MicroRNAs (miRNAs) are important regulators of gene expression programs in the pancreas; however, little is known about the role of miRNA pathways during endocrine cell specification and maturation during neonatal life. In this study, we deleted Dicer1, an essential RNase for active miRNAs biogenesis, specifically from NGN3+ endocrine progenitor cells. We found that deletion of Dicer1 in endocrine progenitors did not affect the specification of hormone-expressing endocrine cells. However, the islets in the mutant mice in the neonatal period exhibited morphological defects in organization and loss of hormone expression, and the mutant mice subsequently developed diabetes. Dicer1-deficient β-cells lost insulin expression while maintaining the expression of β-cell transcription factors such as Pdx1 and Nkx6.1 early in the postnatal period. Surprisingly, transcriptional profiling showed that that the Dicer1-deficient endocrine cells expressed neuronal genes before the onset of diabetes. The derepression of neuronal genes was associated with a loss in binding of the neuronal transcriptional repressor RE-1-silencing transcription factor to its targets in Dicer1-deficient β-cells. These studies suggest that miRNAs play a critical role in suppressing neuronal genes during the maturation of endocrine cells.

A potential therapeutic approach to replenish the pancreatic β-cell mass in diabetic patients involves the transplantation of functional, glucose-responsive β-cells differentiated from human pluripotent stem cells. Several attempts have been made at differentiating β-cells in vitro from stem cells, with limited success, (1,2) because the insulin-expressing cells generated lack the characteristic hallmarks of functionally mature β-cells, such as the ability to regulate glucose-stimulated insulin-secretion. Although many transcription factors and signaling pathways underlying the stepwise cell fate acquisition during β-cell development are known (36), a complete understanding of the molecular basis of β-cell specification and functional maturation is lacking.

Of significant interest is the role of microRNAs (miRNAs) in regulating the pancreatic developmental program. miRNAs are non–protein-coding small RNAs (∼19–25 nucleotides) that negatively regulate gene expression at the post-transcriptional level (7) and have been implicated as important regulators of animal development (8). Newly transcribed miRNAs undergo a series of processing steps that require the RNase III enzymes Drosha and Dicer1 before becoming functional (9,10). Although several miRNAs have been proposed to regulate β-cell transcription factors during development (11), many of these computationally predicted miRNA–mRNA interactions have not been experimentally validated in vivo. The dysregulation of miRNAs through Dicer1 ablation in the early embryonic pancreatic progenitor cells expressing Pdx-1 resulted in severe deficiencies in the formation of all islet cell lineages (12). More recently, it has been shown that deletion of Dicer1 in β-cells leads to loss of insulin expression and to development of diabetes in adult mice (13). Although these studies reveal key functions of miRNA-dependent pathways during early pancreatic development and in adult β-cells, they preclude analysis of the role of miRNAs during the specification of endocrine cells and their functional maturation in postnatal life.

In this study, we used a mouse model where expression of Cre recombinase directed by the Ngn3 promoter conditionally deleted floxed Dicer1 alleles in endocrine progenitor cells. In addition, by crossing these mice onto the R26RYFP reporter line, we were able to trace the lineage of the Dicer1-deficient islet progenitor cells. Our data demonstrate that Dicer1-deficient endocrine progenitors differentiate into hormone-expressing endocrine cells but subsequently lose hormone expression during the neonatal period and develop diabetes. More surprisingly, we found that the Dicer1-deficient islet cells expressed neuronal genes, supporting a model in which miRNA pathways control important transcriptional networks required for suppressing neuronal fate during the maintenance and maturation of newly specified endocrine cells.

Mice and physiology.

Mice were maintained in a 12-h light/dark cycle under standard conditions. Studies involving mice were performed in accordance with National Institutes of Health policies on the use of laboratory animals and approved by the University of California, Los Angeles (UCLA) Animal Research Committee. The mice used in this study are the conditional Dicer1 flox/flox line (14), the Ngn3-Cre (15), and the R26R-YFP (16) lines. The control mice used throughout were heterozygous for the conditional Dicer1 allele and the Ngn3-Cre transgene (NC:Dicer1fl/+). All mice were maintained in the C57BL6 background.

DNA extracted from tails was used for PCR-based genotyping. Blood glucose levels were measured from tail vein blood using a FreeStyle glucometer (Abbot Diabetes Care), and pancreatic insulin content was measured using a mouse insulin ELISA kit (Mercodia) after acid ethanol (0.18 mol/L HCl in 70% ethanol) extraction according to the protocol recommended by the Animal Models of Diabetes Complications Consortium (http://www.amdcc.org/).

Histology and immunohistochemistry.

Pancreatic tissue was processed for immunohistochemical analyses as previously described (17). Briefly, the pancreas was dissected and fixed in 4% formaldehyde for 2 h before being embedded in paraffin. Sections (5-μm thick) were deparaffinized, rehydrated, subjected to antigen retrieval using Antigen Unmasking Buffer (Vector Laboratories), and permeabilized in 0.4% Triton X-100/TBS. Tissues were subsequently blocked with 3% IgG-free BSA (Jackson ImmunoResearch Laboratories). Incubation with primary antibodies was performed overnight at 4°C in blocking solution at the following dilutions: 1:200 guinea pig anti-insulin (Dako), 1:500 rabbit anti-glucagon (Immunostar), 1:200 rabbit anti-amylase (Sigma-Aldrich), 1:100 mouse anti-Pdx1 (DHSB), 1:200 rabbit anti-Pdx1 (Chemicon), 1:250 chicken anti-green fluorescent protein (GPF; Aves Laboratories Inc.), 1:500 rabbit anti-MafA (18), 1:100 rabbit anti-MafB (Bethyl), 1:1000 rabbit anti-Glut2 (Millipore), 1:250 rabbit anti-tyrosine hydroxylase (Th; Millipore), 1:50 mouse anti-Nkx2.2 (BCBC), 1:50 mouse anti-Nkx6.1, and 1:50 mouse anti-Ki67 (BD Pharmingen). Secondary antibodies (Jackson ImmunoResearch Laboratory) conjugated to fluorescein isothiocyanate (1:200 dilution) or Cy3 (1:1000 dilution) were diluted in blocking buffer. In situ cell death detection assay (TUNEL) was performed according to the manufacturer’s instructions (Roche). Slides were mounted with Vectashield with DAPI (Vector Laboratories), and images were obtained using Openlab software (Perkins Elmer) and a Leica DM6000 microscope.

