Single-Cell Sequencing of Developing Human Gut Reveals Transcriptional Links to Childhood Crohn's Disease

Abstract

Human gut development requires the orchestrated interaction of differentiating cell types. Here, we generate an in-depth single-cell map of the developing human intestine at 6-10 weeks post-conception. Our analysis reveals the transcriptional profile of cycling epithelial precursor cells; distinct from LGR5-expressing cells. We propose that these cells may contribute to differentiated cell subsets via the generation of LGR5-expressing stem cells and receive signals from surrounding mesenchymal cells. Furthermore, we draw parallels between the transcriptomes of ex vivo tissues and in vitro fetal organoids, revealing the maturation of organoid cultures in a dish. Lastly, we compare scRNA-seq profiles from pediatric Crohn's disease epithelium alongside matched healthy controls to reveal disease-associated changes in the epithelial composition. Contrasting these with the fetal profiles reveals the re-activation of fetal transcription factors in Crohn's disease. Our study provides a resource available at www.gutcellatlas.org, and underscores the importance of unraveling fetal development in understanding disease.

Keywords: human fetal gut development; inflammatory bowel disease; intestinal organoids; intestinal stem cells; pediatric Crohn's disease; single-cell RNA sequencing; villus formation.

Graphical abstract

Figure 1

Single-Cell Profiling of Embryonic, Fetal and Pediatric Gut (A) Schematic illustration of experimental design. Blue circles mark biopsy location (i.e., terminal ileum). (B) Tissue dissociation and single-cell sequencing strategy. (C and D) (C) and (D) UMAP projection of embryonic/fetal (n = 9 donors) and childhood/adolescence (n = 8 donors) scRNA-seq samples, respectively. (E) Changes in embryonic/fetal cell type abundance (% of cells) at different developmental time points grouped by intestinal regions. Time point annotation colored in yellow and green are embryonic and fetal, respectively. Crypt/absorptive epithelium are SI, colonic epithelium, and uniform progenitors grouped together. (F and G) UMAP plots colored by (F) gut region of the embryonic cells, (G) post-conception week of embryonic and fetal cells as in (C). Circled populations in F are epithelial cells. EC, endothelial cell; FLC, fibroblasts; Epi, epithelium; SI, small intestinal; prog, progenitors; ICC, interstitial cells of Cajal. See also Figure S1; Table S1.

Figure 2

Cell-Type Groups and Their Marker Genes Identified in Fetal and Pediatric Datasets (A and C) UMAP plots of fetal and pediatric datasets (A and C, respectively) broadly grouped into seven groups: epithelial (blue), mesenchymal (dark pink), neural (orange), endothelial (green), immune (light pink), and erythroid lineage (brown). (B and D) Dot plots of relative expression and percentage of cells expressing marker genes in fetal (B) and pediatric (D) datasets. The color bars match the cell-type group colors. Epi, epithelium; FLC, fibroblasts; EC, endothelial cells; ICC, interstitial cells of Cajal.

Figure 3

Epithelial Cell Composition during Villus Formation in Humans (A) Representative hematoxylin and eosin staining of embryonic and fetal ileum at 6 and 10 PCW (n = 3 donors). (B) Sub-clustered epithelial cells from duo-jejunum and ileum colored by cell type. (C) Changes in epithelial cell-type abundance (% of cells) at different developmental time points and in two small bowel regions. Colors match the cell-type annotation in (B). (D) Pseudo-spatial distribution of developing epithelial cells along the crypt-villus (base-top) axis. The axis score was derived by using the expression of selected crypt-villus axis markers as defined by Moor et al. (2018) and Parikh et al. (2019). (E) Dotplot with marker genes used to annotate fetal epithelial cell subtypes. (F) smFISH analysis of MKI67, LGR5, BEX5, and TACSTD2 transcripts in embryonic (pseudostratified) and fetal (vilified) epithelium. (Embryonic: 6 PCW; fetal: 10 PCW.) Zoom-in boxes show channels with and without DAPI, as well as each channel independently. Scale bar: main panel, 100 μm; zoom panel, 50 μm. (G–J) (G) Embryonic and (I) fetal epithelium scVelo graphs with overlaid arrows. Expression of LGR5, BEX5, TACSTD2, and cell cycle phase overlaid on (H) embryonic and (J) fetal epithelial cells shown as feature plots. See also Figure S2.

Figure 4

Cell-Cell Interactions that Support Transition from Embryonic to Fetal Epithelium in Humans (A) Abundance of mesenchymal and neuronal cell subsets (% of cells) in developing gut from small (left panel) and large intestines (right panel). (B) Average expression score of hedgehog (HH) pathway genes. (C) Dot plot with expression of BMP and WNT agonists/antagonist and RSPO genes in all mesenchymal cells. (D) Pseudo-positioning schematic of PDGF and HH receptor expression as well as BMP and WNT ligand expression in the cross-section of the developing small bowel upon villus formation. (E and F) (E) Visualization of PDGF ligand and receptor and (F) HH pathway genes in the embryonic small intestine using smFISH at two developmental time points (left panel: embryonic; right panel: fetal). Scale bar: main panel, 100 μm. PLA2G2A expression marks developing smooth muscle, UPK3B+ serosal cells, and uniform progenitor cells. (G) Dot plot of ligand-receptor interactions between uniform progenitor cells and mesenchymal/endothelial populations as predicted using CellPhoneDB analysis in embryonic (columns marked with E) and fetal (columns marked with F) samples. Point size indicates permutation p value and color indicates the scaled mean expression level of ligand and receptor. The interacting cell type and molecule pair relationship is explained in a schematic, where molecule 1 (black) in cell-type cluster 1 (blue) interacts with molecule 2 (red) in cell-type cluster 2 (yellow). FLC, fibroblasts; SMC, smooth muscle cells; EC, endothelial cells; ICC, interstitial cells of Cajal; PCW, post-conception weeks. See also Figure S3.

Figure 5

Fetal Intestinal Organoids Mature in Culture (A) Schematic representation of WNT3A organoid culture experiment. (B) Brightfield images of fetal organoids grown without (WNT3A−) or with (WNT3A+) conditional medium at passage 2, day 5 post-passage. Scale bar: top panels, 1,000 μm, top panels, 200 μm. UMAP plots of single cells from fetal organoids grown with or without WNT3A. (C–E) Cells are colored by either (C) cell type, (D) condition, or (E) cell cycle phase. (F) Cell-type prediction in WNT3A± organoid culture using logistic regression classifiers trained on all primary small intestinal fetal cells. UMAP plots show overlaid predicted probability for selected cell types. Abundance of cell types (% of cells) in organoids, as confidently predicted (over 80% probability of a single cell type) by the logistic regression classifier. (H) Pseudo-spatial distribution of organoid epithelial cells along the crypt-villus (base-top) axis. (I) Schematic representation of experimental design. The experiment was performed using two independent biological samples (replicate 1 from 5.4 PCW and replicate 2 from 6.4 PCW). (J and K) On day 5 of passage 1 (p1) organoids were split and one fraction was kept in the culture while the other was dissociated and processed using 3′ V2 10× protocol. On day 5 of passage 17 (p17) the organoids were dissociated and processed using the 10× platform again. UMAP plots with single cells colored by either (J) passage or (K) cell cycle phase. (L) Prediction of cell types of p1 and p17 organoids using logistic regression trained on primary fetal cells. UMAP plots represent visualization with overlaid predicted probability for selected cell types as in (F). (M) Predicted cell-type abundance or PCW age of fetal organoids grown for 1 or 17 passages as predicted using logistic classifier. N) Pseudo-spatial distribution of organoid epithelial cells along the crypt-villus (base-top) axis. FLC, fibroblasts; SMC, smooth muscle cells; EC, endothelial cells; ICC, interstitial cells of Cajal; PCW, post-conception weeks; prog, progenitor. Organoids were derived from fetal tissues BRC2038- 6.4 PCW, BRC2039 −5.4 PCW, BRC2206- 6.5 PCW. See also Figure S4.

Figure 6

Epithelial Cell Dynamics in Crohn’s Disease Patients Show Transcriptional Similarities to Developing Epithelium (A and B) (A) and (B) UMAP plots of epithelial cell subtypes in healthy children (n = 8) and patients with CD (n = 7), respectively. (C) Epithelial cell-type changes in pediatric health and CD patients. TA, enterocytes, goblet cells, and tuft cell proportions were changed significantly between control and CD patients (p values indicated, t test). (D) Dot plot with ligand-receptor interactions between stromal (S1-S4 FLC) and endothelial cells (Arterial/venous EC) and selected epithelial cell types. Point size indicates permutation p value (CellPhoneDB). Color indicates the scaled mean expression level of ligand and receptor. FLC, fibroblasts; EC, endothelial cell. (E) Heatmap showing the relative mean expression of transcription factors, which were identified to be differentially expressed in CD epithelium, across epithelium from five groups. “Non-inflamed CD” was a group of patients with minimal epithelial composition changes as in (Figure S5F, arrows). Arrows point to genes discussed in text that either have previously been linked to proliferation (red arrows) or inflammation and/or development (black arrows). Epi, epithelium; CD, Crohn’s disease. See also Figures S5 and S6.