Estrogen, SNP-Dependent Chemokine Expression and Selective Estrogen Receptor Modulator Regulation

Abstract

We previously reported, on the basis of a genome-wide association study for aromatase inhibitor-induced musculoskeletal symptoms, that single-nucleotide polymorphisms (SNPs) near the T-cell leukemia/lymphoma 1A (TCL1A) gene were associated with aromatase inhibitor-induced musculoskeletal pain and with estradiol (E2)-induced TCL1A expression. Furthermore, variation in TCL1A expression influenced the downstream expression of proinflammatory cytokines and cytokine receptors. Specifically, the top hit genome-wide association study SNP, rs11849538, created a functional estrogen response element (ERE) that displayed estrogen receptor (ER) binding and increased E2 induction of TCL1A expression only for the variant SNP genotype. In the present study, we pursued mechanisms underlying the E2-SNP-dependent regulation of TCL1A expression and, in parallel, our subsequent observations that SNPs at a distance from EREs can regulate ERα binding and that ER antagonists can reverse phenotypes associated with those SNPs. Specifically, we performed a series of functional genomic studies using a large panel of lymphoblastoid cell lines with dense genomic data that demonstrated that TCL1A SNPs at a distance from EREs can modulate ERα binding and expression of TCL1A as well as the expression of downstream immune mediators. Furthermore, 4-hydroxytamoxifen or fulvestrant could reverse these SNP-genotype effects. Similar results were found for SNPs in the IL17A cytokine and CCR6 chemokine receptor genes. These observations greatly expand our previous results and support the existence of a novel molecular mechanism that contributes to the complex interplay between estrogens and immune systems. They also raise the possibility of the pharmacological manipulation of the expression of proinflammatory cytokines and chemokines in a SNP genotype-dependent fashion.

The effects of estrogens on the immune system and on immune response are complex (1). We recently performed a genome-wide association study (GWAS) that identified single-nucleotide polymorphisms (SNPs) near the T-cell leukemia/lymphoma 1A (TCL1A) gene that were associated with risk for aromatase inhibitor (AI)-induced musculoskeletal pain (2). Specifically, estradiol (E2) induced TCL1A expression but only for the variant SNP genotype 3′ of the TCL1A gene that created a functional estrogen response element (ERE). In addition, we reported that downstream estrogen-dependent regulation of the expression of selected cytokines and cytokine receptors was also associated with these SNPs (3, 4). In parallel, in a separate GWAS involving the use of selective estrogen receptor modulators (SERMs) for the chemoprevention of breast cancer (5), we observed that SNPs located hundreds of base pairs away could, in a SNP genotype-dependent fashion, dramatically alter ER binding to an ERE as well as subsequent gene transcription but that SERMs could, at times, reverse that SNP dependence, ie, that they could reverse the effect of SNP genotypes on transcription (5). Those observations raised the possibility that SERMs might be used to modify the expression of cytokines in a SNP-dependent fashion. In the present study, we set out to unite and extend these two sets of functional genomic observations to determine how general the effect of SNPs at a distance from EREs on estrogen effects on cytokine and, possibly, chemokine expression might be, and to determine whether SERMs and ERα antagonists and down-regulators like fulvestrant might be able to regulate that process pharmacologically.

Estrogen receptors (ERs) are transcription factors that can regulate the expression of a large number of genes directly or through the effects of a series of coregulators and signaling proteins (6). Drugs such as SERMs and AIs that interfere with estrogen signaling are a mainstay in the therapy of ER-positive breast cancer, as are ER antagonists like fulvestrant (7). SERMs have also been used for breast cancer prevention and for osteoporosis prevention and treatment (5). In a recent GWAS for SERM breast cancer chemoprevention, we observed that SNPs located hundreds of base pairs away from EREs in intron 2 of the ZNF423 gene, a gene that encodes a zinc finger transcription factor, had a striking genotype-dependent effect on ER binding to those EREs (8). However, 4-hydroxytamoxifen (4-OH-TAM), an active metabolite of tamoxifen (9), could reverse the effect of SNP genotype on both ERα binding and the subsequent expression of ZNF423 (8).

ER-mediated transcription has been studied extensively, with a focus on the promoter regions of genes such as TFF1 (10), BRCA1 (11), NRIP1, MADH9, NME3, TPD52L, and ABCG2 (12) on regions 3′ of genes and on distant enhancer regions (13), and it is now clear that primary sites for estrogen-dependent regulation of transcription are not necessarily in promoter regions (13). In the present study, we set out to explore molecular mechanisms underlying SNP genotype-dependent E2 induction of TCL1A expression. In this case, all of the SNPs studied mapped 3′ of TCL1A. We also determined whether SNPs at a distance from ERE motifs 3′ of TCL1A might influence the expression of that gene and, downstream, of genes encoding cytokines and chemokines and, if so, whether those SNP-dependent effects could be altered or reversed by 4-OH-TAM or fulvestrant. The experiments described subsequently were performed using a human variation panel of 300 human lymphoblastoid cell lines (LCLs) obtained from 300 healthy individual subjects, 100 each from subjects of three different ethnicities, a genomic data-rich cell line model system that has repeatedly demonstrated its value for pharmacogenomic studies and that made it possible for us to select LCLs for study with virtually any common genotype or combinations of genotypes, as described subsequently (3, 8, 14–17).

We observed that TCL1A SNPs at a distance from an ERE could alter ERα binding in parallel with changes in TCL1A expression and that 4-OH-TAM or fulvestrant were capable of reversing the SNP genotype-dependent effects on binding and on expression. Variation of TCL1A expression, in turn, was associated with the downstream expression of a series of proinflammatory cytokines, chemokines, and their receptors, including IL-17RA, IL-17A, chemokine receptor 6 (CCR6) and chemokine C-C motif ligand 20 (CCL20), all of which play important roles in immune-mediated disease (18). Furthermore, some of those genes, eg, IL17A and CCR6, also displayed TCL1A-independent E2 induction that was regulated by SNPs at a distance from EREs in the genes encoding the cytokine, chemokine or their receptors and, even more striking, those SNP-dependent effects could also be reversed in a SNP genotype-dependent fashion by exposure to 4-OH-TAM or fulvestrant. Those observations could be explained, at least in part, by SNP-dependent variation in ERα binding to EREs located up to hundreds of base pairs distant from the SNPs.

In summary, the inflammatory response can be strongly influenced by estrogens (19). In the present studies, we set out to determine whether variation of ER binding modulated by SNPs at a distance from ERE motifs might be a general genomic phenomenon, with a focus on genes that encode immune mediators. The functional genomic studies described subsequently demonstrated that this molecular mechanism may play an important role in the complex interplay between the endocrine and immune systems. They also raised the possibility that SERMs might be used to reverse, in a SNP genotype-dependent fashion, the estrogen-dependent expression of proinflammatory cytokines and chemokines.

