The dyslipidemia-associated SNP on the APOA1/C3/A5 gene cluster predicts post-surgery poor outcome in Taiwanese breast cancer patients: a 10-year follow-up study

Background Post-surgery therapies are given to early-stage breast cancer patients due to the possibility of residual micrometastasis, and optimized by clincopathological parameters such as tumor stage, and hormone receptor/lymph node status. However, current efficacy of post-surgery therapies is unsatisfactory, and may be varied according to unidentified patient genetic factors. Increases of breast cancer occurrence and recurrence have been associated with dyslipidemia, which can attribute to other known risk factors of breast cancer including obesity, diabetes and metabolic syndrome. Thus we reasoned that dyslipidemia-associated nucleotide polymorphisms (SNPs) on the APOA1/C3/A5 gene cluster may predict breast cancer risk and tumor progression. Methods We analyzed the distribution of 5 selected APOA1/C3/A5 SNPs in recruited Taiwanese breast cancer patients (n=223) and healthy controls (n=162). The association of SNP (APOA1 rs670) showing correlation with breast cancer with baseline and follow-up parameters was further examined. Results APOA1 rs670 A allele carriage was higher in breast cancer patients than controls (59.64% vs. 48.77%, p=0.038). The rs670 A allele carrying patients showed less favorable baseline phenotype with positive lymph nodes (G/A: OR=3.32, 95% CI=1.77-6.20, p<0.001; A/A: OR=2.58, 95% CI=1.05-6.32, p=0.039) and negative hormone receptor expression (A/A: OR=4.85, 95%CI=1.83-12.83, p=0.001) in comparison to G/G carriers. Moreover, rs670 A/A carrying patients had higher risks in both tumor recurrence (HR=3.12, 95% CI=1.29-7.56, p=0.012) and mortality (HR=4.36, 95% CI=1.52-12.47, p=0.006) than patients with no A alleles after adjustments for associated baseline parameters. Furthermore, the prognostic effect of rs670 A/A carriage was most evident in lymph node-negative patients, conferring to the highest risks of recurrence (HR=4.98, 95% CI=1.40-17.70, p=0.013) and mortality (HR=9.87, 95%CI=1.60-60.81, p=0.014) than patients with no A alleles. Conclusions APOA1 rs670 A/A carriage showed poor post-surgery prognosis in Taiwanese lymph node-negative breast cancer patients, whose prognosis were considered better and adjuvant treatment might be less stringent according to currently available assessment protocols. Our findings suggest that APOA1 rs670 indicate a post-surgery risk of breast cancer disease progression, and that carriers of this SNP may benefit from more advanced disease monitoring and therapy regimens than the current regular standards. Furthermore, control of lipid homeostasis might protect APOA1 rs670 minor allele carriers from breast cancer occurrence and progression.


(Continued from previous page)
Conclusions: APOA1 rs670 A/A carriage showed poor post-surgery prognosis in Taiwanese lymph node-negative breast cancer patients, whose prognosis were considered better and adjuvant treatment might be less stringent according to currently available assessment protocols. Our findings suggest that APOA1 rs670 indicate a post-surgery risk of breast cancer disease progression, and that carriers of this SNP may benefit from more advanced disease monitoring and therapy regimens than the current regular standards. Furthermore, control of lipid homeostasis might protect APOA1 rs670 minor allele carriers from breast cancer occurrence and progression.
Keywords: APOA1/C3/A5, SNP, APOA1 rs670, Lymph node micrometastasis, Post-surgery, Prognosis Background Breast cancer is currently the leading cause of cancer deaths in females worldwide, and also the second most common cancer after lung cancer [1]. The current therapeutic regimens for operable early-stage breast cancers include endocrine therapy, chemotherapy and radiotherapy. The assessment tools for post-surgery planning are currently based on clinicopathological evaluations, but are unsatisfactory requiring finer adjustments [2]. The risk factors for breast cancer poor prognosis include positive sentinel lymph node metastasis, hormone receptor negativity, larger tumor size, younger age, and menopausal status [2]. In contrast to the Western population, the breast cancer incidence peaks at a younger age in Oriental Asians, which include Taiwanese [3,4]. Nevertheless, the westernized dietary pattern and lifestyle in Taiwan in the last two decades has increased the incidences of metabolic disorders, including dyslipidemia, as well as breast cancer in Taiwanese females [4].
