Neural cell adhesion molecule-associated polysialic acid potentiates alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor currents.

The highly negatively charged polysialic acid (PSA) is a carbohydrate predominantly carried by the neural cell adhesion molecule (NCAM) in mammals. NCAM and, in particular, PSA play important roles in cellular and synaptic plasticity. Here we investigated whether PSA modulates the activity of the alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) subtype of glutamate receptors (AMPA-Rs). Single channel recordings of affinity-purified AMPA-Rs reconstituted in lipid bilayers revealed that bacterially derived PSA, called colominic acid, prolonged the open channel time of AMPA-R-mediated currents by severalfold and altered the bursting pattern of the receptor channels but did not modify AMPA-R single channel conductance. This effect was reversible, concentration-dependent, and specific, since monomers of sialic acid and another negatively charged carbohydrate, chondroitin sulfate, did not potentiate single channel AMPA-R currents. Recombinant PSA-NCAM also potentiated currents mediated by reconstituted AMPA-Rs. In pyramidal neurons acutely isolated from the CA1 region of the early postnatal hippocampus, l-glutamate or AMPA (applied in the presence of antagonists blocking voltage-gated Na(+) and K(+) currents and N-methyl-d-aspartate and metabotropic glutamate receptors) induced inward currents, which were significantly increased by co-application of colominic acid. Chondroitin sulfate did not affect AMPA-R-mediated currents in CA1 neurons. The effect of colominic acid was age-dependent, since in pyramidal neurons from adult hippocampus, colominic acid failed to potentiate glutamate responses. Thus, our study demonstrates age-dependent potentiation of AMPA receptors by PSA via a mechanism probably involving direct PSA-AMPA-R interactions. This mechanism might amplify AMPA-R-mediated signaling in immature cells, thereby affecting their development.

Here, we examined a possible link between AMPA-Rs and an interesting carbohydrate, polysialic acid, which is also involved in synaptic plasticity (13). PSA is a highly negatively charged homomeric polymer of sialic acid that can form an unusual ␣2,8-linkage in chains that can be up to 200 residues long. This polymer is predominantly carried by the neural cell adhesion molecule (NCAM) and modulates its functions during cell migration and axonal outgrowth (see Ref. 13 for a review). Furthermore, delivery of PSA-NCAM to the cell surface in neurons and endocrine cells is activity-dependent (14, 15). Both NCAM protein backbone and PSA are expressed pre-and postsynaptically in a subset of spine synapses, as visualized by immunoelectron microscopy (16). The extracellular domain of PSA-NCAM can be cleaved by the tissue plasminogen activatorplasmin system at the cell surface so that it becomes a soluble molecule (17,18). The concentration of soluble NCAM is increased 10-fold after induction of long term potentiation in the dentate gyrus (19). Several other observations provide strong evidence that PSA is required for long term potentiation and depression in the CA1 region of the hippocampus (20 -22). The mechanisms underlying an increase in synaptic efficacy during long term potentiation in this region probably involve changes in the number and functional properties of AMPA-Rs (23)(24)(25)(26).
In this context, we set out to test the effects of bacterially derived PSA, colominic acid, on the activity of purified AMPA-Rs reconstituted in artificial lipid bilayers. This technique was used, since it allowed us to address the possibility of a direct modulation of AMPA receptors by PSA. Our single channel recordings in lipid bilayers demonstrate that colominic acid can dramatically prolong AMPA-R channel open time and increase its bursting activity. Colominic acid also increased AMPA-R currents in immature but not in mature CA1 pyramidal cells. Thus, our data reveal an age-dependent interaction between PSA and AMPA-Rs that may modulate neuronal transmission and plasticity in the developing central nervous system.
Membrane Solubilization-Lysed membranes (1.5 mg of protein/ml) were solubilized in ice-cold buffer A with 4% (w/v) n-octyl-␣-D-glucopyranoside, 10% (w/v) glycerol, and 0.12% (w/v) phosphatidylcholine (PC). The suspension was homogenized in an etched glass Potter-Elvehjem tissue grinder for 30 s at high speed and then placed on ice for 1 h. After centrifugation at 50,400 ϫ g for 40 min at 4°C, the supernatant was stored at Ϫ80°C.
