Regulation of IPSP Theta Rhythm by Muscarinic Receptors and Endocannabinoids in Hippocampus

Christian G. Reich, Miranda A. Karson, Sergei V. Karnup, Lauren M. Jones, Bradley E. Alger


Theta rhythms are behaviorally relevant electrical oscillations in the mammalian brain, particularly the hippocampus. In many cases, theta oscillations are shaped by inhibitory postsynaptic potentials (IPSPs) that are driven by glutamatergic and/or cholinergic inputs. Here we show that hippocampal theta rhythm IPSPs induced in the CA1 region by muscarinic acetylcholine receptors independent of all glutamate receptors can be briefly interrupted by action potential–induced, retrograde endocannabinoid release. Theta IPSPs can be recorded in CA1 pyramidal cell somata surgically isolated from CA3, subiculum, and even from their own apical dendrites. These results suggest that perisomatic-targeting interneurons whose output is subject to inhibition by endocannabinoids are the likely source of theta IPSPs. Interneurons having these properties include the cholecystokinin-containing cells. Simultaneous recordings from pyramidal cell pairs reveal synchronous theta-frequency IPSPs in neighboring pyramidal cells, suggesting that these IPSPs may help entrain or modulate small groups of pyramidal cells.


Neuronal theta rhythm oscillations represent prominent electrophysiological phenomena that have been linked to many behavioral events (Bland 1986; Buszaki 2002). Theta rhythms are especially evident in the hippocampus, and much work has gone into explicating the mechanisms that modulate them. More than one mechanism has been identified and GABAergic synaptic inhibition is a major factor in many cases (for reviews see Buzsaki 2002; Whittington and Traub 2003). For example, hippocampal field theta rhythms in vivo can be generated from intricate interactions among the cholinergic and GABAergic inputs from the medial septum-diagonal band of Broca (MSDB) and glutamatergic inputs from the entorhinal cortex (Buzsaki 2002). Interestingly, in vitro models of field theta rhythms have convincingly shown that circuitry intrinsic to the hippocampus can generate theta oscillations in the absence of extrinsic inputs. These intrahippocampal rhythms are dependent on inhibitory postsynaptic potentials (IPSPs) and are either initiated in CA3 by muscarinic acetylcholine receptors (mAChRs) and projected by excitatory axons onto the CA1 region (Fellous and Sejnowski 2000; Fisahn et al. 1998; Fischer 2004; Palhalmi et al. 2004; Willams and Kauer 1997) or generated within CA1 by strong activation of group I metabotropic glutamate receptors (mGluRs) (Gillies et al. 2002). However, mAChR agonists activate CA1 interneurons (Chapman and Lacaille 1999; McMahon et al.1999; McQuiston and Madison 1999; Pitler and Alger 1992b) even when ionotropic glutamate responses are inhibited, and an independent interneuronal network might be activated by mAChRs in CA1. We were particularly interested in the possibility of theta-rhythm IPSP regulation by endocannabinoids because previous work had demonstrated that exogenous cannabinoids regulate hippocampal gamma rhythm IPSCs initiated by kainic acid (Hajos et al. 2000).

We find that, when activated by mAChRs, interneurons within the CA1 region of rat hippocampal slices generate a theta pattern of IPSPs in pyramidal cells independent of input from CA3 or local mGluR activation. This suggests that a neural substrate within CA1 is capable of generating theta IPSPs when provided with the appropriate input (i.e., septal cholinergic afferents). Importantly, the mAChR-initiated rhythm is reversibly interrupted by the release of endocannabinoids from the pyramidal cells, a process called depolarization-induced suppression of inhibition or DSI (for review see Alger 2002), possibly implicating cholecystokinin (CCK)-containing interneurons in the generation of theta-rhythm IPSPs. Finally, we observe synchronous theta-rhythm IPSPs in simultaneously recorded pyramidal cells, suggesting that these IPSPs may help entrain or modulate theta at least in small groups of pyramidal cells.


