Genome-wide case-control study of musculoskeletal adverse events and functional genomics in women receiving aromatase inhibitors: going beyond associations

In postmenopausal women with early breast cancer, the third-generation aromatase inhibitors (AIs; anastrozole, exemestane, and letrozole) have been proven to be of value in multiple large well-conducted clinical trials as initial adjuvant endocrine therapy, after several years of tamoxifen, and as extended adjuvant endocrine therapy after about 5 years of tamoxifen [1-8]. A recent American Society of Clinical Oncology clinical practice guideline stated that an AI should be used in the adjuvant setting either initially or after some period of tamoxifen therapy [9]. 
 
Despite the proven value of the AIs, many women are not adherent [10]. About one-half of women treated with an AI have new or worsening joint complaints [11]. The importance of musculoskeletal complaints was identified in reviewing the experience with MA.27, a large phase III trial comparing the non-steroidal AI anastrozole with the steroidal AI exemestane as adjuvant therapy for early breast cancer. These musculoskeletal adverse events (MS-AEs) were the most common reason why patients discontinued AI therapy. We had previously demonstrated marked variability in metabolism and pharmacodynamics of one of the AIs, anastrozole [12]. We hypothesized that the variability seen with respect to MS-AEs in women treated with anastrozole or exemestane on MA.27 could be related to genetic variability of the patients. We proceeded to perform a genome-wide association study (GWAS) aimed at identifying SNPs associated with MS-AEs. The results of this GWAS and the functional genomic laboratory studies performed have recently been published [13]. This short communication summarizes the highlights of this work plus a commentary on future pharmacogenomic studies of anti-cancer agents.


Introduction
In postmenopausal women with early breast cancer, the third-generation aromatase inhibitors (AIs; anastrozole, exemestane, and letrozole) have been proven to be of value in multiple large well-conducted clinical trials as initial adjuvant endocrine therapy, after several years of tamoxifen, and as extended adjuvant endocrine therapy after about 5 years of tamoxifen [1][2][3][4][5][6][7][8]. A recent American Society of Clinical Oncology clinical practice guideline stated that an AI should be used in the adjuvant setting either initially or after some period of tamoxifen therapy [9].
Despite the proven value of the AIs, many women are not adherent [10]. About one-half of women treated with an AI have new or worsening joint complaints [11]. Th e importance of musculoskeletal complaints was identifi ed in reviewing the experience with MA.27, a large phase III trial comparing the non-steroidal AI anastrozole with the steroidal AI exemestane as adjuvant therapy for early breast cancer. Th ese musculoskeletal adverse events (MS-AEs) were the most common reason why patients discontinued AI therapy. We had previously demonstrated marked variability in metabolism and pharmacodynamics of one of the AIs, anastrozole [12]. We hypothe sized that the variability seen with respect to MS-AEs in women treated with anastrozole or exemestane on MA.27 could be related to genetic variability of the patients. We proceeded to perform a genome-wide association study (GWAS) aimed at identifying SNPs associated with MS-AEs. Th e results of this GWAS and the functional genomic laboratory studies performed have recently been published [13]. Th is short communication summarizes the highlights of this work plus a commentary on future pharmacogenomic studies of anticancer agents.

Methods
MA.27 is a randomized clinical trial conducted by the Breast Cancer Intergroup of North America that was coordinated by the NCIC Clinical Trials Group [13]. Eligible patients were those with a postmenopausal status and resected American Joint Committee on Cancer (version 6) stage I to III breast cancer that was hormone receptor positive. Patients were randomized to anastro zole or exemestane for a period of 5 years. A total of 6,827 women in North America were randomized and the majority provided DNA and consent for its use in genetic studies.
Cases were defi ned as those patients who developed a MS-AE, as previously defi ned [13], and each case was matched to two controls. Th e design utilized was a nested matched cas e-control study and the primary analyses were based on conditional logistic regression. Th e RIKEN Center for Genomic Medicine performed genotyping with the Illumina Human610-Quad platform. Imputation and fi ne mapping were performed in the region of interest on chromosome 14 containing the SNPs with the smallest Pvalues. Functional genomic studies, relating to the SNPs on chromosome 14, were performed using electrophoretic motility shift (EMS) assays, chromatin immuno preci pitation (ChIP) assays, and transfection studies.

Results
We studied 293 cases and 585 controls, and cases and controls were well balanced for all factors except prior hormone replacement therapy, which was signifi cantly higher in cases than controls (66% versus 44%), and fractures within the past 10 years, which were slightly higher in cases than controls (13% versus 9%). A total of 551,395 SNPs were used in the association analyses after exclusion of genotype failures (n = 11,281), SNPs with a minor allele frequency <0.01 (n = 29,478), and SNPs with a departure (P < 1E-06) from Hardy-Weinberg equili brium (n = 82). Th e conditional logistic regression analy ses were adjusted for population stratifi cation and revealed three SNPs (rs7158782, rs7159713, rs2369049) on chromosome

© 2010 BioMed Central Ltd
Genome-wide case-control study of musculoskeletal adverse events and functional genomics in women receiving aromatase inhibitors: going beyond associations James N Ingle* 14 in high linkage disequilibrium with the smallest Pvalues (7.74E-07 to 2.23E-06), which approach the Bonferroni threshold of 1E-07. Subse quently, imputation and fi ne mapping revealed an additional SNP (rs11849538), also on chromosome 14 and in high linkage disequilibrium with the three geno typed SNPS, with the smallest P-value (6.67E-07). Th ese four SNPs were all found to be close (that is, within 7,109 bp) of the 3' end of the T-cell leukemia 1A (TCL1A) gene with the closest SNP (rs11849538) being only 926 bp away.

