Abnormal exocrine-endocrine cell crosstalk promotes β-cell dysfunction and loss in MODY8

Abstract

MODY8 (maturity-onset diabetes of the young, type 8) is a dominantly inherited monogenic form of diabetes associated with mutations in the carboxyl ester lipase (CEL) gene expressed by pancreatic acinar cells. Patients carrying the mutation develop childhood-onset exocrine pancreas dysfunction followed by the manifestation of diabetes during adulthood. However, it is unclear how CEL mutations cause diabetes. Here we report the transfer of proteins from acinar cells to β-cells as a form of crosstalk between exocrine and endocrine cells. Human β-cells show a relatively higher propensity for internalizing the mutant versus the wild-type CEL protein. Following internalization, the mutant protein forms stable intracellular aggregates leading to β-cell secretory dysfunction. Analysis of pancreas sections from a MODY8 patient revealed the presence of CEL protein in the few extant β-cells. This study provides compelling evidence for the mechanism by which a mutant gene expressed specifically in acinar cells promotes dysfunction and loss of β-cells to cause diabetes.

Introduction

The exocrine pancreas is composed of acinar cells that secrete digestive enzymes and ductal cells secreting bicarbonate and fluid into the duodenum. In contrast, the endocrine compartment, constituting ~2% of the pancreas, primarily consists of cells that secrete hormones into the circulation. While traditionally the endocrine and exocrine pancreas have been considered as two separate entities, the close anatomical relationship between the two tissues suggests they function in close coordination with each other. The regulatory interactions between the cells in the two compartments¹ are exemplified by the effects of insulin on acinar cell function^2,3, and conversely, of the effects of exocrine enzymes on islet cells^4–7 in animal models. Thus, it is not surprising that endocrine and exocrine diseases co-exist in the same patient⁸. For example, alterations in pancreatic exocrine function manifest in patients with diabetes⁹ and conversely an increased risk for diabetes is evident in patients with exocrine pancreatic diseases such as pancreatitis¹⁰, cystic fibrosis¹¹, or pancreatic cancer¹².

In MODY8, the development of exocrine insufficiency predates the development of impaired glucose intolerance and diabetes in adults¹³. This disease is characterized by lipomatosis of the pancreas, low fecal elastase levels, severe pancreatic dysfunction with compensated nutritional status¹⁴, and steatorrhea before the age of twenty followed by reduced pancreas volume, development of pancreatic cysts¹⁵, and diabetes later in life. MODY8 is caused by heterozygous mutations in the CEL gene, also known as bile salt-dependent lipase (BSDL)¹⁶. The causal mutations were found in the first or fourth repeating segment of the variable number of tandem repeats (VNTR) region of the CEL gene. Here, we focused on the mutation in the first repeating segment of VNTR (c.1686delT; p.C563fsX673)^13,16.

CEL is secreted mainly from pancreatic acinar cells and activated by bile salts in the duodenum. It hydrolyzes a wide variety of dietary lipids, cholesteryl esters, cholesterol, ceramides, triacylglycerides, phospholipids, lipid-soluble vitamins^16,17, and the newly discovered branched fatty acid esters of hydroxyl fatty acids (FAHFAs)¹⁸. Earlier studies showed that the mutant protein forms intracellular and extracellular aggregates indicating that MODY8 is a protein misfolding disease^19,20. The single-base deletions in the VNTR region of the CEL gene lead to a frameshift; the new positively charged C-terminal end alters the net charge of the mutant protein allowing it to aggregate¹⁹. The aggregated mutant protein was reported to be present in intracellular and extracellular compartments of HEK293 cells that overexpress the mutant gene in vitro^19,20. While a majority of the studies to date have focused on the effects of the mutant CEL protein in acinar and non-endocrine cells^20–22, a direct demonstration of how CEL mutations cause diabetes in MODY8 patients is lacking. Here, we used pancreatic cell lines, human stem cell-derived cellular models, and humanized mice to investigate the crosstalk between acinar and β-cells that potentially underlies the endocrine dysfunction in MODY8.

Results

Mutant CEL is transferred from acinar to β-cells

We began by testing whether the mutant protein expressed by acinar cells is secreted into culture media and taken up by neighboring human β-cells. To this end, we used a transwell system and co-cultured human pancreatic β-cell line (EndoC-βH1 cells) with acinar cells (266–6) overexpressing either the mutant (MUT) or the wild-type (WT) CEL gene. Acinar cells transfected with an empty vector (EV) were included as a negative control. Western blot data demonstrated that acinar cells were efficiently transfected and overexpressed either WT or MUT CEL gene (Extended Data Fig. 1a-c). In the co-culture model, the two cell types were separated by a membrane bearing small pores (0.4 μm) to allow for the transport of molecules²³ (Fig. 1a). The wild-type protein could be easily detected in culture media bathing the acinar cells expressing the wild-type CEL, compared to the smaller amount of mutant protein in media with the mutant CEL-expressing acinar cells (Fig. 1b, c). This was likely due to poor secretion and retention of the mutant CEL in the acinar cells (Extended Data Fig. 1d, e) consistent with the previous observation that the MODY8 mutation led to a secretion defect²¹. Although both wild-type and mutant proteins were successfully transferred from the acinar cells to β-cells, the amount of mutant protein, despite being less abundant in the media, was transferred significantly more into the β-cells than wild-type protein (Fig. 1c-d). These data indicate that β-cells have a relatively higher propensity towards internalizing the mutant compared to wild-type CEL. To exclude the possibility that contaminating acinar cells in our cultures were a direct source of the detected CEL protein, β-cells were treated with conditioned media obtained from 266–6 acinar cells transiently transfected with EV, WT, or MUT plasmids (Fig. 1e). The conditioned medium was filtered to remove contaminating cells and the concentration of CEL protein was equalized before treating β-cells (Extended Data Fig. 1f-h). We again observed that the amount of mutant protein transferred from acinar to β-cells was higher than wild-type protein (Fig. 1f-h).

Fig. 1 |. Mutant CEL protein is transferred from acinar cells to β-cells.

a, Mouse acinar cells (266–6) overexpressing empty vector (EV), wild-type CEL (WT), or mutant CEL (MUT) were co-cultured with human β-cells (EndoC-βH1) for 24 h in a transwell system. b, Western blots, representative of three independent experiments, showing the amount of V5-tagged human CEL protein variants in media and in β-cell lysates. Note that the mutant protein (~100 kDa) is smaller than the wild-type (~130 kDa) due to the frameshift mutation leading to a premature stop codon resulting in a truncated protein. c, d The quantification of three biologically independent samples. Levels of wild-type and mutant protein in media (c) and in β-cell lysates (d). CEL levels in the β-cell lysates were normalized to β-actin levels. Fold change relative to EV. Data are presented as mean values ± SEM. One-way ANOVA followed by Tukey’s multiple comparison test. e, Recipient cells (EndoC-βH1 β-cells) were treated for one hour with conditioned media collected from 266–6 acinar cells overexpressing EV, WT, or MUT CEL gene. f, Western blots, representative of five independent experiments, showing the amount of V5-tagged CEL protein in media and in β-cell lysates. g, h The quantification of five biologically independent samples. Levels of wild-type and mutant protein in media (g) and in β-cell lysates (h). CEL levels in the β-cell lysates were normalized to β-actin levels. Fold change relative to EV. Data are presented as mean values ± SEM. One-way ANOVA followed by Tukey’s multiple comparison test. Dashed line is added for enhanced comprehension (b, f).

We then used conditioned media obtained from HEK293 cells stably transfected with EV, WT, or MUT plasmids rather than acinar donor cells hereafter since acinar conditioned media contains proteases that could adversely affect cell viability²⁴. Fluorescence-activated cell sorting (FACS) further demonstrated that a significantly higher number of β-cells was taking up the mutant protein compared to wild-type protein 1 h after treatment with conditioned media obtained from HEK293 donor cells stably transfected with EV, WT, or MUT plasmids (Fig. 2a).

Fig. 2 |. Intracellular localization of mutant CEL protein after uptake by β-cells.

a, Percentage of V5+ (CEL+) β-cells detected by FACS one hour after treatment with conditioned media (n=3 biologically independent samples). Data are presented as mean values ± SEM. One-way ANOVA followed by Tukey’s multiple comparison test. b, Confocal images, representative of three biologically independent samples, showing V5 (CEL) immunostaining in β-cells treated with conditioned medium obtained from donor cells for one hour. Nuclei stained with DAPI (blue) and V5-tagged CEL stained in green. Boxed areas are shown at higher magnification in lower panels. Scale bars are 20 μm (upper panels) and 10 μm (lower panels). c, Confocal images, representative of three biologically independent samples, showing EndoC-βH1 cells stained with Proteostat fluorescent dye (red) after overnight treatment with HEK293 donor conditioned medium. MG132 (5 μM) treated cells were used as positive control. Nuclei stained with DAPI (blue). Boxed areas are shown at higher magnification in lower panels. Scale bars are 20 μm (upper panels) and 10 μm (lower panels). d, Quantification of number of aggresomes per cell in β-cells treated with donor conditioned medium treated cells (n=3 biologically independent samples). Data are presented as mean values ± SEM. One-way ANOVA followed by Tukey’s multiple comparison test. e, Electron micrographs, representative of two independent experiments, show EndoC-βH1 cells exposed to conditioned media obtained from HEK293 donor cells overexpressing EV, WT, or MUT protein for one hour. Cells were processed for immunoperoxidase electron microscopy using the anti-CEL As20.1 antibody. Abbreviations: PM, plasma membrane; N, nucleus; G, granule; M, mitochondria. Scale bar is 0.5 μm. f, Confocal images, representative of three biologically independent samples, show EndoC-βH1 cells exposed to mutant conditioned media for an hour. LAMP1+ vacuoles in red and CEL protein in green. Scale bar is 10 μm. The insets show the boxed areas in additional 1.6X magnification.

