The N-methyl-D-aspartate (NMDA) subtype of glutamate receptor is blocked by ketamine, and this action likely contributes to ketamine's anesthetic and analgesic properties. Previous studies suggest that ketamine occludes the open channel by binding to a site located within the channel pore. This hypothesis was examined by investigating the effects of ketamine on single-channel currents from NMDA receptors.
The cell-attached and outside-out configurations of the patch clamp technique were used to study NMDA-activated currents recorded from cultured mouse hippocampal neurons.
In cell-attached patches, NMDA evoked currents that had an apparent mean open time (tau o) of 3.26 ms. The probability of at least one channel being open (Po') was 0.058. The addition of ketamine (0.1 microM or 1 microM) to the pipette solution decreased Po' to 53% and 24% of control values, respectively. At 1 microM ketamine, this reduction was due to a decrease in both the frequency of channel opening and the mean open time (44% and 68% of control values, respectively). Ketamine did not influence channel conductance and no new components were required to fit the open- or closed-duration distributions. Ketamine (50 microM), applied outside the recording pipette, reduced the opening frequency of channels recorded in the cell attached configuration. This observation suggests that ketamine gained access to a binding site by diffusing across the hydrophobic cell membrane. In outside-out patches, ketamine potency was lower than that observed in cell-attached patches: 1 microM and 10 microM ketamine reduced Po' to 63% and 34% of control values, respectively, and this reduction was due primarily to a decrease in the frequency of channel opening with little change in mean open time.
These observations are consistent with a model whereby ketamine inhibits the NMDA receptor by two distinct mechanisms: (1) Ketamine blocks the open channel and thereby reduces channel mean open time, and (2) ketamine decreases the frequency of channel opening by an allosteric mechanism.
Glutamate is the major excitatory neurotransmitter in the mammalian central nervous system. Based on pharmacologic sensitivity, ion permeability, and channel kinetics, at least two major classes of ionotropic glutamate receptors have been identified: the N-methyl-D-aspartate (NMDA) receptors, and the non-NMDA or alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) kainate receptors.  The NMDA receptor is a unique neurotransmitter receptor in that channel activation requires the binding of both glutamate and the coagonist glycine.  The channel is blocked in a voltage-dependent manner by physiologic concentrations of Mg2- and undergoes calcium-dependent and calcium-independent forms of inactivation. [3–5] Receptors for NMDA are thought to play a pivotal role in various physiologic functions, including synaptic plasticity and neuronal development. In addition, NMDA receptors may contribute to pathologic processes such as seizure, ischemia-related neuronal death, and several neurodegenerative diseases. 
Ketamine (2-O-chlorophenyl-2-methylamino cyclohexanone) blocks NMDA receptors, and this action is thought to contribute to ketamine's potent anesthetic and analgesic properties. [7–10] Electrophysiologic and binding studies have revealed that various other antagonists, including phencyclidine, memantine, amantadine, and MK-801, induce a “use-dependent” inhibition of the NMDA receptor. [11–15] The blockade is use dependent in that the rates of onset and the recovery from blockade are increased by applying agonist. These observations have been interpreted as indicating that the channel must be in the open state before ketamine can bind to or dissociates from a blocking site possibly located within the channel pore. This type of blockade is called “uncompetitive inhibition” because it is contingent on the binding of agonist to the receptor.  In contrast, a “noncompetitive” blocker binds independently of the presence of agonist.
Previous single-channel studies of the NMDA channel have provided insights into the mechanisms by which dissociative anesthetics block the NMDA receptor. Heuttner and Bean  examined the effects of MK-801 on NMDA-activated single-channel currents and found that MK-801 produced a dose-dependent decrease in the mean duration of channel opening.  Similarly, phencyclidine decreased the frequency and duration of NMDA channel openings.  These actions are consistent with a sequential open channel blocking mechanism that can be represented by scheme 1 Figure 1.
