Deficiency of G3BP1, the stress granules assembly factor, results in abnormal synaptic plasticity and calcium homeostasis in neurons


Address correspondence and reprint requests to Jamal Tazi, IGMM-CNRS, 1919 Route de Mende, 34293 Montpellier, France. E-mail:


Ras-GAP SH3-domain–binding protein, G3BP, is an important component in the assembly of stress granules (SGs), which are cytoplasmic aggregates assembled following translational stress. To assess the physiological function of G3BP, we generated viable G3bp1-knockout (KO) mice, which demonstrated behavioral defects linked to the CNS-associated with ataxia phenotype. Immunohistochemistry pinpointed high expression of G3BP in the cytoplasm of hippocampal neurons and Purkinje cells of the cerebellum of wild-type mice. Also, electrophysiological measurements revealed that the absence of G3BP1 leads to an enhancement of short-term potentiation (STP) and long-term depression in the CA1 area of G3bp1 KO mice compared with wild-type mice. Consistently, G3BP1 deficiency in neurons leads to an increase in intracellular calcium and calcium release in response to (S)-3,5-Dihydroxyphenylglycine, a selective agonist of group I metabotropic glutamate receptors. These results show, for the first time, a requirement for G3BP1 in the control of neuronal plasticity and calcium homeostasis and further establish a direct link between SG formation and neurodegenerative diseases.

Abbreviations used

Alzheimer's disease




excitatory post-synaptic potential




long-term depression


T cell internal antigen 1



Stress granules (SGs) are cytoplasmic RNA granules, formed following exposure of cells to various stresses (Anderson and Kedersha 2008) and following initiation of translation inhibition (Mazroui et al. 2006). In neurons, SGs are part of neuronal granules found in dendrites and are associated with the inhibition of translation initiation and the disassembly of polysomes (Anderson and Kedersha 2008). In addition to RNAs, SGs contain various proteins, including cellular RNA-binding proteins such as T cell internal antigen 1 (TIA-1) and Ras-GTPase-activating SH3-domain-binding protein 1 (G3BP1) binding to mRNPs. Although the respective roles of these proteins in SG-associated functions have not yet been fully elucidated, G3BP1 has been shown to play a critical role in the assembly of SGs (Tourrière et al. 2003). When over-expressed in somatic cells, G3BP1 induces SG formation in the absence of stress, and this function depends on its N-terminal domain, which is homologous to nuclear transporter factor 2 and its C-terminal RNA recognition motif, which binds to specific RNA motifs (Tourrière et al. 2003).

To study the function of G3BP1 in vivo, we generated G3bp1 knockout (KO) mice in a mixed ‘Balb/c/129/Sv’ background and obtained viable mice. Here, we describe the phenotype of these mice and show for the first time that G3BP1 plays a role in synaptic transmission and plasticity in the hippocampus. We further show an alteration of calcium homeostasis in G3BP1-deficient hippocampal neurons, which can be linked to the defects in neuronal transmissions.

Materials and methods

All animal procedures were conducted in strict adherence with the European Community Council Directive of 24 November 1986 (86–609/EEC). For behavioral tests and electrophysiological recordings, animals were transferred to the university animal facility (CECEMA, University of Montpellier 2) 1 week before the experiments. They were housed in group, allowed food and water ad libitum except during experiments. They were maintained in a controlled environment (22 ± 1°C, 55 ± 5% humidity) with a 12 : 12 h light : dark cycle (light on at 7 : 00 am).

Behavioral tests

Open-field behavior

The general motility of male mice (WT and KO 129/SvBalb/cF1) was examined in a circular gray Polyvinyl chloride (PVC) arena (diameter 75 cm) as previously described (Maurice et al. 2008). Two concentric circles were drawn on the floor (diameter 15 cm and 45 cm, respectively), with the outer ring being divided into eight partitions and the middle ring into four partitions. The open-field session consisted of placing the mouse in the center circle and monitoring its movements for 10 min using a video camera. The following parameters were evaluated: (1) the time taken to move out of the center circle (departure latency (s)); (2) locomotion activity, in terms of number of partitions crossed, then calculated as the distance traveled (m) and locomotion in the five central partitions; (3) immobility duration (s); (4) number of rearings; (5) rearing frequency; (6) number of groomings; (7) locomotor speed (cm/s) (see Fig. 1c). Data (mean ± SEM) were analyzed using Student's t-tests.

Figure 1.

