|Home | About | Journals | Submit | Contact Us | Français|
Although expression of the mammalian RNA-binding protein HuD was considered to be restricted to neurons, we report that HuD is present in pancreatic β cells, where its levels are controlled by the insulin receptor pathway. We found that HuD associated with a 22-nucleotide segment of the 5'-untranslated region (UTR) of preproinsulin (Ins2) mRNA. Modulating HuD abundance did not alter Ins2 mRNA levels, but HuD overexpression decreased Ins2 mRNA translation and insulin production, and conversely, HuD silencing enhanced Ins2 mRNA translation and insulin production. Following treatment with glucose, HuD rapidly dissociated from Ins2 mRNA and enabled insulin biosynthesis. Importantly, HuD-knockout mice displayed higher insulin levels in pancreatic islets, while HuD-overexpressing mice exhibited lower insulin levels in islets and in plasma. In sum, our results identify HuD as a pivotal regulator of insulin translation in pancreatic β cells.
Changes in circulating glucose modulate insulin production by the β cells of the pancreatic islets of Langerhans. In turn, insulin influences glucose uptake in insulin-sensitive peripheral tissues such as fat and muscle, and maintains glucose homeostasis (Rhodes and White, 2002). As a key metabolic factor, insulin levels are tightly regulated by different mechanisms. Insulin is produced by proteolytic cleavage of preproinsulin in pancreatic β cells. Preproinsulin is encoded by insulin mRNA, a highly abundant transcript in β cells (>30% of total mRNA) with an exceptionally long half-life (>24 h) due to the presence of a pyrimidine-rich stretch in its 3'-untranslated region (UTR) (Itoh and Okamoto, 1980; Goodge and Hutton, 2000). Tillmar and Welsh (2002) identified the RNA-binding protein (RBP) polypyrimidine tract-binding protein (PTB) as being responsible for associating with the pyrimidine-rich stretch in insulin mRNA and contributing to its high stability. Increased glucose availability enhanced PTB binding to insulin mRNA and elevated its levels; hours later, insulin mRNA was also transcriptionally upregulated (Jahr et al., 1980).
However, in response to acute elevations in circulating glucose, the necessary and timely rise in insulin production is primarily controlled by rapid increases in the translation of insulin mRNA in β cells. Wicksteed and coworkers (2001) reported that insulin translation was regulated through the cooperative action of a stem-loop in the 5'UTR and the conserved UUGAA sequence in the 3'UTR. A 9-nt element present in the insulin 5'UTR was shown to be responsible for the glucose-dependent translational increase in insulin production (Wicksteed et al., 2007). A 29-nt long element within the rat insulin 5'UTR was also found to contribute to the glucose-triggered translational upregulation (Muralidharan et al., 2007). However, the specific factor(s) that associate with these elements were unknown.
Here, we identify HuD (human antigen D) as an RBP that binds to insulin mRNA and controls its translation. Like two other Hu family members (HuB and HuC), HuD was believed to be expressed specifically in neurons, while the remaining member, HuR was ubiquitous (Hinman and Lou, 2008). However, a recent survey of HuD expression in different tissues (Abdelmohsen et al., 2010), unexpectedly revealed HuD expression in pancreatic β cells. Hu proteins have three RNA recognition motifs (RRMs) through which they associate with mRNAs bearing specific sequences that are often AU- and U-rich. HuD bound to the 3'UTR of target mRNAs and stabilized them, as shown for p21, tau and GAP-43 mRNAs (reviewed in Hinman and Lou, 2008). HuD also modulated target mRNA translation; for example, interaction of HuD with the p27 mRNA disrupted an internal ribosome entry site (IRES) and inhibited p27 translation (Kullmann et al., 2002), while HuD enhanced the stability and translation of Nova1 mRNA (Ratti et al., 2008). Despite the short and unstructured 5'UTRs of the human insulin (INS) mRNA and the mouse ortholog (Ins2 mRNA), HuD binding to the Ins2 5'UTR repressed Ins2 mRNA translation and decreased insulin production. Accordingly, HuD knockout mice expressed higher levels of insulin in β cells, while HuD-overexpressing mice expressed lower insulin levels in β cells and in the circulation.
