Introduction

Abstract Introduction Methods Results Discussion References

Functional NMDA receptors are composed of members from two subunit families, namely NR1 (eight different splice variants) and NR2A-D (reviewed by Mori and Mishina, 1995). The subunit family members show distinct distribution patterns in adult brain and during development (Kutsuwada et al., 1992; Laurie and Seeburg, 1994; Monyer et al., 1994; Mori and Mishina, 1995). Members of the NR1 family are expressed in all brain areas with a differential distribution of splice variants (Laurie and Seeburg, 1994; Nakanishi et al., 1992), whereas NR2 family members exhibit a more selective distribution. In cortex, NR2B subunits are expressed from late embryonic stages up to adulthood, whereas expression of NR2A subunits only becomes detectable at early postnatal stages and increases to adult levels within about 3 weeks of birth (Mori and Mishina, 1995; Portera-Cailliau et al., 1996; Wenzel et al., 1995; Williams et al., 1993). This expression pattern of NR2A and NR2B subunits is replicated during differentiation of cortical neurons in culture (Williams et al., 1993; Zhong et al., 1994).

Studies on different NMDA receptor subunit combinations expressed in oocytes or transfected cell lines as well as studies with native receptors have clearly indicated the existence of NMDA receptor subtypes with distinct pharmacologies (Farrant et al., 1994; Grimwood et al., 1996a, b; Lynch et al., 1995; Monyer et al., 1994; Priestley et al., 1994, 1995; Williams et al., 1993). NMDA receptors are activated by the co-agonists glutamate and glycine (Johnson and Ascher, 1987) and modulated via different sites including polyamine binding sites (Benveniste and Mayer, 1993; Traynelis et al., 1995). Ligands acting at either the glutamate or the glycine binding site show poor selectivity among different subtypes of NMDA receptors (Grimwood et al., 1996a, b; Kendrick et al., 1996; Priestley et al., 1994). In contrast, polyamines preferentially enhance the function of NMDA receptors containing the NR2B subunit in combination with NR1 splice variants lacking an N-terminal insert (Durand et al., 1993; Williams et al., 1994). The antagonist ifenprodil, acting at another allosteric site (Gallagher et al., 1996), also binds with high affinity to NMDA receptors containing the NR2B subunit (Gallagher et al., 1996; Lynch et al., 1995; Williams, 1993, 1995; Williams et al., 1993).

Overactivation of NMDA receptors plays a critical role in animal models of ischemic brain damage, and several different types of NMDA receptor blockers have attracted interest in recent years as neuroprotective compounds (Bullock et al., 1990; Gill et al., 1995; Gotti et al., 1988; Park et al., 1988; Sauer et al., 1993; Warner et al., 1995). In clinical trials, dosing of nonselective NMDA receptor blockers, e.g., glutamate-site antagonists or noncompetitive NMDA receptor channel blockers, is limited by adverse cardiovascular effects, hallucinations and agitation (for review see Muir and Lees, 1995), and accordingly, plasma levels that are protective in animal models of ischemic damage are hard to achieve in man. The atypical noncompetitive NMDA receptor blockers ifenprodil and eliprodil (Carter et al., 1988) were neuroprotective in focal ischemia (Gotti et al., 1988) and eliprodil was reported to be without the expected mechanism-related side-effects described above. The discovery that ifenprodil was selective for the NMDA receptor subtype containing the NR2B subunit (Williams, 1993) provided a possible explanation for these observations. In addition, the recently described activity-dependent nature of ifenprodil's block, which is different from voltage-dependent open-channel block, may also contribute to its attractive in vivo profile (Kew et al., 1996). Eliprodil and ifenprodil, however, are not ideal drugs because they show little selectivity for NMDA receptors over other recognition sites (see Biton et al., 1994; Chenard et al., 1991; McCool and Lovinger, 1995).

In this paper we describe the in vitro profile of Ro 25-6981, a high-affinity, selective, activity-dependent blocker of NMDA receptors containing NR2B subunits, with potent neuroprotective effects in vitro. It is structurally related to ifenprodil. Part of this paper was published in abstract form (Fischer et al., 1996; Kemp et al., 1996; Trube et al., 1996).

