The DE differentiation stage varies from 3- to 4 days in different studies (Sharon et al., 2019a; Hogrebe et al., 2021). To confirm the suitable duration of differentiation to DE cells, we incubated the cells for 3- or 4-day in stage 1 and found no significant difference in differentiation efficiency (Supplementary Figure S1A). Thus, we have chosen the 4-day incubation at stage 1 to align with the backbone 2D protocol we were based on (Hogrebe et al., 2020). IWP2 was introduced at stages 2, 4, and 5, respectively (Figure 1A). The WNT inhibition at stage 2 by IWP2 did not significantly affect PDX1+ cell commitment (Supplementary Figure S1B). In addition, the WNT inhibition at stage 2 or 4 did not substantially affect the PP cell differentiation, which was evaluated by analyzing the PDX1+/NKX6.1+ cell population (Supplementary Figure S1C). Chromogranin A (CHGA), a pan-endocrine cell marker, was detectable at stage 4, and the WNT inhibition at stage 4 but not at stage 2 significantly increased the CHGA + cell population (Figure 1B, Supplementary Figure S1D). However, the NKX6.1 and CHGA co-expression was rarely detectable by immunofluorescence staining and flow cytometry analysis (Supplementary Figures S1D, E), depicting more non-beta progenitor cells’ introduction due to a WNT inhibition at stage 4 but not at stage 2.

It has been reported that WNT inhibition using IWR at stage 5 in 3D culture significant increases CHGA+ and C-peptide/NKX6.1+ cells differentiation (Sharon et al., 2019b). To investigate the effect of WNT inhibition at stage 5 current in our 2D differentiation protocol, we added IWP2 at stage 5. There was no significant difference at stage 5 on cells positive for C-peptide (Cp), Glucagon (GCG), and CHGA, nor when we calculated the percentage of Cp + cells among CHGA + cells at stage 5 (Supplementary Figures S1G, H). Moreover, the WNT inhibition by IWP2 at late stages (stage 4 or 5) induced a reduction in Cp+ and GCG + cell percentages observed among the final products at stage 6, and a reduced Cp+/CHGA + ratio (Figures 1C,D,F). The difference compared to the previous observation of WNT inhibition at stage 5 may be due to the 2D versus 3D culture systems used, as the study shows that the spatial environments can affect the SC-islets differentiation (Sharon et al., 2019b; Carrasco et al., 2022). Thus, our data showed that the WNT inhibition at stage 2, but not at the late stages (stage 4 or 5), generates the highest CHGA + cells while not significantly decreasing Cp + cell yielding.

To evaluate the effect of the WNT inhibition at stages 2, 4, or 5 on the functionality of the SC-islets at stage 6, a static glucose-stimulated insulin secretion (GSIS) measurement was performed respectively. All the SC-islets generated under different conditions showed a significant increase in insulin secretion in response to glucose (Figure 1E). However, the SC-islets with WNT inhibition at stage 5 showed a higher background insulin secretion in 2 mM glucose Krebs buffer (Krb) than those with or without WNT inhibition at stage 2 (Figure 1E’). In summary, our data showed that WNT inhibition at stage 2 compared to untreated or WNT inhibition at stage 4 or 5, enhanced pancreatic endocrine cell commitment without dismissing the Cp + cell’s differentiation efficiency and insulin secretion ability. Therefore, the subsequent investigation was conducted with the SC-islets generated with IWP2 treatment at stage 2.

Within the SC-islets, C-peptide-positive cells constituted a majority, exceeding 50% (Figure 1C). Small amounts of unwanted cell populations such as bi-hormonal cells that are either Cp+/GCG+, Cp+/Somatostatin (SST)+ (Hogrebe et al., 2020; Mahaddalkar et al., 2020), or enterochromaffin-like cells (SC-ECs) represent mis-differentiated byproducts from SC-islets differentiation (Balboa et al., 2022; Sean et al., 2022). To evaluate the undesired poly-hormonal and off-target cells among SC-islets, we found that Cp+/GCG+ and Cp+/SST + cells still exist in the current products at stage 6, day 14 (Figure 1F). SC-ECs (SLC18A1+) were detectable (Figure 1F). In alignment with previous publications (Hogrebe et al., 2020; Mahaddalkar et al., 2020), the SC-islets express core identity markers, including CHAG, CXCR4, PDX1, NKX6.1, MAFA, and MAFB, similar to human islets (Supplementary Figure S2).

