Loss of α1,6-Fucosyltransferase Decreases Hippocampal Long Term Potentiation

Background: High expression levels of core fucosylated N-glycans in brain tissues remain unexplained. Results: Loss of core fucosylation enhanced AMPA receptor heteromerization and decreased long term potentiation. Conclusion: Core fucosylation is required for hippocampal long term potentiation. Significance: Core fucosylation may be very important for the neuronal synaptic plasticity that is required for learning and memory. Core fucosylation is catalyzed by α1,6-fucosyltransferase (FUT8), which transfers a fucose residue to the innermost GlcNAc residue via α1,6-linkage on N-glycans in mammals. We previously reported that Fut8-knock-out (Fut8−/−) mice showed a schizophrenia-like phenotype and a decrease in working memory. To understand the underlying molecular mechanism, we analyzed early form long term potentiation (E-LTP), which is closely related to learning and memory in the hippocampus. The scale of E-LTP induced by high frequency stimulation was significantly decreased in Fut8−/− mice. Tetraethylammonium-induced LTP showed no significant differences, suggesting that the decline in E-LTP was caused by postsynaptic events. Unexpectedly, the phosphorylation levels of calcium/calmodulin-dependent protein kinase II (CaMKII), an important mediator of learning and memory in postsynapses, were greatly increased in Fut8−/− mice. The expression levels of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptors (AMPARs) in the postsynaptic density were enhanced in Fut8−/− mice, although there were no significant differences in the total expression levels, implicating that AMPARs without core fucosylation might exist in an active state. The activation of AMPARs was further confirmed by Fura-2 calcium imaging using primary cultured neurons. Taken together, loss of core fucosylation on AMPARs enhanced their heteromerization, which increase sensitivity for postsynaptic depolarization and persistently activate N-methyl-d-aspartate receptors as well as Ca2+ influx and CaMKII and then impair LTP.

Schizophrenia is a common, chronic, and severe brain disorder that ranks as one of the leading causes of disability worldwide because it afflicts 2% of the world's population (1,2). The disease is typically characterized by a loss of emotional expression and a lack of motivation. The etiology of schizophrenia is only partially understood, but multiepisode patients show significantly disturbed neuronal plasticity, suggesting that synaptic activity and connectivity are altered during the progression of the disease. Recently, cognitive impairments, such as deficits in learning and memory, have also been shown to be a fundamental feature of the disorder (3). Long term potentiation (LTP), 2 a well established model based on the neurophysiological study of learning and memory, has been found to be an important mechanism that underlies synaptic changes and plasticity in schizophrenia (4,5). The induction of LTP begins with the activation of one kind of ionotropic glutamate receptor,␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptors (AMPARs), which induce the opening of N-methyl-D-aspartate receptors (NMDARs), which are another kind of glutamate receptor, and leads to calcium entry that initiates a biochemical cascade through the activation of CaMKII, thereby increasing the number of AMPARs in the postsynaptic density (PSD) area. The end product of this biological reaction is the long lasting potentiation of AMPAR-mediated excitatory postsynaptic current (6,7).
Previous research has focused on identifying genes or protein expression-level changes in schizophrenia and ascertaining the functional effects of those alterations. Recently, the post-translational modification, N-glycosylation, has become a new target of investigation for schizophrenia (8). For example, abnormal N-glycans on AMPAR, NMDAR, and GABA A receptors were found in patients with schizophrenia (9 -11).
The importance of fucosylation has been highlighted by the identification of the monogenetic inherited human disease, congenital disorder of glycosylation IIc (12). Due to defective Golgi GDP-fucose transporter activity, patients present with important clinical symptoms that include mental and growth retardation and severe immunodeficiency (13). Interestingly, the phenotypes of the transporter deletion mice (14) are similar to ␣1,6-fucosyltransferase knock-out (Fut8 Ϫ/Ϫ ) mice (15), suggesting that core fucosylation is a key player among all fucosylations in mammals. Previously, we reported that the Fut8 Ϫ/Ϫ mice exhibited multiple behavioral abnormalities, including decreased prepulse inhibition, increased locomotion, and impaired working memory with a schizophrenia-like phenotype (16). In fact, the unique structure of the core fucosylated N-glycans is highly expressed in brain tissues, and the expression patterns of N-glycans are altered during brain development (17).