Both cohort and case-control epidemiological observations have demonstrated the association of metabolic syndrome, obesity and diabetes with increased breast cancer risk [5,6]. Dyslipidemia often occur in parallel to obesity and diabetes, and is a component of metabolic syndrome [7]. Dyslipidemia in the forms of hypertriglyceridemia, hypercholesterolemia, and low high density lipoprotein-cholesterol (HDL-C) have been observed in breast cancer patients of several ethnic groups [8][9][10][11]. Moreover, in large cohort studies dyslipidemia is associated with increased breast cancer risk and poor prognosis [12][13][14]. The intake of lipid-lowering drugs or agents in women was associated with reduced breast cancer occurrence and recurrence risk [15][16][17]. Consistently, mammary tumor growth and metastasis was accelerated in a hyperlipidemic murine model [18].
In this study, we first tested the correlation of 5 selected well-known SNPs on the APOA1/C3/A5 gene cluster with breast cancer in a case-control manner. Furthermore, we analyzed the effect of the SNPs associated with greatest breast cancer risk on patient baseline tumor characteristics, and their post-surgery outcome after a mean follow-up period of 10.4 years. Finally, we specifically examined the prognostic value of APOA1 rs670 in lymph node-negative breast cancer patients.

Patient recruitment
Taiwanese female breast cancer patients (n=223, 48.4± 10.2 years, ranged 29-75 years) received surgical intervention plus axillary/sentinel lymph node dissection during 1999-2005 at National Cheng Kung University Hospital (NCKUH) and Tainan Hospital, and were followed-up to November 2012. Healthy female controls (n=162, 43.0 ± 8.8 years, ranged 19.0-69.0 years.) were also recruited. This is a continuation study of Hsiao et al., 2004 [30]. Bodyweight (kg) and body height (m) were measured at the time of mastectomy, and used for calculating body mass index (BMI, kg m -2 ). The diagnosis was confirmed by histological examinations of mammary and node specimens. Estrogen receptor (ER) and progesteron receptor (PR) expressions in primary breast tumor were determined as described previously [31]. Individuals with at least one first-degree or second-degree female relatives affected by breast cancer were considered to have a family history. This information was obtained by interview with patients and their family members. The recruited patients received tamoxifen (TAM) (n=92), TAM and chemotherapy (n=56), TAM and radiotherapy (n=10), and triple-therapy (n=55). This study received approval from the local institutional review board (NCKUH IRB) and signed informed consent was obtained from the patients.
Genomic DNA extraction, SNP genotyping and linkage analysis Genomic DNA was extracted from white blood cells using the Puregene DNA Isolation Kit (Gentra Systems, Minneapolis, MN, USA). According to criteria described in Hsu et al. [29], we selected five SNPs on the APOA1/ C3/A4/A5 gene cluster that follows: APOA1 rs670, APOC3 rs2854116, APOC3 rs2854117, APOA5 rs662799 and APOA5 rs2075291. The SNP genotypes were determined using commercial real-time PCR primer and probes from Applied Biosystems (ABI, Foster City, CA, USA) (APOA5 rs662799 and APOA5 rs2075291) and TIB MOBIOL (Berlin, Germany) (APOA1 rs670, APOC3 rs2854116 and APOC3 rs2854117). Fluorescence data from real-time PCRs were collected by a Step-One-Plus Sequence Detection System (ABI) or LightCycler 480 (Roche, St. Louis, MO, USA). Haploview [32] was used for the analysis of SNP linkage disequilibrium, and Hardy-Weinberg equilibrium, and haplotype analysis. SNP linkage disequilibrium test results with logarithm of odds (LOD) socores ≥2 and pair-wise D'>0.80 were considered as significant linkage.

Statistical analysis
The association of SNP with breast cancer risk, baseline clinical parameters and post-surgery progression was analyzed by Chi-squared test. The odds ratio for unfavorable baseline characteristics and events in post-surgery progression was analyzed by binary or multi-nominal logistic analysis. The differences in BMI, age, and mean years in survival were analyzed by one-way ANOVA. The survival curves of APOA1 rs670 genotype carriers were plotted by Kaplan-Meier analysis. The hazard ratio for overall and recurrence-free survival was calculated by Cox proportional-hazards regression analysis. Possible confounders including unfavorable baseline characteristics, age and BMI were adjusted for in regression analysis. Statistical analysis was performed using SPSS 13 (SPSS Inc., Chicago, DE, USA). In all cases, p-values ≤0.05 were considered statistically significant.

Demographic characteristics of recruited breast cancer patients
The baseline characteristics of the recruited breast cancer patients are shown in Table 1. The mean BMI of breast cancer patients at baseline was 23.36±3.79 kg m -2 . The major tumor type of recruited patients was infiltrating ductal carcinoma (86.55%), and the tumor occurrence side was evenly distributed (right breast: 44.84%,  (Table 1).
