Patients with NEUROG3 mutations suffer from diabetes mellitus due to pancreatic endocrinopathy and chronic malabsorptive diarrhea due to enteric endocrinopathy. The role of NEUROGENIN3 (NEUROG3) as a master regulator of pancreatic and intestinal endocrine differentiation has been modeled in detail using transgenic animals and human embryonic stem cells 1-8. Animal models with biallelic deletions of Neurog3 have no pancreatic or enteric endocrine cells and die within hours after birth. Human embryonic stem cell models of NEUROG3-independent differentiation have concluded that very few pancreatic endocrine cells can differentiate in the absence of NEUROG3, and these differentiated cells are likely not functional 1,7,9,10. In contrast, patients with NEUROG3 mutations develop insulin-dependent diabetes, with onset ranging from 13 days to >23 years of age 11. Patients also develop severe generalized malabsorptive diarrhea that fails extensive dietary challenges in the immediate postnatal period. It is thought that the clinical phenotypes are less severe than the deletion models because the hypomorphic clinical mutations retain some functionality lost in the null model systems.
We identified a 16-year-old male with a biallelic frameshift mutation in the NEUROG3 gene (c117delC; P39PfsX38) that resulted in a truncated protein that is functionally null for any biochemical activity 12. At 2 weeks of age he was diagnosed with malabsorption and was temporarily managed with parenteral nutrition. At nearly 3 years of age, he was admitted with a blood glucose of >600 mg/dL and diagnosed with diabetes. He was negative for islet autoantibodies and his c-peptide levels declined over time (Supplementary Fig 1). Immunohistochemical staining of endocrine cells on the subject’s duodenal biopsy was negative, confirming the null phenotype of the patient’s mutation (Figure 1A). The contrast between the phenotype of null model systems and our biochemically null P39PfsX38 subject’s clinical presentation suggests that the molecular mechanisms by which NEUROG3 regulates human pancreatic cell differentiation are complex and require further investigation.
We created induced pluripotent stem cells (iPSCs) from the subject’s fibroblasts to investigate how the P39PfsX38 mutation affected pancreas differentiation. Native iPSC clones were referred to as NEUROG3NULL. We replaced the mutated single exon of NEUROG3 with a wild-type copy of the exon in the endogenous gene locus using CRISPR-Cas9 cellular engineering (Figure 1B) 13. Corrected clones were sequenced to verify biallelic correction of the mutation and were referred to as NEUROG3CORR (Figure 1C). NEUROG3NULL and NEUROG3CORR iPSCs had normal karyotypes, expressed pluripotency markers, and had low expression of common lineage markers (Supplementary Figure 2A-D). NEUROG3NULL and NEUROG3CORR cells appropriately expressed endoderm, ectoderm, and mesoderm markers when subjected to lineage-specific differentiation protocols (Supplementary Figure 3).
To identify the developmental stage at which the NEUROG3-P39PfsX38 mutation disrupted human pancreas differentiation, we assessed the cell fates of NEUROG3NULL and NEUROG3CORR cells at key steps of a reproducible beta cell differentiation protocol 14. Both cell lines differentiated into definitive endoderm at similar rates (Supplementary Figure 4). Next, we assessed the differentiation of mature, competent pancreatic progenitor cells by the co-expression PDX1 and NKX6.1 15,16. Immunocytochemical staining and flow cytometry established that NEUROG3NULL cells differentiated into similar numbers of PDX1+ cells as NEUROG3CORR cells, but significantly fewer PDX1+/NKX6.1+ competent pancreatic progenitor cells (9% vs 24% by flow cytometry, respectively, n=6, p<0.001; Figure 2A-B). Quantitative PCR confirmed that the expression of PDX1 was similar in both cell lines, but NEUROD1 expression, a NEUROG3-dependent transcription factor, was significantly decreased in NEUROG3NULL cells (Figure 2C, p<0.05, n=4). Expression of the NEUROG3-independent pro-endocrine gene MAFB was unchanged (Figure 2C).
