Characterization of triple negative breast cancer based on metabolic profiles using HR MAS MRS
Abstract
Triple negative breast cancers (TNBC) lack estrogen receptors (ER), progesterone receptors (PgR) and human epidermal growth factor receptors 2 (HER-2). Approximately 15% of breast cancer patients are diagnosed with TNBC. It is the most aggressive subgroup of breast cancer with high rate of metastases and is associated with a poor prognosis. Little effect of cytotoxic chemotherapy and lack of targeted therapy have increased the focus on this particular subgroup of breast cancer. Based on histological, clinical, and molecular features; TNBC is likely to represent a heterogeneous mix of breast cancer subtypes. High resolution magic angle spinning magnetic resonance spectroscopy (HR MAS MRS) based metabolomics has been employed to study cancer for more than a decade. The obtained spectra from HR MAS MRS reflect tissue biochemistry and can identify diagnostic and prognostic markers in the cancer tissue. In this study first, the effective echo time for the spin echo sequence was optimized and the changes occurring within the HR MAS probe overtime was evaluated. Then, the metabolic patterns of triple negative tumors were explored and compared to that of triple positive tumors. The influence from individual receptors on metabolic profiles was also analyzed.
Human breast tissue is heterogeneous in nature and can comprise substantial amount of lipids. In MR spectra of breast tissue, signals from small metabolites can be lost due to overlapping signal. Spectral editing using spin echo pulse sequences can be applied to overcome this spectral complexity. However, as all signals are influenced by the spin echo sequence, the optimal effective echo time should be determined. A series of spin echo experiments were carried out, using six different echo times. This was applied for 13 different samples. The resulting spectra showed that spin echo time of 273.5 ms gave significant lipid suppression and proper resolution of the spectra from the breast tumour tissue.
Mapping changes in metabolic profiles due to various factors after tissue excision may reduce the number of unforeseen experimental errors. Therefore, the effect of time in the HR MAS probe on the tissue metabolites was monitored. Spectra of 14 tumor samples were acquired at six different time points of 1.5 hrs intervals. Principal component analysis (PCA) suggested that the changes in metabolites with time were small compared to that of the inter-patient variation in general. The observed metabolic changes with time were increased levels of choline and glycine and decreased levels of glycerophosphocholine (GPC). These findings suggest that the degradation of metabolites during the HR MAS analysis can be considered negligible, and thus have little influence in metabolic characterization of tumor samples. However, we should be aware that metabolic alterations could originate from tissue degradation if spectra are recorded at very different time-points after sample thawing.
In this study the overall aim was to characterize the metabolic profile of TNBC patients. Tumor samples obtained during surgery of breast cancer patients (n=66) were analyzed by HR MAS MRS. Breast tumor spectra of triple positive (n=10) and triple negative (n= 15) patients were analyzed by partial least squares discriminant analysis (PLSDA) in order to discover potential metabolic differences between the patient groups. The discrimination gave a classification accuracy of 71.3% with a sensitivity and specificity of 70.0% and 73.3% respectively, and permutation testing showed that the metabolic differences were significant (p= 0.026). The analysis showed that the triple negative samples appeared to have lower levels of phosphocholine (PCho), creatine, and glycine compared to triple positive samples. Previous studies have shown significant differences in the metabolite profiles of ER+ and ER- breast cancer patients, thus there was a possibility that ER status alone could be responsible for the discrimination of triple negative and triple positive patients. In order to further explore the influence of the individual receptors on the metabolic difference between triple negative and triple positive patients, PLSDA of four additional sample groups were performed: ER status, PgR status, HER-2 status for ER+ samples, and HER-2 status for ER- samples. When looking at the receptors individually, the ER+ samples (n=34) could be discriminated from ER- samples (n=28) by higher levels of taurine, GPC, PCho, and creatine, and lower levels of choline, glycine and lactate (p=0.001). However, for the PgR status more than 50% of samples were classified incorrectly, thus no changes in metabolite levels due to PgR status were detected (p=0.504). There was significantly different metabolic profiles among HER+\-status for ER+ samples (p=0.004), but not among ER- samples (p=0.103). Higher levels of PCho, creatine and glycine appeared among HER2+ breast cancer samples. Thus, this is in accordance with the finding of higher levels of PCho, creatine and glycine among triple positive sample.
In conclusion, the combination of MR spectroscopy and multivariate analysis enabled discrimination of triple negative and triple positive breast cancer samples. TNBCs were characterized by lower levels of PCho, creatine, and glycine compared to triple positive samples. Based on variable importance in projection (VIP) loading, PCho could be regarded as predictive marker for TNBCs. These metabolic differences can partly be explained by metabolic patterns related to ER and HER-2 growth factor receptor status.