ARID1A loss in pancreas leads to islet developmental defect and metabolic disturbance

Summary

ARID1A is a tumor suppressor gene mutated in 7–10% of pancreatic cancer patients. However, its function in pancreas development and endocrine regulation is unclear. We generated mice that lack Arid1a expression in the pancreas. Our results showed that deletion of the Arid1a gene in mice caused a reduction in islet numbers and insulin production, both of which are associated with diabetes mellitus (DM) phenotype. RNA sequencing of isolated islets confirmed DM gene signature and decrease of developmental lineage genes. We identified neurogenin3, a transcription factor that controls endocrine fate specification, is a direct target of Aird1a. Gene set enrichment analysis indicated the enhancement of histone deacetylase (HDAC) pathway after Arid1a depletion and a clinically approved HDAC inhibitor showed therapeutic benefit by suppressing disease onset. Our data suggest that Arid1a is required for the development of pancreatic islets by regulating Ngn3⁺-mediated transcriptional program and is important in maintaining endocrine function.

Graphical abstract

Highlights

• •Depletion of Arid1a in the pancreas leads to islet defect and diabetes phenotype
• •Neurogenin 3 is a downstream mediator of Arid1a to control islet development
• •Expression of histone deacetylases is increased after Arid1a loss
• •Arid1a depletion alters genome-wide occupancy of histone deacetylase 6

Endocrinology; Endocrine regulation; Diabetology; Omics; Transcriptomics

Introduction

Gene transcription in mammalian cells is tightly orchestrated by the recruitment of transcription factors, co-regulators, mediators, transcription initiation factors and RNA polymerase II to the promoter regions.^1,2 ATP-dependent chromatin remodeling complexes are co-regulators that alter chromatin configuration to promote the binding of transcription factors to DNA. This process initiates gene transcription as well as DNA replication and repair.^3,4

There are four major families of ATP-dependent chromatin remodeling complexes in mammals, including the BAF complex. Mutations in the subunits of the BAF complex are found in >20% of human cancers.^5,6 The BAF complex evolved from the yeast SWItch and sucrose nonfermenting (SWI/SNF) complex. Of the BAF complex components, the AT-rich interactive domain-containing protein 1A (ARID1A) is the most commonly mutated subunit.⁷ In addition, loss of ARID1A expression is associated with poor clinical outcome in various cancers.^8,9,10 To clarify the role of Arid1a in pancreatic tumorigenesis, recent studies generated Arid1a-deficient mice and combined a k-ras mutation with a p53 deletion.^11,12,13 These studies concluded that Arid1a maintains acinar cell homeostasis and ductal cell differentiation, and that loss of Arid1a induces intraductal papillary mucinous neoplasms (IPMNs) in mice. The combination of k-ras and p53 dysregulation accelerates the formation of pancreatic cancer from IPMNs. These data suggested that Arid1a functions as a tumor suppressor gene in pancreatic cancer. All of these studies used Ptf1a-driven Cre recombinase to deplete Arid1a gene in various cell types in the pancreas. However, Cre activity is extremely low or undetectable in the islet cells of Ptf1a-Cre mice.^14,15 In line with the finding, all three studies showed that the expression of Arid1a is retained in the islets of the mice.^11,12,13 Thus, whether Arid1a plays a role in islet development and endocrine regulation is still unknown.

Because conventional knockout of the Arid1a gene in mice caused embryonic lethality,¹⁶ the involvement of Arid1a in organ development and functional regulation is also largely unclear at present. A recent study demonstrated that liver-specific deletion of Arid1a induced non-alcoholic steatohepatitis by modulating the genes that control hepatic lipogenesis and fatty acid oxidation via epigenetic regulation.¹⁷ By using the pancreatic and duodenal homeobox 1 (Pdx1)-Cre mouse model, we demonstrate that Arid1a is required for the development of islets in mice and is a key player in controlling the functionality of pancreatic β cells.

Results

Deletion of Arid1a causes DM phenotype in mice

Pdx1 is a master regulator in the developing pancreas and is important for the maturation of β cells. We utilized Pdx1-Cre mice (PC mice) to perform a genetic cross with Arid1a^flox/flox (A) mice to generate PCA^L/L mice that lack Arid1a expression in the pancreas. Arid1a depletion was achieved by deleting the eighth exon of the Arid1a allele in the presence of Cre recombinase activity.¹⁶ The PCA^L/L mice were both viable and fertile; however, their body weight was decreased (Figure 1A). Pancreas size in the PCA^L/L mice was significantly reduced (Figure 1B). We found that blood sugar levels were much higher in the PCA^L/L mice, and the magnitude of this difference increased with age (Figure 1C). Immunohistochemical (IHC) staining confirmed that Arid1a was absent in islet cells and was also reduced in acinar and ductal cells (Figure 1D). The morphology of acinar cells and the expression of amylase were not significantly changed (Figures S1A and S1B). Examination of liver, duodenum, prostate and pituitary gland did not find significant pathological alterations in these organs (Figure S2). However, the number and size of insulin-positive islets were significantly decreased in the pancreas of the PCA^L/L mice (Figure 1E). The number of glucagon-expressing cells was also decreased (Figures 1D and 1E). In line with this finding, insulin levels in the blood were significantly reduced in the PCA^L/L mice (Figure 1F). Glucose tolerance tests showed that blood sugar levels were much higher in the PCA^L/L mice and did not decrease at 2 h following glucose challenge (Figure 1G). Blood chemistry analyses demonstrated significant alterations in liver and kidney functions and lipid metabolism (Figure S3). Total cholesterol and triacylglycerol were dramatically increased whereas total bilirubin and albumin were unchanged. Diabetic retinopathy, characterized by the thinning of inner plexiform and inner nuclear layers (Figures S4A and S4B), and thickening of the basement membrane of the vascular capillary (Figure S4C), was observed in the retinas of the aged (>20 weeks) PCA^L/L mice. These results suggested that deletion of Arid1a impairs pancreas development and causes DM phenotype in mice. The survival of the PCA^L/L mice was poor, with a median survival time of 28.8 weeks (Figure 1H). The clinical relevance of Arid1a expression in human DM was studied using the Human Pancreas Analysis Program Database (https://hpap.pmacs.upenn.edu), which was recently published.¹⁸ Single-cell transcriptome analysis showed a 50% reduction of ARID1A expression in the islet cells of patients with DM, when compared with that of normal individuals (Figure S5A). We also extracted the microarray data from the dataset (GSE106148) reported previously.¹⁹ Downregulation of ARID1A, detected by four independent probes, was found in the pancreases of DM patients (Figure S5B). These data support that ARID1A depletion may lead to pancreas dysfunction.

