A Role for Dendritic Translation of CaMKIIα mRNA in Olfactory Plasticity

Local protein synthesis in dendrites contributes to the synaptic modifications underlying learning and memory. The mRNA encoding the α subunit of the calcium/calmodulin dependent Kinase II (CaMKIIα) is dendritically localized and locally translated. A role for CaMKIIα local translation in hippocampus-dependent memory has been demonstrated in mice with disrupted CaMKIIα dendritic translation, through deletion of CaMKIIα 3′UTR. We studied the dendritic localization and local translation of CaMKIIα in the mouse olfactory bulb (OB), the first relay of the olfactory pathway, which exhibits a high level of plasticity in response to olfactory experience. CaMKIIα is expressed by granule cells (GCs) of the OB. Through in situ hybridization and synaptosome preparation, we show that CaMKIIα mRNA is transported in GC dendrites, synaptically localized and might be locally translated at GC synapses. Increases in the synaptic localization of CaMKIIα mRNA and protein in response to brief exposure to new odors demonstrate that they are activity-dependent processes. The activity-induced dendritic transport of CaMKIIα mRNA can be inhibited by an NMDA receptor antagonist and mimicked by an NMDA receptor agonist. Finally, in mice devoid of CaMKIIα 3′UTR, the dendritic localization of CaMKIIα mRNA is disrupted in the OB and olfactory associative learning is severely impaired. Our studies thus reveal a new functional modality for CaMKIIα local translation, as an essential determinant of olfactory plasticity.


Introduction
Since the seminal observation of polyribosomes localized at the base of dendritic spines [1], local translation in dendrites has been shown to be a major determinant of neuronal plasticity, participating in the synaptic changes that underlie learning and memory [2]. Among the mRNAs that have been clearly shown to be dendritically localized and locally translated is the mRNA encoding the a subunit of the calcium/calmodulin dependent Kinase II (CaMKIIa). CaMKII is a major component of postsynaptic densities (PSD) [3] and is essential to different forms of synaptic plasticity linked to learning and memory [4]. CaMKIIa mRNA is transported into dendrites of hippocampal and cortical neurons [5,6] and this dendritic localization is mediated by its 39UTR [7]. Local translation of CaMKIIa mRNA is found in biochemical fractions enriched for synapses (synaptosomes, SN) [8,9] and in neuronal processes isolated from the soma of hippocampal neurons in culture [10]. In behaving animals, LTP induction in the hippocampus triggers a rapid delivery of CaMKIIa mRNA to dendrites [11] and synaptic sites [12]. Moreover, a 30 min exposure to light of dark-reared rats leads to an increase of CaMKIIa local translation in the visual cortex [13,14,15]. In Drosophila, neural activity drives CaMKIIa mRNA to synaptic sites, where it is rapidly translated and, most importantly, an olfactory associative learning task triggers an odor-specific induction of CaMKIIa mRNA synaptic transport and translation [16]. This strongly suggests that CamKIIa local translation is modulated by neural activity and contributes to the synaptic plasticity associated with learning and memory. To directly test the role of CamKIIa local translation in learning and memory, knocked-in mice were generated, in which CamKIIa 39UTR was replaced by the 39UTR of bovine growth hormone mRNA, a message that is not dendritically localized [17]. These mice display a dramatic reduction of CaMKIIa in PSDs of the hippocampus (HC), a reduction in late-phase long-term potentiation, and impairments of hippocampus-dependent memories. This confirms a role for CaMKIIa local translation in synaptic and behavioral plasticity.
We thus became interested in CaMKIIa local translation in the olfactory system. The olfactory bulb (OB) is the first relay of the olfactory pathway and presents a high level of plasticity in response to olfactory experience [18]. Here, we report that CaMKIIa is expressed by granule cells (GCs) of the OB. CaMKIIa mRNA is transported in GCs dendrites, synaptically localized and might be locally translated at GC synapses. This synaptic localization of CaMKIIa mRNA is regulated by olfactory activity through NMDAR. In mice devoid of CaMKIIa 39UTR [17], the mRNA dendritic localization is dramatically decreased in the OB and olfactory associative learning is impaired. Our work thus suggests a fundamental role for CaMKIIa local translation in olfactory plasticity.