Materials and Methods

Identification of ERE motifs

We obtained the reference genomic sequence across the TCL gene cluster on chromosome 14, a 300-kb genomic area that encodes TCL1A, TCL1B, and TCL6 and scanned that region for full- or half-ERE motifs using FIMO software. (http://meme.nbcr.net/meme/fimo-intro.html) BEDTools (http://gitub.com/arq5x/bedtools2) was then used to calculate distances from SNPs to the locations of putative ERE motifs (20). Two of the SNPs imputed during our original GWAS, rs7160302 (P = 2.52E-06) and rs7359033 (P = 3.89E-06), were in tight linkage disequilibrium with rs11849538, the top hit SNP identified during the GWAS. Both of those SNPs had a putative ERE motif located within ±500 bp and, as a result, could be studied using chromatin immunoprecipitation (ChIP) assays, as described subsequently. We also used the TRANSFAC database and Genomatix software to identify EREs in genes encoding cytokines, chemokines, and their receptors using as the reference sequence human genome build 37. We identified putative EREs within ±500 bp of the IL17A gene rs2275913 SNP and the CCR6 gene rs3093023 SNP, both of which we subjected to functional genomic study, as described in Results.

Human variation panel LCLs

The human variation panel model system consists of LCLs from 300 healthy subjects (100 European-Americans, 100 African-Americans, and 100 Han Chinese-Americans). This panel was generated by the Coriell Institute (Camden, New Jersey). We genotyped all 300 cell lines for genome-wide SNPs using Illumina 550K and 510S SNP BeadChips (Illumina), and the Coriell Institute obtained Affymetrix SNP array 6.0 (Affymetrix) data for the same cell lines. These combined SNP genotype data (∼1.3 million genotyped SNPs) were used to impute a total of approximately 7 million SNPs per cell line (15). Basal levels of gene expression were determined in all of these cell lines using the Affymetrix U133 2.0 Plus GeneChip expression array. This LCL model system has been used repeatedly to generate and/or test pharmacogenomic hypotheses arising from clinical GWAS (3, 8, 14–17). Its use in the present study made it possible to evaluate associations between TCL1A SNP genotypes, TCL1A expression, and the expression of cytokines, chemokines, and their receptors. Genotypes for the SNPs discussed subsequently in all 48 individual LCLs used to perform the functional genomic studies included in this report are listed in Supplemental Table 1.

Restriction fragment length polymorphism (RFLP) assays to validate genotypes

The two imputed SNPs near TCL1A that we studied (rs7160302 and rs7359033) were genotyped by RFLP analyses using NcoI and HinP1I (New England Biolabs), respectively. The PCR primer sequences used in those genotyping studies are listed in Supplemental Table 2. Specifically, PCR products (5 μL) were digested using 1 μL of specific restriction enzyme and 10× NEB buffer for each SNP at 37°C for 1 hour. NcoI digestion of fragments containing the C allele for the rs7160302 SNP generated two fragments 92 and 74 bp in length, whereas the fragment containing the T allele remained undigested by the enzyme. Similarly, HinP1I digestion of fragments containing the C allele for the rs7359033 SNP produced fragments 79 and 60 bp in length, whereas the fragment containing the T allele remained undigested. Sanger DNA sequencing was used to validate the genotyping results.

Chromatin immunoprecipitation assays

ChIP assays were performed with LCLs selected on the basis of genotype using the EpiTect ChIP OneDay kit (QIAGEN). DNA-ERα complexes were immunoprecipitated using antibodies directed against ERα (Santa Cruz Biotechnology) with normal mouse IgG as a control. After purification, DNA was subjected to quantitative PCR (qPCR) using the primer sets listed in Supplemental Table 2. Percentage of ChIP DNA/input was determined by qPCR. The level of enrichment of the DNA sequence was determined relative to the total quantity of input DNA (percentage of input). The level of enrichment was expressed as relative enrichment above background (enrichment relative to IgG control). To account for chromatin sample preparation differences, cycle threshold (Ct) values for each immunoprecipitation (IP) DNA fraction were normalized to the input DNA fraction Ct value: δCt (normalized IP) = (Ct [IP] − Ct [input]-log2 [input dilution factor]). 2^{(-δCt normalized IP)} was used to calculate the percentage input for each IP faction. Data were represented as percentage input, (enrichment relative to IgG control) = percentage input (ERα antibody) − percentage input (IgG).

RNA interference and transfection

Small interfering RNA (siRNA) for TCL1A, CCR6, IL17A, and negative controls were purchased from Dharmacon. To perform siRNA gene knockdown, cells (3 × 10⁶) were resuspended with 100 μL of nucleofector solution (Lonza) and 500 nM of siRNA and were electroporated using Nucleofector program X-001 (Lonza). The cells were then transferred into 12-well plates containing preequilibrated RPMI 1640 culture medium to perform the experiments described subsequently.

Drug treatment

LCLs selected based on genotype were cultured in RPMI 1640 media (Cellgro) supplemented with 15% fetal bovine serum (FBS; Atlanta Biologicals). Cells were cultured in RPMI 1640 media containing 5% (vol/vol) charcoal-stripped FBS for 24 hours and were subsequently cultured in FBS-free RPMI 1640 media for another 24 hours before E2 treatment. Cells were treated with increasing concentrations (10⁻⁵, 10⁻³, and 10⁻¹ nM) of E2 for 24 hours. Total RNA was extracted using the Quick-RNA miniprep kit (Zymo). In some experiments, after an initial 24-hour incubation with 0.1 nM E2, cells were treated with increasing concentrations (10⁻¹⁰ and 10⁻⁹ and 10⁻⁸ and 10⁻⁷μM) of 4-OH-TAM or fulvestrant (formerly ICI 182780) for an additional 24 hours. Cells were then collected.

Quantitative real time PCR (qRT-PCR) and Western blot analysis

RNA was isolated from the cells (Zymo). cDNA was then synthesized from 2 μg of total RNA (Bioline). Gene expression was measured using qRT-PCR. GAPDH and ACTB were used as the reference genes. The PCR mixture contained 3 μL of cDNA, 5 μL of 2× SensiMix SYBR low-ROX kit (Bioline), 1 μL of gene-specific primer (QIAGEN), and distilled water to make a 10-μL reaction. PCRs were performed in duplicate using the Applied Biosystems ViiA 7 real-time PCR system (Life Technologies). The 2^{-δδcycle threshold} method was used for statistical data analysis.

