HNF1A deficiency causes reduced calcium levels, accumulation of abnormal insulin granules and uncoupled insulin to C-peptide secretion in a stem cell model of MODY3


 Mutations in HNF1A cause Maturity Onset Diabetes of the Young type 3 (MODY3), the most prevalent form of monogenic diabetes. Using stem cell-derived pancreatic endocrine cells from human embryonic stem cells (hESCs) with induced hypomorphic mutations in HNF1A, we show that HNF1A orchestrates a transcriptional program required for calcium-dependent insulin secretion. HNF1A-deficient β-cells display a reduction in CACNA1A and intracellular calcium levels, as well as impaired insulin granule exocytosis in association with SYT13 down-regulation. Knockout of CACNA1A and SYT13 reproduce the relevant phenotypes. Retention of insulin is associated with accumulation of enlarged secretory granules, and altered stoichiometry of secreted insulin to C-peptide. Glibenclamide, a sulfonylurea drug used in the treatment of MODY3 patients, increases intracellular calcium, and thereby restores C-peptide and insulin secretion to a normal ratio. While insulin secretion defects are constitutive in cells with complete HNF1A loss of function, β-cells from patients with heterozygous hypomorphic HNF1A mutations are initially normal, but lose the ability to secrete insulin and acquire abnormal stoichiometric secretion ratios, while gene corrected cells remain normal. Our studies provide the molecular basis for the treatment of MODY3 with sulfonylureas, and demonstrate promise for the use of cell therapies for MODY3.


Introduction
Maturity onset diabetes of the young (MODY) is an autosomal dominant form of diabetes with onset typically before the age of 25 years accounting for 1-2% of diabetes incidence (1). There are at least 11 (2) genetically distinct types of MODY, due to derangements in β-cell development or function. MODY3 is caused by mutations in the transcription factor HNF1A (3,4) and accounts for 63% of diagnosed instances of MODY (5). MODY3 patients have normal glucose tolerance during childhood and early adult life, but show progressive reduction of insulin secretion in response to glucose (6)(7)(8)(9). Glycemia typically increases over time, resulting in the need for treatment with the insulin secretory sulfonylurea drugs. Eventually, 30-40% of patients become insulin-dependent due to progressive deterioration of β-cell function.
HNF1A is a 631-amino acid transcription factor with three major domains: dimerization, DNAbinding, and transactivation. Over 200 HNF1A diabetes-related mutations have been identified in all major ethnic groups (10) and the DNA-binding domain has the highest frequency of mutation in MODY3 patients (5). Understanding the role of HNF1A gene and the pathophysiology of MODY3 has been difficult because of the difficulty in studying human islets or mass with these mutations. Mouse models of HNF1A deficiency do not accurately mimic patient phenotypes (11).
A STEM CELL MODEL OF HNF1A DEFICIENCY 3 Stem cell-derived β-cells provide a useful model system, and have been used to study β-cell development in humans (12,13) and to recapitulate disease phenotypes (14,15). Differentiation of pluripotent stem cells to pancreatic endocrine cells can be achieved by a multistep protocol resulting in islet-like clusters containing all endocrine cell types (16)(17)(18). Transplanting these isletlike clusters into mice allows functional testing of stem cell-derived β-cells (scβ-cells) in vivo (19,20). Here we show that human stem cell-based models of HNF1A deficiency shows islet developmental bias towards a-cells, altered insulin granule morphology and the stoichiometry of insulin:C-peptide secretion in vitro and in vivo. We used these models to identify disordered calcium homeostasis and glucose-mediated exocytosis as key mechanisms accounting for the secretory defects observed. This study was designed using two cell-based models: Human embryonic stem cell (hESC) lines null for HNF1A in Fig. 1-5 and Fig. S1-S10; and induced pluripotent stem cell (iPSC) lines with MODY3 patient-specific mutations in Fig. 6 and Fig. S11-S13. Fig. S1 provides a schematic overview of the study.