CaMKIIa Expression in the OB
We first investigated the expression pattern of CaMKIIa in the OB by immunohistochemistry. We observed a strong labeling in the granule cell layer (GCL) and the external plexiform layer (EPL) (Fig. 1A). The GCL contains the cell bodies of GCs, which extend their long apical dendrite into the EPL, where they synapse onto the dendrites of mitral cells, forming reciprocal dendro-dendritic synapses. The staining was absent in the glomerular layer (Fig 1A). At the cellular level, CaMKIIa immunoreactivity surrounds GCs nuclei in their thin rim of cytoplasm and can occasionally be observed in their dendrites, extending towards the EPL ( Fig. 1B and C). In the EPL, however, dendrites could not be individualized and CaMKIIa staining appeared blurry. Mitral cells (MCs), which form a single cell layer composed of larger cells around the GCL, appeared unlabeled (Fig. 1B). Overall, this pattern of immunoreactivity is consistent with previous work [19] and confirms that CaMKIIa is expressed by GCs in the OB and present in both their cell bodies and dendrites.

CaMKIIa mRNA Dendritic Localization and Local Translation in GCs
As CaMKIIa mRNA has been described to be dendritically localized in the cortex and HC [5,6], we investigated its localization in the OB. In situ hybridization (ISH) against CaMKIIa mRNA shows strong staining in multiple regions of the brain, as described ( Fig. 2A). Staining in the HC is in agreement with previous reports: in the dentate gyrus and CA1-CA3, cell bodies are strongly labeled and a more diffuse staining is observed in the dendritic compartments. In the OB, the GCL is strongly stained. Higher magnification of the OB confirms a strong expression of CaMKIIa mRNA in the GCL and reveals a diffuse staining in the EPL, where the GC apical dendrites arborize (Fig. 2B). This suggests that CaMKIIa mRNA is dendritically localized in GCs.
To confirm and refine this result, we prepared OB synaptosomes (SN). SN are isolated resealed-synapses obtained by a biochemical fractionation. mRNAs extracted from these preparations were retro-transcribed and analyzed by quantitative PCR to calculate an index of synaptic localization ''I''. For a given mRNA, this index is the ratio of the quantity of synaptic mRNA over the quantity of this mRNA in total brain extract normalized to HPRT, a transcript restricted to the cell soma. In all analyzed experiments, HPRT mRNA contamination in SN preparation was between 0.1 and 5% (not shown). With this technique, we found that CaMKIIa mRNA is highly enriched in SN (7-fold to HPRT, n = 3), suggesting synaptic localization (Fig. 2C). PSD95 mRNA is known to be synaptically localized [20,21] and was included as a positive control. Its index of synaptic localization was similar that of CaMKIIa mRNA. Taken together, these results strongly suggest that CaMKIIa mRNA is dendritically and synaptically localized in GCs.
We then assessed whether the synaptically localized CaMKIIa mRNA could be locally translated. To this extent, we prepared SN and metabolically labeled them with a mixture of 35 S-Met and 35 S-Cys with or without stimulation by 10 mM glutamate and 50 mM NMDA (Fig.2D). SDS-PAGE and autoradiography of the same quantity of proteins from unactivated or activated SN showed a global 1.7 increase of protein synthesis with stimulation (n = 3, p,0.05). Among the metabolically labeled proteins, one had a size corresponding to CaMKIIa, as revealed by Western Blot and its quantity similarly increased by 1.6 upon Glu/NMDA stimulation. This suggests that CaMKIIa might be newly-synthesized in our SN preparation and thus that the synaptically localized CaMKIIa mRNA might be locally translated.
