INTRODUCTION

TOP ABSTRACT INTRODUCTION IMPACT OF NETWORK ACTIVITY... DEVELOPMENT OF NORMAL BRAIN... REFERENCES

This article emphasizes that network synaptic activities modulate, and often overwhelm, intrinsic neuronal properties. I certainly wish that this claim would become a truism for neuroscientists. However, with the advent of in vitro preparations, which provided not only some technical advantages over the work in vivo but also helped to achieve a better understanding of brain operations, a climate arose in which simplistic concepts sometimes appeared, such as the idea that the firing patterns induced by current pulses, taken to define the electrophysiological properties of a given neuronal type, are inflexible. This belief stands in contrast with data showing that patterns of neuronal activity change at various levels of membrane potential and with synaptic activity during shifts in behavioral states. There is also a tendency toward obtaining pure rhythms arising in simple circuits, whereas the intact brain displays oscillations of different types that are grouped together within complex wave sequences due to interactions between a variety of structures. Some investigators working in brain slices use their data to infer normal and pathological processes that require global operations in an intact brain.

Clearly both simplified preparations and normally operating networks are needed, but so far there are too few attempts to regard isolated networks within the context of the whole brain. The development of methods has succeeded in dissecting the brain and transforming it into reduced neuronal circuits. While behavioral and system neuroscience stands to gain from the achievements of biophysics and molecular biology in simplified preparations, the logic of life requires orchestration of the different parts composing the whole. The goal is to apply the information obtained from studies of isolated neurons and simple networks within the context of an intact brain. I will discuss the impact of synaptic activities on neuronal properties, as well as these effects during normal and paroxysmal oscillations in corticothalamic neuronal loops.

	IMPACT OF NETWORK ACTIVITY ON INTRINSIC NEURONAL PROPERTIES

TOP ABSTRACT INTRODUCTION IMPACT OF NETWORK ACTIVITY... DEVELOPMENT OF NORMAL BRAIN... REFERENCES

The intrinsic properties of neocortical and thalamic neurons were first revealed in brain slices. The major advantages of these simplified preparations are the control of the extracellular ionic environment, the simultaneous exploration of different neuronal compartments, and the possibility of investigating the actions of neurotransmitters on identified neuronal types after blockage of synaptic transmission. Presently, some of these techniques are not possible in vivo. In contrast with the earlier view of nerve cells acting in a purely reflex way, with little consideration for the role of their intrinsic properties, the host of voltage- and transmitter-gated conductances discovered in vitro (Gutnick and Crill 1995; Huguenard 1996; Llinás 1988) have provided new insights into the functions of different brain structures and changed our thinking on the electrical properties of central neurons. The ionic nature of different types of conductances has been investigated in cortical and thalamic neurons (Crill 1996; Gutnick and Mody 1995; Llinás 1988; Schwindt et al. 1988a,b, 1989). Multiple intracellular recordings from various cell types in neocortical slices, and from different compartments of single neurons, have revealed single-axon excitatory and inhibitory postsynaptic potentials (EPSPs and IPSPs) between identified neurons, investigated the mechanisms for coupling the inputs reaching various cortical layers, and demonstrated that synaptic transmission is differentially exerted by the same axon of a pyramidal neuron innervating another pyramidal cell and a local inhibitory interneuron, with synaptic depression in the former case and facilitation in the latter (Markram 1997; Markram et al. 1997, 1998; Thomson and Deuchars 1997; Thomson et al. 1993). These effects are important for understanding the rules underlying frequency-dependent plasticity. Paired-cell recordings revealed networks of electrically and chemically coupled inhibitory interneurons (Galarreta and Hestrin 1999; Gibson et al. 1999).

However, investigators in vitro recommended that the enthusiasm for work in slices must be tempered with caution (Connors and Gutnick 1990). They emphasized the biological and physical reactions occurring in the traumatized tissue and concluded that some neuronal properties described in vitro may be different from those seen in the living organism. Moreover, the overwhelming majority of studies have been conducted on slices from one structure, leaving all related systems aside. The disadvantages of brain slices arise not only because of absence of long-range connectivity but also from the fact that different research groups use animals at various early developmental stages, with different temperature and dissimilar extracellular bathing milieu, which, as shown in the following text (see Neocortex: changing firing patterns during different functional states), may drastically change neuronal properties.

This article addresses the effects exerted by synaptic activity on intrinsic neuronal properties with emphasis on normal and paroxysmal oscillations. In this first section, I will discuss the properties of cortical and thalamic neurons, as investigated in brain slices, when these neurons are embedded in intact-bain circuits and are subject to spontaneous shifts in behavioral states. This part mainly refers to the actions of synaptic activities arising in corticothalamic and generalized modulatory systems on the firing patterns and incidence of some neuronal classes in different types of experiments, membrane potential, apparent input resistance, backpropagation of action potentials, plateau potentials, and regularity of firing patterns. Next I will compare different oscillatory types occurring in the simplified circuits of cortical and thalamic slices with rhythmic activities during natural events that occur in brains with preserved connectivity.

