HuD Interacts with Bdnf mRNA and Is Essential for Activity-Induced BDNF Synthesis in Dendrites

Highly specific activity-dependent neuronal responses are necessary for modulating synapses to facilitate learning and memory. We present evidence linking a number of important processes involved in regulating synaptic plasticity, suggesting a mechanistic pathway whereby activity-dependent signaling, likely through protein kinase C (PKC)-mediated phosphorylation of HuD, can relieve basal repression of Bdnf mRNA translation in dendrites, allowing for increased TrkB signaling and synaptic remodeling. We demonstrate that the neuronal ELAV family of RNA binding proteins associates in vivo with several Bdnf mRNA isoforms present in the adult brain in an activity-dependent manner, and that one member, HuD, interacts directly with sequences in the long Bdnf 3' untranslated region (3'UTR) and co-localizes with Bdnf mRNA in dendrites of hippocampal neurons. Activation of PKC leads to increased dendritic translation of mRNAs containing the long Bdnf 3'UTR, a process that is dependent on the presence of HuD and its phosphorylation at threonine residues 149 and/or 165. Thus, we found a direct effect of HuD on regulating translation of dendritic Bdnf mRNAs to mediate local and activity-dependent increases in dendritic BDNF synthesis.


Introduction
Brain-derived neurotrophic factor (BDNF) is an important molecule involved in learning and memory in the adult brain, with demonstrated roles in regulating synaptic plasticity [1][2][3]. The mechanisms by which BDNF regulates synaptic strength have become increasingly well understood. BDNF stored in dendritic vesicles is released both constitutively and through an activity-dependent pathway. Released BDNF can potentially bind to TrkB receptors on both preand post-synaptic terminals, leading to retrograde and anterograde signaling [4][5][6][7]. Recent studies highlight the significance of regulation of local translation of mRNAs in activated spines as one way in which BDNF can specifically modify synapses in response to stimulation [8][9][10][11]. Despite the increasing understanding of how a number of important genes are regulated by BDNF signaling to modify the structure and composition of synapses, less is known about how BDNF itself is regulated. While recent work has begun to show some of the key mechanisms at play in regulating dendritic BDNF synthesis [12], there is still much to be determined about the various pathways involved.
Our previous studies have shown that sequences in the long Bdnf 3' untranslated region (3' UTR) are sufficient to traffic mRNAs to distal dendrites, and that Bdnf mRNAs lacking the long 3' UTR are primarily restricted to the soma [11], suggesting the cis-acting elements regulating dendritic Bdnf mRNA translation are located in this region. Indeed, it has been found that under basal conditions the long Bdnf 3' UTR is a repressor of translation, and neuronal activity can increase the translation of mRNAs containing this sequence [13]. The trans-acting factors binding to Bdnf mRNAs to regulate their local translation, however, are not known. The neuronal Hu proteins, a family of RNA binding proteins (RBPs), are ideal candidates. They are RBPs that stabilize their target mRNAs, can lead to increased translation in response to neuronal activation, and have been implicated in regulating learning and memory processes [14][15][16]. One of these proteins, HuD, has been the focus of a number of studies [17]. Protein kinase C (PKC) activity is necessary for phosphorylation on HuD to mediate its associated increases in target mRNA stability and translation [18,19]. It has been shown that PKCdependent phosphorylation of HuD and its association with the short Bdnf 3' UTR are important for upregulation of BDNF translation and BDNF-dependent long-term potentiation (LTP) [19,20]. However, these studies did not examine the interaction of HuD with the long Bdnf 3' UTR.
Here we present in vitro and in vivo findings that bring to light a possible mechanism for linking neuronal activation to increased local protein synthesis of dendritically localized mRNAs. Immunoprecipitations of messenger ribonucleoprotein (mRNP) complexes from mice under basal or stimulated conditions reveal an activity-dependent association of Bdnf mRNAs with neuronal Hu proteins. Observations of dissociated rat hippocampal cultures manipulated genetically and pharmacologically show the importance of HuD and PKC in regulating the dendritic translation of mRNAs containing Bdnf 3' UTR sequences, and RNA electrophoretic mobility shift assays (REMSAs) point to a direct interaction between the HuD protein and Bdnf mRNA sequences specific to the long 3' UTR.

Animals and DNA constructs
Pregnant female Sprague Dawley rats were purchased from the Charles River Laboratories (Wilmington, MA, USA). Both genders of mice were used in this study. The Georgetown University Animal Care and Use Committee approved the animal procedures performed in this study.