	Material and Methods

Abstract Introduction Methods Results Discussion References

Materials

Drugs used in binding, electrophysiological and toxicity experiments were from the following sources: D-AP-5 (Tocris Cookson, Bristol, U.K.), flunarizine (RBI, Natick, MA), ifenprodil (RBI), MK-801(dizocilpine, RBI), NBQX (Tocris Cookson), TCP (RBI), TTX (Latoxan, Rosans, France), veratridine (RBI). Ro 25-6981 was synthesized in the Chemistry Department of Hoffmann-La Roche (Nutley, NJ). All other chemicals were obtained from Sigma (St. Louis, MO) or Fluka (Buchs, Switzerland).

Binding Experiments

Binding experiments with rat forebrain membranes and the radioligand ³H-MK-801 (dizocilpine) were performed as described (Ransom and Stec, 1988). MK-801 is a potent, nonselective open-channel blocker of NMDA receptors (Wong et al., 1986, 1988). Membranes were prepared from whole brain of 150- to 200-g male rats, without cerebellum and medulla oblongata by homogenization (Ultra-Turrax maximum speed, 30 s at 4°C in 50 volumes of cold 50 mM Tris-HCl with 10 mM ethylenediaminetetraacetic acid, pH 7.4; wet weight per volume) and centrifugation (48,000 × g for 10 min). The pellet was rehomogenized twice and frozen at 80°C in 35-ml fractions for at least 16 h and not more than 2 weeks. For binding experiments, the membranes were washed three times (homogenization in 25 volumes of cold 5 mM Tris-HCl (pH 7.4) with an Ultra-Turrax at maximum speed for 30 s). The final pellet was rehomogenized in 25 volumes of buffer (original wet weight) and used as such in the assay. The final protein concentration in the assay was 200 µg/ml. The incubation was performed in the presence of 1 nM added glutamate, glycine and spermidine. The ligand, [³H]MK-801, (+)-[3-³H(N)], (NET-972; NEN, Boston, MA), 20 Ci/mmol, was used at 5 nM final concentration. Nonspecific binding was determined in the presence of 100 µM TCP. After 2 h of incubation at room temperature, the suspension was filtered (Whatman GF/B, soaked in 0.1% polyethylenimine for 2 h) and washed five times with 3 ml of cold 5 mM Tris-HCl (pH 7.4). The filters were counted with 10 ml of Ultima-gold (Packard, Rockville, MD) in a Tri-Carb 2500 TR scintillation counter after agitation. DPM values were transformed to % of specific binding. Each experiment was repeated three to four times.

Electrophysiology

Cloned NMDA receptors expressed in Xenopus oocytes. cDNA clones coding for the subunits NR1C (also termed NMDAR1-2a or NR₀₀₁) and NR2A of the rat NMDA receptor were isolated from a rat brain gt11 cDNA library as published elsewhere (Sigel et al., 1994). The clone for the subunit NR2B was kindly provided by S. Nakanishi (Kyoto, Japan). The cDNA encoding the NR1F subunit (also termed NMDAR1-2b or NR₁₀₁, a splice variant of NR1C containing a 63-base-pair insertion near the N-terminus) was engineered by replacing a 354-bp Eco47III-EcoNI fragment of NR1C by a corresponding polymerase chain reaction-amplified fragment containing the insertion. The cDNAs were transcribed, capped and poly(A⁺)-tailed as described previously (Malherbe et al., 1990). Oocytes of South African frogs (Xenopus laevis) at maturation stage V to VI were used for expressing the subunit combinations NR1C & NR2A, NR1C & NR2B or NR1F & NR2B. Approximately 3 fmol of a 1:1 mixture of the respective mRNA species were injected into every oocyte. Three days later the oocytes were defolliculated by collagenase treatment and the electrophysiological experiments were done on the following 2 days as described by Methfessel et al. (1986). During the experiments the oocytes were superfused at room temperature by a salt solution containing (in mM): NaCl, 90; KCl, 1; BaCl₂, 1; HEPES, 5 (pH 7.4, 22°C). The membrane potential was set to 80 mV by a two-microelectrode voltage-clamp (TurboTEC-05, NPI, Tamm, Germany) and the receptors were activated by applying a mixture of the co-agonists L-aspartate and glycine. Agonist concentrations close to the respective EC₅₀ values were chosen for each subunit combination to account for the different agonist sensitivities of the various NMDA receptor subtypes: 70 µM L-aspartate and 2.5 µM glycine for NR1C & NR2A, 15 µM L-aspartate and 0.2 µM glycine for NR1C & NR2B, 40 µM L-aspartate and 0.4 µM glycine for NR1F & NR2B. The agonists were applied for 15-s intervals once every 2.5 min, and the amplitude of the evoked current was measured at the end of each application. Ifenprodil and Ro 25-6981 were added in a cumulative fashion to both the basal and the agonist-containing saline. Five to ten oocytes were tested for each compound and subunit combination.

