The RNA-binding Protein HuD Regulates Autophagosome Formation in Pancreatic β Cells by Promoting Autophagy-related Gene 5 Expression*

Background: Autophagy-related gene 5 (ATG5) has a pivotal role in the formation of autophagosomes. Results: ATG5 expression was increased by the RNA-binding protein HuD at the mRNA level. Conclusion: HuD promotes autophagosome formation by elevating ATG5 expression in pancreatic β cells. Significance: HuD is a novel regulator of autophagosome formation and autophagy flux in pancreatic β cells. Tight regulation of autophagy is critical for the fate of pancreatic β cells. The autophagy protein ATG5 is essential for the formation of autophagosomes by promoting the lipidation of microtubule-associated protein LC3 (light chain 3). However, little is known about the mechanisms that regulate ATG5 expression levels. In this study, we investigated the regulation of ATG5 expression by HuD. The association of HuD with ATG5 mRNA was analyzed by ribonucleoprotein complex immunoprecipitation and biotin pulldown assays. HuD expression levels in pancreatic β cells were knocked down via siRNA, elevated by overexpression of a HuD-expressing plasmid. The expression levels of HuD, ATG5, LC3, and β-actin were determined by Western blot and quantitative RT-PCR analysis. Autophagosome formation was assessed by fluorescence microscopy in GFP-LC3-expressing cells and in pancreatic tissues from WT and HuD-null mice. We identified ATG5 mRNA as a post-transcriptional target of the mammalian RNA-binding protein HuD in pancreatic β cells. HuD associated with the 3′-UTR of the ATG5 mRNA. Modulating HuD abundance did not alter ATG5 mRNA levels, but HuD silencing decreased ATG5 mRNA translation, and, conversely, HuD overexpression enhanced ATG5 mRNA translation. Through its effect on ATG5, HuD contributed to the lipidation of LC3 and the formation of LC3-positive autophagosomes. In keeping with this regulatory paradigm, HuD-null mice displayed lower ATG5 and LC3 levels in pancreatic β cells. Our results reveal HuD to be an inducer of ATG5 expression and hence a critical regulator of autophagosome formation in pancreatic β cells.

Autophagy is a catabolic process that leads to the degradation and recycling of unnecessary or dysfunctional cellular components (1). It is primarily activated in response to stress conditions induced by a multitude of physiological and pathological situations, including starvation, growth factors, oxidative stress, and endoplasmic reticulum stress (2,3). Autophagy also occurs at basal levels in most tissues for the selective elimination of misfolded or aggregated proteins, damaged organelles, and intracellular pathogens, thereby protecting against various diseases. It is a highly regulated process that has key roles in cell survival, development, and homeostasis, helping to maintain a balance between synthesis and degradation in cells undergoing intense metabolic activity and is deregulated in various pathological conditions, including cancer, neurodegenerative diseases, and type 2 diabetes (4,5).
The hallmark of autophagy is the formation of a doublemembrane vesicle with cytosolic contents that is called an autophagosome (6). The formation of the autophagosome and their delivery to lysosomes are controlled by conserved key regulators known as autophagy-related (ATG) proteins including ATG5, ATG7, ATG8 (or LC3, for microtubule-associated protein light chain 3), and ATG12 (7). They mediate the conjugation of LC3I to phosphatidylethanolamine to form LC3II, which specifically associates with autophagosomes, a critical step in autophagy (8,9). LC3 conjugation, which can be monitored by a shift in electrophoretic mobility from the LC3I to the LC3II by Western blot analysis, and LC3-positive puncta are used as well accepted markers for autophagosome formation and autophagy (10). Accumulating data indicate that ATG5 plays critical roles in autophagosome formation under most circumstances by promoting LC3I lipidation to LC3II (11,12). The activity of ATG5 during the autophagic process is regulated by several post-translational modifications including acetylation (13) and phosphorylation (14). Accordingly, deletion of ATG5 suppresses the lipidation of LC3I to LC3II, inhibits autophagy (15,16), and was sufficient to cause pathological conditions including neurodegeneration and neurolog-ical pathology (17). Although the function of ATG5 in autophagosome formation and autophagy is well understood, the mechanisms that regulate the endogenous level of ATG5 are unknown.
