Genetic polymorphisms in RNA binding proteins contribute to breast cancer survival



The RNA-binding proteins TTP and HuR control expression of numerous genes associated with breast cancer pathogenesis by regulating mRNA stability. However, the role of genetic variation in TTP (ZFP36) and HuR (ELAVL1) genes is unknown in breast cancer prognosis. A total of 251 breast cancer patients (170 Caucasians and 81 African–Americans) were enrolled and followed up from 2001 to 2011 (or until death). Genotyping was performed for 10 SNPs in ZFP36 and 7 in ELAVL1 genes. On comparing both races with one another, significant differences were found for clinical and genetic variables. The influence of genetic polymorphisms on survival was analyzed by using Cox-regression, Kaplan-Meier analysis and the log-rank test. Univariate (Kaplan-Meier/Cox-regression) and multivariate (Cox-regression) analysis showed that the TTP gene polymorphism ZFP36*2 A > G was significantly associated with poor prognosis of Caucasian patients (HR = 2.03; 95% CI = 1.09–3.76; p = 0.025; log-rank p = 0.022). None of the haplotypes, but presence of more than six risk genotypes in Caucasian patients, was significantly associated with poor prognosis (HR=2.42; 95% CI = 1.17–4.99; p = 0.017; log-rank p = 0.007). The effect of ZFP36*2 A > G on gene expression was evaluated from patients' tissue samples. Both TTP mRNA and protein expression was significantly decreased in ZFP36*2 G allele carriers compared to A allele homozygotes. Conversely, upregulation of the TTP-target gene COX-2 was observed ZFP36*2 G allele carriers. Through its ability to attenuate TTP gene expression, the ZFP36*2 A > G gene polymorphism has appeared as a novel prognostic breast cancer marker in Caucasian patients.

Breast cancer alone is expected to account for 30% of all new cancer cases among women and 39,520 deaths in 2011 in the United States.1 Breast cancer is a multi-factorial disease with several environmental and genetic factors contributing to its occurrence and progression,2 with ∼28% of familial cases attributed to mutations in breast cancer susceptibility loci.3 In addition, several reports show that the breast cancer incidence, progression and mortality vary between Caucasians and African–Americans, with heightened incidence and lower mortality observed in Caucasian than African–Americans patients.1, 4–6 These observations indicate the need to identify novel genetic factors that can contribute to the occurrence and progression as well as race-specific differential prognosis of breast cancer.

A critical point in the regulation of many pro-inflammatory cytokines, growth factors and proto-oncogenes occurs through post-transcriptional mechanisms that regulate mRNA degradation.7 A prominent cis-acting RNA element present in a majority of these cancer-associated transcripts is the adenylate- and uridylate (AU)-rich element (ARE) contained within the mRNA 3′-untranslated region (3′UTR).8 The importance of this particular RNA element is evident, since estimates ranging from 8 to 16% of all human protein-coding genes contain a 3′UTR ARE sequence.9, 10 AREs mediate their regulatory function through their association with RNA-binding proteins that display high affinity for AREs. The best studied ARE-binding proteins can promote rapid mRNA decay, mRNA stabilization, or translational silencing.7 Through these mechanisms, ARE-binding proteins exhibit wide-ranging effects on gene expression, since a single ARE-binding protein can target multiple distinct transcripts.

The ARE-binding proteins TTP (Tristetraprolin; ZFP36) and HuR (Hu antigen R; ELAVL1) regulate gene expression through opposing post-transcriptional activities. TTP protein is a member of a small family of tandem Cys3His zinc finger proteins and promotes rapid decay of ARE-containing mRNAs.11, 12 In contrast, HuR protein is a ubiquitously expressed member of the ELAV-like family of RNA-binding proteins and can function to stabilize ARE-containing mRNAs when overexpressed in cells.13–15 In the context of cancer, HuR has been demonstrated to promote expression of many breast cancer-related genes including COX-2, VEGF, HIF1α, TSP1, ERα, IL-8, Cyclin D1, Cyclin E1, MMP-9 and BRCA-1.16–22 Whereas, TTP has been shown to promote downregulation of various cytokines (e.g., TNFα, IL-3, IL-6, IL-8, IL-10, IL-12, IL-23 and GM-CSF) and oncogenes and growth factors (e.g., COX-2, VEGF, Cyclin D1, uPA, uPAR, MMP-1 and c-Myc).23–34