Cortical neurons. Cells were used after 5 to 14 days in vitro (see below). Whole-cell voltage-clamp recordings were performed as described previously (Kew et al., 1996). Cells were continuously superfused at room temperature with saline containing (in mM): NaCl, 149; KCl, 3.25; CaCl₂, 2; MgCl₂, 2; HEPES, 10; D-glucose, 11 with pH 7.35 and an osmolarity of 350 mOsm adjusted with sucrose. Patch pipettes had a resistance of approximately 2 to 4 megaohms when filled with a solution containing (in mM): CsF, 120; CsCl, 10; ethyleneglycol-bis(-aminoethyl ether)-N,N,N,N-tetraacetic acid, 11; CaCl₂, 0.5; HEPES, 10 with pH 7.25 and osmolarity adjusted to 330 mOsm with sucrose. Whole-cell recordings were made at a holding potential of 60 mV, unless stated otherwise, with an Axopatch 200A amplifier (Axon Instruments, Foster City, CA). Pipette seal resistances were typically >10 gigaohms and pipette capacitance transients were minimized both before and after membrane breakthrough. No series resistance compensation was applied. Drugs were diluted from concentrated stock into a modified version of the superfusion solution lacking MgCl₂ and including 30 µM glycine. Drugs were applied to cells by fast superfusion from double- or triple-barreled capillary assemblies (internal capillary diameter, 320 µm). Currents were filtered (cut-off frequency, 5 kHz), digitized with a Digidata 1200 (Axon Instruments) and captured on-line by pCLAMP6 software (Axon Instruments).

Cell Cultures

Cortical neurons from 17- to 18-day-old rat embryos were prepared as described for hippocampal neurons (Möckel and Fischer, 1994). They were plated on confluent astrocyte feeder layers either on glass coverslips (15 mm diameter) or in 24-multiwell plates (Nunc, Roskilde, Denmark) with cell densities of either 50,000 cells or 150,000 cells/cm² for electrophysiological or toxicity experiments, respectively. Cells were cultured in DMEM (GIBCO, Grand Island, NY) supplemented with 10% horse serum (Boehringer, Mannheim, Germany) in a 5% CO₂ in air atmosphere in a humidified incubator at 37°C. After 5 DIV, cells were treated with 10 µM cytosine arabinoside (Fluka). After 7 DIV one third of the medium of the low-density cultures on coverslips was exchanged, and the cell culture medium of high-density cultures in 24-multiwell plates was replaced completely by DMEM supplemented with 5% horse serum and 10 µM D-AP-5. The cultures were used for the experiments between 5 and 14 DIV.

Toxicity Experiments

Glutamate toxicity. Cortical neurons cultured for 11 to 12 DIV in 24-multiwell plates were washed once with BME (GIBCO) and incubated for 16 h in 300 µl/well of BME supplemented with 18 mM glucose with or without addition of 300 µM glutamate plus 1 µM glycine and various concentrations of test compounds.

Kainate toxicity. Cortical neurons cultured for 11 to 12 DIV in 24-multiwell plates were washed once with BME and incubated for 16 h in 300 µl/well of BME supplemented with 18 mM glucose with or without addition of 500 µM kainate plus 100 µM D-AP-5 and various concentrations of test compounds. The NMDA receptor blocker D-AP-5 was added to eliminate an indirect contribution to toxicity of NMDA receptors by kainate-induced glutamate release.

