Osteoarthritis-like transcriptional changes induced by IL-17A point to its potential validity as a therapeutic target

For this brief article, I have decided to re-analyse the RNA-seq data from a recently published paper on the osteoarthritis-like (OA-like) transcription changes induced in response to Interleukin-17 (IL-17) (Mimpen et al, 2021) using our Omics Playground platform.  In the original manuscript, the authors exposed human chondrocytes (HC) and synovial fibroblasts (SF) from end-stage OA patients to three members of the IL-17 cytokine  family (namely IL-17A, IL-17AF and IL17-F) and observed the dysregulation of pathways associated with angiogenesis, immunity and the complement system. They also observed an association with arthritis and musculoskeletal disease gene sets and concluded that IL-17A played a role in OA pathophysiology. 

The biological material (HC and SF cells) for the study was obtained from OA patients. The samples were divided into a control group and three treatment groups (IL-17A, IL-17AF and IL-17F) of six samples each for each cell type and then sequenced, for a total of 24 HC and 24 SF RNA-seq samples.

  1. Clustering Analysis

Samples clustered by cell-line (using the tSNE approach), but beyond that it was impossible to distinguish any further sub-structure in the data according to treatment (Fig. 1).

Figure 1. tSNE plot showing the clustering of the samples according to cell type and treatment. Cell type: green, HC; blue, SF. Treatments: circles, control; squares, IL-17A; stars, IL-17AF; triangles, IL-17F.

When cell samples were separated by cell type and grouped by treatment, a marked distinction in gene expression between treated (particularly IL-17A treated) and control groups could be observed in clustered heatmaps (Fig. 2A and 3A). In both cases, gene cluster S3 showed the largest difference in gene expression profiles between IL-17A treated samples and the control group. Annotation of the S3 gene cluster in both cell types yielded similar results, displaying a strong correlation with inflammatory responses, angiogenesis, various immune responses and ,in the case of HC samples, complement activation (Fig. 2B and 3B). This was in line with the main observations of the published article. 

Figure 2. Clustered heatmap of the HC cell type samples grouped by treatment. Genes are divided into four main clusters (S1-4).
Figure 2. Clustered heatmap of the HC cell type samples grouped by treatment. Genes are divided into four main clusters (S1-4).

Figure 3. Clustered heatmap of the SF cell type samples grouped by treatment. Genes are divided into four main clusters (S1-4).
Figure 3. Clustered heatmap of the SF cell type samples grouped by treatment. Genes are divided into four main clusters (S1-4).

  1. Differential Expression Analysis

This analysis was performed using the intersection of  three different gene expression analysis methods (EdgeR (qlf), DESEQ2 (wald) and limma (trend)) with FDR<0.05 and logFC>|0.5|.

When comparing IL-17A treated vs control group by cell type, 306 genes (196 up- and 110 down-regulated) were differentially expressed in HC samples but only 38 (32 up- and 6 down-regulated) in SF samples (Fig. 4A and 4B). 29 genes were present in both cell type comparisons (Fig. 5). Among these, IL-6, four C-X-C motif chemokine ligands, NFKB inhibitor zeta (NFKBIZ), C-C motif chemokine ligand 2 (CCL2), and IRF1 were upregulated. All these genes are involved in immune responses. Another upregulated gene was MAP3K8, which is known to be overexpressed in rheumatoid arthritis (here).

Figure 4. Volcano plots for IL-17A treated vs control pairwise comparisons by cell type. A: HC cell type; B: SF cell type.
Figure 5. Venn diagram of the intersection of the dysregulated genes in the IL-17A vs control comparisons by cell type. Genes are separated into up- (top numbers) and down-regulated genes (bottom numbers). A: SF cell type; B: HC cell type.
Figure 5. Venn diagram of the intersection of the dysregulated genes in the IL-17A vs control comparisons by cell type. Genes are separated into up- (top numbers) and down-regulated genes (bottom numbers). A: SF cell type; B: HC cell type.

  1. Gene Set Enrichment Analysis

The expression profiles of the IL-17A vs control groups for the HC and SF cell types were compared against various public datasets using the fgsea algorithm with an FDR<0.05 and logFC>0.2. When tested against the “DISEASE” database in the platform, both comparisons yielded a positive correlation against osteoarthritis, arthritis, rheumatoid arthritis and inflammatory and autoimmune diseases such as scleroderma, Crohn’s disease and discoid lupus erythematosus. Results are shown for the HC cell type profile only (Fig. 6).

Figure 6. Top 10 most correlated disease gene expression profiles with IL-17A treatment in HC cells.

Performing a KEGG pathway analysis revealed the significant (q<0.01) upregulation of the NOD-like receptor signalling pathway, involved in innate immunity and inflammation,  in IL-17A treated HC samples (Fig. 7). A GO enrichment analysis showed the significant upregulation (q<0.01) of various gene sets associated with immune responses and inflammation in both HC and SF IL-17A treated samples and to a lesser extent in IL-17AF treated HC samples (Fig. 8). 

