Neuronatin, a Downstream Target of BETA2/NeuroD1 in the Pancreas, Is Involved in Glucose-Mediated Insulin Secretion

The pancreas from e14.5 embryos was isolated in RNAlater (Ambion) and kept in Trizol at −80°C. Generation of the BETA2 null mutant and PCR primers for genotyping have been previously published (9). Each embryo was genotyped by PCR, and KO-positive DNA was further confirmed by Southern blot analysis. In all, 10–12 pancreases of the same genotype were pooled and total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) and purified on an RNeasy spin column (Qiagen, Valencia, CA). Total RNA from β-TC3, α-TC1, NIT, and ARIP cells was purified with Trizol. A total of six Affymetrix MG_U74Av2 chips were probed (three wild-type [WT] and three KO chips) with cRNA generated from 10 μg of total RNA. Hybridization and scanning of the chips were performed by the Molecular Genomics Core at the University of Texas Medical Branch (Galveston, TX). Software used for the analysis included DChip (www.dchip.org) and S+ArrayAnalyzer (Insightful).

Total RNA of each genotype was loaded on a denaturing agarose gel containing 0.8 mol/l formaldehyde and transferred to a BioDyne B nitrocellulose membrane. The efficiency of the transfer was assessed by methylene blue staining. This staining is reversible through incubation with the hybridization buffer and does not appear to interfere with the hybridization process. Probes were generated using a NEBlot random labeling kit (New England Biolabs), and hybridization was performed with UltraHyb solution (Ambion) at 50°C. Membranes were washed twice with 2 × sodium chloride−sodium citrate (SSC), 0.1% SDS at 50°C, and twice with 0.1 × SSC, 0.1% SDS (50°C), and then exposed to BioMax film. Densitometry quantification was performed using ImageJ software (National Institutes of Health).

Embryos were fixed overnight in formalin at 4°C and embedded in paraffin. Adult pancreases were fixed in Bouin fixation buffer. Sections were cut at 7 μm. In situ hybridization was performed as previously described using ³⁵S− or ³³P-UTP−labeled probes (20). For the immunocytochemistry studies, NIT cells were grown on microscope slides and fixed with methanol (−20°C) for 10 min at room temperature. After PBS washes, cells were permeabilized with 0.2% Triton-PBS (PBS-X) for 10 min.

Immunohistochemistry was performed as follows. Slides were dewaxed and washed in PBS (three times at 5 min). Microwave antigen retrieval was performed in Antigen Retrieval Citra solution (Biogenex). The slides were permeabilized with PBS-X for 10 min and blocked in 10% normal donkey or goat serum and 1% BSA for 30 min. Controls used to determine signal specificity included either omitting the primary antibody or using the appropriate species IgG. Primary antibodies were incubated overnight at 4°C in PBS containing 1% serum and 0.1% BSA. The dilution and detection of the primary antibodies are summarized in Table 1. Slides were washed in Tris-buffered saline with 0.1% Triton (TBS-X; Sigma) containing TBS-X (three times at 5 min) and incubated with the appropriate secondary antibodies for 30 min at room temperature. Slides were washed in TBS-X (three times at 5 min) and mounted with Slow-Fade with DAPI (Molecular Probes). Tyramide signal amplification was performed as described by the manufacturer (Molecular Probes). Slides were examined under epifluorescence using a Axioskop 2 microscope. Confocal microscopy images were captured with a Zeiss LSM 510 META Confocal Microscope. Images were captured using the AxioVision imaging software and further processed using Photoshop 7.01 (Adobe).