Figure 1.

Pancreas-specific deletion of Arid1a induces DM in mice

(A) Body weight was decreased in the Arid1a-depleted PCA^L/L mice (n = 5).

(B) The pancreas size of the PCA^L/L mice was reduced (n = 5).

(C) Age-dependent increase of blood sugar in the Arid1a-depleted PCA^L/L mice (n = 5).

(D) Representative pictures of Arid1a, insulin and glucagon staining in the pancreases of normal and Arid1a-depleted PCA^L/L mice.

(E) The Arid1a-, insulin- and glucagon-positive areas were expressed as the percentage of total areas measured. Five independent areas were examined in a tissue slide and five tissues were counted in each group.

(F) Blood samples were collected from 13-week-old mice and insulin level was measured by ELISA assays. (n = 5).

(G) Thirteen-week-old mice (n = 5) were fasted for 6 h in the morning and intraperitoneally injected with 2 g/kg of glucose. Blood glucose levels were continuously measured for 120 min using a glucometer.

(H) The survival of control (n = 20) and PCA^L/L mice (n = 41). ∗p<0.05; ∗∗p<0.01; ∗∗∗p<0.001.

Arid1a depletion decreases insulin expression and increases MHC expression in mouse islet cells

Islets from the control and PCA^L/L mice were isolated and compared. The morphology of the islets in culture was strikingly different, and the islet size of the PCA^L/L mice was much smaller (Figure 2A). The insulin levels in the conditioned medium of cultured PCA^L/L islets were significantly reduced (Figure 2B). RNA sequencing identified 4,048 differentially expressed genes (DEGs) with 3,040 upregulated after Arid1a depletion (Figure 2C). KEGG pathway analysis revealed several significantly altered pathways including pancreatic secretion, insulin secretion and DM phenotype (Figure 2D). Upregulations of MHC-I and MHC-II related genes were found in the DM gene signature in the Arid1a-depleted islets and were further confirmed by RT-PCR analysis (Figure S6). Gene set enrichment analysis (GSEA) also showed the downregulation of hallmark of pancreatic β cells and reduced expression of insulin, glucokinase (GCK) and glucose-6-phosphatase catalytic subunit 2 (G6PC2), three β cell-enriched genes (Figure 2E and Table S1). DM has been found to be associated with the upregulation of inflammatory cytokines.²⁰ Consistently, GSEA also demonstrated the activation of interferon α signaling (Figure 2F), which has been strongly implicated in DM.²¹ In addition, we also found the upregulation of interleukin-1 β (IL-1β), another important inflammatory cytokine, in the Arid1a-depleted pancreas (Figure S7). Previous studies demonstrated that MHC-I and MHC-II are upregulated in the islets of DM patients.^22,23 Enhancements of several biological processes in KEGG analysis including leukocyte migration, regulation of defense response and T cell activation were all associated with enhanced immune reaction (Figure S8). Increase of a number of cytokines and chemokines in the isolated Arid1a-depleted islets was confirmed by RT-PCR analysis (Figure 2G). When examining the pancreatic tissues of the PCA^L/L mice, a dramatic increase of T cell infiltration was observed (Figure 2H). Thus, pancreas-specific depletion of Arid1a in mice induces β cell dysfunction and triggers inflammation and immune attack in the pancreas that recapitulates the pathological alterations found in human DM.

Figure 2.

Arid1a depletion impairs islet development

(A) Islet isolation was performed as described in Materials and Methods. The cultured islets were imaged by light microscope and the islet size was calculated. The average diameters of the islets were expressed as Mean ± SD (n = 20).

(B) The conditioned media of control and isolated PCA^L/L islets were collected at 72 h after culture and the levels of insulin were measured by ELISA assays.

(C) Gene expression of isolated control and PCA^L/L islets was investigated by RNA sequencing and the expression profile was presented by heatmap.

(D) The pathways enriched in the PCA^L/L islets studied by KEGG pathway analysis.

(E) GSEA analysis suggested β cell dysfunction in the isolated PCA^L/L islets.

(F) Activation of interferon α signaling was found in the PCA^L/L islets.

(G) Total RNAs were extracted from the isolated islets and the expression of cytokines and chemokines was studied by RT-PCR analysis.

(H) Pancreatic tissues collected from the control and PCA^L/L mice. Infiltration of T cells in the pancreas was studied by IHC staining. (n = 5). ∗∗∗p<0.001.

