| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Brain Research Institute, University of Zurich, CH-8057 Zurich, Switzerland
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
(Received 22 September 2003;
accepted after revision 22 October 2003;
first published online 31 October 2003)
Corresponding author Y. Fischer: Brain Research Institute, University of Zurich, CH-8057 Zurich, Switzerland. Email: kfischer{at}hifo.unizh.ch
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
How oscillations are generated, maintained, synchronized and paced are central questions. Although external sources are suggested to contribute to all these processes (Buzsáki, 2002), intrinsic mechanisms may underlie them. In line with this, it has been shown that the oscillatory mode of activity can be faithfully reproduced in a hippocampal network that lacks extrinsic influences (Fischer et al. 1999, 2002). The network preserved in the in vitro system of hippocampal slice culture (Gähwiler, 1981, 1998) always adopts the oscillatory mode of activity in response to cholinergic activation, and exhibits all the above-mentioned features.
We have previously reported that the septal cholinergic input to the hippocampus is sufficient for the activation of the intrinsic network oscillator (localized in area CA3), and for the expression of concurrent theta and gamma oscillations (Fischer et al. 1999, 2002). Moreover, oscillations develop as a result of network interactions, while the cholinergic input has only a permissive role. Oscillations depend on the diffuse action of acetylcholine (Fischer et al. 1999), which is supported by the findings that the response of pyramidal cells (PCs) to cholinergic afferent activation is slow, i.e. a muscarinic EPSP of 1–3 min in duration appears within 1 s of septal afferent stimulation (Gähwiler & Brown, 1985), and that septal cholinergic afferents form free ending terminals, in addition to classical synapses, in the hippocampus (Frotscher & Léránth, 1985). Furthermore, cell classes that spontaneously discharge and the numbers of action potentials expressed by these cells are comparable to in vivo data (Fischer et al. 2002).
Generation of action potentials by the network is critical for the oscillatory activity, since it is abolished by TTX (Fischer et al. 2002). However, the spontaneous discharge of interneurones, the major cell class to spontaneously discharge during oscillations, is incompatible with the triggering or pacing of the oscillations (Fischer & Dürr, 2003). Therefore, identification of additional sources of action potentials becomes imperative. Natural candidates, to take over these functions are the excitatory cells. Therefore, I set out to examine the activity of individual CA1, CA2, and CA3 PCs, and dentate granule cells (DGCs), during the cholinergically mediated (methacholine; MCh; 10–20 nM) oscillatory mode of activity. Here I determine which excitatory cells discharge, and the consequences of spontaneous action potential discharge to the initiation, maintenance, synchronization, pacing and appearance of hippocampal network oscillations. I also determine whether discharge was a direct result of the cholinergic input, or was due to network interactions. Moreover, I set out to identify additional reactions that may facilitate the expression of oscillatory activity by the hippocampal network.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
All experiments were done according to the guidelines set forward by the Department for Veterinary Affairs of the Kanton of Zurich. Hippocampal slice cultures were prepared from 6-day-old rat pups killed by decapitation, and maintained for 3–6 weeks in vitro (Gähwiler et al. 1998).
Lesions
A lesion between fascia dentata and CA3 stratum pyramidale was made 24 or 48 h prior to recording, in a cutting medium that consisted of a normal tissue culture medium that was supplemented with 0.5 µM TTX and 10 mM MgCl2.
Electrophysiological recordings
For electrophysiological recordings, cultures were transferred to a recording chamber, mounted on an upright or inverted microscope (Zeiss, Jena, Germany), and continuously superfused with warmed (32–33°C) saline containing (mM): Na+ 145, Cl- 149, K+ 2.7, Ca2+ 2.8, Mg2+ 2, HCO3- 7.7, H2PO4- 0.4, glucose 5.6, and Phenol Red (10 mg l-1) at pH 7.4.
Simultaneous electrophysiological recordings were obtained either from individual PCs of areas CA1 and CA3, or from individual PCs of areas CA2 and CA3, or from individual CA3 PCs and DGCs using sharp microelectrodes containing 1 M potassium methylsulphate either alone or with 2% biocytin (electrode resistance: 30–60 M
), as previously described (Fischer et al. 1999; Fischer et al. 2002). Only a fraction of these cell pairs were monosynaptically connected, tested by evoking an action potential (40 ms current pulse) in either of the cells, and by monitoring the postsynaptic response in the other. Recordings obtained from CA3 PCs were also used as a coherent time reference for the oscillatory activity.
Drugs
Methacholine (MCh), carbenoxolone, halothane, ammonium chloride (NH4Cl), atropine and bicuculline were purchased from Sigma-Aldrich (Buchs, Switzerland), 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzo[f]quinoxaline-7-sulfonamide (NBQX), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and tetrodotoxin (TTX) from Tocris-Cookson (Bristol, UK), and biocytin, Lucifer yellow and micro ruby from Molecular Probes (Eugene, OR, USA). (±)-3-(2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) was a gift from Novartis (Basel, Switzerland).