Neocortex: changing firing patterns during different functional states

The morphological diversity of neocortical neurons has been recognized since Ramón y Cajal (1911), and their electrophysiological properties are quite complex (reviewed in Gutnick and Crill 1995). Since the early 1980s, the electrophysiological properties of neocortical neurons were characterized intracellularly by their responses to depolarizing current pulses, first in slices maintained in vitro (Connors et al. 1982; McCormick et al. 1985), thereafter in acutely prepared animals under deep anesthesia (Gray and McCormick 1996; Nuñez et al. 1993; Steriade et al. 1996a, 1998b) and, finally, during chronic experiments in awake cats (Steriade et al. 2001; Timofeev et al. 2001b).

Four cellular types are usually described. 1) Regular-spiking (RS) neurons constitute the majority of cortical neurons. They display trains of single spikes that adapt quickly or slowly to maintained stimulation. 2) Intrinsically bursting (IB) neurons generate clusters of action potentials, with clear spike inactivation, followed by hyperpolarization and neuronal silence. During prolonged depolarizing current pulses, the spike bursts of IB neurons may recur rhythmically at 5-10 Hz. 3) Fast-rhythmic-bursting (FRB) neurons give rise to high-frequency (300-600 Hz) spike bursts recurring at fast rates (generally 30-50 Hz). And 4) fast-spiking (FS) neurons fire thin action potentials and sustain tonically very high firing rates without frequency adaptation.

Generally, RS and IB neurons are pyramidal-shaped neurons, while FS firing patterns are conventionally regarded as local GABAergic cells (but see following text). Neurons displaying FRB firing patterns are either pyramidal-shaped neurons or local-circuit, sparsely spiny or aspiny interneurons. The firing patterns described in cats under anesthesia are similar to those described in vitro or in awake animals. As to the duration of intracellularly recorded action potentials at half-amplitude, measured during the state of natural waking in chronically implanted cats, RS neurons show modes between 0.6 and 1 ms (slightly longer spikes are fired by IB neurons); in contrast, both FRB and FS neurons demonstrate much shorter action potentials, with modes at about 0.3 ms (Steriade et al. 2001).

The preceding classification in four neuronal types had a temporarily heuristic value. However, data discussed below show that this systematization does not implicate strict, distinctly different, categories. Initially, electrophysiologists were impressed by the peculiar properties of some cortical neurons. Because of the virtual absence of spontaneous activity in slices, the characterization of these neurons could not take into consideration the role of synaptic activities generated in neocortex and/or thalamus in modifying the firing patterns resulting from intrinsic cellular properties. The morphological correlate, laminar location, and electrophysiological feature of IB neurons have been thought to be so precise that they were qualified as the signature of layer V and having a unique physiology (Connors and Amitai 1995). This is possibly why consciousness was thought by some theoreticians to result from the activity of a special (bursting) neuronal type (Crick 1994; Crick and Koch 1998). These authors wondered whether the visual representation is largely confined to certain neurons in deep cortical layers and further suggested that there are special sets of awareness neurons in the cortex, specifying layer V bursting cells. It would be a mystery why a peculiar cell class, IB neurons, which do not exceed 5% of cortical neurons in awake preparations (Steriade et al. 2001), would be so privileged that their activity gives rise to global states of awareness and consciousness.

In reality, each of the aforementioned firing patterns does not necessarily apply to a single class of neurons; the electrophysiological characteristics of different cortical cell types are much more flexible than conventionally thought; their location is far from being exclusively confined to distinct cortical layers; and their relative proportions vary with the type of preparation (intact cortex or isolated cortical slabs, anesthetized or nonanesthetized animals).

Thus although FS-firing neurons were previously equated to GABAergic interneurons, it is now known that some local-circuit inhibitory neurons fire like RS or bursting cells (Thomson et al. 1996). That the firing pattern of one neuronal type may be transformed, under certain physiological conditions, into another type became obvious from investigations on various neuronal classes. A reorganization of firing patterns may occur with shifts in the state of vigilance, from deafferented to brain-active behavioral states. The maintained depolarization of IB neurons results in burst inactivation (Mason and Larkman 1990; Timofeev et al. 2000) (Fig. 1A). It was then proposed that thick layer V neurons could operate in two modes, switching between bursts and tonic discharges, as a function of modulatory neurotransmitters (Mason and Larkman 1990). Indeed, the enhanced synaptic activity during brain activation by setting into action the ascending brain stem reticular systems (Steriade et al. 1993a), and in vitro application of some neurotransmitters (Wang and McCormick 1993) released in the intact brain by generalized activating systems (Steriade and McCarley 1990), are all conditions that may transform IB into RS firing patterns (Fig. 1B).

View larger version (51K): [in this window] [in a new window]   Fig. 1. Transformation of bursting to tonic firing patterns in neocortical neurons by changing the membrane potential (Vm) and synaptic activity. A: responses of intrinsically bursting (IB) neuron in isolated cortical slab from suprasylvian gyrus in vivo (cat under ketamine-xylazine anesthesia) to the same intensity of depolarizing current pulse (0.5 nA) at the resting Vm (70 mV) and under slight depolarization (+0.2 nA, 63 mV). A typical burst is expanded at right (). B: area 7 neuron in cat under urethan anesthesia recorded in vivo. An IB cell (as identified by depolarizing current pulses) fired spike bursts during the slow-sleep oscillation and transformed this burst firing into tonic, single-action potentials following brain activation produced by stimulation (horizontal bar, 1.8 s, 30 Hz) of the pedunculopontine tegmental (PPT) nucleus. , an expanded detail showing a spike burst followed by single spikes. A, modified from Timofeev et al. (2000). B, modified from Steriade et al. (1993a).