Generation of isogenic cell lines with HNF1A mutations in hESCs and MODY3 iPSCs
To elucidate the cellular functions of HNF1A, we used CRISPR/Cas9 technology to generate hESC lines (Mel1) harboring non-naturally occurring heterozygous and homozygous null mutations. The  Table S1) because the DNA-binding domain has the highest frequency of mutation in MODY3 patients (5). Transfection of hESC lines with Cas9-GFP and sgRNAs #12 or #14, followed by sorting for GFP (Fig. S2B), achieved 17.3 and 20.4% indel efficiency for the sgRNAs as shown by surveyor assay (Fig. S2C). Gene editing was efficient, resulting in compound heterozygous knockouts and heterozygous mutations ( Fig. S2D) with no off-target mutations detected through targeted sequencing (Table S2). Heterozygous (hESC HNF1A Het) and compound heterozygousnull mutant cell lines (hESC HNF1A KO1 and KO2) with different indels in exon 3 and premature protein termination were chosen for further studies (Fig. S2E). hESC HNF1A Het, KO1 and KO2 lines retained the HNF1A dimerization domain (truncated HNF1A), while hESC HNF1A KO3 line was mutated at the start codon, deleting the entire HNF1A protein (Fig. S2E).
To understand the consequences of MODY3 patient-specific heterozygous mutations, we also generated iPSC lines from MODY3 patient fibroblasts containing heterozygous mutations in the transactivation domain (MODY3 iPSC Het: +/460_461insCGGCATCCAGCACCTGC); and DNA-binding domain (MODY3 iPSC Het: +/R200Q). The R200Q variant has been previously associated with MODY3 (24); however, the precise effect of this missense mutation on HNF1A function is unknown, but very likely pathogenic (24). Using CRISPR/Cas9, we corrected the R200Q mutation from MODY3 iPSC Het (Fig. S2F) with an efficiency of 7.9% (5 clones out of 63) to generate isogenic wildtype cells. In addition to the hESC knockout lines, we also generated iPSCs compound heterozygous knockout lines with an efficiency of 42.9% (27 clones out of 63) (Fig. S2D). An iPSC line with a deletion of the WT allele, but intact patient's mutation (MODY3 iPSC: R200Q/-) was chosen for further studies (Fig. S2G). We also introduced the same R200Q A STEM CELL MODEL OF HNF1A DEFICIENCY 5 mutation into both alleles in the hESC line (hESC Hom: R200Q/R200Q) ( Fig. S2H) with an efficiency of 20.8% (5 clones out of 24) (Fig. S2I). All genetically manipulated cell lines (Fig.   S2J) resulted in modified versions of HNF1A protein ( Fig. S2K and S2L) and had a normal karyotype ( Fig. S2M and S2N) and no off-target mutations (Table S2).

HNF1A is not required to generate pancreatic endocrine cells
To determine the functional consequences of HNF1A deficiency, the first part of this study will focus on the hESC model. Stem cells were differentiated into the pancreatic lineage using a published differentiation protocol (20) (Fig. S3A). Differentiation of wildtype stem cells using this protocol consistently generated 82% PDX1 + /NKX6.1 + cells at the pancreatic progenitor stage (day 11). At the endocrine stage (day 27), we obtained 45% PDX1 + /CPEP + cells with 60% of them coexpressing NKX6.1 (scb-like cells), 30% GCG + cells (sca-like cells) and 10% SST + cells (scδlike cells) (Fig. S3B).