For Western blot studies, protein samples were used to perform electrophoresis, followed by transfer to polyvinyl difluoride membranes. The membranes were probed with primary antibody (TCL1A, TCL1B, IL-17A, IL-17RA, CCR6, and CCL20) purchased from Novus Biologicals at a 1:1000 dilution in blocking buffer with gentle shaking at 4°C overnight. Washed membranes were then incubated with secondary goat antirabbit antibody (Novus Biologicals) at a 1:20 000 dilution in blocking buffer at room temperature for 1 hour with gentle shaking. The washed membranes were subsequently incubated in Pierce enhanced chemiluminescence Western blotting substrate (Thermo Scientific) and were visualized using Geldoc (Bio-Rad Laboratories).

Cloning of TCL1A promoter reporter gene constructs

The TCL1A promoter (1162 bp) was PCR amplified using LCL DNA. Primer sequences are listed in Supplemental Table 2. The amplified PCR product for the TCL1A promoter was cloned into the Kpnl and Sacl sites of the PGL3 basic plasmid to create the initial PGL3-TCL1A construct. The sequence of the construct was verified by DNA sequencing. Subsequently, a 548-bp DNA segment of the region 3′ of the terminus of TCL1A that included wild-type (WT) sequences for the rs11849538 and rs7359033 SNPs was PCR amplified. The WT SNP sequence was amplified using LCL genomic DNA as a template from a cell line that was homozygous WT for these SNPs. This 548-bp amplicon contained the rs11849538 SNP that creates an ERE as well as the rs7359033 SNP that is located 5 bp away from an ERE. Purified PCR product was subcloned into the SaIl and BamHl sites to create the PGL3-TCL1A-WT plasmid. We were unable to include the rs7160302 SNP in this reporter construct because this SNP is located approximately 6.8 kb away from the rs11849538 SNP. Sequences of the constructs were verified by Sanger DNA sequencing.

Site-directed mutagenesis

The PGL3-TCL1A-WT construct was subjected to site-directed mutagenesis using the QuikChange Lightning multisite-directed mutagenesis kit (Agilent Technology). To change the WT alleles to the V genotypes, we designed a mutagenic primer for each of those three TCL1A SNPs, including the rs11849538 and rs7359033 and rs7160302SNPs (see Supplemental Table 2).

Luciferase reporter assay

Human embryonic kidney 293 (HEK293)-T, MCF7, and LCLs cells were used to performe luciferase reporter assays. HEK293T and MCF7 cells were transfected using Lipofectamine 2000 (Invitrogen) with 1 μg of the plasmid DNA and 0.1 μg of pRL-SV40 Renilla luciferase reporter as an internal control (Promega) in a 24-well plate. After 24 hours, the cells were incubated for 24 hours with 0.1 nM E2, followed by the addition of 10⁻⁷ μM 4-OH-TAM or 10⁻⁷ μM fulvestrant. Luciferase assays were performed using a dual-luciferase reporter assay system (Promega). Renilla activity was used to correct for a possible variation in transfection efficiency. PGL3 basic vector without an insert was used as a negative control. Expression was calculated as the relative firefly luciferase activity normalized by use of the activity of the transfection control Renilla luciferase. Results are presented as fold change in the relative luciferase units compared with the PGL3 basic vector. Experiments were repeated three times in triplicate.

Statistics

Data were analyzed using GraphPad Prism Software. A Student's t test was used to compare the relative mRNA expression levels between the different genotypes at the same drug concentration. A value of P < .05 was considered statistically significant.

Results

TCL1A SNP genotypes and SERM-induced reversal of ER binding

We reported previously that the rs11849538 SNP had the lowest P value (6.67E-07) in our GWAS for AI-induced musculoskeletal pain in patients enrolled in the MA.27 adjuvant AI breast cancer clinical trial (3). That SNP mapped between the 3′-terminus of TCL1A and the 3′-terminus of TCL1B, as shown graphically in Figure 1, and it created a functional ERE (3). LCLs homozygous for the variant genotype displayed significantly greater estrogen-dependent induction of TCL1A expression than did cells homozygous for the WT allele (3, 4). In our original GWAS report (3), we focused on the estrogen-dependent induction of TCL1A expression in cells homozygous for the variant rs11849538 SNP genotype. As described subsequently, based on data obtained during the present study, there are at least three SNPs, rs11849538, rs7359033, and rs7160302 (see Figure 2A), all in tight linkage disequilibrium (LD), which are included in the SNP signal on chromosome 14 that is shown in Figure 1 and which appear to act in concert to influence the E2-dependent induction of TCL1A expression. In addition, and of importance for the present study, after performing our MA.27 GWAS with the phenotype of musculoskeletal adverse events, we performed an independent GWAS for the occurrence of breast cancer among high-risk patients enrolled in the P-1 and P-2 SERM breast cancer chemoprevention studies, studies that included approximately 33 000 women at elevated risk for breast cancer (8). Functional genomic pursuit of SNP signals from that GWAS showed that SNPs hundreds of base pairs from an ERE can have a striking influence on ER binding to that ERE and subsequent transcription of an adjacent gene (8). Moreover, the P-1 and P-2 GWAS showed that SERMs could reverse the genotype-dependent effect of those distant SNPs with regard both to ERα binding to EREs and subsequent gene expression, ie, if the WT SNP genotype showed greater binding in absence of a SERM, the variant genotype did in the presence of a SERM (8). In the present study, we determined whether these novel observations might provide insight into the SNP genotype-dependent E2 induction of the expression of TCL1A and the expression of downstream cytokines and chemokines.

Figure 1.

Locus Zoom plot showing the association of SNPs identified during the MA.27 musculoskeletal adverse event GWAS within the TCL gene cluster on chromosome 14.

Figure 2.

