Modeling diabetic alpha cell dysfunction using stem cell-derived alpha cells

Summary

Dysfunction of pancreatic alpha cells contributes to the pathophysiology of diabetes. Features of diabetic alpha cell dysfunction include glucagon hypersecretion, defects in proglucagon processing, and altered transcriptomic profile. The lack of an in vitro human alpha cell model has prevented the investigation, and potential correction, of these dysfunctional phenotypes. Here, we show that induction of endoplasmic reticulum (ER) stress in stem cell-derived alpha (SC-α) cells induces hypersecretion of glucagon. ER stress also increases the secretion of glicentin and the expression of glucagon-like peptide-1 (GLP-1), peptides produced by alternate cleavage of proglucagon by the prohormone convertase 1 (PC1/3) enzyme. Additionally, ER stress establishes a diabetic transcriptional state in SC-α cells characterized by downregulation of MAFB, as well as glycolysis and oxidative phosphorylation pathways. We show that sunitinib, a tyrosine kinase inhibitor, protects SC-α cells against the ER stress-induced glucagon hypersecretion phenotype. Thus, SC-α cell model can advance our knowledge of islets in health and diabetes.

Graphical abstract

Highlights

• •SC-α cells share transcriptional similarities with primary α cells
• •SC-α cells can recapitulate the diabetic phenotype of hypersecretion of glucagon
• •ER stress causes defects in proglucagon processing and alters transcription profile
• •Sunitinib protects SC-α cells against ER stress-induced hypersecretion of glucagon

Shrestha et al. report the development of an in vitro human model of pancreatic alpha cells using pluripotent stem cells (SC-alpha cells). SC-alpha cells recapitulate the gene expression, proglucagon processing, and glucagon secretion of human alpha cells and model diabetes-related defects in these processes in response to ER stress. The authors identify a tyrosine kinase inhibitor that protects SC-alpha cells from ER stress-induced dysfunction. SC-alpha cells provide a valuable model for the field to advance our knowledge of alpha cell physiology.

Introduction

Diabetes represents a global health crisis. In the United States, 38 million people have been diagnosed with diabetes (Parker et al., 2024). Although the most well-studied aspects of diabetes pathology are deficits in the production and release of insulin by pancreatic beta cells and insulin resistance in peripheral tissues, the importance of dysfunction within glucagon-producing alpha cells is increasingly recognized (Ahrén, 2009; Grubelnik et al., 2020). The dysfunction of alpha cells may precede or coincide with beta cell dysfunction (Jamison et al., 2011; Mumme et al., 2017; Doliba et al., 2022). Defects in diabetic alpha cells include improper glucagon secretion, altered proglucagon processing, and transcriptional changes (Reaven et al., 1987; Campbell et al., 2020; Bosi et al., 2022).

In diabetic individuals, alpha cells inappropriately secrete high amounts of glucagon when blood glucose levels are high (Dinneen et al., 1995), exacerbating hyperglycemia (Jamison et al., 2011; Vergari et al., 2019; Asadi and Dhanvantari, 2021). The lack of proper paracrine signaling from beta cells and impaired glucose sensing have been proposed to contribute to the inability of diabetic alpha cells to properly restrain glucagon secretion.

Proglucagon processing is also altered in diabetes (Zhou et al., 1999). In normal alpha cells, proglucagon is cleaved by prohormone convertase 2 (PCSK2; PC2) to produce glucagon (Rouille et al., 1994). In other endocrine cell types including intestinal L-cells and neuroendocrine cells, the same proglucagon peptide is cleaved into glicentin, glucagon-like peptide-2 (GLP-2), and GLP-1 via prohormone convertase 1 (PCSK1; PC1/3) (Lafferty, O'Harte et al., 2021). Interestingly, a subpopulation of alpha cells within normal human islets expresses PC1/3 and produces GLP-1(Marchetti et al., 2012), which acts as a paracrine signal to stimulate insulin secretion and suppress glucagon secretion (Chambers et al., 2017; Traub et al., 2017; Ramracheya et al., 2018). The number of GLP-1-expressing cells is higher in type 2 diabetes (T2D) islets, although it is unclear if this phenotype is a cause or effect of the diabetic condition (Campbell et al., 2020; Wideman et al., 2009; He et al., 2021).

Transcriptional changes within alpha cells are also observed in diabetes. Downregulation of alpha cell-specific transcription factors ARX and MAFB is a common feature of alpha cells from diabetic patients (Fujita et al., 2021; Brissova et al., 2018; Conrad et al., 2016; Artner et al., 2010). Transcriptomics studies have revealed downregulation in metabolic pathways such as glycolysis and oxidative phosphorylation in alpha cells of diabetic patients (Bosi et al., 2022). These changes may be an early marker of alpha cell dysfunction, as they can be observed in individuals who tested positive for a single autoantibody but have not yet developed the clinical features of type 1 diabetes (T1D) (Doliba et al., 2022; Bosi et al., 2022).

