Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Some neuropathological states, like epilepsy, are characterized by excessive activation of receptors and cells, in terms of frequency or duration. Periods of excessive activation, prolonged ischemic depolarization, or multiple spreading depression-like periods also characterize ischemia. Some of the most useful anticonvulsants, like phenytoin, are uncompetitive, or use-dependent antagonists (Meldrum, 1993). Use-dependent compounds can only access their site of action when the receptor/channel is in an active state, resulting in a delay between receptor activation and the antagonist-induced blockade. Therefore, they are less likely to block short periods of activation, which typify normal activity. This characteristic of use-dependent compounds probably accounts for their clinical utility to block convulsant episodes without severely affecting normal function.

Antagonists of the NMDA type of excitatory amino acid receptors are neuroprotective in the in vitro and in vivo models of ischemia (Kemp et al., 1987; Olney et al., 1989). In addition, cerebral ischemia has been associated with an elevation of extracellular concentrations of excitatory amino acids (Benveniste et al., 1984). However, the clinical use of NMDA antagonists has been hampered by the occurrence of side effects that appear to be a class effect (Small and Buchan, 1997). Hallucinations and dysphoria of the type produced by PCP appear to be due to blockade of normal activity mediated by NMDA receptor activation.

NMDA receptor antagonists that act at the PCP site produce a use-dependent block, acting only after the receptor enters the open state (Huettner and Bean, 1988). Unexpectedly, many of these use-dependent antagonists, like MK-801, also have no therapeutic value due to PCP-like neurobehavioral side effects (Leppik et al., 1988). Many of these compounds have high affinity for the receptor, whereas some lower affinity derivatives have reduced side effects. It was hypothesized that the lack of safety of high-affinity compounds was due to the fact that the concentrations needed to reach steady-state block rapidly enough to be therapeutic (via mass action) are much higher than those that already completely block receptor activity (Rogawski, 1993). For the lower affinity compounds, the concentrations that only partially modulate receptor activity are similar to those needed to achieve rapid block. In support of this hypothesis, memantine (Muller et al., 1995; Parsons et al., 1995), amantadine (Parsons et al., 1995), and ADCI (Rogawski et al., 1991, 1995), all of which have relatively low affinity (Chen et al., 1992; Parsons et al., 1993, 1995), lack serious side effects.

However, a comparison of a series of uncompetitive NMDA receptor antagonists for efficacy as antiepileptic drugs demonstrated that some lower affinity antagonists actually had worse therapeutic indices (Parsons et al., 1995), suggesting that other factors are also important. Trapping, a feature of many use-dependent NMDA receptor antagonists, may account for these neurobehavioral side effects. Antagonists such as MK-801, ADCI, memantine, amantadine, ketamine, and AR-R15896AR all exhibit trapping (Jones and Rogawski, 1992; Blanpied et al., 1997; Chen and Lipton, 1997; Mealing et al., 1997, 1999; Lanthorn et al., 2000; Sobolevsky and Yelshansky, 2000). Trapping occurs when agonist is removed while antagonist is present, and the antagonist remains bound to the receptor. During subsequent receptor activation, the trapped antagonist is already present, blocking activity immediately. Thus, a trapped antagonist functionally loses its use-dependence and produces a tonic block. With compounds like MK-801 and ADCI, trapping appears to be complete (Jones and Rogawski, 1992). However, in the case of the clinically safer compound, memantine, about one sixth of the blocked channels release, rather than trap, the blocker (Blanpied et al., 1997). We have recently reported significant differences in the degree of trapping among three low-affinity, use-dependent NMDA receptor antagonists that possess intermediate kinetics (Mealing et al., 1999). Ketamine, which induces psychotomimetic side effects (Ginski and Witkin, 1994; Krystal et al., 1994), showed 86% trapping, whereas memantine and AR-R15896AR, which have improved therapeutic safety profiles (Parsons et al., 1995; Palmer et al., 1996, 1997), trapped only 71% and 54%, respectively. This reduction in trapping, from ketamine to memantine to AR-R15896AR, is in the same order as that for reduction of severity of side effects as measured by behavioral and cognitive tests in rodents. Therefore, use-dependent NMDA receptor antagonists that exhibit less trapping may provide safer compounds for therapeutic use.

