MicroRNA Expression Is Required for Pancreatic Islet Cell Genesis in the Mouse

Mice were housed on a 12-h light/dark cycle in a controlled climate according to the University of California San Francisco regulations. Timed matings were carried out with e0.5 being set as midday of the day of discovery of a vaginal plug. Control mice in this study were double heterozygous for the Dicer1 (12) and Pdx1-Cre (3) alleles. CD-1 mice were obtained from Charles River Laboratories (Wilmington, MA).

A total of 118 embryos from 10 CD-1 mice were removed from pregnant mothers at e14.5, and embryonic pancreata were dissected and isolated from the spleen and other surrounding tissue. RNA was then isolated using Trizol and the manufacturer's protocols (Invitrogen, Carlsbad, CA), with the exception that the RNA pellet was not washed with 70% ethanol following isopropanol precipitation. We obtained 189 μg of RNA from 118 pancreata. miRNA cloning was carried out as previously described by Lau et al. (13). Briefly, linkers were then ligated onto the 3′ and 5′ ends of size-fractionated RNA, and RNA was reverse transcribed and amplified using PCR. The PCR products were restriction digested with BanI and concatermized with T4 DNA ligase, and gel-purified fragments ranging from 600 to 800 bp were TOPO cloned (Invitrogen). Plasmid DNA was isolated from 100 colonies and sequenced using M13 primers. Sequences were deconvolved using a Java applet (available upon request from the authors). Sequences were then aligned against the miRNAome (available at http://microrna.sanger.ac.uk/sequences/), and unknown sequences were aligned against genome. Regions of high sequence homology were then folded using RNA shapes (14), and those that folded into hairpin loops were considered miRNAs.

Embryonic tissue was harvested and immediately fixed at 4°C with 4% paraformaldehyde. Tissues were then dehydrated through 50, 70, 95, and 100% ethanol and then through xylene. Tissues were then impregnated with paraffin (Paraplast; Fisher Life Sciences) at 60°C under a vacuum and then embedded in cassettes. Tissue was then cut into 5 μmol/l sections and adhered to glass slides (Superfrost Plus; Fisher Life Sciences). For immunoanalyses, tissue was first deparafinized with xylene and then rehydrated through a reverse ethanol series to ddH₂0. Antigen retrieval was then carried out by microwave in a citrate buffer (Biogenex, San Ramon, CA). Tissues were then blocked at room temperature with 5% goat serum and goat anti-mouse IgG (1:30; MP Biomedicals/Cappel, Solon, OH) in PBS. Incubations with the primary antibodies were performed overnight at 4°C using standard techniques in PBS containing 1% goat serum with the following primary antisera: 1:2,000 guinea pig anti–pdx-1 (15), 1:3,000 guinea pig anti-glucagon (Linco, St. Charles, MO), 1:4,000 mouse anti-glucagon (Sigma-Aldrich, St. Louis, MO), 1:3,000 guinea pig anti-insulin (Linco), 1:2,000 guinea pig anti-neurogenin3 (15), 1:200 hamster anti-mucin (NeoMarkers, Fremont, CA), 1:200 mouse anti-synaptophysin (Biogenex), 1:1,000 rabbit anti-amylase (Sigma), 1:1,000 rabbit anti-Hes1 (a gift from T. Sudo, Toray Scienfic, Osaka, Japan [16]), 1:150 rabbit anti–activated caspase-3 (Cell Signaling, Danvers, MA), 1:200 rabbit anti-MKi67 (Novocastra, Newcastle upon Tyne, U.K.), or 1:100 rabbit anti-Dicer (Santa Cruz Biotechnology). Secondary antibodies were obtained from Jackson Immunoresearch (West Grove, PA) and used at 1:200 (fluorescein isothiocyanate) to 1:400 (Cy3) dilutions, and coverslips were mounted with Vectashield (Vector Laboratories, Burlingame, CA). Slides were imaged on a Leica TCS SL confocal microscope using sequential scanning of dyes. Morphometric analyses were done by serially sectioning the entire pancreas followed by staining and counting every 12th section for e18.5 embryos or every 8th section for e15.5 embryos. DAPI-stained cell nuclei were counted using the cell-counting application, and cytoplasmic staining was quantified using the integrated metamorphic analysis, both in the Metamorph suite of programs (version 7; Molecular Devices, Union City, CA). Statistical analyses were done using PRISM (version 4; Graphpad Software, San Diego, CA), and either t tests or nonparametric Mann-Whitney U tests were used. P values of <0.05 were considered statistically significant.