Prolonged SC-islet culture is often needed for disease modeling or drug evaluation (Bauer et al., 2017). To this end, we extended the culture for stage 6 SC-islets to 56 days (8 weeks) in vitro without supplements other than culturing in a defined enriched serum-free medium (ESFM). We observed the GCG+/Cp+ and SST+/Cp + bi-hormonal and the Cp + cell population decreased, whereas the GCG+/Cp- and SST+/Cp-single hormonal cell population increased from week 2–8 (Figure 1G, G’). Flow cytometry quantification confirmed a significant drop of bi-hormonal cells from week 2–8, for Cp+/GCG+ from 16.6% to 5.3%, and for Cp+/SST+ from 12.7% to 7.0% (Figure 1H, H’). In addition, the total GCG + cells significantly increased but not the SST + cells under prolonged culture in vitro. The total Cp + cells significantly declined simultaneously, whereas the CHGA + cells remained unchanged (Figure 1H, H’, Supplementary Figures S3A, B). These results provide provide quantitative evidence that the Cp+/GCG + bi-hormonal cells are more likely to develop as GCG + cells under prolonged culture in vitro.

To further understand the details of SC-islets’ single hormonal cell commitment, we performed immunofluorescence staining with Cp, GCG, and SST antibodies. We observed a tri-hormonal cell population that was positive for Cp, GCG, and SST at week 2 of stage 6, and this tri-hormonal cell population was still observable but was less in week 4 of stage 6, as indicated with white arrowheads. (Figure 2A). The flow cytometry data supported the presence of SST + cells among Cp+/GCG + cell subpopulation (Supplementary Figure S3C). Of note, this tri-hormonal cells were also detectable in the control group that without IWP2 treatment at stage 2 (Supplementary Figure S3D). In addition, Cell populations identified with T-Distributed Stochastic Neighbor Embedding (tSNE) and FlowSOM were labeled and manually annotated, with special interests where populations labeled in purple and green shows the tri-hormonal cells (Figure 2B, B’, Supplementary Figure S3E, E’). In which these two clusters represents tri-hormonal cells that is positive for Cp, GCG and SST. The tri-hormonal cells were fuether characterized as GCG high, Cp/SST low population (Pop1, Supplementary Figure S3F) and SST high, Cp/GCG low population (Pop4, Supplementary Figure S3F) with FlowSOM analysis. From week 2–4 of stage 6, the number of tri-hormonal cells consistently decreased from 10% to less than 5%, followed by less than 3% in week 8 (Supplementary Figure S3G).

To investigate the functionality of SC-islets under long-term culturing in vitro, we monitored the GSIS ability at stage 6 days 14 (week 2), day 42 (week 6), and day 56 (week 8) from three independent batches of differentiation. Three independent batches of SC-islets showed a significant increase in insulin secretion in response to 20 mM glucose stimulation from week 2 to week 8 (Figure 2C). We observed a slight but significant decrease in the insulin stimulation index (SI, insulin secreted in high glucose or indecated insulin secretagogues/insulin produced in low glucose) from week 2 to week 6, however, the SI is still maintained at 1.4 at week 8 (Figure 2D). GSIS analysis in week 2 of stage 6 is the most common timepoint selected for quality assessment in use of SC-islets differentiation protocols (Mahaddalkar et al., 2020; Hogrebe et al., 2021; Balboa et al., 2022). Although under debate, the SI > 1, determined from the ratio of insulin produced in high- and low-glucose, is currently a fundamental release criterion for human islets transplantation (Glieberman et al., 2021; Molano et al., 2024). That said, our data supported that the SC-islets can be used for downstream application in a 6-week time window (from weeks 2–8 of stage 6) without significant abnormality in GSIS ability. This time window provides flexibility for downstream interpretation of disease modeling.

To investigate the connections of gene expression to the GSIS function changes, we conducted the gene expression analysis under prolonged culture condition in vitro. The expressions of genes including, Insulin, PDX1, NKX6.1, NKX2.2, MAFB, ISL1, NEUROD1 and ABCC8, were higher compared to human islets, while the disallowed gene such as LDHA is significantly lower (Supplementary Figure S3H). The Insulin expression was significantly increased in week 2 compared to week 1 of at stage 6 but remained unchanged under prolonged culturing (stage 6 weeks 2 vs. 6, Figure 2E). However, the GCG and SST expression was significantly increased under prolonged culturing in week 6 compared to week 2 at stage 6 (Figure 2E). The increase of GCG and SST gene expressions further indicated that the tri-hormonal and bi-hormonal cells may not be as likely to develop into Cp + cells under the current protocol.

The MAFA expression in SC-islets at week 2 of stage 6 was significantly lower than in human islets, with a downregulation from stage 5–6 during the differentiation (Figure 2F, Supplementary Figure S3H). However, under prolonged culture, the MAFA and MAFB mRNA expression levels remain unchanged (Figure 2F). MAFA was located in the nucleus at stages 5 and the first week of stage 6 (Supplementary Figure S4). However, under prolonged culture in vitro, the MAFA’s nucleus localization became less pronounced, whereas the MAFB was presented in nucleus (Supplementary Figure S4). Evidence supports that MAFA is a maturation marker for beta cells and keeps upregulation postnatal, while MAFB keeps downregulation (Balboa et al., 2022). Thus, the loss of MAFA nucleus localization may relate to the fact that SC-islets could not maintain their GSIS responses for more than 6 weeks in vitro.