The siRNA for Nnat were designed by Dharmacon using their Smart Pool service. The Smart Pool siNNAT was introduced into NIT cells using the double electroporation protocol. On day 1, cells were electroporated and allowed to rest for 3 days; on the 3rd day, cells were re-electroporated and again allowed to rest for 3 days, after which a radioimmunoassay (RIA) was performed. Briefly, 10 million cells were resuspended in 1 ml of PBS. Smart Pool siNNAT was added to the cells at a final concentration of 100 nmol/l. One pulse of 0.4 mV (250 μF) was applied. Cells were then plated onto a 10-cm dish containing Dulbecco’s modified Eagle’s medium (10% FCS). The same electroporation protocol was repeated 3 days later. Cells were then plated at 500,000 cells per well in a 12-well plate in triplicate for further studies by RIA. Another 1 million cells were plated onto a well of a 6-well plate, and RNA was extracted from these cells for confirmation of siRNA knockdown by Northern blot.

The RIA for insulin was performed as described by Wang et al. (21). The RIA kit was obtained from Linco Research.

CHIP assays were performed as described in Yoon et al. (22). The CHIP assays were then performed with two antibodies against BETA2 (G20 and N19 from Santa Cruz Biotechnology). For PCR, 1 μl from a 50-μl DNA extraction and 30–35 cycles of amplification were used. The primers used for PCR were as follows: MS Insulin I-S-296, TCAGCCAAAGATGAAGAAGGTCTC; MS Insulin I-AS-126, TCCAAACAGTTGCCTGGTGC; NNAT-S-1, AGCGGACTCCGAGACCAGTA; NNAT-AS+205, ATGGATTAAGGGGTCCCGGG; NNAT-S-644, AAGCTCGCTCCTTTGGCACA; NNAT-AS-360, GGATGGAAAGCCGAAGTTACA; NNAT-S-1,201, TTGCCTTTTGCAAATAGTTCAGC; NNAT-AS-1040, CTTTCATAGACCCATTTCTGTTCAA; NNAT-S-2245, TCTCGGCAGACGTTAGGTGC; NNAT-AS-2082, GGAACGGAAACCTTGCACAA; NNAT-S-3005, ATACTGCACATCTCCCTCCTG; NNAT-AS-2750, TCCTTCGTTCCCTTTTGAGA; PCK-S-434, GAGTGACACCTCACAGCTGTGG; and PCK-AS-96, GGCAGGCCTTTGGATCATAGCC.

BETA2 and E47 were cotranslated in vitro using the TNT Quick Coupled Transcription/Translation Systems (Promega). We incubated 1 μl of the translation product with ∼50,000 cpm of Klenow-labeled, double-stranded oligonucleotides, corresponding to the putative E-boxes, in 15 μl of binding buffer (20 mmol/l Tris-HCl [pH 7.5], 5 mmol/l MgCl₂, 100 mmol/l KCl, 10% glycerol, 0.01% Triton X100, and 0.5 μg polydI-dC). Incubation was performed on ice and, when indicated, 2 μl (400 ng) of goat α-BETA2 (N19; Santa-Cruz) was included in the binding buffer. The binding reaction was loaded onto a 4% polyacrylamide gel with 0.5% glycerol in 0.5 × Tris borate EDTA buffer. The oligonucleotide sequences were as follows: GATCCTTCCGCGTGCTGCTGCAGGTGAGTATGTACCCGGGCTTT, +1/−205-S; GATCAAAGCCCGGGTACATACTCACCTGCAGCAGCACGCGGAAGGATC, +1/−205-AS; GATCCAAGATTCTCACAAAACAGCTGGCATGGGTTTAAAA, −644/−360-S; GATCTTTTAAACCCATGCCAGCTGTTTTGTGAGAATCTTG, −644/−360-AS; GATCCCACAATACCCAAACTCAGTTGGACCATTAAACC, −1,201/−1040-S; GATCGGTTTAATGGTCCAACTGAGTTTGGGTATTGTGG, −1,201/−1040-AS; and GATCCGGATCAGCAGATGGCCAGAGGGGCTA, RIPE3a-AS.