Data analysis
The analog signal was recorded on videotape at 22 kHz and digitized at 1 kHz. Analysis of oscillatory activity was based on windowed fast Fourier transform (FFT) analysis, using 500 ms windows with 50% overlap, as previously described (Fischer et al. 2002). Results of the windowed FFT analysis are presented in the form of spectrograms, which illustrate the Fourier analysis power spectrum (frequencies: 2–500 Hz) as a function of time, using a false colour scale to indicate the power in decibels (dB) (warm colours represent high power and thus the dominant frequencies). Rhythmic activity in PCs was evaluated by calculating the mean of the dominant frequencies obtained from the FFT of each window (during a 10 s period of activity). For assessing gamma activity in the cells, our previously described subtraction method was used, where the slow activity content of the PCs and DGCs signals was calculated and subtracted and windowed FFT was recalculated (Fischer et al. 2002).
Interspike intervals were measured from signals digitized at 1 kHz. Action potential and unitary connection parameters were extracted from signals digitized at 10 kHz.
Phase analysis of CA2 PC firing was performed using the procedure described in Fischer & Dürr (2003). In brief, for each action potential in the CA2 PC (identified by a threshold method), the corresponding oscillatory cycle in the CA3 PC, its duration, and the phase relationship between the two were determined. The beginning of the corresponding oscillatory cycle (in the CA3 PC) was assigned to the onset of the population IPSP, which is zero-phase by definition. The ratio between the spike delay, i.e. the time difference between the action potential and the cycle onset, and the local cycle length (by definition 360 deg) gives the immediate phase of each action potential, which is thereafter represented in firing-phase histograms. Also here analysis was preformed on signals digitized at 1 kHz.
Calculation of the slow depolarization in CA3 and CA2 PCs (Figs 3, 5 and 6) was based on a running average of 4 s without phase shift.
|
|
|
Dye transfer
Dye transfer properties of individual cells of the hippocampal network were measured under three experimental conditions: (a) control, (b) MCh (10–20 nM) and (c) MCh with NH4Cl (2 mM). In each condition, a single cell (per culture) was injected with an intracellular solution containing 1 M potassium methylsulphate, 2% biocytin and 4% micro ruby. Upon penetration, the first 5 min were used to check the electrophysiological condition of the cell and to verify that only a single cell was impaled, on the basis of the fluorescent internal control micro ruby (molecular weight 3000). This was followed by 5 min exposure to control saline, or to MCh, or to MCh and NH4Cl, and an additional 5 min of wash. Following this 15 min (in all conditions) the electrode was withdrawn, and the culture was inspected again to verify (again the fluorescent internal control micro ruby) that only the original cell was injected. For condition C, the ability to transfer Lucifer yellow (4% aqueous) was also tested. At the end of experiments, cultures were fixed with a solution containing 4% paraformaldehyde and 15% (v/v) saturated picric acid in 0.1 M phosphate buffer (PB). After fixation, biocytin injected cultures were transferred to a cryoprotective solution (30% sucrose and 12% glycerol in 0.1 M PB), frozen in dimethylbutane, and stored at –20°C until processing. For the biocytin reaction, cultures were transferred to PB (0.1 M), incubated in avidin-biotinylated horseradish peroxidase complex (ABC, 1 : 200, Vector), and transferred to Tris-buffered saline (TBS, pH 7.4). Tissue-bound peroxidase was visualized by 3,3'-diaminobenzidine 4 HCl (Sigma) as a chromogen (oxidized with H2O2). Following dehydration, cultures were embedded in dammar resin (Fluka, Switzerland). Lucifer yellow injected cultures were embedded in dammar resin, following fixative wash (PB) and dehydration.
Visualization of biocytin injected CA2 PCs and DGCs
Biocytin injected PCs (CA2 and CA3) and DGCs were fixed, cryoprotected, and frozen as described above. To facilitate the visualization of the axonal projections, an additional amplification step was used. Cultures were washed in PB (0.1 M, 3 x 10 min), and non-specific sites were blocked with heat inactivated horse serum (2% in PB, 1 h). Specific sites were then detected by Alexa 488-conjugated streptavidin (1: 250, Molecular Probes, Eugene, OR, USA) (overnight 4°C, in blocking solution). Following washes and blocking (as above), cultures were incubated with a biotinylated antiavidin (1: 1000, Vector) (overnight 4°C, in blocking solution). This was followed by incubation in avidin-biotinylated horseradish peroxidase complex (ABC, 1: 200, Vector), visualized as above, and embedded in slowfade (Molecular Probes, Eugene, OR, USA).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Granule cells of the dentate gyrus give rise to a major excitatory pathway in the hippocampus, the mossy fibre pathway, which innervates area CA3. The involvement of DGCs and mossy fibres in hippocampal oscillations has not been characterized in hippocampal slice cultures thus far. Simultaneous intracellular recording were obtained from DGCs and CA3 PCs (n= 7), whose morphological identities were confirmed following biocytin injection (Fig. 1A). At their resting potentials, CA3 PCs (–59.7 ± 4.1 mV, n= 7) and DGCs (–63.4 ± 2.1 mV, n= 7) exhibited synaptic theta oscillations in response to MCh application (20 nM) (Fig. 1B). In CA3 PCs rhythmic activity was characterized by cycles consisting of population EPSP/IPSP sequences that are usually dominated by population IPSPs (see also Fischer et al. 1999, 2002), while in DGCs EPSPs, of duration and amplitude similar to the unitary connection from CA3 PCs (not shown), prevailed. Dominant frequencies were 7.7 ± 1.5 Hz in DGCs (n= 6), and 7.6 ± 1 Hz in CA3 PCs (n= 6) (P > 0.05, t test). Windowed FFT analysis also showed that activity in the gamma band was 18.5 ± 3.1 dB lower (n= 6) (about 100 times smaller) in DGCs. Cross-correlation analysis revealed that such patterns of activity occurred synchronously at individual cells of both regions (Fig. 1B). Moreover, none of the DGCs discharged during oscillations. These results indicate that the dentate gyrus plays a passive role in the generation of the oscillations. Indeed, making a lesion between fascia dentata and CA3 pyramidale (Fig. 1C) did not impair theta oscillations in CA3 PCs (n= 5), demonstrating that the dentate gyrus is not essential for intrinsically generated oscillations.