Growth abnormalities, motor and behavioral dysfunctions in G3BP1 knockout (KO) mice. (a) Weights of G3BP wild-type (+/+), heterozygote (+/−), and KO (−/−) mice (129/SvBalb/cF1) over a 21-day period. (b) Photographs of 1-month-old G3BP1 KO mice with the incapacitated hind limbs compared with wild-type (WT). (c) Behavioral parameters of 2-month-old G3BP1 KO and WT mice in the open-field test. G3BP1 KO mice show a pattern of alterations, with diminished locomotor ability and grooming behaviors. (d, e) Locomotor activity in the open-field test. Locomotion test was analyzed in terms of time course of the total locomotor activity during the 10-min duration session (d) and locomotion in 2-min time interval (e). (f) When lifted by the tail, WT mice extend their legs, whereas G3BP1 KO mice show a paw-clasping reflex (upper right). Batches of WT, KO-without-hind-limb-paralysis (NA KO, ‘non-affected’), and KO-with-paralysis (A KO, ‘affected’) G3BP mice were subjected to this tail suspension test and scored for abnormal reflex at different ages. Experiment with akinetic limb-paralysed G3BP1 KO mice could not be followed more than 2 months because of a high mortality. (g) Alteration of short-term memory in G3BP1 KO mice. Spontaneous alternation (g, upper panel) and total number of arm entries (g, lower panel) during the 8-min duration in the Y-maze. (h) Step-down latencies (presented as median and interquartile range) in the passive avoidance task. Abbreviations: T1, training session 1; T2, training session 2; R, retention session. Data are represented as means ± SEM. *< 0.05, **< 0.01, ***< 0.001 vs. WT data, Student's t-test in (a–g), Mann–Whitney's test in (h); ##< 0.01 vs. T1 data, Wilcoxon's signed rank test.

Spontaneous alternation in the Y-maze

The spatial working memory was analyzed using the spontaneous alternation performance in the Y-maze as previously described (Maurice et al. 2008). Animals (WT and KO 129/SvBalb/cF1, male) were allowed to explore a Y-shaped gray PVC. As in (Maurice et al. 2008), ‘each arm was 40 cm long, 13 cm high, 3 cm wide at the bottom, 10 cm wide at the top, and converging at an equal angle. Each mouse was placed at the end of one arm and allowed to move freely through the maze during an 8 min session. The series of arm entries, including possible returns into the same arm, was recorded using an Apple IIe computer. An alternation was defined as entries into all three arms on consecutive occasions. The number of maximum alternations was therefore the total number of arm entries minus two and the percentage of alternation was calculated as (actual alternations/maximum alternations) × 100. Data (mean ± SEM) were analyzed using a Student's t-test.'

Step-down-type passive avoidance test

The non-spatial long-term memory of WT and KO male 129/SvBalb/cF1 mice was analyzed using a step-down-type passive avoidance behavior as previously described (Maurice et al. 2008). The apparatus consisted of a transparent acrylic cage (30 × 30 × 40 cm high) with a grid-floor, inserted in a soundproof outer box (35 × 35 × 90 cm high). The cage was illuminated using a 15-W lamp. As in (Maurice et al. 2008), ‘a wooden platform (4 × 4 × 4 cm) was fixed at the center of the grid-floor. Electric shocks (1 Hz, 500 ms, 43 V DC) were delivered to the grid-floor using an isolated pulse stimulator (Model 2100, AM Systems, Everett, WA, USA). The test involved two training sessions, at 90-min time interval, and in a retention session carried out 24 h after the first training. Each mouse was placed on the platform during the training sessions. When it stepped down and placed its four paws on the grid-floor, shocks were delivered for 15 s. Step-down latency and the numbers of vocalizations and flinching reactions were measured. Shock sensitivity was evaluated by summing these last two numbers. Animals which did not step down within 60 s during the second session were considered as remembering the task and taken off without receiving more shocks. The retention session was performed after 24 h, similarly as training, but with no shock. Each mouse was placed again on the platform and the step-down latency was recorded with a cut-off of 300 s. Two parametric measures of retention were analyzed: the latency and the number of animals reaching an avoidance criterion, defined as correct if the retention latency was higher than three-fold the 2nd training latency and at least greater than 60 s. Latencies did not show a normal distribution, as cut-off times were set, and were thus expressed as median and interquartile range. They were analyzed using a non-parametric Mann–Whitney's test and the percentages of animals to criterion were analyzed using a Chi-squared test.'

SDS–PAGE and western blotting

Mice (WT, ± and KO 129/SvBalb/cF1) brain extracts were lysed in lysis buffer (HEPES 50 mM, NaCl 150 mM, glycerol 10%, Triton X-100 1%, Sodium pyrophosphate 10 mM, Sodium Fluorure 100 mM, EGTA 1 mM, MgCl2 1.5 mM, 1 mM Na3VO4), Protease Inhibitor (Calbiochem, San Diego, CA, USA), using a FastPrep tissue homogenizer. Protein concentrations were determined using BCA protein assay kit (Pierce, Rockford, IL, USA). Protein samples boiled in Laemmli sample buffer were resolved by 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes, which were probed with anti-G3BP1 (Abnova, Taipei, Taiwan) (1 : 3000) and anti-β-actin (Sigma, St Louis, MO, USA) (1 : 5000).

