The Chd4 Helicase Regulates Chromatin Accessibility and Gene Expression Critical for β-Cell Function In Vivo

Chromatin is a dynamic molecular structure of DNA and nucleosomes, which are histone octamers that serve to package the eukaryotic genome and regulate accessibility of DNA for recombination, damage repair, replication, and transcription. Since the occupancy of nucleosomes along the DNA strand hinders access of transcriptional regulators, modulation of chromatin accessibility is a major control point in gene expression. Movement of nucleosomes along the DNA strand is facilitated by multiprotein chromatin-remodeling complexes that use the energy derived from ATP hydrolysis to exchange, slide, or evict histones (1). Maintenance of transcriptional regulation by these complexes is largely mediated by recruitment to specific promoter or enhancer genomic loci regions in a transcription factor (TF)-dependent manner.

Transcriptional dysregulation is the basis of numerous diseases, including type 2 diabetes (T2D). It was demonstrated that a subset of failing human islet β-cells from donors with T2D lose activity and expression of islet-enriched TFs, including PDX1, NKX6.1, and MAFA (2). Under physiological conditions, these TFs direct the transcription of genes essential for glucose-stimulated insulin secretion (GSIS), including Ins, components of glucose transport and metabolism (e.g., Slc2a2, Gck), components of K_ATP channels and voltage-gated calcium pumps (e.g., Kcnj11, Abcc8, Vdccβ, Vdccα1D, Serca2b, Serca3), and components necessary for hormone processing and insulin granule exocytosis (e.g., Pcsk1, Pcsk2, Cpe, Stx1a, Stxbp1, Snap25, Vamp2, Vamp3, Syt4, Sytl4) (3). Loss of transcriptional activity of key islet-enriched TFs leads to reduced expression of these components necessary for GSIS, impaired β-cell function, and ultimately overt diabetes. Therefore, it is of critical importance to understand the mechanisms by which TFs control expression of genes to maintain GSIS under physiological conditions and how their activities become disrupted in settings of T2D.

Numerous transcriptional coregulators have been identified to modulate the activity of β-cell–specific TFs and to play a role in maintaining β-cell function in vivo. For example, conditional whole-islet deletion of Ldb1 impaired Isl-1 target gene expression in α- and β-cells and led to hyperglycemia and reduced hormone production (4,5). Additionally, conditional deletion of the Mll3/4 subunit NCoA6 in mouse β-cells impaired transcription of MafA target genes and ex vivo GSIS (6). Similarly, β-cell–specific deletion of Swi/Snf chromatin remodeling activity in mice impaired binding of Pdx1 to Ins2, which blunted insulin production leading to impaired whole-body glucose homeostasis (7).

Recently, our group demonstrated that the chromodomain helicase 4 (Chd4) subunit of the nucleosome remodeling and deacetylase (NuRD) complex is required to maintain β-cell function in vitro (8). We found that Chd4 dynamically interacts with the essential β-cell TF, Pdx1, in mouse and human β-cell lines and primary tissue. In vivo, acute glucose stimulation enhanced Pdx1–Chd4 interactions, and transient reduction of Chd4 in rodent β-cell lines impaired GSIS. Interestingly, interactions between Pdx1 and Chd4 are severely impacted in T2D settings, as β-cells from tissue of mice fed a high-fat diet and tissue of human donors with T2D contained significantly fewer interactions compared with tissue of mice fed a normal diet or tissue of donors without diabetes, respectively.

Based on these findings and the profound observation that loss of CHD4-PDX1 complex formation is a significant feature of T2D pathophysiology (8), further evaluation is necessary to uncover the mechanistic basis by which Chd4 modulates the chromatin landscape and gene expression programs required for optimal β-cell function in vivo. To this end, we generated a tamoxifen-inducible β-cell–specific Chd4-deficient mouse model. We demonstrate that loss of Chd4 from the mature β-cell impairs whole-body glucose homeostasis and islet insulin secretion and leads to increased immature insulin granules and excess proinsulin throughout Chd4-deficient β-cells. These alterations were found to be the consequence of a disordered chromatin landscape and differential gene expression programs critical for normal β-cell function, several of which were confirmed following CHD4 knockdown in human β-cell lines. Collectively, our results highlight a predominant role for Chd4 in controlling gene expression signatures in adult β-cells.