Quantitative real-time PCR
Real-time PCR was performed using 50 cycles with FastStart Universal SYBR Green Master (ROX) (Roche Diagnostics) on the StepOne Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) using oligonucleotide pairs designed on Primer Express 3.0 software (Applied Biosystems) to span introns. The primers used were: F-ExI (ACTGAGTCTCCAGG ACAGCAAAG); F-ExIIc (GTGGTGTAAGCCGCAAAGAA); F-ExIV (CAGAGCAGCTG CCTTGATGTT); F-ExVI (CAGAAGCGTGACAACAATGTGA); R-BDNF (CCTTCATG CAACCGAAGTATGA); F-Rpl10a (GAAGAAGGTGCTGTGTTTGGC); R-Rpl10a (TCGG TCATCTTCACGTGGC). Reactions were carried out in triplicate for each sample. For the association assays ΔC t was calculated by subtracting the C t for Rpl10a, a housekeeping gene whose mRNA does not associate with HuD [22], and ΔΔC t was calculated as the ΔC t for the α-nELAV IP minus the ΔC t for the IgG isotype control IP, with fold-difference determined by 2 -ΔΔCt . For the seizure-induction assays ΔC t was calculated by subtracting the C t for the input (relative percent recovery), and ΔΔC t was calculated as the ΔC t for the pilocarpine treatment minus the ΔC t for the PBS control treatment, with fold-difference determined by 2 -ΔΔCt .

Local translation reporter assays
Primary dissociated hippocampal cultures were established and transfected as described previously [23]. The cultures were transfected with the indicated plasmids at 14 days in vitro (DIV) and assays were performed 48 hours later at 16 DIV. Pre-warmed and CO 2 -equilibrated Neurobasal media was used for all media replacements. For KCl stimulation experiments, media was removed and replaced with Neurobasal containing 30 mM KCl for 5 min, which was then replaced with unsupplemented Neurobasal for an additional 55 min. For PMA stimulation experiments, media was replaced with Neurobasal containing 100 nM Phorbol 12-myristate 13 acetate (PMA) (Enzo Life Sciences, Farmingdale, NY, USA) for 1 hour. For the assays with PKC inhibition, media was replaced with Neurobasal containing 1 μM GF 109203X for 30 min prior to KCl or PMA treatment as above, with 1 μM GF 109203X included in the stimulation media. Cells were fixed and processed as described for in situ hybridization and immunocytochemistry. Neurons having pyramidal-like morphologies were selected, and apical dendrites were chosen for analysis based on structural properties. Images were acquired on a Nikon Eclipse E800 microscope using PictureFrame software (Optronics, Muskogee, OK, USA).

Immunocytochemistry (ICC)
Cultured neurons were fixed with 4% paraformaldehyde, blocked in a blocking buffer (10% BSA, 0.1% Triton X-100 in PBS) for 1 hour at room temperature, and incubated with primary antibodies in a dilution buffer (1% BSA, 0.1% Triton X-100 in PBS) overnight at 4°C. Mouse monoclonal antibody to HuD (16C12, Santa Cruz Biotechnology) was used at a 1:500 dilution for somatic expression analysis and at a 1:250 dilution for dendritic expression analysis. Rabbit polyclonal antibody to GFP (Clontech) was used at a 1:2,000 dilution. After washes, the coverslips were incubated with the appropriate DyLight 488, 594 or 649 secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) at a 1:1,000 dilution, washed, mounted with gelvatol and analyzed by fluorescence or confocal microscopy.

Fluorescent in situ hybridization (FISH)
In situ hybridization of cultured neurons was performed as described [11] with modifications. Antisense and sense RNA probes were synthesized from linearized plasmids using T3 and T7 RNA polymerases (Promega, Madison, WI, USA) and DIG-labeled ribonucleotides. Probe concentrations were 500 ng/ml for Bdnf mRNA and 100 ng/ml for GFP mRNA. For combined FISH and ICC, rabbit polyclonal antibody to GFP (Clontech) was included at a 1:2,000 dilution after the hybridization step. Coverslips were washed four times for 10 min in TNT buffer (100 mM Tris-Cl, pH7.5, 150 mM NaCl, and 0.05% Tween 20), and were incubated with secondary antibody as described above for ICC and washed again with TNT buffer. Then, in situ signals were amplified with the TSA Plus Fluorescein System (PerkinElmer) according to the manufacturer's instructions. The coverslips were mounted onto slides with gelvatol and examined by fluorescence or confocal microscopy. In situ hybridization images from the antisense probe and its sense control probe were taken with the same settings. Green fluorescence from GFP in transfected neurons was not detectable after the hybridization and wash steps.