β-Cell mass analysis.

β-Cell mass was calculated, as previously described (19), by analyzing pancreata from four mice for each age and genotype.

Islet isolation and fluorescence-activated cell sorter sorting.

Islet isolation from mice was performed using a Liberase enzyme blend and purified by centrifugation in a Histopaque gradient, as described previously (17), with slight modifications. Instead of Liberase perfusion through the common bile duct, neonatal pancreata were randomly injected with Liberase blend. To obtain purified pancreatic and endocrine progenitor cells, pancreatic buds from embryonic day (e)12.5 Pdx1-yellow fluorescent protein (YFP) and e14.5 Ngn3-GFP (20) embryos were dissected and dissociated into a single-cell suspension using nonenzymatic cell dissociation buffer (Sigma-Aldrich). GFP+ cells were sorted by fluorescence-activated cell sorter (FACS) (FACSaria BD biosciences). A similar process was followed to obtain β-cells from islets isolated from 2-week-old MIP-GFP (21) mice.

RNA isolation, RT-PCR, and real-time quantitative PCR.

Total RNA from islets, dissected pancreata, and Ngn3-GFP and MIP-GFP cells was isolated using an RNeasy RNA extraction Kit (Qiagen). Single-stranded cDNA was prepared using Superscript III Reverse Transcriptase (Invitrogen) with oligodT priming. Real-time quantitaive (q)PCRs were performed using the 7900HT Fast Real-time PCR system (Applied Biosystems). The expression level of each transcript was normalized to the housekeeping gene Cyclophilin. Data and standard deviations shown were measured from at least three independent biological replicate experiments. All RT-PCR and real-time qPCR primer sequences used are available upon request.

Chromatin immunoprecipitation assays and chromatin immunoprecipitation-qPCR.

To obtain purified β-cells from P7 control and NC:Dicer1fl/fl mice for chromatin immunoprecipitation (ChIP) analyses, dissected pancreata were dissociated into a single-cell suspension. The pancreatic cells were immunostained for insulin, after fixation and permeabilization with BD Cytofix reagent (BD PharMingen), using guinea pig anti-insulin antibody, followed by incubation with a Cy3-conjugated secondary antibody, and sorted by FACS (FACSaria BD Bioscience). Pancreatic cells processed without primary antibody were used as a negative control for FACS. ChIP experiments on purified β-cells were performed using the micro-ChIP protocol, as previously described (22). The sequence-specific primers used to amplify the region around the RE1 sites on the Stmn2, Stmn3, Th, and Syn1 locus are available upon request. The qPCR analysis was performed using the 7900HT Fast Real-time PCR system (Applied Biosystems). The data shown were from independent biological triplicates, and ChIP-qPCR signals are reported as percentage of input.

Microarray gene expression analysis.

Two pairs of independent pancreata from control and mutant NC:Dicer1fl/fl mice were dissected out at postnatal day 7 (P7), and total RNA was extracted using an RNeasy Plus Mini Kit (Qiagen), according to the manufacturer’s protocol. Total RNA quality was assessed using an Agilent 2100 Bioanalyzer and an RNA Integrity Number (RIN) generated using 2100 Expert Software (Agilent Technologies). All RNA samples used had a RIN greater than 7. One microgram of total RNA was processed, labeled, and hybridized to the mouse 1.0ST GeneChip Microarray (Affymetrix), according to manufacturer’s recommendations, by the UCLA DNA Microarray Core Facility.

The CEL files (raw expression measurement data) generated from array image analysis were imported into Partek Genomics Suite Software (23), robust multi-array average (RMA) normalized, and converted to log2 values. The software was further used to perform ANOVA and identify statistically significant differentially expressed genes (P values) between control and NC:Dicer1fl/fl pancreata by applying a false discovery threshold of 5% to the P values.

Statistical analyses.

All data are expressed as means ± SEM. Statistical significance was determined by an unpaired Student t test. A P value <0.05 was used to reject the null hypothesis.

Dicer1-null endocrine precursor cells specify appropriately but display loss of hormone expression during neonatal phase and subsequently develop diabetes.

The catalytic role of Dicer1 is central to the functional maturation of miRNAs from their precursors (9). To investigate in vivo the role of miRNAs during islet cell specification from endocrine precursor cells expressing Neurogenin3 (Ngn3), the Dicer1 conditional allele (Dicer1fl/fl) (14) was deleted specifically from endocrine precursor cells using NC-mediated excision (15) thereby preventing the generation of active miRNAs in these cells. In addition, an R26R-YFP reporter line (16) was crossed onto the NC:Dicer1fl/fl mice to heritably label any cells undergoing recombination and deletion of Dicer1. Co-immunostaining of NC:Dicer1fl/fl pancreatic sections for insulin/YFP (Fig. 1A–F) and glucagon/YFP (data not shown) showed a high degree of overlap of these markers in the mutant islets, therefore confirming that a high percentage of specified islet cells were derived from endocrine precursor cells that had undergone Cre-mediated recombination. Real-time qPCR analysis revealed that the Dicer1 transcript was reduced by 80% in the mutant NC:Dicer1fl/fl islets at 1 week (Fig. 1G), further confirming the deletion of Dicer1 from a majority of islet cells.

FIG. 1.

Mutant NC:Dicer1fl/fl islet cells specify normally but lose hormone expression and develop hyperglycemia during neonatal development. Immunostaining of 1-week-old Cre negative (A–C) and mutant NC:Dicer1fl/fl (D–F) pancreatic sections (n = 3) for insulin (red ) (B and E), YFP that marks recombined cells (green) (A and D), and DAPI to visualize the nuclei (blue). C and F: Most of the insulin+ β-cells colocalize with YFP in the mutant islets but not in the control islets. G: Transcript levels of Dicer1 determined by real-time qPCR using RNA isolated from islets of control and mutant NC:Dicer1fl/fl mice (n = 3) at P7. Expression levels in control were set as one arbitrary unit. Representative pancreatic sections (n = 3 for each age and genotype) from e18.5 and 1- and 2-week-old control (HJ) and mutant NC:Dicer1fl/fl (KM) littermates were immunostained for insulin (green) and glucagon (red). N: Nonfasting blood glucose levels in neonatal control and mutant NC:Dicer1fl/fl mice at 1, 2, and 4 weeks (n = 5 for each age group). NC:Dicer1fl/fl mutant mice are hyperglycemic by 2 weeks. O: Total pancreatic insulin content in control and NC:Dicer1fl/fl mice at 1 and 3 weeks reveals minimal pancreatic insulin content remaining in NC:Dicer1fl/fl mice at 3 weeks (n ≥ 3). PQ: β- and α-cell mass in control and mutant NC:Dicer1fl/fl mice at P1 and 2 weeks was assessed as described in research design and methods (n = 4). The error bars represent the SEM. *P < 0.05, **P < 0.01, ***P < 0.005. In all cases, at least two to three pancreas sections were used for each animal.