To evaluate the in vivo function, we transplanted two batches of the SC-islets (week 2 of stage 6) under the kidney capsule in 5 healthy male Rag1^−/− mice that do not contain mature B and T lymphocyte (Figure 3A). The human Cp was detectable at 2 weeks post-transplantation and increased from day 41 to day 83 post-transplantation (Figure 3B). Alloxan that ablates selectively rodent but not human beta cells was administrated at day 90 post-transplantation to induce diabetic mice. After Alloxan-induced diabetes, the random blood glucose was maintained at a normal range (<11 mM) in the mice with SC-islets grafts compared to control mice without SC-islets (Figure 3C). Furthermore, the oral glucose tolerance test (OGTT) at day 100 showed SC-islets grafted mice had a glucose tolerance similar to the healthy mice (Figure 3D), as well as when calculated as the area under curve (AUC) (Figure 3E). Immunostaining of the retrieved kidney with SC-islets grafts after termination of the studies revealed a large region of Cp+ and GCG + cells, with a few SST + cells presented (Figures 3F,G). These data demonstrated that these SC-islets prevented Alloxan-induced diabetes in mice.

To assess whether the information collected with SC-islets as disease models is translatable to that in human islets, we investigated the insulin secretion of SC-islets (during weeks 2-3 of stage 6) in response to different nutrient sources and benchmarked to human islets. Human primary islets showed a glucose SI of 3.13, significantly higher compared to 1.43 observed from 4 independent batches of SC-islets (Figure 4A, Supplementary Figure S5A, A’). Next, adding 5 mM of leucine alone in Krb without glucose strongly stimulates insulin secretion in human islets, with SI = 2.99, whereas SC-islets were unresponsive (SI = 0.83) (Figures 4B,C, Supplementary Figure S5B, B’). Glutamine is an important energy substrate for many body tissues, and is the most abundant amino acid in the blood, with a normal circulation concentration maintained at 0.6–0.9 mM (Henquin et al., 2006). A 10 mM of glutamine alone in Krb without glucose could not induce insulin secretion neither in human islets (SI = 0.85) nor in SC-islets (SI = 0.78) (Figures 4B,C, Supplementary Figure S5C, C’). The SI in response to a mix of 5 mM of leucine and 10 mM of glutamine without glucose in Krb was 4.71 for human islets, whereas the SI was 1.02 for SC-islets (Figures 4B,C, Supplementary Figure S5D, D’). The glutamine does not significantly increase the glucose stimulated insulin secretion index neither for human islets nor SC-islets, which showed that glutamine is not a primary insulin secretagogue in the categories of small amino acids. Interestingly, leucine stimulated insulin secretion in human islets but not in SC-islets. However, the mixture of 2 mM glucose, 5 mM leucine and 10 mM glutamine increased the SI for SC-islets from unresponsive to 1.20, whereas the SI for human islets was 4.08 (Figures 4B,C, Supplementary Figure S5E, E’). When 20 mM glucose presented together with a leucine-glutamine mixture, the SI for human islets increased to 5.96, whereas the SI for SC-islets increased to 1.53 (Figures 4B,C, Supplementary Figure S5F, F’). It has been reported that low-protein content diets induce mild insulin secretion, whereas a high-protein content diet meal potentiates the insulinemic response (Regazzi et al., 2016). Overall, our data suggested that leucine could stimulate insulin secretion in human islets but not in SC-islets. In addition, 5 mM leucine plus 20 mM glucose showed a trend of potentiated insulin secretion in human islets but not in SC-islets (Figures 4B,C).

Palmitic acid is the most common saturated fatty acid in the human body (Carta et al., 2017). To investigate the ability of SC-islets to secrete insulin in response to fatty acid and fructose, we incubated the cells in 0.2 mM palmitate together with different concentrations of glucose. The SI for human islets incubated with 0.2 mM palmitate alone was 1.28, whereas the SI for SC-islets was 1.18 (Figures 4D,E, Supplementary Figure S5G, G’). When 2 mM glucose presents together with 0.2 mM palmitate, the SI for human islets was 1.88 and for SC-islets increased to 1.49 (Figures 4D,E, Supplementary Figure S4H, H’). When 20 mM glucose presents together with 0.2 mM palmitate, the SI for human islets increased to 5.23, and the SI for SC-islets increased to 1.91 (Figures 4D,E, Supplementary Figure S5I, I’). The fructose was added with different concentrations together with 2 mM glucose. 2-, 5-, 10-, and 20-mM fructose could not induce significant insulin secretion in both human and SC-islets (Supplementary Figure S5J). In summary, palmitate alone could not stimulate insulin secretion in human islets and SC-islets. However, when 0.2 mM palmitate was added to 20 mM glucose in Krb, the SI was significantly higher than 20 mM glucose alone in stimulating insulin secretion in SC-islets (Figures 4D,E). Thus, our data revealed that palmitate could potentiate insulin secretion of SC-islets, consistent with previous observations in human islets that the presence of palmitate plus glucose increases insulin secretion than glucose alone (Parker et al., 2003; Cen et al., 2016; Kristinsson et al., 2017; Kuok et al., 2019).

