Neural Cell Adhesion Molecule-associated Polysialic Acid Inhibits NR2B-containing N-Methyl-d-aspartate Receptors and Prevents Glutamate-induced Cell Death*

The neural cell adhesion molecule (NCAM) and its associated glycan polysialic acid play important roles in the development of thenervoussystemandN-methyl-d-aspartate(NMDA)receptor-dependent synaptic plasticity in the adult. Here, we investigated the influence of polysialic acid on NMDA receptor activity. We found that glutamate-elicited NMDA receptor currents in cultured hippocampal neurons were reduced by ≈30% with the application of polysialic acid or polysialylated NCAM but not by the sialic acid monomer, chondroitin sulfate, or non-polysialylated NCAM. Polysialic acid inhibited NMDA receptor currents elicited by 3 μm glutamate but not by 30 μm glutamate, suggesting that polysialic acid acts as a competitive antagonist, possibly at the glutamate binding site. The polysialic acid induced effects were mimicked and fully occluded by the NR2B subunit specific antagonist, ifenprodil. Recordings from single synaptosomal NMDA receptors reconstituted in lipid bilayers revealed that polysialic acid reduced open probability but not the conductance of NR2B-containing NMDA receptors in a polysialic acid and glutamate concentration-dependent manner. The activity of single NR2B-lacking synaptosomal NMDA receptors was not affected by polysialic acid. Application of polysialic acid to hippocampal cultures reduced excitotoxic cell death induced by low micromolar concentration of glutamate via activation of NR2B-containing NMDA receptors, whereas enzymatic removal of polysialic acid resulted in increased cell death that occluded glutamate-induced excitotoxicity. These observations indicate that the cell adhesion molecule-associated glycan polysialic acid is able to prevent excitotoxicity via inhibition of NR2B subunit-containing NMDA receptors.

Cell interactions play important roles during development and in maintenance and modification of synaptic functions in the adult. The question is whether recognition molecules that specify and modulate contacts between neural cells may influence other parameters essential for nervous system function, such as neurotransmitter release and receptor activity. Among the recognition molecules with widespread functions is the neural cell adhesion molecule (NCAM) 4 that starts to be expressed at the time of neural tube closure and remains detectable at lower levels in the adult. Its importance in shaping synaptic functions in the adult has long been recognized (1)(2). NCAM is unique among recognition molecules in that its adhesive and concomitant signal transduction functions are modified by an unusual glycan, polysialic acid, a highly negatively charged and voluminous carbohydrate, which is regulated by its attachment to the protein backbone (3). PSA is a polymer of ␣2,8-linked sialic acid residues with chain lengths of up to 200 residues. It has not been detected on other recognition molecules. PSA is synthesized by two sialyltransferases, ST8Sia-II and ST8Sia-IV, and attached to the fifth immunoglobulin-like domain of NCAM (4). Even more so than the protein backbone of NCAM, expression of PSA is developmentally regulated, with high expression during embryonic stages and gradual reduction as development proceeds. However, it remains expressed in some areas of the brain during adulthood, including the hippocampus, which undergoes functional changes underlying synaptic plasticity. In particular, PSA-NCAM is required for NMDAR-dependent long-term potentiation and spatial learning (5)(6)(7)(8)(9).
The NMDARs are a subtype of ionotropic glutamate receptors that are found widely throughout the brain. NMDARs are heteromers assembled from the NMDAR subunit NR1 and at least one type of NR2 subunit (10). CA1 pyramidal cells, for instance, express mostly two different NR2 subunits, NR2A and NR2B, and perinatally, NR2D (11). It is believed that the different NR2 subunits confer distinct gating and pharmacological properties to NMDARs (12) and couple them to distinct intracellular signaling mechanisms (13), which may shape their characteristic roles in synaptic plasticity (14 -16). NMDARs are expressed both synaptically and extra-synaptically. The subunit composition at these locations is not uniform; synaptic NMDARs predominantly contain NR2A, whereas extra-synaptic NMDARs contain mostly the NR2B subunit (17), although recent data in mature cultured hippocampal neurons support a view that both subtypes can be located in either synaptic or extrasynaptic compartments (18). Importantly, synaptic and extra-synaptic NMDARs may activate opposing pathways to prevent or induce glutamate-induced excitotoxic cell death, respectively (19).