The rates of channel opening and closing are beta and alpha, the rates of association and dissociation of the blocker are k sub +B and k sub -B, and [A] and [C] are the concentrations of agonist and blocker, respectively.  Because this model assumes that the receptor exists in a limited number of discrete states, the rates governing the transitions between these states can be estimated from the time constants obtained from the dwell time distributions. According to scheme 1, the mean duration of channel opening is decreased by the blocking drug because the open (ion-conducting) state can be terminated by either channel closure (alpha) or transitions into the open, nonconducting state referred to as “blocked”(k sub +B [C]). Thus the mean single-channel open time (tauo) will be inversely related to the sum of the two transition rates leading from the open state; that is, tau sub = 1/(k sub +B [C]+ alpha). Accordingly, open-channel blockade is characterized by a decrease in the mean duration of the open state of the channel.
Scheme 1, however, cannot account for agonist-dependent recovery from channel blockade because the model assumes that the agonist remains bound until the blocker dissociates from the receptor. Thus the rate of recovery from blockade should be not be influenced by the extracellular concentration of agonist. To account for this inconsistency, a model of “drug trapping” was postulated, as summarized in scheme 2 Figure 2. Here the agonist can dissociate and the receptor close with ketamine bound to the receptor. Closing of the channel “traps” ketamine on the receptor, and the probability of dissociation of ketamine remains low until the channel is stimulated to reopen after agonist binding.  In its simplest form, this model proposes that ketamine does not directly influence channel “gating” or the transition rates between the open and closed states. The true affinity of the binding site remains unchanged even though the site is “guarded”(that is, access is restricted by the conformation of the receptor). Scheme 2, which is summarized here, is discussed in detail by MacDonald et al. 
Use-dependent blockade does not necessarily imply occlusion of the open pore or preferential binding to the open state. Receptors for NMDA can exist in several agonist-associated closed states, such as the desensitized form of the receptor. Preferential binding of ketamine to a agonist-associated closed state would also be use dependent (scheme 3 Figure 3). This scheme resembles the modified-receptor model proposed by Hille  that describes the effects of certain local anesthetics on voltage-activated sodium channels. In scheme 3, the normal channel is represented as making transitions between three states: closed, open, and desensitized. The NMDA channel resides in more that three states, but for our purpose the scheme has been simplified. The rates of entry into, and recovery from, a nonconducting desensitized state are indicated by k sub +D and k sub -D, respectively. The rates of association and disassociation of ketamine are assumed to be negligible when the channel is in a non-desensitized state.
Similar to scheme 2, ketamine can be trapped when the agonist is removed and the channel reverts to the non-ligand-bound state so that the onset and recovery from blockade will be agonist dependent. However, in contrast to scheme 2, scheme 3 predicts that ketamine will reduce the number of channels available for opening without influencing mean channel-open time.
The kinetic properties of the ketamine blockade of the NMDA receptor have not been examined previously at the single-channel level. It is plausible that the mechanism of ketamine block differs substantially from that of phencyclidine or MK-801: Ketamine's actions on the nicotinic acetylcholine receptor and non-NMDA (quisqualate)-type glutamate receptor in locust suggest that it interacts with both the open and closed conformational state of these channels. [21–23] Furthermore, the physical site(s) of binding may not be identical for all uncompetitive antagonists of the NMDA receptor. Molecular studies of the binding sites for phencyclidine and MK-801 indicate that the phencyclidine binding site appears to be confined to the putative M2 segment, whereas blockade by MK-801 requires an interplay between residues associated with the M2 and M3 regions of the protein.  Different binding sites and mechanisms of action might account for the diverse clinical properties of uncompetitive NMDA receptor antagonists and represent potential targets for the development of new pharmaceutical agents.
Here we report the effects of ketamine on single-channel currents activated by NMDA and examine the validity of the various kinetic schemes intended to describe the blockade. Our results suggest that there are two mechanisms of inhibition: an open-channel blockade and a closed-channel blockade that involves a site accessible from the hydrophobic membrane phase. Part of this work was published previously in abstract form.*
Materials and Methods
Dissociation and Culture of Hippocampal Neurons
Animal protocols were approved by the University of Toronto Animal Care Committee. Cultures of fetal hippocampal neurons were prepared from Swiss white mice, as previously described.  Briefly, hippocampal neurons were dissected from fetal mice pups. Neurons were dissociated by using enzymatic digestion and trituration. Cells were plated on collagen-coated dishes, incubated at 37 degrees Celsius in 7% carbon dioxide, and maintained in culture for 10–21 days.