Male Sprague–Dawley rats approximately 4–6 wk old (Charles River Laboratories) were deeply anesthetized with halothane and decapitated in accordance with the guidelines set forth by the Institutional Animal Care and Use Committee of the University of Maryland School of Medicine. The brain was rapidly removed and hippocampi were dissected free. Transverse hippocampal slices (400 μm thick) were cut on a Vibratome (Series 1000; Technical Products International). Slices were kept in a holding chamber at room temperature at the interface of physiological medium and a humidified gas mixture of 95 O2-5% CO2 for >1 h. For most experiments, the slices were then transferred to a submersion-type recording chamber (Nicoll and Alger 1981) and perfused with physiological saline warmed to 30°C. For dual recordings from pyramidal cells, the recording chamber was kept at room temperature, and the cells were visualized through differential interference contrast optics. The extracellular medium contained (in mM): NaCl, 120; KCl, 3; MgSO4, 2; NaH2PO4, 1; NaHCO3, 25; CaCl2, 2.5; and glucose, 10 and was saturated with a gas mixture of 95% O2-5% CO2 (pH 7.4). The α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) antagonist 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfoamide(NBQX, 10 μM), the N-methyl-d-aspartate receptor (NMDAR) antagonist dl-(−)-2-amino-5-phosphovaleric acid (APV, 50 μM), the broad mGluR antagonist LY341495 (100 μM), the group I mGluR agonist dihydroxyphenylglycol (DHPG, 50 μM), and the CB1R antagonist AM251 (3 μM) were purchased from Tocris Cookson. The mAChR agonist carbamylcholine [carbachol (CCh)] were purchased from Sigma. All drugs were prepared as stock concentrated solutions and diluted by 1:1,000 in saline. In experiments with AM251, slices were preincubated in the holding chamber for ≥1 h.

Current-clamp recordings from CA1 pyramidal cells were obtained using sharp, glass electrodes (50–150 MΩ) filled with 3 M KCl. High Cl–containing electrodes facilitated the observance of spontaneous inhibitory postsynaptic potentials (sIPSPs) by causing these events to be large (1–12 mV) and depolarizing. Acceptable cells had resting potentials > −60 mV. Whole cell recordings from visualized pyramidal cells were made with pipettes containing (in mM): CsCH3SO3, 90; MgCl2, 1; CsCl, 50; Mg-ATP, 2; Cs4-BAPTA, 0.2; HEPES, 10; Tris-GTP, 0.3 (pH 7.2). Reliable sIPSP activity (seen in nearly 80% of all recorded cells) was observed after a 5- to 20-min bath application of CCh, NBQX, and APV. Action potentials were induced in the CA1 cells either by trains of 10-ms DC pulses (1–50 pulses per train) or by “theta burst” trains of action potentials, each comprising five action potentials elicited by positive DC pulses (each 5 or 10 ms in duration) delivered by the microelectrode at 50 Hz. Three or four such theta bursts were delivered at 200-ms intervals (i.e., at 5 Hz) to elicit DSI. Signals were digitized at 5 kHz with an A/D interface (Digidata 1200, Axon Instruments) and analyzed with pClamp 9.0 software (Axon Instruments). For analysis of theta rhythms, the data were filtered at 200 Hz with a low-pass, eight-pole Bessel filter (Frequency Devices). Power spectrum analysis and autocorrelations were performed using Clampfit 9.0 (Axon Instruments). Peak power was defined as the largest power value between 2 and 50 Hz during 10 s of sIPSP activity; however, because of intercellular differences, comparison of peak power across cells was not a reliable measure. Instead we calculated a proportional metric of theta power {Relative Theta Power = [absolute theta power (4–14 Hz)/absolute total power (2–50 Hz)] × 100, where absolute power = summation of all spectral values in a given range}, for the same period of activity. This calculation resulted in a consistent Relative theta power of approximately 60% across all baseline experimental conditions. For further statistical comparisons, any Relative theta power value after an experimental manipulation was expressed as a percentage of the baseline value. One-way repeated-measures ANOVAs with subsequent Newman–Keuls post hoc tests (P < 0.05) were performed on mean group percentages. IPSP amplitude and kinetic analyses were conducted with MiniAnalysis software (Synaptosoft, Atlanta, GA). Only events with amplitudes ≥1.5 mV and area thresholds ≥10 were considered sIPSPs.


Conventional intracellular sharp electrode recordings were carried out in the CA1 region of acute hippocampal slices from rats. A high concentration of KCl in the recording electrode inverted the IPSPs into depolarizing potentials (Fig. 1) and ionotropic glutamate antagonists APV and NBQX were present in all experiments to block fast glutamate synapses. Under these conditions, bath application of 5 μM CCh initiates ongoing IPSP activity exclusively through muscarinic receptors (Martin and Alger 1999; Pitler and Alger 1992b).

FIG. 1.

