Regulated and adaptive in vivo insulin secretion from islets only containing β-cells

Abstract

Insulin-producing β-cells in pancreatic islets are regulated by systemic cues and, locally, by adjacent islet hormone-producing 'non-β-cells' (namely α-cells, δ-cells and γ-cells). Yet whether the non-β-cells are required for accurate insulin secretion is unclear. Here, we studied mice in which adult islets are exclusively composed of β-cells and human pseudoislets containing only primary β-cells. Mice lacking non-β-cells had optimal blood glucose regulation, enhanced glucose tolerance, insulin sensitivity and restricted body weight gain under a high-fat diet. The insulin secretion dynamics in islets composed of only β-cells was comparable to that in intact islets. Similarly, human β-cell pseudoislets retained the glucose-regulated mitochondrial respiration, insulin secretion and exendin-4 responses of entire islets. The findings indicate that non-β-cells are dispensable for blood glucose homeostasis and β-cell function. These results support efforts aimed at developing diabetes treatments by generating β-like clusters devoid of non-β-cells, such as from pluripotent stem cells differentiated in vitro or by reprograming non-β-cells into insulin producers in situ.

Fig. 1. Generation and characterization of transgenic mice for the simultaneous ablation of islet α-cells, δ-cells and γ-cells.

a, CRISPR–Cas9 strategy to replace the Sst coding sequence (Sst CDS) on mouse chromosome 16 with the human DTR coding region (targeting vector). LHA, left homology arm; RHA, right homology arm; UTR, untranslated region. b, Transgenes required for DTR-mediated non-β-cell ablation. c, PCR products of Gcg-DTR (~800 bp), Sst-DTR (WT 871 bp, KI 483 bp) and Ppy-DTR (WT 389 bp, KI 223 bp) transgenes. WT, wild type; HTZ, heterozygous. ‘No DNA’ is a negative control. d, Immunofluorescence on pancreatic sections from control (Ctrl) or β-only mice 4 wpa. INS, insulin (red); GCG, green (left); SST, green (middle); PPY, green (right). Scale bars, 20 µm. e, Quantification of GCG⁺, SST⁺ and PPY⁺ cells in Ctrl (n = 4) or DT (n = 5) mice. GCG⁺ and SST⁺ cells were scored in the dorsal pancreas; PPY⁺ cells were scored in the ventral pancreas. Statistical tests: two-way ANOVA. P values for Ctrl vs DT: GCG⁺, ****P < 0.0001; SST⁺, **P = 0.0012; PPY⁺, ****P < 0.0001. f, qPCR of Gcg, Sst and Ppy on isolated islets of Ctrl (n = 10) and β-only mice at 1 wpa (n = 4), 2 wpa (n = 8) and 4 wpa (n = 7). Data are normalized to housekeeping genes (Gapdh and β-actin) and shown as relative hormone expression to Ctrl. P values for Gcg: 1 wpa vs 0 wpa, ***P = 0.0001; 2 wpa vs 0 wpa, ***P = 0.0001; 4 wpa vs 0 wpa, ****P = 0.00005; Sst: 1 wpa vs 0 wpa, ***P = 0.0001; 2 wpa vs 0 wpa, **P = 0.007; 4 wpa vs 0 wpa, ****P = 0.00005; Ppy: 1 wpa vs 0 wpa, P = 0.056; 2 wpa vs 0 wpa, ***P = 0.0001; 4 wpa vs 0 wpa, ****P = 0.00005. NS, not significant. g, Pancreatic insulin content of Ctrl (n = 10) and DT (n = 10) at 4 wpa. P value Ctrl vs DT, P = 0.63. All data are shown as mean ± s.e.m. Male and female mice were used for these experiments. Unless otherwise indicated, P values are from two-tailed Mann–Whitney tests. NS, not significant. Source data

Fig. 2. Optimal blood glucose homeostasis in mice whose islets comprise β-cells only.