Although steady improvements have been made, the current SC-islets differentiation protocols are often unable to achieve an equivalent function to that of human primary islets (Davis et al., 2020). We observed a significantly lower GSIS index of SC-islets from our preparations and the different responses to nutrient sources compared to human islets (Figure 4). It has been evidenced that reduced anaplerotic cycling in the mitochondria causes reduced GSIS in SC-islets (Davis et al., 2020). To investigate the metabolism chain sufficiency of our SC-islets, we checked insulin secretion response with several inter-metabolites and benchmarked to human islets.

Glucose 6-phosphate (G6P) is the first intermediate of glucose metabolism, it bears a negative charge and cannot cross the cell membrane (Rajas et al., 2019). The SI for human islets incubated with 10 mM glucose 6-phosphate (G6P) alone was 0.96, whereas the SI for SC-islets was 1.00 (Figures 5A,B, Supplementary Figure S6A, A’). Phosphoenolpyruvate (PEP) is converted into ATP and pyruvate by pyruvate kinase (PK), which underlies β-cell sensing of both glycolytic and mitochondrial fuels (Lewandowski et al., 2020). The SI for human islets incubated with 10 mM PEP alone was 1.22, whereas the PEP significantly increased the insulin secretion for SC-islets compared to glucose SI (SI = 4.13) (Figures 5A,B, Supplementary Figure S6B, B’). It has been reported that SC-islets can actively respond to pyruvate stimulation in a dynamic profusion condition, while human islets are less responsive to pyruvate. This phenomenon has been characterized as an “immaturity” of SC-islets (Balboa et al., 2022). We could not observe a significant response to pyruvate in terms of insulin secretion from both human islets (SI = 0.80) and SC-islets (SI = 0.98) under a static incubation condition (Figure 5C, Supplementary Figure S6C, C’). Under a dynamic profusion condition, 10 mM pyruvate could not induce insulin secretion in both human islets and SC-islets, and the following 20 mM glucose perfusion massively induced insulin secretion in human islets but not in SC-islets (Supplementary Figure S6E, E’).

Unlike pyruvate, which barely triggers insulin secretion, methyl pyruvate (M-Pyr) has been proposed as a potent mitochondrial substrate and is widely used to study beta cell stimulus-secretion coupling (Düfer et al., 2002; Düfer et al., 2002). It has been reported that M-Pyr depolarize beta cells in a concentration-dependent manner from 5 to 20 mM, and stimulates insulin secretion at 20 mM (Zawalich and Zawalich, 1997; Lembert et al., 2001). To keep the equimolar mass as other insulin secretagogues used for this study, we incubated human islets and SC-islets with 10 mM M-Pyr, and observed an increase of insulin secretion in SC-islets with an SI = 1.58, however, no statistic significance when compared to cell-impermeable G6P stimulation (Figure 5B). In addition, 10 mM M-Pyr could not significantly induce insulin secretion in human islets (SI = 0.77) (Figure 5D, Supplementary Figure S6D, D’). Interestingly, 20 mM of M-Pyr induced insulin secretion under a dynamic profusion condition (Supplementary Figure S6F, F’). Almost half of pyruvate is converted to oxaloacetic acid (OAA) (anaplerosis) inside mitochondria in beta cells (Prentki et al., 2013). 10 mM of OAA slightly stimulated insulin secretion for human islets (SI = 1.83), whereas had a higher magnitude effect for SC-islets (SI = 4.9) (Figures 5A,B, Supplementary Figure S6G, G’).

Glucose-induced mitochondrial respiration ability is highly correlated with GSIS ability (Davis et al., 2020). To investigate the glucose metabolism chain, we analyzed the real-time oxygen consumption rate (OCR) during glucose stimulation and different mitochondrial respiration modulators, including Oligomycin, Carbonyl cyanide 3-chlorophenylhydrazone (CCCP), and Antimycin A & Rotenone. We observed that 20 mM of glucose increased mitochondrial respiration in human islets (Figure 5C) and SC-islets (Figure 5D). The area under the curve (AUC) analysis indicated that the OCR of human islets and SC-islets under glucose stimulation has no significant difference (Figure 5E).

