Ebook Current practice of clinical electroencephalography (4th edition): Part 1

(BQ) Part 1 book Current practice of clinical electroencephalography presents the following contents: The cellular basis of EEG, cortical generators and EEG voltage fields, engineering principles, recording techniques, normal adult EEG, normal pediatric EEG-Neonates and children, generalized encephalopathy, progressive childhood encephalopathy.

FOURTH EDITION

Current Practice of Clinical Electroencephalography

Current Practice of Clinical Electroencephalography

FOURTH EDITION

Editor

Professor of Neurology and Director

Adult Epilepsy Center and Clinical

Associate Editor

Aatif M Husain, md

Professor Department of Neurology Duke University Medical Center Director, Neurodiagnostic Center Veterans Affairs Medical Center Durham, North Carolina

E D I T O R S

Senior Product Development Editor: Kristina Oberle

Production Project Manager: Alicia Jackson

Senior Manufacturing Manager: Beth Welsh

Marketing Coordinator: Stephanie Manzo

Production Service: S4Carlisle Publishing Services

© 2014 by Wolters KluWer HealtH

library of Congress Cataloging-in-Publication Data

Current practice of clinical electroencephalography / editors, John S Ebersole, Douglas R Nordli Jr., Aatif M Husain—Fourth edition.

or omissions or for any consequences from application of the information in this book and make

no warranty, expressed or implied, with respect to the currency, completeness, or accuracy of the contents of the publication Application of this information in a particular situation remains the professional responsibility of the practitioner The authors, editors, and publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accordance with current recommendations and practice at the time of publication However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any change in indications and dosage and for added warnings and precautions This is particularly important when the recommended agent is a new or infrequently employed drug.

Some drugs and medical devices presented in this publication have Food and Drug Administration (FDA) clearance for limited use in restricted research settings It is the responsibility of the health care provider to ascertain the FDA status of each drug or device planned for use in their clinical practice.

10 9 8 7 6 5 4

new things on our own; to our colleagues, who supported us during this endeavor; and to our wives, who lovingly tolerated the long process

of completing this volume.

CONTRIBUTORS

A.G Christina Bergqvist, MD

Associate Professor

Department of Neurology and Pediatrics

Perelman School of Medicine at the University

Professor of Neurology and Pediatrics

Perelman School of Medicine

University of Pennsylvania;

Founder and Former director, Pediatric Regional Epilepsy

Program The Children’s Hospital of Philadelphia

New York, New York

Dennis J Dlugos, MD, MSCE

Associate Professor

Department of Neurology and Pediatrics

Perelman School of Medicine at the University of Pennsylvania;

Director, Pediatric Regional Epilepsy Program

Attending Neurologist

Division of Child Neurology

The Children’s Hospital of Philadelphia

Philadelphia, Pennsylvania

François Dubeau, MD

Associate Professor Department of Neurology and Neurosurgery McGill University;

Head, EEG Laboratory and Epilepsy Monitoring Unit Montreal Neurological Hospital and Institute Montréal, Québec, Canada

John S Ebersole, MD

Professor of Neurology and Director Adult Epilepsy Center and Clinical Neurophysiology Laboratories Department of Neurology The University of Chicago Chicago, Illinois 

Lawrence J Hirsch, MD

Professor of Neurology Chief, Division of Epilepsy and EEG; Co-Director, Yale Comprehensive Epilepsy Center

Yale School of Medicine New Haven, Connecticut

Aatif M Husain, MD

Professor Department of Neurology Duke University Medical Center Director, Neurodiagnostic Center Veterans Affairs Medical Center Durham, North Carolina

Philippe Kahane, MD, PhD

Faculty of Medicine, Joseph Fourier University Head, Epilepsy Unit, Neurology Department University Hospital of Grenoble

Grenoble, France

Mohamad Z Koubeissi, MD

Associate Professor Director, Epilepsy Center Department of Neurology The George Washington School of Medicine Washington, DC

Brian Litt, MD

Professor Department of Neurology and Bioengineering University of Pennsylvania;

Director Penn Epilepsy Center Hospital of the University of Pennsylvania Philadelphia, Pennsylvania

Douglas Maus, MD, PhD

Assistant Professor Departments of Neurology and Bioengineering University of Pennsylvania

Epilepsy Division Hospital of the University of Pennsylvania Philadelphia, Pennsylvania

Rodney A Radtke, MD

Professor of Neurology

Chief, Division of Epilepsy and Sleep

Department of Neurology

Duke University School of Medicine

Medical Director, Duke Hospital Neurodiagnostic

Laboratory

Medical Director, Duke Hospital Sleep Laboratory

Duke University Medical Center

Durham, North Carolina

Associate Professor of Neurology

Vice-Chair for Education, Neurology

Duke University Medical Center

Director, Epilepsy Monitoring Unit

Duke University Hospital

Durham, North Carolina

Joseph I Sirven, MD

Professor of Neurology Professor and Chairman Department of Neurology Mayo Clinic Arizona Phoenix, Arizona