a, Glycaemia over 40 wpa in Ctrl (n = 6) and DT (n = 6) mice under random-fed (RF) or fasting (F) conditions. Right panel: area under the curve (AUC). P values (DT vs Ctrl): RF, P = 0.73; F, P = 0.3939. b, Intraperitoneal glucose tolerance test (ipGTT) on Ctrl (n = 33) and DT (n = 34). Right panel: AUC P value (Ctrl vs DT), P = 0.3. c, Plasma insulin 0 min and 15 min after glucose injection along ipGTT in Ctrl (n = 25) and DT (n = 25) mice. Right panel: stimulation index of insulin secretion. P value (Ctrl vs DT), **P = 0.0074. d, Oral glucose tolerance test (oGTT) on Ctrl (n = 7) and DT (n = 11). Right panel: AUC P value (Ctrl vs DT), **P = 0.0028. e, Plasma GLP-1 along oGTT in Ctrl (n = 5) and DT (n = 7) mice. P value (Ctrl vs DT), *P = 0.0303. f, Plasma insulin 0 min and 10 min after glucose gavage along oGTT in Ctrl (n = 8) and DT (n = 11) mice. Right panel: stimulation index; P value (Ctrl vs DT), ***P = 0.0006. g, Insulin tolerance test (ITT) on Ctrl (n = 6) and DT (n = 6) mice. Data as % relative to glycaemia at time 0. Right panel: AUC of ITT; P value Ctrl vs DT, *P = 0.03. h, Radiolabelled 2-deoxy-d-glucose (2DG) uptake by WAT in Ctrl (n = 6) and DT (n = 6) mice. P values (Ctrl vs DT), *P = 0.039. i, Energy expenditure (EE) in Ctrl (n = 9) and DT (n = 7) mice. Right panel: AUC of EE, P value: light period Ctrl vs DT, **P = 0.0012; Ctrl dark vs light period, P < 0.0001; DT dark vs light period, **P = 0.0047. j, Respiratory exchange ratio (RER) in Ctrl (black, n = 9) and DT (red, n = 7) mice. Right panel: AUC of RER. P value: light period Ctrl vs DT, *P = 0.035; Ctrl dark vs light period, P < 0.0003; DT dark vs light period, P = 0.1212. Only male mice were used. All data are shown as mean ± s.e.m.; P values are from two-tailed Mann–Whitney tests. Source data

Fig. 3. The ablation of pancreatic α-cells has no effect on GCG-mediated hepatic function.

a, Glycaemia of control (Ctrl) and DT β-only mice after 5 h (Ctrl, n = 6; DT, n = 6) or 16 h (Ctrl, n = 7; DT, n = 6) fasting. P values (DT vs Ctrl): 5 h fasting, P = 0.093; 16 h fasting, P = 0.7587. b, Pancreatic GCG levels of Ctrl (n = 7) and DT (n = 6) β-only mice after 16 h fasting. P values: DT vs Ctrl, **P = 0.0012. c, Plasma GCG levels of Ctrl and DT β-only mice after 5 h (Ctrl, n = 8; DT, n = 8) or 16 h (Ctrl, n = 9; DT, n = 8) fasting. P values: 5 h fasting DT vs Ctrl, **P = 0.0011; 16 h fasting DT vs Ctrl, P = 0.1447. NS, not significant. d–f, Plasma cortisol (Ctrl n = 7; DT n = 6) (d), growth hormone (Ctrl n = 7; DT n = 4) (e) and epinephrine (Ctrl n = 8; DT n = 6) (f) levels of Ctrl and DT after 16 h fasting. P values DT vs Ctrl: cortisol, P = 0.0688; growth hormone, P = 0.7879; epinephrine, P = 0.1074. g, Pyruvate tolerance test (PTT) after 16 h fasting in Ctrl (n = 13) and DT (n = 16) mice. Right panel: AUC of PTT. P values: DT vs Ctrl, P = 0.7136. h–j, Hepatic glycogen (Ctrl n = 12; DT n = 8) (h), pyruvate (Ctrl n = 15; DT n = 8) (i) and lactate (Ctrl n = 14; DT n = 8) (j) levels after 16 h fasting in Ctrl and DT mice. P values DT vs Ctrl: glycogen, P = 0.69; pyruvate, P = 0.579; lactate, P = 0.713. k, Expression of hepatic gluconeogenic enzymes after 16 h fasting in Ctrl (black, n = 14) and DT (red, n = 9) mice. Phosphoenolpyruvate carboxykinase (Pepck), pyruvate carboxylase (Pcx), glucose-6-phosphatase (G6pase) and glucose transporter 1 (Glut1). P values DT vs Ctrl: Pepck, P = 0.2303; Pcx, P = 0.095; G6pase, P = 0.8775; Glut1, P = 0.0534. Only male mice were used. All data are shown as mean ± s.e.m.; P  values are from two-tailed Mann–Whitney tests. Source data

Fig. 4. Restricted body weight gain and better glucose homeostasis in β-only mice under HFD.