A critical feature impacting the ability of TTP and HuR to function properly occurs through alterations in their respective expression. In various cancer types, including breast cancer, TTP expression is commonly lost and HuR levels are commonly upregulated in tumor tissue.12, 13 Through these combined defects, aberrant mRNA stabilization of ARE-containing mRNAs can occur in breast cancer cells leading to overexpression of growth-promoting genes. However, the genetic factors contributing to the loss of TTP and HuR overexpression in breast cancer are not understood. In this study, we have examined novel genetic polymorphisms in ZFP36 and ELAVL1 genes and determined their possible associations with breast cancer prognosis in two native populations of United States. The genetic variant ZFP36*2 A > G was identified as a marker of poorer overall survival in Caucasian breast cancer patients, with the G allele attenuating TTP gene expression. These findings indicate a causal role of this SNP in prognosis of breast cancer through the suppression of TTP expression and thus allowing for pathogenic gene overexpression during tumor development.

Material and Methods


This study consisted of 251 histologically confirmed primary breast cancer patients (170 Caucasian and 81 African–American). Patient recruitment and demographic/clinical data retrieval was accomplished with University of South Carolina Cancer Research Center Biorepository in collaboration with Palmetto Health Tissue Bank, Columbia, SC. Patients who had received their treatment in Palmetto Health Center during the period from 2001 to 2005 and followed up until March, 2011 were included. Patients were followed at 6-month intervals through Palmetto Health Tumor Bank registry from the time of enrollment until the end of study or the patients' final outcomes (death). All-cause deaths were considered as events for survival analysis. Study approval was obtained from the Institutional Review Board of University of South Carolina.

Available clinicopathological data include: (i) patient-related characteristics (e.g., age, race, family history, tobacco/alcohol intake habit), (ii) clinical follow-up (e.g., treatment regimen, overall survival) and (iii) tumor-based properties (e.g., side of paired organ, pathology, grade, stage, size, tumor marker status), along with tumor marker status including estrogen/progesterone receptor (ER/PR) and human epidermal growth factor receptor (HER)-2/neu. All patients enrolled underwent surgical treatment with or without systemic treatments (including chemo/radio/hormone therapies); delays were reported during diagnosis to first surgical/systemic treatments. The impact of these delays upon patient survival was calculated by adding “delay in surgical treatment” and “delay in systemic treatment” to yield “total delay.”

DNA extraction and genotyping

Human breast tumors and histologically normal tissue were obtained from surgical remnants through the University of South Carolina Cancer Research Center Biorepository. Tissue was snap-frozen in liquid nitrogen and kept at −80°C until processed. Blood was collected from 25 patients (17 Caucasian and 8 African–American), for those the tissue was not available. All patients were informed and had provided written consent. Genomic DNA was extracted from 50 mg of tissue samples (178 histologically normal and 48 tumor samples) and blood samples using Qiagen DNA mini kit according to the vendor's protocol (Qiagen, Valencia, CA) and quantitated using a NanoDrop analyzer (Thermo Scientific, Wilmington, DE).

Genotyping of ZFP36 and ELAVL1 SNPs involved PCR amplification followed by restriction fragment length polymorphism (PCR-RFLP) and/or DNA sequencing. Details for genotyping primers and restriction enzymes used are given in Supporting Information Table 1. As a quality control measure, 5% of cases from each genotype that were assayed by PCR-RFLP were randomly selected for sequencing and the results were in 100% of concordance.

RNA extraction and qPCR

Total RNA was isolated from 50 mg of histologically normal breast tissue samples using Trizol reagent (Invitrogen, Carlsbad, CA). Complementary DNA (cDNA) synthesis was performed using 1 μg of total RNA in combination with oligo(dT) and Improm-II reverse transcriptase (Promega, Madison, WI). Real-time PCR (qPCR) analysis was performed as described25 using Taqman probes for TTP (ZFP36), COX-2 (PTGS2) and GAPDH purchased from Applied Biosystems (Foster City, CA) using the 7300 PCR Assay System (Applied Biosystems); GAPDH was used as control for normalization. Mean of fold changes for all the genotypes were calculated and compared by independent two-sample t-test.