To evaluate the effect of core fucosylation on LTP, we investigated synaptic strength and plasticity in the CA1 subregion of the hippocampus of Fut8 Ϫ/Ϫ mice. We found that high frequency stimulation (HFS)-induced LTP was decreased in Fut8 Ϫ/Ϫ mice. The molecular mechanism could have been caused by the aberrant heteromerization of AMPAR constitutively activating intracellular signaling, which may disturb LTP in Fut8 Ϫ/Ϫ mice.

Experimental Procedures
Mice-The Fut8 Ϫ/Ϫ mice were generated as described previously (16). All experiments were conducted with female and 3-5-week-old Fut8 Ϫ/Ϫ and wild-type (Fut8 ϩ/ϩ ) mice. The mice were housed in a temperature-controlled room with a 12-h dark/12-h light cycle and ad libitum feeding. This study was approved by the Institutional Animal Care and Use Committee of Tohoku Pharmaceutical University, Japan.
Extracellular Recordings and Induction of LTP-Field excitatory postsynaptic potentials were recorded in the stratum radiatum of the CA1 region using glass microelectrodes (tip diameters, 10 m), which were filled with 1.78% Na 2 SO 4 solution. A tungsten bipolar electrode was used to stimulate the Schaffer collateral axons and twin pulses at 200-s duration and 50-ms intervals applied at a frequency of 0.033 Hz, with an intensity of less than 100 A. The field excitatory postsynaptic potential was amplified and digitized using a conventional electrophysiological system (Axon Axopatch 200B plus Digidata1200B, Molecular Devices) and analyzed using Clampex version 10.2 software (Molecular Devices). A baseline recording was performed for at least 1 h. Then E-LTP was induced by applying a single 100-Hz train for 1 s at test strength. Voltage-dependent Ca 2ϩ channel-dependent LTP was induced by applying 25 mM TEA for 10 min (18).
Ca 2ϩ Imaging-Primary neural cells were cultured as described previously (19). After culturing for 14 days on glassbottom dishes (Greiner Bio-one), the cultured media were replaced by a fresh medium containing 2 M Ca 2ϩ indicator Fura-2/AM (Dojindo, Japan) at 37°C for 1 h to load and sensitize the probe by cleaving its AM group with cytosolic esterase.  (20). Then cells were exposed to either 1 or 3 M glutamate plus 10 M glycine containing a balanced salt solution flow (1.75 ml/min, 34°C) for 1 min, respectively, followed by a 3-min balanced salt solution flow wash. The AMPAR-mediated Ca 2ϩ influx was blocked by an AMPA/kainate receptor antagonist, 6-cyano-7-nitroquinoxaline-2,3-dione or D-(Ϫ)-2-amino-5-phosphonopentanoic acid, a competitive NMDA receptor antagonist. The images were acquired at emission wavelengths between 490 and 520 nm while being excited sequentially at 340 (red) and 380 (green) nm (exposure time, 0.1-2 s). The ratio of F 340 /F 380 , indicating an influx efficiency of Ca 2ϩ , was analyzed by using an image processor (Aqua Q, Hamamatsu, Japan) connected to a cooled CCD camera (Orca, Hamamatsu, Japan) (21).
PSD Fractions-All biochemical experiments were carried out either on ice or at 4°C (22). Dissected sections of mouse brains were homogenized with 10 volumes of buffer A (0.32 M sucrose, 10 mM Tris-HCl, 1 mM EDTA, and 1% protease inhibitor mixture) with 15 up and down strokes at 1,000 rpm. The homogenate was centrifuged at 1,000 ϫ g for 10 min to remove the nuclear fraction and large debris. The supernatant (S1, postnuclear fraction) was centrifuged at 10,000 ϫ g for 20 min to yield a crude synaptosome pellet (P2) and the supernatant (S2). The S2 was centrifuged at 105,000 ϫ g for 60 min to yield a light membrane pellet (P3). The P2 fraction was lysed with hypo-osmotic (10% buffer A ϩ 90% H 2 O) buffer and centrifuged at 25,000 ϫ g for 30 min to yield the synaptosomal membrane fraction (LP1). The LP1 fraction was then suspended in 0.5% Triton X-100 in buffer A for 15 min and centrifuged at 105,000 ϫ g for 1 h to yield the PSD fraction. All pellet fractions were dissolved in 0.5% SDS. The protein concentration of each fraction was adjusted to 2 g/l using a Pierce BCA protein assay kit (Thermo).