Several lines of evidence indicate that there are intriguing links between PSA-NCAM and glutamate receptors. NMDARs are co-redistributed with NCAM after induction of long-term potentiation (20), and NCAM associates with the postsynaptic spectrin-based scaffold cross-linking NCAM with the NMDAR and calmodulin kinase II␣ (21). Activity of NMDARs is required for PSA-NCAM-stimulated synaptogenesis (22), and PSA directly increases the probability of the open state of ␣amino-3-hydroxy-5-methylisoxazole-4-propionate subtype of glutamate (AMPA) receptors (23). To address the question of whether PSA may also modulate NMDARs, we examined the effects of PSA on glutamate-evoked NMDAR currents in cultured hippocampal neurons and synaptosomal NMDARs incorporated in artificial lipid bilayers.

EXPERIMENTAL PROCEDURES
Chemicals-Polymers containing 25-50 sialic acid residues were purified from colominic acid (Fluka, Buchs, Switzerland) using anion exchange chromatography on a Hamilton PRPX column dissolved in 20 mM Tris, pH 7.4, with 0 -500 mM NaCl and detected at 214 nm (24). The concentration was determined using a colorimetric resorcinol method adapted to microtiter plate format (25).
Mouse PSA-NCAM-Fc, containing the extracellular domain of NCAM and the Fc portion of human IgG, was produced according to Vutskits et al. (26) using a stably transfected TE671 cell line kindly provided by Genevieve Rougon. Mouse NCAM-Fc was produced using stably transfected Chinese hamster ovary cells as described (27). Polysialylation of PSA-NCAM-Fc and PSA negativity of NCAM-Fc were checked by Western blotting using a monoclonal antibody (clone 735) to PSA (23,28). Recombinant endoneuraminidase-N (Endo-N) (0.2 g/ml), which specifically cleaves the long-chain PSA residues from the NCAM protein backbone (29), was used to remove PSA from cultured neurons. The efficacy of this enzymatic treatment was confirmed by the loss of PSA immunoreactivity using the monoclonal antibody to PSA.
Electrophysiological Recordings in Cultured Neurons-All recordings were performed in the whole-cell patch clamp mode at room temperature in extracellular solution containing 140 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 0.5 mM MgCl 2 , 10 mM HEPES, 30 mM D-glucose, 12 mM D-sucrose, adjusted to pH 7.3 with NaOH. In some experiments, as indicated under "Results," modified solutions with 0 mM Ca 2ϩ or 1.5 mM Mg 2ϩ were used. Patch pipettes were filled with 135 mM CsCl, 10 mM KCl, 10 mM HEPES, 0.2 mM EGTA, 2 mM Mg-ATP, 0.2 mM Na-GTP, 10 mM D-glucose, adjusted to pH 7.2 with CsOH with an osmolarity of 290 -320 mosmol/liter. Fire-polished patch pipettes with a resistance of 2-4 megaohms were pulled on a DMZ-Universal puller (Zeitz, Munich, Germany) using GB150F-8P borosilicate glass (Science Products, Hofheim, Germany). Dissociated hippocampal cultures used for recordings were prepared from 1-3-day-old mice (30) and maintained in vitro for 12-20 days. In these relatively mature neurons, the composition of NMDARs appeared to be stable (18).
To isolate NMDAR currents in neurons, tetrodotoxin (1 M, Alomone Labs, Jerusalem, Israel), tetraethylammonium chloride (10 mM, Sigma-Aldrich), 1,2,3,4-tetrahydrobenzo͓f͔quinoxaline-7-sulfonamide (5 M, Tocris, Bristol, UK), (RS)-methyl-4-carboxyphenylglycine (100 M, Tocris), and picrotoxin (100 M, Tocris) were added to the extracellular solution. These compounds were applied to block Na ϩ and K ϩ voltage-dependent channels and ␥-aminobutyric acid, type A, glycine, and non-NMDA glutamate receptors. Cells were voltage-clamped using an EPC9 amplifier (HEKA, Lambrecht, Germany) at Ϫ60 mV with a series of voltage steps from Ϫ100 to ϩ60 mV every 3 s, with each step lasting 100 ms. NMDAR currents were evoked by glutamate (3, 5, or 30 M), which was applied using a multiple input, single-barrel application system that allowed direct application onto the patched cell. Glutamate rather than NMDA was used in our experiments since it is a natural ligand of NMDARs, and thus, the data obtained are easier to relate to physiological processes in the brain. Control experiments were performed to verify that solutions coming from different inputs did not interfere with each other. Compounds were applied for 10 s with a 60-s washout period between applications. The current/voltage relationships were plotted in Sigma Plot 5.0 (SPSS Inc., Chicago, IL) after subtraction of the leak current and averaging currents elicited by four voltage steps during the application