S H O R T CO M M U N I C AT I O N
Initially, we determined that TCL1A is variably expressed in 288 lymphoblastoid cell lines from three diff erent ethnic groups for which we have expression array and genome-wide SNP data. Functional genomic studies were performed with the three genotyped SNPs and the one imputed/fi ne mapped SNP to examine for any functional signifi cance. Lymphoblastoid cell nuclear extract used in EMS assays showed a 'shift' , that is, protein binding, for three of the SNPs (rs11849538, rs7158782, rs7159713) with less binding by the variant than wild-type sequences. Of particular interest as it relates to a drug that lowers estrogen levels, a TRANSFAC database search predicted that the SNP (rs11849538) with the smallest P-value would create an estrogen response element and this was confi rmed with a ChIP assay utilizing lymphoblastoid cells with known genotype for this SNP that had been transfected with estrogen receptor (ER)α. TCL1A expression was linked to estrogen by exposing U20S cells that had been stably transfected with ERα or ERβ to 0.1 nM E2, and demon strating eightand six-fold increases in TCL1A mRNA expression after 18 hours and 1 hour, respectively. To determine the eff ect of these four SNPs on estrogen-dependent TCL1A expres sion, lymphoblastoid cell lines with genotypes known for the four SNPs were transiently transfected with ERα and exposed to various concen trations of estradiol. All three ethnic groups showed greater TCL1A expression with the variant than the wild-type sequence. Because the clinical picture of the MS-AEs is reminiscent of a chronic infl ammatory state, we utilized the same 288 lymphoblastoid cell lines noted above and determined the correlation between TCL1A expression and expres sion of IL17 and the IL17 receptor A (IL17RA). Remark ably, TCL1A and IL7RA expression were highly correlated (P < 1.9E-10). In studies using U20S cells transfected with either ERα or ERβ, small interfering RNA (siRNA) knockdown of TCL1A resulted in decreased expression of IL17RA but increased expression of IL17, whereas TCL1A overexpression resulted in increased IL17RA expression and decreased expression of IL17.

Discussion
We identifi ed four SNPs on chromosome 14 that were associated with MS-AEs in women receiving AIs as adjuvant therapy for early-stage breast cancer. Given that AIs perturb estrogen levels in these women, we examined the relationship between the SNPs and estrogen action. Of particular note is that the SNP with the smallest Pvalue (rs11849538) created an estrogen response element that was shown by ChIP assay to be functional. Estrogens induced TCL1A expression that was signifi cantly higher in cells with the variant SNPs than those with the wildtype sequence. Expression of TCL1A was directly associated with IL17RA expression in that siRNA knockdown of TCL1A resulted in decreased expression of IL17RA and increased expression of IL17 whereas overexpression of TCL1A resulted in increased expression of IL17RA and decreased expression of IL17.
Th e identifi cation of a relationship between TCL1A expression and the infl ammatory cytokine IL17 is particu larly noteworthy given the recent identifi cation of a third lineage of T-helper cells (in addition to Th 1 and Th 2), that is, Th 17, that is associated with chronic infl ammation [14], and in which IL17 plays a central role. IL17 and IL17RA represent potential targets and eff orts to target these factors in a variety of infl ammatory diseases are in progress (reviewed in [14]).
Th e study reported here demonstrates the value of GWASs. A GWAS is, by defi nition, a 'hypothesis-free' interrogation of the entire genome and, in our study, identifi ed a gene, TCL1A, which was totally unexpected. However, it was clearly necessary to perform the functional genomic studies to begin to understand the potential biologic implications of these fi ndings. Clearly, replication is an important consideration in the performance of a GWAS because of the problem of false positive associations [15]. Th e replication of genomewide pharmacogenomic studies of anti-cancer agents will likely be found to be substantially more problematic than genome-wide studies examining risk of developing a cancer. For example, in a recent study to identify breast cancer susceptibility alleles, a GWAS was performed in 3,659 cases and 4,897 controls, and then promising associations were then studied in 12,576 cases and 12,223 controls [16]. In the case of pharmacogenomic studies of anti-cancer agents, a GWAS will likely be best conducted utilizing patients entered into a prospectively conducted clinical trial. MA.27, the study from which the patients in our GWAS were obtained, is such a trial but is the largest study conducted of AIs alone, taking 8 years to complete at a cost of tens of millions of dollars. Replication will not be possible utilizing a larger cohort of patients as it simply does not exist. Although the conduct of rigorous and insightful functional genomic studies can be viewed as important in pursuing leads from any GWAS, such studies may have greater importance in the case of anticancer agents because of limitations imposed by a relative paucity of patient cohorts, at least compared to studies examining risk, that is, genetic predisposition, of developing cancer.
In conclusion, our GWAS identifi ed four SNPs related to MS-AEs in women receiving AIs as adjuvant therapy for their early breast cancer. Th ese SNPs were related to TCL1A, which in turn was related to the infl ammatory cytokine IL17. Further work is ongoing in our laboratory to expand our understanding of the mechanisms involved in the TCL1A-IL17 relationship. Ultimately, this new know ledge should help identify means to ameliorate the AI-related MS-AEs in order to allow more women with early breast cancer to take these potentially life-saving drugs.

Competing interests
The author declares that he has no competing interests.