FIG. 1.

Mutant NC:Dicer1fl/fl islet cells specify normally but lose hormone expression and develop hyperglycemia during neonatal development. Immunostaining of 1-week-old Cre negative (A–C) and mutant NC:Dicer1fl/fl (D–F) pancreatic sections (n = 3) for insulin (red ) (B and E), YFP that marks recombined cells (green) (A and D), and DAPI to visualize the nuclei (blue). C and F: Most of the insulin+ β-cells colocalize with YFP in the mutant islets but not in the control islets. G: Transcript levels of Dicer1 determined by real-time qPCR using RNA isolated from islets of control and mutant NC:Dicer1fl/fl mice (n = 3) at P7. Expression levels in control were set as one arbitrary unit. Representative pancreatic sections (n = 3 for each age and genotype) from e18.5 and 1- and 2-week-old control (HJ) and mutant NC:Dicer1fl/fl (KM) littermates were immunostained for insulin (green) and glucagon (red). N: Nonfasting blood glucose levels in neonatal control and mutant NC:Dicer1fl/fl mice at 1, 2, and 4 weeks (n = 5 for each age group). NC:Dicer1fl/fl mutant mice are hyperglycemic by 2 weeks. O: Total pancreatic insulin content in control and NC:Dicer1fl/fl mice at 1 and 3 weeks reveals minimal pancreatic insulin content remaining in NC:Dicer1fl/fl mice at 3 weeks (n ≥ 3). PQ: β- and α-cell mass in control and mutant NC:Dicer1fl/fl mice at P1 and 2 weeks was assessed as described in research design and methods (n = 4). The error bars represent the SEM. *P < 0.05, **P < 0.01, ***P < 0.005. In all cases, at least two to three pancreas sections were used for each animal.

To determine whether loss of Dicer1 affected the specification of islet cells, pancreatic tissue isolated from control (heterozygous for the Dicer1 conditional allele NC:Dicer1fl/+) and mutant NC:Dicer1fl/fl mice were co-immunostained for insulin and glucagon at different stages of neonatal development. Immunostaining of pancreata from e18.5 NC:Dicer1fl/fl mice for insulin and glucagon revealed that α- and β-cells specified appropriately in the pancreas, and no morphological differences in islet architecture were apparent (Fig. 1H and K). However, pancreata from 1-week-old NC:Dicer1fl/fl mice exhibited altered islet organization, with many instances of α-cells prevalent within the core of the islet rather than at the periphery, an observation also confirmed by Pdx1 and glucagon staining at the same stage (Fig. 1I and L; Supplementary Fig. 1A and B). Pancreata from 2-week-old NC:Dicer1fl/fl mice displayed severe defects in morphology, and the expression of insulin and glucagon was severely diminished (Fig. 1J and M). Quantification at 1 day after birth revealed similar β- and α-cell mass (Fig. 1P and Q). A similar comparison showed a dramatic reduction in endocrine cell mass in the mutant NC:Dicer1fl/fl mice at 2 weeks after birth. Consistent with the reduction in β-cell mass by 2 weeks (Fig. 1P), an almost total loss of pancreatic insulin content was observed in the mutant NC:Dicer1fl/fl animals (shown at 3 weeks in Fig. 1O). However, only a very modest decrease in insulin content was noticeable in the mutant NC:Dicer1fl/fl animals at 1 week. The mutant NC:Dicer1fl/fl animals displayed an inability to metabolize glucose and developed hyperglycemia and frank diabetes within 2 weeks of birth (Fig. 1N), consistent with a nearly total loss of β-cells by that age. Taken together, our results suggest that Dicer1 in the endocrine progenitors was not required for specification of endocrine cells during embryogenesis but was required postnatally to maintain the expression of hormones and the maintenance of endocrine cell mass. These observations therefore underscore the key role miRNAs play during the neonatal period when endocrine cells became functionally mature and capable of maintaining blood glucose levels.

Endocrine cells from Dicer1-null mice maintain expression of characteristic transcription factors despite losing hormone expression.

Next, we asked whether the loss of hormone expression in mutant NC:Dicer1fl/fl islet cells was due to any potential effect on key islet transcriptional regulators upon ablation of miRNAs. We tested this hypothesis by assessing the expression of characteristic β-cell transcription factors Pdx1, Nkx6.1, and MafA. Immunostaining of pancreata from 2-week-old control mice showed the expected coexpression of all these transcription factors with insulin (Fig. 2A–C and data not shown). In contrast, whereas the pancreata from mutant NC:Dicer1fl/fl littermates showed normal expression levels of Pdx1, Nkx6.1, and MafA in the islets, only a few of these cells were also positive for insulin, with the rest devoid of insulin expression (Fig. 2D–F). Similarly, 2-week-old mutant NC:Dicer1fl/fl pancreatic sections showed a number of cells that expressed MafB but not glucagon (Supplementary Fig. 1C and D). Furthermore, real-time qPCR analyses of a set of key β-cell genes between control and mutant NC:Dicer1fl/fl pancreata revealed a significant reduction in the expression of Insulin1 and Insulin2 mRNA, whereas no differences in the expression of transcription factors markers was observed (Supplementary Fig. 1E and F and Supplementary Fig. 2A). These results suggested that endocrine hormone expression was lost although the transcriptional regulatory genes that characterize endocrine cells were intact in the absence of Dicer1.

FIG. 2.

Loss of insulin expression in mutant NC:Dicer1fl/fl β-cells is not preceded by downregulation of characteristic β-cell transcription factors. Representative pancreatic sections from 2-week-old control and mutant NC:Dicer1fl/fl mice co-immunostained for insulin (Ins; green) and Pdx1 (red) (A and D), Nkx6.1 (red) (B and E), and MafA (red) (C and F). Many instances of insulin-negative β-cells that still express the β-cell transcription factors are observed in the mutant sections. In control animals, β-cells always coexpress insulin with the transcription factors (n ≥ 3 for each group of animals). In all cases, at least two to three pancreas sections were used for each animal.

