Electrophysiology

Gary S. Aston-Jones and George R. Siggins

INTRODUCTION

Neurons are cells specialized for the integration and propagation of electrical events. It is through such electrical activity that neurons communicate with each other as well as with muscles and other end organs. Therefore, an understanding of basic electrophysiology is fundamental to appreciating the function and dysfunctions of neurons, neural systems, and the brain.

The purpose of this chapter is to describe, for the nonelectrophysiologist, the methods used in animal studies to understand the electrical functioning of neurons in the central nervous system (CNS), particularly as related to drug actions and mental function and dysfunction. This chapter is divided into sections devoted to different methods, models, preparations, and concepts used in electrophysiology. These methods differ fundamentally in the level of analysis, and our survey will review them from subcellular (patch-clamping single-ion channels) to behavioral approaches (neuronal recordings in awake primates). Noninvasive electrophysiological techniques, such as electroencephalography (EEG) and event-related potential recordings, are discussed in In Vivo Structural Brain Assessment and Methodological Issues in Event-Related Brain Potential and Magnetic Field Studies. Methods of metabolic imaging to measure neuronal activity are discussed in Positron and Single Photon Emission Tomography: Principles and Applications in Psychopharmacology, Brain Imaging in Mood Disorders, Functional Brain-Imaging Studies in Schizophrenia, Neuroimaging Studies of Human Anxiety Disorders: Cutting Paths of Knowledge through the Field of Neurotic Phenomena, and Anatomic and Functional Brain Imaging in Alzheimer’s Disease).

Within each section we describe techniques, give specific examples of their application, and point out their strengths and limitations. Each section will also discuss how a method's particular procedures and level of analysis relate to neuropsychopharmacology. Substantial technical development in electrophysiology has arisen from studies in nonmammalian species (e.g., aplysia, squid). However, to remain most directly relevant to neuropsychopharmacology in the space available, we will consider only techniques used in mammalian preparations. Unfortunately, because of space constraints we are compelled to neglect several very important but lesser-used methods and models, such as noise analysis of ion channel activity (1), and various forms of electrophysiological analyses in intraocular grafts (2), brain "chunks," or in vitro perfused explants (see, e.g., ref. 3). The reader should refer to the cited references and the excellent "methods" book by Kettenmann and Grantyn (4) for details on these and other novel approaches.

The introductory descriptions of neurons (see Introduction to Preclinical Neuropsychopharmacology) provide a background for the following sections on specific electrophysiological techniques. Throughout this chapter, we will draw upon knowledge reviewed by Barondes (Basic Concepts and Techniques of Molecular Genetics), Watson and Cullinan (Cytology and Circuitry), Zigmond (A Critical Analysis of Neurochemical Methods for Monitoring Transmitter Dynamics in the Brain), and Roth and colleagues (Methodological Issues in Event-Related Brain Potential and Magnetic Field Studies). Readers are referred to those chapters for details. #1

At the cellular level, electrical activity of neurons consists of the movement of charges (ions) through neuronal surface membranes. The major charge carrying ions are sodium (Na+), potassium (K+), chloride (Cl-) and calcium (Ca2+). The surface membranes of neurons are primarily composed of lipids (resistive elements, in electrical terms) which do not allow ionic flow. Instead, these semipermeable membranes are spanned by large specialized protein aggregates that form pores or channels through the lipid membrane. There are specific channel protein assemblies (usually more than one) for each of the ionic charge carriers, as well as those for certain cations in general, that confer a semipermeable nature to the membrane. The ability of these channels to permit ion flow is determined by several factors, most prominently the electrical potential that exists across the membrane, the gradient of ions set up by membrane pumps, and the semipermeable nature of the channels, as well as by responses of receptors, guanosine triphosphate (GTP) binding proteins (termed G proteins), and second messengers to neurotransmitters and hormones. For more detail on these aspects of neuronal membrane and channel properties, (see Cholinergic Transduction, Signal Transduction Pathways for Catecholamine Receptors, and Serotonin Receptors: Signal Transduction Pathways); also see larger works by Siggins and Gruol (5), Hille (6) and Shepherd (7).

There is a long history of electrophysiological studies on artificial lipid bilayer membranes that were designed in large part to determine whether bioelectric events result from membrane ion pores (channels) or transmembrane ion carriers (active transport). The results of these studies were important for substantiating the view cited above of cellular-level events involved in bioelectric activity. Studies on pore-forming antibiotic models by Mueller and Rudin (reviewed in ref. 6) laid a strong biophysical foundation (such as the involvement of water molecules) for subsequent understanding of natural ion channels, and they were largely predictive of subsequent single-channel recordings of biological channels (see below). Later refinements of this method included insertion of natural ion channels prepared from a variety of cell types or organelles. In brief, the artificial bilayer membranes are usually made by forming a sheet of lipid (such as phosphatidylserine) across a partition with a small hole separating two aqueous compartments. A vesicle preparation (e.g., "liposomes") containing either the antibiotic protein (e.g., gramacidin), a fractionated membrane containing ion channels, or (in later studies) reconstituted vesicles with purified channels is added to one of the compartments (see ref. 8 for review); under the right conditions, the vesicles fuse with the lipid bilayer membrane and insert channels. Standard voltage-clamp recordings (see below) are then performed between the two compartments, and drugs and ion changes can be applied to either side of the artificial membrane.

Studies using natural membranes with artificially inserted foreign but natural receptor-channel complexes have given even more validity to the pore theory and provided another powerful model for testing pharmacological agents. The procedure here typically involves injection of a channel preparation directly into a large living cell (tolerant to insertion of large injection and recording pipettes) such as the Xenopus laevis oocyte or several types of cell lines. The channel source is usually either (a) vesicles prepared from fractionated channel-bearing membrane, (b) vesicles with reconstituted, purified channel (glyco)protein, or (c) channels newly synthesized by foreign DNA or RNA injected into the oocyte via large pipettes (). Of course, the more purified the protein or DNA/RNA, the more homogenous will be the channel population eventually inserted.

In terms of their great potential for dissection of the molecular mechanisms of pharmacological action, studies using these artificial, reconstituted or "cloned" channels are still in their infancy. However, there have already been too many fine examples of the use of the methods to adequately recount here. Notwithstanding, we would be remiss not to mention the elegant molecular and pharmacological studies of the gamma-aminobutyric acid (GABA) receptor–ionophore complex by Eric Barnard, Robert MacDonald, and co-workers (9), as well as of the various glutamate receptor-channel subtypes by the Dingledine and Heinemann groups (10, 11) (see also Excitatory Amino Acid Neurotransmission and GABA and Glycine). Moreover, recent studies of ethanol effects on brain GABA and glutamate receptor channels expressed in Xenopus oocytes (12) and cell lines (13) have provided considerable insight into these two major sites of alcohol action (14) and provide prime examples of the potential uses of these methods for the study of the molecular mechanisms of action of psychopharmacological agents. Future studies of this type, in combination with site-directed mutagenesis, will help delineate the molecular site(s) of drug actions on a wide range of neuronal receptors.

The discovery and development, by Sakmann and Neher (15), of the "patch-clamp" method for recording from single-ion channels provided decisive proof of the aqueous pore theory for the origin of neuronal excitability and led to a long string of seminal studies culminating in their Nobel prize. This method originally required fabricating and fire-polishing of specific types of glass micropipettes (with large tip diameters relative to the "sharp" pipettes used in traditional intracellular recording), so that the tips could form a high-resistance (gigohm) seal when pushed onto a cultured or acutely isolated cell. The gigohm seal (and new electronic breakthroughs) essentially allowed the high current gain, low noise amplification necessary for recording the small, brief currents (under voltage-clamp conditions) passing through single ionic channels. Imagine the excitement when Sakmann and Neher saw the now familiar "box-like" currents suggesting the abrupt opening and closing of channels (; see ref. 15). There were other surprises not predicted by the biophysical models, such as the rapid open-and-closed "flickering" and burst-like openings of the channels in many cases.

The patch-clamp method can be applied in at least four configurations (), giving the technique formidable adaptability for testing the molecular mechanisms of receptors and their associated ion channels and second messengers. Three of these configurations (cell-attached patch, inside-out patch, and outside-out patch) allow study of individual ion channels under different conditions. In the cell-attached configuration, after formation of the gigohm seal, recording of single channels is made without disruption of the cell membrane. In the other two single-channel preparations, the membrane patch is detached from the neuron after a gigohm seal is formed, and single-channel activity is recorded in isolation from the cell. In the inside-out configuration, the patch of membrane is gently pulled away from the cell, and the patch remains attached to the pipette with its cytoplasmic surface now exposed to the bathing solution. Preparation of the outside-out patch begins by making a whole-cell configuration (see below) whereby, after forming the gigohm seal, the membrane patch under the pipette is ruptured by applying a strong vacuum through the recording pipette. Then the pipette is gently pulled away from the cell, carrying a piece of membrane with it. The detached membrane seals over the pipette tip during this maneuver, in favorable cases forming a membrane patch in which the extracellular membrane surface is exposed to the bathing solution. The whole-cell configuration will be described below.

