigated associations between rs11515 genotypes and expression of CDKN2A (transcripts, p16 INK4a and p14 ARF ), and CDKN2B in 25 breast tumors (12 from women with the CG genotype and 13 with the CC genotype). Due to the limited availability of tumor tissue for those with the CG genotype, ANRIL was measured in 21 breast tumors (8 from women with the CG genotype and 13 with the CC genotype).

DiscUssiOn
The current study adds breast cancer to a growing list of malignancies associated with the rs11515 minor allele (10)(11)(12)(13)(14)(15)(16)(17). The CG genotype was more frequent in those with breast cancer, associated with an older age and a more invasive tumor, and found with a negative correlation between ANRIL and p16 INK4a . These data suggest the CG genotype is associated with a more aggressive tumor. Reduced p16 INK4a tumor suppressor function in combination with increased ANRIL offers an explanation for increased cancer susceptibility with the rs11515 minor allele. An association between the CG genotype and lower expression of p16 INK4a is consistent with previous studies. In melanoma and colorectal carcinoma, the G allele was associated with loss of p16 INK4a and p14 ARF expression, and in glioblastoma, CG genotype was associated with reduced p16 INK4a and p14 ARF gene dosage (13,14,17). The observation that ANRIL levels were higher and negatively correlated with p16 INK4a in patients with the CG genotype, suggests ANRIL may be upregulated and acting to repress p16 INK4a expression. There is evidence to suggest ANRIL affects expression of 9p21 transcripts by acting as a scaffold to guide chromatinremodeling proteins toward 9p21 (19,21,(35)(36)(37). ANRIL may promote the formation of heterochromatin facilitating reduced expression of p16 INK4a expression. A specific example is the binding of ANRIL to polycomb repressive complexes 1 and 2 (PRC1 and PRC2) that is thought to direct epigenetic regulation and reduced p16 INK4a and p15 INK4b (19,35).
Other data suggest increased ANRIL would be associated with increased rather than decreased expression at 9p21 in breast cancer. Using TCGA expression data for invasive breast cancer, ANRIL and CDKN2A (p14 ARF and p16 INK4a combined) were positively correlated (Pearson correlation 0.588, Figure 4A) (38,39), with a marginal positive correlation between ANRIL and p15 INK4b (Pearson Correlation 0.39, Figure 4B). An analysis of 12 breast tumors by Pasmant et al. (18) found increased ANRIL correlated with increased p14 ARF , p16 INK4a , and p15 INK4b (18). The available TCGA data are not divided based on the rs11515 genotype. Considering most women will be homozygote for the major rs11515 allele, the inconsistency between the TCGA data with that from the current study raises the possibility that the relationship between ANRIL and p16 INK4a may be influenced by the rs11515 genotype, with an inverse relationship for those with the CG but not the CC genotype.
Although we argue that the CG genotype is correlated with increased risk toward breast cancer due to increased nodal involvement, increased ANRIL, and reduced p16 INK4a expression, it could be argued that the CC genotype is the risk genotype. Patients with the CC genotype were younger. In another study, increased rather than decreased expression of p16 INK4a was associated with a higher breast tumor grade and increased proliferative capacity   (40). It could be that the CG genotype protects against tumors initially, but with age tumor surveillance may be compromised leading to a more aggressive tumor. The current study did not investigate ANRIL expression in normal associated breast tissue, so we did not determine whether it is the CG or CC genotype that is aberrant for ANRIL expression in breast tumors. Those with the GG genotype were young and this supports the G rs11515 allele being the risk allele for breast cancer. However, few had the GG genotype and breast tissue from GG homozygotes was not available to determine if ANRIL expression is greatest in these women. The association between the CG genotype and ANRIL did not occur in heart tissue, suggesting the effect of rs11515 alleles is tissue or disease specific. Multiple isoforms of ANRIL exist including both linear and circular forms with tissue-specific expression patterns identified (41,42). ANRIL isoforms with an ALU motif were found to be more effective in recruiting chromatin-remodeling proteins (21). The specific ANRIL isoforms increased in breast tumors were not identified so the tissue-specific differences between breast cancer and myocardium could be attributed to different ANRIL isoforms being present. The rs11515 SNP is predicted to affect the binding site of microRNA-601 (43), so the effect of rs11515 genotypes may be dependent of the expression of other transcripts.
Other 9p21 polymorphisms have been correlated with cancer including breast cancer (rs1011970), glioma (rs1063192, rs2157719, rs1412829, and rs4977756), basal cell carcinoma (rs2151280), nasopharyngeal carcinoma (rs1412829), and breast cancer (rs1011970 and rs3088440) [reviewed in Ref. (34,44)]. Consequently, another polymorphism in linkage with rs11515 could be the causal allele for the increased ANRIL expression. The rs3088440 polymorphism is located 40 bp downstream of rs11515. In a large cohort of over 3000 women with breast cancer, the rs3088440 minor allele was associated with breast cancer (34), and in pancreatic cancer associated with a reduced time to tumor progression and a poorer response to therapy (45). The rs3088440 genotypes were not significantly increased in the breast cancer cohort in the current study; however, a limitation of the current study was a smaller cohort was used (34). The majority of individuals' heterozygote for the rs11515 allele were homozygote for the rs3088440 major allele; therefore, it is possible that a haplotype contributes to the associations found in this study instead of the rs11515 genotype alone. cOnclUsiOn To our knowledge, this is the first report of an association between the rs11515 minor allele with breast cancer and a negative correlation between ANRIL and p16 INK4a levels with the CG genotype. Reduced expression of p16 INK4a may be one mechanism by which the rs11515 CG genotype increases cancer risk.
aUThOr cOnTriBUTiOns JR -designed and supervised experimental studies, and edited the manuscript; AP -performed PCR, data analysis and interpretation, wrote, and edited the manuscript; AA -performed PCR, analyzed data, and edited the manuscript; HM, MC, CM, WS, WHWT -designed research, interpreted data, and edited the manuscript. CF -performed the statistical analyses, interpreted data, and edited the manuscript; IAR -interpreted TCGA data, designed research, and edited the manuscript; NA -conceived the study, interpreted data, and edited the manuscript. TS -conceived the study, performed PCR, designed research, and wrote and edited the manuscript.