Glucose-Dependent miR-125b Is a Negative Regulator of β-Cell Function

Pancreatic β cells are essential regulators of glucose homeostasis, secreting insulin in response to increases in circulating levels of the sugar (1). Nevertheless, chronic hyperglycemia has a negative effect on β-cell function and survival, contributing to the development of the type 2 diabetes (T2D) (2).

miRNAs are noncoding RNAs, ∼21 nucleotides, that silence gene expression posttranscriptionally and are essential for endocrine cell development and function (3). Even though β cells contain hundreds of different miRNAs, the function of only a few has been studied in detail.

MiR-125b (miR-125b-5p) is a highly conserved miRNA widely studied in the context of tumorigenesis as an important regulator of cellular differentiation and apoptosis (4). We have previously shown that miR-125b expression increases in islets after β cell–specific deletion of AMP-activated protein kinase (βAMPKdKO) (5), suggesting that AMPK may act as a negative regulator of miR-125b expression. AMPK activity is suppressed acutely in β cells by high glucose levels and lowered in islets from individuals with T2D (6).

These observations led us to hypothesize that inhibition of AMPK in response to high glucose levels would result in islet miR-125b upregulation, which may mediate some of the deleterious effects of hyperglycemia in β cells. Here, we show that high glucose levels increase islet miR-125b expression and that miR-125b has an important negative effect in β-cell function, in vitro and in vivo, by targeting genes involved in regulating lysosomal and mitochondrial function.

MIN6 and EndoCβ-H1 (7) cells and mouse and human islets were obtained, maintained, and then transfected with Lipofectamine 2000 (ThermoFisher Scientific), as described in Supplementary Research Design and Methods.

EndoCβ-H1 cells were infected with lentiviral vectors expressing two guide RNAs targeting MIR125B-2 and a rat insulin promoter (RIP)-driven hSpCas9. Lentivirus without guide RNAs was used as a control. Integrating EndoCβ-H1 cells were selected with puromycin. See Supplementary Research Design and Methods for more details.

The cassette expressing miR-125b under the control of a reverse tetracycline transactivator (rtTA)-inducible promoter was excised from pBI-LTet-MIR125B and used for pronuclear microinjection of C57BL6/J oocytes at the Centre for Transgenic Models Basel, Switzerland). A founder with a unique copy of the transgene was bred with mice expressing the rtTA under the control of the RIP7 promoter (RIP7-rtTA^+/+). Mice bearing the transgene (MIR125B-Tg: RIP7-rtTA^+/−, MIR125B Tg^+/−) and control (RIP7-rtTA^+/−) were used in subsequent experiments. Doxycycline (0.5 g/L) was continuously administered in the drinking water from the time of mating. Mice had free access to standard chow or, when indicated, ketogenic diet (Ssniff). In vivo procedures were approved by the UK Home Office Animal Scientific Procedures Act, 1986 (Licenses PA03F7F0F and PP7151519).

Quantitative reverse transcription PCR was performed as previously described (5). For miRNA quantification, let-7d-3p (Figs. 1 and 4) and miR-574-3p (Supplementary Figs. 1 and 2) were used as endogenous controls because of their stable expression in our system (5). For mitochondria DNA to nuclear DNA ratio analysis, islet DNA was extracted with 40 μg/mL proteinase K in SNET buffer.

miRNA induced silencing complex (miRISC) immunoprecipitation was performed as previously described (8) with a mouse–anti–AGO2 antibody (clone E12–1C9; Abnova). Library preparation from mRNA enriched from total RNA with a NEBNext Poly(A) mRNA Magnetic Isolation Kit (New England Biolabs) and from miRISC-isolated RNA was performed using a NEBNext Ultra II Directional RNA Library Prep Kit from Illumina. Library preparation from human islets total RNA (200 ng) was performed using a NEBNext Low Input RNA Library Prep kit (New England Biolabs). Libraries were sequenced on a HiSeq4000. Mapping and differential expression analysis were performed with Salmon v1.3.0 (9) and DESeq2 (10).

For miRNA target identification, total mRNA (T-RNA) and immunoprecipitated RNA (RIP-RNA) samples were treated as separate data sets, and the two resulting gene lists of differential analysis were used to calculate the ratio of RIP-RNA to T-RNA for each gene. Additional details are provided in Supplementary Research Design and Methods.