Image analysis
All images were analyzed using NIH ImageJ software. Background was subtracted using a "rolling ball" algorithm. Dendritic processes were straightened and binned into 50 μm sections for ICC or 20 μm sections for combined FISH and ICC. The mean fluorescence for each bin was calculated as the mean of the brightest pixel of each line transecting the dendrite crosswise, emphasizing puncta representing areas of local synthesis. A minimum of 30 dendrites for each of 3 separate experiments were analyzed for each condition. Co-localization experiments were analyzed by an intensity correlation coefficient-based (ICCB) method using the JACoP plugin (http://rsbweb.nih.gov/ij/plugins/track/jacop.html) for ImageJ. Confocal images of Bdnf mRNA FISH and HuD ICC were thresholded to include most of the dendritic process, and Pearson's and Manders' coefficients were calculated.

RNA Electrophoretic Mobility Shift Assays (REMSA)
REMSAs were carried out as previously described [22] with modifications. Sense riboprobes were synthesized from linearized plasmids using T7 RNA polymerases (Promega) and biotinlabeled or unlabeled ribonucleotides (Roche Diagnostics). In vitro transcription reactions were treated with 0.1% SDS and 50 μg proteinase K at 55°C for 30 min, diluted to 130 μl and treated with 15 μl ammonium acetate stop solution (5 M ammonium acetate, 100 mM EDTA), extracted with an equal volume of phenol/chloroform, isopropanol precipitated at-70°C for 20 min, washed with 75% ethanol, and resuspended in 20 μl DEPC-treated ddH 2 O. For standard REMSA assays 2 nM labeled riboprobes were incubated with 2 μM recombinant GST-HuD or GST-HuDpd in 10 μl binding buffer (150 mM NaCl, 20 mM Tris, 1.5 mM MgCl 2 , 0.5 mM DTT, 0.5 μg/μl tRNA) for 20 min at room temperature, and for competition assays 8 nM unlabeled riboprobes were incubated with protein as above for 15 min before adding 2 nM labeled Nova1 A riboprobe and incubating an additional 20 min. Samples were run on a 5% polyacrylamide gel in TBE buffer and nucleic acids were transferred to positively-charged nylon membranes (Roche Diagnostics) followed by UV-crosslinking. Membranes were blocked in Odyssey blocking buffer (LI-COR Biosciences, Lincoln, NE, USA) + 1% SDS for 30 min, incubated with streptavidin DyLight 800 (Thermo Fisher Scientific) diluted 1:10,000 in Odyssey blocking buffer + 1% SDS for 30 min, washed 3X for 5 min with PBS + 0.1% Tween 20 followed by a final wash with PBS, then imaged using the Odyssey infrared imaging system (LI-COR Biosciences).