Immunostaining of human and mouse pancreatic sections detected HuD in insulin-producing, β cells (Fig. 1A); HuD was also expressed in brain, but not in other mouse tissues (Fig. 1B, Fig. S1A,C). By Western blot analysis, HuD levels in immortalized β cells isolated from the pancreas of wild-type (βIRWT) mice were significantly higher and more glucose-inducible than those in β cells isolated from an insulin receptor (IR)-null (βIRKO) mouse (Fig. 1C) (Assmann et al., 2009; Kim et al., 2011); ectopic IR re-expression in βIRKO cells restored HuD abundance under conditions of low glucose and low serum (Fig. 1D). Treatment of βIRWT cells with insulin similarly elevated HuD levels in a dose-dependent manner (Fig. 1E). In mouse insulinoma βTC6 cells, silencing the insulin receptor substrate 2 (Irs2), a downstream effector of IR signaling, lowered HuD levels (Fig. 1F). Likewise, silencing or inhibiting Akt, a kinase that functions downstream of Irs2, also lowered HuD levels (Fig. 1G,H), as did inhibiting the Akt kinase PI3K using LY294002 (Fig. H). Akt phosphorylates and thereby inactivates the transcriptional repressor FoxO1; in keeping with the putative FoxO1 binding site in the HuD promoter (492 bp from the transcription start site), chromatin immunoprecipitation (ChIP) analysis in βTC6 cells revealed an interaction of FoxO1 with the HuD promoter. FoxO1 showed greater association with the HuD promoter in conditions of low glucose and reduced association in high glucose (Fig. 1I. Fig. S1B). As shown, silencing FoxO1 in βIRWT and βTC6 cells augmented HuD expression levels (Fig. 1J,K). Conversely, overexpression of FoxO1 reduced HuD production, while overexpression of the transcription-deficient FoxO1(H215R-537), bearing a mutation in the DNA binding domain (H215R) and C-terminal deletion in the transactivation domain (Zhang et al., 2011) did not (Fig. 1L). Together, these data indicate that HuD is expressed in β cells under control of the pathway IR → Irs2 → PI3K → Akt FoxO1 (→) HuD (Fig. 1M).
We tested the interaction of HuD with the mRNAs encoding mouse insulin (Ins1 and Ins2) in βTC6 cells. By ribonucleoprotein (RNP) immunoprecipitation analysis (RIP) of βTC6 cells, Ins2 mRNA (the most abundant insulin-encoding transcript) was significantly enriched in HuD IP compared with IgG IP (Fig. 2A). To map the RNA region of association, biotinylated segments spanning the 5'UTR, CR, and 3'UTR of Ins2 mRNA were synthesized (Fig. 2B), and the RNP complexes of HuD and biotinylated RNAs was detected by biotin pulldown analysis using streptavidin-coated beads. As shown in Fig. 2C, HuD interacted with the 5'UTR of the Ins2 mRNA, but not with the Ins2 CR or 3'UTR or with a negative control transcript (spanning the 3'UTR of GAPDH mRNA). Biotin pulldown analysis of other RBPs, including HuR, TIAR, hnRNP K, NF90 (Fig. S2A), as well as hnRNP C and FMRP (not shown), did not identify other Ins2 mRNA-interacting RBPs. The fact that argonaute IP failed to show interaction with Ins2 mRNA by either biotin pulldown or RIP (not shown) suggested that microRNAs were unlikely to be major regulators of insulin production. Further subdivision of the 5'UTR revealed that segment 5'D, spanning positions 52–73 and showing high conservation in human and rodents, specifically interacted with HuD (Fig. 2C,D, Fig. S2B). This interaction was specific, as endogenous HuD did not bind to mutant biotinylated 5'D (mut1–mut4) RNAs, none of which showed an affinity for HuD (Fig. 2E); similarly, incubation with recombinant, purified GST-HuD showed selective binding to 5'D, but not to mutant (mut1–mut4) biot-Ins2 RNAs (Fig. 2E).