ApoA1 rs670 was associated with increased breast cancer risk The distributions of the tested SNPs (APOA1 rs670, APOC3 rs2854116, APOC3 rs2854117, APOA5 rs662799 and APOA5 rs2075291) fitted the Hardy-Weinberg equilibrium in both breast cancer patients and healthy controls (Additional file 1). Different linkage patterns of the tested SNPs were observed between breast cancer patients (Additional file 2). Significant linkage (LOD≥ 2, D'> 0.80) was observed among APOA1 rs670, APOC3 rs2854116 and APOC3 rs2854117 (D'=0.84-0.88), and between APOA5 rs662799 and APOA5 rs2075291 in healthy controls (D'=0.86). In contrast, the linkage among APOA1 rs670, APOC3 rs2854116 and APOC3 rs2854117 was lost in breast cancer patients, while significant linkage was only observed between the two APOA5 SNPs (D'=1.00). The frequency of APOA1 rs670 A allele (G/A + A/A) carriers in breast cancer patients was significantly higher than that in healthy controls (59.64% vs. 48.77%, p=0.038, Additional file 3). On the other hand, the genotype and allele frequencies of all other SNPs were comparable between breast cancer patients and healthy controls. The post-surgery therapies received by the different APOA1 rs670 genotype carriers were comparable (p=0.151). Therefore we tested the contribution of APOA1 rs670 to breast cancer in recruited patients cross-sectionally at baseline and longitudinally at follow-up.
APOA1 rs670 was associated with sentinel lymph node-positivity and tumor ER/PR negativity at baseline Upon analysis of APOA1 rs670 against breast tumor clinical parameters, we found the mean age and BMI of different APOA1 rs670 genotype carriers were comparable at baseline (Table 2). In contrast, APOA1 rs670 was associated with nodal involvement as well as tumor hormone receptor expression (Table 2 and 3). Baseline lymph node positivity was more frequent in patients carrying APOA1 rs670 G/A or A/A genotypes than in those carrying G/G genotype (G/A: 51.89%, A/A: 48.15%, G/G: 24.44%, p=0.002) ( Table 2). Moreover, APOA1 rs670 G/ A or A/A carriage also significantly increased the frequency of having tumors double negative for ER/PR expression than G/G carriage in patients (G/A: 24.53%, A/ A: 48.15%, G/G: 14.44%, p=0.006) ( Table 2). In contrast, APOA1 rs670 was not associated with tumor occurrence side, type, size, and stage or a family history ( Table 2). The odds of being lymph node-positive in APOA1 rs670 G/A carrying patients (OR=3.32, 95% CI=1.77-6.20, p< 0.001) and APOA1 rs670 A/A carriers (OR=2.58, 95% CI=1.05-6.32, p=0.039) was higher than G/G carriers (Table 3). On the other hand, APOA1 rs670 A/A carriers have significantly higher odds in having ER/PR doublenegative tumors than their G/G counterparts (OR=4.85, 95% CI=1.83-12.83, p=0.001) ( Table 3).
APOA1 rs670 A allele carriage was associated with poor post-surgery outcomes Higher proportions of APOA1 rs670 G/A and A/A carrying breast cancer patients developed recurrence or death than G/G carriers ( Table 4). The higher incidences of recurrence in APOA1 rs670 G/A and A/A carriers translate higher risks to 2.07-fold (95% CI=1.03-4.14, p=0.041) and 3.44-fold (95% CI=1.33-8.86, p=0.011) risk as compared with G/G carriers, respectively (Table 5). However, the increase in risk of recurrence in APOA1 rs670 G/A carriers may be confounded by their increased odds in worse baseline phenotype. After adjustment for lymph node status, the statistical significance remained in A/A carriers (OR=2.83, 95% CI, 1.05-7.61, p=0.040), while that in G/A carriers were diminished (p=0.238) ( Table 5). The risk of death was significantly increased in APOA1 rs670 A/A carriers (OR=6.03, 95% CI=2.08-17.51, p<0.001) and this persisted after adjustments for baseline confounders of lymph node (p=0.007) and ER/PR (p=0.010) status, while no significant risk was observed in G/A carriers with or without adjustments ( Table 5). The mean years in recurrence-free survival and overall survival of APOA1 rs670 A/A carriers were significantly shorter than that of G/A and G/G counterparts (all p<0.001) ( Table 4). Despite node-positivity, ER/PR-negativity, combined therapy, and APOA1 rs670 A/A carriage showed increased recurrence and mortality risk in unadjusted models ( Table 6). The increased risk of recurrence and mortality observed in APOA1 rs670 A/A carriers remained significant after adjustments for lymph node involvement, ER/PR status, age, BMI, or post-surgery adjuvant (Additional file 4). When testing the prognostic effects of lymph node, ER/PR, postsurgery adjuvant and APOA1 rs670 A/A carrying status in an adjusted model, the significance of lymph node and APOA1 rs670 A/A was stronger than of ER/PR and postsurgery adjuvant (Table 6). Lymph node positivity showed  (Table 6).