Combined oxygen and glucose deprivation. Cortical neurons cultured for 11 to 12 DIV in 24-multiwell plates were washed twice with a BSS (pH 7.2) containing (in mM): NaCl, 120; KCl, 5.4; CaCl₂, 1.8; MgCl₂, 0.8; Tris-HCl, 7.5; NaHCO₃, 17.5. Cultures were washed once with oxygen-free BSS supplemented in addition with 15 mM sorbitol, placed in an anaerobic chamber (<0.3% O₂, 85% N₂, 10% H₂ and 5% CO₂) and washed again. After an incubation for 60 min at 37°C the cells were washed once with BME supplemented with 18 mM glucose and returned for 20 h in the normal incubator in the same medium. Test compounds were added during OGD as well as the recovery phase.

Quantification of cell death. Neuronal cell death was assessed qualitatively by phase-contrast microscopy and quantified by determination of LDH activity in the cell culture supernatant (Koh and Choi, 1987). At least three independent experiments were performed in quadruplicate. Arithmetic means of the quadruplicates were calculated for each experiment.

Intracellular Free Na ⁺ and Ca⁺⁺ Measurements

Cells cultured for about 12 to 14 DIV in 24-multiwell plates were either loaded with fura-2AM (Grynkiewicz et al., 1985) or with SBFI-AM (Minta and Tsien, 1989) for measurement of calcium or sodium, respectively. For dye loading, cells were incubated at 37°C with 20 µM fura-2AM (Molecular Probes, Eugene, OR) for 30 min or with 17 µM SBFI-AM (Molecular Probes, Eugene, OR) plus 100 µM D-AP-5 for 4 h in the presence of 0.025% Pluronic F-127 (Molecular Probes, Eugene, OR) in BSS with 25 mM glucose. The cultures were then washed with BSS and de-esterification was allowed to proceed for at least 30 min at 37°C before starting the experiments. For calcium imaging the cells were stimulated at room temperature for 10 min with 50 mM potassium in BSS (partial replacement of NaCl by KCl) supplemented with 25 mM glucose with or without test compounds. For sodium imaging the cells were stimulated in BSS at room temperature for 10 min with 10 µM veratridine, an alkaloid known to prevent Na⁺ channel inactivation (Catterall, 1980). Several drugs were used for validation of these tests, namely, TTX as a selective blocker of Na⁺ channels (Catterall, 1980); flunarizine as a potent, mixed blocker of Na⁺ and Ca⁺⁺ channels (Akaike et al., 1989; Pauwels et al., 1991); and MK-801 as a specific NMDA receptor blocker. Experiments were performed in duplicate. Imaging measurements were made on an inverted microscope with a long distance 40× dry objective (Axiovert 405 M, Zeiss, Oberkochen, Germany). A cooled CCD camera system (CH-250, Photometrics, Tucson, AZ), as described (Müller and Conner, 1991), was controlled by a Macintosh IIci computer to acquire image pairs (one per well) at 340 and 380 nm excitation wavelengths with dark correction. Exposure times were 400 ms. The intrinsic fluorescence in cells not dye-loaded was less than 5% and did not contribute a significant error to the measurements. Ratio values were calculated (Grynkiewicz et al., 1985). At least 20 individual cells were analyzed per image for changes in the ratio values and the mean response was calculated for every well. Mean values were calculated from the duplicates for every experiment.

Data Analysis

In the binding assays the concentration-response relations from individual experiments were fitted by the function:

(1)

with y, labeled receptor; P_low, proportion of low-affinity binding sites; K_dlow, affinity of low-affinity binding sites; K_dhigh, affinity of high-affinity binding sites.

Relations from the functional experiments were fitted by the function:

(2)

with y, response; x, antagonist concentration; y, response at saturating antagonist concentration; IC₅₀, 50% inhibitory concentration; H, Hill coefficient. For the descriptive statistics of IC₅₀, K_dlow and K_dhigh we calculated the decadian logarithms of the values from the individual experiments and then arithmetic mean values (M_log) and standard deviations (S.D._log) from the logarithms. Values in the text are expressed as M (l, h), where M, l and h are the antilogarithms of M_log, M_log  S.D._log and M_log + S.D._log, respectively. The statistical analysis of all other parameters was done as usual without conversion to logarithms and results are given as arithmetic means ± S.D.

