Journal of Neurophysiology

Recruiting Extrasynaptic NMDA Receptors Augments Synaptic Signaling

Alexander Z. Harris, Diana L. Pettit


N-Methyl-d-aspartate receptor (NMDAR) activation may promote cell survival or initiate cell death, with the outcome dependent on whether synaptic or extrasynaptic receptors are activated. Similarly, this differential activation has been proposed to govern the direction of plasticity. However, the physiological parameters necessary to activate extrasynaptic NMDARs in brain slices remain unknown. Using the irreversible use-dependent NMDAR antagonist MK-801 to isolate extrasynaptic NMDARs, we have tested the ability of short-stimulation trains from 5 to 400 Hz to activate these receptors on CA1 hippocampal slice pyramidal neurons. Frequencies as low as 25 Hz engage extrasynaptic NMDARs, with maximal activation at frequencies between 100 and 200 Hz. Since similar bursts of synaptic input occur during exploratory behavior in rats, our results demonstrate that “extrasynaptic” NMDARs regularly participate in synaptic transmission. Further, 175-Hz-stimulation trains activate all available synaptic and extrasynaptic dendritic NMDARs, suggesting these NMDARs act as synaptic receptors as needed, transiently increasing synaptic strength. Thus extrasynaptic NMDARs play a vital role in synaptic physiology, calling into question their status as “extrasynaptic.”


N-Methyl-d-aspartate receptors (NMDARs) are a critical component of excitatory transmission in the CNS. Activating these receptors can trigger both long- and short-term plasticity, promote cell survival, and initiate cell death (Malenka and Nicoll 1993; Soriano and Hardingham 2007). Recent evidence has suggested that activation of synaptic NMDARs can produce an enhancement of transmission, whereas activation of extrasynaptic receptors depresses transmission (Lu et al. 2001; Massey et al. 2004). Although these results suggest a vital role for extrasynaptic NMDARs in neuronal communication during physiological and pathologic conditions, the physiological parameters necessary to activate these NMDARs in intact tissue remain unclear. Although inactive during low-frequency input, extrasynaptic NMDARs could contribute to postsynaptic physiology during periods of increased activity (Faber and Korn 1988). Under these conditions, neurotransmitters can spill over from the synaptic membrane, and the high affinity of NMDARs makes them uniquely suited for detection of that spillover (Kullmann and Asztely 1998).

We have examined the parameters for physiological activation of extrasynaptic NMDARs in acutely dissected CA1 hippocampal slice pyramidal neurons after the removal of synaptic NMDARs with MK-801 (using 0.1-Hz stimulation). Short synaptic stimulation bursts from 5 to 400 Hz were used to test for activation of extrasynaptic NMDARs. We found that short bursts of presynaptic input at frequencies as low as 25 Hz produce substantial extrasynaptic NMDAR activation. Peak activation of all available synaptic and extrasynaptic dendritic NMDARs occurs at a stimulation frequency of 175 Hz. These data suggest that extrasynaptic NMDARs participate in neuronal signaling under physiological conditions and regularly contribute to the physiology of the cell.


Synaptic transmission

Animals were anesthetized with trifluoroethane and decapitated to prepare 300- to 400-μm-thick coronal slices from the hippocampus of postnatal day (P) 14 to P22 rats (Yang et al. 2006). Whole cell patch-clamp recordings were made from CA1 pyramidal neurons. The patch pipettes were filled with a cesium gluconate solution containing (in mM): 123 Cs gluconate, 8 NaCl, 1 CaCl2, 10 EGTA, 10 HEPES, 10 glucose, 5 ATP, 0.4 GTP, and 1 QX-314 (pH 7.2; 280–290 mOsm). Slices were superfused (3–4 ml/min) at 34°C with oxygenated physiological saline (in mM: 119 NaCl, 2.5 KCl, 1.3 or 0.1 MgCl2, 2.5 CaCl2, 1 NaH2PO4, 26.2 NaHCO3, and 11 glucose). Recordings were rejected if the holding current exceeded −100 pA when pyramidal cells were voltage clamped at −60 mV. To maintain voltage control, series resistance was kept as low as possible, averaging 10.6 ± 3 MΩ (SD; n = 32). Experiments showing any evidence of loss of voltage control were discarded. NMDA receptor currents were isolated by including 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX, 5 or 10 μM) and picrotoxin (100 μM) in the external solution. NMDA currents were recorded by relieving Mg2+ block with depolarization to −20 mV.