Induction of theta rhythm inhibitory postsynaptic potentials (IPSPs) by muscarinic acetylcholine receptor (mAChR) activation. Carbachol (CCh, 5 μM), 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfoamide (NBQX, 10 μM), and dl-(−)-2-amino-5-phosphovaleric acid (APV, 50 μM) present in all experiments. A, left to right: 10-s raw trace of spontaneous inhibitory postsynaptic potential (sIPSP) activity, power spectrum showing peak at 7.63 Hz, autocorrelogram demonstrating the intrinsic theta rhythmicity (see methods for calculation of relative theta power). B: theta sIPSPs persist in the presence of metabotropic glutamate receptor (mGluR) antagonist LY341495 and subsequent addition of mGluR agonist dihydroxyphenylglycol (DHPG) does not affect theta activity in LY341495. Data are from a different cell than that in A. Slight increase in the secondary peak at 10 Hz in the spectrum to the right was not a consistent observation. C: theta sIPSPs are abolished in the presence of mGluR agonist DHPG.

Because of the appearance of rhythmicity in the mAChR-induced IPSPs, we performed power spectral analyses of their activity. The power spectra revealed major peaks in the theta range (4–14 Hz), about 8 Hz in the example of Fig. 1A. In 19 cells, the peak power was at 6.27 ± 0.33 Hz. Atropine (1 μM) completely abolished this activity (data not shown), as expected from previous work (e.g., Martin and Alger 1999). Becauses pacemaker activity from CA3 is normally delivered by ionotropic glutamatergic (iGluR) synapses (e.g., Fisahn et al. 1998; Fischer 2004), and these synapses were blocked in the present experiments, this oscillatory IPSP activity appeared to be intrinsic to CA1. Indeed, surgical removal of either CA3 (n = 4) or CA3 plus subiculum (n = 4) did not prevent the occurrence of rhythmic IPSP activity, which had spectral peaks at 6.50 ± 1.21 and 4.85 ± 0.83 Hz, respectively, in the CA1 region of these slices. Thus interneuronal activity confined to CA1 is capable of generating theta IPSPs.

Field theta rhythms dependent on IPSPs can be initiated within isolated CA1 by the group I mGluR agonist DHPG or the mAChR agonist CCh (Gillies et al. 2002). However, CCh has not been shown to induce intrinsic CA1 theta in the presence of both mGluR and iGluR blockade, and therefore a cooperative role for mGluRs remained possible. We find that a nonselective dose (100 μM) of the mGluR antagonist LY341495 does not prevent mAChR-induced theta IPSPs (e.g., Fig. 1B); Relative theta power in control conditions was 60.6 ± 2.45 and, in the presence of LY341495, it was 60.0 ± 1.02 (n.s., P > 0.1), indicating these two mechanisms for IPSP theta are independent. The inhibition of IPSPs normally produced by 50 μM DHPG [Relative theta power reduced to 12.9 ± 7.98 (P < 0.001), Fig. 1C] was fully prevented by pretreatment of LY341495 (Relative theta power was 63.1 ± 0.88, n.s., P > 0.1, n = 3, Fig. 1B), showing the antagonist was effective. These results imply that the mAChR-activated IPSP activity is independent of all glutamate receptors.

In previous work, we demonstrated that during bath application of CCh, a persistent barrage of large IPSCs could suddenly “switch on” (Martin et al. 2001). However, the rhythmic nature of the activity was not investigated. Martin et al. (2001) also reported that in some cells apparently similar sudden increases in sIPSPs could occur in noradrenaline or 8 mM [K+]o perfusion, although in these cases the activity was rarely as robust or sustained as in CCh. We now report that the theta IPSPs appear quite abruptly in the presence of CCh. During baseline recording conditions overall spectral power was very low with no outstanding peaks (Relative theta power = 50.54 ± 4.08, n = 5; data not shown). sIPSP activity increased in the presence of both CCh and 8 mM [K+]o, although a selective increase in the theta power occurred only with CCh (Relative theta power = 60.64 ± 2.46 vs. 50.56 ± 2.38, 8 mM [K+], P < 0.01 t-test; data not shown). Figure 2A illustrates a typical example of abrupt onset in CCh, and the two power spectra graphs on the top right confirm that the activity is in the theta range. In past work we reported that after the abrupt onset, sIPSP activity in CCh remained constant in about 50% of the tested cells (72/140) for periods of ≫1 h (Martin and Alger 1999; see also Martin et al. 2001). The persistence of this activity for long periods of time meant that receptor desensitization was not a prominent factor. Figure 2B shows an example of a cell recorded for >1 h in which the peak of the theta activity remained relatively constant, although the power declined slightly with time (graphs to the right of the traces). The group data in Fig. 2C show that the percentage of baseline theta remained very stable for periods up to 1 h. The cells in Fig. 2C for which theta power is reported for lesser times were subject to additional experimental treatments or were lost, and so CCh-induced theta could not be tracked further.

