The SNAG Domain of Insm1 Regulates Pancreatic Endocrine Cell Differentiation and Represses β- to δ-Cell Transdifferentiation

Pancreatic endocrine cells are differentiated from Ngn3⁺ progenitor cells and undergo terminal differentiation to form mature α, β, δ, pancreatic polypeptide (PP), and ε cells via complex regulatory networks (1,2). During differentiation, glucagon-positive α-cells and PP-positive cells are first detected at embryonic day 10.5 (E10.5), and the coexpression of these two hormones is observed following the early developmental stages (3). Insulin expression in β-cells is initially detected at E11.5 (3), whereas somatostatin-positive δ-cells appear at ∼E13.5 (3). Ghrelin expression can be detected as early as on E10.5 (2). From E15.5 to E18.5, endocrine cells undergo a rapid expansion; however, β-cells expand much more rapidly than the other types of endocrine cells and become the major population of pancreatic endocrine cells (3,4). By E18.5, islets of Langerhans start to form and the expression of hormone acquires the characteristic one cell–one hormone pattern (3).

The differential expression of distinct transcription factors determines the allocation and specification of the pancreatic endocrine cell lineages (5). Among these, Ngn3 is transiently expressed in endocrine progenitor cells and determines their differentiation fate (1,6,7). Furthermore, Pdx1 is initially expressed in the pancreatic anlagen, and its expression is essential for the development of all pancreatic lineages (8). During development, the expression of Pdx1 is gradually enriched in pancreatic endocrine cells, particularly in β-cells (4). The combinatorial expression of Pdx1, Nkx2.2, and Nkx6.1 has been shown to be essential for β-cell differentiation (4,9,10). Additional factors, such as Pax4, promote the specification of β- and δ-cell fate and suppress α-cell fate (11–13), whereas Pax6 and Arx are required for α-cell differentiation (12,14). In addition, MafB is required for both α- and β-cell differentiation (15,16). Lastly, hematopoietically expressed homeobox (Hhex) is the only identified δ-cell–specific factor that is required for δ-cell differentiation (17).

Functionally, α- and β-cells inversely regulate blood glucose levels and maintain euglycemia by tightly regulating glucagon or insulin secretion. In contrast, δ-cells tonically inhibit glucagon and insulin secretion and contribute to the therapeutic effects observed in patients with diabetes (18,19). In mouse islets, β-cells are located in the central region, whereas the other endocrine cells locate in the periphery. Morphologically, δ-cells display a complex morphology with long neurite-like processes and can directly come into contact with other endocrine cells in the islets. This suggests that δ-cells are an important regulator and a novel pharmacological target in the islets (20). Furthermore, δ-cells rapidly contribute to the increase in the number of β-cells in individuals with type 1 diabetes (21).

The insulinoma-associated protein 1 (Insm1) gene encodes a DNA-binding protein that contains the SNAG domain and five zinc finger domains (22). Insm1 null mutation results in delayed pancreatic α-cell differentiation and interferes with terminal differentiation of β-cells. In addition, decreased numbers of δ- and ε-cells are found in Insm1 null mutants, whereas an increase in PP cell numbers is detected in these mutant animals (23,24). Insm1 also plays an essential role in β-cell function in adults. Mutation of Insm1 in adult β-cells, disturbing the function of the Insm1–NeuroD1–FoxA2 complex, leads to an immature β-cell phenotype and hyperglycemia (25). Haploinsufficiency of Insm1 delays cell cycle progression in β-cells during the early postnatal period, resulting in decreased β-cell numbers in adults (26).