Altered metabolism in Arid1a-depleted islet cells

In lung, Arid1a-deficiency leads to enhanced glycolysis pathway by enhancing transcription of PGAM1, PKM2, and PGK1.²⁴ In liver, Arid1a depletion impairs fatty acid oxidation by downregulating PPARα.²⁵ These studies suggested that a lack of Arid1a changes cell metabolism. To explore the cellular metabolic changes in islet cells underArid1a-knockout condition, we isolated the islet cells from normal and PCA^L/L mice and performed a comprehensive metabolomic analysis by capillary electrophoresis time-of-flight mass spectrometry. Principal component analysis revealed significant differences in metabolic pathways between the control and PCA^L/L islet cells (Figure 3A). Differentially expressed metabolites were displayed by heatmap (Figure 3B). Top ten upregulated or downregulated metabolites were shown (Figures 3C and 3D). The metabolite with the greatest increase was 2-oxoisovaleric acid, an intermediate highly upregulated in streptozotocin-induced DM.²⁶ In addition, elevated levels of 2-oxoglutaraic acid, pyruvic acid, glyceraldehyde-3-phosphate, lactic acid, and 3-hydroxybutyrate (a ketone body) were found in the PCA^L/L islets, consistent with a diabetic phenotype (Figure 3E). Several amino acids including asparagine, aspartate, glutamate and alanine were also enhanced (Figure 3F). Of interest, N¹-methylnicotiamide, a metabolite that has been reported to be upregulated in the serum of DM patients,²⁷ was significantly elevated (Figure S9A). Other altered pathways included those involved in the biosynthesis of polyamine, purine and pyrimidine (Figure S9B). These results suggest that Arid1a depletion alters metabolism in the islet cells.

Figure 3.

Metabolic alteration in the islets of the PCA^L/L mice

(A) The metabolite profiles of the islets isolated from the control and PCA^L/L mice were compared by principal component analysis.

(B) Differentially expressed metabolites were demonstrated by heatmap.

(C) Top 10 upregulated metabolites in the isolated PCA^L/L islets were shown as fold change (Arid1a/control).

(D) Top 10 downregulated metabolites in the isolated PCA^L/L islets were shown as fold change (Arid1a/control).

(E) Alteration of carbohydrate metabolism and increase of glycolytic intermediates in the isolated PCA^L/L islets. The metabolite levels in the control islets (red color) and PCA^L/L islets (blue color) were presented.

(F) The levels of various amino acids in the control islets (red color) and PCA^L/L islets (blue color) were shown and the increase of alanine, aspartic acid, asparagine, glutamate, proline was found. ∗p<0.05; ∗∗p<0.01.

Developmental defect following Arid1a depletion

Significant reductions in body weight, pancreas size and islet number were found early in some mice at 5–6 weeks after birth. In addition, a number of transcription factors enriched in β cells were decreased in the Arid1a-depleted islets (Table S1). We tested whether Arid1a affects pancreas development. RT-PCR assays confirmed the data of RNA sequencing and demonstrated the downregulation of developmental lineage markers including Pdx1 (a key regulator of pancreatic-specified endoderm development), Hnf1B (a crucial transcription factor in multipotent progenitors) and neurogenin 3 (Ngn3, a master controller of endocrine fate specification) in the Arid1a-depleted islets (Figure 4A). Because Ngn3 is a key transcription factor involved in the generation of endocrine progenitor cells, we studied whether Ngn3 is a direct target gene of Arid1a. Bioinformatics analyses predicted four potential Arid1a binding sites in the Ngn3 promoter (Figure 4B). Chromatin immunoprecipitation (ChIP) assays revealed the binding of Arid1a to two upstream regions of the Ngn3 gene in the control islet cells but not in the Arid1a-depleted islet cells and the region 3 (Ngn3-3) was the major binding site of Arid1a in the Ngn3 promoter (Figure 4B). We next performed rescue assay by expressing Arid1a and Ngn3 in the isolated mouse islet cells (Figure 4C). Our results showed that expression of Ngn3 alone could not significantly enhance insulin expression because of the lack of the SWI/SNF complex to appropriately remodeling the chromatin in the Arid1a-depleted islet cells. In addition, expression of Arid1a alone also could not fully activate insulin expression because the transfection efficiency of primarily isolated islet cells was very low (around 10%) and could not effectively trigger Ngn3 expression in these cells. However, co-expression of Ngn3 and Arid1a indeed upregulated insulin levels by three-fold in Arid1a-depleted islet cells (Figure 4C). Of interest, we found that mouse NIT-1 insulinoma cells expressed high levels of Arid1a and knockdown of Arid1a by shRNA decreased Ngn3 and insulin expression (Figure 4D). We next used human 1.2B4 cell line, an insulin-secreting hybrid cell line generated by electrofusion of a primary culture of human pancreatic islets with HuP-T3 human pancreatic carcinoma cells, to verify our finding. Knockdown of ARID1A also reduced the expression of Ngn3 and insulin in 1.2B4 cells, accompanying with decreased insulin level in the cultured medium (Figure 4E). These results suggest that Arid1a depletion leads to developmental defect of β cells, and Ngn3 is one of the downstream targets for Arid1a to control β cell differentiation and maturation.

Figure 4.

Ngn3 is a direct target of Arid1a

(A) The islets were collected from the control and PCA^L/L mice and the expression of endocrine developmental lineage genes was studied by RT-PCR. ∗p<0.05; ∗∗p<0.01.

(B) Genomic DNAs were collected from the control and PCA^L/L islets and ChIP assay was performed to study the binding of Arid1a to the predicted regions (marked from region 1 to 4 upstream of the transcriptional start site) in the Ngn3 gene promoter.

