The miR-203/ZBTB20/MAFA Axis Orchestrates Pancreatic β-Cell Maturation and Identity During Weaning and Diabetes Free

Pancreatic β-cells are specialized for lowering blood glucose levels by secreting insulin. Glucose-stimulated insulin secretion (GSIS) follows a biphasic time course consisting of a rapid and transient first phase, followed by a slowly developing and sustained second phase (1). This highly sensitive GSIS capacity is acquired by postnatal β-cell maturation, a cell autonomous process triggered in mice by weaning from high-fat maternal milk to a carbohydrate-rich chow diet (2). Intriguingly, recent literature has provided considerable evidence that metabolic stresses are able to reverse the specialized β-cells into immature neonatal-like ones, leading to insulin secretion defect and diabetes onset (3–6). The stressed β-cells share profiles with postnatal β-cells in cellular metabolism and gene transcription that are massively controlled by mTORC1 to AMPK switching (7). However, mechanisms responsive to the reversible β-cell remodeling at weaning and in diabetes remain largely undefined.

PDX1, NEUROD1, and MAFA are β-cell-specific transcriptional factors (TFs) unique to β-cell development and functional maturity. They separately or synergistically bind to the enhancer/promoter elements of β-cell functional genes (8). They can also transcriptionally inhibit the expression of disallowed genes that lead to β-cell metabolic remodeling and de-/transdifferentiation (9). Human genetic studies have shown that mutations of β-cell TFs cause maturity-onset diabetes of the young (MODY), a monogenic form of diabetes (10). Selective decreases in these TFs are found in diabetes islets, related to β-cell dysfunction and diabetes onset (11). Among these TFs, MAFA, but not PDX1, is more specific for postnatal β-cell maturation (12,13).

miRNAs can play important roles in virtually all aspects of physiological and pathological processes, including β-cell maturation (14). miR-203, an intergenic miRNA, is transcribed from as a primary transcript (pri-miR-203), which then undergoes a two-step cleavage to form the mature miR-203. Reports have shown that the mature form of miRNA (miR-203) has critical roles in neurodegenerative diseases and metabolic disorders. For example, it is transcriptionally regulated by TCF7L2 and participates in antiobesity effects in mice fed a high-fat diet (HFD) (15). In pancreatic islets, the expression level of miR-203 changes both at weaning and in type 2 diabetes rodents (16,17), opposite to the changes in MAFA. However, the in vivo role of miR-203 in β-cell biology is not yet appreciated.

In the current study, we used both gain-of-function and loss-of-function miR-203 to identify the role of miR-203 in β-cell function and maturity. We showed a novel miR-203/ZBTB20/MAFA axis orchestrating β-cell function. The miR-203/MAFA pathway is also conserved in humans. Our results highlight a potential epigenetic strategy for boosting functional β-cell numbers during weaning and diabetes.

ROSA^mT/mG (JAX: 007576) and miR-203^fl/fl (JAX: 028362) C57BL/6J mice were purchased from The Jackson Laboratory. RIP2-Cre transgenic and miR-203–knockout (KO) mice were from GemPharmatech Co., Ltd. (Nanjing, China). β-Cell–specific miR-203 transgenic (βTG), RIP2-Cre;ROSA^mT/mG;βTG (βm green fluorescent protein [GFP];βTG), RIP2-Cre;ROSA^mT/mG;wild-type (βmGFP;WT), and RIP2-Cre;miR-203^fl/fl (βKO) mice were constructed by GemPharmatech Co., Ltd. All animal studies were approved by the Research Animal Care Committee of Nanjing Medical University, China (permit number: IACUC-NJMU 1707009). Mice were housed at 23°–25°C using a 12-h light/12-h dark cycle with ad libitum access to water and food.

Primary islets were cultured in medium (Connaught Medical Research Laboratories [CMRL]-1066 for human; RPMI-1640 for murine) containing 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. Mouse insulinoma MIN6 cells (passage 21–35) (18) were cultured in DMEM (Invitrogen, Grand Island, NY) with 15% FBS (Gibco, Burlington, Ontario, Canada), and rat insulinoma INS-1 cells were cultured in RPMI-1640 medium (Invitrogen) containing 8% FBS, adding 100 units/mL penicillin, 100 μg/mL streptomycin, 10 mmol/L HEPES, and 50 μmol/L β-mercaptoethanol (Sigma-Aldrich, St Louis, MO). The islets and insulinoma cells were incubated at 37°C in a suitable atmosphere containing 95% O₂ and 5% CO₂.