Once again, examples of the use of single-channel patch-clamp methods are far too numerous to itemize here. However, psychopharmacologists interested in the use of these methods in elucidating the pharmacology of NMDA, GABA, and opiate receptors may wish to consult the work of Barker and colleagues (16), MacDonald and colleagues (17), and North and co-workers (18). These papers contain abundant detail on the elegant methods used to record single channels and test the action of various psychopharmacological agents.

The whole-cell clamp, illustrated in , is a form of the cell-attached configuration that uses the same pipette type and gigohm seal method described above but that, by rupturing the membrane under the tip, allows recording of the "macroscopic" or summed currents flowing through all channels in the entire cellular membrane, rather than through a single channel. In this configuration, after the pipette is sealed to the membrane, another slightly stronger vacuum is applied to the pipette tip (via the tube attached to the pipette holder) to rupture the membrane under the tip without disrupting the gigohm seal or cell viability. During current-clamp recording, successful "break-in" is signaled by a negative shift in the recorded potential (to the cell resting membrane potential) and a large reduction in the input resistance of the system (now due only to the series resistance of the pipette and the input resistance of the total cell). In this configuration, the diffusable contents of the pipette then exchange over time with those of the cell.

Again, there are numerous examples in the literature over the last 10 years of whole-cell patch recording of neurons in isolated systems and an ever-increasing number of whole-cell patch-clamp studies of neurons in slice preparations (see below). From a pharmacological viewpoint, the whole-cell patch studies of the MacDonald (17) and North groups (18) on second messenger and G-protein mediation or regulation of GABA, opiate, catecholamine, and somatostatin effects deserve special reference. See the chapter by Foote and Aston-Jones for details on such studies in rat locus coeruleus neurons.

Over the last 30 years, a variety of nonelectrophysiological (and sometimes even noninvasive) methods have been explored to measure the membrane properties and ionic constituents of excitable cells. Purely optical recording methods (without dyes) were originally applied to isolated axons (e.g., from squid or crab) and revealed changes in both light scattering and membrane birefringence during stimulus-evoked action potentials. Later studies infused a potentiometric merocyanine dye intracellularly and used signal averaging methods to show that optical absorption or fluorescence changes closely followed the time course of the action potential (19), indicating that these were useful methods for studying membrane potential by relatively noninvasive methods. The development of newer, more effective voltagesensitive dyes offer pharmacologists the opportunity to follow membrane potential changes optically, without penetration or other disruption of the neuronal membrane other than extracellular treatment with the dye. The reader is referred to recent reviews (e.g., refs. 20 and 21) for details on the types and mechanisms of these membrane potential indicators.

Dyes can also be used to measure the intracellular and extracellular concentrations of certain free ions (or better, their activities). While there are now a variety of indicator dyes relatively selective for several different ions (including H+, Na+, and Cl-), by far the most research has been done with Ca2+-sensitive dyes (see ref. 21 ). Metallochromatic indicator dyes like arsenazo III were found useful for the measurement of cytosolic free Ca2+, as first tested in squid giant axons. The luminescent photoprotein aequorin also had its heyday as an indicator of intracellular free Ca2+. However, most of these methods involved the injection of the indicator into the cell, therefore requiring the study of larger neurons. The newer fluorescent probes (quin-2, fura-2, indo-1, and fluo-3) based on the Ca2+-chelator ethyleneglycol bis(aminoethyl ether)tetraacetate (EGTA) model (see ref. 21) also generally requires the use of fluorescent (ultraviolet or near-ultraviolet) illumination on relatively isolated neurons (e.g., neuronal cultures, very thin slices, or acutely isolated neurons) or isolated axon bundles. However, these newer indicators can now be loaded into neurons without injection or penetration by using their hydrolyzable esters such as acetoxymethyl (MA) ester. These esterified indicators can merely be applied extracellularly; the ester confers hydrophobicity, allowing the indicator to pass through the membrane into the cytoplasm, where the ester is removed by endogenous esterases, trapping the indicator inside. Furthermore, the recent development of scanning confocal microscopy (which can optically section a neuron without contamination by out-of-focus objects), in combination with these Ca2+ indicators, has made it possible to observe spatial or compartmental changes in intracellular free Ca2+ even in neurons within relatively thick preparations (21, 22).

However, as implied above, unless a confocal microscope is available (still a rather large expense), all of these methods require a certain degree of isolation of the neurons under study. Therefore, if the chosen model is an in vivo or thick-slice preparation, the use of ion-sensitive electrodes containing ion-exchange resins, although difficult to implement (see ref. 23 for details on electrode fabrication), can be of considerable advantage. These electrodes can be inserted blindly into thick brain slices, or even into brain regions in vivo, to record absolute values or changes in ion activities. These electrodes achieve their ionic selectivity by virtue of the resin which generates a current flow in the electrode in proportion to the concentration of a specific ionic species. K+ and Ca2+ ion activities are the most often measured. Whereas the ion-exchange resin method is best at measuring extracellular ionic activities (because the high resistance of the resin usually requires the use of rather large-tipped micropipettes), under the right conditions intracellular measures can also be obtained with this method (23).

Considerable information about neuronal and synaptic mechanisms and the effects of drugs on these mechanisms has been obtained with either the extracellular or intracellular application of ion-sensitive microelectrodes. For example, Lux, Heinemann, and colleagues (see ref. 23) have used these microelectrodes in various preparations to follow the extracellular K+ and Ca2+ levels with epileptiform activity or synaptic action via stimulation of afferent pathways.

As for the ion-sensitive fluorescent probes, fura-2 has been used to measure intracellular Ca2+ in innumerable studies. However, the ability of this indicator to measure compartmentally (spatially) distinct and time-dependent changes of intracellular calcium levels in several neuron types, with alteration of these changes by neurotransmitters, drugs (e.g., caffeine), and ion changes, is an especially exciting use of this method (21). This method also has great applicability for non-neuronal cells: Holliday and Gruol (24) recently used fura-2 imaging to show that the cytokine interleukin-1b dramatically increases intracellular Ca2+ levels in cortical astrocyte cultures and also enhances the increased Ca2+ evoked by the glutamate receptor agonist quisqualate.

The artificial membrane, reconstituted receptor/channel, and patch-clamp methods have allowed the use of isolated membrane components, isolated mRNA, foreign (non-neural) cells, and acutely isolated (enzymatically dissociated) or cultured neurons. As a result, there are several global advantages for these methods. First, these isolated preparations represent greatly simplified systems whose electrophysiological responses are not confounded by uncontrolled synaptic or hormonal inputs. Indeed, the total external and internal milieu of the cells and channels can be controlled by the experimenter, thus providing unprecedented capabilities for testing the influence of countless influences and variables. For example, the researcher can adjust the ionic compositions on either side of the membrane or channel so that the voltage differential is exactly opposite (i.e., positive on the inside surface) to that in normal cells; this facilitates greater isolation of membrane conductances involved in certain membrane and synaptic functions.

Another advantage of these isolated membranes and systems is their suitability for use of molecular and genetic techniques (described by Barondes). Thus, as implied above, one can easily test the effects of minute modifications of the molecular structure of receptors and their associated channels, as well as the individual components (subunits) of G proteins and second messenger systems (described by Zigmond). As a result, researchers can now answer many physiological questions at the molecular level. For example, one can determine exactly where [i.e., at what codon(s) in the RNA or amino acid(s) in the protein] in the channel molecule or subunit that an antagonist or drug acts, or at what point protein phosphorylation (e.g., via protein kinases) can modulate receptor function. Thus, using combined electrophysiology and molecular biology, for the first time we can begin to make a definitive connection between structure and function at the molecular level.

The several disadvantages of these models are principally derived from the isolation process itself. Hence, these isolated preparations, because they lack normal synaptic connections with other neurons, may be functioning under conditions greatly different than normally seen in the living organism, and thus may provide answers not relevant to the "real world." In addition, one is never certain that the "extracellular" and "intracellular" media used do not constitute a totally artificial environment that would not be relevant to a living neuron in vivo. The lack of normally circulating agents such as steroids, hormones, plasma proteins, and other colloidal substances could lead to drastic changes in the function of the molecules and channels under study. Finally, this inability to examine receptor and channel function in the context of an intact functioning system may cloud the application of findings derived from these models to the behavior of living organisms.

