Abstract
The astrocyte brain fatty acid binding protein (Fabp7) has previously been shown to have a coordinated diurnal regulation of mRNA and protein throughout mouse brain, and an age-dependent decline in protein expression within synaptoneurosomal fractions. Mechanisms that control time-of-day changes in expression and trafficking Fabp7 to the perisynaptic process are not known. In this study, we confirmed an enrichment of Fabp7 mRNA and protein in the astrocytic perisynaptic compartment, and observed a diurnal change in the intracellular distribution of Fabp7 mRNA in molecular layers of hippocampus. Northern blotting revealed a coordinated time-of-day-dependent oscillation for the Fabp7 mRNA poly(A) tail throughout murine brain. Cytoplasmic polyadenylation element-binding protein 1 (CPEB1) regulates subcellular trafficking and translation of synaptic plasticity-related mRNAs. Here we show that Fabp7 mRNA coimmunoprecipitated with CPEB1 from primary mouse astrocyte extracts, and its 3′UTR contains phylogenetically conserved cytoplasmic polyadenylation elements (CPEs) capable of regulating translation of reporter mRNAs during Xenopus oocyte maturation. Given that Fabp7 expression is confined to astrocytes and neural progenitors in adult mouse brain, the synchronized cycling pattern of Fabp7 mRNA is a novel discovery among known CPE-regulated transcripts. These results implicate circadian, sleep, and/or metabolic control of CPEB-mediated subcellular trafficking and localized translation of Fabp7 mRNA in the tripartite synapse of mammalian brain.
Introduction
Many behavioral and physiological processes exhibit diurnal control. Sleep propensity (Fuller et al., 2006), feeding (Saper, 2006), long-term potentiation (Harris and Teyler, 1983; Raghavan et al., 1999; Chaudhury et al., 2005), and memory formation (Chaudhury and Colwell, 2002; Lyons et al., 2005, 2006a,b; Lyons and Roman, 2009) follow diurnal rhythms. Circadian genes affect complex physiological manifestations in many organisms (Andretic et al., 1999; Garcia et al., 2000; Sarov-Blat et al., 2000; Abarca et al., 2002; Dudley et al., 2003; Sakai et al., 2004; McClung et al., 2005; Turek et al., 2005; Feillet et al., 2006; Eckel-Mahan and Storm, 2009; Gerstner and Yin, 2010; Kondratova et al., 2010), suggesting that clock molecules influence various plasticity and neurophysiological processes (Herzog, 2007; Roth and Sweatt, 2008; Eckel-Mahan and Storm, 2009; Gerstner et al., 2009; Gerstner and Yin, 2010; Masri and Sassone-Corsi, 2010). Here we focus on time-of-day control of brain fatty-acid binding protein (Fabp7). Fabp7 has a unique synchronized pattern of global diurnal expression in astrocytes (Gerstner et al., 2006, 2008), and Fabp7 knock-out mice have altered NMDA-receptor-dependent current responses and memory (Owada et al., 2006), indicating astrocyte Fabp7 expression is critical to normal neurophysiological mechanisms and behavior.
Astrocytes are a major player in brain plasticity (Halassa and Haydon, 2010), and modulate neurophysiological mechanisms, such as LTP and LTD (Nishiyama et al., 2002; Yang et al., 2003; Pascual et al., 2005), and homeostatic synaptic scaling (Panatier et al., 2006; Stellwagen and Malenka, 2006; Turrigiano, 2006). Astrocytes have been implicated in the regulation of complex behavior (Halassa and Haydon, 2010), such as circadian rhythms (Li et al., 2002; Suh and Jackson, 2007), sleep (Halassa et al., 2009a), and learning and memory (Nishiyama et al., 2002; Gibbs et al., 2006; Bracchi-Ricard et al., 2008; Gibbs and Hertz, 2008; Suzuki et al., 2011). Astrocytic adenosine contributes to sleep disruption of hippocampal-dependent memory and synaptic plasticity (Florian et al., 2011), and its release is controlled by the circadian clock (Marpegan et al., 2011). These data suggest astrocytic cellular and molecular processes contribute significantly to linking behavioral and physiological processes.
Fabp7 is enriched near synapses. Its abundance in synaptoneurosomal fractions declines with age (Pu et al., 1999). We found Fabp7 abundance cycles in synaptoneurosomes, suggesting daily variations in trafficking machinery. One potential mechanism is through mRNA-binding proteins, such as cytoplasmic polyadenylation element (CPE)-binding protein (CPEB). One such protein, CPEB1, is present in astrocytes (Jones et al., 2008), but its role in targeting mRNAs to perisynaptic processes in astrocytes has not been explored. CPE-mediated mechanisms exist for trafficking and local translation of neuronal mRNAs, so we conducted a bioinformatic search that revealed phylogenetically conserved CPE sequences in Fabp7 mRNA. We hypothesized that Fabp7 CPEs are functional and able to mediate translation, predicting Fabp7 mRNA could undergo time-of-day subcellular translocation and polyadenylation. Our novel findings confirm that, in addition to neurons, cytoplasmic polyadenylation processes exist for astrocytic CPE-containing mRNA (Jones et al., 2008) and establish the existence of coordinated control for mRNA trafficking and subsequent translation at the perisynaptic astrocytic process compartment across the day.
Materials and Methods
Subjects and handling
Male mice (C57BL/6) were either purchased directly from Harlan Sprague Dawley (adult) or taken at specific ages from a breeding colony. Animals were entrained to a 12 h light, 12 h dark (12:12) schedule for a minimum of 2 weeks. Zeitgeber time (ZT) 0 was lights-on, and ZT12 was lights-off, on this diurnal schedule. For tissue collection, animals were killed by decapitation, their brains dissected, flash frozen at −30°C in 2-methylbutane, and stored at −80°C. For tissue punches of specific regions, brains were embedded in optimal cutting temperature media and sectioned on a Leica (CM3050) cryostat at −25°C as previously described (Gerstner et al., 2006, 2008). Punches from the same brain region were pooled from each of the animals within the same group, and stored at −80°C until RNA isolation. All animal care and use procedures were in strict accordance with University of Wisconsin Institutional Animal Care and Use Committee and National Institutes of Health guidelines for the humane treatment of laboratory animals.