To further define the effects of the lack of BETA2 in the developing pancreas, we performed a detailed immunostaining study for important pancreatic transcription factors. We have previously shown that in the absence of BETA2, a decrease in endocrine cells is observed around e15.5 and culminates with an ∼50–60% decrease in endocrine cells at e17.5-P0 due to extensive apoptosis. Thus, e14.5 pancreases from WT and KO embryos were stained with BETA2, β-galactosidase, Pdx1, Ngn3, Pax6, Nkx6.1, and Nkx2.2. At this stage, BETA2 was detected in few cells adjacent to epithelial duct cells in the WT but not the KO pancreas (Fig. 1A and B) and, conversely, β-galactosidase, which replaces the BETA2 gene locus in the KO, was detected only in the KO and not the WT pancreas (Fig. 1C and D). We did not observe any major differences in staining between the WT and KO embryos in all the other transcription factors studied (Fig. 1E–N). In addition, a counting of Ngn3-positive (an upstream gene of BETA2) and Pax6-positive (a potential downstream gene of BETA2) cells did not reveal any statistical differences (data no shown). Furthermore, immunostaining for two cell adhesion molecules (Ep-CAM and E-CAD) did not reveal any difference between the WT and KO embryos at e14.5 (data not shown). These results suggest that e14.5 is an ideal age at which to identify potential target genes of BETA2 without the negative aspect of apoptosis.

DNA microarray analysis was performed to identify potential BETA2 target genes by comparing RNA samples from e14.5 BETA2 WT with those from KO pancreases. Consistent with the function of BETA2 in endocrine differentiation, BETA2, insulin, and glucagon transcripts were found to be downregulated, which seems to validate this approach for BETA2 target gene discovery. Among the genes identified, Nnat was dramatically reduced in the KO samples and was chosen for further studies, in light of its expression in neurons and β-cell lines.

The validity of Nnat as a target of BETA2 was assessed by in situ hybridization and Northern blot analysis. From in situ hybridization, Nnat transcripts were detected in WT pancreases from both e14.5 and e16.5, whereas no transcripts were observed in KO pancreases (Fig. 2A–D and I–L). In KO embryos, the absence of Nnat transcripts in the pancreas appears to be specific to this tissue, as Nnat was detected in the developing pituitary gland and cerebellum at levels comparable with those observed in WT embryos (Fig. 2E–H and data not shown). To confirm that the absence of Nnat transcript is not due to the absence of endocrine cells on the section analyzed, the adjacent section was hybridized with a probe encoding for GLUT2, a marker for endocrine cells in the pancreas. GLUT2 transcript was detected on both the WT and the KO pancreases (Fig. 2M−P), with no apparent difference in expression. In addition, Nnat transcripts were clearly absent from KO RNA samples at e16.5, as measured by Northern blot analysis (Fig. 3A). This decrease was paralleled by a marked reduction in insulin and glucagon transcripts, two known target genes of BETA2 (Fig. 3B), whereas the expression of the exocrine marker amylase was unaltered (Fig. 3C). Furthermore, transcripts for Nnat were also reduced in the eye of postnatal day 1 pups. These results indicated that Nnat is downregulated in two tissues where BETA2 has important developmental function.

A CHIP assay was performed to determine if Nnat is a direct target of BETA2. The cell line βTC3 was used because it expresses high levels of Nnat mRNA (Fig. 4A). The CHIP assay was performed with two antibodies against BETA2 (G20 and N19), with a similar outcome. Specific PCR amplification was obtained after precipitation with anti-BETA2 antibodies, but not when goat IgG was used. All regions selected for PCR contained potential E-box (CANNTG) binding sites. Only the proximal Nnat promoter was efficiently precipitated and amplified (Fig. 4B, lanes NNAT[−1, +205], NNAT[−664, −360], and NNAT[−1,201, −1,040]), whereas the distal Nnat promoter was not amplified (Fig. 4B, lanes NNAT[−2,245, −2082] and NNAT[−3,005, −2,750]). As a positive control, the insulin promoter, which contains BETA2 binding sites, was precipitated and amplified, whereas a nonregulated promoter, the PEPCK gene, was not. To further confirm the CHIP results, we performed gel shift assay using the in vitro–translated BETA2/E47 protein complex and the E-boxes from the three positive regions identified by the CHIP assay. Strong binding to the E-box located in the NNAT(−664, −360) region was seen, whereas weak or no binding was observed for the E-boxes present in the NNAT(−1, +205) and NNAT(−1,201, −1,040) regions (Fig. 4C). In addition, gel shift experiments with the E-boxes that were not positive in the CHIP assay were also negative for binding to BETA2 in the gel shift assay (data not shown). The presence of BETA2 in the retarded complex was confirmed by antibody supershift (Fig. 4C, “+” lanes), whereas control goat IgG did not (Fig. 4C, “−” lanes). It was interesting to note that the intensity of the binding for the NNAT (−664, −360) E-box was similar to that of the classical E-box present on the RIPE3a enhancer of the insulin promoter.

