MSACL 2023 Abstract

Introduction
Based on our growing understanding of the molecular heterogeneities in the central nervous system cancers, the recent ‘World Health Organization Handbook for Diagnosis of Central Nervous System (CNS) Cancers’ proposes an ‘integrated’ molecular approach to supplement morphometric indicators traditionally used in rapid intraoperative CNS diagnosis methods. The bulk of the proposed molecular diagnosis methods, however, are based on genomic or immunohistochemistry assays and as such cannot be used on intrasurgical timescales to drive personalized cancer resections in the operating theatre. Therefore, there is a need for a unified methodological platform that can deliver rapid CNS cancer diagnosis based on both morphometric and molecular indices of interest on intrasurgical timescales. This is crucial for CNS cancers where preoperative biopsies are uncommon and there is a growing body of retrospective data suggesting benefit to maximal resection in patients with cancers of certain molecular types, currently only actionable in revision surgeries.
Capitalizing on the close coupling between lipid metabolism and cancer formation & progression, untargeted mass spectrometry profiling of the tissue lipidome using ambient ionization methods has constituted an attractive strategy for rapid determination of cancer types through comparing rapidly acquired MS1 profiles of a query specimen to a validated library of known cancer lipid fingerprints. One such method is Picosecond InfraRed Laser Mass Spectrometry (PIRL-MS) that has been successful in discriminating molecular subgroups of medulloblastoma cancers in 10 seconds with minimal tissue consumption (< 1mm3) and no thermal damage outside the sampling zone.

Methods
A local tissue bank containing over 3,000 frozen specimens of over 20 different classes of adult and pediatric brain cancers has been subjected to analysis by PIRL-MS towards building a molecular library of PIRL-MS signatures of morphometrically or molecularly distinct classes of CNS cancers. 10-second MS1 PIRL-MS spectra are collected on a Waters Xevo G2 XS quadrupole Time of Flight (qTOF) mass spectrometer. Multivariate modeling of the MS1 spectra using both supervised dimensionality reduction methods as well as unsupervised clustering are conducted to validate the molecular models using blind sample predictions for the determination of sensitivity and specificity of use, and to inform of potential lead mass-to-charge (m/z) values most important for cancer class differentiations. These lead values are subjected to rigorous determination of their molecular identities using targeted high-resolution tandem mass spectrometry after chromatography on a Synapt qTOF (Waters). The resultant list of the characterized tissue metabolites will be tested again with a ‘sparse’ multivariate clustering approach for the determination of the list’s success rate in 10-second classification of morphometric and molecular classes of common CNS cancers. The use of a well-characterized metabolite array for MS1 cancer type classification is expected to shield against noise that has been reported in many untargeted MS explorations of cancer molecular fingerprint and in related metabolomics endeavours as a confounding factor.

Results
We have investigated 150 pediatric brain cancers (3 morphometrically distinct types and 7 molecularly distinct types) and close to 500 adult brain cancers (14 morphometrically distinct classes and so far only 2 molecularly distinct type of one class). The sensitivity and specificity of the integrated morphometric and molecular diagnosis for pediatric brain cancers with 10-second PIRL-MS was > 94%. This classification used 18 tissue lipids whose identities were determined with targeted chromatography and tandem mass spectrometry. The classification of the adult brain cancers with PIRL-MS is currently ongoing but preliminary data suggests > 84% sensitivity and specificity across 14 morphometric classes and select molecular types of isocitrate dehydrogenase-1 or BRAF-V600E mutations with only 10 seconds of data analysis and multivariate model comparisons. Further refinement of the model and blind sample tests will take place in the near future. We will also run post PIRL MS sampling immunohistochemistry to confirm the tumour class. We suspect our numbers will improve after this is done. Some of the 16% that were not classified correctly may not be the correct tumour in the frozen sample.

Conclusions
The integrated diagnosis workflow proposed by WHO heavily emphasizes the use of multiple instruments with varying turnaround times. The demonstrated capability of PIRL-MS in delivering 10-second diagnosis of both morphometrically distinct as well as molecularly different (sub)types of CNS cancers contributes to its potential to form a unified platform for integrated morphometric and molecular diagnoses of CNS cancers on the same instrumental platform. This integration of diagnostic workflow that now can use a single instrument for diagnosis based on molecular and morphometric indicators (from their correlated downstream molecular metabolites) is expected to save costs and improve the current standard of care, offering benefits to both patients and hospitals that may adopt it upon authorization for future clinical use.