The patch-clamp method is nearly ideal for the study of the mechanisms of drug action at the single-channel level. Some of the advantages of this method include the following: (a) Ions, toxins, neurochemicals, and other pharmacological agents can be applied easily (either in the bath or in the pipette), in defined concentrations, to both the external and internal surfaces of the membrane; (b) several chemicals or ions can be tested on one patch (or channel) either together or in sequence; and (c) several different drug or ion concentrations can be tested on the same membrane patch, thus facilitating generation of dose–response curves. In addition, the gating mechanism(s) behind the opening of channels can be pharmacologically tested more easily in a patch configuration.

The single-channel, cell-attached, and inside-out patch-clamp methods are particularly well-suited for the study of second messenger systems, and particularly for those systems (such as G-protein-mediated events) that are "membrane-delimited" (see, e.g., ref. 25). Thus, in the cell-attached configuration, if a transmitter or other receptor regulator (first messenger) applied externally in the bathing solution—but not when applied from within the pipette—alters single-channel function (e.g., opening or closing of the channel recorded), the receptor must be remote from the recorded channel and mediation of the event by a diffusible second messenger is suspected. If the transmitter or regulator alters channel function even when applied via the recording pipette to an inside-out patch (where a soluble second messenger would diffuse away from the system), a membrane-delimited system is a likely candidate (see ref. 18) for an elegant example of this approach in opiate responses of locus coeruleus neurons.

A major disadvantage of the single-channel patch-clamp method is the necessity to use cultured or acutely isolated cells; such models seem to allow formation of better gigohm seals, probably because the relative lack of overlying glia or other supporting cells facilitates close apposition and seal of the pipette to the neuronal membrane. In addition, compared to cells lying within a slice preparation, the thin or nonexistent level of tissue and/or fluid overlying the recorded neuron reduces the capacitance in the recording system and allows better recording characteristics for the small-current signals generated by single channels. Therefore, most single-channel studies to date have been limited to cultured or isolated neurons. However, new refinements have allowed whole-cell patch (and some single-channel) recording in brain slices, and even in vivo (see below).

In part because of the free exchange of contents between the cell and the pipette, the whole-cell method has several additional advantages and disadvantages. First, unlike the single-channel methods, the whole-cell method can be used routinely in brain-slice preparations (see below and refs. 26 and 29) as well as in vivo. Also, with this method one can adjust the ionic balance and contents of the cell merely by placing the appropriate buffers and salts in the pipette solution. Furthermore, second messengers and drugs affecting them can be placed directly in the pipette solution for diffusion into the cell. For selective blockade of certain channel types (e.g., K+-selective channels), toxins (e.g., tetraethylammonium) or appropriate ions (e.g., Cs+) can also be placed in the pipette. And finally, dyes such as lucifer yellow or biocytin can be placed in the pipettes for quicker and more complete filling of the neurons than with the standard intracellular "sharp" pipettes.

The major disadvantages of the whole-cell method usually stem from the same properties that confer the advantages. Thus, because it is usually not possible to know the exact ionic or second messenger composition of the normal resting cell, there is always the risk that essential cell constituents (e.g., cyclic AMP) will diffuse out of the cell into the pipette (which usually constitutes a much larger volume than that of the cell). It is thought that such diffusion accounts for the slow "run-down" in some cells of certain currents such as the L-type Ca2+ current. Attempting to replace such lost constituents (e.g., cyclic AMP, ATP, GTP) by adding them to the pipette solution can help alleviate some of these problems. The use of a nystatin "barrier" in the pipette tip (see ref. 27), which allows passage of small monovalent ions (e.g., for current injection) but prevents the diffusion of most divalent and large nonionic constituents (e.g., proteinaceous buffers and components of phosphorylation systems) from the cell, is another procedure for reducing these problems.

As alluded to above, there are many advantages of the various imaging approaches, the most important being the ability to measure ionic or potential changes via relatively noninvasive methods. However, it should be remembered that most of these methods require dyes and/or light exposure of some sort that can distinctly alter cell function over time. In addition, most of these methods require relatively isolated or thin preparations, so that microscopic observation can be performed.

Implicit in these studies of ion channels and their related receptors is the idea that a substantial molecular and electrophysiological understanding of the function of each structural element or subelement will allow the rational development of more effective and selective pharmacological agents and, ultimately, better therapeutic drugs. For example, understanding what part (amino acid residues, phosphorylated region, etc.) of the N-methyl-D-aspartate (NMDA) receptor-channel complex is affected by alcohol could lead to the development of a drug to blunt the alcohol antagonism of NMDA receptor function. Because the anti-NMDA effect of ethanol is thought to account for many aspects of alcohol intoxication (14), this drug could be the "silver bullet," sought by alcohol researchers for years, for rapidly reversing ethanolinduced intoxication. In addition, all of the single- channel methods have the capability of being applied to human samples (e.g., slices or cultures from biopsies), both from normal tissue and from diseased brains. The single-channel (and molecular) data from these two sources could then be compared for indications of the structural source of the abnormalities or malfunction; with this sort of knowledge base, more rational drug development or intervention procedures would lead to new, more effective treatments for the disease.

The above discussion concerns parts of cells or individual channels. However, there are many motives for electrophysiological study of intact neurons in isolation or semi-isolation. Activity in a single channel typically has little influence on the overall electrical activity of a postsynaptic cell. A neuron usually receives synapses from thousands of other neurons, and many of those inputs may be active at approximately the same time (synchrony). The way in which this postsynaptic cell integrates these numerous inputs and alters its own electrical activity is the fundamental basis of neuronal integration and information processing in the brain. The researcher may need to understand the integration between different parts of the neuron, or how an action at one locus (e.g., an ion channel) can lead to transmission of information to another locus (e.g., by a second messenger), without interference from outside events. To accomplish this understanding, the electrophysiologist uses the numerous tools and models that have been developed to study intact neurons.

The most completely isolated yet relatively intact cellular preparation is the acutely isolated neuron model. Such isolated neurons are prepared by mild enzymatic digestion and gentle agitation and trituration of brain slices (28). Neurons can be isolated from most brain regions; when prepared correctly, they have most of the properties of neurons in culture and slice preparations, but are free of the glial investments and debris that prevent formation of gigohm seals in patch-clamp studies. Although the distal dendrites can be truncated by this procedure (with possible loss of important channels or receptors), the reduced dendritic length greatly reduces space-clamp difficulties in voltage-clamp recordings (see below).

Chronically cultured neurons and cell lines (such as those also used for single-channel studies) have been a major cellular source for electrophysiological studies (see, e.g., refs. 9 and 16), as have single neurons within brain slices (see below for methods of preparation). Acutely isolated or cultured neurons have the advantage of direct microscopic visualization and manipulation; however, neurons in brain slices (described later) also can be visualized now using new optical methods such as Nomarski microscopy combined with sensitive video imaging (29, 30). Another relatively new model, the slice culture, has been developed by the Gähwiler group (31). Slice cultures are prepared by long-term incubation of brain slices in a rotating drum ("roller tubes"), where the slices develop into a monolayer of neurons that retain their local cytoarchitectonic relationships and that can be individually manipulated under microscopic visualization.

Although there are usually synaptic connections among neurons of all these models except acutely isolated neurons and some cell lines, for purely cellular studies the neurons can be chemically isolated by blocking synaptic transmission with Na+ or Ca2+ channel antagonists such as tetrodotoxin or low concentrations of Ca2+ with high concentrations (8–12 mM) of Mg2+. The researcher can then study the neurons individually with little influence from remote synaptic effects. Another method of electrical isolation is to bathe the slice in isotonic sucrose (without salts) and then apply a small amount of (conductive) Ringer's solution via a small-bore pipette to the neuronal region under study; this is a cellular application of the old "sucrose gap" method. Methods such as these should always be used as controls in any study testing the effects of exogenously applied drugs or transmitters: if the isolation procedure blocks the drug effects, then indirect (remote) effects are suspected.

For details on the preparation and use of these and other cellular models, we direct the reader to two excellent books on methods, edited by Shahar et al. (32) and by Kettenmann and Grantyn (4).

Both of these methods involve recording membrane potential with an electrode inserted into, or in contact with, the intracellular compartment. Such intracellular recordings are required for several types of analyses—including, for example, observation of pacemaker activity or subthreshold synaptic effects.