Glucose sensing, followed by uptake and oxidative metabolism, is tightly regulated by the beta cells in islets to maintain glucose homeostasis in the body (Davis et al., 2020). To investigate the glucose uptake and oxidative metabolism in SC-islets, we traced glucose metabolism by 0.2 mM D-[¹⁴C] glucose with our previously developed method (Supplementary Figure S6H) (Wensaas et al., 2007). The completed glucose oxidation measured by the trapped ¹⁴CO₂ was similar in SC-islets to that in human islets (Figure 5F). In addition, once a mitochondrial pyruvate carrier (MPC) blocker, UK-5099, was added to the incubation buffer, the glucose oxidation was significantly inhibited in SC-islets, whereas only a decreasing trend observed in human islets (Figure 5F), indicating that the ¹⁴CO₂ production was highly dependent on following mitochondrial metabolism after glycolysis in SC-islets and human islets.

Cell-associated D-[¹⁴C] glucose measurement revealed a lower glucose accumulation in SC-islets than in human islets (Figure 5G). The MPC blocking induced a higher glucose accumulation that is three times higher in human islets than in SC-islets. In contrast, glucose accumulation induced by MPC blockers slightly decreased in SC-islets (Figure 5G). The total glucose uptake for SC-islets was similar to human islets, (Figure 5H), which is consistent with previous observation (Davis et al., 2020). However, MPC blocking increased the total glucose uptake in human islets and significantly decreased glucose uptake in SC-islets (Figure 5H). The glucose oxidation fraction (“completed glucose oxidation”/(“completed glucose oxidation” + “cell-associated glucose”)) was lowered by MPC blocking in both human islets and SC-islets (Figure 5I). Surprisingly, the glucose oxidation fraction was significantly higher in SC-islets than in human islets (Figure 5I). Previous data revealed insufficient glycolysis in SC-islets is a bottleneck that reduces insulin secretion compared with human islets (Davis et al., 2020). The pyruvate-mitochondrial transportation cutting off dismissed the glucose uptake and accumulation in SC-islets but not in human islets, supported that the glycolysis may not be sufficient even at a basal respiration condition (0.2 mM glucose).

To investigate how the SC-islets repones to potential antidiabetic drugs, we tested a classical glucokinase activator, MK-0941 (Meininger et al., 2011), and a recently developed pyruvate kinase activator (PKa), TEPP46 (Lewandowski et al., 2020) in SC-islets. Glucokinase (GK) is expressed in beta cells and sets the metabolic flux rate to enable an instant response to blood glucose changes (Lewandowski et al., 2020). Thus, glucokinase activation (GKa) was initially predicted can enhance insulin section (Ashcroft et al., 2023). Pyruvate kinase (PK) converts ADP and PEP into ATP and pyruvate. Recent preclinical data showed that PK activation improved beta-cell function in vitro and in vivo (Abulizi et al., 2020; Corkey, 2020). Consistent with previous reports (Lewandowski et al., 2020), we observed an acute response to PKa and GKa in human islets regarding increased insulin secretion under glucose stimulation (Figure 6A). However, GK and PK activation could not significantly increase insulin secretion for SC-islets (Figure 6B, B’, Supplementary Figure S7D and E). By incubating the SC-islets with PEP and PKa to investigate if the PK activation can increase insulin secretion by bypassing the early glycolysis clog, we still cannot observe a significant change in SI (Figure 6C and Supplementary Figure S7F). The D-[¹⁴C] glucose oxidative analysis showed that PKa did not impact SC-islets’ oxidative status when PKa was added into the Krb buffer (Figure 6D).

Animal models and T2D patients with GKa treatments have been reported hypoglycaemic (Meininger et al., 2011), whereas PK activation improved whole-body metabolic health in diabetic rat (Abulizi et al., 2020), which demonstrated that PK activation by TEPP46 could be a potentially better antidiabetic drug. Thus, we incubated the human islets and SC-islets (during week 2–3 of stage 6) with PKa for 6 days to investigate the long-term effects. 6 days of incubation with PKa did not significantly alert SC-islets glucose oxidation (Figure 6D), accumulation (Figure 6E), uptake (Figure 6F), and the fraction of oxidized glucose status (Figure 6G), although upward trends were observable (Figures 6D,E). 6 days of incubation PKa did not increase the insulin secretion index for human islets (Figure 6H, Supplementary Figure S7G). Unlike the actuate incubation with PKa, the insulin secretion indexes were significantly increased in SC-islets after 6 days incubation with Pka (Figure 6H’, Supplementary Figure S7H). The mRNA expression analysis revealed that the beta cell maturation makers, such as NKX2.2, MAFA, and Glut1 were significantly upregulated after the 6-day incubation with PKa (Figure 6I). It was not surprising that PKa failed to increase insulin secretion in SC-islets in acute situations as PEP can successfully elevate insulin secretion, which means the PK activity may already be sufficient in SC-islets (Figure 6J). Nonetheless, the long-term incubation with PKa significantly increased SI in SC-islets, indicating that the PK activation may have a different regulation mechanism for SC-islets other than directly modulating PK activity.

