Abstract
Recently, we identified a novel breast cancer (BC) susceptibility locus at 6q22.33 following a genome-wide association study (GWAS) in the Ashkenazi Jewish (AJ) genetic isolate. To replicate these findings, we performed case-control association analysis on 6q22.33 (rs2180341) in additional 487 AJ BC cases and in an independent non-Jewish (non-AJ), predominantly European-American (EU-Am), populations of 1,466 BC cases and 1,467 controls. We have confirmed the 6q22.33 association with BC risk in the replication cohorts (per-allele OR=1.18, 95%CI 1.04–1.33, p=0.0083) with the strongest effect in the aggregate meta-analysis of 3,039 BC cases and 2,616 AJ and non-AJ controls (per-allele OR=1.24, 95%CI 1.13–1.36, P=3.85×10−7).
We have also shown that the association was slightly stronger with ER positive tumors (per-allele OR=1.35, 95%CI 1.20–1.51, p=2.2×10−5) compared to ER negative tumors (per-allele OR=1.19, 95%CI 0.97–1.47, p=0.1). Furthermore, this study provides a novel insight into the functional significance of 6q22.33 in BC susceptibility. Due to stronger association of 6q22.33 with ER-positive BC we examined the effect of candidate genes on ER response elements (ERE). Upon transfection of overexpressed RNF146 in the MCF-7 BC cell line, we observed diminished expression of an ERE reporter construct. This study confirms the association of 6q22.33 with BC, with slightly stronger effect in ER positive tumors. Further functional studies of candidate genes are in progress and a large replication analysis is being completed as part of an international consortium.
Keywords: Ashkenazi Jews, Breast Cancer, Genome-wide association studies, SNPs, estrogen receptor
INTRODUCTION
It is estimated that up to 30% of BC cases may be caused by genetic factors (1–4). Family history of BC is responsible for the greatest increase in risk, however, the high-penetrant genes that have been identified, such as BRCA1, BRCA2, PTEN and p53, only explain 20–25% of familial BC (5) and 5% of all BC (6). Recently, genome wide association studies (GWAS’s) have proven to be useful in identification of additional genetic factors responsible for BC susceptibility (7–10). These large independent studies have reported genetic variation in fibroblast growth factor receptor 2 (FGFR2) conferring 1.2–1.4 increased risk to BC in populations with different genetic ancestries. Besides FGFR2, several other loci have been identified with a low penetrant effect on BC risk. However, the replication of these additional candidates has varied between the studies, very likely as a result of statistical power limitations and population stratification. Therefore, large replication studies complemented with functional evidence will be needed to confirm reported associations with BC risk.
In our recent GWAS, performed in the Ashkenazi Jewish (AJ) population, we have identified a novel candidate region on 6q22.33 associated with ~ 1.3 fold increased risk to BC. We utilized a two stage design; first by analysis of 250 samples from AJ cases with a marked family history of BC compared to 300 AJ controls and then, in a second stage by replication analysis of 384 loci in ~ 1,000 unselected AJ BC cases and 1,000 AJ controls (10). We have shown that the strongest association mapped within a 200kb region on 6q22.33, where two candidate genes are located: enoyl Coenzyme A hydratase domain containing 1 (ECHDC1) and ring finger protein 146 (RNF146). Prior evidence links both genes to BC tumorigenesis. ECHDC1 has been suggested to play a major role in mitochondrial fatty acid oxidation and it is well established that endogenous fatty acid synthetic activity is abnormally elevated in a subset of breast carcinomas (11, 12). RNF146, a.k.a. dactylidin, encodes a polypeptide containing an amino-terminal C3HC4 RING finger domain, characteristic of the ubiquitin proteasome system, which regulates such processes as cell cycle, apoptosis, transcription, protein trafficking, DNA replication and repair, and angiogenesis (13, 14).
In order to validate the association of 6q22.33 with increased BC risk, in this study we have performed a replication analysis on independent cohorts of cases and controls of both AJ as well as non-AJ, predominantly EU-Am populations. In addition to epidemiological observations supporting the involvement of 6q22.33 in BC susceptibility, we also provide evidence of a possible functional mechanism accounting for the association of the 6q22.33 locus and BC risk.
MATERIALS AND METHODS
Subjects
Cases and controls for association analyses were identified from two populations – AJ and non-AJ populations of predominantly European (EU) ancestry. For replication analysis on AJ population, we used 487 BC patients, ascertained by the Clinical Genetics Service at Memorial Sloan-Kettering Cancer Center; a substantial proportion of these cases were described in the context of prior epidemiologic studies (15). All these cases tested negative for AJ founder mutations in BRCA1 and BRCA2 genes. Cases were compared to 1,149 healthy AJ controls used in our previous study (10). Replication population of non-AJ cases included: 1) 171 familial BC cases of EU ancestry ascertained from clinical protocols at MSKCC. Eligible cases had 3 or more individuals with BC present in a single lineage. 2) 751 non-AJ sporadic BC cases unselected for a family history of the disease and collected as a part of a separate protocol at MSKCC. 3) 544 non-AJ sporadic BC cases unselected for a family history of the disease, which were ascertained from anonymized protocols at MSKCC. Overall, as illustrated in Table 1A, although all non-AJ cases were predominantly EU-Am (n=1604), other ancestries were present in this ascertainment, such as African-American (Af-Am) (n=167), Hispanic (n=117), Asian (n=58) and other populations (n=7).
Table 1.