To determine the timing of HNF1A expression, we performed qPCR at different stages of differentiation. Insulin mRNA was detected at the endocrine stage (day 20) (Fig. S3C), whereas HNF1A deficiency did not affect the proportion of scδ-like cell fate as shown by flow cytometry In order to assess the developmental requirements of HNF1A in endocrine cells in vivo, we transplanted pancreatic islet-like clusters derived from hESC HNF1A WT and hESC HNF1A KO lines with GAPDH Luciferase/wt and INS GFP/wt dual-reporter (21). Mice received ~180 clusters of stem cell-derived islet-like cells (Fig. S7A) containing similar amount of CPEP/PDX1 positive cells across genotypes ( Fig. S7B and S7C); but with different amounts of bihormonal GCG/CPEP positive cells (Fig. 2D). Cell Transplantation was done by injection into the ventral and medial muscle groups of the left quadriceps in NOD SCID gamma immunodeficient mice (NSG mice) A STEM CELL MODEL OF HNF1A DEFICIENCY 11 (Fig. 2F). The skeletal muscle was chosen for its dense vasculature and easy access for surgical procedures. Skeletal muscle has been used for other endocrine transplants, including for parathyroid auto-transplantation in patients undergoing parathyroidectomy with a 93% success rate (33). Analysis of engraftment efficiency was evaluated by bioluminescence intensity (BLI) several weeks post-transplantation. Mice with successful engraftment showed a two-fold increase in BLI 4 to 6 weeks post transplantation, while those with failed engraftment showed a decrease ( Fig. S7D). Transplantation was successful in 79% (31/39) of mice, independent of HNF1A genotype (Fig. S7E). 92.5% (49/53) of mice transplanted with hESC lines remained teratoma-free ( Fig. S7F). In those mice, the graft explant (30 weeks post-transplantation) (Fig. S7G) was a vascularized tissue of about ~220 mg (Fig. S7H). According to luminescence intensity, there was a gradual increase in BLI from week 0 to 30 post-transplantation without significant differences by HNF1A genotype (Fig. 2G). Grafts remained localized for up to 50 weeks post-transplantation, and in no case (0/39) was luminescence detected in an ectopic location. These results show that skeletal muscle is a stable and suitable transplantation site for SC-derived islet-like cells.
Thirty weeks post-transplantation, we isolated grafts from normoglycemic animals. Quantification of scb-mass as determined by GFP fluorescence did not differ by HNF1A genotype (Fig. 2H-2I and S7I-S7J). The presence of exclusively mono-hormonal endocrine cells, including scb-cells (CPEP + /PDX1 + /NKX6.1 + ), as well as glucagon-and somatostatin-positive cells were apparent in both hESC HNF1A WT and hESC HNF1A KO grafts (Fig. 2J). No double hormone-positive cells were identified in cells of either genotype. Consistent with the in vitro studies, the percentage of sca-cells in HNF1A KO compared to HNF1A WT islet-like structures were increased by 24% ( Fig.   2K and S7K), leading to a significant increase in GCG + /CPEP + cell type ratio (Fig. 2L); similar 12 results as found in cadaveric human islets of a MODY3 donor (+/T260M) (29) (Fig. 2K).
Consistent with this observation, ex vivo analysis of isolated hESC HNF1A KO grafts showed an increase in glucagon protein content by ELISA ( Fig. 2M and S7L-S7M). Glucagon secretion was higher in hESC HNF1A KO grafts when stimulated with glibenclamide, a second-generation sulfonylurea drug compared to hESC HNF1A WT grafts (Fig. 2N). Similarly, elevated glucagon secretion upon KCl stimulation was detected in vitro from hESC HNF1A KO-derived endocrine cells (Fig. S7N). These results point not only to a gain of a-cell number, but also enhanced a-cell function due to HNF1A deficiency. This is consistent with the up-regulation of genes (PKM, CALM1 and CALM2) involved in glucagon signaling pathways in sca-like cells.