Microscopic analysis revealed intracellular aggresomes of the mutant protein compared to a homogenous distribution of the wild-type protein in the cytoplasm of β-cells (Fig. 2b). The aggresomes were visualized by staining with the thioflavin T derivative, Proteostat, which specifically binds to misfolded proteins²⁵ (Fig. 2c). β-cells treated with the proteasome inhibitor MG132, which induces protein misfolding, served as a positive control. We observed multiple Proteostat-positive structures in the cytoplasm of β-cells exposed to mutant proteins compared to β-cells exposed to media obtained from EV or WT donor cells (0.82 ± 0.07 versus 0.1 ± 0.01 and 0.23 ± 0.05 aggresome per β-cell, respectively) (Fig. 2d). Electron and confocal microscopy analysis of β-cells exposed to conditioned media from mutant cells localized the mutant CEL proteins to intracellular vacuoles (Fig. 2e, f). Taken together, these findings indicate that the intracellular mutant CEL proteins accumulate as insoluble aggregates within the cytoplasm of β-cells.

Next, we undertook experiments to determine the mechanism underlying the transfer of CEL protein (Extended Data Fig. 1i, j). We noted that cold treatment, a classical endocytosis blocker, as well as treatment with the PI3K inhibitor Wortmannin²⁶ (Extended Data Fig. 1k) each independently reduced internalization of both wild-type and mutant proteins significantly and confirmed that internalization of CEL protein is mediated by endocytosis. In contrast, treatment of donor cells with an inhibitor of neutral sphingomyelinase, GW4869²⁷, to reduce exosome release, did not alter the levels of either the wild-type or mutant proteins in media, ruling out a mediation of the transfer by exosomes (Extended Data Fig. 1l). These results indicate that the mutant protein was internalized by endocytosis and formed intracellular aggregates within the cytoplasm of β-cells.

Mutant CEL accumulates in β-cells as an insoluble fraction

Next, we examined the internalization kinetics of the secreted proteins in β-cells exposed to conditioned media obtained from HEK293 donor cells overexpressing either wild-type or mutant CEL gene over time (Fig. 3a). Both wild-type and mutant proteins were detected in the detergent-soluble fractions of the total cell lysates within minutes after treatment with conditioned media. Importantly, the EndoC-βH1 cells also internalized the mutant protein effectively compared to wild-type protein (Fig. 3b, c). The accumulation of mutant protein in the soluble fraction of β-cells peaked at 6 h, and then decreased, possibly due to degradation of internalized proteins in the lysosomes after entering the endosomal pathway²⁰ (Fig. 3c). In contrast, a significant amount of mutant protein remained even after 12 h, suggesting it accumulates as an insoluble fraction after internalization by β-cells (Fig. 3d).

Fig. 3 |. Mutant CEL protein forms stable insoluble aggregates after internalization by pancreatic β-cells.

a, Time course treatment of EndoC-βH1 β-cells with conditioned media obtained from HEK293 donor cells stably transfected with WT or MUT plasmids. b, Western blots, representative of three independent experiments, showing CEL protein levels in detergent-soluble and detergent-insoluble fractions of β-cells exposed to donor conditioned medium for various time durations. α-tubulin was used as a loading control. c, d The quantification of three biologically independent samples. Fold change relative to WT 1 min. Data are presented as mean values ± SEM. Two-tailed multiple t-tests. e, EndoC-βH1 β-cells treated with HEK293 donor conditioned medium for one hour, washed with DPBS twice, and cultured in CEL-free growth medium over time to determine degradation kinetics of mutant protein. f, Western blots, representative of three independent experiments, show CEL protein levels in detergent-soluble and detergent-insoluble fractions of β-cells 0–24 h after removal of donor conditioned medium. g, h The quantification of three biologically independent samples. Fold change relative to WT 0 h. Data are presented as mean values ± SEM. Two-tailed multiple t-tests. Dashed line is added for enhanced comprehension (b, f).

To determine the degradation kinetics of internalized wild-type versus mutant protein, β-cells were exposed to conditioned media obtained from HEK293 donor cells for an hour, washed twice, and then maintained in growth media for 24 h (Fig. 3e). As evident in the detergent-soluble fractions of the total cell lysates, EndoC-βH1 cells internalized more mutant protein than wild-type protein 1 h after conditioned medium treatment. Internalized wild-type and mutant proteins were degraded by β-cells and completely disappeared from the soluble fraction 2 h and 6 h after removal of conditioned media, respectively (Fig. 3f, g). In contrast to the barely detectable wild-type protein in the insoluble fraction, the mutant protein accumulated and remained stable within the insoluble fraction for more than 24 h (Fig. 3f, h). Using Proteostat dye we observed that only mutant proteins formed aggregates after internalization by β-cells (Extended Data Fig. 1m). These data suggest that mutant proteins that accumulate in the insoluble fraction are resistant to degradation, whereas they are degraded and eliminated in the soluble fraction.

Mutant CEL protein impairs function of β-cells

Progressive accumulation of protein aggregates is linked to the pathogenesis of several neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease, and Huntington’s disease²⁸. To test the hypothesis that dysfunctional protein clearance and accumulation of insoluble aggregates in pancreatic β-cells are involved in the development of diabetes in MODY8 we assessed the effects on cell viability. We therefore cultured β-cells for 10 days in conditioned media obtained from HEK293 donor cells overexpressing either the MUT or the WT CEL gene (Fig. 4a) and then determined the number of viable cells by quantifying intracellular ATP levels²⁹. EndoC-βH1 cells treated with conditioned medium obtained from empty vector-transfected HEK293 donor cells were included as control. While exposure to wild-type conditioned media for 10 days did not significantly alter the viability of EndoC-βH1 β-cells, the exposure to mutant protein led to fewer viable β-cells compared to the WT and EV groups (Fig. 4b). Interestingly, proliferation was also significantly reduced in the β-cells exposed to mutant conditioned media compared to the WT and EV groups (Fig. 4c), but apoptosis was not altered (Extended Data Fig. 2a). Consistent with the decreased proliferation, several key cell-cycle genes (CDK1, CDK2, CDK4)³⁰ were downregulated (Fig. 4d) while an increase in p16 and a decrease in p21 protein levels pointed to cellular senescence³¹ in the β-cells exposed to the mutant protein (Fig. 3e, f). RNA sequencing validated that several genes involved in cellular senescence were altered in β-cells exposed to mutant conditioned media (Extended Data Fig. 2b).

Fig. 4 |. Accumulation of mutant CEL protein aggregates reduces proliferation and impairs function of β-cells.

a, EndoC-βH1 β-cells were cultured for 10 days in conditioned media obtained from HEK293 donor cells stably transfected with EV, WT, or MUT plasmids. b, Cell viability was determined by measuring intracellular ATP levels at the end of 10 day-treatment period (EV n=4, WT n=6, MUT n=6 biologically independent samples). Fold change relative to EV. c, Percentage Ki67+ proliferating cells of β-cells (n=3 biologically independent samples). d, Gene expression levels of cell-cycle genes in β-cells. e, f Western blot data showing p16 (e) and p21 (f) levels in β-cells. g, h Gene expression levels of β-cell identity (g) and β-cell function (h) genes in β-cells. i, Glucose-stimulated insulin secretion (GSIS) assay. Fold change relative to EV 3.3 mM (n=8 biologically independent samples). j, The stimulation index was calculated as the fold increase in insulin release measured in 16.7 mM over 3.3 mM glucose (n=8 biologically independent samples). k, Gene expression levels of spliced XBP1/unspliced XBP1 in β-cells. l-o Western blot data showing pIRE1α (l), BiP (m), GRP94 (n), peIF2α (o) levels in β-cells. o, Schematic drawing of the proposed mechanism of mutant protein uptake by β-cells and accumulation of aggregates leading to several changes in β-cells. Data are presented as mean values ± SEM. One-way ANOVA followed by Tukey’s multiple comparison test applied to b, c, e, f, j, l-o, Two-tailed multiple t-tests applied to d, g, h, Two-way ANOVA followed by Sidak’s multiple comparison test applied to i, and one-tailed t-test applied to k. The gene expression analysis in d, g, h, k show fold change relative to EV (n=4 biologically independent samples). Dotted line indicates gene expression levels in EV-treated control β-cells set to 1 (d, g, h, k). β-actin was used as a housekeeping gene control. Western Blot data in e, f, l-o show fold change relative to EV (n=4 biologically independent samples). Protein level is normalized to α-tubulin. Dashed line is added for enhanced comprehension (e, f, l-o).