of glutamate. The leak current was the average of the five voltage steps before agonist application. Currents were normalized using the value recorded in the presence of glutamate at ϩ60 mV. Statistical comparisons of the current/voltage curves were done in Statistica 5.0 (StatSoft Inc., Tulsa, OK) using analysis of variance with "voltage" and "treatment" as repeated measures. To evaluate whether PSA induces changes in the conductance or gating of NMDAR channels in cultures, the fluctuation analysis of glutamate-activated NMDAR currents was performed (31,32). The coefficient of variation was calculated for glutamate-evoked NMDAR currents at Ϫ60 mV as the following ratio: CV ϭ [(variance (glutamate response) Ϫ variance (base line)] 1/2 /[mean (glutamate response) Ϫ mean (base line)]. This measure is useful in light of the following considerations. The mean and variance of current generated by n independent channels, which are open with a probability p and have a conductance q, are described by a binomial law and equal to q⅐n⅐p and q 2 ⅐n⅐p⅐(1 Ϫ p), respectively. Then CV Ϫ2 ϭ n⅐p/(1 Ϫ p) is a measure of variability that does not depend on q and changes only when the number of channels or their open probability is changed. Any changes in q alone would not affect this measure. Using these equations one can derive that the ratio between mean currents before and after inhibition of channel activity (x) would be proportional to the ratio between corresponding CV Ϫ2 (y). The x,y values would lie along the line of identity (y ϭ x) if inhibition is exclusively due to reduction in number of channels (Fig. 2B). If inhibition of channel activity is due to reduction in open probability, the x,y values would be below y ϭ x. The x,y values would lie along the line y ϭ 1 if reduction in q mediates inhibition.
Synaptosomal Preparation-The synaptosomes were prepared as described elsewhere (33) with minor modifications. The hippocampi were isolated from 7-10 days old Sprague-Dawley rats and homogenized in an Eppendorf tube with 400 l of ice-cold 95% O 2 , 5% CO 2 -aerated Krebs-Henseleit buffer consisting of 118.5 mM NaCl, 4.7 KCl, 1.18 mM MgSO 4 , 2.5 mM CaCl 2 , 1.18 mM KH 2 PO 4 , 24.9 mM NaHCO 3 , 10 mM dextrose, with the pH adjusted to 7.4. Leupeptin (0.01 mg/ml), pepstatin A (0.005 mg/ml), aprotinin (0.10 mg/ml), adenosine deaminase (10 mg/ml), and benzamide (5 mM) were added to the buffer to minimize proteolysis. The homogenate was diluted with 1.60 ml of additional Krebs buffer after being homogenized with 5 turns of a handheld pestle. The mixture loaded into the 1-cc tuberculin syringe was forced through three layers of nylon (Tetko, 100-m pore size) pre-wet with 150 l of Krebs buffer and collected in an Eppendorf tube. After another filtering with a pre-wet Millex filter (5-m pore size polyvinylidene difluoride Millipore filter), the filtrate was centrifuged at 1000 ϫ g for 15 min in a microcentrifuge at 4 C. After removing the supernatant, the pellet, which contained the synaptosomes, was resuspended in 100 l of Krebs buffer for electrophysiological recordings.
Reconstitution of Synaptosomal NMDA Receptors in Lipid Bilayers-Incorporation of synaptosomal NMDARs in artificial lipid bilayers was carried out using the "tip-dip" method (34). The phospholipid bilayer was formed at the tip of a polished glass pipette (World Precision Instruments Inc., Sarasota, FL). The P-2000 laser micropipette puller (Sutter Instrument Co., Novato, CA) was used to pull pipettes with 100-megaohm resistance. The synthetic phospholipids were prepared by dissolving 1,2diphytanoyl-sn-glycero-3-phosphocholine (Avanti Polar-Lipids Inc., Alabaster, AL) in hexane (Sigma-Aldrich) to obtain a concentration of 1 mg/ml. About 3-5 l of this phospholipid preparation was delivered into 300 l of bath solution. The bilayer formation was initiated by successive transfer of two monolayers onto the tip of the patch pipette in an asymmetric saline condition with "outside-out" configuration. The bath solution contained pseudoextracellular fluid composed of 125 mM NaCl, 5 mM KCl, 1.25 mM NaH 2 PO 4 , and 5 mM Tris HCl. The pseudointracellular fluid consisting of 110 mM KCl, 4 mM NaCl, 2 mM NaHCO 3 , 1 mM MgCl 2 , 0.1 mM CaCl 2 , and 2 mM MOPS was used as the pipette solution. After forming a stable membrane, a 3-5-l suspension of synaptosomes was transferred to the pseudoextracellular fluid. Gentle stirring facilitated fusion of synaptosomal fragments into the bilayer.