To determine whether the downregulation of PAX4 seen in HNF1A mutant cells (Fig. 1H) could reproduce the bias towards the a-cells, we generated hESC PAX4 KO cell lines using CRISPR/Cas9 (Fig. S7O). The PAX4 gene is an essential gene for differentiation of insulinproducing b-cells in the mammalian pancreas (34). Knockout of PAX4 in rabbits induces a decreased in number of b-cells and increased number of a-cells (26), whereas overexpression of PAX4 reduces glucagon expression in differentiating hESCs (35). Endocrine cells differentiated from an hESC PAX4 KO line were characterized by significant decrease in scb-like cells compared to hESC WT and HNF1A KO lines. In the hESC PAX4 KO lines, sca-like cells were increased to levels similar to hESC HNF1A KO cells by immunohistochemistry (Fig. 2O) and flow cytometry ( Fig. S7P). Therefore, in HNF1A KO lines, down-regulation of PAX4 and de-repression of glucagon gene are associated with the developmental bias towards a-cell fate.
A STEM CELL MODEL OF HNF1A DEFICIENCY

13
In summary, HNF1A is not required for development of mono-hormonal endocrine cells (a-, b and δ-cells). But restriction of HNF1A activity increases the number and proportion of a-cells in association with PAX4 downregulation. As HNF1A clearly regulates a network of genes, a role of additional HNF1A target genes in b-and a-cell fate is likely.

HNF1A deficiency affects glucose-mediated exocytosis of insulin granules
To determine the consequences of HNF1A deficiency on β-cell function, we measured glucosestimulated insulin secretion. hESC HNF1A KO scb-like cells had reduced basal insulin secretion ( Fig. S8A) and impaired glucose-stimulated insulin secretion compared to WT scb-like cells and compared to human pancreatic islets in vitro (Fig. 3A). These differences in hormone secretion between HNF1A mutant and HNF1A WT scβ-like cells were not due to a reduction in insulin content (Fig. S8B). In contrast to glucose, treatment with the insulin secretagogues, tolbutamide or potassium chloride resulted in insulin secretion in hESC HNF1A KO cells comparable to hESC HNF1A WT b-like cells (Fig. 3B). Thus, membrane depolarization enables recruitment of an insulin granule reservoir that is abnormally retained in HNF1A mutant b-cells and fails to be secreted in response to glucose. The ability of sulfonylureas in releasing insulin from β-cell granules is consistent with the clinical efficacy of sulfonylurea drugs in MODY3. These defects are likely intrinsic to the b-cell, and not due to paracrine effects of glucagon (36).
As calcium signaling is a mechanism for the release of insulin through granule exocytosis, we measured intracellular calcium levels in dispersed scb-like cells in vitro. Both genotypes had glucose-stimulated calcium responses (Fig. 3C). However, hESC HNF1A KO b-like cells had significantly lower calcium levels relative to hESC HNF1A WT scb-like cells as measured by A STEM CELL MODEL OF HNF1A DEFICIENCY 14 Fura-2 fluorescence (Fig. 3C). This was further supported by measurements of absolute intracellular calcium concentrations by Fura-2 fluorescence (Fig. 3D). Among the HNF1A target genes, reduced expression of CACNA1A is potentially involved in reducing intracellular calcium levels, and SYT13 may be required for efficient exocytosis-mediated insulin secretion. CACNA1A encodes a voltage-dependent calcium channel mediating the entry of calcium into excitable cells and is involved in calcium-dependent insulin secretion and type 2 diabetes (37). Synaptotagmins are calcium sensors localized in the β-cell insulin granules, and are required for vesicle fusion and glucose-stimulated insulin release (38-40). SYT13 is a member of the synaptotagmin family and is predicted to be involved in calcium-regulated exocytosis. SYT13 is down-regulated in human T2D islets and silencing of SYT13 impairs insulin secretion in INS1-832/13 cells (41). To determine whether down-regulation of CACNA1A and SYT13 seen in HNF1A mutant cells (Fig. 1H) reproduced the reduced intracellular calcium levels and reduced glucose-stimulated insulin secretion, we generated hESC CACNA1A KO and SYT13 KO cell lines using CRISPR/Cas9. Sanger sequencing revealed homozygous mutations resulting in frame shifts (Fig. S8C).
