Stomach-derived human insulin-secreting organoids restore glucose homeostasis

Abstract

Gut stem cells are accessible by biopsy and propagate robustly in culture, offering an invaluable resource for autologous cell therapies. Insulin-producing cells can be induced in mouse gut, but it has not been possible to generate abundant and durable insulin-secreting cells from human gut tissues to evaluate their potential as a cell therapy for diabetes. Here we describe a protocol to differentiate cultured human gastric stem cells into pancreatic islet-like organoids containing gastric insulin-secreting (GINS) cells that resemble β-cells in molecular hallmarks and function. Sequential activation of the inducing factors NGN3 and PDX1-MAFA led human gastric stem cells onto a distinctive differentiation path, including a SOX4^High endocrine and Galanin^High GINS precursor, before adopting β-cell identity, at efficiencies close to 70%. GINS organoids acquired glucose-stimulated insulin secretion in 10 days and restored glucose homeostasis for over 100 days in diabetic mice after transplantation, providing proof of concept for a promising approach to treat diabetes.

Extended Data Fig. 1 |. Optimizing conditions to induce insulin-expressing cells from cultured human gastric stem cells.

a, Top: stomach samples from 3 different donors; middle, primary hGSC colonies derived from the stomach samples; bottom, immunofluorescence of passage-10 hGSC colony staining for SOX9 and KI67. b, Growth kinetics of hGSCs from three different donors. c, Co-expression of Ngn3, Pdx1, and Mafa using a polycistronic inducible construct in hGSCs yielded low levels of insulin expression (n=3 biological independent samples). Ubc: ubiquitin promoter. d, To optimize the relative timing of Ngn3 and Pdx1- MafA expression, we expressed a Ngn3ER fusion protein in which Ngn3 activity was induced by 4-OH Tamoxifen (4-OH-TAM). Polycistronic PDX1 and MAFA co-expression was controlled by rtTA-TetO and activated by the addition of Doxycycline in the culture medium. Higher INS expression was achieved by sequential activation of the transcription factors NGN3 and PDX1-MAFA. n=3 independent experiments. e, Comparison of PDX1-MAFA with the other transcription factor combinations in insulin induction. 2- day Ngn3ER induction (by 4-OH-TAM) preceded the other TFs, or alternatively, co-expression cassettes were used. ND: not detected. n=3 independent experiments. c, d, e, Data presented as mean ± s.d.; two-tailed unpaired t-test (c), or one-way ANOVA with Dunnett multiple comparisons test (d, e).

Extended Data Fig. 2 |. Formulating chemically defined serum-free medium for GINS organoid differentiation.

a, Experimental design for the supplement screen. Ngn3ER: fusion gene in which NGN3 activity was induced by 4-OH-TAM. Polycistronic PDX1 and MAFA co-expression was controlled by rtTA-TetO and activated by the addition of Doxycycline in the culture medium. mCherry was co-expressed with Ngn3ER in the cell line. Ngn3 was activated from day 0 to day 2. Culture medium was switched to a basal serum-free medium (advanced DMEM/F12, 10 mM HEPES, 1X GlutaMAX, 1X B-27, 1X N-2, and 500 μM N-Acetyl-L-Cysteine) on day 2 with addition of a single supplement and Doxycycline. b, The list of supplements that were screened and the pathways they targeted. Up and down arrows indicate agonists or antagonists, respectively. c, Relative expression of INS mRNA on day-7 post differentiation in comparison with no supplement control. N = 3 (treatments) or 5 (control) independent samples. Nicotinamide and Y-27632 treatment significantly up-regulated INS. d, Spontaneous clustering of cells was evaluated by mCherry live imaging on day-7 post differentiation. Select conditions were shown. A8301 treatment had the most observable clustering effect on the nascent GINS cells. Repeated independently 3 times with similar results. e, Relative expression of β-cell markers measured on day-7 post differentiation in comparison with no supplement control. N = 6 independent samples. c, e, Data presented as mean ± s.d.; one-way ANOVA (c), or two-way ANOVA (e) with Holm–Sidak’s multiple comparisons test.

Extended Data Fig. 3 |. Molecular and functional characterization of GINS organoids derived from multiple donors.

