Single Cell Transcriptome and Surface Epitope Analysis of Ankylosing Spondylitis Facilitates Disease Classification by Machine Learning

• Published in: Frontiers in immunology Vol. 13; p. 838636
• Main Authors: Alber, Samuel, Kumar, Sugandh, Liu, Jared, Huang, Zhi-Ming, Paez, Diana, Hong, Julie, Chang, Hsin-Wen, Bhutani, Tina, Gensler, Lianne S, Liao, Wilson
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Media S.A 13.05.2022
• Subjects: ankylosing spondylitis; CITE-seq; Epitopes; genomics; Humans; Immunology; Leukocytes, Mononuclear - pathology; Machine Learning; single cell sequencing; Spondylitis, Ankylosing - diagnosis; spondyloarthritis; Transcriptome; spondyloarthritis; genomics; single cell sequencing; machine learning; CITE-seq; ankylosing spondylitis
• ISSN: 1664-3224; 1664-3224
• DOI: 10.3389/fimmu.2022.838636

Ankylosing spondylitis (AS) is an immune-mediated inflammatory disorder that primarily affects the axial skeleton, especially the sacroiliac joints and spine. This results in chronic back pain and, in extreme cases, ankylosis of the spine. Despite its debilitating effects, the pathogenesis of AS remains to be further elucidated. This study used single cell CITE-seq technology to analyze peripheral blood mononuclear cells (PBMCs) in AS and in healthy controls. We identified a number of molecular features associated with AS. CD52 was found to be overexpressed in both RNA and surface protein expression across several cell types in patients with AS. CD16 monocytes overexpressed and IL-18Rα in AS, while CD8 T cells and natural killer cells overexpressed genes linked with cytotoxicity, including , and . Tregs underexpressed CD39 in AS, suggesting reduced functionality. We identified an overrepresented NK cell subset in AS that overexpressed CD16, CD161, and CD38, as well as cytotoxic genes and pathways. Finally, we developed machine learning models derived from CITE-seq data for the classification of AS and achieved an Area Under the Receiver Operating Characteristic (AUROC) curve of > 0.95. In summary, CITE-seq identification of AS-associated genes and surface proteins in specific cell subsets informs our understanding of pathogenesis and potential new therapeutic targets, while providing new approaches for diagnosis machine learning.