Neuronal Hu proteins are associated with Bdnf mRNAs in the mouse forebrain
Mouse Bdnf mRNAs can have either a short (0.35 kb) or long (2.85 kb) 3' UTR, depending on which of two polyadenylation sites is used for processing [24]. The long Bdnf 3' UTR has several AU-rich elements that could serve as binding sites for proteins that modulate RNA metabolism and translation. The neuronal ELAV (nELAV) family of RNA-binding proteins (RBPs), consisting of HuB, HuC and HuD, has been previously demonstrated to interact with several mRNAs to regulate their expression [21,25,26], and putative binding sites for these proteins are present within the long Bdnf 3' UTR, making them potential candidates for the trans-acting factors regulating local BDNF synthesis.
We immunoprecipitated mRNP complexes from lysates prepared from mouse forebrain using an antibody that recognizes all 3 nELAV RBPs. Real-time quantitative PCR analysis of the co-precipitated mRNAs was used to identify whether specific Bdnf mRNA isoforms were associated with the immunoprecipitated mRNPs, using Rpl10A mRNA for normalizing between samples, as it has been previously shown not to interact with nELAV proteins [22]. Primers for Bdnf mRNA isoforms I, IIc, IV and VI were used, as these species are found in relative abundance in adult mouse brain. ΔCt values were found to be significantly lower for mRNP IPs using the pan-nELAV antibody as compared to those using the mouse IgG isotype control, corresponding to a 12.4-, 16.9-, 13.3-and 17.3-fold change, respectively (Fig. 1A), demonstrating an in vivo association of nELAV RPBs with all 4 Bdnf mRNA isoforms examined.
It has been previously demonstrated that BDNF expression can be regulated by neuronal activity [24,27], so we decided to test whether the association of nELAV proteins with Bdnf mRNAs could be regulated in a similar way, as might be expected if this interaction is important in mediating activity-dependent changes in BDNF expression. We used intraperitoneal (IP) injection of seizure-inducing doses of pilocarpine to elevate neuronal activity and prepared forebrain lysate 30 minutes post-injection for use in mRNP immunoprecipitations as described above (Fig. 1B). At 30 minutes post pilocarpine injection, the abundance of Bdnf mRNA is not changed [13], which would simplify the interpretation of the data from mRNP immunoprecipitation. Ct values for samples from pan-nELAV antibody immunoprecipitations were normalized to Ct values for the corresponding total mRNP sample (pre-immunoprecipitation), and found to be lower when mice were injected with pilocarpine as compared to negative control mice injected with PBS, resulting in a 1.8-, 2.0-, 2.0-and 2.2-fold change in percent recovery of Bdnf mRNA isoforms I, IIc, IV and VI, respectively (Fig. 1C). Since protein kinase C (PKC) has been shown to phosphorylate nELAV RBPs to regulate their trafficking and mRNA stabilizing properties [18], we repeated the pilocarpine stimulation experiment in mice pre-treated with Ro-32-0432, a systemic inhibitor of protein kinase C. This resulted in failure of pilocarpine stimulation to increase the recovery of Bdnf mRNA-containing mRNPs (Fig. 1D), implicating PKC activity as an important mediator of this effect.