|
One identified source for action potentials among excitatory cells, during oscillations, are the CA3 PCs (Fischer et al. 1999, 2002). In general, discharge of CA3 PCs was associated with a slow transient depolarization of their membrane potential (Fig. 2). Firing itself was limited, and individual cells exhibited few action potentials during oscillatory activity (Fig. 2). Moreover, this firing preferentially occurred during the initial phase of the slow depolarization, coinciding with the onset of oscillations (Fig. 2), and subsided sharply afterwards. Overall 27%(n= 57/206) of CA3 PCs exhibited some form of action potential discharge during oscillations; in 84%(n= 48/57) firing was, and in 16%(n= 9/57) was not, associated with the transient membrane depolarization.
|
Whether the slow depolarization results from the direct activation of muscarinic receptors expressed by CA3 PCs or from network interactions was tested next. The slow depolarization, that accompanied the oscillatory activity (Fig. 3A) (n= 93), was abolished by atropine (0.1 µM) (Fig. 3B) (n= 7/7), and was insensitive to TTX (0.5 µM) (Fig. 3C) (n= 6/6). Furthermore, atropine (0.1 µM, n= 22) and TTX (0.5 µM, n= 14) abolished the network oscillations. Thus, the slow transient depolarization observed in CA3 PCs, resulted from the direct activation of muscarinic receptors by MCh.
The role of CA2 PCs in oscillations
In the hippocampus there exists an additional class of PCs that forms area CA2. This area can be identified following TIMM staining (Fig. 4C), whose reaction product defines a narrow clear region between the end of area CA3 (defined by the end of the mossy fibers projection) and the beginning of area CA1 (see also Baimbridge & Miller, 1982). I was interested in characterizing the activity of
|
In practical terms, it is relatively easy to localize CA2 PCs for recordings under visual inspection. These cells are situated after the end of CA3 pyramidale, and are identifiable by somata that are distinctively smaller than that of CA3 PCs and bigger than that of CA1 PCs. Morphological features of these PCs were revealed by the injected marker biocytin. CA2 PCs are characterized by an elongated dendritic arbor (Fig. 4A). The axonal arbor followed the patterns previously described (Lorente de Nó, 1934; Tamamaki et al. 1988), and a schematic representation is provided in Fig. 11. The axon originated in the basal dendrites and projected, in all labelled cells, a main branch toward area CA1 at stratum (str.) oriens and the alveus. Commonly, the main axon would give rise to an opposite branch directed to area CA3a, at the same strata. In some cases the main axon branched in str. oriens to send a collateral that crossed str. pyramidale in a right angle and then branched at 90 deg toward CA1 at the outer edge of str. radiatum, toward and within str. lacunosum-moleculare. Occasionally this branching would also project toward CA3a in a similar manner. However, the most extensive axonal plexus was usually observed in the vicinity of the injected cell, in areas CA2 and CA3a (see schema in Fig. 11), in str. oriens and pyramidale. Although projecting toward area CA1, CA2 PCs do not give rise to Schaffer collaterals (Fig. 1); their axon collateral that projects in the outer edge of str. radiatum is situated closer to the location of the perforant path input, over the outer apical dendrites of CA1 PCs (see also Lorente de Nó, 1934; Tamamaki et al. 1988).
|
|
In eight out of these nine CA2 PCs, action potential firing accompanied the slow depolarization. In five cells, discharge was extensive and lasted for minutes; median firing rates were between 0.56 and 3.7 Hz (Fig. 5B). Analysis of the spontaneous discharge in those five CA2 PCs showed that the sequence in which action potentials occurred was structured to exhibit acceleration in the interspike intervals (Fig. 5C), corresponding to the temporal modulation of the slow depolarization (Fig. 5B). Instantaneous frequencies of action potentials varied between 0.01 and 100 Hz (Fig. 5C). Moreover, instantaneous firing frequencies exhibited, in the log domain, a unimodal normal distribution whose peak appeared in the theta band (Fig. 5C).