FIG. 2.

Loss of insulin expression in mutant NC:Dicer1fl/fl β-cells is not preceded by downregulation of characteristic β-cell transcription factors. Representative pancreatic sections from 2-week-old control and mutant NC:Dicer1fl/fl mice co-immunostained for insulin (Ins; green) and Pdx1 (red) (A and D), Nkx6.1 (red) (B and E), and MafA (red) (C and F). Many instances of insulin-negative β-cells that still express the β-cell transcription factors are observed in the mutant sections. In control animals, β-cells always coexpress insulin with the transcription factors (n ≥ 3 for each group of animals). In all cases, at least two to three pancreas sections were used for each animal.

Deletion of Dicer-1 upregulates neuronal genes in islet cells.

To investigate the molecular changes in endocrine cells of mutant NC:Dicer1fl/fl animals that could be responsible for the loss of hormone expression in endocrine cells, we performed genome-wide transcription profiling using microarrays on two independent pairs of control and mutant NC:Dicer1fl/fl pancreata isolated at P7. The microarray was performed at this stage, which precedes any apparent physiological changes. Transcription profiling analysis revealed 162 differentially expressed genes with at least a twofold change at a false discovery rate of 5%, of which 145 genes were upregulated and 17 were downregulated. A surprisingly disproportionate number of neuronal genes were upregulated in the NC:Dicer1fl/fl pancreata, whereas the expression of transcription factors and other characteristic islet cell markers did not exceed the twofold change threshold (Fig. 3A–C). The upregulation of these neuronal genes in NC:Dicer1fl/fl pancreata was further validated by performing real-time qPCR analysis from the microarray samples (Fig. 3E) as well as on a triplicate of samples independent of those used for the microarray analysis (Supplementary Fig. 2A).

FIG. 3.

Unchanged endocrine gene expression but increased neuronal gene expression in mutant NC:Dicer1fl/fl islet cells at the presymptomatic stage (P7). A: Representative scatter plot of global gene expression obtained from microarray analysis of control and mutant NC:Dicer1fl/fl pancreas at P7. A high similarity (R2) in gene expression is observed between control and mutant NC:Dicer1fl/fl samples. Red lines indicate a twofold difference in expression. B: The expression levels of many neuronal markers as identified by the Gene Ontology database are upregulated in the mutant NC:Dicer1fl/fl samples. Key upregulated neuronal genes (red dots) are circled. The expression levels of most endocrine cell markers are unchanged (black dots). C: List of key neuronal genes (circled in B) upregulated in microarray ChIP analyses of mutant NC:Dicer1fl/fl vs. control samples. D: Gene ontology analysis displaying biological processes associated with gene function of the differentially expressed genes with fold-change greater than two. The x-axis values are in logarithmic scale and correspond to raw binomial P values. E: Real-time qPCR validation of representative upregulated neuronal gene transcripts Nefm, Stmn2, Stmn3, Syn1, Npy, Th, and Dbh using RNA isolated from pancreata of control and mutant NC:Dicer1fl/fl mice. Expression levels in control pancreata were set as one arbitrary unit. All data points represent means ± SEM of at least three biologically independent experiments. *P < 0.05, **P < 0.01, ***P < 0.005. (A high-quality color representation of this figure is available in the online issue.)

FIG. 3.

Unchanged endocrine gene expression but increased neuronal gene expression in mutant NC:Dicer1fl/fl islet cells at the presymptomatic stage (P7). A: Representative scatter plot of global gene expression obtained from microarray analysis of control and mutant NC:Dicer1fl/fl pancreas at P7. A high similarity (R2) in gene expression is observed between control and mutant NC:Dicer1fl/fl samples. Red lines indicate a twofold difference in expression. B: The expression levels of many neuronal markers as identified by the Gene Ontology database are upregulated in the mutant NC:Dicer1fl/fl samples. Key upregulated neuronal genes (red dots) are circled. The expression levels of most endocrine cell markers are unchanged (black dots). C: List of key neuronal genes (circled in B) upregulated in microarray ChIP analyses of mutant NC:Dicer1fl/fl vs. control samples. D: Gene ontology analysis displaying biological processes associated with gene function of the differentially expressed genes with fold-change greater than two. The x-axis values are in logarithmic scale and correspond to raw binomial P values. E: Real-time qPCR validation of representative upregulated neuronal gene transcripts Nefm, Stmn2, Stmn3, Syn1, Npy, Th, and Dbh using RNA isolated from pancreata of control and mutant NC:Dicer1fl/fl mice. Expression levels in control pancreata were set as one arbitrary unit. All data points represent means ± SEM of at least three biologically independent experiments. *P < 0.05, **P < 0.01, ***P < 0.005. (A high-quality color representation of this figure is available in the online issue.)

Gene ontology analysis for biological processes associated with the dysregulated genes revealed a significant enrichment of genes involved in nervous system development, synaptic transmission, transmission of nerve impulse, and neurogenesis (Fig. 3D). The upregulated genes included pan-neuronal markers such as Scg10, Stmn3, and the neurofilament markers Nefl and Nefm, as well as molecular markers of noradrenergic neurons, such as the neurotransmitter-synthesizing enzymes Th and dopamine β hydroxylase (Dbh), the vesicular transport molecule–vesicular monoamine transporter 2 (Vmat2), and the plasma membrane transporter–norepinephrine transporter (Net). In fact, many components of the noradrenergic program were upregulated. A number of these neuronal genes are expressed to some extent early during endocrine pancreas development, but their expression declines significantly by 3 to 4 weeks of age (24). For example, real-time qPCR analysis indicated that the expression of neuropeptide Y (NPY), which is normally downregulated upon maturation, was expressed at high levels in the mutant NC:Dicer1fl/fl pancreata.