Bdnf 5' UTR sequences negatively regulate mRNA translation
To establish an in vitro system to study the effects of nELAV RBPs on the regulation of mRNAs containing Bdnf sequences, we decided to include both 5'-and 3'-UTR sequences in a reporter construct that expresses destabilized GFP modified to include a myristoylation peptide Figure 1. Quantitative RT-PCR analysis of in vivo mRNPs containing neuronal Hu proteins. (A) mRNPs were immunoprecipitated from mouse whole brain lysates using either the HuC/HuD neuronal protein antibody (16A11) or mouse IgG isotype control, and subjected to quantitative RT-PCR analysis using primers specific to Bdnf mRNA isoforms I, II, IV and VI, as well as primers for Rpl10a as reference control. Fold change was calculated as 2 (-ΔΔCt) , using values from the mouse IgG IP as control (Mean ± SEM; n = 3; *, p<0.05; **, p<0.01; student's t-test). (B) Schematic showing the drug injection protocol for seizure induction. Atropine MeNO 3 was injected alone or with Ro-32-0432 (systemic PKC inhibitor), followed 30 minutes later by injection of pilocarpine for seizure induction or PBS as a negative control, and mice were euthanized 30 minutes later, collecting forebrain tissue for isolation of mRNPs. (C) mRNPs were immunoprecipitated using lysates from mice injected with either pilocarpine or PBS. Percent recovery versus control was calculated as the ratio of the Ct for the IP sample to the Ct for the pre-IP sample, normalized to the value obtained from the PBS-injected mice (Mean ± SEM; n = 3; *, p<0.05; **, p<0.01; student's t-test). (D) The experiment from (C) was repeated with the modification that all mice were pre-injected with Ro-32-0432, a systemic PKC inhibitor (Mean ± SEM; n = 3; student's t-test).
doi:10.1371/journal.pone.0117264.g001 and a nuclear localization sequence (NLS), restricting movement of GFP signal distal to the site of synthesis and serving as a more accurate indicator of local translation ( Fig. 2A).
Inclusion of the 5' UTRs of Bdnf mRNA isoforms I, IIc, IV and VI reduced expression of a local translation reporter containing the long Bdnf 3' UTR with the first polyadenylation site mutated (AÃB) so that only the long 3' UTR species is generated (Fig. 2). The pattern of reduced expression of the local translation reporter was consistent between somatic and dendritic compartments (Fig. 2B-E), suggesting decreased translation or stability of the mRNAs containing the different Bdnf 5' UTRs. The 5' UTR from Bdnf mRNAs transcribed from promoter VI resulted in the largest decreases in somatic and dendritic reporter expression, 96% and 95%, respectively, while the 5' UTR from Bdnf mRNAs transcribed from promoter IV resulted in the smallest somatic and dendritic decreases, 25% and 39%. To potentially minimize confounds resulting from negative regulation of local translation acting on 5' UTR sequences we used the Bdnf 5' UTR from isoform IV in all further local translation reporter constructs.
PKC activation increases dendritic translation of mRNAs with the long Bdnf 3' UTR in an HuD-dependent manner It has previously been demonstrated that neuronal activity can increase local translation of Bdnf mRNAs in dendrites. This has been shown in hippocampal culture using chemical LTP induction through tetraethylammonium (TEA) treatment [13]. In order to gain more insight into the interaction between neuronal Hu RBPs and Bdnf mRNAs we investigated the role of PKC activation in local translation. The PKC pathway is known to be activated by TEA induction of LTP [28], and is important for the transport of nELAV RBPs to their sites of action [18], and thus may play a role in mediating any effects they may have on regulating local translation of Bdnf mRNAs.
Rat dissociated hippocampal cultures were transfected at 14 DIV with local translation reporter constructs containing either the short (ExIV A) or long (ExIV AÃB) Bdnf 3' UTRs ( Fig. 3A) and treated 2 days later for 1 hour with 30 mM KCl or 100 nM PMA. Neurons transfected with ExIV AÃB, but not ExIV A, showed significant increases in local dendritic translation compared to controls for both treatments (Fig. 3B-E). Furthermore, the effect of PMA treatment was completely abrogated in cultures pre-treated with the PKC inhibitor GF 109203X, whereas the effect of KCl stimulation was only partially reduced (Fig. 3F, G), suggesting other pathways besides PKC are involved in depolarization-dependent regulation of mRNAs with the AÃB sequence.
HuD, one of the more well-studied members of the nELAV RBP family, has been found to be upregulated in rat hippocampus after spatial learning, concomitant with increases in association with and stability of target mRNA, leading to increased translation [16,29]. We investigated the importance of HuD in the PKC-mediated upregulation of the local reporter construct detailed above by co-transfecting hippocampal cultures with a plasmid encoding an shRNA specific to rat HuD mRNA. HuD protein levels as detected by immunocytochemistry using a monoclonal antibody against HuD were decreased by 40% in the soma of co-transfected neurons (Fig. 4B, C), though there may be some cross-reactivity with the other nELAV proteins, in which case the actual reduction in HuD would be even greater than estimated. In cultures co-expressing the local synthesis reporter construct and the rat HuD shRNA, PMA treatment failed to increase dendritic reporter expression (Fig. 4D, E), suggesting an important role for HuD in mediating this effect. Interestingly, the basal level of dendritic reporter expression was increased compared to controls with normal HuD protein levels. Recovery experiments were conducted by co-transfecting the reporter and shRNA constructs with plasmids overexpressing either wild-type mouse HuD (mHuD) or a putative phosphorylation-deficient mutant (mHuDpd). In silico prediction of threonine phosphorylation sites in mouse HuD revealed 5 candidate residues, 2 with very high confidence scores, Thr149 and Thr165, which were substituted with alanine residues to generate mHuDpd (Fig. 4A). Overexpression of mHuD or mHuDpd in the presence of the shRNA lead to an increase in somatic HuD levels to 268% or 309% of control, respectively. Recovery with mHuD, but not with mHuDpd, restored the PKC-mediated increase in dendritic synthesis of the local reporter (Fig. 4D, E), indicating that the effect of the shRNA is specifically due to reduced HuD, and that one or both of the substituted threonine residues is a phosphorylation site of PKC necessary for mediating the ability of HuD to increase translation of its target mRNA. Recovery with mHuD resulted in a decrease in basal dendritic translation of the local reporter, while recovery with mHuDpd resulted in levels that were no different than those from the knockdown of HuD alone. These findings suggest that non-phosphorylated HuD inhibits translation of its target mRNAs, and this repression is relieved by PKC phosphorylation.