These alpha cell-specific changes in hormone secretion, proglucagon processing, and transcriptional profile may play a role in diabetes etiology. To date, the lack of a suitable human cell model has prevented a better understanding of these phenotypes. Current models such as rodent islets and immortalized cell lines do not fully reflect the metabolism and functionality of human islets (Tellez et al., 2020; Riahi et al., 2023).

To generate a more relevant model of human alpha cells, we have generated stem cell-derived alpha (SC-α) cells (Peterson et al., 2020). We identify a diabetogenic stressor that, when applied to SC-α cells in vitro, results in hypersecretion of glucagon, incorrect proglucagon processing, and altered transcriptomic profile. We also show a proof-of-concept study using the drug sunitinib to correct these phenotypes. Together, these findings establish a new model of human alpha cell dysfunction and demonstrate its utility in identifying compounds that can restore function in diabetes-like conditions.

Results

SC-α cells share transcriptional similarities with human islet alpha cells

In order to develop a model of diabetic alpha cell dysfunction, we used our previously reported 6-stage differentiation protocol to generate a population of approximately 30% mono-hormonal glucagon-positive SC-α cells from human embryonic stem cells (Figures 1A, 1B, and S1) (Peterson et al., 2020). To assess whether SC-α cells resemble bona fide alpha cells from human pancreatic islets (HI-α), we performed single-cell RNA sequencing (scRNA-seq) on these cells and compared the results to published data from primary human islets (Xin et al., 2018). Uniform manifold approximation and projection (UMAP) and cluster analysis revealed that the SC-α protocol generates 68% glucagon-expressing SC-α cells (Figures S1G and S1H), in addition to other cell types (Figures 1C and 1D). To determine how closely SC-α cells resemble HI-α cells, we compared the average gene expression in these two cell types (Figure 1E; Table S1). Glucagon (GCG) was the most abundant transcript in both. Among the top 100 most abundantly expressed genes in each group, 77 genes were shared between SC-α cells and HI-α cells including chromogranin A (CHGA), transthyretin (TTR), and carboxypeptidase-e (CPE) (Figure 1F; Table S1).

Figure 1.

SC-α cells share transcriptional similarities with human islet alpha cells

(A) Schematic of directed differentiation of SC-α cells with stages representing embryonic stem cells (ES), definitive endoderm (DE), gut tube endoderm (GTE), pancreatic progenitor (PP), endocrine progenitor (EP), pre-alpha (PA), and SC-α.

(B) Representative image of flow cytometry of SC-α cells expressing insulin (x axis) and glucagon (y axis) (n = 9).

(C) UMAP showing different cell types generated from SC-α protocol (n = 3).

(D) UMAP showing different cell types present within human donor islets (n = 12).

(E) Scatterplot comparing normalized RNA counts in log scale in scRNA-seq of SC-α cells (x axis) and alpha cells from human donor islets (y axis).

(F) Venn diagram highlighting 77/100 genes being shared between the top 100 genes in alpha cells from SC-α protocol and the top 100 genes in alpha cells from human donor islets.

(G) Bar plot showing fold change in expression of transcripts associated with alpha cell identity and function in alpha cells from SC-α protocol and alpha cells from human islets.

We further assessed the relative expression of selected genes known for maintaining alpha cell identity (Figure 1G) (Brissova et al., 2018; van Gurp et al., 2022). HI-α displayed higher expression of FAP, GPX3, and TM4SF4 (log₂FC < −1, p < 0.0001), while SC-α cells had higher expression of CHGA, IRX1, IRX2, and ARX (log₂FC >+1, p < 0.0001). GCG, MAFB, PAX6, RFX6, TTR, and ISL1 were expressed similarly in both HI-α and SC-α cells. These findings revealed that SC-α cells are transcriptionally similar to HI-α cells and thus represent a good model in which to assess perturbations to the alpha cell transcriptome.

SC-α cells as an in vitro model to explore proglucagon processing

To investigate the utility of SC-α cells to study proglucagon processing via PC2 and PC1/3, we first compared transcripts of PSCK2 and PCSK1 in HI-α and SC-α cells. Both cell types abundantly expressed PSCK2, whereas PCSK1 transcripts were nearly undetectable (Figures S2A and S2B). Immunostaining for PC2 and PC1/3 showed that, as expected, PC2 colocalized with GCG in both HI-α and SC-α cells (Figures 2A and S2C). In HI, PC1/3 protein colocalized only with insulin (INS) (i.e., beta cells) and was excluded from HI-α cells (Figures 2B and S2D). In SC-α cell clusters, we identified four cell populations expressing PC1/3: PC1/3⁺GCG⁻INS⁺ (beta-like), PC1/3⁺GCG⁺INS⁻ (alpha-like), PC1/3⁺GCG⁺INS⁺ (pre-alpha-like), and PC1/3⁺GCG⁻INS⁻ (immature endocrine) (Figures 2B and 2C). Expression of PC2 and PC1/3 appeared to be mutually exclusive in SC-α cells (Figure S2E). Despite the presence of PC1/3⁺GCG⁻INS⁺ cells in the SC-α population, insulin secretion was not detected (Figure S2F).

