Fragile X syndrome (FXS) is caused by loss of the FMR1 gene product FMRP (fragile X mental retardation protein), a repressor of mRNA translation. According to the metabotropic glutamate receptor (mGluR) theory of FXS, excessive protein synthesis downstream of mGluR5 activation causes the synaptic pathophysiology that underlies multiple aspects of FXS. Here, we use an in vitro assay of protein synthesis in the hippocampus of male Fmr1 knock-out (KO) mice to explore the molecular mechanisms involved in this core biochemical phenotype under conditions where aberrant synaptic physiology has been observed. We find that elevated basal protein synthesis in Fmr1 KO mice is selectively reduced to wild-type levels by acute inhibition of mGluR5 or ERK1/2, but not by inhibition of mTOR (mammalian target of rapamycin). The mGluR5-ERK1/2 pathway is not constitutively overactive in the Fmr1 KO, however, suggesting that mRNA translation is hypersensitive to basal ERK1/2 activation in the absence of FMRP. We find that hypersensitivity to ERK1/2 pathway activation also contributes to audiogenic seizure susceptibility in the Fmr1 KO. These results suggest that the ERK1/2 pathway, and other neurotransmitter systems that stimulate protein synthesis via ERK1/2, represent additional therapeutic targets for FXS.
Fragile X syndrome (FXS) is caused by the loss of the FMR1 gene product FMRP (fragile X mental retardation protein) (Verkerk et al., 1991). Converging lines of evidence suggest that FMRP represses mRNA translation in neurons and that cerebral protein synthesis is elevated in the absence of FMRP (Laggerbauer et al., 2001; Li et al., 2001; Huber et al., 2002; Aschrafi et al., 2005; Qin et al., 2005; Dölen et al., 2007; Bolduc et al., 2008). Group 1 (Gp1) metabotropic glutamate receptors (mGluR1 and mGluR5) stimulate mRNA translation at synapses (Weiler and Greenough, 1993; Weiler et al., 1997) and many lasting physiological consequences of Gp1 mGluR activation require rapid synaptic protein synthesis (Merlin et al., 1998; Huber et al., 2000; Raymond et al., 2000; Karachot et al., 2001; Vanderklish and Edelman, 2002; Banko et al., 2006). Based initially on the finding that mGluR-dependent long-term synaptic depression (mGluR-LTD) is exaggerated in the hippocampus of Fmr1 knock-out (KO) mice (Huber et al., 2002), the proposal was made that many of the symptoms of FXS might plausibly be explained by excessive protein synthesis downstream of Gp1 mGluR activation (Bear et al., 2004). The prediction that multiple aspects of fragile X can be corrected by reducing or inhibiting mGluR5 has been confirmed in numerous studies in several species (for review, see Dölen and Bear, 2008).
Although it is now clear that mGluR5 participates in the pathogenesis of FXS, at least in animal models, it is still poorly understood how Gp1 mGluRs trigger protein synthesis and how this process is altered in the absence of FMRP to disrupt synaptic function. Several studies have examined this issue in the hippocampus and cortex, but no clear consensus has emerged (Weiler et al., 2004; Hou et al., 2006; Muddashetty et al., 2007; Kim et al., 2008; Park et al., 2008; Ronesi and Huber, 2008; Sharma et al., 2010). One source of confusion may be that proxy measures of protein synthesis, such as mGluR-LTD or phosphorylation of signaling molecules, have been used in intact hippocampal slice preparations, whereas metabolic labeling experiments have been performed in synaptoneurosome preparations of cortex that are not easily related to altered hippocampal synaptic plasticity.
In the current study, we reexamine the question of how protein synthesis is elevated in the Fmr1 KO using a metabolic labeling approach in hippocampal slices maintained under the same conditions that revealed altered mGluR-dependent synaptic plasticity in previous studies from our laboratory (Huber et al., 2002; Auerbach and Bear, 2010). A strong rationale for taking this approach is that slice has been shown to accurately reproduce the in vivo phenotype of elevated basal protein synthesis in the Fmr1 KO hippocampus (cf. Qin et al., 2005; Dölen et al., 2007). Further, in addition to reproducing this core biochemical phenotype, the slice has the advantage that it enables pharmacological and biochemical access that is not possible in vivo. Our data suggest that elevated protein synthesis in the Fmr1 KO is due to saturation of mRNA translation downstream of the MAP kinase ERK1/2 which is basally activated by mGluR5.
Materials and Methods
Fmr1 KO (Jackson Labs) and wild type littermates were kept on the C57BL/6J background, group housed, and maintained in a 12 h light/dark cycle. All animals were treated in accordance with NIH and MIT guidelines. All experiments were performed blind to genotype. On each day of slice experimentation, 4 animals from each genotype were killed in an interleaved fashion and slices were prepared as rapidly as possible (≤5 min) as described below. This procedure yielded yoked, same-day controls for genotype and drug treatments.
(R,S)-3,5-Dihydroxyphenylglycine (DHPG), 2-methyl-6-(phenylethynyl)pyridine hydrochloride (MPEP), 1,4-diamino-2,3-dicyano-1,4-bis[2-aminophenylthio]butadiene (U0126), actinomycin D (ActD), and α-[amino[(4-aminophenyl)thio]methylene]-2-(trifluoromethyl) benzeneacetonitrile (SL 327) were obtained from Tocris Bioscience. DHPG and MPEP stocks were freshly prepared in ddH2O on the day of the experiment. ActD stock was prepared in ACSF + 0.5% DMSO and kept at −20°C. Anti-TrkB (R&D Systems), BDNF (Peprotech), insulin (Sigma), rapamycin (EMD Biosciences), cycloheximide (EMD Biosciences), U0126, and SL 327 were reconstituted according to manufacturer's instructions and either used immediately or stored at −20°C. For all slice experiments, the final concentration of DMSO was <0.1%.