Synaptoneurosome preparation
Synaptoneurosomes (SNs) were prepared from mouse whole brain as described previously (Westmark and Malter, 2007; Westmark et al., 2011). Briefly, mice were asphyxiated with carbon dioxide followed by removal of the brain. Brains were washed in ice-cold gradient medium (GM) buffer (0.25 m sucrose, 5 mm Tris, pH 7.5, and 0.1 mm EDTA), transferred to a glass Dounce homogenizer containing ice-cold GM buffer, and gently homogenized with five strokes of the loose pestle followed by five strokes of the tight pestle. The homogenate was spun at 1000 g for 10 min at 4°C in round-bottom tubes to pellet cellular debris and nuclei. The supernatant (2 ml of aliquots) was applied to Percoll gradients (layers, 2 ml each of 23%, 15%, 10%, and 3% isomotic Percoll) and spun at speed (32,500 g) for 5 min at 4°C. The third band from the top of the gradient (23%–15% interface) containing intact SNs was removed and pooled for the experiments.
Immunocytochemistry and electron microscopy
Animals were perfused through the heart with 0.1% heparin in 0.01 m PBS, pH 7.4, until liver cleared, followed by 2% paraformaldehyde, 1% glutaraldehyde in 0.1 m phosphate buffer (PB), pH 7.4. Sections were cut at 50 μm on a vibratome and stored in 0.01 m PBS and 0.01% azide solutions until processing. The hippocampal section was prepared from a mouse killed at ZT12.
Following fixation, the slices were rinsed in 0.1 m PB and then treated with freshly prepared 0.1% sodium borohydride in 0.1 m PB for 10 min to quench unbound aldehydes. Next, the slices were permeabilized for 30 min with 0.1% Triton X-100 in PBS followed by three 10 min rinses in PBS. The slices were then blocked for nonspecific antibody binding with Aurion goat-blocking agent (Cat. #25596, containing 5% BSA, 0.1% cold water fish gelatin, and 5% normal goat serum in PBS, pH 7.4) for 30 min. The slices were then rinsed three times for 10 min each in incubation buffer containing PBS + 0.1% BSA-C (Aurion Immuno Gold Reagents, Cat. #25558). Finally, the samples were incubated in a mixture of incubation buffer and rabbit primary Fabp7 antibody (Millipore) at a dilution of 1:1000 overnight at 4°C. The following day, the slices were rinsed in incubation buffer every 10 min for 2 h (12 rinses total), and then incubated overnight in a 1:100 dilution of ultrasmall gold conjugate [F(ab′) 2 fragment of goat-anti-rabbit IgG (heavy and light chains), Aurion Immuno Gold Reagents, Cat. #25360] at 4°C. After the overnight incubation, the slices were rinsed in incubation buffer 6× for 10 min each, and then rinsed in PBS 6× for 5 min each. The slices were postfixed in 2% glutaraldehyde in 0.1 m PB for 30 min and rinsed 2× for 5 min each in PB.
The use of ultrasmall gold conjugates requires a silver-enhancing step to increase the size of the subnanometer gold particles to more easily visualize the particles by transmission electron microscopy. After postfixation, the slices were rinsed 3× for 10 min each in Enhancing Conditioning Solution (ECS) (Aurion). Next, the slices were developed in Silver Enhancement Solution (Aurion) for 1.25 h. To terminate silver enhancement, the slices were exposed to a solution of 0.3 m sodium thiosulfate in ECS for 5 min. The slices were then rinsed twice in 10 min intervals in ECS and finally processed routinely for electron microscopy.
After silver enhancing, the slices were fixed in a solution of 0.5% osmium tetroxide in 0.1 m PB for 30 min. Following OsO4 postfixation, the samples were dehydrated in a graded ethanol series, then further dehydrated in propylene oxide and flat-embedded in Epon epoxy resin. The specific regions used for the analysis were trimmed from the flat-embedded sections and flat-mounted on blank epoxy resin stubs. Excess epoxy resin was then sectioned away in 1 μm increments with a Diatome Histo diamond knife using a Reichert–Jung Ultracut-E ultramicrotome until the surface of the tissue was exposed. Samples were then sectioned using a Diatome Ultra 45 diamond knife with the same microtome and contrasted with Reynolds lead citrate and 8% uranyl acetate in 50% EtOH. Ultrathin sections were observed with a Philips CM120 electron microscope and images were captured with a MegaView III side-mounted digital camera.
Northern blot analysis
Northern blotting was performed as described previously (Gerstner et al., 2006). Briefly, brains were dissected and total RNA was isolated using TRIzol (Invitrogen), according to the manufacturer's specifications, and stored at −80°C. Before loading, each sample was incubated with sample buffer [7.5% formaldehyde, 43% formamide, 12% 10× 3-(N-morpholino)propanesulfonic acid, 0.12% ethidium bromide] for 5 min at 56°C, and then cooled on ice for 5 min. Northern blots consisted of 1 μg of total RNA per lane and were electrophoresed on a 1.2% agarose gel. RNA was transferred overnight onto a sheet of GeneScreen Plus nitrocellulose (NEN Life Science Products). DNA templates generated by PCR were labeled with 32P using the Megaprime DNA labeling system (GE Healthcare). The following primers were used: Fabp7 forward primer, AGACCCGAGTTCCTCCAGTT, reverse primer, CCTCCACACCGAAGACAAAC, Cpeb1 forward primer, CATCTTGGGACCTTCTTGGA, Cpeb1 reverse primer, AAGACCCAAGGGATTACC, Cpeb2 forward primer, GGTCGCTCTTCCCTATTTCC, Cpeb2 reverse primer, TCCAAAGGCTGAGAACCATC, Cpeb3 forward primer, CCCTTCTCCAGCAATGTGAT, Cpeb3 reverse primer, CTCGTTCCCCATTTTGACAT, Cpeb4 forward primer, CCTCACTGCTTCACTCACCA, Cpeb4 reverse primer, AACAAAGCGGGCACTGATAG. Probes were mixed with hybrisol and incubated overnight at 42°C. Blots were washed 3× with 100 ml of 2× SSC, 1% SDS, at room temperature, then twice for 30 min each with 100 ml of 2 × SSC, 1% SDS, at 50°C, and finally once for 30 min with 100 ml of 0.5 × SSC, 1% SDS, at 60°C. Following washes, blots were exposed to a phosphoscreen for 3 d and image analysis was performed using the Storm 860 and ImageQuant 5.2 software (Molecular Dynamics). A 983 bp PCR product for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was generated as described previously (Gerstner et al., 2006).