Solutions and Single-channel Current Recordings
Before recording, cells were rinsed thoroughly with a standard extracellular recording solution containing 140 mM NaCl, 1.3 mM CaCl2, 5.4 mM KCl, 25 mM N'-2-hydroxy-ethylpiperazine-N'-2-ethanesulphonic acid, 33 mM glucose, and 300 nM tetrodotoxin. The solution was buffered to a pH of 7.4 with NaOH and the osmolality was adjusted to 297–300 mOsm. Glycine (1 micro Meter) was added to the solution because it is an essential co-agonist for gating of the NMDA receptor.  Patch electrodes were constructed from thin-walled borosilicate glass (1.5 mm outer diameter; World Precision Instruments) using a two-stage vertical puller (Narishige PP-83). Pipettes were coated near the tips with Sylgard 184 (Dow Corning, Midland, MI) and fire polished to a final resistance of 2–5 M Omega. The electrodes used for cell-attached recordings were filled with a solution containing 70 mM NaCl, 70 mM Na2SO4, 1 mM CaCl sub 2, 10 micro Meter NMDA, and 1 micro Meter glycine. Ketamine (0 micro Meter, 0.1 micro Meter, or 1.0 micro Meter) was added to the pipette solution. In a second series of cell-attached experiments, ketamine (50 micro Meter) was added to the bath solution. For outside-out patch recordings, the pipettes were filled with solutions containing 140 mM CsF, 10 mM N'-2-hydroxy-ethylpiperazine-N'-2-ethane-sulphonic acid, 11 mM ethylene glycol-O,O'-bis(2-aminoethyl)-N,N,N',N'-tetraacetic acid, and 1 mM CaCl2. The pH was adjusted to 7.4 with CsOH. Patches were voltage clamped using an Axopatch 1B or Axopatch 200 amplifier (Axon Instruments). For the outside-out patch recordings, drugs and agonist were applied through a perfusion barrel that was positioned close to the membrane patch.  Patches were exposed to NMDA for approximately 5–15 min before records were sampled for kinetic analysis. We assumed that channels were functioning under near steady-state conditions and the extent of glycine-insensitive desensitization had stabilized. [25,26] All experiments were conducted at room temperature (20–25 degrees Celsius).
Single-channel currents were filtered at 2 kHz and stored on VHS tape for off-line analysis. Data were digitized (T1–1 interface, Axon Instruments, Inc., Foster City, CA) and sampled every 100 micro s using pCLAMP software (Axon Instruments, Inc., Foster City, CA). Current records were selected for detailed kinetic analysis based on the stability of the recording and the number of channels present in the patch. Only patches that remained stable over a 5- to 45-min period were analyzed. It was assumed that patches contained at least as many channels as conductance levels observed and that each channel behaved independently. Only records with infrequent (< 10%) openings to a second open level were used for the analysis. However, most patches had at least two open levels.
The threshold method was used to detect transitions between the open and closed states.  The threshold was set at one half the maximal current amplitude, and all records were monitored visually to ensure stability of the baseline. To minimize the effect of false events (such as transitions due to random noise), the minimum event duration was set at 300 micro s. This strict criteria, as well as the limitations of the recording system, would inevitably result in “missed” or undetected events. For example, if the duration of channel closure was shorter than the rise time of the recording system or the specified minimum event duration, closed events would be missed. [27,28] This would result in an overestimation of channel open time. Conversely, missed brief openings would result in an overestimation of closed times. Methods to correct for missed events were not used. However, the effect of missed events on our interpretation of ketamine's effects was minimized by using identical analysis criteria for records obtained in the absence and presence of ketamine. Furthermore, all records were analyzed at least twice to ensure that a substantial number of brief events were not excluded from the analysis. Data were binned into open- and closed-interval histograms and plotted using the pSTAT program (pCLAMP 6.0, Axon Instruments, Inc., Foster City, CA). We observed that the duration of closed events spanned several orders of magnitude. To detect multiple kinetic components of channel closure, closed-interval histograms were plotted with logarithmically scaled bin widths. Open interval histograms were constructed with a linearly scaled bin width because no new components were evident when histograms were plotted on a linear-log scale. The time constants and relative areas were determined by fitting the sums of exponential functions to the plotted data using the method of maximum likelihood.  The number of exponential components was increased until the improved fit was not statistically significant according to the F-value (where the F-value compares the sum of the squared errors extracted from the fit for the various models). Critical values of F corresponding to P < 0.05 were obtained from standard tables. Results are presented as means +/- SEM unless otherwise stated.