a, Schematic diagram and representative images of cells at key stages in GINS organoid formation. Ngn3ER-hGSCs: human gastric stem cells that incorporated a Ngn3 and estrogen receptor (ER) fusion gene (Ngn3ER); 4-OH-TAM: 4-OH Tamoxifen; Lenti-CMV-PM, lentiviral integration of a polycistronic Pdx1-Mafa co-expression cassette. b, Representative immunofluorescent staining of corpus GINS organoids derived from three different donors, and co-localization of INS and CPPT in GINS cells. a, b, Repeated independently 3 times with similar results. c, To assess CPPT⁺ mono-hormonal cells, a cocktail of GCG, SST and GHRL antibodies were stained together with CPPT in day-21 GINS organoids. Right panel shows immunofluorescent staining of CPPT (red) and a combination of GCG, GHRL and SST (green). left panel shows quantification of mono-hormonal CPPT⁺ cells (n = 10 organoids from donor #6). d, Relative expression of endocrine hormone genes including INS, GCG, SST, and GHRL in day-18 GINS organoids derived from four different donors in comparison with human islets. n = 4 (donor#6) or 3 (donor#7, #9, #10, or islet donors) separate batches of samples for each donor. e, Relative expression of key β-cell markers in GINS organoids derived from different donors in comparison with human islets. n = 4 (donor#6) or 3 (donor#7, #9, #10, or islet donors) separate batches of samples for each donor. f, Glucose-stimulated insulin secretion of GINS organoids at different time points (n = 3 independent samples from donor #6 for each batch of differentiation) or donor #9 (n = 3 independent samples, day-18). g, Insulin secretion of day-18 GINS organoids from donor #6 incubated with the indicated concentrations of glucose with or without 10 nM glibenclamide (Glib) or 0.5 mM diazoxide (Dzx) as indicated (n = 4 independent samples). c-g, Data presented as mean ± s.d.; one-way ANOVA (d, e, g) or repeated-measures two-way ANOVA with Holm–Sidak’s multiple comparisons test for different time points (f) or one-tailed paired t-test for donor#9 (f).

Extended Data Fig. 4 |. Characterizing GINS organoid cells with scRNA-seq.

a, Top UMAP, cells sampled from hGSC cultures (blue, n = 1, donor #6) or GINS organoids (red, n = 1, donor #6); middle UMAP, hGSC cultures included both stem cells (stem) and mucus-secreting cells (mucus) spontaneously differentiated from hGSCs. GINS organoids contained four endocrine cell types. Cells are colored according to cell types; bottom UMAP, relative expression of cell type-specific markers. b, Relative expression of endocrine cell type-specific markers. The shading displays scaled average gene expression, and diameter denotes fractional expression. c, Comparison of GINS β-like cells (n = 2 independent batch of organoids, one representative batch shown, donor #6) and islet β-cells (n = 4 independent donors, an integration of all samples) in expression profiles of key genes for β-cell function, identity, metabolism, and exocytosis. MODY: Maturity Onset Diabetes of the Young. d, Relative expression of disallowed genes in the indicated cell types (n = 1, donor #6). e, Violin plots showing the expression levels of proliferative markers in the indicated cell types (n = 1, donor #6).

Extended Data Fig. 5 |. scRNA-seq comparison of GINS organoids derived from human antrum vs corpus stomach.

a, Diagram of human stomach. b, Immunofluorescence of antral GINS organoid (donor#6) stained for CPPT and MAFA. Repeated independently 5 times with similar results. c, Comparison of corpus and antral GINS organoids (both from donor#6) in the expression of β-cell marker genes. n=3 independent experiments. Data presented as mean ± s.d.; one-way ANOVA with Dunnett multiple comparisons test comparing GINS with islets. d, t-distributed stochastic neighbor embedding (t-SNE) plots of integrated corpus and antral GINS organoids. Cells are colored according to cell types. G-like: G-like cells that expressed gastrin (GAST). Pie charts indicate cell-type proportions. e, Comparison of antral and corpus GINS β-like cells expression profiles of key genes for β-cell function and identity. Red, antral GINS β-like cells; Blue, corpus GINS β-like cells. f, Glucose-stimulated insulin secretion of antral GINS organoids at different time points (days post differentiation) or from different batches. n=4 independent groups of GINS organoids for the time course GSIS. n=5 independent groups of GINS organoids for batch 1, n=4 independent groups of GINS organoids for batch 2 and 3. Two-way repeated-measures ANOVA with Holm–Sidak’s multiple comparisons test.

Extended Data Fig. 6 |. GINS organoids are not fully mature.

a, Volcano plot comparing gene expression of GINS β-like cells (n = 2 independent batch of organoids, one representative batch is shown, donor #6) versus islet β cells (n = 4 independent donors, an integration of all samples) identified in Fig. 3a. The number of differentially expressed genes (DEGs) enriched in either cell group is shown in the plot. Threshold of DEGs: adjusted-P < 0.01 and log₂ fold-change > 1. P-value calculated by Wilcoxon Rank Sum test and adjusted based on Bonferroni correction. b, Gene Ontology (GO) analysis of DEGs enriched in GINS β-like cells (blue) or islet β-cells (red). P-value calculated by hypergeometric distribution followed by Benjamini-Hochberg adjustment.