FIG. 2.

CCh-induced theta rhythm IPSPs commence suddenly and persist. A: representative 10-s traces of baseline activity directly before adding 5 μM CCh to the bath solution (top) and about 15 min later (bottom). Sudden appearance of large, rhythmic IPSPs at about this time is typical. Power spectra (right) of the 2 traces confirm the sudden onset of theta activity. Data were taken from last 5 s of each trace. B: raw traces taken at the indicated times and corresponding power spectra from a cell with persistent theta activity for 1 h. C: group data (numbers of cells reported for each time point) showing percentage Relative theta power calculated at 5-min intervals. Numbers of cells decrease with time because of either cell loss or the presentation of additional experimental manipulations, not the cessation of theta IPSP activity in a given cell.

Hippocampal field gamma oscillations induced by kainic acid are attenuated by exogenous cannabinoids (Hajos et al. 2000). It was suggested that this resulted from activation of CB1 receptors on CCK–GABAergic presynaptic terminals, and might be caused by endocannabinoids as well, although this was not tested. We hypothesized that endocannabinoids released from pyramidal cells would modulate mAChR-induced theta IPSPs. Depolarization-induced suppression of inhibition (DSI) is a form of presynaptic retrograde inhibition mediated by endocannabinoids released from pyramidal cells that act at GABAergic synapses (Ohno-Shosaku et al. 2001; Wilson and Nicoll 2001). To test whether DSI could modulate theta IPSPs, we used three “theta-burst” trains of action potentials generated in current clamp to induce DSI. Each theta burst comprised five action potentials elicited by positive constant-current pulses, 5 or 10 ms in length, delivered through the microelectrode at 50 Hz, and the three bursts were delivered at 200-ms intervals (i.e., at 5 Hz). This procedure readily induced DSI (about 80% suppression, n = 10) that was prevented by pretreating slices (see methods) with the CB1 receptor antagonist AM251 (Fig. 3, B and C, n = 4). Importantly, interruption of theta rhythm was also readily apparent in the power spectral analysis that shows an almost complete elimination of theta power during DSI (Fig. 3, B and C). Activation of group I mGluRs also releases endocannabinoids through a G-protein–dependent, but Ca2+-independent, biochemical pathway (Maejima et al. 2001; Varma et al. 2001). As noted above, we found that application of 50 μM DHPG abolished theta IPSPs (Fig. 1C) and occluded DSI.

FIG. 3.

Depolarization-induced suppression of inhibition (DSI) disrupts theta IPSP activity. A: continuous raw trace of sIPSP activity illustrating DSI induced by 3 trains of theta-burst stimulation. B: power spectra of the trace in A demonstrating reduction of theta power during DSI and subsequent recovery. C: group data showing that theta power is almost abolished during DSI but significantly recovers to baseline levels afterward (P < 0.0001). D: CB1 antagonist AM251 (3 μM), which prevents DSI, preserved theta IPSP activity even after 4 more intense trains of theta-burst stimulation.

Repetitive theta bursts constitute a relatively vigorous mode of stimulation and thus we also used simple trains of action potentials induced by DC injection to induce DSI of theta sIPSPs. We varied the number of 10-ms DC pulses given at 50 Hz from 1 to 50 (1, 2, 3, 5, 10, 15, 20, 30, 50) in eight cells (every cell received the entire sequence) that showed large ongoing theta IPSPs in the presence of CCh. For these experiments, we integrated the voltage traces over 6-s intervals before and after the action potential train to assess the degree of IPSP activity, and thereby determine the magnitude of DSI. Figure 4 shows that DSI was detectable after two or three action potentials, as previously reported (Pitler and Alger 1992a), and appeared to be maximal after 10–15 action potentials. Maximal DSI represented a 77% reduction of sIPSP activity.

FIG. 4.

Action potential dependency of DSI on theta IPSPs. A: representative trace of sIPSPs and their suppression by brief trains of action potentials (trains delivered at large upward deflections; numbers of action potentials in each train indicated in parentheses). B: complete data showing the degree of sIPSP suppression for each cell (n = 8). Voltage traces were integrated over 6-s intervals before (control) and after the action potential train, and the data are expressed as percentage control as a function of the number of action potentials in each train. C: data shown in B expressed as mean ± SE. Data were well fit (r = 0.979) by the function y = 94.8e−0.12/x[r] + 23.1. Curve suggests that at its maximum DSI produces a 77% reduction in sIPSP activity.