Immunoblot Analysis and Immunoprecipitation-Cells cultured under different conditions were washed with PBS and lysed with lysis buffer that contained 1% Triton X-100, 10 mM Tris-HCl, 150 mM NaCl, and 1% protease inhibitor mixture. Cell lysates were separated in SDS-PAGE gels under reducing conditions and were then transferred to PVDF membranes. For Western blot, the membranes were blocked in 5% dried skimmed milk for 1 h at room temperature and probed with specific primary antibodies followed by incubation with appropriate secondary antibodies that had been conjugated with HRP. Finally, specific proteins were visualized using an ECL system (Amersham Biosciences). These membranes were stripped and reprobed with an antibody against the corresponding total proteins to confirm equal loading. For lectin blot, the membranes were blocked in 3% BSA and then detected with A. aurantia lectin. The immunoreactive bands were visualized using a Vectastain ABC kit (Vector Laboratories).
For the immunoprecipitation, all of the experiments were carried out on ice or at 4°C; 2 g of anti-GluA2 antibodies were attached to 5 l of magnetic protein A beads in 100 l of PBS plus 0.1% Triton X-100. Protein A beads were pelleted and incubated with the membrane lysates or cell lysates for 40 min with constant rotation. The beads were then washed three times with lysis buffer, and the immunoprecipitate was dissolved in 30 l of SDS-PAGE sample solution.
Statistical Analysis-All of the electrophysiological data in this study are expressed as means Ϯ S.E., and others are expressed as means Ϯ S.D. Either a two-way analysis of variance or an unpaired Student's t test was used to analyze the statistical significance of these results.

The Basal Synaptic Transmissions Were Increased, but HFSinduced LTP Was Decreased in Fut8
Ϫ/Ϫ Mice-Previously, we reported that the Fut8 Ϫ/Ϫ mice exhibited a schizophrenia-like phenotype (16). Synaptic activity and connectivity played important roles during the progression of the schizophrenia (3). Therefore, we analyzed basal synaptic neurotransmission. First, we recorded the field excitatory postsynaptic potential (fEPSP) in the hippocampus CA1 area (Fig. 1A) and found that the responses to stimuli were more sensitive in Fut8 Ϫ/Ϫ mice than in Fut8 ϩ/ϩ mice, which was confirmed by the input-output curve (Fig. 1B). However, the paired pulse ratio, which indicated presynaptic facilitation, showed no significant difference between Fut8 ϩ/ϩ and Fut8 Ϫ/Ϫ mice (Fig. 1C). On the other hand, based on our previous results from the Y-maze task, we knew that learning and memory could be impaired in Fut8 Ϫ/Ϫ mice (16). We further examined the formation of E-LTP, which is generally used as a primary experimental model of memory formation in neuronal circuits. E-LTP was triggered by the application of HFS, and we found that the scale of E-LTP was significantly decreased in Fut8 Ϫ/Ϫ mice, compared with Fut8 ϩ/ϩ mice (Fig. 1D), but the TEA-induced (voltage-gated Ca 2ϩ channel-dependent) LTP showed no significant difference (Fig. 1E), suggesting that the decline in the E-LTP in Fut8 Ϫ/Ϫ mice was mainly caused by the events in postsynaptic neurons rather than in the presynaptic regions. Above all, Fut8 Ϫ/Ϫ mice showed hyperactivity in basal synaptic transmissions and a decrease in E-LTP in postsynaptic neurons.