MIP-Cre^ERT (9) mice were used to remove the LoxP sites flanking exons 12 and 21 of the Chd4 locus (Chd4^f/f [10]) and the stop cassette in the Rosa26-Loxp-Stop-Loxp-tdTomato lineage reporter (R26^LSL-tdTomato [11]). The following genotypes were used: control (MIP-Cre^ERT;Chd4^f/+;R26^{LSL-tdTomato/+}) and Chd4^Δβ (MIP-Cre^ERT;Chd4^f/f;R26^{LSL-tdTomato/+}). Cre^ERT-mediated recombination of Chd4^f/f and the R26^LS-LtdTomato was achieved through administration of 100 mg/kg tamoxifen (T5648; Sigma-Aldrich) by oral gavage once per day for a 5-day period.

Mice (n = 10–18) were given an injection of d-glucose in PBS (2 mg/g body wt i.p.) after a 6-h fast. Blood glucose was measured with an AimStrip Plus Blood Glucose Meter (Germaine Laboratories). Plasma insulin under ad libitum feeding conditions or following a 6-h fast was collected and measured by radio immunoassay at the Translation Core at Indiana University School of Medicine. For glucose-stimulated plasma insulin and proinsulin measurements, plasma was collected 2 min following injection with d-glucose (2 mg/g body wt) and measured by radio immunoassay or ELISA.

Whole mouse pancreata were isolated and fixed for 4 h in 4% (v/v) paraformaldehyde and then washed three times in PBS and placed in 30% sucrose overnight at 4°C. The following day, pancreata were embedded in optimal cutting temperature embedding medium and frozen in a −80°C freezer. Sections were generated at 6-μm thickness on a cryostat (Leica Biosystems). We performed immunofluorescence staining by incubating slides with primary antibodies, described in Supplementary Table 1.

Images were acquired on a Zeiss LSM 800 confocal laser scanning microscope and processed with ImageJ software. For β-cell area quantitation, six sections (∼300 µm apart) from control and Chd4^Δβ mice were analyzed for insulin staining with use of 3,3′-diaminobenzidine substrate and counterstained with eosin. The percentage of insulin⁺ area relative to pancreas area was calculated. For quantitation of immunofluorescence images, we restricted the quantitation to the Tomato⁺ area using a mask function in Image J and subsequently measuring the pixel intensities of MafA, CgA, and CgB staining.

Islets were isolated (12) and perifusion was performed by the Islet and Physiology Core at Indiana University School of Medicine. Insulin secretion samples were measured with ELISA by the Translation Core at Indiana University School of Medicine. Insulin secretion was normalized to DNA content as measured with the Quant-iT PicoGreen dsDNA Assay Kits and dsDNA Reagents (P7589; Invitrogen).

For measurement of intracellular Ca²⁺, isolated islets were incubated in 5.5 mmol/L glucose media for 48 h and loaded with the ratiometric Ca²⁺ indicator Fura-2 acetoxymethyl ester (Fura-2, AM) (8 μmol/L, F-1221; Invitrogen) and 0.04% Pluronic F127. Islets were transferred for imaging to a glass-bottom plate containing Hanks’ balanced salt solution supplemented with 0.2% BSA, 1.0 mmol/L Mg²⁺, 2.0 mmol/L Ca²⁺, and 5.5 mmol/L glucose for baseline measurements. Plates were mounted on the stage of a Zeiss Axio Observer microscope and incubated at 37°C with use of an in-line heater (Pecon) for 10–15 min before being perifused with either 16.7 mmol/L glucose or 30 mmol/L KCl. Excitation light from a xenon burner was supplied to the preparation via a light pipe and filter wheel (Sutter Instrument Company). Images were taken sequentially under conditions of 340 nm and then 380 nm excitation with a Hamamatsu ORCA-ER camera for production of data representing each intracellular Ca²⁺ ratio from emitted light at 510 nm.

Approximately 1,000 isolated islets from four control and five Chd4^Δβ animals were suspended separately in 100 μL media and 100 μL 2% glutaraldehyde + 4% paraformaldehyde in 0.1 mol/L sodium cacodylate for 10 min. Supernatant was removed and 500 μL fixative was added to the islets for 1 h. Fixative was removed and replaced with fresh fixative, and islets were resuspended. The fixed islets were sent to the Advanced Electron Microscopy Core Facility at the University of Chicago for further processing and imaging.

Mature and immature insulin granules were quantitated with the ImageJ Cell Counter plugin. Granules were quantitated as a percentage of mature, immature, or rod like divided by the total number of counted granules in each image. A total of 20 images were quantitated per group.

Isolated islets were dispersed into a single-cell suspension (Accumax, A7089; Sigma-Aldrich), stained with DAPI, and sorted by gating for Tomato⁺DAPI⁺ cells with FACS at the Indiana University School of Medicine Flow Cytometry Resource Facility. RNA was isolated from FACS-purified β-cells (mean ± SEM 30,648 ± 6,565 cells [control; n = 4] and 18,725 ± 3,161 cells [Chd4^Δβ; n = 4]) with use of the RNAqueous-Micro kit and analyzed on 2100 Bioanalyzer (Agilent Technologies). Only samples with an RNA integrity number >7.5 were used for cDNA synthesis and library preparation.

