Metabolic intratumor heterogeneity of breast cancer.

Abstract

e23095 Background: Breast tumors are highly heterogeneous due to subpopulations of cancer cells that differ in genetic and phenotypic characteristics. Tumor heterogeneity has been associated with treatment resistance and relapse. Therefore, it can be questioned how representative one biopsy is for the whole tumor. Tumor phenotypic heterogeneity cannot be solely attributed to genetic differences, as epigenetics and interaction with the tumor microenvironment also contribute. In this study we have examined intra-tumor heterogeneity by measuring metabolite expression in breast cancer tissue compared with fibroadenomas. Methods: Fresh frozen tissue slices from surgically removed breast tumors were used. Five cores from different areas of the slices were drilled out from 10 tumors; 6 invasive ductal carcinomas grade 2-3, and 4 fibroadenomas. Histological examination of HES-stained sections from each core was done, and metabolic profiling was performed by magnetic resonance spectroscopy (MRS). The relative concentrations of 23 metabolites were quantified. Metabolic heterogeneity was assessed by coefficient of variation (CoV) and PLSDA classification was used for prediction of tumor origin. Results: Cancer tissue showed significantly higher heterogeneity than fibroadenomas for 16/23 metabolites (mean CoV range: 0.15-0.94 for cancer samples, 0.09-0.37 for fibroadenomas, p < 0.05). However, 23/50 samples did not contain tumor tissue on histological examination. After exclusion of tumor-free samples, the heterogeneity of 3 metabolites (glycine, glycerophosphocholine (GPC) and phosphocholine (PCho) remained significantly different between cancer and fibroadenomas (mean CoV range: 0.12-0.65 for cancer, 0.07-0.42 for fibroadenomas, p < 0.05). GPC and PCho are involved in building of cell membranes and may reflect cell-turnover. Multivariate classification could correctly predict which patient a sample belonged to with 78% accuracy. Conclusions: Metabolic heterogeneity could partly be explained by differences in tumor cell and stromal content, and the origin of an unknown sample could be successfully predicted, showing that metabolic intratumor heterogeneity is smaller than the heterogeneity between patients.