Non-Coding Polymorphisms in Nucleotide Binding Domain 1 in ABCC1 Gene Associate with Transcript Level and Survival of Patients with Breast Cancer

Objectives ATP-Binding Cassette (ABC) transporters may cause treatment failure by transporting of anticancer drugs outside of the tumor cells. Multidrug resistance-associated protein 1 coded by the ABCC1 gene has recently been suggested as a potential prognostic marker in breast cancer patients. This study aimed to explore tagged haplotype covering nucleotide binding domain 1 of ABCC1 in relation with corresponding transcript levels in tissues and clinical phenotype of breast cancer patients. Methods The distribution of twelve ABCC1 polymorphisms was assessed by direct sequencing in peripheral blood DNA (n = 540). Results Tumors from carriers of the wild type genotype in rs35623 or rs35628 exhibited significantly lower levels of ABCC1 transcript than those from carriers of the minor allele (p = 0.003 and p = 0.004, respectively). The ABCC1 transcript levels significantly increased in the order CT-GT>CC-GT>CC-GG for the predicted rs35626-rs4148351 diplotype. Chemotherapy-treated patients carrying the T allele in rs4148353 had longer disease-free survival than those with the GG genotype (p = 0.043). On the other hand, hormonal therapy-treated patients with the AA genotype in rs35628 had significantly longer disease-free survival than carriers of the G allele (p = 0.012). Conclusions Taken together, our study shows that genetic variability in the nucleotide binding domain 1 has a significant impact on the ABCC1 transcript level in the target tissue and may modify survival of breast cancer patients.


Introduction
Breast cancer (OMIM: 114480) is the most common malignancy affecting female population worldwide. Despite early detection and improved understanding of molecular mechanisms of this disease, it is still the second leading cause of cancer death in women [1].
Multidrug resistance (MDR) represents a major obstacle to successful therapy of tumors. MDR was first described in 1970 [2] as a cross-resistance to structurally and functionally different anticancer drugs. Most MDR is caused by enhanced expression of membrane-bound ATP-Binding Cassette (ABC) transporters [3,4]. ABC transporters pump drugs outside of the cells into the extracellular space, thus reducing their cytotoxic effect [5][6][7].
A recent tissue microarray study has concluded that high ABCC1 protein expression is a negative prognostic marker as it has been found in highly aggressive molecular subtypes of breast carcinoma [12]. Significant overexpression of ABCC1 transcript in both pre-chemotherapy (n = 100) and post-chemotherapy (n = 68) tumors compared with adjacent non-neoplastic tissues from breast carcinoma patients and associations of intratumoral transcript levels with tumor grade and expression of estrogen receptor, proliferative marker Ki67, and p53 protein have been recently reported [13].
A high number of single nucleotide polymorphisms (SNPs) in ABCC1 have been identified in different human populations and their haplotypes were examined [14,15]. ABCC1 has high haplotype diversity with significant differences across ethnic groups [16]. Very recently convincing association between rs4148350, rs45511401, and rs246221 SNPs in ABCC1 and risk of febrile neutropenia in breast cancer patients treated by 5fluorouracil, epirubicin and cyclophosphamide (FEC regimen) has been shown [17]. Several studies suggested in vitro functional effects of SNPs in ABCC1. For instance, Gly671Val (dbSNP: rs45511401) SNP located near the nucleotide binding domain 1 (NBD1, Figure 1) which is important for the ATPase activity was associated with reduced levels of ABCC1 transcript [14]. Serine at position 433 (rs60782127) significantly increased the resistance to doxorubicin [18] whereas serine at position 43 (rs41395947) enhanced expression and altered ABCC1 protein trafficking to the plasma membrane [19]. Moreover, several ABCC1 SNPs including Arg723Gln (rs4148356) located between the Walker A and B motifs in NBD1 have been shown to affect the resistance to a number of anticancer drugs [20].
The present study investigated the effect of tagged haplotype of the ABCC1 gene covering NBD1 with adjacent sequences at ABCC1 transcript level in tumor and non-neoplastic tissues from breast cancer patients. In addition, we also addressed the prognostic and predictive significance of genetic variability of ABCC1.