Recent studies have identified microRNAs that post-transcriptionally regulate ATG transcripts including ATG5 mRNA (18 -20). Here, we describe the identification of HuD/human antigen D/embryonic lethal abnormal vision-like 4 (ELAVL4) as a new autophagy regulatory RNA-binding protein that associates with ATG5 mRNA and controls its translation in pancreatic ␤ cells. Like other Hu/ELAV family members (HuR, HuB, and HuC), HuD has three RNA recognition motifs that mediate its association with the UTR of mRNAs bearing specific sequences that are often AU-and U-rich (21). Through these associations, HuD can modulate the stability or/and the translation of target mRNAs, often enhancing it, but other times repressing it (22)(23)(24). The roles of HuD in the regulation of neuronal development and plasticity were characterized using animal models (25,26). We previously found that HuD is also expressed in pancreatic ␤ cells in addition to neurons (27). In this study, we showed that the binding of HuD to the 3Ј-UTR of ATG5 mRNA enhances ATG5 mRNA translation, thereby promoting the conversion of LC3I to LC3II and leading to an increase of LC3-positive autophagosomes. Because of this regulatory process, HuD-null (HuD Ϫ/Ϫ ) mice expressed lower levels of ATG5 and LC3 in ␤ cells.
RNA Analysis and Ribonucleprotein Complex Immunoprecipitation-Total RNA was prepared from whole cells using TRIzol (Invitrogen). After RT using random hexamers and reverse transcriptase (Toyobo), the abundance of transcripts was assessed by quantitative PCR (qPCR) analysis using the SYBR green PCR master mix (Kapa Biosystems) and gene-specific primer sets (Table 1). RT-qPCR analysis was performed on Applied Biosystems model 7300 instruments.

For cloning of HuD fusion plasmid
Mouse Ribonucleoprotein complex immunoprecipitation analysis was performed using anti-HuD primary antibodies or control IgG (Santa Cruz Biotechnology) (27). In brief, RNP complexes were immunoprecipitated using HuD or control IgG antibodies and incubated with DNase I and proteinase K; RNA in the immunoprecipitation (IP) samples was isolated and further analyzed by RT-qPCR using the primers listed (Table 1).
Biotin Pulldown Analysis-To synthesize biotinylated transcripts, PCR fragments were prepared using forward primers that contained the T7 RNA polymerase promoter sequence (T7, CCAAGCTTCTAATACGACTCACTATAGGGAGA). The primers used to prepare biotinylated transcripts spanning the ATG5 mRNA (NM_053069.5) are listed in Table 1. After purification of the PCR products, biotinylated transcripts were synthesized using the MaxiScript T7 kit (Ambion) and biotin-CTP (Enzo Life Sciences). Whole cell lysates (200 g/sample) were incubated with 1 g of purified biotinylated transcripts for 30 min at room temperature, and then complexes were isolated using streptavidin-coupled Dynabeads (Invitrogen). Proteins present in the pulldown material were studied by Western blot analysis as described in Ref. 27.
Polysome Analysis-48 h after transfection of siHuD and pHuD with appropriate controls, ␤TC6 cells were preincubated with cycloheximide (100 g/ml, 15 min) and lysed with polysome extraction buffer containing 20 mM Tris-HCl, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 0.5% (v/v) Nonidet P-40, 1ϫ protease inhibitor mixture, and RNase inhibitor, followed by centrifugation at 10,000 ϫ g for 10 min. Lysates were further fractionated by ultracentrifugation through linear sucrose gradients as described in Ref. 29. RNAs from each fraction were isolated, and cDNA was synthesized as described above. The relative levels of ATG5 and GAPDH mRNAs were analyzed by RT-qPCR using specific primer set.
Immunostaining-HuD Ϫ/Ϫ mice and wild-type littermates (HuD ϩ/ϩ ) have been described previously (25) and were used at 14 weeks of age. Pancreases from fasted HuD ϩ/ϩ and HuD Ϫ/Ϫ mice were fixed in 4% paraformaldehyde, immersed in 20% sucrose before freezing, and then sectioned at a thickness of 7 m. After antigen unmasking, the slides were incubated with primary antibodies (anti-insulin (Sigma) or anti-ATG5 and anti-LC3 (Abcam)), followed by secondary antibodies (Invitrogen). Slides were viewed using a LSM-710 confocal microscope (Carl Zeiss MicroImaging). For quantification, digital images of multiple sections from three mice per group, separated by at least 200 m from each section, at a magnification of 40ϫ were obtained, and signal intensity was determined using LSM Image Browser software (Carl Zeiss).