(C) Arid1a-depleted islet cells were infected with control (left) or Ngn3+Arid1a viral expression vectors. After 48 h, the conditioned media were collected and the insulin levels were determined by ELISA assay. (n = 5). ∗p<0.05.

(D) Mouse NIT-1 insulinoma cells were transfected with Arid1a shRNA and the protein levels of Arid1a, insulin and Ngn3 were analyzed by Western blotting. The expression levels of E-cadherin (E-Cad) and MafA were also examined.

(E) Human 1.2B4 insulin-secreting hybrid cells were transfected with Arid1a shRNA and the protein levels of ARID1A, Insulin and NGN3 were analyzed by Western blotting. The expression levels of HDAC6 and MafA were also examined. The insulin level in the conditioned media of cells transfected with control or shArid1a was determined by ELISA assay. (n = 3). ∗p<0.05.

Upregulation of histone deacetylases (HDACs) in Arid1a-depleted pancreas

An enriched gene signature in our RNA sequencing data is the HDAC pathway (Figure 5A). We found the expressions of HDAC4, 5 and 6 were upregulated whereas the expression of HDAC1 was reduced in Arid1a-depleted pancreas tissues (Figure 5B). In agreement with this data, HDAC6 expression was increased after Arid1a depletion in human 1.2B4 islet cells (Figure 4E). Because HDAC6 has been shown to be a repression target of Arid1a,²⁸ we performed chromatin immunoprecipitation-sequencing analysis (ChIP-seq) to study genome-wide HDAC6 occupancy in the control and Arid1a-depleted islet cells. Ablation of Arid1a increased the binding of HDAC6 to proximal gene promoters (<3 kb) whereas it decreased HDAC6 occupancy on 5′UTR and gene bodies (Figure 5C). Volcano plot showed 136 significantly altered genes in Arid1a-depleted islet cells (Figure 5D). The preferential binding motifs were also predicted (Figure 5E). Gene ontology analysis demonstrated the enrichment of genes in diverse biological processes (Figure 5F). Pathway analysis showed differentially expressed genes in insulin signaling, glycolysis/gluconeogenesis, and pyruvate metabolism, consisting with the diabetic phenotypes found in the PCA^L/L mice (Figures 5G and 5H). Among the top 30 altered genes, two critical genes, Ngn3 and Pdx1, involving in pancreas development and endocrine regulation were identified as HDAC6 targets (Figure 5I). ChIP assay confirmed that the binding of HDAC6 to the Arid1a-binding Ngn3-3 region in the Ngn3 promoter was enhanced and the acetylation of histone H3 (activation marks) in this region was attenuated (Figure 5J). These results suggested that Arid1a ablation increases HDAC6 expression and changes its genome-wide occupancy to impair islet development.

Figure 5.

Upregulation of HDACs expression in Arid1a depletion islet

(A) GSEA analysis revealed the upregulation of HDAC pathway in the PCA^L/L islets.

(B) Islets were isolated from the pancreas of the control and PCA^L/L mice and the protein levels of HDACs were determined by Western blotting.

(C) ChIP-sequencing was performed to identify genome-wide binding of HDAC6 in control and PCAL/L mice. Distribution figure of peaks on various functional elements of genes.

(D) Differential volcano plot (Red dots represent peaks that are significantly up-regulated and blue dots represent those that are significantly down-regulated. X axis:log2 fold change of peaks. Y axis: statistical significance of the differential in log10(q-value).

(E) The most significant motif sequence. The abscissa is motif sequence. The ordinate is percentage of each base.

(F) GO enrichment histogram (X axis: number of differential peak-associated genes in this GO category. Color code is to distinguish the categories - biological processes, cellular components and molecular functions.

(G) KEGG enrichment histogram (X axis: gene number. Y axis: pathway term).

(H) Scatterplot of peak-associated genes KEGG enrichment (X axis: RichFactor. Y axis specify KEGG pathways. The size of the dot is positively correlated with the number of peak-associated genes in the pathway. Color code is to indicate Q-value ranges.

(I) Depicted are the top30 gene targets with minimum p-values identified by ChIP-Seq peak calling analysis.

(J) Genomic DNAs were collected from the control and PCAL/L islets and the binding of HDAC6 to the Arid1a-occupied region 3 (Ngn3-3) in the Ngn3 gene promoter was studied by ChIP assay. Acetylation of histone H3 around the Ngn3-3 region in the Ngn3 gene promoter was also studied by ChIP assay.

Therapeutic effect of HDAC inhibitor on Arid1a depletion-induced pancreas dysfunction

Vorinostat (also known as suberanilohydroxamic acid, SAHA), a clinically approved HDAC inhibitor, was given to 7-week-old PCA^L/L mice, and blood sugar levels were continuously monitored. Treatment with SAHA did not significantly change body weight of the mice (Figure S10). However, SAHA treatment increased pancreas size, and decreased blood sugar level in the PCA^L/L mice (Figures 6A and 6B). In addition, insulin levels in the blood and the number and size of insulin-producing islets were increased (Figures 6C and 6D). Moreover, the number of Ngn3⁺ progenitor cells was upregulated (Figures 6D and 6E). Conversely, infiltration of CD3⁺T cells and F4/80⁺ macrophages in the pancreas was significantly decreased (Figures 6F and 6G). These data suggest that HDAC inhibitors show therapeutic benefit on Arid1a depletion-induced pancreas dysfunction by reducing immune attack of islet cells and amplifying the pool of Ngn3⁺ progenitor cells.

Figure 6.

HDAC inhibitor suppresses the onset of DM induced by Arid1a depletion

(A) Seven-week-old PCA^L/L mice were administered twice weekly by intraperitoneal injections of 50 mg/kg of SAHA or PBS vehicle for 4 weeks (n = 5, per group). Mice were sacrificed, and the weights of pancreases were measured. In addition, blood sugar levels were measured at different time points before animal sacrifice.