Differentiation of hESC CACNA1A KO and hESC SYT13 KO generated INS-GFP organoids ( Fig.   3E) comprised of ~51% PDX1/CPEP and ~16% GCG positive cells ( Fig. 3F and S8D) with no differences compared to isogenic WT cells. Dispersed hESC CACNA1A KO b-like cells (Fig. S8E) had significantly reduced intracellular calcium levels compared to hESC WT cells ( Fig. 3G  to HNF1A KO scb-like cells (Fig. 3H). Treatment with tolbutamide or KCl stimulated insulin secretion in all KO lines comparable to the corresponding WT lines. Impairments in insulin release due to reduced levels of these molecules can be overcome by elevating calcium beyond physiological levels using tolbutamide or KCl. The milder phenotypes of cells deficient for single target genes compared to HNF1A KO cells suggests that the combined effects on several HNF1A target genes may be responsible for reduced glucose-stimulated insulin secretion.

HNF1A deficiency alters the stoichiometry of insulin to C-peptide secretion
To interrogate the function of hESC HNF1A KO b-cells in vivo, we monitored circulating human insulin and C-peptide concentrations in euglycemic mice transplanted with sc-islet-like clusters.  (Fig. 4B) despite high plasma human C-peptide levels (>300 pM).
These differences in hormone secretion between HNF1A mutant and HNF1A WT scβ-cells were not due to reduced scβ-cell mass in vivo (Fig. 2I). These findings in vivo led us to quantify the A STEM CELL MODEL OF HNF1A DEFICIENCY 16 secretion ratio of those hormones in scb-cells in vitro. For equivalent secretion of human C-peptide at basal glucose condition (Fig. S8G), secretion of human insulin was significantly reduced in hESC HNF1A KO lines (Fig. S8H), leading to decreased insulin:C-peptide secretion ratios in vitro ( Fig. S8I). Thus, the absence of insulin in the plasma was not a limitation of the sensitivity of the assay but a defect in the stoichiometry of insulin to C-peptide in circulation. While plasma of mice grafted with hESC HNF1A WT b-cells consistently showed an insulin:C-peptide molar ratio of 0.22±0.11 from week 4 to 30 post-transplantation, mice grafted with hESC HNF1A KO β-cells had an 18-fold lower ratio (Fig. 4C). This decrease in circulating insulin:C-peptide ratio was reciprocal to a 3-fold increase in the insulin:C-peptide ratio of intracellular content from hESC HNF1A KO isolated grafts (Fig. 4D), demonstrating complementary imbalance of insulin to C-peptide. The altered insulin:C-peptide ratio was not due to differences in insulin processing since insulin:proinsulin ratios in hESC HNF1A KO and hESC HNF1A WT grafts were identical in vivo ( Fig. 4E) and in vitro ( Fig. S8J-S8L) and no differences were found in the transcript levels of processing genes (PC1/PC3) in RNA sequencing analysis (Table S3-S5). Impaired stoichiometry of circulating insulin to C-peptide was also observed in mice transplanted with two additional homozygous mutant hESC lines (Fig. S8M). Therefore, HNF1A deficiency not only impairs insulin secretion, but also the stoichiometry of circulating insulin to C-peptide ratios.
To evaluate glucose-stimulated insulin secretion, we performed an intraperitoneal glucose tolerance test (iPGTT) in transplanted mice. During fasting and following an iPGTT, mice transplanted with hESC HNF1A WT β-cells displayed normal human insulin and C-peptide secretion profiles (Fig. 4F-4H). In contrast, hESC HNF1A KO β-cells were non-responsive to glucose: plasma human C-peptide was not decreased by fasting and did not increase after glucose injection with concentrations significantly lower than in mice transplanted with hESC HNF1A WT β-cells (Fig. 4G). Remarkably, circulating human insulin remained undetectable in animals engrafted with hESC HNF1A KO β-cells despite high plasma levels of human C-peptide (Fig. 4H).