Patients
The study included a total of 540 breast cancer patients (C50 according to ICD-10) of Caucasian origin diagnosed in Motol Faculty Hospital, Institute for the Care for Mother and Child, BIOLAB Praha k.s., and Faculty Hospital Kralovske Vinohrady in Prague, Czech Republic between February 2000 and December 2010 (for study flow diagram see Figure S1 in File S1). Blood samples were available from all patients. Tumor tissue samples were collected during the primary surgery from subgroups of patients. First subgroup of patients (n = 60) underwent preoperative neoadjuvant chemotherapy regimens based on 5-fluorouracil/ anthracyclines/cyclophosphamide (FAC or FEC) and/or taxanes. The second subgroup was treated by adjuvant chemotherapy and/ or hormonal therapy after surgery (n = 89). Paired samples of adjacent non-neoplastic tissues as controls were available from 67 patients. In the whole set, patients with metastatic disease treated by first line palliative therapy were also included (for all treatments see Table S1 in File S1). Collection and pathological processing of tissue samples and retrieval of data was performed as described before [13,21]. Expression of receptors for estrogen (ER) and progesterone (PR) was evaluated as positive when at least 10% of cell nuclei showed staining by routine immunohistochemistry. HER2 (ERBB2, OMIM: 164870) status was defined as positive in samples with immunohistochemical score 2+ or 3+ confirmed by SISH analysis. For expression of the p53 (OMIM: 191170) protein, 50% cut off was used (negative ,50% vs. positive $50%, see Material and Methods S1 and References S1 in File S1). Patients were experimentally divided into groups according to molecular subtypes of their tumors (Luminal A = ER+/HER22 and grade 1 or 2, Luminal B/HER2-= ER+/HER22 and grade 3, Luminal B/HER2+ = ER+/HER2+, HER2+ = ER2/HER2+, and triple negative = ER2/PR2/HER22) according to [22]. All patients after primary chemotherapy and surgery were followed for local or distant relapse or in the case of palliative setting for disease progression by regular visits every 3 months during the first 3 years, twice a year during the next 2 years and yearly then after. During the visits mammography, chest X ray, skeletal survey, and abdominal ultrasound was performed yearly and clinical examination together with tumor markers (CEA and CA 15-3) was performed during every visit. In the case of clinical uncertainty, additional tests and examinations were performed to rule out possible disease relapse or progression.

Ethics statement
All patients were asked to read and sign an informed consent and the study was approved by the Ethical Commission of the National Institute of Public Health in Prague.

DNA extraction
Blood samples were collected during the diagnostic procedures using tubes with K 3 EDTA anticoagulant. Genomic DNA was isolated from human peripheral blood lymphocytes by the standard phenol/chloroform extraction and ethanol precipitation method [23]. DNA samples were stored in aliquots at 220uC prior to analysis.

ABCC1 genotyping
DNA sequence covering coding exons 15-19 (NBD1), interspersed introns, and sequences surrounding both 59-and 39untranslated regions (Chr16:16,076,000-16,091,000, NCBI Build 36.3 version) were analyzed by HaploView v4.2 program [24]. Together nine SNPs tagging common haplotypes at r 2 .0.8 and Role of ABCC1 Genetic Variability in Breast Cancer PLOS ONE | www.plosone.org minor allele frequency (MAF).0.05 in HapMap CEU sample with minimally 75% genotype data were identified. The ABCC1 region containing eight selected SNPs (rs35623, rs4148351, rs35626, rs11075295, rs3851711, rs3888565, rs35625, and rs4148350) was then divided into four regions. Inside of these regions, we also analyzed additional four SNPs (rs35628, rs11866794, rs4148353, and rs4148356). All analyzed SNPs are characterized in Table S2 in File S1. For each region pair of forward and reverse primers with M13 sequence adaptors was designed using the Primer3 software [25]. Oligonucleotide primers were synthesized by Generi Biotech (Hradec Kralove, Czech Republic). Twelve SNPs were then determined by direct sequencing. PCR products were generated using 50 ng of genomic DNA in a 25 ml final volume containing 2.5 ml of 106 reaction buffer consisting of 0.8 (region 1) or 1.6 mM MgCl 2 (regions 2-4), 0.25 mM dNTPs, 0.2 mM of each primer, and 0.5 ml of Taq DNA polymerase, 1 U/ml (all chemicals except for dNTPs from Top-Bio, Vestec, Czech Republic). Primer sequences and optimized conditions for PCR cycling are specified in Table S3 in File S1. The PCR products were resolved and analyzed on 2% agarose gel containing ethidium bromide and visualized by ultraviolet light. All samples containing the PCR products were then sequenced by using the BigDye Terminator v3.1 Cycle Sequencing Kit (Life Technologies) with 5 ng of PCR product and 2 pmol of universal M13 sequencing primer in a 10 ml final reaction volume. PCR conditions for sequencing reactions were as recommended by the producer (Life Technologies). Separate sequencing reaction included a control template pGEM-3Zf(+) under the same conditions as above. Sequencing products were purified by EDTA/sodium acetate/ethanol precipitation. DNA sequencing was performed on Applied Biosystems 31306L Genetic Analyzer and the results were evaluated by Sequencing Analysis Software v5.2 (Life Technologies). About 10% of samples were resequenced with 100% conformity of the results.