Senescence during aging is known to impair the expression of genes related to β-cell identity and function in rodent β-cells³². Consistently, PAX6, NKX2.2, INS1 were all significantly downregulated and expression of the transcription factors, NKX6.1, PDX1, MAFA also tended to be downregulated in the β-cells treated with mutant conditioned media compared to treatment with wild-type conditioned media (Fig. 4g). Additionally, among the critical β-cell function genes, ABCC8³³ expression was significantly decreased while others (SLC30A8, GCK) showed a trend towards reduction (Fig. 4h). Consistently, RNA sequencing data indicated that genes involved in insulin secretion were significantly altered in β-cells exposed to mutant conditioned media (Extended Data Fig. 2c). Of functional relevance, long-term culture of β-cells in mutant conditioned media showed a failure to respond to glucose stimulation compared to WT or EV groups (Fig. 4i, j). Furthermore, basal insulin levels tended to be higher in the mutant group (Fig. 4i), a finding that is consistent with senescent β-cells³⁴. To investigate whether impairment of glucose-stimulated insulin secretion was associated with altered mitochondrial biology³⁵, we analyzed the expression of genes involved in cellular energy metabolism and measured oxygen consumption rates (OCR). Indeed, expression of phosphoenolpyruvate carboxykinase 2 (PCK2)³⁶ and enolase 2 (ENO2)³⁷, genes involved in glycolysis with implications for insulin secretion were decreased and several genes involved in oxidative phosphorylation were significantly altered in β-cells exposed to mutant conditioned media (Extended Data Fig. 2d). We also observed that the β-cells exposed to mutant conditioned media tended to display lower oxygen consumption and lower maximal and basal respiration capacity compared to cells exposed to WT or EV conditioned media (Extended Data Fig. 2e).

We next considered whether the accumulation of the mutant aggregates perturbs endoplasmic reticulum (ER) homeostasis to trigger the unfolded protein response (UPR) in β-cells³⁸. The ratio of spliced versus unspliced XBP1 showed a tendency to increase in the β-cells exposed to mutant conditioned media for 10 days compared to the wild-type group (Fig. 4k) and could be due to elevated levels of pIRE1α suggesting activation of the IRE1α branch of the UPR³⁸ (Fig. 4l). Additionally, the major chaperones BiP (Fig. 4m) and GRP94 (Fig. 4n) were both increased in the mutant group; the lack of significant changes in eIF2α phosphorylation suggested activation of the ATF6 but not the eIF2α branch (Fig. 4o). These data indicate that accumulation of protein aggregates after long-term exposure to the mutant protein activates ER stress to adversely affect the function, growth, and viability of β-cells (Fig. 4p).

Development of β- and acinar-like cells is normal in MODY8

To examine the translational relevance of our findings, we undertook a stem cell approach since it allows directed differentiation of human pluripotent stem cells (hPSCs) into disease-relevant cell types to identify cellular mechanisms underlying human diseases³⁹. We used the H1 iCas9 hESC line⁴⁰ to introduce the patient-specific mutation (c.1686delT)¹³ into the VNTR of the CEL gene. Additional mutant cell lines that harbor random deletions in CEL gene resulting in frameshift were also generated in order to control for potential off-target effects and clonal variations and to increase the sample size (Extended Data Fig. 3). We also generated MODY8 patient-derived human induced pluripotent cells (hiPSCs) by integration-free episomal reprogramming of skin fibroblasts⁴¹ obtained from the original MODY8 family with the c.1686delT mutation (Supplementary Table 1, Extended Fig. 4a-c). Examination of patient-derived hiPSC lines ∼13 weeks after implantation into immunodeficient mice showed that all lines exhibited a normal karyotype, expressed the pluripotency markers OCT4, SOX2, and SSEA4, and as expected, formed teratomas containing the three germ layers (Extended Data Fig. 4d-i).

To model human pancreatic development using hPSCs, we employed an established pancreatic differentiation protocol with modifications (Extended Data Fig. 5a)^42,43. Gene expression and immunostaining analysis confirmed that hPSC-derived S6 β-like cells expressed key β-cell marker genes (Extended Data Fig. 5b-c). Next, we differentiated isogenic hESCs and MODY8 patient-derived hiPSCs. Both hPSCs with or without mutation differentiated into definitive endoderm (DE) expressing CXCR4 and SOX17 (Extended Data Fig. 6) and pancreatic progenitors (PP) co-expressing PDX1 and NKX6.1 (Extended Data Fig. 7) with comparable efficiencies as demonstrated by FACS and immunostaining. Further differentiation of pancreatic progenitors to S6 β-like cells resulted in generation of C-peptide (CPEP)+ β-like cells at similar proportions in wild-type and mutant groups with an efficiency ranging from 15–49% CPEP+ and 11–29% CPEP+ β-like cells derived from hESCs and hiPSCs, respectively (Extended Data Fig. 8a-c). Both the mutant and wild-type CPEP+ β-like cells showed similar glucose-stimulated insulin secretion patterns (Extended Data Fig. 8d,e). Thus, analyses of mutant hPSC lines ruled out a developmental or functional defect upon differentiation into β-like cells. We also considered directed differentiation towards the exocrine lineage and generated ductal-like and acinar-like organoids from isogenic hESCs and patient-derived hiPSCs to explore whether exocrine differentiation is impaired (Extended Data Fig. 9a, b). Both wild-type and mutant hESCs and hiPSCs differentiated into exocrine organoids with comparable efficiencies. Gene expression analysis showed that the wild-type and mutant organoids both expressed high levels of acinar genes such as CPA1, CTRB1, PTF1A, SPINK, MIST1, AMY2A, and the ductal cell markers KRT19, CA2 (Extended Data Fig. 9c) indicating that exocrine pancreas development is not significantly affected in the mutant cells.

Function of MODY8-specific β-like cells is impaired

To further investigate the development and functional maturation of β-like cells, the S4 pancreatic progenitor cells derived from wild-type or mutant hPSCs were transplanted under the kidney capsule of immunocompromised NOD/SCID-γ (NSG) mice (Fig. 5a). S4 cells are pancreatic progenitors and their in vivo engraftment is an established test to induce β-cell maturation^42,44. We used S4 cells to allow endocrine and exocrine cells to develop together to monitor functional consequences of their potential interactions. Mice transplanted with S4 cells derived from wild-type or mutant hESC clones were followed for 20 weeks after transplantation. Unfortunately, the mice transplanted with hiPSC-derived S4 cells developed teratomas by 10 weeks post-transplantation and did not secrete detectable levels of human C-peptide. On the other hand, circulating human C-peptide levels were easily detectable in the serum of mice bearing either the wild-type or mutant hESC-derived S4 cells 12 weeks post-transplantation, and the levels significantly increased 16 weeks post-transplantation (Fig. 5b). Upon glucose challenge, the mice bearing the wild-type hESC-derived S4 graft showed a robust and significant increase in human insulin levels 20 weeks post-transplantation. In contrast, the mice transplanted with mutant hESC-derived S4 cells failed to respond (Fig. 5c). Removal of the grafts and subsequent immunohistochemical analysis demonstrated formation of both INS+ endocrine and TRYP+ or CK19+ exocrine cells in hESC-derived S4 grafts (Fig. 5d). WT and MUT hESC-derived S4 cells that differentiated into pancreatic INS+ β-like and CEL+ acinar-like cells resembled cells developing in the human fetal pancreas (Fig. 5e). INS+ β-like cells derived from both WT and MUT S4 grafts expressed PDX1, NKX6.1, NKX2.2, and NEUROD1 (Extended Data Fig. 10). Analysis of the proportions of INS+ β-like cells and CEL+ acinar-like cells revealed fewer INS+ cells in the MUT grafts (Fig. 5e, f) that was, in part, due to lower proliferation compared to that in wild-type grafts (Fig. 5g, h) likely as a consequence of uptake of mutant CEL protein (Fig. 5i). We also observed less proliferation in CEL+ acinar-like cells in MUT grafts compared to that in WT grafts (Fig. 5j, k) and a greater number of aggresomes in the cytoplasm of MUT CEL+ acinar-like cells (Fig. 5l, m).