Therefore, HNF1A deficiency affects glucose-stimulated insulin secretion, and C-peptide is released in a constitutive manner independent of metabolic state, resulting in an altered ratio of insulin to C-peptide.

Glibenclamide restores insulin secretion in HNF1A deficient scβ-cells
In contrast to glucose, intra-peritoneal injection of the sulfonylurea drug, glibenclamide, in mice transplanted with hESC HNF1A KO β-cells showed a significant increase in C-peptide 10 minutes after glibenclamide injection (Fig. 4I). Surprisingly, human insulin concentrations in mice transplanted with hESC HNF1A KO β-cells increased from undetectable levels to 14±21 mU/L, reaching levels comparable to those in the control group (Fig. 4J). Insulin:C-peptide ratios in plasma of animals transplanted with hESC HNF1A KO β-cells were increased 10-fold by glibenclamide, equal to the ratios in mice transplanted with hESC HNF1A WT β-cells (Fig. 4K).
Clearance of insulin from the circulation occurred with the same kinetics in mice grafted with hESC HNF1A WT and hESC HNF1A KO β-cells 20 minutes after glibenclamide administration ( Fig. 4K), excluding insulin stability or clearance in plasma as the cause for the low insulin levels in mice transplanted with HNF1A mutant scβ-cells. To further test whether the reduced insulin:Cpeptide secretion ratios from hESC HNF1A KO β-cells were due to secretory defects and not insulin stability or clearance, we isolated the hESC HNF1A KO grafts from mice and identified a significant decrease in insulin:C-peptide secretion ratios compared to HNF1A WT grafts ex vivo.
This ratio was restored after exposure of the grafts to glibenclamide (Fig. 4L).
A STEM CELL MODEL OF HNF1A DEFICIENCY 18 Thus, membrane depolarization initiated by closure of the ATP-sensitive K + (K ATP ) channels (target of glibenclamide) and high intracellular calcium levels allow the secretion of insulin granules that are abnormally retained in HNF1A mutant b-cells.

HNF1A deficiency causes abnormal insulin granule structure
To identify the basis of the insulin secretory defect in hESC HNF1A KO β-cells we examined insulin granule morphology by electron microscopy from 30-week explants of normoglycemic mice. hESC HNF1A WT grafts were comprised of scβ-cells with a characteristic high-electron density core separated from limiting membranes by a "halo" (Fig. 5A and S9A). These morphologies are consistent with electron microscopy of normal human β-cells. In contrast, hESC HNF1A KO islets-like were comprised of scβ-cells where the majority (>90%) of the insulin granules were abnormally enlarged ( Fig. 5A-5C) with a diffuse electron-light insulin core ( Fig.   5D-5E and S9A), characteristic of immature secretory granules. Enlarged insulin granules were further confirmed by immunogold staining for human C-peptide (Fig. 5F). Similar results were observed in vitro, where hESC HNF1A mutated cells had increased insulin granule diameter with a diffuse light insulin core ( Fig. S9B and S9C). For sca-cells, no differences in morphology, structure or granule size between HNF1A genotypes was seen (Fig. S9D). While these abnormal insulin granules are not secreted in response to glucose, they can be secreted in response to sulfonylureas. These results show that HNF1A is required for normal insulin granule formation and function.