Juvenile [postnatal day 25 (P25)–P30] male littermate wild-type (WT) and Fmr1 KO mice were given an overdose of Nembutal, and the hippocampus was rapidly dissected into ice-cold artificial CSF (ACSF) (in mm: 124 NaCl, 3 KCl, 1.25 NaH2PO4, 26 NaHCO3, 10 dextrose, 1 MgCl2, 2 CaCl2, saturated with 95% O2 and 5% CO2). Slices (500 μm thick) were prepared using a Stoelting Tissue Slicer and transferred into 32.5°C ACSF (saturated with 95% O2 and 5% CO2) within 5 min. Unless indicated otherwise, slices were incubated in ACSF undisturbed for 3.5–4 h to allow for recovery of protein synthesis (Sajikumar et al., 2005). ActD (25 μm) was then added to the recovery chamber for 30 min to inhibit transcription. For DHPG (100 μm) and whole-slice MPEP (50 μm) experiments, slices were incubated in 10 μCi/ml 35S-Met/Cys (express protein labeling mix, PerkinElmer) ± drug for 5 min, and transferred to fresh ACSF with 10 μCi/ml 35S-Met/Cys for another 25 min to measure protein synthesis. For cycloheximide (60 μm; performed in WT), CA1 MPEP (10 μm), U0126 (5 μm), and rapamycin (20 nm) experiments, slices were incubated ± drug during ActD exposure (30 min), and transferred to fresh ACSF with 10 μCi/ml 35S-Met/Cys ± drug for another 30 min. For TrkB activation experiments, slices were incubated ± 1 μg/ml anti-TrkB for 30 min, then 25 μm ActD ± 1 μg/ml anti-TrkB for 30 min, and protein synthesis measured with 10 μCi/ml 35S-Met/Cys ± 1 μg/ml anti-TrkB for 1 h. After labeling, slices were either snap frozen in liquid nitrogen or processed immediately.
With the exception of CA1 MPEP experiments, slices were homogenized in ice-cold homogenization buffer (10 mm HEPES pH 7.4, 2 mm EDTA, 2 mm EGTA, 1% Triton X-100, protease inhibitors (cocktail III, EMD Biosciences), and phosphatase inhibitors (cocktails I + II, EMD Biosciences), and incubated in trichloroacetic acid (TCA; 10% final) for 10 min on ice to precipitate radiolabeled proteins. Samples were then spun at 21,000 × g for 10 min, and the pellet was washed with ice-cold ddH2O and resuspended in 1 N NaOH until dissolved. After adjustment to a neutral pH with HCl, triplicate aliquots of each sample were added to scintillation cocktail (HiSafe II, PerkinElmer) and read with a scintillation counter, and also subjected to a protein concentration assay (Bio-Rad). Averaged triplicate cpm values were divided by protein concentrations, resulting in cpm per μg protein. To control for daily variation in incorporation rate, the values obtained on each day were normalized to the 35S-Met/Cys ACSF used for incubation, and the average incorporation of all slices analyzed in that experiment, as described by Lipton and Raley-Susman (1999).
For CA1-MPEP experiments, slices were briefly thawed and CA1 regions were dissected. To obtain autoradiographs, aliquots of homogenized CA1 were taken before TCA precipitation and boiled in Laemmli sample buffer. Samples were then resolved on SDS polyacrylamide gels, transferred to nitrocellulose, and stained for total protein (Memcode staining kit, Pierce). Blots were exposed to a PhosphorImager screen for 24–72 h, and the screen was read with a PhosphorImager (Fujifilm). To quantify autoradiographs and Memcode-stained blots, a line-scan of each lane was measured and quantified using the gel analyzer tool in ImageJ. Protein synthesis was calculated by normalizing data from autoradiographs to Memcode staining data obtained from the same lanes.
Slices were prepared and allowed to recover as for metabolic labeling, incubated in 100 μm DHPG for exactly 5 min or 1 μm insulin for 10 min, then snap frozen in liquid nitrogen. Frozen slices were either immediately homogenized in Laemmli sample buffer containing phosphatase inhibitors, or briefly thawed and microdissected in homogenization buffer with protease and phosphatase inhibitors (minus detergent). Microdissected regions were homogenized in sample buffer containing phosphatase inhibitors immediately following dissection.
Synaptoneurosomes were isolated from sets of 4 slices essentially as described previously (Chen and Bear, 2007). Slices were homogenized on ice using 2 ml glass Dounce homogenizers (Wheaton), passed through 2 × 105 μm meshes, followed by 1 × 5 μm mesh, and the resulting samples spun at 1,000 × g for 10 min at 4°C. Pellets were washed, spun at 1000 × g, and processed for SDS-PAGE.
Samples were boiled in Laemmli sample buffer, resolved on SDS polyacrylamide gels, transferred to nitrocellulose, and stained for total protein. Immunoblotting was performed with the following primary antibodies: from Cell Signaling Technology, p-ERK1/2 (Thr202/Tyr204), ERK1/2, p-p38 (Thr180/Tyr182), p38, p-Akt (Ser473), p-mTOR (mammalian target of rapamycin) (Ser2448), mTOR, p-PTEN (Ser380/Thr382/383), PTEN, p-p70S6K (Thr389), p70S6K, p-S6 (Ser235/236), S6 p-Trk (Tyr490), and GAPDH; other: TrkB (BD Biosciences) and α-Ca2+/calmodulin-dependent kinase kinase II (α-CaMKII) (Sigma). After incubation in primary antibody overnight at 4°C, immunoblots were either incubated with fluorophore-conjugated secondary antibodies and imaged with the Odyssey imaging system (LiCor Biosciences), or incubated with HRP-conjugated secondary antibodies (GE Healthcare), developed with ECL plus (GE Healthcare) and exposed to film. Densitometry was performed on scanned blot films or LiCor images using Quantity One software (Bio-Rad). Data were expressed as the value of the phosphorylation signal divided by the value of the total protein signal in the same lane. To correct for blot-to-blot variance, each signal was normalized to the average signal of all lanes on the same blot. All gels were loaded and analyzed blind to genotype and treatment.
WT and Fmr1 KO (P25–P32) male littermates were killed by rapid decapitation, and hippocampi were rapidly dissected into ice-cold homogenization buffer. Tissue was homogenized on ice using 2 ml glass Dounce homogenizers, and processed for SDS-PAGE.
