Integrative epigenomics, transcriptomics and proteomics of patient chondrocytes reveal genes and pathways involved in osteoarthritis

Abstract

Osteoarthritis (OA) is a common disease characterized by cartilage degeneration and joint remodeling. The underlying molecular changes underpinning disease progression are incompletely understood. We investigated genes and pathways that mark OA progression in isolated primary chondrocytes taken from paired intact versus degraded articular cartilage samples across 38 patients undergoing joint replacement surgery (discovery cohort: 12 knee OA, replication cohorts: 17 knee OA, 9 hip OA patients). We combined genome-wide DNA methylation, RNA sequencing, and quantitative proteomics data. We identified 49 genes differentially regulated between intact and degraded cartilage in at least two -omics levels, 16 of which have not previously been implicated in OA progression. Integrated pathway analysis implicated the involvement of extracellular matrix degradation, collagen catabolism and angiogenesis in disease progression. Using independent replication datasets, we showed that the direction of change is consistent for over 90% of differentially expressed genes and differentially methylated CpG probes. AQP1, COL1A1 and CLEC3B were significantly differentially regulated across all three -omics levels, confirming their differential expression in human disease. Through integration of genome-wide methylation, gene and protein expression data in human primary chondrocytes, we identified consistent molecular players in OA progression that replicated across independent datasets and that have translational potential.

Figure 1

Overview of the genes identified in each –omics experiment and their overlap. (a) Schematic view of the 3 functional genomics experiments identifying the number of genes shortlisted for each. (b) Venn diagram identifying the number of overlapping shortlisted genes from each individual experiment.

Figure 2

Comparison of changes identified in the –omics experiments. (a) Comparison of the log-fold-changes between all genes identified in both the proteomics and RNA-seq experiments. Each gene is represented as a single point, and the colour corresponds to whether the gene is identified as differentially expressed using edgeR in the RNA-seq or proteomics experiments, or both. The trend lines are derived from a linear regression in each subset. Positive fold changes indicate increased expression in degraded samples. (b) Comparison of RNA-seq log-fold-change and mean promoter region methylation change. The trend lines are derived from a linear regression in each subset. Genes are coloured according to the results of the RNA-seq and the promoter-region analyses analogously to Fig. 2a.

Figure 3

Replication of gene expression and methylation changes. (a,b) Replication of gene expression changes (a) and CpG methylation changes (b) in independent datasets of samples from patients with knee (left) and hip (right) OA. Differentially expressed genes (a, DEGs) and differentially methylated probes (b, DMPs) from the discovery data are marked in black; DEGs and DMPs that additionally show nominal significance in the replication data are marked in red. Inset: correlation between log-fold-changes in discovery and replication data. All: all 14,762 genes (a) or 416,437 probes (b) that passed QC in the knee discovery, knee replication, and hip replication data. DEGs: 332 of the 349 genes with FDR ≤ 5% in the knee discovery data that also pass QC in both replication datasets; DMPs: 9,723 of the 9,867 probes with FDR ≤ 5% in the knee discovery data that also pass QC in both replication datasets. All correlation values shown have p < 10⁻¹⁵. (c,d) Directional concordance of changes between discovery and replication for the differentially expressed genes (c, DEGs) and differentially methylated probes (d, DMPs) from the discovery data. Same dir: proportion of DEGs or DMPs with same direction of change in the replication data. Same dir and p ≤ 0.05: among DEGs or DMPs with same direction of change in the replication data, proportion of those that reach nominal significance in the replication dataset.

Figure 4

Significantly enriched gene sets in the integrative analysis. Only gene sets that contain at least 5 genes from 2 experiments are shown. Asterisks indicate significant (5% FDR) enrichment for each individual gene-set in an individual –omics experiment. (a) Fold enrichment for KEGG & Reactome pathways significant at 5% FDR in the integrative analysis. (b) Fold enrichment for GO terms significant at 5% FDR in the integrative analysis.