We describe an integrative genomics approach that combines GWAS information with gene expression data to identify molecular signatures, biological pathways and gene regulatory networks, which are dysregulated in response to tamoxifen treatment in ER+ breast cancer patients. Our work has the clinical goal of better understanding the molecular mechanisms underlying variable response to tamoxifen treatment in ER+ breast cancer patients. Key findings from this study can be summarized as follows: (a) Genes containing SNPs associated with risk for breast cancer are dysregulated in response to tamoxifen treatment and could distinguish treated from untreated breast cancer patients. (b) Within tamoxifen treated patients, genes containing SNPs associated with risk for breast cancer were able to stratify patients into two groups. This is an important finding in that it could guide stratification of patients into poor and good responders a key step in guiding patient treatment at point of care. This finding could also allow identification of genes which contribute to variable response to tamoxifen treatment a major step in identifying molecular mechanisms which may contribute to patients’ resistance to treatment and guide personalized treatment. (c) Pathways and gene regulatory networks containing SNPs associated with risk for breast cancer were found to be dysregulated in response to tamoxifen treatment. This is an important finding in that such candidate pathways if confirmed could be targeted for therapy. To the best of our knowledge such findings have not been previously reported in breast cancer.
Clearly, application of GWAS information in clinical practice has been complicated by the fact that studies of risk alleles for breast cancer have shown that although GWAS can identify novel genes that contribute to risk, the odds ratios and effect sizes as determined by P
-values are relatively small.10,11
However, the results in this study demonstrate that genes containing SNPs with small P
-values and SNPs replicated in multiple independent studies interact with genes containing SNPs with moderate P
-values in biological pathways which influence response to tamoxifen treatment. This finding has two important implications: first, it allows identification of potential candidate genes and candidate pathways that could serve as drug targets if confirmed. Second, it demonstrates that genes containing SNPs with small P
-values act in concert with those containing SNPs with moderate P
-values. This is an important aspect of these results given that most of the loci found to date are small and replication in most GWAS studies tends to be elusive.
Currently, few clinically relevant genome-wide association studies of drug response phenotypes on breast cancer have been reported that it is impossible to effectively compare our results. However, the results reported in this study are consistent with the 12 pharmacogenomics-based GWAS on various diseases reported and summarized by Crowley et al.36
The main difference between our study and these studies is that rather than identifying polymorphisms altering drug response, our focus was on identifying genes and key biological pathways containing SNPs, which are key drivers of response to tamoxifen treatment. The results demonstrate that integration of GWAS information with gene expression data on breast cancer patients uniformly treated with tamoxifen is an essential tool to identifying genes and key biological pathways in which polymorphisms might affect the disposition or response to tamoxifen treatment. Although this study focuses on tamoxifen, this approach could be applied to any given drug where GWAS information and gene expression are available.
The results in this study show that response to tamoxifen involves many genes and multiple pathways. These results are consistent with earlier reports based on expression profiling,6,8
and those reported recently by Mendes-Pereira et al.9
However, one caveat is important in this study. The results in this study do not show how individual SNPs or alleles contribute to variable response to tamoxifen treatment. This is a key limitation of this study and is acknowledge here. However, the results of this study provide a proof-of-concept and information about the biological pathways in which SNPs associated with risk for breast cancer operate. The practical clinical utility of that type of information is that it could guide future experimental designs to identify candidate genes and candidate pathways containing SNPs, which are key drivers of drug response and could be potential targets for therapy. In fact, although we did not investigate allele- specific expression or the effect of genetic variants on gene expression, previous studies have demonstrated that individual alleles could affect gene expression in humans.37,38
For example, a recent study showed that allele specific up-regulation of the FGFR2 (the most replicated gene in GWAS) increased susceptibility to breast cancer.39
The potential clinical relevance of the results reported in this study can be summarized as follows: (1). High-throughput SNP-mapping combined with transcription profiling could potentially allow cancer associated drug targets to be identified thereby reducing attrition in early-phase clinical trials. For example, the cost of additional clinical trials might be reduced if the population of responders and nonresponders could be segmented on the basis of their genetic profiles in early phases of clinical trials.40
SNPs identified from candidate genes in early phases of clinical trials could allow non-responders to be excluded from subsequent clinical trial studies, therefore potentially allowing enriched, smaller, faster, less expensive clinical studies on patients with better chance of responding favorably.40
(2) Although classic response to tamoxifen treatment has been assessed by polymorphisms in the CYP2D6 gene, the results from this study and others studies,6,8
show that response to tamoxifen treatments is under polygenic control. The identification of multiple pathways that are dysregulated in response to tamoxifen treatment tends to suggest that global pathway crosstalk may be involved. (3) The results show that integrative analysis combining GWAS information with gene expression data could potentially identify candidate genes and potential drug targets that lie outside of the current range of knowledge. This approach could also potentially provide novel biological insights into the mechanisms of drug action and resistance. However, further studies will be required to determine how individual SNPs influence gene expression and response to tamoxifen treatment before firm conclusions of the practical utility of GWAS information in a clinical setting can be drawn. Such investigation though warranted, was beyond the scope of this report, but is the subject of our future studies.