Hippocampal slices (5–8 per animal) were prepared as described above from WT and Fmr1 KO mice and metabolically labeled with 50 μCi/ml 35S-Met/Cys for 1 h to ensure visualization of individual target proteins. Immunoprecipitation (IP) experiments were performed on yoked WT and Fmr1 KO slices essentially as described by Osterweil et al. (2005) and Kundel et al. (2009). Briefly, slices were homogenized in ice-cold homogenization buffer with 200 mm NaCl, spun at 2000 × g for 5 min, and the supernatant was adjusted to 400 mm NaCl. Samples were then spun at 16,000 × g for 30 min, precleared with 1/10 volume protein A-Sepharose (GE Healthcare) for 1 h at 4°C, and incubated with 10 μg/ml nonimmune mouse IgG (Santa Cruz Biotechnology), mouse anti-GAPDH (Millipore), or mouse anti-α-CaMKII (Millipore) overnight at 4°C. Samples were then incubated with 1/10 volume protein A-Sepharose for 2 h at 4°C, and the IPs were washed 5 times in 1 ml of homogenization buffer with 400 mm NaCl. IPs were resuspended in an equal volume of Laemmli sample buffer, resolved on SDS polyacrylamide gels, transferred to nitrocellulose, and exposed to a PhosphorImager screen for 2–3 weeks. The same membranes were then immunoblotted for α-CaMKII and GAPDH. For each sample, the ratio of 35S-incorporated/total was calculated by dividing the density of the band seen by autoradiography to the density of the band seen by immunoblot (in the same lane). Experiments were performed and analyzed blind to genotype.
Hippocampal neurons were prepared from embryonic day 18 rat embryos as described previously (Krueger and Nairn, 2007), and cultured for 21 d. Thirty minutes before stimulation, the medium was removed and replaced with 0.75 ml of medium (50% conditioned, 50% fresh) to control for volume. TrkB stimulation was performed with a 15 min incubation of vehicle or 1 μg/ml anti-TrkB, followed by application of vehicle or 100 ng/ml BDNF for 5 min. Plates were then washed with ice-cold PBS, and cells were lysed in buffer containing 50 mm Tris, pH 7.4, 1 mm EDTA, 1 mm EGTA, 1% SDS, protease inhibitors, and phosphatase inhibitors. Whole-cell lysates were processed for SDS-PAGE and immunoblotted.
Experiments were performed essentially as described previously (Dölen et al., 2007). Male WT and Fmr1 KO littermates (P18–P22) were injected intraperitoneally with SL 327 (100 mg/kg; dose based on Zhong et al., 2009), rapamycin (6 mg/kg; dose based on Ehninger et al., 2008 and Meikle et al., 2008) or an equal volume of vehicle (50% DMSO in ddH2O for SL 327 experiments; 100% DMSO for rapamycin experiments), and returned to their home cage for 1 h. Each testing session contained at least one set of vehicle-treated controls from each genotype. Mice were then transferred to a transparent plastic test chamber and, after at least 1 min of habituation, exposed to an alarm (modified personal alarm, 125 dB RadioShack model 49–1010 or 130 dB Samfe model SWPDAL-130, powered from a DC converter) for 2 min. For each group, incidence of the following stages of audiogenic seizures (AGS) was calculated: wild running, clonic seizure, tonic seizure, and death. All mice were injected and scored blind to genotype.
For AGS experiments, significance was determined using two-tailed Fisher's exact test (appropriate for analyzing nominal datasets). For recovery time course experiments, significance was determined using repeated-measures ANOVA, followed by post hoc two-tailed unpaired Student's t tests. For all other experiments, outliers (±2 SDs from the mean) were removed, and significance between more than two groups was determined by two-way repeated measures mixed model ANOVA. If significant effects were found by ANOVA, post hoc analyses were performed to compare individual groups using two-tailed paired Student's t tests. For datasets that contained only two groups, significance was determined by two-tailed paired Student's t tests.
Basal protein synthesis is elevated in the Fmr1 KO hippocampus
The molecular mechanisms underlying dysfunctional protein synthesis in the Fmr1 KO are largely unknown. To explore this question, we used an in vitro assay designed to measure protein synthesis in acute hippocampal slices. To directly relate our observations to the aberrant physiology seen in the Fmr1 KO, we examined slices isolated from P25–P30 dorsal hippocampus, as this is when and where exaggerated mGluR- and protein synthesis-dependent LTD is observed (Huber et al., 2002). Slices were prepared from dorsal hippocampus, and immediately transferred to 32.5°C ACSF (see Materials and Methods). These slices were then exposed to ActD (25 μm) for 30 min to prevent new transcription, and protein synthesis was measured over 30 min via incorporation of a 35S-labeled methionine/cysteine mix (10 μCi/ml 35S-Met/Cys). To verify that our measurements accurately reflect global protein synthesis, slices were incubated ± 60 μm cycloheximide for 30 min, and protein synthesis measured ± 60 μm cycloheximide for an additional 30 min. This experiment confirmed that exposure to cycloheximide, a potent inhibitor of mRNA translation, completely eliminates 35S-Met/Cys incorporation (supplemental Fig. S1, available at www.jneurosci.org as supplemental material).
A number of previous studies suggest that a long (>2–4 h) postslicing recovery period is necessary to achieve stability in metabolic function (i.e., ATP and creatine levels) (Whittingham et al., 1984), signaling cascades (Ho et al., 2004), dendritic spine density (Kirov et al., 1999), and protein synthesis-dependent synaptic plasticity (Huber et al., 2001; Sajikumar et al., 2005). We confirmed and extended this conclusion using our assay of basal protein synthesis in hippocampal slices. We found that optimal and stable protein synthesis recovers 4–6 h after preparing slices (protein synthesis expressed as percentage ± SEM of 4 h value: 0.5 h, 76 ± 7%; 1 h, 72 ± 8%; 2 h, 88 ± 9%; 4 h, 100 ± 8%; 6 h, 99 ± 8%; ANOVA time p < 0.05; n = 10; Fig. 1A). This observation is consistent with electrophysiological measurements of optimal protein synthesis-dependent synaptic plasticity in the hippocampus (Sajikumar et al., 2005), including optimal mGluR-LTD (Huber et al., 2001). Therefore, for all experiments reported here, the hippocampus was allowed to recover for at least 4 h following slicing.