Current-clamp is a method of intracellular recording involving measurement of the voltage difference across the cellular membrane while injecting constant positive or negative current (as "square" d.c. pulses) into the cell. Voltage recording without current injection or other perturbation will usually tell the researcher only what the membrane potential is (usually around -60 to -80 mV in resting neurons). However, by injecting repetitive constant current pulses (or steps) into the cell and using appropriate "bridge" methods to balance out the resistive influence of the recording micropipette, the electrophysiologist can obtain from the voltage response a relative measure of the resistance (or, inversely, of conductance: g) of the membrane. Ohm's law (E = IR) can be applied here to obtain a simple relationship between the current injected (I), the voltage recorded (E), and the "input resistance" (R) of the membrane. If a drug or transmitter is then applied to the cell, a change in the size of the voltage response to the current pulse indicates a change in ionic conductance (g = 1/R). By incrementally varying the amplitudes of the current steps over a wide range (typically from 0.1 to 1 nA in mammalian CNS neurons), a family of voltage responses can be obtained for construction of a voltage–current (V–I) curve (where voltage is typically plotted as a function of injected current; see . This curve reveals much about the "macroscopic" currents (that is, the aggregate currents flowing through many ionic channels) passing through the neuronal membrane at different membrane potentials. Any drug treatment that alters ionic conductance will also alter the slope and shape of the V–I curve. Thus, a reduction in the slope of the V–I curve indicates increased ionic conductance, whereas a steeper slope indicates decreased conductance.

In practice, current-clamp recording is usually performed by inserting a single sharp micropipette into a neuron while recording voltage and injecting current through the same pipette. Penetration of the cell is signaled by an abrupt transition to a large negative voltage (about -70 mV), accompanied by an increase in input resistance (as typically reflected in the voltage deflection produced by a current pulse). After successful settling ("sealing") of the pipette into the membrane, control V–I curves and synaptic activations can be generated (usually nowadays by sophisticated computer methods); drug administration by superfusion or pipette application (see below) is then followed by repeated V–I and synaptic measures for statistical comparison to the control measures. Reversal of any drug effects by washout with the vehicle (artificial cerebrospinal fluid) alone assures the researcher that any changes are not merely the result of a rundown (e.g., slow death) of the cell or a slowly improving penetration "seal." The usual measures taken in current clamp include: resting membrane potential, input resistance, I–V curves, and voltage responses (excitatory postsynaptic potentials or inhibitory postsynaptic potentials) to activation of inhibitory or excitatory synaptic afferents. In addition, much information about membrane and drug properties can be obtained from the rebound voltage responses (so-called "anodal break" depolarizations, due to activation of several possible currents) immediately following strong hyperpolarizing current steps, or from the prolonged hyperpolarizations [afterhyperpolarizations (AHPs) due to Ca2+-dependent K+ conductances] following the burst of spikes evoked by strong depolarizing current steps. Many neurotransmitters have been shown to potently alter the latter measure (see ref. 5).

In voltage-clamp recording, the investigator measures the current required to hold a neuron at a constant voltage; voltage "commands" are typically applied as steps or as a steady voltage (termed "holding potential"; see ). A major advantage of this method over current-clamp recording is that the investigator can directly measure ionic currents, and not just the reflection of currents passing through channels (i.e., voltage response, which may vary with other changes). In addition, by using abrupt voltage-command jumps, one can measure the changes (and their kinetics) in the currents flowing at the original potential as the channels slowly adjust (open or close) to the new potential. Thus, with this method, voltage-dependent and time-dependent ionic conductances, and the effects of drugs on these conductances, can be directly monitored.

In practice, voltage-clamp recording of mammalian CNS neurons in vitro involves the same sorts of micropipettes and other methods as those used in current-clamp recording (most high-performance commercial headstage amplifiers allow switching between the two modes). In fact, one usually penetrates a neuron with the micropipette under current-clamp mode. Then after the cell stabilizes, a series of adjustments of the recording characteristics can allow stable switching to voltage clamp.

Until recently, most research on brain-slice preparations (described below) has applied either extracellular recording or standard intracellular recording with "sharp" pipettes. Because these pipettes usually have the unfortunate feature of high electrical resistance (because of the need for fine tips to penetrate small cells with little injury), recording properties were not optimal. Because early voltage-clamp methods generally required insertion of two pipettes into large (invertebrate) cells, early studies of brain-slice neurons typically employed "current-clamp" methods (described above). Later, a novel "switch-clamp" method was developed whereby a single pipette is used to switch rapidly (at about 2–7 kHz) back and forth between current injection and voltage measurement modes (see ref. 33). As with the two-electrode voltage-clamp method, the measured voltage is compared to a desired voltage (the "holding" or "command" potential) and the difference is used during the next brief current-injection cycle to inject the appropriate current to bring the membrane to the desired voltage. However, the switch-clamp method has several disadvantages, including the need to keep pipette resistance and capacitance to a minimum (usually 50–70 MW or less) so that the cell and not the pipette potential is clamped. Another disadvantage is the difficulty in clamping remote membrane areas (e.g., in the dendrites and other processes), which can lead to "space-clamp" artifacts.

Some of these disadvantages can be minimized with new methods recently devised to perform whole-cell patch-clamping in slice preparations (so-called "patch-slice" methods; see above for whole-cell recording techniques). Prior to about 1988 this feat was considered impossible because of the postulated "fouling" of the pipette tip with cellular debris arising from the tissue in the slice overlying the neuron to be recorded. One method (29) got around this problem by observing the slice under Nomarski optics and "cleaning away" the cells and debris overlying the target neuron with puffs of saline applied from a large-bore pipette. However, a more straightforward method (26) entailed the slow, "blind" penetration of the slice with a non-fire-polished patch pipette until the telltale small increase in resistance occurred. Then a gigohm seal was formed exactly as in the patch-clamp methods described above. With the single-pipette, whole-cell method, the pipette tip resistance is low (3–8 MW) and continuous-mode clamping or a faster (5–10 kHz) form of switch-clamping can be performed, with consequent reduction in artifactual clamping of the pipette tip potential. In addition, with whole-cell clamp it is much easier to inject ions and toxins into the cell to eliminate major sources of large conductances, thus helping to reduce the remote or space-clamp problem.

An example of the use of voltage clamp is provided by analysis of the M-current. The M-current is a voltage-dependent conductance that is only active (but persistently so) at membrane potentials slightly depolarized from resting potential (-60 to -10 mV); if one holds the membrane potential at about -40 mV (where the M-current is permanently "on") and then applies a hyperpolarizing step towards -60 mV or so, a slowly changing inward current develops (actually, a reduction of an outward K+ current; the M-current "relaxation") that signals the slow closing of M-channels caused by bringing the membrane potential out of the range of M-current activation. shows the enhancing effects of the opioid peptide dynorphin A on this current in hippocampus (34). Without voltage-clamp methods, this unusual but important conductance (or channel type), and its alteration by transmitters, probably would not have been discovered (reviewed in ref. 5).

These cellular methods have added powerful tools to the armamentarium of the neuropsychopharmacologist. First, because they use isolated preparations or relatively simple systems, they allow study of neuronal properties and cellular integrative mechanisms without the confounds of outside influence (e.g., from other neurons or hormones, etc.). Second, the ability to use known concentrations of drugs (with application by perfusion) allows the researcher to attempt to (i) mimic the concentrations of neurotransmitters released synaptically, (ii) adjust applied drug concentrations to those known to exist in blood or brains of humans with systemically administered drugs, or (iii) use drug concentrations within the known range for selective action at target receptors. For example, opiates or alcohol can be perfused onto neurons over a range of concentrations known to cause intoxication in humans. In addition, unlike single-channel studies where a large sample of channels must be studied one at a time, these cellular methods allow a more rapid survey of the effects of drugs on many types of conductances. Finally, synaptic events are also available for study with these methods (provided chemical isolation is not used); even in the acutely isolated neuron, intact synaptic boutons can remain attached and exhibit spontaneous transmitter release (35).

Obviously, some of the same advantageous features of the isolated preparations also confer disadvantages. Thus, the lack of normal connections between sets of functioning neurons lessens the utility of these models for the study of hodology, neuronal networks, or the normal interactions between brain regions. Still, some of these confounds can be overcome to a degree by the proper preparation of brain slices or brain "chunks": cutting large slices of tissue in the proper orientation (e.g., nucleus accumbens with attached A10 area or cortex). Still, the fact that the cells are always maintained in a somewhat artificial environment—or, in the case of cultures, may have abnormal developmental properties—are problems not so easily circumvented in these preparations.

These cellular-level approaches are crucial to drug development. As our knowledge of the structure, function, and enormous specificity of receptor proteins increases, we are better able to design drugs that have specific targeted effects at the cellular level on identified neurons; for example, the design of neuroleptics that interfere with a particular subset of dopamine receptors on cortical neurons is one currently feasible goal. Such specific drug design demands neuropharmacological testing at the cellular level using the techniques described above.

A considerable advantage of these methods for clinical work is that in many cases human samples can be directly explored. For example, it is feasible to extract mRNA of a receptor protein from diseased human brain, express it in a test cellular system (e.g., oocyte), and determine its functional status ().