	Results

Abstract Introduction Methods Results Discussion References

Effects of Ro 25-6981 on ³H-MK-801 binding. Ro 25-6981 (fig. 1) showed a biphasic concentration-effect relationship in inhibiting ³H-MK-801 binding to rat forebrain membranes (fig. 2) similar to ifenprodil, a compound known to preferentially inhibit NMDA receptors containing the NR2B subunit (Williams, 1993). The proportion of high-affinity binding sites was similar for Ro 25-6981 and ifenprodil, being 55 ± 13% and 57 ± 12%, respectively, of total specific ³H-MK-801 binding. However, Ro 25-6981 showed a greater separation between the two binding sites than ifenprodil. Mean K_d values obtained by nonlinear regression analysis were 0.003 µM and 0.105 µM for the high-affinity binding and 149 µM and 62 µM for the low-affinity binding of Ro 25-6981 and ifenprodil, respectively (table 1), which suggests a higher potency and selectivity of Ro 25-6981 for a subpopulation of NMDA receptors. MK-801 showed the expected monophasic inhibition (data not shown) with a K_d value of 0.034 (0.032, 0.035) µM.

View larger version (6K): [in this window] [in a new window]   Fig. 1.   Structure of Ro 25-6981.

View larger version (20K): [in this window] [in a new window]   Fig. 2.   Inhibition of 3H-MK-801 binding to rat forebrain membranes by Ro 25-6981 (open circles) and ifenprodil (squares). A biphasic function (see "Materials and Methods") was fitted to the data to describe the concentration dependence of inhibition. Values represent means from triplicates.

Blockade of NMDA receptors by Ro 25-6981: electrophysiological characterization. Ifenprodil binds preferentially to NMDA receptors containing the NR2B subunit (Williams, 1993). To confirm that this was also the case for Ro 25-6981 we expressed either the NR2B or the NR2A subunit together with the NR1 subunit in Xenopus oocytes and measured the membrane current induced by L-aspartate plus glycine. Ro 25-6981, at a relatively low concentration (0.01 µM), slowly inhibited about 50% of the current in oocytes which had been injected with NR2B and NR1C mRNAs (fig. 3, A and B). At a 10-fold higher concentration the effect was more rapid and more pronounced (fig. 3B). Similar experiments were performed with oocytes expressing the NR2B and NR1F or the NR2A and NR1C subunits. The NR1F splice variant of NR1 differed from NR1C by the presence of a 21-amino-acid insert close to the N-terminus. This insert is known to abolish the potentiating effect of polyamines on NMDA receptors (Durand et al., 1993; Williams et al., 1994), however, its presence only slightly reduced the sensitivity for Ro 25-6981 and ifenprodil (fig. 3C). The IC₅₀ values of ifenprodil for antagonism of the NR1C & NR2B or NR1F & NR2B receptors were about 25-fold higher than those of Ro 25-6981 (fig. 3C, table 1). Oocytes which had been injected with the NR1C and NR2A mRNAs were much less sensitive for both Ro 25-6981 and ifenprodil than those expressing the NR2B subunit (table 1).

