diaPASEF Proteomics and Feature Selection for the Description of Sputum Proteome Profiles in a Cohort of Different Subtypes of Lung Cancer Patients and Controls

Abstract

The high mortality, the presence of an initial asymptomatic stage and the fact that diagnosis in early stages reduces mortality justify the implementation of screening programs in the populations at risk of lung cancer. It is imperative to develop less aggressive methods that can complement existing diagnosis technologies. In this study, we aimed to identify lung cancer protein biomarkers and pathways affected in sputum samples, using the recently developed diaPASEF mass spectrometry (MS) acquisition mode. The sputum proteome of lung cancer cases and controls was analyzed through nano-HPLC-MS using the diaPASEF mode. For functional analysis, the results from differential expression analysis were further analyzed in the STRING platform, and feature selection was performed using sparse partial least squares discriminant analysis (sPLS-DA). Our results showed an activation of inflammation, with an alteration of pathways and processes related to acute-phase, complement, and immune responses. The resulting sPLS-DA model separated between case and control groups with high levels of sensitivity and specificity. In conclusion, we showed how new-generation proteomics can be used to detect potential biomarkers in sputum samples, and ultimately to discriminate patients from controls and even to help to differentiate between different cancer subtypes.

Keywords: adenocarcinoma; diaPASEF; lung cancer; proteomics; sputum.

Figure 1

Differential abundance of selected proteins in the sputum proteomes of cases vs. control groups. (a) Normalized abundance levels of IGHV3−49, SERPINA1, LSP1, PRKAR1A, CRP, and ENO1. (b) Heatmap of the five proteins showing differential abundance in sputum. IGHV3−49, immunoglobulin heavy variable 3−49; SERPINA1, serpin family A member 1; LSP1, lymphocyte specific protein 1; PRKAR1A, protein kinase cAMP-dependent type I regulatory subunit Alpha; CRP, C−reactive protein and protein kinase; ENO1, enolase 1.

Figure 2

Functional analysis of the quantitative results. (a) Functional analysis in STRING, showing enriched terms (FDR < 0.01) for different categories (GO Process, Go Component, STRING clusters, KEGG, WikiPathways and UniProt Keywords). (b) STRING interaction network for the mapped differentially abundance proteins. Nodes represent proteins, edges represent protein–protein interactions (line thickness indicates the strength of data support for each interaction). Highlighted nodes: complement and coagulation cascades (KEGG pathway hsa04610) in red; immune response (GO Biological Process GO:0006955) in green.

Figure 3

Effect of lung cancer alteration on the sputum proteome, on the complement and coagulation cascades pathway (KEGG:05171), highlighting protein perturbation according to our quantification results and showing coherent cascades.

Figure 4

Feature selection with sPLS−DA in mixOmics for (a–c) the lung cancer vs. control groups comparison and (d–f) adenocarcinoma, squamous and microcytic groups using one-vs-all comparisons. (a,d) Preliminary analysis with PCA; (b,e) sPLS−DA sample plot with (b) sample prediction area or (e) confidence ellipse plots; (c,f) ROC curves and AUC.

Figure 5

Proteomics workflow. (a) Sample preparation and LC-MS acquisition using diaPASEF, and (b) data analysis.