RNase H and poly(A) analysis
The RNase H digestion procedure was modified from previously published methods (Goodwin and Rottman, 1992). Briefly, 5 μg of total RNA was incubated at 80°C for 10 min, rapid cooled to 20°C, and combined with or without 0.15 μg of Oligo-dT(12–18) in 1 μm EDTA for 15 min. The addition of 50 U RNase H (Promega) was added to RNase H buffer (20 mm HEPES-KOH, pH 7.8, 50 mm KCl, 10 mm MgCl2) at a final volume of 50 μl, and digested at 37°C for 30 min. RNA was isolated from the reactions by phenol-chloroform extraction, ethanol was precipitated, and Northern blotting was performed as above.
Densitometric analysis of the poly(A) tail was performed using ImageQuant software. Briefly, a 1 × 3 pixel window (i.e., the width of a lane in the Northern blot × 3 pixel window) was generated, and the starting spot was determined at ZT 12 by the lane that contained the time point with the lowest poly(A) tail size. The starting spot within this lane was located above the resolved band with a signal intensity value nearest the background value. This height on the gel was determined to be the starting spot across all lanes. Identical boxes were placed at this level in adjacent lanes for all ZT time points, and a signal intensity value was determined within each box. The boxes were then moved up a single pixel, and the new signal intensities tabulated (single rung). Signal intensity values for each rung were done repeatedly until the lane with the last value closest to the background value was tabulated (equal to longest poly(A) tail). Tabulated values for each rung were averaged, and this value was used to generate a normalized value across all time points within a given single rung. These normalized values within a given time point were averaged across all rungs to control for relative signal intensity. This normalization controls for time-of-day changes in transcript intensity of the band, but retains a signal above noise/background for calculating band length.
Quantitative reverse transcriptase PCR
All quantitative reverse transcriptase PCR (qRT-PCR) was performed at the University of Wisconsin-Madison Genome Center. Complementary DNA was generated using Superscript II Reverse Transcriptase (SSRII; Invitrogen) as follows: 5 μg of total RNA was added to 100 pmol of T7 oligo-dT primer [GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-d(T)24] and the reaction volume was brought to 12 μl of total volume with PCR grade water. The reaction was incubated at 70°C for 10 min in a Mastercycler ep thermocycler (Eppendorf) and placed on ice for 2 min. Five times SSRII buffer (4 μl), 2 μl of 0.1 m DTT, and 1 μl of 10 mm deoxyribonucleotide triphosphates were added and the reaction was incubated in a thermocycler for 2 min at 42°C. After incubation, 1 μl of SSRII was added and the reaction was incubated at 42°C for 1 h. CDNA was then quantified using an ND-1000 Spectrophotometer (NanoDrop Technologies) and normalized to 20 ng/μl with PCR grade water.
PCR methods
Individual reactions contained 1× Probes MasterMix (F. Hoffmann-La Roche), 0.05 U uracil-N-glycosylase, 1× LCGreen Plus (Idaho Technology), 200 nm each primer, and 20 ng of cDNA. The following primers were used: Fabp7 forward primer, CTCTGGGCGTGGGCTTT, reverse primer, TTCCTGACTGATAATCACAGTTGGTT, β2M forward primer, CATGGCTCGCTCGGTGACC, β2M reverse primer, AATGTGAGGCGGGTGGAACTG, β-actin forward primer, AGAGGGAAATCGTGCGTGAC, β-actin reverse primer, CAATAGTGATGACCTGGCCGT, GAPDH forward primer, CATGGCCTTCCGTGTTCCTA, GAPDH reverse primer, ATGCCTGCTTCACCACCTTCT. The total reaction volume of 5 μl was overlaid with 10 μ of Chillout Wax (Bio-Rad Laboratories). All reactions were run on a LightCycler 480 Real-Time PCR System (F. Hoffmann-La Roche). The cycling parameters were 50°C for 5 min, 95°C for 10 min, and then 50 cycles of 95°C for 15 s and 60°C for 30 s. Fluorescence data were collected at a wavelength setting of 450–500 nm. Analysis was performed using both LightCycler 480 proprietary software and qBASE v1.3.5 (Jan Hellemans, Center for Medical Genetics, Ghent University, Belgium).
In situ hybridization
In situ hybridization was performed as previously described (Gerstner and Landry, 2007). Briefly, postfixed, cryostat sections (20 μm) were pretreated with proteinase K (0.2 μg/ml; Promega) and hybridized overnight at 55°C in 150 μl of 35S-labeled antisense riboprobe (10,000 cpm/μl). Following posthybridization washes, sections were exposed to a phosphoscreen for 6 d. Image analysis was performed using the Storm 860 and ImageQuant 5.2 software (Molecular Dynamics). For densitometric analysis of in situ hybridization data, specific regions from a minimum of four sections were averaged per animal per time point, and normalized to background as described previously (Gerstner and Landry, 2007).
Antisense (35S-labeled) was made from Fabp7 PCR templates derived from T7-tagged reverse or forward primers, respectively. The following primers were used: Forward primer: AGACCCGAGTTCCTCCAGTT, reverse primer, CCTCCACACCGAAGACAAAC. Fabp7 template used for in vitro transcription was generated using standard PCR conditions. T7 sequence tag extending the reverse primer was GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG. In vitro transcription reactions were performed as described by Promega Biotech.