We next performed a detailed developmental analysis of Nnat expression in the pancreas by in situ hybridization and Northern blot. At e12.5, Nnat was detected in a few pancreatic cells (Fig. 5A), whereas the number of Nnat-expressing cells increased dramatically by e14.5 (Fig. 5B). Of interest, Nnat-expressing cells were found adjacent to ductal epithelial cells (Fig. 5B, red arrow) or in cell aggregates (Fig. 5B, black arrow), suggesting that Nnat is expressed in differentiating endocrine cells and early aggregating islet cells. Nnat expression gradually increases by e16.5 (Fig. 5C), paralleling the development of endocrine cells. By e18.5 (Fig. 5D), Nnat expression was restricted to the central core of the forming islet (data not shown). Immunohistochemical localization demonstrated the presence of NNAT in the central core of an islet of Langerhans in the adult pancreas; such a pattern indicates that it is located in β-cells similar to BETA2. Northern blot analysis using embryonic pancreatic and postnatal RNA samples was performed and confirmed the steady increase of Nnat during pancreas development (Fig. 5F). Thus, the expression profile of Nnat parallels BETA2’s profile and suggests a potential role for Nnat in the development of endocrine cells.

To gain further insight into the function of NNAT, we performed immunocytochemistry to localize the NNAT protein in the insulinoma cell line NIT, which expresses high levels of Nnat RNA (Fig. 4A). Using confocal microscopy, NNAT was localized to the cytoplasm (Fig. 6B). Insulin was also consistently colocalized to the cytoplasm (Fig. 6A). The merging of the two images suggests a potential colocalization of NNAT and insulin, but further experiments are required to definitely confirm the colocalization. Furthermore, attempts to detect NNAT proteins in the medium were unsuccessful, indicating that NNAT is unlikely to be a secreted protein (data not shown).

We next tested the potential role for NNAT in β-cell’s main function, that is, insulin secretion in response to glucose. To test our hypothesis, we performed siRNA experiments to knockdown Nnat in NIT cells. The introduction of siRNA against Nnat (siNNAT) into NIT cells led to the knockdown of Nnat mRNA when compared with controls (siGFP) (Fig. 7A). As a control for the specificity for the knockdown, Northern blot for insulin transcripts did not change between control (siGFP) and siNNAT (Fig. 7A).

Electroporation had no effect on NIT’s ability to secrete insulin, as unelectroporated cells responded to similar levels as the electroporated cells (Fig. 7B, compare electroporation control vs. untreated cells). When NIT cells were challenged with glucose, a 7- to 15-fold increase in secreted insulin was observed (Fig. 7B, siGFP, electroporation control, and untreated cells), whereas only a 2.7-fold increase was observed when Nnat was knocked down (Fig. 7B, siNNAT). Hence, it seems that the inhibition is specific to the downregulation of NNAT, rather than to nonspecific effects impacting on granule transport or exocytic events. Taken together, these results suggest an important function for NNAT in insulin secretion.

