Cellular-Synaptic Generation of EEG Activity

Cellular-synaptic generation of EEG activity

György Buzsáki1, Roger D. Traub2 and Timothy Pedley3

1Center for Molecular and Behavioral Neuroscience, Rutgers,

The State University of New Jersey, 197 University Avenue,

Newark, NJ 07102

2 Department of Physiology, Neuroscience Unit

Medical School, University of Birmingham

Birmingham B15 2TT, U.K.

and

3Department of Neurology

Columbia University, New York, NY 10032

Correspondence:György Buzsáki

Center for Molecular and Behavioral Neuroscience, Rutgers University

197 University Avenue

Newark, NJ 07102

Tel: (973) 353-1080 ext. 3131

Fax: (973) 353-1588

E-mail:

Key words: EEG, cellular activity, synchrony, extracellular currents, intrinsic oscillations, synaptic activity, current-source density analysis

Supported by NIH (NS34994, MH54671) and the Wellcome Trust.

Introduction

To date, three methods can provide high temporal resolution of neuronal interactions at the network level: electric field recording (EEG), magnetoencephalogram (MEG; 51, 70) and optical imaging (32, 86). Each of these have their advantages and shortcomings. MEG is not practical for experimental work on freely moving subjects due to the large size of magnetic sensors. A major obstacle of the optical imaging method is that its "view" is confined to surface events. Since most of the network interactions occur in the depth of the brain at the level of the synapses, a search for alternative methods is warranted. In addition, research in both MEG and optical imaging fields face the same fundamental questions as those arose decades ago in connection with scalp-recorded EEG: the "reverse engineering" problem of signal interpretation (cf. 10, 31, 63).

Membrane currents generated by neurons pass through the extracellular space. These currents can be measured by electrodes placed outside the neurons. The field potential (i. e., local mean field), recorded at any given site, reflects the linear sum of numerous overlapping fields generated by current sources (current from the intracellular space to the extracellular space) and sinks (current from the extracellular space to the intracellular space) distributed along multiple cells. This macroscopic state variable can be recorded with electrodes as a field potential or electroencephalogram (EEG) or with magnetosensors (SQUIDs) as a magnetoencephalogram (MEG). These local field patterns, therefore, provide experimental access to the spatio-temporal activity of afferent, associational and local operations in a given structure. To date, field potential measurements provide the best experimental and clinical tool for assessing cooperative neuronal activity at high temporal resolution. However, without a mechanistic description of the underlying neuronal processes, the scalp or depth EEG simply remains a gross correlate of brain activity rather than a predictive descriptor of the specific functional/anatomic events. The essential experimental tools for the exploration of EEG generation have yet to come. In this chapter we provide a basic description of field potential generation in the mammalian archicortex and neocortex and summarize recent progress and future directions.

A straightforward approach to deconvolve the surface (scalp) recorded event is to simultaneously study electrical activity on the surface and at the sites of the extracellular current generation. Electrical recording from deep brain structures using wire electrodes is one of the oldest methods in neuroscience. Local field potential measurements or "micro-EEG" (66) combined with recording of neuronal discharges is the best experimental tool available to study the influence of cytoarchitectural properties, such as cortical lamination, distribution, size and network connectivity of neural elements on electrogenesis. However, large number of observation points combined with decreased distance between the recording sites are required for high spatial resolution and for making interpretation of the underlying cellular events possible. Progress in this field is expected to be accelerated by the availability of micromachined silicon-based probes with numerous recording sites (60). The information obtained from the depth of the brain will then help interpret the surface-recorded events. Clearly, such a task requires collaborative work among neuroscience, silicon nanotechnology, micromachinery, electric engineering, mathematics and computer science. The stake is high, since interpretation of macrosignals such as those obtained with EEG, MEG, fast MRI, PET or optical imaging methods will still require network (submillimeter) level interpretation of the cellular- synapic interactions.