Population substructure of non-AJ cases and controls in the study. A. Age and ethnicity breakdown of replication cases and controls. B. Population structure of non-AJ replication cases and controls with rs2180341 allele and genotype frequencies used in the study. Minor allele frequencies (MAF) are from control data only.
| Table 1A. | ||||
|---|---|---|---|---|
| Replication Population | ||||
| Cases | Controls | |||
| AGE | N | % | N | % |
| <45 | 533 | 27 | 330 | 22 |
| 45–54 | 613 | 31 | 433 | 30 |
| 55–64 | 450 | 23 | 405 | 28 |
| >65 | 357 | 18 | 299 | 20 |
| Total | 1953 | 1467 | ||
| RACE | ||||
| European* | 1604 | 82 | 1337 | 91 |
| African-American | 167 | 9 | 51 | 3 |
| Hispanic | 117 | 6 | 58 | 4 |
| Asian | 58 | 3 | 18 | 1 |
| Other | 7 | 0 | 3 | 0 |
| Total | 1953 | 1467 | ||
| Table 1B | ||||
|---|---|---|---|---|
| Cases | Controls | MAF | ||
| European | ||||
| AA | 612 (55) | 783 (58) | 0.26 | |
| AG | 425 (38) | 492 (37) | ||
| GG | 80 (7) | 62 (5) | ||
| Total | 1117 | 1337 | ||
| African-American | ||||
| AA | 66 (40) | 23(46) | 0.37 | |
| AG | 78 (47) | 22 (42) | ||
| GG | 22 (13) | 6 (12) | ||
| Total | 166 | 51 | ||
| Hispanic | ||||
| AA | 73 (63) | 37 (65) | 0.22 | |
| AG | 35 (30) | 18 (30) | ||
| GG | 8 (7) | 3 (5) | ||
| Total | 116 | 58 | ||
| Asian | ||||
| AA | 33 (57) | 8(41) | 0 26 | |
| AG | 20 (34) | 10 (59) | ||
| GG | 5 (9) | 0 (0) | ||
| Total | 58 | 18 | ||
European ancestry also includes replication AJ cohorts (n=487)
For non-AJ controls, we ascertained 630 cancer-free women, who participated in the New York Cancer Project (NYCP), an ongoing cohort study (16), all of whom were of EU ancestry. A second group of non-AJ controls comprised 837 women who were either participating in cancer screening and were cancer free, or who were spouses of patients with prostate cancer, and who did not have a personal or family history of BC. The population structure of non-AJ controls was similar to the non-AJ sporadic BC group, as detailed in Table 1A: EU ancestry (n= 1337), Af-Am (n=51), Hispanics (n=58), Asians (n=18) as well as other ancestries (n=3). Previously published association data from our GWAS (phase 1 and phase 2) were included in a final aggregate meta-analysis only, with a detailed description of BC cohorts and controls published in that prior study (10).
Genotyping
Genomic DNA was prepared using Gentra Autopure system, according to manufacture’s protocol (Qiagen, Valencia, CA). Other DNA extraction procedures were performed as previously described (17). Genotyping of rs2180341, rs6569479, rs6569480 and rs7776136 was performed by the TaqMan allelic discrimination procedure using assays by design under standard conditions (Applied Biosystems, Foster City, CA). In order to avoid potential bias by inclusion of data from samples previously genotyped by other methods (Affymetrix 500K and Illumina GoldenGate assasy), we have re-genotyped all published sample populations by conventional TaqMan allelic discrimination. All genotypes showed 100% concordance. The clustering of genotype calls was performed by SDS 2.1 software (Applied Biosystems, Foster City, CA).
Statistical methods
Deviations of the genotype frequencies in the controls from those expected under Hardy-Weinberg equilibrium were evaluated by chi-square tests (1 degree of freedom). BC risk associated with rs2180341 was estimated as odds ratios (OR) and 95% confidence intervals (CI) using unconditional logistic regression with multiple genetic models including the genotype model (separate indicators for heterozygotes and rare homozygotes), dominant model (indicator for heterozygotes and rare homozygotes combined), recessive model (indicator for homozygotes) and log additive (‘per-allele’) model (each copy of rare allele) with the common homozygote as the reference category. For cases with no apparent trend in ORs, in all analysis we have also used a two degrees of freedom (2df) test (genotype test). All models were adjusted for continuous age at diagnosis (cases) or at the time of inclusion in the study (controls) and ethnicity (EU-Am, Af-Am, Hispanic etc).
Sequencing
Coding regions of ECHDC1 and RNF146 were sequenced by ABI3700 capillary sequencing. The primers were designed to cover entire transcribed regions of both genes and to capture ~50bp of sequence on both sides of each exon (primer sequences available upon request). Sequencing was performed from both directions and sequence data were analyzed by both Sequencher (Genes Codes, Ann Arbor, MI) and Mutation surveyor (Softgenetics, State College, PA) software packages.
Cell Culture
MCF-7 cells (American Type Culture Collection, Manassas, VA) were grown in normal DMEM supplemented with 10% (v/v) FCS, 0.01 mg/ml bovine insulin and antibiotics at 37°C in a humidified atmosphere of 95% air and 5% CO2.
Immunoblot assay
MCF7 cells were grown in a 35 mm dish to about 80% confluence. 4 µg pCMV6-XL5/RNF146 or pCMV6-XL5 (Origen Technologies Inc, Rockville, MD) with 8 µl FuGene HD transfection reagent (Roche Applied Science, Indianapolis, IN) were mixed and added into the cultured cells according to Roche’s protocol. After 48 or 72 hours transfection, the cells were washed with 1X PBS and lysed with 0.5 ml lysate buffer (Pierce Biotechnology, Rockford, IL) containing a mammalian protease inhibitor mixture (Sigma-Aldrich, St. Louis, MO). The cell lysates were centrifuged at 13200 rpm in eppendorf-centrifuge, 4°C for 10 min. The supernatants were used for Western blot analysis as described by Chen et al.(18). RNF146 antibody dilution was as recommended by the Vander (Santa Cruz Biotechnology, Inc.). Horseradish peroxidase-conjugated secondary antibodies and an ECL chemiluminescent kit were purchased from Amersham Biosciences (GE Healthcare, Piscataway, NJ).