BETA2 is an important factor for pancreatic endocrine cell terminal differentiation and survival (9). BETA2 is believed to regulate genes important for mature β-cell function. It is not surprising that BETA2 is an important regulator of insulin and glucagon gene transcription (1,23). Furthermore, surviving KO mice have an impaired glucose tolerance test, suggesting a defect in either glucose sensing or insulin secretion (10). The importance of BETA2 is highlighted by the discovery that it is the sixth maturity-onset diabetes of the young (MODY6) gene identified (24). In addition, many genetic studies in various populations have linked mutations in BETA2 to susceptibility to type 1 (25–27) and type 2 (28,29) diabetes. Consequently, target genes of BETA2 will likely affect the etiology of diabetes linked to BETA2. To identify potential target genes of BETA2 in the pancreas, we performed a microarray screening of embryonic pancreatic RNA. Based on our previously described results, endocrine cell development is perturbed during late gestation (9); thus, we used e14.5 as a working time point for the microarray studies. We found 216 genes that varied by at least 1.5-fold, with 66 genes downregulated and 150 genes upregulated in the KO samples. We also found many metabolic hormones (insulin, glucagon, gastrin, and neuropeptide Y) downregulated in the KO pancreases. We observed that 11 transcription factors were higher in the WT samples, including BETA2, Nfatc1, Kruppel-like factor 7, Pbx1, and Paraxis. Of interest, Pbx1 was recently shown to be required for pancreas development (30). Numerous genes involved in cytoskeleton/intracellular/intracellular matrix formation were also downregulated in the mutant pancreases (e.g., lamin-B1, actinin-α1, matrilin 4); this observation is consistent with the high state of proliferation and migration that occurs at this age. Finally, many molecules involved in signaling were identified, including Notch1, IRS 2, Rap1, and Nnat. We were surprised to identify more genes upregulated in the KO samples acting on diverse biological pathways. Of interest, numerous genes involved in apoptosis were upregulated in the KO samples, such as Bnip2, Bnip3l, and Pdcd6/Alg2. The upregulation of pro-apoptosis genes is consistent with the KO phenotype.

The identification of Nnat as a target gene of BETA2 fits well with this premise, as it appears to play a role in insulin secretion. NNAT shows weak homology to phospholamban and PMP1, two proteolipid family members that function as a Ca²⁺-ATPase and H⁺ATPase, respectively (31,32).

In the pancreas, Nnat is present as early as e12.5 in a few cells and increases gradually; from e14.5 to e18.5, there is a dramatic expression in endocrine cells (Fig. 5A–D). This increase is reminiscent of the appearance of BETA2 and insulin in the pancreas (9). Furthermore, NNAT was detected by immunohistochemical methods in islets of the adult pancreas (Fig. 5F). NNAT was detected in the central core of the islet, which is indicative of its presence in mostly β-cells; however, its presence in the other endocrine cell types cannot be excluded. In addition, Nnat mRNA expression persists in the postnatal and adult pancreas (Fig. 5G).

The expression of Nnat was assessed in BETA2 WT and KO pancreas from e14.5 and e16.5 by in situ hybridization and Northern blot analysis. By in situ hybridization, Nnat was readily detected in the pancreas of WT embryos at e14.5 and e16.5, whereas it was undetectable in KO embryos (Fig. 2A–D and I–L). Similar results were obtained by Northern blot analysis (Fig. 3A). Furthermore, GLUT2 was detected in both WT and KO e16.5 embryos, thereby confirming the presence of endocrine cells in the KO sections (Fig. 2M−P). The decrease of Nnat observed by Northern blot analysis also parallels the decrease in insulin and glucagon transcripts, two known targets of BETA2, in the KO samples, whereas the level of amylase, a marker for exocrine cells, was unchanged in the KO pancreas (Fig. 3C). In addition, Nnat transcripts were also reduced in the developing eye (Fig. 3D), whereas no difference was observed in the developing cerebellum (data not shown). Although the function of NNAT in the retina is unknown, its temporal and expression patterns in the retina (33) suggest that Nnat is expressed in amacrine cells, where BETA2 is known to play a role in cell fate specification (34,35). The observation of the downregulation of Nnat in two BETA2-expressing tissues indicates the relevance of this target gene. Finally, the immunohistochemistry of numerous transcription factors important for endocrine cells genesis and/or differentiation does not suggest a general downregulation of endocrine cells at e14.5 (Fig. 1), but does indicate that the downregulation of Nnat is specific.