Fig. 5 |. Function of β-like cells derived from mutant hPSC lines is impaired.

a, Outline of the transplantation strategy to induce differentiation of S4 cells to endocrine and exocrine cells in vivo. b, c Random fed human C-peptide levels 8, 12, and 16 weeks post-transplantation (b), human C-peptide levels before and 30 min after intraperitoneal glucose administration 20 weeks post-transplantation (c). Data are represented as median with 25% to 75% percentile box and min/max whisker plots. One-way ANOVA with Tukey correction analyzing 16 versus 8 weeks post-transplant (b), two-tailed multiple t-test (b) and two-tailed paired t-test (c) analyzing the difference between WT and MUT. d, Confocal images of sections obtained from wild-type and mutant grafts stained for insulin (green), trypsin (red), and CK19 (gray). Nuclei stained with DAPI (blue). e, Representative images of graft sections showing expression of CEL (green) and INS (red). Human fetal pancreas tissue (35 weeks, n=1) was used as control. Donor information is given in Supplementary Table 1. Nuclei stained with DAPI (blue). f, Proportion of INS+ and CEL+ cells in graft sections. g, Representative images of graft sections stained for INS (green), Ki67 (red), DAPI (blue). h, Percentage of Ki67+ proliferating β-like cells in graft sections. i, Confocal image of INS (red) and CEL (green) double positive cells in MUT S4 graft. The white arrowhead shows a double positive cell. Scale bars are 50 μm (left panel) and 10 μm (right panel). j, Representative images of graft sections stained for CEL (green), Ki67 (red), DAPI (blue). k, Percentage of CEL+Ki67+ acinar-like cells in graft sections. l, Representative images of graft sections stained for CEL (white) and aggresome (red). White arrows point to aggresomes in CEL+ acinar-like cells. m, Quantification of number of aggresome per CEL+ acinar-like cells in graft sections. Four (b-k) or three (l, m) biologically independent animals. Data are presented as mean values ± SEM. Two-tailed t-test (f, h, k, m). Boxed areas are shown at higher magnification (d, e, i). Scale bars are 50 μm (d), 20 μm (e, g, j, l).

Collectively, these results demonstrate that the MODY8-causing CEL mutation does not significantly affect the developmental stages of embryonic pancreas formation but impairs the function and growth of β-cells.

Mutant CEL protein co-localizes with β-cells in human islet

To highlight the translational significance of the findings observed in EndoC-βH1 β-cell line and MODY8-specific β-like cells, we used primary human pancreatic islets and cultured them in conditioned media from donor cells overexpressing either an EV, WT, or MUT protein (Fig. 6a, b). Several CEL+ β-cells were detected in the islets cultured in mutant conditioned media compared to the β-cells exposed to wild-type conditioned media (Fig. 6c-d). The functional consequences of the uptake of mutant CEL protein in other islet cells such as α-cells and δ-cells require further investigation (Fig. 6c,d). These data indicate that primary human islet cells preferentially take up the mutant protein in vitro that is consistent with previous results obtained from EndoC-βH1 cells (Fig. 2b). In order to test whether CEL uptake by primary human β-cells also occurs in vivo, we undertook a transplantation approach wherein human islets were co-transplanted with human acinar cells that were overexpressing either the wild-type or mutant CEL gene (Fig. 7a, b). Assessment of grafts two weeks post-transplantation revealed several CEL+ β-cells in the islets transplanted with mutant acinar cells compared to significantly fewer in the islets transplanted with wild-type acinar cells (Fig. 7c, d).

Fig. 6 |. Mutant CEL protein is taken up by primary human pancreatic islet cells.

a, Cadaveric human islets were exposed to conditioned media from donor cells overexpressing either the EV, WT, or MUT protein for 10 days. b, CEL (V5) protein levels in conditioned media (n=3 biologically independent samples). c, Representative immunostaining images of human islets cultured in EV (n=3), WT (n=3), or MUT (n=3) conditioned media for 10 days showing co-localization of CEL (V5) protein with INS (red), PDX1 (green), GCG (green), or SST (green). Nuclei stained with DAPI (blue). Scale bar is 50 μm. The insets show the boxed areas in additional 2X magnification. d, Percentage of CEL+ islet cells (n=3 biologically independent samples). Data are presented as mean values ± SEM. One-way ANOVA followed by Tukey’s multiple comparison test.

Fig. 7 |. Analysis of the islet and acinar grafts.

a, Fluorescent images, representative of three biologically independent samples from one human donor, show human acinar cells transduced with lentiviral vectors encoding GFP. The image was taken 24 h after transduction. Donor information is given in Supplementary Table 1. Scale bar is 100 μm. b, The transplantation approach where human islets were co-transplanted with human acinar cells that were overexpressing either the wild-type or mutant CEL gene. The grafts were analyzed two weeks post-transplantation. c, Representative images show immunostaining of INS (red) and V5 (green) in the graft sections (n=3 biologically independent animals). Arrowheads show CEL+ positive β-cells (INS+CEL+). Arrows show CEL (V5) expressing acinar cells (INS-CEL+). Nuclei were stained with DAPI (blue). Scale bar is 50 μm. d, Percentage of INS-CEL+ and INS+CEL+ cells in the grafts (n=3 biologically independent animals). Data are presented as mean values ± SEM. Two-tailed t-test.

MODY8 patient β-cells contain CEL protein

Finally, we analyzed the rare pancreatic tissues obtained from a rare donor with MODY8, undergoing pancreatectomy, to directly examine the changes in the β-cells. Biopsies from the non-neoplastic pancreas of the MODY8 patient exhibited significantly reduced β-cell area and a relative increase in α-cell numbers compared to controls (Fig. 8a-c). The few islets that were detected in the MODY8 pancreas were surrounded by pronounced fibrotic tissue with some of the islets interspersed in fat-cell rich area (Fig. 8d, e). We observed strong BiP staining in the remaining acinar tissue and a few BiP+ β-cells in the MODY8 pancreas indicating ER stress (Fig. 8f). Interestingly, immunostaining of pancreas sections from MODY8 patients exhibited inconsistent NKX6.1 expression in insulin+ β-cells compared to relatively uniform expression in control β-cells (Fig. 8g). Most importantly, we observed several double positive (INS+CEL+) cells in MODY8 islets compared to virtually none in the control pancreatic sections (Fig. 8h-j). A possible explanation for this finding is that the mutant CEL protein was taken up and remained insoluble in β-cells.

Fig. 8 |. Histological analysis of MODY8 donor pancreas.

a, Representative confocal images of control and MODY8 pancreas sections co-immunostained with INS (red), GCG (green), and DAPI (blue). Donor information is given in Supplementary Table 1. b, Percent of INS+ area in the islets from control (n=15 islets/3 donors) and MODY8 (n=25 islets/5 sections/1 donor) pancreas. c, Percent GCG+ cells in the islets from control (n=35 islets/3 donors) and MODY8 (n=57 islets/5 sections/1 donor) pancreas. d, e Representative images of hematoxylin and eosin staining of MODY8 donor pancreas. f, Representative images of INS (green) and BiP (red) staining in control and MODY8 donor pancreas. The arrowhead shows BiP+ positive β-cells. Scale bar is 50 μm (top images) and 10 μm (bottom images). g, Representative images show immunostaining of INS (green) and NKX6.1 (red) in control and MODY8 donor pancreas. h, Representative confocal images of control and MODY8 pancreas sections stained with INS (green), CEL (red), and DAPI (blue). i, Representative confocal images of MODY8 pancreas sections co-immunostained with INS (green), CEL (red), and DAPI (blue). j, Percent of islets containing CEL+ β-cells in control (n=45 islets/3 donors) and MODY8 pancreas (n=90 islets/9 sections/1 donor). Control (n=3 donors) and MODY8 (n=1 donor) pancreas sections were analyzed (a-j). Data are presented as mean values ± SEM (b, c, j). Two-tailed t-test (b, c, j). Boxed areas are shown at higher magnification (f, h, i). Scale bar is 50 μm (a, g, h, i), 100 μm (d), 250 μm (e).

Discussion

Clinical investigations on MODY8, a rare, dominantly inherited monogenic form of diabetes, have reported that pre-diabetic mutation carriers exhibited significantly reduced first-phase insulin response with early β-cell dysfunction prior to development of diabetes¹³. The lack of a phenotype in mutant CEL transgenic mice⁴⁵ prompted us to use a variety of approaches to dissect the mechanism(s) underlying the disease. Our studies using both in vitro and in vivo hPSC-based disease models and pancreatic cell lines demonstrate that an abnormal relationship between pancreatic exocrine and endocrine cells underlies the β-cell dysfunction in MODY8.

