Pleural fluid proteomics from patients with pleural infection shows signatures of diverse neutrophilic responses: The Oxford Pleural Infection Endotyping Study (TORPIDS-2)

Abstract

Background: Pleural infection is a complex disease with poor clinical outcomes and increasing incidence worldwide, yet its biological endotypes remain unknown.

Methods: We analysed 80 pleural fluid samples from the PILOT study, a prospective study on pleural infection, using unlabelled mass spectrometry. A total of 449 proteins were retained after filtering. Unsupervised hierarchical clustering and Uniform Manifold Approximation and Projection analyses were used to cluster samples and pathway analysis was performed to identify the biological processes. Protein signatures as identified by the pathway analysis were compared to microbiology as defined by 16S rRNA next-generation sequencing. Spearman and exact Fischer's methods were used for correlation assessment.

Results: Higher neutrophil degranulation was correlated with increased glycolysis (odds ratio (OR) 281, p<2.2×10^-16) and pentose phosphate activation (OR 371.45, p<2.2×10^-16). Samples dominated by Streptococcus pneumoniae exhibited higher neutrophil degranulation (OR 12.08, p=0.005), glycolysis (OR 11.4, p=0.006) and pentose phosphate activity (OR 12.82, p=0.004). Samples dominated by anaerobes and Gram-negative bacteria exhibited lower neutrophil degranulation (OR 0.15, p=0.01), glycolysis (OR 0.14, p=0.01) and pentose phosphate activity (OR 0.07, p=0.001). Increased activity of the liver and retinoid X receptors pathway was associated with lower risk of 1-year mortality (OR 0.24, p=0.04).

Conclusions: These findings suggest that pleural infection patients exhibit diverse responses of neutrophil-mediated immunity, glycolysis and pentose phosphate activation, which are associated with microbiology. Therapeutic targeting of the liver and retinoid X receptors pathway with agonists is a possible treatment approach.

Overview of the study. LXR-RXR: liver and retinoid X receptor.

FIGURE 1

Pleural fluid samples derived from pleural infection patients showed diverse proteomic expression. a) Uniform Manifold Approximation and Projection (UMAP) plot for the whole proteome after protein filtration and normalisation. Each dot represents a sample. The samples were clustered into two separate and independent cohorts (Group 1 and Group 2). b) Differential protein expression analysis. Volcano plot presenting the differential protein expression analysis for Group 1 compared with Group 2 samples as separated in the UMAP above. Green dots represent proteins significantly decreased in Group 1 compared to Group 2; red dots represent proteins significantly increased in Group 1 compared to Group 2; black dots represent proteins that are below the fold-change (FC) threshold and nonsignificant. Statistical significance was determined using the Mann–Whitney U test. c) Significant pathways as identified by the pathway analysis of the significant proteins as identified in b. The colour represents the p-value and the count the number of identified proteins.

FIGURE 2

Pleural infection patients exhibited different levels of neutrophil degranulation activity. a) Heatmap of unsupervised hierarchical clustering for the identified proteins of the neutrophil degranulation pathway. Each row is a protein of the pathway and each column a sample. The colours of the cells represent the relative protein abundance and the colours at the top represent the groups as defined in figure 1. b–e) Uniform Manifold Approximation and Projection plots for the proteins of the neutrophil degranulation pathway. Each dot is a sample, and the colour represents the groups as defined in figure 1 (b), dominance of Streptococcus pneumoniae (c), dominance of Streptococcus anginosus (d) and dominance of a mixture of strict anaerobes and Gram-negative bacteria (e).

FIGURE 2

Continued.

FIGURE 3

Pleural infection patients showed different levels of glycolytic metabolism. a) Heatmap of unsupervised hierarchical clustering for the identified proteins of the glycolysis pathway. Each row is a protein of the pathway and each column a sample. The colours of the cells represent the relative protein abundance and the colours at the top the groups as defined in figure 1. b–d) Uniform Manifold Approximation and Projection plot for the identified proteins of the glycolysis pathway. Each dot is a sample, and the colours represent the groups as defined in figure 1 (b), dominance of Streptococcus pneumoniae (c) and dominance of a mixture of strict anaerobes and Gram-negative bacteria (d).

FIGURE 4

Pleural infection patients displayed different levels of pentose phosphate metabolism. a) Heatmap of unsupervised hierarchical clustering for the identified proteins of the pentose phosphate pathway. Each row is a protein of the pathway and each column a sample. The colours of each cell represent the relative protein abundance and the colours at the top the groups as defined in figure 1. b–d) Uniform Manifold Approximation and Projection plot for the identified proteins of the pentose phosphate pathway. Each dot is a sample. The colours represent the groups as defined in figure 1 (b), dominance of Streptococcus pneumoniae (c) and dominance of a mixture of strict anaerobes and Gram-negative bacteria (d).

FIGURE 5

Pleural infection patients exhibited various levels of the liver and retinoid X receptor (LXR-RXR) activation pathway. a) Heatmap of unsupervised hierarchical clustering for the identified proteins of the LXR-RXR activation pathway. Each row is a protein of the pathway and each column a sample. The colours of the cells represent the relative abundance of the proteins and the colours at the top the groups as defined in figure 1. b–d) Uniform Manifold Approximation and Projection plot for the identified proteins of the LXR-RXR activation pathway. Each dot is a sample. The colours represent the groups as defined in figure 1 (b), 1-year survival (c) and dominance of a mixture of strict anaerobes and Gram-negative bacteria (d).