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Inhibition of cloned NMDA receptors expressed in Xenopus oocytes. (A) Representative current traces recorded from an oocyte injected with mRNA encoding the NR1C and NR2B subunits. Currents were evoked by 15-s stimuli of L-aspartate (15 µM) and glycine (0.2 µM). The last response before the addition of Ro 25-6981 (ctrl.) and selected responses taken 1, 6 and 19 min after adding the compound (0.01 µM) are shown. (B) Time course of NMDA receptor inhibition. Mean current amplitudes measured before and during the exposure to Ro 25-6981 (0.01 µM and 0.1 µM) are plotted versus time. Current amplitudes were expressed as a percentage of the last control response (filled circle). Error bars indicate S.D. from five oocytes. Coexpression of NR1C and NR2B. (C) Comparison of the effects of ifenprodil (squares) and Ro 25-6981 (circles) applied to oocytes expressing the subunits NR1C and NR2B (open symbols, lines) or NR1F and NR2B (filled symbols, dashed lines). Sigmoidal curves were obtained from equation 2 ("Materials and Methods") by nonlinear regression analysis; the mean IC50 values are given in table 1. Estimates of the parameters H (Hill coefficient) and y (base line) ranged from 0.8 to 1.1 and 4 to 7%, respectively.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Activity-dependent action of Ro 25-6981 on NMDA receptors. NMDA receptors were stimulated in 10- to 14-day-old cultured cortical neurons by either 100 µM NMDA (A) or 1 µM NMDA (B) at a saturating glycine concentration of 30 µM. Currents were measured by whole-cell voltage-clamp recordings. The effect of 1 µM Ro 25-6981 depended on the level of NMDA receptor activation and as such potentiated currents elicited by very low NMDA concentrations whereas profoundly inhibiting currents elicited by higher NMDA concentrations. Representative recordings are shown. Application of NMDA and Ro 25-6981 is indicated by bars.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   Comparison of the kinetics of inhibition of steady-state inward currents elicited by either 10 or 100 µM NMDA after fast application of Ro 25-6981. NMDA receptors were stimulated in 10- to 14-day-old cultured cortical neurons. Currents were measured by whole-cell voltage-clamp recordings. Single exponential curves were fitted to the digitized data and yielded on-rate time constants for the block of the steady-state current as shown. Ro 25-6981 (3 µM) showed a faster onset and higher efficacy of NMDA receptor blockade at the higher level of receptor activation. Current amplitudes were normalized to facilitate comparison. Representative recordings are shown.

Effects of Ro 25-6981 on neurotoxicity in vitro: comparison with ifenprodil and MK-801. The in vitro neuroprotective properties of Ro 25-6981 and ifenprodil were compared in two toxicity models with cultured cortical neurons which mimic critical aspects of ischemic brain damage, namely, exposure to glutamate (fig. 6) or OGD (fig. 7). Astrocytes were not sensitive to these toxicity tests, as verified by morphological observation, but almost all neurons (>95%) died. Ro 25-6981 and ifenprodil protected the neurons in a concentration-dependent manner. Ro 25-6981 was more potent than ifenprodil in both toxicity tests. The IC₅₀ values calculated from curve fits for Ro 25-6981 (n = 3) and ifenprodil (n = 4) were 0.4 µM and 3.5 µM in the glutamate toxicity test and 0.04 µM and 3.2 µM in the OGD test (n = 3 for both compounds), respectively (table 1). The IC₅₀ values for MK 801 (n = 3) were 0.2 (0.17, 0.32) µM and 0.06 (0.04, 0.09) µM for glutamate toxicity and OGD, respectively.

View larger version (18K): [in this window] [in a new window]   Fig. 6.   Protection against glutamate toxicity in cultured cortical neurons. The concentration-dependent protection of Ro 25-6981 (circles) and ifenprodil (squares) is shown. After 12 days in vitro cultured cortical neurons were exposed to glutamate (300 µM) and glycine (1 µM) with or without addition of drugs. For quantification of neuronal cell death LDH was measured in the culture supernatant after 16 h. The difference between control cultures and glutamate-exposed cultures was taken as 100% LDH release. Each experiment was performed in quadruplicate. Means ± S.D. are shown for Ro 25-6981 (n = 5) and ifenprodil (n = 4).

View larger version (19K): [in this window] [in a new window]   Fig. 7.   Protection against combined oxygen and glucose deprivation in cultured cortical neurons. The concentration-dependent protection of Ro 25-6981 (circles) and ifenprodil (squares) is shown. After 12 days in vitro cultured cortical neurons were deprived of oxygen and glucose for 60 min with or without addition of drugs. For quantification of neuronal cell death LDH was measured in the culture supernatant after a recovery period of 16 h. The difference between control cultures and deprived cultures was taken as 100% LDH release. Each experiment was performed in quadruplicate. Means ± S.D. (n = 3) are shown.