Embryos (e14.5) were harvested and fixed overnight at 4°C in 4% paraformaldehyde. Following fixation, embryos were briefly washed with PBS and transferred to 30% sucrose in diethylpyrocarbonate-treated PBS and kept for an additional 24 h at 4°C. Embryos were then mounted in Tissue-Tek O.C.T. (Sakura Finetek, Torrance, CA) and frozen on dry ice. Embryos were sectioned with a cryostat into 10-μm sections and adhered to slides (Superfrost, Fisher Scientific), and slides were kept at −80°C until use. Slides were removed from freezer and dried for 1 h at room temperature. Sections were then refixed for 10 min at room temperature in 4% paraformaldehyde and then washed with RNase-free PBS. Slides were then incubated for 2 min with 1 μg/ml proteinase K (Roche Applied Science, Indianapolis, IN) in 50 mmol/l Tris, pH 7.5, containing 5 mmol/l EDTA. Slides were then washed again in PBS and acetylated (1.2% triethanolamine, 0.018N HCl, and 0.25% acetic anhydride) for 10 min at room temperature. Slides were then washed again before prehybridization in 50% formamide, 5 × sodium chloride–sodium citrate, 0.1% Tween-20, 9.2 mmol/l citric acid, 50 μg/ml heparin, and 500 μg/ml yeast transfer RNA for 2 h at 60°C. Antisense locked nucleic acid (LNA) probes were ordered from Exiqon (Vedbaek, Denmark) and were subsequently labeled using the DIG Oligonucleotide 3′-End Labeling Kit (Roche Applied Science) and the manufacturer's protocol. Labeled LNAs were then diluted to 50 μl with RNase-free water and separated from unincorporated label using Microspin G-25 columns (Amersham Biosciences, Piscataway, NJ). A total of 1 pmol of DIG-labeled LNA was then diluted into 250 μl of hybridization buffer, briefly heated to 80°C, and then iced. Labeled probes were then applied to the slides and allowed to hybridize overnight at 60°C. Slides were then washed for 3 h at 60°C in 0.2 × sodium chloride–sodium citrate and incubated with alkaline phosphatase–conjugated sheep anti-DIG antibody (Roche Applied Sciences) in 100 mmol/l Tris, pH 7.5, 150 mmol/l NaCl with 1% goat serum (Invitrogen) overnight at 4°C. Slides were then washed extensively in the aforementioned buffer and then equibrated with 100 mmol/l Tris, pH 9.5; 100 mmol/l NaCl; and 50 mmol/l MgCl2 before alkaline phosphatase reaction in the same buffer with 1:50 dilution of NBT/BCIP stock (Roche Applied Sciences) for 1–3 days. For Pdx1 staining, sections were then washed with PBS and endogenous peroxidases were quenched with a 15-min incubation in methanol containing 0.6% hydrogen peroxide. Slides were then incubated with a 1:3,000 dilution of guinea pig anti–Pdx-1 overnight at 4°C. Slides were then washed and incubated with 1:200 dilution of biotinylated goat anti–guinea pig secondary antibody (Vector Labs, Burlingame, CA). Slides were then incubated with ABC solution and then reacted with diaminobenzidine (VECTASTAIN; Vector Labs). Sections were dehydrated and coverslips mounted and imaged on a Zeiss Axio Imager.A1 equipped with an AxioCam MrC5 and using Zeiss AxioVision software (Thornwood, NY).

Embryonic pancreata were harvested at e12.5 and immediately placed in 0.5 ml of Trizol (Invitrogen), and RNA was extracted using the manufacturer's protocols. A total of 250 ng of total RNA was subjected to reverse transcription using random hexamers, Superscript III (Invitrogen), and the manufacturer's suggested reaction conditions. Complementary DNA was then amplified in quadruplicate using Taqman with an intron-spanning primer–probe set for Hes1 (forward, AAG GCA GAC ATT CTG GAA ATG AC; reverse, CGC GGT ATT TCC CCA ACA C; and probe, AGA TGA CCG CCG CGC TCA GC). Hes1 expression was normalized to Taqman quantified mouse glucuronidase expression in the individual pancreatic cDNA samples (17).