To examine whether protein synthesis is elevated in Fmr1 KO hippocampus under conditions where exaggerated mGluR-LTD is observed, we performed metabolic labeling on juvenile, dorsal Fmr1 KO slices. Our results reveal a significant elevation of basal protein synthesis in Fmr1 KO hippocampus compared with wild type (WT) controls (WT 100 ± 3%, KO 119 ± 5%; t test p < 0.02; n = 13; Fig. 1B). The magnitude of this increase in protein synthesis is in quantitative agreement with in vivo measurements (Qin et al., 2005), and with our previous results in adult, ventral hippocampus (Dölen et al., 2007).
Autoradiographs of Fmr1 KO versus WT slices show an increased 35S-Met/Cys incorporation into proteins of multiple molecular weights, which supports the proposal that FMRP functions as a rather general repressor of translation (Fig. 1B) (Laggerbauer et al., 2001; Li et al., 2001; Mazroui et al., 2002; Aschrafi et al., 2005). To confirm that our measurement of increased total protein synthesis reflects derepression of FMRP-regulated translation, we performed IP experiments to isolate newly made α-CaMKII, a validated target mRNA for FMRP (Brown et al., 2001; Zalfa et al., 2003; Ferrari et al., 2007; Muddashetty et al., 2007). α-CaMKII was immunoprecipitated from WT and Fmr1 KO slices that had been metabolically labeled with 50 μCi/ml 35S-Met/Cys for 1 h (see Materials and Methods). To ensure the efficacy of our IP experiments, we quantified the amount of α-CaMKII (assessed by immunoblot) observed in anti-α-CaMKII IPs versus IgG IPs from the same lysates (supplemental Fig. S2A, available at www.jneurosci.org as supplemental material). Our results confirm that α-CaMKII is significantly enriched in the anti-α-CaMKII IPs from both WT and Fmr1 KO slices (WT IgG 100 ± 12%, WT CaMK 600 ± 73%, KO IgG 70 ± 25%, KO CaMK 544 ± 45%; t test WT IgG vs CaMK *p < 0.002, t test KO IgG vs CaMK *p < 0.0001; n = 6) (supplemental Figure S2A, available at www.jneurosci.org as supplemental material). To quantify newly translated protein, 35S-incorporated α-CaMKII was measured by autoradiography, and normalized to total α-CaMKII in the same IP (Fig. 1C). A comparison of these values in WT versus Fmr1 KO reveals that α-CaMKII is excessively translated in hippocampal slices from the Fmr1 KO (WT 100 ± 12%, KO 124 ± 11%; t test *p < 04; n = 6; Fig. 1D). This increase in newly translated α-CaMKII is similar in magnitude to the increase seen in total protein synthesis (Fig. 1B).
We also performed the same IP experiments for glyceraldehyde phosphate dehydrogenase (GAPDH), a non-FMRP target (Brown et al., 2001). Analysis of GAPDH levels confirmed a significant enrichment in anti-GAPDH IPs versus IgG IPs in both WT and Fmr1 KO slices (WT IgG 100 ± 19%, WT GAPDH 1832 ± 268%, KO IgG 99 ± 26%, KO GAPDH 2250 ± 314%; t test WT IgG vs GAPDH *p < 0.003, t test KO IgG vs GAPDH *p < 0.002; n = 5) (supplemental Fig. S2B, available at www.jneurosci.org as supplemental material). However, a comparison of 35S-incorporated/total ratios reveals that GAPDH is not excessively translated in Fmr1 KO slices (WT 100 ± 9%, KO 88 ± 6%; t test p = 0.31; n = 5; Fig. 1D). These findings suggest that the elevated total protein synthesis we detect in the Fmr1 KO reflects derepression of FMRP-regulated translation.
Excessive protein synthesis in the Fmr1 KO is corrected by acute, pharmacological antagonism of mGluR5
To see whether acute pharmacological inhibition of mGluR5 could effectively reduce the excess protein synthesis seen in juvenile Fmr1 KO hippocampus, we exposed slices to the mGluR5 antagonist, MPEP. Interestingly, our initial experiments revealed that a 5 min application of 50 μm MPEP, which had no significant effect on WT protein synthesis, was sufficient to reduce the protein synthesis in Fmr1 KO slices back to WT levels (WT control 100 ± 1%, KO control 117 ± 2%, WT MPEP 104 ± 2%, KO MPEP 95 ± 2%; ANOVA genotype × treatment p < 0.05; n = 8; Fig. 1E). Because we are particularly interested in the protein synthesis related to mGluR-LTD, we repeated the experiment using an MPEP treatment that has been shown to block LTD (Hou and Klann, 2004) and measured changes in protein synthesis specifically in the CA1 region of dorsal hippocampus. Slices were preincubated ± 10 μm MPEP for 30 min, then protein synthesis was measured for another 30 min ± 10 μm MPEP. Measurements restricted to microdissected area CA1 showed that although MPEP had no effect on protein synthesis under basal conditions in WT slices, it was sufficient to correct elevated protein synthesis in Fmr1 KO slices (WT control 100 ± 4%, KO control 117 ± 6%, WT MPEP 107 ± 7%, KO MPEP 104 ± 6%; ANOVA genotype × treatment p < 0.02; n = 8; Fig. 1F). These data represent the first demonstration that excessive protein synthesis in the Fmr1 KO hippocampus can be rapidly corrected by acute pharmacological inhibition of mGluR5. Thus, the elevated protein synthesis in the Fmr1 KO is downstream of constitutive mGluR5 activation.
mGluR-mediated protein synthesis is mimicked and occluded in the Fmr1 KO
In hippocampal slices from Fmr1 KO mice, LTD induced by application of DHPG is greater in magnitude (Huber et al., 2002) and qualitatively unlike WT because it is not reduced by protein synthesis inhibitors (Hou et al., 2006; Nosyreva and Huber, 2006). These findings have led to the suggestion that proteins responsible for long-term expression of LTD are rapidly synthesized in response to mGluR activation in WT, and constitutively overexpressed in Fmr1 KO hippocampus. Gp1 mGluR activation has been shown to lead to increased protein synthesis in a variety of systems (Weiler and Greenough, 1993; Raymond et al., 2000; Job and Eberwine, 2001; Todd et al., 2003; Shin et al., 2004; Muddashetty et al., 2007). However, whether activation of mGluRs under LTD conditions leads to an increase in protein synthesis in hippocampal slice has not been directly tested.