Protein analysis

Western blots were performed as described25 using a polyclonal anti-TTP antibody (ab36558; Abcam, Cambridge, MA). Blots were stripped and then probed with β-actin antibody (Clone C4; MP Biomedicals, Aurora, OH). Detection and quantitation of blots were carried out as previously described.25 Cell lysates (50 μg/sample) were obtained from Caucasian normal breast tissue samples (ZFP36*2 AA and GG genotypes). Tissue was homogenized in M-PER mammalian protein extraction reagent (Thermo Scientific, Wilmington, DE) supplemented with protease inhibitors (Sigma, 50X protease inhibitor cocktail). Samples were dounced in microcentrifuge tubes, kept on ice for 30 min and centrifuged at 13,000 rpm for 30 min at 4°C.

Statistical analysis

A comparison between two populations for different variables (clinical, genetic and exposure with environmental risk factors) were performed by cross tabulation and chi-square test. Mean age of onset and total delay in treatment were compared through independent two-sample t-test. Demographic and clinical characteristics of patients were stated as percentages or summary measures. The primary outcome for this study was overall survival (OS) which was estimated using the Kaplan-Meier method. A log-rank test was used to assess the association between the factors and OS. Univariate Cox's-regression analysis was used to assess the association between each potential prognostic factor and OS. Factors found to be relatively significant (p < 0.1) in the univariate analysis were included in the multivariate Cox's proportional hazards regression model to evaluate the effect of different variables on OS with adjustments for age and known prognostic factors of tumor. The relative risk (hazard ratio [HR]) and 95% CI were calculated from the Cox model for all significant predictors from cancer diagnosis to the end point of study (event). Analyses were also conducted after stratifying the data by cancer prognostic factors to examine the potential interactive effects. A two-tailed p-value of < 0.05 was considered significant. Due to the exploratory nature of this study, no attempt was made to correct for multiplicity of analyses and nominal p values were reported.

Statistical tests for survival analyses were performed using SPSS software version 15.0 (SPSS, Chicago, IL). Haplotypes were constructed, linkage disequilibria were measured and D′ values were calculated to measure indices of linkage disequilibrium (LD) using SNPAnalyzer Version 1.0 (ISTECH). Haplotypes were compared between dichotomized patients (with an OS time of 5 or less years versus patients with an OS time of more than 5 years).

To maintain quality control, Levene's test for equality of variance was performed before the comparison of means to assess the assumption of the equality of variances in different samples. In addition, the “proportional hazard model” assumption by “log-minus-log” survival plot for Cox-regression was evaluated and found that survival lines do not intersect indicating that the “proportional hazard assumption” was satisfied and therefore this study was not subjected to time dependent correlation for Cox regression to analyze the data.


ARE: adenylate- and uridylate-rich element; HR: hazard ratio; htSNP: haplotype-tag SNP; HuR: Hu antigen R; LD: linkage disequilibrium; OS: overall survival; qPCR: real-time PCR; RFLP: restriction fragment length polymorphism; SNP: single nucleotide polymorphism; TTP: tristetraprolin; 3′UTR: 3′-untranslated region


Survival analysis and comparison of clinical characteristics between Caucasian and African–American breast cancer patients

The distribution of demographic and clinical characteristics in breast cancer patients are summarized in Table 1. There were 170 Caucasian patients and 81 African–American patients and the mean age of onset was significantly higher in Caucasian than African–American breast cancer patients (60.42 years vs. 52.54 years, respectively; p = 1.80 × 10–6). However, we found a significant higher mean of total delay in treatment in African–American than Caucasian breast cancer patients (p = 0.009). Even though the status of survival (live vs. dead) and 5-year survival (≤5 year survival vs. >5 year survival) was similar between both of the populations, the median survival was poorer in African–American than Caucasian breast cancer patients (116 months vs. 124 months). In addition, various other factors such as higher tumor grade, ER/PR negativity, eligibility for chemotherapy, non-eligibility for hormone therapy, less frequency of familial cancer, nondrinking habit of alcohol and higher Elston histological score were significantly more prevalent in African–American than Caucasian breast cancer patients (Table 1).