ABCC1 gene expression
Total RNA was isolated from frozen tissues, stored, and characterized as described [21]. cDNA was synthesized using 0.5 mg of total RNA and random hexamer primers with the help of RevertAid TM First Strand cDNA Synthesis Kit (MBI Fermentas, Vilnius, Lithuania). Quality of cDNA in terms of DNA contamination was confirmed by PCR amplification of ubiquitin C [26]. Quantitative real-time PCR (qPCR) of ABCC1 and reference genes EIF2B1 (OMIM:606686), MRPL19 (OMIM:611832), IPO8 (OMIM:605600), and UBB (OMIM:191339) was performed in RotorGene 6000 (Corbett Research, Sydney, Australia) as described [13]. Reference genes for data normalization were selected using software programs geNorm (version 3.5) and NormFinder (version 19) (see File S1). The qPCR study design adhered to the MIQE Guidelines (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) [27].

Statistical analyses
The following differences in distribution of genotypes were evaluated: wild type vs. minor allele carrier (dominant model) and rare genotype carrier vs. wild type allele carrier (recessive model).
The additive model was also tested. Haplotypes were evaluated using HaploView software program version 4.2 [24]; phasing of haplotypes prior to a block selection was done using the E-M algorithm and the block selection was based on confidence intervals [28]. Associations between categorized values as genotypes or haplotypes and clinical-pathological data were analyzed using the two-sided Fisher's Exact test. Clinical and pathological

Patients' characteristics
Clinical characteristics of patients are presented in Table 1.

Distribution of genotypes and haplotypes
Twelve SNPs selected with the help of HaploView v4.2 pairwise tagging algorithm of the region including NBD1 and surrounding sequences of ABCC1 were genotyped in 540 breast carcinoma patients. The rate of missing genotype data due to DNA of insufficient quality or quantity did not exceed 3.5% in particular SNPs.

Associations of ABCC1 SNPs and diplotypes with transcript levels
The ABCC1 transcript level was previously assessed in tumors and non-neoplastic control tissues from breast cancer patients [13]. A subset of these patients with complete genotype data was included into this study (n = 149) and associations between genotypes, predicted diplotypes, and transcript levels were analyzed by Mann-Whitney or Kruskal-Wallis tests. Associations of all SNPs and frequent diplotypes with expression levels were analyzed but to retain concise style only significant results are reported ( Table 4 and Figure 3). Tumors from carriers of the wild type genotype in rs35623 or rs35628 expressed significantly lower ABCC1 transcript levels than those with the minor allele (p = 0.003 and p = 0.004, respectively; q = 0.008, both significant; Table 4 and Figures 3A, B). A significant upward trend in the ABCC1 transcript level in the order CT-GT.CC-GT.CC-GG (p = 0.023; q = 0.017, non-significant; Table 4 and Figure 3C) for the rs35626-rs4148351 diplotype was observed. Non-neoplastic control tissues from carriers of the wild type genotype in rs11866794 expressed lower ABCC1 transcript levels than those with the minor allele (p = 0.017; q = 0.004, non-significant; Table 4 and Figure 3D). The rs4148356 SNP was predicted to be benign with a score of 0.014 by PolyPhen-2 and tolerated with a score 0.30 by SIFT programs. From synonymous SNPs, rs35626 was classified as likely to affect binding and linked to expression of a gene target (score 1f), rs35625, and rs11866794 as likely to affect binding (2c) by the Regulome DB program. The rest of SNPs was classified as having minimal binding evidence (4-6; Table S4 in File S1).

