University Medical Center Groningen Groningen, Netherlands
Disclosure information not submitted.
Paulina Czarnota1, Rick Wilbrink2, Bhuwan Khatri1, Anna Stolarczyk1, Cherilyn Pritchett Frazee1, Chuang Li1, Caleb Marlin1, Kyle Wright3, Kandice Tessneer1, Lida Radfar4, Judith James1, Robert Scofield1, Indra Adrianto5, Astrid Rasmussen1, Joel Guthridge1, A. Darise Farris1 and Christopher Lessard1, 1Oklahoma Medical Research Foundation, Oklahoma City, OK, 2University Medical Center Groningen, Groningen, Netherlands, 3University of Oklahoma Health Sciences Center, Oklahoma City, OK, 4University of Oklahoma College of Dentistry, Oklahoma City, OK, 5Henry Ford Health, Detroit, MI
Background/Purpose: 10X Visium spatial transcriptomics evaluates gene expression in a 50μm tile coordinate of a sectioned tissue, yielding heterogeneous cell sampling. Spatial PCA algorithm was developed using homogenous tissue types with distinct boundaries to identify like tissue regions and determine the cellular context of spatial coordinates for spatial transcriptomic analyses [1]. However, it is less effective at differentiating like tiles from heterogenous tissue types such as the minor salivary gland; the target tissue of Sjögren’s disease (SjD). This study introduces a novel approach, HistoPCA, that connects pathologically annotated and segmented images with spatial coordinates derived from minor salivary gland biopsies to group like tiles based on similar cellularity.
Methods: Histopathological images of minor salivary gland biopsies were annotated based on tissue type (fibrosis, glandular, inflammatory, fat), then annotations were combined with spatial coordinates using the novel HistoPCA algorithm. Image segmentation was used to obtain cell numbers in tiles and combined with pseudo-color percentages associated with certain tissue types to perform unsupervised KMeans clustering. Differential expression (DE) between SjD cases and controls was analyzed using pseudo-bulk gene expression, then analyzed by Ingenuity Pathway Analysis (IPA).
Results: HistoPCA and UMAP with KMeans clustering outperformed Spatial PCA (Adjusted Rand Index (ARI) of 0.52 vs 0.51), identifying four distinct clusters in minor salivary gland. Despite distinct features such as inflammation in Cluster 1 (21% inflammatory, 32% fibrosis, and 23% glandular tiles) and increased glandular involvement in Cluster 3 (0% inflammatory, 48% fibrosis, and 41% glandular tiles), DE and pathway analyses showed similar patterns of dysregulation in SjD cases (n=28) compared to controls (n=16). As a positive control, elevated interferon signatures were observed in all SjD cases and anti-Ro positive SjD cases (n=19) compared to controls across all clusters [2]. Additional immune pathways (T cell receptor, Th1, Th2, and IL-4 signaling, etc.) were also upregulated in SjD cases while CTLA4, IL-10, and IL-12 signaling pathways were downregulated.
Conclusion: HistoPCA is a novel approach that uses histopathological and segmentation data to successfully group like tiles from spatial transcriptomic analysis of heterogeneous tissues, e.g., minor salivary gland. Cluster annotation of HistoPCs (principal components), followed by pseudo-bulk DE and pathway analyses identified four clusters with dysregulated immune pathways in SjD.Further, dysregulation of immune pathways in Cluster 3, which was devoid of tiles associated with inflammation, indicates potential glandular involvement in SjD pathogenesis. Future analyses within spatial contexts will provide additional insights into the role of different cell/tissue types in SjD pathobiology of the salivary gland.
References: [1] Shang L, et al. Nat Commun. (2022); 13:7203 [2] Hjelmervik TOR, et al. Arthritis Rheum (2005);52(5):1534-44
P. Czarnota: None; R. Wilbrink: None; B. Khatri: None; A. Stolarczyk: None; C. Pritchett Frazee: None; C. Li: None; C. Marlin: None; K. Wright: Scopio Labs, 1; K. Tessneer: None; L. Radfar: None; J. James: Bristol-Myers Squibb(BMS), 5, GlaxoSmithKlein(GSK), 2, Novartis, 2, Progentec Biosciences, 5; R. Scofield: Merck/MSD, 2, Pfizer, 1; I. Adrianto: None; A. Rasmussen: None; J. Guthridge: None; A. Farris: Janssen Research and Development, LLC, 5; C. Lessard: Janssen, 5.