For emulsion autoradiography, sections processed for in situ hybridization (ISH) were dipped in NTB 3 liquid emulsion (Eastman Kodak) under safelight conditions and stored at 4°C for 6 weeks. Processing was as described by the manufacturer. For analysis of diurnal time points, all sections within a series were processed under identical conditions in the same run (N = 2–5 animals per group). Images of ISH emulsion-dipped hippocampal sections were taken using darkfield microscopy, and saved as Tagged Image File Format (TIFF) files for analysis. The region of hippocampus was analyzed using ImageJ software (National Institutes of Health) extending from interaural plane 1.62 to 2.22 mm as defined by Paxinos and Franklin (2003). Counts were tabulated, background was subtracted, and averaged. Briefly, a standardized ∼230 × 230 μm box was made that covered the precursor layer and extended through the granular cell layer and into the molecular layer of the dentate gyrus of the hippocampus. Each individual animal had ∼4 sections that were analyzed per time point and these were then averaged to generate a single number that was used as an N = 1. Final counts for each time point were calculated by averaging across all animals in each group. Density measures were used to calculate distribution of silver grains by normalizing signal intensity values of the molecular layer against the highest average background-subtracted value within the 230 × 230 μm box. These silver grain distribution values (background-subtracted pixel counts per maximal pixel count) were then averaged and divided by the square area (μm2) analyzed. Individual samples were averaged within groups, plotted with SEM bars, and subjected to ANOVA statistical analysis.
Western blotting
Samples prepared from whole brain were subjected to SDS-PAGE by separating 10 μg of protein per lane on a 10–20% gradient precast gel (Biorad). Protein was transferred to 0.2 μm Protran nitrocellulose membranes (PerkinElmer), blocked in 5% dried milk powder in 50 mm Tris-HCl, pH 7.5, 15 mm NaCl, 0.5% Tween-20 (TBST), washed briefly in TBST and incubated in primary antibody in TBST overnight at 4°C. Fabp7 antibody (Millipore) was used at 1:1000 dilution in TBST; antibody to β-actin (Imgenex) was used at 1:10,000 dilution; and anti-glyceraldehyde-3-phosphate dehydrogenase (Imgenex) was used at 1:5000 dilution. Blots were then washed three times in TBST, incubated for 1 h in antirabbit horseradish peroxidase-conjugated secondary antibody (1:7500 in TBST; KPL), washed three times in TBST and incubated in ECL plus Western blotting detection reagent for 5 min based on the manufacturer's instructions (GE Healthcare). For time-of-day synaptoneurosomal blots, Fabp7 antibody was used at 1:2000 dilution, PSD-95 (MAB1598, Millipore) was used at 1:1000, and normalization to total protein was evaluated on the same blots using India ink (Higgins Black Magic, No. 44011). Visualization was performed by chemoluminescence scanning (medium sensitivity, photomultiplier voltage, 600 V) on a Typhoon 9410 PhosphorImager (GE Healthcare). Densitometric analysis was performed using ImageQuant version 5.2.
Tissue culture
Dissociated primary cortical astrocyte culture tissue was a gift from the laboratory of Dr. Dandan Sun (University of Wisconsin-Madison) and prepared as described previously (Su et al., 2002). Briefly, cerebral cortices were removed from 1-d-old mice. The cortices were incubated in a solution of 0.25 mg of trypsin per milliliter of HBSS for 25 min at 37°C. The tissue was then mechanically triturated and filtered through nylon meshes. The dissociated cells were rinsed and resuspended in Eagle's minimal essential medium (EMEM) containing 10% fetal bovine serum. Viable cells (1 × 104 cells per well) were plated in 24-well plates coated with collagen type 1. Cultures were maintained in a 5% CO2 atmosphere at 37°C. To obtain morphologically differentiated astrocytes, confluent cultures (days 12–15 in culture) were treated with EMEM containing 0.25 mm dibutyryl cAMP (DBcAMP) for 7 d to induce differentiation.
Immunoprecipitation
To assess the relative binding of Fabp7 mRNA to CPEB, coimmunoprecipitation experiments were done according to those previously described by Brown et al. (2001). Briefly, mouse brain astrocyte culture was lyzed in 1 ml of ice-cold buffer (10 mm HEPES, pH 7.4, 200 mm NaCl, 30 mm EDTA, and 0.5% Triton X-100) with 2× complete protease inhibitors (Boehringer Mannheim) and 400 U/ml RNasin (Promega). Nuclei were pelleted, the supernatant was precleared, and an aliquot (20%) of this precleared supernatant (“input”) was saved for RNA extraction. The remaining lysate was immunoprecipitated with anti-CPEB antibody (Affinity BioReagents). The immunoprecipitate was resuspended in DEPC-treated water for protein analysis and RNA extraction.
Xenopus oocyte maturation assay
Reporter constructs.
The 277 nt long Fabp7 3′UTR cDNA sequence (as in NCBI accession no. BC055280) was directionally cloned downstream of the luciferase open reading frame in pLuci-XlGLD-2A(S) (Rouhana and Wickens, 2007) using XhoI and SpeI restriction endonucleases. For mutational analysis of the Fabp7 3′UTR, sense and antisense oligonucleotides corresponding to the last 100 nt of the Fabp7 3′UTR wild-type sequence and mutations (3XCPE mutant and hexanucleotide mutant) were synthesized (Integrated DNA Technologies), annealed, and ligated downstream of the luciferase ORF as above. Control constructs for Gld-2 have been previously described (Rouhana and Wickens, 2007), as have the construct used for transcription of β-galatosidase mRNA (Gray et al., 2000).
In vitro transcription of capped reporter mRNAs.
DNA templates were linearized with SpeI or BamHI and used as substrate for transcription of capped mRNAs using Megascript T7 RNA polymerase (Ambion) in the presence of 6 mm m7GpppG (New England Biolabs). Transcripts were DNase treated, ethanol precipitated, and resuspended in RNase-free water.
Injection of mRNA and manipulation of Xenopus oocytes.