Histological analyses

Brains removed from killed WT and akinetic KO 129/SvBalb/cF1 mice were fixed in 10% buffered formalin (Sigma) and cut into serial 5-μm paraffin sections, then stained with hematoxylin and eosin.


Mice tissues were fixed by intracardiac perfusion of 4% paraformaldehyde of anesthetized mice (129/SvBalb/cF1, male and female). Brains were included in 4% agarose to make 100-μm vibratome sagittal sections, and immunofluorescences were carried out on free-floating sections. Concerning neurons in culture, fixation with 4% PFA/4% sucrose was performed after 14 days of culture, after 1 h of sodium arsenite treatment. After permeabilization in Triton 1% (slices) or 0.5% (neurons), slices/neurons were blocked in fetal bovine serum 5% and incubated with primary antibodies overnight. Mouse anti-G3BP1 (Abnova), 1 : 50 and rabbit anti-G3BP1 (Novus, Littleton, CO, USA), 1 : 100, anti-TIA-1 kindly provided by N. Kedersha and P. Anderson, 1 : 200, anti-Caprin-1 kindly provided by Y.Wang, 1 : 500, and rabbit anti-PCP2 (Abnova), 1 : 100, were used, with goat anti-rabbit and anti-mouse Alexa 488/546 secondary antibodies (1 : 1000; Molecular Probes, Eugene, OR, USA). Nuclear staining was obtained using Hoechst. Images were acquired using a confocal microscope (Zeiss, LSM 510 META, Zeiss, Oberkochen, Germany, and Leica, SP5, Leica, Nanterre, France) and the images panels were constructed using Image J software.


Experiments were carried out on freshly prepared hippocampal slices (350 μm) obtained from 3- to 8-week-old G3BP1 KO (akinetic) or WT mice (129/SvBalb/cF1, male and female). After decapitation, brains were quickly dissected and placed in ice-cold buffer containing 124 mM NaCl, 3.5 mM KCl, 25 mM NaHCO3, 1.25 mM NaH2PO4, 1 mM CaCl2, 2 mM MgSO4, and 10 mM glucose (infused with O2/CO2: 95/5%). Slices were then prepared using a vibratome (VT1000S Leica) and maintained at 25°C for at least 1 h before electrophysiological recording in the same buffer supplemented with 1 mM CaCl2. The latter buffer was used for further recordings. Hippocampal slices were transferred to a submerged-style recording chamber positioned under an upright microscope (DMF, Leica) and superfused with the extracellular medium (flow rate, 2 mL/min). Field excitatory post-synaptic potential (EPSPs) were recorded in the dendritic field of CA1 hippocampal neurons with glass microelectrodes (4–5 M resistance) filled with extracellular medium described above. Field EPSPs were evoked by test frequency stimulation (0.066 Hz) of the Schaffer collateral–commissural pathway using a bipolar electrode. Input/output were generated by plotting fEPSP amplitude, obtained at different stimulation intensities (from 30 to 150 μA), against the corresponding fiber volley amplitude. The statistical significance was calculated using multivariate analysis of variance (anova). Differences were considered significant when p < 0.05.

Long-term depression (LTD) was triggered by delivering either 900 shocks at 1-Hz frequency or 900 paired shocks (with 50-ms interval) at 1-Hz frequency. Electrophysiological signals were amplified (Axopatch 200B; Axon Instruments, Jakarta, Indonesia) and digitized (Digidata 1200A; Axon Instruments). Field EPSPs were further recorded and analyzed using Dr. W. Anderson's LTP software (Anderson and Collingridge 2007). The statistical significance was calculated using multivariate analysis of variance (anova). Differences were considered significant when p < 0.05.

Neuron culture and intracellular calcium concentration measurement

Cultures of dissociated primary hippocampal neurons were prepared as previously described (Dotti et al. 1988). Briefly, hippocampi from E18.5 (embryonic day 18.5) embryos (129/SvBalb/cF1) extracted from ± females (heterozygous 129/Sv or Balb/c crossed with heterozygous Balb/c or 129/Sv males, respectively) were removed and digested in 0.25% trypsin (in HEPES-buffered Hanks' balanced salt solution) at 37°C for 15 min. Cells dissociated by pipetting up and down were plated at a concentration of 5000–15000 cells/cm2 on poly-l-lysine-coated coverslips (Sigma), in Dulbecco's modified Eagle's medium-10% fetal bovine serum. Three to four hours after seeding, the media was replaced by neurobasal medium supplemented with Glutamine and NS21 [prepared as described in (Chen et al. 2008)]. Embryos were genotyped by PCR on genomic DNA extracted from a piece of their tails, following the same conditions and using the same oligonucleotides as in (Zekri et al. 2005). Neurons from genotyped WT and KO embryos were used.