One nanogram of total RNA per sample was used for library preparation. cDNA was first synthesized with use of SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara Bio). A dual indexed cDNA library was then prepared with use of Nextera XT DNA Library Prep Kit (Illumina). Each library was quantified and its quality accessed with Qubit and Agilent Bioanalyzer, and multiple libraries were pooled in equal molarity. The average size of the library insert was ∼300–400 base pairs (bp). The pooled libraries were then denatured and neutralized before loading to the NovaSeq 6000 sequencer for 100b paired-end sequencing (Illumina). Approximately 30–40 × 10⁶ reads per library were generated. A Phred quality score (Q score) was used to measure the quality of sequencing. More than 95% of the sequencing reads reached Q30 (99.9% base call accuracy). The generated FASTQ files were processed with the Genialis visual informatics platform (https://www.genialis.com).

A total of 27,000–80,000 Tomato⁺ mature β-cells were collected for assay for transposase-accessible chromatin with sequencing (ATAC-Seq). Cells were first lysed to collect nuclei through incubating with 100 μL lysis buffer (10 mmol/L Tris-HCl, 10 mmol/L NaCl, 3 mmol/L MgCl₂, 0.10% Tween-20, 0.10% Nonidet P40 Substitute, 0.01% Digitonin, 1% BSA, 1 mmol/L dithiothreitol [DTT], 40 units/μL RNase Inhibitor) for 3 min, centrifuging at 300 rcf for 5 min at 4°C, and discarding the supernatant. The resulting nuclei were then washed with 1 mL wash buffer (10 mmol/L Tris-HCl, 10 mmol/L NaCl, 3 mmol/L MgCl₂, 0.10% Tween-20, 1% BSA, 1 mmol/L DTT, 40 units/μL RNase Inhibitor) and centrifuged at 200 rcf for 5 min at 4°C two to three times. The final nuclei, ranging from 14,000 to 45,000 in number, were used for tagmentation and library preparation following procedures previously described (13). In brief, the nuclei were immediately suspended in 50 μL tagmentation reaction with use of 2× TD Buffer (FC-121-1030; Illumina) and Nextera Tn5 Transposase enzyme (FC-121-1030; Illumina). Fragments >600 bp were excluded in library preparation. Each resulting indexed library was quantified and its quality assessed with Qubit and Agilent Bioanalyzer, and multiple libraries were pooled in equal molarity. The pooled libraries were then denatured and neutralized before loading to NovaSeq 6000 sequencer at 300 pmol/L final concentration for 100b paired-end sequencing (Illumina). Approximately 100 × 10⁶ reads per library were generated. A Phred quality score (Q score) was used to measure the quality of sequencing. More than 90% of the sequencing reads reached Q30 (99.9% base call accuracy).

HOMER (Hypergeometric Optimization of Motif EnRichment) (14) was used to perform motif enrichment analysis on those differentially accessible chromatins (DAC) and enhancer regions that were found in Chd4^Δβ β-cells. The search lengths of the motifs were 10 bp. P values were calculated through comparison of the enrichments within the target regions with those of a random set of regions (background) generated by HOMER.

Human EndoC-βH1 cells (15) were cultured at 37°C in 5% CO₂ in low glucose (1 g/L) DMEM, 2% albumin from bovine serum fraction V, 50 μmol/L 2-mercaptoethanol, 10 mmol/L nicotinamide, 5.5 μg/mL transferrin, 6.7 ng/mL sodium selenite, and 1% PSA with passage numbers ranging between 85 and 95. siRNA knockdown in EndoC-βH1 cells was achieved with use of ON-TARGETplus siRNAs targeting human CHD4 (no. L-009774-00; Dharmacon). Targeting siRNA or a nontargeting control (no. D-001810-10; Dharmacon) was diluted in Opti-MEM and incubated with Lipofectamine RNAiMAX (LMRNA015; Invitrogen) at a 1:3 ratio for 5 min. Cells suspended in Opti-MEM were combined with the mixture of siRNA and Lipofectamine RNAiMAX at a density of 2 × 10⁶ cells in a 6-well dish until adherent, and then media was replaced with EndoC-βH1 growth media. GSIS was performed and RNA and nuclear extracts were collected 72 h after transfection. Nuclear extracts were resolved by SDS-PAGE and immunoblotted with CHD4 and β-actin antibodies. RNA was purified per the manufacturer’s instructions (D7001; Zymo Research) and cDNA prepared (4368814; Applied Biosystems). Quantitative PCR reactions were performed with the gene primers listed in Supplementary Table 2 on a QuantStudio 3 Real-Time PCR System (A28567; Applied Biosystems). Gene expression changes were analyzed with the 2^−ΔΔCT method (16) with use of GAPDH for normalization. For GSIS, cells were treated with baseline glucose solution at 1 mmol/L glucose for 1 h. Cells were then treated with either 2.8 mmol/L glucose (low) or 16.7 mmol/L glucose (high) for 1 h. Secretion media were collected, and cells were treated with acid/ethanol solution for collection of contents. Human insulin and proinsulin ELISAs were performed by the Translation Core at Indiana University School of Medicine. Insulin secretion samples were normalized to insulin content.

Statistical significance was determined with the two-tailed Student t test for comparison of two experimental groups or one-way ANOVA with Tukey post hoc analysis for comparing more than two groups. Data are presented as means ± SEM. A threshold of P < 0.05 was used to declare significance.

All animal studies were reviewed and approved by the Indiana University Institutional Animal Care and Use Committee. Mice were housed and cared for according to the Indiana University Laboratory Animal Resource Center and the Institutional Animal Care and Use Committee/Office of Animal Welfare Assurance standards and guidelines.

Raw and analyzed RNA-sequencing and ATAC-Seq data sets have been deposited in Gene Expression Omnibus (GEO) (accession no. GSE217446). All noncommercially available resources generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

To evaluate the contributions of Chd4 in controlling β-cell function in vivo, we crossed transgenic mice containing a tamoxifen-inducible, β-cell–specific Cre recombinase (mouse Ins1 enhancer/promoter [MIP]-driven Cre^ERT [9]) and the Rosa26-Loxp-Stop-Loxp-tdTomato (R26^LSL-tdTomato [11]) lineage reporter with mice containing LoxP sites flanking exons 12–21 of the Chd4 gene (i.e., Chd4^Δβ [10]) (Supplementary Fig. 1A and B). All experimental and control (MIP-Cre^ERT;Chd4^f/+) animals contain the MIP-Cre^ERT transgene and received tamoxifen, as this line has been shown to augment islet β-cell mass and function independently (17,18). Removal of Chd4 was achieved through administration of five doses of tamoxifen over a 5-day period. Four weeks following the last tamoxifen dose, immunofluorescence analysis confirmed removal of Chd4 protein from the majority of β-cells throughout the islets of Chd4^Δβ mice (Supplementary Fig. 1C and D).

Male Chd4^Δβ mutants displayed impaired glucose tolerance 4 weeks following the last tamoxifen treatment with no alterations in body weight (Fig. 1A and Supplementary Fig. 2A). Importantly, we compared MIP-Cre^ERT–positive mice (MIP-Cre^ERT;Chd4^+/+) with our controls (MIP-Cre^ERT;Chd4^f/+), which revealed no change in glucose tolerance between the Cre-alone animals and heterozygous controls (Supplementary Fig. 2B). Based on the breeding strategies, efforts to use control and experimental littermates, and the similar phenotypes between MIP-Cre^ERT–positive and MIP-Cre^ERT;Chd4^f/+ mice, we selected the latter as the control cohort throughout. Female Chd4^Δβ mutants also displayed the glucose intolerance phenotype at 4 weeks post–tamoxifen treatment, although not as severe as male littermates (Supplementary Fig. 2C). Moreover, the phenotype did not appear to become worse as the animals age, as at 8 weeks post–tamoxifen treatment glucose tolerance in female Chd4^Δβ mutants was similar to that at 4 weeks (Fig. 1A and Supplementary Fig. 2C and D). Ad libitum feeding blood glucose levels were elevated in male and female Chd4^Δβ mutants, and plasma insulin levels were reduced in male, but not female, mutants (Fig. 1B and C and Supplementary Fig. 2E and F). The defect in glucose homeostasis and reduced plasma insulin levels suggested that there could be a decrease in islet β-cell area; however, analyses from 4 weeks post–tamoxifen treatment in male mutants show no difference (Fig. 1D).

As β-cell area was unchanged, we queried whether insulin secretion from male Chd4^Δβ mutant islets was impaired. Islet perifusion analysis was used to examine the dynamics of insulin secretion ex vivo at rest (2.5 mmol/L glucose), in response to 16.7 mmol/L glucose, and in response to 2.5 mmol/L glucose with direct depolarization by 30 mmol/L KCl. As expected, based on in vitro Chd4 knockdown followed by GSIS in rat INS-1 832/13 β-cell lines (8), Chd4^Δβ mutant islets have a marked reduction in insulin secretion at steady state, 1st phase, and 2nd phase and under KCl-induced depolarizing conditions (Fig. 2A and B).

The profound defects in insulin release following glucose stimulation and KCl-induced depolarization suggested that stimulation of Ca²⁺ influx through voltage-sensitive Ca²⁺ channels could be altered. Therefore, Ca²⁺ imaging experiments using the ratiometric Ca²⁺ indicator Fura-2, AM, were conducted in islets from male control and Chd4^Δβ mutants. Remarkably, while phase 1 duration was modestly expedited in Chd4^Δβ islets, no alterations were observed in baseline Ca²⁺ flux prior to glucose stimulation or phase 1 amplitude, oscillatory amplitude, or oscillatory duration after glucose stimulation (Supplementary Fig. 3). Moreover, islets treated with 2.5 mmol/L glucose and 30 mmol/L KCl presented with no change in cytosolic Ca²⁺ (Fig. 2C). Collectively, these observations indicate that the insulin secretion defects are largely independent of Ca²⁺ flux dynamics.

To define the molecular influence of Chd4 in controlling gene expression programs within the β-cell, we performed RNA sequencing on FACS Tomato⁺ β-cells from dispersed male control and Chd4^Δβ islets. Using a twofold cutoff and false discovery rate of <0.05, we found 438 down- and 1,093 upregulated genes in Chd4^Δβ β-cells (Fig. 3A and Supplementary Table 3). Several of the genes with reduced expression include those encoding the MafA TF (MafA), islet secretory granule molecules chromogranin A (Chga) and chromogranin B (Chgb), the Ca²⁺-binding synaptotagmin 10 (Syt10), the primary glucose transporter Glut2 (Slc2a2), and roundabout receptor 2 (Robo2) and its ligand Slit1—all critical factors in β-cell physiology (19). We confirmed the reduction of several of these important modulators of insulin production and secretion at the protein level, including MafA, CgA, and CgB (Fig. 4). Additionally, disallowed genes Hk2, Pdgfra2, and Mycl, and numerous Serpina and Serpine genes, were upregulated in Chd4^Δβ β-cells (Fig. 3A and B and Supplementary Table 3). The Serpina family of genes cluster together on chromosome 12 and are downregulated in the late stages of the secondary transition, a temporary point critical for endocrine cell development (20). SERPINE2 expression has been shown to be upregulated in T2D donor islets (21), further highlighting that the transcriptional programming governed by Chd4 is closely linked to diabetes pathogenesis. Moreover, we observed no significant alterations in expression of fundamental islet-enriched TF mRNAs (e.g., Foxo1, Hnf1b, Isl1, Mnx1, Neurod1, Nkx2.2, Pax6, Pdx1) (Supplementary Fig. 4A and B). Lastly, we found no increased levels of glucagon or somatostatin in Tomato⁺Chd4^Δβ β-cells (Supplementary Fig. 4B), supporting that loss of Chd4 does not lead to aberrant expression of α-cell– and δ-cell–enriched genes.

As Chd4 has been shown to interact with Pdx1 and play instrumental roles in maintaining Pdx1 function (8), we compared those genes differentially expressed in Chd4^Δβ β-cells with genes bound by Pdx1 in mouse islets (22), which revealed a partial overlap of 385 genes. Gene ontology (GO) analysis of these genes with use of the Database for Annotation, Visualization, and Integrated Discovery (DAVID) led to the identification of pathways linked to hormone secretion and regulation of K⁺ ion transmembrane transport (Fig. 3C), the latter of which is supported by the significant defect in insulin secretion when Chd4^Δβ islets are treated under KCl-mediated depolarizing conditions (Fig. 2A).

The reductions in CgA and CgB could partially explain the defects in insulin secretion from Chd4^Δβ β-cells. Loss of CgB obstructs proinsulin processing leading to accumulation of proinsulin content in β-cells and causes an accumulation of immature insulin secretory granules (23,24). Therefore, we used transmission electron microscopy to provide detailed visualization of the cytoarchitecture of male control and Chd4^Δβ islets. We quantitated the number of mature, immature, and rod-like granules throughout and observed an increased percentage of immature insulin granules in male Chd4^Δβ islets (Fig. 5A and B). Accumulation of immature secretory granules could be resultant from defects in proinsulin conversion to insulin. In support of this idea, we found that male Chd4^Δβ islet contents have increased proinsulin-to-insulin ratios (Fig. 5C). Next, plasma collected from male Chd4^Δβ mutant mice 2 min following a glucose injection revealed that Chd4^Δβ mutants secreted significantly more proinsulin into the bloodstream and that overall proinsulin-to-insulin ratios were increased in Chd4^Δβ mutants (Fig. 5D and E).

To interrogate chromatin architecture changes in the Chd4^Δβ mutants, we performed ATAC-Seq on flow-sorted Tomato⁺ cells from dispersed male control and Chd4^Δβ islets. Interestingly, of the DAC regions identified in Chd4^Δβ β-cells, 872 resided within promoters (±1 kb from transcription start site [TSS]) (Fig. 6A), corresponding to 81 genes with closed and 791 genes with open chromatin in Chd4^Δβ β-cells. Of the 791 genes with open chromatin at the promoter, 218 were upregulated, and 12 of the 81 genes with closed chromatin were downregulated, in Chd4^Δβ β-cells. In GO pathway analyses, these concordant overlapping genes were linked to pathways associated with regulation of insulin secretion and K⁺ export across plasma membrane (Fig. 6A and Supplementary Table 4). We next aligned DAC regions from Chd4^Δβ β-cells with those defined to be functional enhancers within islets (25). There were 283 defined islet-enhancer regions that were differentially accessible in Chd4^Δβ β-cells. Of the genes associated with those enhancers, 12 were differentially expressed in Chd4^Δβ β-cells, including MafA (Supplementary Table 5).

Previously, we established that Chd4 binds to an upstream MafA regulatory sequence termed region 3 (base pairs −8118 to −7750 relative to the TSS [26]) in rodent β-cell lines (8). In support of this, across replicate samples, we observed MafA region 3 is in a more closed chromatin state (Fig. 6B). Moreover, we found that Slc2a2 (encoding the Glut2 primary glucose transporter in the rodent β-cell) expression is reduced (Fig. 3A and B), with chromatin accessibility being compromised at the promoter and intragenic regions (Supplementary Fig. 5). In a concordant manner, accessible chromatin signals were enhanced at promoter sites of genes upregulated in Chd4^Δβ β-cells, including Kcnk5 and Serpina1a (Fig. 6C and Supplementary Fig. 5). Interestingly, we found differential regulation of alternate Chd members of the same class II Chd family. Whereas Chd5 was found to be upregulated and promoter/intragenic regions showed increased chromatin accessibility in Chd4^Δβ β-cells (Supplementary Fig. 5), transcript levels of Chd3 were unchanged in Chd4^Δβ β-cells. Interestingly, we observed that both Chd3 and Chd5 proteins are elevated in Tomato⁺Chd4^Δβ β-cells (Supplementary Fig. 4B). These data suggest partial compensation or possible functional redundancy by alternate Chd helicase subunits in the absence of Chd4, albeit to a degree that does not fully rescue the loss of Chd4.

We observed a relatively small number of differentially expressed genes (DEGs) in Chd4^Δβ β-cells that overlap with Pdx1-bound genes (Fig. 3C). Those DEGs not bound by Pdx1 suggest that the differences could be driven by either secondary effects, or, most likely, other islet-enriched TFs that are modulated by Chd4 activity. To this end, we performed HOMER analysis on our ATAC-Seq data set. As expected, a conserved homeobox-domain binding site (TAAT) was found to be a region with DAC at promoter regions (Supplementary Fig. 6A). There are several islet-enriched TFs, including Pdx1, that bind to this consensus sequence in the islet (Supplementary Fig. 6A). Additionally, binding sites for NeuroD1 and Smad proteins were identified at both promoter and enhancer regions (Supplementary Fig. 6A and B), opening the possibility that Chd4 is being recruited by these TFs to genomic control regions. NeuroD1 is of particular interest, as global deletion leads to arrest of islet development, loss of final β-cell mass, severe hyperglycemia, and perinatal death (27). To further interrogate the relationship between Chd4 and NeuroD1 in the β-cell, we analyzed ChIP-sequencing data sets for NeuroD1-occupied enhancer loci in mouse islets, which were also flanked by H3K4me1-marked nucleosomes (28), to evaluate their overlap with Chd4^Δβ DEGs and those bound by Pdx1 (from mouse islets [22]). We revealed a notable overlap of genes bound by both Pdx1 and NeuroD1, with 74 of Chd4^Δβ DEGs found in all three gene sets (Supplementary Fig. 7). Those genes in all three data sets include β-cell functional genes such as Robo2, Slc2a2, and Chga (Supplementary Table 6). There was a relatively small number of Chd4^Δβ DEGs that reside in NeuroD1-bound genes (23 genes), suggesting other islet-enriched TFs are also likely being governed by Chd4.

We next assessed the functional role of human CHD4 using siRNA in the well-characterized human β-cell line model, EndoC-βH1 cells. CHD4 levels were reduced ∼50% from EndoC-βH1 cells following siRNA-mediated depletion (Fig. 7A). Analyses of various genes linked to β-cell function and those differentially expressed in Chd4^Δβ β-cells revealed significant reductions in MAFA, G6PC2, and INS (Fig. 7B). The reduced expression of INS was not expected, as levels of Ins1 and Ins2 are unchanged in both Chd4^Δβ β-cells and rodent β-cell lines following transient knockdown of Chd4 (8). GSIS following CHD4 knockdown in EndoC-βH1 cells revealed a marked reduction in insulin secretion, with no significant alteration in insulin content (Fig. 7C and D). In agreement with observations from Chd4^Δβ islets, we reveal a trending elevation of proinsulin and proinsulin-to-insulin ratios in EndoC-βH1 cell contents following knockdown of CHD4 (Fig. 7E and F).

Blood glucose homeostasis requires adapted responses from hormone-secreting islet cells and hormone-sensitive peripheral tissues such as liver, muscle, and adipose cells. The adaptation by the islet β-cell is largely regulated at the transcriptional level, through interaction among DNA-bound TFs, recruited transcriptional coregulators, and the basal transcriptional machinery. Although numerous coregulators with diverse enzymatic activities exist in mammalian cells, few have been characterized to govern the activities of islet-enriched TFs. We previously showed that interactions between Pdx1 and the Chd4 subunit of the NuRD complex are amplified under acute glucose stimulatory conditions and are significantly reduced in pathophysiological conditions associated with diabetes (8), suggesting that Chd4 is a modulator of Pdx1 activity and, therefore, β-cell function. Here, we report the importance of Chd4 in governing β-cell gene expression, chromatin accessibility, β-cell function, and mature insulin granule production in vivo.

In this regard, conditional removal of Chd4 from mature islet β-cells leads to disruptions of gene expression programs that induce glucose intolerance and impaired insulin secretion (Figs. 1 and 2). For example, we demonstrate that Chd4 has a prominent role in regulating MafA, a critical TF important for insulin secretion, by augmenting chromatin accessibility at Pdx1-regulated MafA region 3 enhancer (Figs. 3 and 6). Furthermore, we identified a role for Chd4 in positively regulating the genes encoding CgA and CgB, which are essential for appropriate granule maturation, and in maintaining appropriate proinsulin-to-insulin ratios in the β-cell (23,24,29). Interestingly, in addition to reduced mature granule density, we also found elevated proinsulin-to-insulin ratios in Chd4^Δβ islets (Fig. 5). Given the history of the MIP-Cre^ERT and other transgenic Cre-driver lines in augmenting β-cell function (30), all control and experimental animals included MIP-Cre^ERT and received tamoxifen. While we controlled for factors that could confound the results, this model is not without its limitations, as we are unable to distinguish any potential role of Chd4 in controlling Cre expression/activity or Tamoxifen actions within the β-cell.

Deletion of Pdx1 from mature β-cells leads to loss of β-cell function and identity, where Pdx1-deficient cells acquire molecular and transcriptional signatures of α-cells, including expression of Gcg and MafB (31). This suggests that Pdx1 activates β-cell functional genes and represses non–β-cell genes. As Chd4 has been shown to maintain identity of a variety of cell types (e.g., cardiac muscle cells, skeletal muscle cells, embryonic stem cells), we expected to observe an increase in non–β-cell genes in Chd4-deficient β-cells. However, we did not observe upregulation of genes encoding non–β-cell hormones (e.g., Gcg, Sst, Ppy, Ghrl) (Supplementary Fig. 4), but we did detect changes in several disallowed genes (Hk2, Pdgfra2, Mycl, Serpina and Serpine family members) (Fig. 3), suggesting a partial reprogramming of the β-cell transcriptional state on Chd4 loss. Moreover, this indicates that other coregulators are contributing to the maintenance of Pdx1-repressive functions within the β-cell.

We found that β-cell–specific deletion of Chd4 leads to increased chromatin accessibility and upregulation of an alternate Chd isoform, Chd5, and slightly elevated protein levels of Chd3 (Supplementary Fig. 4). This observation is consistent with our recent findings that transient knockdown of Chd4 in mouse β-cell lines leads to upregulation of both Chd3 and Chd5 (8). Similarly, a satellite cell-specific deletion of Chd4 led to upregulation of both Chd3 and Chd5, which were additionally found to be bound directly by Chd4 (32). Interestingly, Chd5 has not garnered as much attention as Chd4, potentially due to having low expression in tissues outside the nervous system and testis (33). Whereas Chd5 might not play major roles outside of these systems under physiological conditions, it could become important in situations where Chd4 activity is altered, such as the multisystemic neurodevelopmental disorder Sifrim-Hitz-Weiss, caused by missense mutations in CHD4 (34). Therefore, the roles of other Chd subunits when Chd4 activity is altered will be an important avenue for future study.