a, Experimental design; wdc, weeks of diet challenge. b, Percentage of body weight change in Ctrl + chow (n = 10), Ctrl + HFD (n = 10), DT + chow (n = 8) and DT + HFD (n = 11). Right panel: AUC of body weight change. P values: Ctrl + chow vs Ctrl + HFD, ****P < 0.0001; Ctrl + chow vs DT + chow, P = 0.44; DT + chow vs DT + HFD, *P = 0.03; Ctrl + HFD vs DT + HFD, **P = 0.007. c, Body composition (fat and lean mass, as % of body weight (BW)) in Ctrl + chow (gray, n = 10), Ctrl + HFD (black, n = 10), DT + chow (pink, n = 8) and DT + HFD (red, n = 11). P values: fat mass, Ctrl + chow vs Ctrl + HFD, ****P < 0.0001; Ctrl + HFD vs DT + HFD, P = 0.0003; lean mass, Ctrl + chow vs Ctrl + HFD, ****P < 0.0001; Ctrl + HFD vs DT + HFD, P < 0.0001. d, Fasting glycaemia of Ctrl + chow (gray, n = 10), Ctrl + HFD (black, n = 10), DT + chow (pink, n = 8) and DT + HFD (red, n = 11). Right panel: AUC of fasting glycaemia. P values: Ctrl + chow vs Ctrl + HFD, **P = 0.0089; Ctrl + HFD vs DT + HFD, *P = 0.0207. e, oGTT at 24 wdc of Ctrl + chow (gray, n = 10), Ctrl + HFD (black, n = 10), DT + chow (pink, n = 8), DT + HFD (red, n = 11) mice. Right panel: AUC of oGTT. P values: Ctrl + chow vs Ctrl + HFD, **P = 0.0015; DT + chow vs DT + HFD, *P = 0.0068; Ctrl + HFD vs DT + HFD, **P = 0.0028. f, Plasma insulin along oGTT. Left panel, plasma insulin 0 and 10 min after glucose injection in Ctrl + chow (n = 10), Ctrl + HFD (n = 8), DT + chow (n = 6), DT + HFD (n = 9). Right panel: stimulation index of insulin release. P values: Ctrl + chow (0 vs 10 min), P = 0.0052; Ctrl + HFD (0 vs 10 min), P = 0.0006; DT + chow (0 vs 10 min), P = 0.0043; DT + HFD (0 vs 10 min), P = 0.004. Ctrl + HFD vs DT + HFD at 0 min, **P = 0.0056; Ctrl + HFD vs DT + HFD at 10 min, **P = 0.0025. g, ITT at 24 wdc. Left panel, ITT as % of change relative to time 0 in Ctrl + chow (gray, n = 9), Ctrl + HFD (black, n = 10), DT + chow (pink, n = 6) and DT + HFD (red, n = 10). Right panel: AUC of ITT. P values: Ctrl + chow vs DT + chow, *P = 0.0496; Ctrl + HFD vs DT + HFD, ***P = 0.0005. Body weight changes, body composition, fasting glycaemia, oGTT, plasma insulin levels and ITT were obtained from the same mice. Only male mice were used. All data are shown as mean ± s.e.m.; P values are from two-tailed Mann–Whitney tests. Source data

Fig. 5. Monotypic β-only islets have normal dynamic insulin secretion and islet respirometry.

a, Representative picture of a β-only mouse pancreas slice. b, Dynamic insulin secretion profiling on pancreatic slices from Ctrl (n = 5) and DT (n = 3) mice. Gray background area indicates the first and second phase of insulin release. c, Representative picture of isolated islets of β-only mice. d, Dynamic glucose-stimulated insulin secretion profiling on isolated islets. Absolute insulin secretion from Ctrl (n = 9) and DT (n = 8) mice. Gray background area represents the first and second phase of insulin release. e, Stimulation index (as fold change to baseline values) of in vitro glucose-stimulated insulin secretion in isolated islets of Ctrl (black, n = 7) and DT (red, n = 9) mice in the presence or absence of the GLP-1 agonist Ex-4. Low glucose is 3 mM; high glucose is 16.7 mM. P values: Ctrl with vs without Ex-4, *P = 0.048; DT with vs without Ex-4, *P = 0.018. f, Islet respirometry by Seahorse XF96 bioanalyser. Figure shows representative OCR traces of islets from Ctrl (n = 8 mice) and DT (n = 7 mice) sequentially exposed to 16.7 mM glucose (G16.7), 2.5 µM oligomycin (Oligo), 2 µM FCCP and 3 µM Rot/AA at the indicated time points. Basal measurement was performed at 3 mM glucose before injections. OCR was normalized by insulin content of each well and represented as % of basal. Two to four wells per mouse, eight to ten islets per well. Two-way ANOVA with Sidak’s multiple comparisons test. Only male mice were used. All data are shown as mean ± s.e.m. Unless otherwise indicated, P values are from two-tailed Mann–Whitney tests. Source data

Fig. 6. Monotypic β-only islets have a normal response to extreme insulin demand.