In principle, every event associated with membrane potential changes of individual cells (neurons and glia) should contribute to the perpetual voltage variability of the extracellular space. Until recently, synaptic activity has been viewed as the exclusive source of extracellular current flow or EEG. As will be discussed below, however, synaptic activity is only one of the several membrane voltage changes that contributes to the measured field potential. Progress during the past decade revealed numerous sources of relatively slow membrane potential fluctuations, not directly associated with synaptic activity. Such non-synaptic events may contribute also significantly to the generation of local field potentials. These include calcium spikes, voltage-dependent oscillations and spike afterpotentials observed in various neurons.

Sources of extracellular current flow

1. Fast (Na+) action potentials

The largest amplitude intracellular event is the sodium-potassium spike, referred to as the fast (Na+)action potential intracellularly and as unit activity extracellularly. Individual fast action potentials are usually not considered to contribute significantly to the scalp recorded EEG, mainly because of their short duration (<2 msec). An additional factor is the high-pass frequency filtering (capacitive) property of the extracellular medium, which attenuates spatial summation of high frequency events. As a result, the amplitude of extracellular unit activity decreases much more rapidly with distance between the cell membrane and the recording site than is the case for slower membrane events. However, when a microelectode is placed close to the cell body layer of cortical structures the recorded field potentials contain both extracellular units and summed synaptic potentials. Furthermore, when action potentials from a large number of neighboring neurons occur within a short time window, e. g., in response to electrical stimulation of afferents, during epileptic activity or even during synchronous physiological patterns, these "population spikes" can be recorded with relatively large size electrodes and in a larger volume (Fig. 1; refs. 4, 9, 25).

2. Synaptic activity

In most physiological situations, synaptic activity is clearly the most significant source of extracellular current flow or EEG. The notion that synaptic potentials contribute to the generation of EEG stems from the recognition that for the summation of extracellular currents from numerous individual compartments, the events must be relatively slow (39). The dendrites and soma of a neuron form a tree made of an electrically conducting interior surrounded by a relatively insulating membrane with tens of thousands of synapses on it. Each synapse acts as a small battery to drive current, always in a closed loop. Depending on the chemical nature of the neurotransmitter released in the synaptic cleft, the postsynaptic membrane is depolarized (EPSP) or hyperpolarized (IPSP). Excitatory currents, involving Na+ or Ca2+ ions, flow inwardly at an excitatory synapse (i. e., from the activated postsynaptic site to the other parts of the cell) and outwardly away from it. Such an outward current is referred to as a passive return current from the intracellular milieu to the extracellular space. Inhibitory loop currents, involving Cl- or K+ ions, flow in the opposite direction. The current flowing across the external resistance of the cortex sums with the loop currents of neighboring neurons to constitute a local mean field (Fig. 1). Viewed from the perspective of the extracellular space, membrane areas where current flows into or out of the cells are termed sinks or sources, respectively. The active or passive nature of the sinks and sources are ambiguous, unless the location and types of synapses, involved in the current generation, are identified. Supplementary information may come from simultaneous intracellular recording from neurons dominantly involved in the current generation. Alternatively, extracellular recording of the action potentials and their cross-correlation with the laminar distribution of the field event can provide the necessary clues for the identification of an sink as opposed to a passive return (inward) current of an active inhibitory source (outward). Cross-correlation of the interneuronal discharges with the field potential in question may further decrease the ambiguity regarding the passive versus active nature of the sink-source dipole (16).