Synaptic local translation requires that functional translation machinery is located close to the synapse. The association of polyribosomes with dendritic spines of granule cells of the dendate gyrus has been clearly documented by electron microscopy [1]. We thus verified by EM that some polyribosomes could also be found associated with GCs spines forming dendrodendritic synapses in the OB. Indeed, we found polyribosomal rosettes at the base of, or in close proximity to, the spine neck (Fig. 2E), similar to those observed in the HC. In addition, we quantified EM sections of GCs dendrites in the EPL for the density of ribosome clusters (ranging in size from 2 to 20) along with the density of synaptic appositions. We found that the density of ribosome clusters correlated with the density of synapses (mean number of ribosomes 6.79+/20.66 per 80 mm 2 versus mean number of synapses 8.76+/20.45 per 80 mm 2 , Spearman's correlation r = 0.55, p,0.0008, n = 33 fields of 80 mm 2 ). This correlation suggests that the presence of a translational machinery in GCs dendrites depends on the presence of synapses, raising the possibility that GCs dendrites are functional for synaptic local translation.
We thus tested the effects of olfactory activity on CaMKIIa dendritic localization in the OB through a simple olfactory enrichment protocol. Groups of 10 mice were first habituated to a clean new environment for one hr. They were then presented for 15, 30 or 60 min with a tea-ball containing a cocktail of unfamiliar odors (garlic and tarragon). Control groups were presented with an empty tea-ball and represent the basal conditions. SN were then prepared and mRNAs quantified, as described above. After 15 min of enrichment, we found no increase in the quantity of CaMKIIa mRNA in SN as compared to basal conditions (Fig. 3A). However, after 30 min of enrichment, CaMKIIa mRNA levels in SN were dramatically increased by 3.4, as compared to basal conditions (Fig. 3A), with an index of synaptic localization rising from 7 in basal conditions to 26 in enriched conditions (n = 3; p,0.02, t-test) (Fig. 3B). After one hr of enrichment, CaMKIIa mRNA levels in SN returned to basal levels (Fig. 3A). Enrichment had no effect on CaMKIIa transcription, as CaMKIIa mRNA levels in total OB extracts remained unchanged (data not shown). These results suggest an activity-dependent re-localization of CaMKIIa mRNA in GCs dendrites, during a restricted timewindow of olfactory enrichment.
We then verified whether this increase in dendritic CaMKIIa mRNA could be visualized by ISH. Indeed, the dendritic staining in the EPL appeared more intense after a 30 min enrichment (Fig. 3C). Quantification of the signal showed that it was increased by 2.3 upon enrichment (Fig.3D, n = 2, p = 0.006).
To test whether the increased dendritic CaMKIIa mRNA could lead to its increased local translation, we quantified CaMKIIa protein in SN after enrichment and saw that there was an increase of 1.5 as compared to basal conditions (n = 4, p,0.05) (Fig. 3E). This could be the consequence of an increased local translation of CaMKIIa mRNA in some GC spines.
NMDA receptors are pivotal in activity-dependent mRNA synaptic localization [23]. Moreover, they are essential to the dendrodendritic inhibition exerted by GCs onto mitral cells [24,25]. To test their role in the activity-dependent dendritic localization of CaMKIIa mRNA in the OB, we injected mice with CPP ((RS)-3-(2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid, 10 mg/kg by intraperitoneal injection), an NMDA-receptor antagonist, 30 min prior to olfactory enrichment. In these conditions, 30 min of olfactory enrichment had no effect on synaptic CaMKIIa mRNA levels that remained at basal level in SN (n = 3; p = 0.6, t-test) (Fig. 4A), contrary to what is seen in untreated animals (Fig.3B). NMDA-receptors blockade thus prevents the activity-induced increase of CaMKIIa mRNA levels in SN. To fully test the role of NMDA-receptors in the dendritic localization of CaMKIIa mRNA in the OB, we treated mice with D-cycloserine ((R)-4-Amino-3-isoxazolidone, 4-Amino-3-isoxazolidinone, 20 mg/kg by intraperitoneal injection), an agonist of NMDA receptors, 30 min before sacrifice. Control mice were injected with saline. In D-cycloserine injected mice, CaMKIIa mRNA levels in SN were increased 3.4 fold as compared to control (n = 3; p = 0.02, t-test) (Fig.4B). Activating NMDAreceptors through the use of D-cycloserine thus recapitulates the increase in CaMKIIa mRNA levels in SN, consecutive to a 30 min olfactory enrichment. Neither CPP nor D-cycloserine injections affected levels of CaMKIIa mRNA in total OB extracts (data not shown). Our data strongly suggests that olfactory activity regulates CaMKIIa mRNA synaptic localization in GCs through NMDA receptors.

