Deciphering Multi-Dimensional Proteomics From Resected Lung Adenocarcinomas

Abstract

Lung adenocarcinomas often display a high degree of intratumoral heterogeneity as manifested both by histology and by dynamic changes in protein expression. Clinically, this underlying heterogeneity can drive tumor progression towards a more aggressive neoplastic state and is associated with a worse prognosis. Furthermore, the presence of individual tumor sub-populations may differ in their sensitivity to anti-neoplastic agents, limiting the effectiveness of treatment. Therefore, understanding the diverse composition of a tumor both for risk stratification and for designing novel therapies is of critical importance. To explore whether intratumoral heterogeneity could be used to construct a differentiation-based intracellular hierarchy in a novel manner, we conducted a low-cost quantitative proteomics analysis using MALDI-TOF mass spectrometry (MS) on different histological regions from 35 lung adenocarcinomas, and from 3 mesenchymal stem cell (MSC) samples and 3 basal cell samples. After demonstrating feasibility of this analysis, we collected similar proteomics data on histological regions from 59 additional specimens. In both cohorts, the histologies identified included areas of the neoplasm (acinar, lepidic, complex gland, micropapillary, papillary, solid subtypes) and normal tissue (bronchial epithelium and normal alveoli). Patient-specific information including survival, sex, age, smoking status, glucose uptake, and tumor size were obtained. We compared the proteomes derived from each histology and compared these to the proteomes derived from the MSC and normal lung samples. Using bioinformatics techniques, we attempted to organize the subtypes into a phylogenetic tree. We hypothesized that by mapping tumor progression based on the distance from MSCs, we could identify tumor sub-populations associated with clinical variables. Using co-expression network analysis, we also identified significantly dysregulated co-regulatory protein networks within each histology. We further incorporated the spatial information contained within our mass spectrometry analysis into a novel algorithm to combine proteomics and spatial information to significantly predict prognosis; the hypothesis was that tumors with a high concordance in intra-tumoral molecular and spatial distances would be less aggressive as these tumors were more likely to grow in a highly regulated fashion. Ultimately, we found that low-cost proteomic profiling in solid tumors can be used to better characterize and understand intratumoral heterogeneity and allow for identification of peptides that are crucial to cancer progression. Our results may help inform the field of targeted broad-scale proteomics profiling for therapeutic use.