TCL1A SNPs regulate ERα binding to ERE motifs 3′ of TCL1A. A, Schematic diagrams of two SNP, rs7359033 and rs7160302, in tight LD with rs11849538, the top hit signal from the MA.27 GWAS. Locations of EREs are shown as boxes for these three SNPs that map between the 3′-termini of TCL1A and TCL1B. Distances between the SNPs and EREs are shown and blue arrows indicate the locations of primers used to conduct the ChIP amplifications. B and C, SNP- and estrogen-related variation in TCL1A mRNA expression in LCLs. The cells were treated with varying concentrations of E2 or with 0.1 nM E2 plus increasing concentrations of ICI-182780 (ICI) or 4-OH-TAM (4OH) for an additional 24 hours. Eight cell lines homozygous for the variant (V) genotypes for all three of these SNPs and eight cell lines homozygous for WT genotypes were used in these experiments. D–F, ChIP assays showing the effect of ERα binding to EREs (note that the WT sequence for the rs11849538 SNP did not encode an ERE, so there was no binding; see panel D) in or near the SNPs in response to E2 (0.1 nM), E2 (0.1 nM) plus 4-OH-TAM (10⁻⁷ μM), or ICI (10⁻⁷ μM) treatment of LCLs homozygous for either WT or V genotypes for the rs11849538, rs7359033, and rs7160302 SNPs. Percentage of ChIP DNA/input was determined by qPCR. The level of enrichment was expressed as relative enrichment above background (enrichment relative to IgG control). Data are represented as percentage input (enrichment relative to IgG control) = percentage input (ERα antibody) − percentage input (IgG). *, P < .05, **, P < .005 comparing WT and V SNP genotype cell lines at the same concentrations of E2 and ICI. Three independent experiments were performed (n = 8 for each genotype). G, TCL1A luciferase activity as a function of TCL1A SNP genotypes. HEK293T cells were transfected with the PGL3 basic vector containing the TCL1A promoter and a 548-bp fragment of DNA containing homozygous WT sequences for the rs11849538 and rs7359033 SNPs was subcloned into SaIl and BamHl sites and to create a PGL3-TCL1A-WT reporter construct (referred to as PGL3-TCL1A-WT). This construct was used to perform site-directed mutagenesis and was mutated to variant sequences for rs11849538 and rs7359033 SNPs (referred as PGL3-TCL1A-V). The rs7160302 SNP was 7173 bp distant from TCL1A, so we were unable to include this SNP in the reporter gene construct. Twenty-four hours after transfection, cells were treated with either vehicle or 0.1 nM E2 for 24 hours, followed by 4-OH-TAM or ICI for an additional 24 hours. Luciferase activity was measured 72 hours after transfection. The firefly luciferase activity was normalized by the use of Renilla luciferase to correct for possible variation in transfection efficiency. Results were expressed as fold change in relative luciferase units (RLU) compared with the PGL3 basic vector. Experiments were repeated three times in triplicate.

The studies described subsequently were performed using a genomic data-rich panel of 300 human LCLs that has proved to be a powerful system for the generation and testing of genomic hypotheses (3, 8, 14–17). It is understood that LCLs, although derived from B lymphocytes, are not identical to all cells involved in immune responses, but, as will become clear subsequently, this cell line model system offers a tremendous advantage in that is it possible to study the relationship between virtually any common genetic variant and a cellular phenotype expressed in these cells because we have 300 cell lines from subjects of three major ethnic groups, all with dense genomic data. Specifically, we determined whether SNPs at a distance from EREs near TCL1A as well as genes encoding cytokines, chemokines, and their receptors might influence the effect of E2 on the expression of those genes and, if so, whether treatment with 4-OH-TAM or the ER antagonist and down-regulator fulvestrant might alter ERα binding to those ERE motifs as well as the effect of E2 treatment on gene transcription in a SNP genotype-dependent fashion.

The first set of experiments used eight LCLs homozygous for the TCL1A rs11849538 GWAS top hit WT SNP sequence and eight LCLs homozygous for the variant SNP genotype. As a first step, we exposed these LCLs to increasing concentrations of E2 and then added increasing concentrations of either 4-OH-TAM (Figure 2B) or of fulvestrant (Figure 2C) while maintaining the E2 concentration constant to mimic the in vivo situation in patients treated with these drugs. As can be seen in Figure 2, B and C, the phenomenon of reversal of the SNP genotype effect by SERMs that we had witnessed during our GWAS study of SERM breast cancer prevention (8) was also observed for TCL1A expression. As a control, we tested the possible effect of the treatment of these cell lines with increasing concentrations of 4-OH-TAM or ICI alone and, although there was a slight increase in TCL1A expression for LCLs with the WT genotype, it was only approximately 10% of the increase seen in Figure 2, B and C (see Supplemental Figure 1).

We next performed ChIP assays using LCLs with known genotypes for all three of the linked SNPs near the 3′-terminus of TCL1A after exposure to 0.1 nM E2, 0.1 nM E2 plus 10⁻⁷ μM 4-OH-TAM (4OH in Figure 2) or 0.1 nM E2 plus 10⁻⁷ μM fulvestrant (ICI in Figure 2). The TCL1A top hit rs11849538 SNP variant sequence displayed ERα binding to the ERE created by the variant in response to E2 treatment, but this phenomenon was inhibited, as anticipated, when 4-OH-TAM or fulvestrant was present (Figure 2D). Because there was no ERE motif in the absence of the variant nucleotide, no binding was observed for the WT SNP genotype. These results confirmed our previous observation that the variant genotype for this SNP created a functional ERE, but they did not explain reversal of the expression phenotype when 4-OH-TAM or fulvestrant was added to cells homozygous for the WT genotype for this SNP (see Figure 2, B and C).

Therefore, we extended our experiments to include additional polymorphisms in the cluster of imputed SNPs near TCL1A that were in LD with rs11849538 (R²>0.85) (Figure 1). Two additional SNPs (rs7160302 and rs7359033) with low P values (P = 2.52E-06 and P = 3.89E-06, respectively) were identified. These two imputed SNPs (rs7160302 and rs7359033) were in tight LD (R² = 0.92 and R² = 0.98, respectively) with rs11849538 and were near, but not within, putative ERE motifs. These putative EREs were predicted, as described in Materials and Methods, to be located 20 bp downstream of rs7160302 and 5 bp downstream of rs7359033 (see Figure 2A). We next asked whether these two SNPs might influence estrogen-dependent ERα binding to the ERE motifs located near rs7160302 and rs7359033 and, if so, whether that binding might contribute to the regulation of TCL1A expression. Because the genotypes for rs7160302 and rs7359033 in the LCLs had been imputed, genotypes for both SNPs in the LCLs used in these experiments were confirmed by RFLP and DNA sequencing. Both of the putative EREs near these SNPs were functional based on results of the ChIP assays shown in Figures 2, E and F, but in a striking SNP genotype-dependent fashion. Specifically, even though neither of these SNPs was within a canonical ERE sequence motif, as shown diagrammatically in Figure 2A, enhanced ERα binding was observed in cells homozygous for variant SNP genotypes in the presence of E2, similar to what we had observed for rs11849538 (Figure 2D). However, in the presence of 4-OH-TAM or fulvestrant, the SNP and estrogen-dependent pattern of ERα binding was reversed for both rs7160302 and rs7359033 (Figure 2, E and F), very different from what we observed for the top hit rs11849538 SNP. These findings were compatible with the induction of TCL1A expression in the presence of E2 only in cell lines homozygous for the variant genotypes for all three of these SNPs, but the addition of 4-OH-TAM or the ER antagonist fulvestrant resulted in the reversal of the TCL1A expression, with the induction of TCL1A expression only in the cells homozygous for the WT SNP genotypes in the presence of these drugs (Figure 2, E and F).