Free intracellular calcium concentration ([Ca2+]i) was measured with the fluorescent indicator Fura-2 (Grynkiewicz et al. 1985). For this purpose, hippocampal cells grown on glass coverslips were loaded with fura-2 by incubation for 30 min at 37°C with the extracellular solution: 125 mM NaCl, 5 mM KCl, 26 mM NaHCO3, 2 mM CaCl2, 1 mM MgCl2, 10 mM d glucose, 10 mM HEPES containing 5 μM fura-2AM and 0.02% Pluronic. [Ca2+]i was monitored by videomicroscopy. A coverslip was transferred to the recording chamber mounted on a microscope (Leica, DMIRB). Fura-2 emission was obtained by exciting alternatively at 340 and 380 nm with a rotating filter wheel (Sutter Instruments, Novato, CA, USA) and by monitoring emissions (F340 and F380) at 510 nm. The ratio of emissions at 510 nm (F340/F380) was recorded every 30 or 60 s. Fluorescent signals were collected using a CCD camera (Hamamatsu Photonics, Massy, France), digitized, and analyzed using image analysis software (Acquacosmos, Hamamatsu). The coverslips were continually superfused with extracellular solution. Drug application was performed with a gravity-fed system. Concerning the basal intracellular Ca2+ concentration, the ratio values were converted to estimated measurements using the equations of Grynkiewicz. The calibration was performed in situ, in which Rmax, Rmin, and β values were determined using solutions containing a Ca2+ ionophore (10 μM 4Br-A23187) alone or with Ca2+-free medium plus 1 mM EGTA. For internal calcium release analysis, (S)-3,5-dihydroxyphenylglycine (DHPG) was applied at a concentration of 10 μM for 30 s. The responsiveness of neuronal cells were checked by applying 30 mM of KCl.


Phenotypic characterization and behavioral examination of G3BP1 KO mice

Homozygous null mutations of the G3bp1 gene in 129/Sv mice induce embryonic lethality (Zekri et al. 2005). To circumvent the potential influence of genetic background, mice heterozygous for a null mutation in G3bp1 gene were crossed with Balb/c mice, for 10 generations. Resultant heterozygote mice were intercrossed with heterozygote 129/Sv mice. Viable homozygous knockout mice with a mixed genetic background (50% 129/Sv/50% Balb/c: 129/SvBalb/cF1) were thus obtained. Offspring homozygous for the targeted mutation were detected at a lower frequency than predicted by Mendelian laws (33% WT, 58% heterozygous, 9% KO), indicating that G3bp1−/− is lethal during fetal development. Moreover, although KO mice were viable, less than 20% of G3bp1−/− mice lived longer than 2 months (male and female). On average, the body mass of G3BP1-deficient mice was also half to three-fourth that of WT mice, and this growth retardation was increasingly apparent with age (Fig. 1a). In addition, G3BP1 KO mice (male and female) demonstrate a striking akinesia; most KO mice dragged their limbs behind them when walking (Fig. 1b).

Prior to any further behavioral investigation, we examined the general motility of male WT and G3BP1 KO mice in an open-field paradigm. As summarized in Fig. 1c, G3BP1 KO mice showed an important pattern of alterations. G3BP1 KO mice presented a decreased locomotor ability, a diminished locomotor speed and an increased immobility duration. It is important to notice that this diminution in general activity was not restricted to mobility, as decreased number of rearing and grooming behaviors were also measured. Analysis of the locomotor time-course pattern during the 10-min session showed that the cumulative activity, measured in 2-min time intervals, regularly increased for both WT and KO mice, but remained significantly lower in KO mice at all time points (repeated measure anova,< 0.0001: F(4,49) = 69.1, n = 10, for WT; F(4,59) = 16.8, n = 12, for KO, Fig. 1d). The activity analyzed within 2-min time interval decreased significantly for WT animals (F(4,49) = 6.71, < 0.001; Fig. 1e), but on the contrary, increased for KO animals (F(4,59) = 6.04, < 0.001; Fig. 1e), suggesting an alteration of the spontaneous reaction to a novel environment (a putative lack of motivation to explore, for instance).

We have also analyzed the motor functions of G3BP1−/− mice. When suspended by the tail toward the floor (limb clasping test), WT mice (male and female 129/SvBalb/cF1) extend their limbs, whereas G3BP1−/− mice showed paw clasping reflex close to a bat-like posture (Fig. 1f, upper right pictures). This test was performed with limb-paralyzed and non-paralyzed KO mice, to assess if this response was only because of the physical incapacitation of the mice or if it was because of an alteration of the CNS. Quantitation of this behavior in mice ranging from 15 days to 2 years of age by scoring the position of their limbs for 15 s (Fig. 1f) demonstrated that KO mice with paralysis (‘affected’, n = 4) have the highest score, but that ‘unaffected’ KO mice (n = 4) progressively lose normal reflex with age, rapidly acquiring higher scores than WT mice (n = 6). Given that no anomalies were detected in muscles or motor neurons from G3BP1 KO mice (data not shown), it seems more likely that the ataxia-like phenotype is because of a dysfunction of the CNS.