The purpose of this study was to examine a series of structurally related compounds (analogs of AR-R15896AR) to see if differences could be found in the degree of trapping block. We then looked for correlations between these differences and the physical and structural characteristics of the compounds.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Compound Selection. Starting from a library of 131 compounds, a diverse subset of clusters of compounds structurally related to the NMDA receptor antagonist, AR-R15896AR (formerly called ARL 15896AR or FPL 15896AR), were generated. The diversity metrics used for creating these clusters were two-dimensional fingerprints, molecular weight, and dipole total. These parameters were calculated from the compounds' chemical structures using SYBYL 6.7 software (Tripos, Inc., St. Louis, MO). LogP and pK_a values, computed using ACD Software (Advanced Chemistry Development, Inc., Toronto, Ontario, Canada), were also included in the cluster creation criteria. Clusters were generated using Wards clustering algorithm (JMP software; SAS Institute, Inc., Cary, NC). Twenty-one compounds, each one selected arbitrarily from a different cluster, were tested to determine the degree of trapping they exhibited. Compounds were not selected from all clusters generated because of their unsuitability (for example, their weak antagonism). Two additional compounds, AR-R15808 and AR-R26952, were subsequently chosen to examine the impact of specific structural features (the presence of two basic nitrogen atoms) on trapping. All compounds tested were previously synthesized during the development of AR-R15896AR.

Chemicals and Reagents. Dulbecco's phosphate-buffered saline, HEPES, Eagle's minimal essential medium (MEM), poly-L-lysine, tetrodotoxin, and strychnine hydrochloride were purchased from Sigma Chemical (St. Louis, MO). Heat-inactivated fetal bovine serum was purchased from Life Technologies (Gaithersburg, MD), and heat-inactivated horse serum was purchased from Hyclone Laboratories, Inc. (Logan, UT). NMDA, (±)-2-amino-5-phosphonopentanoic acid, and ketamine were purchased from RBI/Sigma (Natick, MA). EGTA was purchased from Fluka Chemical Corp. (Ronkonkoma, NY). NMDA receptor antagonists were synthesized by AstraZeneca USA (Worcester, MA).

Cell Culture. Rat cortical neurons isolated from E18 fetuses were grown in primary culture as previously described (Black et al., 1995). Timed-pregnant Sprague-Dawley rats were purchased from Charles River Canada (St. Constant, PQ, Canada). After killing the mother by cervical dislocation under halothane anesthesia, the fetuses were removed from the uterus on day E18, their brains removed and placed in ice-cold phosphate-buffered saline, and the cortices isolated. The cortical neurons were dispersed by triturating with a 10-ml pipette, and the cells were centrifuged at 250g for 5 min at 4°C. The cells were gently resuspended in plating medium and viable cells, as determined by trypan blue exclusion, were counted. The cells were then plated at 10⁵ cells/cm² on poly-L-lysine-coated 35 mm culture dishes (Nunc, Roskilde, Denmark) in 2 ml of plating medium at 37°C in an atmosphere of 5% CO₂/95% air. Plating medium consisted of 80% MEM, with 20 mM glucose, 10% heat-inactivated fetal bovine serum, and 10% heat-inactivated horse serum containing 2 mM L-glutamine. Since the cultures contained both neurons and glial cells, they were treated with 15 mg/ml of 5-fluoro-2'-deoxyuridine and 35 mg/ml on day 5 to minimize glial cell proliferation. On day 6, half of the medium was removed and replaced with growth medium consisting of 90% MEM and 10% horse serum. Experiments were performed on cultures after 14 to 19 days in vitro.

Whole-Cell Recording. Cortical neurons grown on 35-mm culture dishes were mounted on the stage of an inverted Zeiss microscope equipped with Hoffman modulation optics in a perfusion system flowing continuously at 1 ml/min at 22°C. The bathing solution contained 140 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 10 mM HEPES, and 3 mM glucose at pH 7.4. Bathing solutions also contained 1 µM tetrodotoxin to block Na⁺ currents, 10 µM glycine to saturate the glycine site on NMDA receptors, and 1 µM strychnine to block glycine-activated Cl currents.