Overall, these cellular level methods have enormous potential for the future of neuropsychopharmacology by virtue of the fact that they interface extremely well with molecular biological manipulations. Thus, new developments in the genetics or molecular biology of mental disease can find direct application using these cellular electrophysiological techniques.

McIlwain first discovered that if a brain is rapidly removed from the skull and rapidly cooled, a "slice" or slab of brain tissue could be cut that would survive for many hours in the proper organ culture environment in vitro. Whereas originally developed for studies of metabolism, neurochemistry, and neurotransmitter release from the cerebral cortex (see A Critical Analysis of Neurochemical Methods for Monitoring Transmitter Dynamics in the Brain), this technique has now been widely applied to many brain regions for the study of electrophysiology of neurons and local brain circuits.

With minor variations for different brain areas, the method for removal and incubation of a brain slice is straightforward. The techniques for slices of various types from various brain areas are described in detail elsewhere (reviewed in ref. 4). The brain is rapidly removed from the skull and placed in ice-cold saline or artificial cerebrospinal fluid (ACSF) saturated with carbogen (95% O2, 5% CO2) gas. Using either a vibrating microtome or a tissue chopper, a thin (usually 100–400 mM) slice of fresh brain is cut through the area of interest. This slice is rapidly placed in cold carbongenated ACSF and (perhaps later) transferred to a slice recording chamber containing ACSF. Recording is usually performed after 1–2 hr of incubation to allow recovery from the insult of the surgery. Although recording chambers vary somewhat in their design, they all have the ability to continuously perfuse the slice with fresh ACSF and to add drugs to the ACSF perfusate at known concentrations (as well as by the pipette methods described in the section on in vivo studies). The two most common slice techniques are (i) submerged slices, in which the tissue is fully submerged in the bath with continuous superfusion, or (ii) "interface" slices, in which the bath fluid extends just to the upper surface of the tissue (to interface with a layer of warm, moist carbogen gas). While in the recording chamber, the temperature is best maintained in the physiological range so that normal processes may be studied. With practice, the brain-slice preparation in most cases will remain viable and yield excellent electrophysiological recordings as well as neurochemical measurements for 12 hr or more.

Finally, we would be remiss to omit the isolated brain in vitro (36). Seeming like science fiction, in this method the entire brain (or brainstem plus cerebellum) is removed and kept alive in an incubation environment where various electrophysiological experiments can be performed. This exciting preparation has many advantages of the slice in terms of recording stability and ease of intracellular recordings, yet is nearly fully intact like the in vivo brain. While technically difficult and not yet used extensively for pharmacology questions, this approach holds great promise for future neuropharmacology studies. A related but technically more feasible preparation is the isolated brainstem–spinal cord, which has been used to great advantage in studies of the physiology of respiration (37).

The brain-slice technique has been widely used over the last two decades, so that slices of most brain areas have been studied. One of the first, and by far the most widely studied, of these preparations has been the hippocampal slice. This preparation has been a key factor in working out mechanisms underlying long-term potentiation, postsynaptic effects of various transmitters at the membrane level, and actions of a variety of drugs. Although far too numerous to list in full, the reader is referred to work by Madison and Nicoll (38) and by Siggins and colleagues (34, 39) for specific examples of the utility of this preparation.

Another preparation that has been used to great advantage in understanding the cellular effects of opiates is the locus coeruleus (LC) slice. The extensive studies of opiates in LC slices by Aghajanian and colleagues and by North, Williams, and co-worker are described in more detail in Foote and Aston-Jones).

The brain-slice method allows the repeatable application of known concentrations of drugs to the cells being studied, an important advantage in drawing conclusions concerning receptor identity and postsynaptic mechanisms. A second important advantage compared to most in vivo methods is the relative ease of obtaining long-term, stable intracellular recordings without anesthetics or immobilizing agents, so that the effects of drugs on membrane properties of identified neurons can be directly ascertained. An equally important advantage of the brain slice for electrophysiological studies is that the local circuits and cytoarchitecture of the tissue are relatively intact. This allows straightforward identification of the neuron being studied by visualization of its position with respect to known landmarks and other characteristics (e.g., the LC is clearly evident in the slice as a translucent group of cells adjacent to the fourth ventricle). This contrasts with studies in cultured neurons, where the neurochemical or nuclear identity of the neuron under study may be difficult or impossible to determine. Finally, the relatively intact local anatomy of the slice preparation also allows one to study synaptic responses of brain neurons as in some culture preparations (see above). Finally, the physiology and pharmacology of neurons can be studied in semi-isolation from the confounds of the ongoing behaviors of a freely moving animal.

Perhaps the most significant disadvantage here again is that the slice is a relatively isolated preparation, and neurons in the slice lack many normal afferent inputs and efferent targets. Of course, neurons in the slice are also isolated from any circulating influences such as hormones or steroids. Thus, properties observed in slice studies must always be considered with the caveat that results may reflect the artificial nature of the preparation and may differ from those obtained in the intact organism. Similarly, the slice is, of necessity, situated in an artificial environment rather than the natural and more complex milieu of the brain. The properties of neurons observed vary widely with minor changes in the slice environment, so that results may be heavily biased by the particular experimental conditions employed in an individual lab (for example, depending upon whether interface or submersion slice chambers are used). Also, by being isolated from the behaving organism, neurons within the slice are not amenable to study in the intact, functioning circuits in which they normally reside. Responses of neurons to transmitters are often best revealed when the neuron is challenged by other afferents, which may be lacking in the slice. Similarly, one cannot use this preparation to test the role of a set of neurons in a particular circuit function or behavior. Thus, slice studies are, of necessity, limited to the cellular level of analysis. However, when used in combination with studies in the intact organisms (reviewed next), experiments in the brain slice provide a powerful adjunctive analysis of an important range of phenomena.

The same advantages for neuropsychopharmacology listed above for cellular techniques apply to approaches using brain slices, because these techniques also allow a cellular-level examination of function. However, slices have the additional advantage that neuronal structure is better preserved than in isolated or culture situations, so that drug development can be carried out more easily at identified neurons and synapses.

Slices can also take advantage of animal models of disease to examine underlying cellular changes in identified neurons and synapses. One example here are models of drug abuse (using chronic drug administration) where slice studies have provided important insights into underlying neuronal changes (e.g., see Electrophysiological Properties of Midbrain Dopamine Neurons, Dopamine Autoreceptor Signal Transduction and Regulation, and Mesocorticolimbic Dopaminergic Neurons: Functional and Regulatory Roles). Other animal models (e.g., transgenic mice) could also be profitably explored using slice methods.

The intact, functioning brain is readily explored with microelectrodes in anesthetized animals. In this approach, the animal is anesthetized, most commonly with a barbiturate, urethane, chloralose, or halothane. The animal is then placed in a stereotaxic instrument which positions the skull in an exact position and orientation with respect to submillimeter scales in three dimensions on the instrument. By positioning the microelectrode tip at a desired coordinate along these scales, determined by reference to a stereotaxic atlas of the brain of that species, any site within the brain can be found and cellular activity recorded. X-ray or magnetic resonance imaging methods may also be used for this purpose in human studies.

In these experiments, impulse activity of neurons is typically recorded extracellularly, in contrast to the intracellular recordings discussed above. In extracellular recordings, the tip of a microelectrode (typically 1–10 mm in diameter) is positioned immediately adjacent to, but outside of, a neuron. When in close proximity to the neuron, current fields generated by action potentials in that cell are detected by the microelectrode as small voltage deflections (typically 0.1–1 mV).

There are many experimental applications of in vivo single-cell electrophysiology. Below we briefly describe three: iontophoresis and local drug application, stimulation recording, and antidromic activation.

In neuropharmacology experiments it is often useful to study the direct effects of neurotransmitter agents or drugs on neurons in the intact brain to mimic or alter responses to synaptically released transmitters. The ability of an exogenously applied agent to mimic the actions of the endogenously released transmitter is one of the cardinal criteria for establishing the identity of a neurotransmitter at a particular synapse (5). Also, such direct application of drugs obviates interpretive problems of systemic drug application, where direct effects might be confounded with indirect effects mediated by the drug acting at multiple sites in the CNS or periphery.