Luciferase assay
MCF-7 cells were grown in 24 well plates to near confluence. 200 ng of DLC1 promoter-Luciferase construct (19) or pCMV-3XERE-Luciferase plasmid (superArray Biosience, Frederick, MD) and 1 µg pCMV6-XL5/RNF146 or pCMV6-XL5 with 3 µl FuGene HD were mixed and added to plate wells for transfection. After a 24 hour incubation, Estradiol (E2)(Sigma-Aldrich, St. Louis, MO) was added to 10 nM final concentration. Luciferase activities of cell lysates were measured according to the manufacturer’s instructions for the Dual Luciferase Assay System (Promega) in a Moonlight 2010 luminometer (Analytical Luminescence laboratory, San Diego, CA), (Turner Designs, Sunnyvale, CA, USA). The experiments were carried out in triplicates and the statistical analysis was performed on the mean values from three independent measurements, given as the mean ± standard deviation (SD).
RESULTS
Replication of 6q22.33 association with BC risk
In this study, we tested the association of a locus at 6q22.33 with BC in independent populations of BC cases and controls. We have previously shown that in AJ population there is a strong linkage disequilibrium (LD) between four SNPs (rs2180341, rs6569479, rs6569480, rs7776136), corresponding to a BC association signal on 6q22.33 (10). Because strong LD among these 4 SNPs was also confirmed in other populations, such as CEU population in HapMap, for all replication screens in this study we have used rs2180341 which tags to an LD block including the associated locus (data not shown). First, we have expanded the initial AJ set used in the previous study by additional 487 clinically ascertained AJ BC cases that were previously tested negative for BRCA1 and BRCA2 mutations (Table 2A, left panel). These were compared to 1,149 Ashkenazi healthy controls used in our previous study (10). The analysis confirmed a significant increase of BC risk assuming a dominant genetic model (age-adjusted OR=1.28, 95%CI 1.03–1.59, p=0.024), however, the association was statistically marginal by per-allele test (age-adjusted OR= 1.18, 95% CI 0.99–1.41, P=0.066), possibly due to small sample size of this replication case group (n=487). The aggregate AJ analysis consisting of a total of 1,565 cases and 1,149 controls from present and prior studies, however, confirmed the strong association of 6q22.33 with BC risk (per-allele OR= 1.32, 95% CI 1.15–1.50, P=0.000057) (Table 2A, right panel).
Table 2.
Replication of 6q22.33 association with breast cancer risk: Odds ratios (OR) and 95% confidence intervals (CI) of rs2180341 and breast cancer risk in AJ controls versus A) AJ BC cases BRCA1/2 negative (left panel) and aggregate AJ cases (right panel) B) non-AJ controls of mixed ethnicities versus non-AJ familial cases of EU ancestry (left panel) and aggregate of all non-AJ BC cases of mixed ethnicities (right panel) C) BC cases and controls of EU ancestry that are a subset of non-AJ populations D) aggregate of replication population of AJ and non-AJ cases and controls, excluding previously published data E) aggregate analysis of all case/control ascertainments.
| A. | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| *AJ Controls | AJ (BRCA1/2 negative) cases | OR | 95% CI | P-value | *All AJ cases | OR | 95% CI | P-value | |
| n=1149 | n=487 | n=1565 | |||||||
| A/A | 710 (61.8%) | 271 (55.6%) | 1 | 837 (53.5%) | 1 | 1 | |||
| A/G | 382 (33.2%) | 192 (39.4%) | 1.31 | 1.05–1.64 | 631 (40.3%) | 1.41 | 1.20–1.67 | ||
| G/G | 57 (5%) | 24 (4.9%) | 1.09 | 0.66–1.79 | 97 (6.2%) | 1.02 | 1.02–2.08 | ||
| 2 df test | 0.06 | 1.00E-04 | |||||||
| dominant | 1.28 | 1.03–1.59 | 0.024 | 1.42 | 1.21–1.67 | 0.000043 | |||
| recessive | 0.98 | 0.60–1.60 | 0.93 | 1.27 | 0.90–1.80 | 0.17 | |||
| Per allele | 1.18 | 0.99–1.41 | 0.066 | 1.32 | 1.15–1.50 | 0.000057 | |||
| B. | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| All non-AJ Controls | non-AJ familial cases | OR | 95% CI | P-value | All non-AJ cases | OR | 95% CI | P-value | |
| n=1467 | n=171 | n=1466 | |||||||
| A/A | 854 (58.2%) | 91 (53.2%) | 1 | 789 (53.8%) | 1 | ||||
| A/G | 542 (37%) | 71 (41.5%) | 1.31 | 0.93–1.83 | 561 (38.3%) | 1.09 | 0.93–1.27 | ||
| G/G | 71 (4.8%) | 9 (5.3%) | 1.07 | 0.51–2.25 | 116 (7.9%) | 1.63 | 1.19–2.24 | ||
| 2 df test | 0.3 | 0.0087 | |||||||
| dominant | 1.28 | 0.92–1.76 | 0.14 | 1.15 | 0.99–1.34 | 0.062 | |||
| recessive | 0.96 | 0.47–1.99 | 0.92 | 1.57 | 1.15–2.15 | 0.0039 | |||
| Per allele | 1.17 | 0.90–1.52 | 0.24 | 1.18 | 1.04–1.33 | 0.0081 | |||
| C. | |||||
|---|---|---|---|---|---|
| EU controls | EU cases | OR | 95% CI | P-value | |
| n=1337 | n=1117 | ||||
| A/A | 783 (58.6%) | 612 (54.8%) | 1 | ||
| A/G | 492 (36.8%) | 425 (38%) | 1.1 | 0.94–1.31 | |
| G/G | 62 (4.6%) | 80 (7.2%) | 1.63 | 1.15–2.32 | |
| 2 df test | 0.016 | ||||
| dominant | 1.17 | 1.00–1.37 | 0.055 | ||
| recessive | 1.57 | 1.11–2.21 | 0.0096 | ||
| Per allele | 1.19 | 1.04–1.35 | 0.01 | ||
| D. | |||||
|---|---|---|---|---|---|