a, Glycaemia before and 2 weeks after S961 treatment on Ctrl (n = 10), Ctrl + S961 (n = 9), DT (n = 9) and DT + S961 (n = 10). Right panel: AUC of glycaemia. P values: Ctrl vs DT, ***P = 0.001; Ctrl vs Ctrl + S961, ****P < 0.0001; DT vs DT + S961, ****P < 0.0001; Ctrl + S961 vs DT + S961, P = 0.1903. b, Plasma insulin release under random-fed conditions in Ctrl (gray, n = 10), Ctrl + S961 (black, n = 9), DT (pink, n = 9) and DT + S961 (red, n = 9). P values: Ctrl vs DT, **P = 0.0015; Ctrl vs Ctrl + S961, ****P < 0.0001; DT vs DT + S961, ***P = 0.0002; Ctrl + S961 vs DT + S961, P = 0.9182. c, Immunofluorescence on pancreatic sections from Ctrl or DT-treated β-only mice with or without S961 treatment. INS, green; Ki67, gray; DAPI, blue. Scale bars, 20 µm. d, Percentage of double-positive INS⁺KI67⁺ in INS⁺ population in Ctrl (n = 4), Ctrl + S961 (n = 4), DT (n = 4) and DT + S961 (n = 4). P values: Ctrl vs DT, P = 0.6857; Ctrl vs Ctrl + S961, *P = 0.286; DT vs DT + S961, *P = 0.286; Ctrl + S961 vs DT + S961, *P = 0.286. Only male mice were used. All data are shown as mean ± s.e.m. P values are from two-tailed Mann–Whitney tests. Source data

Fig. 7. Human monotypic β-pseudoislets maintain glucose-stimulated insulin secretion and islet respirometry.

a, Generation of human pseudoislets. Islets from the same donor were dissociated and sorted by FACS to obtain highly pure β-cells that were aggregated into pseudoislets. Control pseudoislets comprising all cell types were dissociated and aggregated directly. b, qPCR of INS, GCG, SST and PPY on control and β-pseudoislets. Data are normalized to housekeeping genes (RN18S); n = 4 donors. P values (Ctrl vs β): INS, P = 0.200; GCG, *P = 0.0286; SST, *P = 0.0286; PPY, P = 0.1714. NS, not significant. c, Immunofluorescence on control pseudoislets (Ctrl) or monotypic β-pseudoislets (β). INS, red; GCG, blue; SST, green (left); INS, red; GCG, blue; PPY, green (right). Scale bar, 20 µm. d, In vitro insulin secretion normalized by total insulin content in Ctrl and monotypic β-pseudoislets assessed at basal (3 mM, low glucose (LG)) or stimulatory (16.7 mM, high glucose (HG)) glucose concentrations in the presence or absence of the GLP-1 agonist Ex-4. n = 4 donors. P values (Ctrl vs β): LG, P = 0.1831; HG, P > 0.999; HG + Ex-4, ***P = 0.0002. NS, not significant. e, Islet respirometry by Seahorse XF96 bioanalyser. OCR traces of human Ctrl or β-only pseudoislets. Glucose (16.7 mM; G16.7), oligomycin (2.5 µM; Oligo), FCCP (2 µM) and Rot/AA (3 µM) were injected at the indicated time points. Basal measurement was performed at 3 mM glucose before injections. OCR was normalized by insulin content of each well and represented as % of basal. n = 3 independent donors. Two to five wells per condition per donor, ten pseudoislets per well; Two-way ANOVA with Sidak’s multiple comparisons test. Both male and female human donors were used. All data are shown as mean ± s.e.m. Unless otherwise indicated, P values are from two-tailed Mann–Whitney tests. Source data

Extended Data Fig. 1. Long-term characterization of aged mice 40-weeks after non-β-cell ablation.

a. Immunofluorescence on pancreas tissue sections from control (Ctrl) or DT-treated mice analyzed 40 weeks post-ablation (wpa). INS: red; GCG: green (left); INS: red; SST: green (middle); INS: red; PPY: green (right). Scale bar: 20 µm. b. Quantification of GCG+, SST+, and PPY+ cells in Ctrl (black) or DT (red) mice 40wpa. GCG+ and SST+ (Ctrl n = 7, DT n = 5) cells were scored in the dorsal region of the pancreas; PPY+ (Ctrl n = 3, DT n = 5) cells were scored in the ventral region of the pancreas. P-values Ctrl vs DT: GCG + P = 0.0025, SST + P = 0.0025, PPY + P = 0.036. c. Pancreatic insulin content (ng/mg) of Ctrl (black, n = 6) and DT (red, n = 6) mice 40wpa. P-value Ctrl vs DT = 0.937. Only male mice were used. All data are shown as mean ± sem. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 2. Somatostatin-producing intestinal D-cell ablation and regeneration in β-only mice.