To test whether the increase in transcript levels of the neuronal genes was associated with a corresponding increase in protein levels, we immunostained 1 week control and NC:Dicer1fl/fl pancreatic sections for insulin and Th. Very few Th and insulin double-positive cells were observed in control sections (Fig. 4A–C). In contrast, the mutant NC:Dicer1fl/fl sections displayed a larger percentage of cells that coexpressed insulin and Th (Fig. 4D–F). Thus, the upregulation of mRNA of neuronal genes due to the absence of Dicer1 also results in a concomitant increase at the protein level. To assess whether the upregulation of Th was limited to β-cells, we costained Th with Pdx1. All Th-positive cells in the NC:Dicer1fl/fl islets also coexpressed Pdx1 (Fig. 4H–M), suggesting that β-cells upregulated the neuronal marker Th in the absence of Dicer1. To verify the endocrine lineage of these Th-positive cells, we performed immunostaining for the lineage trace marker YFP and Th in the mutant NC:Dicer1fl/flpancreata. All of the Th-positive cells were also positive for YFP staining (Fig. 4G and J), therefore confirming that the Th-positive cells were derived from endocrine precursor cells that had undergone recombination and lost Dicer1 expression.

FIG. 4.

Expression of Th in mutant NC:Dicer1fl/fl pancreas. Pancreatic sections from 1-week-old control and mutant NC:Dicer1fl/fl littermates showing immunostaining for insulin (Ins; green) (A and D), Th (red) (B and E), and merged view (C and F) reveals sharp upregulation of Th expression in the mutant β-cells (n = 3). Immunostaining of 1-week-old control (G and H) and NC:Dicer1fl/fl mutant (J and K) pancreatic sections for Th (red)/YFP (green) and Th (red)/Pdx1 (green; n = 3). Most Th expression is limited to Pdx1-positive cells. I and L: Magnified inset of an area in H and K is shown. M: Quantification of Th-positive cells expressed as percentage of insulin-positive and Pdx1-positive cells in 1-week-old mice. Data points represent means ± SEM of at least three independent experiments. ***P < 0.005. Immunostaining of 2-week-old control and mutant NC:Dicer1fl/fl littermates for insulin (Ins; green) (N and Q), Th (O and R), and merged view (P and S) indicates the presence of many insulin-negative β-cells that express Th (n = 3). In all cases, at least two to three pancreas sections were used for each animal.

FIG. 4.

Expression of Th in mutant NC:Dicer1fl/fl pancreas. Pancreatic sections from 1-week-old control and mutant NC:Dicer1fl/fl littermates showing immunostaining for insulin (Ins; green) (A and D), Th (red) (B and E), and merged view (C and F) reveals sharp upregulation of Th expression in the mutant β-cells (n = 3). Immunostaining of 1-week-old control (G and H) and NC:Dicer1fl/fl mutant (J and K) pancreatic sections for Th (red)/YFP (green) and Th (red)/Pdx1 (green; n = 3). Most Th expression is limited to Pdx1-positive cells. I and L: Magnified inset of an area in H and K is shown. M: Quantification of Th-positive cells expressed as percentage of insulin-positive and Pdx1-positive cells in 1-week-old mice. Data points represent means ± SEM of at least three independent experiments. ***P < 0.005. Immunostaining of 2-week-old control and mutant NC:Dicer1fl/fl littermates for insulin (Ins; green) (N and Q), Th (O and R), and merged view (P and S) indicates the presence of many insulin-negative β-cells that express Th (n = 3). In all cases, at least two to three pancreas sections were used for each animal.

We also examined Th expression in the mutant NC:Dicer1fl/fl islets at 2 weeks. Consistent with observations in islets from 1-week-old control mice, costaining of Th with insulin (Fig. 4N–S) revealed that control islets from 2-week-old mice also typically displayed only a few Th-positive cells that also stained for insulin (Fig. 4N–P). In contrast, mutant NC:Dicer1fl/fl islets displayed a massive upregulation of Th staining (Fig. 4Q–S). A few insulin and Th-copositive islet cells were evident, but most of the Th-positive cells did not show any insulin staining, suggesting that neuronal gene upregulation persists in mutant islet cells that have already downregulated endocrine hormones. Taken together, our results suggest that miRNAs likely play an important role in suppressing a neuronal gene program during the maturation phase of endocrine islet cells.

Loss of RE1-silencing transcription factor–mediated repression of neuronal targets in Dicer1-null β-cells.

We hypothesized that the acquisition of a neuronal transcriptional profile by the mutant NC:Dicer1fl/fl islet cells could possibly be attributed to the dysregulation of transcriptional programs dedicated toward the maintenance of islet cell identity. To investigate transcription factors regulating neuronal gene repression in islet cells, we used the MetaCore Interactome characterization tool (GeneGo, Inc., http://www.genego.com) to identify the top transcription factors that exhibited significantly large numbers of interactions with the dysregulated genes in the mutant NC:Dicer1fl/fl islet cells. The Interactome analysis revealed that 16% of the dysregulated genes were targets of the neuronal transcriptional repressor RE1-silencing transcription factor (REST), also known as neuron restrictive silencer factor (Fig. 5A). REST is a repressor of neuronal genes that binds in a sequence-specific manner to the RE1 elements of its target genes and mediates transcriptional repression (25,26).

FIG. 5.

Loss of REST binding to neuronal targets in mutant NC:Dicer1fl/fl β-cells. A: MetaCore Interactome analysis (GeneGo, Inc.) revealed that 16% of the dysregulated genes (more than twofold change) are targets of the neuronal transcriptional repressor REST. B: Real-time qPCR analysis of Ins1, Ngn3, and REST mRNA transcript levels in adult lung tissue and brain tissue isolated at e14.5, FACS-purified Ngn3-positive pancreatic endocrine progenitor cells at e14.5, and MIP-GFP β-cells at 2 weeks. C: ChIP analysis of FACS-purified control and mutant NC:Dicer1fl/fl mice β-cells at P7. REST binding to the RE1 regions of Stmn2, Stmn3, Th, and Syn1 locus was reduced in mutant β-cells. ChIP data are reported as fold-change of REST binding at the RE1 locus of the respective genes relative to a negative control region and represent means ± SEM of at least three biologically independent experiments. D: Real-time qPCR analysis of REST, Pdx1, and Th mRNA transcript levels in isolated islets from pancreata of control and mutant NC:Dicer1fl/fl mice at P7. E: Transcript levels of Phox2a and Phox2b as determined by real-time qPCR using RNA isolated from pancreata of control and mutant NC:Dicer1fl/fl mice at P7. Expression levels in the control were set as one arbitrary unit. All data points represent means ± SEM of at least three biologically independent experiments. *P < 0.05, **P < 0.01, ***P < 0.005. (A high-quality color representation of this figure is available in the online issue.)