HuD can directly regulate local protein synthesis
If HuD interacts with Bdnf mRNAs to regulate their dendritic translation, they should co-localize in these compartments. Hippocampal cultures were transfected to express a GFP-mHuD fusion protein, and co-localization of this protein with Bdnf mRNA was assessed using dual fluorescent in situ hybridization (FISH) and immunocytochemistry. Intensity correlation coefficient-based (ICCB) analysis yielded a pearson's coefficient of 0.864, with 95% of the GFP-mHuD signal overlapping with Bdnf mRNA (Fig. 5A), suggesting a high degree of dendritic association even in the absence of stimulation. If these mRNP complexes of Bdnf mRNA with HuD represent RNA transport granules, HuD could be involved in regulating target mRNA trafficking to dendrites. Thus altering HuD levels through shRNA knockdown and heterologous overexpression would affect the dendritic localization of target mRNAs, potentially resulting in the observed differences in basal dendritic expression of the local reporter under decreased or elevated HuD levels. However, FISH with a probe directed against the Bdnf coding region showed no difference in Bdnf mRNA levels in the dendrites of hippocampal neurons after HuD knockdown or recovery (Fig. 5B, C), indicating that HuD is not essential for the dendritic trafficking of Bdnf mRNAs.
Increased neuronal activation has been shown to increase Bdnf mRNA trafficking to dendrites [30], and it is possible that PKC activation induced by PMA treatment could result in the observed increase in dendritic reporter synthesis as a consequence of an increased pool of dendritic reporter mRNAs, rather than by directly affecting local translation of pre-existing dendritic mRNAs. To distinguish between these two mechanisms we performed FISH on control or PMA-treated cultures transfected with the local reporter, using probes directed against the GFP coding sequence. After 1 hour of PMA treatment, the length of time used in previous experiments to assess local reporter synthesis, there was no detectable increase in dendritic reporter mRNA levels (Fig. 5D, E), indicating the increased local reporter synthesis after this treatment is a result of increased translation of pre-existing mRNAs. However, 3 hours of PMA treatment led to significantly increased dendritic reporter mRNA levels, suggesting a complex role for PKC in regulating dendritic translation of mRNAs with the Bdnf long 3' UTR, mediating both rapid, short-term increases by acting directly on local translation and slower, longterm increases through increased trafficking of mRNAs to sites of activity.

HuD dendritic localization requires threonine phosphorylation
The trafficking of HuD has been reported to depend on its phosphorylation by PKC [18]. Immunocytochemistry with a monoclonal antibody directed against HuD revealed an increase in dendritically-localized HuD after overexpression of mHuD, but not mHuDpd (Fig. 6A, B), which is consistent with the observed changes in basal dendritic translation of the local reporter if HuD is a repressor of translation under resting conditions. Dendritic levels of HuD were further increased after PMA treatment only in neurons overexpressing mHuD (Fig. 6A, B), further demonstrating the likelihood that one or both of the substituted threonine residues in mHuDpd is phosphorylated by PKC to regulate its translocation to dendrites and relieve repression of its target mRNAs.
HuD interacts directly with Bdnf mRNA sequences mRNP complexes can be composed of multiple proteins and distinct mRNAs, and while the previous data has indicated an association of HuD with Bdnf mRNAs [19], whether HuD interacts with the long Bdnf 3' UTR is unknown. REMSAs were carried out with recombinant GST-HuD or GST-HuDpd and labeled RNA probes corresponding to the coding sequence and 3' UTR of Bdnf (Fig. 7A, B). All of the probes corresponding to long Bdnf 3' UTR sequences, B1, B2 and B3, showed an upward shift in the presence of GST-HuD or GST-HuDpd, with the greatest effect observed for probes B1 and B3, while the Bdnf coding sequence probe, CDS, and the short 3' UTR probe, A, showed little or no shift (Fig. 7C, E). To complement these results we performed competition REMSAs, adding a labeled probe corresponding to the Nova1 3' UTR, a known target of HuD [22], to GST-HuD or GST-HuDpd that was pre-incubated with unlabeled Bdnf probes. Only the long Bdnf 3' UTR probes were able to compete for binding with the recombinant proteins and prevent an upward shift of the labeled Nova1 3' UTR probe (Fig. 7D, F). These findings demonstrate a direct interaction between HuD and sequences in the long Bdnf 3' UTR, and further show that substitution of the two threonine residues with alanines has no effect on the mutant protein's ability to interact with its target mRNAs, validating its use in earlier experiments to determine the relevance of the two residues as potential PKC phosphorylation sites.

