Detecting breast tumor markers with a fast turnaround time from frozen sections should foster intraoperative histopathology in breast-conserving surgery, reducing the need for a second operation. Hence, rapid label-free discrimination of the spatially resolved molecular makeup between cancer and adjacent normal breast tissue is of growing importance. We performed desorption electrospray ionization mass spectrometry imaging (DESI-MSI) of fresh-frozen excision specimens, including cancer and paired adjacent normal sections, obtained from the lumpectomy of 73 breast cancer patients. The results demonstrate that breast cancer tissue posits sharp metabolic upregulation of diacylglycerol, a lipid second messenger that activates protein kinase C for promoting tumor growth. We identified four specific sn-1,2-diacylglycerols that outperformed all other lipids simultaneously mapped by the positive ion mode DESI-MSI for distinguishing cancers from adjacent normal specimens. This result contrasts with several previous DESI-MSI studies that probed metabolic dysregulation of glycerophospholipids, sphingolipids, and free fatty acids for cancer diagnoses. A random forest-based supervised machine learning considering all detected ion signals also deciphered the highest diagnostic potential of these four diacylglycerols with the top four importance scores. This led us to construct a classifier with 100% overall prediction accuracy of breast cancer by using the parsimonious set of four diacylglycerol biomarkers only. The metabolic pathway analysis suggested that increased catabolism of phosphatidylcholine in breast cancer contributes to diacylglycerol overexpression. These results open up opportunities for mapping diacylglycerol signaling in breast cancer in the context of novel therapeutic and diagnostic developments, including the intraoperative assessment of breast cancer margin status.
Read full abstract