Western immunoblotting was performed with EndoCβ-H1 total protein (5–15 μg) or and extract of 10–20 islets prepared with radioimmunoprecipitation assay. Islets and slides from isolated pancreata for immunohistochemistry were prepared, visualized, and quantified as previously described (11). Additional details on the approaches used for blotting, β-cell mass, proliferation, apoptosis, and cathepsin D/LAMP1 measurements, and an exhaustive list of all antibodies and kits used can be found in Supplementary Research Design and Methods.

MIN6 and EndoCβ-H1 cells and whole isolated islets were fixed and prepared as described by Jones et al. (12). Ultrathin 70 nm sections were examined in Tecnai Spirit or JEOL 1400plus transmission electron microscopes. Images at ×2,000 magnification were analyzed and quantified blindly in ImageJ, as described by Pasquier et al. (13).

Mice fasted overnight (for insulin secretion measurement) or for 5 hours (for insulin tolerance tests) were submitted to intraperitoneal glucose tolerance tests or insulin tolerance tests after administration of 1 g/kg body weight of glucose or 1.2 U/kg insulin (NovoRapid), respectively, as previously described (14). To measure insulin secretion, mice were injected with 3 g/kg glucose and blood insulin levels were measured using an Insulin ELISA kit (Crystal Chem), following manufacturer’s instructions.

Cell lines and islets were submitted to glucose- or KCl-stimulated insulin secretion as previously described (7, 14). Secreted and total insulin levels were quantified using an HTRF insulin kit (Cisbio) in a PHERAstar reader (BMG Labtech), following the manufacturer's guidelines. Insulin and protein contents were measured by HTRF and bicinchoninic acid (Pierce), respectively, in radioimmunoprecipitation assay cellular lysates or in full pancreas homogenates, as previously described (11).

Ca²⁺ and ATP imaging was performed in whole islets with Cal-520 (4.5 μM; Stratech) and adenovirus-encoded Perceval probe, respectively, and total internal reflection of fluorescence was performed in dissociated islets infected with an adenovirus construct for NPY-Venus, as described by Martinez-Sanchez et al. (14) and Georgiadou et al. (15). Mitochondria morphology analysis was performed with Mitotracker Green (ThermoFisher) as described by Georgiadou et al. (15). See Supplementary Research Design and Methods for additional details.

Voltage-clamp electrophysiology was performed in dissociated β cells using patch electrodes and generating voltage-dependent Ca²⁺-channel (VDCC) currents by application of sequential 10-mV depolarizing steps as detailed in Supplementary Research Design and Methods.

Oxygen consumption rate was determined using the XFe96 Extracellular Flux Analyzer and a XFe96 FluxPak (Seahorse Bioscience) in groups of 3–6 size-matched islets per well embedded in Matrigel and incubated at 3 mmol/L glucose followed by 17 mmol/L glucose and oligomycin (5 μmol/L), per manufacturer’s instructions.

Statistical significance was evaluated with GraphPad Prism 9.0 software, as indicated in the figure legends. Correlation data were estimated using R software with the ppcor package (16). All data are shown as mean ± SEM. P < 0.05 was considered statistically significant unless otherwise specified.

All mouse in vivo procedures were conducted in accordance with the UK Home Office Animal (Scientific Procedures) Act of 1986 (project license PA03F7F0F to IL) and approved by the Imperial College Animal Welfare and Ethical Review Body. For human islet experiments, islets were isolated from the locations indicated in Supplementary Table 9, with full approval and informed consent.

All data generated or analyzed during this study are included in the published article and its online supplementary files. No applicable resources were generated or analyzed during this study.

To determine whether glucose regulates miR-125b expression in islets, we measured miR-125b in mouse and human islets cultured at increasing glucose concentrations. MiR-125b expression was significantly increased after culture at high (11–25 mM) versus low (3.5–5.5 mmol/L) glucose concentrations (Fig. 1A, Supplementary Fig. 1A). Consistent with the possible in vivo relevance of these findings, expression of miR-125b in human islets was positively correlated with the BMI of donors but not with their age or sex (Fig. 1B, Supplementary Fig. 1B and C).