Only the data exhibiting long stretches of single channel current transition without base-line drifts were chosen for quantitative analysis. All point current amplitude histograms were constructed and fitted with Gaussian curves to identify the individual conductances. The single channel conductance of NMDA receptors were obtained by plotting current as a function of membrane voltage, and the conductance was determined according to the equation g ϭ I/(V Ϫ V 0 ), where I is the single channel current, V is the voltage, and V 0 is the reversal potential. The single channel open probability was estimated as where R c and R o stand for the areas under the current-amplitude histogram corresponding to close and open states, respectively.
Excitotoxicity Assay-Cell survival assays were performed using dissociated hippocampal cultures maintained in vitro for 7 days (in glass-bottomed 96-well plates, which are well suited for pharmacological experiments) or for 12 days (in cloning cylinders on glass coverslips, which are superior in comparison to plates in terms of cell survival after 9 days in vitro). The conditioned culture medium (70 l/well) was removed from wells and saved for the later use (see below). Compounds used during the induction of excitotoxicity were applied in Lockes buffer (154 mM NaCl, 5.6 mM KCl, 10 mM CaCl 2 , 3.6 mM NaHCO 3 , 5 mM HEPES, and 5.5 mM glucose, pH 7.4. Lockes buffer (70 l) alone or Lockes buffer containing the compounds of interest was then added to and immediately removed from the cells twice. A further 70 l of the respective solutions was then added, and the cells were returned to the incubator for 10 min. For washing, solutions were exchanged 3 times by replacing 70% of the total volume at each step to avoid drying out the cells during the washing procedures and to obtain a Ͼ95% solution exchange after the last step. After this, the conditioned medium (saved as culture supernatant) was returned to the cells. Cell survival was measured 20 -24 h after treatment using the live cell marker, calcein (0.5 M, Molecular Probes, Eugene, OR), and the dead cell marker, propidium iodide (3 M, Sigma-Aldrich). Three visual fields were counted per well, and there were three wells per treatment group per culture preparation. Excitotoxicity was calculated as the ratio of dead cells to total cells. The t test was used to determine the statistical differences between the experimental groups.

Reduction of NMDAR Currents by PSA in Cultures of Dissociated Hippocampal
Neurons-Whole-cell voltage clamp recordings were used to measure the NMDAR currents in pyramidal-like neurons from 12-20-day-old cultures (Fig. 1A). The currents were evoked by the application of glutamate (3 M) and recorded at different membrane potentials from Ϫ100 to ϩ60 mV. Because previous experiments showed an effect of 40 g/ml bacterially produced PSA, colominic acid, on AMPA receptor currents in hippocampal neurons (23), we tested whether this concentration of colominic acid would affect NMDAR activity. Co-application of colominic acid with glutamate reduced the NMDAR currents by ϳ30% (n ϭ 40, F 8,312 ϭ 60.34, p Ͻ 0.000001) at all tested membrane potentials (Fig. 1B). After wash-out of colominic acid, the NMDAR currents in seconds returned to control levels (Fig. 1A).
To verify that the reduction of NMDAR currents was specific for colominic acid, a series of control experiments was performed. First, to show that the effect was due to PSA and not impurities in the sample, polymers of sialic acid with 25-50 residues were purified. Co-application of the purified PSA (10 g/ml) with glutamate produced a reduction of NMDAR currents of ϳ20% (n ϭ 4, F 8,24 ϭ 4.28, p Ͻ 0.005) in comparison to glutamate alone. This effect was not different from the effect of unpurified colominic acid (Fig. 1D).
Because PSA is a polymer of sialic acid, there was the possibility that the effect was not specific for the polymeric and highly negatively charged PSA but could also be induced by equally overall negatively charged monomers. To investigate this possibility, sialic acid monomers (40 g/ml) were co-applied with glutamate. This, however, did not produce any significant change in the glutamate-induced NMDAR currents compared with glutamate alone (n ϭ 4, F 8,24 ϭ 2.12, p Ͼ 0.05, Fig. 1D). As another control, colominic acid was treated with the enzyme endo-N, which is known to specifically cleave ␣2,8linked PSA. Endo-N-treated colominic acid did not change NMDAR currents when co-applied with glutamate (n ϭ 6, F 8,56 ϭ 0.98, p Ͼ 0.05, Fig. 1, C and D).