The idea of transformation from IB into RS firing patterns, based on previous results from anesthetized animals (Steriade et al. 1993a), is now substantiated by a similar transformation during shifts from the natural state of slow-wave sleep to either wakefulness or rapid-eye-movement (REM) sleep when the membrane potential of cortical neurons is slightly depolarized (Steriade et al. 2001). Figure 2 shows different (IB and RS) firing patterns of the same neuron, evoked by depolarizing current pulses applied during slow-wave and REM sleep, respectively. Also, the mode of interspike intervals during the spontaneous activity in slow-wave sleep was at 3-3.5 ms, reflecting the presence of spike bursts, while this mode was absent in REM sleep, and there were many more longer intervals (20-100 ms) during REM sleep, reflecting the single spike firing in the latter state.

View larger version (41K): [in this window] [in a new window]   Fig. 2. Changes in firing patterns of an IB cortical neuron from area 7 during slow-wave sleep (SWS) and rapid-eye-movement (REM) sleep in chronically implanted cat. Top: electroencephalo- and electromyographic (EEG and EMG) patterns characterizing the 2 states as well as intracellular recording of this neuron together with 3 depolarizing current pulses (indicated by current monitor). Below: responses to depolarizing current pulses (the 1st response is indicated by * in the top panel). Note spike doublets in SWS and single spiking in REM sleep. Bottom: examples of spontaneous firing of this neuron during SWS and REM sleep. The interspike interval histograms in each state show a mode at 3-3.5 ms in SWS (reflecting bursting activity), absence of this mode in REM sleep, and many more longer intervals (20-100 ms) in REM sleep, reflecting single spike firing. Modified from Steriade et al. (2001).

Moreover, antidromically identified and intracellularly stained corticothalamic (glutamatergic and excitatory) neurons, recorded in vivo, may fire like FRB neurons in response to depolarizing current pulses, but below that level they fire like RS neurons and, at more depolarized levels, like FS neurons (Fig. 3A). Similar voltage-dependent changes, from RS to FRB and further to FS firing patterns, are observed in formally identified local-circuit basket cells (Fig. 3B) (Steriade et al. 1998b). Work in cortical slices also showed that RS neurons may develop their type of discharges into those of FRB neurons by repeated application of depolarizing current pulses (Kang and Kayano 1994). The transformation of output pattern, from RS single-spike firing to FRB burst discharges (Fig. 3), may render unreliable cortical synapses reliable (Lisman 1997). In vitro, FRB neurons are not seen in animals that are <4 mo of age (Brumberg et al. 2000). An additional factor that complicates the recording of this neuronal type in cortical slices is the composition of the ionic medium that requires 1.2 mM [Ca²⁺]_o, while most in vitro studies use [Ca²⁺]_o of 2 mM or more. Therefore a certain level of increased excitability in neuronal tissue by decreasing [Ca²⁺]_o (Hille 1992) may lead to the transformation of neuronal firing patterns, from an RS into an FRB type. With an ionic composition in vitro closer to that in the intact brain, FRB neurons could eventually be recorded in slices (Brumberg et al. 2000).

View larger version (40K): [in this window] [in a new window]   Fig. 3. Corticothalamic neurons and local-circuit (basket-type) neurons display fast-rhythmic-bursting (FRB) firing patterns that develop into fast-spiking (FS) patterns. Intracellular recordings in cats under ketamine-xylazine anesthesia. A: corticothalamic neuron from area 7, projecting to the lateroposterior (LP) nucleus. Depolarizing current pulses at different intensities (shown) elicited changing patterns, from single spikes (1) to spike-bursts at 25-35 Hz (2 and 3) and, eventually, FS patterns (4). A depolarizing afterpotential (DAP) is indicated by  in 1. Below: antidromic identification of a corticothalamic neuron, displaying the same changes in firing patterns. Stimulus () was applied to the thalamic LP nucleus. Note failure of antidromic response membrane potentials more negative than 58 mV and appearance of excitatory postsynaptic potentials (EPSPs). This is an example of neuron interposed in a corticothalamocortical loop. B: morphologically local-circuit (basket-type) cell located in layer III of area 7. Spontaneous action potentials showed their very brief duration (0.3 ms at half-amplitude; not depicted). Changes in firing patterns, from regular-spiking (RS in 1) to FRB (2-3) and finally to FS patterns (4), similarly to those shown in A for a corticothalamic neuron. DAPs are marked () in 2 and 3. Right: camera-lucida reconstruction of this neuron (see photomicrograph in Steriade et al. 1998b). Modified from Steriade et al. (1998b).