Learning and memory were then analyzed using two tests assessing different memory processes (Maurice et al. 2008): namely, the spontaneous alternation behavior in the Y-maze assessing spatial working memory, and passive avoidance response assessing non-spatial long-term memory. G3BP1 KO mice showed a significant impairment of spontaneous alternation, as compared with WT animals, which showed a correct percentage of alternation (62.6 ± 3.4%, n = 10, for WT mice vs. 46.5 ± 4.9%, n = 11, for G3BP1 KO mice; Fig. 1g, upper panel), suggesting working memory impairment. Interestingly, the total number of arm entries did not differ significantly between WT and KO mice (Fig. 1g, lower panel), suggesting that maze exploration was not significantly affected, contrary to what could be expected from the open-field test results. During the first passive avoidance training session (T1), G3BP1 KO mice did not show a significant difference in shock sensitivity (18 ± 3, n = 11, for WT; 13 ± 2, n = 13, for KO, > 0.05, Student's t-test), but showed a significantly higher step-down latency (Fig. 1h) that could be related to their decrease in general activity (Fig. 1c). G3BP1 KO mice showed, however, a similar significant increase in latency during T2 (Fig. 1h) and a correct learning, as the retention session latency (Fig. 1h) and percentage of animals-to-criterion (72% in WT and 82% in KO reaching an avoidance criterion, not shown) did not differ among groups.

Thus, G3BP1 appears to contribute to spatial working memory, but not to long-term memory as assessed in the passive avoidance procedure. However the general physical ability of KO mice rendered difficult a more detailed analysis of their long-term memory ability in the water-maze test.

G3BP1 localization in mouse brain

To further investigate the role of G3BP1 in different structures of the CNS, we established its expression profile by immunohistochemistry. We first confirmed its expression in WT brain and its absence in KO by western blotting (Fig. 2a), and observed that the hippocampal and cerebellar structures were maintained in the akinetic KO (Fig. 2b). In the cerebellum, however, Purkinje cells (PCs) appear affected in G3BP1 KO mice, as there was a slight Purkinje cells degeneration (assessed visually and by counting the number of Purkinje cells, PCs, by μm, with a p-value of 0.1) (Fig. 2b, lower right panel). Given that no anomalies were detected in muscles or motor neurons from G3BP1 KO mice and that PCs are the sole output of the cerebellum involved in the control of motor activity (Cheron et al. 2008), the ataxia-like phenotype might be linked to cerebellum dysfunction. However, other motor-related brain regions may also be involved in this phenotype and need to be further studied.

Figure 2.

Distribution of G3BP1 in mouse brain. (a) Western blots and immunoblotting by anti-G3BP1 antibody confirm the absence of the protein in KO brain extracts and a decrease in heterozygous extracts. Anti-β-actin antibody is used as a loading control. (b) Histological staining shows conserved anatomy of hippocampus in wild-type (WT) and akinetic G3BP1 KO mice (top panels). The cerebellum architecture is also conserved in the akinetic KO (lower panels) even if there is some loss of Purkinje cells [stained with an antibody against PCP-2 (Purkinje cell protein 2)] (lower panel, right). Scale bars = 100 μm.. The box plot represents the number of Purkinje cells per micrometer, with + the median and − the mean. ml: molecular cell layer, gl: granule cell layer, PC: Purkinje cells layer. (c) G3BP1 distribution in the hippocampus (i), with magnifications of CA1 (ii) and CA3 areas (iii), in the frontal cortex (iv), and the cerebellum (v–vi, with magnification in vi). Vibratome slices of 2-month-old WT mice, 100 μm thick, were labeled with anti-G3BP1 (green). DNA was counter-stained in blue with Hoechst. Scale bars represent 50 μm (i and v), 20 μm (ii–iv), and 25 μm (vi). (d) G3BP1 is present in stress granules formed in neurons under arsenite treatment. MAP2 (microtubule-associated protein 2) staining confirms the neurons identity. Caprin-1 and T cell internal antigen 1 (TIA-1) colocalization with G3BP1 confirms that G3BP1-containing granules are SGs. Concordant with previous studies, TIA-1 is cytoplasmic and nuclear. Scale bar = 10 μm.