Identification of synaptic currents in the archicortex

Figure 1 illustrates the necessary steps in the identification of network mechanisms of evoked and spontaneous field events. The example is taken from the hippocampus, because it is a simple, three-layered structure consisting of orderly arranged principal cells (pyramidal and granule cells) and interneurons. Therefore, the synaptic interpretation of the extracellular current is much simpler than in multilayered structures. The termination zones of the excitatory paths and the inhibitory connections are also well studied in the hippocampus (10, 84). Activation of the excitatory associational input by indirect, trisynaptic electrical stimulation will depolarize the mid-apical and basal dendrites of pyramidal cells (shown in blue in Fig. 1). The passive return current will flow out of the cells at the level of the neuronal bodies and distal apical dendrites (shown in red in Fig. 1). This change in voltage is reflected by the characteristic distribution of field potentials in different depths. The extracellular voltage is negative close to the excitatory synapse and positive in the cell body layer. The reason for this is the large depolarization of the dendrite and the gradual decrease of intracellular depolarization towards to soma. This synaptic activity-induced intracellular voltage difference between the dendrites and soma (a "dipole") will result in a current flow across the membrane (arrows in Fig. 1F). Simultaneous events in many neighboring pyramidal cells will linearly summate and produce an extracellular voltage fluctuation which can be measured with closely-spaced electrodes. After determining the impedance characteristics of the extracellular space, the voltage change can be converted into current change (28).

Increased afferent discharge also activates interneurons, some of which terminate on the cell bodies of the pyramidal cells. The discharging basket cells release GABA and activate Cl- channels with resulting hyperpolarization of the pyramidal cell somata. Somatic hyperpolarization, in turn, creates a voltage gradient between the soma and dendrites (inhibitory dipole). The created intracellular voltage difference is the driving force of charges across the cell membrane and the consequent spatially distributed current flow in the surrounding extracellular fluid (Fig. 1). Note that the direction of current flow is the same as in the case when the driving force is apical dendritic depolarization (active sink). Since the direction of current flow is identical for dendritic excitation and somatic inhibition, the excitatory and inhibitory currents will sum in the extracellular space, resulting in large amplitude field potentials.

The contribution of GABAa-mediated inhibitory currents, however, is believed to be small, because the Cl- equilibrium potential is close to the resting membrane potential. Thus, the change of the transmembrane voltage is limited. However, in actively spiking neurons, when the cell body is depolarized, the transmembrane potential, mediated by GABAa synapses can be large. Another cautionary note is that inhibition may operate also on the dendrites, causing current flow opposite to the direction of excitatory currents. For the identification of excitatory and inhibitory components, represented by the extracellular current flow, a precise knowledge about the anatomical network is essential. Physiological experiments, including recordings from interneurons and pyramidal cells as well as differential pharmacological blockage of the excitatory and inhibitory synapses, can then provide the necessary knowledge for the proper interpretation of the observed sinks and sources. When all this knowledge is in place, the extracellular events can be interpreted unambiguously.

Provided that dendritic excitation is strong enough to override somatic inhibition, the cells may discharge. In the simplest case, a Na+ spike will be generated in the initial segment of several pyramidal neurons. The large inward current associated with the spike is reflected by a negative change of the extracellular voltage at the level of the axon initial segment/cell body accompanied by a smaller amplitude, extracellular positive deflection out in the dendritic regions for the same reasons as described above for the EPSPs. However, since the spatial location of this event is opposite to the afferent excitation of dendrites, the direction of the extracellular current flow will also be opposite. The contribution of fast spikes to the extracellular mean field is due to the hypersynchronous discharge of many pyramidal neurons (population spike) as a result of artificial stimulation of an afferent bundle. Interpretation of the extracellular events after the population spike is not straightforward, however, due to complex feedback effects of the network and other non-synaptic events (see below).

Once a circuitry, such as shown in Fig. 1, has been "calibrated" by electrically evoked potentials, one can move to the next step: network level description of the generation of spontaneous EEG events. The tutorial example is an intermittently occurring, large amplitude hippocampal sharp wave (SPW). SPWs are present during immobility, consummatory behaviors and slow wave sleep. It is important that the events to be analyzed are clearly separable from other waveforms. After extracting the invariable features of this EEG pattern by averaging or other pattern recognition methods, the simultaneous voltage measurements are converted into a current-source density map (Fig. 1D and E). Note that the distribution of the sinks and sources of SPW is strikingly similar to the potentials elicited by stimulation of the associational/commissural inputs to the pyramidal cells. Indeed, experimental work revealed that SPW in the hippocampus arise from the quasi-synchronous discharge of CA3 pyramidal neurons, the source of associational and commissural afferents to the CA1 region (9, 13). Temporally overlapping activation of converging activity on single CA1 pyramidal cells results in a large depolarization of the dendrites, similar to the depolarization of these cell when the associational pathways are electrically activated. These extra- and intracellular events therefore provide circumstantial evidence that the same neuronal machinery is activated during spontaneously occurring SPWs as during electrical stimulation of the associational afferents.