Associations between clinical characteristics, therapy outcome, and ABCC1 SNPs and diplotypes
Associations of all SNPs and frequent diplotypes with clinical data were analyzed but to retain concise style only significant results are reported ( Table 5). The ABCC1 SNP rs3888565 was significantly associated with expression of estrogen receptor ( Table 5). Carriers of the AA genotype had more frequently tumors without ER expression than carriers of the G allele (p = 0.003; q = 0.004, significant). Moreover, G allele in this SNP was associated with triple-negative disease exhibiting the worst prognosis of all molecular subtypes of breast carcinoma (p = 0.008; q = 0.004, non-significant). Regarding rs4148350, patients with stages II-IV (advanced disease) or lymph nodes affected by metastasis had a greater incidence of the T allele than those with early stage I or metastasis-free lymph nodes (p = 0.005 and p = 0.028, respectively; q = 0.004 and q = 0.008, both nonsignificant). Similarly, patients with HER2-positive tumors carried more frequently the T allele in rs4148350 than those without HER2 expression (p = 0.014; q = 0.004, non-significant). The T allele in rs4148353 also predisposed patients to tumors with ER expression in comparison with wild type carriers (p = 0.049; q = 0.004, non-significant). On the other hand, tumors of the T allele carriers in respect to rs4148353 were usually HER2-negative (p = 0.001; q = 0.004, significant; Table 5). Advanced stages II-IV, similarly as tumors with grades 2 or 3 occurred more frequently in carriers of the C allele in rs35625 than in those with the wild type TT (p = 0.040, p = 0.029, respectively; q = 0.004 and q = 0.008, both non-significant). Carriers of the C allele in rs3851711 had more frequently tumors of histological type other than ductal and exhibited more frequently triple-negative molec- ular subtype histology than those with the TT genotype (p = 0.040 and p = 0.039, respectively; q = 0.004 and q = 0.008, both nonsignificant).
No association between age at diagnosis, menopausal status, tumor size, expression of progesterone receptor, and p53 and the SNPs followed was found (results not shown). Large tumor size (pT2-4), presence of lymph node metastasis (pN1-3), lack of expression of hormonal receptors (ER and PR), and triple-negative molecular subtype were significant predictors of poor prognosis, i.e. short DFS in the set of chemotherapy-treated patients (p, 0.001, p = 0.001, p = 0.011, p = 0.001, and p = 0.003, respectively). Large tumor size (pT2-4), presence of lymph node metastasis (pN1-3), and lack of expression of PR were significant predictors of poor prognosis, i.e. short DFS in the set of hormonal therapytreated patients (p = 0.001, p,0.001, and p = 0.031, respectively). Chemotherapy-treated patients carrying T allele in the rs4148353 SNP had longer DFS than those with wild type GG genotype in univariate analysis (n = 271, p = 0.043; Figure 4A). On the other hand, hormonal therapy-treated patients with the wild type AA genotype in the rs35628 had longer DFS than patients carrying the G allele (n = 353, p = 0.012; Figure 4B). Multivariate analysis using the Cox regression hazard model with pT, pN, ER, and PR expression, triple-negative molecular subtype, and individual SNPs has not confirmed association with DFS for rs4148353 (n = 252, p = 0.116). However, for rs35628 the association observed in univariate model remained significant in multivariate model with pT, pN, and PR expression (n = 323, p = 0.008). Survival analysis was not corrected for multiple testing.