The SNAG domain, first identified in Snail/Slug and Gfi1/Gfi1b zinc finger transcription factors, functions as a molecular hook to directly recruit the histone demethylase Lsd1 (Kdm1a). Lsd1 participates in multiple histone modification complexes, including the Lsd1–Rcor1–Hdac1/2 complex (27–29). Lsd1 itself can specifically remove the mono- or dimethyl groups from the fourth amino acid lysine on the histone 3 tail (H3K4me1/2). The histone H3K4me1 and H3K4me2 provides the active chromatin loci for transcriptional activity, removing these modification results in transcriptional repression on associated genes (30). The function of the N-terminus SNAG domain of Insm1 has been studied in pituitary endocrine cells (31). This study showed that Insm1SNAG can directly recruit Lsd1 to regulate the specification of pituitary endocrine cells; in addition, similar deficits in pituitary differentiation have been observed in Insm1 null and Insm1 SNAG domain–specific mutant mice (31). Whether this is a cell-type–specific function of the Insm1 SNAG domain or a broad function in endocrine cells has yet to be determined. In the current study, we used Insm1 SNAG mutant and Insm1 null mutant mice to explore the SNAG domain–specific function of Insm1 in the regulation of pancreatic endocrine cell differentiation.

Comparison analysis was performed in Insm1delSNAG mutant (Insm1^delSNAG/lacZ), Insm1 null mutant (Insm1^lacZ/lacZ), and heterozygous control (Insm1^+/lacZ) mice. The Insm1^+/+ and Insm1^{delSNAG/delSNAG} mice were used to verify the phenotypes observed in the comparison analysis. The β-cell conditional knockout (InsCre;Insm1^f/delSNAG; CKO), Cre-heterozygous control (InsCre;Insm1^f/+), and Cre control (InsCre;Insm1^+/+) mice were used to investigate the transdifferentiation of β- to δ-cells. The mTmG transgene was introduced into the CKO and the Cre-heterozygous control mice for tracing analysis.

The development of Insm1^+/lacZ, Insm1^+/delSNAG, and Insm1^+/flox mice has been described previously (25,26,31). InsCre mice were provided by Dr. Pedro Luis Herrera (Department Genetic Medicine and Development, University of Geneva, Geneva, Switzerland) (32). mTmG (007576; The Jackson Laboratory) mice were provided by Dr. Carmen Birchmeier (Developmental Biology/Signal Transduction, Max Delbrück Center for Molecular Medicine). All animals were housed in a 12-h light/dark environment at a temperature of 21–23°C, with ad libitum access to food and water. Pregnant Insm1^+/lacZ or Insm1^+/delSNAG females who had been mated with Insm1^+/lacZ or Insm1^+/delSNAG males were provided drinking water containing 0.1 mg/mL phenylephrine hydrochloride, 0.1 mg/mL isoprenaline hydrochloride, and 0.1% ascorbic acid to prevent fetal death in the mutants (33). Mice numbers are indicated in each figure legend. All animal experiments were approved by the Institutional Animal Care and Use Committee of Jinan University (IACUC-20201231-03).

Immunofluorescence was performed as described previously (26). Rabbit anti-insulin (20056,1:1,000; Immuno‐Star), guinea pig anti-insulin (4011-01F,1:100; LINCO), rat anti-insulin (MAB1417, 1:500; R&D Systems), rabbit anti-glucagon (20076,1:2,000; ImmunoStar), guinea pig anti-glucagon (1:500; LINCO), rabbit anti-somatostatin (A0566 1:500, DakoCytomation; and HPA019472, 1:2,500, Sigma-Aldrich), rabbit anti-PP (AB939,1:500; Chemicon International), chicken anti–β-galactosidase (β-gal) (ab9361,1:1,000; Abcam), mouse and rabbit anti-ghrelin (ab57222, 1:100, Abcam; and H-031–31, 1:1,000, Phoenix Pharmaceuticals), rabbit anti-MafA (34), rabbit anti-MafB (HPA005653,1:5,000; Sigma-Aldrich), goat anti-Pdx1 (ab47383,1:5,000; Abcam), mouse anti-Isl1/2 (39.4D5, 1:50; DSHB), mouse anti-Ngn3 (F25A1B3, 1:50; DSHB), mouse anti-Pax6 (PAX6-S, 1:20; DSHB), mouse anti-Nkx6.1 (F55A10-C, 1:200; DSHB), and rat anti-Ki67 (14–5698, 1:50; Invitrogen) were used as primary antibodies. Secondary antibodies coupled to Cy3, Cy2, or Cy5 against rabbit, rat, mouse, chicken, or guinea pig (Jackson ImmunoResearch Laboratories) were used. Fluorescence imaging was performed using a Leica confocal microscope, and image processing was performed using Adobe Photoshop software.