Table 1. Univariate survival analysis of clinical characteristics in Caucasian (CA) and African–American (AA) breast cancer patients and comparision between both races
inline image

Patients' survival was taken as continuous variable in the analysis with clinical characteristics (log-rank test); however, for survival analysis of haplotypes dichotomous survival data was used. “Higher extent of disease at diagnosis,” “higher tumor grade,” and “higher Elston's score” were found to impart significant negative effect on survival of breast cancer patients in both Caucasian and African–American populations. Whereas, “higher AJCC staging,” “ER negativity,” “no hormone therapy” and “no radio therapy” variables were significantly associated with poor survival in Caucasian breast cancer patients (Table 1).

ZFP36 (TTP) and ELAVL1 (HuR) gene polymorphisms in Caucasian and African–American breast cancer patients and their association with survival outcome

In this study, ZFP36 and ELAVL1 SNPs were selected based on their minor allele frequencies (MAFs) in normal Caucasian and African–American populations according to dbSNP, SNP's for which stratified data was unavailable, global MAFs were used. Based on this, three ZFP36 gene polymorphisms and seven ELAVL1 gene polymorphisms that had >5% variant allelic frequency in corresponding control population were chosen (Supporting Information Table 1) and examined in genomic DNA extracted from breast tumor and histologically normal tissue obtained from surgical remnants; genomic DNA extracted from blood samples was used in 25 cases where tissue was not available. To our knowledge, there are no reports to indicate these SNPs result as a tumor-associated somatic mutation and DNA sequencing of PCR products derived from tumor tissue found no tumor-associated somatic mutations to occur in the ZFP36*2 polymorphic site or the surrounding sequence constituting the restriction site (data not shown).

The distributions of ZFP36 and ELAVL1 genotypes of all selected SNPs were consistent with Hardy-Weinberg equilibrium in both Caucasian and African–American breast cancer patients, except for ELAVL1 rs35986520 genotypes in Caucasian breast cancer patients. Genotypic frequencies of all selected SNPs are shown in Table 2. We compared genotypic frequencies between the two populations studied to see the ethnic variability for selected SNPs and found that distribution of genotypes for most of the SNPs (rs251864, rs17879933, rs12983784, rs14394, rs12985234 and rs2042920) was significantly different in both populations (Table 2). One of the SNPs in ELAVL1 (rs74369359) was detected to be monomorphic in all Caucasian and African–American breast cancer patients. Genotypic frequencies of one ZFP36*8 (rs3746083) and two ELAVL1 gene polymorphisms (rs35986520 and rs10402477) were not significantly different in Caucasian or African–American breast cancer patients.

Table 2. Distribution of ZFP36 (TTP) and ELAVL1 (HuR) gene polymorphisms in Caucasian (CA) and African–American (AA) breast cancer patients and univariate survival analysis
inline image

To check the independent effect of each SNP on survival of breast cancer patients, we performed log-rank test and estimated hazard for death using univariate Cox regression analysis. Since the frequencies of homozygous variants were low, especially with final outcome of disease, minor allele-containing genotypes were grouped according to the dominant model and the data was analyzed according to both log-additive as well as dominant models (Table 2). Out of all 10 SNPs evaluated in ZFP36 and ELAVL1 genes, only one ZFP36 gene polymorphism ZFP36*2 (rs251864) was found to be significantly associated with poor survival of breast cancer patients in Caucasian population but not with African–American population. As shown in Figure 1a and Table 2, Caucasian breast cancer patients who are carriers of ZFP36*2 G alleles (AG + GG) were found to be at a two-fold more hazard for death than those carrying AA genotypes (HR = 2.03; 95% CI = 1.09–3.76; p = 0.025; log-rank p = 0.022). This in contrast to African–American breast cancer patients where this effect of the presence of ZFP36*2 G allele was not observed to impact overall survival (HR = 1.20; 95% CI = 0.45–3.23; p = 0.711; log-rank p = 0.710; Figure 1b).

Figure 1.

Kaplan-Meier survival curves in Caucasian and African–American breast cancer patients according to ZFP36*2 A > G genotypes: AA versus AG + GG genotype. Vertical ticks show censored cases and each step down represents an event (death). (a) Survival curves for Caucasian breast cancer patients. (b) Survival curves for African–American breast cancer patients. Median survivals could not be calculated for African–American breast cancer patients so mean survivals are indicated.