(+)-5-Methyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine maleate (MK-801, 50 μM in bath or 500 μM via picospritzer) and both l- and d-2-amino-5-phosphonovaleric acids (l-APV and d-APV, 250 and 50 μM, respectively) were used to block NMDA receptors. Synaptic stimulation was achieved by placing a monopolar glass stimulating electrode adjacent to the dendrite to activate synapses (every 20–30 s; 100–200 μs). Repeated electrical stimulation in the presence of MK-801 was used to block synaptic NMDARs (0.1 or 0.2 Hz; 40–200 μA). Stimulus bursts used to activate extrasynaptic NMDARs were delivered every 30 s. Data were collected and analyzed off-line with Igor Pro (WaveMetrics, Lake Oswego, OR).


Individual neurons were visualized with a ×40 water-immersion objective (Olympus). For the photolysis of caged glutamate (200 μM, MNI-glutamate; Tocris), the output of a continuous emission 5-W krypton ion laser (Coherent, Innova 302) with a 351-nm line was delivered, via a multimode optical fiber, through an Olympus ×40 water-immersion objective to form an uncaging spot about 6 μm wide with a final power of 5–10 mW (Pettit et al. 1997; Yang et al. 2007). We estimate this will produce a glutamate concentration of 100 μM (Canepari et al. 2001; Wang and Augustine 1995). An acousto-optical modulator was used to vary the duration of the light pulse between 1 and 2 ms. The uncaging spot was positioned over a cellular process by including Oregon Green (200 μM; Molecular Probes, Eugene, OR), in the patch-pipette solution and then visualizing the cell with an Olympus Fluoview 300 confocal microscope.


Activation of extrasynaptic NMDARs by synaptic spillover

Previous reports have identified a substantial pool of extrasynaptic NMDARs constituting about one third of the total dendritic population (Harris and Pettit 2007; Rosenmund et al. 1995; Thomas et al. 2006). Although extrasynaptic NMDARs are inactive during low-frequency synaptic stimulation, increased synaptic input might activate these receptors, allowing them to participate in postsynaptic physiology. The CA3 neurons that activate CA1 pyramidal cells can fire at low frequencies or in short bursts, reaching firing frequencies of 400 Hz (Kandel and Spencer 1961; Wong and Prince 1981). We therefore tested the ability of short, 10-pulse bursts of electrical stimuli at 5–400 Hz to activate extrasynaptic NMDARs. Because these experiments test for synaptic spillover, they rely on temperature-dependent glutamate transporter function (Asztely et al. 1997). As a result, experiments were performed at 34°C to maximize transporter function. The peak amplitude of currents elicited by these stimulus bursts represents a mixture of synaptic and any potential extrasynaptic NMDAR activation.

After establishing a baseline of stable synaptic NMDAR current amplitudes (30-s intervals), we stimulated the presynaptic inputs with 5- to 400-Hz 10-pulse bursts. Stimulation strength was set to 40–200 μA for 100–200 μs, which elicited current amplitudes from 129 ± 20 pA (0.1 Hz) to 1,049 ± 155 pA (175 Hz; Fig. 1A, n = 10). Each frequency was tested three times, separated by 30 s, and presented in random order to avoid any systematic errors that might be introduced by stimulating with multiple high- or low-frequency bursts. We measured the peak amplitude of each burst and normalized by the largest-amplitude excitatory postsynaptic current (EPSC) to allow for comparison between cells (Fig. 1, B and C, arrows). Peak current amplitude increased with frequency, reflecting greater temporal summation, and reached a plateau roughly tenfold higher than baseline at frequencies between 100 and 200 Hz (Fig. 1C; n = 10). Stimulation frequencies >200 Hz elicited reduced current amplitudes corresponding to 80% of maximum.

FIG. 1.

High-frequency stimulation increases synaptic current amplitude 10-fold. A: a plot of the peak current amplitude for 10-pulse bursts delivered at 5–400 Hz (n = 10). Stratum radiatum inputs were stimulated every 30 s. B: individual whole cell N-methyl-d-aspartate receptor (NMDAR) currents from a single cell stimulated at multiple frequencies. C: a plot of the normalized peak current amplitude for all cells (n = 10). Current amplitude increased with frequency to a maximum that dropped off about 20% at frequencies >200 Hz. Inset: 175-Hz trace from B on an expanded timescale. Arrows indicate location of amplitude measure. D: a plot of the 10–90% decay times for currents elicited with bursts from 5–400 Hz. To compare the decay times across frequencies, we measured the decay times for each burst within a cell and normalized by the longest decay time. Decays increased with frequency so that decays for bursts were about double that of single stimulations (single = 76.6 ± 6.6 ms, 125 Hz = 138 ± 20.4 ms).