Tail vein blood glucose was measured using a Glucometer Elite monitor (Abbott, Maidenhead, U.K.). Intraperitoneal (i.p.) glucose tolerance tests (IPGTTs) were performed after overnight fasting. Insulin tolerant tests (ITTs) were performed after 4-h fasting. Serum levels of insulin or C-peptide were measured separately by insulin (Mercodia, 10-1247-01) or C-peptide (ALPCO Diagnostics, 80-CPTMS-E01) ELISA kits.

Adenovirus (Ad) pAd-Mafa, Pdx1, Ngn3 (MPN)-mcherry plasmid was purchased from Addgene (cat no. 61041). Ad-Zbtb20 (Prof. Zhang Wei-ping), Ad-Mafa (19), and Ad-MPN were amplified and purified by Hanbio Biotechnology (Shanghai, China).

Mice were anesthetized, and an incision was made to expose the biliary duct. A 31-gauge blunt-end cannula was inserted at the beginning of the pancreatic duct branch for ∼0.5–0.8 cm. The cannula was secured, and the microinjection pump was opened for adenovirus infusion (5 × 10⁷ plaque forming units [pfu]/g body wt).

MIN6 cells and islets were transfected with miR-203 or anti–miR-203 by using Lipofectamine 2000 (Invitrogen), as we previously described (20). miRNA transfection efficiency of primary islets was evaluated (Supplementary Fig. 1).

The Ad-GFP or Ad-Zbtb20 adenoviruses were diluted with a serum-free medium at a concentration of 1 × 10⁷ pfu/mL for islets and 2 × 10⁶ pfu/mL for MIN6 cells culturing in 3.5-cm dishes. The adenovirus-containing culture solution was added for 2 h and replaced with fresh complete medium for the indicated time. Adenovirus infection efficiency of islets was evaluated (Supplementary Fig. 2).

For perfusion, islets were collected in a syringe filter (Millex-GP; Millipore) and perfused with Krebs-Ringer bicarbonate solution (KRB) with low glucose (2 mmol/L glucose) and high glucose (20 mmol/L glucose) for the indicated time (21).

For GSIS, islets or MIN6 cells were incubated in Krebs-Ringer bicarbonate solution with low and high glucose for 1 h. Insulin contents were extracted using an acid-ethanol solution (0.15 mol/L HCl in 75% ethanol in H₂O) overnight at 4°C. Insulin levels in supernatants and insulin contents were measured by radioimmunoassay (21).

Pancreas samples were rinsed in cold PBS and fixed overnight in 4% (g/v) paraformaldehyde. The samples were then processed and embedded in paraffin, and consecutive sections were obtained for incubation with indicated antibodies. Images were captured and analyzed by a confocal laser scanning microscope (Olympus FV1200).

Western blotting was performed as previously described (21). Primary antibodies and secondary antibodies were incubated for indicated time, and the strips were measured by ImageJ.

The treated cells were fixed with 10% formalin and then cleaved in SDS lysis buffer. The resultant lysate was collected for chromatin immunoprecipitation (ChIP) assay with indicated antibodies following the manufacturer’s protocols (Millipore). The immunoprecipitated and input samples were evaluated by semiquantitative PCR or quantitative PCR (qPCR).

The dispersed islet cells were collected and examined by cell counter (Ruiyu Biotech, Shanghai, China) to ensure >90% viability. The 10x Genomics Chromium platform was used to prepare a single-cell RNA sequencing (scRNA-seq) library. The resulting libraries were sequenced on an Illumina Hiseq PE150 platform. The data cleaning, normalization, and scaling were handled by Cell Ranger algorithm (Novogene, Beijing, China), as previously described (22).

Islets and MIN6 cells were obtained for total RNA extraction by using Trizol reagent (Invitrogen). Total islet RNAs from lines 290, 214, and wild-type controls were sent for RNA-seq (Novogene). The reverse transcriptions of primary miRNA (pri-miRNA), miRNA, and mRNA were obtained as previously described (20).

Statistical analysis was performed by GraphPad Prism 9.0 software (San Diego, CA). Comparisons between two groups were done using the Student t test, whereas the comparisons among multiple groups were analyzed using one-way or two-way ANOVA. Data are presented as mean ± SD, and a P < 0.05 was considered statistically significant.