CHIP assay suggests that indeed Nnat is a direct target of BETA2 (Fig. 4B). BETA2 was able to bind to multiple regions proximal to the transcriptional start site, but failed to precipitate the more downstream chromatin regions NNAT(−2,245, −2082) and NNAT(−3,005, −2750). All the E-boxes located in the targeted regions contained the core consensus E-box CANNTG (Fig. 4D). Thus, other determinants, such as the sequence of the two variable nucleotides in the central core of the E box consensus and flanking sequences of the E-box, will likely confer specificity and affinity. Gel shift experiments were performed with the E-boxes from the three regions that were positive on the CHIP assay. Strong binding to the E-box located in the NNAT(−664, −360) region was observed, comparable with the classic RIPE3a binding site in the insulin promoter. Only weak or no binding was observed for the E-boxes present in the NNAT(−1, +205) and NNAT(−1,201, −1040) regions, respectively. The weak or lack of binding of these two E-boxes identified by the CHIP assay could be due to 1) the E-boxes and their flanking sequences, 2) the requirement for a specific interacting partner/environment present only in vivo, or 3) the limitation of the CHIP assay in defining the exact E-box when they are in close proximity (i.e., 400–800 bp), which is directly linked to the chromatin size generated by the sonication step. One interesting observation was the correlation between the CHIP and gel shift assay results, in that a strong binding in gel shift corresponded to an efficient CHIP assay (Fig. 4B and C, lanes NNAT[−644, −360] and insulin/RIPE3a).

Immunocytochemistry for NNAT was performed on NIT cells. NIT cells express high levels of Nnat mRNA (Fig. 4A). By confocal microscopy, insulin and NNAT were localized in the cytoplasm (Fig. 6A and B). Although the merge figure (Fig. 6C) suggests the possible colocalization of NNAT with insulin, the definite colocalization will require a more detailed analysis using electron microscopy. We tested if NNAT was important for secretion; to do so, we knocked down Nnat using siRNA. Knockdown of Nnat in NIT cell did not have a significant effect on basal insulin secretion, but did lead to impaired response in the presence of glucose (Fig. 4B). The result indicates a potential role for NNAT in insulin secretion in response to glucose. This is the first report of a physiological function for NNAT.

BETA2 is a transcription factor important for neuronal and endocrine cell terminal differentiation and/or survival. Thus, it is quite appropriate that Nnat is a target gene for BETA2, given that Nnat is expressed in both neurons and β-cells of the pancreas. NNAT does indeed appear to play a role in mature β-cells, whose main function is insulin secretion in response to glucose. The latter finding is consistent with the impaired glucose tolerance test observed in the surviving KO mice (10). It is interesting that in naturally occurring mice that lack Nnat (MatDp.dist2.T26H), there is a decrease in insulin staining with an accumulation of insulin signal adjacent to the nucleus in newborn pancreas when compared with controls (K.C., M.-J.T., C.V. Beechey, J. Peters, unpublished observations); this finding is suggestive of incorrect sorting or impaired trafficking of secretory vesicle. Furthermore, in a rat model with impaired glucose mediated insulin secretion, Nnat has been identified as downregulated in islets (36). The authors of the latter study overlooked the importance of such a finding, as no function was assigned to Nnat.

In conclusion, the present data show that Nnat is a direct target gene of BETA2 and, more importantly, its role in glucose-mediated insulin secretion. The potential role for NNAT in diabetes should be an attractive avenue for further studies.

M.-J.T. is supported by National Institutes of Health Grant HD-17379.

We thank Eric Némoz-Gaillard, Debra Bramblett, Luoping Li, and Sophia Tsai for a critical reading of the manuscript and helpful discussions.