To further characterize currents arising from high-frequency bursts, the current decay was analyzed for all frequencies. If high-frequency trains increase glutamate concentrations and there is diffusion to adjacent extrasynaptic receptors, we would expect an increase in the decay time. Because currents elicited by the high-frequency bursts could not be fit by single or double exponentials (see Fig. 1C, inset), we measured the 10 to 90% decay times. To compare the decay times across cells, we measured the decay times for each burst within a cell and normalized by the longest decay time (Fig. 1D). Decay times grew longer with increases in burst frequency relative to those measured for the single stimulation (Single = 76.6 ± 6.6 ms vs. 125 Hz = 138 ± 20.4 ms; n = 10). Surprisingly, we observed an increase in decay time with 10-Hz stimulation bursts that returned to that of a single stimulation at very high frequencies (400 Hz; Fig. 1D).

As can be seen in Fig. 1A, current amplitudes for bursts of 100–200 Hz can reach 1–2 nA. With currents this large, it is possible that voltage-clamp errors could interfere with our decay measurements. However, it should be noted that the decays increase for 10-Hz bursts whose current amplitudes average 300 pA. With currents this small there should be no loss of voltage control (Arnth-Jensen et al. 2002; Diamond 2001). Finally, we compared the impact of low concentrations of the high-affinity NMDAR antagonist (±)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP, 250–500 nM) on synaptic and 200-Hz burst amplitudes and found no significant difference (current remaining after CPP application: synaptic = 54 ± 5%, bursts = 61 ± 12%, paired Student's t-test; P > 0.5; n = 4). The 10–90% decay was also measured for currents elicited by 200-Hz bursts before and after application of CPP. Again, there was no significant difference in the decays (201 ± 48 ms before and 210 ± 43 ms after; paired Student's t-test; P > 0.25; n = 4). These results also suggest no loss of voltage control.

Although increases in decay time are consistent with increased glutamate spillover to extrasynaptic receptors (Arnth-Jensen et al. 2002; Diamond 2001; Lozovaya et al. 2004; Scimemi et al. 2004), this is a somewhat indirect measurement. An additional complication is that the increase in decay time could represent spillover to adjacent synapses rather than to extrasynaptic receptors. To properly address this issue it is necessary to remove synaptic receptors so that extrasynaptic NMDAR activation can be examined in isolation.

Conversion of extrasynaptic NMDARs to synaptic NMDARs

To directly test whether high-frequency bursts activate extrasynaptic NMDARs, we blocked synaptic NMDARs by stimulating at low frequency (0.1 Hz; Fig. 2 Supplemental Fig. S1)1 in the presence of MK-801 (500 μM) applied by picospritzer. Following the convention of physiologists in the field, synaptic NMDARs were defined as the functional receptors that respond to glutamate released during low-frequency (0.1-Hz) synaptic events (Chen and Diamond 2002; Clark and Cull-Candy 2002; Lozovaya et al. 2004; Scimemi et al. 2004; Tovar and Westbrook 2002). Extrasynaptic NMDARs do not respond to this glutamate within the time frame of our experiments, although they can be activated by the application of exogenous glutamate (Hardingham et al. 2002; Harris and Pettit 2007; Thomas et al. 2006). Because NMDARs are subject to random stochastic channel openings, MK-801 blocking time was kept to a maximum of 30 min. Average time in MK-801 was 17 ± 2.5 min (range 7–27 min, n = 10). We previously determined that there was no relationship between the time in MK-801 and the size of the unblocked extrasynaptic receptor pool within this time frame (Harris and Pettit 2007).

FIG. 2.

Increasing stimulation strength recruits additional fibers in a linear manner. A: fiber volley (FV) traces at 2 stimulation strengths before and after application of tetrodotoxin (TTX, 1 μM). Each trace is an average of 5 trials. Post-TTX FVs were subtracted from pre-TTX traces and the subtraction traces are shown on the right. B: a plot of FV amplitude vs. time for a range of stimulation strengths. C: a plot of normalized FV amplitude vs. the change in stimulation strength across all experiments (n = 4). Data were normalized for the lowest stimulation strength tested and pooled.

Following block of synaptic NMDARs, we attempted to activate extrasynaptic NMDARs by stimulating the presynaptic inputs with 5- to 400-Hz 10-pulse bursts (Fig. 3A). Since we wished to test the effect of multiple frequencies in single cells, we applied small quantities of MK-801 (500 μM) via picospritzer to facilitate MK-801 washout. In separate experiments we determined that 20 min was sufficient for complete MK-801 washout, as evidenced by stable current amplitudes post wash (Harris and Pettit 2007) (Supplemental Fig. 1).

FIG. 3.