To test if the interaction of HuD with the Ins2 5'UTR was functional, we prepared reporter constructs derived from plasmid pEGFP in which the entire 5'UTR [p(5')EGFP], a single copy of fragment D [p(5'D)EGFP] or 3 copies of fragment D [p(3×5'D)EGFP] were inserted before the translation start site of the EGFP coding region (Fig. 2F, left). After transfection of each reporter construct into βTC6 cells expressing normal (Ctrl siRNA) or silenced HuD (HuD siRNA), EGFP expression was assessed by Western blotting. As observed, silencing HuD selectively increased EGFP production in p(5')EGFP, p(5'D)EGFP, and in p(3×5'D)EGFP, but not in the control reporter group (pEGFP). This increase in EGFP protein (Fig. 2F, `relative to each Ctrl') did not arise from changes in EGFP mRNA (Fig. 2G). Moderate differences in transfection of each plasmid led to differences in basal expression of reporter EGFP mRNA and EGFP protein (Fig. 2F `relative to pEGFP Ctrl', Fig. 2G, inset graph). The `translation efficiency index' (which compares changes in protein relative to changes in mRNA) supports the view that translation was robustly increased in HuD siRNA cells (Fig. S2D).
Likewise, endogenous Ins2 mRNA levels in βTC6 cells did not change after silencing HuD, while expression of the encoded insulin precursor (proinsulin) increased markedly (Fig. 2H versus 2I) without changes in proinsulin protein stability (Fig. S2C). These data indicate that HuD associates with the Ins2 mRNA 5'UTR, and that this interaction represses insulin production, likely by inhibiting its translation.
More direct evidence that HuD represses translation of the Ins2 mRNA was obtained by polysome analysis in βTC6 cells expressing different HuD levels. Cytoplasmic extracts from control and HuD-silenced cells were fractionated through sucrose gradients, with the lightest components sedimenting at the top (fractions 1,2), small (40S) and large (60S) ribosomal subunits and monosomes (80S) in fractions 3–5, and progressively larger polysomes, ranging from low- to high-molecular-weight (LMWP, HMWP) in fractions 6–12 (Fig. 2J, left). While in Ctrl siRNA cells, polysome-associated Ins2 mRNA peaked in fractions 6 and 7, silencing HuD increased the peak size of Ins2 mRNA polysomes to fraction 8 (Fig. 2J, right). The distribution of the housekeeping Actin mRNA largely overlapped between the two groups (Fig. 2J, right). Additional evidence that HuD modulated insulin translation was gained from nascent translation analysis. The incorporation of 35S-Met and 35S-Cys into newly synthesized proinsulin during a brief time period (15 min), immediately followed by IP using anti-insulin antibody, showed enhanced translation of proinsulin in the HuD siRNA group, while translation of the housekeeping protein GAPDH was unchanged between the two groups (Fig. 2K; signals were charactersitically faint, reflecting the short labeling time). Together, these findings strongly indicate that HuD represses Ins2 production by reducing Ins2 mRNA translation.
Analysis of HuD-Ins2 mRNA complexes in βTC6 cells revealed a rapid and robust release of Ins2 mRNA from HuD complex within 30 min of glucose stimulation (25 mM glucose; Fig. 3A). In keeping with the view that HuD represses insulin production, glucose treatment (15 mM, 20 min) elevated proinsulin abundance in control cells (vector), but this increase was largely lost in cells overexpressing HuD (Fig. 3B, left; Fig. S2A). These changes occurred without changes in Ins2 mRNA levels (Fig 3B, right) but were accompanied by changes in nascent translation of proinsulin (Fig. 3C). Conversely, glucose treatment (15 mM, 20 min) elevated proinsulin abundance in cells transfected with Ctrl siRNA, while HuD-silenced cells expressed significantly higher levels of proinsulin, as assessed by Western blotting (Fig. 3D) and by ELISA measurement of intracellular and secreted insulin (Fig. 3E). Again, changes in proinsulin and insulin production occurred without changes in Ins2 mRNA levels (Fig. 3D, graph). Similar trends were observed for C-peptide (a linker between the A and B chains of insulin) in βTC6 cells (Fig. 3F) and for insulin and Ins2 mRNA measured in other pancreatic β cells (Fig. S3B,C).