To determine their expression profile and identify novel miRNAs, small RNAs were cloned and sequenced from the pancreata of e14.5 mouse embryos using standard miRNA cloning and sequencing protocols (13). In total, 107 known mouse miRNAs were cloned, five of which were evolutionarily conserved but had not previously been found in the mouse (online appendix Table 1 [available at http://dx.doi.org/10.2337/db07-0175]). Unknown sequences were then aligned with the genome, and the surrounding, conserved genomic regions were folded into hairpins (online appendix Table 2). Using this approach, 18 novel mouse miRNAs were discovered, including some that lie in potential multicystronic miRNA transcripts (online appendix Table 2). In addition, significant differences in a number of miRNA sequences were observed from those annotated in miRBase (18). For example, multiple copies of mmu-miR-720 and mmu-miR-411 differed significantly from the sequences in miRBase (Table 1). Differences identified in the 5′ end of several miRNA are of particular significance because the majority of miRNA target prediction algorithms rely on Watson-Crick base pairing of nucleotides 2–7 from the 5′ end of the miRNA; thus, one –or two–base pair differences at the 5′ end could completely change the target profile of any miRNA (19,20).

Many of the miRNAs that were cloned here have not been previously detected at high levels in other tissues; therefore, we hypothesized that they might play specific roles in pancreas development. To assess the expression domain of these miRNAs during pancreas development, in situ hybridization with highly specific LNA probes was carried out for a few of these highly expressed miRNAs (Fig. 1) (21). We first tested mmu-miR-375 because it was overrepresented in our cloning experiment (online appendix Table 1) and has previously been demonstrated to be expressed in, and important for, pancreatic β-cell function. In situ analyses showed localization of miR-375 in the pancreatic ductal epithelium (i.e., the progenitor cells from which the endocrine and exocrine cells differentiate) at e14.5 (Fig. 1A and B). Furthermore, Fig. 1B demonstrates colocalization of miR-375 with Pdx-1 in the e14.5 pancreas; thus, in addition to regulating insulin secretion in mature β-cells (22), miR-375 expressed in the progenitor cells may also be important during normal development. We also performed in situ analyses for three other overrepresented miRNAs. Both miR-503 and miR-541 were expressed in a pattern similar to that in miR-375 in progenitor cells in e14.5 pancreata, while miR-214 was expressed throughout the gut (Fig. 1C–H).

To determine the importance of miRNAs in pancreatic development, the second RNase III domain of Dicer1 (3,12) was conditionally ablated in all pancreatic cells starting at e9.5 using Pdx1-Cre–mediated deletion (3). Biogenesis of miRNAs requires the sequential activity of two double-stranded RNA–specific endonucleases, Drosha and Dicer1, that act in the nucleus and cytosol, respectively (23). In mammals, removal of the premiRNA hairpin loop by a single Dicer homolog is a prerequisite for interaction of the mature sequence with RNA-induced silencing complex. Therefore, loss of the RNase III domain of Dicer1 blocks the formation of miRNAs (12,24).

The pancreas-specific Dicer1 knockout mice survived until birth but failed to grow and died by P3 (Fig. 2A and B). Pancreata from e18.5 null mice had a dramatic reduction in the ventral pancreas, as well as a reduction in the overall epithelial contribution to the dorsal pancreas (Fig. 2D). The remaining tissue in the Dicer1-null pancreata was comprised of clumps of disorganized epithelium interspersed with large vacuous areas and dilated ducts filled with impactions of debris (Fig. 2F and H).

Cell type–specific antisera were utilized to determine the effects of Dicer1 loss on specific pancreatic lineages. Costaining of the e18.5 pancreas for synaptophysin, an endocrine marker, and the exocrine enzyme amylase revealed reductions in both the number of stained cells and the intensity of staining for both markers (Fig. 3A and B). Immunoreactivity for the duct marker mucin1 remained unchanged in the knockout animal at this stage, but the debris filling the dilated ducts stained strongly for both mucin and amylase (Fig. 3B and J, arrow heads). Staining for specific hormones demonstrated a significant loss (94%) (Fig. 3K) of insulin-producing β-cells in the knockout animals and 79, 95, and 86% losses of α, δ, and pancreatic polypeptide cells, respectively (Fig. 3K), but no compensatory increase in ghrelin-producing ε-cells (Fig. 3C–J) (25). The differences in the reductions of the various endocrine cell types were significant, as demonstrated by significant changes in the ratios of distinct endocrine cell types in the Pdx-Cre Dicer1^flox/flox relative to their wild-type littermates (Fig. 3L). Pdx-1 immunoreactivity, which is restricted predominantly to β-cells at this stage, was also decreased within e18.5 knockout pancreata (cf. Fig. 3E and F).