Newly formed β-cells are surrounded by emerging acinar cells in the developing pancreas⁴ and the close anatomical location likely enables long-term exposure of β-cells to the mutant CEL protein synthesized by acinar cells in MODY8. Indeed, the co-localization of insulin with CEL protein in the mutant β-like cells generated in our study and the MODY8 patient β-cells supports a crosstalk between exocrine and endocrine compartments. While in vitro co-culture experiments and in vivo transplantation studies could only test the effects of such a crosstalk for relatively short duration, the actual effects in patients is for much longer duration of interaction between acinar and beta cells. While it cannot be formally excluded that these double-positive cells exhibit a dual identity, it is worth noting that a study reported the co-appearance of granules that are characteristic of exocrine and endocrine compartments in the same cell in human pancreas with more frequency in type 1 diabetes pancreas⁴⁶. Thus, the CEL protein could be taken up by β-cells by a paracrine and/or endocrine pathways, and previous reports support the involvement of the former^6,7,47–49. For example, acinar-like cell clusters have been reported to adhere closely to islet cells via desmosomal and coated-pit-like structures that would potentially allow the exchange of molecules^5,6. Pancreases in diabetic mice have been reported to exhibit a ‘zonation’ of acinar cells surrounding the islets⁴⁹ that is induced by CCK secreted from islet cells⁷. In humans, an attenuation in the endocrine to exocrine communication and alterations in the islet exocrine interface has been suggested to underlie pancreatic insufficiency in prediabetes and overt type 2 diabetes⁴⁸. On the other hand, a recent study showing bidirectional blood flow between endocrine and exocrine pancreas⁵⁰ supports the involvement of the endocrine pathway. Whether circulating CEL in human plasma is preferentially taken up by β-cells⁵¹ compared to other cell types requires further investigation.

Notwithstanding the mode of internalization, the accumulation of the mutant CEL protein over the long-term likely leads to toxic effects on β-cells due to ER stress and the infectious effects of senescence-associated secretory phenotype (SASP)⁵². Together with ER stress the acquisition of a SASP could induce bystander senescence in neighboring cells^32,52 and contribute to a significantly reduced β-cell mass in MODY8. Another point is that potential transmission of ER stress between cells⁵³ could contribute to the phenotype observed in β-cells. While non-β cells could also take up mutant CEL in vitro (Fig. 6c, d), the reason of observing loss and dysfunction selectively in β-cells might be related to expression of disallowed genes⁵⁴ in β-cells versus non-β cells. Furthermore, the development of exocrine insufficiency in pre-diabetic mutation carriers may also adversely affect the microenvironment to compromise the growth and secretory function of β-cells. These observations have implications for type 1 diabetes. In a recent study, several newly identified risk variants of type 1 diabetes mapped at cis-regulatory elements of genes with pancreatic exocrine expression, including one variant that mapped close to the CEL gene⁵⁵. While direct involvement of enzymatic activity of the mutant CEL protein was not the focus of this study, it is possible that accumulation of the mutant protein leads to changes in the endogenous levels of CEL substrates such as FAHFAs¹⁸. Whether this pathway also participates in the development of diabetes in MODY8 patients requires further investigation.

In summary, the diabetic symptoms exhibited by MODY8 patients could be due to impaired secretory function and a reduced β-cells mass made worse by the naturally slow proliferative nature of the cells containing the mutant CEL protein.

Methods

Animals.

All mice were housed four to five mice per cage at a relative humidity of 30–70% and room temperature of 22.2 ± 1.1 °C at the Animal Facility of the Joslin Diabetes Center on a 12 h light–12 h dark cycle with free access to food and water.

Plasmids.

Cloning of cDNAs encoding wild-type and mutant (c.1686delT) CEL into the pcDNA3.1/V5-His vector (Invitrogen) in-frame with a C-terminal V5/His tag procedure is described in detail in Johansson et al¹⁹. Cells transfected with pcDNA3.1/V5-His without any insert (empty vector, EV) were included in all experiments as a negative control.

Cell lines.

The EndoC-βH1 cell line, purchased from Univercell-Biosolutions (CVCL_L909), was cultured and passaged as previously described⁵⁶. HEK293 cell lines stably transfected with EV, WT, or MUT plasmids were described in detail in Johansson et al¹⁹. The 266–6 mouse acinar cell line was obtained from ATCC (CRL-2151) and cultured in DMEM (glucose 4.5 g L⁻¹; Gibco) supplemented with 10% FBS (Gibco).

Primary human cells.

Cadaveric human islets and acinar tissues were obtained from either the Integrated Islet Distribution Program (IIDP) or Prodo Laboratories. Upon receipt, human islets or acinar tissues were centrifuged, resuspended in fresh Miami Media #1A (Cellgro) or RPMI media with 10% FBS, respectively. Cells were transferred to Petri dishes and cultured o/n in 5% CO₂ at 37°C before any treatment. Establishment of primary human skin fibroblasts is described in detail in Teo et al⁵⁷. Genome-edited H1 hESC lines and patient-derived hiPSC lines were cultured on Vitronectin (VTN-N, Gibco) coated tissue culture plates in Essential 8 medium (Gibco, A1517001). Cells were split using 0.5 mM EDTA at 1:10–1:20 ratio every 4–6 days. The cells were routinely tested for mycoplasma contamination. All studies and protocols used were approved by the Joslin Diabetes Center’s Committee on Human Studies (CHS#5–05).

Co-culture experiments.

The 266–6 mouse acinar cell line was obtained from ATCC (CRL-2151) and cultured in DMEM (glucose 4.5 g L⁻¹; Gibco) supplemented with 10% FBS (Gibco). The cells were transfected with 1 μg plasmid (EV, WT, or MUT plasmids) in a 6-well plate using Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer’s instructions. Acinar cells were seeded onto the inserts with 0.4 μm pore size (Corning, 353493) 24 h after transfection. EndoC-βH1 cells were seeded at the bottom of the 6-well plate one day before and the inserts with 266–6 cells were placed on top of each well. Fresh medium was added to the wells and the cells were co-cultured for 24 h.

Treatment with conditioned medium.

HEK293 cells stably transfected with EV, WT, or MUT plasmids (donor cells) or 266–6 acinar cells transiently transfected with EV, WT, or MUT plasmids were washed with DPBS twice and incubated in EndoC-βH1 growth medium (for the treatment of EndoC-βH1 cells) or in Miami Medium (for the treatment of human islets) for 3 h. Conditioned medium was collected, filtered using a 0.2 μm filter to remove detached and dead cells, and stored at −20°C. The concentration of CEL protein was determined by quantitative immunoblotting (Extended Fig. 1f-h). Increasing amounts of recombinant mouse CEL proteins (0 to 10 ng protein, R&D, 5658-CE) and conditioned media were loaded onto the same SDS-PAGE gel. Proteins were transferred to nitrocellulose membrane and then labeled with anti-CEL (1:1000, HPA052701; Sigma-Aldrich). The concentrations of wild-type and mutant CEL proteins were estimated by preparing a standard curve. Approximately 2 μg ml⁻¹ CEL protein concentration was used for each conditioned medium treatment. EV conditioned medium was used to dilute WT or MUT conditioned medium to adjust CEL protein concentrations. EndoC-βH1 cells or human islets were treated with conditioned medium for the times indicated in the figure legends. For long-term treatments, conditioned medium was refreshed every 24 h.

Uptake and degradation kinetics.

For uptake experiments, EndoC-βH1 cells were exposed to a conditioned medium time course (1 min, 0.5 h, 2 h, 6 h, 12 h, 24 h). Cells were washed with cold PBS twice and lysed in RIPA buffer. For clearance experiments, EndoC-βH1 cells were exposed to conditioned medium for one hour, washed with DPBS twice, and incubated in a fresh growth medium time course (0 h, 0.5 h, 2 h, 6 h, 12 h, 24 h). Cells were washed with cold PBS twice and lysed in RIPA buffer for Western blotting.

Western blotting.

Media were collected and centrifuged at 200 g for 5 min. Supernatant was transferred to a clean tube, mixed with 1x Laemmli buffer containing 10% glycerin, 2% 2-mercaptoethanol, 1.5% SDS, 60 mM Tris HCl pH 6.8, 0.05% Coomassie G-250, and heated to 96°C for 5 min. Cells were washed with cold DPBS (Invitrogen) twice and lysed in ice-cold RIPA buffer (pH 7.4) containing 50 mM Tris HCl pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% Na-deoxycholate, 5 mM EDTA, 0.1% SDS, 2 mM Na₃VO₄, 10mM NaF, 10 mM Na₄P₂O₇, protease inhibitor cocktail (P8340; Sigma), phosphatase inhibitor 2 (P5726; Sigma) and phosphatase inhibitor 3 (P0044; Sigma). The cells were then incubated on a rotating wheel for 30 min at 4°C and centrifuged at 13,800 g for 15 min at 4°C. The supernatant (detergent-soluble fraction) was collected, and the pellet was washed twice with cold PBS. The pellet (detergent-insoluble fraction) was resuspended in Laemmli buffer and heated to 96°C for 5 min. The supernatant was mixed with 1x Laemmli buffer and then heated to 56°C for 15 min after total protein concentration was determined by Pierce BCA Protein Assay Kit (23225; Thermo). Supernatant and pellet proteins were run in 6% SDS-PAGE and transferred to nitrocellulose membrane (Millipore). Membranes were blocked for 1 h at room temperature (RT) with 5% milk and were incubated overnight at 4°C with primary antibodies. After three washing series (10 min), the membranes were incubated for 1 h at RT with secondary antibodies. After three 10-min washes, signals were visualized via Pierce ECL Western blotting substrate (PI32106; Thermo). Western blot image quantification was performed using Image J software (1.51s). See also the Supplementary Table 2 for detailed antibody information.