FIG. 5.

Loss of REST binding to neuronal targets in mutant NC:Dicer1fl/fl β-cells. A: MetaCore Interactome analysis (GeneGo, Inc.) revealed that 16% of the dysregulated genes (more than twofold change) are targets of the neuronal transcriptional repressor REST. B: Real-time qPCR analysis of Ins1, Ngn3, and REST mRNA transcript levels in adult lung tissue and brain tissue isolated at e14.5, FACS-purified Ngn3-positive pancreatic endocrine progenitor cells at e14.5, and MIP-GFP β-cells at 2 weeks. C: ChIP analysis of FACS-purified control and mutant NC:Dicer1fl/fl mice β-cells at P7. REST binding to the RE1 regions of Stmn2, Stmn3, Th, and Syn1 locus was reduced in mutant β-cells. ChIP data are reported as fold-change of REST binding at the RE1 locus of the respective genes relative to a negative control region and represent means ± SEM of at least three biologically independent experiments. D: Real-time qPCR analysis of REST, Pdx1, and Th mRNA transcript levels in isolated islets from pancreata of control and mutant NC:Dicer1fl/fl mice at P7. E: Transcript levels of Phox2a and Phox2b as determined by real-time qPCR using RNA isolated from pancreata of control and mutant NC:Dicer1fl/fl mice at P7. Expression levels in the control were set as one arbitrary unit. All data points represent means ± SEM of at least three biologically independent experiments. *P < 0.05, **P < 0.01, ***P < 0.005. (A high-quality color representation of this figure is available in the online issue.)

The upregulation of REST target genes in NC:Dicer1fl/fl islet cells suggested an important role for REST during islet cell development and maturation. We first investigated whether REST was expressed at different stages of islet cell development. Real-time qPCR analysis of REST expression showed that REST transcripts were present in FACS-purified Ngn3-positive endocrine precursor cells at e14.5 and in β-cells at P14 at levels comparable to those expressed in e14.5 brain tissue and adult lung tissue where REST is known to be highly expressed (27,28) (Fig. 5B). Although microarray and real-time qPCR analysis (data not shown and Fig. 5D) did not reveal any changes in REST mRNA levels between control and NC:Dicer1fl/fl islets, we hypothesized that changes in miRNAs could potentially be regulating molecular events that may alter recruitment of REST protein and/or members of the REST repressive complex.

To analyze whether REST binding to the RE1 sites of its target genes was lost in the mutant NC:Dicer1fl/fl β-cells, ChIP analysis on a subset of the upregulated REST target genes showed that REST was bound to the RE1 sites of the four genes examined in FACS-purified control β-cells at P7. In contrast, REST binding to its target genes was significantly reduced in the mutant NC:Dicer1fl/fl β-cells (Fig. 5C), and a concomitant increase in mRNA expression of these genes was observed in the mutant NC:Dicer1fl/fl pancreata (Fig. 3E). This suggests that miRNAs play an important role in regulating the recruitment of the REST repressive complex, which in turn represses the transcription of its target neuronal genes.

In addition to REST repression of the neuronal genes, we also identified other potential candidate transcription factors using the Interactome analysis, including the neuronal homeobox transcription factors Phox2a and Phox2b, which are important regulators of the noradrenergic phenotype in vertebrates (2931). Real-time qPCR analysis indicated that both transcripts were significantly upregulated in the mutant NC:Dicer1fl/fl pancreata (Fig. 5D). Taken together, our results reveal an important role of miRNAs in regulating key transcriptional programs required for maintaining islet cell identity during neonatal development.

Dicer1-null islet cells die by early adulthood.

To determine the ultimate fate of the Dicer1-null islet cells that had acquired an altered transcriptome, we immunostained the islets of control and NC:Dicer1fl/fl pancreatic sections for the islet hormone markers at 6 weeks. By 6 weeks, most insulin expression (Fig. 6A and B, Supplementary Fig. 3A and B) and other hormone expression in the islet was lost (data not shown). To determine whether the hormone-negative Dicer1-null islet cells survived into adulthood, immunostaining of control and NC:Dicer1fl/fl pancreatic sections with the lineage trace marker YFP revealed that by 6 weeks, very few YFP-positive cells remained in the mutant NC:Dicer1fl/fl pancreatic sections compared with the littermates (Fig. 6C and D). Because the genetic ablation of Dicer1 is associated with apoptosis during development (14,32,33), we assessed the apoptosis rate in control and mutant NC:Dicer1fl/fl islets using TUNEL immunostaining. A significant increase (11.7-fold) in apoptosis was observed in mutant NC:Dicer1fl/fl β-cells at P12 (Fig. 6E–G) as well as other islet cells (data not shown), therefore suggesting that Dicer1 was ultimately responsible for the survival of mature islet cells.

FIG. 6.

Dicer1-null islet cells undergo apoptosis during neonatal period. AD: Representative pancreatic sections (n = 3) from 6-week-old control and mutant NC:Dicer1fl/fl mice co-immunostained for insulin (green)/amylase (red) (A and B) and YFP (green)/DAPI (blue) (C and D). E and F: Immunostaining of pancreatic sections (n = 3) from 12-day-old control and mutant NC:Dicer1fl/fl mice for insulin and TUNEL reveals an increase in TUNEL-positive apoptotic cells in the mutant mice. G: Quantification of TUNEL-positive cells expressed as percentage of β-cells in 12-day-old mice. Data points represent means ± SEM of at least three independent experiments. ***P < 0.005. In all cases, at least two to three pancreas sections were used for each animal.

FIG. 6.

Dicer1-null islet cells undergo apoptosis during neonatal period. AD: Representative pancreatic sections (n = 3) from 6-week-old control and mutant NC:Dicer1fl/fl mice co-immunostained for insulin (green)/amylase (red) (A and B) and YFP (green)/DAPI (blue) (C and D). E and F: Immunostaining of pancreatic sections (n = 3) from 12-day-old control and mutant NC:Dicer1fl/fl mice for insulin and TUNEL reveals an increase in TUNEL-positive apoptotic cells in the mutant mice. G: Quantification of TUNEL-positive cells expressed as percentage of β-cells in 12-day-old mice. Data points represent means ± SEM of at least three independent experiments. ***P < 0.005. In all cases, at least two to three pancreas sections were used for each animal.