In summary, in the presence of only E2, all three SNPs showed higher ER binding only for the variant SNP genotypes (Figure 2, D–F). However, after ER blockade using either fulvestrant or 4-OH-TAM, the rs11849538 top hit SNP showed decreased binding for the variant sequence, as anticipated, but the other two SNPs inverted their binding patterns to show increased binding in cells homozygous for WT SNP genotypes, thus helping to explain the reversal in genotype-dependent TCL1A expression shown graphically in Figure 2, B and C. To pursue these observations further, we also performed luciferase reporter gene assays to provide direct evidence for the functional role of these SNPs in the regulation of TCL1A transcription. Specifically, the TCL1A promoter was amplified from an LCL and was cloned into the Kpnl and Sacl sites of the PGL3 basic vector to create an initial PGL3-TCL1A promoter-containing construct. This construct was then used to make a PGL3-TCL1A WT reporter construct. We subcloned a PCR fragment (548 bp) that included the homozygous WT sequence for the rs11849538 and rs7359033 SNPs, SNPs that are in tight LD, into the SaIl and BamHl sites of the PGL3-TCL1A promoter-containing construct. These two SNPs were then mutated to variant genotype by site-directed mutagenesis. The rs7160302 SNP is located approximately 7 kb away from TCL1A, as shown in Figure 2A. Therefore, we were unable to include this SNP in the reporter gene construct. We observed that the PGL3-TCL1A-V reporter construct displayed substantially increased luciferase activity in response to E2 treatment in HEK293T cells (fold change 22.8, P ≤ .005), and we once again observed the reversal of this effect when 4-OH-TAM or ICI was present, with increased luciferase activity for the WT genotype (fold change 15.1, P ≤ .005) and decreased activity for the variant genotype (Figure 2G). We were also able to replicate these reporter gene results using MCF7 and LCL cells (see Supplemental Figure 2).

This series of observations suggests a complex interplay among the effects of these SNPs, with E2 alone apparently having similar effects on TCL1A transcription for all of the SNPs but with a genotype-dependent reversal of that effect for rs7160302 and rs7359033 in the presence of a SERM or ER antagonist. These results also raised the possibility of a novel mechanism by which drugs might be used to regulate genetically variable estrogen-dependent induction of cytokine and chemokine expression, as described subsequently.

SERMs modulate CCR6 and CCL20 expression in a TCL1A SNP-dependent manner

We reported previously that the SNPs near TCL1A that we identified during our initial MA.27 GWAS (8) were associated with variation in estrogen-dependent TCL1A expression and, downstream, the expression of a series of cytokines and cytokine receptors as well as the transcriptional activity of nuclear factor-κB (NF-κB) (4), with obvious implications for the pathophysiology of immune-mediated diseases (18). However, chemokines are chemotactic cytokines that are key mediators of cell migration during the inflammatory process, and we had not addressed chemokines during our previous studies (18, 21). Excessive expression of chemokines and chemokine receptors can pathologically stimulate symptoms of various immune-mediated diseases (18). Therefore, we determined whether TCL1A might also modulate the expression of chemokines and chemokine receptors in a SNP- and estrogen-dependent fashion. As a first step, we used our LCL model system to determine the relationship between basal level of TCL1A expression and basal expression of chemokines and chemokine receptors across all 300 cell lines. Highly significant correlations (P < 1E-08) were observed between TCL1A expression and the expression of the chemokines CCL20 and CCL5 and the chemokine receptors CCR6, CCR7, and CCR1 (Table 1). The next series of experiments was performed to determine whether the expression of those genes might also be SNP and SERM dependent using the same cell lines from which the data shown in Figure 2 were obtained.

Table 1.

Correlations of TCL1A Expression With Chemokine and Chemokine Receptor Expression in the Human Variation Panel of 300 LCLs

Gene	P Value	Spearman Correlation
CCR7	1.98E-20	−0.515
CCL20	1.07E-19	−0.506
CCR6	1.84E-14	0.436
CCR1	4.31E-12	0.398
CCL5	2.89E-08	−0.323

Basal levels of expression, in the absence of E2, of TCL1A, CCL20, CCL5 CCR6, CCR7, and CCR1 showed no significant differences between cell lines homozygous for all three TCL1A SNP WT genotypes when compared with those homozygous for the variant genotypes. However, CCR6 expression increased approximately 3-fold in LCLs with TCL1A variant SNP genotypes after E2 treatment (Figure 3A), whereas the expression of CCL20, the only known ligand for CCR6 (22), increased significantly in LCLs homozygous for WT SNP sequences (Figure 3B). When 4-OH-TAM was present with E2, a reversal of the pattern of CCR6 expression was observed, with a 3.5-fold increase in cells with WT SNP genotypes and with a decrease to baseline levels in cells with variant SNP genotypes. In the same LCLs, CCL20 expression was up-regulated by E2 and down-regulated when 4-OH-TAM was added in the presence of E2 in cells homozygous for WT TCL1A SNP genotypes, but it remained near baseline levels in cells homozygous for variant genotypes with E2 exposure and increased with 4-OH-TAM exposure, a mirror image of the behavior of LCLs with variant SNP genotypes (Figure 3, A and B). No significant changes were observed in the expression of the CCR1, CCR7, or CCL5 genes during estrogen treatment (data not shown). Because of our previous findings with respect to the relationship between TCL1A SNPs and the expression of IL-17A and IL-17RA (4), we also examined the effect of estrogens and 4-OH-TAM on the expression of those genes in cells with WT and variant SNP genotypes. Those experiments showed that 4-OH-TAM could reverse the effect of the SNPs on the induction of IL-17A in the presence of E2, and a reciprocal effect was observed for IL-17RA (Figure 3, C and D), observations similar to those for CCR6 and CCL20 (Figure 3, A and B). We were also able to replicate these results using an ER blocker, fulvestrant (formerly ICI 182780) (see Supplemental Figure 3).

Figure 3.

E2 induction of TCL1A expression is associated with the expression of CCR6 (A) CCL20 (B), IL-17RA (C), and IL-17A (D). *, P < .05 comparing cell lines homozygous for TCL1A WT and V SNP genotypes at the same concentrations of E2 and 4-OH-TAM. Three independent experiments were performed. Eight cell lines of each genotype were used in each experiment.