The difficulty in maintaining the strict classification in four distinctly separate cortical cell classes (RS, IB, FRB, and FS) also stems from the fact that neurons with thin (<0.5 ms) action potentials and tonic firing without frequency adaptation (like FS-firing cells), conventionally regarded as local GABAergic neurons, were actually identified as corticothalamic cells (see Fig. 3A). The transformation from FRB to FS patterns was similarly demonstrated in intracellularly stained corticothalamic and local-circuit aspiny or sparsely spiny basket (presumably inhibitory) neurons (Fig. 3, A and B). Note that these transformations in discharge patterns, from those defining the firing of a cell type into another, are not just the result of delivering current pulses because similar changes in membrane potential occur spontaneously when an animal passes from natural slow-wave sleep, characterized by prolonged hyperpolarizing episodes, to either waking or REM sleep (Steriade et al. 2001) (see also following text, Fig. 16). The difficulty of considering a simple dichotomy between the two major cell groups, long-axoned pyramidal (RS) and GABAergic local-circuit (FS) neurons, arises not only from the diversity of inhibitory interneurons in the neocortex (Jones 1988, 1995), expressing different electrophysiological features in at least five anatomical classes (Gupta et al. 2000), but also from the fact that intracellularly stained cells, with the same FRB firing pattern, proved to be either deeply lying pyramidal cells or local-circuit basket cells (Steriade et al. 1998b) (Fig. 3). As remarked in a study by Markram's group working in cortical slices (Gupta et al. 2000), the usual classification of RS, IB, and FS neurons, stemming from previous in vitro experiments (Connors et al. 1982; McCormick et al. 1985), is too vague to encompass the diversity of responses.

Network activity during various functional states is decisive in altering the firing patterns generated by intrinsic neuronal properties. Thus typical FRB patterns, which are evoked during the silent background activity of interspindle lulls (as in slices), are dramatically changed during epochs with rich synaptic activity produced by intracortical or thalamocortical volleys (Fig. 4) and develop into patterns resembling the FS firing (Steriade et al. 1998b).

View larger version (39K): [in this window] [in a new window]   Fig. 4. Changes in responses of a corticothalamic neuron from area 21 (antidromically identified from the LP nucleus) to depolarizing current pulses with different intensities during periods poor and rich in synaptic activity. Cat under barbiturate anesthesia. Field potentials were simultaneously recorded from the related thalamic LP nucleus and from the depth of cortical areas 5, 7, and 21 (the latter in the immediate vicinity of the impaled neuron). Depolarizing current pulses (duration, 200 ms) with 3 intensities (0.4, 1 and 1.2 nA) were applied during interspindle lulls, with negligible or absence of synaptic activity (as in slices), and during spindle sequences, rich in synaptic activity generated by thalamocortical volleys. Note the transformation from rhythmic (35 Hz) spike bursts into tonic firing (450 Hz) without frequency adaptation during neuronal silence, and disruption of intrinsically generated rhythmic spike bursts by network synaptic activity. Modified from Steriade et al. (1998b).

To fully realize the importance of synaptic activity in a living animal, compared with slices, and the striking difference in connectivity as well as incidence of synaptic potentials between slightly different sizes of slices, here are the results of two in vitro studies on sensorimotor neocortex (Thomson 1997; Thomson et al. 1996). Out of 595 dual recordings in which an interneuron was recorded simultaneously with a pyramidal neuron in slices 400-µm thick, 39 yielded monosynaptic, single axon IPSPs, i.e., an average probability of 1:15 of each recorded inhibitory interneuron contacting a neighboring pyramidal cell; however, with slices 500-µm thick, the probability rose about three times. Thomson also reported a significantly higher incidence of connections and an increase in spontaneous activity in 500-µm, compared with 400-µm, slices. The dramatic increase in connectivity on increasing the slice thickness by just 0.1 mm may explain the differences between some results from slices, compared with those from the intact brain. Other dissimilarities, from works in the cerebral cortex, thalamus and related systems, are fully discussed elsewhere (Steriade 2001).

The preceding data show that changes in membrane potential and a high degree of synaptic activity in the intact brain decisively modulate, and even transform, the firing patterns due to intrinsic neuronal properties and expressed by responses to direct depolarization. The ideas on the discrete laminar localization of various neuronal types in neocortex also evolved. IB neurons were initially found in layer V, but IB neurons were later recorded also from layers IV and III (Connors and Amitai 1995; Montoro et al. 1988; Steriade et al. 1993e). The FRB neurons (also termed "chattering") were described as exclusively located in supragranular layers II-III of the visual cortex (Gray and McCormick 1996), but the same type of rhythmic bursting neurons was found in all cortical layers, from II to VI, of sensory-motor and association areas (Steriade et al. 1998b); their deep location is corroborated by antidromic identification as corticothalamic neurons (see Fig. 3A).

The proportions of FS and IB firing patterns in nonanesthetized, awake animals (Steriade et al. 2001) are quite different from those found in anesthetized animals with intact cortex (Nuñez et al. 1993; Steriade et al. 1998b) or small isolated slabs in vivo (Timofeev et al. 2000); the latter type of experiments partially reproduce the in vitro condition. Neurons displaying the firing patterns of FS neurons are much more numerous in naturally alert animals (24%) than in the intact cortex of anesthetized animals (12%) or in small isolated cortical slabs (4%). The FS (putative inhibitory) neurons have been implicated in the generation of fast (20-40 Hz) rhythms (Buzsáki and Chrobak 1995; Llinás et al. 1991; Lytton and Sejnowski 1991; Traub et al. 1999) that characterize the spontaneous activity in the waking state and dreaming mentation in humans and animals (Llinás and Ribary 1993; Steriade et al. 1996a,b). These states of network activity, accompanied by relatively depolarized levels of membrane potential, may transform neurons with other firing patterns (i.e., FRB) into FS-type neurons (see Fig. 3). This would result in an increased proportion of neurons identified as FS. On the contrary, neurons displaying IB firing patterns are found in <5% of neurons of awake animals (Steriade et al. 2001), whereas they represent ~15% of neurons in anesthetized animals (Nuñez et al. 1993; Steriade et al. 1998b) and may reach 40% of neurons in isolated cortical slabs in vivo (Timofeev et al. 2000) or in some studies in vitro that reported proportions of 64% IB neurons (Yang et al. 1996). The strikingly diminished proportion of IB firing patterns in the alert condition is likely due to the enhanced synaptic activity and increased release of some modulatory neurotransmitters, i.e., conditions that may transform IB into RS firing patterns (Steriade et al. 1993a; Wang and McCormick 1993).