Nascent Translation Assay-Nascent translation of EGFP was studied by incubating ␤TC6 cells with 1 mCi L-[ 35 S]methionine and L-[ 35 S]cysteine (PerkinElmer Life Sciences) per 35-mm plate for 15 min. After lysis, IP reactions were carried out using mouse IgG or anti-GFP (Santa Cruz Biotechnology), and reactions were resolved by SDS-PAGE, transferred to PVDF membranes, and visualized using a PharoseFX Plus system (Bio-Rad) analyzed using Quantity One software (Bio-Rad).
Electron Microscopy-For transmission electron microscopic observation, ␤TC6 cells were fixed with 1% osmium tetroxide and embedded using Epon 812. Ultrathin sections were observed with a transmission electron microscope (JEM 1010).

RESULTS
HuD Binds to the 3Ј-UTR of ATG5 mRNA-During experiments that identified HuD as an RNA-binding protein that binds to insulin mRNA and controls its translation (27), we also identified ATG5 mRNA as another target of HuD in pancreatic ␤ cells. We confirmed the interaction between HuD and ATG5 mRNA by ribonucleoprotein complex immunoprecipitation with an anti-HuD antibody. ATG5 mRNA associated with HuD in the RNP complexes was further analyzed by RT followed by qPCR analysis in mouse insulinoma ␤TC6 cells. Ins2 mRNA was used as a positive control for the HuD IP (27). As seen in Fig. 1A, ATG5 mRNA is significantly enriched in the anti-HuD IP relative to the control IgG IP (Fig. 1A), supporting the idea that ATG5 mRNA is a part of HuD RNP complexes. We further identified a potential HuD binding region in ATG5 mRNA using biotin-labeled RNA probes (27). Biotinylated segments of the ATG5 mRNA spanning the 5Ј-UTR, coding region, and 3Ј-UTR (3U1 and 3U2) were synthesized (Fig. 1B, top), and the RNP complexes containing HuD and biotinylated RNAs were detected using streptavidin-coated beads. As shown in Fig. 1B, HuD has specific affinity for the 3U2 fragment of the ATG5 mRNA, but not for the ATG5 5Ј-UTR, coding region, 3U1, or a negative control RNA, the 3Ј-UTR of GAPDH mRNA. Experiments to examine the association of other RNA-binding proteins and ATG5 mRNA reveal that ATG5 mRNA associates with HuR but not with AUF1, hnRNP L, TIAR, and NF90 (Fig.  1B). Further assays using fragments of the ATG5 3U2 reveal that segment B, spanning positions 819 -1038, predominantly interacts with HuD (Fig. 1C). In sum, HuD associates with the ATG5 mRNA through its 3Ј-UTR. We then set out to study the functional link between HuD and ATG5 mRNA.
HuD Promotes the Translation of ATG5 mRNA by Interacting with the ATG5 3Ј-UTR-Next, we tested whether the 3U2 region of ATG5 mRNA is responsible for the regulation of ATG5 expression by HuD in ␤TC6 cells. We generated EGFP reporter constructs harboring the 3U1 or 3U2 segment of ATG5 mRNA downstream of the coding region of EGFP mRNA (pATG5 3U1 or pATG5 3U2; Fig. 2A) and transfected the reporter constructs into siRNA-transfected (siCtrl or siHuD) or plasmid-transfected (pcDNA or pHuD) ␤TC6 cells. As observed in Fig. 2B, EGFP mRNA remains unchanged after both silencing and overexpression of HuD, as measured using RT-qPCR (Fig. 2B). However, HuD silencing leads to selectively decreased expression of EGFP reporter in the pATG5 3U2 group, but not in the control reporter group (pEGFP), as assessed by Western blot analysis. Conversely, HuD overexpression up-regulates the expression of EGFP reporter in the pATG5 3U2 group, but not in the control group (Fig. 2C). However, EGFP expression levels are not changed by HuD silencing or overexpression in the groups transfected with pATG5 3U1 (Fig. 2D), the region that shows no direct interaction with HuD (Fig. 1B).