Receptor Purification-AMPA-Rs were partially purified by a sequence of chromatographic steps: DEAE anion exchange, wheat germ lectin affinity, and polyethyleneimine anion exchange as described below. Solubilized membranes were diluted with an equal volume of buffer A with 10% glycerol and applied to a DEAE-Sepharose column equilibrated at 4°C with buffer A containing 1% n-octyl-␣-D-glucopyranoside, 10% glycerol, and 0.05% PC (buffer B). After the column was washed with 3 column volumes of buffer B at a flow rate of 2 ml/min, AMPA-Rs were eluted with a 500-ml linear salt gradient from 0 to 2000 mM KSCN in buffer B. The AMPA-R fractions were applied at a 1 ml/min flow rate to a wheat germ lectin affinity column in buffer B containing 0.1 M NaCl. The column was washed, and AMPA-Rs were eluted with 0.7 M N-acetylglucosamine in buffer B. An aliquot of AMPA-Rs was injected into the HPLC polyethyleneimine column in buffer B with a flow rate of 1 ml/min. After wash with buffer B and a linear salt gradient from 0 to 350 mM KSCN, fractions were tested with anti-GluR1 immunoblotting (28), and the most GluR1-enriched fractions were pooled. Finally, equal volumes of buffer A were added to portions of each AMPA-R pool to dilute the n-octyl-␣-D-glucopyranoside to 0.5%, and pure PC (1 mg/ml) and phosphatidylserine (50 g/ml) were added.
AMPA-Rs were immunoprecipitated from the solubilized membranes with anti-GluR1 antibodies, as previously described (28). Anti-GluR1 antibodies, covalently cross-linked to protein A-Sepharose CL-4B (Sigma), were incubated with solubilized membranes for 6 -8 h at 4°C. The immobilized receptors were washed thoroughly in 30 mM Tris, 0.5 M NaCl, 0.2% Triton X-100; extracted from the immunosupport with diethylamine/deoxycholate at pH 11.5; and immediately neutralized with 0.5 M Tris, pH 6. Purity of the preparation was verified by silver staining and Western blotting.
The purified AMPA-Rs retained many of the properties of the receptors in their native membrane microenvironment (27)(28)(29)(30). Like native AMPA receptors, reconstituted purified AMPA receptors are activated by glutamate and AMPA, although with higher affinity. They are potentiated by ampakine and blocked by CNQX. The open channel probability obtained for purified receptors is similar to that observed in native retinal cells, and their conductance corresponds to maximal conductances characteristic of native AMPA receptors (31,32).
Production and Purification of PSA-NCAM-Fc-Mouse PSA-NCAM-Fc was produced according to Vutskits et al. (33) using a stably transfected TE671 cell line kindly provided by Dr. G. Rougon (Laboratoire de Genetique et Physiologie du Developpement, Marseille, France). Coomassie staining and Western blotting of purified molecules showed a single broad band with a molecular mass above 200 kDa that is immunopositive with NCAM and PSA antibodies (Fig. 1A).
The coefficient of variation was calculated for glutamate responses at Ϫ70 mV as the following ratio: CV ϭ (variance(glutamate response) Ϫ variance(base line)) 1 ⁄2 /(mean(glutamate response) Ϫ mean(base line)). Two hundred data points were used for calculation of mean and variance values.
Reconstitution of AMPA Receptors in Lipid Bilayers-To reconstitute affinity-purified AMPA-Rs and analyze single channel currents, the "tip-dip" method was used (36 -38). The formation of artificial lipid bilayers was initiated by forming small bilayers on the tips of polished glass patch pipettes (World Precision Instruments Inc., Sarasota, FL). The BB-CH-PC electronic micropipette puller (Mechanex S.A., Switzer- land) was used to pull patch pipettes with 100-M⍀ resistance. The patch bilayer was formed in asymmetric saline conditions with the pseudoextracellular fluid (ECF) on one side, and the pseudointracellular fluid (ICF) on the other side. The ECF contained 125 mM NaCl, 5 mM KCl, 1.25 mM NaH 2 PO 4 , and 5 mM Tris-HCl (pH 7.4), whereas the ICF contained 110 mM KCl, 4 mM NaCl, 2 mM NaHCO 3 , 0.1 mM CaCl 2 , 1 mM MgCl 2 , 2 mM MOPS (pH 7.4) (Fisher). Lipid bilayers in the "outside-out" configuration were initiated by lowering both the patch pipette, which was filled with ICF, and the reference electrode into the bathing solution using the attached micromanipulator. The bathing solution consisted of 300 l of the ECF. Formation of lipid bilayers was initiated by adding 5 l of the synthetic phospholipid 1,2-diphytanoyl-sn-glycero-3phosphocholine (Avanti Polar Lipids Inc., Alabaster, AL) to the ECF. The synthetic phospholipid was prepared by dissolving it in anhydrous hexane (Aldrich) at a concentration of 1 mg/ml (39,40).