Curtis first used the iontophoretic technique, showing that charged drug molecules in a solution would be carried out of the tip of a micropipette by electrical current flow of the same polarity as the charge on the drug ion. Thus, by passing current through a glass micropipette, one can apply drug into the local area of the neuron being simultaneously recorded by another, adjacent pipette. This technique, denoted initially as microelectrophoresis but later as iontophoresis, was further developed in the CNS by Krnjevic and Salmoiraghi and colleagues (see ref. 5 for review). In brief, a multibarrel glass micropipette is manufactured so that 5–7 tips are adjacent to one another. A single micropipette is used for recording neural impulses extracellularly; this may be one barrel of the multibarrel pipette assembly, or (better) it may be an adjacent pipette affixed (glued) to the iontophoretic multibarrel electrode so as to protrude 10–20 mm, as illustrated in The iontophoretic barrels are filled with drug or salt solutions. It is important that one barrel be filled with NaCl, so that current of opposite polarity to that being passed through a drug barrel can be applied at the same time as the drug ion, to neutralize stray currents and minimize artifacts. Once stable impulse activity from a neuron is recorded, current (usually 5–200 nA) of appropriate polarity is then applied to the barrel containing the drug of interest and the resulting effect on neural activity is monitored, usually using a ratemeter-type recording of firing rate (). When the drug is not being applied (during control conditions), a "backing" or "holding" current of appropriate intensity (usually 5–20 nA) and polarity to attract the drug ion is applied to prevent unwanted leakage of drug from the tip. It is most advantageous to use an iontophoretic device designed to apply the drugs at precise intervals and automate the control procedures.

An important variant of iontophoresis is local application of drugs from micropipettes by pressure (denoted micropressure application). In this method, a multibarrel micropipette similar to the iontophoretic assembly is employed, but controlled automated pneumatic pressure instead of electrical current is used to eject drug from the tip. This method is often necessary to apply large or uncharged molecules (e.g., large peptides) that do not readily move with iontophoresis. It is possible to configure the pipette so that both iontophoresis and micropressure techniques can be used from the same barrel. If similar results are obtained with both methods of local drug delivery, it is less likely that the results are due to artifacts associated with either technique alone. See ref. 40 and Physiological and Anatomical Determinants of Locus Coeruleus Discharge: Behavior and Clinical Implications for recent examples of this combined method.

Local micropressure application has several advantages over iontophoresis. One of the major drawbacks of iontophoresis is that one does not know the concentration of drug applied. This is because the drug is carried by current, and for most drug solutions the physiochemical properties determining the relative transport of drug molecules in an electrical field in the micropipette glass are not known. It is possible that very high concentrations are ejected even with low currents (for easily ionized drugs), while even high currents may eject very little of another, poorly ionizable drug. There is also considerable release variability across micropipettes. In contrast, with local pressure application the solution ejected is the same concentration as that in the pipette. While the absolute concentration at the recorded cell is uncertain due to diffusion and dilution in the extracellular milieu, at least the highest possible concentration is known. This is important, because the receptor specificity of drugs are dependent upon their use within a certain concentration range. By increasing the volume of solution ejected, pressure application may also allow a larger area of tissue to be infused than with iontophoresis. For this reason, pressure is typically the method of choice when trying to locally antagonize synaptically mediated events that may reflect inputs onto distal or remote dendrites of the neuron being recorded. However, there are caveats with this method, such as artifacts due to pressure (movement), pH, or osmolarity changes; iontophoresis usually allows a greater range of drugs to be applied to the same neurons, and it is often associated with more stable and successful recording.

It is important for both microiontophoresis and micropressure application of drugs that the proper controls for current, pH, and volume effects be conducted, and that drug application follows a regularly timed protocol to minimize "warm-up" effects and possible experimenter bias (see ref. 5 for review).

This is the simple but requisite procedure for discerning the functional effect of an afferent input to a neuron. There are two methods available for this purpose: (i) the classical approach of electrically stimulating the afferents while recording the target neuron and (ii) a more recent method (especially important for studies in such complex tissues as brain) using local chemical (instead of electrical) stimulation to activate the input source.

With the former method, pulses of electrical stimulation are applied to a stimulating electrode to activate neurons that project to the area where a target cell is recorded. Extracellular recordings are typically used to measure the functional effect of the input. Most commonly, responses are measured in displays called peri-stimulus time histograms (PSTHs), where neural activity recorded for many successive stimulus trials is accumulated, synchronized with the stimulus presentation (). By accumulating activity in such a histogram, even relatively weak responses can be revealed due to the summation over many trials. This type of analysis allows quantitation of response magnitude, onset latency, and duration. These parameters can then be compared before and after drug administration to determine, for example, the effect of a particular receptor antagonist on the response to activation of an input and thereby help determine the likely transmitter candidate in that afferent. The reader is referred to the review by Ranck (41) for a detailed treatment of factors determining types of neural elements activated with different parameters of electrical stimulation.

A major drawback of electrical stimulation is that both cell bodies in the area of the stimulating electrode, and fibers of passage derived from cells located in other areas, will be activated by the stimuli applied. Therefore, the origin of the responses obtained is uncertain. This problem is surmounted by using local chemical stimulation, which activates the input neurons by infusion of a neural activator such as glutamate or one of its analogues into the area of the cell bodies or dendrites. Because stimulation by this method relies on receptor activation, and receptors are thought to reside only on somata and dendrites of neurons, this approach does not activate passing axons that originate from neurons elsewhere. However, while the origin of responses are better identified with this method, the temporally imprecise activation by chemical microinfusion does not allow accurate determination of response latencies. A second important limitation of chemical stimulation is that neurons can be inactivated by stimulating chemicals such as glutamate, if too high a concentration is applied. That is, too much of a chemical activator (e.g., glutamate) can depolarize neurons into a state of depolarization block, where the neuron is maintained in a depolarized state and thus cannot generate action potentials due to persistent inactivation of Na+ channels. Thus, what is thought to be stimulation can actually inhibit neurons of interest. A procedure that minimizes such concerns is to apply a range of concentrations of the chemical activator and examine the corresponding dose–response curve in the target cell. This should reveal the minimum dose for obtaining an effect, which with glutamate is presumably due to excitation of neurons near the site of infusion. Another way around this potential problem is to record the response of neurons in the infusion site during application of the chemical stimulant. This not only allows direct confirmation of the effect of the infused agent on those cells, but also gives the time that activity in the local neurons is affected; this can greatly improve the temporal accuracy of such stimulation–response studies. Without such procedures, results obtained with local chemical stimulation are difficult to interpret.

A similar method is used in combination with stimulation-recording experiments to test the involvement of a particular brain region as a circuit element mediating an evoked response. Instead of infusing an excitatory neurochemical, however, the solution microinfused is one that inactivates or synaptically decouples local neurons. As illustrated in , a composite recording/infusion pipette can be used so that neuronal recordings help localize the area desired for infusion as well as verify the effect of the infusion on neurons at the infusion site. One approach is to locally infuse a local anesthetic (e.g., lidocaine) to block local activity while conducting the stimulation-recording experiment (42). However, anesthetics block impulse conduction in passing fibers as well as in local somata, so that this approach does not test the role of local neurons as a possible relay in the circuit response being examined. Alternatively, the infusion solution contains either (i) a strongly inhibitory neurotransmitter agent (e.g., GABA, or a GABA agonist such as muscimol) or (ii) a synaptic decoupling agent (e.g., divalent ions such as low Ca2+/high Mg2+, Cd2+/Mn2+, or Co2+). By inhibiting neurons in the infusion area, the first approach prevents them from responding to synaptic inputs, while in the latter approach interference with Ca2+-dependent neurotransmitter release "synaptically decouples" the local area infused. Because passing fibers are not thought to be sensitive to these agents, in either case the result is that local neurons but not passing fibers are functionally removed from circuit activity, and their role in the circuit response being examined can be directly tested. See and (43) for an example of the use of these methods to investigate the role of the ventral medulla in sensory responses of locus coeruleus neurons.

Antidromic activation has at least two major uses: (i) to confirm a projection to an area and (ii) to determine the time required for conduction of an impulse along a projection pathway. In this procedure, an electrical stimulation electrode is placed in the projection area of a neuron while a microelectrode in the soma region records neurons that are "backfired" from the target area. This technique is based upon the fact that axons will conduct impulses in both directions. Although under most conditions impulses are generated at the source of the axon (soma) and only travel orthodromically, antidromic ("backwards") conduction is a powerful and convenient property. By stimulating the target area and recording an antidromically driven response elsewhere, one can conclude that the neuron recorded sends an axon to the region being stimulated (). This is often done to confirm anatomical evidence of a projection. Note, however, that because this employs electrical stimulation a positive response does not indicate that the recorded neuron terminates at the site of stimulation, only that the neuron sends a fiber (possibly en route elsewhere) to that region; additional (orthodromic) tests must be performed to indicate a synaptic input (e.g., as described under Stimulation Recording, above). In addition, such activation yields the time required for the impulse to travel the length of the axon (conduction velocity is the same in both directions along the axon). This latency of impulse conduction is useful in interpreting stimulation-recording experiments, because synaptic (orthodromic) responses at a similar latency may be due to a direct monosynaptic (versus indirect, polysynaptic) projection from the input neuron stimulated.