Altered CaMKIIa mRNA Dendritic Localization Disrupts Olfactory Associative Learning
To test a role of CaMKIIa local translation in olfactory functions, we took advantage of the knocked-in mice where CaMKIIa 39UTR has been replaced by the 39UTR of an unlocalized mRNA [17]. By ISH, we found strongly reduced dendritic labeling in the HC of the 39UTR mutants as described, and also observed a similar reduction in the OB (Fig. 5A). Quantification of this staining in the EPL showed that it was reduced by 4 in the mutants (WT = 18,161.6; 39UTR = 4.461.3, mean intensity 6 sem, n = 2, p = 0,02). This decreased localization of CaMKIIa mRNA in OB dendrites was confirmed through SN preparation followed by RTqPCR: the index of synaptic localization for CaMKIIa mRNA normalized over HPRT was reduced by 65% in the 39UTR mutants as compared to the WT (39UTR, I = 2.160.6 versus WT, I = 660.18, p = 0.026, n = 3) (Fig.5B), whereas this index was not significantly changed for PSD95 mRNA.
We then tested the olfactory capacity of the mutants. First we used the habituation/dishabituation paradigm to assess olfactory habituation and discrimination between two odorants (Carvone and Pentanol). When presented with the habituation odorant, both control and mutant mice investigation time decreased across the four habituation trials (control: F(3,27) = 30.452, p,0.001; mutant: F(3,27) = 20.469, p,0.001; trial effect) (Fig. 5C). When the test odor was then presented to the mice, the time of odor investigation significantly increased compared to the last habituation trial (Hab4) in both groups of mice (control: p,0.00005; mutant: p,0.00005, t-test) (Fig.5C). These results indicate that the mice detected the odor, habituated to it and were able to discriminate the test odor from the habituation one, regardless of their genotype. Taken together, these data indicate that the mutant mice have normal basic olfactory function.
In order to test if the reduction of CaMKIIa mRNA in dendrites affects olfactory learning capacity, mice were submitted to an olfactory associative learning task. The mice were trained to use an olfactory cue to find a reward hidden in a hole (see Material and Methods). This paradigm was chosen because it is hippocampus-independent [26]. The percentage of correct choice in the control group increased significantly across blocks (one block = 4 trials) (F(5,49) = 3.391, p,0.01; ANOVA, block effect) (Fig. 5D). This indicates that control mice learn to associate the odorant to the reward. Learning was confirmed by the decrease in latency across trials (F(5,49) = 4.037, p,0.005; ANOVA, block effect) (Fig. 5E). In contrast, the success rate of the mutants did not increase with blocks, nor did the latency decrease (success rate: This indicates that the 39UTR mutants could not learn the odor-reward association, and suggests a role for CaMKIIa mRNA dendritic localization and local translation in olfactory associative learning.

CaMKIIa Localization in the OB
Our immunocytochemical localization of CaMKIIa shows that CaMKIIa is exclusively expressed by GCs in the OB. As described before [19], CaMKIIa is not expressed by glutamatergic mitral cells, whereas it is typically associated with glutamatergic synapses in other brain regions. The immunostaining in the EPL suggests its presence in the dendro-dendritic reciprocal synapses between GC dendritic spines and mitral cells, which was indeed previously observed by immuno-electron microscopy [19].
Interestingly, we could also detect CaMKIIa mRNA in the EPL. CaMKIIa mRNA has been among the first mRNAs to be described in dendrites of the HC and cortex [5,6], but the authors did not address localization in the OB. By ISH, we could clearly see CaMKIIa strong expression in cell bodies of the GCL and a lighter staining extending in a gradient into the EPL, where GCs project their apical dendrites. This dendritic labeling appears to be specific, as it disappears in the knocked-in mice, where CaMKIIa 39UTR is ablated [17] (see Fig. 5A).