To sum up, the bursting and regular (tonic) firing patterns represent a continuum of discharge properties and the electrophysiological distinctions between various neuronal classes are much less clear-cut in nonanesthetized animals than were conventionally thought in the early studies on cortical slices or in anesthetized preparations.

The impact of spontaneous synaptic activity on intrinsic neuronal properties was further studied with emphases on the membrane potential (V_m), the apparent input resistance (R_in, a measure resulting from passive electrical neuronal properties and balanced changes in excitatory and inhibitory inputs from specific and modulatory pathways), backpropagation of action potentials from the axonal initial segment to dendrites, and plateau potentials after blockage of K⁺ currents. These issues are discussed below.

MEMBRANE POTENTIAL AND INPUT RESISTANCE. The isolated cortical slab in vivo (10 × 6 mm) is a new preparation that was introduced to examine the necessary number of interconnected neurons for the presence of sleep-like oscillations and that has the advantages of both in vitro and in vivo preparations; that is, it does not drastically change the milieu of the neurons in the network (Timofeev et al. 2000). In this preparation, triple intracellular recordings have been first performed in vivo. The mean V_m in small isolated neocortical slabs in vivo is 70 mV and the R_in is 49 M, whereas in intact (adjacent) cortical areas of the same animal the values are 62 mV and 22 M, respectively. In another study (Paré et al. 1998b), the differences between in vivo and in vitro recordings of the same type of pyramidal neurons are as follows: the standard deviation of the intracellular signal is 10-17 times lower in vitro than in vivo and the R_in measured in vivo during relatively quiescent periods (37 ± 3.9 M) is reduced by 70% during epochs associated with intense synaptic activity, and increases by 70%, approaching the in vitro values (66.14 ± 1.3 M), after tetrodotoxin (TTX) application in vivo (Fig. 5).

View larger version (61K): [in this window] [in a new window]   Fig. 5. Comparison of spontaneous synaptic activity displayed by neocortical neurons in vivo and in vitro and impact of synaptic activity on the resting properties of neocortical neurons. A: intracellular recording of an infragranular RS neuron from cat suprasylvian gyrus under barbiturate anesthesia, together with EEG activity (rest, 64 mV). B: intracellular recording of an infragranular neuron from a slice of cat suprasylvian gyrus, recorded at 34°C (rest, 76 mV). C: intracellularly stained suprasylvian cortical neuron with Neurobiotin in vivo. D: effect of tetrodotoxin (TTX) dialysis in vivo on apparent input resistance (Rin) of an infragranular RS neuron and on amplitude of response to a cortical stimulus (D1) and voltage response to a hyperpolarizing current pulse of constant amplitude (0.2 nA; D2).  the onset of TTX dialysis. Insets: comparison of cortically evoked response and voltage response to current pulses before and 20 min after onset of TTX dialysis (averages of 20 sweeps, same scaling). Modified from Paré et al. (1998b).

View larger version (38K): [in this window] [in a new window]   Fig. 6. Apparent input resistance (Rin) of neocortical neurons during natural states of vigilance in chronically implanted cat. Top: 3 periods of intracellular recording from the same RS neuron during SWS, REM sleep and waking. Rin was measured by applying 0.1-s hyperpolarizing current pulses, every 0.5 s. Bottom: averages of responses of this neuron during different epochs in the 3 states of vigilance (note differences between the hyperpolarizing and depolarizing phases of the slow oscillation in SWS and between epochs with and without ocular saccades in REM; see text). Modified from Steriade et al. (2001).

BACKPROPAGATION OF ACTION POTENTIALS. The backpropagation of action potentials (APs) has two aspects. The first refers to the propagation of APs generated at ectopic sites (regions remote from the axon hillock) toward the soma. Ectopically generated APs generally occur in pathological conditions, such as thalamic (Gutnick and Prince 1972; Schwartzkroin et al. 1974) or callosal (Schwartzkroin et al. 1975) neurons projecting to cortical epileptic foci. The second aspect of backpropagation, which is still a disputable issue because of differences in results from in vitro and in vivo experiments (see following text), is the initiation of APs in the axon hillock and its propagation into the dendrites of various neuronal types, thus providing a retrograde signal of neuronal output to the dendritic tree (Häusser et al. 2000; Stuart and Sakmann 1994; Stuart et al. 1997). The functional consequences of APs, backpropagated from the axon hillock to dendrites, may be an influx of Ca²⁺ without evoking a Ca²⁺ AP (Larkum et al. 1999). This would imply a facilitation of the initiation of Ca²⁺ APs when backpropagating APs coincide (within a time window of ~10 ms) with distal dendritic inputs and is regarded as a mechanism for coupling inputs reaching cortical neurons at different layers. The backpropagating APs may signal the level of neuronal output to the dendritic sites receiving synaptic inputs, thus serving as a link between output and input.