Oocyte removal, injection, and the induction of meiotic maturation were performed as previously described (Ballantyne et al., 1997) with the following minor modifications. MRNAs were injected at ∼2–3 fmol of the luciferase reporter mRNA, along with 12 fmol of β-gal reporter mRNA in 50 nL of RNase-free water. Oocytes were then allowed to recover for 2 h and left intact or stimulated for meiotic maturation with 10 μg/ml of progesterone overnight at room temperature, except for time points indicated in Figure 5C.
Luciferase and β-galactosidase measuring assay.
Groups of five oocytes were pooled and analyzed for luciferase and β-galactosidase activity as previously described (Gray et al., 2000).
Results
Location
Fabp7 tripartite synapse enrichment
The localization of Fabp7 in the mammalian adult CNS is not well characterized. One previous study described the decline of both Fabp7 and Fabp3 protein abundance within brain synapstosomal fractions of aged mice (Pu et al., 1999). Because Fabp7 knock-out mice have been shown to influence NMDA-receptor-dependent physiology and memory behavior (Owada et al., 2006), we decided to examine Fabp7 synaptic localization in further detail. A significant enrichment of Fabp7 protein was observed in whole brain synaptoneurosomal fractions compared with β-actin (t test, p < 0.05, N = 3) as measured by Western blot (Fig. 1A), confirming its presence in synaptic preparations as previously reported (Pu et al., 1999). Since Fabp7 protein expression is confined to astrocytes and neural progenitors in adult mammalian brain, this suggests that it is enriched within the perisynaptic compartment of fine astrocytic processes that surround synapses, which together comprise the tripartite synapse (Halassa et al., 2007). We validated the presence of Fabp7 protein in the tripartite synapse using electron microscopy (EM) following immunocytochemistry on mouse brain tissue sections. Analysis of nanogold-labeled anti-Fabp7 immunoreactive (IR) protein with EM showed a clear localization of the IR product in the fine glial processes of hippocampal (Fig. 1B) and cerebellellar tripartite synapses (Fig. 1C). Fabp7 IR was not observed in neuronal compartments and nanogold labeling was nonspecific in control sections lacking primary Fabp7 antibody (Fig. 1D).
Fabp7 mRNA and protein enrichment cycles in tripartite synapses
Circadian cycling of core clock components and clock-controlled genes are well known in terms of transcriptional autoregulatory processing in the nuclear compartment of cells. In contrast, very little is known about time-of-day changes of cycling molecules in other cellular locations, including the synapses in brain. We therefore examined the localization of Fabp7 mRNA in synaptoneurosomal fractions over the light/dark cycle. We observed dramatic time-of-day changes in Fabp7 mRNA in synaptoneurosomal fractions compared with total homogenate, with higher levels in the synaptoneurosome during the dark phase compared with the light phase (Fig. 2A). We also observed diurnal changes in GAPDH and β-actin mRNA expression in synaptoneurosomal fractions compared with total homogenate. However, these differences did not compare with the ∼1000-fold time-of-day differences observed for Fabp7. We next examined whether time of day changes Fabp7 protein levels in synaptneurosomes. We observed >2-fold changes in the relative abundance of Fabp7 protein in synaptoneurosomal fractions compared with total homogenate (Fig. 2B). Since β-actin protein is not as enriched in the synaptoneurosomal fraction as Fabp7 (Fig. 1), and since β-actin mRNA abundance cycles in the opposite direction as Fabp7 mRNA (Fig. 2A), we compared relative Fabp7 protein levels in synaptoneurosomes to the postsynaptic density protein 95 (PSD-95), a member of the membrane-associated guanylate kinase family that aids in receptor and ion channel clustering at synapses. We observed >4-fold changes in the relative abundance of PSD-95 protein in synaptoneurosome versus total homogenate (Fig. 2B), with lower levels at lights on (ZT0). PSD-95 was more enriched in synaptic fractions at ZT6 and ZT12, and the ratio of its relative abundance in synaptoneurosomal versus total homogenate was similar to the ratio of Fabp7 at ZT0 and ZT18. Together, these data show that there are dynamic time-of-day-regulated changes in the relative levels of Fabp7 mRNA and protein in synaptic compartments.
Fabp7 mRNA is trafficked in adult hippocampus
Previously, we have shown that time-of-day oscillations in Fabp7 mRNA abundance occur broadly throughout murine brain, including various layers of adult murine hippocampus (Gerstner et al., 2008). Given that Fabp7 protein is enriched in perisynaptic processes of astrocytes, and that Fabp7 mRNA is increased in the hippocampus following intense neuronal stimulation (Owada et al., 2006), we more closely examined circadian changes of Fabp7 mRNA in hippocampus. We specifically determined whether there is a time-of-day change on the distribution pattern of mRNA in molecular layers of the hippocampus. A more focal distribution of Fabp7 mRNA near astroglial somata would suggest recently synthesized transcript from the nucleus, whereas a more scattered distribution would reflect mRNA trafficking away from the cell body into the cytoplasm of astrocyte processes. Previously we reported significant time-of-day changes in Fabp7 mRNA levels in various brain regions of mouse brain (Gerstner et al., 2006, 2008). Quantification of Fabp7 mRNA by in situ hybridization using radio-labeled antisense riboprobes on this tissue subjected to liquid emulsion in the dentate gyrus of hippocampus revealed a significant time-of-day change in mRNA levels with peak levels of expression observed at ZT0 and ZT6 (Gerstner et al., 2008), in phase with other brain regions (Gerstner et al., 2006, 2008). We next examined the density of silver grains over the light/dark cycle to determine whether diurnal changes exist in histological localization for Fabp7 mRNA. Dispersal of silver grains is greater at the light/dark transition (ZT12), compared with a less distributed pattern between ZT0 and ZT6 (Fig. 3A,B). Analysis of silver grain density shows statistically significant difference in mRNA distribution based on time of day (one-way ANOVA, p = 0.0156; Fig. 3B). The more somatic and punctate distribution observed at the ZT0–6 time period (Fig. 3A) corresponds with times of higher mRNA levels in the hippocampus (Gerstner et al., 2008), likely reflecting localization of recently transcribed Fabp7 mRNA in the cell bodies of astrocytes (Gerstner et al., 2008). Changes in the temporal distribution of Fabp7 mRNA were not specific to hippocampus. Similar patterns of Fabp7 mRNA localization were observed in other brain regions, such as the cortex, thalamus, amygdala, and hypothalamus (data not shown). Together, these data indicate a coordinated process in directed Fabp7 mRNA trafficking throughout brain.