TCL1A modulates CCR6 and CCL20 expression

The next series of experiments was designed to determine whether the association between TCL1A and chemokine or chemokine receptor expression (Table 1) might be causal, ie, whether TCL1A was upstream of CCR6 and CCL20. To test that possibility, TCL1A was knocked down using four different individual siRNA and one pooled siRNA in eight LCLs with variant genotypes and eight with WT genotypes for the TCL1A SNPs (Supplemental Figure 4). In Supplemental Figure 4, we have shown both the mean values for each of the eight LCLs of each genotype but also data for each individual cell line. No SNP-dependent difference was observed when TCL1A was knocked down in the absence of E2, ie, at baseline (Figure 4A). We had shown previously that knockdown of TCL1A resulted in a significant decrease in the expression of the cytokine receptor IL-17RA and a significant increase in the expression of its ligand, the cytokine IL-17A at both mRNA and protein levels (4). In the present study, when TCL1A was knocked down to 15% of its baseline level, in a very similar fashion, expression of the chemokine receptor CCR6 decreased approximately 50% and, simultaneously, expression of its ligand, the chemokine CCL20, was up-regulated approximately 2.7-fold (P < .05). We also replicated our previous observations for IL-17A and IL-17RA and observed that the changes in mRNA levels were paralleled by the results of Western blot analysis (Figure 4, A and B). These observations confirmed that TCL1A expression was associated with the expression of IL-17RA, IL-17A, CCR6, and CCL20, as suggested by the relationship of TCL1A SNPs to estrogen-dependent gene expression (Figure 3, A–D). The expression of CCR1, CCR7, and CCL5 did not change significantly when TCL1A was knocked down (Figure 4A).

Figure 4.

TCL1A is associated with the expression of CCR6 and CCL20. A, Relative mRNA expression of TCL1A, TCL1B, TCL6, CCR6, CCL20, CCR1, CCL5, CCRT, IL-17RA, and IL-17A after knockdown of TCL1A using pooled siRNA in lymphoblastoid cell lines. Eight cell lines of each genotype were used in the experiment. *, P < .05. B, Western blot analysis of TCL1A, IL-17A, IL-17RA, CCR6, CCL20, TCL1B, and ACTB in LCLs after TCL1A was knocked down using pooled TCL1A siRNA. Three independent experiments were performed.

TCL1B cannot regulate the expression of IL-17RA, IL-17A, CCR6, and CCL20

The top SNP in our original GWAS, rs11849538, is within the TCL gene cluster on chromosome 14 and maps between the 3′-terminus of TCL1A and the 3′-terminus of TCL1B (Figure 1). TCL1A and TCL1B display a high degree of amino acid sequence homology (23). Therefore, we wanted to determine whether the two genes in this cluster other than TCL1A, TCL1B and TCL6, might also display SNP genotype and estrogen-dependent expression. The E2 induction of TCL1B expression was most prominent in the cells homozygous for the TCL1A variant SNP sequence, and the reversal of TCL1B expression was observed when 4-OH-TAM was present (Figure 5A). TCL6 is a nonprotein coding gene. Although we observed significant induction of TCL6 mRNA expression in response to E2 treatment, that induction was not SNP dependent (Figure 5B). Because both TCL1A and TCL1B can be up-regulated by E2 in a SNP-dependent fashion (Figures 2A and 5A), we determined whether TCL1B might also participate in the regulation of the expression of IL-17RA, IL-17A, CCR6, and CCL20. When TCL1B was knocked down to less than 23% of baseline in LCLs, TCL1A expression did not change significantly, confirming the specificity of the siRNA (Figure 5C). However, no significant changes in the protein expression of IL-17RA, IL-17A, CCR6, or CCL20 were observed when TCL1B was knocked down (Figure 5C), indicating that TCL1B expression did not appear to be directly associated with the expression of IL-17RA, IL-17A, CCR6, and CCL20.

Figure 5.

TCL1A SNP- and estrogen-dependent variation in TCL1B (A) and TCL6 (B) mRNA expression in LCLs. The figure shows the effect of 24-hour incubations with varying concentrations of E2 or 0.1 nM E2 plus increasing concentrations of 4-OH-TAM for an additional 24 hours. A panel of eight homozygous WT and eight homozygous variant (V) LCLs was used to perform the study. *, P < .05 comparing WT and V SNP genotype cell lines at the same concentrations of E2 and 4-OH-TAM. Data represent mean ± SEM. Three independent experiments were performed. Eight cell lines of each genotype were used in the experiment. C, Western blot analysis of TCL1A, IL-17A, IL-17RA, CCR6, CCL20, TCL1B, and ACTB in LCLs after TCL1B was knocked down.

Direct SNP-E2-SERM effects on CCR6 and IL-17A expression

In the preceding experiments, we had observed a SNP-dependent and 4-OH-TAM- or fulvestrant-dependent reversal of the expression of TCL1A and, downstream, the expression of proinflammatory cytokines and chemokines (Figures 2 and 3). This indirect (indirect because it was mediated through TCL1A expression), SNP genotype-dependent regulation of cytokine and chemokine expression in the presence of SERMs might have clinical implications. For example, anticytokine therapies have been used to treat rheumatic diseases, and IL-17 is a potential druggable target for the treatment of rheumatoid arthritis (RA) and psoriatic arthritis (24). In a similar fashion, SNPs in the CCR6 gene have been reported to represent risk factors for the occurrence of RA (25), and the preceding experiments demonstrated indirect (once again mediated through TCL1A) SNP and SERM-dependent regulation of the expression of CCR6 (see Figure 3).