We observed that the same treatment that induces LTD in hippocampal slices (100 μm DHPG, 5 min) also significantly increases protein synthesis in WT slices (Fig. 1G). Interestingly, however, the same treatment does not stimulate protein synthesis over the elevated baseline level in Fmr1 KO hippocampal slices (WT control 100 ± 2%, WT DHPG 119 ± 2%, KO control 110 ± 5%, KO DHPG 112 ± 4%; ANOVA genotype × treatment p < 0.02; n = 8; Fig. 1G). Thus, genetic deletion of FMRP appears to mimic and occlude the effect of DHPG on protein synthesis.
No detectable difference in basal mGluR signaling under conditions of elevated protein synthesis
Previous studies have shown that DHPG stimulation leads to activation of the MAPK/ERK1/2 pathway (Ferraguti et al., 1999; Gallagher et al., 2004; Hou and Klann, 2004). More recently, activation of the PI3K-Akt-mTOR pathway has been reported (Hou and Klann, 2004; Sharma et al., 2010). Both pathways are linked to the initiation of 5′ cap-dependent translation of mRNAs (Fig. 2A). ERK1/2 activates the MAPK-interacting kinase (Mnk), which leads to phosphorylation of eukaryotic initiation factor 4E (eIF4E) (Proud, 2007). Akt facilitates the activation of mTOR, which derepresses eIF4E by phosphorylating eIF4E binding proteins (4EBPs) (Gingras et al., 1999). mTOR also initiates translation of 5′ TOP mRNA, which is linked to the activation of ribosomal protein S6 kinases (i.e., p70S6K), and the subsequent phosphorylation of S6 (Proud, 2007). Phosphorylation of both 4EBP and p70S6K are considered to be reliable and equivalent readouts of mTOR activation (Hara et al., 1998; Avruch et al., 2001; Wang et al., 2005).
Our results showing that the excessive protein synthesis in the Fmr1 KO hippocampus is reversible with MPEP application suggest one of two things: (1) mGluR5 signaling is excessive in the Fmr1 KO hippocampus, or (2) the Fmr1 KO hippocampus is hypersensitive to normal constitutive mGluR5 signaling. In an attempt to differentiate between these two options, we measured ERK1/2, p38 MAPK, and Akt activation in the same hippocampal slices from WT and Fmr1 KO that had been used to measure protein synthesis. Interestingly, we observed no basal increase in either ERK1/2 or Akt activation in Fmr1 KO slices (ERK1/2: WT 100 ± 6%, KO 92 ± 7%; t test p = 0.22; Akt: WT 100 ± 6%, KO 92 ± 4%; t test p = 0.12; n = 27; Fig. 2B), despite detecting the robust increase in protein synthesis (Fig. 1B). There was, however, a small but significant decrease in phosphorylated p38 in the Fmr1 KO (t test *p < 0.05; n = 10; Table 1). This difference may reflect a compensatory downregulation in response to elevated protein synthesis, but is unlikely to be a cause of the increased protein synthesis.
To determine whether upstream or downstream effectors of these pathways were hyperactive, we examined the phosphorylation states of PTEN, a negative regulator of the PI3K pathway (Stambolic et al., 1998), as well as of the downstream components mTOR, p70S6K, and S6. In all proteins tested, no significant increase in activation state was observed in Fmr1 KO hippocampus (Table 1).
Evoked mGluR-ERK1/2 signaling appears normal in the Fmr1 KO
Results from untreated slices suggested that basal mGluR5 signaling is not excessive in the Fmr1 KO hippocampus. However, we could not exclude the possibility that we were missing a change that can only be seen with mGluR5 activation or in a specific biochemical fraction. We therefore measured activation of these pathways after 5 min of stimulation with 100 μm DHPG, the protocol for mGluR-LTD induction. Our results revealed that there were no significant differences in activation of ERK1/2, p38, PTEN, Akt, mTOR, p70S6K, or S6 between WT and Fmr1 KO (Table 2). Interestingly, while we observed a robust activation of ERK1/2 in both WT and Fmr1 KO slices (WT control 100 ± 6%, WT DHPG 132 ± 8%, KO control 90 ± 6%, KO DHPG 135 ± 6%; ANOVA treatment p < 0.0001; n = 13; Fig. 2C) we did not observe a significant activation of Akt in these slices (WT control 100 ± 5%, WT DHPG 104 ± 6%, KO control 100 ± 5%, KO DHPG 100 ± 6%; ANOVA treatment p = 0.57; n = 14; Fig. 2C). Further investigation revealed that none of the PI3K pathway members examined (PTEN, mTOR, p70S6K and S6) were activated by DHPG in either WT or Fmr1 KO slices (Table 2).
The exaggerated mGluR-LTD phenotype in the Fmr1 KO has been described in CA1 (Huber et al., 2002), and we wanted to confirm our results in this area of the hippocampus. Consistent with results from whole slices, we found no hyperactivation of ERK1/2 or Akt in area CA1 microdissected from Fmr1 KO hippocampus, nor was there any occlusion of an mGluR-stimulated response (ERK1/2: WT control 100 ± 3%, WT DHPG 164 ± 6%, KO control 79 ± 7%, KO DHPG 145 ± 16%; ANOVA treatment p < 0.0001, genotype × treatment p = 0.95; n = 8; Akt: WT control 100 ± 7%, WT DHPG 106 ± 6%, KO control 95 ± 6%, KO DHPG 99 ± 9%; ANOVA treatment p = 0.55, genotype × treatment p = 0.89; n = 8; Fig. 2D). Interestingly, a slight but significant decrease in ERK1/2 phosphorylation was seen in Fmr1 KO CA1 suggesting the possibility of compensatory downregulation in response to elevated protein synthesis (t test *p < 0.008; n = 8; Fig. 2D). These results support the conclusion that ERK1/2 activation is neither elevated nor saturated in Fmr1 KO hippocampus. Consistent with our previous results, no significant mGluR-mediated activation of Akt was observed in either WT or Fmr1 KO CA1 (Fig. 2D).
