Phosphoproteomic profiling of early rheumatoid arthritis synovium reveals active signalling pathways and differentiates inflammatory pathotypes

Abstract

Background: Kinases are intracellular signalling mediators and key to sustaining the inflammatory process in rheumatoid arthritis (RA). Oral inhibitors of Janus Kinase family (JAKs) are widely used in RA, while inhibitors of other kinase families e.g. phosphoinositide 3-kinase (PI3K) are under development. Most current biomarker platforms quantify mRNA/protein levels, but give no direct information on whether proteins are active/inactive. Phosphoproteome analysis has the potential to measure specific enzyme activation status at tissue level.

Methods: We validated the feasibility of phosphoproteome and total proteome analysis on 8 pre-treatment synovial biopsies from treatment-naive RA patients using label-free mass spectrometry, to identify active cell signalling pathways in synovial tissue which might explain failure to respond to RA therapeutics.

Results: Differential expression analysis and functional enrichment revealed clear separation of phosphoproteome and proteome profiles between lymphoid and myeloid RA pathotypes. Abundance of specific phosphosites was associated with the degree of inflammatory state. The lymphoid pathotype was enriched with lymphoproliferative signalling phosphosites, including Mammalian Target Of Rapamycin (MTOR) signalling, whereas the myeloid pathotype was associated with Mitogen-Activated Protein Kinase (MAPK) and CDK mediated signalling. This analysis also highlighted novel kinases not previously linked to RA, such as Protein Kinase, DNA-Activated, Catalytic Subunit (PRKDC) in the myeloid pathotype. Several phosphosites correlated with clinical features, such as Disease-Activity-Score (DAS)-28, suggesting that phosphosite analysis has potential for identifying novel biomarkers at tissue-level of disease severity and prognosis.

Conclusions: Specific phosphoproteome/proteome signatures delineate RA pathotypes and may have clinical utility for stratifying patients for personalised medicine in RA.

Fig. 1

Clustering of differentially expressed phosphosites and proteins between the lymphoid and myeloid pathotypes. A Immunohistochemistry of synovial biopsies for CD3, CD20, CD68L, CD68SL and CD138 cell surface markers to classify samples as lymphoid (B cell aggregates present), myeloid (sublining macrophage infiltration). B Hierarchical clustering using the most significant phosphosites that were differentially expressed between the pathotypes (q < 0.01) and (C) the most significant proteins that were differentially expressed between the pathotypes (q < 0.05). Phosphosites with the similar levels were grouped for the visualisation of the heatmap. D Volcano plots of differentially expressed substrates between the pathotypes using limma comparing. E Boxplots showing semiquantitative histological scores of CD3, CD20, CD68L, CD68SL and CD138 for lymphoid and myeloid pathotypes. For statistical analysis, two-sided Wilcoxon signed-rank test was used. ns p > 0.05, * p < 0.05, ** p < 0.01

Fig. 2

Receptor and ligand expression levels and predicted kinase activity between lymphoid and myeloid pathotypes. A Boxplots of the differentially expressed ligands myocilin; (MYOC), decorin (DCN) and Von Willebrand Factor (VWF) and the differentially expressed receptors; Platelet Endothelial Cell Adhesion Molecule (PECAM1), Glycoprotein Ib Platelet Subunit Beta (GP1BB) and Integrin Subunit Beta 3 (ITGB3). B Scatterplot illustrating the predicted differential kinase activities between lymphoid and myeloid pathotypes

Fig. 3

Pathway and transcription factor enrichment analysis of the differential phosphosite and protein levels between the lymphoid and myeloid pathotype. A Dotplot of the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis using the differential phosphosite levels between the two pathotypes and (B) the differentially expressed proteins between the two pathotypes. C Heatmap of the predicted transcription factor activity based of the differentially expressed proteins between the pathotypes and a barplot of the mean transcription factor expression for each pathotype

Fig. 4

Correlation between clinical endpoints and phosphosite levels. A Correlation-based similarities of clinical variables. B Heatmap of the Pearson correlations between phosphosite levels and clinical endpoints (ESR: erythrocyte sedimentation rate; CRP: C-reactive protein; Ultrasound ST/PD BJ: ultrasonographic scores (ST, synovial thickness; PD, power doppler) at the biopsy joint (BJ); Ultrasound ST/PD 12: ultrasonographic scores across 12 representative joints, disease activity scores: DAS28-CRP/ESR baseline and DAS28 (baseline, after 6 months and the delta between baseline and 6 months). Correlation pairs with p < 0.05 and an absolute correlation coefficient higher than 0.9 are shown. Clinical endpoints at the x-axis are grouped and ordered based on similarities among them. C Pairs of clinical endpoints and phosphosites with the highest positive and negative correlation coefficients. D Inverse correlation between DAB2-Ser723 with delta DAS28. Two missing delta DAS28 data points were not shown in this plot. E Correlation of FLNA-Ser1459 with ultrasound score synovial thickening and inflammatory score