a, Body weight over 40weeks on Ctrl (black, n = 6) and DT (red, n = 6) mice in random fed (RF, continuous line) and fasting (F, dashed line). Right panel: Area under the curve (AUC). P-values (DTvsCtrl): RF P = 0.3312, F P = 0.3939. b, Body composition in Ctrl (black, n = 8) and DT (red, n = 8) mice. P-values CtrlvsDT: Lean mass P = 0.7209, Fat mass P = 0.083. c, Immunofluorescence on SST (red) in stomach (upper micrographs) and GLP-1 (red) in colon (lower micrographs) from Ctrl or β-only mice. Scale bar: 20 µm or 10 µm (insets). d, Duodenum qPCR of Sst of Ctrl (black, n = 4) and β-only mice at 7 dpa (pink, n = 7), 15 dpa (light red, n = 7), and 30 dpa (red, n = 6). Data are normalized ct values relative to β-actin and Gapdh and shown as relative hormone expression to Ctrl. P-values: 7 dpa vs Ctrl P = 0.0061, 15 dpa vs Ctrl P = 0.0606, 30 dpa vs Ctrl P = > 0.999. e, Plasma somatostatin in Ctrl (n = 16) and β-only after 15 dpa (pink, n = 8) and 30 dpa (red, n = 8). P-values: 15 dpa vs Ctrl P = 0.0016, 30 dpa vs Ctrl P = 0.1090, 30 dpa vs 15 dpa P = 0.0468. f, Brain qPCR on Sst of Ctrl (black, n = 4) and β-only (red, n = 6). Data are normalized ct values relative to β-actin and Gapdh and shown as relative hormone expression to Ctrl. P-values Ctrl vs DT: P = 0.0095. g, Brain somatostatin of Ctrl (n = 8) and β-only (n = 7). P-values: DT vs Ctrl P = 0.0128. h-j, Food intake (h; gr/gr), total movement (i; counts, infrared breaks in 15 minutes) and running wheel activity (j) in Ctrl (black, n = 9) and DT (red, n = 7). Right panel: AUC. Food intake (h) P-value (CtrlvsDT): light period P = 0.2523; dark period P = 0.6065; 24 h P = 0.4079. Total movement (i) P-value (CtrlvsDT): light period P = 0.2105; dark period P = 0.8371; 24 h period P = 0.7577. Running wheel (j) P-value (CtrlvsDT): light period P = > 0.99; dark period P = 0.9182; 24 h period P = > 0.99. Only male mice were used. All data are shown as mean ± sem. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 3. Improved blood glucose homeostasis in female β-only mice.

a. Glycemia over 40 weeks post-ablation of Ctrl (black; n = 9) and DT (red; n = 10) female mice in random-fed (RF, continuous line) or fasting (F, dashed line). Right panel: area under the curve (AUC) of glycemia. P-values (DT vs Ctrl): RF P = 0.0038; F P = 0.067. b. Body weight over 40 weeks post-ablation of Ctrl (black; n = 9) and DT (red; n = 10) female mice in RF (continuous line) or F (dashed line). Right panel: AUC of body weight. P-values DT vs Ctrl: RF P = 0.468; F P = 0.2013. c. Intraperitoneal glucose tolerance test (ipGTT) at 4wpa. Left panel: ipGTT on Ctrl (black; n = 17) and DT (red; n = 18). Right panel: AUC of ipGTT; P-value (Ctrl vs DT) P = 0.3142. d. Plasma insulin along ipGTT. Left panel: plasma insulin at 0 and 15 minutes after glucose injection in Ctrl (black; n = 12) and DT (red; n = 7) mice. Right panel: stimulation index (S.I.) of insulin secretion. P-value (Ctrl vs DT) P = 0.0646. e. Oral glucose tolerance test (oGTT) at 30 dpa. Left panel: oGTT on Ctrl (black; n = 10) and DT (red; n = 7). Right panel: AUC of oGTT. P-value (Ctrl vs DT) = 0.0431. f. Plasma insulin along oGTT. Left panel: plasma insulin at 0 and 10 minutes after glucose gavage in Ctrl (black; n = 10) and DT (red; n = 6) mice. Right panel: stimulation index (S.I.) of insulin secretion. P-value Ctrl vs DT P = 0.2424. g. Insulin tolerance test (ITT) at 30 dpa. Left panel: ITT on Ctrl (black; n = 8) and DT (red; n = 7) mice. Data are shown as % relative to glycemia at time 0. Right panel: AUC of ITT. P-value Ctrl vs DT P = 0.0093. Error bars denote s.e.m. Two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 4. Blood glucose homeostasis in mice without somatostatin-expressing δ-cells.