Data Analysis. The block of NMDA-evoked currents was calculated according to the formula B = [(I  I_B) / I] × 100, where I was determined by curve fitting the decay of the NMDA-evoked current during the NMDA application and extrapolating the fit to the end of the NMDA antagonist coapplication, and I_B was the current measured at the end of NMDA/blocker coapplication. Current decays were fit to first-order exponential curves using a Chebyshev fit method and pClamp software.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

The Physiological Relevance of Trapping. In general, trapping experiments were conducted with antagonist removed between the initial agonist/antagonist coapplication and the later agonist application. However, in a therapeutic situation, antagonist would be present continuously, making it important to demonstrate that antagonist could still escape from the channel in this situation. As is shown in Fig. 1, AR-R15896AR does escape from the channel whether or not it is present in the bath between applications of NMDA. There was no significant difference in trapping whether antagonist was continually present (61 ± 2%, n = 3) or not (54 ± 3%, n = 9). This demonstrates that the partial trapping of an antagonist is a physiologically relevant phenomenon.

View larger version (17K): [in this window] [in a new window]   Fig. 1.   Partial trapping occurs regardless of whether the antagonist is continuously present during the 30 s between agonist applications (middle trace) or not (top trace). The bottom current trace (in red) shows the onset and immediate relief of block using the same concentration of antagonist used in the top traces. The section of this trace between the dashed lines corresponds to the antagonist off-rate and has been superimposed (in red) where agonist was reapplied in the upper traces. This illustrates the difference between the similar partial trapping observed in the top two traces and their appearance if trapping were complete. Open bars indicate application of 10 µM NMDA, and shaded bars indicate application of 50 µM AR-R15896AR. Scale, 100 pA; 5 s.

On-/Off-Rates and Trapping of Compounds Structurally Related to AR-R15896AR. The concentration of each compound that was required to produce a block of NMDA-induced whole-cell current, not significantly different from the block of 81 ± 1% produced by 50 µM AR-R15896AR, was determined. Then on-/off-rates and degree of trapping of the initial block were measured using the appropriate concentration of antagonist. Representative current traces for selected compounds that trap significantly less than, equivalent to, or more than AR-R15896AR are shown in Fig. 2, top, middle, and bottom, respectively. On-/off-rates were measured by curve fitting the development of block, or relief from it, as neurons were exposed for 5 s to 10 µM NMDA, followed by a 30-s coapplication of NMDA + antagonist and a 50-s re-exposure to NMDA. Trapping was determined by exposing cells to 10 µM NMDA for 5 s, followed by a 30-s coapplication of NMDA + antagonist (in black), and then 120 s later reapplying 10 µM NMDA for 20 s (shown superimposed in red upon the initial response). The degree of trapping and the on-/off-rates for each compound at the appropriate concentration are summarized in Table 1.

View larger version (20K): [in this window] [in a new window]   Fig. 2.   The left current trace for each compound shows the response to a 5-s exposure to 10 µM NMDA, followed by a 30-s coapplication of NMDA + antagonist and a 50-s re-exposure to NMDA. The right set of two current traces shows an initial response (in black) to a 5-s exposure to 10 µM NMDA, followed by a 30-s coapplication of NMDA + antagonist. Superimposed (in red) upon this initial response is the response to a 20-s exposure to 10 µM NMDA exactly 120 s later. Arrows indicate the initial level of current upon reapplication of agonist. Scale, 100 pA; 10 s.

View larger version (15K): [in this window] [in a new window]   Fig. 3.   A, plot of Logoff versus percent block trapped. Linear fit shows a significant positive correlation (p < 0.0001; r = 0.85). B, plot of ACD_LogP versus percent block trapped. Linear fit shows no significant correlation (p = 0.18; r = 0.28). C, plot of ACD_LogD versus percent block trapped. Linear fit shows a significant positive correlation (p < 0.003; r = 0.67). D, plot of ACD_LogpKa versus percent block trapped. Linear fit shows a significant negative correlation (p < 0.003; r = 0.59).

Correlation between Physical Characteristics and Trapping. There was no significant correlation between lipophilic character and trapping, as estimated by ACD_LogP (Fig. 3B). A negative correlation would be expected if partial trapping was due to lipophilic escape from the closed channel. On the contrary, a significant positive correlation was found between trapping and ACD_LogD (Fig. 3C), a parameter that takes into account the ionic character of compounds at physiological pH. This result is mirrored by a significant negative correlation between ACD_pK_a values and trapping (Fig. 3D). In addition, compounds that exhibited the least trapping contained two basic nitrogen atoms that were significantly protonated at physiological pH, thus increasing the hydrophilic nature of the molecule.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

This study demonstrates that partial trapping of block during repetitive stimulation can occur even in the continuous presence of antagonists. Therefore, the degree of trapping of an antagonist is a relevant factor to consider in the physiological situation. It is likely that trapping of channel block contributes to the side effects that have limited the clinical development of use-dependent NMDA receptor antagonists. Our studies confirm and extend the observation that trapping is not an all-or-nothing phenomenon. Among the 24 compounds tested, we observed levels of trapping that varied over a broad range, from essentially complete (93%) to at least as low as 40%. In fact, we measured trapping of only 8% for AR-R15808. However, this value may be an underestimate, since this compound's off-rate exceeded the resolution of our system. If the factors that led to these differences in degree of trapping were amenable to drug design, compounds with increased safety could be synthesized.