Next, we examined the effect of another negatively charged carbohydrate polymer, chondroitin sulfate. Co-application of chondroitin sulfate (40 g/ml) with glutamate resulted in NMDAR currents that were not different from those induced by glutamate alone (n ϭ 7, F 8,48 ϭ 0.35, p Ͼ 0.05, Fig. 1D). Thus, the influence of colominic acid on NMDARs is specific to the unique PSA polymer of sialic acid.
Because these experiments were performed using a bacterially produced PSA, we next addressed the question of whether mammalian PSA associated with NCAM would affect NMDARs. Co-application of glutamate with mouse PSA-NCAM-Fc (10 g/ml), containing the extracellular domain of NCAM in fusion with the Fc portion of human IgG, reduced NMDAR currents in the same manner as colominic acid (n ϭ 7, F 8,48 ϭ 19.38, p Ͻ 0.000001). Application of non-polysialylated NCAM-Fc (10 g/ml) did not affect NMDAR currents (n ϭ 3, F 8,16 ϭ 2.41, p Ͼ 0.05). Thus, PSA carried by NCAM, but not the NCAM protein backbone that carries glycan chains not related to PSA, inhibits NMDAR currents (Fig. 1D).
Pharmacological and Fluctuation Analyses of the Effects of PSA on NMDARs-Because divalent cations forming a complex with PSA may modify properties of the PSA chains (35,36) and regulate the activity of NMDARs (12, 37), we tested whether modulation of NMDAR currents by colominic acid depends on the extracellular concentrations of either Ca 2ϩ or Mg 2ϩ . Therefore, the effects of colominic acid on NMDAR currents were compared in the same cells for two Ca 2ϩ concentrations, 0 or 2 mM, or for two Mg 2ϩ concentrations, 0.5 or 1.5 mM ( Fig.  2A). There was no difference in modulation of NMDAR currents by colominic acid under these conditions. These experiments suggest that association of the negatively charged PSA with positively charged divalent cations is not critical for the effects of PSA on NMDARs. Also, the addition of the NMDAR agonist spermidine (100 M) or omitting 2 M glycine normally included in the perfusion solution did not alter the inhibition of NMDAR currents by colominic acid (data not shown). However, increasing the concentration of glutamate from 3 to 30 M abolished the inhibition of NMDAR currents by colominic acid ( Fig. 2A), suggesting that PSA acts as a competitive antagonist, possibly at the glutamate binding site.
To investigate whether PSA reduces the conductance of NMDARs or the number of functional channels and probability of NMDAR channel opening, a fluctuation analysis was performed. Only recordings with a low level of background noise (S.D. Ͻ 2.5 pA) were selected for this analysis to provide a reliable estimate of the coefficient of current variation. Fig. 2B presents a summary of experiments performed at the two concentrations of colominic acid, 0.01 and 40 g/ml, and 10 g/ml PSA-NCAM-Fc, showing the relationship between changes in glutamate-induced current (I, recorded at Ϫ60 mV), and CV Ϫ2 , where CV is the coefficient of current variation (Fig. 2B). Obvi- ously, the ratio between amplitudes of currents recorded before and after application of 0.01 g/ml colominic acid is close to 1, i.e. there is no effect of the glycan, in contrast to a strong reduction of current amplitude elicited by 40 g/ml colominic acid and 10 g/ml PSA-NCAM-Fc. Because values of parameters are grouped around the line of identity y ϭ x, one can conclude that reduction in the current amplitude is mostly mediated by a decrease in number of functional channels.
The NR2B Subunit Is Necessary for Inhibition of NMDA Receptors by PSA-The results derived by the fluctuation analysis and the fact that PSA only partially inhibited NMDAR currents suggest that PSA inhibits a fraction of NMDARs at the cell surface. To test whether NMDARs containing NR2B subunit are affected by PSA, the NR2B specific antagonist, ifenprodil (10 M), was applied to cultures of hippocampal neurons. Ifenprodil reduced NMDAR currents by ϳ40% (n ϭ 6, F 8,40 ϭ 11.42, p Ͻ 0.000001), which is a level of inhibition similar to that generated by PSA when applied to these cells (Fig. 3A). Furthermore, ifenprodil fully occluded the inhibition of NMDAR currents elicited by colominic acid (Fig. 3A). Similarly, another potent antagonist of NR2B-containing NMDARs, Ro 25-6981 (0.5 M), mimicked the effects of colominic acid (Fig. 3C).