CPEB machinery
Phylogenetic conservation of CPE sequences in the Fabp7 3′UTR
CPEs represent one mechanism that controls trafficking of mRNA in neurons. The CPEs interact with CPEB1 within neurons following increased activity to regulate experience-dependent local translation (Wu et al., 1998). CPEB1 has also been shown to operate on β-catenin mRNA in astrocytes to regulate translation and subsequent migration (Jones et al., 2008). Given the diurnal changes in tripartite synaptic enrichment and trafficking observed in Fabp7 mRNA, we first bioinformatically searched for cis-acting elements within the 3′UTR that are involved in signaling pathways known to influence trafficking and synaptic targeting. The consensus CPE sequence UUUUAU and likely similar elements were therefore bioinformatically searched in the Fabp7 mRNA 3′UTR of various species. Multiple phylogenetically conserved CPEs were observed in all species examined (Table 1). The relative distance and loci of the CPEs from the hexanuculeotide (AAUAAA) are also conserved in multiple species (Table 1), suggesting functional regulatory mechanisms underlying Fabp7 mRNA translational control conserved across species.
CPEB machinery exists in astrocytes and binds Fabp7 mRNA
Four CPEB proteins have been characterized in adult mammalian brain (Theis et al., 2003), and CPEB1 was previously shown to be expressed in astrocytes (Jones et al., 2008). To confirm the presence of CPEB1, and determine whether the other CPEBs exist in astrocytes, Northern blot analysis of mRNA from cultured murine primary astrocytes was performed. The presence of CPEB1 and CPEB4, but not CPEB2 or CPEB3 mRNA, was identified in astrocyte culture homogenates (Fig. 4A). All CPEBs were observed in total RNA from whole brain (Fig. 4A), confirming previous studies (Theis et al., 2003; Rouhana et al., 2005).
Given murine Fabp7 mRNA contains CPEs, we performed coimmunoprecipitation experiments on lysate from mouse primary astrocyte culture to assess whether CPEB1 binds to Fabp7 mRNA. This method selectively restricts interactions of CPEB protein with Fabp7 mRNA expressed specifically in astrocytes, eliminating the possibility for false-positive interactions from whole-brain lysate. There was a ∼12-fold enrichment of Fabp7 mRNA in the eluate of anti-CPEB1-conjugated protein-A beads compared with input (Fig. 4B). Previously the CPE-containing β2 microglobulin mRNA was identified in a microarray screen search for CPEB1-bound mRNAs from mammalian brain (Du and Richter, 2005). β2 microglobulin is also present in astrocytes (Diedrich et al., 1993; Vanguri, 1995), and therefore we used it as a positive control for a CPEB-interacting mRNA. QRT-PCR analysis of CPEB1 co-IP showed a ∼3-fold enrichment of β2 microglobulin mRNA compared with input, confirming previous findings of binding to CPEB1 (Du and Richter, 2005). A statistically significant elevation in Fabp7 mRNA compared with β2 microglobulin mRNA co-IP was observed (t test, p < 0.01), suggesting that Fabp7 mRNA binds to CPEB with greater affinity.
CPE function
CPE mediates cytoplasmic regulation of Fabp7 mRNA translation
Since Fabp7 mRNA binds CPEB1 in cultured astrocytes (Fig. 4B), we were interested in determining whether the CPEs are functional. To do so, we tested whether the Fabp7 mRNA 3′UTR would control translation during Xenopus oocyte maturation. Legitimate CPEs from neuronal mRNAs are commonly active in this assay, enhancing translation and promoting polyadenylation during oocyte maturation (Wu et al., 1998; Kim and Richter, 2008; Novoa et al., 2010; Igea and Mendez, 2010).
The murine Fabp7 3′UTR was subcloned downstream of the luciferase open-reading frame, and the resulting mRNA injected into Xenopus oocytes, half of which were then induced to mature using progesterone (Rouhana and Wickens, 2007). A β-galactosidase reporter without the Fabp7 3′UTR was coinjected for normalization. A 2.4-fold induction of luciferase activity was observed following progesterone stimulation, demonstrating that translational stimulation elements operate on the Fabp7 3′ UTR (Fig. 5). An mRNA carrying the cyclin B1 3′UTR served as a positive control, and exhibited translational repression in resting oocytes and translational activation after maturation, as expected (de Moor and Richter, 1999; Barkoff et al., 2000).
A time course analysis showed a temporal window of luciferase/Fabp7 mRNA 3′UTR translation peaking at 7 h following progesterone stimulation (Fig. 6). Mutations in either the CPEs or the cleavage and polyadenylation specificity hexanucleotide, both of which are required for cytoplasmic polyadenylation (Fox et al., 1989), blocked the translational stimulation mediated by the Fabp7 3′UTR (Fig. 6). Mutations in the CPEs release repression of translation, leading to an increase in luciferase activity in unstimulated oocytes compared with wild type. These observations are similar to other validated CPE-containing 3′UTRs using this assay, such as in the 3′UTR of CPEB4 mRNA (Igea and Mendez, 2010). We conclude that the Fabp7 3′UTR possesses functional CPEs.