Therefore, to specifically rule out confounding from baseline lymph node involvement in predicting prognosis, we tested the prognostic effect of APOA1 rs670 A/A carriage in baseline lymph node status-stratified patient groups. The prognostic effect of APOA1 rs670 A/A carriage was only observed in the lymph node-negative patients (n=121, Table 7 and Figure 1) with higher hazard risks than non-stratified group (n=223, Table 4 and   (Table 7).

Discussion
Dyslipidemia is a known complication of TAM-treatment in breast cancer patients, and increases risk of coronaryartery diseases in patients [33]. Of equal importance, dyslipidemia has been observed in breast cancer patients prior to treatment, and this correlation is supported by epidemiological studies [8][9][10][11]. Concomitantly, usage of lipid-lowering drugs such as statin and niacin has been shown to associate with decreased breast cancer recurrence  and risk [15][16][17]. SNPs on the APOA1/C3/A5 gene cluster, which is involved in lipid metabolism, are highly associated with dyslipidemia, metabolic syndrome, insulin resistance and responsiveness to relative treatments [19]. However, the contribution of SNPs on the APOA1/C3/A5 gene cluster to breast cancer is yet to be defined. In this study, we tested the effects of dyslipidemia-associated SNPs on the APOA1/C3/A5 gene cluster: APOA1 rs670, APOC3 rs2854116, APOC3 rs2854117, APOA5 rs662799 and APOA5 rs2075291 on breast cancer progression in a Taiwanese patient group with mainly operable early-stage tumors, and a 10-year follow-up interval. We showed that only APOA1 rs670 out of the 5 tested SNPs was correlated to breast cancer, lymph node-positivity, ER/PR doublenegativity at baseline. Furthermore, carriers of both minor alleles on APOA1 rs670 had shortest survival time and highest risk in disease progression independent of baseline characteristics. Moreover, the prognostic value of APOA1 rs670 A/A carriage in the worse post-surgery outcomes was most evident in lymph node-negative patients. The rs670 SNP contains a G-to-A substitution at 75bp upstream of APOA1 transcriptional start site, and is within a MSPI restriction enzyme recognition [19]. A cross-sectional study focusing on the association of APOA1 SNPs with breast cancer and patient phenotype reported that the -75G/A polymorphism correlated with breast cancer risk at baseline [34]. However, the slight discrepancy in associations of -75G/A polymorphism with ER status observed by us and Hamrita et al. is likely due to ethnic (Taiwanese vs. Tunisian) and age differences (62.7% vs. 38.9% <50years) [34]. Nevertheless, in our longitudinal study we found that APOA1 rs670 predicted worse outcome after adjustment for ER/PR status, indicating alternative contributions from APOA1 rs670 in cancer progression. ApoA1 is the structural protein of HDL, and interacts with lecithin cholesterol acyl transferase, which controls the limiting steps in HDL maturation and therefore reverse cholesterol transport [19]. The minor allele of APOA1 rs670 has been correlated to altered HDL-C, diabetes and coronaryartery disease: lower levels of HDL-C was observed in Northern Indians [20], and correlated with severe forms of cardiovascular diseases in Northern Indians, Caucasians in Spain and Australia [20,24,25]. Moreover, APOA1 rs670 A/A Spanish carriers had a higher risk of diabetes than their non-A/A counterparts, though their HDL-C did not differ significantly [21,22]. As plasma apoA1 level is regulated by estrogen and thus TAM treatments, the timing of plasma apoA1 measurement in breast cancer patients is critical. However, we did not have access to TAM treatment-naïve plasma of the patients recruited in this study. Nevertheless, the disease phenotype associated APOA1 rs670 remains significant without HDL-C differences, as has been reported in a number of studies on metabolic/cardiovascular diseases [21,24,25].