To confirm and quantify CaMKIIa mRNA synaptic localization, we performed RTqPCR on mRNAs from OB SN. The index of synaptic localization was indeed 7 times higher than HPRT mRNA, an unlocalized mRNA. This index was similar to the one we quantified in the HC (not shown) and to previous studies [12,27], thus validating our SN preparation for the analysis of CaMKIIa mRNA localization. Of importance, the synaptically localized CaMKIIa mRNA in the OB might be translated, since we could metabolically label newly-synthesized CaMKIIa protein in SN (see Fig. 2D). Accordingly, our EM observation showed that polyribosome could be found in GCs dendritic shaft in close proximity to spines forming dendrodendritic synapse and that the number of polyribosomes in GCs dendrites is correlated to the number of synaptic appositions. This is consistent with previous observations made in other systems [1] and suggests for the first time that GCs synapses are functional for local translation.
Activity-dependent Transport and Translation of CaMKIIa mRNA in the OB Here, we show that synaptic CaMKIIa mRNA is dramatically increased by a 30 min exposure to new odors. In cultures of hippocampal neurons, depolarization increases the motility of CaMKIIa mRNA containing granules [28,29,30]. In the HC of freely moving rats, LTP induction triggers delivery of CaMKIIa mRNA in dendrites [11] and synaptic sites [12]. Contrary to what we observe in GC dendrites, this transport appeared to be NMDA independent [12]. This might reflect a difference in the regulation of CaMKIIa mRNA transport between OB and HC but also a difference in the timing of CPP administration (30 min before enrichment in our study versus 2 hrs in the HC study). In our experiment, NMDA proves to be necessary and sufficient to support the increase of synaptic CaMKIIa mRNA localization induced by olfactory enrichment. Interestingly, glutamate acting through NMDA-receptors is essential to the proper functioning of dendrodendritic synapses, as the release of GABA from GCs spines depends on NMDAR activation [24,25]. CaMKIIa mRNA transport and local translation might play a role in this process, as CaMKII can regulate neurotransmitter release at presynaptic sites [5] and as NMDAR can be phosphorylated by CaMKII [31].   The increased transport of CaMKIIa mRNA upon olfactory enrichment is accompanied by a 1.5 fold-increase in synaptic CaMKIIa protein. This suggests an increase in CaMKIIa mRNA local translation upon olfactory enrichment. This increase is reminiscent of what was observed in the visual cortex of darkreared rats exposed to light for 30 min [13]. Interestingly, the response in visual cortex was also NMDAR dependent, since it was inhibited by CPP [15]. An NMDA-dependence of CaMKIIa mRNA local translation has also been observed ex vivo in SN [8,9].

CaMKIIa mRNA Local Translation is Necessary for Olfactory Associative Learning
We analyzed the function of CaMKIIa mRNA local translation in the olfactory pathway using mutant mice expressing a form of CaMKIIa mRNA without its 39UTR [17]. These mice were originally created to analyze the functional role of CaMKIIa mRNA local translation in the HC and showed its importance in synaptic and behavioral plasticity. The mice displayed a reduction in late-phase LTP and impairments in spatial memory, associative fear conditioning and object recognition memory. The authors reported a massive reduction of CaMKIIa mRNA in dendrites and SN from the HC, which we also observe in the OB. In this initial study, an unexpected 50% reduction of CaMKIIa protein in whole hippocampal extracts was also observed in the mutants. We observe a similar reduction in the OB (not shown). One can thus not formally exclude that the phenotype we observe could be partly due to this reduction.
In a subsequent article concerning the same mice [32], quantitative proteomics of HC synapses showed that the synaptic protein constituents were not substantially altered in the mutant mice, apart from CaMKIIa itself. Of importance, in the OB, we see that the GCL size is unaltered in the mutants, suggesting that GCs survival is normal. Moreover, in an olfactory habituation/ dishabituation test (Fig. 5C) [33], the mutant mice displayed normal basic olfactory function in terms of detection, habituation and discriminations of odors. Together, these data suggest that the reduction of CaMKIIa mRNA local translation does not trigger major compensatory changes that could affect the OB functioning. Nevertheless, we report here that, in an olfactory associative learning paradigm [33,34,35], these mice were incapable of learning. Noticeably, the HC is not required for acquisition of this type of olfactory non-spatial associative task [26,36].