Persistent Activation of CaMKII and Increased AMPARs in the PSD Area of Fut8 Ϫ/Ϫ Mice-To examine the events in the postsynaptic regions, we analyzed the activation levels of ionotropic glutamate receptors and CaMKII, which are mainly mediated in the induction and maintenance of E-LTP. We found that the CaMKII, a necessary component of the cellular machinery underlying learning and memory, was greatly activated in Fut8 Ϫ/Ϫ mice (Fig. 2, A and B). Phosphorylation at Thr-286 of CaMKII plays essential roles in LTP, whereas LTP is attenuated in T286A (Thr-286 non-autophosphorylatable) mutant mice (23,24). However, the expression of a constitutively active form of CaMKII into CA1 neurons enhances postsynaptic transmission and prevents further potentiation via synaptic stimulation (25,26) because the CaMKII and LTP enhance synaptic transmission by the same mechanism (27). Thus, to be consistent with those observations, we thought that the persistent activation of CaMKII might lead to a decrease in E-LTP in Fut8 Ϫ/Ϫ mice.
Ionotropic glutamate receptors, which can be subdivided into NMDAR, AMPAR, and kainate receptors, mediate most excitatory neuronal transmission (28). Among these receptors, NMDAR channels have several unique features, including a voltage-sensitive block by extracellular Mg 2ϩ and a high permeability to Ca 2ϩ (29). The Mg 2ϩ block acts as a molecular   JULY 10, 2015 • VOLUME 290 • NUMBER 28 switch, with the removal of Mg 2ϩ from the pore of the channel when postsynaptic cells are depolarized. The relief of the block leads to the Ca 2ϩ influx through the NMDAR channel that regulates synaptic strength through Ca 2ϩ -activated signaling cascades. The phosphorylation of NMDAR is known to alter the channel properties (30). Consistent with the activated CaMKII in Fut8 Ϫ/Ϫ mice, in the present study, the expression levels of phosphorylated GluN2A at Tyr-1246 and GluN2B at Tyr-1472 were also increased (Fig. 2, A and B). The rise of intracellular Ca 2ϩ level causes reversible inactivation through the phosphorylation of receptors to prevent more synaptic stimulation (31). For example, phosphorylation of Tyr-1472, the major phosphorylation site by Fyn on GluN2B (32), disrupts its binding to an AP-2 adaptor, thereby resulting in the inhibition of GluN2B-mediated endocytosis.

FUT8 Plays Important Roles in Neurons
When postsynaptic cells receive stimulation, the activated CaMKII phosphorylates TARPs binding to PSD95, thereby regulating AMPAR numbers in the PSD area (6,33). However, with the decay of LTP, the AMPARs leave the PSD area and prepare for the next stimulation (7). The trafficking of AMPARs to regulate the number of receptors at the synapse plays a key role in various forms of synaptic plasticity. Interestingly, the expression levels of GluA1, GluA2, and GluA3, but not GluN2A and CaMKII, in the PSD area of brain tissues were increased in Fut8 Ϫ/Ϫ mice (Fig. 2, C and D). Furthermore, Stargazin (␥2), the main member of the TARP family, accumulated in the PSD area and acted as auxiliary subunits of AMPARs (Fig. 2, C and  D). The distribution of TARPs has been reported (33), ␥2 is abundant throughout the brain, and ␥8 is mainly expressed in the hippocampus. Unfortunately, we failed to detect subunits of ␥8 using commercially available antibody.
The Enhanced Heteromerization of AMPARs in Fut8 Ϫ/Ϫ Mice-To investigate the underlying mechanism of the persistent activation of CaMKII in Fut8 Ϫ/Ϫ mice, we analyzed the heteromerization of AMPARs. Among the ionotropic glutamate receptors that mediate fast excitatory synaptic transmission, most of the biochemical cascades were initiated by the AMPAR-mediated depolarization of postsynaptic cells. Because AMPARs are the primary requirement in the expression of LTP, the modulation of AMPARs could affect synaptic transmission. In fact, several studies have reported that the biological functions of AMPARs are affected by post-translational modification (34). For example, human natural killer (HNK1) glycoepitope regulates the stability of GluA2 on the neuronal cell surface and dendritic spine maturation (22,35). We found that there were no significant differences in either the total or the membrane expression levels of NMDARs and AMPARs in the hippocampi of the three genotypes of the mice (Fut8 ϩ/ϩ , Fut8 ϩ/Ϫ , and Fut8 Ϫ/Ϫ ) (Figs. 2A, 3A, and 3C), whereas the core fucosylation on these receptors was completely blocked in Fut8 Ϫ/Ϫ mice (Fig. 3B). Then the heteromerization of AMPARs was analyzed by using a pull-down assay. The association levels of GluA1/2 and GluA2/3 were greatly increased in Fut8 Ϫ/Ϫ mice compared with that in Fut8 ϩ/ϩ mice (Fig. 3, C-E).