Discussion
Multidrug resistance frequently causes cancer treatment failure. Numerous in vitro and in vivo data revealed that multidrug resistance is often due to enhanced expression ABC transporters [30]. Thus, in depth analysis of ABC transporters appears inevitable for individualization of treatment.
The multidrug resistance-associated protein 1 encoded by the ABCC1 gene is one of the most studied ABC transporters. Very recently, we demonstrated significant overexpression of ABCC1 transcript in tumors compared to adjacent non-neoplastic tissues from breast cancer patients and suggested its intratumoral levels as potential modifiers of breast carcinoma progression [13]. Another contemporary study has suggested that a high ABCC1 protein expression is a negative prognostic marker, as it has been found in highly aggressive molecular subtypes of breast carcinoma [12]. Despite already accumulated knowledge on ABCC1 there are significant gaps in understanding its role in cancer therapy and prognosis which preclude clinical applications.
NBD1 of ABCC1 contains several functional motifs, ATPbinding site, Walker A/P-loop, Q-loop/lid, ABC transporter signature, Walker B, D-loop, and H-loop/switch (Figure 1). Unlike most ABCC proteins, NBD1 of ABCC1 binds ATP with All SNPs and frequent diplotypes were analyzed but to retain concise style only significant associations are reported. Footnotes: *Analyzed by Mann-Whitney test. The higher is the rank the lower is the normalized expression ABCC1/reference genes. # Result passed FDR analysis for multiple testing [29]. doi:10.1371/journal.pone.0101740.t004 high affinity but has low hydrolytic activity, while the reverse is true of NBD2 [31]. Besides this functional asymmetry, it seems obvious that NBD1 and NBD2 cooperate together and with surrounding transmembrane cytoplasmic loops may influence substrate selectivity and the proper assembly and trafficking of ABCC1 to the plasma membrane [32,33]. Moreover, very recently, virtual screening of X-ray crystal structure of ABCC1 NBD1 [34] revealed that about 5% of the National Cancer Institute compounds possessed lower docking scores than ATP in ABCC1 NBD1 and it has been suggested that the compounds identified may be potential inhibitors of ABCC1 and require further pharmacological analysis [35]. Apparently, the role of ABCC1 as predictive biomarker and potential drug target in human cancers raises further interest. The present study addressed yet unexplored associations between genetic variability in NBD1 and adjacent sequences of ABCC1 and clinical course of breast cancer. Further, it evaluated relations between genotype and phenotype represented by transcript levels in tissues of breast cancer patients. Analogously with the ABCB1/P-glycoprotein [36], in the present study, we have observed associations between genetic variability in ABCC1 and its transcript levels in tissues from breast cancer patients. Carriers of the wild type genotype in rs35623 or rs35628 SNPs had significantly lower ABCC1 levels in their tumors than the rest of patients, suggesting their potential as predictors of treatment outcomes. The association of the rare CT-GT diplotype rs35626-rs4148351 with ABCC1 transcript levels observed in the uncorrected analysis has not passed the FDR test and should be replicated on a larger sample size.
The analysis of associations between ABCC1 SNPs and transcript levels could be confounded by the fact that some patients received pre-operative chemotherapy. Some ABC transporters may be induced by chemotherapy [37]. However, in our study the difference in ABCC1 transcript levels between post-and pre-treatment patients was non-significant (p.0.05).
Most interestingly, the rs35628 SNP significantly influenced DFS of patients treated by hormonal therapy in both univariate and multivariate analysis. Taken together, patients with the wild type genotype AA in rs35628 SNP had lower ABCC1 levels in tumors and better survival rates after hormonal therapy than those with the G allele. Cell lines with overexpression of ABCC1 are resistant to anticancer drugs [30] and high expression of ABCC1 protein was associated with shorter DFS [38]. The role of ABCC1 in the efflux of anticancer drugs has been recently proposed [39,40]. On the other hand, the intratumoral ABCC1 transcript level did not modify DFS of unselected patients (n = 88) or patients stratified according to the therapy type, referring to a more complex phenomenon [13]. We also have not found significant association between transcript and protein levels of ABCC1 (n = 30) in the previous study on independent set of patients [13].
Tamoxifen metabolites endoxifen and 4-hydroxy-tamoxifen are substrates of the ABCB1 transporter in vitro [41], but the association of genetic variation in ABCB1 and tamoxifen effectiveness is unknown. The role of ABCC1/MRP1 in the transport of tamoxifen, its metabolites or aromatase inhibitors also remains vastly unexplored [42] and thus it is currently impossible to draw any conclusive remarks from our observations in hormonally-treated breast cancer patients.
We have also found out that chemotherapy-treated carriers of the T allele in rs4148353 SNP had significantly better DFS than those with the wild type GG genotype. However, this association has not been confirmed by multivariate analysis. The association of rs4148353 with DFS could be modulated by the significant associations of this SNP with ER and mainly HER2 which were shown to be the best predictors of chemotherapy response in breast carcinoma [43]. Patients carrying the T allele had higher frequency of ER-positive or HER2-negative tumors when compared with wild type carriers in the present study and the lack of expression of hormonal receptors or the triple-negative molecular subtype of breast cancer were indeed significant predictors of poor DFS.
The lack of validation study in independent sample set may be seen as limitation of the present study. By searching Catalog of Published Genome-Wide Association Studies at NHGRI (www. genome.gov) and GWAS Central (www.gwascentral.org) we have found no supportive data for associations between ABCC1 SNPs with breast carcinoma survival or therapy response that could support our results. Microarray study that explored associations of transcript levels with SNP markers from the International HapMap Project in lymphoblastoid cells of 57 unrelated CEPH individuals has not observed such association(s) for ABCC1 [48].
Significant associations of non-coding SNPs with expression and clinical phenotype observed by the present study were not confirmed by in silico analyses or additional experimental data. This fact limits the interpretation of the results before complex functional study is completed.
In conclusion, according to our present data, SNPs rs35623 and rs35628 in non-coding regions around NBD1 may modulate ABCC1 transcript levels in breast tumors, thus contributing to a complex pattern of chemotherapy resistance by so far unknown mechanism. Associations of rs35628 and rs4148353 with DFS of breast cancer patients warrant further studies aimed at validation or disqualification of these putative prognostic markers.

Supporting Information
File S1 Contains the following files: Material and Methods S1. References S1. Table S1: Chemotherapy and hormonal therapy regimens. Table S2: Positions of the analyzed SNPs in ABCC1. Table S3: Sequencing primers and PCR conditions for assessment of polymorphisms in NBD1 of ABCC1. Table S4: In silico analysis of functional significance of all studied polymorphisms in NBD1 of ABCC1. Figure S1: Flow diagram of the study. (DOC)