(B) Blood glucose levels in the PCA^L/L mice treated with or without SAHA were measured every 14 days (n = 5).

(C) Blood insulin levels in the PCA^L/L mice treated with or without SAHA were measured by ELISA assays. (n = 5).

(D) The pancreases of vehicle- or SAHA-treated PCA^L/L mice were harvested for IHC staining to detect the expression of insulin, glucagon and neurogenin3 (Ngn3). Five independent areas were examined in a tissue slide and five tissues were counted in each group.

(E) Areas of positive insulin, glucagon and neurogenin3 immunostaining were quantified by ImageJ software.

(F) Infiltration of CD3⁺T cells and F4/80⁺ macrophages in the pancreas of vehicle- or SAHA-treated PCA^L/L mice was studied by IHC staining.

(G) Areas of positive CD3⁺ and F4/80⁺ immunostaining were quantified by ImageJ software. ∗p<0.05; ∗∗p<0.01.

Discussion

Although the tumor-suppressive activity of Arid1a has been demonstrated in various cancers including pancreatic cancer, the functional role of Arid1a in organ development is largely unclear because conventional knockout of this gene causes embryonic lethality.¹⁶ Recently, three studies generated pancreas-specific deletion of Arid1a in mice.^11,12,13 However, all these studies utilized Ptf1a-Cre for gene deletion. Because the expression of Ptf1a is very low in islet cells especially in β cells, Arid1a is retained in the islets of the mice, and no alteration of endocrine function has been noted. By using Pdx1-driven Cre model, we provide the first evidence that Arid1a is required for islet development.

Pdx1 is a critical transcription factor that controls the development, maturation and regeneration of β cells. Deletion of the Pdx1 gene in β cells induces reprogramming of gene transcription, resulting in loss of β cell identify and enhancement of α cell phenotype.²⁹ Of interest, inactivation of Pdx1 causes the development of DM.^29,30 However, the effects of ablation of Arid1a and Pdx1 are different because pancreas-specific deletion of Arid1a affects multiple islet cell types and reduces the expression of insulin and glucagon simultaneously, whereas ablation of Pdx1 triggers α cell transcriptional program and glucagon production. Although Pdx1 may partially contribute to the developmental defect seen in Arid1a-depleted mice, it does not confer all biological effects of Arid1a in pancreas development. Our study suggests that Arid1a plays a crucial role in the regulation of pancreatic progenitor cell proliferation. One potential mediator by which Arid1a regulates the fate and maturation of islet cells is Ngn3. This transcription factor is required for endocrine lineage formation of the pancreas by controlling various downstream transcriptional factors like NeuroD1, Nkx2-2, Pax4 and Nkx6-1.³¹ Our RNA sequencing data confirmed that all these transcription factors were decreased in Arid1a-depleted islets (Table S1). In human, development of endocrine pancreas is also dependent on Ngn3.³² We showed that Arid1a directly binds to the Ngn3 promoter to increase its expression and co-expression of Ngn3 and Arid1a increases insulin production in Arid1a-depleted islet cells. Our results suggest that Ngn3 is one of the downstream targets of Arid1a to controls pancreas development. However, other transcriptional factors like Pdx1, Hnf1b and Rfx6 (Figure 4A) may also contribute to Arid1a-mediated effect.

A complex interaction between β cells and immune cells has been implicated in the development of DM. External insults like viral infection and immune stimulators have been shown to induce local inflammation via a self-defense mechanism in the pancreas. However, whether some intrinsic mediators may promote inflammatory responses in the pancreas to promote DM pathogenesis are relatively unclear. A potential role of chromatin regulators including DNA methyltransferase, nucleosome remodeling complexes, histone acetyltransferases/deacetylases and polycomb group complex proteins have been implicated in pancreas development and DM pathogenesis.³³ However, a genetically engineered animal model to address the effect of chromatin regulators on the development and function of pancreatic islet cells is still lacking. A recent study demonstrated that embryonic deletion of the ATPase subunit Brg1 caused pancreatic hypoplasia whereas removal of another ATPase subunit Brm did not.³⁴ Our results showed that deletion of Arid1a activates interferon α signaling in the islets (Figure 2F). This cytokine has been shown to increase the expression of MHC genes in human β cells which results in enhanced immune attack and islet dysfunction.^21,22,23 Thus, Arid1a inactivation could be an intrinsic factor to modulate DM phenotype. Our hypothesis is supported by the following observations. First, bioinformatics analysis demonstrates that ARID1A is significantly downregulated in the pancreas or isolated islets of DM patients in independent databases (Figure S5), suggesting the phenotypic alterations found in our animal model may recapitulate at least in part of the disease progression in DM patients. Second, in addition to mutation, promoter hypermethylation of ARID1A gene has been reported in various human diseases.^35,36,37 It is possible that the decrease of ARID1A in DM patients can be induced by aberrant epigenetic regulation. Third, ARID1A reduction may contribute to β cell senescence. Recent studies demonstrated that senescence of β cells is significantly increased in the pancreas in DM patients and elimination of senescent cells improves the severity of diabetes.^38,39 Of interest, we also found the increase of cell senescence and p21 expression in Arid1a-depleted pancreas (Figures S11A and S11B).