Effects of Ro 25-6981 on Na⁺ and Ca⁺⁺ channels. The effects of Ro 25-6981 on Na⁺ and Ca⁺⁺ channels were quantified with imaging methods with ion-sensitive fluorescent dyes, by measuring relative increases of cytoplasmic ion concentrations after stimulation of cortical neurons for 10 min with 10 µM veratridine or 50 mM K⁺, respectively. Compared with the potencies in blocking NR2B-containing NMDA receptors in the different tests (see table 1), under these conditions, Ro 25-6981 was only a weak blocker of Na⁺ channels [IC₅₀ = 18 (16, 21) µM; n = 3] and almost inactive at Ca⁺⁺ channels (23 ± 20% block at 100 µM, n = 3). Ifenprodil was also a weak blocker of Na⁺ channels [IC₅₀ = 17.8 (7.7, 41.3) µM; n = 3] and was less active in blocking Ca⁺⁺ channels (48 ± 17% block at 100 µM; n = 4). For comparison, MK-801 at up to 30 µM did not significantly affect the increase in intracellular Na⁺ (stimulation with veratridine) or Ca⁺⁺ (stimulation with K⁺), which indicates that activation of NMDA receptors did not significantly contribute. TTX and flunarizine blocked the Na⁺ increase in a concentration-dependent manner with IC₅₀ values of 0.02 (0.015, 0.035) µM (n = 6) and 2.5 (1.8, 3.3) µM (n = 3), respectively. TTX (1 µM) did not block the increase in intracellular Ca⁺⁺, whereas flunarizine blocked the increase with an IC₅₀ value of 2.8 (2.1, 3.5) µM (n = 3).

	Discussion

Abstract Introduction Methods Results Discussion References

The results show that Ro 25-6981 is the most potent and selective blocker of NMDA receptors containing the NR2B subunit described to date. Its mode of action at NMDA receptors seems to be similar to that of ifenprodil. The slow onset of and recovery from receptor blockade produced by Ro 25-6981 is in agreement with its high affinity in binding experiments as well as its high potency in blocking NMDA receptors in electrophysiological experiments. Ro 25-6981 potently protected cultured cortical neurons against glutamate toxicity as well as OGD.

The NMDA receptor subtype selectivity of Ro 25-6981 was first indicated by the biphasic inhibition of ³H-MK-801 binding to adult rat forebrain membranes (fig. 2) which contain different NMDA receptor subtypes. In situ hybridization experiments (Kutsuwada et al., 1992; Laurie and Seeburg, 1994; Monyer et al., 1994) and immunostaining (Portera-Carllian et al., 1996; Wenzel et al., 1995) of brain slices indicate the presence of different NR1 splice variants as well as predominantly NR2A and NR2B subunits in cortical and hippocampal areas which mostly contribute to this membrane preparation. Competitive glutamate and glycine site-directed antagonists show a poor selectivity, differences within one order of magnitude, among different subtypes of NMDA receptors (Grimwood et al., 1996a, b; Priestley et al., 1995). A biphasic inhibition of ³H-MK-801 binding was first described for ifenprodil (Reynolds and Miller, 1989; Williams, 1993), which has been identified as a NR2B selective blocker with about 400-fold higher selectivity for NR2B than for NR2A (Lynch et al., 1995; Williams, 1993; Williams et al., 1993). The high-affinity portion of the biphasic inhibition (about 55%) is similar for ifenprodil (Williams et al., 1993) and for Ro 25-6981. However, Ro 25-6981 is approximately 25-fold more potent than ifenprodil in blocking NMDA receptors containing NR2B subunits, whereas similar IC₅₀ values were found for receptors containing NR2A (table 1). Based on IC₅₀ values of 0.009 µM and 52 µM (table 1) the selectivity of Ro 25-6981 for NR2B over NR2A is about 5000-fold. The effect on receptors containing NR2C or NR2D subunits has not yet been tested, but it should be noted that ifenprodil does not block these subtypes of NMDA receptors (Williams, 1995).

The mode of action of Ro 25-6981 at NMDA receptor subtypes seems to be distinct from that of open-channel blockers, but similar to that of ifenprodil, which was recently characterized as an "activity-dependent" blocker (Kew et al., 1996). Ifenprodil binds with 35- and 50-fold higher affinities to the activated and desensitized states of the receptor, respectively, in comparison with the resting state. It increases the affinity of glutamate (measured at saturating glycine concentrations), potentiates NMDA receptor currents at very low agonist concentrations and reduces markedly the open probability of the channel without completely blocking the receptor. The potentiation of NMDA receptor currents at very low agonist concentrations by Ro 25-6981 (fig. 4), the marked inhibition of NMDA-evoked currents at higher agonist concentrations (figs. 4 and 5), the increased efficiency and the relatively faster onset of block at higher agonist concentrations (fig. 5) all support a similar mode of action for Ro 25-6981. The block of NMDA receptors by Ro 25-6981 appears to be voltage-independent (table 2), which together with its potentiating effects at very low NMDA concentrations suggests that it is not an open-channel blocker.