Flow cytometry.

Cells were harvested, washed with DPBS, and stained with Zombie NIR^™ Viability dye (Biolegend) for 15 min at RT. Cells were washed first with medium containing 10% FBS followed by PBS wash. Cells were then fixed in 4% paraformaldehyde (PFA) for 15 min at RT. Cells were spun and washed with cold FACS buffer (5% FBS in PBS). Permeabilization and blocking was carried out on ice for 30 min in FACS buffer with 0.1% Triton X-100. Antibody staining was performed on ice for 30 min followed by incubation with secondary antibody for 30 min on ice. Cells were washed and resuspended in 250 μl FACS buffer and filtered through 30 μm filter before analysis by LSR II (BD Biosciences, Joslin Flow Cytometry Core). For apoptosis detection, cells were stained with Zombie NIR Viability dye and resuspended in 1X binding buffer containing APC Annexin V (1:20, BD Biosciences). Cells were incubated for 15 min at RT and analyzed by FACS Aria (Joslin Flow Cytometry Core). Apoptotic cell rate was determined as percentage of AnnexinV+ and Zombie NIR- cells. Data was analyzed by using FlowJo 10.7.1. Gating was determined according to the secondary-only and isotype controls. See also the Supplementary Table 2 for detailed antibody information.

Immunofluorescence.

Cells growing in chamber slides were fixed in 4% PFA for 15 min at RT and washed with PBS three times. Cells were then permeabilized with PBS containing 0.25% Triton X-100 for 30 min at RT and blocked with PBS containing 0.25% Triton X-100 and 5% donkey serum for 1 h at RT. Primary antibody was diluted in antibody dilution buffer (Dako) and added to the wells for overnight incubation at 4°C. Cells were washed three times with PBS and the secondary antibody, diluted in PBS, was added to the wells for 1 h at RT. Cells were washed three times with PBS and 4,6-diamidino-2-phenylindole (DAPI) was added to the wells. Images were captured using an Olympus IX51 Inverted Microscope or Zeiss LSM 710 NLO confocal laser scanning microscope. See also the Supplementary Table 2 for detailed antibody information.

Endocytosis assay.

For cold-mediated endocytosis blockage, β-cells were treated with warm or cold conditioned medium for 30 min at 37°C or at 4°C. For inhibitor treatment, EndoC-βH1 cells were pretreated for 30 min with DMSO or 100 nM Wortmannin (InvivoGen) in BSA-free EndoC-βH1 growth medium and then treated 30 min with DMSO or Wortmannin in BSA-free conditioned medium.

Exosome assay.

Donor cells overexpressing wild-type or mutant CEL were treated overnight with DMSO or 5 μM GW4869 (Sigma #567715) after 4 h pretreatment in serum-free medium. Media were collected and centrifuged at 200 g for 5 min. Supernatants were analyzed by Western blotting.

Aggresome detection.

EndoC-βH1 cells were seeded in chamber slides and exposed to conditioned medium overnight. A proteasome inhibitor, MG132 (5 μM), was added to positive control wells overnight. Aggresomes were visualized with the Proteostat Aggresome detection kit (catalog no. ENZ-51035; Enzo Life Sciences) according to the manufacturer’s instructions²⁵. Cells were analyzed with a Zeiss LSM 710 NLO confocal laser scanning microscope. Intracellular darkly colored red dots are considered as aggresomes. Number of aggresomes per β-cell was calculated by counting number of aggresomes in approximately 125 β-cells for each group.

Protein aggregation assay.

Conditioned medium collected from donor cells was tested for protein aggregation using the Proteostat Protein Aggregation assay (catalog no ENZ-51023; Enzo Life Sciences) according to the manufacturer’s instructions. Fluorescence was detected using a Promega fluorescence microplate reader with 520 nm excitation and 580–640 nm emission filters.

Immunoperoxidase electron microscopy.

EndoC-βH1 cells grown in 6-well-plates were exposed to conditioned media for an hour, washed with DPBS, and fixed with paraformaldehyde fixative. Cells were labelled with anti-mouse CEL (As 20.1) and anti-mouse IgG (HRP) before embedding in Epoxy resin^20,58. Thin sections (1 micron) of the immunolabelled cells were stained with lead citrate and viewed under the Philips 301 transmission electron microscope.

Cell viability assay.

The assay was performed using a Cell Titer Glo 2.0 luminescent cell viability kit (Promega) according to the manufacturer’s instructions. 5×10⁴ EndoC-βH1 cells were seeded in clear bottom black 96-well plates and grown in the conditioned medium for 10 days. Conditioned medium was refreshed every day (100 μl well⁻¹). On the day of analysis, 100 μl of Cell Titer Glo 2.0 reagent was added to the cells and the content was mixed for 2 min on an orbital shaker. After 10 min of incubation at RT, luminescence was recorded by Promega GloMax luminometer with an integration time of 1 sec per well.

Insulin secretion assay.

EndoC-βH1 cells were starved overnight in 2.8 mM glucose followed by 1 h incubation in Krebs Ringer Buffer (KRB). Static insulin secretion assays were then initiated by adding KRB containing 3.3 mM or 16.7 mM glucose for 1 h. β-like cells were starved 1h in KRB containing 0.5 mM glucose. Insulin secretion was stimulated by adding KRB containing 1 mM glucose (LG) or 20 mM glucose (HG) for 1 h⁵⁶. Aliquots of supernatants were removed for later analysis and ice-cold acid ethanol was added to extract insulin content from cells. Insulin secretion and content were measured by the human insulin ELISA (Mercodia) according to manufacturer’s instructions. Human insulin secretion was normalized to the total number of INS+ β-like cells.

RNA extraction and analysis.

Cells were lysed in RLT buffer and then RNA was extracted using Qiagen RNeasy kit according to the manufacturer’s instructions. The RNA quality and quantity were analyzed using a NanoDrop 1000 Spectrophotometer (Thermo Fisher). cDNA was generated from 1 μg RNA using Multiscript cDNA synthesis kit (Applied Biosystems) according to the manufacturer’s protocol for real-time quantitative PCR with SYBR Green detection (Applied Biosystem) by ABI7900 Real Time PCR System (Joslin), and gene expression was calculated using the ΔΔCt method. Data were normalized to a housekeeping gene (β-actin). See the Supplementary Table 3 for primer sequences.

Generation of mutant hESC lines.

H1 iCas9 hESCs were treated with doxycycline one day before and at the time of gRNA transfection. Cells were dissociated into single cells using TrypLE Select and transfected in suspension with gRNA and ssDNA for homology directed recombination using Lipofectamine RNAiMAX (Life Technologies) as described previously⁴⁰. Three days after transfection, cells were harvested and approximately 2,000 cells were seeded per 10 cm dish. Cells were allowed to grow for 10–14 days. Fifty to one hundred colonies were picked and transferred into a 96-well plate. DNA was isolated for Sanger sequencing to identify mutant clones. PCR and sequencing primers are provided in Supplementary Table 3. Clonal hESC lines carrying the mutation and wild-type controls were expanded. Mutation was further validated by TOPO cloning followed by DNA sequencing according to manufacturer’s instructions (Invitrogen). We analyzed four heterozygous lines and one compound heterozygous mutant line carrying frameshift mutations generated using two gRNAs targeting different sequences and compared them to four isogenic wild-type control H1 lines for in vitro experiments.

Episomal Reprogramming.

Human fibroblasts were cultured in DMEM 1 g L⁻¹ glucose supplemented with 10% FBS, 1% MEM-NEAA, and 1% sodium pyruvate. Cells were lifted using 0.05% trypsin and 500,000 cells were nucleofected with a mixture of three plasmids pCXLE-hOCT3/4-shp53-F (Addgene, 27077), pCXLE-hSK (Addgene, 27078), pCXLE-hUL (Addgene, 27080) using P22 and NHDF Nucleofector Kit (Lonza, VAPD-1001) according to manufacturer’s instructions⁴¹. Nucleofected cells were immediately seeded on 6-well plates and medium was changed every other day. On day 5, cells were harvested and reseeded on mouse embryonic fibroblast (MEF) cells. Cell medium was switched from DMEM to hESC medium which is DMEM/F12 (Stem Cell Technologies, 36254) supplemented with 10 ng ml⁻¹ FGF2 (Miltenyi Biotech, 130–093-838), 20% knockout serum replacement (KOSR, Invitrogen, 10828–028), 1% L-glutamine, 1% MEM-NEAA on day 7. Media was changed every 24 h. Growing hiPSC colonies were picked between days 20 and 40 and seeded on fresh MEFs and expanded for characterization. More than four hiPSC clones were established from each individual. Considering line-to-line and clonal variation in hiPSC studies, we used at least two independent hiPSC clones from each individual for in vitro experiments.

DNA Sequencing.

Genomic DNA was isolated from human fibroblasts or hPSCs using DNeasy Blood & Tissue Kit (Qiagen). CEL exon 11 sequences were amplified using TaKaRa LA Taq with GC-buffer (Clontech, RR02AG). Primer sequences are provided in Supplementary Table 3. PCR products were extracted from gel, purified, and submitted to DNA Resource Core (Dana Farber/Harvard Cancer Center) for Sanger sequencing.