Extrasynaptic NMDARs can participate in synaptic transmission. A: to obtain a measure of extrasynaptic current amplitude, synaptic currents were blocked at 0.1 Hz in the presence of MK-801. Following washout of MK-801, 10 pulses were delivered at intervals between 5 and 400 Hz. B: an individual experiment showing blockade of synaptic NMDARs and activation of extrasynaptic NMDARs. For purposes of clarity only a subset of tested frequencies are plotted. Traces correspond to currents from the baseline, just prior to MK-801 block and postblock (averages of 3 trials). Post MK-801 amplitudes are averages of 3 trials. C: individual whole cell NMDAR currents from a single cell stimulated at multiple frequencies. Extrasynaptic currents were elicited with 25-Hz stimulation bursts. D: a plot of the normalized peak extrasynaptic current amplitude for all cells (n = 10). Significant activation was observed at 25 Hz (ANOVA, P < 0.001), with peak activation between 100 and 200 Hz. Arrows indicate location of amplitude measure. E: the amplitude of extrasynaptic NMDAR current normalized to the amplitude of the single synaptic current. Maximal extrasynaptic current was 33 ± 8.5% at 200 Hz. F: extrasynaptic activation was independent of stimulation strength. Current amplitudes for a single stimulation were placed in 50-pA bins and plotted against extrasynaptic amplitude at 25 and 175 Hz (black and gray bars, respectively) as a percentage of the single synaptic amplitude. Traces are averages of 3 trials.

It is possible that high-frequency bursts could depolarize additional axon fibers that were subthreshold during low-frequency stimulation. If these new axon fibers were to participate at high frequencies, they might elicit synaptic NMDAR currents not blocked by MK-801 during low-frequency stimulation. To address this concern we increased stimulation strength twofold during synaptic blockade. We confirmed that this approach engaged additional fibers by measuring the size of fiber volleys elicited in this manner (Fig. 2, A and B). Fiber volleys represent the sum of extracellular voltage change produced when axons fire action potentials and are linearly related to the number of axons firing. Volley amplitude was quantified by subtracting post-tetrodotoxin (TTX, 1 μM) traces for each stimulation strength. Within the range of stimulation used for these experiments, fiber volley amplitudes doubled when either the duration or magnitude of our stimulation was doubled (Fig. 2C). Although we cannot guarantee that every axon activated by high-frequency trains was engaged by doubling stimulation strength, this approach limits the problem.

Bursts of stimulation elicited NMDAR currents that increased in amplitude with frequency (Fig. 3, B and C). We found that significant extrasynaptic NMDAR activation occurred at 25 Hz (Fig. 3D; ANOVA, P < 0.001; n = 10). Average current amplitude for 25-Hz bursts was 20 ± 4 pA (n = 10). Although our decay data suggested activation of extrasynaptic NMDARs at 10 Hz, the difference between single-stimulation amplitudes (4 ± 1 pA) and 10-Hz bursts (10 ± 2.5 pA) was not significant. Even though we used 10-pulse bursts, extrasynaptic currents developed by the second or third stimulation (Fig. 3C). We confirmed that these were NMDAR currents by blocking them with the NMDAR antagonist APV (50 μM; Supplemental Fig. 1).

Stimulation bursts between 100 and 200 Hz produced maximal extrasynaptic NMDAR current amplitudes (65 ± 13 pA). Surprisingly, extrasynaptic NMDAR activation decreased by about 50% at higher frequencies (400 Hz). This is consistent with the decreased decay time associated with currents at this frequency (Fig. 1D), suggesting decreased diffusion to extrasynaptic NMDARs. In seven cases it was possible to test both the synaptic and extrasynaptic activation in the same cell. These results were in agreement with between-cell comparisons.

To quantify the impact of unmasking extrasynaptic receptors, we compared the amount of current elicited with high-frequency stimulation (5–100 Hz) to baseline (0.1-Hz) synaptic stimulation. The amplitude of extrasynaptic NMDAR current elicited by high-frequency bursts was normalized to the amplitude of the single synaptic current. The maximal extrasynaptic current elicited in this manner was 33 ± 8.5% (Fig. 3E; n = 10), suggesting that extrasynaptic receptors can produce a substantial increase in synaptic power. Activation of extrasynaptic receptors was independent of stimulation strength because our baseline synaptic currents ranged from 33 to 275 pA (Fig. 3F). The relatively low frequencies that engage extrasynaptic NMDARs suggest that these receptors regularly contribute to synaptic communication.

Irreversibility of MK-801 blockade

To ensure that blocked synaptic receptors remained unavailable for the duration of our experiments, single stimulations were interleaved with stimulus trains in a subset of experiments (Fig. 4, A and B). If spontaneous or burst-induced MK-801 unblocking occurred we would expect that single stimulations would begin to elicit current. Across all experiments there was no significant increase in residual current over the course of the experiment (0.7 ± 0.3 pA; paired Student's t-test, P < 0.3; n = 8; Fig. 4, C and D). These results are consistent with our previous findings that show no current recovery of synaptic NMDARs up to 1 h after MK-801 blockade (Harris and Pettit 2007).

FIG. 4.