Total RNA was isolated from E18.5 pancreata using the TRIzol reagent (Invitrogen). For RNA-sequencing (RNA-seq) analysis, four animals per genotype, Insm1^+/lacZ (control), Insm1^lacZ/lacZ (Insm1 null), and Insm1^delSNAG/lacZ (Insm1delSNAG), were collected. Sequencing libraries were generated using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina (New England Biolabs). Sequencing was performed on an Illumina NovaSeq 6000, and 150-nucleotide paired-end reads were generated. At least 6 GB of clean data with >94% of them above Q30 were produced for each sample. HISAT2 (35) and StringTie (36) were used to align the reads and to analyze the transcripts. The DEGseq R package was used to identify differentially expressed genes (37). The whole analysis was performed on BMKCloud (www.biocloud.net).

For real-time RT-PCR analysis, cDNA was synthesized using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara Bio, Beijing, China), and real-time RT-PCR was performed using the SYBR Fast qPCR Mix (Takara Bio) on a CFX96 RT-PCR system (Bio-Rad Laboratories). Expression levels were determined using the 2^−ΔΔCt method. Actb was used as the internal control, and the values are presented relative to the heterozygous control. The primer sequences are listed in Supplementary Table 1.

Immunoprecipitation and Western blotting were performed as described previously using SJb β-cells (25,31). Anti-Insm1 (25), anti-Lsd1 (ab17721; Abcam), anti-NeuroD1 (cs-4373; Cell Signaling Technology), anti-FoxA2 (sc-6554X; Santa Cruz Biotechnology), and anti-Flag (F1804; Sigma-Aldrich) antibodies were used. ImageQuant LAS 500 (GE Healthcare) was used for chemiluminescent Western blot imaging.

Chromatin immunoprecipitation (ChIP) was performed as previously described (25). SJb cells and freshly isolated islets were used for the ChIP assays. Insm1 mutation in SJb cells was introduced by transfecting a Cre expression plasmid into the SJb cell line containing the Insm1^flox/flox alleles. A wild-type control and InsmdelSNAG mutant islets were used for ChIP. The Amy1 gene promoter region was used as the negative control for ChIP-PCR. Anti-Lsd1 (ab17721; Abcam), anti-H3K4me1 (ab8895; Abcam), anti-H3K4me2 (ab32356; Abcam), and anti-Insm1 (25) antibodies were used. The primers used for ChIP-PCR are listed in Supplementary Table 1.

Data are expressed as mean ± SD. Pairwise comparisons were made using the two-tailed unpaired Student t test. A P value of <0.05 was considered statistically significant. All graphical and statistical analyses were performed using Prism 8 software (GraphPad Software, San Diego, CA) and Microsoft Excel. For the apoptosis analysis, we presented the counting results without a further statistical analysis, due to the very low rates of apoptosis in pancreatic endocrine cells.

The data sets generated during the current study are available from the corresponding authors on reasonable request.

The Insm1 null mutation results in severe developmental phenotypes in pancreatic endocrine cells (23,24). However, whether the SNAG domain of Insm1 accounts for the dysfunctions observed in Insm1 mutant mice is presently unknown. In this study, we aimed to determine the specific function of the SNAG domain of Insm1 in the development of pancreatic endocrine cells by comparing the developmental consequences of Insm1delSNAG and Insm1 null mutations. In particular, we analyzed Insm1 ^delSNAG/lacZ (Insm1delSNAG mutation), Insm1^lacZ/lacZ (Insm1 null mutation), and Insm1^+/lacZ (control) littermate embryos by immunofluorescence.

