Imaging Mass Spectrometry (IMS) to Discriminate Neoplastic Diseases

Introduction Cancer therapy is primarily directed by tumor origin, making correct pathologic diagnosis imperative for proper patient management. Molecular methods for histopathology are today increasingly considered to support diagnosis together with classical methods. MALDI imaging mass spectrometry (IMS) is one of the most striking technique providing an overall insight of the molecular context of a given physiological or pathological status that cannot not be obtained with classical histopathological examinations. Objective In this study, we introduce MALDI IMS as a new and valuable approach to classify tumor types in routine histopathological formalin-fixed paraffin-embedded (FFPE) tissue specimens, towards improved cancer diagnosis and management. Materials & methods We investigated FFPE tissue microarrays (TMAs) consisting of needle core tumor biopsies from breast, pancreas, lung, and colon. Each TMA section was subjected to deparaffination, heat-induced epitope retrieval, in-situ trypsin digestion and deposition of a matrix solution for MALDI IMS analysis. MALDI MS data were acquired at a spatial resolution of 100 µm using a Bruker Autoflex Speed TOF/TOF system, and analyzed and imaged with FlexImaging (Bruker Daltonik) and SCiLS Lab (SCiLS GmbH) software. TMA spectra datasets each containing a specific tumor type were used to train linear discriminant analysis (LDA) model for classification analysis. Another independent subset of spectra was used to validate the LDA algorithm. Immunohistochemistry was performed for validation of results. Results Class prediction models were developed that could discriminate all different tumor types presented in this study with a cross validation accuracy of Patients with lung, pancreas, colon, and breast tumor were correctly classified in 92, 75, 97, and 91% of cases respectively. In addition, the LDA model was successfully applied to distinguish the primary tumors from distant metastases. Class images successfully resembled the histopathological diagnosis made by pathologists thus providing good correlation with tumor/organ specific profile. Conclusion Our study represents an initial but important attempt to describe an unbiased molecular diagnostic tool capable of tumor classification in addition to other molecular techniques