A suspension of affinity-purified AMPA-Rs was fused into the lipid bilayer on the tip of the patch pipette by gentle stirring of the extracellular fluid using a small magnetic stirring bar placed at the bottom of the microbeaker.
Single Channel Recordings-Single channel activity evoked by 290 nM AMPA alone and in the presence of 0.5, 1, 2, 3, and 10 g/ml colominic acid was recorded on a videotape (Sony Corp., New York) using a VR-10B Digital Data recorder (Instrutech Corp., Port Washing-ton, NY). For single channel analysis, data were digitally lowpass filtered at 2 kHz and compressed to final sampling rates of 20 kHz and read in pClamp software (Axon Instruments, Union City, CA). Data processing of the patch clamp recordings was essentially as described previously (41). All of the digitized traces were carefully inspected for artifacts and base-line drift before any quantitative analysis was performed. Transitions to and from the major conductance level were analyzed. The various peaks in the amplitude histogram were fitted with Gaussian curves to determine the maxima.
AMPA-R currents were plotted as a function of membrane voltage. The channel conductance was determined according to the equation g ϭ I/(V Ϫ V 0 ), where I is the current, V is the voltage, and V 0 is the reversal potential estimated from the I/V graph. To calculate the equivalent valence of the AMPA-R channel gate (z) in the presence and absence of colominic acid, we used the method described by Sinnarajah and coworkers (41) Open and closed time distributions (or the probability density functions, pdf) were assumed to be a sum of several exponential functions two exponential functions using the Marquardt least squares method (PSTAT module of Axon pClamp 6.0 software). For burst analyses, only experiments in which patches exhibiting single channel current levels were chosen. The full-size openings and the burst delimiter were determined from the plot of burst delimiter versus closings per burst as previously described (42). Burst duration time constants were derived from exponential fits obtained using the maximum likelihood method.
The estimated rate constants 1 , 2 , and 3 depend on fundamental transition rate constants k between connected states in the ion channel gating scheme. In order to derive individual rate constants for steps underlying activation of AMPA-Rs by colominic acid, we used the QuB program (available on the World Wide Web at www.qub.buffalo.edu). The digitized data were first idealized using the segmental k-means algorithm. Segmental k-means uses hidden Markov models to both find the most likely sequence of events in the data set and estimate model parameters. The maximum likelihood interval analysis program was used to compute the likelihood of the experimental series of open and closed times for a given set of trial rate constants and to search for the rate constants maximizing the likelihood (43).

Modulation of AMPA Receptors Reconstituted in a Lipid
Bilayer-To investigate the physiological properties of isolated AMPA-Rs, these were purified from brain homogenates by a sequence of chromatographic steps, culminating with immunoprecipitation using anti-GluR1 antibodies. Silver staining and Western blot analysis of purified proteins revealed a single band with a molecular mass of 105 kDa corresponding to the cognate molecular mass of the GluR1 subunit (Fig. 1B, arrow). A similar staining was found with anti-GluR2/3 antibody. Thus, the immunoprecipitated material probably represents the native heteromeric composition of AMPA-R subunits in the adult brain with an endogenous glycosylation pattern that makes it preferable to the use of recombinant AMPA receptors.
Isolated receptors reconstituted in a lipid bilayer expressed single channel currents upon application of 290 nM AMPA (Fig.  2). This activity could be blocked by the antagonist of the AMPA/kainate glutamate receptors, CNQX (1 M). When 290 nM AMPA was applied together with colominic acid (at 0.5, 1, 2, 3, and 10 g/ml), a considerable increase in channel activity was observed (Figs. 2 and 3). The effect of colominic acid was reversible; the open probability increased from 0.23 Ϯ 0.04 to 0.83 Ϯ 0.03 after co-application of 2 g/ml colominic acid and 290 nM AMPA and returned to 0.23 Ϯ 0.05 after a 2-min washout of colominic acid (n ϭ 6). Colominic acid induces no significant changes in the single channel conductance, but there was a profound increase in probability of channels to stay in the open state. Open probability, burst duration, and interburst intervals saturated at concentrations of colominic acid between 3 and 10 g/ml (Fig. 4). To verify the specificity of colominic acid effects, we tested monomers of sialic acid (5 g/ml) and another negatively charged polymeric carbohydrate, chondroitin sulfate A (5 g/ml). Co-application of these compounds with 290 nM AMPA did not potentiate channel activity (open probability of 0.27 Ϯ 0.06 (n ϭ 5) for AMPA with sialic acid and 0.24 Ϯ 0.05 (n ϭ 4) for AMPA with chondroitin sulfate A versus 0.23 Ϯ 0.04 (n ϭ 9) for AMPA alone), implying that only polymers of sialic acid, and not monomers or any negatively charged sugar polymers, modulate AMPA-R activity.