HNF1A deficient scβ-cells are unable to maintain glucose homeostasis in diabetic mice
To test the ability of hESC HNF1A KO β-cells to maintain in vivo normoglycemia, endogenous murine β-cells were ablated by the administration of streptozotocin (STZ) 22 weeks posttransplantation. STZ is a drug that, by virtue of species-related difference in glucose transporters, specifically targets murine β-cells without affecting human β-cells (42). We first confirmed that HNF1A deficiency did not increase sensitivity of human cells to STZ. There was no change in cell death ( Fig. S10A) or C-peptide secretion (Fig. S10B) when HNF1A WT or KO scβ-like-cells were exposed to STZ in vitro. In vivo, circulating plasma human C-peptide from mice transplanted with HNF1A WT or HNF1A KO cells was unchanged or increased post STZ treatment (Fig. 5G). Mouse C-peptide was undetectable by ELISA ( Fig. S10C) and immunohistochemistry of the pancreas ( Fig. S10D) demonstrating successful and specific ablation of mouse β-cells. Mice transplanted with the hESC HNF1A WT β-cells were normoglycemic (~100 mg/dl and HbA1C ~5%) for at least 10 weeks post STZ injection ( Fig. 5H and S10E) and had normal human insulin and Cpeptide secretion profiles during an iPGTT ( Fig. S10F-S10H). In contrast, mice transplanted with hESC HNF1A KO β-cells did not increase human C-peptide secretion (Fig. 5G) after becoming severely diabetic (blood glucose >500 mg/dl) within one week following STZ injection (Fig. 5H).
Consistent with in vitro studies, hESC HNF1A KO β-cells are unresponsive to blood glucose even at high levels, failing to maintain systemic glucose homeostasis in a mouse model of β-cell deficient diabetes.

HNF1A R200Q homozygous mutation is pathogenic and causes a developmental bias towards the a-cell fate in vitro
These studies demonstrate the molecular, cellular and functional consequences of complete HNF1A deficiency, a situation that does not occur clinically, but was created experimentally to amplify the effects of HNFIA functional hypomorphism in order to enable mechanistic studies.  S11C). In contrast to HNF1A truncated lines, the heterozygous point mutation R200Q of MODY3 iPSC line did not affect the total amount of HNF1A protein, suggesting that the mutant R200Q protein was produced (Fig. S11C). No significant differences were found in endocrine cell types (sca-, scb-and scδ-like cells) between HNF1A WT and heterozygous cell lines at day 27 of differentiation by immunohistochemistry (Fig. S11D).
To determine the consequences of heterozygosity for HNF1A in our MODY3 patient lines for the bias towards a-cell fate, we measured the percentage of glucagon cells co-expressing C-peptide in two MODY3 iPSC Het (+/460ins and +/R200Q) lines in vitro. We found no significant differences between MODY3 iPSC lines and control iPSC R200Q-corrected WT lines (Fig. S11E). To determine the pathogenicity of the HNF1A R200Q mutation, we knocked out of the HNF1A A STEM CELL MODEL OF HNF1A DEFICIENCY 21 wildtype allele in a MODY3 iPSC Het line (R200Q/-), which resulted in a significant ~70% increase in GCG and CPEP co-expressing cells compared to the R200Q-corrected WT control lines ( Fig. S11E and S11F). We also detected a significant increase in the percentage of GCG and CPEP co-expressing cells in hESC HNF1A R200Q homozygous (R200Q/R200Q) and hESC HNF1A Het (+/-) lines compared to hESC WT (+/+) lines ( Fig. S11G and S11H)

HNF1A haploinsufficiency gradually impairs scβ-cell functional capacity in the context of metabolic stress
In order to assess the developmental potential of MODY3 patient-specific mutations in vivo, we transplanted pancreatic islet-like clusters derived from iPSCs using identical methods as previously described. While 87.9% (51/58) of mice transplanted with isogenic MODY3 iPSC Het lines (+/R200Q) and R200Q-corrected WT (+/+) were teratoma-free; the majority (61.5%, 8/13) of the MODY3 iPSC Het (460ins) and control iPSC WT lines showed teratoma formation (Fig.   S11I). Because of variable teratoma formation among different iPSC lines (20), only teratomafree mice were used for further analysis. Teratoma-free transplanted mice had 67.3% (35/52) A STEM CELL MODEL OF HNF1A DEFICIENCY 22 engraftment efficiency. Thirty weeks post-transplantation, we isolated grafts from normoglycemic animals ( Fig. S11J) and found no significant differences in glucagon-to-insulin content ratios or endocrine cell types between MODY3 iPSC Het (+/R200Q) and MODY3 iPSC R200Q-corrected WT (+/+) control grafts ( Fig. S11K and S11L). These results are consistent with our iPSC in vitro results; however, a recent study from cadaveric MODY3 human islets with a T260M heterozygous mutation in DNA-binding domain (29), showed a bias of endocrine cells toward the a-cell fate Five to six weeks post STZ treatment, we found that the ratio of circulating insulin to C-peptide fell by 55% in mice transplanted with MODY3 iPSC R200Q Het (+/R200Q) cells. In contrast, insulin:C-peptide ratios remained constant in mice transplanted with MODY3 R200Q-corrected WT (+/+) cells (Fig. 6F). The decrease in circulating insulin:C-peptide ratios was reciprocal to a significant increase by 74% in the insulin:C-peptide ratios of intracellular content from isolated HNF1A mutated grafts compared to isogenic controls (Fig. 6G), demonstrating complementary imbalance of insulin to C-peptide secreted. These results show that HNF1A R200Q haploinsufficiency gradually impairs the stoichiometry of circulating insulin:C-peptide, and that gene correction of the R200Q mutation protects MODY3 iPSC β-cells from acquiring this imbalance.