SC-islets were generated from human embryonic stem cell line H1 (Coriell Institute for Medical Research). Undifferentiated H1 cells were cultured in Essential 8™ media (A1517001, ThermoFisher) on tissue culture plates coated with Geltrex™ (A1413201, ThermoFisher) in a humidified incubator containing 5% CO₂ at 37°C. The H1 cell clones were passaged as clumps in every 2–4 days by 0.5 mM EDTA dissociation. To initiate SC-islets differentiation, H1 cells were counted with NucleoCounter NC-202 cell counter (ChemMetec) and seeded as single cells at 2 × 10⁵ cells/cm² in 1:50 diluted Geltrex™ coated cell culture plates for 24 h. Cells were then washed with DPBS and incubated with stage 1 media supplemented with 3 μM CHIR99021 (S2924, Selleckchem) and 100 ng/mL Activin A (120–14, PeproTech) for day 1 of stage 1. On days 2–4, the cells were incubated with stage 1 media supplemented with 100 ng/mL Activin A. At stage 2, the cells were incubated for 2 days with stage 1 basal media supplemented with 0.25 mM vitamin C (A4544, Sigma), 1.25 µM IWP2 (3533, Bio-Techne) and 50 ng/mL KGF (AF-100–19, PeproTech). At stage 3, the cells were incubated for 2 days with stage 3 basal media supplemented with 0.2 µM LDN193189 (6053, Tocris), 0.2 µM TPPB (HY-12359, MedChemExpress), 2 µM Retinoid Acid (R2625, Sigma), 0.25 µM SANT1 (C23H27N5, Sigma) and 50 ng/mL KGF. At stage 4, the cells were incubated for 4 days with stage 3 basal media supplemented with 0.1 µM Retinoid Acid, 0.25 µM SANT1, 0.2 µM LDN193189, 0.2 µM TPPB and 50 ng/mL KGF. At stage 5, the cells were incubated with stage 5 basal media supplemented with 0.1 µM Retinoid Acid, 0.25 µM SANT1, 0.2 µM LDN193189, 10 µM ALK5 inhibitor II (ALX-270–445-M005, Enzo Life Sciences), 1 µM T3 (T6397, Sigma), 1 µM Xxi (565789, Sigma). 1 μM Latrunculin A (3973, Bio-Techne) was added during the first day of stage 5. At stage 6, the cells were cultured in enriched serum-free medium (ESFM) with media change every day for 7 days. On day 7 of stage 6, the cells were dissociated with TrypLE Express (12604013, Gibco) and 4 million cells in 4 mL ESEM media were seeded in each well of ultra-low attachment cell culture plate (CLS3261, Sigma). The cells were then aggregated and maintained as spheroids on an orbit-shaker (Thermo Fisher) at 100 RPM for over 7 days till analysis or drug treatment. The media was changed every other day after day 7 of stage 6. The basal media were as follows. Stage 1: MCDB 131 (10372–019, Invitrogen) supplemented with 1% P/S, 100x GlutaMAX (35050079, Invitrogen), 0.5% BSA (68700, Proliant), 1.174 g/L NaHCO₃ (S6297, Sigma), 4.5 mM glucose (G7021, Sigma). Stage 3: MCDB 131 supplemented with 1% P/S, 100x GlutaMAX (35050079, Invitrogen), 2% BSA, 1.154 g/L NaHCO₃, 4.5 mM glucose, and 0.25 mM vitamin C. Stage 5: MCDB 131 supplemented with 1% P/S, 1x GlutaMAX, 2% BSA, 1.154 g/L NaHCO₃, 14.5 mM glucose, 0.25 mM vitamin C, 10 μg/mL Heparin (H3149, Sigma) and 200x ITS-X (51500056, ThermoFisher). ESFM: MCDB 131 supplemented with 1% P/S, 100x GlutaMAX, 100x NEAA (11140050, Gibco), 2% BSA, 1000x Trace Elements A (Corning, 25-021-CI), 1000x Trace Elements B (Corning, 25-022-CI), and 10 µM ZnSO4 (Z0251, Sigma). The WNT inhibition was conducted by adding 0.25 µM of IWP2 in the completed differentiation either at stages 2, 4 or 5.