The temporal relationship between the spontaneous discharge of CA2 PCs and the immediately relevant theta cycles in the corecorded CA3 PCs was determined (n= 5). In contrast to spontaneous firing in CA3 interneurones (Fischer & Dürr, 2003), discharge in all five CA2 PCs was not temporally structured to occur at specific and limited portions of the oscillatory cycle. An apparent modulation in the firing phase histogram (Fig. 5C) was not a consistent feature, and in the five experiments somewhat elevated discharge at 30, 80, –80, 10 and 50 deg accompanied consistent contributions at all phases of the cycle. Therefore, firing occurred almost equally at all portions of the oscillatory cycle, thus providing a temporally uniform source of action potentials (Fig. 5C).
Additional reactions
Thus, the patterns of action potential firing, in the excitatory cells, do not lend themselves to a ‘simple’ explanation of the oscillatory activity, as their numbers are low (see Discussion) and as they are temporally uncorrelated with the oscillatory cycles. Therefore additional interactions must exist in order to translate the kernel of action potentials, i.e. the pool of all PC-generated, action potentials available at any given time, into the oscillatory patterns. Moreover, such interactions should occur in a manner that would provide the ‘clock’ function, i.e. the property of the network to synchronize and pace the rhythmic synaptic activities, such that theta and gamma oscillations of the patterns and properties found would emerge (given the action potential kernel constraints). I therefore decided to test whether it is possible to identify interactions between the discharging cells which may facilitate the expression of oscillatory activity, in particular the possibility that axo-axonal communication, mediated by gap junctions, may provide this function.
Additional reactions: pharmacology. The first approach used was pharmacological. Several compounds, such as halothane, octanol and carbenoxolone, are reported to block gap junction communication (Johnston et al. 1980; Leslie et al. 1998), and to diminish oscillatory activity (Draguhn et al. 1998; Traub et al. 2000). The effect of halothane and carbenoxolone on the cholinergically mediated oscillations was evaluated in simultaneously recorded CA3 and CA1 PCs. Application of halothane (5 mM, n= 9) strongly hyperpolarized the membrane potential of the cells by about 20 mV. The hyperpolarization developed gradually, over a couple of minutes, and prevented the expression of oscillatory activity. Likewise carbenoxolone application (100 µM, n= 2; 200 µM, n= 6; 20 µM, n= 4) resulted in slowly developing hyperpolarization. Due to these ‘non-specific’ effects, such compounds cannot be used for the investigation of the involvement of gap junction communication in the oscillatory response.
Another chemical manipulation designed to target gap junction communication during oscillations was to use the alkalizing agent ammonium chloride (NH4Cl). Gap junction opening is suggested to be enhanced during the alkalization of the cytosol by NH4Cl, and upon removal of NH4Cl to be consequently suppressed, due to acidification of the cytosol and the closure of gap junctions (Spray et al. 1981; Thomas, 1984; Draguhn et al. 1998). At the commonly used concentration of 10 mM (n= 5), and at 5 mM (n= 2), NH4Cl was found to have deleterious effects on CA1 and CA3 PCs and caused instability to the cells and their ultimate death. Therefore the concentration of NH4Cl was reduced to 2 mM, a concentration in which no deleterious effects on PCs of areas CA1 and CA3 were observed. In six simultaneously recorded CA1 and CA3 PCs (Fig. 7) the effects of NH4Cl (2 mM) application were examined after confirming the control oscillatory theta response (MCh, 20 nM) (Fig. 7B) (Fischer et al. 1999). Application of NH4Cl, in the control saline, had no apparent effect on the basal activity of the recorded cells (for a period of 7–10 min), nor did it elicit oscillatory activity (n= 6) (Fig. 7C). Adding MCh in the presence of NH4Cl again elicited oscillatory activity, but the weight of IPSPs in the rhythmic sequence of CA3 PCs decreased and the weight of EPSPs increased (Fig. 7D). At the same time, the amplitude of the population EPSP theta sequence in the corecorded CA1 PCs increased, in a time-variable manner, 2- to 4-fold (Fig. 7D). Both effects, on the oscillatory activity of CA3 and CA1 PCs, were immediately apparent upon expression of oscillations by the network. At this point NH4Cl was removed, still in the presence of MCh. This resulted in the rapid (within 2 min) blockade of oscillatory activity in both cells (Fig. 7E). Such blockade of the cholinergically mediated oscillatory activity was never observed under control conditions. Reapplication of MCh, after wash, elicited the control theta oscillations in both cells (Fig. 7F), which subsided following 5–10 min wash (Fig. 7G). Such effects are in accordance with the time course of alkalization, acidification, and recovery by NH4Cl, where cells maintain an alkalized state for the duration of NH4Cl application used here, acidify almost immediately (about 1 min) upon removal of NH4Cl and recover (due to homeostasis) in about 15 min (Thomas, 1984). Moreover, none of the CA1 PCs used in this experiment, or others, discharged during oscillations (author's unpublished results, n= 108). These results indicate that the opening and closure of gap junctions has immediate and profound effects on the oscillatory activity.