MK-801 blockade is irreversible for the duration of our experiments. To test for spontaneous unblocking by MK-801, we interleaved single-pulse synaptic stimulation with high-frequency stimulus bursts. A: traces before and after MK-801 blockade. Black traces are single stimulations (averages of 3 trials), whereas the grey trace corresponds to a representative 175-Hz burst (average of 3 trials). Numbers correspond to locations in B. B: when current amplitude vs. time is plotted for an individual experiment, no current recovery was observed, confirming that synaptic NMDARs remained blocked. C: a plot of current amplitude as a percentage of baseline immediately after MK-801 washout (early) and at the end of the experiment (late) (n = 8). Values were calculated from averages of the first 3 single stimulations after wash and the last 3 single stimulations. D: a plot of current amplitude for the data summarized in C (n = 8). No significant increase in current amplitude was observed (paired Student's t-test; P > 0.3).

Presynaptic axons fail at high frequencies, limiting extrasynaptic NMDAR activation

We found the large decrease in extrasynaptic NMDAR activation at 300–400 Hz puzzling since a reduction in the peak current amplitude implies less glutamate availability and spillover during these bursts. Because the presynaptic terminals supply the glutamate during these bursts, we looked for a presynaptic cause for this decline in glutamate. The amplitude of the fiber volley (i.e., the extracellular potential generated by the summed activity of axons firing action potentials) represents one measure of the activity of presynaptic axons. We determined whether high-frequency stimulation changed the ability of the axon to fire action potentials by comparing the amplitude of the first fiber volley in a burst to the average of the subsequent fiber volleys induced by pulses 2–10 (Fig. 5A). Blocking excitatory and inhibitory postsynaptic currents with NBQX (5 μM), APV (50 μM), and picrotoxin (100 μM) allowed us to isolate the fiber volley. Figure 5A shows that amplitudes of the first and tenth fiber volleys are similar for low frequencies such as 5 Hz. However, with high-frequency bursts (e.g., 400 Hz), the fiber volley is reduced by an average of 50% (Fig. 5, B and C). Because there is a relationship between fiber volley amplitude and glutamate release, 400-Hz bursts are likely to release substantially less glutamate than 200-Hz bursts. Therefore at 400 Hz there is less glutamate available for spillover to extrasynaptic receptors. At frequencies >200 Hz we observed a sawtooth pattern in the fiber volley amplitude (Fig. 5B), which suggests that alternating groups of fibers are refractory from pulse to pulse. It is apparent that action potentials are diminished during high-frequency stimulation bursts (Fig. 5C; n = 5). The reason for this failure is probably a combination of Na+ channel inactivation, decreases in driving force, depletion, and/or failure of release machinery. That this decrease in the fiber volley has such a dramatic effect on the extrasynaptic NMDARs, while leaving the synaptic NMDAR response relatively unscathed, suggests that the extrasynaptic receptors are comparatively unsaturated and remain acutely sensitive to changes in presynaptic glutamate concentration.

FIG. 5.

Transmission fails at high-input frequencies. A: to test for the ability of presynaptic fibers to follow stimulation bursts, glutamate transmission was blocked and extracellular fiber volleys were recorded from stratum radiatum. Stimulus artifacts have been removed for purposes of display. Fiber volleys 1 and 10 of a 5-Hz burst are the same size. B: an example of the fiber volley elicited during a 400-Hz burst. Fiber volley amplitude is substantially decreased by the second stimulation. C: a plot of the ratio between volley 1 and the average of fiber volleys 2–10. Fiber volley amplitudes are relatively stable at ≤200 Hz.

Extrasynaptic NMDARs are exposed to less glutamate than synaptic NMDARs

To examine the extent of occupancy of the extrasynaptic receptor pool, we have tested the ability of the low-affinity NMDAR antagonist l-APV to block extrasynaptic and synaptic currents. If the extrasynaptic NMDARs are exposed to a lower concentration of glutamate than the synaptic receptors, low doses of l-APV (250 μM) will produce significant blockade of extrasynaptic current with little effect on synaptic currents (Chen and Diamond 2002). Two sets of experiments were performed. First, we examined the effect of l-APV on whole cell NMDAR currents elicited by synaptic stimulation (Fig. 6, A and C; n = 5). On average, l-APV reduced synaptic currents by 23 ± 5% (n = 5). To test the relative receptor occupancy underlying extrasynaptic currents, MK-801 was used to eliminate all synaptic NMDARs. Following washout of MK-801, 10-pulse bursts at 200 Hz were used to activate extrasynaptic NMDARs. Subsequent l-APV (250 μM) application reduced extrasynaptic currents significantly more (50.1 ± 4%; n = 5; Fig. 6, B and C) than synaptic currents (P < 0.0025; Student's t-test). Finally, d-APV was used to block all remaining current (Fig. 6B).

FIG. 6.