Data described so far provide evidence that exogenous soluble PSA and PSA-NCAM inhibit NMDARs. Next, we asked whether endogenous PSA-NCAM expressed by hippocampal neural cells may regulate the activity of NMDARs. We used endo-N as the most specific treatment to remove PSA. As a complementary approach, we used neurons derived from NCAM-deficient mice (38). We, thus, treated wild type neurons with endo-N or used neurons derived from NCAM-deficient mice to estimate the total, ifenprodil-sensitive (NR2B subunit-mediated) and insensitive (NR2A subunit-mediated) NMDAR current components in neurons deficient in PSA or both polysialylated and non-polysialylated forms of NCAM. There was no difference between control and endo-N-treated neurons in the amplitude of the total, ifenprodil-sensitive and insensitive components of NMDAR currents (Fig. 3B), showing that removal of PSA does not affect NMDAR currents. However, the magnitudes of the total NMDAR current and its components were increased 2-fold in NCAM deficient neurons. Enzymatic removal of PSA or ablation of NCAM did not affect the response of NMDARs to PSA (Fig. 3C). Thus, our data with endo-N treatment show that only PSA and PSA-NCAM presented to neurons as soluble compounds but not PSA associated with NCAM at the cell surface of neurons affect NR2B-containing NMDARs. An increase in NMDAR currents in NCAM-deficient, although not in endo-N-treated neurons, suggests that expression of NCAM at the neuronal cell surface is required for normal activity of NMDARs via a PSA-independent mechanism(s) that operates in addition to soluble PSA/PSA-NCAM.

PSA Reduces Open Probability of NR2B-containing Synaptosomal NMDA Receptors Reconstituted in Lipid Bilayers-We
previously developed a method for recording single synaptosomal glutamate receptors (34) and utilized it here to study the modulation of NMDARs by PSA. Single-channel currents mediated by NMDARs were isolated pharmacologically by blocking other major receptors and channels. Fig. 4A shows sample traces with currents from NMDA receptors reconstituted in lipid bilayers and activated by 2 M glutamate. Channel activity is evident by upward transitions of the current representing the open state. Their amplitudes were approximately of 4 pA. To verify the identity of recorded channels, the specific antagonist of NR2B-containing NMDARs, Ro 25-6981, was applied at the end of each experiment, and only channels that were fully blocked by Ro 25-6981 were included in the analysis (Fig. 4, E and H). Application of colominic acid to a single NMDAR at increasing concentrations (1, 2, and 3 g/ml) produced a progressive inhibition of NMDAR activity (Fig. 4,  B-D). Analysis of amplitude histograms showed that colominic acid affected exclusively open probability rather than single-channel conductance (see Fig. 6, A  and B). When colominic acid was applied to NMDARs at 0.01 and 0.1 g/ml, it produced no significant effects (see Fig. 6, A and B). Also chondroitin sulfate (even at 10 g/ml) did not affect activity of NMDARs, confirming the specificity of PSA action (see Fig. 6, A and B).
Because experiments with hippocampal neurons revealed that PSA inhibits only NMDARs activated by low glutamate concentrations, we investigated this aspect also using single-channel recordings. Application of glutamate at increasing concentrations (2, 4, and 6 M) to a single channel in the presence of colominic acid (3 g/ml) resulted in restoration of channel openings (see Fig. 6C). The single channel conductance was not affected at any tested concentration of glutamate (see Fig. 6D). In summary, these data demonstrate that PSA affects gating of NR2B-containing NMDAR at low glutamate concentrations.
To verify whether effects of PSA are specific for NR2B-containing NMDARs, we performed analysis of single channel activity in the presence of Ro 25-6981 (Fig. 5A). There were no significant effects of colominic acid on opening and conductance of Ro 25-6981-insensitive channels (Figs. 5B and 6, E and F). Because these channels were blocked by DL-2-amino-5phosphonopentanoic acid (APV) but not Ro 25-6981, they are NMDARs that lack the NR2B subunit. Thus, PSA specifically inhibits NR2B-containing NMDARs.