There are three tests that should be performed to confirm that a driven impulse is antidromic: (i) It should have a very constant latency of activation (synaptically driven responses usually exhibit a few milliseconds of "jitter"); exceptions to this constant latency rule can occur for very-small-diameter unmyelinated fibers (e.g., see ref. 44); (ii) the driven response should faithfully follow highfrequency activation above 100 Hz (e.g., two stimuli at a 5-msec interpulse interval should drive two spikes); and (iii) orthodromic (spontaneously occurring) spikes should collide with and eliminate the driven spike (collision test).

Because these methods have been in wide use for several decades, there are literally hundreds of specific examples that could be described. The reader is referred to the text by Shepherd (7) for additional references and descriptions of applications of these methods in sensory and motor systems of the brain. These methods have been used to determine the origin of the hyperactivity of locus coeruleus neurons during opiate withdrawal (45) and to demonstrate that systemic nicotine potently activates the LC indirectly (46) (see also Physiological and Anatomical Determinants of Locus Coeruleus Discharge: Behavior and Clinical Implications). These results could not be realized in cultured LC neurons, or in LC slice preparations.

The advantages of in vivo electrophysiology compared to the in vitro methods described previously are obviously due to the more intact preparation in vivo. With these in vivo methods, one can study brain regions or neurons in their intact state with its normal complement of inputs and targets, and in their natural milieu of circulating hormones and factors. The cells being studied usually have not been severed or damaged, as is almost always the case with slice studies, and have developed normally in the intact organism, in contrast to the culture preparation. These considerations lend additional credibility and fewer caveats to results concerning neuronal activity in vivo.

There are several experimental questions that require an intact organism, and they cannot be pursued in vitro. For example, to mimic the clinical situation it is important to determine the effect of a systemically administered drug (e.g., abused drugs like ethanol or opiates) upon activity in a particular brain region. In this way even if the drug has several sites of action in the brain, one sees the "net effect" of human-like drug exposure on the neurons of interest. The intact in vivo preparation is also necessary for determining the effect of certain organismic physiological manipulations on particular neurons (e.g., effects of changes in cardiovascular activity or steroid levels). Similarly, the effects of functionally defined inputs typically must be examined in the intact organism (e.g., sensory or painful stimuli). Finally, a significant advantage of the in vivo preparation for electrophysiology is that it is more readily correlated with anatomical studies than in in vitro models. Antidromic activation can more directly confirm projections found in anatomical experiments, and stimulation-recording studies can establish the functional effect and neurotransmitter of a pathway, again confirming results seen with anatomical tract-tracing and immunohistochemical experiments (for anatomical approaches to these issues see Cytology and Circuitry).

However, there are also several disadvantages of in vivo preparations. In addition to the relative difficulty in performing many of the intracellular and whole-cell studies described above (and therefore in obtaining data on membrane mechanisms of drug action), the researcher does not have as much knowledge as in the in vitro preparations of actual drug concentrations at the cell under study. Therefore, drug and transmitter responses are less confidently identified with a specific receptor or channel. In addition, there may be other confounds, such as the presence of anesthetics (or in awake animals, immobilization stress) that could alter the normal electrophysiological responses to drugs and transmitters.

Several applications of in vivo electrophysiological methods lend themselves particularly well to clinically relevant questions of special interest to neuropsychopharmacology. In addition to the points above, the in vivo preparation allows the study of neural activity and drug responses in animal models of human disorders. Studies of locus coeruleus activity in opiate withdrawal (see Physiological and Anatomical Determinants of Locus Coeruleus Discharge: Behavior and Clinical Implications, Intracellular Messenger Pathways as Mediators of Neural Plasticity, and Opioids) are obvious examples of this application, but others abound. There are animal models of other abused drugs (e.g., alcohol and cocaine) and several disorders, including schizophrenia, depression, and anxiety, all of which promise to make (or already have made) significant contributions to neuropsychopharmacology, and which require electrophysiological testing in the intact organism (for descriptions of animal models see Behavioral Techniques in Preclinical Neuropsychopharmacology Research, Central Norepinephrine Neurons and Behavior, Serotonin and Behavior: A General Hypothesis, Adaptive Processes Regulating Tolerance to Behavioral Effects of Drugs, Animal Models of Psychiatric Disorders, and Genetic Stategies in Preclinical Subtsance Abuse Research). In addition, because these techniques are needed for study of effects of systemically administered drugs, they can be an important step in new drug development.

In addition to interactions between individual neurons, there are other, more complex organizations in the nervous system. Neurons are typically associated in functionally related groups and circuits. Function at the behavioral level is a product of these neuronal networks rather than simply the product of properties of individual neurons. There are networks and circuits specialized for sensory and motor functions, and others specialized for associative activities. It seems highly likely that the elements of such neuronal networks have evolved within the context of network function(s) to have specific and perhaps unique properties tailored for that network.

As has been stated above for in vivo techniques, many questions concerning neuropsychopharmacology require experiments in the intact animal. This is perhaps most true for questions regarding cognition. While molecular and cellular experiments are important for understanding details of processes involved in mental dysfunction or drug responses, they are unable to integrate such results to ultimately and completely explain cognitive functions such as attention, perception, emotion, or memory. An analogous relationship exists between physics and chemistry: While the principles of physics are critical to our understanding of chemistry, they are not sufficient to fully understand or predict the properties of chemical reactions. It is fundamentally necessary to conduct experiments in chemistry per se or, as is the case at hand, in cognitive neuroscience.

Hence, most studies in the electrophysiology of cognitive processes involve recording single neurons in behaving animals. These methods will be briefly described below, followed by a description of methods used to locally manipulate neurons in behaving animals to test hypotheses generated by correlations found with recording experiments.

These methods rely on the same principles as described above with some modifications for behaving animals. Most such studies employ extracellular recordings from a metal microelectrode held in a miniature micropositioner on the animal's head. As illustrated in , the micropositioner is typically attached to a chronically implanted steel chamber or cylinder that is stereotaxically positioned and permanently cemented to the skull in a prior surgery. These chambers often allow lateral repositioning of the recording microelectrode for multiple penetrations, while the micropositioner allows the electrode to be lowered into the brain to the desired target along a particular track. Additional details on such methods can be found in refs. 47, 48, 49.

The most common types of microelectrodes used for recording from neurons in behaving animals are (a) etched tungsten or platinum–iridium wires, insulated with either glass or lacquer except for ~20 mm from the tip, or (b) thin microwires that are typically 25–62 mm in diameter and lacquer-insulated except for the bluntly cut tip. Neurons of different brain areas are recorded more easily with one or another type of electrode; for example, locus coeruleus neurons in awake rats and monkeys (43) are more easily recorded using the more flexible microwires. In general, microwires are advantageous for experiments entailing long-term recordings from neurons in deep structures in behaving animals, whereas etched, stiff microelectrodes are advantageous for studies where penetration of the dura mater is needed or where numerous penetrations in a small area are desired.

Two general approaches are used with microwire electrodes. Perhaps most commonly the microwires are simply implanted and glued in place with no further movement of the wires possible after surgery. In this method, a large number of wires (often more than 40) are implanted and each wire is monitored daily over the course of months for unit activity. Activity occurs on enough wires over time that many long-term and very stable recordings can often be obtained (e.g., see refs. 50 and 51). The second method is to attach a small number of microwires (two to six) to a movable microdrive which allows the wires to be repositioned vertically, and sometimes laterally as well, for new penetrations after surgery (e.g., see ref. 43). This approach has the advantage of obtaining many more recordings from a single subject than the fixed wire approach, an important consideration when subjects are in limited supply (e.g., monkeys) or when extensive training is required for each animal recorded.

High-gain amplification of signals from the head of a moving animal often yields considerable movementrelated electrical artifacts; these are the bane of a behavioral electrophysiologist. These problems are typically overcome by including a miniature first-stage amplifier in the fixed implant on the animal's head, so that recordings from the microelectrode can be amplified and converted to low-impedance signals before traveling over long distances of flexing cables.

As with all extracellular recordings, it is important to know whether activity seen on an individual electrode is generated from one neuron only, or from several nearby neurons simultaneously recorded. Results of the latter, termed multiple-cell recording or multi-unit recording, are more difficult to interpret because neurons in the multiple cell population may be physiologically heterogeneous. In that case, opposite changes in different cells recorded may appear as no change in the multiple-cell data. In addition, it is more difficult to ensure the stability of the recorded signal over time with multiple-cell activity.

In recordings in awake monkeys, the animal's head is usually fixed in place by a post that is cemented to the skull and anchoring screws, so that such movement artifacts are minimized (). This technique also allows precise measurement of the direction of gaze by monitoring eye position, an especially valuable aspect in studies where monitoring and controlling attention is important. However, this approach does not allow free movement of the animal and may produce uncontrolled effects of immobilization stress.

A key element in behavioral electrophysiology is the computer system that is used to acquire and analyze the data. Because so many events are typically recorded simultaneously (e.g., two or more cells, EEG, EMG, markers for different sensory stimuli and behavioral events, video time markers, X and Y eye positions), it is necessary to have a system that can rapidly record large amounts of data on-line with millisecond temporal resolution. In the last few years, affordable microcomputers with sufficient speed and disk storage have become available to accomplish this data storage task. In addition, sophisticated software is required so that neuronal activity or other data can be tabulated (typically in PSTHs) with respect to an arbitrarily chosen type of event out of the many recorded. For example, it may be desired to construct PSTHs of neuronal activity synchronized with a particular type of sensory stimulus out of the many presented, or synchronized with a particular type of sensory stimulus that elicits a specific behavioral response (e.g., to examine neuronal activity associated with non-target conditioned cues for trials in which the animal mistakenly elicited a behavioral response). The most challenging (but important) aspect of developing such software is to make it easy to use and easy to abstract results from very large data files, but also make it sufficiently flexible so that new subroutines can be written and integrated to analyze electrophysiological data. The latter is an ever-present need, so that typically the behavioral physiologist must be quite computer-literate!

Recordings of neuronal activity in animals performing structured behavioral tasks have proven invaluable in understanding the neural bases of various types of cognitive activity. Of the many studies using this approach, one example for activity of locus coeruleus neurons in a monkey performing an attention task is found in ref. 47 (see also Pharmacology and Physiology of Central Noradrenergic Systems, Physiological and Anatomical Determinants of Locus Coeruleus Discharge: Behavior and Clinical Implications, Noradrenergic Neural Substrates for Anxiety and Fear: Clinical Associations Based on Preclinical Research, and Serotonin and Behavior: A General Hypothesis). Similarly, recent studies by Schultz and colleagues recording putative dopamine neurons in waking monkeys (52) hold great promise for further understanding this system and its importance in neuropsychopharmacology. For other examples of the use of behavioral electrophysiology to decipher the neurobiology of cognition, the reader is referred to work by Goldman-Rakic and colleagues (47) and by Wise, Desimone, and colleagues (48, 49).

Studies of the causal role of a brain area in a specific behavior or cognitive process commonly employ lesions of one type or another. The simplest approach is to remove or destroy the area of brain by excision or electrolytic lesion and tissue coagulation. However, these approaches have serious drawbacks that limit the interpretability of results obtained, most prominently (i) passing fibers originating from neurons located elsewhere are lesioned and (ii) recovery of function may occur (via adaptive changes in remaining brain structures) during the weeks needed for the animal to recuperate from such a gross insult. This effect may lead to false-negative results concerning the role of the lesioned brain structure. The reader is referred to the elegant studies of Newsome and colleagues (53) for an example of recovery of function following chronic lesion manipulations.

A better approach to such causal behavioral studies is to (a) acutely manipulate (e.g., activate or inactivate) the neurons in question using local infusions of selective chemical agents and (b) examine the effect on behavior during the acute effect. This approach avoids the two concerns listed above. These studies employ techniques similar to those described under Stimulation Recording (above) to locally infuse drugs into a target group of neurons. Optimally, the investigator employs a combination recording/infusion electrode that allows neuronal recording from cells in the immediate area during infusions of chemical agents (e.g., see ). This permits several advantages: (i) The desired infusion site can be precisely localized by recording characteristic neuronal activity prior to infusion, (ii) the time of onset and offset of altered activity in the target neurons can be monitored and compared to the timing of behavioral changes, and (iii) individual infusion trials can be directly confirmed as to their effectiveness in altering neuronal activity at the infusion site in the desired fashion.

One example of this approach is found in recent studies by Foote, Valentino, and colleagues in studies of the impact of activity in locus coeruleus neurons on EEG in rats. As described in more detail in Pharmacology and Physiology of Central Noradrenergic Systems, local intra-coerulear infusions of the cholinergic drug pilocarpine to activate locus coeruleus neurons also activated the EEG, whereas local infusions of clonidine (that powerfully and selectively attenuated locus coeruleus activity) led to EEG synchronization. Recordings from locus coeruleus neurons during the infusions confirmed the effectiveness of the manipulations and revealed that the EEG changes were closely time-locked to the changes induced in LC activity.

A second example of the power of this approach to test hypotheses generated by electrophysiology in behaving animals is found in the work of Goldman-Rakic and colleagues. Recording experiments suggested that the dorsolateral prefrontal cortex may be involved in spatial working memory (47). To test the role of dopamine in this proposed function, Sawaguchi and Goldman-Rakic (54) infused selective D1 antagonists into this area and found that animals apparently could not remember a specific location for objects; the same objects in surrounding locations were not affected. Hence, this method can also be used to test the role of transmitter influences in a particular brain area on behavioral functions.

The advantages of behavioral electrophysiological methods for neuropsychopharmacology are numerous. These are the only methods whereby one can directly correlate neural activity with behavior. This can be especially powerful when the behavior being measured is itself a measure or reflection of a cognitive process. In addition, there is no confounding effect of anesthesia on neuronal activity recorded. Finally, because these studies take place in the intact animal, one can relate results to anatomical and neurochemical properties of the relevant circuits.

Extending electrophysiology with acute chemical manipulations of specific brain neurons can directly test the causality of hypotheses generated from the correlative results of behavioral electrophysiology experiments. This approach obviates problems of nonspecificity and recovery of function encountered with more conventional lesion manipulations. By using local infusions of neurotransmitters or related drugs, this method also provides results that can be directly related to anatomical and neurochemical results on the same system. Together, these techniques provide a powerful means of investigating the neural basis of cognitive function.

Disadvantages of this approach are also considerable. First, because these experiments take place in the behaving animal, it is difficult (if not impossible) to control all of the possible relevant variables that may affect the activity being recorded (e.g., behavioral state, stress, training differences, individual differences in task ability, etc.). Second, the behavioral measures obtained may not be temporally precise, or may only indirectly reflect the process of interest (e.g., attentional studies). Third, these studies are slow, technically difficult, and require tedious and long (often months-long) training of the animals before recording experiments can even begin. Finally, elegant though they are, these behavioral-recording studies yield only correlative data. It is necessary to extend such experiments with manipulations of the systems of interest using activation or inactivation of select groups of neurons (employing methods described below) to test causal hypotheses of their roles in specific behaviors.

The importance of these approaches to neuropsychopharmacology lies in the fact that many mental disorders are problems of complex cognitive function; the neural bases of these normal cognitive functions are only partly understood (if at all). Substantial progress in understanding and developing new treatments for disorders of memory, attention, drug craving, and the like require a more complete understanding of the underlying biological processes. As stated above, such an understanding will require experiments combining neurobiology and cognitive testing, particularly electrophysiology and local acute manipulations in animals performing sophisticated cognitive tasks. In addition, studies by Georgopoulous et al. (55) and by Houk et al. (56), among others (50), are revealing the power of analyzing activity in networks of neurons to understand the neural bases of cognition and behavior. If combined with drug administration and testing, such experiments analyzing neuronal populations in behaving animals could prove to be very valuable for future neuropsychopharmacological analyses. Such models of normal function can be used not only to understand the neural processes involved and to develop new drug treatments of related disorders, but also to test putative drug treatments in animals in a clinically relevant manner before their application to the human patient.

In this review we have only briefly surveyed the myriad electrophysiological methods available to study nervous system function. Advances in these methods have been very rapid, and it is expected that this pace will continue. The direction of future studies will be determined to a great extent by technical advances. One example: The recent rapid advances in speed and storage capabilities of microcomputer systems have greatly aided (and in some cases enabled) the above experimental procedures. Moreover, the reduced cost of these systems and commercial availability of experimentally oriented software packages has markedly increased access to sophisticated electrophysiological analyses.

It is increasingly clear that any one method or level of analysis is insufficient to provide a detailed and yet complete understanding of neural function as needed to drive rapid progress in neuropsychopharmacology. Integration among these methods and levels of analysis will prove to be very important in increasing not only our understanding of neuropsychological function, but also our ability to pharmacologically manipulate it and treat its disorders.

We thank S. Aston-Jones and C. Chiang for artwork. This work was supported by PHS grants NS24698, DA06241, DA03665, MH44346, AA06420, and MH47680 and by the Air Force Office of Scientific Research, Air Force Systems Command, USAF, under grant numbers AFOSR-90-0147 and F49620-93-1-0099.

published 2000