Large sIPSPs with fast kinetics are produced by basket and bistratified cell synapses on perisomatic regions of pyramidal cells (Banks et al. 1998; Buhl et al. 1995; Freund and Buzsaki 1996; Soltesz et al. 1995). To determine whether perisomatic-targeting interneurons contribute to the production of theta IPSPs, we first measured the IPSP properties and found that they tended to be large and have relatively rapid rise and decay times, as shown in the cumulative-frequency plots of Fig. 5B. Although these properties are consistent with the theta being generated by perisomatic inputs, a more stringent test is provided by surgical cuts in stratum radiatum, which isolate the pyramidal cell somata and proximal portion of the apical dendrites. In three slices with cuts within 150 μm of stratum pyramidale (e.g., Fig. 5A), we applied CCh and recorded rhythmic IPSPs. As shown in the cumulative-frequency plots of Fig. 5B, the IPSPs from these slices were somewhat smaller, but had rise and decay times that were not significantly different from IPSPs in the control slices. There was a slight but significant difference in the amplitude distributions from cut and control slices [4.24 ± 0.21 vs. 6.90 ± 0.24 mV, P < 0.005, Kolmogorov-Smirnov (K-S) test]. The Relative theta power in these cells (n = 3) did not differ from that in five randomly selected control cells (57.64 ± 3.26 vs. 50.04 ± 3.26, P > 0.30). The average peak power was 5.39 ± 1.63 Hz (n = 3), although the cell illustrated in Fig. 5 showed a slowing in the peak power (2.54 Hz). This could reflect a modulatory influence of the distal inputs on the activity of the theta interneurons (cf. Mann et al. 2005) or perhaps a depressant effect of tissue damage resulting from the cuts. Nevertheless, the data clearly indicate that IPSPs from distal apical regions are not required for the generation of rhythmic IPSPs in CA1 and are consistent with a perisomatic origin of the IPSPs.

FIG. 5.

Theta IPSPs can be recorded in CA1 pyramidal cell somata that are surgically isolated from their apical dendrites. A: a hippocampal slice in which a cut (line indicated by arrows) had been placed in stratum radiatum about 50 μm from stratum pyramidale. A pyramidal cell was recorded at the position indicated by the schematic electrode in the presence of 5 μM CCh and the rhythmic sIPSPs shown in the bottom trace were detected. Note the occurrence of DSI caused by theta-burst action trains. Graphs to the right show the power spectra obtained from this cell before (baseline), during, and after DSI (recovery). B: cumulative-frequency plots of data obtained from 5 cells from control slices and 3 cells from cut slices, showing IPSP amplitude and both rise and decay time data. Note that the IPSPs in the cut slices have somewhat smaller amplitudes, but not significantly different rise or decay times than those of IPSPs in control slices. During DSI the amplitude distribution shifts to smaller IPSPs, with slightly slower rise times for both groups. Note: ≥100 events were used for these analyses during 10-s periods of activity.

In the cut slices, DSI continued to decrease the larger IPSPs without a substantial change in their rise time, although decay times of IPSPs during the DSI period were somewhat faster in the cut slices. During the DSI period the amplitude distributions were significantly shifted to the left for both cut (3.61 ± 0.18 mV) and control (5.02 ± 0.24 mV) cells (P < 0.005, K-S test), and the groups significantly differed (P < 0.005). No differences were found in the 10–90% rise times of theta IPSPs of cut and control cells before (4.42 ± 0.11 vs. 4.91 ± 0.11 ms, P > 0.03) or after DSI (5.26 ± 0.18 vs. 5.13 ± 0.14 ms).