Another important finding of this study is that pan-HDAC inhibitors show therapeutic benefit on Arid1a depletion-induced pancreas dysfunction. Our RNA sequencing data revealed the upregulation of HDAC pathway after Arid1a depletion (Figure 5A). A previous study also demonstrated that gene expressions of most of HDACs are declined during pancreas development and HDAC inhibitors may amplify endocrine progenitor pool in mice.⁴⁰ Extended from this work and our RNA sequencing data, we found that Vorinostat increases islet numbers and insulin secretion, and represses the onset of DM phenotype in Arid1a-depleted mice. In addition, Vorinostat also reduces inflammation and infiltration of immune cells in the islets which may prevent or attenuate immune attack and β cell death. Collectively, our study reveals the importance of Arid1a in pancreatic β cell development and in metabolic homeostasis.

Limitations of the study

We establish an animal model to study the role of Arid1a in pancreas development and reveal the importance of Arid1a in the regulation of pancreatic β cell function. However, validation of the expression of Arid1a in the pancreas of DM patients is difficult because only limited RNA sequencing and single-cell transcriptome results were available in public database. In addition, immune cell infiltration in the pancreas in Arid1a-depleted mice found in our study needs to be confirmed in human samples.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Arid1a	Cell Signaling Technology	Cat# 12354; RRID:AB_2637010
Glucagon	Cell Signaling Technology	Cat# 2760; RRID:AB_659831
Insulin	Cell Signaling Technology	Cat# 3014; RRID:AB_2126503
Sox9	Abcam	Cat# ab185966; RRID:AB_2728660
CD3	Abcam	Cat# ab5690; RRID:AB_305055
Neurogenin-3	Millipore	Cat# AB5684; RRID:AB_11213023
E-Cadherin	BD Biosciences	Cat# 610182; RRID:AB_397581
Arid1a	Abcam	Cat# ab182560; RRID:AB_2313773
MafA	Santa Cruz	Cat# sc-390491; RRID:AB_2313773
GAPDH	Abcam	Cat# ab9482; RRID:AB_307272
HDAC1	Cell Signaling Technology	Cat# 5356; RRID:AB_10612242
HDAC4	Cell Signaling Technology	Cat# 5392; RRID:AB_10547753
HDAC5	Cell Signaling Technology	Cat# 2082; RRID:AB_2116626
HDAC6	Cell Signaling Technology	Cat# 7612; RRID:AB_10889735
Acetyl-Histone H3	Cell Signaling Technology	Cat# 7627; RRID:AB_10839410
Neurogenin-3	Abnova	Cat# PAB1916; RRID:AB_1577683
F4/80	Cell Signaling Technology	Cat# 70076; RRID:AB_2799771
CD31	Abcam	Cat# ab28364; RRID:AB_726362
IL1-beta	Abcam	Cat# ab8319; RRID:AB_306472
p16 INK4a	Santa Cruz	Cat# sc-468; RRID:AB_632103
p21	Santa Cruz	Cat# sc-469; RRID:AB_632121
p53	Santa Cruz	Cat# sc-126; RRID:AB_628082
Chemicals, peptides, and recombinant proteins
SAHA	Cayman	Cat# 10009929
Lipofectamine™ 3000 Transfection Reagent	Thermo Fisher Scientific	Cat# L3000008
cOmplete™, Mini Protease Inhibitor Cocktail	Millipore Sigma	Cat# 4693124001
ProLong™ Glass Antifade Mountant with NucBlue™ Stain	Thermo Fisher Scientific	Cat# P36981
DMEM, high glucose	Thermo Fisher Scientific	Cat# 11965118
Critical commercial assays
Mouse Insulin ELISA KIT	Mercodia	Cat# 10-1247-01
High-Capacity cDNA Reverse Transcription Kit	Thermo Fisher Scientific	Cat# 4368813
Deposited data
Analyzed data	NCBI	GEO: GSE106148
Raw and analyzed data	This paper	GEO: GSE220638, Mendeley Data: https://data.mendeley.com/datasets/zybcfhdb58
Experimental models: Cell lines
1.2B4	Merck Sigma	Cat# 10070103-1VL RRID:CVCL_2258
Experimental models: Organisms/strains
Arid1atm1.1Zhwa/J	The Jackson Laboratory	RRID:IMSR_JAX:027717
B6.FVB-Tg(Pdx1-cre)6Tuv/Nci	Frederick National Laboratory	RRID:IMSR_NCIMR:01XL5
Oligonucleotides
Primers for qPCR, see Table S2	see Table S2	Primers for qPCR, see Table S2
Software and algorithms
GraphPad Prism v9	GraphPad	RRID: SCR_00279
Living Image version 4.5	Perkin Elme	https://www.perkinelmer.com/product/spectrum-200-living-image-v4series-1-128113
GSEA	Subramanian et al.41	https://www.gsea-msigdb.org/gsea/index.jsp
ImageJ	Goto,42	https://imagej.nih.gov/ij/index.html
Primer3	Koressaar et al.43	http://primer3.ut.ee/

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Wen-Chun Hung (hung1228@nhri.org.tw).

Materials availability

This study did not generate new unique reagents.

Experimental model and subject details

Genetically modified mice

Pdx-1Cre and Arid1a^tm1.1Zhwa mice were obtained from the Mouse Models of Human Cancers Consortium (MMHCC) and The Jackson Laboratory under material transfer agreements. Mice were genotyped as described by the MMHCC and The Jackson Laboratory (JAX) PCR protocols for strains 01XL5 and strain Arid1a^tm1.1Zhwa. Experiments were performed on off-spring male mice (4 weeks-old) derived from homozygous mating. The age of mice used in this study is from 4-weeks-old (for drug treatment) to 1-year-old (for medium survival determination). Animal studies were approved by the Institutional Animal Care and Use Committee of the National Health Research Institutes.