| All replication controls | All replication cases | OR | 95% CI | P-value | |
| n=1467 | n=1953 | ||||
| A/A | 854 (58.2%) | 1060 (54.3%) | 1 | ||
| A/G | 542 (37%) | 753 (38.6%) | 1.09 | 0.93–1.28 | |
| G/G | 71 (4.8%) | 140 (7.2%) | 1.62 | 1.18–2.23 | |
| 2 df test | 0.01 | ||||
| dominant | 1.15 | 0.99–1.34 | 0.06 | ||
| recessive | 1.56 | 1.14–2.13 | 0.0048 | ||
| Per allele | 1.18 | 1.04–1.33 | 0.0083 | ||
| E. | |||||
|---|---|---|---|---|---|
| *Aggregate controls | *Aggregate cases | OR | 95% CI | P-value | |
| n=2616 | n=3031 | ||||
| A/A | 1564 (59.8%) | 1626 (53.6%) | 1 | ||
| A/G | 924 (35.4%) | 1192 (39.3%) | 1.25 | 1.11–1.40 | |
| G/G | 128 (4.9%) | 213 (7%) | 1.51 | 1.19–1.92 | |
| 2 df test | 3.56E-07 | ||||
| dominant | 1.28 | 1.15–1.43 | 3.30E-06 | ||
| recessive | 1.38 | 1.09–1.75 | 0.0065 | ||
| Per allele | 1.24 | 1.13–1.36 | 3.85E-07 | ||
indicates inclusion of AJ datasets from previous study (10) (n=1,149 AJ controls and n=1,078 AJ cases) All statistical tests were adjusted for age and ethnicity as detailed in Materials and Methods.
Second, we performed a replication analysis in three sets of non-AJ BC cases compared to 1,467 non-AJ controls from the same geographical region and including predominantly EU-Ams (Table 2B). To replicate the 6q22.33 association from phase I of our previous study, in which we performed GWAS on 250 affected AJ probands from families with 3 or more BC cases, we genotyped rs2180341 in the cohort of 171 non-AJ cases enriched for family history of BC (Table 2B, left panel). Although statistically insignificant (per-allele p=0.24), the genotype frequencies of G/G and A/G (5.3% and 41.5% in cases versus 4.8% and 37% in controls) showed a minor trend toward association that is also reflected in per allele OR (OR=1.17, 95%CI 0.9–1.52). To increase the statistical power, we have genotyped additional populations of 1,295 non-AJ consecutive BC cases consisting of non-overlapping groups of 751 and 544 non-AJ BC patients collected at the same cancer center. As seen in right panel of table 2B, after adjustment for age and ethnicity, the aggregated analysis of replication non-AJ BC cases (n=1,466) and controls (n=1,467) showed a significant association in all modes of analysis with the strongest association signal assuming recessive model (OR= 1.57, 95% CI 1.15–2.15, p=0.0039). Although the non-AJ cohorts in this study were derived predominantly from a EU background, a small fraction of the study population represented other ancestries (see table 1A and 1B). With the exception of Af-Am population, where we noted differences for rs2180341, the allele and genotype frequencies did not significantly differ in other ancestries (Table 1B). Nevertheless, to correct for population stratification the statistical analyses of non-AJ cohorts were adjusted for age and ethnicity in multivariate models.
Due to slightly different allele frequencies in Af-Am, and in order to verify whether rs2180341 can be used to tag the association signal in the non-AJ population, we examined the LD structure of 6q22.33 in non-AJ by genotyping all four SNPs from our prior study in a subgroup of 847 non-AJ controls and have confirmed strong LD (D’=0.98) across all four subpopulations in this set, that was comparable to observations in AJ (data not shown). Because of the population substructure of the non-AJ cases and controls used in this study, we have performed a separate analysis on the population of EU ancestry only and the association remained statistically significant (per-allele OR=1.19, 95%CI 1.04–1.35, p=0.01)(Table 2 C). The replication of association of 6q22.33 with BC risk was further demonstrated by pooled analysis of all replication ascertainments, excluding the data from previous AJ association study (Table 2D). As shown, the significant associations were observed under all models of analysis with perallele OR=1.18 (95% CI 1.04–1.33, p=0.008).
Finally, we performed an aggregate meta-analysis, including all aforementioned cases (n=3,031) and controls (n=2,616) from this and prior GWAS study. As shown in Table 2E, the association, adjusted for age and ethnicity, was consistent with previous findings and strongly significant (per-allele OR=1.24, 95% CI 1.13–1.36, p=3.85×10−7).
Sequencing and expression analysis of candidate genes in 6q22.33
We have previously shown that rs2180341 tags an LD block constituted of four SNPs that were equally associated with BC risk covering a region of ~ 200 kb. Two candidate genes map within this region: ECHDC1 and RNF146. To further investigate functional consequences of 6q22.33 association we sequenced the transcribed regions of both genes. Besides several rare polymorphisms, for which no significant differences were observed after testing them independently in a subset of case-control population, we did not find any potentially pathogenic mutation that would explain the association signal (Supplemental Table 1).