PSA Reduces Excitotoxic Cell Death Induced by Low Micromolar Concentrations of Glutamate-To measure the proportion of live cells in cultures, dead cells were labeled by propidium iodide and live cells were labeled by calcein, and the ratio between the number of live cells to the total number was calculated. Exposure of hippocampal cultures to 5 or 30 M glutamate for 10 min induced a 20 and 40% increase, respectively, in cell death within 24 h (Fig. 7, A-C). We chose these concentrations of glutamate since colominic acid significantly inhibited NMDAR currents activated by 5 M glutamate but failed to inhibit NMDAR currents activated by 30 M glutamate ( Fig. 2A). Application of colominic acid, the general NMDAR antagonist APV, or the antagonist of NR2B-containing NMDARs, Ro 25-6981, abolished the excitotoxic effect of 5 M glutamate (Fig. 7B). Although both APV and Ro 25-6981 also abolished the excitotoxic effect of 30 M glutamate, colominic acid failed to do so (Fig. 7C). These results show that glutamateinduced excitotoxic cell death depends on NMDARs containing the NR2B subunit and that colominic acid is able to prevent this effect at low glutamate concentrations. The failure of colominic acid to inhibit the excitotoxicity induced by a higher concentration of glutamate is in accordance with the electrophysi- ological data showing that NMDAR currents elicited by 30 M glutamate are insensitive to colominic acid ( Fig. 2A).
The described experiments were performed with neural cells maintained in vitro for 7 days in 96-well plastic plates. After 9 days their survival significantly decreased, making subsequent analysis impossible. However, physiological recordings were performed using neurons that could be maintained on glass coverslips for 12 days and longer. To verify whether PSA would prevent excitotoxicity under these conditions, we performed an additional series of experiments and found that application of 5 M glutamate to hippocampal cultures maintained for 12 days in vitro induced cell death that could be prevented by colominic acid, Ro 25-6981 or APV (Fig. 7D).
To investigate the role of endogenous PSA in cell survival, hippocampal cultures were pretreated with endo-N for 3 h before exposure to glutamate. Endo-N pretreatment by itself increased cell death to a level similar to that observed after application of 5 M glutamate (Fig. 7E). Pretreatment of cultures with endo-N occluded the induction of further cell death by 5 M glutamate and its block by co-application of colominic acid (Fig. 7E). To investigate if survival-promoting effects of endogenous PSA are mediated by inhibition of NR2B-containing NMDARs, we tested whether co-application of Ro 25-6981 would reverse the effect of endo-N. In accordance with this notion, co-treatment of neurons with endo-N and Ro 25-6981 restored neural cell survival to control levels, which was the survival seen in the absence of additives (Fig. 7E).

DISCUSSION
Our study shows that both bacterially produced PSA (colominic acid) and eukaryotically produced PSA-NCAM specifically inhibit the NR2B subunit-containing NMDARs of cultured hippocampal neurons. Furthermore, PSA inhibits the activity of single synaptosomal NR2B-containing NMDARs reconstituted in lipid bilayers. Because 10 M ifenprodil occluded effects of PSA in hippocampal neurons, it is likely that PSA acts on NR1/NR2B receptors rather than on triheteromeric NR1/NR2A/NR2B receptors that are more resistant to ifenprodil (39). Effects of PSA are prominent at the rather low concentration of 1 g/ml that corresponds to 44 nM, assuming a chain length of 73 sialic acid residues that is the estimated average size of colominic acid used in this study. Because the inhibition of NMDARs by PSA is immediate and similar in hippocampal neurons and lipid bilayers, we take this as an indication that PSA may act directly on NMDARs. The same concentrations of PSA as used in the present study have been previously shown to potentiate purified AMPA receptors reconstituted in lipid bilayers and AMPA receptor-mediated currents in cultured astrocytes and immature hippocampal neurons (23). The fact that PSA inhibited NMDAR currents at lower but not at higher concentrations of glutamate suggests that PSA competes with glutamate in binding to positively charged amino acids in the S1 and S2 extracellular domains of NR2 subunits, which form the glutamate binding site of NMDARs (40). Computer modeling predicts that there are only 4 -6 NR2 subunit-specific amino acid residues exposed to the glutamate binding pocket of NR2 subunits (41). These residues are located at the edge of the glutamate binding pocket, indicating that only larger sized antagonists may provide subtypespecific inhibition of NMDARs. This notion is in line with our observation that only polymers but not monomers of sialic acid inhibited NR2B-containing NMDARs. Because inhibition of NMDARs by exogenous PSA was observed in endo-N-treated and NCAM-deficient neurons, it is not mediated by neuronal cell surface-expressed PSA-NCAM or NCAM. Although the data obtained, particularly in artificial lipid bilayers, suggest that PSA directly interferes with binding of glutamate to NMDARs, we cannot exclude the possibility that this influence is indirect, for instance, via interaction with a lipid membrane  near the receptor that could modify its configuration and, thus, indirectly affect ligand binding.