Figure 2.

SC-α cells as an in vitro model to explore proglucagon processing

(A) Immunofluorescence staining showing PC2 (green), GCG (red), and DAPI (blue) in human islets (left) and SC-α cells (right). Scale bar, 50 μm.

(B) Immunofluorescence staining showing PC1/3 (green), GCG (red), INS (white), and DAPI (blue), in human islets (left) and SC-α cells (right). Scale bar, 50 μm.

(C) Immunofluorescence staining showing different PC1/3-expressing cells in SC-α cells (left to right), PC1/3⁺GCG⁻INS⁻, PC1/3⁺GCG⁻INS⁺, PC1/3⁺GCG⁺INS⁻, and PC1/3⁺GCG⁺INS⁺.

(D) Glucagon secretion in human islets (n = 3) and SC-α cells (n = 5) at 3.3 and 16.3 mM glucose.

(E) Glicentin secretion in human islets (n = 3) and SC-α cells (n = 5) at 3.3 and 16.3 mM glucose.

(F) GLP-2 secretion in human islets (n = 3) and SC-α cells (n = 5) at 3.3 and 16.3 mM glucose. Scale bar, 50 μm at 20× magnification.

(D–F) Unpaired Mann-Whitney comparison with p values indicating non-significant (ns) (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; and ^∗∗∗∗p < 0.0001. Error bars represent mean ± SEM.

To determine whether proglucagon cleavage products produced by PC2 and PC1/3 are released by HI and SC-α cells, we measured the secretion of peptides cleaved by PC2 (glucagon) and PC1/3 (glicentin, GLP-2, and GLP-1). Of these, glucagon was the most abundant (Figure 2D). Surprisingly, we observed measurable quantities of glicentin and GLP-2 in both human islets and SC-α cells (Figures 2E and 2F). Further refinement of these secretion studies using four concentrations of glucose revealed that glucagon secretion in SC-α cells followed a V-shaped curve; no glucose-dependent trends were observed for glicentin, GLP-2, and GLP-1 (Figure S2I). SC-α also secreted glucagon in response to arginine, which is consistent with known alpha cell physiology (Figure S2G).

GLP-1 peptide was detected at extremely low levels in both SC-α cells and human islets (Figures S2H and S2I). The dipeptidyl peptidase 4 (DPP4) enzyme, expressed by most islet cells, rapidly degrades GLP-2 and GLP-1 but not glucagon and glicentin (Acosta-Montalvo et al., 2020). We found that DPP4 protein colocalized with glucagon-expressing SC-α cells (Figures S2J and S2K). Since DPP4 may degrade GLP-1 and GLP-2, glicentin secretion is a more reliable readout of PC1/3 proglucagon processing.

ER stress results in glucagon hypersecretion in SC-α cells

Hypersecretion of glucagon is a well-known feature of T2D (Dinneen et al., 1995; Asadi and Dhanvantari, 2021). To investigate the ability of SC-α cells to model this phenotype, we subjected SC-α cells to the diabetogenic stressors hyperlipidemia (1 μM oleic acid), glucotoxicity (30 mM glucose), and endoplasmic reticulum (ER) stress (5 μM tunicamycin) (Figure 3A), which have previously been used to study beta cell dysfunction (Leite et al., 2022). Following a 48-h treatment, we quantified glucagon secretion in both low- and high-glucose conditions. ER stress was the most effective at increasing glucagon secretion (Figure 3B).

Figure 3.

ER stress established the hypersecretion of glucagon in SC-α cells

(A) Schematic of SC-α cells treated with vehicle (DMSO, n = 4), and physiological stressors of lipidemia (1 μM oleic acid; n = 3), glucotoxicity (30 mM glucose; n = 4), and ER stress (5 μM tunicamycin; n = 4) for 48 h in 30 mL bioreactor at 40 rpm.

(B) Glucagon secretion after 48 h of treatments with vehicle or physiological stressors at 3.3 and 16.3 mM glucose.

(C) Glucagon secretion after 48 h of treatments with vehicle or SERCA inhibiting ER stressor (thapsigargin: 100 nM) or N-glycosylation inhibiting ER stressor (5 μM tunicamycin) at 3.3 and 16.3 mM glucose.

(D) Relative mRNA expression (2^−ddCt) of GRP78, IRE1-α, PERK, and ATF6 in SC-α cells treated with vehicle or TM (n = 5).

(B and C) Mixed Model ANOVA with Dunnett’s multiple comparison test and (D) two-tailed t test (unpaired) with p values indicating non-significant (ns) (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; and ^∗∗∗∗p < 0.0001. Error bars represent mean ± SEM.