We next asked whether similar mechanisms might play a direct role in the regulation of the expression of important cytokines and chemokines. As a first step, we screened for possible ERE motifs within the IL17RA, IL17A, CCR6, and CCL20 genes as well as the promoter regions of those genes by use of the TRANSFAC database and Genomatix software. Three putative half-palindromic ERE motifs were found in the IL17A promoter that were located 106, 13, and 85 bp, respectively, away from the rs2275913 SNP, as shown graphically in Figure 6A. This IL17A SNP had previously been reported to be associated with a risk for rheumatic diseases such as RA (26). At least some of the putative EREs near this SNP were functional based on the results of ChIP assays, as shown in Figure 6B, assays that displayed differential ERα binding in a SNP- and estrogen-dependent fashion. Because our LCL cell model system makes it possible to select LCLs for study with virtually any common genotypes or combinations of genotypes, the cell lines used to perform the experiments shown in Figure 6 were homozygous for the IL17A SNP genotypes being studied, but they were also all homozygous WT for the three TCL1A SNPs shown in Figure 2A to prevent confusion based on any possible effects of the TCL1A SNPs. The IL17A promoter SNP (rs2275913) was not within a canonical ERE sequence motif but was at a distance away from the three EREs shown graphically in Figure 6A. However, we observed significantly enhanced ERα binding to these EREs in cells homozygous for the variant IL17A SNP genotype as compared with cell lines homozygous for the WT genotype in the presence of 0.1 nM of E2 (Figure 6B). Furthermore, the pattern of ER binding was reversed in the presence of fulvestrant or 4-OH-TAM, consistent with our observations of IL17A SNP- and estrogen-dependent variation in gene expression as shown in Figure 6, C and E. We were unable to identify any putative ERE motifs within the IL17RA gene or its promoter. However, IL17RA expression was altered in an IL17A SNP and estrogen-dependent fashion, but with the opposite genotype dependence, ie, the WT genotype, not the variant, was associated with increased E2-dependent expression, whereas the variant genotype displayed increased expression when a 4-OH-TAM or ICI was added (Figure 6, D and F). Furthermore, after IL-17A was knocked down with two independent siRNAs, we observed an induction of IL-17RA protein expression (Figure 6G), compatible with the changes that we observed in IL-17RA mRNA expression shown in Figure 6, D and F.

Figure 6.

IL17A SNPs are associated with ERα binding to ERE motifs. A, Schematic diagram of the IL17A promoter rs2275913 SNP. The locations of three EREs (referred to as ERE1, ERE2, and ERE3) that mapped near the SNP are indicated. Distances between the SNP and EREs are listed under the ERE boxes, and blue arrows indicate the locations of primers used to perform the ChIP amplifications. B, ChIP assays showing ERα binding to EREs near the SNP in response to E2 (0.1 nM), E2 (0.1 nM) plus 4-OH-TAM (4OH in the figure) (10⁻⁷ μM), or ICI (10⁻⁷ μM) treatment of LCLs homozygous for either WT or V SNP genotypes for rs2275913. C–F, Effects of E2 and E2 plus 4-OH-TAM or ICI on the expression of IL-17A (C and E) and IL-17RA (D and F). A panel of eight homozygous WT and eight homozygous variant (V) LCLs was used to perform this study. All LCLs were homozygous for WT genotypes for the SNPs 3′ to TCL1A. *, P < .05, **, P < .005 comparing WT and V SNP genotype cell lines at the same concentrations of E2 and 4-OH-TAM or ICI. G, Western blot analysis of IL-17A, IL-17RA, and ACTB in LCLs after IL-17A was knocked down using two individual siRNAs. Three independent experiments were performed.

When we took a similar approach to the CCR6 gene, two putative EREs predicted by the TRANSFAC database and Genomatix software were present at a distance from the rs3093023 SNP in intron 1 of CCR6. These ERE motifs were located 261 bp upstream and 188 bp downstream of rs3093023, as shown graphically in Figure 7A. Of interest is the fact that this SNP has been associated with an RA risk on the basis of the results of GWA studies (25, 27, 28). LCLs homozygous for the two genotypes for this SNP were used to determine whether it might influence ER binding to the putative EREs as we had observed for TCL1A and IL-17A. Once again, the cell lines used in these experiments were homozygous for WT genotypes for the TCL1A and IL17A SNPs described earlier. ChIP assays showed a significant increase in ERα binding to both EREs in cells homozygous for CCR6 variant SNP genotypes as compared with cell lines homozygous for the WT sequences in the presence of 0.1 nM of E2 alone. However, reversal of the ERα binding patterns for both EREs was observed when 4-OH-TAM or the ER antagonist fulvestrant was present (Figure 7, B and C), a pattern that was associated with a dose-dependent variation in CCR6 expression in a SNP and estrogen-dependent fashion (Figure 7, D and F). Finally, CCL20 is the only known ligand for CCR6, and the expression of CCL20 was also altered in a CCR6 SNP and estrogen-dependent fashion but, once again, with genotype dependence that was opposite to that seen for CCR6 itself (Figure 7, E and G). In addition, CCL20 protein expression was up-regulated when CCR6 was knocked down with two different siRNAs (Figure 7H). Those results paralleled our observations with regard to the reciprocal relationship between the expression of ligand (IL-17A) and receptor (IL-17RA) for IL-17A/IL-17RA (Figures 6G and 7H).

Figure 7.

CCR6 SNPs are associated with ERα binding to ERE motifs. A, Schematic diagram of the two EREs near the CCR6 rs3093023 SNP. Distances between the SNP and the EREs are listed under the ERE boxes, and blue arrows indicate the locations of primers used to conduct the ChIP amplifications. B and C, ChIP assays showing ERα binding to the two EREs (referred to as ERE1 and ERE2) near the SNP in response to E2 (0.1 nM), E2 (0.1 nM) plus 4-OH-TAM (10⁻⁷ μM), or ICI (10⁻⁷ μM) treatment of LCLs homozygous for either WT or V SNP sequences for rs3093023. D–G, Effects of E2 and E2 plus 4-OH-TAM or fulvestrant (ICI-182780) on the expression of CCR6 (C and F) and CCL20 (D and G). A panel of eight homozygous WT and eight homozygous variant (V) LCLs was used to perform the study. All LCLs were homozygous for WT genotypes for the SNPs 3′ to TCL1A. *, P < .05, **, P < .005 comparing WT and V SNP genotype cell lines at the same concentrations of E2 and 4-OH-TAM or ICI. H, Western blot analysis of CCR6, CCL20, and ACTB in LCLs after CCR6 was knocked down using two individual siRNAs. Three independent experiments were performed.

Discussion

In the present study, we have demonstrated the SNP and estrogen-dependent induction of the expression of TCL1A and, downstream, of immune mediators that include CCR6, CCL20, IL-17RA, and IL-17A, induction that could be reversed by the SERM 4-OH-TAM and by estrogen receptor blockade with fulvestrant. These findings complement and greatly extend our previous reports of a novel genetic mechanism for the estrogen-dependent regulation of the expression of inflammatory mediators (3, 4). Even more striking, the present study showed that SERMs could reverse this SNP genotype-dependent ERα binding, that is, they could reverse the effects of the SNP genotypes, raising the possibility of pharmacological manipulation of ER binding and subsequent phenotypes. Specifically, if the SNP genotypes were known, the molecular phenotypes, in this case downstream expression of a chemokine receptor (CCR6) or a cytokine receptor (IL-17RA), might be altered by drug (SERM) therapy.