The mean duration of channel opening and closure were extracted from the event lists. Unfortunately, the number of channels in the patch cannot be determined accurately if the channels take long sojourns in desensitized or inactivated states.  Our records reflect the steady-state activity of a population of receptors, and prolonged closings of individual channels cannot be distinguished. Because we could not determine accurately the number of channels present in each patch, we defined a modified open probability (Po') in which Po' is the probability that at least one channel is open. Po' was calculated as to/tiwhere towas the total dwell time at the open level and tiwas the time interval over which Po' is measured. The frequency of channel opening was determined by the number of open events divided by the duration of a specified time interval. A discrepancy between Po' and the open probability of a single channel (calculated by multiplying the frequency of channel opening and mean open time) is expected because Po' does not distinguish between single or double openings, whereas the mean open time was calculated from single open events.
Drugs and Other Chemicals
Ketamine is an arylcyclohexylamine that contains an asymmetric carbon atom. It exists as two isomers: the (R) and (S) form, where the (S) isomer is the more potent general anesthetic and NMDA receptor antagonist. [30,31] The commercially available preparations of ketamine used in these experiments are racemic mixtures (Parke-Davis, Scarborough, Canada). All other agonists and compounds were purchased from Sigma Chemical Co. (St. Louis, MO) unless otherwise specified.
NMDA-activated Currents Recorded in the Cell-attached Configuration
The addition of NMDA (10 micro Meter) to the pipette solution resulted in single-channel openings, as illustrated in Figure 4. With the membrane voltage clamped at the resting potential of the cell (pipette potential = 0 mV), currents were characterized by bursts of openings interrupted by brief and long channel closures. The mean current amplitude for seven patches was 4.33 +/- 0.32 pA. The current-voltage relation, measured at pipette holding potentials ranging from -30 to +30 mV, was linear with an estimated slope conductance of 43.2 +/- 9.4 pS.
We observed a large variation in the frequency of channel opening among different patches (1.66 sup -1 to 55.5 s sup -1) However, in individual patches, the frequency of channel opening remained stable during the duration of each recording. Therefore we assumed that this variability resulted from differences in the number of channels present in each membrane patch. The probability of at least one channel being open (Po') was 0.058 +/- 0.015. The mean frequency of channel opening was 13.4 +/- 4.4 s sup -1, and the mean open time (tauo) was 3.26 +/- 0.25 ms (n = 13).
The duration of channel openings and closings were measured and event lists were used to construct open- and closed-duration histograms. Table 1summarizes the mean time intervals, time constants, and relative areas estimated from the dwell-time distributions.
Nowak and Wright  reported that membrane hyperpolarization caused a decrease in the probability of NMDA channel opening. To hyperpolarize the membrane patch, a positive holding potential of +30 mV was applied to the recording electrode. Conversely, the membrane was depolarized by applying a pipette potential of -30 mV. The value of Po' was 0.03 +/- 0.02 (n = 8) at a pipette potential of +30 mV and 0.07 +/- 0.04 (n = 6) at a pipette potential of -30 mV. The mean channel-open time also decreased with membrane hyperpolarization (V pipette =+30 mV, tauo= 2.85 +/- 0.48 ms; V pipette =-30 mV tauo= 9.97 +/- 2.00 ms).
Ketamine Modulation of Single-channel Currents from Cell-attached Patches
Ketamine was shown previously to reversibly inhibit NMDA-evoked whole-cell currents recorded from hippocampal and striatal neurons. The median inhibitory concentration for blockade is reported to range from 0.43 to 10 micro Meter. [8,12,33] To determine if this inhibition by ketamine resulted from a decrease in channel conductance, frequency of channel opening or channel-open time, we studied, in the cell-attached configuration, the effects of ketamine on single NMDA receptors. We selected ketamine concentrations in the range of previously reported IC sub 50 (concentration that inhibits 50% of the maximal response) of whole-cell currents.
Inclusion of ketamine (0.1 micro Meter or 1.0 micro Meter) in the recording electrode produced a pronounced reduction in the frequency of channel opening but no change in single-channel conductance (Figure 5). We also tried to record single-channel events with 10 micro Meter ketamine in the pipette solution, but channel openings were so infrequent that kinetic analysis of these records was not possible.
(Table 1) summarizes the distributions of open and closed intervals measured with ketamine (0 micro Meter, 0.1 micro Meter, or 1 micro Meter) in the pipette. Ketamine did not significantly influence channel conductance as the mean current amplitude of NMDA channels exposed to ketamine (0.1 micro Meter or 1 micro Meter) was 3.82 +/- 0.16 pA (n = 8) and 3.96 +/- 0.73 pA (n = 5), respectively. Ketamine (0.1 micro Meter or 1 micro Meter) reduced Po' to 53%(Po' = 0.031 +/- 0.010, n = 8) and 24%(Po' = 0.014 +/- 0.007, n = 5, P < 0.05, by the Mann-Whitney test) of control values. The frequency of channel opening (f) was also decreased to 68%(f = 9.1 +/- 4.11 s sup -1) and 44%(f = 5.9 +/- 24.4 s sup -1) of control values (P < 0.05, by the Mann-Whitney test). A significant decrease in channel-open time was observed at the higher concentration of ketamine. With 1 micro Meter ketamine, channel-open time was reduced to 68% of control. The difference in the fractional reduction of Po' compared with f with 1 micro Meter ketamine can be attributed to the decrease in channel-open time.
The concentration-dependent decrease in mean open time is consistent with a bimolecular interaction between ketamine and a site on the open channel, as illustrated in scheme 1. The relation between drug concentration (ketamine 0 micro Meter, 0.1 micro Meter, and 1 micro Meter) and channel-open time was used to approximate the forward rate constant (k sub +B) of ketamine blockade (Figure 6). The rate of association, calculated from the slope of the plot of 1/tau o, was 1.4 x 108m sup -1 [center dot] s sup -1.
In the Cell-attached Configuration, Addition of Ketamine to the Bath Solution Inhibits NMDA Channels Recorded in the Cell-attached Configuration
The uncharged form of ketamine is highly lipid soluble.  Thus it is plausible that ketamine can diffuse into the cell membrane and gain access to a hydrophobic binding site on the NMDA receptor. Therefore we tried to determined if ketamine, added to the bath solution, could inhibit NMDA receptors isolated in the tip of the recording electrode.
Stable single-channel currents were recorded with 10 micro Meter NMDA in the pipette solution. Adding 50 micro Meter ketamine to the bath solution caused a gradual, reversible inhibition of channel opening (Figure 7). The probability of channel opening was significantly reduced compared with control values (Po', control = 0.046 +/- 0.010, Po', ketamine = 0.051 +/- 0.003, n = 5, P < 0.05, by the Wilcoxon signed rank test).
This inhibition was primarily characterized by a decrease in the frequency of channel opening, rather than by a reduction in the mean channel open time (Table 2). Three of the five patches remained stable for a prolonged period, and perfusion of the cell with the ketamine-free solution resulted in a gradual increase in the probability of channel opening (Po', recovery = 0.011 +/- 0. 002).
Ketamine Inhibition of NMDA Channels Recorded in the Outside-out Configuration
A major advantage of the cell-attached recording method is that the plasma membrane is not disrupted after seal formation. This results in minimal disruption of cytosolic factors that might influence channel gating or drug sensitivity.  However, the pipette solution cannot be easily exchanged, so the same population of receptors cannot be examined before and after exposure to a known concentration of drug. To study the same receptor population in the absence and presence of ketamine, we examined the effects of ketamine on NMDA receptors in outside-out membrane patches (Figure 8).
In patches held at -60 mV, applications of NMDA (10 micro Meter) evoked channel openings with a mean current amplitude of 2.69 +/- 0.11 pA (n = 9); the estimated single channel conductance was 44.8 pS. Application of ketamine (1 micro Meter or 10 micro Meter) did not change the single-channel current amplitude (2.47 +/- 0.13 pA n = 5, and 2.50 +/- 0.16 pA, n = 4, respectively), (Figure 9).
As summarized in Table 3, ketamine (1 micro Meter or 10 micro Meter) reduced Po' to 63% and 34% of control values, respectively. For five patches, the steady-state probabilities of at least one channel being open, measured before and after exposure to ketamine (1 micro Meter), were 0.087 +/- 0.028 and 0.056 +/- 0.019 (P < 0.05, by the Wilcoxon signed rank test), respectively. Thus 1 micro Meter ketamine reduced Po' to 64% of control, and 10 micro Meter ketamine decreased Po' to 23% of control. These reductions in Po' were due primarily to a decrease in the frequency of channel opening, rather than to a reduction in the mean channel-open time (Table 3). Applications of the competitive NMDA receptor antagonist, DL-2-amino-5-phosphovaleric acid (10–100 micro Meter), reversibly inhibited residual channel openings (n = 3), indicating that the currents were indeed mediated by the NMDA receptor.
Similar to the dwell-time histograms constructed for currents recorded in the cell-attached configuration, open-interval histograms were best fit with the sum of two exponential functions, whereas the closed-duration histograms were best fit with the sum of three exponential functions (Figure 10). A possible decrease in channel open time was apparent with the higher concentration of ketamine (10 micro Meter), but this effect was not significant (P = 0.437, by the Wilcoxon signed rank test).
Single-channel studies were undertaken to identify the mechanisms underlying ketamine blockade of the NMDA receptor. In the absence of ketamine, channel conductance was similar to that previously reported for NMDA receptors in hippocampal neurons. [36,37] Furthermore, the duration of NMDA channel openings was within the range reported for native hippocampal receptors and cloned NR1a-NM epsilon 1 and NR1-NR2A subunits expressed in HEK 293 cells and Xenopus oocytes. [36–41] The effects of ketamine on single-channel NMDA currents are summarized here.
In cell-attached patches, the inclusion of ketamine (0.1 micro Meter or 1 micro Meter) in the pipette solution caused a concentration-dependent decrease in both the duration and frequency of NMDA channel openings. These actions are similar to the effects of phencyclidine and its analogs and are consistent with an open-channel mechanism of blockade.  However, the decreases in channel-open times induced by 0.1 micro Meter and 1 micro Meter ketamine (6% and 32%, respectively) were small relative to the reduction in open-channel frequency (32% and 56%, respectively). Furthermore, in cell-attached patches, externally applied ketamine (50 micro Meter) caused an 89% decrease in the frequency of channel opening without a significant change in channel-open time. Ketamine blockade of receptors recorded in the outside-out patch configuration was similarly characterized by a decrease in the frequency of open events with little change in open time.
Mechanisms of Ketamine Blockade
As discussed previously, several mechanisms could account for ketamine blockade of the NMDA receptor. Schemes 1, 2, and 3 will now be considered in light of our observations. Scheme 1 is not consistent with the use-dependent recovery from block previously described for ketamine.  In addition, scheme 1 predicts that a new closed state with a mean duration of 1/kBshould be evident in the distribution of the closed times. We observed no change in the fast or intermediate components of the closed-time distributions and no new closed-time components. Thus we rejected scheme 1 as a possible model of ketamine blockade. Both scheme 2 and 3 are consistent with use-dependent recovery from blockade; however, scheme 2 predicts a reduction in mean open time, whereas scheme 3 does not.
In the cell-attached configuration with ketamine added to the recording electrode, there was a concentration-dependent decrease in channel-open time, consistent with scheme 2. The rate of association of ketamine with the open channel was estimated to be 14 x 108m sup -1 [center dot] sec sup -1. This value is considerably greater than that previously reported for ketamine inhibition of whole-cell currents (1.7 x 104m sup -1 [center dot] sec sup -1 and 9 x 104m sup -1 [center dot] sec sup -1). [7,33] However, the time course of inhibition of whole-cell currents depends not only on the association rate constant but also on the probability of channel opening (Po) where kon= 1/Po (alpha + k sub +B [B]). The maximum probability of NMDA channel opening is not known, and an overestimation of Po results in an underestimation of k sub +B. Thus molecular rate constants are more directly determined from single-channel experiments. Although our value of 1.4 x 108m sup -1 [center dot] sec sup -1 is presented only as an approximation, it suggests that the rate of ketamine binding is of the order of magnitude expected for a fast channel blocker. However, the observed decrease in channel-open time accounted, only in part, for the decrease in Po', suggesting that the channel could close with ketamine bound to the receptor. This closed-blocked state would reduce the number of channels available for activation and thereby reduce the frequency of channel opening. These observations are consistent with scheme 2.
In outside-out patches or cell-attached patches with ketamine added outside the recording electrode, a large reduction in opening frequency was observed with no significant change in mean open time (despite the presence of high concentrations of ketamine). Detailed single-channel studies on NMDA receptor function suggest that the receptor kinetics are characterized by several agonist-bound inactive states. [42–47] Any one of these states might preferentially enhance ketamine binding and lead to blocked states. This would reduce the number of channels available for opening but produce no change in mean open time. We observed that only the longest closed time was prolonged by ketamine, as expected from a decrease in the number of channels available for normal activation. The lack of change in the open-time constants and the fast and intermediate closed-time constants suggests that ketamine, acting through a membrane delimited pathway does not modulate channel gating.
Our data suggest that ketamine acts by at least two mechanisms: an open-channel block, as evidenced by the decrease in tauo, and a closed block characterized by a decrease in open frequency with no change in open time. It is not known if distinct sites mediate open and closed blockade. Furthermore, the kinetics of the two mechanisms appear to depend on the configuration of the membrane patch. There was an apparent decrease in drug sensitivity after patch excision. Factors that could contribute to changes in drug sensitivity include disruption of the cytoskeleton and alterations of cytosolic factors that influence the sensitivity of the receptor to antagonists. 
Open and closed mechanisms of blockade were previously proposed for ketamine inhibition of the nicotinic acetylcholine receptor. [21,22,47–50] Ketamine, as well as various other intravenous and volatile anesthetics, selectively bind to the closed nicotinic acetylcholine receptor receptor. [17,22] These anesthetics drive the receptor into a nonactivatable configuration that is similar to the agonist-induced desensitized state (for a review of this, see Forman and Miller ). Furthermore, binding studies indicate that the site that facilitates desensitization of the nicotinic acetylcholine receptor by [sup 3 H]phencyclidine is a hydrophobic region of the nicotinic acetylcholine receptor receptor that is not directly accessible from the aqueous phase. 
Ketamine Gained Access to a Binding Site through a Hydrophobic Pathway
In addition to the mechanisms of ketamine blockade, the site(s) of antagonism are of interest. It was previously suggested that ketamine binds in the vicinity of the extracellular mouth of the NMDA channel pore.  However, we observed that ketamine added to the bath solution inhibited NMDA receptors recorded in the cell-attached configuration. It is unlikely that ketamine gained access to the extracellular mouth of the channel by diffusing across the high-resistance seal (> 10 G Omega) and accumulating in the tip of the recording electrode. Ketamine is formulated as a hydrochloride salt and is highly water soluble, with apKa (negative logarithm of the acid ionization constant) of 7.5.  Under physiologic conditions, a large fraction of the drug exists in the lipid-soluble form (44%). The octanol/buffer partition coefficient was reported to be 398 (pH 7.4).** Therefore the concentration of ketamine in the lipid phase is several orders of magnitude greater than in the aqueous phase. Our results suggest that ketamine gained access to a blocker site associated with the lipid membrane or the lipid-protein interface, as indicated in Figure 11. Indeed, the presence of a hydrophobic pathway for dissociated anesthetics was predicted from binding studies of the antagonist [sup 3 H]MK-801.  Javitt and Zukin  postulated that [sup 3 H]MK-801 gained access to the slow hydrophobic pathway, which was independent of channel gating and a fast hydrophilic pathway associated with channel opening.
The site(s) of blockade by uncompetitive antagonists, including ketamine, have been investigated using cloned NMDA receptor subunits and site-directed mutagenesis. [53,54] Dissociative anesthetics are thought to bind to a site associated with the Mg2+ binding location on the putative transmembrane domain. Replacement of the conserved asparagine 598 residue in the M2 region by glutamine caused a decrease in the receptor sensitivity to MK-801 and Mg2+.  However, examination of a series of NMDA subunit heteromers revealed that not all combinations of NMDA receptor subunits were equally sensitive to dissociative anesthetics: Not all the dissociative anesthetics bind to an identical site on the NMDA receptor, and their effects were mediated by a site other than the Mg2+ binding domain. 
Our results are somewhat analogous to the actions of local anesthetics on voltage-activated sodium channels.  This family of compounds induce both a use-dependent or “phasic” block and “tonic” inhibition of sodium channels (for a review, see Butterworth and Strichartz ). Local anesthetics are thought to dissolve in the cytoplasm and reach the channel via the hydrophylic pathway. In addition, they may diffuse into the membrane and bind via a hydrophobic pathway. Clues to the location of binding sites of local anesthetics have been obtained by examining the effects of structurally diverse local anesthetics and their effects under different recording conditions (such as changes in membrane potential or pH). Similarly, the location of the ketamine-binding site(s) might be elucidated by examining the stereoselectivity and voltage sensitivity of the closed- and open-channel blockade by ketamine of NMDA receptors.
Ketamine has unique clinical properties and is the anesthetic of choice in specific situations, such as in patients with a compromised hemodynamic status or asthma.  More recently, ketamine has been used in subanesthetic doses as a potent analgesic. In humans, inhibition of the NMDA receptor is thought to mediate ketamine's clinical properties. [57,58] During anesthesia, the peak plasma concentration in humans is approximately 8.5–9.5 micro Meter, whereas analgesia is associated with plasma concentrations of approximately 0.55 micro Meter. [10,59] In rats, the peak plasma concentration associated with anesthesia is less than 36 micro Meter.  Ketamine in the plasma exists in two states: a free unbound form and a form bound to plasma proteins. [61,62] The free fraction of ketamine determines the rate of diffusion to a site of action. In humans, as much as 47% of ketamine is bound to plasma proteins.  However, ketamine is highly lipid soluble and the brain-to-plasma ratio for ketamine is estimated to be 6.5:1, suggesting that ketamine preferentially accumulates in the brain.  Thus it is possible that, during anesthesia, the concentration of ketamine present at the NMDA receptor is considerably higher than the plasma concentration.
We observed that clinically relevant concentrations of ketamine induced both an open and closed blockade of the NMDA receptor. It is of interest to determine which of these mechanisms contribute to ketamine's clinical actions. The predominance of closed-channel blockade observed at low concentrations of ketamine suggests that ketamine's analgesic properties might result from closed-rather than open-channel blockade. In this regard it is notable that the anti-Parkinsonian drugs, memantine and amantadine, have no appreciable anesthetic or analgesic properties and inhibit the NMDA receptor by open-channel blockade.  It will be of interest to determine whether the two modes of blockade can be pharmacologically dissociated. [39,64–66]
Our results are consistent with a dual mechanism of ketamine blockade: occlusion of the open channel from the aqueous phase and closed-channel blockade from the membrane phase.
The authors thank Lydia Brandes and Claire Bartlett for technical assistance, Valerie Oxorn for graphic design, and Patricia Watson for reviewing the manuscript.
*Orser BA, MacDonald JF: Ketamine blocks NMDA channels via a hydrophobic pathway. (Abstract) Can J Anaesth 1992; 39:A12.
**Correspondence from Parke-Davis, Scarborough, Canada.