Extended Data Fig. 7 |. Phenotypic characterization of GINS grafts.

a, Quantification from immunofluorescent staining of marker proteins, n = 3 independent experiments. Representative image showing co-expression of INS and CPPT in the GINS graft. b, Electron microscopy imaging of GINS graft. The electron-dense core granules were partially condensed. Repeated independently 3 times with similar results. c, SLC30A8 relative expression levels in GINS organoids, GINS grafts, and human islets. n=3 independent groups of GINS organoids, independent GINS grafts from different mice, or independent human islets from different donors. Data presented as mean ± s.d.; one-way ANOVA with Dunnett multiple comparisons test comparing GINS with islets. d, Images of the kidney from mice transplanted with GINS cells on day 0 and day 110 post transplantation. e, mCherry-labeled hGSCs (0.5 x 10⁶) transplanted under the renal capsule and visualized under fluorescent microscope on day 0 and day 80 post transplantation. No Cherry⁺ cells were found at day 80. Repeated independently 5 times with similar results. f, tSNE projection of integrated GINS organoids and grafts. Cells are colored according to cell types. Horizontal bars indicate cell type ratios. g, Violin plots showing the expression levels of select ribonucleoproteins. h, Relative expression of select genes in the pathways elevated in cultured GINS β-like cells compared with human islet-β cells.

Extended Data Fig. 8 |. Dynamic gene and signaling pathway activations in hGSC differentiation to GINS organoid.

a, Relative expression levels of cell type-specific markers in UMAP. b, Expression of select genes shown along pseudotime in GINS organoid formation. Each dot represents a cell. a, b, n = 1, donor #6. c, Heat map showing transcription factor expression clusters along the pseudotime trajectory from hGSCs to GINS organoid cells (n = 1, donor #6) and islet β cells (n = 1, donor #1). Density plot on the top showing cell populations along pseudotime. Stem: hGSCs; Endo 1: endocrine progenitors 1; Endo 2: endocrine progenitors 2; GINS pre: GINS precursors.

Extended Data Fig. 9 |. Characterization of the developmental path of GINS organoids.

a, Heatmap showing waves of transcription factor regulon activations. Key regulons are labeled on the right with the number of their predicted target genes. Stem: hGSCs; Endo 1: endocrine progenitors 1; Endo 2: endocrine progenitors 2; GINS pre: GINS precursors. b, Select regulon activity overlaid on UMAP. c, RNA velocity and pseudotime trajectory analysis in UMAP showing the developmental path from GINS precursors to endocrine cells in GINS organoids. a, b, n = 1, donor #6.

Fig. 1 |. Generation of Gastric Insulin-Secreting (GINS) organoids from human stomach samples.

a, Diagram showing key steps of GINS organoid generation. Medium supplements for hGSC (human gastric stem cell) culture or GINS organoid induction are indicated, with up and down arrows indicating agonists or antagonists, respectively. b, Representative images of human stomach samples, hGSC colonies, and GINS organoids. 1000 cells per organoids. c, d, Immunofluorescent staining and quantification of day-21 GINS organoids for C-peptide (CPPT), MAFA, glucagon (GCG), somatostatin (SST), and ghrelin (GHRL). For CPPT⁺ cell quantification, n = 33 independent organoids (donor #6), for other hormone quantification, n = 6 independent organoids (donor #6). e, Quantitative flow cytometry of day-21 GINS organoids (donor #10) with negative control. f, Insulin content of day-18 GINS organoids (n = 11 different batches of organoids from donor #6) and human islets (n = 3 samples from independent donors). Data are mean ± s.d.; two-tailed unpaired t-test.

Fig. 2 |. GINS organoids secret human insulin in response to glucose and GLP-1 analogue.

a, Glucose-stimulated insulin secretion of GINS organoids from donor #6 at different time points (days post differentiation, n = 4 independent groups of same batch organoids per time point), from different batches (n = 4 (batch1-3) or n = 3 (batch 4) independent groups of organoids per batch), or from donor #10 (n = 4 independent batches of organoids). For GINS organoids from donor #6, two-way repeated-measures ANOVA with Holm–Sidak’s multiple comparisons test. For GINS organoids from donor #10 and islets, one-tailed paired t-test. b, Insulin secretion of day-18 GINS organoids in response to sequential glucose challenges and KCl depolarization (n = 5 groups of organoids from donor #6, from two independent experiments). One-way repeated measures ANOVA with Holm–Sidak’s multiple comparisons test. c, Dynamic glucose-stimulated insulin secretion by perifusion assay of day-21 GINS organoids derived from donor #6 and #10 (n = 4 independent groups of organoids for each donor). a, b, c, Data presented as mean ± s.d.