Our study illustrates an essential role of Dicer1-dependent miRNA pathways in the suppression of a subset of neuronal markers and the maintenance of endocrine hormone expression during the maturation of newly specified islet cells in neonatal development. Surprisingly, we found that deletion of the miRNA-processing enzyme Dicer1 in islet progenitors cells (NC:Dicer1fl/fl mice) did not affect the specification of endocrine cells. However, these newly specified endocrine cells acquired an altered transcriptome in which neuronal genes were upregulated and endocrine hormone expression was lost, resulting in failure to regulate blood glucose and onset of diabetes by age 2 weeks. One possible explanation for this is that the loss of miRNAs early in endocrine development may alter later developmental pathways involved in the maturation of endocrine cells.

Neuronal and β-cells both share the expression of a large number of proteins, and many similarities exist in the transcription factors involved in their differentiation programs (34). The neuronal genes could be direct targets of miRNAs, such that Dicer1 ablation would result in their derepression in the mutant NC:Dicer1fl/fl islet cells. Alternatively, it is a more likely scenario where transcriptional regulators of the neuronal markers that are direct or indirect targets of a combination of miRNAs could be deregulated in the NC:Dicer1fl/fl islet cells, resulting in the derepression of neuronal transcripts. The decreased binding of the neuronal transcriptional repressor REST to target neuronal genes, and increased levels of the neuronal transcriptional activators Phox2a and Phox2b in mutant NC:Dicer1fl/fl β-cells, suggests that these miRNA-controlled transcriptional regulators are likely part of a broader transcriptional program that actively suppresses neuronal differentiation genes to maintain islet cell identity.

In vivo genome-wide mapping of REST binding sites by ChIPSeq analysis has previously identified the critical islet-cell development transcription factors neurogenic differentiation 1 (NEUROD1), hepatocyte nuclear factor (HNF)4α, HNF6, Hes1, and Ngn3 as targets of REST (35). Interestingly, genes encoding these endocrine cell regulators were bound relatively poorly by REST compared with terminally differentiated neuronal genes such as Stmn2, Stmn3, and the neurofilament genes, which bound with greater affinity. This suggests that REST might not play a major role during islet cell differentiation during embryogenesis, but may instead control terminally differentiated neuronal genes that might be involved in maturation of endocrine cells. Dicer1 deletion results in deregulation of REST recruitment, resulting in the derepression of its target neuronal genes. The eventual loss of islet cells is not surprising given that the loss of Dicer1 expression is associated with massive cell death in a range of different tissues during development (14). Our results are supported by a recent study on the deletion of Dicer1 in β-cells using the transgenic RIP-Cre line, which showed that the neonatal pancreas contained ultrastucturally normal β-cells, although these mice developed progressive hyperglycemia and full-blown diabetes in adulthood (36). The derepression of neuronal genes in β-cells, however, was not explored.

Given the similarities in the neuronal and islet cell differentiation programs, our studies demonstrate a novel role of miRNA pathways in maintaining islet cell identity through the repression of neuronal markers while maintaining the expression of endocrine hormones. Identifying the individual miRNAs involved in the maturation and maintenance of newly specified islet cells, particularly β-cells, will aid in the development of therapeutic strategies to generate functionally mature replacement β-cells. These results also have important implications for engineering β-cells from embryonic stem cells.

This study was supported by National Institutes of Health Grant DK-083762 to M.G.M. and by a grant from the Helmsley Trust and Juvenile Diabetes Research Foundation to A.B.

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

M.K. designed and performed research, analyzed data, and wrote the manuscript. M.G.M. designed research and reviewed and edited the manuscript. A.B. designed research, analyzed data, and wrote the manuscript. A.B. 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.

The authors acknowledge technical help from Jiafang Wang, Lendy Le, and Emily Snyder of UCLA and helpful discussions, advice, and criticism from Dr. Peter C. Butler (UCLA) as well as members of the Bhushan laboratory.