Multivariate analysis for survival outcome

Based on the univariate survival analysis indicating the presence of multiple factors which could affect patient survival, multivariate Cox regression analysis was employed to identify important factors associated with overall survival in Caucasian patients (Table 3). Significant modulators of survival arrived through multivariate analysis were ZFP36*2 A > G gene polymorphism, age of disease diagnosis, tumor grade and AJCC staging and hormone therapy received. Furthermore, interactions between all significant variables were performed but none were detected as significant, indicating that these are independent prognostic factors for breast cancer in Caucasian patients (Table 3). When performed in African–American breast cancer patients, multivariate Cox regression did not identify any factors to be significantly associated with patient survival (data not shown).

Table 3. Multivariate survival analysis for Caucasian breast cancer patients
inline image

ZFP36*2 A > G gene polymorphism and its effect on gene expression

ZFP36*2 polymorphism exists within the promoter region of TTP and the presence of the minor G allele can inhibit promoter activity,35 suggesting that ZFP36*2 genetic variation could be a factor contributing to the loss of TTP expression. To assess this, normal tissue of Caucasian breast cancer patients were genotyped for ZFP36*2, and TTP protein and mRNA levels were assayed. As shown in Figure 2a, TTP protein was detected in three per four samples with the ZFP36*2 AA genotype, whereas limited expression was observed in tissues bearing the ZFP36*2 GG genotype. In agreement, there was a significant decrease in TTP mRNA in tissue that correlated with the presence of the G allele. Heterozygote carriers for the ZFP36*2 AG genotype and homozygotes for the ZFP36*2 GG genotype were found to express 0.41- and 0.31-fold less TTP mRNA compared with the ZFP36*2 AA genotypes, respectively (Fig. 2b).

Figure 2.

Relative TTP and COX-2 expression among different genotypes of ZFP36*2 A > G polymorphism. (a) Protein lysates isolated from normal Caucasian breast tissue genotyped for the ZFP36*2 A > G polymorphism were assayed for TTP expression by western blot. Actin was used as a loading control. (b) Total RNA was isolated from normal Caucasian breast tissue genotyped for the ZFP36*2 A > G polymorphism (n = 5 samples of each genotype) and assayed for TTP and COX-2 mRNA expression by qPCR. Relative mRNA levels were normalized to GAPDH internal control. (a) *p = 0.0419 and 0.0206 for AA vs. AG and AA vs. GG genotypes, respectively. (b) *p = 0.016 for AA vs. GG genotypes. (c) Schematic representation of the ZFP36 promoter containing the ZFP36*2 A > G SNP. Transcription factor binding sites containing the ZFP36*2 A allele (shown in bold) were identified36 and indicated with the consensus binding motif shown in uppercase.

Previous work has demonstrated overexpression of the prostaglandin synthase COX-2 to be a factor in breast cancer pathogenesis.37, 38 The COX-2 mRNA contains an ARE within its 3′UTR39 and based on our previous findings demonstrating the ability of TTP to target COX-2 mRNA for rapid degradation,25 we hypothesized that COX-2 expression levels would be inversely correlated with ZFP36*2 genotype. Shown in Figure 2b, COX-2 mRNA levels were increased in tissue samples from ZFP36*2 G allele carriers, with ZFP36*2 AG heterozygotes and GG homozygotes showing a 2.5-fold and 5.2-fold increase in COX-2 expression compared with the ZFP36*2 AA genotypes, respectively. These findings indicate that the presence of ZFP36*2 G allele attenuates TTP expression in breast tissue allowing for enhanced expression of the TTP target gene COX-2.

ZFP36 and ELAVL1 haplotypes and their impact on survival

Since the ZFP36 and ELAVL1 genes are located on the same chromosome, we tested LD for all possible pairs of loci (Supporting Information Table 2). Both populations had variable LD scores and significance. Intragenic loci displayed a high degree of LD with greater significance, whereas only one intergenic locus showed significant LD. Haplotype frequencies were estimated and compared between patients with an OS time of 5 or less years versus patients with a greater than 5 year OS. Five haplotypes were constructed for ZFP36 gene, but none were found to be associated with OS (global p = 0.1089 and 0.1337 for Caucasian and African–American breast cancer patients, respectively). Similarly, for the 17 haplotypes constructed with ELAVL1, none of them was associated with OS (global p = 0.8978 and 0.4923 for Caucasian and African–American breast cancer patients, respectively; Supporting Information Table 3).

