A validated single-cell-based strategy to identify diagnostic and therapeutic targets in complex diseases

Abstract

Background: Genomic medicine has paved the way for identifying biomarkers and therapeutically actionable targets for complex diseases, but is complicated by the involvement of thousands of variably expressed genes across multiple cell types. Single-cell RNA-sequencing study (scRNA-seq) allows the characterization of such complex changes in whole organs.

Methods: The study is based on applying network tools to organize and analyze scRNA-seq data from a mouse model of arthritis and human rheumatoid arthritis, in order to find diagnostic biomarkers and therapeutic targets. Diagnostic validation studies were performed using expression profiling data and potential protein biomarkers from prospective clinical studies of 13 diseases. A candidate drug was examined by a treatment study of a mouse model of arthritis, using phenotypic, immunohistochemical, and cellular analyses as read-outs.

Results: We performed the first systematic analysis of pathways, potential biomarkers, and drug targets in scRNA-seq data from a complex disease, starting with inflamed joints and lymph nodes from a mouse model of arthritis. We found the involvement of hundreds of pathways, biomarkers, and drug targets that differed greatly between cell types. Analyses of scRNA-seq and GWAS data from human rheumatoid arthritis (RA) supported a similar dispersion of pathogenic mechanisms in different cell types. Thus, systems-level approaches to prioritize biomarkers and drugs are needed. Here, we present a prioritization strategy that is based on constructing network models of disease-associated cell types and interactions using scRNA-seq data from our mouse model of arthritis, as well as human RA, which we term multicellular disease models (MCDMs). We find that the network centrality of MCDM cell types correlates with the enrichment of genes harboring genetic variants associated with RA and thus could potentially be used to prioritize cell types and genes for diagnostics and therapeutics. We validated this hypothesis in a large-scale study of patients with 13 different autoimmune, allergic, infectious, malignant, endocrine, metabolic, and cardiovascular diseases, as well as a therapeutic study of the mouse arthritis model.

Conclusions: Overall, our results support that our strategy has the potential to help prioritize diagnostic and therapeutic targets in human disease.

Keywords: Biomarker and drug discovery; Network tools; scRNA-seq.

Fig. 1

scRNA-seq analysis of a mouse model of antigen-induced arthritis (AIA). a An overview of the AIA mouse model. b Representative joint images from naïve mice and arthritic joints after hematoxylin and eosin (H&E) staining. B, bone marrow; S, synovial cavity; C, cartilage. Arrows indicate (1) infiltration of inflammatory cells to the synovium, (2) cartilage/bone destruction, and (3) hyperplasia of the synovial lining. c A schematic overview of seq-well scRNA-seq and cell type identification using reference component analysis (RCA). d t-SNE plot of 7086 healthy and RA joint cells (n = 4 healthy mice samples and 5 sick mice samples), and 1333 healthy and AIA lymph nodes cells (n = 4 healthy mice samples and 5 sick mice samples), colored by RCA clusters

Fig. 2

Multicellular disease models (MCDMs) from a mouse model of AIA. MCDMs were constructed based on scRNA-seq data by connecting differentially expressed genes in each cell type with predicted upstream regulators in all other cell types. Cell type size corresponds to centrality score. Numbers indicated by the nodes denote the number of identified cells of specific type (for example in RA joint, we have identified 4258 granulocytes). a An MCDM of lymph nodes from arthritic mice. b An MCDM from arthritic joints. c Multicellular model of a healthy mouse joint (lymph node model is not shown because there was only one predicted interaction). Gene names of predicted upstream regulators are indicated on arrows. Treg, T regulatory cells. d Correlation between centrality score of cell types and enrichment of genes harboring genetic variants identified by GWAS and expert curated repositories among differentially expressed genes (the genes were derived from DisGeNet and the analysis based on the mouse orthologues of the human genes)

Fig. 3

Network models of disease-associated cell types. a 24 cell types and subsets that were significantly enriched for GWAS-enriched epigenetic markers associated with RA. Cell type size corresponds to association −ln (p value). b Network model of cell types associated with human rheumatoid arthritis (RA). Nodes correspond to cell types, node size corresponds to significance of association (−log10 RA GWAS-epigenetic marker enrichment p value). Cell types with potential spatial interactions are linked, and cell type position depends on the centrality score as indicated by the rings in the background. c Bar plot of cell type classes ordered by significance of association with 175 human diseases (Fisher combined GWAS-enriched epigenetic markers – disease association p value calculated for each cell type class). d Network model of cell types associated with 175 diseases, based on the same parameters as in b (for details see results)

Fig. 4

Diagnostic potential of CD4+ T cells based on clinical profiling studies of 13 diseases. a Toy model of a disease module. Disease-associated genes (red) are mapped on proteins (blue) in the human protein-protein interaction network. Disease-associated genes that co-localize form a module. b Overview of the module-based analyses. First step is the identification of disease modules for each of the 13 diseases profiled in the prospective microarray study of CD4+ T cells. For each disease module, genes separate patients from healthy controls. For pairwise comparison of the diseases, genes in the union of two respective modules separate patients with different diseases; for example, genes in influenza and asthma modules separate patients with influenza from patients suffering from asthma with AUC of 0.99, p = 3.3 × 10⁻⁵, as shown in c. c Heatmap presenting area under the curve (AUC) values of 13 disease classifications based on the module intersections genes, using elastic net. d An independent validation study of classification accuracy of breast cancer patients (n = 24) and healthy subjects (n = 14) based on previously preselected biomarkers (genes) measured in CD4+ T cells. Classification was performed with elastic net, preserving same lambda (λ) value as estimated for the original study. e–j Potential diagnostic classification of IBD patients based on six secreted plasma proteins identified in the intersection of ulcerative colitis (UC) and Crohn’s disease (CD) modules. These proteins could separate patients from healthy controls (HCs). e CXCL11; f CCL25; g CXCL1; h CXCL8; i IL1B; j TNF. k Crohn’s disease and ulcerative colitis patients’ classification based on normalized protein levels of CXCL1 and CXCL8. UC, ulcerative colitis; CD, Crohn disease; HC, healthy controls. Star denotes p value < 0.05. d–k The bars in the boxes represent median and 25th and 75th percentiles, while whiskers extend to ± 2.7σ (see the “Methods” section)

Fig. 5

Bezafibrate protects against antigen-induced arthritis (AIA). Female mice with mBSA-induced arthritis were intraperitoneally (i.p.) treated with bezafibrate (n = 4) or mock (AIA control, n = 5). a Arthritis severity was scored based on histopathology day 28 in the two groups (H&E staining, vertical bars indicate median, differences between groups evaluated using the Mann-Whitney U test, *p < 0.05). b Representative H&E joint image from the bezafibrate-treated mice. c Antigen recall response of CD4+ helper T cells among spleen and lymph node cells isolated from mock- (AIA control, n = 5) or bezafibrate-treated (n = 4) mice; vertical bars indicate mean ± SEM, differences between groups evaluated using the two-sided Mann-Whitney U test *p < 0.05)