Extrasynaptic NMDARs are less saturated than synaptic NMDARs. A: synaptic NMDAR currents elicited by single stimulations were relatively insensitive to blockade by the weak competitive NMDAR antagonist l-2-amino-5-phosphonovaleric acid (l-APV, 250 μM; in this example 23%). Inset: averages of 3 traces before and after l-APV application. B: synaptic currents were eliminated with MK-801, followed by 20-min washout. Extrasynaptic NMDAR currents were elicited by 10-pulse bursts of synaptic stimulation (200 Hz). l-APV (250 μM) blocked a substantial amount of extrasynaptic current (44%). Subsequent application of d-2-amino-5-phosphonovaleric acid (d-APV, 50 μM) blocked the remaining current. Inset: single representative traces from each condition. Peak amplitude was measured at the location indicated by the dashed line. C: across all experiments extrasynaptic NMDAR currents were significantly more sensitive to blocking by l-APV (50.1 ± 4%; n = 5) than single synaptic currents (23 ± 5%; n = 5; P < 0.0025; Student's t-test).

l-APV was also applied to NMDAR currents elicited by 10-pulse bursts of synaptic stimulation at 200 Hz. Although we expect activation of extrasynaptic NMDARs with this stimulation protocol, the majority of the current is generated by synaptic receptors, which should be less susceptible to low-affinity antagonists. l-APV had little effect on these currents reducing peak amplitude by 9.6 ± 3% (n = 5; data not shown). Although the degree of block for single synaptic and burst synaptic currents was not significant, extrasynaptic currents were reduced significantly more than either synaptic current (ANOVA, P < 0.0001; n = 5). These data indicate that the extrasynaptic NMDARs activated by high-frequency bursts are exposed to less glutamate than are synaptic receptors. If bursts of stimulation recruited new synaptic inputs, we would expect an equivalent blockade by l-APV. Because these receptors behave differently from synaptic receptors they are likely activated by glutamate spillover.

Presynaptic release maximally activates extrasynaptic NMDARs

Our data suggest that extrasynaptic NMDARs participate in synaptic communication as short bursts of relatively low frequency (25 Hz) stimulation activate extrasynaptic NMDARs. However, it was unclear how much of the pool of total dendritic NMDARs could be activated by high-frequency stimulation. If only a subpopulation of extrasynaptic receptors can be activated by synaptic stimulation the remainder might function only under pathological conditions. To test this possibility, we blocked all accessible receptors with MK-801 and probed for a residual population of extrasynaptic NMDARs with exogenous application of glutamate. We previously showed that the photolysis of caged glutamate accesses all synaptic and extrasynaptic NMDARs (Harris and Pettit 2007). As a result, we used local photolysis (6 μm) of glutamate (200 μM; Tocris) to determine whether any extrasynaptic NMDARs remain unactivated by bursts of synaptic activity. After baseline photolytic currents were established, we used 175-Hz stimulation (the point of maximal extrasynaptic activation) to activate all available NMDARs, in the presence of MK-801 (50 μM in bath). In a separate set of experiments we determined that a large region of the apical dendrite is activated by this protocol (>100 μm; Supplemental Fig. 2). Because the uncaging area was 6 μm, localizing the uncaging beam in the middle of the stimulation area ensures overlap between the two regions (Fig. 7A). Following MK-801 (50 μM in bath) blockade, we sampled for unblocked receptors with photolysis. As can be seen in Fig. 7B, <2% of the initial photolytic excitatory postsynaptic current (phEPSC) remains after MK-801 blockade with 175-Hz stimulation (1.8 ± 1.1%; n = 5). Photolysis at sites distant from electrical stimulation produce current, indicating that high-frequency stimulation produces blockade of extrasynaptic NMDARs specifically on the dendrites receiving input (Fig. 7 B, current 3). These data suggest that under the right conditions, the presynaptic terminals can activate the vast majority of extrasynaptic NMDARs.

FIG. 7.

Synaptic stimulation at 175 Hz activates all accessible dendritic NMDARs. A: a live confocal image of an individual CA1 pyramidal neuron filled with Oregon Green-BAPTA (200 μM). Note the stimulating pipette (S) adjacent to the secondary apical dendrite. A circle (1) and square (3) indicate the size and locations of the photolytic areas. B: individual currents generated at the time points indicated in C. C: a plot of current amplitude vs. time for synaptic and photolytic excitatory postsynaptic currents (phEPSCs). To test for maximal extrasynaptic NMDAR activation synaptic currents (filled circles) were blocked using MK-801 (50 μM in bath) while stimulating at 175 Hz. Following MK-801 washout no current could be elicited by photolysis (open circles), indicating that all receptors were blocked (compare currents 1 and 2 in B). Subsequent photolysis at a different location (open square in A) elicits current (open square, current 3 in B). Inset: bar graph showing the average current remaining for all experiments (1.76 ± 1.14%).