All raw data, including scRNA-seq and bulk RNA-seq generated for this study, are publicly available at Gene Expression Omnibus accession GSE232473, GSE232476, and GSE232474. All resources, including mouse strains, generated for this study are available upon request.

We confirmed the decreased level of mature miR-203 in mouse islets from weaning and HFD-fed mice (Supplementary Fig. 3A and B), consistent with a previous report. The expression levels of pri-miR-203 in islets was significantly lower in weaning mice (Fig. 1A), and was higher in HFD-fed mice (Fig. 1B). High-glucose treatment caused an increase in GSIS capacity and a decrease in pri-miR-203 and miR-203 expression (Fig. 1C and Supplementary Fig. 3C and D). Moreover, miR203 transcripts also increased in islets from rodents and patients with type 2 diabetes (Fig. 1D–F and Supplementary Fig. 3E).

We mapped the promoter region of miR203 with β-cell–specific TFs (MAFA, PDX1, and NEUROD1). Expression levels of TF in islets were increased in weaning mice but decreased in type 2 diabetes rodents (Supplementary Fig. 3F–J). JASPAR prediction discovered several potential binding sites for these TFs, whereas only MAFA inhibited miR203 promoter activity (Supplementary Fig. 3K and Fig. 1G). Moreover, interference by Mafa upregulated while overexpressing Mafa dowregulated miR203 transcripts (Fig. 1H–J and Supplementary Fig. 3L and M). These data demonstrated that nutrient transition regulates miR-203 expression.

Next, βTG mice with three different expression levels—290 (low), 214 (high), and 312 (high)—were used to investigate the role of miR-203 (Fig. 1K and Supplementary Fig. 4A–C).

Overall, the male βTG mice were born heathy. Line 214 and line 312 mice gained less body weight and had shorter body lengths after weaning (Fig. 1L and Supplementary Fig. 4D). Blood glucose levels of line 214 and line 312 mice significantly increased after weaning (Fig. 1M). Line 290 mice showed only slight hyperglycemia (Fig. 1M). We used line 214 and line 290 mice subsequent investigations due to the similarity between line 214 and line 312.

IPGTT assays showed line 214 mice exhibited a much higher glucose excursion curve (Fig. 1N). Line 290 mice exhibited slight glucose intolerance (Fig. 1N). The impaired IPGTT results of βTG mice were attributed to the insufficiencies in insulin secretion without changes in insulin sensitivities (Fig. 1O and P). Likewise, the serum level of C-peptide was also decreased in line 214 mice (Supplementary Fig. 4E). The growth retardant and diabetes phenotypes were reproduced in female βTG mice (Supplementary Fig. 4F–J).

The decreased levels of insulin in 5-week-old βTG mice were not associated with the changes in islet growth (Supplementary Fig. 4K–O). The reduced islet mass in 8-week-old line 214 mice probably reflected an inhibition of β-cell proliferation, as shown by the decrease in Ki-67–positive β-cells (Supplementary Fig. 4P). No between-group alteration of insulin secretion occurred at low-glucose; however, high-glucose no longer triggered more insulin secretion in line 214 islets (Fig. 1Q and R). The insulin contents were dramatically reduced in both male and female line 214 mice (Supplementary Figs. 1S and 4Q). Collective data demonstrated that overexpression of miR-203 in β-cells could cause growth retardation and diabetes due to deficiencies in GSIS function and insulin biosynthesis.

Although line 290 mice showed only slight glucose intolerance in the chow-diet condition, metabolic stresses, such as a HFD and aging, could significantly impair glucose tolerance and GSIS function but not insulin sensitivity in the line 290 mice (Supplementary Fig. 4A–J). Thus, miR-203 in β-cells could drive the development and progression of diabetes under metabolic stress.

RNA-seq identified 764 altered genes in line 290 mice and 1,292 altered genes in line 214 mice (Supplementary Fig. 6A and B). Downregulated genes in line 290 islets were linked to MODY, type 2 diabetes, and the insulin-signaling pathway, whereas upregulated genes were enriched in glycine, serine, and threonine metabolism, among others (Fig. 2A).

The downregulated genes in line 214 islets were also linked to MODYs (Supplementary Fig. 6C). The heat map showed significant decreases in genes associated with β-cell maturation (Ucn3) (23,24) and glucose sensing and increases in non–β-cell islet hormones and endocrine-cell and β-cell dedifferentiation markers (Fig. 2B). The alterations in gene expression were confirmed by qRT-PCR (Fig. 2C and D). The heat map also revealed decreased levels of oxidative phosphorylation and mitochondrial genes (Supplementary Fig. 6D).