Fig. 3 |. Four endocrine cell types identified in GINS organoids by scRNA-seq.

a, UMAP visualization of integrated GINS organoid (day 21, n = 1, donor #6) and primary human islet single-cell transcriptomes (n = 4 independent donors). b, Relative expression of endocrine cell type-specific markers. The shading displays scaled average gene expression, and diameter denotes fractional expression. c, Violin plots of key β-cell function and identity markers in GINS β-like and islet β cells. d, Cell identity scoring with gastric and β-cell signatures (868 and 1,034 genes, respectively). Mucus: mucus-secreting cells spontaneously differentiated from hGSCs in culture; Stem: hGSCs; β-like: GINS β-like cells; islet-β: islet β-cells.

Fig. 4 |. Transplanted GINS organoids secrete human insulin and reversed diabetes in mice.

a, b, GINS cells (0.8 x 10⁶, corpus donor #6) were transplanted under the renal capsule of non-diabetic NSG mice, which yielded grafts containing CD31⁺ vascular cells and predominant INS⁺ cells that co-expressed PAX6, MAFA, NKX2-2, PCSK1, and the mature β cell-associated marker ENTPD3 (6 months post transplantation). A small number of SST⁺, GCG⁺, and GHRL⁺ were also found. Repeated independently 3 times with similar results. (a). Human insulin from individual mouse was measured at 7 weeks (cohort-1) and 11 weeks (cohort-2) post transplantation, grafted with different batches of cells, after overnight fasting (blue bar) and 60 minutes after glucose challenge (red bar) (b). c, d, STZ-induced diabetic NSG mice were transplanted with 6–8 million GINS cells from donor #6 (GINS#6, n = 5 independent batches of organoids transplanted into 5 individual mice) or 6 million GINS cells from donor #10 (GINS#10, n = 3 independent batches of organoids transplanted into 3 independent mice) or without transplantation (Sham). Random-fed blood glucose levels (c) and intraperitoneal glucose tolerance tests (IPGTT, 2 weeks post transplantation) (d) showed significant improvement for both transplanted groups. Data presented as mean ± s.e.m.; two-way repeated-measures ANOVA. Tx: transplantation. Nx: nephrectomy. e, 24-hour pulse-chase with EdU revealed no proliferative cells in GINS graft but abundant replicating cells in the intestine of the same engrafted mouse. Repeated independently 3 times with similar results. f, Violin plots of select genes in GINS β-like cells before and after transplantation and comparison with islet β cells. g, Assessment of transcriptomic heterogeneity among GINS β-like cells before and after transplantation and among primary human islet β cells with correlation coefficient analysis across 2,000 top variable genes.

Fig. 5 |. Developmental trajectory of GINS cells.

a, Experimental design of sampling cells with scRNA-seq at key stages of hGSC differentiation to GINS organoids. UMAP showing clustering of hGSCs (day 0, yellow, n = 1, donor #6), endocrine progenitors (day 2, green, n = 1, donor #6), GINS precursors (day 6, blue, n = 1, donor #6), GINS organoids (day 20, red, n = 1, donor #6) and primary human islets (gray, n = 1, donor #1). b, UMAP of cell types colored according to their identities. Stem: hGSCs; Endo 1: endocrine progenitors 1; Endo 2: endocrine progenitors 2; GINS pre: GINS precursors. c, Relative expression of select markers across cell types. d, Pseudotime trajectory analysis of cell types. e, Violin plot showing the expression levels of the Pdx1-Mafa transgenes in different cell types. f, Heat map showing gene-expression clusters along the pseudotime trajectory from hGSCs to GINS cells and human islet β cells. Select significant GO terms enriched in each gene cluster are shown. P-value calculated by hypergeometric distribution followed by Benjamini-Hochberg adjustment.

Fig. 6 |. Galanin⁺ precursors give rise to GINS cells.

a, Relative expression of INS and GAL in GINS precursors, GINS β-like cells and human islet β cells. UMAP subset of Fig. 5b. b, Purification of GFP^high GINS precursors differentiated from GAL-GFP hGSC reporter line at day 7 post differentiation. c, d, Quantification of GAL⁺ and CPPT ⁺ cell percentage in organoids and their average staining intensity at day 1 or 14 post sorting. Data presented as mean ± s.d. (GAL⁺ at day 1: n = 9; GAL⁺ at day 14: n = 10; CPPT⁺ at day 1: n = 13; CPPT⁺ at day 14: n = 8; all independent organoids from donor #6); two-way ANOVA with Sidak’s multiple comparisons test. e, Representative images of immunofluorescent staining for GAL and CPPT at day 1 and day 14 post sorting. f, Representative immunofluorescence of GAL and INS in primary human islet grafts in NSG mice. Arrows indicate GAL⁺INS⁺ cells. g, Violin plot showing scRNA levels of GAL in GINS β-like cells before and after transplantation and in human islet β cells. h, Proposed model for GINS organoid formation, with rerouting of the hGSC developmental trajectory by NGN3, PDX1 and MAFA.