Additional reactions: dye transfer. The second approach used was to test whether the gap junction-mediated interactions between cells could be visualized by monitoring the dye transfer properties of a single CA3 cell (per culture) under various conditions. Experiments were preformed under three conditions: control (n= 26), MCh (20 nM) (n= 34), and coapplication of MCh (20 nM) and NH4Cl (2 mM) (n= 18). Estimation of the dye transfer rates was based on the visualization of the marker biocytin, coinjected with micro ruby acting as an internal control (a combination permissive for electrophysiological measurements). Results are summarized in histograms (Fig. 8) that depict the number of additional cells in which biocytin reaction product was observed. At control, in 13 of the 26 cultures there was no dye transfer, while in 12 cultures transfer to one to three additional cells was detected, and in one case to eight additional cells, giving an average rate of 1.11 ± 1.7 cells. During MCh application only 5 of the 34 cultures did not exhibit dye transfer, and in the remaining 29 cultures transfer to one to eight additional cells was detected, resulting in an average rate of 2.26 ± 2.3 cells. When MCh and NH4Cl were coapplied, only in a single culture dye transfer was not observed, while in the other 17 cultures transfer to 1–20 additional cells was measured, resulting in an average rate of 7 ± 5.8 cells. Dye transfer rates in each condition were significantly different (P < 0.05, t test). With the exception of two cases, in which 20 adjacent cells were revealed after application of MCh and NH4Cl, dye transfer was characterized by specific labelling of well defined separate cells that were not immediately adjacent to the injected cell (Figs 8 and 9), including PCs and interneurones. Furthermore, no dye transfer was detected when the marker Lucifer yellow was used for MCh and NH4Cl coapplication (n= 12, not shown), a dye that in contrast to biocytin does not label the axonal compartment, but mainly labels the somatic and dendritic compartments, of the injected cells (Buhl & Lübke, 1989), thus indicating that the dye transfer reaction occurs in the axonal compartment of the cells. Thus, MCh specifically triggered the axonal coupling via gap junctions.
|
|
|
Lastly, I report of experiments showing that the intrinsic oscillator, in the terms described in the Summary and Discussion, and illustrated in Fig. 11, is functional under conditions where fast excitatory and inhibitory transmission are eliminated from the network. Under such conditions, the core element is expected to remain active through the suggested mechanisms of (a) the creation of axonal activation pathways in the axonal network of the intrinsic oscillator and (b) the formation of a reverberatory system that maintains a critical number of action potentials in the core element. In three experiments clear effects were observed. First, the response of individual CA3 PCs (five cells) to application of MCh (20 nM) was examined, to which the cells exhibited the control oscillatory response (Fig. 12). From this activated state of the network, and still in the presence of MCh, fast glutamatergic (CPP 10 µM, CNQX 20 µM or NBQX 20 µM) and GABAergic (bicuculline 20 µM) transmission were blocked. As all synaptic contributions are eliminated, traces depict the residual activity, which consisted of spikelets (indicative of axonal origin) (Fig. 12). These spikelets were found to be sensitive to TTX (0.5 µM), which led to the almost full elimination of the residual activity (Fig. 12). Thus it can be concluded that the residual activity depends on the generation and transmission of TTX-sensitive action potentials and spikelets in the axonal network that forms the intrinsic oscillator. In an additional four similar experiments spikelets were not observed. This may have been due either to the lack of functional coupling between the recorded cells and the core element, or to functional coupling that occurs at an electrical distance that is not detectible by the recording method. Moreover, such patterns of spikelets do not demonstrate the rhythmicity itself, although they may provide an estimate to the range in which individual cells could contribute to the oscillatory pattern by the new mechanism, and they suggest that single cells are unlikely to pace the network.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The action potential kernel during oscillations
Generation of action potentials is critical for the expression of oscillatory activity by the hippocampal network, as their elimination (by TTX) abolishes the oscillatory activity. Therefore it was necessary to identify the cells that are spontaneously discharging, to determine whether the discharge was a direct result of cholinergic activation, and to establish the relation of the discharge to the oscillatory activity.
Three major cell classes are identified as spontaneously discharging during cholinergically mediated oscillatory activity, CA3 and CA2 PCs and CA3 interneurones.
CA3 PC discharge was consistent with participation in both the triggering and maintenance of oscillatory activity. However, discharge of individual cells was scarce, such that the precise timing of their action potential in relation to the oscillatory cycles was not determinable, and was observed in about 25% of the population. Given the fraction of discharging cells (25%), the population size (5000 cells) and the firing rate (0.03 Hz, Fischer et al. 2002), it is calculated that CA3 PCs contribute three to four action potentials per theta cycle in total. Discharge resulted from a cholinergically induced slow transient membrane depolarization, and was correlated and associated with its waveform. Nevertheless, expression of the slow transient depolarization translated to firing only in about 50% of the cases. Moreover, some CA3 PCs discharged at their resting membrane potential, without the activation of this process. Why only a fraction of the CA3 pyramidal cells transiently depolarize, and only a fraction of those discharge, may be due in part to differences in (a) the receptor subtypes and/or subunit composition, (b) the apparent affinity of the muscarinic receptors, or (c) the transduction mechanism underlying the expression of the slow transient depolarization.