Previously, miR-181 was shown to be a negative regulator of ATG5 expression by targeting 3Ј-UTR (20). To determine whether miR-181 competes with HuD the regulation of ATG5 expression, we examined the level of EGFP in miR-181-and/or HuD-transfected cells. Whereas the level of EGFP expressed by the control (pEGFP) is not affected by miR-181 or HuD expression, expression of EGFP by pATG5 3U2 is down-regulated by miR-181 transfection and restored by HuD overexpression (Fig.  2E).
Likewise, endogenous ATG5 mRNA levels in ␤TC6 cells remained unchanged after both silencing and overexpression of HuD (Fig. 3A). However, the levels of ATG5 protein were significantly reduced by silencing HuD via siRNA and, conversely, increased by ectopic expression of HuD (pHuD) (Fig. 3B). Collectively, these results indicate that HuD can associate with the 3Ј-UTR in ATG5 mRNA and that HuD increases ATG5 abundance, likely by enhancing its translation.
To investigate whether HuD regulates translation of ATG5 mRNA, we tested the relative association of ATG5 mRNA with polyribosomes by polysome fractionation in ␤TC6 cells after HuD silencing or HuD overexpression as described previously (29). Cytoplasmic lysates were fractionated through sucrose gradients, with the lightest components sedimenting at the top (fractions 1 and 2), small (40 S) and large (60 S) ribosomal subunits and monosomes (80 S) appearing in fractions 3-5, and progressively larger polyribosome appearing in fractions 6 -12 (Fig. 3C). Polysome profiles are not affected by HuD silencing or HuD overexpression (Fig. 3C). Whereas in siCtrl cells, the distribution of polysome-associated ATG5 mRNA levels peaks in fraction 7, HuD silencing reduces the average peak size of ATG5 mRNA polysomes such that most of the ATG5 mRNA is found in the lighter molecular weight polysomes in fraction 6 ( Fig. 3D, left top). In HuD-overexpressing cells, polysome-associated ATG5 mRNA is found among the larger polysomes in fractions 7 and 8 (Fig. 3D, right top). The distribution of the housekeeping GAPDH mRNA associating with polysomes largely overlaps between two groups (Fig. 3D, bottom). The rel-ative distribution of ATG5 mRNA is significantly enriched in the polysome fraction (fractions 7-11) compared with the free subunits/monosome fraction (fractions 1-6; Fig. 3E). To provide additional evidence for the translational regulation of ATG5 mRNA by HuD, the distribution of EGFP mRNA in the polysome fractions is analyzed in Fig. 3F. Polysome-associated EGFP mRNA containing ATG5 mRNA 3Ј-UTR is found in their heavier polysome fractions in HuD-overexpressing cells, whereas is found in the lighter molecular weight polysomes in HuD-silenced cells (Fig. 3F). Together, these findings suggest that HuD increases the association of the ATG5 mRNA with high molecular weight, actively translating polysome fractions, and therefore that HuD promotes ATG5 expression by enhancing ATG5 mRNA translation.
To further confirm the translational regulation of ATG5 mRNA by HuD, we also studied the nascent translation of ATG5; however, the ATG5 antibody used in this study did not work for IP (data not shown). Instead, we measured the de novo synthesis of the reporter protein expressed from pATG5 3U2 using an EGFP antibody. Briefly, following transfections, cells were incubated for a short time period (20 min) with [ 35 S]Met and [ 35 S]Cys, whereupon lysates were prepared, and the abundance of newly synthesized [ 35 S]EGFP was measured by IP. As shown Fig. 3G, nascent EGFP translation is enhanced by HuD overexpression but down-regulated by HuD silencing, which suggests that HuD enhances the translation of ATG5 mRNA.