To test whether differences in mGluR-mediated signaling could be seen specifically at the synaptic level, we measured activation of ERK1/2 and Akt in synaptoneurosome fractions isolated from slices stimulated ± 100 μm DHPG for 5 min. Consistent with results from whole slices, we find that mGluR-mediated activation of ERK1/2 is preserved in Fmr1 KO synaptoneurosomes (WT control 100 ± 8%, WT DHPG 135 ± 8%, KO control 98 ± 10%, KO DHPG 139 ± 9%; ANOVA treatment p < 0.0001, genotype × treatment p = 0.68; n = 10 sets of slices from 7 animals; Fig. 2E). No activation of Akt was observed in either WT or Fmr1 KO synaptoneurosomes (WT control 100 ± 9%, WT DHPG 102 ± 10%, KO control 98 ± 13%, KO DHPG 93 ± 13%; ANOVA treatment p = 0.79, genotype × treatment p = 0.60; n = 10 sets of slices from 7 animals; Fig. 2E).
Our data suggest that the PI3K-Akt pathway is not basally hyperactive under conditions in which we observe excessive protein synthesis. To ensure that our assay was sensitive to changes in Akt activation, we exposed slices to insulin, a potent activator of the PI3K pathway (Proud, 2007). Results from these experiments show that a 10 min application of 1 μm insulin leads to a robust activation of Akt in both Fmr1 KO and WT slices (WT control 100 ± 6%, WT insulin 191 ± 16%, KO control 113 ± 21%, KO insulin 205 ± 20%; ANOVA treatment p < 0.0005, genotype × treatment p = 0.979; n = 8; Fig. 2F). These results suggest that the PI3K-Akt pathway is preserved, and not saturated, in Fmr1 KO slices.
Together, the results from these experiments show that differences in basal or evoked activation of mGluR5-mediated signaling do not parallel—and are therefore unlikely to account for—the differences in basal and DHPG-evoked protein synthesis in Fmr1 KO hippocampus.
Inhibition of ERK1/2, but not mTOR, corrects excessive protein synthesis in the Fmr1 KO
Our results show that the excessive protein synthesis in the Fmr1 KO can be corrected by inhibition of mGluR5, and that activation of Gp1 mGluRs leads to robust activation of ERK1/2. We therefore hypothesized that inhibition of ERK1/2 could, like MPEP, correct the excessive protein synthesis in the Fmr1 KO. This hypothesis was tested using the MEK1/2-ERK1/2 inhibitor U0126, which we found robustly decreased ERK1/2 activation in WT and Fmr1 KO at 5 μm (WT control 100 ± 15%, WT U0126 15 ± 3%; KO control 86 ± 8%, KO U0126 12 ± 2%; ANOVA treatment p < 0.0001; n = 4; Fig. 3B). Slices were preincubated ± 5 μm U0126 for 30 min, and protein synthesis measured ± 5 μm U0126 for an additional 30 min. Our results reveal that U0126 corrects protein synthesis in the Fmr1 KO down to WT levels (WT control 100 ± 6%, KO control 115 ± 4%, WT U0126 94 ± 6%; KO U0126 91 ± 6%; ANOVA genotype × treatment p < 0.03; n = 9; Fig. 3B). These results provide the first evidence that downregulation of the ERK1/2 pathway is effective in correcting a core Fmr1 KO phenotype.
In light of a recent study proposing that the Akt/mTOR pathway contributes to the exaggerated mGluR-LTD in the Fmr1 KO (Sharma et al., 2010), we examined whether application of the mTOR antagonist rapamycin could rescue the excess protein synthesis in the Fmr1 KO. Hippocampal slices were incubated ± 20 nm rapamycin for 30 min, then metabolically labeled ± 20 nm rapamycin for 30 min. This treatment failed to correct the excess protein synthesis seen in the Fmr1 KO (WT control 100 ± 5%, KO control 115 ± 6%, WT rapamycin 103 ± 3%, KO rapamycin 125 ± 8%; ANOVA genotype p < 0.002, ANOVA genotype × treatment p = 0.51; n = 13; Fig. 3C). Rapamycin was confirmed to produce robust inhibition of mTOR by monitoring the phosphorylation of p70S6K, a direct downstream target (WT control 100 ± 19%, WT rapamycin 24 ± 5%; KO control 94 ± 14%, KO rapamycin 15 ± 4%; ANOVA treatment p < 0.0001, ANOVA genotype × treatment p = 0.90; n = 7; Fig. 3C). These data suggest that the Akt-mTOR pathway does not contribute directly to the excessive protein synthesis seen in the Fmr1 KO and might influence LTD by actions other than regulation of protein synthesis.
TrkB-mediated protein synthesis is occluded in the Fmr1 KO
Results from our protein synthesis and signaling pathway experiments suggest that ERK1/2 links mGluR5 to protein synthesis, and that protein synthesis in the Fmr1 KO is hypersensitive to the activation of this pathway (Fig. 3D). This model would account for both the selective reduction of protein synthesis in the Fmr1 KO with MPEP and U0126, and the selective increase in protein synthesis in the WT with DHPG stimulation (Fig. 3D). A prediction of this model is that synaptic protein synthesis in response to any activator of ERK1/2 will be saturated in the Fmr1 KO hippocampus. To test this idea, we looked at protein synthesis downstream of the BDNF receptor, TrkB, in WT and Fmr1 KO slices. Activation of TrkB has been shown to result in ERK1/2-dependent protein synthesis at the synapse, and this is thought to play a role in sustained LTP (Kang and Schuman, 1996; Schratt et al., 2004; Kanhema et al., 2006).
Our goal was to test TrkB-mediated translation in our hippocampal slices; however, the penetration of BDNF in brain slices has been shown to vary considerably depending on experimental conditions (Kang et al., 1996). In addition, BDNF is known to activate p75-NTR, a receptor involved in cytotoxicity (Chao, 1994). Given these limitations, we chose to employ an antibody-based TrkB activation strategy. Dimerization of Trk receptors in response to agonist binding leads to autophosphorylation and initiation of downstream signaling cascades (Jing et al., 1992). Trk antibodies lead to the same activation, and have been shown to initiate the downstream signaling and functional consequences to the same degree as agonist application for both TrkA and TrkB (Clary et al., 1994; Qian et al., 2006).