The neuroprotective potential of Ro 25-6981 was characterized in two different in vitro toxicity models with cortical neurons, namely (1) permanent exposure to glutamate for 16 h and (2) combined oxygen and glucose deprivation for 60 min. In both models nonselective NMDA receptor blockers are highly protective (Choi et al., 1988; Goldberg and Choi., 1993), and ifenprodil is protective against short-term NMDA and glutamate exposure of cortical neurons (Graham et al., 1992; Shalaby et al., 1992; Tamura et al., 1993). In agreement with its higher potency in blocking NR2B-containing NMDA receptors (fig. 2, table 1), Ro 25-6981 was found to be more potent as a neuroprotectant in these tests than ifenprodil (figs. 6 and 7). The higher potencies of Ro 25-6981 and MK-801 for protection against OGD in comparison with glutamate toxicity might be explained by the different strength of the toxic stimulus, which might require a higher percentage block of NMDA receptors to achieve protection against glutamate toxicity. The reason for the relatively low potency of ifenprodil in protecting against OGD is not clear. The high efficacies of protection against permanent glutamate toxicity (80%) and combined oxygen and glucose deprivation (90%) are in good agreement with the preferential expression of NR2B rather than NR2A subunits in cortical neurons during the first 2 weeks in culture (Zhong et al., 1994), which parallels the expression profile of these subunits during early postnatal development of cortical structures (Mori and Mishina, 1995; Williams et al., 1993; Zhong et al., 1995). In adult brain NR2A expression becomes more prominent in cortical and other brain areas (Mori and Mishina, 1995; Wenzel et al., 1997). The formation of heteromeric NMDA receptors containing both NR2A and NR2B subunits is supported by immunoprecipitation experiments (Sheng et al., 1994). However, the portion of these heteromeric receptors is under debate (Blahos Wenthold, 1996; Luo et al., 1997). Ifenprodil may block heteromeric receptors with an affinity similar to those containing NR2B as the sole NR2 subunit (Hess et al., 1996). If the biphasic block of ³H-MK-801 binding shown in fig. 2 is predictive for block of NMDA receptors in adult rat cortex, about 55% could be blocked with high affinity.

The lack of protection with Ro 25-6981 in the kainate toxicity test strongly supports its selectivity for NMDA receptor subtypes among ionotropic glutamate receptors. Blockade of other ion channels such as voltage-operated sodium or calcium channels does not appear to contribute to the neuroprotective potential of Ro 25-6981. It blocked Na⁺ channels only weakly in comparison with its potency in blocking NR2B-containing NMDA receptors (at least 50-fold less potent). Effects on Ca⁺⁺ channels were assessed by measuring the cytoplasmic increase of Ca⁺⁺ after stimulation of cortical neurons for 10 min with 50 mM K⁺. Under these conditions Ca⁺⁺ influx into cells may occur predominantly through noninactivating L-type channels. Ro 25-6981 was several hundredfold less potent in blocking these channels in comparison with its potency in blocking NR2B-containing NMDA receptors. Compared with its potency at NMDA receptor subtypes, ifenprodil was also only a weak blocker of Ca⁺⁺ channels, in agreement with electrophysiological studies on cortical and hippocampal neurons (Netzer et al., 1992; Church et al., 1994).

In summary, our results show that Ro 25-6981 is a high-affinity NMDA receptor subtype selective blocker with preference for NR2B subunits. It lacks significant activity at kainate/AMPA receptors, Na⁺ and Ca⁺⁺ channels at concentrations showing maximal protection in neurotoxicity tests. This selectivity together with its activity-dependent mechanism of action make Ro 25-6981 an attractive neuroprotectant. For a therapeutically useful compound NMDA receptors should be blocked markedly only at unphysiologically high levels of activation as occurs, e.g., during brain ischemia. Notably, Ro 25-6981 was found to be protective in a model of permanent focal brain ischemia in rats without marked side-effects (Fischer et al., 1996).