Islets from βAMPKdKO mice had significantly higher miR-125b levels than did control islets when cultured at a low (5.5 mM) glucose concentration (Fig. 1C, Supplementary Fig. 1D), whereas miR-125b expression remained unchanged in islets cultured at high glucose concentration (25 mM). To determine whether this glucose/AMPK–dependent regulation of miR-125b also occurred in vivo, we fed control, βAMPKdKO or βLKB1KO [i.e., mice with a β cell–specific deletion of the AMPK-upstream kinase liver kinase B1, and LKB1/STK11 (17)] animals a ketogenic (low-sugar) diet. As previously shown (5), βAMPKdKO islets had higher levels of miR-125b than did controls and the same was observed in βLKB1KO islets (Fig. 1D). Feeding a low-sugar diet resulted in a significant decrease in islet miR-125b in control animals but not in islets from βAMPKdKO and βLKB1KO mice (Fig. 1D). Moreover, culture of human islets with the selective AMPK activators ¹³C and 991 (5) or with adenovirus encoding a dominant-negative form of the enzyme caused a small but significant reduction and increase, respectively, in miR-125b expression (Fig. 1E, Supplementary Fig. 1E). Taken together, these results indicate that high glucose concentration stimulates miR-125b expression in both mouse and human islets by inhibiting AMPK activity.

The most important characteristic of the β cell is its capacity to respond to high levels of circulating glucose by secreting insulin (1). Glucose-stimulated insulin secretion (GSIS) in MIN6 cells with ∼40-fold increased miR-125b (Supplementary Fig. 2A) after transient transfection with miR-125b mimics was lower than in control cells, and miR-125b overexpressing cells contained 25% less insulin (Fig. 1F, Supplementary Fig. 2B). Conversely, inhibition of miR-125b function with an adenovirus encoding for an miR-125b inhibitor resulted in increased GSIS and insulin content (Fig. 1G, Supplementary Fig. 2C).

To explore whether miR-125b regulated insulin secretion in a human β-cell line, we next generated human EndoCβ-H1 cells (18) with miR-125b loss of function using CRISPR-Cas9 (EndoCβ-H1-MIR125B2-KO; Fig Supplementary Fig. 3A). EndoCβ-H1-MIR125B2-KO cells contained ∼80% less mature miR-125b than did controls (Supplementary Fig. 3B and C), showed similar viability (Supplementary Fig. 3D) and, as anticipated, secreted more insulin in response to high glucose levels than did controls, although only a small, nonsignificant increase was detected in insulin content (Fig. 1H, Supplementary Fig. 2D).

To further explore the genes and molecular pathways regulated by miR-125b, as well as its mechanism of action, we sought to explore the impact of the miRNA on β-cell gene expression and to identify direct targets in a high-throughput manner. miRNAs guide the miRISC, including an Argonaute protein, to the target mRNAs, which results in mRNA destabilization and/or inhibition of translation (19). Thus, we used an experimental approach that combined overexpression of miR-125b in MIN6 cells, immunoprecipitation of miRISC and target mRNAs using an antibody against AGO2 and high-throughput sequencing and differential analysis of cellular total and immunoprecipitated RNAs (20) (Supplementary Fig. 4A). As a consequence of miR-125b overexpression, we expected miR-125b targets to be reduced or unchanged (if repression occurs exclusively at the level of translation) in the T-RNA fraction (Supplementary Table 1) and increased in the RIP-RNA fraction (Supplementary Table 1). Thus, for miR-125b direct targets, we expect a RIP-RNA to T-RNA ratio >1 after miR-125b overexpression. To validate our approach we sorted all genes according to RIP-RNA to T-RNA ratio (Supplementary Table 1) and performed unbiased motifs enrichment analysis using cWords (21). As anticipated, cWords found a significant enrichment of miR-125b seed-matching sequences in both the 3′ nontranslated (3′UTR) and coding sequences (CDSs) of highly ranked (i.e., high RIP-RNA to T-RNA ratio) genes (Supplementary Fig. 4B), suggesting that miR-125b–mRNA target interactions occur through both these regions. A total of 180 mRNAs were enriched in miRISC after miR-125b overexpression, with an RIP-RNA to T-RNA ratio >1.5 (Fig. 2A, Table 1). Of these, 60 were significantly downregulated at the mRNA level after miR-125b overexpression (Table 1, adjusted P value [Padj] < 0.1) and 87 contained predicted binding sites in their 3′UTRs, according to TargetScan, ∼50% (n = 49) of which are conserved in humans (Table 1).