PLATEAU POTENTIALS. Plateau potentials in neocortical neurons are elicited after blockage of K⁺ currents and are due to a class of high-voltage-activated Ca²⁺ channels in dendrites (Reuveni et al. 1993; Yuste et al. 1994). The high background activity in vivo may block the Ca²⁺ plateau potentials. Synaptic inputs lead to the termination of plateaus (Paré et al. 1998a). Using dual intracellular recordings in vivo, with one pipette filled with potassium acetate to control network activity and the other pipette filled with cesium acetate to block K⁺ currents, we showed that synaptic inputs, generated by corticipetal volleys during thalamically generated spindle oscillations, consistently shut off plateaus (Contreras et al. 1997c). Similarly, PSPs evoked by electrical thalamic stimulation blocked the Cs⁺-induced plateaus (Fig. 7). The dendritic Ca²⁺ electrogenesis in cortical neurons (Kim and Connors 1993; Llinás 1988) may play an important role in synaptic plasticity (Swanson 1989). The fact that Cs⁺-induced plateaus are blocked (but sometimes triggered) by synaptic inputs suggests that coherent oscillations in thalamocortical networks may have a role in plasticity by modifying Ca²⁺ electrogenesis.

View larger version (35K): [in this window] [in a new window]   Fig. 7. Thalamic-evoked PSPs block the plateau potentials evoked by blockage of K+ currents. Cat under barbiturate anesthesia. A: depolarizing current pulses of constant amplitude and duration were applied at rest to 2 simultaneously recorded cells, 1 with cesium acetate (cell 1)- and the other with potassium acetate (cell 2)-filled pipettes. Stimulation of thalamic LP, delivered during the depolarizing pulse, blocked the plateaus in cell 1, while eliciting an EPSP and a spike burst in cell 2. Consecutive sweeps were displaced vertically and horizontally for clarity. B: 5 depolarizing pulses of increasing amplitude were applied to both cells. In cell 1, pulses triggered plateaus with progressively shorter latency that were shut off by thalamic stimulation. In cell 2, responses to pulses were passive except for a highest-amplitude pulse that triggered a direct spike; thalamic-evoked EPSPs became suprathreshold with increasing depolarization (, response during highest amplitude pulse). Pulse protocols indicated in A and B. From Contreras et al. (1997c).

REGULAR FIRING PATTERNS DURING BRAIN-ACTIVE BEHAVIORAL STATES. The reliability of firing patterns increases with fluctuating current waveforms resembling synaptic activity (Mainen and Sejnowski 1995). Suprathreshold current pulses to the soma elicit spike trains with a progressive lack of reliability in precise timing, whereas fluctuating current waveforms evoke precise and stable timing throughout the length of the trials (Fig. 8). In the latter condition, the action potentials may be separated by 100 ms, yet they occur with the precision of most responses in the range of 1-2 ms (see also Nowak et al. 1997). These data suggest that currents resembling synaptic inputs may be repeatedly encoded into spike patterns with millisecond precision. The results regarding the regularity of firing evoked by depolarizing current pulses with added fluctuating waveforms, which simulate synaptic activity, fit well with the relative regularity of firing seen during behavioral states of vigilance associated with a high degree of synaptic activity as in wakefulness and REM sleep (Evarts 1964; Steriade 1978; Steriade et al. 1974). The firing regularity during these two brain-active states is expressed by Gaussian-like interspike interval histograms with virtual absence of very short (<25 ms) and very long (>150 ms) intervals (see Fig. 7 in Steriade et al. 1974). During slow-wave sleep, when the cerebral cortex is disconnected from the outside world because of the blockade of synaptic transfer within the thalamus (Steriade et al. 1969; Timofeev et al. 1996), there is a much greater irregularity of firing patterns because the presence of spike bursts, reflected by very short interspike intervals, interspersed with long periods of silence.

View larger version (42K): [in this window] [in a new window]   Fig. 8. Reliability of firing patterns of cortical neurons evoked by constant and fluctuating current. A: a suprathreshold DC current pulse (150 pA, 900 ms) evoked trains of action potentials (~14 Hz) in a RS neuron from layer V of rat visual cortex, in vitro. Responses are shown superimposed (1st 10 trials, top) and as a raster plot of spike times over spike times (25 consecutive trials, bottom). B: the same cell as in A was again stimulated repeatedly, but this time with a fluctuating stimulus (Gaussian white noise). From Mainen and Sejnowski (1995).

Thalamus: effects of synaptic inputs on neuronal properties and local oscillations

There are three main classes of thalamic neurons: thalamocortical (TC), which are all glutamatergic, thus excitatory; thalamic reticular (RE), which send their axons to the dorsal thalamus and are all GABAergic, thus inhibitory; and local-circuit GABAergic neurons whose axonal domain is confined within the limits of the thalamic nucleus where their somata are located. With few exceptions, the three types of thalamic neurons are homogenous in terms of morphology (Jones 1985) and electrophysiological properties (Steriade et al. 1997b). This stands in contrast with the variety and complexity of neocortical neurons.