Elson L So, MD

Director, Section of Electroencephalography Department of Neurology

Mayo Clinic Rochester, Minnesota

James Tao, MD, PhD

Associate Professor Director of Electroencephalography Laboratory Department of Neurology

The University of Chicago Chicago, Illinois

William O Tatum IV, DO

Professor of Neurology Mayo Clinic College of Medicine Mayo Clinic Florida

Jacksonville, Florida

Andrew Trevelyan, MD, DPhil

Senior Lecturer in Network Neuroscience Institute of Neuroscience

Newcastle University Medical School Newcastle upon Tyne, United Kingdom

Elizabeth Waterhouse, MD

Professor Department of Neurology Virginia Commonwealth University School

of Medicine Richmond, Virginia

PREFACE

This volume represents the fourth iteration of Current Practice of Clinical EEG As such, we hope

it reflects the progressive changes and improvements in EEG and evoked potential recording and interpretation that have occurred since the publishing of the third edition 10 years ago The fourth edition features two new associate editors, with expertise complementary to mine, and 12 new chapter authors, who are expert in their own right Our goal was to assemble a group of nation- ally recognized authors who would produce a substantial, yet practical, compendium of EEG know-how to serve as a reference for students, physicians-in-training, researchers, and practicing electroencephalographers in the 21st century.

In addition to updating areas of clinical EEG that are well established, we wanted to emphasize its neurophysiologic bases in order to promote a deeper understanding of EEG, rather than sim- ply reemphasize a recognition of its patterns We also expanded the discussion of rapidly evolving areas in clinical neurophysiology, including intraoperative monitoring, ICU EEG, and advanced digital methods of EEG and EP analysis It is our hope that EEG interpretation will be appreci- ated again as a science and not simply as a clinical art As a field of endeavor, EEG is not stagnant, nor has it reached the end of its evolution; rather, there is much remaining to learn and much to be done to exploit to the fullest these electrical signals for the benefit of our patients.

John s ebersole, MD

ACKNOWLEDGMENTS

A number of individuals contributed to this volume both directly and indirectly through the

software that they developed, which we used to create figures These include Patrick Berg (Dipole

Simulator), Michael Scherg (BESA), Manfred Fuchs and Michael Wagner (Curry) We sincerely

thank them.

Chapter 1 The Cellular Basis of EEG 1

Catherine A Schevon and Andrew J Trevelyan

Chapter 2 Cortical Generators and EEG Voltage Fields 28

John S Ebersole

Chapter 3 Engineering Principles 45

Douglas Maus and Brian Litt

Chapter 4 Recording Techniques 78

Saurabh R Sinha

Chapter 5 Normal Adult EEG 90

William O Tatum IV

Chapter 6 Normal Pediatric EEG: Neonates and Children 125

Robert R Clancy, A.G Christina Bergqvist, Dennis J Dlugos and Douglas R Nordli Jr.

Chapter 7 Generalized Encephalopathy 213

Elizabeth Waterhouse

Chapter 8 EEG in Focal Encephalopathies: Cerebrovascular Disease,

Neoplasms, and Infections 238

Joseph I Sirven

Chapter 9 Progressive Childhood Encephalopathy 258

Douglas R Nordli Jr and Darryl C De Vivo

Chapter 10 Pediatric Epilepsy Syndromes 283

Douglas R Nordli Jr.

Chapter 11 EEG in Adult Epilepsy 315

Mohamad Z Koubeissi and Elson L So

Chapter 12 EEG Voltage Topography and Dipole Source Modeling

of Epileptiform Potentials 338

John S Ebersole

Chapter 13 Subdural Electrode Corticography 367

James Tao and John S Ebersole

Chapter 14 Intracerebral Depth Electrode Electroencephalography

(Stereoencephalography) 393

Philippe Kahane and François Dubeau

Chapter 15 Evoked Potentials Overview 442

Aatif M Husain

Chapter 16 Neurophysiologic Intraoperative Monitoring 488

Aatif M Husain

Chapter 17 Continuous EEG Monitoring in the Intensive Care Unit 543

Saurabh R Sinha and Lawrence J Hirsch

Chapter 18 Sleep Disorders: Laboratory Evaluation 599

Rodney A Radtke

Index 631

CATHERINE A SCHEVON • ANDREW J TREVELYAN

The Cellular Basis of EEG

1

Introduction Electrical Flow in the Brain

Action PotentialsSynaptic CurrentsActive ConductancesGap-Junction CouplingNonneuronal Currents