We confirmed the reduced amount of MAFA protein in both line 290 and line 214 islets. Notably, neither mouse line showed alterations in islet PDX1 protein levels (Fig. 2E and F). These data revealed that forced expression of miR-203 in β-cells caused MODY-like diabetes due to inhibition of β-cell maturation.

As βTG mice developed diabetes when they were weaned to a carbohydrate-rich diet (Fig. 1M), we considered diet composition might affect β-cell function. The βTG mice and their littermates were subjected to weaning and suckling treatments (Fig. 3A). The body weights were greater in the weaning mice without genotype differences (Fig. 3B). Weaning caused dramatic hyperglycemia that was prevented by suckling in βTG mice (Fig. 3C).

Suckling-to-weaning transition influences postnatal β-cell differentiation and GSIS function by reprogramming β-cell–maturation genes (2), which intriguingly were suppressed in βTG islets (Fig. 2B–D). Weaning triggered dramatic increases in the expression levels of islet hormone genes and β-cell maturation genes in wild-type mouse islets. By contrast, β-cell maturation–associated genes were significantly repressed by miR-203 elevation but were increased by miR-203 deletion after weaning (Fig. 3D and E and Supplementary Fig. 7A).

Weaning forces cells to use glucose that is critical for postnatal β-cell differentiation and maturation (7). We therefore treated the mouse islets with high glucose to mimic β-cell functional maturation at the in vitro level. High-glucose treatment significantly enhanced the expression levels of β-cell maturation genes in wild-type islets, but not in βTG islets from both postnatal day 18 and mature stage (Supplementary Fig. 7B and Fig. 3F and G). The protein levels of islet hormones were also significantly enhanced after weaning in wild-type mice but not in βTG mice (Fig. 3H).

We used a lineage tracing technique to label β-cells with GFP in a RIP2-Cre–dependent manner (Supplementary Fig. 7C–E). Costaining of GFP-positive cells with MAFA or PDX1 confirmed that weaning triggered a much greater loss of MAFA signals in βTG islets (Fig. 3I and Supplementary Fig. 7F). Numbers of MAFA-positive cells were also reduced in βTG islets (Fig. 3J–K). These data demonstrated that miR-203 suppressed the weaning- and glucose-guided β-cell maturation.

scRNA-seq was used to explore the islet cell compositions. Overall, 4,540 islet cells from wild-type mice and 4,602 islet cells from βTG mice met our quality control metrics. After adjusting the k-means value, the following eight islet cell populations were clustered (Supplementary Fig. 8A): β1-cells (Ins) (Supplementary Fig. 8Bb), β2-cells (Ins) (Supplementary Fig. 8B, b), α/pancreatic polypeptid (PP) cells (Gcg [Supplementary Fig. 8Bc] and Ppy [Supplementary Fig. 8Be]), δ-cells (Sst) (Supplementary Fig. 8Bd), CgB^Hi cells (Chgb) (Supplementary Fig. 8Bf), macrophages, endothelial cells, and stromal cells. All the endocrine cell types showed high expression of CgA (Chga) (Supplementary Fig. 8Ba), an endocrine cell marker gene.

The heat map displayed markers that defined these eight cell types (Fig. 4A). Separate images of wild-type and βTG islets indicated the cell types and numbers (Fig. 4B). The CgB^Hi cells increased in βTG islets still expressed insulin genes (Fig. 4C), but not β-cell maturation gene Ucn3 (Supplementary Fig. 8Bg) or glucose sensor gene Slc2a2 (Supplementary Fig. 8Bh). The CDK4/CCND2 and BACE2/TMEM27 pathways are both involved in β-cell proliferation (25,26). Ccnd2 was uniformly expressed in endocrine cells, whereas Bace2 and Tmem27 were enriched in CgB^Hi cells (Supplementary Fig. 8Bj–l). Thus, the CgB^Hi cells are immature and proliferative β-cells.