Most schemes of thalamic functioning include only TC and RE neurons. However, all dorsal thalamic nuclei of cats and primates (and the lateral geniculate nucleus of rodents) also possess an important proportion (25%) of local-circuit inhibitory interneurons (Jones 1985). About 8-10% of RE neurons project to local thalamic interneurons (Liu et al. 1995), and although apparently minor, this GABAergic-to-GABAergic projection may produce significant effects on the ultimate targets, TC neurons, eventually leading to their disinhibition. Indeed, a greatly increased incidence of IPSPs in TC neurons was observed after destruction of RE neurons, reflecting the release from the inhibition of local interneurons after the excitotoxic lesion of RE perikarya (Steriade et al. 1985). The connection between the two types of thalamic GABAergic cells, RE and local-circuit interneurons, may be important for focusing attention to relevant signals (Steriade 1999). Figure 9 illustrates this hypothesis. The top RE neuron, part of the RE pool that is directly connected to the top TC (Th-cx) neuron, contributes to enhancement of relevant activity by inhibiting the appropriate pool of local-circuit elements. Simultaneously, the activity in adjacent RE areas is suppressed by RE-to-RE GABAergic contacts within the nucleus. The consequence would be the disinhibition of related local interneurons (bottom L-circ cell) and the inhibition of weakly excited TC neurons in areas adjacent to the active focus.

View larger version (154K): [in this window] [in a new window]   Fig. 9. Relations between GABAergic thalamic reticular (RE) and local-circuit (L-circ) neurons, and their effects on thalamocortical (Th-cx) neurons. The top Th-cx in the figure receives prevalent excitation from the afferent fiber (Aff.) while the bottom Th-cx receive less collaterals from the Aff. axon. The RE neurons, which are directly connected to the top Th-cx neuron (the top RE neuron is part of this pool), contribute to further enhancing the relevant activity by inhibiting the pool of L-circ elements (the top L-circ neuron is part of this pool). Simultaneously, the activity in adjacent RE areas (bottom RE neuron) is suppressed by axonal collateralization and dendro-dendritic synapses within the RE nucleus. The consequence would be the released activity of target L-circ neurons (bottom L-circ cell) and inhibition of weakly excited Th-cx neurons (bottom Th-cx neurons) in areas adjacent to the active focus. This hypothesis derived from a study on the activity of RE neurons during the natural waking-sleep cycle of chronically implanted cats (Steriade et al. 1986). The circuit was proposed in Steriade (1991) and was redrawn by E. G. Jones.

Here, I will focus on the impact exerted by synaptic activity, arising locally or in distant structures, on a major intrinsic property of thalamic neurons (the low-threshold spike, LTS) as well as on related oscillatory phenomena in these neurons, the network generated sleep spindles and the intrinsically generated clock-like (delta) rhythm. Other intrinsic properties of TC and RE neurons, their ionic bases, and the biophysical models of ionic currents, are discussed in two recent monographs (Destexhe and Sejnowski 2001; Steriade et al. 1997b).

THE LTS. The ability of thalamic neurons to display a paradoxical form of excitation resulting from their hyperpolarization was known since the late 1960s (Andersen and Andersson 1968; see also Maekawa and Purpura 1967), but systematic studies on the postinhibitory rebound and the discovery of the Ca²⁺-dependent low-threshold current (I_T) underlying this intrinsic neuronal property were only possible with the advent of slice studies (Jahnsen and Llinás 1984a,b; reviewed in Huguenard 1996; Llinás 1988). More recent studies reported that the LTS of TC neurons also contains a component mediated by a persistent Na⁺ current (I_Na(p)) (Parri and Crunelli 1998).

View larger version (35K): [in this window] [in a new window]   Fig. 10. Low-threshold spikes (LTSs) in cat thalamocortical (TC) neurons are graded in amplitude during spindle oscillation. A: ketamine-xylazine anesthesia. Intracellular recording from the thalamic ventrolateral (VL) nucleus. Fluctuations in time to peak and amplitude of LTS at the break of the threshold hyperpolarizing current pulse (0.8 nA, 0.1 s). Conditioning Vm is the Vm just before the end of the current pulse; amplitude of maximal depolarization was calculated from baseline Vm. B: barbiturate anesthesia. Simultaneous field potential from cortical area 4 and VL nucleus, together with intracellular from VL nucleus. Right: an expanded spindle sequence, further expanded below (). Modified from Timofeev et al. (2001a).