Another important finding of this study is that, within the initial group of 21 compounds selected, none trapped significantly less than AR-R15896AR, whereas 15 of them trapped significantly more. AR-R15896AR was chosen as a clinical candidate over other compounds in this group based strongly on its lack of PCP-like effects in rats (Palmer et al., 1999). Taken together, these observations are consistent with the hypothesis that trapping is an important factor in the PCP-like behaviors seen with many use-dependent NMDA receptor antagonists.

Specific structural features influence the degree of trapping. From examining structures of the 21 compounds in the initial group selected, the presence and nature of two basic nitrogen atoms appeared to be important. AR-R15896AR itself contains a pyridine and is less trapped than compounds containing a phenyl group in the analogous position. This suggested that a pyridine or other heterocyclic group might reduce trapping in other compounds. Two additional compounds, AR-R26952 and AR-R15808, were specifically chosen to test this hypothesis. Like AR-R15896AR, AR-R26952 contains a pyridine instead of a phenyl group. AR-R26952 trapped only 40% of its initial block, which was significantly less than that of AR-R15896AR. In the case of AR-R15808, changing the pyridine of AR-R15896AR to a piperidine ring increased the basicity of this nitrogen. This substitution resulted in a compound with the lowest apparent degree of trapping and the fastest off-rate of all compounds tested. At the other end of the spectrum, the four highest trapping compounds have a fully saturated carbon atom in the  position. In fact, 8 of the 15 high trapping compounds have a fully saturated carbon atom in the  position, whereas only two of the seven intermediate trapping and none of the low trapping compounds have a fully saturated carbon atom in the  position. Thus, changes in structural features influence the degree of trapping, suggesting that a specific binding site is responsible for differences in trapping and that partial trapping is an inherent feature of the trapping mechanism in NMDA receptors.

A simple mechanism that could account for differences in the degree of trapping is closed-channel egress via a lipophilic pathway. However, there was no significant correlation between the extent of trapping and ACD_LogP among the 24 compounds examined in this study. This observation is supported by our previous report that lipophilicity did not correlate with trapping among AR-R15896AR, memantine, and ketamine (Mealing et al., 1999). Furthermore, that study also demonstrated that trapping was stable over relatively long periods of time, which would not be predicted if lipophilic diffusion was occurring. Cumulatively, this evidence strongly argues that lipophilic diffusion through the receptor protein or lipid bilayer is not responsible for partial trapping.

Our data indicate a positive linear correlation between Log_off and degree of trapping, which is consistent with antagonist unbinding being an exponential process. Furthermore, among three NMDA receptor antagonists with intermediate kinetics that we previously used to investigate trapping, a similar trend is evident (Mealing et al., 1999). We also noted that the degree of trapping was the same across a range of concentrations of antagonist. This finding is consistent with the correlation between off-rate and degree of trapping, because off-rate is concentration-independent. Compounds with a fast off-rate could escape being trapped simply because they exit faster than the receptor closes. This could possibly be described by modifications to existing linear models of trapping block. However, a sequential or linear mechanism of trapping block would predict inward tail currents, developing upon removal of agonist following onset of block, since the blocked receptor must return to a closed state via an open state (Lingle, 1983). In this study, tail currents were not observed with any of the compounds tested. Using similar methodology, we have previously demonstrated tail currents using the sequential channel blocker, 9-AA (Mealing et al., 1999). However, failure to observe tail currents with 23 of the compounds tested is not compelling evidence against a sequential mechanism of block. These compounds all have a much slower off-rate than 9-AA. Consequently, tail currents spanning a longer duration with smaller amplitude could have escaped detection. However, AR-R15808, whose off-rate, like that of 9-AA, was beyond our resolution, also failed to show tail currents. Therefore, although we cannot rule out a sequential mechanism of block for compounds that exhibit partial trapping, we can offer no evidence to support it.