CA2 PC discharge was consistent with being involved in the maintenance of oscillatory activity. In about 30% of the population relatively extensive firing was observed, and although discharge was mainly in the theta band instantaneous firing frequencies spanned a large range (0.01–100 Hz). Moreover, firing was largely not correlated with the oscillatory cycles. Thus, CA2 PCs produced a uniform source of action potentials. Discharge resulted from a cholinergically induced gradually developing sustained depolarization, and followed its time course. Nevertheless, area CA2 is limited in length (100–150 µm; Swanson et al. 1978) and the total number of action potentials produced is relatively low. Assuming a population size of 300 cells, it is estimated, on the basis of the cell fraction that discharges and the range of median firing rates, that CA2 PCs contribute a total of 5–30 action potentials per theta cycle.
CA3 interneurones form the largest source of action potentials in the hippocampal network during oscillations, where basket, O-LM, bistratified, str. lucidum-specific and multisubfield cells (Fischer et al. 2002) provide a total of 10–80 action potentials per theta cycle (Fischer & Dürr, 2003). Their discharge was found to determine the apparent oscillatory sequence of CA3 PCs, and to dynamically regulate the network interactions (Fischer & Dürr, 2003). However, the discharge of interneurones did not result from direct cholinergic activation, the duration of individual interneurones' synaptic output did not determine the frequencies of oscillatory activity, the time at which action potentials occurred was incompatible with pacing or triggering of the oscillations, and their firing activity was governed by non-interneuronal network interactions (Fischer & Dürr, 2003). Finally, it is hard to explain how interneurone discharge develops, as they do not depolarize during oscillations (Fischer et al. 2002; Fischer & Dürr, 2003).
These properties of discharge mean that the firing of CA2 and CA3 PCs is critical for the expression of oscillatory activity. However, this also illustrates the difficulty in explaining the oscillatory activity in conventional terms: a cell fires an action potential that travels along its axon to produce release of transmitter that produces an oscillatory cycle. The difficulty appears in several critical aspects. (a) Even if we are to assume a connectivity rate of 50% in area CA3 and of 10–20% from area CA3 to area CA1, then there are not enough action potentials to produce just a simple sequence of consecutive synaptic events. (b) The excitatory oscillatory patterns are based on population event, so it becomes even more difficult to explain such output, not to mention the issue of synchronization, with the number of action potentials generated by principal cells. (c) Principal cell discharge does not explain at all how discharge in interneurones originates and how IPSPs are generated.
How to proceed?
The key issue is how the network utilizes the action potentials generated by the excitatory cells for the expression of synchronous and coherent oscillatory activity. As excitatory action potentials were scarce and/or not correlated with the oscillatory cycle, additional interactions are required. The cholinergically mediated and newly identified reaction of facilitating axo-axonic interactions, by enhancing gap junction opening, may serve this function. Several lines of evidence supported this. Dye transfer experiments indicated that the site of interaction was axonic. Moreover, spikelets, with a waveform indicative of axonic origin (Traub et al. 1999; Traub & Bibbig, 2000), were observed in CA2 and CA3 PCs and in CA3 interneurones during oscillations. Furthermore, alkalization and acidification of the cytosol, considered to target gap junctions (Spray et al. 1981; Thomas, 1984; Draguhn et al. 1998), effectively influenced the oscillatory activity. Thus, gap junctions are the likely candidates to mediate these axo-axonic interactions.
The role of gap junctions in oscillations
How does enhancement of gap junction opening in the axonal network lead to the expression of coherent and synchronous oscillatory activity? Since the oscillatory patterns are based on synaptic sequences, oscillations depend on the creation of activation pathways that would produce them. A ‘simple’ explanation is that the chance of creating consecutive activation sequences is enhanced by the opening of axo-axonic gap junctions. As a result, action potentials are transmitted to axons of inhibitory and excitatory cells that otherwise would not reach the discharge threshold, thereby allowing the creation of activation sequences of specific synaptic patterns. By coordinating the opening and closing of these transmission lines, coherent and synchronous patterns are achieved.
To ensure reliability in this reaction, only a fraction of the available gap junctions are activated at any given time during oscillations. Dye transfer experiments showed an average opening of 2.26 gap junctions per cell during oscillations (assuming a 1-to-1 ratio of coupling and molecules), out of seven functionally available gap junctions per cell, which doubles the basal level of 1.11 gap junctions per cell. These experimental values (1.11 and 2.26) correspond to (a) the percolation limit, i.e. the theoretical number of functional gap junctions per cell above which a significant effect on the network activity can be expected, and (b) the theoretical estimation of 2.5 functional gap junctions per axon required for the expression of hippocampal oscillatory activity (Traub et al. 1999, 2000; Traub & Bibbig, 2000).
The second functional consequence of gap junction opening, which promotes the expression of oscillatory activity, is the creation of a reverberatory system. As a result of the opening of axonal pathways between cells, and depending on the probability of maintaining an activation pathway, action potentials can ‘circulate’ in the axonal network and thus form a reverberatory system. The benefit of such a system is that the rarely generated action potentials can extend their normal time span, thereby, ensuring a critical number of available action potentials, and increasing the chance to express synchronous and coherent activation sequences.
Both functions of gap junction opening, i.e. the activation pathways and the reverberatory system, create a new source of action potentials in the network, in addition to the somatic one.