We further validated the six targets (Taz, M6pr, Tor2a, Mtfp1, Tnks1bp1, and Gnpat) with the highest RIP-RNA to T-RNA ratios using luciferase reporter–based assays (20). As expected, cotransfection of miR-125b mimics reduced Firefly/Renilla activity ratios in cells transfected with constructs containing the Taz, M6pr, Tor2a, and Mtfp1 3′UTRs, Gnpat CDS, or three perfectly complementary sequences to miR-125b (positive control), downstream of the Firefly open reading frame in the bicistronic plasmid pMiRGlo, demonstrating that miR-125b silences gene expression through those sequences (Fig. 2B). In contrast, miR-125b mimics did not affect Firefly/Renilla activity of the empty vector or in the presence of the 3′UTR of Gnpat, which lacked predicted miR-125b binding sites. We failed to validate miR-125b action through Tnks1bp1 CDS, perhaps because the location of the binding site downstream of the luciferase CDS and not within the CDS itself (22).

To identify biological pathways and functions regulated by miR-125b, we submitted the list of genes dysregulated at the RNA level (T-RNA, n = 306 and 317 genes down- and upregulated, respectively; Padj < 0.1; Supplementary Table 1) as well as those identified as miR-125b direct targets with a RIP-RNA to T-RNA ratio > 1.5 (Table 1) to the Database for Annotation, Visualization and Integrated Discovery (23). Functional annotations with high enrichment scores included mitochondrion, tricarboxylate cycle, sterol biosynthesis, protein transport, Golgi apparatus, and endosome (Fig. 2C, Supplementary Table 2), all closely associated with the capacity of the β cells to produce and secrete insulin in response to glucose.

The gene at the top of our miR-125b targets list, the lysosomal- and Golgi-associated M6pr (manose-6-phosphate receptor–cation dependent) controls trafficking of lysosomal hydrolases from the Golgi via endosomes (24) and thus may influence protein turnover and lysosomal function. Luciferase assays further confirmed the interaction of miR-125b via a predicted, conserved miR-125b binding site in its 3′UTR (Fig. 3A), and Western immunoblotting of EndoCβ-H1 cells overexpressing miR-125b and EndoCβ-H1-MIR125B2-KO cells showed a strong down- and upregulation in M6PR protein, respectively (Fig. 3B), without significant changes at the mRNA level (Fig. 3C, Supplementary Table 3). These results confirm that miR-125b targets M6PR to repress protein production in human EndoCβ-H1 cells.

We hypothesized that M6PR downregulation after miR-125b overexpression may lead to a defect in lysosomal hydrolases and an accumulation of lysosomal structures with nondigested cargos. Accordingly, electron microscopy revealed an accumulation of aberrant, enlarged lysosomal structures in MIN6 cells overexpressing miR-125b (Fig. 3D).

Gene ontology (GO) analysis of genes dysregulated upon miR-125b overexpression also revealed a significant overrepresentation of genes involved in mitochondrial function (Supplementary Table 2, Fig. 2C), including one of the top miR-125b direct targets, mitochondrial fission process 1 (Mtfp1, also known as MTP18). Luciferase experiments further confirmed the interaction of miR-125b via a predicted conserved binding site in Mtfp1 3′UTR (Fig. 3A). MTFP1 has been implicated in mitochondrial fission and apoptosis in mammalian cells (25). To explore a possible role for miR-125b in controlling mitochondrial morphology, we stained MIN6 cells transfected with control or miR-125b mimics with a mitochondria-targeted green fluorescent probe (Mitotracker green). MIN6 cells overexpressing miR-125b contained the same number of mitochondria and overall mitochondrial area, though their mitochondria were slightly more elongated and less circular (Fig. 3F, Supplementary Fig. 5A). In contrast, EndoCβ-H1-MIR125B2-KO cells revealed a marked reduction in mitochondrial area and the number of mitochondria with a smaller perimeter, which were also less elongated and more circular (Fig. 3F, Supplementary Fig. 5B). Further emphasizing the importance of miR-125b for mitochondrial homeostasis, GO analysis following RNA sequencing (RNA-seq) of EndoCβ-H1-MIR125B2-KO confirmed a strong enrichment in dysregulated genes associated with mitochondrial function (Supplementary Fig. 5C, Supplementary Table 4), including MTFP1 (1.5-fold, P = 0.006, Padj < 0.1). These findings suggest that relief from miR125-mediated repression leads to disruption of the tubule-vesicular network and to profound changes in the expression of genes involved in mitochondrial homeostasis, which may underlie the observed improvement in insulin secretion.