a, Pancreas immunofluorescence from control (Ctrl) or δ-cell ablated mice. INS: red; SST: green. Scale bars: 20 µm. b. Quantification of SST+ cells per islet section in Ctrl (black, n = 3) or DT (red, n = 3) mice (dorsal pancreas). P-values CtrlvsDT P = 0.0007. c, Plasma somatostatin (ng/ml) in Ctrl (black, n = 11) and DT (red, n = 8). P-values CtrlvsDT: P = 0.0157. d, Glycemia of Ctrl (black, n = 6) and DT (red, n = 6) in RF (continuous line) or F (dashed line). Right panel: area under the curve (AUC). P-values: RF DTvsCtrl P = 0.0043; F DTvsCtrl P = 0.8182. e, Body weight change (% from day 0) of Ctrl (black, n = 6) and DT (red, n = 6) in random-fed. Right panel: AUC. P-values: RF DTvsCtrl P = 0.0043. f, Intraperitoneal glucose tolerance test (ipGTT) on Ctrl (black, n = 9) and DT (red, n = 10). Right panel: AUC; P-value CtrlvsDT P = 0.0279. g, Plasma insulin at 0 and 15 minutes after glucose in Ctrl (black, n = 8) and DT (red, n = 10). P-value Ctrl (0’ vs 15’) = 0.0140; DT (0’ vs 15’) = 0.0015. h, Insulin tolerance test (ITT) on Ctrl (black, n = 9) and DT (red, n = 10). Data are shown as % relative to glycemia at time 0. Right panel: AUC; P-value CtrlvsDT=0.0789. i, Oral glucose tolerance test (oGTT) on Ctrl (black, n = 8) and DT (red, n = 10). Right panel: AUC; P-value (CtrlvsDT) = 0.0055. j, Plasma insulin at 0 and 10 minutes after glucose in Ctrl (black, n = 7) and DT (red, n = 10). P-value Ctrl (0’ vs 10’) = 0.0041; DT (0’ vs 10’) = 0.0011. Right panel: stimulation index (S.I.) of insulin; P-value (CtrlvsDT)=0.9623. k, Plasma GLP-1 along oGTT in Ctrl (black, n = 9) and DT (red, n = 9). P-value (CtrlvsDT)=0.0056. Only male mice were used. Data are shown as mean ± sem. Statistical tests are two-tailed Mann-Whitney. Source data is available. Source data

Extended Data Fig. 5. Hyperinsulinemic-euglycemic clamp after non β-cell ablation.

a, Experimental time-line for the hyperinsulinemic-euglycemic clamp. 2.5mUI/kg/min of insulin were continuously infused during the insulin clamp. Blood samples were taken at multiple time for further analyses. b-c, Glucose infusion rate (GIR, b) and hepatic glucose production (HGP, c) measured during clamp studies in Ctrl (black, n = 7) and DT (red, n = 6) mice. P-values: CtrlvsDT HGP P = 0.007. d-f, Radiolabeled 2-deoxy-D-glucose (2DG) uptake by adipose tissue in Ctrl (black, n = 7) and DT (red, n = 6) mice. P-values (Ctrl vs DT): Brown Adipose Tissue (BAT) P = 0.073; subcutaneous White Adipose Tissue (sWAT) P = 0.008; inguinal White Adipose Tissue (iWAT) P = 0.005. Insulin clamp and 2DG uptake were obtained from the same mice. Only male mice were used. All data are shown as mean ± sem. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 6. High-fat diet administration to β-only mice.

a, Food intake (gr) of Ctrl+Chow (grey, n = 6), Ctrl+HFD (black, n = 5), DT+Chow (pink, n = 4), DT + HFD (red, n = 6) over the last 4-weeks of high-fat diet (HFD). Right panel: area under the curve (AUC). P-values: Ctrl+chow vs Ctrl+HFD P = 0.7922, Ctrl+chow vs DT+chow P = 0.4762, Ctrl+chow vs DT + HFD P = 0.0649, Ctrl+HFD vs DT+chow P = 0.9048, Ctrl+HFD vs DT + HFD P = 0.6623; DT+chow vs DT + HFD P = 0.4762. b, Random fed glycemia of Ctrl+Chow (grey, n = 10), Ctrl+HFD (black, n = 10), DT+Chow (pink, n = 8), DT + HFD (red, n = 11) over 24 weeks of HFD. Right panel: AUC. P-values: Ctrl+chow vs Ctrl+HFD P < 0.62, Ctrl+chow vs DT+chow P = 0.12, Ctrl+chow vs DT + HFD P = 0.13, Ctrl+HFD vs DT+chow P = 0.03, Ctrl+HFD vs DT + HFD P = 0.04; DT+chow vs DT + HFD P = 0.97. c, Intraperitoneal glucose tolerance test (ipGTT) 24-weeks after HFD of Ctrl+chow (grey, n = 10), Ctrl+HFD (black, n = 10), DT+chow (pink, n = 8), DT + HFD (red, n = 11). Right panel: AUC. P-values: Ctrl+chow vs Ctrl+HFD P = 0.0073, Ctrl+chow vs DT+chow P = 0.0.0038, Ctrl+chow vs DT + HFD P = 0.9404, Ctrl+HFD vs DT+chow P < 0.0001, Ctrl+HFD vs DT + HFD P = 0.0012; DT+chow vs DT + HFD P = 0.0129. d, Plasma insulin release at 0 and 15 minutes after glucose injection in Ctrl+chow (grey, n = 9), Ctrl+HFD (black, n = 10), DT+chow (pink, n = 8), DT + HFD (red, n = 11). P-values: Ctrl+Chow (0’vs15’) P = 0.0039; Ctrl+HFD (0’vs15’) P = 0.0355; DT+Chow (0’vs15’) P = 0.0281; DT + HFD (0’vs15’) P = 0.0192. Ctrl+HFD vs DT + HFD at 0’ P = 0.0357; Ctrl+HFD vs DT+Chow at 15’ P = 0.0044; Ctrl+HFD vs DT + HFD at 15’ P = 0.0101. Only male mice were used. All data are shown as mean ± sem. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 7. Improved glucose homeostasis after the specific ablation of the pancreatic α-, δ-, and γ-cells.