Combined effect of variant genotypes on patient survival

To detect the combined effect of multiple risk genotypes of ZFP36 and ELAVL1 gene polymorphisms, we categorized patients according to number of risk genotypes present in each patient. Genotypes showing HR > 1 in dominant model were considered as risk genotypes (Table 2). For ZFP36 gene polymorphisms, patients were categorized into “0–1” vs. “2–3” risk genotype carriers, however for ELAVL1 gene polymorphisms, patients were categorized into “0–3” vs. “4–7” risk genotype carriers. Furthermore, patients were also compared as “0–2” vs. “3–5” or “6–8” total risk genotype carriers (Table 4). We found that, in Caucasian breast cancer patients “2–3” ZFP36 risk genotype carriers show poorer survival (HR = 1.79; 95%CI = 1.04–3.11; p = 0.037; log-rank p value = 0.034) than “0–1” ZFP36 risk genotype carriers and this risk increases in patients who carry total number of “6–8” risk genotypes of ZFP36 and ELAVL1 gene polymorphisms (HR = 2.42; 95% CI = 1.17–4.99; p = 0.017; log-rank p = 0.007). The P-trend analysis also showed that with increased number of risk genotypes, a respective increase in Caucasian breast cancer patient HR was observed (ptrend = 0.0109). The African–American breast cancer patients did not show similar trend of poor prognosis with increased number of total risk genotypes, however, when we compared “0–3” versus “4–7” ELAVL1 risk genotype carriers, synergic effect of ELAVL1 gene polymorphisms (more than three risk genotypes) appeared to have borderline modest effect in prognosis of African–American breast cancer patients (HR = 2.18; 95% CI = 0.94–5.08; p = 0.070; log-rank p = 0.063).

Table 4. Number of risk genotypes and survival of breast cancer patients
inline image


Breast cancer is the most common type of malignant cancer among women, with various factors contributing to its high mortality rate.1 While current efforts utilizing gene expression profiling (e.g., Oncotype DX and MammaPrint) are gaining acceptance as clinical predictors,40 further insight into the genetic causes underlying pathogenic gene expression in breast cancer is needed. Various gene products associated with promoting the various facets of tumorigenesis are frequently overexpressed in cancer cells. A consistent feature present within these gene transcripts is the ARE sequence. However, the ability of the ARE to target these mRNAs for post-transcriptional regulation is defective in tumor cells, allowing for aberrant gene overexpression and the acquisition of neoplastic traits during breast cancer development.8

Through their ability to bind ARE sequences, the RNA-binding proteins TTP and HuR are pleiotropic regulators of several genes associated with breast cancer.17–20, 34 While changes in the expression pattern of HuR and TTP are commonly observed during tumorigenesis resulting in enhanced mRNA stabilization,12, 13 the underlying causes promoting these changes are not well understood. Here, we examined whether 10 common genetic polymorphisms present in HuR and TTP genes (ELAVL1 and ZFP36, respectively) play a pivotal role influencing corresponding gene expression and more significantly, impact disease outcomes using a cohort of 251 breast cancer patients of Caucasian and African–American origins. To our knowledge, this is the first study exploring the association of genetic variations in ELAVL1 and ZFP36 genes with cancer patient outcomes.

Several studies have shown that African–Americans have poor prognosis in comparison with Caucasian breast cancer patients. Various reasons underlying this difference can be attributed to social, environmental and genetic factors in African–American breast cancer patients.4, 5, 41–44 In this study, we also found poor survival in African–Americans than Caucasian breast cancer patients (116 months vs. 124 months) and comparison between clinical characteristics of patients suggested that treatment delay, higher grade and extent of tumor and ER/PR negativity were primary clinical factors contributing to poor African–American patient survival outcomes. Our findings are consistent with previous reports4–6 except that there was a significant difference between mean age of disease onset in patients of both races (60.42 years in Caucasians vs. 52.54 years in African–Americans; p = 1.80 × 10–6).