It is well established that population theta rhythms in the CA1 region are dependent on the activation of glutamatergic excitatory postsynaptic potentials (EPSPs) projecting from the CA3 region, or on the activation of mGluRs within CA1 by bath-applied mGluR agonists (Gillies et al. 2002). Nevertheless, interneurons—particularly those targeting perisomatic pyramidal cell regions—synapse on hundreds, even thousands, of pyramidal cells because of their profuse axonal projections (Freund and Buzsaki 1996). It seemed likely therefore that if such interneurons are involved at least small groups of pyramidal cells should receive synchronous IPSPs. We tested this idea by making simultaneous recordings from visualized pyramidal cells in CA1. The whole cell pipettes contained a dye that permitted confirmation of the cell type by morphology and location of the soma. We recorded from eight such pairs, the somata of which were within 100 μm of each other. An example of dual pyramidal cell recording is shown in Fig. 6. Images of the cells are shown at the top under fluorescence (left) or DIC illumination (right); each cell is visualized in its optimal focal plane in the top or bottom row. The traces in Fig. 6A are samples of synchronous, muscarinic-induced (8 μM CCh) sIPSPs in the two cells. The autocorrelation functions (Fig. 6B) in each cell show IPSP oscillations occurring at about 3 Hz, and the cross-correlogram (Fig. 6C) reveals a high degree of synchronous activity, with subsidiary peaks at about 300-ms intervals. Figure 6D shows that, as anticipated, the spectra of the autocorrelation functions and cross-correlogram have very similar shapes and peak at about 3 Hz. The dual pyramidal cell recordings were made at about 22°C, compared with the 30°C for the sharp electrode experiments, and so the somewhat lower-peak sIPSP frequency is not unexpected. We have observed results similar to those in Fig. 6D in four other pyramidal cell pairs. Nevertheless, it is worth indicating that, even though robust sIPSPs were recorded in all cells, simultaneous IPSP activity was not always present. An example of data from a pyramidal cell pair in which sIPSPs were not correlated is shown in Fig.6, E and F.

FIG. 6.

Synchronous sIPSP activity recorded simultaneously in 2 CA1 pyramidal cells. Images of both cells shown in top rows. Both cells were filled with Alexa dye (left) and were impaled at the indicated depths below the slice surface, although each can be seen when the other is in focus (arrowheads); also shown in differential interference contrast (DIC, right). All data in the figure were taken from these cells. A: sample simultaneous raw traces of depolarizing sIPSPs in current-clamp mode [8 μM CCh, 10 μM NBQX, and 50 μM APV present]. B: autocorrelation functions (ACFs) for c1 and c2. C: cross-correlation function (CCF) of the same activity across c1 and c2. D: spectral power for ACFs and CCF. Rhythmic activity peaked at 3 Hz in all analyses. E: example of a cross-correlogram of uncorrelated sIPSP activity in a pair of pyramidal cells; the confidence intervals are ±0.01. F: cross-correlation coefficients calculated in 10-s bins for 300 s for the cells in E, showing that there were no epochs of significantly correlated activity.

Notably, in pyramidal cell pairs with synchronous sIPSPs, we observed that an interictal-spike–like event lasting ≳1.5 s generally produced DSI of synchronous sIPSPs (e.g., Fig. 7), in agreement with LeBeau and Alger (1998). Of 32 bursts recorded in four pyramidal cell pairs showing good synchronous sIPSP activity, 27 bursts occurred in only one cell of the pair at a time. Each burst caused marked DSI in the bursting cell (to 16.53 ± 8.74% of control, n = 7 bursts), but no suppression of sIPSPs in the paired cell (117.14 ± 15.10%; compare top and bottom records in Fig. 7). This agrees with a report by Wilson and Nicoll (2001), who showed that DSI in pyramidal cell pairs did not “spread” beyond a very narrow radius from its initiation point (the second cell had to be impaled within 20 μm of the first cell). Because the cells in our pairs were considerably farther apart than this, DSI in the secondary cell would not have been expected. Nevertheless, this issue will require further detailed investigation. In the remaining five of 32 cases, bursts in both cells occurred nearly synchronously (within about 30 ms of each other), suggesting that large synchronous, depolarizing sIPSPs may have helped trigger both bursts. In any event, the clear susceptibility of synchronous sIPSPs to DSI is consistent with their origin from perisomatic interneurons expressing CB1 receptors. Unfortunately, because of the difficulty in maintaining the paired recordings, we were unable to do these experiments in the presence of AM251.

FIG. 7.

DSI induced by intertical spikes may not affect sIPSPs in neighboring pyramidal cells. Sample traces from 2 simultaneously recorded CA1 pyramidal cells in the presence of 8 μM CCh. Left: pair of traces showing that a large, spontaneously generated interictal spike in one cell induced DSI of the sIPSPs in that cell for 5–10 s, and no effect on the sIPSPs in the paired cell. Both cells were capable of initiating DSI. Similar effects were seen in a total of 32 bursts from 4 cell pairs. Top right: pair of traces illustrating that, on occasion, bursts were initiated nearly simultaneously in both cells, and pronounced DSI was induced. Small dotted vertical lines in the bottom expanded traces highlight the synchrony of sIPSPs occurring in both cells. Large sIPSP marked by the vertical dashed line on the right may have triggered the simultaneous bursts in the top right graph. Cell 1 was held at −85 mV with negative DC to reduce spontaneous action potential generation; the trace for cell 1 was offset for display purposes. Cell 2 was near its resting potential of −70 mV.