Functional evaluation of 6q22.33 association with ER positive tumors
To evaluate whether rs2180341 is more strongly associated with ER positive compared to ER negative tumors, we have stratified the model by ER status (Table 3). We were able to retrieve ER status data on 1,658 BC cases that were genotyped in this or previous AJ study. These included 248 ER negative and 979 ER positive tumors from AJ patients and 100 ER negative and 331 ER positive tumors from non-AJ cases. The data were compared to 1,779 EU-Am controls representing the studies for which the ER data on BC cases were available, excluding a subset of non-AJ, ethnically mixed controls (n=837) with no annotations for ER status. Each copy of the rare allele of rs2180341 was associated with slightly stronger risk of ER positive tumors (per-allele OR =1.35, 95%CI 1.20–1.51 p=2.2×10−5) compared to ER negative tumors (per-allele OR=1.19, 95%CI 0.97–1.47, p=0.1)(Table 3A). To eliminate a possibility of potential sampling bias in these estimates as suggested in other studies (20), we have compared ORs of the cases for which ER status data were available to overall ORs of all BC cases used for the study and did not note any significant difference (data not shown).
Table 3.
Odds ratios (OR) and 95% confidence intervals (CI) of rs2180341 and breast cancer risk by ER status. Aggregate controls in the analysis include AJ ascertainments from prior GWAS study (10) (n=1,149 AJ controls and n=1,078 AJ cases). All statistical tests were adjusted for age and ethnicity as detailed in Materials and Methods.
| A | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| *Controls | ER + tumors | OR | 95% CI | P-value | ER - tumors | OR | 95% CI | P-value | |
| n=1779 | n=1310 | n=348 | |||||||
| A/A | 1095 (61.5%) | 679 (51.8%) | 1 | 192 (55.2%) | 1 | ||||
| A/G | 600 (33.7%) | 538 (41.1%) | 1.44 | (1.23–1.68) | 136 (39.1%) | 1.28 | 0.99–1.67 | ||
| G/G | 84 (4.7%) | 93 (7.1%) | 1.78 | (1.30–2.46) | 20 (5.8%) | 1.21 | 0.69–2.14 | ||
| 2 df test | 0.00001 | 0.17 | |||||||
| Dominant | 1.48 | (1.28–1.72) | 0.00001 | 1.28 | 0.99–1.64 | 0.05 | |||
| Recessive | 1.54 | 1.13–2.11 | 0.0067 | 1.1 | 0.63-1.93 | 0.81 | |||
| per-allele | 1.39 | 1.23-1.57 | 0.00001 | 1.19 | 0.97-1.47 | 0.09 | |||
| B | |||||
|---|---|---|---|---|---|
| ER − tumors | ER + tumors | OR | 95% CI | P-value | |
| n=348 | n=1310 | ||||
| A/A | 192 (55.2%) | 679 (51.8%) | 1 | ||
| A/G | 136 (39.1%) | 538 (41.1%) | 1.12 | 0.87–1.43 | |
| G/G | 20 (5.8%) | 93 (7.1%) | 1.34 | 0.80–2.24 | |
| 2 df test | 0.42 | ||||
| Dominant | 1.15 | 0.90–1.46 | 0.26 | ||
| Recessive | 1.28 | 0.78–2.12 | 0.32 | ||
| per-allele | 1.14 | 0.94–1.38 | 0.19 | ||
Although the per allele ORs for ER positive tumors versus ER negative tumors were not statistically different (per-allele P=0.19)(Table 3B), a trend resulted from a slightly elevated genotype frequency of G/G (7.1% ER+ versus 5.8% ER-) and A/G (41% ER+ versus 39.1% ER-) in ER positive tumors, corresponding to increased ORs (per-allele OR=1.14, 95% CI 0.94–1.38). We, therefore, further investigated the functional consequences of ER positive tumor association with BC risk in RNF146. We found that over-expression of RNF146 under control of a dynein light chain 1 (DLC1) promoter with a luciferase reporter in MCF-7 cells (Figure 1A) resulted in remarkably decreased luciferase activity (Figure 1B). The DLC1 promoter contains half of an estrogen receptor binding site (18). This implies that RNF146 plays a role in estrogen mediated cellular activities. To verify this possibility, we transfected RNF146 and pCMV-3XERE-Luciferase-luciferase constructs into MCF-7 cells. Similarly in this experiment the luciferase activity was significantly inhibited (Figure 1C). The pCMV-3XERE-luciferase contains only a minimal CMV promoter, 3 copies of an estrogen receptor binding element (ERE) and a luciferase reporter. In summary, these experiments show diminished luciferase activity from the ERE reporter vector in the presence of RNF146 overexpression, suggesting that RNF146 plays a role in downregulation of estrogen response elements.
Figure 1.
A. Immunoblot analysis of RNF146 in MCF7 cell. MCF7 cells were transfected with pCMV6-XL5/RNF146 and pCMV6-XL5 plasmids by FuGene HD. About 0.5 ml the cell lysates were subjected to immunoprecipitation and Western blot analysis with RNF146 antibody. For immunoprecipitation, RNF146 antibody was diluted 500X, and for Western blot, the antibody was diluted 10000X. Lane 1, pCMV6XL5/RNF146 transfected; Lane 2, pCMV6-XL5 transfected.
B. RNF146 suppresses ERE-Luciferase activity. After 48 hours transfection, the MCF7 were washed with 1XPBS buffer and lysed with 100 µl lysate buffer. The cell lysates were centrifuged at 12000rpm in 4°C for 10 min. 10 µl supernatant was used for Luciferase activity measurement using DLC1 promoter-Luciferase with RNF146 construct assay and
C. pCMV-3XERE-Luciferase with RNF146 construct asssay. pCMV-3XERE contains 3 copies of an estrogen receptor binding element (ERE) and a luciferase reporter.