Extrasynaptic receptors are of limited interest unless they can be activated under physiological conditions. Here we show that short, physiological stimulation bursts activate these receptors at relatively low frequencies (25 Hz). Stimulation at higher frequencies recruited additional extrasynaptic receptors with maximal activation occurring between 100 and 200 Hz. Because the CA3 axons that activate CA1 pyramidal cells fire in this range of frequencies (Kandel and Spencer 1961), it is clear that extrasynaptic receptors routinely participate in neural communication. In fact, similar bursts of synaptic input occur during exploratory behavior in rats (Otto et al. 1991).

These results are consistent with findings from α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)–only synapses in the cerebellum and retina, where stimulating at increased frequency (Clark and Cull-Candy 2002) or increasing the quantal content of synaptic release (Chen and Diamond 2002) activates extrasynaptic NMDARs. Previous work with mixed NMDA/AMPA-receptor synapses has shown that high-frequency trains or blockade of glutamate transporters can increase the decay time constant of synaptic currents (Arnth-Jensen et al. 2002; Lozovaya et al. 2004). However, increases in decay time can be interpreted in a number of ways, including diffusion to adjacent synapses rather than to extrasynaptic NMDARs. In fact, we observed an increase in decay time with 10-Hz stimulation bursts, which did not produce significant extrasynaptic NMDAR current following the removal of synaptic NMDARs. The dissociation between prolonged decay and extrasynaptic receptor activation highlights the importance of directly studying extrasynaptic NMDARs.

The requisite increase in nonsynaptic glutamate that results in extrasynaptic NMDAR activation could be a consequence of presynaptic facilitation and/or pooling of glutamate during high-frequency trains. Glutamate transporter clearance rates have been measured at 8 ms (Diamond and Jahr 2000). Because the interstimulus interval for a 200-Hz train is 5 ms, our high-frequency stimulus trains must produce pooling of glutamate. Synaptic NMDAR currents from CA1 hippocampal synapses have been shown to produce facilitation of the second response in a train followed by depression for subsequent events in a train (Dittman et al. 2000). Since our extrasynaptic currents continue to increase in amplitude after the third stimulation, extrasynaptic NMDAR activation likely results from a combination of presynaptic facilitation and glutamate buildup with repetitive stimulation.

It should be noted that in vitro stimulation protocols synchronously activate groups of axons synapsing in close proximity on the dendrite. In the intact brain, pyramidal neurons are subject to intense synaptic bombardment (Paré et al. 1998) that may arrive in a more asynchronous and distributed manner. As a result, it could be more difficult to elicit spillover to synaptic receptors in vivo than in vitro. Although it would be interesting to know whether the activity of a single axon can activate extrasynaptic NMDARs, our maximum extrasynaptic current amplitude represents 33 ± 8.5% of the single synaptic current amplitude (Fig. 3E). Assuming that single synaptic NMDA conductance is about 3 pA (Pankratov and Krishtal 2003), we need to activate 10–20 synapses (or 30–60 pA) to resolve the extrasynaptic currents, making examination of a single synapse impractical. Nonetheless, we observe substantial extrasynaptic activation with stimulation strengths that elicit 30–70 pA synaptic currents. These currents represent the coactivation of a relatively small number of synapses and demonstrate substantial spillover to extrasynaptic receptors at the limit of our resolution. Further, previous work has demonstrated that groups of pyramidal neurons fire during sharp-wave electroencephalographic activity in the CA1 region (Buzsáki et al. 1983). These waves have recently been shown to result from the synchronous firing of CA3 pyramidal neurons (Csicsvari et al. 2000). As a result, stimulating coherent groups of axons simulates the natural synchronous input from CA3 neurons.

Although we have argued that the NMDAR current in response to high-frequency stimulation bursts represents activation of extrasynaptic NMDARs, it is possible that these receptors are actually unblocked synaptic receptors. This could occur if high stimulation frequencies recruited additional axon fibers that were subthreshold at low-frequency stimulation. To avoid this complication we doubled stimulation strength during MK-801 blockade and returned to the lower stimulation strength to probe for extrasynaptic activation (Fig. 3A). This approach activates a larger population of axon fibers during MK-801 blockade and reduces the probability of recruiting new fibers during stimulation bursts. Further, these receptors behave differently from synaptic receptors because they are more sensitive than synaptic currents to blockade by the low-affinity competitive antagonist l-APV (Fig. 6). Because glutamate must spill over and diffuse to extrasynaptic receptors they should be exposed to a lower concentration of glutamate than synaptic receptors and more sensitive to l-APV. Finally, we have previously shown that our stimulation protocol activates all functional synapses on the targeted dendritic region (Harris and Pettit 2007). These experiments used two stimulating pipettes, one placed 100 μm from the test dendrite and a second placed as close as possible to the dendrite. Stimulation strength was adjusted so that the proximal pipette was activating a subset of the fibers activated by the distal electrode. All synaptic current generated by the distal electrode was then blocked with MK-801. If our approach activated and blocked only a fraction of the synapses on the target dendrite, we should have been able to elicit synaptic currents with the proximal stimulating electrode. This electrode was sufficiently close to the dendrite to activate all local functional synapses by direct depolarization. Because no synaptic current was elicited by the proximal electrode, our MK-801 protocol eliminates all functional synapses on the targeted dendritic region.