Fabp7 mRNA poly(A) tail is diurnally regulated in mouse brain
CPE-containing mRNAs are known to regulate polyadenylation and subsequent translation (Wu et al., 1998; Wells et al., 2001; Du and Richter, 2005; Igea and Mendez, 2010). Since Fabp7 mRNA binds to CPEB1, is translationally regulated by existing CPEs in its 3′UTR, and trafficked in astrocytes in a time-of-day-dependent manner, a corresponding diurnal change in the poly(A) tail can be predicted. To determine whether Fabp7 mRNA has a corresponding change in polyadenylation, we examined the pattern of mRNA length over the light/dark cycle. We previously reported that the abundance of Fabp7 mRNA varies with time of day in multiple brain regions (Gerstner et al., 2006, 2008). Reanalysis of that data revealed that the lengths of the Fabp7 mRNA, and not merely its abundance, vary in each of these regions of the brain. This is a specific effect, since Fabp5 mRNA length (data not shown), another Fabp of similar size expressed in CNS glia (Veerkamp and Zimmerman, 2001), does not vary. To confirm that this change in Fabp7-banding pattern was a result of alterations in poly(A) tail length, we used RNase H analysis. The RNase H enzyme hydrolizes phosphodiester bonds of RNA hybridized to DNA, and is used for determining polyadenylation (Goodwin and Rottman, 1992). Using an oligo-dT DNA primer and hybridizing it to total brain poly(A) RNA, a consistent mobility of digested Fabp7 mRNA products were observed following digestion in the presence of the RNase H enzyme at both ZT12 and 24 time points in whole-brain homogenates examined by Northern blotting (data not shown). Closer examination of RNase H digested poly(A) mRNA products using oligo-dT at multiple time points confirmed a single predicted Fabp7 mRNA band at 762 bp (Fig. 7A). Synchronization of this banding pattern was observed in multiple regions throughout murine brain (Gerstner et al., 2006, 2008), including regions involved in circadian rhythms (suprachiasmatic nucleus) and sleep/wake regulation (ventrolateral preoptic nucleus, tuberomammalary nucleus, lateral hypothalamus, locus ceruleus, pons) (Saper et al., 2010). A comparison of the changes in mRNA level and polyadenylation shows an increased length in poly(A) tail when the Fabp7 mRNA is low, during the dark period, at ZT18 (Fig. 7B).
Discussion
The results presented here are the first to show that circadian changes in polyadenylation occur in a non-neuronal mRNA in mammalian brain, and that a diurnally controlled mRNA is post-transcriptionally regulated by CPEB-mediated cytoplasmic polyadenylation. Fabp7 mRNA polyadenylation is mediated, at least in part, through cis-regulatory elements in the Fabp7 mRNA's 3′UTR that are phylogenetically conserved. Further, CPEs in the Fabp7 3′UTR are necessary for repression of translation in the Xenopus oocyte maturation assay and Fabp7 transcript is bound by CPEB1 in mouse astrocytes. These data provide a novel mechanism for coordinated CPEB-mediated translational processing of mRNA in astrocytes over the circadian cycle in mammalian brain.
An aging-dependent decline in Fabp7 protein in synaptoneurosomal fractions of mouse brain was previously observed (Pu et al., 1999). We confirmed that the Fabp7 was present in the fine perisynaptic processes of astrocytes using electron microscopy, and observed a time-of-day-dependent alteration in Fabp7 mRNA localization in synaptoneurosomal fractions. Exactly whether an aging-dependent decrease in Fabp7 protein in synaptoneurosomal fractions is due to an overall decrease in Fabp7 levels, or some deficit in subcellular localization and/or polyadenylation mechanisms remains to be determined. However, since Fabp7 binds the ω-3 polyunsaturated fatty-acid, docosahexaenoic acid (DHA), to high affinity (Xu et al., 1996) and DHA has been implicated in Alzheimer's disease and cognitive function (Morley and Banks, 2010), it is possible that alterations in Fabp7 processing or targeting contributes to age-related cognitive decline. This is an intriguing possibility since amyloid-precursor proteins, which are believed to contribute to Alzheimer's disease etiology (Tanzi and Bertram, 2005; Haass and Selkoe, 2007), have been shown to act as membrane anchors for the cytoplasmic translational activity of CPEB (Cao et al., 2005). It is possible that misregulation of CPEB-mediated Fabp7 mRNA post-transcriptional events could contribute to an aging-dependent decline in cognitive processing and/or neurodegenerative disease related to fatty-acid signaling (Di Paolo and Kim, 2011).
CPEBs are mRNA-binding proteins involved in subcellular trafficking and local translation of targeted mRNAs at synapses (Huang et al., 2003). We confirmed that CPEB1 also exists in astrocytes as described previously (Jones et al., 2008), and characterized functional CPE-sites in the Fabp7 3′UTR. However, the physiological signals that direct Fabp7 mRNA trafficking and polyadenylation remain unknown. Since Fabp7 has been shown to regulate cellular outgrowth (Feng et al., 1994; Arai et al., 2005) and motility (Mita et al., 2010), CPE-mediated polyadenylation and targeting of Fabp7 may be involved in regulating astrocytic process extension in plasticity-related contexts. It is not yet known, however, whether changes in neuronal activity and/or synaptic plasticity affect trafficking and polyadenylation of Fabp7 mRNA, leading to morphological changes in astrocytic processes.
Our results show that Fabp7 mRNA undergoes longer poly(A) extension during a time window when the base levels of mRNA are lower. Interestingly, the shortest poly(A) tail was observed at ZT12, the time point of greatest mRNA dispersal, suggesting either dormant or degraded Fabp7 mRNA is directed away from cell bodies at the light-to-dark transition. Dramatic increases in levels of Fabp7 heteronuclear RNA transcripts were previously observed at the dark-to-light transition (Gerstner et al., 2008), suggesting a major transcriptional event and reflected by the concentrated somatic localization of silver grain density observed in astrocytes during the light period (Fig. 3A). This suggests that increases in Fabp7 poly(A) tail length at times of lower mRNA levels are likely not derived from nuclear transcriptional events per se, but instead are controlled post-transcriptionally through mechanisms involving CPEB-mediated translational activation (see Fig. 8). Protein kinase activity is thought to direct phosphorylation of CPEB and to regulate polyadenylation of target transcripts (Richter, 2007). One prediction is that similar kinase activation of CPEB exists in astrocytes, gated by time-of-day alterations in neuronal activity, metabolism, or behavioral state. This might be expected at light/dark transition times, when animals switch between dormant and active states, where sleep/wake neurophysiology could signal changes in Fabp7 mRNA perisynaptic localization, polyadenylation, and subsequent translation. Therefore, one prediction would be that activity state at specific times of day contributes to Fabp7 mRNA trafficking, polyadenylation, and protein accumulation in perisynaptic astroglial processes. Whether CPEB activation fits this model (Fig. 8) or state-dependent signaling pathways and kinases mediate this process remains to be determined.