HuD Increases Autophagosome Formation-ATG5 is responsible for LC3I conjugation to phosphatidylethanolamine to form LC3II, as well as for the membrane localization of autophagic machinery and the formation of autophagosomes (7). Because LC3II is localized in autophagosome-associated puncta, assessing LC3 lipidation is a useful way to monitor autophagy in mammals (10). By monitoring the change in LC3II levels in ␤TC6 cells expressing different levels of HuD, we investigated whether HuD-mediated modulation of ATG5 FIGURE 1. HuD associates to ATG5 mRNA. A, ␤TC6 cell lysates were subjected to ribonucleoprotein complex immunoprecipitation followed by RT-qPCR analysis to measure the enrichment of ATG5 mRNA in HuD IP compared with control IgG IP. All mRNA amounts were normalized to the levels of GAPDH mRNA (2 Ϫ[Ct(ATG5 or Ins2)-Ct(GAPDH)] ). The data represent means Ϯ S.E. from three independent experiments. *, p Ͻ 0.05 B, top, schematic description of the ATG5 mRNA 5Ј-UTR (5U), coding region (CR), and 3Ј-UTR (3U1 and 3U2), as well as the biotinylated transcripts synthesized for biotin pulldown analysis. Bottom, each biotinylated transcript was incubated with ␤TC6 cell lysates, and the interaction between RNA-binding proteins and biotinylated transcripts was analyzed by Western blot analysis using HuD, HuR, AUF1, hnRNPL, TIAR, and NF90 antibodies. C, biotinylated subdivided regions of ATG5 3U2 were synthesized, and fragments A, B, and C were further incubated with ␤TC6 cell lysates. The interaction between HuD and biotinylated transcripts was analyzed by Western blot analysis using HuD antibody. nt, nucleotide. expression is involved in the formation of autophagosomes. As shown in Fig. 4A, ectopic expression of HuD increases the levels of LC3II, which, conversely, is reduced by HuD silencing. Because autophagy induction is mediated by AMPK activation, we next treated U2OS cells stably expressing GFP-LC3 with AICAR, which is an AMPK activator, or colchicine, which is an autophagic-lysosomal delivery inhibitor, for 16 h and examined whether HuD affects GFP-LC3II conversion, as well as the formation of GFP-positive puncta, using Western blot analysis and fluorescence microscopy, respectively (Fig. 4, B and C). AICAR or colchicine treatment leads to elevated levels of GFP-LC3II in control cells (transfected with pcDNA), whereas HuD-overexpressing cells (pHuD) have higher levels of GFP-LC3II, as assessed by Western blot analysis (Fig. 4B). Consistent with these results, HuD overexpression facilitates the formation of GFP-positive puncta in both untreated cells and AICAR/colchicine-treated cells (Fig. 4, C and D).
The ratio of the number of cells having GFP-LC3 puncta relative to the number of GFP-positive cells is shown in Fig. 4D. AICAR or colchicine treatment increases GFP-LC3II accumulation in both control and HuD-overexpressing cells (Fig. 4D), indicating that HuD-mediated conversion of LC3I to LC3II contributes to the formation of autophagosome. We also examined the effect of HuD in autophagome formation by transmission electron microscopic analysis (Fig. 4E). As expected, the number of autophagomes is reduced in HuD silencing cells. The essential role of ATG5 in the regulation of autophagy has previously been demonstrated by showing impaired LC3 conversion and autophagosome formation in ATG5-deficient mice (11). To test whether HuD-mediated modulation of LC3 con-FIGURE 2. HuD regulates ATG5 expression through its 3-UTR. A, schematic of reporter plasmids: parent vector (pEGFP), a plasmid expressing only one copy of fragment ATG5 3Ј-UTR1 (pATG5 3U1), and a plasmid expressing only one copy of fragment ATG5 3Ј-UTR2 (pATG5 3U2). B, 48 h after transfection of either siCtrl or siHuD or 24 h after transfection of either pcDNA or pHuD, together with each reporter plasmids, the levels of EGFP mRNAs were analyzed by RT-qPCR. All mRNA amounts were normalized to the levels of GAPDH mRNA (2 Ϫ[Ct(EGFP)-Ct(GAPDH)] ). The data represent means Ϯ S.E. from three independent experiments. C and D, 48 h after transfection of either siCtrl or siHuD or 24 h after transfection of either pcDNA or pHuD, reporter plasmids (pATG5 3U2 for C; pATG5 3U1 for D) were sequentially transfected. The level of EGFP, HuD, and loading control ␤-actin were assessed by Western blot analysis. Relative density of the indicated proteins from three independent experiments is shown on the bottom. E, 24 h after transfection of pcDNA or pHuD, miR-181b or control miRNA was transfected into HCT-116 cells, followed by transfection with reporter plasmids (pATG5 3U2 or pEGFP). The expression levels of EGFP and ␤-actin were analyzed by Western blot analysis.