Together with the comparison of breast cancer patients' prognosis between two races, this study examined associations between genetic variations in ELAVL1 and ZFP36 genes and survival outcomes. The ELAVL1 gene has more than 400 SNPs, whereas ZFP36 has 49 SNPs according to dbSNP ( With the majority of ELAVL1 SNPs having less than 5% MAF, 7 ELAVL1 SNPs were selected that had a MAF >5%. An established functional role for these selected SNPs has yet to be determined, although their location within the ELAVL1 mRNA suggests a possible role in mRNA stability, microRNA recognition and splicing regulation. Previously, Carrick et al. had explored genetic polymorphisms and haplotypes of ZFP36 gene and identified four specific haplotype-tag SNPs (htSNPs) for Caucasians and five for African–Americans that were predicted to distinguish 95% or more of the haplotypes.45 However, out of these htSNPs only three polymorphisms (ZFP36*2, ZFP36*8 and ZFP36*10) have >5% MAF and therefore chosen in this study. ZFP36*2 is a promoter region polymorphism previously shown to impact TTP promoter activity using a luciferase reporter assay, with the presence of the G allele inhibiting TTP promoter activity two-fold.35 In order to predict the impact of ZFP36*2 SNP on TTP gene expression, the ZFP36 promoter sequence surrounding this SNP was identified to bind several putative transcription factors whose binding could be negatively impacted due to the ZFP36*2 G allele (Fig. 2c). For instance, liver X receptor (LXR), a member of the nuclear receptor family of transcription factors that plays a role in lipid metabolism, has a putative binding site in the wild-type sequence that could be disrupted when the SNP is present. This is interesting since LXR agonists have been shown to inhibit expression of inflammatory mediators in cultured macrophages and be used to limit inflammation.46 In contrast, the ZFP36*8 polymorphism in protein coding domain is not predicted to alter the amino acid sequence of TTP, and its functional consequence impacts protein translation presumably through a rare codon phenomenon.45, 47

The genotypic frequencies were consistent with Hardy-Weinberg equilibrium but differ significantly between both races for 60% SNPs selected indicating strong ethnic variability. On analyzing the independent effect of each SNP on patient survival, ZFP36*2 G allele carriers were found to have significantly lower median survival (101 months vs. 132 months) and higher risk for death (HR = 2.03; 95% CI = 1.09–3.76; p = 0.025; log-rank p = 0.022) in comparison with ZFP36*2 AA genotype carriers in Caucasian race. These results were still valid for multivariate analysis, however this polymorphism did not show any significant interaction with other factors, indicating ZFP36*2 A > G gene polymorphism as independent prognostic factor for Caucasian breast cancer patients. By contrast, this polymorphism was not found to be associated with survival outcomes of African–American patients.

Suppressed expression of TTP is associated with poor prognosis of breast cancer,29, 48 indicating that the poor prognosis in ZFP36*2 “AG + GG” genotype carriers may be due to lower TTP expression. Our results showed a significant difference between mRNA expression between ZFP36*2 AA vs. ZFP36*2 AG and GG genotypes. Furthermore, ZFP36*2 genotype-dependent loss of TTP expression was reflected in enhanced expression of the TTP-target mRNA COX-2 and may additionally may allow for upregulation other TTP-target genes encoding pro-inflammatory cytokines, oncogenes and growth factors. These findings are in agreement with a study examining rheumatoid arthritis (RA) where a trend was observed with the ZFP36*2 GG genotype to have an early age of disease onset compared to the AA/AG genotypes, however no significant differences were observed in ZFP36*2 allele frequencies between healthy individuals and RA patients.35 Taken together, these findings indicate the ability this SNP to modulate disease activity by negatively impacting expression of TTP on a transcriptional level and advocate ZFP36*2 A > G polymorphism as a new prognostic marker for breast cancer patients.

This was the first study exploring the role of common genetic variations in ELAVL1 and ZFP36 genes in prognosis of breast cancer patients. While the findings are comprehensive and make a causal link between pathogenic gene expression and a regulatory SNP in the mRNA decay factor TTP, some limitations should be noted. The main limitation of our study was low sample size especially in African–American breast cancer patients. Along with this limitation was missing data for some variables such as menopausal status, progression free survival and specific therapeutic details. Nonetheless, the novel findings presented here provide the basis for future similar and replicative studies in larger cohorts. In conclusion, ZFP36*2 A > G gene polymorphism has emerged as novel prognostic marker for Caucasian breast cancer patients and ELAVL1 gene polymorphisms may have some contributing role in determining survival outcome of breast cancer patients.


We thank Dr. Kristin Wallace and Dr. Edsel Pena for critical review of this manuscript and helpful comments and the University of South Carolina Cancer Research Center Biorepository for providing study samples and clinical details.