To distinguish between a role in β-cell genesis and in maintenance or proliferation of mature β-cells, we also specifically removed Dicer1 from differentiated β-cells by crossing the Dicer1^flox mice with mice carrying a Cre allele driven by the rat insulin promoter 2 (RIP2) (26) (Fig. 4). β-Cell–specific Dicer1 knockout animals survived and their islets cells appeared morphologically normal at 8 months of age (cf. Figs. 4A and B). In addition, to exclude the possibility that Dicer1 loss specifically impacts the maintenance or proliferation of mature β-cells in the embryo, β-cell counting was performed also at e18.5. Control and RIP2-Cre Dicer^flox/flox animals at this developmental stage contained equal numbers of insulin-immunoreactive β-cells (Fig. 4C). Finally, we also conditionally ablated Dicer1 using a neurogenin3-Cre transgene, which causes recombination in all endocrine cells. These animals also exhibit normal islet cell hormone staining at e18.5 (data not shown).

To identify the origins of the Pdx-1–mediated Dicer deletion phenotypes that were observed at e18.5, earlier time points were examined. At e12.5 staining for Pdx-1 and glucagon, markers of undifferentiated pancreatic epithelial progenitors and the earliest differentiated cells in the pancreas, respectively, was not different between the mutant and wild-type pancreata (Fig. 5A and B). However, by e15.5 knockout embryos lacked Dicer1 immunoreactivity in the Pdx-1–expressing cells (online appendix Fig. 1) and had significant reductions in pancreatic size and ductal branching (Fig. 5C–H). At this stage in development, Pdx-1 antisera weakly stains the majority of cells lining the ducts and strongly stains the nascent β-cells, and neurogenin3 expression marks the committed endocrine progenitors. In the knockout animals, a decrease in the number of cells staining brightly for Pdx-1 (Fig. 5D) and a significant (89%) reduction in neurogenin3-expressing cells was observed (Fig. 5D, F, H, and I), indicating a defect in the endocrine differentiation program. Ductal Pdx-1 and mucin1 staining was not grossly altered in the Dicer1-null animals at e15.5 (Fig. 4D and H).

Notch signaling inhibits the expression of neurogenin3 and thereby constrains endocrine differentiation in the developing pancreas in a manner analogous to lateral inhibition in the developing nervous system (5,27,28). The notch target gene Hes1 is expressed in the majority of early pancreatic epithelial cells and strongly represses Neurog3 gene transcription within those cells (8,16). We hypothesized that specific miRNAs may activate endocrine differentiation by inhibiting components of the Notch signaling pathway, thereby relieving the constraint on neurogenin3 expression. We tested this hypothesis by assessing Hes1 expression using Taqman and found a modest increase in Hes1 mRNA (35%) at e12.5 (Fig. 6K). Because Hes1 is expressed in the pancreatic mesenchyme as well as the Pdx-1–expressing epithelium, and because Hes1 may be a target for translational regulation, we also stained e12.5 pancreata for Hes1 and observed an increase in Hes1 expression specifically within the Pdx-1 expression domain in Dicer1-deficient pancreata (Fig. 6).

In addition to the loss of neurogenin3, starting at e15.5 there was an overall decrease in cells derived from the pancreatic epithelium in the knockout animals (Fig. 2 and data not shown). Since miRNAs can regulate apoptosis during organ development (12,29,30), rates of programmed cell death and cell proliferation were assessed in the pancreatic-specific Dicer1 knockout animal, using specific antisera for activated caspase 3 and Ki-67 (MKi67), respectively. A clearly significant increase (4.6-fold) was observed in total apoptotic cells at e15.5 in pancreata from knockout animals, and this increase appeared to be limited to the epithelial cells, as apoptosis was unchanged in the surrounding mesenchymal cells, which still expressed Dicer (Fig. 6H–J). Cell proliferation was unchanged (online appendix Fig. 2; counting data not shown).