Indeed, G3BP1 is found in different parts of the brain in WT, with high expression in the cerebellum and hippocampus, but also the cortex (Fig. 2c, panels a–f). G3BP1 demonstrated strong expression in the soma of the dentate granule cell layer and CA pyramidal regions of hippocampal formation (Fig. 2c, panel a, b, and c). In the cerebellum, it is predominantly found in the Purkinje cells layer (Fig. 2c, panel e–f).

To assess the formation of G3BP1-positive stress granules in neurons, we analyzed the intracellular localization of endogenous G3BP1 by performing immufluorescence in cultured hippocampal neurons. In untreated neurons, G3BP1 is distributed in large granular structures in the soma (Fig. 2d). However, when the neurons are stressed with arsenite treatment for 1h, G3BP1 becomes localized in discrete cytoplasmic structures that resemble previously described G3BP1-positive SGs (Fig. 2d). Furthermore, these granules colocalize with Caprin-1, partner of G3BP also involved in SGs assembly, and with TIA-1, a known marker of SGs. G3BP1, which is present in cells important in plasticity like in the hippocampus, might be an actor in the processes of learning/memory in mouse.

Synaptic plasticity in hippocampal CA1 neurons of mice lacking G3BP1

In view of the spatial memory impairment of the KO mice, we next characterized synaptic transmission by evoking excitatory synaptic transmission between Schaffer collaterals and CA1 neurons in 1- to 2-month-old KO mice (presenting an akinetic phenotype) and their WT littermate controls (129/Sv/Balb/cF1).

First of all, basal synaptic transmission was examined. The Input–Output curve was constructed by plotting the fEPSP amplitude obtained at various stimulation amplitudes as a function of the corresponding fiber volley. It appears that in KO mice, the Input–Ouput curve had a greater slope than that obtained with WT [5.26 ± 0.46 (r2 = 0.98; n = 4) vs. 4.98 ± 0.49 (r2 = 0.99; n = 4); p < 0.001 (anova)] (Fig. 3a). This indicates a facilitatory effect of basal transmission in KO mice.

Figure 3.

Synaptic plasticity is modified in G3BP1 knockout (KO) mice. (a) Input–output relationship in wild-type (WT) and akinetic KO mice. Graphs were generated by plotting the field excitatory post-synaptic potential (EPSP) amplitude against the corresponding fiber volley amplitude obtained for different stimulation values. Data are averages of both fEPSP and fV amplitudes. (b) Short-term synaptic plasticity in KO and WT mice. Paired-pulse facilitation was elicited by pairing stimulation pulses with an interval ranging from 25 to 500 ms. In the graph, data are means ± SEM of four independent experiments conducted with akinetic KO mice and five independent experiments with WT mice. Samples traces extracted from recordings of synaptic facilitation obtained at an interpulse interval of 300 ms in WT and KO mice are shown. (c) Depression was not sustained in WT or KO mice when stimulations of 1 Hz were done during 15 min. (d) However, paired-pulse stimulations induced a robust long-term depression (LTD) in KO mice, not observed in WT. This LTD was abolished by 2-Methyl-6-(phenylethynyl)pyridine, an inhibitor of metabotropic glutamate receptors.

Short-term synaptic plasticity was then studied, by applying paired stimuli with increasing interval from 25 to 500 ms. Paired-pulse facilitation ratio was significantly higher in KO mice (n = 4) than in WT (n = 5) at all the intervals considered (anova, p < 0.001) (Fig. 3b), demonstrating an enhancement of short-term synaptic plasticity. Thus, the invalidation of G3BP1 gene leads to a pre-synaptic facilitatory effect on synaptic transmission.

Long-lasting synaptic changes were then studied. We first examined the induction of long-term potentiation (LTP) by one train at 100 Hz (1 s) and four trains (100 Hz, 1 s) with 5-min interval between trains. However, both protocols led to transient potentiations in both WT and KO mice. This negative result can be attributed to the mouse strain used here. Indeed, although there are no data available regarding synaptic plasticity in the mouse strain Balb/c backcrossed with 129/Sv, different recent reports indicate that 129/Sv mice are unable to produce conventional LTP (Nguyen et al. 2000; Blackshaw et al. 2003). However, a substantial synaptic enhancement can be obtained by delivering a 4 × 100 Hz stimulus protocol with a 3-s interval between each 100-Hz train (Nguyen et al. 2000). We have tried such a protocol with this mouse strain: it consistently led to the occurrence of epileptic-like activity in both WT and KO mice (data not shown). LTD was also examined. In both WT and KO, conventional LTD elicited by 900 shocks delivered at 1 Hz for 15 min was not observed (Fig. 3c). However, applying paired-pulse shocks at 1 Hz for 15 min (paired-pulse Low-Frequency Stimulation, ppLFS), which did not trigger any LTD in WT (n = 6), surprisingly elicited a consistent depression in KO (n = 5) (Fig. 3d). This paired-pulse protocol is thought to involve the activation of the group 1 metabotropic glutamate receptors mGluR1/5, mGluR1 in the cerebellum and mGluR5 in the hippocampus (Collingridge et al. 2010; Lüscher and Huber 2010). The use of 2-Methyl-6-(phenylethynyl)pyridine, a specific antagonist of mGluR5, abolished the LTD observed in KO (Fig. 3d), confirming the involvement of mGluR5 in this significant induction of LTD. Thus, in absence of G3BP1, a mechanism involved in the expression of a form of LTD dependent on mGluR is exacerbated in CA1 area of the hippocampus.