version is dependent on ATG5, HuD (pHuD) was ectopically expressed in U2OS/GFP-LC3 cells (28) with either siCtrl or ATG5 siRNA (siATG5), and then the cells were incubated with AICAR or colchicine (Fig. 5). As anticipated, siATG5 transfec-   (pHuD). B, lysates from ␤TC6 cells that were transfected as described in A were prepared to assess the level of ATG5, HuD, and loading control ␤-actin by Western blotting. Relative density for the indicated proteins is shown on the bottom. The data represent means Ϯ S.E. from three independent experiments. **, p Ͻ 0.01. C and D, 48 h after transfection of siCtrl or siHuD or 24 h after transfection of pcDNA or pHuD, lysates were fractionated through sucrose gradients (C), and the relative distribution of ATG5 mRNA and GAPDH mRNA in each 12 fractions was studied by RT-qPCR analysis (D). 40S, small ribosome subunits; 60S, large ribosome subunits; 80S, mono-ribosome; Polysomes, polyribosome. The data are representative of three independent experiments. E, relative ATG5 mRNA distributions between free subunits/monosomes (fractions 1-6) and polysomes (fractions 7-11) were represented from three independent experiments. *, p Ͻ 0.05. F, after co-transfection of reporter construct containing ATG5 3Ј-UTR (pATG5 3U2) in siHuD or pHuD transfected cells, the distribution of EGFP mRNA and GAPDH mRNA was analyzed as described in D. G, nascent EGFP production was monitored by a brief (20-min-long) incubation with L-[ 35 S]methionine and L-[ 35 S]cysteine 48 h after transfection of siCtrl or siHuD, or 24 h after transfection of pcDNA or pHuD. Lysates were prepared and subjected to IP using anti-EGFP or anti-GAPDH antibodies, and the incorporation of radiolabeled amino acids into newly synthesized EGFP and GAPDH proteins was assessed by SDS-PAGE and visualized using a PharoseFX Plus.   Together, these data suggest that HuD contributes to the lipidation of LC3II and to autophagosome formation by directly regulating ATG5 expression.
Decreased ATG5 and LC3 Levels in Pancreatic ␤ Cells of HuD Ϫ/Ϫ Mice-HuD Ϫ/Ϫ mice are known to show impaired neuronal differentiation and decreased proliferation of neural stem/progenitor cells (25). In our previous report (27), we found increased insulin levels in pancreatic ␤ cells from HuD Ϫ/Ϫ mice, compared with HuD ϩ/ϩ mice. Because HuD increased ATG5 levels and affected LC3 conversion in cultured cells, we tested the possibility that HuD also elicits this effect in pancreatic ␤ cells in vivo. Analysis of pancreases from HuD Ϫ/Ϫ mice reveals lower expression of ATG5 in the pancreatic ␤ cells from HuD Ϫ/Ϫ mice compared with HuD ϩ/ϩ mice, as assessed by immunofluorescence (Fig. 6A). Furthermore, the perinuclear accumulation of LC3, which is apparent mainly during longer starvation periods (30) and is considered a reliable marker of autophagy (31), is reduced in pancreatic ␤ cells from HuD Ϫ/Ϫ mice, compared with those from HuD ϩ/ϩ mice (Fig.  6B). Thus, our data suggest that HuD plays an important role in autophagosome formation by regulating ATG5 translation in pancreatic ␤ cells in vivo.

DISCUSSION
In type 2 diabetes, it is well known that an increased number of autophagic vacuoles and increased cell death is linked to pancreatic ␤ cell dysfunction. ATG5 is a critical factor for autophagosome formation under most circumstances by facilitating the conversion of LC3I to LC3II. In this study, we identified HuD as a novel regulator of autophagosome formation, a role that is linked to the binding of HuD to the ATG5 mRNA 3Ј-UTR, which leads to enhanced ATG5 expression. In pancreatic ␤ cells, HuD overexpression elevated the level of ATG5 and enhanced the conversion of LC3I to LC3II, whereas HuD silencing lowered ATG5 levels and suppressed LC3II conversion. Moreover, ectopic expression of HuD in U2OS cells, which do not express endogenous HuD, enhanced the conversion of GFP-LC3I to GFP-LC3II and the formation of GFP-LC3 puncta after AICAR or colchicine treatment. Taken together, these data support the hypothesis that HuD is one of the pivotal regulators of autophagosome formation in pancreatic ␤ cells.