Bdnf mRNAs interact directly with HuD
The neuronal Hu proteins are likely candidates for being trans-acting factors regulating the stability and translation of Bdnf mRNAs, based on the presence of HuD-binding motifs in the long Bdnf 3' UTR [31], association of HuD with Bdnf mRNA in cultured hippocampal neurons [19], and related functional consequences arising from perturbations in neuronal Hu protein levels or Bdnf mRNA expression [32,33]. We performed quantitative RT-PCR analysis on RNAs co-immunoprecipitated from mouse forebrain lysate with a pan neuronal Hu antibody and found enrichment of all Bdnf mRNA isoforms examined, indicating their in vivo association with neuronal Hu proteins.
We next sought to uncover whether this association could be modulated by neuronal activation, as might be expected if HuD were involved in mediating activity-dependent changes in BDNF expression. We repeated the quantitative RT-PCR experiments using tissue from animals with elevated neuronal activity as a result of seizures induced by pilocarpine injection, in the presence or absence of a systemic inhibitor of PKC. Elevated neuronal activation significantly increased the percent of Bdnf mRNAs co-immunoprecipitated as a fraction of input RNA compared to animals with basal neuronal activity, and systemic inhibition of PKC completely reversed this increase, indicating the activity-dependent modulation depends on PKC phosphorylation. This increased co-immunoprecipitation could result from increased formation of mRNP complexes containing neuronal Hu proteins and Bdnf mRNAs, or the stabilization of existing complexes following HuD phosphorylation by PKC.
To determine whether there is a direct interaction between neuronal Hu proteins and Bdnf mRNA we performed REMSAs using purified recombinant HuD and labeled RNA probes corresponding to sequences in the coding region and 3' UTR of Bdnf. We also performed competition assays using unlabeled Bdnf probes and a labeled probe corresponding to a sequence in the Nova1 3' UTR previously shown to interact with HuD [22]. Only probes corresponding to sequences in the long Bdnf 3' UTR were found to interact strongly with HuD in both assays, indicating the presence of multiple HuD binding sites throughout this region. These findings do not contradict with the observation that HuD interacts with the short Bdnf 3' UTR [19], as we also observed this interaction, although it is much weaker than the interaction between HuD and the B segment of the long Bdnf 3' UTR.
The physiological relevance of these results is highlighted by the similar subcellular distributions of Bdnf mRNAs and HuD, with both found either in translationally repressed mRNP complexes or associated with polysomes, sites of active translation [13,34]. Activation of PKC results in the nucleocytoplasmic shuttling of Hu proteins, their threonine phosphorylation, and increased association of the cytoskeleton and polysomes [18], establishing a pathway whereby neuronal activity could lead to HuD-mediated increase in BDNF synthesis.