a. Immunofluorescence from control (Ctrl, left panel), mice injected with 120 ng of DT (middle panel) or 5 ng of DT (right panel). Top row: Brain somatostatin (green). Mid row: islet insulin (red), glucagon, somatostatin or Ppy (green). Bottom row: stomach somatostatin (green). Scale bars: 20 μm or 10 μm (insets). b. Brain qPCR on SST in Ctrl (black; n = 6) and 120ngDT (grey; n = 3) or 5ngDT (red; n = 5). Data are normalized ct values relative to β-actin and Gapdh. P-values: Ctrlvs5ngDT P = 0.6623, Ctrlvs120ngDT P = 0.0238, 5ngDTvs120ngDT P = 0.0357. c. Quantification of pancreatic GCG+, SST+, and PPY+ cells in Ctrl (black; n = 5), 120ngDT (grey; n = 4) or 5ngDT (red; n = 4). P-values: GCG+ Ctrlvs5ngDT P < 0.0001, Ctrlvs120ngDT P < 0.0001, 5ngDTvs120ngDT P = 0.5783. SST+ Ctrlvs5ngDT P = 0.0001, Ctrlvs120ngDT P < 0.0001, 5ngDTvs120ngDT P = 0.9852. PPY+ Ctrlvs5ngDT P < 0.0001, Ctrlvs120ngDT P < 0.0001, 5ngDTvs120ngDT P = 0.9964. d. Stomach qPCR on SST in Ctrl (black; n = 14) and 120ngDT (grey; n = 4) or 5ngDT (red; n = 4). Data are normalized ct values relative to β-actin and Gapdh. P-values: Ctrlvs5ngDT at 15 dpa P = 0.0013; Ctrlvs5ngDT at 30 dpa P = 0.7209; Ctrlvs120ngDT at 15 dpa P = 0.0006; Ctrlvs120ngDT at 30 dpa P = 0.0676; 5ngDT 15dpavs30dpa P = 0.0286; 120ngDT 15dpavs30dpa P = 0.0043. e-f. Glycemia (e) and body weight (f, % of change) of Ctrl (black; n = 12), 120ngDT (grey; n = 7) or 5ngDT (red; n = 8). Small panel: area under the curve (AUC). P-values: Glycemia Ctrlvs5ngDT P = 0.0017; Ctrlvs120ngDT P < 0.0001; 5ngDTvs120ngDT P = 0.003. Body weight Ctrlvs5ngDT P = 0.1196; Ctrlvs120ngDT P < 0.0001; 5ngDTvs120ngDT P = 0.003. g. Oral glucose tolerance test (oGTT) in Ctrl (black; n = 14), 120ngDT (grey; n = 10) or 5ngDT (red; n = 8). P-values: Ctrlvs5ngDT P = 0.0003; Ctrlvs120ngDT P < 0.0001; 5ngDTvs120ngDT P = 0.4215. h. Stimulation index (S.I.) of insulin in Ctrl (black; n = 12), 120ngDT (grey; n = 10) or 5ngDT (red; n = 8). P-values Ctrlvs5ngDT P = 0.0011; Ctrlvs120ngDT P = 0.0206; 5ngDTvs120ngDT P = 0.2844. i. Insulin tolerance test (ITT) in Ctrl (black; n = 14), 120ngDT (grey; n = 9) or 5ngDT (red; n = 8). Data are shown as % relative to glycemia at time 0. Right panel: AUC. P-values Ctrlvs5ngDT P = 0.0072; Ctrlvs120ngDT P = 0.0005; 5ngDTvs120ngDT P = 0.7979. j. Graphical summary. Only male mice were used. Error bars denote s.e.m. Two-way ANOVA and two-tailed Mann-Whitney test are used. Source data is available. Source data

Extended Data Fig. 8. Ex vivo dynamic insulin secretion in pancreatic slices and isolated islets.