HuD has been extensively studied in neuronal system (25,(32)(33)(34)(35)(36)(37). It has been reported to associate with AU-rich RNA elements in 3Ј-UTRs from several mRNAs, and this association often results in mRNA stabilization (29,38,39). However, our previous studies found that HuD was also expressed in pancreatic ␤ cells and repressed insulin translation through binding to the insulin mRNA 5Ј-UTR (27), in turn lowering insulin production. A similar function for HuD as a translational repressor FIGURE 6. HuD-null mice show abnormal expression of ATG5 and LC3. The expression level of ATG5 (A) and LC3 (B) were analyzed by immunostaining in pancreatic islets of WT and HuD-null mice. The slides were stained with antibodies against insulin (green) and ATG5 (red) or LC3 (red). Scale bars, 50 m. For graph, multiple sections from three mice per genotype, separated by at least 200 m from each section, were imaged, and the relative signal intensity for the indicated proteins was assessed using Image Browser software (Carl Zeiss). *, p Ͻ 0.05. has also been identified previously, with HuD being shown to repress the translation of p27 target mRNA and other RNAs (38 -40). In contrast, HuD binding to the Nova1 mRNA enhanced its stability and translation (41). The highly conserved protein HuR, a ubiquitous member of the Hu family, also associates with the ATG5 mRNA 3Ј-UTR (Fig. 1B). The possibility that HuR may also be involved in the regulation of ATG5 expression awaits further investigation.
The results presented in this study show that the binding of HuD to ATG5 mRNA facilitates ATG5 translation by enhancing the assembly of large polysomes to increase ATG5 levels, which promotes the conversion of LC3I to LC3II and the formation of autophagosomes. A recent study by Tekirdag et al. (20) showed that miR-181a affects starvation-and rapamycininduced autophagy by down-regulating ATG5 levels in human MCF, Huh7, and K562 cells. Overexpression of miR-181a resulted in decreased autophagy activity during starvation or rapamycin treatment by targeting the ATG5 mRNA 3Ј-UTR. Another study reported that mimics and inhibitors of miR-181a modulated ATG5 expression levels as well as those of a luciferase reporter containing the ATG 3Ј-UTR (42). HuD and miR-181 have different binding sites in 3Ј-UTR of ATG5 mRNA (Fig.  1C) and regulate ATG5 expression in opposite ways; HuD enhances ATG5 expression, whereas miR-181 inhibits it, which indicates that they might competitively regulate ATG5 expression (Fig. 2E). Further studies will be needed to elucidate the molecular mechanisms underlying the regulation of ATG5 expression through various post-transcriptional regulators.
Our results add to a small but emerging body of evidence that ATG5 expression is regulated at the post-transcriptional level in pancreatic ␤ cells. The fact that HuD Ϫ/Ϫ mice express lower levels of ATG5 and reduced perinuclear accumulation of LC3 in pancreatic ␤ cells compared with HuD ϩ/ϩ mice (Fig. 6) leads us to hypothesize that HuD Ϫ/Ϫ mice may show altered autophagy activity. Further studies are underway to monitor autophagy influx in HuD Ϫ/Ϫ animals after stress responses with heavy involvement of ␤ cell activity, including endoplasmic reticulum stress and changes in circulating glucose. This expansion of the work will provide critical insight into the possible pathological roles of HuD during stress-induced autophagy leading to ␤ cell dysfunction.
Tight regulation of autophagy influx is critical for maintaining normal pancreatic ␤ cell homeostasis and preserves pancreatic function, structure, and mass in response to stresses such as diabetes-induced oxidative stress, which causes the abnormal cytoplasmic accumulation of proteins (43). It has been shown that the loss of autophagy (ATG7 Ϫ/Ϫ ) in ␤ cells increases the accumulation of ubiquitinated proteins and damaged organelles in the cytoplasm (44), and ATG7-deficient pancreatic ␤ cells showed reduced ␤ cell mass caused by increased apoptosis and decreased cell proliferation. Moreover, ATG7 f/f : RIP-Cre mice developed diabetes with obvious ␤ cell loss when fed a high fat diet (45). Our studies add to an expanding body of evidence that underscores the complex function of HuD in preserving glucose homeostasis in pancreatic ␤ cells.