Proteome analysis of non-small cell lung cancer cell line secretomes and patient sputum reveals biofluid biomarker candidates for cisplatin response prediction

Molecular markers are urgently needed to select non-small cell lung cancer (NSCLC) patients most likely to benefit from platinum-based chemotherapies. Of particular interest are proteins that can be found in biofluids like sputum for non-invasive detection. Therefore, we profiled the secretomes of 6 NSCLC cell lines with varying IC50-values for cisplatin, using label-free GeLC-MS/MS-based proteomics. Out of a total dataset of 2610 proteins, 304 proteins showed significant differences in expression levels between cisplatin sensitive and insensitive cell lines. Functional data mining revealed that the secretion of typically extracellular factors was associated with a higher sensitivity towards cisplatin, while cisplatin insensitivity correlated with increased secretion of theoretically intra-cellular proteins. Stringent statistical analysis and quantitative filtering yielded 58 biomarker candidates, 34 of which could be detected in clinical biofluids of lung cancer patients such as sputum using label-free LC-MS/MS-based proteomics. To assess performance of these biofluid biomarker candidates, we correlated protein expression with patient survival using a publically available clinical gene expression data set (GSE14814). We thus identified 3 top candidates with potential predictive value in determining cisplatin response (UGGT1, COL6A1 and MAP4) for future development as non-invasive biomarkers to guide treatment decisions. Significance: Platinum-based chemotherapies are still the standard of care for NSCLC and other lung cancer types in the clinic today. However, due to chemoresistance, many patients suffer from the toxic side effects of these treatments without gaining any benefit in terms of survival. To date, no molecular biomarkers are available to predict clinical outcome of platinum-based chemotherapy. Because proteins present the functional read-out of genetic, epigenetic and translational events in the cell, a protein test is likely to be particularly suitable for response prediction. Of high relevance are proteins that are shed or secreted from cells, for example at primary tumor sites, and can be found in easily accessible biofluids like sputum for non-invasive detection. Here, we report the proteome profiling of the conditioned media (secretomes) of a panel of NSCLC cell lines in relation to cisplatin IC50 values, as a pre-clinical model, and of patient sputum as a clinical, lung cancer relevant biofluid. Using this approach in conjunction with exploration of the predictive potential in a transcriptome lung cancer patient dataset, we reveal biofluid biomarker candidates that, with further validation, may be used for non-invasive cisplatin response prediction in the future