The We analyzed further the changes in open and closed times after application of colominic acid. Our analysis revealed that the histograms for open and closed duration were best fitted by three exponentials (Fig. 5, Table I For the longest open state, the changes were particularly striking and were accompanied by an increase in the occurrence of this state. Colominic acid also decreased the duration of the longest closed state. At a concentration of 3 and 10 g/ml colominic acid, the longest closed state was completely abolished. We also tested whether polysialylated NCAM (PSA-NCAM) could modulate AMPA-Rs similarly to bacterially derived colominic acid. PSA-NCAM was generated as a recombinant chi-meric protein composed of the extracellular part of NCAM coupled to the Fc portion of human IgG in mammalian cells expressing polysialyltransferase PST-1 (Fig. 1B). Similar to colominic acid, PSA-NCAM-Fc (20 g/ml) did not affect single channel conductance but increased the open time of AMPAinduced single channel currents from 1.185 Ϯ 0.002 ms to 6.525 Ϯ 0.023 ms (n ϭ 5) in a reversible manner (Fig. 6). These currents were not affected by the kainate receptor antagonist SYM 0821 (1 M, n ϭ 5) but were blocked by the AMPA/kainate receptor antagonist CNQX (1 M, n ϭ 5), confirming that they were mediated by AMPA-Rs rather than kainate receptors (Fig. 6). Treatment of PSA-NCAM-Fc with endoneuraminidase-N, which specifically degrades ␣2,8-linked PSA, com-pletely inhibited the effects of PSA-NCAM-Fc on AMPA-Rs, confirming that the potentiating effect of PSA-NCAM-Fc was due to the PSA moiety rather than to the protein backbone or  Modulation of AMPA Receptors by Colominic Acid in Neurons of the Hippocampal CA1 Region-We next aimed to verify whether colominic acid also affects native AMPA-Rs in neurons. CA1 pyramidal neurons were freshly isolated from the hippocampus at postnatal days 1-4 or 60. As expected from previous experiments (33), cellular maturation was accompanied by a significant increase in membrane capacitance, C m (immature cells: C m ϭ 15.3 Ϯ 5.6 pF (n ϭ 16); adult: C m ϭ 23.1 Ϯ 6.0 pF (n ϭ 6)). To isolate AMPA-R-mediated currents, recordings were performed in the presence of blockers of voltage-gated Na ϩ and K ϩ channels and N-methyl-D-aspartate receptors (0.5 M TTX, 4 mM 4-AP, 50 M APV) (Fig. 7A). Starting from a holding potential of Ϫ70 mV, the membrane was repetitively depolarized and hyperpolarized between ϩ100 and Ϫ100 mV for 100 ms (test pulse interval 100 ms). Bath application of glutamate (20 M) activated receptor currents that were isolated by subtracting the control currents recorded before application from the responses obtained in the presence of the agonist at corresponding membrane potentials (Fig. 7B). The I/V relation of the glutamate responses displayed a slight outward rectification and currents reversed at about 0 mV (Fig. 7C).
In immature neurons from mouse hippocampus, glutamate (20 M) evoked current densities of 1.2 Ϯ 0.7 pA/pF (n ϭ 9) at Ϫ70 mV. After washout, co-application of glutamate (20 M) together with colominic acid (40 g/ml) to the same cells led to a significant increase in glutamate responses in 9 of 11 cells (2.3 Ϯ 1.1 pA/pF) (Fig. 7, B-D), whereas the AMPA-R currents of two neurons were not affected. In single cells, the colominic acid-induced potentiation of the glutamate responses was variable, ranging between 1.3-and 7.7-fold (2.7 Ϯ 2, n ϭ 9). A similar increase of 20 M glutamate-evoked currents by colominic acid (40 g/ml) was observed in age-matched neurons isolated from rat hippocampus (increase in current density by a factor of 2.6 Ϯ 0.4, V ϭ Ϫ70 mV, n ϭ 4). In four experiments in mouse neurons, in which the S.D. of noise was below 2.5 pA, colominic acid potentiated glutamate responses 2.5-fold, and this increase was accompanied by a decrease in the coefficient of variation from 17.4 Ϯ 1.8 to 10.3 Ϯ 0.7% (p Ͻ 0.05, t test). This reduction in current variability is consistent with results from single channel recordings from purified AMPA-Rs, showing that colominic acid increases open probability rather than single channel conductance.
