The single-cell transcriptional landscape of innate and adaptive lymphocytes in pediatric-onset colitis

Abstract

Innate lymphoid cells (ILCs) are considered innate counterparts of adaptive T cells; however, their common and unique transcriptional signatures in pediatric inflammatory bowel disease (pIBD) are largely unknown. Here, we report a dysregulated colonic ILC composition in pIBD colitis that correlates with inflammatory activity, including accumulation of naive-like CD45RA⁺CD62L^- ILCs. Weighted gene co-expression network analysis (WGCNA) reveals modules of genes that are shared or unique across innate and adaptive lymphocytes. Shared modules include genes associated with activation/tissue residency, naivety/quiescence, and antigen presentation. Lastly, nearest-neighbor-based analysis facilitates the identification of "most inflamed" and "least inflamed" lymphocytes in pIBD colon with unique transcriptional signatures. Our study reveals shared and unique transcriptional signatures of colonic ILCs and T cells in pIBD. We also provide insight into the transcriptional regulation of colonic inflammation, deepening our understanding of the potential mechanisms involved in pIBD.

Keywords: IBD; ILC; gut immunity; pediatric inflammatory bowel disease; single-cell RNA sequencing; tissue-resident T cells.

Graphical abstract

Figure 1

Dysregulated colonic ILC composition correlates to inflammatory activity in pIBD colitis (A) Bar graphs showing the frequencies of total ILCs, T cells, and NK cells (gated as shown in Figure S1A) in non-IBD (n = 10), endoscopically non-inflamed (n = 12), and endoscopically inflamed (n = 15) pIBD colon samples. A total of 8 samples were derived from paired inflamed and non-inflamed samples. Frequencies shown out of CD45⁺ lymphocytes. (B) Bar graphs showing the frequencies of ILC1s, ILC2s, NKp44⁺ ILC3s, and NKp44⁻ ILCs (gated as shown in Figure S1A) in non-IBD (n = 10), endoscopically non-inflamed (n = 12), and endoscopically inflamed (n = 15) pIBD colon samples. Frequencies shown out of total ILCs. (C) Correlation plots between frequencies of ILC1s, ILC2s, NKp44⁺ ILC3s, and NKp44⁻ ILCs and the endoscopic and histological scores determined for each of the endoscopically non-inflamed (n = 12) and endoscopically inflamed (n = 15) pIBD colon samples. Statistical significances in (B) were detected with Mann-Whitney test. Bar graphs are shown as median ± interquartile range (IQR). Spearman’s rank correlation test was applied for assessing correlations between variables. In (A)–(C), cells from each donor were analyzed in independent experiments.

Figure 2

Single-cell transcriptional profiling of colonic ILCs, NK cells, and T cells in pIBD colitis (A) Schematic illustration of the scRNA-seq workflow of the total 12 paired colonic biopsy samples obtained from the endoscopically least and most inflamed areas of the 6 pediatric patients with CD-colitis (n = 6 donors). (B) UMAP visualization of the scRNA-seq data showing the unbiased clustering of total 44 717 ILCs, NK cells, and T cells. (C) Dotplot displaying expression of lineage-identity genes for the annotation of the clusters in (B). In (A)–(C), cells from each donor were analyzed in independent experiments.

Figure 3

Gene module analyses of colonic innate and adaptive lymphocytes reveal subset-specific and shared programs of co-expressed genes (A) Summary flow chart used for the analysis of scRNA-seq data and subsequent gene module and nearest-neighbor analysis. (B) Circular dendrogram illustrating the number and composition of each module and the contribution of each module to discriminating cell clusters and donors. Connecting lines represent modules with highly correlated expressions (kNN [k nearest neighbor]). Modules selected for analysis and described in the text are marked with an asterisk. (C–I) A selection of genes in each module marked with an asterisk in (B) illustrated by dotplots. The expression of the modules across the annotated clusters is shown in Figure S3.

Figure 4

Gene modules associate with ILC subclusters and identify quiescent/naive-like ILCs (A) UMAP visualization of ILC subclusters after reclustering main cluster 1 shown in Figure 2B. (B) Dotplot showing selected ILC-module genes (Figures 3D–3I). The expression is shown within the annotated subclusters of ILCs depicted in (A). (C–G) UMAP visualization of selected ILC-module genes in the ILC subclusters. Dotted lines mark the subcluster where the depicted gene is highly and differentially expressed. A list of differentially expressed genes (DEGs) per ILC subcluster is shown in Figure S4A. (H) Developmental trajectories of ILC subclusters performed with Monocle 3. Subcluster 4 (dotted circle) was used as the root. (I) Frequencies of CD45RA⁺ ILCs among ILCs expressing combinations of HLA-DR and NKp44 in inflamed pIBD (n = 12–14) and non-IBD (n = 8) colon biopsies (bar graphs are shown as median). The samples were analyzed in independent experiments. (J) Representative flow cytometry plots of CD45RA versus CD62L (marked in red is the CD45RA⁺ ILC subset), gated out of NKp44⁻HLA-DR⁻ ILCs, in inflamed pIBD and non-IBD colon biopsies. (K) Frequency of CD45RA⁺ ILCs out of CD161⁺ ILCs in non-ΙBD (n = 8), endoscopically non-inflamed (n = 10), and endoscopically inflamed pIBD (n = 14) colon biopsies. Mann-Whitney test (p < 0.05), mean with SD. The samples were analyzed in independent experiments. (L) Correlation plots (Spearman, p < 0.05) between the frequency of CD45RA⁺ ILCs and the endoscopic and histological scores determined for each of the endoscopically non-inflamed (n = 8) and endoscopically inflamed (n = 14) pIBD colon samples.

Figure 5

Nearest-neighbor-smoothed histological scores reveal the most and least inflammatory cells (A) Schematic illustration showing the histological scores from the endoscopically least and most inflamed paired colonic biopsies used for scRNA-seq. (B) Pie chart showing the total number of cells from areas of different histological scores from all 12 pIBD colon biopsy samples. (C) The raw histological score of each cell derived from each sample depicted onto the main UMAP from Figure 2B. (D) Schematic illustration of the Euclidean neighbor smoothing. Each cell was given the average histological scores of its 100 transcriptionally closest neighbors. These smoothed histological scores were used for the subsequent analysis. (E) Histogram showing the cell type distribution across the histological scores derived from the nearest-neighbor analysis. Cells were categorized into 3 groups: ≤4 (least inflamed), 5–8 (intermediate), and ≥9 (most inflamed). (F) Visualization of cells from the nearest-neighbor smoothing on the main UMAP with cells derived from the ≥9 histological (histo) score (most inflamed) and ≤4 histo score (least inflamed) groups, which also have neighbors from at least 5 samples. (G and H) Violin plots showing the expression level of selected DEGs (log2 fold change [FC] > 0.10 and adjusted p value [adj. p value] < 0.05) upregulated in the most (G) and least (H) inflamed cells versus the rest of the cells in each cluster. (I) Frequencies of CD4⁺CD69⁺, CD4⁺CD161⁺CD69⁺, CD8⁺CD69⁺, and CD8⁺CD161⁺CD19⁺ cells in colonic adult non-IBD (n = 8) and IBD samples (n = 8). Mann-Whitney test (p < 0.05), mean with SD. The samples were analyzed in two independent experiments.