Teratoma assay.

Approximately 1×10⁶ hiPSCs were harvested as cell clumps and resuspended in a 2:1 mixture of hESC media:Matrigel in a final volume of 150 μl⁵⁷. Cell clumps were injected intra-muscularly (gastrocnemius) into 4-week-old male NSG mice (Jackson Laboratories). When tumors were detected by palpation (approximately 2 cm diameter), teratomas were excised, sectioned, and stained with hematoxylin and eosin (HE) to identify embryonic tissues.

In vitro differentiation.

Differentiation was initiated 24 h to 48 h after plating when the culture was 90% in confluency as described⁵⁶. The differentiation condition and media recipes are presented in the Supplementary Table 4. Pancreatic exocrine progenitor organoids and differentiation were conducted following protocols reported previously⁵⁹. S3 cells were digested with Accutase (Sigma) for 5 minutes and centrifuged at 220 g for 5 min. Dissociated cells were resuspended in DMEM medium with 5% Matrigel, 1% B27, 50 μg/ml ascorbic acid, 20 μg/ml insulin, 0.25 μg/ml hydrocortisone, 100 ng/ml FGF2, 100 nM all-trans retinoic acid and 10 μM Y267632. The cell suspension was seeded on chamber slides pre-coated with Matrigel at 5,000 cells per well (400 μl). From day 8, organoids were incubated with differentiation medium based on DMEM, containing 5% Matrigel, 1 % B27, 300 μM 2-phospho ascorbic acid, 100 ng/ml FGF7, 10 ng/ml EGF, 1 μM A8301 and 1 μM DBZ. Differentiation medium was changed every two days.

Transplantation.

Eight-to-10-week-old female NSG mice were obtained from Jackson Laboratories. Mice were anaesthetized with ketamine and xylazine. Approximately 5×10⁶ S4 cells were transplanted under the kidney capsule as described previously⁶⁰. Glucose-stimulated human C-peptide secretion in mice was assessed by collecting blood samples after an overnight fast (16 h) and 30 min following an intraperitoneal glucose bolus (2 g kg⁻¹; 30% solution). Serum was stored at −80°C and later assayed using a human C-peptide ELISA (Mercodia). For primary human islet and acinar transplantations, 500 IEQ and 200 acinar cell clusters were handpicked, transplanted under the kidney capsule, and grafts were harvested 2 weeks post-transplantation. Islet and acinar tissues were obtained from the same donor. Donor information is provided in Supplementary Table 1. Three days before transplantation, acinar cells were cultured o/n in RPMI media supplemented with 10% FBS, washed with PBS, and transduced with lentiviral vectors at 50 MOI in serum free RPMI media containing 10 μg/ml polybrene for 4 h, and RPMI media with 10% FBS was added. Transduction efficiency was confirmed by observing GFP expression at 24 h and 48 h. Lentiviral vectors pLV-EGFP-T2A-Puro-EF1A-mCherry, pLV-Puro-CMV-hCEL-WT-V5, pLV-Puro-CMV-hCEL-MUT-V5 were ordered from VectorBuilder.

Immunohistochemistry.

At the end of the follow-up period, mice were killed, and kidneys with grafts were collected, fixed, and embedded in paraffin. Sections were stained using antibodies to INSULIN, GLUCAGON, CEL, TRYPSIN, CK19, Ki67, NKX6.1, NKX2.2, NEUROD1, PDX1. Islet and acinar graft sections were stained using antibodies to V5 (to detect exogenous CEL) and INSULIN. Appropriate secondary antibodies were used, and sections were counterstained with DAPI. See also Supplementary Table 2 for detailed antibody information. Images were captured using an Olympus IX51 Inverted Microscope or Zeiss LSM 710 NLO confocal laser scanning microscope. For MODY8 patient pancreas samples, SP8 AOBS confocal microscope (Leica Microsystems) and Leica Application Suite X (LAS X) software (Version: 3.5.7.23225) were used to capture images. To measure the percent of α and β area, contours of islets were demarcated manually. The percentage of INS+ area was quantified automatically using QuPath software v0.2.2 by dividing the number of INS+ area by the total islet area. Percent GCG+ cells in the islets was quantified by dividing the number of GCG+ cells by the total cell number in each islet.

Statistics and reproducibility.

All statistics were performed using GraphPad Prism software (GraphPad Software Inc., La Jolla, CA). Specific statistical tests for each experiment are described in the figure legends. All values are ± SEM, and statistical significance was set at P < 0.05.

Study approval.

All animal experiments were conducted in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care. All protocols were approved by the Institutional Animal Care and Use Committee of the Joslin Diabetes Center following NIH guidelines. All human studies and protocols used were approved by the Joslin Diabetes Center Committee on Human Studies (CHS, 5–05). Informed consent was obtained from control and MODY8 patients with no participant compensation. All studies and protocols used were approved by the Regional Ethical Committee of Western Norway for the MODY8 case and controls (REK Vest 2013/1772). All stem cell-related experiments were approved by the Embryonic Stem Cell Research Oversight Committee (ESCRO, 2010–01).

Extended Data

Extended Data Fig. 1 |. Reduced secretion of mutant CEL protein from 266–6 acinar cells and endocytosis-mediated uptake by β-cells.

a, Mouse acinar cells (266–6) transfected with empty vector (EV), wild-type CEL (WT), or mutant CEL (MUT) plasmids. Cell lysates (L) and media (M) were collected and labeled with anti-V5 antibody to measure V5-tagged CEL levels. b, Western blots, representative of three independent experiments, show low molecular mass forms of wild-type and mutant human CEL proteins in lysates and the fully glycosylated, high molecular mass forms secreted into media. c-e, The quantification of three biologically independent samples. Fold change relative to EV. Lysate CEL levels were normalized to β-actin. f, Determination of the concentration of CEL protein in conditioned medium by quantitative immunoblotting (n=2 biologically independent samples) using recombinant mouse CEL proteins (rmCEL). g, A standard curve for rmCEL. h, The concentrations of wild-type and mutant CEL proteins were estimated according to the standard curve. i, Recipient β-cells were treated with conditioned medium obtained from HEK293 donor cells for 30 min at 37C or at 4C (n=3 biologically independent samples). j, The quantification of three biologically independent samples. Fold change relative to WT 37C. V5-tagged CEL levels were normalized to-tubulin levels. k, Percentage of CEL+ β-cells detected by immunostaining after treatment with Wortmannin to block endocytosis (n=3 biologically independent samples). l, Western blot quantification of β-cells treated with conditioned medium obtained from HEK293 donor cells treated with DMSO or exosome inhibitor GW4869. Fold change relative to WT DMSO (n=3 biologically independent samples). m, Conditioned media collected from HEK293 donor cells stably transfected with EV, WT, or MUT plasmids was subjected to protein aggregation assay (n=9 biologically independent samples). Fluorescence signal generated by Proteostat detection dye was measured. Aggregated lysozyme (20 g) and monomeric lysozyme (20 g) were used as positive and negative controls, respectively (n=3 independent samples). Data are expressed as fold change relative to WT. Data are presented as mean values SEM. One-way ANOVA followed by Tukey’s multiple comparison test (c, d, j, k-m), two-tailed t-test used for (e). Dashed line is added for easy comprehension (b, f, i).

Extended Data Fig. 2 |. Accumulation of mutant CEL protein aggregates reduces proliferation and impairs function of β-cells.

a, Representative FACS plots showing percentage of AnnexinV and Zombie Near Infrared (NIR) stained β-cells treated with conditioned medium for 10 days to assess apoptosis levels (n=3 biologically independent samples). b-d, Heatmap showing differentially expressed genes involved in cellular senescence (b), Insulin secretion (c), glycolysis/gluconeogenesis, oxidative phosphorylation (d) in β-cells treated with conditioned medium for 10 days (EV n=9, WT n=10, MUT n=10 biologically independent samples). e, Mitochondrial respiration profile of EndoC-βH1 cells exposed to conditioned media for 10 days. Cells were challenged with oligomycin (15 μM), FCCP (10 μM), and Rotenone+actinomycin (1.1 μM + 25 μM) (top panel). Quantification of maximal and basal respiration capacity (EV n=5, WT n=4, MUT n=5 biologically independent samples). Data are presented as mean values SEM. One-way ANOVA followed by Tukey’s multiple comparison (a, e).