We stained β-gal (expressed by lacZ allele) to indicate Insm1 expression in these control and mutant embryos. We observed a decreased number of insulin-positive β-cell as assessed by insulin and β-gal antibody costaining of the pancreas of Insm1delSNAG mutants at E18.5, albeit this deficit was less severe when compared with that observed in Insm1 null mutant mice (Fig. 1A–A’’’). No change of α-cell numbers was observed in Insm1delSNAG or Insm1 null mutants as compared with the control (Fig. 1B). Comparable increased numbers of PP and decreased numbers of ε-cells, indicated by PP and ghrelin staining, respectively, were observed in Insm1delSNAG and Insm1 null mutants when compared with the control and to each other (Fig. 1D and E). However, the number of somatostatin-positive δ-cells was significantly increased in the Insm1delSNAG mutants and decreased in Insm1 null mutants as compared with the control mice (Fig. 1C–C’’’).

To investigate whether these deficits occur in an earlier developmental stage (i.e., E15.5), the end of the first wave of endocrine cell development and the start of the rapid expansion of the endocrine cells, we examined the pancreatic endocrine lineages at E15.5. We observed a similar decrease in the number of insulin-, glucagon-, PP-, and ghrelin-positive cells in Insm1delSNAG and Insm1 null mutants as compared with the control and to each other (Fig. 2). However, the number of somatostatin-positive cells was not altered in either Insm1delSNAG or Insm1 null mutants compared with the control mice at E15.5 (Fig. 2C–C’’’). The expression of hormone genes was further verified by real-time RT-PCR (Supplementary Fig. 1). Therefore, the increased δ-cell number in Insm1delSNAG mutants occurred at E18.5 but not at E15.5.

To exclude the contribution of lacZ allele or heterozygous Insm1 to the deficits, we examined the phenotypes using the Insm1^{delSNAG/delSNAG} mutants and wild-type control (Insm1^+/+) mice at both E15.5 and E18.5. We observed similar alteration of the endocrine cell lineages as that observed in Insm1delSNAG mutants versus the heterozygous control at both development stages (Supplementary Figs. 2 and 3 vs. Figs. 1 and 2). Thus, the SNAG domain mutation in Insm1 results in the deficits observed in the endocrine cell lineage differentiation.

We next investigated the expression of islet-specific transcription factors in Insm1delSNAG mutants. To this end, we introduced both the wild-type and heterozygous as controls and costained Insm1 with the transcription factors MafA, MafB, Pdx1, Pax6, Ngn3, and Isl1/2 (Figs. 3 and 4). We detected dramatically decreased MafA and MafB expression and mild decreased Pdx1 expression in islets of Insm1delSNAG mutants compared with wild-type or heterozygous controls at both E18.5 (Fig. 3A–C) and E15.5 (Fig. 4A–C). The expression of Pax6 was decreased at E18.5 but had little change at E15.5 (Figs. 3D and 4D). The expression of Isl1/2 and the endocrine cell progenitor marker Ngn3 had no alteration at both stages (Figs. 3E and F and 4E and F). Therefore, Insm1delSNAG mutation results in decreased expression of the α- and β-lineage–specific transcription factors but not the progenitor or pan-islet factors.