Several studies from our own group10,11
and others breast cancer41,42
have reported pathway-based approaches to dissection of the genetic susceptibility architecture of breast cancer. To our knowledge, this is the first study to associate GWAS information with response to tamoxifen treatment and to identify genes and biological pathways dysregulated in response to tamoxifen treatment. Recently, Genomic Health (Redwood City, CA) developed the Oncotype DX diagnostic assay based on candidate gene selection (not genome wide) approach.7
The multiplex 21-gene test includes genes associated with proliferation, estrogen, and HER2 action, invasion, and five control genes. This 21-gene recurrence score assay provides a recurrence score for node-negative breast cancer patients with ER+
tumors who have received adjuvant tamoxifen.7
The association between this 21-gene recurrence assay and risk of locoregional recurrence in node-negative estrogen receptor-positive breast cancer has been established using results from NSABP B-14 and NSABP B-20.43
The utility of the Oncotype DX risk estimates in clinically intermediate risk hormone receptor-positive, HER2-normal, grade II, lymph node-negative breast cancers has also be been established.44
Although our study was not designed to develop a therapeutic index or to evaluate the Oncotype DX assay, the findings are consistent with those reported in the Oncotype DX assay in that regulation and response to tamoxifen is under polygenic control. In fact, we indentified two genes ESR1 and ERBB2, which are also found in the Oncotype DX assay.
One of the major challenges in endocrine therapy is resistance to tamoxifen treatment, and many patients who respond tend to relapse. In this study, 71 patients relapsed out of the 298 treated with tamoxifen, while the rest exhibited significant variation in response to tamoxifen. The genetic mechanisms underlying resistance to tamoxifen treatment remain poorly understood. Our analysis revealed that genes containing SNPs are co-regulated with genes that are predictive of tamoxifen. This is a significant finding given the urgent need to identify predictive markers and potential targets for developing novel therapeutic strategies. The association of tamoxifen resistant and SNP-containing genes is consistent with recent studies.20,21
However, more research is needed to ascertain the role of SNPs in tamoxifen treatment. Although such a study would provide more insights about the genetic mechanisms underlying variability and resistance to tamoxifen treatment, it is beyond the scope of this paper, but is the subject of our ongoing investigation and will be reported elsewhere.
The results reported in this study explain the broader context in which genes containing SNPs associated with risk for breast cancer operate in response to tamoxifen treatment. However, the limitations of the study must be acknowledged. First, our study relies on use of publicly available data, which could have some deficiencies, such as sampling errors, genetic and phenotypic heterogeneity, and environmental factors which were not corrected for. In addition, the data did not include other factors such as age, tumor grade, tumor size. Therefore, these results cannot be generalized and their interpretation should be conservative. Majority of the GWAS studies and gene expression data used in this study are based on Caucasian population. Given the emerging evidence that genetic susceptibility loci may confer population-specific risk,45
and the plausibility that response to tamoxifen could potentially differ between populations, these results cannot be generalized to all populations. Use of association results diagnostically will require explicit evaluation of how well they can be transferred across different population groups. Additionally, further research is needed to determine the effects of genetic variants on gene expression in different populations. One such approach would involve assessment of allelic variation in gene expression among breast cancer patients.31,39
Such analysis was beyond the scope of this report.
However, despite these limitations, the results from this study provide insights about the global biological context in which SNPs associated with risk for breast cancer operate in tamoxifen treated patients. This is a major step towards translating GWAS discoveries into clinical practice. The results of this study could guide future experimental designs in breast cancer to identify targets for the development of more effective therapeutic strategies.
In conclusion, our data shows that combining gene expression profiling with GWAS information provides a unified approach to identifying candidate genes and candidate pathways containing SNPs, which are dysregulated in response to tamoxifen treatment in ER+ breast cancer. Furthermore, our analysis demonstrates that genes containing SNPs act in concert with experimentally confirmed tamoxifen resistant genes to confer resistance to tamoxifen. Additional studies are needed to determine how individual or all polymorphisms collectively contribute to variability to endocrine therapy in different ethnic populations, and to determine whether these polymorphisms could serve as potential biomarkers for stratifying breast cancer patients to individualized therapies.