Given the regulatory effect of miR-125b in vitro, we decided to study the role of miR-125b in vivo by generating a mouse with β cell–specific, doxycycline-inducible overexpression of miR-125b using a RIP7-rtTA promoter (14). MIR125B-Tg (MIR125B-Tg: RIP7-rtTA^+/−, MIR125B Tg^+/−) animals contained ∼60-fold more islet miR-125b than did controls (RIP7-rtTA^+/−) (Fig. 4A), whereas statistically significant changes were not detected in miR-125b levels in the hypothalamus, a potential site for activity of the RIP promoter (26) (Supplementary Fig. 6A).

MIR125B-Tg mice had indistinguishable weight gain compared with control mice (Supplementary Fig. 6B), whereas random-fed glycemia incidence was significantly higher in both male and female MIR-125B-Tg mice (Fig. 4B). Consistently, MIR125B-Tg mice were highly glucose intolerant (Fig. 4C and D), though fasting glycemia was unaffected. No changes in glucose tolerance were observed in animals bearing the transgene in the absence of rtTA (-rtTA control: RIP7-rtTA^−/−, MIR125B OE^+/−; Supplementary Fig. 6C), demonstrating that the defects observed were not due to an off-target genomic event resulting from transgene integration. Blood glucose levels were efficiently reduced by administration of exogenous insulin, excluding insulin insensitivity as a major player in the glucose intolerance observed in MIR125B-Tg animals (Supplementary Fig. 6D). Moreover, plasma insulin did not increase in MIR125B-Tg mice in response to an intraperitoneal injection of glucose (Fig. 4E), suggesting impaired β-cell secretory function. β-cell mass and β- to α-cell ratio, as well as proliferation and apoptosis, remained unchanged in the transgenic animals (Fig. 4F, Supplementary Fig. 6E and F). Taken together, these data suggest that increased miR-125b in the β-cell results in hyperglycemia and glucose intolerance due to impaired β-cell secretory function.

To further explore a cell autonomous defect in insulin secretion, we measured GSIS in isolated islets from MIR125B-Tg mice, which showed a strong reduction in insulin secretion in response to glucose or depolarization by KCl (Fig. 5A, Supplementary Fig. 6G). Consistent with our previous data from MIN6 cells, MIR125B-Tg islets contained less intracellular insulin (Supplementary Fig. 6H), as further confirmed by Western blot (Fig. 5B, Supplementary Fig. 6I), revealing lower levels of both intracellular proinsulin and insulin. Accordingly, MiR125B-Tg pancreata contained less insulin (Supplementary Fig. 6J). These data indicate a reduced capacity of MIR125B-Tg β cells to produce and secrete insulin.

Upon entry in the β cell, glucose is rapidly metabolized by mitochondria, causing a sharp increase in intracellular ATP to ADP ratio, closure of ATP-sensitive K⁺ channels, plasma membrane depolarization, and Ca²⁺ influx into the cytosol to trigger insulin exocytosis (1). That secretory responses to both high glucose and KCl levels were impaired in MIR125B-Tg islets suggested defects downstream of glucose metabolism. Accordingly, the glucose-mediated ATP to ADP ratio increase and the basal ATP to ADP ratio were similar in MIR125B-Tg and control isolated islets (Fig. 5C and D). Oxygen consumption rate increases induced by high glucose levels and in response to the inhibitor of mitochondrial ATP synthase oligomycin were also comparable (Fig. 5E), confirming similar mitochondrial capacity to generate ATP. In contrast, both basal and oligomycin-insensitive respiration were significantly elevated in MIR125B-Tg islets (Fig. 5E), suggesting increased inner mitochondrial membrane proton conductance. Additionally, MIR125B-Tg islets had a strong reduction in glucose-induced changes in intracellular Ca²⁺ (Fig. 5F). KCl-induced increases in Ca²⁺ were similar between MIR-125B-Tg and control islets (Fig. 5F), suggesting that MIR125B-Tg β cells contained functional voltage-gated calcium channels. This was further confirmed in single β cells by whole-cell voltage clamp that showed similar VDCC currents (Fig. 5G). The kinetics of individual fusion events were also identical in cells from MIR125B-Tg or control islets (Fig. 5H), as assessed by total internal reflection of fluorescence imaging of neuropeptide Y (NPY)-Venus–expressing vesicles near the plasma membrane (27), indicating that vesicle fusion itself was not affected by miR-125b overexpression.