Human primary islets from non-diabetic brain-dead multiorgan donors were obtained from the Nordic Network for Islet Transplantation (Oslo University Hospital, Rikshospitalet) after appropriate informed consent from relatives for multi-organ donation and use in research, and approved by the Regional Ethics Committees (REK 67671 and REK 270665). Human islets were cultured in CMRL 1066 (15-110-CV, Corning) supplemented with 5% human AB serum (000921, Milan Analytica AG), 20 mM HEPES, and 1% P/S in 6 well ultra-low attachment plates. Islet donor characteristics are listed in Supplementary Table S1. The human islets were maintained in a humidified incubator containing 5% CO₂ at 37°C for a maximum of 3 weeks and the analysis and drug treatment were conducted within these 3 weeks. The media were exchanged every two to 3 days.

SC-islets and human islets were harvested and fixed in 4% PFA for 30 min at 4°C. After blocking with blocking buffer (DPBS with 1% BSA, 0.1% Triton-X100) for 30 min at room temperature (RT), the spheroids were washed with DPBS with 1% BSA and placed in 30% sucrose (S0389, Sigma) at 4°C overnight. The spheroids were then embedded in O.C.T (23730571, Fisherscientific) and preserved at −80°C. The spheroids were cut into 10 µm thick slides with Cryostats (CM 1950, Leica). Cells during differentiation were fixed in 4% PFA for 15 min at RT, permeabilized and blocked with a blocking buffer for 30 min at RT. Animals were euthanized and grafts retrieved within 230 days of implantation. Mouse kidneys grafted with SC-islets were fixed in formalin and embedded in paraffin, and then cut 8 µm thick by Oslo university hospital. The paraffin was removed with Xylene (534056, sigma) and rehydrated with ethanol, and antigens retrieval was done with target retrieval solution (S236784-2, Dako) and boiling water bath. Primary antibodies were incubated overnight at 4°C, and secondary antibodies were incubated for 1 h at RT. Antibodies information used for immunofluorescence staining is listed in Supplementary Table S2. The images were taken with Zeiss LSM 700 confocal microscopy and Nikon Eclipse Ts2R-FL microscopy.

Cells during differentiation at each stage or the SC-islets were dissociated as single cells with TrypLE Express for 3 min at 37°C, and then fixed in 4% PFA for 15 min at RT, permeabilized and blocked with blocking buffer for 30 min at RT. Primary antibodies were incubated overnight at 4°C, and secondary antibodies were incubated for 1 h at RT. Cells after immunostaining were analyzed with LSR Fortessa II (BD Biosciences). Flow cytometry data were analyzed with FlowJo (Version 10.8.1). Over 50,000 cells were acquired for tSNE analysis (Maaten and Hinton, 2008). For cross sample comparism, the data set for each sample were then sampled with Downsample plugin (Version 3.3.1, Flowjo-exchange) to equally assign the same number of events for each sample (20,000 events after single cell gating), and then the three-channel fluorescence data were de-dimensioned with build in tSNE function with the Exact and FFT interpolation algorithm selected, and cell subpopulation was then clustered with FlowSOM plugin (Version 3.0.18) (Van Gassen et al., 2015).

Total RNA from the differentiated cells and human primary islets was isolated using TRIzol by following the manufacturer’s instructions. The cDNA was synthesized with a High-Capacity cDNA Reverse Transcription Kit (4368814, Applied Biosystems). PowerUp SYBR Green (A25780, Applied Biosystems) based real-time PCR was performed in Viia 7 real-time PCR system (Applied Biosystems) with the standard model. TBP (TATA box binding protein) was taken as an internal normalization control. Primers were provided in Supplementary Table S3.

Animals were housed with no more than 5 mice per cage, under a 12 h light-dark cycle with free access to food and water except during fasting. 10-12-week-old male BALB/c Rag 1^−/− immunodeficient mice (C.129S7(B6)-Rag1 ^tm1Mom /J, stock 003145, The Jackson Laboratory) were transplanted with 600–700 SC-islets under the kidney capsule by using a 25G butterfly needle. The mouse’s blood glucose was monitored every week with blood glucose test strips (06453970170, Aviva). Approximately 60 µL of mice blood was taken in every 2 weeks for human C-peptide measurement with the human C-peptide ELISA kit (10–1136-01, Mercodia). At day 90 post transplantation 70 mg/kg alloxan (A7413, Sigma Aldrich) was administrated intravenously followed by monitoring of blood glucose weekly for the first 3 months of alloxan administration and every other week for the next 3 months. Mouse oral glucose tolerance tests (OGTT) were conducted 12 days after alloxan administration. Blood glucose was measured at 15, 30, 45, 60, and 120 min after glucose administration. The in vivo experiments were approved by the Norwegian National Animal Research Authority (FOTS ID 14263) and performed according to the guidelines for care and use of laboratory animals published by the US National Institutes of Health (NIH Publication, 8th Edition, 2011), and Norwegian Animal Welfare Act.