To confirm that an optimal ER stress induction method had been identified, we compared the effect of two commonly used ER stressors, tunicamycin (TM) and thapsigargin (TG), on SC-α cells (Liang et al., 2006; Weldemariam et al., 2022). Only TM resulted in increased glucagon secretion (Figure 3C). Quantitative reverse-transcription PCR (RT-PCR) analysis confirmed that treatment with TM, but not TG, resulted in robust upregulation of ER stress response genes (Figures 3D and S3A) (Oslowski and Urano, 2011). XBP-1, a downstream target of the ER stress sensing protein, IRE1-α, was also more downregulated by TM than TG (Figure S3B). Across a range of doses, we found that 5 μM TM resulted in hypersecretion of glucagon without compromising cell viability (Figures S3C and S3D), and this drug and dosage were selected for further use.

ER stress increases PC1/3-mediated processing of proglucagon in SC-α cells

PC1/3-mediated proglucagon processing is increased in diabetes (Campbell et al., 2020; Cani et al., 2005). Because its stability is not impacted by DPP4, we analyzed glicentin secretion to explore the effect of TM-induced ER stress on PC1/3-mediated proglucagon processing. We measured secreted glicentin in vehicle-treated and ER-stressed SC-α cells. ER stress increased glicentin secretion from SC-α cells at both low and high glucose (Figures 4A and S4A), with the maximal effect observed at 5 μM TM (Figure S4D). Secretion of GLP-1 and GLP-2 was not affected (Figures S4B and S4C). To investigate whether ER stress increased the secretion of glicentin and glucagon equally, we compared the ratio of glicentin to glucagon in the KCl supernatants of matched SC-α cells. The ratio of glicentin to glucagon was higher in ER-stressed SC-α cells than in vehicle-treated SC-α cells (Figure 4B), indicating that ER stress increased the proportion of proglucagon processed by PC1/3.

Figure 4.

ER stress increased PC1/3-mediated proglucagon processing peptides in SC-α cells

(A) Glicentin secretion at 3.3 and 16.3 mM glucose in vehicle-treated and ER-stressed SC-α cells (n = 5).

(B) Bar graph showing the ratio of glicentin secretion to glucagon secretion in KCl-induced depolarized samples of vehicle-treated and ER-stressed SC-α cells (n = 10).

(C) Immunofluorescence staining for total GLP-1⁺ (green) and DAPI (blue) cells in vehicle-treated (top) and ER-stressed SC-α cells (bottom). Scale bar, 50 μm.

(D) Quantification of total GLP-1⁺ cells normalized to the area of the section (n = 6).

(E) Immunofluorescence staining showing PC1/3 (green), GCG (red), INS (white), and DAPI (blue). Scale bar, 50 μm.

(F) Quantification of PC1/3⁺GCG⁺INS⁻cells normalized to the area of the section (n = 4).

(G) Quantification of PC1/3⁺GCG⁻INS⁻ cells normalized to the area of the section (n = 4).

(A) Two-way ANOVA with Sidak’s multiple comparison test, (B) Ratio paired t test, (D, F, and G) paired two-tailed t test with p values indicating non-significant (ns) (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; and ^∗∗∗∗p < 0.0001. Error bars represent mean ± SEM.

Because no suitable antibody exists to detect glicentin by immunofluorescence, we used alternative approaches to evaluate PC1/3-mediated proglucagon processing at the level of an individual cell. First, we performed immunostaining for GLP-1 (Figures 4C and S4E) and found that ER stress increased the total number of GLP-1⁺ cells compared to the vehicle (Figure 4D). Next, we quantified PC1/3⁺GCG⁺INS⁻ and PC1/3⁺GCG⁻INS⁻ cells in ER stress vs. control conditions (Figures 4E–4G and S4F). The number of cells expressing PC1/3 was higher in ER-stressed SC-α cells, and ER-stressed SC-α cells exhibited approximately 2-fold more PC1/3⁺GCG⁻INS⁻ cells than the vehicle-treated SC-α cells. Together, these findings show that the ER stress increased alternate proglucagon processing in SC-α cells, as is observed in alpha cells from diabetic patients.

ER stress establishes T2D metabolic signatures in SC-α cells

To gain a broader understanding of the impact of ER stress on SC-α cells, we performed bulk RNA sequencing (RNA-seq). Compared to vehicle-treated cells, ER-stressed SC-α cells displayed 199 upregulated and 179 downregulated genes (log₂FC >│1│, FDR< 0.05) (Figures 5A and S5A). Most of the upregulated genes were involved in the unfolded protein response (UPR) (Chen et al., 2023). Of interest, the upregulated genes included glucose sensors (GPR119 and GPR146), amino acid transporters (SLC7A1, SLC7A11, and SLC7A5), mitochondrial respiration genes (PDK4), PCSK1 regulators (CREB3L1), and alpha cell identity genes (PROM1) (Greenwood et al., 2020; van Gurp et al., 2019). Downregulated genes included transcription factors associated with pancreatic cell lineage (PTF1A, PAX4, NEUROG3, and PDX1) and glucose sensing/tolerance (IRS2, ADH1B, and PPARG).

Figure 5.

ER stress established T2D metabolic signatures in SC-α cells

(A) Volcano plot of differentially expressed genes in bulk RNA sequencing in vehicle-treated vs. ER-stressed SC-α cells (n = 6 per group, FDR < 0.05, log₂FC > |1|).