Our previous experiments in cell lines pointed to miR-125b as a regulator of mitochondrial and lysosomal function, which could contribute to defective insulin secretion. In contrast to our observations in cell lines, but consistent with an efficient capacity to generate ATP in response to glucose, mitochondrial area and morphology and the levels of mitochondrial DNA (Fig. 6B) were comparable in MIR125B-Tg and control islets (Fig. 6A and B). Even though lysotracker staining revealed no differences in lysosomal area or numbers (Supplementary Fig. 6K), transmission electron microscopy (TEM) identified the presence of enlarged lysosomes and autophagosomes in the MIR125B-Tg islets, which were rarely found in control samples (Fig. 6C and D). Moreover, islet coimmunostaining identified a significant reduction in the lysosomal protease cathepsin D localized to (LAMP1-positive) lysosomes (Fig. 6F). TEM also revealed a sharp reduction in the number of dense-core insulin granules and an increase in granules with defective crystallization (light-core, abnormal rod-like, or empty) (Fig. 6C and E). Importantly, fewer secretory granules were present within 200 nm of the plasma membrane in MIR125B-Tg than in control islets (Fig. 6E).

To shed more light on the molecular mechanisms underlying the secretory defects of miR-125B-Tg islets, we performed RNA-seq on MIR125B-Tg and control islets. We found 320 and 398 genes were significantly (Padj < 0.1) down- and upregulated, respectively (Supplementary Table 5).

GO analysis of downregulated genes identified a strong enrichment within pathways and processes associated with mature β-cell function, such as regulation of insulin secretion and endocrine pancreas development (Supplementary Table 6, Fig. 7A and B, Supplementary Fig. 7A), with a significant reduction of several important β-cell identity genes, such as Ucn3, Pdx1, NeuroD1, Nkx2.2, Slc2a2, and Nkx6.1. These genes either do not contain predicted miR-125b binding sites or had been identified as direct miR-125b targets in our RNA-immunoprecipitation sequencing (RIP-seq) experiments, suggesting an indirect effect. Further supporting a loss of β-cell identity, upregulated genes were significantly associated with neuronal features (Supplementary Table 6, Fig. 7A,Supplementary Fig. 7B). We also observed an upregulation of many glycoprotein genes and of genes associated with cellular and focal adhesion, signaling pathways and voltage-gated and potassium channels. Golgi-associated genes were both up- and downregulated (Supplementary Table 6, Fig. 7A, Supplementary Fig. 7B).

To further explore the relevance of miR-125b in human β cells, we performed GSIS assays in human islets infected with Ad-MIR125B, achieving a ∼4.5-fold increase in miR-125b levels (Supplementary Fig. 8A). MiR-125b–overexpressing islets secreted ∼50% less insulin than did controls, though their insulin content remained unchanged (Fig. 7C, Supplementary Fig. 8B). Additionally, we performed RNA-seq in dissociated human islets infected with these viruses, achieving an approximately threefold increase in miR-125b (Supplementary Fig. 8C). Principal component analysis showed that most of the variation was due to the donor origin of the islets (Supplementary Fig. 8D); thus, this experiment was underpowered (∼20% to detect a twofold change, with P < 0.01, calculated using the Scotty web tool (28)). Nevertheless, the expression of hundreds of genes tended to be altered (P < 0.05; Supplementary Table 7) and gene set enrichment analysis of all the genes ranked by gene expression fold change identified a significant enrichment in several biological pathways (Supplementary Fig. 8E and F, Supplementary Table 8) notably including lysosomal (P < 0.0001; false discovery rate = 0.07) and calcium signaling (P < 0.0001; false discovery rate = 0.06). Further suggesting a role for miR-125b in lysosomal function, a downregulation in M6PR protein was observed in dissociated human islets infected with Ad-MIR125B (Fig. 7D).

Even though elevated levels of circulating miR-125b in association with higher HbA_1c in type 1 diabetes (29) and T2D (30) have been previously identified, this is the first study, to our knowledge, to demonstrate that miR-125b expression is induced by glucose in mouse and human islets in an AMPK-dependent manner. Although our capacity to demonstrate a correlation between hyperglycemia and islet miR-125b levels in human islets in vivo was limited by the lack of HbA_1c data, we found a significantly positive correlation with donor BMI, which often is associated with HbA_1c (31).