1.
Kroon
E
,
Martinson
LA
,
Kadoya
K
, et al
.
Pancreatic endoderm derived from human embryonic stem cells generates glucose-responsive insulin-secreting cells in vivo
.
Nat Biotechnol
2008
;
26
:
443
452
[PubMed]
2.
Assady
S
,
Maor
G
,
Amit
M
,
Itskovitz-Eldor
J
,
Skorecki
KL
,
Tzukerman
M
.
Insulin production by human embryonic stem cells
.
Diabetes
2001
;
50
:
1691
1697
[PubMed]
3.
Habener
JF
,
Kemp
DM
,
Thomas
MK
.
Minireview: transcriptional regulation in pancreatic development
.
Endocrinology
2005
;
146
:
1025
1034
[PubMed]
4.
Bonal
C
,
Herrera
PL
.
Genes controlling pancreas ontogeny
.
Int J Dev Biol
2008
;
52
:
823
835
[PubMed]
5.
Jensen
J
.
Gene regulatory factors in pancreatic development
.
Dev Dyn
2004
;
229
:
176
200
[PubMed]
6.
Ackermann
AM
,
Gannon
M
.
Molecular regulation of pancreatic beta-cell mass development, maintenance, and expansion
.
J Mol Endocrinol
2007
;
38
:
193
206
[PubMed]
7.
Lagos-Quintana
M
,
Rauhut
R
,
Lendeckel
W
,
Tuschl
T
.
Identification of novel genes coding for small expressed RNAs
.
Science
2001
;
294
:
853
858
[PubMed]
8.
Ambros
V
.
MicroRNA pathways in flies and worms: growth, death, fat, stress, and timing
.
Cell
2003
;
113
:
673
676
[PubMed]
9.
Denli
AM
,
Tops
BB
,
Plasterk
RH
,
Ketting
RF
,
Hannon
GJ
.
Processing of primary microRNAs by the Microprocessor complex
.
Nature
2004
;
432
:
231
235
[PubMed]
10.
Han
J
,
Lee
Y
,
Yeom
KH
,
Kim
YK
,
Jin
H
,
Kim
VN
.
The Drosha-DGCR8 complex in primary microRNA processing
.
Genes Dev
2004
;
18
:
3016
3027
[PubMed]
11.
Joglekar
MV
,
Parekh
VS
,
Hardikar
AA
.
New pancreas from old: microregulators of pancreas regeneration
.
Trends Endocrinol Metab
2007
;
18
:
393
400
[PubMed]
12.
Lynn
FC
,
Skewes-Cox
P
,
Kosaka
Y
,
McManus
MT
,
Harfe
BD
,
German
MS
.
MicroRNA expression is required for pancreatic islet cell genesis in the mouse
.
Diabetes
2007
;
56
:
2938
2945
[PubMed]
13.
Melkman-Zehavi
T
,
Oren
R
,
Kredo-Russo
S
, et al
.
miRNAs control insulin content in pancreatic β-cells via downregulation of transcriptional repressors
.
EMBO J
2011
;
30
:
835
845
[PubMed]
14.
Harfe
BD
,
McManus
MT
,
Mansfield
JH
,
Hornstein
E
,
Tabin
CJ
.
The RNaseIII enzyme Dicer is required for morphogenesis but not patterning of the vertebrate limb
.
Proc Natl Acad Sci U S A
2005
;
102
:
10898
10903
[PubMed]
15.
Schonhoff
SE
,
Giel-Moloney
M
,
Leiter
AB
.
Neurogenin 3-expressing progenitor cells in the gastrointestinal tract differentiate into both endocrine and non-endocrine cell types
.
Dev Biol
2004
;
270
:
443
454
[PubMed]
16.
Srinivas
S
,
Watanabe
T
,
Lin
CS
, et al
.
Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus
.
BMC Dev Biol
2001
;
1
:
4
[PubMed]
17.
Zhong
L
,
Georgia
S
,
Tschen
SI
,
Nakayama
K
,
Nakayama
K
,
Bhushan
A
.
Essential role of Skp2-mediated p27 degradation in growth and adaptive expansion of pancreatic beta cells
.
J Clin Invest
2007
;
117
:
2869
2876
[PubMed]
18.
Nishimura
W
,
Kondo
T
,
Salameh
T
, et al
.
A switch from MafB to MafA expression accompanies differentiation to pancreatic beta-cells
.
Dev Biol
2006
;
293
:
526
539
[PubMed]
19.
Georgia
S
,
Bhushan
A
.
Beta cell replication is the primary mechanism for maintaining postnatal beta cell mass
.
J Clin Invest
2004
;
114
:
963
968
[PubMed]
20.
Lee
CS
,
Perreault
N
,
Brestelli
JE
,
Kaestner
KH
.
Neurogenin 3 is essential for the proper specification of gastric enteroendocrine cells and the maintenance of gastric epithelial cell identity
.
Genes Dev
2002
;
16
:
1488
1497
[PubMed]
21.
Hara
M
,
Wang
X
,
Kawamura
T
, et al
.
Transgenic mice with green fluorescent protein-labeled pancreatic beta-cells
.
Am J Physiol Endocrinol Metab
2003
;
284
:
E177
E183
[PubMed]
22.
Dahl
JA
,
Collas
P
.
A rapid micro chromatin immunoprecipitation assay (microChIP)
.
Nat Protoc
2008
;
3
:
1032
1045
[PubMed]
23.
Downey
T
.
Analysis of a multifactor microarray study using Partek genomics solution
.
Methods Enzymol
2006
;
411
:
256
270
[PubMed]
24.
Blum
B
,
Hrvatin
SS
,
Schuetz
C
,
Bonal
C
,
Rezania
A
,
Melton
DA
.
Functional beta-cell maturation is marked by an increased glucose threshold and by expression of urocortin 3
.
Nat Biotechnol
2012
;
30
:
261
264
[PubMed]
25.
Chong
JA
,
Tapia-Ramírez
J
,
Kim
S
, et al
.
REST: a mammalian silencer protein that restricts sodium channel gene expression to neurons
.
Cell
1995
;
80
:
949
957
[PubMed]
26.
Schoenherr
CJ
,
Anderson
DJ
.
The neuron-restrictive silencer factor (NRSF): a coordinate repressor of multiple neuron-specific genes
.
Science
1995
;
267
:
1360
1363
[PubMed]
27.
Palm
K
,
Belluardo
N
,
Metsis
M
,
Timmusk
T
.
Neuronal expression of zinc finger transcription factor REST/NRSF/XBR gene
.
J Neurosci
1998
;
18
:
1280
1296
[PubMed]
28.
Kreisler
A
,
Strissel
PL
,
Strick
R
,
Neumann
SB
,
Schumacher
U
,
Becker
CM
.
Regulation of the NRSF/REST gene by methylation and CREB affects the cellular phenotype of small-cell lung cancer
.
Oncogene
2010
;
29
:
5828
5838
[PubMed]
29.
Pattyn
A
,
Morin
X
,
Cremer
H
,
Goridis
C
,
Brunet
JF
.
The homeobox gene Phox2b is essential for the development of autonomic neural crest derivatives
.
Nature
1999
;
399
:
366
370
[PubMed]
30.
Pattyn
A
,
Goridis
C
,
Brunet
JF
.
Specification of the central noradrenergic phenotype by the homeobox gene Phox2b
.
Mol Cell Neurosci
2000
;
15
:
235
243
[PubMed]
31.
Apostolova
LG
,
Thompson
PM
,
Rogers
SA
, et al
.
Surface feature-guided mapping of cerebral metabolic changes in cognitively normal and mildly impaired elderly
.
Mol Imaging Biol
2010
;
12
:
218
224
[PubMed]
32.
Harris
KS
,
Zhang
Z
,
McManus
MT
,
Harfe
BD
,
Sun
X
.
Dicer function is essential for lung epithelium morphogenesis
.
Proc Natl Acad Sci U S A
2006
;
103
:
2208
2213
[PubMed]
33.
O’Rourke
JR
,
Georges
SA
,
Seay
HR
, et al
.
Essential role for Dicer during skeletal muscle development
.
Dev Biol
2007
;
311
:
359
368
[PubMed]
34.
Wilson
ME
,
Scheel
D
,
German
MS
.
Gene expression cascades in pancreatic development
.
Mech Dev
2003
;
120
:
65
80
[PubMed]
35.
Johnson
DS
,
Mortazavi
A
,
Myers
RM
,
Wold
B
.
Genome-wide mapping of in vivo protein-DNA interactions
.
Science
2007
;
316
:
1497
1502
[PubMed]
36.
Kalis
M
,
Bolmeson
C
,
Esguerra
JL
, et al
.
Beta-cell specific deletion of Dicer1 leads to defective insulin secretion and diabetes mellitus
.
PLoS ONE
2011
;
6
:
e29166
[PubMed]

Supplementary data