(B) Negatively enriched biological processes in ER-stressed SC-α cells, (x axis: normalized enrichment score [NES]; y axis ranked by adjusted p values).

(C) Enriched KEGG pathways in ER-stressed SC-α cells (x axis: NES; y axis ranked by adjusted p values).

(D) Relative mRNA expression (2^−ddCt) of transcripts associated with alpha cell identity normalized to the vehicle (n = 4).

(D) Unpaired t test with p values indicating non-significant (ns) (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; and ^∗∗∗∗p < 0.0001. Error bars represent mean ± SEM.

To determine the implications of these changes, we conducted five different gene set enrichment analyses covering Gene Ontology biological processes (GOBP), molecular function (GOMF), cellular compartments (GOCC), Kyoto Encyclopedia of Genes and Genomes (KEGG), and REACTOME. Across methodologies, we found 207 terms exhibiting positive enrichment and 212 terms showing negative enrichment in ER-stressed SC-α cells (Table S2). Positively enriched terms and pathways were related to unfolded protein response. Additionally, amino acid transport pathways were consistently enriched in all analyses (lead genes: SLC7A3, SLC1A4, and SLC38A1) (Table S2).

Conversely, negatively enriched pathways were associated with metabolism and cellular respiration. GOBP showed negative enrichment of aerobic respiration, oxidative phosphorylation, and mitochondrial electron transport complex (Figure 5B, lead genes: ENO3, LDHA, and PKM). KEGG pathway pointed to negative enrichment of glycolysis and mitochondrial respiration in ER-stressed SC-α cells (Figure 5C). Negatively enriched pathways in REACTOME analysis included the tricarboxylic acid cycle, respiratory electron transport, cellular response to hypoxia, and NOTCH4 signaling (Table S2). These findings indicated that ER stress can establish diabetes-like changes in SC-α cells with downregulation of glycolysis, oxidative phosphorylation, and mitochondrial respiration (Bosi et al., 2022).

To directly query the effects of ER stress on proglucagon processing and the regulation of alpha cell identity, we performed RT-PCR on proglucagon processing-related genes (GCG, PCSK1, and PCSK2) and alpha cell identity-related genes (ARX, IRX1, IRX2, and MAFB) in vehicle-treated and ER-stressed SC-α cells. This analysis revealed downregulation of MAFB in the ER-stressed SC-α cells (Figure 5D) (Conrad et al., 2016; Katoh et al., 2018). Because MAFB is upstream of other key pancreatic transcription factors, this suggests a possible mechanism for the downregulation of genes such as PTF1A, PAX4, NEUROG3, and PDX1 observed in our bulk RNA-seq experiments.

Sunitinib, a tyrosine kinase inhibitor, attenuates ER-stress-induced dysfunction in SC-α cells

Prior work in our laboratory identified sunitinib, a tyrosine kinase inhibitor, as a small molecule that increases the number of GCG⁺ cells produced by the SC-α protocol. Intriguingly, studies in human patients with T1D have found that sunitinib lowers blood glucose and supports insulin independence (Oh et al., 2012; Gitelman et al., 2021; Louvet et al., 2008; Huda et al., 2014). Based on these observations, we hypothesized that sunitinib might more generally promote alpha cell identity and function.

To test this hypothesis, we pretreated SC-α cells with either vehicle or sunitinib and subsequently exposed them to ER stress (Figure S6A). Compared to the vehicle-pretreated cells, sunitinib-pretreated cells had a higher percentage of GCG⁺ cells following ER stress (Figure S6B). To test whether sunitinib inhibited ER stress, we treated SC-α cells with either vehicle, sunitinib, or the ER stress inhibitor tauroursodeoxycholic acid (TUDCA) and compared mRNA expression related to UPR, proglucagon processing, and alpha cell identity. Unlike TUDCA, sunitinib treatment on SC-α cells did not reduce ER stress. However, TUDCA also downregulated alpha cell identity markers (ARX and MAFB), suggesting that the direct inhibition of ER stress could also be detrimental to SC-α cells (Figure S6C).

Next, we treated SC-α cells with vehicle, sunitinib, TM, or both TM and sunitinib and quantified glucagon secretion (Figure 6A). Sunitinib treatment alone did not impact glucagon secretion. However, SC-α cells treated with both TM and sunitinib had much lower glucagon secretion at both low glucose and high glucose than cells exposed to TM alone, suggesting that sunitinib treatment mitigated ER stress-induced glucagon hypersecretion (Figure 6B). SC-α cells treated with both TM and sunitinib also had lower total glucagon content than SC-α cells treated with TM alone (Figure S6D).

Figure 6.

Sunitinib, a tyrosine kinase inhibitor, attenuated ER-stress-induced dysfunction in SC-α cells

(A) Schematic of SC-α cells treated with vehicle, ER stressor (TM: 5 μM), sunitinib (25 μg), or ER stressor (TM: 5 μM) and sunitinib (25 μg).