Despite being highly expressed in β cells, the function and mechanism of action of miR-125b in these cells have remained elusive until now. Here, we show that elevated expression of this miRNA in β cells impairs their secretory function in vitro and in vivo.

We have generated a transgenic model capable of β-cell selective overexpression of miR-125b. These animals were hyperglycemic and strongly glucose intolerant and presented a drastic reduction in circulating insulin after a glucose challenge.

Our findings suggest two main causes for the secretory defects in the MIR125b-Tg mice: strong impairment in GSIS and reduced insulin content. Unlike ATP, levels of cytosolic Ca²⁺ increased after glucose stimulation were substantially lower in MIR125b-Tg islets, though, paradoxically, similar levels of Ca²⁺ were observed upon membrane depolarization with KCl, even though KCl-stimulated insulin secretion was also strongly impaired. This points toward defects at different levels on the secretory pathway, as supported by the striking changes in gene expression. First, ATP-sensitive K⁺–independent amplification pathways (32) might be impaired in MIR125B-Tg islets, which display lower levels of genes encoding enzymes limiting the production of guanine nucleotides and glutamine, such as Imphd1 and GLUL, that can promote insulin exocytosis in an ATP- and ADP-independent manner (33, 34). Second, voltage-clamp electrophysiology experiments confirmed the presence of functional VDCC in MIR125B-Tg β cells. MIR125B-Tg islets expressed normal levels of Ca²⁺ channels but higher levels of PMCA3 and the Na⁺/Ca²⁺-exchanger NCX3, members of protein families responsible for Ca²⁺ extrusion in β cells, which could potentially contribute to reduced levels of cytosolic calcium. Third, TEM revealed fewer granules in close proximity to the plasma membrane in MIR125B-Tg β cells, and RNA-seq identified a strong alteration in genes involved in cellular granule docking and fusion, such as granuphilin (Sytl4), the SNARE protein Vamp3, and synaptotagmins (Syt1, Syt12, and Syt17) (35, 36), pointing toward defects in the final steps of exocytosis. Finally, MIR125B-Tg islets present defective granule crystallization, which might be partially due to reduced levels of the zinc transporter Znt8 mRNA (37). Loss of β-cell identity might further contribute to these secretory defects, though some of these transcriptional changes may occur as a consequence of the moderate fed hyperglycemia in these animals (38) or be residual from impaired maturation, because RIP7-rtTA–driven transgene expression is expected to occur in 11.5-day-old transgenic embryos.

Here, we unbiasedly identified dozens of miR-125b targets through the integration of RIP-seq and RNA-seq data (8,20) of miR-125b–overexpressing MIN6 cells. Given the supraphysiological levels of miR-125b achieved in these experiments, additional loss-of-function experiments will be key to confirm the targeting. It remains to be studied whether some of these targets are shared with the family member miR-125a, also expressed in β cells. Additionally, some of these targets might not be conserved in humans, and this may explain the fact that miR-125b modulation affects insulin content in mouse β cells but less so in humans. Intriguingly, disheveled binding antagonist of β catenin 1 (Dact1) was not in this list and remained unchanged upon miR-125 manipulation in all our experimental models. Dact1 was previously identified as an miR-125b target in β cells, perhaps because of the presence of other pancreatic, highly proliferative cells in the samples (39).

One of the genes at the top of our list of miR-125b targets, M6pr, encodes the cation-dependent manose-6-phosphate protein MP6R (CD-M6PR). Even though the function of CD-M6PR remains unknown in β cells, this receptor is necessary for adequate lysosomal targeting of several hydrolases, such as cathepsin D (40). Remarkably, MIR125B-Tg islets contained less cathepsin D in their lysosomes, which appeared enlarged similarly to those observed by Masini et al. (41) in T2D islet micrographs, suggesting the accumulation of undigested cargos. Defects on the lysosomal or trafficking machinery can also affect the maturation, packaging, and location of the secretory granules within the cell (42) and might contribute to the reduced insulin content and crystallization defects observed in the MIR125B-Tg mice. Whether, and how, elevated miR-125b during hyperglycemia or loss of AMPK activity contributes to these processes in T2D remains to be studied.