The importance of CaMKII mRNA local translation during associative olfactory learning was previously noted in Drosophila [16]. Indeed, using fluorescent reporters of translation, the induction of synaptic CaMKII synthesis was observed in several brain centers following a training paradigm of repetitive odor, paired with electric shock. Remarkably, this synaptic induction was odor specific. This drew a strong correlation between CaMKII local translation and the synaptic changes underlying olfactory learning and memory in Drosophila.
Our results confirm and extend these data to the mammalian olfactory pathway. The 39UTR mutant mice could not learn to associate an odor with a reward, which indicates that CaMKIIa local translation is essential to the synaptic modifications associated with olfactory associative learning. This defective learning might be the consequence of disrupted CaMKIIa local translation in multiple regions of the olfactory pathway. Among them, the OB is likely to play an essential role since i) we show an activity-regulated CaMKIIa mRNA transport and translation in the OB and ii) olfactory learning-induced changes can occur within the OB itself. For instance, the oscillatory response of the whole OB network evolves with learning, in close correlation with behavioral performances [37]. In addition, the response of MCs to an odor depends on the odor association with a positive or negative reinforcement [38]. Most importantly, the synchronized firing of MCs during olfactory associative learning conveys information on the odor association to a reward, so that information on stimulus reward is encoded within the OB itself [39]. The neural representation of the odors in the OB is thus highly modulated by olfactory learning [18] and CaMKIIa local translation might play a particularly meaningful role in the synaptic modifications underlying these changes.
Finally, GCs undergo an extreme form of plasticity, since they are constantly renewed throughout life. We have recently shown that the Fragile X Mental Retardation Protein (FMRP) is a master regulator of neo-GCs morphogenesis [40]. Interestingly, CaM-KIIa mRNA is one of FMRP's mRNA targets [41]. In addition to regulating synaptic strength, CaMKIIa regulates structural plasticity by controlling spine size and density [42,43,44], activity-dependent filopodia growth and spine formation [45] and by stabilizing dendritic arbor structure [14]. It is thus tempting to speculate that CaMKIIa local translation might also play a fundamental role in new GCs integration.

Animals
C57Bl6, CD1 or 39UTR mutants and control littermates (in a C57bl6 genetic background) 2-month old males were used in all experiments. This study was carried out in strict accordance with the recommendations of the CNRS «Formation à l'Expérimentation Animale ». The protocol was approved by the Comité Régional d'Ethique en Expérimentation Animale Nu3 of the région Ile de France (File number p3/2008/047).
For quantification, photomicrographs were taken at a magnification of 40X and the intensity of labeling was quantified with and 39UTR mutants. CaMKIIa mRNA index is significantly decreased in 39UTR mutants as compared to WT (n = 3, p = 0.026) C, Behavioral habituation/ dishabituation paradigm: both WT and 39UTR mutants mice habituated to the habituation odor, as shown by decreased investigation time across the four trials (Hab1-Hab4) (WT: F(3,27) = 30.452, p,0.001; 39UTR: F(3,27) = 20.469, p,0.001; ANOVA, trial effect) and discriminated. Upon presentation of the test odor (Test), the investigation time increased when compared to Hab4, indicating that the animals could discriminate the test odor from the habituation odor (WT: *p,0.00005; 39UTR: *p,0.00005, t-test). D, Olfactory associative learning: mice were conditioned to associate an odor stimulus to a food reward. Success rate results are presented as the percentage of correct choice across consecutive blocks of trials (1 block = 4 trials). Conditioned WT mice learned the task as indicated by the increase in correct choice throughout blocks (F(5,49) = 3.391, p,0.01; ANOVA, block effect). In contrast, conditioned 39UTR mutants did not show an increase in success rate across trials (F(5,54) = 1.078, p.0.1; ANOVA, block effect), showing that impaired CaMKIIa mRNA dendritic localization disrupts olfactory-associative learning. E, Latencies: conditioned WT mice showed a decrease in latency confirming that they learned the task (F(5,49) = 4.037, p,0.005; ANOVA, block effect). In contrast, latencies of conditioned 39UTR mutants did not significantly change (F(5,54) = 0.635, p.0.5; ANOVA, block effect), indicating that the mutants were not able to associate odor and reward. doi:10.1371/journal.pone.0040133.g005 ImageJ on defined surfaces of the EPL and normalized to unstained regions of the periglomerular area.

Synaptosome Preparation for RNA Analysis
Synaptosomes were isolated as described previously [27] with some modifications. Briefly, the OBs from 10 adult mice per condition were homogenized in 9 ml of 320 mM sucrose, 1 mM EDTA, supplemented with protease (Complete tablets, Roche) and RNase inhibitors (20 U/ml RNasin; Promega) at pH7.4. The homogenate was centrifuged at 1000 g for 10 min at 4uC, and the resulting supernatant (S1) was subjected to a discontinuous Percoll-sucrose gradients (3,10,15, and 23% Percoll from GE Healthcare, prepared in 320 mM sucrose, 1 mM EDTA supplemented with 0.25 mM DTT), centrifuged at 32000 g for 10 min at 4uC in a SW41 rotor. The material located at the 10-15% interface was collected, washed for 8 min at 4uC at 12000 g in 320 mM sucrose, 5 mM Hepes pH7.4. The resulting pellet was resuspended in an Optiprep Working Solution (OWS) (50% Optiprep [Sigma], 65 mM sucrose, 10 mM HEPES) and separated on an Optiprep-sucrose gradient (9,12.5,15,25,and 35% Optiprep). Optiprep solutions were prepared by diluting the corresponding volume of OWS in 300 mM sucrose, 10 mM HEPES, pH 7.4. Finally, the material located at interface 15-25% was collected and used for RNA isolation. RNA was purified using the PureLink RNA Mini Kit (Invitrogen) according to the manufacturer's protocol. 50 ng of RNA were retro-transcribed using the Quantitect kit (Qiagen) according to the manufacturer's protocol. cDNAs were subjected to quantitative PCR on a LightCycler apparatus using the LC480 kit (Roche) according to the manufacturer's protocol. Each sample was made in triplicate and controls without retrotranscription were routinely added to the PCR. In each experiment, an internal standard was added and quantities of cDNAs were calculated taking the efficiency of each primer into account.
Results are presented as levels of mRNA in SN over quantities in S1, representing the synaptic localization index (I). Results were normalized to HPRT, a transcript restricted to the soma of neurons, with between 0,1 and 5% of total mRNA being found in SN preparation.

Synaptosome Preparation for Metabolic Labeling and Immunoblotting
The previous protocol of SN preparation leads to collection of SN in an Optiprep solution. This made the pelleting of SN difficult, especially since the quantity of material was very low after the 2 gradients. For protein analysis, we thus used an alternate protocol derived from [46]. The SN obtained with both protocols were comparable, since the measured CaMKIIa mRNA synaptic localization was similar and similarly increased by olfactory enrichment.
Briefly, the OBs from 10 adult mice per condition were homogenized in 7 ml of Homogeneization Buffer (TpH) consisting in 320 mM sucrose, 4 mM Hepes pH 7.4, supplemented with protease and RNase inhibitors. The homogenate was centrifuged at 1000 g for 6 min at 4uC. The resulting supernatant (S1) was centrifuged for 10 min at 12500 g at 4uC. The pellet was resuspended in 1 ml TpH, layered on top of a discontinuous sucrose gradient (0.8 and 1.2 M) and centrifuged for 1 h10 min at 16000 rpm at 4uC in a SW41 rotor. The synaptosome rich 1.2M/ 0.8M interface was collected, pelleted and resuspended in 100 ml Synaptosome Buffer (SnB) containing 10 mM Tris pH 7.5, 2.2 mM CaCl 2 , 0.5 mM Na 2 HPO 4 , 0.4 mM KH 2 PO 4, 4 mM NaHCO 3 and 80 mM NaCl, supplemented with 100 mg/ml chloramphenicol and protease inhibitor. After preincubation at 37uC for 10 min, SN were incubated for 45 min at 37uC with 50 mCi of EasyTag Express Protein Labeling Mix (Perkin Elmer) with or without 10 mM glutamate and 50 mM NMDA. Samples were pelleted, and subjected to SDS-PAGE. Gels were dried and radioactive proteins were detected with a Biorad Personal Molecular Imager FX System.

Olfactory Habituation/Dishabituation Task
To assess basic olfactory function, mice were submitted to a habituation/dishabituation task in a clean standard home cage. A test session consisted of one 50sec presentation of plain mineral oil (MO) and then four 50 sec odor presentations of the habituation odor (inter trial interval = 5 min) (Hab 1-4) followed by one 50 s presentation of test odor (Test). Investigation time was measured as active sniffing within 1 cm of the odorant stimuli. A decrease in odor investigation across the habituation trials indicated that the animals detected the odor and remembered it on a short-term basis, from one trial to another. An increase in investigation time upon the presentation of the test odor indicated that it was discriminated from the habituation odor [47].
Odorants used were Pentanol and Carvone diluted to 1 Pa in mineral oil. Each odorant of the pair was randomly used as habituation odor or test odor. Odorants were presented by placing 100 mL of the odor stimulus onto a polypropylene swab in a tea ball hanging from the ceiling of the cage. Data analysis. Investigation time was averaged within groups for each trial. The data were analysed by one-way ANOVA repeated measures across habituation trials to assess habituation and by t-tests between the last habituation trial (Hab4) and the test trial to assess discrimination using Systat statistical software. Results are presented as group mean ± sem. The level of significance was set to 0.05.

Associative Learning
To assess the learning abilities of the animals, mice were submitted to an olfactory conditioning.
Experimental setup. The mice were tested on a computerassisted 2-hole board apparatus (40640 cm) as previously described [33]. The trial started by placing the mouse on the board facing the holes (3 cm diameter, 4.5 cm deep). Latency and sequence of holes visited were measured by a specific homemade software. Holes contained a polypropylene swap impregnated with 20 mL of +limonene (purity 97%, Sigma-Aldrich, Saint-Louis, MO, USA) or mineral oil. Prior to the olfactory learning experiments, the mice were deprived of food (220% daily consummation, leading to a 10% reduction in body weigh) starting 5 days before shaping.
Shaping. During the shaping (12 trials with 15 min between each trial) the mice were trained to retrieve a reward (a small piece of sweetened cereal, Kellogs, Battle Creek, MI) by digging through the bedding with no odor. The reward was first placed on the top of the bedding and after several retrievals, the reward was buried deeply. The shaping was considered to be complete when a mouse successfully retrieved a reward that was deeply in the bedding with a score of 80% correct choices (8 to 12 trials).
Conditioning. Conditioning consisted of 24 trials of 2 min with 15 min inter-trial interval [34]. For half of the mice, the reward is systematically associated with +Limonene. The position of the reinforced hole was randomized to avoid spatial learning. The other hole contained no odor or reward.
Data analysis. A successful trial was recorded when the mouse first visited the odorized hole (nose poking). For each behavioural session, mean success rates was calculated on blocks of 4 trials and averaged between groups. Mean latencies (time to find the reward) were calculated for each trial and averaged within groups. Results are presented as group mean ± sem and analyzed using ANOVA (Systat).

Electron Microscopy
Tissue prepared for electron microscopy followed previously described procedures [48]. Briefly, mice were perfused with 0.1 M PBS containing 1 unit/mL heparin, followed by cold fixative containing 4% PFA and 2% glutaraldehyde in 0.1 M PBS. Following dissection brains were postfixed for 4 h at 4uC prior to sectioning coronally on a vibratome (50 mm). Sections were incubated in 2% osmium tetroxide for 1 h, then dehydrated and stained with 1% uranyl acetate in 70% ethyl alcohol for 1 h. The tissue was embedded in EPON and thin 70-100 nm, sections were cut to include the external plexiform layer (EPL). Ribbons of thin sections were examined on a JEOL 1200 transmission electron microscope and photographed at a primary magnification of 10,000X.
To quantify the frequency of ribosome clusters and dendrites, we used 80 mm 2 fields from electron micrographs taken at a primary magnification of 10,000X. Within the field, we counted all of the ribosome clusters (a minimum of 2 ribosomes was required) that occurred within GCs dendrites (recognized by their relatively electron dense appearance and irregular contours). We then counted all of the synapses within the same field, including both asymmetrical (mitral cell dendrite to granule spine) and symmetrical (granule cell dendritic spine to mitral cell dendrite).