To investigate the specific gene expression in the pancreas of Insm1delSNAG mutants, we performed RNA-seq analysis using the E18.5 pancreas of Insm1delSNAG, Insm1 null, and control mice. We observed that the expression of 231 genes was significantly dysregulated in Insm1delSNAG mutant pancreas as compared with the control using a cutoff of 1.5-fold and false discovery rate of <0.05 (Supplementary Table 2). Among these genes, those participating in development and hormone secretion were identified (Supplementary Table 3). We observed a significant overlap in the altered gene expression of Insm1delSNAG and null mutations, particularly in the genes that code for hormones and those involved in glucose metabolism and secretion regulation (Fig. 5A and B and Supplementary Tables 3 and 4). Approximately 72.7% of the differentially expressed genes in the Insm1delSNAG mutants were also identified dysregulated in the Insm1 null mutants. However, the Insm1 null mutation resulted in more pronounced changes in both gene expression and numbers compared with Insm1delSNAG mutation (Supplementary Table 2 vs. Supplementary Table 5). These results are consistent with the less severe defects observed in β-cells in Insm1delSNAG mutants compared with those in Insm1 null mutants.

Among the cell lineage–specific genes, the δ-cell–specific genes somatostatin (Sst), Hhex, and retinol binding protein 4 (Rbp4) were upregulated in the pancreas of Insm1delSNAG mutant mice compared with control or Insm1 null mutant mice (Fig. 5C). This is consistent with the increase in the number of δ-cells observed in immunostaining analysis (Fig. 1C). Most of the dysregulated genes can be verified by real-time RT-PCR using the RNA isolated from an independent set of animals (Fig. 5D).

Because the increased δ-cell numbers occurred after E15.5 in the Insm1delSNAG mutants (Figs. 1C and 2C), we investigated whether it is the δ-cell proliferation that contributed to this increase. We performed Ki67 and somatostatin costaining in Insm1delSNAG mutant mice at E15.5 and E18.5. We observed comparable proliferation rates in δ-cells, as indicated by the ratio of Ki67⁺Sst⁺/Sst⁺ in control and Insm1 null or Insm1delSNAG mutants (Supplementary Fig. 4A and B).

We also examined the possibility that a lower apoptotic rate in δ-cells could account for their increased numbers in Insm1delSNAG mutants. However, the apoptotic rate of somatostatin-positive cells was very low, and no difference was detected among Insm1delSNAG, Insm1 null mutants, and the control mice (Supplementary Fig. 5).

The costaining of somatostatin with insulin or glucagon was seldom observed, and no difference was detected between Insm1delSNAG mutants and controls (Supplementary Fig. 6A–D). However, increased costaining of Sst- and β-cell–specific transcriptional factor Nkx6.1 was detected in Insm1delSNAG mutants at both E15.5 and E18.5 (Fig. 6A and B and Supplementary Fig. 6E). The coexpression of Sst and Nkx6.1 indicates the possibility that the increased δ-cells may convert from the β-cells in Insm1delSNAG.

We therefore used InsCre;Insm1^delSNAG/flox (CKO) mice to investigate whether the δ-cell number was increased in animals harboring a β-cell–specific SNAG mutation. We observed a significant increase in δ-cell numbers in pancreas of the neonatal CKO mice (Fig. 7A). Therefore, β-cell–specific Insm1delSNAG mutation results in increased δ-cell numbers. We further used a β-cell–specific tracing mouse model mTmG;InsCre;Insm1^delSNAG/flox (tracing-CKO) to investigate the transdifferentiation of β- to δ-cells. This allowed us to trace the Insm1delSNAG mutant β-cells using GFP. We observed significant costaining of GFP with Sst in the pancreas of β-cell–specific Insm1delSNAG mutants but significantly less costaining in the control mTmG;InsCre;Insm1^+/flox mice (Fig. 7B). Therefore, β-cell to δ-cell transdifferentiation occurs in Insm1delSNAG mutant pancreatic β-cells.