Figure 7. Graphical representation of the KEGG NOD-like receptor signaling pathway. Genes coloured in red are upregulated, in blue are downregulated in the IL-17A vs control comparison in HC.
Figure 7. Graphical representation of the KEGG NOD-like receptor signaling pathway. Genes coloured in red are upregulated, in blue are downregulated in the IL-17A vs control comparison in HC.
Figure 8. Activation matrix showing the most significant up- (red) and down-regulated (blue) GO terms across all pairwise comparisons.
Figure 8. Activation matrix showing the most significant up- (red) and down-regulated (blue) GO terms across all pairwise comparisons. 

Next, I used the drug connectivity panel in the platform to compare the pairwise comparisons against the L1000 drug activity database. I focused on drugs generating significantly (q<0.05) inversely correlated gene expression profiles that could act as potential inhibitors of the IL-17 treatment. The most prominent mode of action identified among potential inhibitors was the inhibition of p38 MAPK (p38 mitogen-activated protein kinases), followed by CDK (cyclin dependent kinases) and MEK (mitogen-activated protein kinase kinases)  inhibition (Fig. 9A). MEKs play a central role in mediating OA pathology and have long been proposed as a therapeutic target (here). The p38 MAPK pathway has been strongly implicated in the pathology of rheumatoid arthritis (here) and its inhibition suppressed the expression of proinflammatory cytokines in cartilage tissues from OA patients (here). Finally, CDK9 inhibition has been shown to have beneficial effects both in vitro and in mouse models of OA (here and here). 

When looking at individual drugs, I observed a significant negative correlation of all IL-17 treatments with several compounds, of which five stood out (Fig. 9B). The first is fostamatinib, an SYK inhibitor which was trialed on OA patients (here) but appears to have limited efficacy ex vivo (here). Compound AZ-628, a receptor-interacting protein kinase-3 (RIP3) inhibitor, also showed an antagonist profile across the board and has been recently shown to abolish OA pathogenesis (here). Alvocidib was shown to inhibit the development of post-traumatic OA in mice (here), while triptolide has been extensively studied as a treatment for OA (here). Finally, PP-2, an Src kinase inhibitor, has been shown to regulate chondrocyte differentiation, indicating the inhibition of Src kinase as a potential therapeutic approach for skeletal diseases such as OA (here).

Figure 9. Drug connectivity analysis. A: Most significant drug modes of action (MOA) observed in the drug connectivity map analysis. Negative normalised enrichment scores (NES) indicate MOA that counteract the IL-17 induced profile and can thus have therapeutic applications. The most negatively correlated MOA are highlighted. B: Activation matrix of the most negatively correlated drug profiles across all pairwise comparisons. Three compounds tested in OA patients or OA models are highlighted.
Figure 9. Drug connectivity analysis. A: Most significant drug modes of action (MOA) observed in the drug connectivity map analysis. Negative normalised enrichment scores (NES) indicate MOA that counteract the IL-17 induced profile and can thus have therapeutic applications. The most negatively correlated MOA are highlighted. B: Activation matrix of the most negatively correlated drug profiles across all pairwise comparisons. Three compounds tested in OA patients or OA models are highlighted. 

  1. Biomarkers Discovery

Using the “Find Biomarkers” panel in tha platform, I produced a decision tree to help find genetic markers for different treatment groups (Fig. 10A). While this is only one of the possible decision trees, it indicates how the activation of components of the immune system is a hallmark of OA-like changes. Chemokine CXCL1 is able to distinguish treated and untreated samples and the difference in expression across all treatment groups compared to the control group is pronounced (Fig. 10B). The same is true for the NFKB Inhibitor Zeta gene (NFKBIZ) and, to a lesser extent, CXCL3 and IL-6. In order to distinguish between the different IL-17 treatments, expression levels of CXCL1, CXCL3, IL-6 and NFKBIZ appear to be indicative and point to the greater impact that IL-17A has on gene expression changes than the alternative isoforms. 

Figure 10. Biomarkers discovery. A: One of the possible decision trees for identifying biomarkers distinguishing the various sample groups in the dataset. B: Expression boxplots of the potential biomarker genes identified by the platform.
Figure 10. Biomarkers discovery. A: One of the possible decision trees for identifying biomarkers distinguishing the various sample groups in the dataset. B: Expression boxplots of the potential biomarker genes identified by the platform.

  1. Conclusion

Consistent with the main finding from the original article, the re-analysis of the dataset by Mimpen and colleagues identified OA-like signatures in the transcriptomes of chondrocytes exposed to IL-17A. The signatures were characterised by the upregulation of genes associated with immune responses and inflammation and matched the public expression profiles of various arthritic and autoimmune conditions. The upregulation of MAP3K8 was also observed. This was correlated to the results from the drug connectivity analysis, which suggested MEK inhibitors as potential therapies for the OA-like profiles. Indeed, MEK inhibition as well as suppression of p38 MAPK  have been tested as treatments for OA, which point to the reliability of IL-17A treatment in mimicking OA transcriptional changes. Finally, several chemokines and immune-related genes were highlighted as potential biomarkers of IL-17 exposure and also confirmed the greater transcriptional responses induced following IL-17A treatment compared to the IL-17AF and IL-17F isoforms.

Share this content
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •