A Proximal-to-Distal Survey of Healthy Adult Human Small Intestine and Colon Epithelium by Single-Cell Transcriptomics

Abstract

Background & aims: Single-cell transcriptomics offer unprecedented resolution of tissue function at the cellular level, yet studies analyzing healthy adult human small intestine and colon are sparse. Here, we present single-cell transcriptomics covering the duodenum, jejunum, ileum, and ascending, transverse, and descending colon from 3 human beings.

Methods: A total of 12,590 single epithelial cells from 3 independently processed organ donors were evaluated for organ-specific lineage biomarkers, differentially regulated genes, receptors, and drug targets. Analyses focused on intrinsic cell properties and their capacity for response to extrinsic signals along the gut axis across different human beings.

Results: Cells were assigned to 25 epithelial lineage clusters. Multiple accepted intestinal stem cell markers do not specifically mark all human intestinal stem cells. Lysozyme expression is not unique to human Paneth cells, and Paneth cells lack expression of expected niche factors. Bestrophin 4 (BEST4)⁺ cells express Neuropeptide Y (NPY) and show maturational differences between the small intestine and colon. Tuft cells possess a broad ability to interact with the innate and adaptive immune systems through previously unreported receptors. Some classes of mucins, hormones, cell junctions, and nutrient absorption genes show unappreciated regional expression differences across lineages. The differential expression of receptors and drug targets across lineages show biological variation and the potential for variegated responses.

Conclusions: Our study identifies novel lineage marker genes, covers regional differences, shows important differences between mouse and human gut epithelium, and reveals insight into how the epithelium responds to the environment and drugs. This comprehensive cell atlas of the healthy adult human intestinal epithelium resolves likely functional differences across anatomic regions along the gastrointestinal tract and advances our understanding of human intestinal physiology.

Keywords: BEST4; Cell Atlas; Intestinal Stem Cell; Paneth Cell; scRNAseq.

Graphical abstract

Figure 1

Sample processing. (A) Schematic for isolating single epithelial cells from 6 intestinal regions for 3 donors and then using hashtag antibodies to sequence cells from all regions side-by-side. (B) UMAP of all analyzed cells in 25 lineage clusters. See Figure 19 for clearly marked clusters. (C–E) UMAP of all cells colored by (C) donor or (D and E) region. (F and G) Heatmaps showing unique markers for major lineages in (F) SI and (G) colon. See Supplementary Table 1 for total DEGs for each lineage. Schematics in panel A were created with BioRender.com. Duo, Duodenum; FACS, fluorescence-activated cell sorter; Ile, Ileum; Interm, intermediate; Jej,Jejunum; Sec. Prog., Secretory Progenitor; UMAP, Uniform Manifold Approximation and Projection.

Figure 2

Patient characteristics and cell counts. (A) Donor information. (B) Cells collected per donor region. (C) Small intestinal lineages collected per donor. (D) Colonic lineages collected per donor. (E) Small intestinal lineages per donor region. (F) Colonic lineages per donor region. Abs., Absorptive; Asc., Ascending; BMI, body mass index; Desc., Descending; Trans., Transverse.

Figure 3

Determining final lineage clusters. (A) Initial Leiden clustering for all cells. (B) Splitting EEC and secretory progenitors by organ. (C and D) An ITLN1-high cluster, all from SI, contains cells expressing PC markers (DEFA5, DEFA6, ITLN2, LYZ) along with cells expressing the GC marker MUC2. (C) cluster defined by ITLN1; (D) UMAP expression of PC and GC markers within the ITLN1-high cluster. (E) Subclustering to define Paneth and Goblet cells. (F) Dotplot showing expression of classic PC and GC genes across the new PC and GC clusters. UMAP, Uniform Manifold Approximation and Projection.

Figure 4

Proliferative crypt populations. (A) Heatmap of DEGs in ISCs vs other lineages (top; red: classic markers), SI vs colon ISCs (middle), and colon vs SI ISCs (bottom). (B) UMAP of LGR5, OLFM4, and RARRES2 expression. (C) Dotplot showing expression of LGR5, OLFM4, and RARRES2 across proliferative lineages of the SI (left) and colon (right). (D) Venn diagram showing overlap between our human ISC signature and a previously described murine signature. (E) Heatmap of DEGs in TA cells vs other lineages (top), SI vs colon TA cells (middle), and colon vs SI TA cells (bottom). (F) Dotplot showing DEGs defined in SI- or colon-specific mature lineages expressing within organ-delineated ISCs and TA cells. (G and H) PAGA showing connectivity between major lineages in (G) SI and (H) colon to infer the maturation trajectory. Line thickness represents connectivity strength. (I–K) Regional cell-cycle phase distribution in (I) ISCs, (J) TA cells, and (K) secretory progenitors. C, Colon; Duo, Duodenum; eACC, Early Absorptive Colonocytes; eAE, Early Absorptive Enterocytes; Gob, Goblet; iAE, Intermediate Absorptive Enterocytes; Ile, Ileum; Jej, Jejunum; mACC, Mature Absorptive Colonocytes; mAE, Mature Absorptive Enterocytes; Pan, Paneth; Sec. prog., Secretory Progenitor.

Figure 5

DEG dotplots for each lineage. Dotplots showing expression of top 20 DEGs for each lineage, as sorted by expression fold-change above the cluster with the next highest expression. DEGs included are genes significantly enriched in the lineage in both the SI and colon (if applicable).

Figure 6

Organ-specific lineage DEGs. Relating to Figure 4F, UMAPs showing expression of DEGs from mature lineages found to be more highly enriched in SI or colon ISCs and TA cells. UMAP, Uniform Manifold Approximation and Projection.

Figure 7

Paneth cells. (A) Heatmap of DEGs in PCs vs other lineages (red: classic markers). (B) Dotplot showing lysozyme mRNA expression across FAE, BEST4, Paneth, and tuft cell lineages. (C) Dotplot showing expression of PC, goblet, and BEST4⁺ cell classic markers across the PC, goblet, and BEST4⁺ cell clusters. (D) Dotplot showing growth factors shown to be expressed in murine PCs in previous literature across human SI lineages. (E) Dotplot showing all members of major intestinal growth factor families that show detectable expression in PCs across SI lineages. (F) Heatmap showing PC (top) and ISC markers (bottom) across all SI ISCs (left) and PCs (right). (G) Immunofluorescence staining for LYZ protein (magenta), in situ hybridization showing SMOC2 mRNA (white), and nuclei (blue) in a human ileum crypt base. Maximum projection of eight 0.5-μm optical slices. Scale bar: 20 μm. (H) Dotplot showing expression of all Frizzled family receptors across SI lineages. (I) Dotplot showing the 10 highest-expressed antimicrobial peptides across SI lineages.

Figure 8

BEST4⁺cells. (A) Heatmap of DEGs in BEST4⁺ cells vs other lineages (top; red: classic markers), SI vs colon BEST4⁺ cells (middle), colon vs SI BEST4⁺ cells (bottom). (B) UMAPs showing all SI (left) and colon (right) BEST4⁺ cells colored according to predicted diffusion pseudotime or expression of BEST4 mRNA. (C and D) In situ hybridization showing BEST4 mRNA (magenta), immunofluorescence staining CFTR protein (white), and nuclei (blue) in human (C) jejunum and (D) colon. Scale bars: 100 μm; 20 μm (zoomed panels). (E) Dotplot showing secreted genes (top) and their receptors (bottom) across lineages. (F) UMAPs of BEST4⁺ cells showing predicted diffusion pseudotime and expression of secreted peptides. (G) Expression of NPY, SI, and APOA1 across regions for each donor. (H) Dotplot showing genes involved in metal binding and endocytosis across lineages. UMAP, Uniform Manifold Approximation and Projection. CFTR, Cystic Fibrosis Transmembrane Conductance Regulator; GUCA2, Guanylate Cyclase Activator 2A; NPY, Neuropeptide Y.

Figure 9

Tuft cells. (A) Heatmap of DEGs in tuft cells vs other lineages (top; red: classic markers), SI vs colon tuft cells (middle), and colon vs SI tuft cells (bottom). (B) Dotplot showing tuft cell enrichment of genes specific to taste signal transduction. (C) Organ-specific signal transduction in SI vs colon tuft cells. (D) Dotplot showing tuft cell–enriched genes enabling interactions with innate and adaptive immune system. (E) Dotplot showing 10 highest-expressed antimicrobial peptides across colon lineages. (F) Dotplot showing tuft cell–specific genes for producing acetylcholine, eicosanoids, and prostaglandins. Schematics in panel C were created with BioRender.com. C, Colon; GPCR, G-protein Coupled Receptor; MHC, major histocompatibility complex; TLR, Toll-like receptor.

Figure 10

Goblet cells. (A) Heatmap of DEGs in GCs vs other lineages (top; red: classic markers), SI vs colon GCs (middle), and colon vs SI GCs (bottom). (B) Dotplot showing expression of the 9 highest-expressed mucins across GCs and proliferative and absorptive lineages of the SI and colon (blue: gel-forming mucins). (C) Dotplot showing the 9 highest-expressed mucins across GCs by region. (D) Dotplot showing expression of the 9 highest-expressed mucins in all absorptive enterocytes and colonocytes by intestinal region. (E) Leiden subclustering of colon GCs. (F) Diffusion pseudotime of colon GCs. (G) UMAP of MUC2 expression in colon GCs. (H) Dotplot showing markers of murine GC subpopulations in the human colon GC subclusters defined in panel E. (I) Immunofluorescence staining for protein expression of RAB27A (white), MUC2 (magenta), and nuclei (blue) in human colon (2-μm optical slice). (J) Dotplot showing expression of mucins in colonic icGCs, crypt-resident goblet cells (crGCs), and early goblet cells. (K) Immunofluorescence staining for protein expression of MUC5B (white), MUC2 (magenta), and nuclei (blue) in human colon (2-μm optical slice). Scale bars: 50 μm. UMAP, Uniform Manifold Approximation and Projection; Duo, Duodenum; Jej, Jejunum.

Figure 11

SI goblet cell subclustering. (A) Left: Leiden subclustering of SI goblet cells, with subclusters named according to genes with high expression. Middle: UMAP of SI goblet cells marked by diffusion pseudotime. Right: UMAP of SI goblet cells marked by MUC2 expression. (B) Dotplot showing expression of mucins in SI GC subpopulations. (C) Dotplot showing expression of mouse-implicated markers of GC subpopulations in human SI GC subclusters. UMAP, Uniform Manifold Approximation and Projection.

Figure 12

Enteroendocrine cells. (A) Heatmap of DEGs in EECs vs other lineages (top, red: classic markers), SI vs colon EECs (middle), and colon vs SI EECs (bottom). (B) Dotplot of EEC regional hormone gene expression. (C) Dotplot of EEC expression of select receptors by region. (D) Dotplot showing expression of DEGs of SI or colon EECs that are present in the GOCC_Presynapse gene list. (E) Pie chart of all genes within the GOCC_Presynapse gene list shown by lineage in which they have the highest expression (SI and colon lineages are combined when applicable). (F) Heatmap showing hormone expression in each individual EEC. Duo, Duodenum; G/K, G-cells/K-cells; Ile, Ileum; Jej, Jejunum; M/X, M-cells/X-cellsD.

Figure 13

Absorptive cells. (A) Heatmap of DEGs in absorptive cells vs other lineages (top), AEs vs ACCs (middle), and ACCs vs SI AEs (bottom). (B) UMAPs showing the AE2 Leiden cluster (top) and cells by region (bottom). (C) Dotplot of classic mature AE markers and top 10 DEGs for AE2 cluster. (D) Dotplots showing regional expression of genes involved in digestion and absorption in all AEs and ACCs. (E) Dotplots showing the 20 highest-expressed cell junction genes in AEs and ACCs by region. (F) Dotplots showing regional aquaporin expression in AEs and ACCs. (G) Dotplot showing aquaporin expression across lineages. (H) UMAPs of late ACCs showing predicted diffusion pseudotime (left) and AQP8 expression (right). UMAP, Uniform Manifold Approximation and Projection; Duo, Duodenum; Ile, Ileum; Jej, Jejunum.

Figure 14

Follicle-associated epithelium. (A) Top: Dotplot showing expression of EPCAM and conserved M-cell markers and other genes known to interact with the immune system across lineages. Bottom: Genes implicated in mouse M cells that are not specific to human FAE. (B) Dotplot showing expression of the top 20 FAE DEGs across lineages. EPCAM, Epithelial Cell Adhesion Molecule; C, Colon; MHC, major histocompatibility complex; TNFA, Tumor Necrosis Factor Alpha.

Figure 15

Extrinsic receptors and drug targets. (A) Dotplot showing expression of the 5 highest-expressing members of each major receptor family by lineage. Top: Small intestinal lineages. Bottom: Colonic lineages. (B) Pie chart showing receptor genes expressed in the intestinal epithelium by lineage with the highest expression. NOD, Nucleotide-binding Oligomerization Domain Sec. Prog., Secretory Progenitor; TNF, tumor necrosis factor.

Figure 16

Drug targets. (A) Pie charts of primary targets of all approved drugs (left) and all phase I (center) and phase II (right) drug metabolism genes expressed in the intestinal epithelium shown by lineage with highest expression. (B) Primary targets of drugs used to treat IBD found to have expression in the intestinal epithelium. (C) Dotplots showing expression of primary targets of drugs used to treat IBD by lineage split into high-, middle-, and low-expressing tables for better visualization. Note scaling changes between tables. (D) Left: In situ hybridization showing BEST4 RNA (white) and immunofluorescence staining for FKBP1A protein (magenta) and nuclei (blue) in human jejunum (top) and colon (bottom). Center: Immunofluorescence staining for IMPDH2 protein (magenta), KI67 (white), and nuclei (blue) in human jejunum (top) and colon (bottom). Right: Immunofluorescence staining for DHFR protein (magenta), KI67 (white), and nuclei (blue) in human jejunum (top) and colon (bottom). Optical slice was 2 μm for all. Scale bars: 50 μm. (E) Dotplot showing expression of the top 3 highest-expressed targets of IBD drugs in the intestinal epithelium across regions and split by donor. FKBP1A, FKBP Prolyl Isomerase 1A; IMPDH2, Inosine Monophosphate Dehydrogenase 2; JAK, Janus Kinase; Sec. Prog., Secretory Progenitor.

Figure 17

Tissue histology. H&E-stained tissues from each region for all 3 donors. Scale bars: 200 μm.

Figure 18

Fluorescence-activated cell sorter (FACS) strategy. FACS strategy for gating out cell fragments, likely doublets, and dead cells. APC, Allophycocyanin.

Figure 19

Final clusters shown by organ.Top: Final lineage clusters used for the rest of the analyses in our study. Bottom: Lineage clusters split by region. C, Colon; Interm., intermediate; Sec. Prog., Secretory Progentior; UMAP, Uniform Manifold Approximation and Projection.

Figure 20

Hashtag deconvolution. (A) Per donor hashtag noise distributions. Blue dotted lines indicate the 99th percentile values for noise. Values above this line were called positive for a specific hashtag. (B) Left: K-medoid clustering for each donor based only on hashtag reads. Cells positive (P < .01) for multiple hashtags were removed as likely multiplets. Cells were called as negative if they did not surpass the noise threshold for all hashtags. Right: K-medoid clustering with final hashtag labeling for nonmultiplet cells.

Figure 21

Filtering for cell quality. Total counts, N genes, and mitochondrial gene percentages shown for each donor before and after filtering; prefiltering (top) and postfiltering (bottom) by (A) donor and (B) region. Note differences in the Y axes between prefiltering and postfiltering rows. Duo, Duodenum; Ile, Ileum; Jej, Jejunum.