To confirm that our monoclonal antibody activated TrkB, we applied 1 μg/ml to mature cultured hippocampal neurons for 15 min. Western blotting for p-Trk Tyr490 reveals that this treatment robustly activates TrkB, and occludes further activation via BDNF (control 100 ± 20%, control + BDNF 553 ± 53%, TrkB 849 ± 79%, TrkB + BDNF 897 ± 14%; ANOVA, BDNF p < 0.002, TrkB p < 0.0001, BDNF × TrkB, p < 0.005; n = 4 sets of cultures) (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). Importantly, this treatment also results in strong ERK1/2 activation that mimics and occludes activation via BDNF (control 100 ± 3%, control + BDNF 173 ± 12%, TrkB 252 ± 20%, TrkB + BDNF 221 ± 22%; ANOVA BDNF p = 0.187, TrkB p < 0.002, BDNF × TrkB p < 0.02; n = 4 sets of cultures) (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). We therefore used this approach to measure TrkB-mediated translation in hippocampal slices. Slices were preincubated ± 1 μg/ml anti-TrkB for 30 min, then ActD ± 1 μg/ml anti-TrkB for another 30 min, and protein synthesis measured ± 1 μg/ml anti-TrkB for 1 h (Kelleher et al., 2004). This treatment led to a significant increase in protein synthesis in WT slices, but failed to raise protein synthesis over the elevated basal level in Fmr1 KO slices (WT control 100 ± 6%, WT TrkB 132 ± 6%, KO control 121 ± 6%, KO TrkB 113 ± 9%; ANOVA genotype × treatment p < 0.02; n = 6; Fig. 4). These results suggest that, similar to mGluR-mediated protein synthesis, TrkB-mediated protein synthesis is occluded (and therefore dysregulated) in the Fmr1 KO hippocampus. This finding is interesting in light of a recent study showing that BDNF-TrkB facilitated LTP is deficient in the Fmr1 KO (Lauterborn et al., 2007). It is possible that the inability of TrkB to elicit further protein synthesis in the Fmr1 KO is related to this phenotype.
Inhibition of ERK1/2 eliminates audiogenic seizures in the Fmr1 KO
If hypersensitivity to ERK1/2 pathway signaling is a core cause of pathological changes in the Fmr1 KO, then treatment with an inhibitor of the ERK1/2 pathway might be expected to correct other phenotypes in the Fmr1 KO. Enhanced susceptibility to AGS is one of the most robust phenotypes observed in the Fmr1 KO mouse, and models the epilepsy seen in FXS patients (Berry-Kravis, 2002; Yan et al., 2005). Acute injection of MPEP has been shown to ameliorate the AGS phenotype in Fmr1 KO mice (Yan et al., 2005), and we wanted to test whether acute injection a brain-penetrant MEK1/2-ERK1/2 inhibitor (SL 327), could do the same. Fmr1 KO and WT mice were injected intraperitoneally with 100 mg/kg SL 327 (based on Zhong et al., 2009) or 50% DMSO vehicle, and returned to their home cage for 1 h. To initiate AGS, mice were transferred to a plastic test cage and exposed to a seizure-inducing alarm for 2 min. During stimulus presentation, mice were scored for four stages of AGS: wild running, clonic seizure, tonic seizure, and death (Yan et al., 2005; Dölen et al., 2007; Zhong et al., 2009). All animals were injected and scored blind to genotype.
Consistent with previous studies, we observed that vehicle-treated Fmr1 KO mice exhibited a 73% incidence of AGS, in stark contrast to the 0% incidence observed in vehicle-treated WT mice (Fisher's exact test *p < 0.03; Table 3) (Yan et al., 2005; Dölen et al., 2007). Strikingly, SL 327 treatment completely eliminated AGS in the Fmr1 KO, dropping the incidence to 0% (Fisher's exact test *p < 0.03; Table 3). These results provide the first evidence that inhibition of ERK1/2 can correct in vivo phenotypes observed in the Fmr1 KO.
We also tested the hypothesis that rapamycin can inhibit AGS, based on recent evidence that mTOR signaling is altered in some biochemical preparations of Fmr1 KO brain (Sharma et al., 2010). Fmr1 KO and WT mice were injected intraperitoneally with 6 mg/kg rapamycin (based on Ehninger et al., 2008; Meikle et al., 2008) or 100% DMSO vehicle, and tested for AGS after 1 h. Unlike the ERK inhibitor, rapamycin failed to significantly reduce the incidence of AGS in the Fmr1 KO (Fisher's exact test KO vehicle vs WT vehicle *p < 0.005, KO vehicle vs KO rapamycin p = 0.372; Table 4). Together, these results suggest that acute inhibition of ERK1/2, but not mTOR, can correct the AGS phenotype in the Fmr1 KO.
Although reduction of mGluR5 activity has been shown to correct multiple phenotypes in the Fmr1 KO mouse (Aschrafi et al., 2005; Yan et al., 2005; Tucker et al., 2006; Dölen et al., 2007; Nakamoto et al., 2007; de Vrij et al., 2008; Qiu et al., 2009), the molecular basis has been unclear. We show here, for the first time, that acute pharmacological inhibition of either mGluR5 or the ERK1/2 signaling pathway is sufficient to normalize protein synthesis in Fmr1 KO hippocampus to WT levels. The elevated protein synthesis observed in the Fmr1 KO mimics and occludes any further increases with DHPG stimulation. However, neither basal nor stimulated activation of either the MAPK or PI3K signaling pathways were found to be increased in the Fmr1 KO. Thus, the cause of increased protein synthesis in the Fmr1 KO brain appears to be increased sensitivity of the protein synthetic machinery to mGluR5-ERK1/2 signaling, rather than increased mGluR5-ERK1/2 signaling per se (Fig. 5). Consistent with this model, we find that protein synthesis stimulated via another synaptic activator of ERK1/2, the TrkB receptor, is also saturated in Fmr1 KO hippocampus. Hypersensitivity to ERK1/2 pathway activation appears to be relevant to the disorder in vivo as we found that pharmacological antagonism of this pathway completely eliminates the AGS phenotype in the Fmr1 KO.