View larger version (28K): [in this window] [in a new window]   Fig. 11. Anterior thalamic (AT) nuclei of cat are devoid of afferences from the RE nucleus and, despite the fact that the intrinsic property of LTS is present in AT neurons, they do not display spindles because of absence of synaptic connections from the pacemaking RE nucleus. Left: the RE projections to various dorsal thalamic nuclei, as resulting from retrograde tracing experiments in cats. Heavy lines indicate prominent projections to intralaminar nuclei. Note absence of projections to AT nuclei. AD, AM, and AV: anterodorsal, anteromedial, and anteroventral nuclei; CA, caudate nucleus; CL-PC, central lateral and paracentral (rostral intralaminar) nuclei; CM-PF, center median-parafascicular (caudal intralaminar) nuclei; MD, mediodorsal nucleus; PARA, paraventricular nucleus; VA, ventroanterior nucleus; VL, ventrolateral nucleus; VM, ventromedial nucleus; VB, ventrobasal complex; RE, reticular nucleus; V3, third ventricle. Right: simultaneous recordings of field potentials (filtered for spindles, Sp., between 7 and 14 Hz) from CL and AV nuclei in cat. Unanesthetized cerveau isolé (collicular-transected) preparation. Abscissae indicate real time (hr, min, s). Data were obtained by applying each filtered EEG signal (see above CL trace, filtered EEG spindles, the 1st sequence corresponding to the 1 depicted below) to a full-wave rectifier, a voltage-controlled oscillator, and to a laboratory computer (see technical details in that paper). Note regularly recurring spindle sequences in CL nucleus and absence of spindles in the AV nucleus. Bottom: intracellular recording of an AT neuron, showing tonic firing at a relatively depolarized Vm (60 mV), LTSs crowned by spike bursts under steady hyperpolarization when the Vm reaches 72 mV, and recovery of tonic firing at 60 mV. Modified from Steriade et al. (1984) and Paré et al. (1987).

EFFECTS OF SYNAPTIC ACTIVITY IN ASCENDING AND CORTICOTHALAMIC PROJECTIONS ON TERMINATION AND WIDE SYNCHRONIZATION OF THALAMICALLY GENERATED SPINDLE OSCILLATIONS. Spindles (7-14 Hz) arise within the thalamus even after decortication and high brain stem transection (Morison and Bassett 1945). The RE nucleus is the pacemaker of spindle oscillations as demonstrated by the abolition of spindles in target thalamic nuclei and corresponding cortical areas after disconnection of TC neurons from the RE nucleus (Steriade et al. 1985) and the preservation of spindles in the RE nucleus disconnected from the remaining thalamus (Steriade et al. 1987). The different reasons explaining the failure to obtain spindles in the isolated RE nucleus of thalamic slices (Von Krosigk et al. 1993), among them the requirement of a larger and more intact collection of RE neurons than usually found in thalamic slices (see Steriade et al. 1993d) and the absence of some brain stem modulatory systems (Destexhe et al. 1984), are discussed elsewhere (Steriade et al. 1997a). Although this oscillation was recorded intracellularly within the thalamus, in the absence of cortex, both in vivo (Deschênes et al. 1984; Steriade and Deschênes 1984; Timofeev and Steriade 1996) and in vitro (Bal et al. 1995a,b; Von Krosigk et al. 1993), the neocortex contributes to the termination of individual spindle sequences, but on the other hand, it plays an important role in the synchronization of spindle sequences over widespread thalamic and cortical territories. Thus long-range projections in corticothalamic systems influence a thalamically generated oscillation, and although spindles arise in local intra-RE and RE-TC circuits, network activities originating in distant cortical areas are powerful enough to change the duration and synchronization patterns of this sleep oscillation. These data are discussed below.

View larger version (53K): [in this window] [in a new window]   Fig. 12. Cortical spindle sequences occur nearly simultaneously during natural sleep of humans and cats, but decortication disorganizes the widespread coherence of thalamic spindles. Top: illustrating natural sleep (human), spindles were recorded from 6 standard EEG derivations (indicated in the schematic at right, ) in a normal subject, during sleep stage 2. Cross-correlations of individual spindle sequences (n = 15) were calculated between C3A2 and each 1 of the other channels. Averaged correlations (cross) showed rhythmicity at 14 Hz and central peak values between 0.7 and 0.9. Bottom: spindles are simultaneously recorded from 7 leads in the thalamus of intact-cortex cat under barbiturate anesthesia. Note the virtual simultaneity of spindle sequences. After decortication (see scheme), recordings from virtually same thalamic sites show disorganization of spindle simultaneity. From Contreras et al. (1996a, 1997a).

CORTICAL SYNCHRONIZATION OF AN INTRINSIC (CLOCK-LIKE DELTA) THALAMIC OSCILLATION. The other thalamically generated oscillation is the clock-like delta rhythm (usually 2-4 Hz), due to the interplay between two currents, I_H and I_T, that are activated and de-inactivated, respectively, by membrane hyperpolarization (Curró Dossi et al. 1992; Leresche et al. 1990, 1991; McCormick and Pape 1990; Soltesz et al. 1991). This oscillation is modulated by different substances that act on purinergic and adrenergic receptors and up- or downregulate the H current (Pape 1996; Pape and Mager 1992; Pape and McCormick 1989; Yue and Huguenard 2001). Although intrinsic to TC neurons (Fig. 13A), this oscillation, which represents only one component of delta waves seen on the EEG during slow-wave sleep, is subject to influences arising in neocortex. Corticothalamic volleys synchronize TC neurons (Fig. 13B) by primarily exciting GABAergic RE neurons that fulfill two basic requirements: they set the V_m of TC neurons at the appropriate level of hyperpolarization for the appearance of the two currents (I_H and I_T) and they project to different dorsal thalamic nuclei, thus synchronizing not only nearby but also distant TC neurons (Steriade et al. 1991).

View larger version (25K): [in this window]