a, Pancreatic insulin content (ng) of slices from control (Ctrl, black, n = 5) and DT (red, n = 3). P-value CtrlvsDT=0.250. b, Area under the curve (AUC) of the first phase (grey area, 14- and 24-minutes) and second phase (dashed grey area, 30- and 58-minutes). P-values (CtrlvsDT): first phase P = 0.393; second phase P = > 0.999. c-d, Islet equivalent count (IEQ, c) and ratio insulin content (ng) and IEQ count (d) of Ctrl (black, n = 9) and DT (red, n = 8). P-value: IEQ count (CtrlvsDT) P = 0.2359; ratio insulin content and IEQ (CtrlvsDT) P = 0.743. e, AUC of first phase (P-value CtrlvsDT P = 0.393) and second phase (P-value CtrlvsDT P = > 0.999) of insulin secretion in isolated islets of Ctrl (black; n = 9) and DT (red; n = 8) mice. f, Total insulin content (ng) of isolated islets from Ctrl (black, n = 7) and DT (red, n = 10) mice with or without Exendin-4 (Ex-4). P-value without Ex-4 CtrlvsDT P = 0.601; P-value with Ex-4 CtrlvsDT P = 0.3031. g-h. Islet respirometry parameters. g. Glucose-stimulated OCR (dashed lined column), proton leak and maximal respiration (grey column) in islets from Ctrl (black, n = 8) and DT (red, n = 7) mice. P-values (CtrlvsDT): Glucose-stimulated OCR P = 0.3285; Proton leak P = 0.2046; Maximal respiration P = 0.999. h. Basal respiration measured at 3 mM glucose from Ctrl (black, n = 8) and DT (red, n = 7). 2-4 wells/mice, 8-10 islets/well. P-values Ctrl vs DT P = 0.281. i. Total insulin content of islets from Ctrl (black, n = 8) and DT (red, n = 7) after OCR measurements. P-values CtrlvsDT P = 0.3357. Each data point (g-i) represents the average values per mouse. Only male mice were used. All data are shown as mean ± sem. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 9. Transcriptomic analysis of β-cells isolated from control and β-only mice.

a, Experimental design. b, Representative gating strategy used for FACS sorting the mCherry+ insulin-expressing cells in Ctrl and β-only mice for RNA-seq analysis. c, Principal component analysis (PCA) of RNA-seq samples. Each dot represents one sample. Four different conditions: Control Fasting (Ctrl F, green, n = 5), DT Fasting (DT F, red, n = 6), Control Refed (Ctrl RF, blue, n = 5) and DT Refed (DT RF, yellow, n = 6). d-e, MA plot of the DEGs (differentially expressed genes) between DT and Ctrl during 8-hour fasting (d) and after 2-hour refeeding (e). Each dot represents one gene. Red dots are the genes with p-values < 0.05 and a log fold change >0.5 or <−0.5. f, Heatmap showing scaled expression (blue, high expression; white, low expression) of representative β-ID genes. The dendrogram represents the clustering of each condition analyzed based on β-ID genes expression. g, Expression levels of key identity (top panel) and functional (bottom panel) β-cell genes during fasting in controls and DT-treated mice after 8-hour fasting or 2-refeeding. P-values for all genes and conditions are non significant (P > 0.05). Only male mice were used. Data are shown as mean ± sd. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data

Extended Data Fig. 10. Stimulatory index and glucose-stimulated oxygen consumption are maintained in human monotypic β-pseudoislets.

a. FACS strategy used to purify human β-cells. Representative plots of cells labelled with the pan-endocrine marker HIC1-2B4, the non-β-cell endocrine marker H1C3-2D12 and the δ-cell enriched marker CD9. b. Stimulation index of absolute insulin secretion at 16.7 mM glucose in presence/absence of the GLP-1 agonist Exendin-4 (Ex-4, 50 nM) in pseudoislets composed by all cell types (Ctrl, black) and only β-cells (red). Values are expressed as fold change to baseline values measured at 3 mM glucose (low glucose, LG). N = 4 donors. c. Total insulin content of control (black) and β-only pseudoislets (red) measured after stimulation at 16.7 mM glucose (high glucose, HG). N = 4 donors. P-values (Ctrl vs β): HG P = 0.9708, HG+Ex4 P = 0.4343. P-values Ctrl HG vs HG+Ex4 P = 0.8848. P-values β HG vs HG+Ex4 P = 0.9843. d-f. Pseudoislet respirometry parameters from Fig. 6e. N = 3 independent donors. 2-5 wells/condition/donor, 10 pseudoislets/well. Each data point represents the well average per donor. d. Glucose-stimulated OCR (dashed lined column), proton leak and maximal respiration (grey column) in control (black) and monotypic β-pseudoislets. P-values (Ctrl vs β): Glucose-stimulated OCR P = 0.872; Proton leak P = 0.539; Maximal respiration P = 0.982. (red). e. Basal respiration measured at 3 mM glucose before injections. P-values Ctrl vs β P = 0.4. f. Total insulin content of control (Ctrl) and β-pseudoislets (red) after OCR measurements. P-values Ctrl vs β P = 0.7. Both male and female human donors were used. All data are shown as mean ± sem. All statistical tests are two-tailed Mann-Whitney test. Source data is available. Source data