The influence of HuD on insulin expression was further studied by immunofluorescence. HuD subcellular distribution was visualized relative to the distribution of two subcellular RNA granules associated with gene repression, processing bodies (PBs), detected using an antibody that recognizes the decapping protein Dcp1, and stress granules (SGs), detected using an antibody that recognizes TIAR. As shown in Fig. 3G, HuD was abundant in the cytoplasm of βTC6 cells, in keeping with its distribution in neurons. In untreated cells, PBs were readily visible while no SGs were detected; interestingly, by 15 and 30 min after glucose stimulation, HuD colocalized in part with cellular PBs and with a few visible SGs (Fig. 3G, Fig. S3D, S3E).
Mice bearing with deletions in both HuD alleles (HuD−/−) or one HuD allele (HuD+/−) showed impaired neuronal differentiation (Pascale et al., 2004; Akamatsu et al., 2005), in keeping with the neuronal presence of HuD. Analysis of pancreas from these mice revealed that insulin levels were highest in pancreatic β cells from HuD−/− mice, moderately lower in HuD+/− mice, and lowest in HuD+/+ mice, as assessed by immunofluorescence (Fig. 4A), and by Western blot and ELISA analysis of insulin production (Fig. 4B,C, Fig. S4A). Thus, in HuD−/− mice, the absence of HuD contributed to elevating constitutive insulin and proinsulin levels in β cells.
To further test the possibility that HuD controls insulin production in pancreatic β cells in vivo, and to extend the analysis to circulating insulin plasma levels, transgenic mice (HuD Tg) were used in which HuD was transcribed from the CamKII promoter and hence was overexpressed in pancreatic β cells [as well as brain (Bolognani et al., 2006), Fig. S4B]. Compared with the pancreatic β cells of wild-type mice (HuD wt; Fig. 4F), pancreatic β cells of HuD Tg mice expressed higher levels of the myc-tagged HuD transgene and correspondingly lower insulin levels, as assessed by immunofluorescence in islets and by Western blot analysis (Fig. 4G,H). The response to changing circulating glucose levels was then assessed in HuD wt and HuD Tg mice by using the intraperitoneal glucose tolerance test (IPGTT). Importantly, HuD Tg mice displayed impaired glucose clearance from blood, as shown by the markedly higher levels of plasma glucose following the glucose challenge (Fig. 4I, top); at the same time, plasma insulin levels in HuD Tg mice were lower throughout the IPGTT period (Fig. 4I, bottom). These results reveal that HuD Tg mice had less readily reseasable insulin in β cells compared with HuD wt mice, which contributed to defective glucose homeostasis. Taken together, our data indicate that HuD represses insulin production in β cells in vivo and thereby contributes to maintaining homeostatic levels of plasma insulin.
We report that HuD, previously considered a neuronal protein (Hinman and Lou, 2008), is also present in mammalian pancreatic β cells, where it regulates insulin levels through the Ins2 5'UTR. By interacting with the p27 5'UTR, a ~400-nt long, highly structured, IRES-containing RNA, HuD repressed p27 translation, likely by disrupting the IRES (Kullmann et al., 2002). In contrast, the 5'UTR of mouse Ins2 mRNA is short (73 nt) as is the 5'UTR of human INS mRNA (59 nt), suggesting that HuD does not repress insulin translation by blocking IRES activity. It is also plausible that simply occupying the Ins2 5'UTR HuD blocks translation initiation.
The release of Ins2 mRNA from HuD upon treatment with glucose is reminiscent of the release of HuR-associated mRNAs upon exposure to stress (Abdelmohsen et al., 2007). Only protein kinase C (PKC) was reported to phosphorylate HuD (Pascale et al., 2005), but neither the specific phosphorylation site of HuD nor its influence on binding to target mRNAs is known. It will be important to elucidate whether PKC or other kinases are responsible for releasing HuD-bound mRNAs and mobilizing HuD to SGs and PBs following glucose challenge.
It is interesting that basal HuD expression is positively regulated by IR signaling, while Ins2 mRNA is rapidly released from HuD after acute exposure to glucose (Fig. 3A), enabling translation of Ins2 mRNA. The distinct mechanisms of HuD and insulin regulation underscore the complexity of posttranscriptional processes, where RBP levels, binding to target mRNA, and localization are governed separately. These multiple events are likely necessary to respond with appropriate kinetics and magnitude of stimulation. Accordingly (Fig. 4J), we propose that positive regulation (green arrows) triggered by glucose or insulin can facilitate insulin production post-transcriptionally in an acute time-frame (minutes) by releasing HuD from Ins2 mRNA and/or relocalizing HuD. The same trigger can balance these effects via the negative regulation of insulin biosynthesis by enhancing HuD levels transcriptionally (red arrows). The coordination of these positive and negative influences helps to restore insulin to homeostatic levels (Fig. 4J).