Here, we demonstrated that the developing pancreas expresses a unique complement of miRNAs and that Dicer1, the enzyme that generates these miRNAs, is essential for normal pancreatic development. Many of the miRNAs that were highly expressed in the developing pancreas have not been previously cloned in high numbers or from specific tissues (e.g., miR-503 and miR-541). High expression in the developing pancreas could suggest that these miRNAs play important roles in the developmental program; although, to date we have not identified any targets of these genes with known roles in pancreas development. In contrast, the novel miRNAs identified in this study all had fairly low expression levels as gauged by the number of clones obtained. Despite their low overall abundance, these miRNAs could still play important roles at specific target genes or in cell types at low abundance in the e14.5 pancreata.

Although exocrine and duct cells were also affected, loss of miRNA processing uniquely impaired the development of the endocrine lineage, especially the β- and δ-cells. Two lines of evidence suggest that miRNAs are important for β-cell genesis rather than β-cell maintenance. First, a dramatic reduction in the number of neurogenin3-positive endocrine progenitor cells from which all β-cells are derived was observed at e15.5. Consistent with a reduction in islet cell progenitors, glucagon-, somatostatin-, and pancreatic polypeptide–expressing endocrine cells were also reduced, although ghrelin-expressing cells were not. The relatively less severe α-cell phenotype may be solely a timing issue, as Pdx-1–mediated loss of Dicer1 may not be complete until after the first glucagon-positive cells differentiate, and Pdx-1–null animals do have glucagon-positive cells (31). Alternatively, miRNAs may act downstream of neurogenin3 to influence cell fate decisions, and, in the absence of miRNAs, the few remaining neurogenin3-expressing progenitors may preferentially differentiate into α- and ε-cells.

Second, when Dicer1 was ablated using the RIP2-Cre (or the Neurog3-Cre) transgene (Fig. 4), we saw little effect on pancreatic islet cell morphology and no apparent effect on β-cell development (Fig. 4C) or β-cell maintenance. A previous study (22) demonstrated that miR-375 normally constrains β-cell insulin secretion in a mytrophin-dependent fashion. Our data here focused on the role of miRNAs in the formation of β-cells and are not at odds with these previous data on islet function. We have shown that islet and β-cell development remains unaffected in the RIP2-Cre Dicer1^flox/flox mice, but we have not directly tested the glucose-stimulated insulin secretory response of these islets. It is possible that there could be a functional defect in the islets of these animals, although given the proposed inhibitory role of miR-375 on insulin secretion such defects might not result in glucose intolerance or overt diabetes. Furthermore, glucose metabolism defects might be confounded by the expression of the RIP2-Cre transgene in cells in the hypothalamus that also impact glucose metabolism (26).

The large number of miRNAs expressed in the developing pancreas, and the large number of potential targets for each of these complicate the task of identifying the miRNAs that normally maintain neurogenin3 expression and endocrine cell generation, but the increase in the neurogenin3 inhibitor Hes1 suggests that Hes1 or upstream genes in the notch signaling pathway could be miRNA targets in pancreatic progenitor cells. Previous studies (32,33) have implicated Hes1 protein as a target of miR-23a, and miR-23a was expressed in the developing pancreas (Table 1). (Although, this article was withdrawn because the authors mistakenly used the wrong gene sequence for their analyses, the observation that miR-23a can influence the expression of Hes1 remains a possibility from this publication.) In addition, it seems likely that miRNA may target other notch pathway components upstream of Hes1, yielding an increase in Hes1 expression in the absence of Dicer1.

In summary, these data demonstrate for the first time that a unique set of miRNAs emerge during pancreas development and regulate pathways important for normal ductal, exocrine, and endocrine development. Exploration of the roles of individual miRNA will provide a more complete understanding of pancreatic development and the differentiation of the distinct cell types, especially β-cells, and help in developing methods for generating these cells for therapeutic uses.

This work was supported by Larry L. Hillblom Foundation Grant no. 2002/1(E), Juvenile Diabetes Research Foundation (JDRF) Innovative Grant no. 5-2006-125, and Cores Laboratories supported by the National Institutes of Health Grant P30 DK63720. Personnel were supported by JDRF Fellowship Award 3-2004-276 (to F.C.L.) and an American Diabetes Association Mentor-Based Postdoctoral Fellowship Grant (to Y.K.).

We thank Dr. Doug Melton, Dr. Mark Magnuson, and Dr. Pedro Herrera for generously providing the Pdx1-Cre, RIP2-Cre, and Neurog3-Cre mice, respectively. We thank members of the German, Hebrok, and McManus Laboratories for help, advice, and criticism.