Altered intracellular calcium in G3BP1 KO neurons

Synaptic transmission involves calcium transients in pre-synaptic and post-synaptic sites. We measured intracellular calcium concentration ([Ca2+]i) in primary cultures of WT and KO hippocampal neurons (from E18.5 embryos) after 12 days of culture, using the Fura-2 probe. [Ca2+]i was higher in G3BP1-deficient cells than in WT, with half of the values above 112 nM (median) for KO, against 83 nM for the WT (mean ± SEM: 98.34 nM ± 5.85 in WT vs. 167.1 nM ± 15.68 in KO) (Fig. 4a, left panel). Consistent with the fact that calcium homeostasis is disturbed in aged neurons in the hippocampus (Hajieva et al. 2009), basal [Ca2+]i was higher at later stages of in vitro development [17 days in vitro (DIV)], and the increase in calcium in KO neurons compared with WT was even more important when considering only the measurements performed at this stage (Fig. 4a, right panel).

Figure 4.

Intracellular calcium is affected in G3BP1-deficient neurons. (a) Intracellular calcium concentration in cultured hippocampal neurons (from E18.5 embryos), calculated from Fura-2 emission ratio (F340/F380) at 510 nM when exciting at 340 and 380 nM, using the equation of Grynkiewicz. Knockout (KO) neurons have higher basal [Ca2+]i than wild-type (WT) neurons (left), even more in older neurons (17 DIV, 17 days in culture) (right). (b) When stimulated with (S)-3,5-dihydroxyphenylglycine (DHPG), KO hippocampal neurons elicit a higher calcium release from intracellular stores than WT neurons, as shown by the ratio of emissions of Fura-2 when excited at 340 and 380 nm (F340/F380) (top left). A typical response to DHPG in WT (top middle) and KO (top right) neuron. The increase in response is even more significant at 17 DIV (bottom). Data are represented as box plots depicting the four quartiles. n: total number of neurons measured, from at least four different embryos from three different litters. **p ≤ 0.005, ****p ≤ 0.00005, Mann–Whitney test.

Thus, in the absence of G3BP1, calcium homeostasis is altered in hippocampal neurons. Pre-synaptic Ca2+ concentration directly regulates basal transmission and short-term plasticity by modulating glutamate release. The defects in basal synaptic transmission and STP observed in KO can therefore be associated to this altered calcium homeostasis.

LTD induced in KO hippocampal slices was dependent on mGluR activation. In neurons, DHPG application induces the production of inositol-1,4,5-triphosphate (IP3) under mGluR activation, followed by a release of calcium from internal stores after binding of IP3 to inositol-1,4,5-triphosphate receptors (ITPR or IP3R) on internal compartments membrane. Indeed, DHPG treatment (Fig. 4b, top panels) induced an increase in the ratio of emission fluorescence (F340/F380). Interestingly, this increase tended to be significantly higher in KO neurons compare to WT, with half of the values between 1.3 and 2.1 times increase of the basal ratio for the KO, compared with WT neurons where the increase never exceeds 1.59 times. This suggests an increase in ITPR response to DHPG treatment. Again, the increase in calcium release was even more significant at 17 DIV (Fig. 4b, bottom panel).


In this study, we have generated a viable mouse knockout of G3BP1 that demonstrated major alteration of the CNS, as shown by impaired motor coordination and dysfunctions of synaptic transmission and plasticity at Schaffer collaterals—CA1 synapses. Calcium homeostasis is altered in KO and can be linked to the defects observed in synaptic transmission and potentiation, pre-synaptically and also post-synaptically, especially in the mechanisms of LTD consolidation. Given the deficits in working memory, it is interesting to link this altered calcium homeostasis in hippocampal neurons to neurocognitive defects, and to do the parallel with neurodegenerative diseases appearing with aging in humans, like Alzheimer's disease (AD).