As a first step in these studies, we examined the correlation in the human variation panel of baseline TCL1A mRNA expression with that of a series of genes encoding chemokines and chemokine receptors. We observed highly significant correlations with the expression of several of those genes, including CCR6 and its ligand CCL20 (Table 1). We also determined the effect of varying doses of E2 on the expression of mRNA for the three TCL genes on chromosome 14, TCL1A, TCL1B, and TCL6. As shown in Figures 2 and 5, the expression of all three genes was induced by estrogen exposure, but only TCL1A and TCL1B displayed significant genotype-dependent induction by E2, with higher expression when the variant genotype for the rs11849538 SNP, the genotype that created an ERE, was present.

Similar experiments were then performed for the CCR6, CCL20, IL17RA, and IL17A genes (Figure 3). All of these genes displayed TCL1A SNP genotype-dependent induction by E2. It should be emphasized that the SNPs used to select the cell lines for study were those 3′ of TCL1A. The SNPs were not in the CCR6, CCL20, IL17RA, or IL17A genes themselves. It should also be emphasized that the genotypes associated with enhanced induction for receptor and ligand genes differed, with the two receptor genes, CCR6 and IL17RA, being induced by E2 in the presence of the variant TCL1A SNP genotypes, whereas the genes encoding their ligands were induced in cells homozygous for the WT genotypes. These observations were confirmed by TCL1A knockdown and Western blot studies (Figure 4). At this point, we had not only confirmed that variant genotypes for SNPs 3′ of the TCL1A gene on chromosome 14 were associated with enhanced E2 induction of the expression of that gene but also that the induction of the expression of downstream genes that encoded receptors for an important chemokine receptor (CCR6) and an important cytokine receptor (IL-17RA) were induced in parallel with TCL1A (ie, increased expression was associated with the variant TCL1A SNP genotypes). All of these results were compatible with a role for genetic variation near the TCL1A gene in the regulation of the estrogen-dependent induction of selected chemokine and cytokine receptors. The next experiments were performed in an effort both to explore mechanism underlying these SNP effects and to move toward the possibility of using these observations to design therapies that might be used to regulate chemokine and cytokine expression.

We had previously demonstrated a novel molecular mechanism by which SNPs that map hundreds of base pairs away from an ERE can alter ERα binding to that response element and influence transcription regulation (8). We understand that there have been previous reports that sequence variation within an ERE can alter receptor binding, but those reports rarely focused on common genotypes for SNPs at a distance from the ERE motif (29). We have now identified several examples of this mechanism using our genomic data rich LCL model system. For example, the IL17A promoter polymorphism rs2275913 has been associated with giant cell arthritis and breast cancer (30, 31), and the CCR6 rs3093023 polymorphism that we studied has been reported to be relevant for the pathogenesis and/or risk for RA (25, 27, 28), all of which are diseases that occur predominantly in women. Although none of the SNPs included in the present study were located within known DNA transcription factor binding motifs, these SNPs influenced ER binding and transcriptional regulation, perhaps by altering the binding of coregulators (32).

The role of estrogens in immunity is complex (33). SERMs have antiinflammatory effects in microglial cells in rodent models (34) with neuroprotective effects on microglia and astroglia during local inflammation in the brain (35) as a result of regulating ER-mediated mechanisms and by inhibiting NF-κB-induced transcription of proinflammatory cytokines and chemokines (36). We have determined that similar genomic mechanisms might contribute to variation in the expression of cytokines, chemokines, and their receptors. The genotype-dependent SERM-induced reversal of the expression of genes encoding proinflammatory cytokines and chemokines provides a possible mechanism for SNP genotype-dependent cytokine and chemokine transcriptional regulation before and after ER blockade that might be relevant to the immunopathology of disease. CCR6, CCL20, IL-17RA, and IL-17A all play important roles in inflammation (37, 38), and this series of experiments has demonstrated that genetic and estrogen-dependent variation in the expression of these inflammatory mediators might potentially be manipulated by drug therapy. Future clinical studies will be required to determine whether pharmacologic modulation of this mechanism might be useful for either the prevention or treatment of disease.

In this series of experiments, we have performed functional genomic studies using an LCL model system that consists of 300 cell lines from individual healthy subjects for which dense genome-wide genomic data have been obtained (15). This model system makes it possible to select cell lines for study with virtually any common genotype or combinations of genotype, and it has already proven to be a powerful tool for both generating and testing functional genomic hypotheses (3, 8, 14–17). However, we understand that this system, like any model system, has limitations, so the results reported here will have to be replicated by future studies conducted with additional cell lines and clinical samples. We began by studying EREs within ±500 bp of the SNP by performing ChIP assays, assays that place a limit on PCR amplicons that ranges from 100 to 500 bp in length. We have shown that SNPs hundreds of base pairs away from an ERE can have striking effects on ER binding and on subsequent gene transcription and that this process can be regulated by SERMs. It is possible that SNPs greater than 500 bp away from EREs could also regulate SNP- and estrogen-dependent ER binding and subsequent transcription. It has been estimated that there are more than 10 000 ERE binding motifs within the human genome, so the possible implications of these observations could be significant (39–41). It remains to be determined what proportion of EREs might display similar SNP- and SERM-dependent differences in ERα binding, and/or subsequent transcriptional regulation that might contribute to a clinical phenotype. Of particular importance, the role of “coregulators” in the function of ERα as a transcription factor might be also involved in mechanisms underlying the SNP-estrogen-SERM-dependent effects reported here (42). Finally, we showed that the TCL1A SNP-estrogen-dependent effects on gene expression were clearly due to the SNPs located 3′ of TCL1A on the basis of the activity of luciferase reporter gene constructs (Figure 2G).

In summary, we initiated this series of studies by performing a GWAS using DNA from postmenopausal women receiving AIs to treat ER-positive breast cancer in an effort to identify SNPs related to musculoskeletal pain during AI therapy. We observed that SNPs 3′ of TCL1A were associated with this phenotype and that they could influence TCL1A estrogen-dependent gene expression in a SNP-dependent fashion, with downstream effects on the expression of a series of cytokines and chemokines. The functional studies described here have demonstrated what appears to be a general genomic phenomenon, striking differential SNP effects on ERα binding to EREs at a distance from the SNPs, with SERM-induced reversal of both ER binding and downstream phenotypes, including the expression of inflammatory mediators, not merely affecting TCL1A but also, independently, IL17A and CCR6. These results greatly extend our original observations and highlight a novel molecular mechanism that may play an important role in the complex interplay between estrogens and immune systems. Our findings also raise the possibility of pharmacological regulation of the expression of inflammatory mediators in a SNP genotype-dependent fashion.

Additional material

Supplementary data supplied by authors.