DISCUSSION
This study represents a replication of the association of the 6q22.33 locus with increased risk of BC. We confirmed a significant association utilizing an expanded set of additional AJ cases as well as pooled analysis of AJ case-control data from current and previous studies (Table 2A). The strong association observed in AJ was driven by frequency differences in distribution of G/G (6.2% in cases, 5% in controls) and A/G (40.3% in cases and 33.2% controls) genotypes. When we compared the observed genotype frequencies in AJ with reported data on 60 CEU individuals from HapMap (G/G 6.7%, A/G 41.7%), we suspected a potential sampling bias due to lower allele frequencies of rs2180341 in the AJ control population used in our previous study that might have contributed to false association discovery. The replication analysis in non-AJ populations, however, showed that the allele and genotype frequencies do not deviate from those observed in AJ control cohorts, which were consistent throughout the study in both non-AJ control ascertainments. The significant association observed in the aggregate analysis of non-AJ cases and controls provides the first evidence that genetic variation in 6q22.33 may contribute to BC risk in populations other than AJ (table 2B). Analysis of a small subset of non-AJ cases derived from families with a strong history of BC did not reveal a significantly stronger association with the 6q22.33 locus than that observed for sporadic BC cases unselected for a family history of the disease, consistent with our previous observations in the AJ population (10). Although there was a trend for a slight increase of minor allele homozygotes and heterozygotes for the non-AJ familial cases, as seen in Table 2B, the addition of this “enriched” group did not offer an obvious advantage in powering this study.
Because in the replication analysis we have used populations of mixed ethnicities, there is the potential for population admixture, or confounding by ancestry. As we have detailed in Table 1A, besides being predominantly of EU ancestry, a small fraction of our sample collections represents other ancestries. This may in turn affect overall association findings if the frequencies of rs2180341 differ between these populations or if LD structure of the region varies substantially. After quantification of admixture in non-AJ case/controls populations based on allele and genotype frequencies (see Table 1B) we did not note significant differences between EU, Asian or Hispanic populations. We noted, however, a slightly elevated minor allele frequency of this SNPs in a subset of Af-Am ancestry (37% vs. 26% in EUs). These findings prompted us to control all statistical analyses for ethnicity. The different allele frequencies in AAs were, however, present equally in controls as well as cases which predicts no dramatic effect on overall association statistics. The conclusion that population stratification in this study does not affect association findings was further supported when the analysis was restricted to EU ancestry only (Table 2C). The significant BC risk effect associated with 6q22.33 in the EU-Am subset was no different compared to the overall association statistics in the entire ethnically mixed non-AJ set (Table 2B, left panel) as well as pooled analysis of all replication case/controls population (Table 2D). This suggests that population substructure does not significantly confound association findings in this study and that adjustment for ethnicity provided sufficient correction for population stratification. Nevertheless, a preliminary analysis of small Af-Am subset demonstrated a trend towards the association of 6q22.33 with BC risk in AA, with higher minor allele frequencies for both cases and controls compared to samples of EU ancestry (OR=1.16, 95% CI 0.727–1.856, p=0.53, complete data not shown). However, these data were limited by very small sample size and a separate association study in this population is currently underway.
The population differences in prior GWAS studies to date (7, 9) may explain the failure to detect the association with 6q22.33, as well as the inconsistent replication of other loci, for example the chromosome 2 loci detected in the Iceland cohorts but not in the initial U.K led consortium (8). Besides global population stratification, other factors may account for these differences between studies. These include sample size differences, geographical differences within the same population and possible sample ascertainment bias that may result in different pools of cancer predisposing alleles in different studies. Although findings from both AJ and non-AJ populations in our study replicate the association of 6q22.33 with BC risk when analyzed separately as well as in aggregate (Table 2E), we have observed a significantly stronger effect in the AJ population as compared to non-AJ, predominantly EU case-control ascertainments. This may indicate population differences in LD structure of the region, even though the LD block of 200kb defined by perfectly correlated SNPs (rs2180341, rs6569479, rs6569480, rs7776136) is comparable in both population ascretaiments. In non-AJ, however, there may be differences in the extent of LD beyond the conserved 200kb locus. Other variants near this region will need to be genotyped in order to better define the association signal outside AJ ancestry.
As the case for other loci mapped by GWAS, sequencing of coding regions of two candidate genes in the implicated region failed to identify any mutation or variant linked to the associated allele. It is possible that the BC-causative variant in 6q22.33 maps outside the 200kb core region. If the extent of LD is reduced in the non-AJ population, this will also cause a reduction of the association signal. A more distant variant linked to the 200kb core region on 6q22.33 could be involved in distant transcription regulation of either of the two candidate genes in this locus; further fine mapping will be needed to address these points. Such long-range effects could in turn distally affect the expression of two candidate genes mapping in this region, ECHDC1 and RNF146. A detailed expression analysis in large collection of pairs of clinically well-annotated normal tissue specimens as well as corresponding breast tumor counterparts is currently underway to correlate gene expression of both candidate genes with the presence of high risk allele at 6q22.33 locus.
There was also evidence for a possible mechanism of the putative association of the 6q22.33 locus with BC risk suggested by analysis of the ER status in our case populations. These findings were similar to other GWAS reports in which novel BC susceptibility loci were slightly strongly associated with ER positive than ER negative tumors (21, 22). In the case of 6q22.33, we found a statistically significant association with risk of ER positive tumors (Table 3A); however, the differences in the associations for risk of ER positive tumors did not differ from ER negative tumors in our study. A likely explanation was the limited power of this comparison due to small sample size for ER negative cancers in this and other analyses. The association of 6q22.33 with ER positive tumors would also predict an association with greater age at onset in cases; however, no such difference was noted (data not shown).