The core element underlying oscillations
To be operative the anatomical structure of the network must support these functions. A special anatomical structure, the hippocampal associative pathway (Lorente de Nó, 1934; Swanson et al. 1978), is formed by axons of discharging PCs. The elaborate plexus formed in str. oriens, pyramidale and radiatum of areas CA2 and CA3a by axon collaterals of CA3 and CA2 PCs (not all cells) (Fig. 11) creates the hippocampal associative pathway. Contributing CA3 PCs are organized in a reverse gradient to those that give rise to Schaffer collaterals, such that the closer a CA3 PC is to CA2 the more likely it is to contribute to the core element (Lorente de Nó, 1934). I suggest this structure to constitute the core element of the intrinsic oscillator. Such a location of the core element is consistent with the findings that (a) PC discharge is critical for the oscillation, and (b) gap junction-mediated axo-axonic interactions are essential. The full oscillator is formed through the interaction of the active core element with associated axons of interneurones, and their consequent activation. The function of the interneurones is therefore ‘secondary’.
The mechanisms underlying oscillatory activity
How do the oscillations work? The cholinergic input serves only as an enabling agent and oscillations depend on the networks synaptic interactions. Activation of the core element by the acetylcholine-mediated reactions of (a) generation of action potentials in CA2 and CA3 PCs and (b) facilitating gap junction-mediated axo-axonic interactions, together with consequent activation of interneurones, provides the minimal and sufficient conditions for the hippocampal network to adopt the oscillatory mode of activity. The functional consequences of these reactions are the formation of activation pathways (responsible for the individual patterns), and the creation of a reverberatory system (to ensure critical number of action potentials). Pacing is achieved through the probabilistic process of creating and maintaining these activation pathways. In turn, a robust ‘clock’ is formed, and as long as activation pathways are formed and sufficient number of action potentials are provided, within a safety range, the oscillatory repertoire would be faithfully reproduced. This ‘clock’ is not of fixed identity, but depends on the activity of the probabilistic core element. Moreover, such a mechanism can function under conditions of low and tightly regulated numbers of action potentials.
The probabilistic formation of activation pathways arises from the fact that gap junctions are not necessarily open continuously during oscillations. This is quite important for the expression of multirhythmicities by the hippocampal network. This probabilistic feature allows the activation of different activation pathways, of similar and dissimilar electrotonic distances, which in turn would determine the frequencies and pattern. Such a process is evident in the oscillatory sequences of all cells, where almost no two consecutive identical cycles are seen, meaning that different activation pathways are activated at different times in a probabilistic manner.
Functional activation of the processing in the axonal compartment, i.e. the core element and associated axons of other cells, during the oscillatory mode of activity forms a probabilistic clock. Formation of the clock is not a result of fixed hardwiring, but of a distributed network (axonal) processing that changes its inter- and intraconnectivity, such that there exists many and different options to reach the same solution, due to the probabilistic and distributed nature of the oscillator. There is no unique solution, but a single global solution (the activation of the distributed and probabilistic oscillator). Thus, a robust clock that faithfully produces the oscillatory mode of activity as long as its activation conditions are met is formed.
The oscillator in action
Not only does such a mechanism account for the phenomena and patterns observed during oscillations, it could also explain some basic features of the oscillatory activity such as why the oscillations are based on a frequency band rather than a single frequency, multirhythmicity and more. Some interesting and difficult to resolve issues are discussed below.
Several of our results are supported and accounted for by such a mechanism. The oscillatory patterns in interneurones, DGCs and CA1 PCs are governed by a synchronized excitatory input (Fischer et al. 1999; Fischer et al. 2002). In the case of CA1 PCs, oscillatory activity was strongly influenced by the coupling between the oscillator and Schaffer collaterals. As a result of enhancement in gap junction coupling (NH4Cl experiments) CA1 PCs exhibited excitatory theta oscillations of increased amplitude, while an opposite effects was observed when the inhibitory tone was increased (Fischer & Dürr, 2003), both without affecting either the pattern or the frequency. Within the scope of the presented framework, I interpret the decrease in CA1 PCs' theta amplitude during the increase in inhibitory tone to represent an effect of inhibition to decouple the Schaffer collaterals from the oscillator. In addition, decrease in gap junction-mediated Schaffer collateral–core element coupling can lead to (a) blockade of Schaffer collateral output to CA1 (when inhibitory tone was increased) (Fischer & Dürr, 2003) and (b) the blockade of oscillatory activity when NH4Cl was removed. Moreover, while excitatory activation pathways account for the oscillatory pattern of interneurones, coupling