The oscillation in trafficking observed within the hippocampus introduces an additional level of complexity for Fabp7 post-transcriptional regulation in astrocytes. Since CPEB is thought to stimulate trafficking and translation of plasticity-related mRNAs, such as αCaMKII, to recently activated synapses (Wu et al., 1998; Bramham and Wells, 2007; Richter, 2007), it is tempting to speculate that the diurnal trafficking observed for Fabp7 mRNA may be mediated by CPEB through astrocytic responses to neuronal activity. Fabp7 mRNA increases following neuronal activity within hippocampus following kainate stimulation (Owada et al., 1996). While mechanisms underlying post-transcriptional processing of clock and clock-controlled genes are beginning to be understood (Kojima et al., 2011), to the best of our knowledge, a role for CPEB in the regulation of post-transcriptional processing of diurnally regulated mRNA has not yet been examined. Our data presented here indicate that astrocytic Fabp7 transcription and subsequent trafficking to perisynaptic process compartments oscillate based on time of day through a putative CPEB-mediated mechanism (Fig. 8).
In addition to the presynaptic and postsynaptic neuronal terminals, an astroglial process, which surrounds the synapse, is an active participant in neural transmission and defines a unique compartment called the tripartite synapse (Halassa et al., 2007, 2009b). Astrocytes, through these perisynaptic processes, are believed to both sequester neurotransmitters and release gliotransmitters and factors to modulate synaptic activity (Halassa and Haydon, 2010). Previously, the glial fibrillary acidic protein (GFAP) mRNA was identified in SN fractions (Rao and Steward, 1993), molecular evidence for the presence of astroglial processes of the tripartite synapse. Glial cells also respond to neuronal activity. For example, an increase in GFAP expression has been observed in rat hippocampus following electrically induced seizures (Steward et al., 1991). A fivefold increase of GFAP mRNA was also observed at 24 h following seizure stimulation before returning to baseline levels 4 d later (Torre et al., 1993). Interestingly, Fabp7 mRNA has a similar hippocampal expression profile following kainate injection, where it is elevated 24 h later, reaching peak expression after 48 h (Owada et al., 1996), a time that corresponds to peak GFAP protein expression (Torre et al., 1993). Since Fabp7 mRNA is enriched in dendritic layers of hippocampus (Zhong et al., 2006), and present in synaptoneurosomes (Pu et al., 1999) (Figs. 1, 2), these data suggest that Fabp7-targeted perisynaptic translocation and subsequent translation within astrocytes may have some function related to changes in neuronal activation or synaptic homeostasis. We did not observe time-of-day-dependent changes in polyadenylation of the Fabp7 mRNA in primary astrocyte culture (data not shown), suggesting that the presence of nonastroglial factors are responsible for changes in Fabp7 post-transcriptional processing. Future work will determine whether neuronal activity is required for the observed effects on Fabp7 mRNA regulation.
Since we show a time-of-day change in Fabp7 mRNA localization to synapses and subcellular trafficking in hippocampus, another intriguing possibility is that Fabp7 function at synapses may be related to time-of-day-dependent changes observed in synaptic plasticity and memory formation (Gerstner et al., 2009; Gerstner and Yin, 2010). DHA is an ω-3 polyunsaturated fatty-acid known to bind Fabp7 to high affinity (Xu et al., 1996), is abundant in the synaptic compartment (Breckenridge et al., 1972; Salem et al., 2001), regulates synaptic plasticity (Young et al., 1998; Itokazu et al., 2000; Fujita et al., 2001; Mirnikjoo et al., 2001), and influences learning and memory behavior (Moriguchi et al., 2000; Moriguchi and Salem, 2003; Fedorova et al., 2007, 2009). DHA-stimulated NMDA-receptor-dependent current is blocked in neurons derived from Fabp7 KO mice (Owada et al., 2006), suggesting that Fabp7 localization in tripartite synapses influences neurophysiological function. Such Fabp7-dependent modulation of synaptic efficacy may in turn regulate memory processes. This is evident from observations of significant alterations in later periods of memory behavior in a passive avoidance-conditioning task in Fabp7 KO mice compared with wild-type mice (Owada et al., 2006). Recently we have shown that Fabp7 overexpression in Drosophila leads to changes in daytime sleep behavior and increases in memory consolidation (Gerstner et al., 2011a,b). These data implicate a functional role for subcellular localization of Fabp7 in mediating neural plasticity and time-of-day-dependent changes in complex behavior.
In conclusion, we provide the first evidence of circadian changes in polyadenylation for the astrocyte Fabp7 mRNA in mammalian brain. We characterize poly(A) tail processing and translational regulation of Fabp7 mRNA through CPE-sites in its 3′UTR, and describe an interaction with CPEB1 in astrocytes. We show the time-of-day-dependent trafficking of Fabp7 mRNA in the mouse hippocampus. We also provide biochemical and ultrastructural evidence of Fabp7 protein in perisynaptic processes of the tripartite synapse, where mRNA abundance changes logarithmically based on time of day. These changes are novel for known circadian genes and implicate Fabp7 post-transcriptional processing in the temporal regulation of the astrocytic component of the tripartite synapse.
Footnotes
This work was supported by National Institutes of Heath Grant DA13780 to C.F.L. L.R. is supported by National Science Foundation Minority Postdoctoral Research Fellowship 0804021. C.J.W. is supported by the FRAXA Research Foundation. J.R.G. is currently supported by National Institutes of Heath Grant T32 HL07713.
We thank Josh Smith and Sandra Splinter BonDurant and the University of Wisconsin-Madison Genome Center for expert qRT-PCR services. We thank Ben August at the University of Wisconsin Medical School Electron Microscope Facility for expert EM services. We thank Laura Roethe, Quentin Bremer, and Andrew Brienen for excellent technical assistance.
- Correspondence should be addressed to Jason R. Gerstner at the above address. gerstner{at}upenn.edu