Although a larger series will be needed to confirm the 6q22.33 association with ER positive tumors, this finding provides an avenue for further biological investigation. Among the two candidate genes at 6q22.33, RNF146 was found to downregulate an estrogen response driven reporter gene in vitro in MCF-7 BC cells (Figure 1). We interpret this to mean that modest upregulation of RNF146 expression associated with high risk allele could interfere with estrogen signaling. Although the current experimental evidence from the literature describing the function of RNF146 is limited, the protein “dactylidin” has been shown to posses E3 ubiquitin ligase activity. These enzymes were observed to be highly overexpressed in many cancers and their overexpression was often associated with poor prognosis (23–25). Moreover, many E3 ligases are targets for small molecules in anticancer therapies (26). Several recent studies have shown that nuclear E3 ubiquitin ligases may posses other essential functions besides their role in ubiquitination, i.e. in modulating the estrogen receptor pathway (27, 28). Our data suggest that RNF146 may be another negative regulator of estrogen response in breast cancer tissue. We hypothesize that 6q22.33 high risk allele harbors a genetic variation that may impact gene expression of RNF146. Thus, overexpressed RNF146 subsequently diminishes the estrogen receptor response upon activation by estrogen, perhaps due to competition with estrogen via ERE regulatory elements. This observation is in accordance with the slightly stronger association of 6q22.33 in ER-positive tumors. A minor allele at this locus, associated with breast cancer risk in ER positive tumors, may also be associated with elevated expression of RNF146. Testing this hypothesis will require additional investigation using approaches such as full sequencing of the region to identify putative functional genetic variation associated with 6q22BC risk alleles, as well as correlated genotype/expression profiling in tumors and normal tissues.
Overall, the finding presented here adds further support for the 6q22.33 locus as a candidate for further investigation in BC genetic epidemiologic investigations. The 30% increased risk associated with this allele in final aggregate analysis (Table 2E) is comparable to confirmed BC loci identified in recent whole genome scans. As a single marker, rs2180341 is of limited clinical or predictive utility. Also, as demonstrated recently, multiplicative models including the several BC risk variants discovered to date (9) do not yield a significant improvement in individual predictive efficacy, although public health uses for genetic panels of such predictive markers may emerge in the future (29). Eventually, a large-scale analysis of all significant “hits” from these studies, including modeling of potential gene-gene and gene-environment interactions, may provide more clinically relevant risk prediction models. Additional approaches, such as epigenome mapping and tumor expression profiling, may also help account for the remainder of the hereditary fraction of BC risk. For the 6q22.33 and other loci there also remains the need for additional epidemiological and clinical information to adjust the associations for other potential confounding variables, i.e. traditional breast cancer risk factors. Because of small BC risk effect predicted for this locus, this would require a large association analysis of well-annotated case/control collections. For that purpose, to precisely estimate the association of 6q22.33 and BC risk and to evaluate the possible effects of population stratification, a large international pooled analysis of case-control studies is underway in the Breast Cancer Association Consortium (BCAC). This study involves the analysis of 6q22.33 genotypes in more than 50,000 BC cases and controls.
Supplementary Material
ACKNOWLEDGMENTS
The authors wish to thank Mathew Danzig for review of charts for estrogen receptor data. Control samples were provided by the New York Cancer Project supported by the Academic Medicine Development Company of New York. This work was supported in part by federal funds from the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research, and from SAIC-Frederick under contract # N01-CO-12400. The content of this publication does not necessarily reflect the views of the Department of Health and Human Services nor does its mention of trade names, commercial products or organizations imply endorsement by the U.S. Government. This research was also supported by the Breast Cancer Research Foundation grant 13740 (KO), the Susan Komen Foundation RO1-BCTR0601361 (KO), the Lymphoma Foundation, and the Niehaus, Southworth, Weissenbach Cancer Research Fund.
REFERENCES
- 1.Lichtenstein P, Holm NV, Verkasalo PK, et al. Environmental and heritable factors in the causation of cancer--analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med. 2000;343:78–85. doi: 10.1056/NEJM200007133430201. [DOI] [PubMed] [Google Scholar]
- 2.Ahlbom A, Lichtenstein P, Malmstrom H, Feychting M, Hemminki K, Pedersen NL. Cancer in twins: genetic and nongenetic familial risk factors. J Natl Cancer Inst. 1997;89:287–293. doi: 10.1093/jnci/89.4.287. [DOI] [PubMed] [Google Scholar]
- 3.Verkasalo PK, Kaprio J, Pukkala E, Koskenvuo M. Breast cancer risk in monozygotic and dizygotic female twins: a 20-year population-based cohort study in Finland from 1976 to 1995. Cancer Epidemiol Biomarkers Prev. 1999;8:271–274. [PubMed] [Google Scholar]
- 4.Locatelli I, Rosina A, Lichtenstein P, Yashin AI. A correlated frailty model with long-term survivors for estimating the heritability of breast cancer. Stat Med. 2007;26:3722–3734. doi: 10.1002/sim.2761. [DOI] [PubMed] [Google Scholar]