Method details

Isolation of mouse pancreatic islets

To isolate mouse islets, pancreases were perfused through the common bile duct with a Hanks’ balanced salt buffer (HBSS)-collagenase solution (1.4 g/L; collagenase type 4) and digested in the same solution in a 37°C water bath for 26–28 min. After shaking for 15 seconds, pancreases were washed three times with HBSS supplemented with 0.5% bovine serum albumin (BSA) and filtrated through 500 μm and 70 μm cell strainers (Corning). Islets were retained in the 70 μm cell strainer while the cell mixture passing through the 70 μm cell strainer containing the exocrine stoma were discarded. Islets were collected and pooled into a Petri dish with RPMI-1640 (GIBCO) containing 11.1 mM glucose, 100 units/ml penicillin, 100 μg/mL streptomycin, 2 mM Glutamax, 50 μg/mL gentamycin, 10 μg/mL Fungison and 10% fetal calf serum (Invitrogen). Islets were used directly for RNA isolation or overexpression of Ngn3 and Arid1a cultured in a humid environment containing 5% CO₂ for transfection assay.

Lentivirus production and shRNA for gene knockdown

The plasmids required for shRNA lentivirus production were purchased from the National RNAi Core Facility (Academia Sinica, Taiwan). The pLKO.1-shRNA vectors used for knockdown of ARID1A were TRCN0000059090 and TRCN0000059091 (Human), TRCN0000071396 and TRCN0000071397 (Mouse). The pLKO.1-emptyT control plasmid was TRCN0000208001. To generate recombinant lentivirus, 293T cells were co-transfected with package plasmid (pCMV8.91), envelop VSV-G plasmid (pMD.G), and shRNA expressing construct. The virus-containing supernatant was harvested at 48 h after transfection. For infection, NIT-1 cells were incubated with the virus-containing medium supplemented with 8 μg/mL of polybrene and the infected cells were selected by 2 μg/mL of puromycin for 72 h before harvesting for western blot analysis.

Transcriptomic profiling by RNA sequencing

RNAs were isolated from islet tissues using Trizol extraction followed by purification with the Qiagen RNAeasy kit. Quantity and purity of RNAs were assessed at 260 nm and 280 nm using a Nanodrop (ND-1000; Labtech International). Three hundred ng of each sample was amplified and labeled using the GeneChip WT Sense Target Labeling and Control Reagents (900652) for expression analysis. Hybridization was performed against the Affymetrix GeneChip MoGene 1.0 ST array. The arrays were hybridized at 45°C for 17 hat 60 rpm. The arrays were washed and stained with streptavidin-phycoerythrin (GeneChip® Hybridization, Wash, and Stain Kit, 900720). The signal intensities on the arrays were scanned using Affymetrix GeneChip® Scanner 3000. The results were analyzed using Expression Console software (Affymetrix) and Transcriptome Analysis Console software (Affymetrix) with default RMA parameters. Differential expressed genes (DEGs) were defined as the genes with a 2.0-fold change and p value <0.05. The DEG lists were uploaded from a Microsoft Excel spreadsheet onto Metacore 6.13 software (GeneGo pathways analysis) (http://www.genego.com). Pathway analyses of DEGs were also performed using GSEA (www.gsea-msigdb.org) and KEGG (www.genome.jp/kegg).

Chromatin immunoprecipitation (ChIP) assay

ChIP assays with the anti-Arid1a antibody (Abnova, MAB15809) were performed using the EZ-ChIP Kit (Merck Millipore) according to the manufacturer’s instruction. Cellular lysates were subjected to five sets of sonication on wet ice with a 60 Sonic Dismembrator (Fisher Scientifi c). Each set consisted of 8 seconds of sonication separated by 1-minute intervals on ice and sonicated to an average DNA length between 300 and 700 bp. Chromatin concentrations were determined using a BCA protein assay kit (Pierce). Chromatin samples were diluted to 300 μg per 200 μL of nuclear lysis buffer and then cleared by centrifugation at 10,000g. Precleared chromatin samples (200 μL each) were added to 800 μL (1:5 dilution) of ChIP dilution buffer (50 mM Tris-Cl (pH 7.5); 150 mM NaCl; 5 mM EDTA; 0.5% IGEPAL CA-630; 1% TX-100) and incubated with 10 μg of prebound ARID1A Protein G beads (Invitrogen) overnight at 4°C. The Protein G bead-anti-ARID1A immune conjugates were washed twice with 1 mL ChIP dilution buffer, three times with 1 mL ChIP dilution buffer supplemented with 500 mM NaCl, twice with 1 mL ChIP dilution buffer, followed by a final wash in 1 mL low-salt TE buffer (10 mM Tris-Cl (pH 8.0); 1 mM EDTA; 50 mM NaCl). The immunoprecipitated samples were eluted in 100 μL of elution buffer (50 mM NaHCO3; 1% SDS) at 62 °C for 20 min. using an Eppendorf tube mixer. The crosslinks were reversed by adding NaCl to a final concentration of 0.2 M followed by an overnight incubation at 65 C. The samples were then deproteinated with Proteinase K and purified using the ChIP DNA Clean & Concentrator kit (Zymo Research) according to the manufacturer’s Instructions. Precipitated DNA was analyzed using PCR primers for Ngn3 promoter listed in Table S2. Real-time quantitative PCR using Ssofast PCR master mix (Biorad) and a IQ5 thermocycler (Biorad).