To test for the specificity of the effect, chondroitin sulfate A (40 g/ml) was co-applied with glutamate (20 M) after washout of colominic acid to the same cell. No increase in glutamate responses was observed under these conditions (1.3 Ϯ 1.2 pA/ pF, n ϭ 3). The subsequent addition of the AMPA/kainate receptor antagonist NBQX (10 M, n ϭ 3) or the AMPA receptor-specific antagonist GYKI 53655 (50 M, n ϭ 4) completely blocked the receptor currents elicited by glutamate alone or by co-application of glutamate and colominic acid (not shown). Colominic acid by itself (20 or 40 g/ml, n ϭ 6) did not have any Starting from Ϫ110 mV, currents were elicited at membrane potentials ranging between ϩ70 and Ϫ160 mV. B, subsequently, glutamate (Glu; 20 M) and colominic acid (CA; 40 g/ml) were applied to the same cell as indicated, whereas the membrane potential was clamped between Ϫ100 and ϩ100 mV (V H ϭ Ϫ70 mV). Glutamate-induced currents were increased upon co-application with colominic acid but not with chondroitin sulfate A (CS; 40 g/ml). C, I/V curves of the respective responses (squares, Glu; triangles, Glu with colominic acid; diamonds, Glu with chondroitin sulfate A). D, a significant increase in current density (at Ϫ70 mV) was registered in the presence of colominic acid, whereas chondroitin sulfate A did not affect glutamate-induced responses. Cell numbers are given in parentheses.
effects on the residual control currents recorded in the presence of TTX, APV, (S)-␣-methyl-carboxyphenylglycine and 4-AP.
In contrast to the findings in immature cells, neurons freshly isolated from the adult mouse hippocampus were insensitive to colominic acid. The current density estimated in mature neurons in response to glutamate application, using the protocols described above, was 1.8 Ϯ 0.9 pA/pF as compared with 2.0 Ϯ 1.2 pA/pF (n ϭ 6) upon co-application of glutamate and colominic acid. These differences were not statistically significant. As expected, bath application of GYKI 53655 (50 M, n ϭ 3) completely inhibited the responses evoked either by glutamate alone or by co-application of glutamate and colominic acid (not shown).

Mechanisms of Interaction between Polysialic Acid and
AMPA Receptors-In our single channel study using purified receptors, 290 nM AMPA was sufficient to elicit single channel activity with a 50 Ϯ 10-pS conductance, as was previously reported for reconstituted AMPA-R channels (29,30). Considering that recordings were obtained at positive voltages (ϩ71 mV) and AMPA-Rs often display outward rectification, this value roughly corresponds to the largest conductance level reported for native receptors (32). Strikingly, colominic acid increased burst duration and mean open time within bursts elicited by 290 nM AMPA and decreased the interburst intervals and mean close time in a concentration-dependent manner. These effects saturated at concentrations around 2 g/ml, corresponding to 85 nM. Both bacterially produced colominic acid and eukaryotically produced PSA-NCAM potentiated AMPA-Rs. Since we used purified AMPA-Rs for our experiments, the data strongly suggest that colominic acid and PSA-NCAM directly interact with AMPA-Rs. Since 1) similar levels of potentiation of AMPA-R currents by colominic acid were observed in lipid bilayers and isolated hippocampal neurons and 2) potentiation of currents in neurons was accompanied by a decrease in the coefficient of current variation (a measure inversely dependent on the open probability), it is conceivable that AMPA-Rs in lipid bilayers and native cells were modulated by the same mechanism(s).