To understand the global impact of the P39PfsX38 mutation on the differentiation of pancreatic progenitor cells, we assessed the RNA-transcriptome of NEUROG3NULL and NEUROG3CORR pancreatic progenitor cells. We normalized the expression profiles against pluripotent H1 embryonic stem cells. As expected, pluripotency genes were enriched in the H1 cells, whereas genes involved in endodermal fate commitment were similarly enriched in both NEUROG3NULL and NEUROG3CORR cell lines. However, the expression profiles of NEUROG3NULL and NEUROG3CORR cells diverged among genes associated with the pancreatic progenitor cell fate (Figure 2D).
Next, we attempted to differentiate both NEUROG3NULL and NEUROG3CORR cell lines into the pancreatic beta-like cell lineage (ßLCs) (Figure 3A). Immunocytochemistry revealed that NEUROG3NULL cells could not generate significant numbers of ßLCs in vitro, while NEUROG3CORR cells differentiated readily (Figure 3B). Rare NEUROG3NULL endocrine cells were polyhormonal and did not coexpress c-peptide and NKX6.1, (Figure 3B) 15,16. We quantified the number of c-peptide+ cells present in both the NEUROG3NULL and NEUROG3CORR ßLCs. Less than 1% of NEUROG3NULL ßLCs were c-peptide+, while 22.24% of NEUROG3CORR ßLCs were c-peptide+ (Figure 3C, p<=0.02, n=5). INS transcripts were very low in NEUROG3NULL ßLCs, but were highly expressed in NEUROG3CORR ßLCs (Figure 3D, p<0.001, n=3-5). We measured the insulin secretory index, which is the ratio of insulin secreted at high glucose and low glucose, in NEUROG3NULL and NEUROG3CORR ßLCs cultured as 3D clusters for 3 days in vitro using a static glucose-stimulated insulin secretion assay. We then transplanted clusters under the kidney capsule of NSG mice to facilitate maturation. After 12 weeks, we measured human insulin levels in the plasma of mice that were fasted then subjected to a glucose challenge (Figure 3E). Before transplantation, neither NEUROG3NULL nor NEUROG3CORR cell clusters appropriately increased insulin secretion in response to glucose (see pre-implant bars, Figure 3E). 12 weeks after transplantation, NEUROG3CORR cells mounted a mature human insulin secretory response to glucose, while NEUROG3NULL cell response was still immature (see post-implant bars in Figure 3E).
Because NEUROG3NULL iPSCs did not differentiate efficiently into ßLCs and did not express mature pancreatic progenitor cell markers, we hypothesized that these cells might have a diminished capacity to differentiate into the pancreatic exocrine lineages. To this end, we assessed exocrine differentiation. NEUROG3CORR cells were able to differentiate into acinar and ductal cells, while far fewer NEUROG3NULL cells were positive for the exocrine lineages (Supplementary Figure 5). This raised the possibility that the patient’s mutation might compromise pancreatic progenitor cell contribution to organogenesis in vivo and that the subject may have undiagnosed exocrine pancreatic insufficiency (EPI). Chronic malabsorptive diarrhea is a shared clinical manifestation of enteric anendocrinosis and EPI.
EPI’s formal diagnosis includes chronic malabsorptive diarrhea, low fecal elastase levels, and diminished exocrine secretory response to hormone stimulation. Diagnostic secretin-enhanced magnetic resonance cholangiopancreatography (MRCP) of the subject’s pancreas indicated it was hypoplastic (Fig 4A, Supplementary video 1). Quantification of pancreatic volume from MRCP revealed a 5-fold decrease in the patient’s pancreatic volume index compared to the average index of age-matched control subjects (Figure 4B). The patient’s fecal elastase level was very low and exocrine pancreatic function tests (EPFT) revealed decreased exocrine pancreatic enzyme levels after secretin stimulation (Figure 4B) 17-19. The subject was diagnosed with clinical EPI and started pancreatic enzyme replacement therapy (PERT) to treat EPI. He demonstrated weight gain, fat absorption, and resolved fat-soluble vitamin deficiency with PERT treatment.