Identification of synaptic currents in the neocortex

The strategy described above is, in principle, applicable to any other a priori identified rhythmic or sparse EEG event. Complications arise when several dipoles are involved in the generation of the same EEG patterns, especially when these dipoles are phase-shifted, as is the case in the generation of numerous neocortical patterns (11, 80, 81).

Of the neocortical EEG patterns, two conspicuous low-frequency (<15 Hz) rhythms, the physiological sleep spindles and spike-and-wave discharges, associated with petit mal epilepsy, have been studied most extensively (11, 14, 44, 55, 77, 80). It is widely accepted that the source of rhythm generation for both patterns is the interplay between the GABAergic reticular nucleus and corticopetal nuclei of the thalamus (11, 14, 79, 80). It is less clear, however, whether synaptic currents of the thalamocortical afferents can fully account for these rhythms or whether intracortical circuitries are significantly involved in their generation (40). Initially, the "recruiting" response, evoked by repetitive stimulation of intralaminar thalamic nuclei, was thought to be the evoked equivalent of spontaneous spindle waves and spike-and-wave patterns (19, 24, 41, 59, 67). Subsequent studies, however, have suggested that spindle waves are more similar to the "augmenting" response; a pattern evoked by repetitive simulation of sensorimotor thalamic nuclei (58, 75, 76). From the point of EEG generation, this distinction is important since recruiting and augmenting responses have different voltage-versus-depth profiles in the cortex. Thus, a critical issue is the identity of synapses and neurons involved in the generation of these rhythmic patterns. If the thalamocortical synapses are the major source of the extracellular synaptic current then the major sinks are expected to correspond to the anatomical targets of the corticopetal thalamic fibers.

Using the approach described above for the hippocampus helped clarify these issues (Fig. 2; 44). The most striking aspects of the experiment shown is Fig. 2 is the general similarity of the spontaneous and evoked field events, independent of the initiating conditions. The spatial position of the major current sinks are sources are similar, independent which thalamic nucleus is being stimulated or which hemisphere. The differences are expressed mainly in the latencies of the large sink-source pairs. Therefore, the similar spatio-temporal distribution of the main sinks and sources suggest that the major current flow derives from the activity of the intracortical circuitry. The neocortex, in essence, functions as a powerful amplifier during these oscillatory events. Because the thalamocortical network is in a metastable state during reduced activities of the brainstem and basal forebrain (55; 77), a weak thalamic or callosal input is capable of recruiting a large population of intracortical neurons. The triggering input may even remain undetectable in the field and the spread of activity reflects primarily the connectivity and excitability of the cortical circuitry rather than the nature of the initiating input (16, 17).

The CSD map and the associated multiple-site unit analysis also revealed that at least three dipoles were involved in the generation of the rhythmic field events (Fig. 3; 44). The most consistent dipole was characterized by a major sink in layer IV (dipole 2). When a surface-positive field component was present, it was associated with a major sink in layers V-VI and a source in layers II-III (dipole 1). The third, delayed dipole was represented by a surface-negative spike component and a corresponding sink in layers II-III (dipole 3). The relative strength of these respective sinks varied within single episodes of HVS (Fig. 3). Although the numerous cell types and the complexity of the intracortical circuitry makes identification of the cellular-synaptic origin of neocortical EEG less accessible, these recent findings indicate that the use of simultaneous recording of field and unit activity is a proper method for the revelation of the synaptic-cellular mechanisms of extracellular current flow in the neocortex.