30 SC-islets or 20 human primary islets were hand-picked into Transwell cell culture inserts (CLS3414, Sigma) placed in 24 well cell culture plates. The spheroids were washed three times and equilibrated in 1 mL Krebs-Ringer buffer (KRB: 128 mM NaCl, 5mM KCl, 1.2 mM MgSO₄, 2.7 mM CaCl₂, 1.2 mM KH₂PO₄, 1 mM Na₂HPO₄, 5 mM NaHCO₃, 10 mM HEPES and 0.1% BSA) containing 2 mM glucose for 1 h at 37°C. The SC-islets or human islets were then incubated sequentially in KRB containing 2 mM glucose, KRB containing different nutrients or inter-metabolites, and KRB containing 2 mM glucose with 30 mM KCl for 60 min each. The supernatant after incubating in each step was harvested for detecting insulin production. Insulin was measured by human insulin ELISA kit (10-1113-10, Mercodia). Different chemicals used for insulin stimulation were listed in Supplementary Table S4.

The oxygen consumption rate (OCR) analysis was performed under the Seahorse XFe24 analyzer (Agilent). 30 differentiated SC-islets or human islets were hand-picked and seeded in the XF24 24-well cell culture microplates, the spheroids were washed 3 times with RPMI-1640 media (R6504, Sigma) containing 0.1% BSA, and then equilibrated in assay media (RPMI-1640 media supplemented with 0.1% BSA and 2 mM glucose) for 2 h at 37°C in air. The basal respiration was determined in assay media and data was measured for 5 cycles with a 6-min interval. Followed by incubating the cells with 20 mM glucose to determine the OCR change upon glucose stimulation. The oxygen consumption at 20 mM glucose condition was measured for 6 cycles with a 6-min interval. The final concentration at 5 μM of oligomycin (9996L, Cell Signaling Technology), CCCP (Carbonyl cyanide 388 3-chlorophenylhydrazone, C2920, Sigma), Rotenone (R8875, Sigma) and Antimycin A (A8674, Sigma) were added sequentially. The oxygen consumption was then measured in 5 cycles with a 6-min interval. The OCR values were normalized to the averaged baseline values.

To examine glucose metabolism, human or SC-islets were differentiated as described previously and cultured in an ultra-low attachment treatment (ULA) Nunclon™ Sphera™ 96-well plate with D-[¹⁴C(U)]glucose (0.5 μCi/mL, 200 µM prepared in DPBS with BSA (15 μM) and HEPES (10 mmol/L)) (PerkinElmer NEN^®, Boston, MA, United States) for 4 h following the previously described method (Supplementary Figure S6H) (Wensaas et al., 2007). Briefly, 1 M NaOH was used to activate a 96-well UniFilter^® microplate (PerkinElmer, Shelton, CT, United States) for capturing CO₂ from metabolism and then placed on top of the 96-well plate during the incubation time. After 4 h, islets were then placed in small tubes (200 µL), washed with PBS, and harvested in 0.1 M NaOH. The ¹⁴CO₂ trapped in the top filter (complete glucose oxidation), resulting from glycolysis and cellular respiration, and the cell-associated radioactivity were quantified by adding scintillation fluid (Ultima Gold XR, PerkinElmer) and counted using a 2450 MicroBeta2 scintillation counter (PerkinElmer). All results were adjusted for protein content, determined by Bio-Rad protein assay in VICTOR™ X4 Multilabel Plate Reader (PerkinElmer). The combined value of ¹⁴CO₂ and the remaining cell-associated (CA) radioactivity were considered as an indicator of the total glucose uptake, represented as CO₂ + CA. The fractional glucose oxidation was calculated as CO₂/(CO₂ +CA). Fractional oxidation represents the proportion of uptake that undergoes oxidation and may or may not correlate with complete oxidation.

The “n” separates experiments from different biological replicates or independent stem cell differentiations or independent human donors; “N” separates experiments from technical replicates. Data is presented as mean ± standard deviation (SD) unless otherwise stated and GraphPad Prism software (10.2.0) was used for data analysis. The normal distribution of the data was assessed by the Shapiro-Wilks test before statistical analysis. If the data did not pass the normality test, the non-parametric option of the appropriate statistical test was used. Any outlier values in insulin secretion measurements were identified using ROUT with Q = 1% and excluded from statistical analysis and figure representation. Differences among the three groups were evaluated by one-way ANOVA or a two-way ANOVA followed by Dunnett’s or Turkey’s multiple comparison post hoc test. An unpaired t-test or a Mann-Whitney U test (for non-parametric) was performed for differences analysis between the two groups. Significance was set at p < 0.05.