(B) Glucagon secretion at 3.3 and 16.3 mM glucose challenge after 48 h (n = 6).

(C) Glicentin secretion at 3.3 and 16.3 mM glucose challenge after 48 h (n = 6).

(D) Relative mRNA expression (2^−ddCt) of transcripts associated with unfolded protein response, proglucagon processing, and alpha cell identity (n = 6).

(B and C) Two-way ANOVA with Tukey’s multiple comparison test, (D) Unpaired t tests with p values indicating non-significant (ns) (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; and ^∗∗∗∗p < 0.0001. Error bars represent mean ± SEM.

We next tested whether sunitinib protected SC-α cells against ER stress-induced defects in proglucagon processing. Glicentin secretion was attenuated in SC-α cells treated with both TM and sunitinib compared to TM alone (Figure 6C). Numbers of GLP-1⁺ cells and PC1/3⁺GCG⁻INS⁻ cells (Figure S6E-H) were not significantly impacted.

Lastly, we used RT-PCR to evaluate markers of the UPR, proglucagon processing, and alpha cell identity in this system (Figure 6D). The addition of sunitinib did not change UPR and proglucagon processing markers. Interestingly, the alpha cell identity makers IRX1 and MAFB were upregulated in TM and sunitinib co-treated cells compared to TM alone. Together, these results identify sunitinib as a promising compound that reinforces alpha cell identity and attenuates ER stress-induced glucagon hypersecretion in SC-α cells.

Discussion

Current models for studying alpha cells (Campbell et al., 2020; Riahi et al., 2023; Tellez et al., 2020) are limited by issues of abundance and relevance to clinical translation. Here, we demonstrate the utility of our scalable differentiation protocol for the modeling of human alpha cell biology in normal and disease states. The glucagon-positive SC-α cells produced by the protocol have a similar transcription profile to primary human alpha cells and secrete glucagon upon glucose or arginine challenge (van Gurp et al., 2022). Induction of ER stress in these cells results in phenotypes congruent with alpha cells from diabetic patients, including glucagon hypersecretion, alterations in proglucagon processing, and changes in transcription profile.

Multiple lines of evidence indicate the relevance of ER stress in the pathophysiology of diabetic alpha cells. The diabetic environment results in the production of reactive oxygen species and the induction of ER stress (Eizirik et al., 2008; Ghemrawi et al., 2018), with increased expression of ER stress markers such as GRP78 and XBP1 in the islets of patients with T2D (Hakonen et al., 2018; Papa, 2012). In beta cells, ER stress results in impaired proinsulin processing and improper insulin secretion (Leite et al., 2020; Mustapha et al., 2021; Chen et al., 2022; Brusco et al., 2023; Iida et al., 2023; Meier et al., 2022). Here, we find that ER stress results in the hypersecretion of glucagon by SC-α cells, generating the first in vitro human cell model for a phenotype observed in patients with T1D and T2D (Omar-Hmeadi et al., 2020; Riahi et al., 2023).

The induction of ER stress also led to an increase in PC1/3-mediated proglucagon processing. Previous studies have reported that proglucagon processing by PC1/3 is increased in T2D islets (Campbell et al., 2020; Cani et al., 2005). PC1/3-expressing alpha cells improve glucose handling and lower fasting glucose levels in high-fat-fed and db/db mice (Wideman et al., 2009). Furthermore, GLP-2 (a PC1/3 cleavage product) is secreted by human islets and is increased by diabetogenic stressors including glucose and palmitate (He et al., 2021). Our work has demonstrated that glicentin is more amenable to mechanistic studies than other PC1/3-mediated proglucagon cleavage products because its longer half-life allows for a more accurate measurement of its production and secretion from alpha cells. The use of the SC-α cell model system, coupled with glicentin as a readout, will enable a more detailed understanding of altered proglucagon processing in diabetes.

SC-α cells provide a model to assess the impact of various perturbations on the alpha cell transcription profile. Induction of ER stress resulted in the downregulation of MAFB (Artner et al., 2010; Brissova et al., 2018), a phenotype observed in diabetic alpha cells (Bosi et al., 2022). Islets require ATP-regulated potassium channels for accurate glucose responsiveness (Rorsman et al., 2014). Altered glycolytic pathway and oxidative phosphorylation can disrupt the regulation of these channels and establish mitochondrial dysfunction (Haythorne et al., 2022), leading to impaired glucagon secretion and defects in glucose responsiveness (Grubelnik et al., 2020; Zhang et al., 2013; Grubelnik et al., 2020). Upregulation in stress markers such as DDIT3 has been reported as transcriptomic signatures of alpha cells in both patients with T1D and T2D (Bosi et al., 2022). In response to ER stress, SC-α cells also displayed alterations in metabolic pathways such as glycolysis, aerobic respiration, and oxidative phosphorylation. These similarities to primary human islet alpha cells isolated from diabetic patients enable the use of our in vitro system for the study of alpha cell dysfunction in diabetes (Doliba et al., 2022; Bosi et al., 2022).