Although it is well known that there are very few GluA1/3 complex formations in the hippocampus (36), the loss of core fucosylation could have altered the AMPAR complex formation. Thus, we checked the GluA1/3 complex formation as well as FIGURE 3. Enhanced AMPAR heteromerization in Fut8 ؊/؊ mice. A, expression levels of GluN1 and GluAs between Fut8 ϩ/ϩ , Fut8 ϩ/Ϫ , and Fut8 Ϫ/Ϫ mice in hippocampus lysate. ␣-Tubulin was used as a loading control. B, comparison of core fucosylation levels on GluNs and GluAs between Fut8 ϩ/ϩ and Fut8 Ϫ/Ϫ mice. The same amounts of hippocampus lysates were immunoprecipitated (IP) with P. squarrosa lectin-agarose and blotted with anti-GluNs and anti-GluAs. C, membrane proteins were isolated as described under "Experimental Procedures." The same amounts of hippocampus membrane proteins were immunoprecipitated with anti-GluA2 and blotted with anti-GluA1, -GluA2, and -GluA3 (bottom panels). The total expression levels of GluA1, GluA2, and GluA3 and synapsin in hippocampus membranes were used as a loading control (top panels). D, -fold heteromerization levels of GluA1 and GluA2 (right) or GluA2 and GluA3 (left) were compared with those in Fut8 ϩ/ϩ mice, which were set as 1. The quantitative data were obtained from three independent experiments. Data represent the mean Ϯ S.D. (error bars); *, p Ͻ 0.05, unpaired Student's t test. E, the same amounts of hippocampus membrane proteins were immunoprecipitated with anti-GluA1 (top panels) or anti-GluA3 (bottom panels) antibodies and then blotted with anti-GluA2, anti-GluN1, and anti-GluA3 or anti-GluA1 antibodies, respectively.

FUT8 Plays Important Roles in Neurons
the interactions with GluA1. As shown in Fig. 3E, the levels of GluA1/3 complex formations could not be detected in the immunoprecipitates of either GluA1 (top panels) or GluA3 (bottom panels). Likewise, GluN1 expression was not detectable in either the GluA3 or GluA1 complexes. These data suggest that a loss of core fucosylation on AMPARs enhanced the heteromerization of their subunits, which might have subsequently depolarized the neuronal cell membrane to induce Ca 2ϩ influx and CaMKII phosphorylation. Based on the observations in Fig. 3, C and E, we speculated that those heteromeric formations were mainly increased in the PSDs, as shown in Fig.  2C. Although the expression levels of homomeric receptors of membrane fractions in the KO mice were theoretically less than those in Fut8 ϩ/ϩ or Fut8 ϩ/Ϫ mice, it is quite difficult to completely exclude the possibility that those homomeric receptors were also increased in Fut8 Ϫ/Ϫ mice. In fact, it has been reported that homomeric GluA1 is first recruited into PSD following glutamate stimulation (37).
To further confirm the effects of core fucosylation on the complex formation for AMPARs, we expressed recombinant mouse GluA1 and GluA2 heteromeric combinations in wildtype and Fut8-knock-out (Fut8-KO) 293T cells, which were generated using the CompoZr Knock-out Zinc Finger Nucleases (ZFN) kit (Sigma), which was confirmed by an A. aurantia lectin blot (Fig. 4A) and RT-PCR (Fig. 4B). The coimmunoprecipitation experiments showed that the association levels of GluA1 and GluA2 were much higher in Fut8-KO cells compared with that in the wild-type 293T cells (Fig. 4, C  and D). Taken together, these results strongly suggested that the loss of core fucosylation on AMPARs enhanced their subunit complex formation in vitro.

AMPAR-mediated Ca 2ϩ Influx Was Enhanced in Primary Neurons Obtained from Fut8
Ϫ/Ϫ Mice-Furthermore, to clarify the functions of the core fucosylation of AMPARs with respect to potentials, we examined the calcium influx in the 293T cells transfected with GluA1 alone compared with the cells co-transfected with GluA1 and GluA2. Unfortunately, when using the Fura-2 system, the induction of calcium influx upon glutamate stimulation was too small to measure in either the wild-type or the Fut8-KO 293T cells. Considering the possibility that an expression of NMDAR is required for the induction, we used primary cultured neurons as a model system to record the Ca 2ϩ influx to examine the AMPAR and NMDAR potentials. Compared with primary neurons from Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ mice, the Ca 2ϩ influxes were significantly increased in the neurons from Fut8 Ϫ/Ϫ mice (Fig. 5, A and B). To check whether the increased Ca 2ϩ influx in Fut8 Ϫ/Ϫ was mediated by an AMPAR-NMDAR sequential response (38) (i.e. to determine whether AMPAR-induced postsynaptic depolarization can activate the NMDAR and then induce Ca 2ϩ influx), both antagonists were added to the balanced salt solution bath flow. As shown in Fig.  5C, both 6-cyano-7-nitroquinoxaline-2,3-dione, an AMPAR blocker, and D-(Ϫ)-2-amino-5-phosphonopentanoic acid, a NMDAR inhibitor, could eliminate the difference in Ca 2ϩ influx between Fut8 ϩ/ϩ and Fut8 Ϫ/Ϫ mice, further suggesting that the loss of core fucosylation enhances the heteromerization of AMPARs and their downstream signaling in primary cultured neurons.

Discussion
In the present study, we investigated the role of core fucosylation on E-LTP in the hippocampus and found that the HFSinduced LTP was dramatically decreased in Fut8 Ϫ/Ϫ mice. A molecular mechanism could be postulated whereby the lack of core fucosylation in AMPARs results in aberrant heteromerization, which persistently activates cellular signaling, such as an increase in CaMKII phosphorylation levels at the basal level. Although the underlying molecular mechanism requires further study, it is reasonable to speculate that a persistent activation of CaMKII may interfere with some of the normal signal pathways (or feedback loops), which would decrease the response for LTP stimulation. In fact, the expression of a constitutively active form of CaMKII enhanced postsynaptic transmission and prevented further LTP (25). A similar phenomenon has also been observed in MAPK pathways. An activation of the MAPK/ERK cascade (short term) is commonly thought to be important for cell proliferation, but a persistently activated ERK (long term) reverses and down-regulates cell growth (39,40).
Social dysfunction related to the loss of neocortical excitation/inhibition balance is an important symptom in schizophrenia (41). In the present study, we found an increase in synaptic transmission in the hippocampal CA1 area, which indicated that the balance of excitation/inhibition might be impaired in Fut8 Ϫ/Ϫ mice. On the other hand, several studies have provided direct neurophysiological evidence for disrupted LTP-like plasticity in schizophrenia (4,5,42). The evaluation of experimental LTP could be used to analyze the schizophrenialike phenotype in Fut8 Ϫ/Ϫ mice. It is well known that LTP is not a solitary phenomenon but contains various forms that can be categorized by the general mechanisms in vivo. For example, ionotropic glutamate receptors, metabotropic type glutamate receptors, and dopamine receptors (D1-D5) mediate LTP through different pathways (43). The LTP induced by HFS might be neither protein synthesis-dependent (D1-D5) nor metabotropic type glutamate receptor-dependent based on previous observations (43)(44)(45). Therefore, in the present study, we mainly focused on the molecules mediated in the ionotropic glutamate receptor-dependent pathway, including the NMDAR, AMPAR, and CaMKII in postsynaptic regions.
CaMKII activation plays essential roles in LTP because LTP induction that is produced by tetanus can be blocked by postsynaptic application of the peptide inhibitors of the kinase (46) or by a knock-in mutation that encodes T286A in CaMKII ␣ (24). Based on these findings, we speculated that a decrease in FIGURE 5. AMPAR-mediated Ca 2؉ influx was enhanced in primary neurons from Fut8 ؊/؊ mice. A, primary neurons from Fut8 ϩ/ϩ , Fut8 ϩ/Ϫ , and Fut8 Ϫ/Ϫ mice cultured for 14 days in vitro. Cells were exposed to either 1 or 3 M L-glutamic acid plus 10 M glycine, respectively. The quantitative data were obtained from three independent experiments. One experiment comprised 40 cells in different fields of view. Data represent the mean Ϯ S.E. An unpaired Student's t test was used for comparison of the peak reaction of independent groups; *, p Ͻ 0.05. B, a representative Ca 2ϩ influx was visualized using an image processor connected to a cooled CCD camera. control, the image at baseline; 3 M Glu: the image of the peak during the stimulation of 3 M L-glutamic acid plus 10 M glycine. C, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) or D-(Ϫ)-2-amino-5-phosphonopentanoic acid (DAP-5) was used to block the AMPAR-NMDAR-mediated Ca 2ϩ influx. The quantitative data were obtained from three independent experiments. One experiment comprised 40 cells in different fields of view. Data represent the mean Ϯ S.E. Student's unpaired t test was used for comparison of the peak reaction of independent groups; *, p Ͻ 0.05. recorded LTP in Fut8 Ϫ/Ϫ mice could be accompanied by a dysfunction in the activation of CaMKII. Unexpectedly, the phosphorylation levels of CaMKII in Fut8 Ϫ/Ϫ mice were significantly higher than those in Fut8 ϩ/ϩ mice ( Fig. 2A). However, it is known that a constitutively active form of CaMKII is sufficient to augment synaptic strength and then prevents further tetanus-induced LTP (25). The hippocampal slices from Fut8 Ϫ/Ϫ mice showed steeper input-output curves (Fig. 1C), which could strongly support potentiated synaptic transmission. Thus, we speculated that the absence of LTP in Fut8 Ϫ/Ϫ mice could be due to a prior maximal activation rather than to a block of LTP. Altered hippocampal synaptic plasticity is likely to affect memory processing, and therefore any such pathology may contribute to the cognitive symptoms of schizophrenia. In a maternal immune activation animal model of schizophrenia, the enhanced persistence of dentate LTP has been associated with reduced behavioral flexibility, which may be related to the alteration of synaptic plasticity (3). In addition, Angelman disease, a form of intellectual disability, is associated with persistent activation of CaMKII, and a loss of LTP underlies this learning deficit in humans (47). These findings support our hypothesis that increased CaMKII activation in postsynaptic cells triggers LTP maximally and then prevents LTP induction, which may relate to the schizophrenia-like phenotype in Fut8 Ϫ/Ϫ mice.
The AMPAR-mediated postsynaptic depolarization can activate the NMDAR, as well as Ca 2ϩ influx and CaMKII. Curiously, the phosphorylated forms of NMDARs were greatly increased in Fut8 Ϫ/Ϫ mice, compared with those in wild type mice (Fig. 2). This seems contradictory because the phosphorylated NMDARs suppress CaMKII activation (48). However, we speculated that a persistent CaMKII activation could promote the phosphorylation of NMDARs via a novel negative feedback mechanism to inhibit CaMKII activation. Therefore, the activation of AMPAR might determine the persistent activation of CaMKII in Fut8 Ϫ/Ϫ mice. The ability of AMPARs to form a complex regulates their biological functions (49). Consistently, this heteromerization was greatly enhanced in the brain tissue of Fut8 Ϫ/Ϫ mice, compared with that in Fut8 ϩ/ϩ mice, which was further confirmed by the 293T overexpression system. Taken together, these studies strongly suggest that the enhanced AMPAR heteromerization may play a key role in augmenting the synaptic strength in Fut8 Ϫ/Ϫ mice.