Atropine-sensitive, hippocampal CA1 field theta rhythm in vitro arises largely as a result of synaptic activity in the CA3 region. It can be prevented by antagonists of AMPA and kainate receptors, or by cutting the axonal connections between CA3 and CA1 (Buzsaki 2002; Fisahn et al. 1998; Fischer 2003; Williams and Kauer 1997). Excitatory input from CA3 entrains CA1 theta by activating both pyramidal cells and interneurons. In vivo, septal cholinergic input can induce theta rhythms in pyramidal cells, but this effect depends on rhythmic GABAergic inhibition (Buzsaki 2002). Such a circuit cannot account for our observations because GABAergic input from the septum does not exist in the slices. Furthermore, our data show that the CCh-induced IPSP theta existing within CA1 (Gillies et al. 2002) is independent of an mGluR-activated circuit. Based on data from both pharmacologically and surgically reduced slices, we believe the theta IPSP activity arises from interneurons within CA1. Because of their highly divergent output and prominent somatic IPSPs, single basket cells can synchronize and entrain the activity of pyramidal cells (Cobb et al. 1995). Such cells may form a substrate for CA3 glutamatergic, septal cholinergic, and GABAergic projections, thereby contributing to CA1 theta rhythms without being capable on their own of generating population theta when isolated from these inputs.

We have shown that either action potentials delivered in a burst pattern or as brief trains can interrupt the theta sIPSPs by an endocannabinoid-dependent mechanism. This is the first evidence that endocannabinoids can modulate an intrinsic, behaviorally relevant neuronal rhythm. Exogenous cannabinoids can disrupt kainic acid–induced gamma rhythms (Hajos et al. 2000), implying that the same cells may be active in the generation of both of these important forms of neuronal oscillation. It is interesting that we found that the magnitude of endocannabinoid-induced suppression of CA1 sIPSPs saturated with about 10 action potentials. This is essentially the same number required for maximal endocannabinoid suppression of presynaptic calcium increases in cerebellar parallel fibers (Brown et al. 2003) or of sIPSCs in neocortical cells (Fortin et al. 2004). This consistency across different cell types and physiological readouts suggests a common underlying biochemical mechanism is responsible.

In addition to DSI, we demonstrated that group I mGluR activation abolishes theta IPSP activity. This result is consistent with previous studies that mGluR stimulation releases endocannabinoids by a G-protein, Ca2+-independent pathway (Maejima et al. 2001; Varma et al. 2001). Interestingly, tonic activation of mAChRs by CCh also causes a persistent endocannabinoid release, although release by mGluRs appears much more vigorous and may involve a different mechanism (Kim and Alger 2002). This may explain in part why mAChR-induced theta IPSPs persist despite a concomitant tonic release of endocannabinoids. Tonic mAChR endocannabinoid release may even actually fine-tune theta IPSP activity by providing a dynamic low-pass filtering of γ-aminobutyric acid (GABA) release from interneuron terminals. Full resolution of these issues, however, is beyond the scope of the present study and will be the focus of future work.

The exquisite sensitivity of the theta IPSPs to endocannabinoids also provides indirect evidence regarding the identity of the theta-generating IPSPs. In CA1, perisomatic synapses are made mainly by mutually exclusive interneuron classes that contain either parvalbumin (PV) or CCK (Freund and Buzsaki 1996; Pawelzik et al. 2002). The great majority of PV cells (>95%) do not express cannabinoid CB1 receptors, whereas the great majority of CB1 receptors (95%) are found on the CCK-cell axon terminals (Freund et al. 2003). The parsimonious explanation for the susceptibility of theta sIPSPs to DSI is that they originate from the CCK interneurons. Paired recordings from theta IPSP–generating cells and pyramidal cells, followed by immunohistochemical staining for CCK, will be required for a detailed test of this hypothesis.