Finally, we have shown that our system can be used to identify small molecules that correct ER stress-induced phenotypes. Tyrosine kinase inhibitors have recently been reported to correct hyperglycemia in diabetic patients (Agostino et al., 2011; Althubiti 2022). While one study has reported that sunitinib acts directly via beta cells (Lutz et al., 2017), another reported similar effects of sunitinib in NOD mice, in which the beta cells have been largely destroyed (Louvet et al., 2008). Sunitinib also lowers blood glucose in type 2 diabetic rats (Mahdi et al., 2022). In our study, sunitinib reduced glucagon hypersecretion and increased transcript levels of MAFB and IRX2 under ER stress (Katoh et al., 2018; Shrestha et al., 2021). Expansion of these studies using additional tyrosine kinase inhibitors will refine our understanding of the interplay between reduced hypersecretion of glucagon and reduced processing of GLP-1, highlighting the utility of this approach in regulating alpha cell identity and function.

In summary, we report that SC-α cells are similar to human alpha cells and are a promising tool to study alpha cell function and dysfunction. Here, we have modeled alpha cell dysfunction with ER stress and demonstrated the ability of a tool compound to partially rescue the resultant diabetic phenotypes. These results will serve as a foundation for in-depth investigations of alpha cell dysfunction and its resolution, with the promise of improved therapeutic options for diabetic patients.

Methods

Embryonic stem cell culture

HUES8-NIHhESC-09-0021 was used to generate 3D spheroids of SC-α cells for all the experiments in this study (Peterson et al., 2020). HUES8 cells (1.5e8) were seeded in mTeSR 1 (STEMCELL Technologies) with 10 μΜ Rhoki (Y27632, Tocris) in 500 mL spinner flasks (Corning). Flasks were maintained at 70 rpm, 37°C, and 5% CO₂. mTeSR1 media (no Y27632) was replenished at 48 h. At 72 h, directed differentiation was initiated. Differentiation to SC-α cells was conducted according to our previously published protocol (Peterson et al., 2020) over the course of 28 days in 6 stages.

Flow cytometry

3D spheroids were collected, washed with PBS, incubated in TrypLE at 37°C for 10 min, dispersed into single cells, and quenched with a complete medium. Dispersed cells were centrifuged at 200g for 5 min, washed with PBS, and fixed using 4% paraformaldehyde (PFA) for 1 h at room temperature. Fixed samples were dispensed through a 35 μm filter, washed once with PBS, and incubated in blocking buffer (PBS + 0.1% Triton X-100 + 5% donkey serum) for 1 h at room temperature. After blocking, samples were incubated with primary antibody at 4°C overnight. Cells were washed three times with PBST (PBS + 0.1% Triton X-100) and incubated in secondary antibody solutions at 4°C for 1 h. After three washes with PBST, samples were suspended in PBS, and data were collected using an Attune NXT instrument. Data were analyzed using FlowJo 10.8.1.

scRNA sequencing

scRNA-seq was performed on 10× Genomic Chromium system using the 3′ v.3 library kit. Three independent differentiations of SC-α cells were collected at stage 6, day 28 for library preparation. Samples were sequenced on the HiSeq 4000 platform. Sequenced samples were aligned and annotated using Cell Ranger (v.4.0.0). Cell Ranger outputs were analyzed in Seurat (v.5.1.0) (Hao et al., 2024). Data from 12 human donors were obtained from Gene Expression Omnibus (GEO: GSE114297) (Xin et al., 2018). Cells with less than 500 detected genes, more than 25% mitochondrial content, or less than 50,000 RNA transcripts were removed from the data before normalization and integration. All datasets were integrated into a single object with IntegrateLayers() using reciprocal principal component analysis. UMAP reduction was performed using the first 30 dimensions, and clusters were manually identified based on gene expression (Lawlor et al., 2017; Stuart et al., 2019).

Immunofluorescence

Primary human islets (BioChain) or SC-α cell clusters were fixed with 4% PFA, washed three times with PBS, embedded in Histogel, and subsequently embedded in paraffin. For immunostaining, 4 μM slide-mounted sections were deparaffinized. Antigen retrieval was performed by boiling slides in sodium citrate buffer at pH 6.5 and followed by 45 min at RT. Samples were incubated with blocking buffer in a humidifying chamber for 1 h at RT, followed by incubation with primary antibodies overnight at 4°C. The next day, after multiple washes with PBST, slides were incubated in secondary antibodies and DAPI (4′,6-diamidino-2-phenlindole) for 2 h at 4°C. Slides were washed with PBST, mounted using Fluoromount-G, and imaged and analyzed with Zeiss Widefield Microscope and Zen2 software. Cell counts and area of each section were quantified using Zen software, and cell counts were normalized to the total area of the section.

Hormone secretion

SC-α cells or human islets (Prodo) were treated with vehicle, stressor, or sunitinib in 30 mL bioreactors maintained at 40 rpm at 37°C for 48 h. Cells were then washed twice with high-glucose (16.3 mM) Krebs Ringer Buffer (KRB) (128 mM NaCl, 5 mM KCl, 2.7 mM CaCl₂, 1.2 mM MgSO₄, 1 mM Na₂HPO₄, 1.2 mM KH₂PO₄, 5 mM NaHCO₃, 10 mM HEPES, and 0.1% BSA), followed by a 1-h equilibration in that buffer in 24-well ultra-low attachment plate with transwell inserts. Samples were then incubated for 1 h in either low (3.3 mM) or high-glucose (16.3 mM) KRB. Finally, cells were transferred to KRB containing 3 mM KCl for 1 h. Hormone detection was performed by ELISA: glucagon (R&D Systems); GLP-2 (Eagles-Biosciences); and glicentin, GLP-1, and insulin (Mercodia).

RT-PCR

RNA was isolated using the RNeasy Mini Kit RNA (QIAGEN) and quantified using NanoDrop One (Thermo Scientific). 400 ng of RNA was used to generate cDNA using the iScript cDNA synthesis kit. Primers used (Integrated DNA Technologies) are listed in Table S1. RT-PCR was performed on a Lightcycler96 (Roche) using SYBR green PCR master mix (Applied Biosystems). Relative gene expression was assessed using the 2^-ΔΔCt method.

Bulk RNA-seq

Samples were submitted in two batches to GENEWIZ for RNA-seq, each containing three replicates of vehicle-treated and tunicamycin-treated SC-α cells. GENEWIZ further assessed RNA quality using Qubit RNA Assay and TapeStation analyses. Samples that passed RNA quality metrics (RIN ≥ 6 and DV200 ≥ 70) were used for library preparation with polyA removal. RNA-seq was performed on an Illumina HiSeq instrument (paired-end, 150 base pair) with >16 million reads per sample. Quality standards for RNA-seq samples were a mean quality score of >35 and a Q30 score of >90. Adapter content was trimmed from RNA-seq reads using cutadapt (v.4.4) and confirmed using FastQC (v.0.12.1). Sequencing reads were mapped to the hg38 human reference genome using HISAT2 (v.2.1.0), counted using featureCounts (v.2.0.6), and stored in a count matrix. The count matrices of the two RNA-seq batches were combined to create a dataset with 6 samples in each experimental group. Batch was defined as a factor in an additive linear model, and differential expression comparisons were performed between the treatment groups. The edgeR package (v.3.42.4) was used to normalize the gene count matrix to counts per million reads (CPM) with the trimmed mean of the M-values method. Genes were filtered such that only genes expressed with a minimum of 1 CPM in 50% or more samples were used in analyses. Differential analysis was performed with the quasi-likelihood F-test. Differentially expressed genes were considered statistically significant with a false discovery rate (FDR) adjusted p value <0.05 and a log₂FC >│1│. Volcano plots of differentially expressed genes were generated with EnhancedVolcano (v.1.18.0). Heatmaps of normalized gene expression with dendritic clustering were generated with ComplexHeatmap (v.2.16.0) with default parameters. Gene set enrichment analysis (GSEA, v.4.3.2) was performed using GSEApy (v.1.13) to assess the enrichment of KEGG pathways, Gene Ontology (GO): biological processes, GO: molecular functions GO: cellular components, and Reactome gene sets. GSEA analysis was performed using gene set permutations and associated default settings. Enriched gene sets met the significance threshold of FDR <0.05.

Statistics

Data represented are mean ± standard error of the mean. Each data point is an independent differentiation of SC-α cells that was tested for at least three technical replicates. Analysis was performed in GraphPad Prism v.10.2.3 software using one-way ANOVA, two-way ANOVA, or t test with p values indicating non-significant (ns); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; and ^∗∗∗∗p < 0.0001.

Resource availability

Lead contact

Further information and requests should be addressed to the lead contact, Quinn P. Peterson (peterson.quinn@mayo.edu).

Materials availability

This study did not generate any new material or reagent.

Data and code availability

All original data are deposited at GEO: GSE293267, scRNA sequencing (GEO: GSE293265), and bulk RNA sequencing (GEO: GSE293266) and are publicly available as of the date of publication. The original codes are available upon request.

Acknowledgments

The authors thank Zenith Khashim, Michael Slama, and Anna-Marie Schornack for technical assistance with different methods used in the manuscript, Dena E. Cohen for editorial contributions during manuscript preparation, and all members of Peterson lab for their assistance in embryonic stem cell cultures and helpful discussions. S.S. was partially supported by the Anita and Gary Klesch Fellowship.

Author contributions

S.S. and Q.P.P. conceptualized and designed all the experiments and prepared the manuscript. S.S. performed all the experiments and statistical analysis. L.T.J. contributed to the sample preparation and processing and analysis of bulk RNA sequencing data. K.K. contributed to analysis of single-cell RNA sequencing data. S.B.S. contributed to analysis of insulin secretion. All authors have read and approved the final manuscript.

Declaration of interests

Q.P.P. serves on the scientific advisory board of Mellicell, Inc., and is listed as an inventor on intellectual property licensed by Vertex.

Published: May 8, 2025