Bone Metastasis in Advanced Breast Cancer: Analysis of Gene Expression Microarray

In this study we analyzed gene expression profiles in advanced breast cancer patients to elucidate genes that can be used to predict bone metastasis. A combination of the identified 3 genes in bone metastasis and other organs and 8 genes in only bone metastasis showed better prediction, and can be used as a training set. Background: Approximately 30% to 40% of breast cancer recurrences involve bone metastasis (BM). Certain genes have been linked to BM; however, none have been able to predict bone involvement. In this study, we analyzed gene expression profiles in advanced breast cancer patients to elucidate genes that can be used to predict BM. Patients and Methods: A total of 92 advanced breast cancer patients, including 46 patients with BM and 46 patients without BM, were identified for this study. Immunohistochemistry and gene expression analysis was performed on 81 formalinfixed paraffin-embedded samples. Data were collected through medical records, and gene expression of 200 selected genes compiled from 6 previous studies was performed using NanoString nCounter. Results: Genetic expression profiles showed that 22 genes were significantly differentially expressed between breast cancer patients with metastasis in bone and other organs (BMthorn) and non-BM, whereas subjects with only BM showed 17 significantly differentially expressed genes. The following genes were associated with an increasing incidence of BM in the BMthorn group: estrogen receptor 1 (ESR1), GATA binding protein 3 (GATA3), and melanophilin with an area under the curve (AUC) of 0.804. In the BM group, the following genes were associated with an increasing incidence of BM: ESR1, progesterone receptor, B-cell lymphoma 2, Rab escort protein, N-acetyltransferase 1, GATA3, annexin A9, and chromosome 9 open reading frame 116. ESR1 and GATA3 showed an increased strength of association with an AUC of 0.928. Conclusion: A combination of the identified 3 genes in BMthorn and 8 genes in BM showed better prediction than did each individual gene, and this combination can be used as a training set.