Our analyses of sIPSP amplitudes and kinetics provide additional evidence concerning the identity of the interneurons that generate mAChR-induced theta. They do not seem to be the LM cells that show cholinergic induction of theta-frequency oscillations (Chapman and LaCaille 1999). The LM cells target apical dendrites and produce slow IPSPs in pyramidal cells; the sIPSPs in the present investigation were large and have the rapid kinetics that make them likely to be perisomatic in origin (Banks et al. 1998; Buhl et al. 1995; Martin et al. 2001; Soltesz and Mody 1995). We also found that theta IPSPs persist in pyramidal cells from which the apical dendrites have been largely removed. The LM cells inhibit pyramidal cells when the interneurons are given minimal extracellular electrical stimulation; it has not been shown that they produce persistent ongoing sIPSPs in the presence of CCh. Finally, IPSPs with slow kinetics are not inhibited by exogenous or endogenous cannabinoids (Wilson et al. 2001) or DSI (Martin et al. 2001), whereas the theta IPSPs are. As noted above the CCK cells could be the generators of theta IPSPs because of the highly significant morphological association between CB1 receptors and these interneurons (Freund et al. 2003). Furthermore, the lesion experiments would also be consistent with the identification of the CCK basket cells as major contributors to the rhythm, although a contribution of dendrite-targeting CCK cells (Cope et al. 2002; Pawelzik et al. 2002) is not ruled out because these may express CB1 as well. Indeed, we have previously reported that DSI can be recorded in distal apical dendrites, >300 μm from stratum pyramidale, under conditions where somatic IPSPs were blocked by local application of tetrodotoxin (Morishita and Alger 2001). Dendritic IPSPs regulated by endocannabinoids would probably be produced by the dendrite-targeting CCK neurons. In any event, our interpretation that perisomatic targeting interneurons are particular targets of endocannabinoid inhibition is in keeping with the evidence put forward by Trettel et al. (2004) for this conclusion in the neocortex.

Synchronous sIPSPs in neighboring pyramidal cells could be produced by single interneurons that innervate nearby pyramidal cells. This appears to be the case in the cerebellum where the sIPSPs are also susceptible to DSI (Vincent and Marty 1993). On the other hand, neocortical networks of PV-containing, CB1R-lacking interneurons generate theta rhythms (Blatow et al. 2003). An important hypothesis states that the PV system is specialized to provide a clocklike timing function, whereas another network of CCK- and CB1R-containing interneurons could also generate rhythmic oscillations that will be readily modifiable by neuronal modulators and thus capable of subserving distinct behavioral roles (Freund 2003). Strong evidence for an electrically coupled population of CCK-containing interneurons has been reported in the neocortex (Galareta et al. 2004) and suggested in the hippocampus (Meyer et al. 2002). However, Buhl et al. (2003) recently reported that hippocampal theta rhythms were intact in connexin 36 knockout mice. Theta IPSPs can be suddenly switched on by mAChR agonists and switched off by neuronally released endocannabinoids, behavior that could reflect either activity within a network of interneurons or all-or-none activity produced by single, potent interneurons. Pyramidal cell pairs reveal that synchronous theta IPSPs do occur in neighboring cells, but the data also suggest that this synchronous activity is not universal. It is possible that information could be represented by small “modules” of entrained pyramidal cells. Our pair recordings also demonstrated that a spontaneous burst of action potentials in one cell can induce DSI of inputs to itself, but not to the other cell, whereas simultaneous bursts in both cells produced DSI in both cells. This suggests that DSI may allow a single cell and possibly its nearby neighbors to transiently escape entrainment. An obvious caveat is that our dual pyramidal cell recordings were obtained in slices; thus the extent of the simultaneous inputs cannot be determined with certainty because of the truncation of extensive interneuronal branching.

Disruption of theta rhythms might be one mechanism by which cannabinoid drugs cause cognitive dysfunction (Ameri 1999). Decreased performance on human word-recognition tasks caused by marijuana can be accompanied by decreased EEG theta rhythms (Ilan et al. 2004). Exogenous cannabinoids could interrupt theta in part by inhibiting the theta-rhythm sIPSPs, although suppression of glutamatergic transmission by activation of a putative “CB3” receptor on the glutamatergic nerve terminals (Hajos et al. 2001) could also be involved. On the other hand, beneficial behavioral functions could be served when endocannabinoids interrupt theta activity in individual pyramidal cells. We have shown directly that single pyramidal cells can regulate sIPSPs that occur synchronously in nearby pyramidal cells. When single cells are disinhibited by DSI, they are more susceptible to the induction of long-term potentiation than are cells in the general population (Carlson et al. 2002). Population theta activity could provide a background against which information coded by single cells would be differentiated and thereby enhanced.


This work was supported by National Institutes of Health Grants NS-30219 and DA-14625 to B. E. Alger. C. G. Reich and M. A. Karson were supported in part by the Cellular and Integrative Neuroscience Training Grant to the University of Maryland (T32 NS-07375). L. M. Jones was a student in the University of Maryland Program in Neuroscience.


We thank S. Thompson and A. Keller for comments on a draft of this manuscript.


  • 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