of interneurones to the core element can account for their discharge activity. Such coupling, through spike transmission, explains how interneurones, and some CA3 PCs, discharge without being depolarized, and how the discharge of interneurones, at the level of individual cells and populations, is synchronized in the phase domain (Fischer & Dürr, 2003). Under control conditions the process of spike transmission from PCs to interneurones is infrequent and depends on excitatory transmission (Miles, 1990; Traub & Miles, 1995; Marshall et al. 2002), while during oscillation I assume (within the scope of the presented framework) that this process is enhanced, and additionally occurs through the newly generated pathways of gap junctions. Thus, the postulated gap junction-mediated axonal coupling between pyramidal cells, which constitute the core element, and interneurones may account for a significant part of their activity during oscillations. An additional consequence of spike transmission is the appearance of spikelets. Moreover, as a consequence of interneurone function to dynamically regulate the oscillator (Fischer & Dürr, 2003), I assume (within the scope of the presented framework) that an increase in the inhibitory tone leads to the dilution of the excitatory input to interneurones (Fischer & Dürr, 2003) through a decrease in the availability of activation pathways. Furthermore, shutdown of the oscillator by acidification of the cytosol (NH4Cl experiments) strongly supports the involvement of gap junction-mediated interactions in the generation and pacing functions. Lastly, experiments presented in Fig. 12 and accompanying text demonstrate, at least in principle, that the intrinsic oscillator is functionally active and that a reverbatory system is formed, just by satisfying the conditions of cholinergic activation, generation of action potentials and enhancement of gap junction opening in the hippocampal associative pathway – the intrinsic oscillator.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Baimbridge KG & Miller JJ (1982). Immunohistochemical localization of calcium-binding protein in the cerebellum, hippocampal formation and olfactory bulb of the rat. Brain Res 245, 223–229.[CrossRef][Medline]
Bragin A, Jandó G, Nádasdy Z, Hetke J, Wise K & Buzsáki G (1995). Gamma (40–100 Hz) oscillations in the hippocampus of the behaving rat. J Neurosci 15, 47–60.[Abstract]
Buhl EH & Lübke J (1989). Intracellular lucifer yellow injection in fixed brain slices combined with retrograde tracing, light and electron microscopy. Neuroscience 28, 3–16.[CrossRef][Medline]
Buzsáki G (2002). Theta oscillations in the hippocampus. Neuron 33, 325–340.[CrossRef][Medline]
Draguhn A, Traub RD, Schmitz D & Jefferys JG (1998). Electrical coupling underlies high-frequency oscillations in the hippocampus in vitro. Nature 394, 189–192.[CrossRef][Medline]
Fischer Y & Dürr R (2003). Inhibitory control of intrinsic hippocampal oscillations?Brain Res 982, 79–91.[CrossRef][Medline]
Fischer
Y, Gähwiler
BH
&
Thompson
SM (1999). Activation of intrinsic hippocampal theta oscillations by acetylcholine in rat septo-hippocampal cocultures. J Physiol
519, 405–413.
Fischer
Y, Wittner
L, Freund
TF
&
Gähwiler
BH (2002). Simultaneous activation of gamma and theta network oscillations in rat hippocampal slice cultures. J Physiol
539, 857–868.
Freund TF & Buzsáki G (1996). Interneurons of the hippocampus. Hippocampus 6, 347–470.[CrossRef][Medline]
Frotscher M & Léránth C (1985). Cholinergic innervation of the rat hippocampus as revealed by choline acetyltransferase immunocytochemistry: a combined light and electron microscopy study. J Comp Neurol 239, 237–246.[CrossRef][Medline]
Gähwiler BH (1981). Organotypic monolayer cultures of nervous tissue. J Neurosci Meth 4, 329–342.[CrossRef][Medline]
Gähwiler BH & Brown DA (1985). Functional innervation of cultured hippocampal neurones by cholinergic afferents from co-cultured septal explants. Nature 313, 577–579.[CrossRef][Medline]
Gähwiler BH, Thompson SM, McKinney RA, Debanne D & Robertson RT (1998). Organotypic slice cultures of neuronal tissue. In Culturing Nerve Cells, ed. Banker, G &.Goslin, K, pp. 461–498. MIT Press, Cambridge, MA, USA.
Gray JA (1971). Medial septal lesions, hippocampal theta rhythm and control of vibrissal movement in freely moving rat. Electroencephal Clin Neurophysiol 30, 189–197.[CrossRef][Medline]
Hirase
H, Leinekugel
X, Czurkó
A, Csicsvari
J
&
Buzsáki
G (2001). Firing rates of hippocampal neurons are preserved during subsequent sleep episodes and modified by novel awake experience. Proc Natl Acad Sci U S A
98, 9386–9390.
Johnston MF, Simon SA & Ramon F (1980). Interaction of anaesthetics with electrical synapses. Nature 286, 498–500.[CrossRef][Medline]
Jung R & Kornmüller AE (1938). Eine methodik der ableitung lokalisierter potential schwankungen aus subcorticalen hirngebeiten. Arch Psychiat Nervenkrankh 109, 1–30.[CrossRef]
Kocsis
B, Bragin
A
&
Buzsáki
G (1999). Interdependence of multiple theta generators in the hippocampus: a partial coherence analysis. J Neurosci
19, 6200–6212.
Lee MG, Chrobak JJ, Sik A, Wiley RG & Buzsáki G (1994). Hippocampal theta activity following selective lesion of the septal cholinergic system. Neuroscience 62, 1033–1074.[CrossRef][Medline]
Leslie J, Nolan MF, Logan SD & Spanswick D (1998). Actions of carbenoxolone on rat sympathetic preganglionic neurones in vitro. J Physiol P, 146P.
Lorente de Nó R (1934). Studies on the structure of the cerebral cortex II. Continuation of the study of the ammonic horn. J Psychol Neurol 46, 113–177.