Th β-cell ratio was significantly decreased, but the CgB^Hi cell ratio was increased to 40.81% in βTG islets (Fig. 4C). Ratios of α/PP cells, δ-cells, and stromal cells were also significantly increased, whereas ratios of macrophages and endothelial cells were decreased in βTG islets (Fig. 4D). Merging all the cell types together clearly demonstrated that βTG islet cells had significantly increased numbers of CgB^Hi cells (Fig. 4E). Together, miR-203 elevation led to dedifferentiation of immature β-cells to CgB^Hi endocrine cells.

We identified potential target genes of miR-203 by comparing the genes predicted by two databases with the downregulated genes of line 290 islets (Supplementary Fig. 9A). Zbtb20 was the only overlapping gene (Supplementary Fig. 9A), and contained three miR-203 responsive elements (MREs) (Fig. 5A). miR-203 regulated two of the three MREs (Fig. 5B). Overexpression of miR-203 suppressed the expression of ZBTB20 at both the mRNA and protein levels (Fig. 5C and Supplementary Fig. 9B), whereas miR-203 interference enhanced ZBTB20 amount (Fig. 5D). We also confirmed the regulatory role between miR-203 and ZBTB20 at both the mRNA and protein levels in βTG islets (Fig. 5D and E), confirming ZBTB20 as an authentic target of miR-203.

As Mafa was the most inhibited gene in βTG islets, we hypothesized that MAFA might be regulated by ZBTB20 and miR-203. The confirmed Mafa enhancer/promoter region by TFs (e.g., FOXA2) (Supplementary Fig. 9C) (27) contained five potential ZBTB20 binding sites. Using effective ZBTB20 antibody verified greater binding of ZBTB20 to the Mafa enhancer region (Supplementary Fig. 9D and E and Fig. 5F) (28). Moreover, overexpression of miR-203 significantly inhibited ZBTB20-enriched enhancers (Fig. 5F). Further ChIP-qPCR demonstrated a 60% decrease in ZBTB20 binding to this region by miR-203 elevation (Fig. 5G).

Moreover, adenovirus-based ZBTB20 overexpression significantly enhanced MAFA levels in both MIN6 cells and INS-1 cells (Fig. 5H and I). Furthermore, elevation of miR-203 reduced ZBTB20 and MAFA protein levels in MIN6 cells (Fig. 5J). miR-203 reduced MAFA protein amount and also decreased the recognition of its target gene Ins1 (Fig. 5K and L).

We also confirmed the regulation between miR-203 and MAFA in human islets (Fig. 5M), indicating an evolutionary conservation among species. These data revealed the presence of an miR-203/ZBTB20/MAFA regulatory axis in β-cells.

As Zbtb20 is an authentic target of miR-203, we tested whether the reconstitution of ZBTB20 in βTG islets would mitigate miR-203–caused impairments (Fig. 6A). Although the initial blood glucose levels differed, βTG mice infected with Ad-Zbtb20 showed significantly reduced blood glucose levels compared with those infected with Ad-GFP (Fig. 6B). Fasting blood glucose levels were also lower in Ad-Zbtb20–infected βTG mice due to increased serum insulin levels (Fig. 6C and D). The recovery of ZBTB20 expression also enhanced the expressions of MAFA and INSULIN in βTG β-cells (Fig. 6E and F).

The rescue effects of ZBTB20 on β-cell function were also tested at the ex vivo level (Fig. 6G). Whole-islet staining showed much more enhanced INSULIN signals in Ad-Zbtb20–infected βTG islets (Fig. 6H). The decreased GSIS function and insulin content in βTG islets were also recovered by overexpression of ZBTB20 due to the increased levels of gene expression associated with β-cell maturation and function (Fig. 6I–K).

Ad-MPN viruses (29) also significantly reduced the blood glucose levels of the miR-203–overexpressing mice by restoring insulin secretion and β-cell maturation (Supplementary Fig. 10). Taken together, our findings revealed that miR-203 caused β-cell dysfunction that was partially dependent on the ZBTB20/MAFA pathway.

Next, βKO mice were fed chow or a HFD, using RIP2-Cre and miR-203^fl/fl mice as controls (Fig. 7A). No between-group alterations were observed in body weight, glucose tolerance, or insulin sensitivity under the chow-diet condition (Supplementary Fig. 11A–C). However, βKO islets had the best ability to perform the first and second phases of insulin secretion, whereas the RIP2-Cre islets displayed the worst (Supplementary Fig. 11D), consistent with other literature showing mild β-cell dysfunction with RIP2-Cre transgene (30).

Placed on HFD feeding, the metabolic data between RIP2-Cre and miR-203^fl/fl mice revealed negligible changes in their body weights, glucose metabolism, and insulin secretion (data not shown). miR-203 KO in β-cells did not affect body weights in HFD-fed mice (Fig. 7B and C). Blood glucose levels were initially normal but gradually higher in the control mice than in βKO mice during 20-week HFD feeding (Fig. 7D). IPGTTs showed better glucose disposal by βKO mice (Fig. 7E). Interestingly, insulin tolerance also improved in βKO mice (Fig. 7F), probably due to the resistance to hyperinsulinemia caused by HFD feeding (Fig. 7G and H). The decreased insulin level in βKO mice had no concern with islet mass (Supplementary Fig. 11E–G). Interestingly, long-term HFD feeding always showed a disappearance of the first phase of insulin secretion (4), whereas miR-203–KO mice largely preserved this ability (Fig. 7I).

Gene expression levels of Zbtb20 and Mafa were significantly higher in βKO islets than in either RIP2-Cre or miR-203^fl/fl islets (Fig. 7J). The β-cell maturation genes also showed enhanced expression, whereas non–β-cell hormonal genes showed no alterations among the three groups (Fig. 7J and K). We confirmed a dramatic decrease of miR-203 in βKO islets (Fig. 7L). Thus, deletion of miR-203 in β-cells prevented HFD-induced metabolic abnormalities due to enhanced expressions of ZBTB20 and β-cell maturation genes.

Molecules that are positioned on the hub of transcriptional regulatory and metabolic networks on β-cell functional maturation are of substantial research interest. Numerous studies that have focused on genetic susceptibility genes, modifications of histone, DNA, and RNA, and maternal nutritional status have shed light on reliable mechanisms to explain postnatal β-cell maturation (31–35). In this study, we demonstrate that miR-203 may define the β-cell maturation status, as its decrease at weaning promotes the establishment of GSIS function, whereas its increase under metabolic stress conditions regresses GSIS function to the neonatal stage. Neonatal and stressed β-cells share an equal miR-203/ZBTB20/MAFA axis that compromises β-cell maturity (Graphical abstract). Importantly, the negative association between miR-203 and β-cell maturation is conserved in both rodents and humans, inevitably leading to impaired GSIS capacity and loss of β-cell identity.

We observed high transcription of miR-203 in both neonatal and diabetes conditions, prompting the hypothesis that miR-203 is associated with β-cell immaturity. Indeed, unable to be suppressed by weaning in βTG mice, the highly expressed miR-203 compromised GSIS capacities and insulin biosynthesis. On the contrary, βKO mice had better insulin secretion abilities and overcame the HFD-caused β-cell dysfunction. Our data clearly showed that a decrease in miR-203 level is indispensable for postnatal β-cell maturation and that metabolic stress adopted this neonatal pathway to cause β-cell immaturity. Our findings argue that miR-203 is a β-cell immaturity driver, rather than merely a marker, mirroring its known function in skin (36).

Considerable literature now points to common features between neonatal and metabolically stressed β-cells, such as lack of responsiveness to glucose changes, alterations in epigenetic markers, and lower expressions of β-cell maturity genes (37,38), all of which favor anaerobic glycolysis for ATP generation. This ineffective energy generation fails to close K_ATP channels that are required for opening L-type voltage-dependent calcium channels, eliciting Ca²⁺ inflow, and thus promoting insulin granule exocytosis (39).

Genes related to the above processes are all important for fulfilling the GSIS function. Reports have shown that mutant K_ATP channels maintain an open state to suppress insulin secretion, leading to neonatal diabetes (40). Succinate dehydrogenase deficiency compromises GSIS function and β-cell growth, thereby initiating the onset of pubertal diabetes (41). These two abnormalities manifest a diabetes phenotype correlated with GSIS defects; however, they are uncoupled by the weaning action, in contrast to our findings that elevation of miR-203 causing mouse diabetes after weaning. Our bulk RNA-seq results largely supported the postweaning diabetes phenotype by verified gene alterations related to glucose uptake (Glut2), glycolysis/gluconeogenesis (Ldha, Pfkl, Pcx, Pck1), glucose oxidative phosphorylation (Cox6a2, Atp5k), and insulin granule biosynthesis (Ins, Pcsk1, and Sytl4), but not those related to K_ATP channels (Abcc8, Kcnj11), or L-type voltage-dependent calcium channels (Cacna1c, Cacna1d), hinting a correlation exists between miR-203 and the nutritional transitions during weaning and diabetes.

