RAPID EVAPORATIVE IONIZATION MASS SPECTROMETRY-BASED LIPIDOMIC FOR IDENTIFICATION OF CANINE MAMMARY PATHOLOGY

Rapid Evaporative Ionization Mass Spectrometry (REIMS) is an emerging ambient ionization technique which allows the identification of tissues in real-time through the analysis of the informative aerosols generated during their intraoperative electrosurgical dissection. The coupling of the REIMS method with a classic surgical electrocautery is known as the intelligent knife (iKnife) [1, 2]. The MS spectra obtained are used to build a predictive statistical model (PCA/LDA, Principal Components Analysis/Linear Discriminant Analysis), which allows the immediate identification and differentiation of tissues based on their differences in the lipidomic profiles [1, 2]. Currently, REIMS technology has been validated for metabolic phenotyping tumors intraoperatively in human surgical oncology, exhibiting high diagnostic accuracy, improving margin assessment and surgical outcomes [3, 4]. The ability to rapidly identify tissues based on their metabolic phenotypes using a REIMS approach may represent an important advantage over routine methods in monitoring tissue resection in the field of veterinary medicine. Therefore, this study aimed to investigate REIMS-TOF-MS technology in identifying pathological canine mammary tissues, differentiating them from normal mammary gland based on different metabolic phenotypes, using histologically validated ex-vivo samples. Thirty-nine mammary gland samples were specularly sectioned and routinely processed for histology (HE) and stored at -80°C for REIMS-TOF-MS analysis. Histological examination classified our samples in 4 classes: normal (n=12), hyperplastic (n=6), inflammatory (n=7), and neoplastic (n=14). The neoplastic samples included 6 benign and 8 malignant canine mammary tumors of different grade of malignancy [5]. Healthy mammary tissues were obtained from necropsied dogs. Mass spectrometric data generated from histologically validated samples (n=20) representative for each histological type were used to build the multivariate statistical model, further validated for robustness and predictive capability by performing two different in silico validation tests (1-file out and 5-fold validation) using LiveID software version 1.2 (Waters Corporation, UK). The confusion matrix of both approaches revealed a very low rate of failures and outliers, with classification accuracy of 90% and 95%, respectively. The validated model showed a distinct clusterization of the samples based on histological classification, in particular between healthy and pathological/neoplastic tissues along the main linear discriminations. A second MS analysis of mammary samples, with known histological diagnosis and not previously included in the statistical model, was performed to test the recognition capability in real-time. All samples were correctly identified with very high correctness score (98-100%). The REIMS method has proved to be a valid, fast and accurate technology in the differentiation of pathological canine mammary tissues based on differences in lipid metabolism. Real-time intraoperative tissue identification would offer a significant advantage especially in the complete surgical excision of tumors. Further validation studies on a larger number of samples and different histotypes will be required before applying REIMS technology to intraoperative veterinary diagnostics.