The single channel currents induced by AMPA and modified by colominic acid exhibited three closed and three open states. That the closed time distributions are best described by three exponentials is in agreement with a previous study that used native neuronal receptors (44). The slight difference in the rates estimated in our study and in the previous study may be due to the fact that we used AMPA instead of glutamate to activate the receptors. We were able to identify three distinct open states, and the value obtained for the longest open time was similar to that reported for receptors expressed in HEK cells (45). In the presence of higher concentrations of colominic acid, two exponentials gave the best fit for closed times, indicating the existence of only two closed states (Table I). Using our single channel data, a kinetic model for the AMPA-R was constructed to interpret our observations (Fig. 8). Our gating scheme obtained for AMPA receptors is different from the one reported previously (32,46), probably since purified receptors exhibit higher affinity for AMPA than the native ones (47), making possible a rapid occupation of all binding sites. Therefore, the transition from the closed to fully open states was achieved very rapidly, and we were unable to observe intermediate steps involving subconductance levels. In the presence of 290 nM AMPA, we did not observe time-dependent changes in conductance levels. Despite this, purified AMPA-Rs were potentiated by cyclothiazide, 2 a known inhibitor of AMPA-R de-sensitization (48), indicating that purified AMPA-Rs, like the native ones (49), exhibit desensitization under conditions of prolonged agonist application.
The rate constants derived from this model demonstrated that in the presence of 290 nM AMPA, the probability of entry into open state O 3 from O 2 was low, and the entry into closed state C 3 from C 1 also occurred infrequently, as shown in Fig.  8A. When 10 g/ml colominic acid was co-applied with AMPA, the transition from C 1 to C 3 became very unlikely, explaining why two rather than three distinct closed states were observed in the dwell time histogram (Fig. 5) (Fig. 8B).
Our experiments revealed that the colominic acid interaction with AMPA-Rs resulted in voltage-dependent channel openings. Calculations showed an ϳ67% increase of equivalent charge of the AMPA channel gate in the presence of 3 g/ml colominic acid together with AMPA compared with that of AMPA alone. Colominic acid did not activate the receptors in the absence of AMPA, suggesting that an open channel conformation may be a prerequisite for the interaction of colominic acid with AMPA-Rs. One possible explanation is that the highly polyanionic colominic acid may interact with the positively charged amino acid residues of the open channel pore of AMPA-Rs. Such a mechanism has been previously suggested for another polyanionic carbohydrate, heparin (41,50). receptor into the desensitized state can result in a decrease of interburst interval and/or increased burst duration. Indeed, single channel burst analysis of our data revealed both a dosedependent decrease in interburst interval and an increase in burst duration with colominic acid.
Age Dependence of Polysialic Acid Effects on Native AMPA Receptors-Although PSA had no effect on AMPA-R-mediated responses in pyramidal neurons isolated from the adult hippocampus, profound modulatory effects were seen with immature hippocampal neurons. This age-dependent effect with cells may indicate that PSA-dependent modulation critically depends on the way the receptors are organized/anchored in their membrane environment. Once removed from such an environment, as in the reconstitution system, age dependence disappears, since robust potentiation was observed in AMPA-Rs purified from the adult brain (this study) or in reconstituted synaptosomes isolated from the adult brain (54). This may happen due to a bias in receptor composition possibly introduced by the purification procedure, for instance by selecting for particular glycan structures by lectin column chromatography. This bias may mask developmental changes in splicing or subunit composition of AMPA-Rs in neural cells (35,55) possibly underlying the age-dependent differences in PSA-mediated modulation. Alternatively, removal of AMPA-R-associated molecules during the purification has to be considered (i.e. neurons isolated from the adult brain might express molecules preventing interaction between PSA and AMPA-R). Particularly, molecules at the cell surface and extracellular matrix components could be involved, such as members of the family of transmembrane AMPA-R regulatory proteins and Narp-NP1 pentraxin complexes (56,57). However, it is also conceivable that cytoskeletal proteins interacting with AMPA-Rs, including members of the GRIP family and SAP97 (58), may modify the receptor conformation, thereby affecting AMPA-R interactions with PSA.
Since the early postnatal brain possesses high levels of PSA-NCAM (13,59), it is possible that the interplay between PSA-NCAM and AMPA-Rs at this developmental stage amplifies neuronal membrane depolarization and Ca 2ϩ entry, thus affecting their development. For instance, AMPA-R activity has been reported to be important for synapse maturation and development of the dendritic architecture (60,61). Available data support the view that PSA has a morphogenic activity and is required for correct lamination and synaptogenesis of mossy fibers (62) and structural plasticity in the hypothalamus (63). It remains, however, to be elucidated to which degree the morphogenic role of PSA is related to its modulation of AMPA-Rs.