The findings that HuD−/− mice express higher insulin while HuD Tg mice express lower insulin support the notion that HuD regulates insulin production in vivo. In light of this result, it will be important to test if HuD function is aberrant in patients with type 2 diabetes. The constitutive levels of HuD, its subcellular distribution, the release of HuD from the INS mRNA in response to glucose or insulin stimulation, all of these parameters warrant study in diabetic and non-diabetic patient populations. For example, in type 2 diabetes, insulin secretion from β cells is severely defective in response to rising circulating glucose, especially in the first ten minutes after a glucose load, and therefore it will be important to test if HuD release from INS mRNA is also slow.
Finally, it is worth noting that the microRNA miR-375 represses HuD production by lowering both the levels and translation rates of HuD mRNA (Abdelmohsen et al., 2010). Interestingly, miR-375 regulates insulin secretion in pancreatic β cells, and miR-375 knockout (375KO) mice have a significantly reduced pancreatic β-cell mass compared with WT mice (Poy et al., 2009). The authors propose that the hyperglycemia of 375KO mice is largely due to the relative larger α-cell compartment (which secretes glucagon) in 375KO mice. However, in 375KO mice, β cells may have higher HuD levels, which lowers insulin production, consistent with the hyperglycemic phenotype of 375KO mice. Whether miR-375/HuD levels contribute to diabetes also deserves future investigation.
βIRWT and βIRKO cells (Kulkarni et al., 1999; Assmann et al., 2009; Kim et al., 2011) and mouse insulinoma βTC6 cells were cultured in high-glucose DMEM (Invitrogen) supplemented with 10% FBS. Cells were cultured for 16 h in low-glucose (2 mM) DMEM before stimulation with the indicated glucose concentrations and in low glucose with 0.1% FBS before stimulation with insulin for 24 h. Chemicals and transfection of siRNA and plasmids are described in supplemental text.
Immunoprecipitation (IP) of endogenous RNP complexes from whole-cell extracts and biotin pulldown analysis were performed as described (supplemental text) using anti-HuD or control IgG antibodies (Santa Cruz Biotech.). The RNA isolated from IP was further assessed by RT-qPCR analysis. Chromatin immunoprecipitation (ChIP) from βTC6 cells was performed using the EZ-ChIP chromatin IP kit (Millipore, Billerica, MA) following the manufacturer's protocol, using anti-FoxO1 antibody (Abcam) and HuD promoter-specific primers (SA Bioscience). Further description in supplemental text.
Pancreatic sections of normal (wild-type) C57B1/6J mice were used for detection of insulin, HuD, and glucagon. HuD−/−, HuD+/− and HuD−/− mice (Akamatsu et al., 2005) were used at 14 wks of age. HuD Tg mice were described (Bolognani et al., 2006). Mice were anesthetized and tissues were quickly excised and frozen in liquid nitrogen for western blot analysis and ELISA, or fixed in 4% paraformaldehyde for immunostaining. For IPGTT, glucose (1 g/kg body weight) was given intraperitoneally to fasted, 3-month-old female mice for measurement of blood glucose and plasma insulin levels at the indicated times (details in supplemental text).
Quantitative data are presented as the mean ± SEM and compared statistically by Student's t-test, using Graphpad Prism (GraphPad Software). A p value of <0.05 was considered statistically significant.
We thank H. J. Okano, M. Igarashi, R. Selimyan, D. Nines, and D. Boyer for assistance, and H. Huang for reagents. This work was supported in part by the NIA-IRP, NIH. R.N.K. is supported by NIH RO1 DK 67536 and 68721. H.O. and W.A. are funded by the Japanese Ministry of Education, Science, Sports, Culture and Technology. E.K.L. is funded by the Korean Ministry of Education, Science and Technology (5-2011-A0154-00046).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.