Metabolome profiles of mice islet medium by capillary electrophoresis-time of flight (CE-TOF) analysis

The conditioned medium of islets isolated from control and Arid1a-depleted mice were used for metabolome analysis. Samples (80 μL) were mixed with 20 μL of Milli-Q water containing internal standards. Metabolome analysis was performed by Basic Scan package of Human Metabolome Technologies (HMT) using CE-TOF (HMT, Tokyo, Japan). Hierarchical cluster analysis (HCA) and principal component analysis (PCA) were performed by statistical analysis software (HMT). The profiles of peaks with putative metabolites were presented on metabolic pathway maps using VANTED (Visualization and Analysis of Networks containing Experimental Data) software.

Glucose tolerance test

Thirteen-week-old mice (n = 5) were fasted for 16 h in the morning and intraperitoneally injected with 2 g/kg of glucose. Blood glucose levels were continuously measured for 15, 30, 60, and 120 min using a glucometer (Accu-Chek; Roche Diagnostics). In addition, blood samples were collected at the indicated time points for later plasma insulin determination by using an insulin ELISA kit (Mercodia, Uppsala, Sweden) following manufacturer’s protocol.

Immunohistochemistry

Pancreatic specimens were fixed with 10% buffered formalin, dehydrated in ethanol, embedded with paraffin, and stained with hematoxylin and eosin. Standard procedures for IHC analysis were performed according to manufacturer’s protocol. Deparaffinize and rehydrate formalin-fixed paraffin-embedded tissue section. Add enough drops of Hydrogen Peroxide Block to cover the sections. Apply Protein Block and incubate for 5 minutes at room temperature to block nonspecific background staining. Apply primary antibody and incubate according to manufacturer's protocol. Apply Biotinylated goat anti rabbit IgG(H+L) and incubate for 10 minutes at room temperature. Apply Streptavidin Peroxidase and incubate for 10 minutes at room temperature. Add 20ul DAB Chromogen to 1 mL of DAB Substrate, mix by swirling and apply to tissue. Add enough drops of Hematoxylin to cover the section. Rinse 7–8 times in tap water. Add Mounting Medium to cover the section. The image of the IHC stained slides were captured using a Carl Zeiss Axioskop 2 plus microscope (Carl Zeiss, Thornwood, NY). The following primary antibodies were used: Anti-glucagon (Cell Signaling Technology, Cat# 2760); Anti-cytokeratin 19 (GeneTex, GTX112666); Anti-insulin (Cell Signaling Technology, Cat# 3014); Anti-ARID1A (Abnova, MAB15809); Anti-IL-1β (Abcam, ab9722); Anti-F4/80 (Cell Signaling Technology, Cat# 70076).

Western blot analysis

For Western blot analysis, cells were harvested by RIPA lysis buffer (150 mmol/L NaCl, 10 mmol/L Tris, pH 7.5, 1% NP40, 1% deoxycholate, 0.1% SDS, protease inhibitor cocktail). Cellular proteins were resolved by the 10% Bis-Tris gradient gel (Invitrogen), transferred to the polyvinylidene difluoride membrane, blocked in 5% non-fat milk in PBS/Tween-20, and probed by the following primary antibodies: anti-insulin (Cell Signaling Technology, Cat# 3014); anti-ARID1A (Abnova, MAB15809); anti-Neurogenin3 (Santa Cruz Biotechnology, sc-374442), and anti-MafA (Santa Cruz Biotechnology, sc-390491). Chemiluminescence substrate was applied using SuperSignal West Pico Chemiluminescent Substrate (#34080, Thermo Fisher Scientific) or SuperSignal West Femto Maximum Sensitivity Substrate (#34095, Thermo Fisher Scientific) and blots were analyzed using the ChemiDoc Touch Imaging System (#1708370, Bio-Rad).

Real-time-quantitative PCR analysis (RT-qPCR)

Total RNAs were extracted from tissues and cells using TRIzol reagents (Invitrogen, USA) according to the manufacturer’s protocol and quantified by Nanodrop (Thermo Fisher Scientific, USA). Synthesis of cDNAs was executed using the GoScript RT system (Promega). Quantitative PCR was done as previously described [18]. The qRT-PCR analysis was performed using PowerTrack SYBR Green Master Mix (Applied Biosystems) on an IQ5 RT-PCR system (BioRad, USA). 18S or GAPDH was used as the internal control to normalize the data to determine the relative expression of the target genes by using the 2−ΔΔCt method. The primers for RT-qPCR were listed in Table S2.

SAHA treatment

For in vivo treatment, SAHA was first dissolved in DMSO at a dose of 50 mg/mL, which was then diluted with phosphate-buffered saline to a final concentration of 5 mg/mL directly before use. Seven-week-old PCA^L/Lmice were administered twice weekly by intraperitoneal injections of 50 mg/kg of SAHA or phosphate-buffered saline vehicle for 4 weeks (n = 5, per group). Blood sugar levels were continuously monitored every two weeks for 8 weeks. At the end of the experiment, blood samples were collected and insulin levels were quantified by ELISA kit. Mice were sacrificed, and the pancreases were collected for histological analyses.

Quantification and statistical analysis

All experiments were performed at least in triplicate, and representative data were shown. Statistical analysis was performed by one- or two-way ANOVA using Prism 5.0 software to identify differences among different experimental groups. The median survival was estimated using the Kaplan-Meier method. Quantitative data were expressed as mean±SEM and p value less than 0.05 was considered significant. The significance was presented as ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001; and no significant difference was presented as ns.

Published: January 20, 2023

Data and code availability

• •Next-generation sequencing data generated in this study have been deposited in the NCBI GEO with accession number GSE220638. Metabolism data, ChIP-seq data and RNA-seq data reported in this paper have been deposited at Mendeley Data and are publicly available with accession number https://data.mendeley.com/datasets/zybcfhdb58.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.