As in a number of neurodegenerative diseases, the hippocampus and Purkinje cells (PCs), important sites in plasticity events, are affected under G3BP1 depletion. Our data establish that G3BP1 is highly expressed in PCs, in which the various forms of synaptic plasticity involve calcium ions to an important extent. Indeed, even if cerebellar Purkinje cells are less subject to the effects of aging and neurodegeneration than the hippocampus, calcium concentration is highly regulated in Purkinje cells, being involved in cerebellar dysfunctions and normal long-term synaptic plasticity events important in motor learning (Rinaldo and Hansel 2010; Lamont and Weber 2012). Given the phenotype of ataxia observed in G3BP1 KO mice, it would be interesting to study LTD events in the cerebellum of these mice. Indeed, stimulation of mGluR1 (from the mGluR1/5 family) in the cerebellum at parallel fiber-Purkinje cells synapses, also leads to IP3-mediated release of Ca2+ from intracellular stores and triggering of LTD. However, some neurodegeneration could be observed in KO brain, especially a loss of Purkinje cells to some extent. This might be attributable to an excess of calcium leading to cell death, but may render the electrophysiological measurements difficult and irrelevant. It is interesting to notice that most but not all the KO mice develop akinesia-linked limb paralysis, suggesting different degrees of alteration in the central nervous system, maybe in the cerebellum. This might be partly because of various levels of loss of Purkinje cells in these animals. However, we have not fully evaluated other brain regions related to motor activity, like the sensorimotor cortex, the thalamic motor nuclei, or the associated tracts, and these regions may also be altered in the absence of G3BP1 (all the more that G3BP1 is substantially present in the cerebral cortex). The study and quantitation of neurons of these regions to see if there is neurodegeneration and loss of cells would be interesting to more completely assess the phenotype of G3BP1 KO mice.

In hippocampal neurons in culture, we showed that in KO there was an increase in calcium release through ITPR under mGluR activation. Consistent with this observation, proteomic analysis of proteins associated with G3BP in the brain (Martin et al., unpublished) revealed the association of G3BP with two proteins involved in neurotransmitter release, LTD, and calcium release (NSF and ITPR1). An increase in calcium release from the internal stores can induce a more elevated concentration of calcium in the cytoplasm of neurons. Furthermore, the intracellular calcium concentration can positively or negatively influence the activity of ITPR channels and consequently the release of calcium from the internal compartments (Joseph et al. 2005). It seems likely that a combination of elevated response mechanisms under mGluR stimulation and basal high calcium concentration leads to an exacerbated internal release of calcium in KO, even more pronounced at later stages of neuronal development in vitro.

This study has further highlighted the potential importance of G3BP1 in neuron physiopathology. Indeed, recent reports indicate that SGs containing G3BP1 become abundant and widely distributed in tauopathies and increase with disease severity (Vanderweyde et al. 2012). However, in these diseases, G3BP1 is present in neurons that do not accumulate tau aggregates which are specific markers of affected neurons characterized by tau phosphorylation and PHF-1 staining (Vanderweyde et al. 2012). This observation raises the possibility that neurons accumulating G3BP SGs are somehow protected against pathological tau and related microtubule dysfunction. Consistent with this possibility, we found that G3BP1 modulates mGluR-induced LTD events, which include rapid translation at the synapse and therefore will require microtubule function. Also, G3BP1 associates with the ubiquitin-specific protease, USP-10, a component of the deubiquitinating complex (Soncini et al. 2001) and thereby may participate in protein turnover at the synapse. As G3BP1 assembles SGs under oxidative stress and oxidative stress is central to AD, the assembly of G3BP1 SGs would be part of the neuroprotective response recently attributed to granulovacuolar degeneration within the pyramidal neurons in ADs (Castellani et al. 2011).

We have recently analyzed the proteins and RNAs present in the G3BP1 complex in brain, to shed light on the mechanisms of actions of G3BP1 in CNS functional maintenance and calcium regulation. The G3BP complex binds preferentially to intron-sequence-retaining transcripts and long non-coding RNAs derived from at least 5800 unique loci. Among the enriched transcripts, we found an overrepresentation of genes involved in synaptic transmission, especially glutamate-related neuronal transmission. Transcripts with retained introns appear to be more stable in the cerebellum than in the rest of the brain and G3BP1 depletion was correlated with decreased intron retention and change in the stability of the mature RNA in the cerebellum. The stability of G3BP targets help to explain the phenotypes associated with G3BP1 deficiency including ataxia and exacerbated neuronal responses and hippocampal plasticity.


This work was supported by a Fondation pour la Recherche Médicale (FRM) grant (Equipe FRM 2011 -n°DEQ20111223745).J.T. is a member of the Institut Universitaire de France. S.M. was supported by a graduate fellowship from the Ministère Délégué à la Recherche et aux Technologies and FRM. A.M. was supported by a fellowship from FRM. L.Z. was a recipient of Ligue Contre le Cancer (Comité de l'Herault) fellowship. We thank Julian Venables for critical reading of the manuscript. The cognitive phenotyping was a project (#04) of the CompAn behavioral phenotyping facility (Montpellier, France). We thank the Montpellier RIO Imaging (MRI) platform for the imaging analysis.

The authors declare to have no conflict of interest.