A well documented mechanism of translational control is the modulation of mRNA availability to the ribosome (Richter, 2007). Given that FMRP has been estimated to bind 4% of all brain mRNAs (Ashley et al., 1993), it is possible that derepression and “leaky” translation of these mRNAs in response to basal ERK1/2 signaling causes the elevated protein synthesis seen in the Fmr1 KO. Consistent with this idea, the mRNA granule population, thought to represent translationally dormant mRNAs, is reduced in the brains of Fmr1 KO versus WT mice (Aschrafi et al., 2005). Furthermore, acute in vivo administration of MPEP in the Fmr1 KO animals was sufficient to shift the mRNA granule population closer to the WT value. Considered together with the current findings, the data suggest that mGluR5 is a major initiator of activity-regulated synaptic protein synthesis in the brain, and that constitutive mGluR5-ERK1/2 signaling is responsible for much of the excessive protein synthesis in the Fmr1 KO under basal conditions.
Basal protein synthesis in the WT hippocampal slice preparation was not significantly reduced by either U0126 or rapamycin (Fig. 3), but was eliminated by cycloheximide (supplemental Fig. S1, available at www.jneurosci.org as supplemental material). Although it is possible that increased exposure to the inhibitors might reveal some inhibition of protein synthesis (e.g., Kelleher et al., 2004; Nie et al., 2010), the treatments we used were sufficient to produce substantial inhibition of MEK and mTOR enzymatic activity as assayed by reduced levels of phosphorylated ERK and p70S6K, respectively. Thus, the data suggest that under the conditions of our experiments, basal ERK and mTOR signaling contribute little to basal protein synthesis in the WT. The residual protein synthesis in the presence of inhibitors is either constitutive or driven by other signaling pathways.
Activation of mGluRs has been shown to stimulate the ERK1/2 pathway in a wide variety of systems, including cell lines (Ferraguti et al., 1999), cultured neurons (Mao et al., 2005), striatum (Choe and Wang, 2001), spinal cord (Adwanikar et al., 2004), hippocampal slice (Berkeley and Levey, 2003; Gallagher et al., 2004), and retinal pigment epithelial cells (García et al., 2008). Although recent studies have shown an activation of Akt (Hou and Klann, 2004; Sharma et al., 2010) and mTOR (Antion et al., 2008; Ronesi and Huber, 2008; Sharma et al., 2010) in response to Gp1 mGluR activation, we did not observe this under the conditions of our experiment. It is possible that this pathway is activated by DHPG at earlier or later time points following stimulation than what we examined. However, the data do indicate that this pathway is not hyperactive under basal conditions in Fmr1 KO slices that exhibit excessive protein synthesis. Failure to observe increased activation of either ERK1/2 or Akt pathways under basal conditions is not accounted for by an insensitivity of our assay to detect increases, as evidenced by the effects of DHPG on ERK and insulin on Akt (Fig. 2).
Our findings contrast with a recent study by Sharma et al. (2010) who showed increased Akt-mTOR pathway activation in the Fmr1 KO hippocampus, attributed by the authors to be caused by elevated expression of the PI3K enhancer protein PIKE (Sharma et al., 2010). One key difference between our study and theirs is the preparation of the Fmr1 KO tissue. Our preparation was optimized to measure differences in protein metabolism in a physiologically stable tissue slice whereas theirs was optimized to capture early postmortem differences in protein phosphorylation. To ensure that we could replicate their findings, we probed the status of Akt and mTOR phosphorylation in homogenates from rapidly dissected hippocampus and similarly observed an increase in p-Akt and p-mTOR in the Fmr1 KO relative to WT (supplemental Fig. S4, available at www.jneurosci.org as supplemental material). This difference observed in freshly dissected tissue could reflect the status in vivo or, equally likely, a differential response to postmortem stress such as anoxia or rapid cooling. Regardless, our results strongly suggest that elevated Akt-mTOR signaling is not a cause of elevated protein synthesis in the Fmr1 KO (Fig. 5). First, we do not see evidence for increased Akt-mTOR signaling under experimental conditions that reveal an elevation in hippocampal protein synthesis quantitatively identical to the status in vivo (Qin et al., 2005). Second, the mTOR inhibitor rapamycin does not affect the basal increase in protein synthesis in the Fmr1 KO slices. As Sharma et al. suggest, increased activation of the Akt-mTOR pathway could be a consequence of overexpression of the regulatory protein PIKE. In this case, it seems more appropriate to view differences in the mTOR pathway as a consequence rather than a cause of increased mGluR5-regulated protein synthesis in the Fmr1 KO. That aberrant mTOR pathway activation may be distal to increased protein synthesis in fragile X in no way diminishes the possible therapeutic significance of that discovery. However, our finding that rapamycin treatment fails to prevent AGS (Table 4) suggests that elevated mTOR activity may not be centrally involved in epileptogenesis associated with FXS.
Our results showing that protein synthesis downstream of TrkB activation is saturated in the Fmr1 KO, and that inhibition of ERK1/2 signaling can correct excessive protein synthesis, suggest that a core defect in FXS is leaky translation in response to ERK1/2 activity. Supporting this, we show that acute pharmacological antagonism of ERK1/2 completely rescues the AGS phenotype in the Fmr1 mouse (Table 3). Although the major regulator of the relevant protein synthesis activity at excitatory synapses is mGluR5, it seems likely that other neurotransmitter systems that signal via ERK1/2 can contribute to the pathophysiology of FXS. This insight suggests the possibility of additional therapeutic targets in addition to mGluR5 for the treatment of the core pathophysiology of this disorder in humans.
This work was supported in part by grants from FRAXA, the National Institute of Mental Health, the National Institute of Child Health and Human Development, the Simons Foundation, and the Hilibrand Foundation. For excellent technical support and helpful discussions we thank the following: Kathleen “Barbara” Oram, Suzanne Meagher, Erik Sklar, Arnold Heynen, Genevieve Conley, Cornelia Hall, Eugenia Gisin, Stephanie Lacy, Charlotte Yang, Lena Khibnik, Monica Linden, Rahmat Muhammad, Bridget Dolan, Gül Dölen, and Gordon Smith.
- Correspondence should be addressed to Mark F. Bear, The Picower Institute for Learning and Memory, MIT 46-3301, 43 Vassar Street, Cambridge, MA 02139.