On the other hand, cyclic models, like those of Gurney and Rang (1984) or Benveniste and Mayer (1995), explicitly allow for the possibility of the return from a closed, blocked state directly to the closed state. These models do not predict tail currents prior to channel closure. Changes in the rate constant for the escape of antagonist from the closed channel, or "closed-channel egress", could allow partial trapping to occur. However, these models give no insight into mechanisms that could mediate closed-channel egress.

A physical description of trapping channel block involves closing of the channel that is associated with trapping of the antagonist molecule. The question arises as to where the blocker is trapped. It is unlikely that these use-dependent antagonists can traverse the NMDA channel. Therefore, closing of the channel cannot prevent the antagonist from rejoining the extracellular fluid. We can visualize the channel to be a "lower barrier" through which antagonists cannot pass into the cell (Fig. 4). To account for antagonists being trapped, there must be an additional, more extracellularly located, "upper barrier", whose closure restricts the antagonist from rejoining the extracellular space. In this scheme, once the upper barrier closes, the antagonist is trapped in a "vestibule", formed by the space between these two barriers, and the receptor is in the closed, blocked state. This scheme is similar to one proposed for quaternary ammonium compounds block of potassium channels (Armstrong, 1971). Barring lipophilic diffusion, the receptor cannot return to the closed, unblocked state without going through a blocked state in which the upper barrier is open. Partial trapping could occur if the upper barrier closes more slowly than the channel. In large, structurally complex molecules like receptor channels, this is not improbable. Compounds with fast off-rates could escape trapping because they escape the vestibule faster than the upper barrier closes. Therefore, the term closed-channel egress would be more accurately defined as closed-channel/open-vestibule egress.

View larger version (27K): [in this window] [in a new window]   Fig. 4.   Cartoon illustrating closed-channel trapping and egress. In the upper left-hand corner, a closed channel is shown whose conformation does not permit ion flow. Application of agonist causes channel opening (upper right). Concurrent application of agonist and antagonist allows antagonist into a "vestibule" to act at the channel pore, resulting in an open channel block (lower right). Removal of the agonist results in channel closure. If the "upper barrier" of the vestibule closes more slowly than the "lower barrier", the channel pore, the antagonist may escape, resulting in closed channel egress (lower left). However, if the upper barrier closes more rapidly than the antagonist can leave the channel pore, trapping results (bottom).

An interesting corollary to this model is that the antagonist should also be able to enter the vestibule during the closed-channel/open-vestibule state. A significant amount of entry would reduce apparent closed-channel egress. In our experiment to test the physiological relevance of closed-channel egress, we did not detect any significant differences suggestive of closed-channel/open-vestibule entry. However, since macroscopic on-rate is a function of concentration, it is possible that this effect would be prominent only at higher concentrations of antagonist. It is also possible that the presence of antagonist in the vestibule retards closing, but it can then proceed rapidly following egress. In addition, physicochemical factors might favor antagonist egress over re-entry into the vestibule.

We have proposed that closed-channel egress provides a means to escape the problem that trapping presents to use-dependence. We noted that AR-R15896AR had the lowest trapping of the compounds chosen using objective selection criteria. Other post hoc comparisons (e.g., AR-R18526) suggest that compounds with high degrees of trapping produce significant amounts of PCP-like behaviors in rats at lower doses (G. Palmer, personal communication). Therefore, we would expect compounds with small degrees of trapping to have low liabilities for producing PCP-like behaviors. Unfortunately, behavioral studies using the two lowest trapping compounds, AR-R15808 and AR-R26952, could not be completed due to complications with bioavailability and specificity. Therefore, the correlation between degree of trapping and safety remains a plausible hypothesis, but presently untestable with the present inventory of compounds.

Rogawski (1993) proposed that low-affinity NMDA receptor antagonists would be safer because they prevent supramaximal block and the delayed recovery resulting from it. However, trapping of antagonist can override the benefits of low-affinity and use-dependence. Partial trapping permits at least some degree of use-dependence to continue. We suggest a model in which partial trapping occurs because of very fast off-rates of compounds. Since fast off-rate is one determinant of low affinity, safety in use-dependent NMDA receptor antagonists can be accounted for by a modified version of Rogawski's proposal. Parsons et al. (1995) have also noted the importance of fast off-rate kinetics to compound safety, especially during strong depolarization, as would occur during excessive receptor activation.