An increase in total, ifenprodil-sensitive and -insensitive NMDAR currents in NCAM-deficient neurons as compared with wild type neurons suggests that NCAM is involved in regulation of NR2B-and NR2A-containing NMDAR activity. Because endo-N pretreated neurons have normal NMDAR currents, this regulation appears to be PSA-independent but NCAM glycoprotein backbone-dependent. These results are supported by biochemical analyses showing that NR1 levels are increased in NCAMϪ/Ϫ brain homogenates, indicating that the overall expression of NMDA receptors in NCAMϪ/Ϫ brains is increased (21).
The observation that pretreatment of neurons with endo-N does not lead to changes in NMDAR currents demonstrates that only PSA added to neurons in the form of soluble PSA, as colominic acid or as recombinant PSA-NCAM, has access to NMDARs. Although we did not observe any influence of endo-N treatment on currents mediated by NR2B subunits in electrophysiological recordings, endo-N treatment promoted NR2B-mediated excitotoxicity. This difference between the electrophysiological and excitotoxicity assays can be accounted for by the following argument; soluble molecules are likely to be washed out in the perfusion chamber in which electrophysiological recordings are performed, whereas soluble NCAM remains present in the neural culture supernatant (42) and, thus, may interfere with glutamate-induced cell death in the closed volumes of small wells used for the excitotoxicity assay. Indeed, we were able to detect PSA by Western blotting using PSA-specific antibodies in the medium of cultures maintained for 7 days in vitro. 5 This observation is noteworthy because soluble PSA-NCAM has also been detected in brain tissue and cerebrospinal fluid (43,44). Furthermore, it is interesting in the context of the biological significance of our present observations that expression and cleavage of the extracellular domain of NCAM/PSA-NCAM is regulated by metalloproteinase activity (45) and activation of NMDARs and the plasmin/tissue plasminogen activator system (43,46). Thus, inhibition of NMDAR by PSA-NCAM released by neuronal activity from the cell surface may provide feedback to reduce NMDAR currents and excitotoxic damage of the activated neurons.
Our analysis of glutamate-induced excitotoxicity revealed that application of PSA prevents cell death, whereas removal of neuronal cell surface-expressed PSA promotes cell death, occluding the excitotoxic effects of glutamate. These results suggest that it is the same population of neural cells that is susceptible to 5 M glutamate and endo-N-induced cell death. It is noteworthy in this respect that removal of PSA with endo-N has previously been shown to reduce survival of cultured cortical neurons due to reduced brain-derived neurotrophic factor-mediated signaling (26). Supplementation of culture medium with an excess of brain-derived neurotrophic factor in these experiments was able to rescue survival of endo-N treated cortical neurons. Under our experimental conditions, however, the addition of brain-derived neurotrophic factor to endo-N-treated hippocampal neurons did not affect endo-N-induced cell death, 5 suggesting that other mechanisms are operant.
Excitotoxicity contributes to neuronal degeneration in many traumatic insults to the nervous system, such as ischemia, amyotrophic lateral sclerosis, and epilepsy (47). Excitotoxic activity of NR2B-containing NMDARs is likely to mediate selective neurodegeneration of striatal medium-sized spiny projection neurons in Huntington disease (48), neurotoxicity associated with alcohol-withdrawal (49), and neuronal cell death after transient cerebral ischemia (50). A recent study highlights the particular vulnerability of immature (1-week-old neurons, as in the present study) neurons to NMDA-induced toxicity, which is caused by activation of NR2B-but not NR2Acontaining NMDARs (51). Because PSA inhibits NR2B-containing NMDARs and is highly expressed during early development and up-regulated during synaptic activity (3,52), our present results suggest that PSA is well in place to prevent excitotoxic neuronal cell death during development and under pathological conditions, resulting in glutamate release, at least in cases when glutamate is accumulated in the extracellular space at low micromolar concentrations. These concentrations are physiological and found in normal and epileptic brains but may be far exceeded during transient cerebral ischemia (53)(54)(55)(56).