A STEM CELL MODEL OF HNF1A DEFICIENCY

25
We then performed iPGTTs two, four, six and eight weeks post STZ administration. MODY3 iPSC R200Q Het (+/R200Q) scβ-cells showed progressive impairment of glucose-stimulated insulin secretion: mild at two weeks post STZ ( Fig. S13A and S13B) and completely unresponsive insulin release to glucose challenges four weeks post STZ ( Fig. S13C and S13D). Mice transplanted with MODY3 scβ-cells fail to clear glucose from the circulation during a glucose tolerance test at six to eight weeks post STZ treatment ( Fig. 6H and S13E). HNF1A R200Q heterozygous (+/R200Q) scb-cells had no glucose-stimulated insulin secretion, whereas MODY3 R200Q-corrected WT (+/+) scβ-cells were sensitive to fasting, remained glucose responsive six to eight weeks post STZ treatment, and cleared glucose from circulation ( Fig. 6H-6I and S13E-S13F). Mice transplanted with MODY3 iPSC R200Q Het (+/R200Q) scβ-cells became diabetic as shown by elevated blood glucose (Fig. S13G) and HbA1c levels ( Fig. 6J) compared to control mice. These results show that MODY3 iPSC β-cells (+/R200Q) fail to compensate for higher insulin demands, and that the hyperglycemia is due to gradual development of insulin secretory defects, characterized by a disruption of the stoichiometry of insulin and constitutive C-peptide release. Correction of R200Q mutations protects MODY3 iPSC β-cells from acquiring these abnormal insulin secretion profiles.

Discussion
Here we report the use of stem cell-derived islets containing HNF1A mutations to elucidate the molecular basis for apparent β-cell dysregulation of insulin secretion in HNF1A hESC null and In silico gRNA design. sgRNAs (Table S1) were designed using an online CRISPR design tool Dynamic insulin secretion. Perifusion was performed as described previously (49) 1). To note that GRCh38 reference genome was applied for the whole analysis. "cellranger count" with "--expect-cells=3,000" was run for each library respectively and "cellranger aggr" with default parameters was run to aggregate results generated from multiple libraries. A normalized umi-count matrix for all the cells was then generated. As for the quality control, cell and gene filtering process, cells with umi-count<5,000 and umi-count>40,000 and cells with <2,000 detected genes were removed. Also, genes with <~1% (25) non-zero counts across all the cells were eliminated from the analysis. To get rid of the Mitochondria cells, cells with more than 10% Mitochondria genes' count were also removed. All the downstream analysis was done with Seurat package in R-3.4.1. Every batch is an independent differentiation (biological replicates). n=2 for each genotype.       statistical significance (-log 10 p-value, y-axis) for differential gene expression (down-regulated in green; up-regulated in red) by bulk RNAseq (see also Table S3)