To further elucidate the molecular function of the SNAG domain of Insm1, we investigated the proteins that interact with Insm1delSNAG via coimmunoprecipitation. We detected that Lsd1 could not be pulled down by the Flag antibody in SJb β-cells transfected with Insm1delSNAG-Flag (Fig. 8A). This result is consistent with a previous observation in pituitary cells (31). Insm1 can interact with NeuroD1 and FoxA2 in pancreatic β-cells (25). Therefore, we performed immunoprecipitation using NeuroD1 and FoxA2 antibodies in the same cell lysate containing the Insm1delSNAG-Flag or Insm1-Flag protein. We observed that mutation in the SNAG domain of Insm1 did not prevent the interaction between Insm1 and NeuroD1 or FoxA2 (Fig. 8A). Therefore, Insm1delSNAG mutation disrupts the interaction between Insm1 and Lsd1 but not between Insm1 and NeuroD1 or FoxA2.

Analysis of our ChIP-sequencing (ChIP-seq) data previously performed in a pancreatic β-cell line (25) revealed that Insm1 can bind to the Hhex locus on an intron and 18-kb downstream of the gene (Fig. 8B). An H3K4me1-enriched region surrounds these binding sites (Fig. 8B, top panel). These features are not observed on the locus of Amy1, a pancreatic exocrine cell–specific gene (Fig. 8B, bottom panel). To investigate the site-specific DNA binding of Lsd1 recruited by Insm1, we performed ChIP-PCR using antibodies against Insm1 and Lsd1. We observed enrichment of the Lsd1 binding at the Insm1 biding sites, the intron, and the 18-kb downstream region of Hhex1, as compared with Amy1 locus, while decreased Lsd1 binding at these two sites were observed in Insm1 mutant SJb cells (Fig. 8C and D). Lsd1 specifically exhibits histone demethylase activity in histones H3K4me2 and H3K4me1 (38). H3K4me1 indicates the enhancer sequences while H3K4me2 is enriched around the promoter and gene body regions (39,40). To investigate the histone modification of H3K4me1 and H3K4me2 mediated by Insm1SNAG-Lsd1, we performed ChIP-PCR using antibodies against H3K4me1 and H3K4me2 in control and Insm1delSNAG mutant islets. We detected two- to threefold enrichment of H3K4me2 at the intron and 18-kb downstream sites of Hhex gene and a twofold enrichment of H3K4me1 at the 18-kb downstream site of Hhex in Insm1delSNAG mutant islets than in wild-type islets (Fig. 8E and F). Therefore, the Insm1delSNAG mutation disrupts the recruitment of Lsd1 to Insm1-binding loci and results in the increased H3K4me1/H3K4me2 levels, particularly on the δ-cell–specific gene Hhex.

To investigate whether the pan-Lsd1 inhibition resembles the Insm1delSNAG-mediated Lsd1 ablation, we treated isolated neonatal pancreatic islets with Lsd1 inhibitor and analyzed the gene expression. We observed that the expression of hormone gene Ins1 was downregulated and Ppy was upregulated. However, the expression of Ins2, Gcg, and Sst had no changes. The β-cell–specific transcription factors Pdx1 and Pax6 were downregulated, as observed in Insm1delSNAG mutation, whereas the expression of Mafa and Nkx6.1 was upregulated. The expression of δ-cell–specific factor Rbp4 was increased but Hhex only increased slightly and did not reach a statistical significance (Supplementary Fig. 7). These data indicated that pan-Lsd1 inhibition results in only partial similarity of the gene expression changes as those observed in Insm1delSNAG mutants (Fig. 5D and Supplementary Fig. 7).

In the current study, we showed that mutation in the SNAG domain of Insm1 results in similar developmental deficits in α, PP, and ε cells but less severe deficits in β-cells compared with that in Insm1 null mutation. The SNAG mutation in Insm1 impairs the interaction between Insm1 and Lsd1. However, the interaction of Insm1-NeuroD1 or Insm1-FoxA2 is preserved. NeuroD1 and FoxA2 play essential roles in the development of pancreatic β-cells; the maintenance of the interaction between Insm1delSNAG and these factors partially explains the less severe developmental deficits in β-cells of Insm1delSNAG mutant mice compared with that in the Insm1 null mutants.