The significance of plasma HDL-C levels in breast cancer has long been investigated and debated [35]. The current held view is that higher HDL-C levels, which are preventive for cardiovascular events, are also protective for breast cancer. The majority of reports supporting this view were of case-control design, and found lower total or lipid-rich HDL, or higher total cholesterol/HDL-C ratio in breast cancer patients as compared with noncancerous or other cancer controls in various ethnic groups [8,10,11]. A 17.2-year prospective cohort study carried out in Norway found that naive females with the lowest quartile of HDL-C had highest risk of later developing breast cancer [14]. Moreover, the same Norwegian group found the breast cancer patients with highest total cholesterol or lowest plasma HDL-C levels had highest risk in mortality [13]. The female breast is an active site of both lipid uptake and lipid secretion during periods of milk secretion, thus it is readily acceptable that apoA1 is present in human milk [36]. Supportively, apoA1 was also identified in the breast cancer tissues and ex vivo cultured medium. ApoA1 was found in breast tumors, and its amount was correlated positively with chemotherapy resistance in malignant tumors [37]. In contrast, breast tumors later found responsive to chemotherapy secreted higher amount of apoA1 than tumors nonresponsive to chemotherapy during short term ex vivo culture, while non-cancerous tissues secreted highest amount of apoA1 [38]. Furthermore, a successful reduction of breast cancer growth in mice by vitamin D treatment was accompanied by decrease of apoA1 production in tumor tissues [39]. It is at this stage unknown how much apoA1 in circulation is contributed by the mammary tissue, but the retention of apoA1 in breast cancer tissues is in accordance with the observation of lower plasma HDL/apoA1 in breast cancer patients. Though the association of plasma lipid profiles and breast cancer incidence or progression is strongly and continuously suggested, the molecular mechanisms of HDL or apoA1 in promoting breast cancer risk are still unclear. Plasma apoA1/lipid-poor HDL bind to ATP-binding cassette (ABC) lipid transporters during maturation, and the intracellular cholesterol is exported to load onto HDL, facilitating reverse cholesterol transport [40]. Several members of the ABC family were found in mammary tissues, but only those associated with drug resistance were increased in breast cancers as compared with normal tissues [41]. In contrast, ABC transporters involved in lipid transport, including ABC-A1 and ABC-A3, were decreased in breast cancer tissues as compared with healthy mammary glands [42]. In vitro studies found that over-expression of ABC-G1 in breast cancer cell line induces cellular cholesterol efflux and apoptosis [43]. Moreover, ABCA-or ABCG-HDL binding concomitantly induces intracellular signaling and suppresses inflammatory reactions, which is known to be a key factor in tumorigenesis. Binding of ABCA or ABCG by apoA1 induces the recruitment of phosphrylated STAT-3 and JAK2, and thus suppresses the inflammatory pathways including TL4-signalling [40]. Furthermore, ABCA induces Cdc42 and PAK-1 signaling, which in turn are regulative for breast cancer ER/PR expression in vitro and in vivo [44,45]. Therefore, the altered HDL/apoA1 levels observed in APOA1 rs670 A carriers may participate in the tumorigenesis, survival, ER/PR status of breast cancer cells. In a previous published study, we found that genetic and intra-tumoral ER polymorphisms were correlated with breast cancer in Taiwanese females [30,31]. It would be of interest how APOA1 rs670 and HDL-C interact with ER polymorphisms both at a systemic and local (mammary tissue) level in these patients.
Clinically, our finding of worst outcome in APOA1 rs670 A/A breast cancer patients as compared to their non-A/A counterparts provides pivotal implications in personalized treatment regimen and therapeutic strategies. The median recurrence-free and overall survival years of APOA1 rs670 A/A carrying breast cancer patients were both longer than the standard chemotherapy duration (~1 year) and the advised 5-year limit for continuous TAM usage. As the patients recruited in this study were non-obese and of early tumor-staging, the administration of adjuvant therapy may be less intensive. Furthermore, the prognostic value of APOA1 rs670 was even more evident in the lymph nodenegative patients, who were given less adjuvant therapies under the current assessment guidelines. Thus APOA1 rs670 A/A carriers may benefit from advanced therapies. Recent studies demonstrated that administration of apoA1 mimetic peptides reduced the development of cancer in murine models [46,47]. Furthermore, a large cohort study on the effect of post-diagnosis statin use in breast cancer females reported a lowered risk in cancer recurrence [48]. These reports suggest that raising HDL-C or apoA1 may be preventive or therapeutic for breast cancers, especially in the APOA1 rs670 A/A carrying patients. On the other hand, due to the low plasma HDL-C and tumor ER/PR negativity in the APOA1 rs670 A/A carriers, the usage of TAM in these breast cancer patients may be controversial and require further evaluation.