Rapid Evaporative Ionisation Mass Spectrometry (REIMS) is an ambient ionisation method developed for the analysis of aerosols formed on cold or thermal ablation of biological tissues. REIMS data features primarily lipid species from fatty acids through phospholipids to triglycerides among others. While REIMS has been proven to be an excellent tool for the analysis of surgical aerosols and developing mass spectrometry-guided surgical approaches, these tools were proven to be impractical for laboratory testing of cells or tissues. More recently we have developed picosecond mid-infrared resonant laser systems using four wave mixing (FWM) and continuous wave-seeded optical parametric amplifier technologies to obtain lasers operating close to 3 µm wavelength in the picosecond pulse width range with beam profiles allowing to focus them below spot sizes comparable to the wavelength. These lasers were shown to be able to produce meaningful mass spectra by evaporating 30-100 pg of biological tissue corresponding to subcellular amounts (for comparison a single HeLa cell is estimated to weigh 500 pg).
We demonstrated the applicability of LA-REIMS integrated in an automated platform for the high-throughput lipidomic and metabolomic profiling of living and frozen cells, with minimal or no pre-treatment. We performed validation experiments with 5 breast and 5 colorectal cancer cell lines, to establish the classification performance and the molecular coverage of the method. Classification accuracy using an independent dataset was found to be 99.5%, while the method was able to detect more than 400 lipid species, including fatty acids, glycerophospholipids, diacyl and triacylglycerides, and cardiolipins. Additionally, the method was tested against an already proven biological hypothesis, shown enhanced cPLA2 activity and arachidonic metabolism in PIK3CA mutant breast cancer cell lines. In a separate experiment we have tested 70 cancer cell lines and 8 matched patient-derived organoids. The mass spectrometric data revealed 0.97 area under curve results in ROC analysis for the detection of PIK3CA mutations in this case, where the organoids were all correctly classified using a support vector machine model generated from the cell line data. The therapy resistance to various cytostatic agents was predicted with >75% accuracy, with very long chain fatty acids and lysophospholipids driving the classification.