We showed that mutation in the SNAG domain of Insm1 results in β- to δ-cell transdifferentiation. We identified that the SNAG domain of Insm1 interacts with and recruits Lsd1 onto DNA. Mutation of the SNAG domain interrupts the recruitment of Lsd1 and results in increased H3K4me1 and H3K4me2 levels at the locus of the δ-cell–specific gene Hhex. Therefore, we conclude that during pancreatic β-cell development Insm1 represses δ-cell fate partially through the recruitment of Lsd1 via its SNAG domain. However, considering the various histone modification complexes that can be further recruited by Lsd1 in a temporal- and cellular-dependent manner, the molecular mechanism we identified may reveal only partial regulatory scenarios.

Vinckier et al. (41) identified a regulatory role of Lsd1 in pancreatic endocrine cell development via repressor activity in a human embryonic stem cell differentiation system in vitro. During the differentiation of embryonic stem cells to pancreatic endocrine cells, they observed that Lsd1 plays a strong role in the early differentiation stages and appears to have little effect at later stages (41). However, the decreased expression, though not significant, of the β-cell–specific genes Ins and Pdx1 was observed upon Lsd1 inhibition in the later stage of development (41). In the current study, we observed that inhibition of Lsd1 by its inhibitor in isolated neonatal pancreatic islets results in significantly decreased expression of some of the β-cell–specific genes, including Ins1 and Pdx1, but increased expression of δ-cell–specific gene Rbp4. However, not all of the examined genes had the identical changes as those observed in pancreas of Insm1delSNAG mice. This indicated that inhibition of Lsd1 influences global Lsd1 target genes, which may directly or indirectly affect Insm1 target genes, and thus shares only partial similarity in the regulation of gene expression as that in the Insm1delSNAG-mediated Lsd1 ablation. Our data extend the knowledge of Lsd1 function, which is mediated by the SNAG domain of Insm1.

Welcker et al. (31) showed that Insm1delSNAG mutation results in very similar developmental deficits during pituitary endocrine lineage differentiation as that observed in Insm1 null mutation. In the current study, we identified a similar developmental phenotype in α, polypeptide, and ε cells but less severe deficits in β-cells in Insm1delSNAG as compared with Insm1 null mutants. We also observed the opposite phenotypes in δ-cells in Insm1delSNAG mice compared with Insm1 null mutant mice. Therefore, the Insm1delSNAG mutation shows different results with respect to endocrine cell regulation compared with the Insm1 null mutation in different tissues and cell types.

In summary, we showed that the SNAG domain of Insm1 is essential for the differentiation of all pancreatic endocrine cells, particularly for the repression of β- to δ-cell transdifferentiation.

Acknowledgments. The authors thank Yanrui Deng (Jinan University) for expert animal husbandry support. The authors thank Carmen Birchmeier (Developmental Biology/Signal Transduction, Max Delbrück Center for Molecular Medicine) for the sharing of the Insm1 mutant mouse strains and mTmG mice. The authors also thank Dr. Pedro Luis Herrera (Department Genetic Medicine and Development, University of Geneva) for the agreement for using the InsCre mice and Dr. Francesca M. Spagnoli (Max Delbrück Center for Molecular Medicine) for sharing of the mice.

Funding. This work was supported by the National Natural Science Foundation of China (grant 81900702 to W.T. and 81770771 to S.J.), the Natural Science Foundation of Guangdong Province (grant 2017A030313527 to S.J.), and the Guangzhou Science and Technology Program (grant 201704020209 to S.J.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. X.L. and H.D. performed the experiments and analyzed the data. Y.M. performed the animal mating, immunostaining, cell counting, and most of the experiments during the revision. U.K. contributed to the preliminary data production and analysis. Y.W., F.D., and H.L. contributed to the animal mating, immunostaining, and cell counting. J.Z. performed the immunoprecipitation experiments. C.W., L.R.H.-M., and W.T. contributed to the study and experiments design. S.J. supervised the project and wrote the manuscript. S.J. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.