Bdnf 5' UTRs repress translational efficiency
A recent study found that 5' UTRs play an important role in regulating translation of Bdnf mRNAs in SH-SY5Y cell line [35]. Our study indicates that 5' UTRs also control translation of Bdnf mRNA in cultured primary neurons. We found that including 5' UTRs from Bdnf isoforms I, IIc, IV and VI in a local reporter construct decreased both somatic and dendritic reporter expression, with the relative changes being similar between the two compartments. The observed translational repressive effect of exons I, IIc, IV and VI is in agreement with the recently published analysis of rat 5' UTR sequences on translatability of Bdnf mRNA [35]. The isoform IV 5' UTR conferred the least reduction in translation, while the isoform VI 5' UTR almost completely blocked expression. These 2 Bdnf isoforms are among the most abundant in adult rat hippocampus [13,36], with isoform IV expression being more activity-dependent than that of isoform VI [24,37]. It has also been reported that isoform VI, but not isoform IV, is targeted to dendrites after neuronal activation [38]. It seems contradictory that the Bdnf mRNA isoform least increased by activity would be most localized to distal dendrites, where activity-dependent local translation could be an important process leading to synaptic remodeling.
PKC activation can increases local dendritic translation of RNAs with the long Bdnf 3' UTR Expression of a local reporter with the long, but not the short, Bdnf 3' UTR was increased in the distal dendrites of hippocampal neurons stimulated either by KCl depolarization or PMAmediated PKC activation. This result is in agreement with a previous finding that chemical LTP induction through TEA treatment can selectively increase dendritic expression of a local translation reporter with the long, but not the short, Bdnf 3' UTR [13]. Pre-treatment of cultures with an inhibitor of PKC completely abolished the increased dendritic translation after PMA treatment, as expected if PMA is acting through PKC, and was able to reduce, but not block, the increased reporter synthesis after KCl stimulation, suggesting that depolarization can trigger PKC-dependent and-independent pathways to increase translation of mRNAs with the long Bdnf 3' UTR. Dendritic reporter translation could be modulated either by a change in the pool of dendritic reporter mRNAs or by a direct effect on the rate of translation of pre-existing dendritic mRNAs. While we found that 3 hours of PMA treatment was able to increase reporter mRNA levels in dendrites, the 1 hour treatment used in reporter assay experiments did not result in any significant difference, suggesting the PKC activity-dependent increase in reporter expression is due to a mechanism operating to increase the translation of pre-existing local mRNAs.
The regulation of BDNF synthesis in dendrites by PKC may be an important pathway involved in learning and memory. PKC activity has been shown to be increased during both the induction and maintenance phases of LTP at Schaffer collateral-CA1 synapses [39], and the constitutively active, atypical PKC isozyme, PKM(, is necessary and sufficient for the maintenance phase of LTP [40]. Since LTP induction does not require protein synthesis, the relevance of PKC-mediated increases in dendritic BDNF translation is likely during LTP maintenance. PKM (activity during this period could increase BDNF synthesis in stimulated spines, leading to enhanced TrkB activation and downstream signaling through Akt-mTOR and Ras-ERK pathways to remodel the spine and increase synaptic strength. Despite the lack of a requirement for protein synthesis during the induction phase of LTP, synthesis of BDNF in activated spines during this period may be physiologically relevant to prime these spines with additional stores of BDNF to facilitate the later, protein synthesis-dependent maintenance phase of LTP.
PKC activity can relieve HuD-mediated translational repression of mRNAs with the long Bdnf 3' UTR Knocking down endogenous HuD in rat hippocampal cultures blocked the ability of PKC activation to increase dendritic translation of mRNAs with the long Bdnf 3' UTR. Unexpectedly, this also resulted in a significant increase in dendritic expression of the local reporter in unstimulated cultures. Rescue with wild-type mouse HuD increased dendritic HuD above control levels and restored the response of the local reporter to PKC activation, while decreasing its basal dendritic translation. These data are consistent with HuD either restricting dendritic trafficking of mRNAs with the long Bdnf 3' UTR or inhibiting their translation under basal conditions. Rescue with a putative phosphorylation-deficient mutant of HuD failed, resulting in elevated basal reporter expression that is insensitive to PKC activation, similar to HuD knockdown alone. One or both of the threonine residues substituted in the mutant are likely PKC phosphorylation sites, as PKC activation was able to increase dendritic localization of the overexpressed wild-type HuD, but not the mutant. Its failure to decrease basal dendritic reporter translation, despite retaining its ability to bind Bdnf mRNA sequences and its high somatic expression, argues against a role for HuD in regulating dendritic mRNA trafficking. Furthermore, we did not observe any changes in the amount of Bdnf mRNA in the dendrites of hippocampal neurons after knockdown or overexpression of HuD.
Our data suggests a model where HuD can repress translation of mRNAs with the long Bdnf 3' UTR, and PKC activation can relieve this repression, possibly by direct phosphorylation of HuD itself (Fig. 8). While many reports on HuD find that it increases translation of its target mRNAs [22,41,42], it has also been found to inhibit translation of p27 [43] and tau [44], and 2 of its RNA binding sites have been identified as the most overrepresented motifs in the 3' UTRs of mRNAs downregulated after expression of miRNAs, and in the 3' UTRs of mRNAs bound by Argonaute [45], an integral component of P-bodies, sites of translational repression [46]. Thus HuD could be an inhibitor of translation in certain contexts, such as for genes regulated by miRNA inhibition, like Bdnf [47,48]. Notably, the long Bdnf 3' UTR has been found to repress basal translation [13], and our results indicate HuD has multiple binding sites throughout this region. Further studies are necessary to demonstrate whether the absence of HuD binding can increase the proportion of Bdnf mRNAs associated with polysomes, and if PKC phosphorylation on either of the two threonine residues mutated in this study is the signal leading to polysome association.
Neuronal activity should regulate dendritic BDNF synthesis through other signaling pathways in addition to the PKC-HuD pathway, because we found that a PKC inhibitor was able to reduce, but not block, the increased dendritic synthesis of a reporter with the long Bdnf 3' UTR after KCl-induced depolarization. Activation of the NMDA receptor has been shown to stimulate dendritic synthesis of Ca 2+ /calmodulin-dependent protein kinase II α subunit through phosphorylation of translation elongation factor eEF2 or cytoplasmic-polyadenylation-element-binding protein [49,50]. It would be interesting to investigate whether these signaling pathways also regulate dendritic BDNF synthesis in future studies.