There are two major complex populations of AMPARs in adult hippocampal neurons: GluA1/2 and GluA2/3 (50). The GluA1/2 complex is recruited to dendritic spines in an activitydependent manner, which is associated with CaMKII and requires an interaction with postsynaptic density proteins (35,51). In contrast, GluA2/3 appears at the synapse, and its delivery is dependent on an interaction with cytoskeletal proteins, such as the N-ethylmaleimide-sensitive factor, and, therefore, GluA2/3 neurons are activated in a constitutively synaptic stimulation-independent manner (52). We focused on the function of the GluA1/2 complex in the present study because activity-dependent changes in the strength of excitatory synapses are the most important cellular mechanism for the plasticity of neuronal networks (53,54). However, we did not exclude the possibility that the aberrant enhancement of the heteromerization of GluA2/3 might also contribute to a loss of LTP in Fut8 Ϫ/Ϫ mice.
Furthermore, in CA1 pyramidal neurons, which serve as a model for understanding LTP, most receptors are known to exist as GluA1/2 heteromers with only a minor contribution from GluA2/3 complexes (36,55). The enhancement of the heteromerization of GluA2 with GluA1 and GluA3 should generate calcium-impermeable low conductivity AMPARs and reduce both the general currents and the chances of postsynaptic depolarizations, resulting in a reduction in the LTP observed in Fut8 Ϫ/Ϫ mice. It is clear that our observations in the present study could not be explained by the preceding notion. Based on the observation shown in Fig. 1, the basal transmission was increased in Fut8 Ϫ/Ϫ mice, which suggests that the changes mainly occurred in the postsynaptic region. Furthermore, the phosphorylation levels of CaMKII were increased in Fut8 Ϫ/Ϫ mice. Therefore, we speculated that an aberrant increase in GluA1/2 and GluA2/3 heteromeric formation induces a hypersensitive response for glutamate, which leads to the induction of postsynaptic depolarization, a persistent activation of NMDAR, and an increase in CaMKII phosphorylation levels, which might reverse and decrease the response for LTP stimulation in Fut8 Ϫ/Ϫ mice, as discussed above.
The present study clearly demonstrated that a loss of core fucosylation could enhance the heteromerization of AMPARs and downstream signaling, but the molecular mechanism for the negative regulator of core fucosylation in the heteromerization remains unclear. Recently, two structural biology research groups analyzed the crystal structures of glycosylated Fc␥RIIIa and human core fucosylated or afucosylated Fc of IgG (56,57). Interestingly, core fucose depletion can increase the incidence of the active conformation of the Tyr-296 of Fc, thereby accelerating the high affinity heteromerization with its receptor. Those studies clearly revealed why a lack of core fucosylation on IgG1 can dramatically enhance antibody-dependent cellular cytotoxicity activity (58,59). In addition, some important amino acid residues in the N-terminal domains of AMPARs reportedly influence their biological functions by regulating the balance between the homo-and heteromeric associations (60). It should be noted here that the heteromeric formation of AMPAR is believed to take place in the endoplasmic reticulum (61), whereas core fucosylation occurs in the medial Golgi apparatus. The present finding that a loss of core fucosylation could enhance the heteromerization of AMPARs reflect the possibility that the regulation of AMPAR heteromerization could be in the Golgi network or at the cell surface as well as in the endoplasmic reticulum. We speculated that the presence of core fucose could have regulated the strength of the interaction in AMPAR complexes.
In fact, regulation of the AMPAR function is a complex process that involves many essential factors (53). For example, the TARP family members can associate with AMPARs and affect their expression levels in PSD, which would then regulate intracellular signaling for synaptic plasticity. Coincidentally, the higher expression levels of stargazin (TARP ␥2) were increased in the PSD area in Fut8 Ϫ/Ϫ mice (Fig. 2). The underlying molecular mechanism will require further study.
Recently, Tucholski et al. (9) reported that a high mannose type of N-glycans in AMPARs was significantly decreased in patients with schizophrenia, which, together with the results of the present study, further suggests that N-glycosylation can affect the functions of AMPARs. In conclusion, our study directly demonstrated the function of core fucosylation for learning and memory in hippocampal neurons, and the possible underlying mechanism for a schizophrenia-like phenotype was shown in Fut8 Ϫ/Ϫ mice.