Pdx1 interacts with a variety of transcriptional coregulators with various enzymatic functions in the β-cell (35). Loss of Chd4 leads to loss of expression of key β-cell functional genes, including MafA and Slc2a2, genes that are also regulated by the Swi/Snf chromatin remodeling complex (7). However, in contrast to Swi/Snf, whose activity is shown to promote Pdx1 occupancy to the Ins2 locus and robust expression of insulin (7), Chd4 does not appear to alter expression of Ins1 or Ins2 or alter chromatin accessibility of their genomic loci (Supplementary Figs. 4 and 5). These findings underscore how the recruitment of different coregulators to distinct loci is imperative for proper gene regulation. Moreover, it raises the question as to how the selective recruitment of coregulators by Pdx1 to specific genomic loci is regulated. Whether potential posttranslational modifications of Pdx1 influence its ability to interact with specific coregulators remains to be established, as Pdx1 has been shown to undergo glycosylation, SUMOylation, ubiquitination, and phosphorylation (36–39), modifications that could alter its binding capacity and transcriptional activity.

We evaluated how transient loss of CHD4 in human β-cell lines would impact gene expression signatures and islet function. Similar to Chd4^Δβ β-cells, we found reductions in MAFA and SLC2A2 (Fig. 7). Interestingly, in EndoC-βH1 cell lines, reduced expression of CHD4 also led to reduced expression of INS, which was unchanged in Chd4^Δβ mouse β-cells. Moreover, and in agreement with our mouse model findings, we revealed transient reduction in CHD4 from EndoC-βH1 cells trends toward elevated proinsulin levels. However, in contrast to Chd4^Δβ islets, EndoC-βH1 insulin content was unchanged following knockdown. We propose that the differences likely arose from the transient siRNA knockdown in EndoC-βH1 cells with insufficient time having passed to reduce insulin content. However, while gene expression programs differ between mice and humans insufficient or deficient for Chd4, β-cell function was impaired in both models, demonstrating the importance for Chd4 in governing β-cell function across species.

Overall, our results demonstrate a prominent role of Chd4 in controlling insulin secretion and modulating a subset of gene targets within the β-cell to maintain its function. Our findings are the first to demonstrate that Chd4 plays an essential role in maintaining mature β-cell function in vivo. The regulation we characterized by Chd4 opens potential therapeutic opportunities to manage β-cell dysfunction associated with T2D.

Acknowledgments. The authors thank Lori Sussel (University of Colorado), Dylan Sarbaugh (University of Colorado), and Chad Hunter (University of Alabama at Birmingham) for critically reading the manuscript. The authors thank Katia Georgopoulos (Massachusetts General Hospital) for generously sharing the Chd4 floxed mice. The authors acknowledge the support of the Islet and Physiology and Translation Cores of the Indiana University School of Medicine Diabetes Research Center (National Institutes of Health grant P30-DK097512). Sequencing was carried out in the Center for Medical Genomics at Indiana University School of Medicine, which is partially supported by the Indiana University Grand Challenges Precision Health Initiative. The authors thank the members of the Indiana University Melvin and Bren Simon Comprehensive Cancer Center Flow Cytometry Resource Facility, which is partially supported by National Institutes of Health grants P30-CA082709 and U54-DK106846. The authors thank the Advanced Electron Microscopy Facility at University of Chicago for help in sample preparation and collection of transmission electron microscopy data.

Funding. This work was supported by grants from the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases (K01-DK115633, R03-DK127129, and R01-DK129287 to J.M.S.; F31-DK128918 to R.K.D.; R01-DK121929 and R01-DK133881 to E.K.S.; and R01-DK093954, R01-DK127236, and R01-DK127308 to C.E.-M.), the U.S. Department of Veterans Affairs Merit Award (I01-BX001733 to C.E.-M.), an award from the Ralph W. and Grace M. Showalter Research Trust and Indiana University School of Medicine (J.M.S.), bridge funding from the Herman B Wells Center for Pediatric Research (J.M.S.) and an Early Career Development Award from the Central Society for Clinical and Translational Research (J.M.S.).

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

Author Contributions. R.K.D. and J.M.S. designed the study and drafted the manuscript. R.K.D., S.K., T.K., J.X., M.O., R.N.B., N.C., and J.M.S. executed the experiments. R.K.D., S.K., and J.M.S. performed data analysis. W.W. performed bioinformatics analysis. S.K., W.W., T.K., J.X., M.O., R.N.B., C.E.-M., and E.K.S. edited the manuscript. All authors read and approved the final manuscript. J.M.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract and oral presentation form at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.