- 5.Peto J, Collins N, Barfoot R, et al. Prevalence of BRCA1 and BRCA2 gene mutations in patients with early-onset breast cancer. J Natl Cancer Inst. 1999;91:943–949. doi: 10.1093/jnci/91.11.943. [DOI] [PubMed] [Google Scholar]
- 6.Antoniou AC, Pharoah PD, McMullan G, et al. A comprehensive model for familial breast cancer incorporating BRCA1, BRCA2 and other genes. Br J Cancer. 2002;86:76–83. doi: 10.1038/sj.bjc.6600008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hunter DJ, Kraft P, Jacobs KB, et al. A genome-wide association study identifies alleles in FGFR2 associated with risk of sporadic postmenopausal breast cancer. Nat Genet. 2007;39:870–874. doi: 10.1038/ng2075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Stacey SN, Manolescu A, Sulem P, et al. Common variants on chromosomes 2q35 and 16q12 confer susceptibility to estrogen receptor-positive breast cancer. Nat Genet. 2007;39:865–869. doi: 10.1038/ng2064. [DOI] [PubMed] [Google Scholar]
- 9.Easton DF, Pooley KA, Dunning AM, et al. Genome-wide association study identifies novel breast cancer susceptibility loci. Nature. 2007;447:1087–1093. doi: 10.1038/nature05887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Gold B, Kirchhoff T, Stefanov S, et al. Genome-wide association study provides evidence for a breast cancer risk locus at 6q22.33. Proc Natl Acad Sci U S A. 2008 doi: 10.1073/pnas.0800441105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Menendez JA, Lupu R. RNA interference-mediated silencing of the p53 tumor-suppressor protein drastically increases apoptosis after inhibition of endogenous fatty acid metabolism in breast cancer cells. Int J Mol Med. 2005;15:33–40. [PubMed] [Google Scholar]
- 12.Zhou W, Simpson PJ, McFadden JM, et al. Fatty acid synthase inhibition triggers apoptosis during S phase in human cancer cells. Cancer Res. 2003;63:7330–7337. [PubMed] [Google Scholar]
- 13.Mani A, Gelmann EP. The ubiquitin-proteasome pathway and its role in cancer. J Clin Oncol. 2005;23:4776–4789. doi: 10.1200/JCO.2005.05.081. [DOI] [PubMed] [Google Scholar]
- 14.Chen C, Seth AK, Aplin AE. Genetic and expression aberrations of E3 ubiquitin ligases in human breast cancer. Mol Cancer Res. 2006;4:695–707. doi: 10.1158/1541-7786.MCR-06-0182. [DOI] [PubMed] [Google Scholar]
- 15.Shaag A, Walsh T, Renbaum P, et al. Functional and genomic approaches reveal an ancient CHEK2 allele associated with breast cancer in the Ashkenazi Jewish population. Hum Mol Genet. 2005;14:555–563. doi: 10.1093/hmg/ddi052. [DOI] [PubMed] [Google Scholar]
- 16.Mitchell MK, Gregersen PK, Johnson S, Parsons R, Vlahov D. The New York Cancer Project: rationale, organization, design, and baseline characteristics. J Urban Health. 2004;81:301–310. doi: 10.1093/jurban/jth116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kirchhoff T, Kauff ND, Mitra N, et al. BRCA mutations and risk of prostate cancer in Ashkenazi Jews. Clin Cancer Res. 2004;10:2918–2921. doi: 10.1158/1078-0432.ccr-03-0604. [DOI] [PubMed] [Google Scholar]
- 18.Chen ZQ, Dong J, Ishimura A, Daar I, Hinnebusch AG, Dean M. The essential vertebrate ABCE1 protein interacts with eukaryotic initiation factors. J Biol Chem. 2006;281:7452–7457. doi: 10.1074/jbc.M510603200. [DOI] [PubMed] [Google Scholar]
- 19.Rayala SK, den Hollander P, Balasenthil S, Yang Z, Broaddus RR, Kumar R. Functional regulation of oestrogen receptor pathway by the dynein light chain 1. EMBO Rep. 2005;6:538–544. doi: 10.1038/sj.embor.7400417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Gaudet MM, Chanock S, Dunning A, et al. HSD17B1 genetic variants and hormone receptor-defined breast cancer. Cancer Epidemiol Biomarkers Prev. 2008;17:2766–2772. doi: 10.1158/1055-9965.EPI-07-2891. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Garcia-Closas M, Hall P, Nevanlinna H, et al. Heterogeneity of breast cancer associations with five susceptibility loci by clinical and pathological characteristics. PLoS Genet. 2008;4:e1000054. doi: 10.1371/journal.pgen.1000054. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Stacey SN, Manolescu A, Sulem P, et al. Common variants on chromosome 5p12 confer susceptibility to estrogen receptor-positive breast cancer. Nat Genet. 2008;40:703–706. doi: 10.1038/ng.131. [DOI] [PubMed] [Google Scholar]
- 23.Tamm I, Kornblau SM, Segall H, et al. Expression and prognostic significance of IAP-family genes in human cancers and myeloid leukemias. Clin Cancer Res. 2000;6:1796–1803. [PubMed] [Google Scholar]
- 24.Fong WG, Liston P, Rajcan-Separovic E, St Jean M, Craig C, Korneluk RG. Expression and genetic analysis of XIAP-associated factor 1 (XAF1) in cancer cell lines. Genomics. 2000;70:113–122. doi: 10.1006/geno.2000.6364. [DOI] [PubMed] [Google Scholar]
- 25.Qiu XB, Ouyang SY, Li CJ, Miao S, Wang L, Goldberg AL. hRpn13/ADRM1/GP110 is a novel proteasome subunit that binds the deubiquitinating enzyme, UCH37. EMBO J. 2006;25:5742–5753. doi: 10.1038/sj.emboj.7601450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lakshmanan M, Bughani U, Duraisamy S, Diwan M, Dastidar S, Ray A. Molecular targeting of E3 ligases--a therapeutic approach for cancer. Expert Opin Ther Targets. 2008;12:855–870. doi: 10.1517/14728222.12.7.855. [DOI] [PubMed] [Google Scholar]
- 27.Karamouzis MV, Konstantinopoulos PA, Badra FA, Papavassiliou AG. SUMO and estrogen receptors in breast cancer. Breast Cancer Res Treat. 2008;107:195–210. doi: 10.1007/s10549-007-9552-5. [DOI] [PubMed] [Google Scholar]
- 28.Byun B, Jung Y. Repression of transcriptional activity of estrogen receptor alpha by a Cullin3/SPOP ubiquitin E3 ligase complex. Mol Cells. 2008;25:289–293. [PubMed] [Google Scholar]
- 29.Pharoah PD, Antoniou AC, Easton DF, Ponder PA. Polygenes, risk prediction, and targeted prevention of breast cancer. N Engl J Med. 2008;358:2796–2803. doi: 10.1056/NEJMsa0708739. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.