Abstract 1943: Elucidating early tumor microenvironmental changes due to immunotherapy with label-free Raman spectroscopy and machine learning

Abstract

Abstract Clinical benefit is observed in only a limited fraction of patients that undergo immunotherapy for a variety of cancer types. Therefore, the determination of robust biomarkers for accurate prediction of response to immunotherapeutic agents remains one of the grand challenges of 21st century. Our primary goal was to investigate if phenotypic differences due to treatment with ICI can be recognized with minimally perturbative molecular tools and guide personalization of immunotherapy. In this work, we provide first-in-class evidence that Raman spectroscopy (an optical method) and machine learning allows sensitive detection of early changes in biomolecular composition of tumors in response to immunotherapy with immune checkpoint inhibitors (ICI). By studying the widely investigated CT26 murine model of colorectal cancer treated with anti-CTLA-4 (n = 8) and anti-PD-L1 (n = 7) ICIs and controls (n = 10), we revealed new biological insights into the nature and degree of microenvironmental modifications induced by exposure to clinically relevant doses of ICI. Multivariate curve resolution-alternating least squares (MCR-ALS) decomposition of Raman spectral datasets revealed early changes in lipid, nucleic acid, and collagen content of the tumors following therapy. We trained supervised classification models - support vector machines and random forests on the spectral datasets and showed that the models provided excellent prediction accuracies for response to both ICIs and delineated spectral markers specific to each therapy, consistent with their differential mechanisms of action. On the basis of these findings, we sought to determine if the tumor microenvironment changes delineated by Raman spectroscopy correspond to proteomic alterations via quantitative mass spectrometry. Of the more than 6600 proteins identified, about 136 proteins were found to be significantly different in treated tumors relative to controls (p < 0.05 and log2 fold change of > 2). A subset of these differentially expressed proteins is known to regulate lipid metabolism and extracellular matrix composition, while others are known to control transcriptome dynamics or are associated with response to ICI therapy, thereby corroborating our Raman spectroscopic measurements. Our observation of biomolecular changes in the TME should catalyze detailed investigations for translating such markers and label-free Raman spectroscopy for clinical monitoring of immunotherapy response in cancer patients. Citation Format: Santosh K. Paidi, Joel R. Troncoso, Narasimhan Rajaram, Ishan Barman. Elucidating early tumor microenvironmental changes due to immunotherapy with label-free Raman spectroscopy and machine learning [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1943.