Even though we had identified several genes encoding mitochondrial proteins as miR-125b targets in MIN6 cells, no mitochondrial morphological defects were detected in the transgenic islets, suggesting that endogenous miR125b levels are sufficient to exert maximal effects on mitochondrial structure. However, several mitochondrial genes were dysregulated in MIR125B-Tg islets and seahorse experiments showed a significant increase in basal and oligomycin-insensitive respiration, suggesting a certain extent of mitochondrial dysfunction.

On the other hand, deletion of miR-125b in EndoCβ-H1 cells resulted in a strong alteration of mitochondrial morphology and in the expression of mitochondrial proteins, including the miR-125b targets Mtfp1 and Gnpat. Mtfp1 is targeted by miR-125b in monocytes to promote elongation of the mitochondrial network (43). GNPAT stabilizes DRP1 (44), which is required for mitochondrial fission and deletion of which impaired insulin secretion (45, 46). In line with these earlier studies, EndoCβ-H1-MIR125B2-KO cells contained shorter mitochondria, which correlated with increased GSIS. It has been contested, however, whether mitochondrial fragmentation promotes apoptosis or correlates with hyperglycemia or diabetes (15, 47). Importantly, we did not observe any differences in cellular death or number in EndoCβ-H1-MIR125B2-KO or in Ddit3 (CHOP) mRNA levels, suggesting an absence of mitochondrial-induced endoplasmic reticulum stress in these cells. Experiments are required to determine whether the positive effects of miR-125b elimination in β-cell function persist under conditions of cellular stress.

In summary, our results suggest that β-cell miR-125b has the potential to act as a glucose-regulated metabolic switch between the lysosomal system and mitochondria dynamics. Whole-body MIR125B-2 knockout mice eating a high-fat diet develop insulin resistance and glucose intolerance, possibly because of the white fat accumulation (48). Nevertheless, insulin secretion or an effect in other metabolic organs were not assessed and our study provides compelling evidence of a role for miR-125b in controlling insulin secretion and strongly indicates that nontargeted administration of miR-125b mimics may reach the islet and lead to defective β-cell function and worsen diabetes.

Studies focused on β cell–specific inhibition of miR-125b and on the function of specific targets will be essential to confirm a potential therapeutic benefit of targeting this miRNA or its targets for the treatment of diabetes.

Acknowledgments. The authors thank Stephen M. Rothery from the Facility for Imaging by Light Microscopy at Imperial College London for support with confocal and widefield microscopy image acquisition and analysis.

Funding. A.M.-S. was supported by an MRC New Investigator Research Grant (MR/P023223/1). G.A.R. was supported by a Wellcome Trust Senior Investigator Award (WT098424AIA) and Investigator Award (212625/Z/18/Z) and MRC Programme grants (MR/R022259/1, MR/J0003042/1, MR/L020149/1). A.To. was funded by the MRC (MR/R010676/1). I.L. was supported by a Diabetes UK project grant (16/0005485). This project has received funding from the European Union’s Horizon 2020 research and innovation program via the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement 115881 (RHAPSODY) to G.A.R. and P.M. Human islet preparations (Milan, Italy) were obtained from the European Consortium for Islet Transplantation; the Human Islet Distribution program was supported by Juvenile Diabetes Research Foundation Grant 3-RSC-2016-160-I-X.

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

Author Contributions. R.C. and G.P. performed most experiments, designed research studies, and obtained and analyzed data. P.C. performed research studies (total internal reflection of fluorescence) and designed microscopy macros. D.R. assisted with computational analysis of RNA-sequencing and RNA-immunoprecipitation sequencing data. A.To. and T.B. performed transmission electron microscopy experiments. Z.W., A.S., A.Th., A.M., and Y.C. performed research studies (insulin secretion, β-cell mass and apoptosis, and Western blot, respectively). M.-S.N.-T. assisted with animal studies (in vivo insulin secretion). M.T.D. and D.A.J. performed whole-cell voltage clamp experiments. P.M., K.B., J.S., L.P., and E.d.K. provided human islets. I.L. assisted with maintenance of animal colonies. K.S. and D.M.S. provided C13 and C991, respectively. G.A.R. contributed to the study design, with reagents and AMPK and LKB1 KO animals, and in writing the manuscript. A.M.-S. conceived the study, designed and performed experiments, analyzed data, and wrote the manuscript. All authors read and approved the manuscript. A.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 at the 79th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 7–11 June 2019.