The nature of miR-203 responsiveness to nutritional shifts is worth further investigation (16,17). Currently, miR-203 is reported to be positively regulated by C/EBPβ in adipocytes, repressed by β-catenin in human melanoma, and epigenetically silenced by DNMT3B in neural crest cells (42–44). miR-203 expression in islets ranked second among regular tissues, implying β-cell–specific TFs might contribute to miR-203 regulation. Although β-cell–specific TFs contained cis-acting elements on the miR-203 promoter, only MAFA was verified negatively regulating miR-203. Interestingly, MAFA is the most important TF that responds to glucose and functions in weaning-triggered β-cell maturation, opposite to the effects of miR-203. Recovery of the MAFA level protected βTG mice from diabetes phenotypes by rescuing insulin biosynthesis, supporting a role for a MAFA–miR-203 negative feedback loop in β-cell maturation.

We identified Zbtb20 as a miR-203 target gene during β-cell maturation. Zinc finger and BTB domain-containing protein 20 (ZBTB20) plays important roles in the pancreas, liver, and brain (45–47). Global Zbtb20-KO mice show growth retardation, and hypoinsulinemia (48). β-Cell ablation of Zbtb20 causes glucose intolerance due to insulin insufficiency (49), akin to the phenotypes of our miR-203–overexpressing mice. ZBTB20 is considered to function as a physiologic facilitator of GSIS in β-cells through transcriptional repression of Fbp1 (49). As a gluconeogenic enzyme, Fbp1 is critical for glucose metabolism and insulin secretion in β-cells (50). We also observed an increase of Fbp1 in βTG islets, confirming the existence in the miR-203/ZBTB20 axis. Moreover, we verified Mafa as a new target of ZBTB20, which explains the transcriptional repression of Mafa in βTG islets.

The GSIS dysfunctional β-cells in βTG islets were further dedifferentiated into CgB^Hi endocrine cells, as suggested by our scRNA-seq data. These CgB^Hi cells retained low expression levels of insulin, glucose-sensing, and metabolism genes, while having comparable levels of pan-endocrine genes (CgA, Nkx6-1, and Neurod1). Although the gene profiles of dedifferentiated β-cells somewhat resembled those of neonatal β-cells, they acquired less proliferative ability than neonatal β-cells, judging by the reduced islet mass in 8-week-old βTG mice. The CgB^Hi cells expressed intact cell-cycle progression genes (Ccnd2, Cdk4) and growth hormonal receptor genes (Insr, Ghr, Egfr), which were important for β-cell proliferation; thus, the antiproliferative mechanism is still undetermined.

In conclusion, our findings provide comprehensive evidence that miR-203 is an immature driver of β-cells. The inability to suppress miR-203 after weaning restrains the β-cell maturation process via the miR-203/ZBTB20/MAFA pathway. Metabolically stressed β-cells adopt this pathway, thereby compromising β-cell maturity and identity. Ablation of miR-203 protects β-cells from HFD-induced GSIS defects; therefore, therapeutic approaches aimed at blocking miR-203 effects may be valuable tools for boosting functional β-cell numbers in diabetes.

Acknowledgments. The authors thank Prof. Wei-ping Zhang (Department of Pathophysiology, Second Military Medical University, Shanghai, China) for providing the ZBTB20-overexpression adenoviruses (Ad-Zbtb20), antibodies against ZBTB20, and siRNA sequences specific for mouse Zbtb20 gene. The authors also thank Prof. Zhang and his team for their valuable discussion and comments for the study.

Funding. This study was supported by research grants from the National Natural Science Foundation of China (82330027, 82070843, 82270844, and 82200919). X.H. and Y.X. are fellows at the Collaborative Innovation Center for Cardiovascular Disease Translational Medicine.

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

Author Contributions. Y.L., Y.Y., Y.S., L.H., L.Z., and H.S. performed all of the experiments and analyzed the data. Y.L., Y.Y., Y.S., and Y.Z. contributed to the data discussion. X.C. contributed to the confocal images. R.L. and S.W. performed human islet experiments and analyzed the data. S.W. and X.H. edited the manuscript. S.W., X.H., and Y.Z. designed the project, supervised research, and coordinated the execution of the experimental plan. Y.Z. wrote the manuscript. All authors reviewed and commented on the manuscript. X.H. and Y.Z. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.