It is also possible that the currents elicited by high-frequency bursts arise exclusively from cut or nonfunctioning synapses in our preparation rather than from extrasynaptic NMDARs. However, electron microscope analysis has demonstrated that NMDARs are present in both synaptic and extrasynaptic locations (Kohr et al. 2003; Takumi et al. 1999). Further, we limited the possibility of encountering a large population of nonfunctional synapses by choosing neurons deep in the slice to avoid cells with damaged dendritic trees. Using this approach we have previously shown that extrasynaptic NMDARs constitute about one third of the total dendritic pool (Harris and Pettit 2007). This result is similar to that seen in autaptic cultures (synaptic pool sizes of 81 ± 4 and 71 ± 3%) in which the number of nonfunctional synapses should be very limited (Rosenmund et al. 1995; Thomas et al. 2006). Although we cannot rule out the possibility that there is spillover to some unblocked synaptic receptors, these receptors are no more likely to be activated than the extrasynaptic receptors at functional synapses, and it should be noted that activation of adjacent synapses is also a method of synaptic signal amplification (Barbour 2001; Faber and Korn 1988; Scimemi et al. 2004).

Our experiments testing how much of the extrasynaptic population can be accessed by synaptic glutamate demonstrate that all dendritic NMDARs can be recruited to the synaptic compartment. When short, 175-Hz bursts (point of maximal extrasynaptic activation) are delivered in the presence of MK-801, all dendritic NMDAR current is blocked (Fig. 7). Because extrasynaptic NMDARs are immobile (in that they do not move into the synaptic compartment) for the period of our experiments in acute slices (Harris and Pettit 2007), spillover with presynaptic bursting represents a physiological method to activate these receptors. The ability to increase the synaptic receptor pool provides a mechanism to transiently increase synaptic strength as needed. These results suggest that current concepts about “extrasynaptic” NMDARs should be revisited.

At first glance, the notion that extrasynaptic NMDARs on CA1 dendrites participate regularly in synaptic communication seems at odds with previous studies reporting that extrasynaptic and synaptic receptors are differentially coupled to downstream signaling pathways (Hardingham and Bading 2002; Hardingham et al. 2002). However, in those studies, bath application of agonist activated extrasynaptic NMDARs located on both the dendrite and the soma. As a result, it is possible that the activation of extrasynaptic NMDARs located on the cell body is necessary to initiate extrasynaptic NMDAR-specific pathways. Alternatively, extrasynaptic receptor-mediated pathways may occur only when activated in specific spatiotemporal patterns. Thus, only continuous activation of extrasynaptic receptors for minutes, or the simultaneous activation of extrasynaptic receptors along the entire cell, would trigger apoptosis. Finally, although burst activity activates extrasynaptic receptors, it likely preferentially activates synaptic NMDARs. As a result, synaptically coupled pathways may predominate under these conditions (Hardingham et al. 2001; Liu et al. 2004).

In addition to direct synaptic communication, extrasynaptic receptors may also participate in neural signaling by “volume transmission” or the diffusion of neurotransmitter through the extracellular space (Agnati et al. 1995). Whereas recent estimates of the ambient glutamate concentration in brain tissue are low (25 nM), this level of glutamate could provide for significant activation of these receptors (Herman and Jahr 2007). In fact, extrasynaptic NMDARs have been shown to contribute a small tonic conductance in pyramidal neurons (Herman and Jahr 2007; Le Meur et al. 2007; Sah et al. 1989). Extrasynaptic NMDARs may also play a role in synchronizing hippocampal neural activity through glutamate release by glia (Angulo et al. 2004; Fellin et al. 2004). In addition, our data suggest that glutamate released during routine synaptic signaling can also active extrasynaptic receptors. Therefore extrasynaptic NMDARs have two methods of signaling. They can transmit low temporal resolution volume information and can be recruited by high-frequency bursts to encode temporally precise synaptic information.


This work was supported by National Institutes of Health Grants R21 NS-051536 and K22 ES-00359 and the Whitehall Foundation.


We thank Drs. M. V. L. Bennett, R. Carroll, and K. Khodakhah and B. Heifets for helpful discussions.


  • 1 The online version of this article contains supplemental data.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract