Background: Rheumatic diseases, such as rheumatoid arthritis (RA) and ankylosing spondylitis (AS), are marked by disrupted bone metabolism and chronic inflammation. Programmed cell death (PCD), such as ferroptosis, Cuproptosis and pyroptosis, has been reported to participate in various rheumatic diseases[1]. Pyroptosis, a crucial biological event, has been associated with bone metabolism in rheumatic diseases[2], yet the predictive value of pyroptosis in these conditions remains unclear.

Objectives: To investigate the correlation between pyroptotic genes and bone metabolism in rheumatic immune diseases like RA and AS. Utilizing public datasets, RNA-seq, and ELISA assays, this study aims to assess the potential of pyroptotic genes as early indicators of bone damage, incorporating a novel machine learning approach. The findings are intended to support the advancement of precision medicine.

Methods: Prior to the substantive study, we conducted an analysis of different PCD gene sets and bone mineralization/resorption gene sets across 3 public datasets (GSE15258, GSE25101, and GSE73754). In addition, we also used CIBERSORT and single-cell transcriptome sequencing to confirm the cell subpopulation distribution of pyroptotic gene expression. We then utilized RNA-seq to profile the whole-blood transcriptomes of RA and AS patients, as well as healthy volunteers to validate the inference from public data. Concomitantly, ELISA assays were used to evaluate pertinent bone metabolism markers. We further incorporated public datasets and made use of blending machine learning methods to investigate the correlation between pyroptotic gene expression and bone metabolism. The first layer of this blending machine learning model consists of XGBoost, Logistic, and LightGBM, and the second layer is a random forest model. The training set accounted for 85%, the verification set accounted for 15%, and the second layer model selected all samples as the test set.

Results: Our investigation identifies a substantial correlation between pyroptotic gene expression (compared with Apoptosis, Ferroptosis, Autophagy, Necroptosis, Cuproptosis and Parthanatos) and bone metabolism in rheumatic diseases (Figure 1). The distribution of pyroptotic genes was mainly concentrated in macrophage subpopulation. Notably, the pyroptotic genes - TNF, IRF2, CASP8, PYCARD, and NLRC4 successfully predicted the bone resorption score in AS patients (Test set AUC: 0.871, Accuracy: 0.833, Figure 2B), however, fell short in predicting bone mineralization scores (Test set AUC: 0.586, Accuracy: 0.583, Figure 2C). For RA patients, however, these genes were good predictors of both bone resorption (Test set AUC: 0.908, Figure 2D) and bone mineralization scores (Test set AUC: 0.859, Figure 2E). Furthermore, we confirmed a certain correlation between pyroptotic-related gene expression and the ELISA test indicators(including calcium, phosphorus, OPG, RANKL and CTX-I) in our samples.

Conclusion: By integrating RNA-seq profiling, ELISA assays, and blending machine learning analysis, our study emphasizes the complexity of the interplay between pyroptosis and bone metabolism in rheumatic diseases. The pyroptosis-related indicators we discovered allow for early prediction of bone metabolism in rheumatic diseases.

REFERENCES:

[1] Zhao J, Jiang P, Guo S, Schrodi SJ, He D. Apoptosis, Autophagy, NETosis, Necroptosis, and Pyroptosis Mediated Programmed Cell Death as Targets for Innovative Therapy in Rheumatoid Arthritis. FRONT IMMUNOL. 2021 2021/1/20;12:809806.

[2] Zhuang L, Luo X, Wu S, Lin Z, Zhang Y, Zhai Z, et al. Disulfiram alleviates pristane-induced lupus via inhibiting GSDMD-mediated pyroptosis. CELL DEATH DISCOV. 2022 2022/9/3;8(1):379.

Figure 1.

Pyroptosis gene expression demonstrates stronger correlation with bone mineralization and bone resorption indicators when compared to other forms of PCD.

Figure 2.

Blending Machine Learning Algorithm Unveils the Predictive Value of Pyroptotic Genes

Acknowledgements: NIL.

Disclosure of Interests: None declared.