Extended Data Fig. 3 |. Generation of hESC lines expressing mutant CEL.

a, Schematic of CRISPR-Cas9 strategy for generation of CEL mutant hESC lines. Exons and introns are represented by boxes and lines, respectively. A 100-nt single strand DNA (ssDNA) carrying the patient specific mutation (c.1686delT) and two gRNAs (gRNA1, gRNA2) were used to target the first repeat of VNTR in the CEL gene exon 11. PAM sequences in the gRNAs are indicated in purple. Heterozygous and compound heterozygous mutant cell lines were generated. Deletions are indicated in red boxes. b, Alignment of the C-terminal end of WT and MUT CEL proteins. WT and MUT amino acid sequences are indicated in blue and red, respectively. Small deletions cause frameshift in the first repeat of VNTR (Rep1) and create a shorter protein. Each asterisk denotes deletion of an amino acid. c, DNA sequencing of isogenic hESC lines carrying deletion mutations in the CEL gene. Asterisks show the mutation sites. d, Normal karyotype of WT and MUT hESC lines generated by CRISPR-Cas9 technology.

Extended Data Fig. 4 |. Generation and characterization of MODY8 disease-specific hiPSCs.

a, MODY8 family pedigree. Solid symbols denote diabetes. NN, no mutation; NM, mutation. b, Outline of the episomal reprogramming approach. Details are given in the Methods section. c, Representative bright-field images of hiPSC colonies derived from family and non-family controls (Fam Ctr n=4, and Non-Fam Ctr n=4 independent clones) and from mutation carriers with or without diabetes (Mut+ Dia+ n=8, Mut+ Dia- n=6 independent clones). Scale bar is 1 mm. d, Normal karyotype of hiPSCs derived from controls and MODY8 patients (Non-Fam Ctr n=2, Fam Ctr n=2, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples). e, DNA sequencing confirmed presence of c.1686delT deletion (asterisk) in hiPSCs derived from MODY8 patients (Non-Fam Ctr n=2, Fam Ctr n=2, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples). f, Immunostaining for pluripotency markers OCT4 (red), SOX2 (green), SSEA4 (red) in hiPSCs derived from controls and mutation carriers (Non-Fam Ctr n=2, Fam Ctr n=2, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples). Nuclei stained with DAPI (blue). Scale bar is 500 m. g, Control and MODY8 hiPSCs formed teratoma approximately 13 weeks after implantation into immunodeficient mice. Non-Fam Ctr n=3, Fam Ctr n=3, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples. Data are presented as mean values SEM. The difference between control vs mutant lines is not significant. Two-tailed t-test. h, Control and mutant hiPSCs formed teratoma (approximately 2 cm diameter) after injection into right leg muscle of immunodeficient mice (Non-Fam Ctr n=3, Fam Ctr n=3, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples). i, Representative images of teratomas stained with hematoxylin and eosin (HE) (Non-Fam Ctr n=3, Fam Ctr n=3, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples). Arrows show differentiated tissues including neural rosettes (ectoderm), cartilage (mesoderm), epithelial tissue (endoderm). Scale bar 50 m.

Extended Data Fig. 5 |. Differentiation of control hiPSCs to generate S6 -like cells.

a, Differentiation protocol used to differentiate hPSCs towards insulin expressing -like cells. Details are given in Supplementary Table 4. b, Control hiPSCs (family and non-family control) were differentiated to S6 and expression levels of genes specific to β-cells were measured. Two differentiation batches of control hiPSCs (Batch1 n=4, Batch2 n=2 biologically independent samples) and human islets collected from five donors were used. Donor information is provided in Supplementary Table 1. -actin was used as a housekeeping control. Fold change relative to S0. Data are presented as mean values SEM. c, Control hiPSCs (N65–51 family control, n=2 biologically independent samples) were differentiated to S6 and co-stained for several pancreatic markers; INS/GCG, NKX6.1/PDX1, CHGA/GCG, PDX1/ISL1, NEUROD1/PDX1, NKX2.2/PDX1. Nuclei stained with DAPI (blue). Scale bar is 500m.

Extended Data Fig. 6 |. Differentiation of gene-edited hESCs and patient-derived hiPSCs to definitive endoderm stage.

a, FACS analysis of wild-type and mutant hESC lines differentiated to S1. +/+ n=12, +/1686delT n=3, +/1698delA n=3, +/1690_1702del n=3, +/1683_1704del n=3, 1686delT/1702delG n=3 biologically independent samples. Data are presented as mean values SEM. No difference was detected in wild-type vs. each mutant line by two-tailed multiple t-test corrected using Holm-Sidak method. b, FACS analysis of control and mutant hiPSC lines differentiated to S1. Control n=8, Mut+ Dia- n=6, Mut+ Dia+ n=7 biologically independent samples. Data are presented as mean values SEM. No difference was detected in control vs Mut+ Dia- and vs. Mut+ Dia+ by two-tailed multiple t-test corrected using Holm Sidak method. c, Immunostaining images, representative of three biologically independent samples, showing DE cells stained for SOX17 (green) and OCT4 (red). Nuclei stained with DAPI (blue). Scale bar is 500 m.

Extended Data Fig. 7 |. Differentiation of gene-edited hESCs and patient-derived hiPSCs to pancreatic progenitor stage.

a, FACS analysis of wild-type and mutant hESC lines differentiated to S4. +/+ n=16, +/1686delT n=4, +/1698delA n=4, +/1690_1702del n=4, +/1683_1704del n=4, 1686delT/1702delG n=4 biologically independent samples. Data are presented as mean values SEM. No difference was detected in wild-type vs. each mutant line by two-tailed multiple t-test corrected using Holm-Sidak method. c, Representative FACS plots of control and mutant (with or without diabetes) hiPSC lines differentiated to S4. b, FACS analysis of control and mutant hiPSC lines differentiated to S4. Control n=6, Mut+ Dia- n=5, Mut+ Dia+ n=4 biologically independent samples. Data are presented as mean values SEM. No difference was detected in control vs Mut+ Dia- and vs Mut+ Dia+ by two-tailed multiple t-test corrected using Holm-Sidak method. c, Immunostaining images, representative of three biologically independent samples, showing PP cells for PDX1 (green) and NKX6.1 (red). Nuclei stained with DAPI (blue). Scale bar is 500 m.

Extended Data Fig. 8 |. Differentiation of gene-edited hESCs and patient-derived hiPSCs to -like cells.

a, FACS analysis of wild-type and mutant hESC lines differentiated to S6. +/+ n=4, +/1686delT n=4, +/1698delA n=3, +/1690_1702del n=4, +/1683_1704del n=4, 1686delT/1702delG n=4 biologically independent samples. Data are presented as mean values SEM. No difference was detected in wild-type vs. each mutant line by two-tailed multiple t-test corrected using Holm-Sidak method. b, FACS analysis of control and mutant hiPSC lines differentiated to S6. Control n=8, Mut+ Dia- n=6, Mut+ Dia+ n=6 biologically independent samples. Data are presented as mean values SEM. No difference was detected in control vs. Mut+ Dia- and vs. Mut+ Dia+ by two-tailed multiple t-test corrected using Holm-Sidak method. c, Immunostaining images, representative of three biologically independent samples, showing-like cells for CPEP (green) and GCG (red). Nuclei stained with DAPI (blue). Scale bar is 500m. d, GSIS was performed by stimulating wild-type or mutant S6 cells with 1 mM low glucose (LG) or 20 mM high glucose (HG) for an hour. The stimulation index was calculated as the fold increase in human C-peptide release measured in 20 mM over 1 mM glucose. +/+ n=4, +/1686delT n=4, 1686delT/1702delG n=4 biologically independent samples. Data are represented as median with 25% to 75% percentile box and min/max whisker plots. Two-tailed multiple t-tests followed by Holm Sidak’s multiple comparison test. e, Stimulation index of iPSC-derived -like cells. Control n=4, Mut+ Dia- n=4, Mut+ Dia+ n=4 biologically independent samples. Data are represented as median with 25% to 75% percentile box and min/max whisker plots. Two-tailed multiple t-tests followed by Holm Sidak’s multiple comparison test.

Extended Data Fig. 9 |. Differentiation of gene-edited hESCs and patient-derived hiPSCs to acinar-like organoids.

a, Differentiation protocol used to differentiate hPSCs towards exocrine organoids. Details are given in the Methods section. b, Bright field images, representative of three biologically independent samples, show organoids that were derived from gene edited hESCs. Scale bar is 100m. c, Expression levels of genes specific to exocrine pancreas were measured by RT-PCR (WT n=5, MUT n=5 biologically independent samples). Data are presented as mean values SEM. No difference was detected in wild-type vs. mutant by two-tailed t-test. Similar results were observed using three differentiation batches of hESC or hiPSC lines. -actin was used as a housekeeping control. Fold change relative to S0.

Extended Data Fig. 10 |. Analysis of WT and MUT S4 graft sections.

Representative immunostaining images of grafts derived from mutant (n=4) or wild-type (n=4) S4 cells stained for β-cell markers such as INS in green, PDX1, NKX2.2, NEUROD1, and NKX6.1 in red. Human fetal pancreas (34 weeks, n=1 donor) was used as control. Donor information is given in Supplementary Table 1. Nuclei stained with DAPI (blue). Scale bar is 20m.

Data availability

RNA-sequencing data in EndoC-βH1 cells have been deposited under accession code GSE185430. Source Data are provided with this paper. Other data that support the findings of this study are available from the corresponding author upon reasonable request.