The Anatomical Organization of Cortical Currents

Hippocampal AnatomyBasket Cells

Neocortical Anatomy and Thalamic Connections

Oscillations

The Structure of EEG

The Relationship between Oscillations and Cellular Activity

Hierarchical Phase-Amplitude Coupling

Cellular Basis of Epileptiform Activity

Neural Activity during Epileptiform DischargesLessons from Microelectrode Arrays: Ictal Discharges and the “Ictal Penumbra”

Surround InhibitionEEG Markers of Ictal Territories: High-frequency Oscillations

Conclusion Acknowledgments References

InTROduCTIOn

EEG remains, as it has been since Berger made his first recordings in the

1920s (1), a pivotal diagnostic clinical tool for assessing brain activity

Tra-ditionally, EEG is interpreted by visual inspection of the signal traces, using

a set of qualitative rules developed through collective clinical experience, to

define features of the EEG that are associated with particular brain states

These qualitative properties of the signal include the structure and try of prominent spontaneous oscillations such as the posterior dominant rhythm and sleep spindles, the relative mixture of frequencies and their spa-tial organization, and the presence of paroxysmal waveforms such as epilep-tiform discharges

symme-This empirical approach to EEG interpretation, in which certain features

of the EEG signal have become associated with particular brain states, has

served generations of neurologists and neurophysiologists well Historically,

the reason this approach has been predominant is obvious: Berger and his

early followers quickly realized the significance of certain characteristic

fea-tures of EEG long before we had tools that could make sense of the signals

Over the ensuing decades, clinicians have learned to use these EEG features

as powerful indicators of focal or widespread cortical abnormalities These

diagnostic associations have proved robust; they are standing the test of

time well, and will be covered extensively in later chapters

Is it enough then simply to understand EEG as a “black box,” and go

no further? We believe not There are powerful arguments that we should

strive for deeper understanding, asking why these associations exist One

reason is that technology is advancing rapidly, allowing us to record and

manipulate brain signals in ways that will shed new light on the meaning

of old data sets Another motivation is that in many ways, the EEG is an

impoverished signal, which prevents us from differentiating between various

possible underlying activities This leads to these different activities being

pooled together, which then weakens what associations we can draw

An-other important reason to explore the neural basis of these signals is that

there are interpretative pitfalls For instance, some uses of EEG, such as

the localization of activity, are fraught with the problem of circular logic, if

independent measures of activity patterns are lacking It is important to be

able to recognize these cases

There is a tendency to rest on the simple intuition that the EEG signal is

merely a weighted average of everything that goes on beneath the electrode

In some ways, this is a truism, but the danger comes from confusing the

different levels of understanding the brain The EEG derives from currents

flowing in and out of cells, but the activity that most neuroscientists and

neu-rologists are ultimately interested in is neuronal firing behavior, because that

is how information is thought to be processed One of our aims in this

chap-ter is to show that the ionic level and the firing levels do not always relate in a

simple manner Most importantly, we want to stress that these issues are not

merely of academic interest, but may seriously impact on clinical practice

In this chapter, we will describe findings in basic neuroscience that link

these EEG signal features to the sources of electrical activity in the brain

Our goal is to provide electroencephalographers with an understanding of

the cellular activity that contributes to the electrical signals measured by

EEG, the sources and structure of cortical oscillations, and the network

be-haviors contributing to normal and pathologic EEG activity

We are at an exciting time in neuroscience, when new ways of recording,

analyzing, and even manipulating neuronal activity are being developed at

great pace Recent neuroengineering advances have resulted in clinical sensors capable of high spatiotemporal resolution recordings that augment standard EEG with information about neuronal firing This makes available a broad array of neurophysiological data types, ranging from oscillations synchro-nized over large brain areas to unit firing in a single cortical macrocolumn, which electroencephalographers should develop familiarity with, as the new technologies find increased use in the clinical setting There is also increas-ing use of sensors recording from below the dura or within the brain paren-chyma, which not only afford ever greater diagnostic possibilities but create new opportunities for the investigation of human cognitive function, a ven-ture that requires close cooperation between neuroscientists and clinicians

ElECTRICAl FlOw In ThE BRAIn

Changes in the extracellular field potential arise because ions flow in and out of cells at very focal sites, thereby creating ionic flow also in the ex-tracellular space (Fig 1.1) The driving force for the initial movement is invariably the electrochemical gradients that exist across the cell membrane

By far, the largest currents flow through ion channels; indeed a simple lipid bilayer without channels is essentially impermeant to ions There are also small currents associated with electrogenic ion pumps—proteins that gener-ally are working to restore ion balances and thus are working against the electrochemical gradient Examples of these are the various ATP-dependent pumps that restore Na+, K+, Ca2+, and H+ across various membranes These can ultimately shift the membrane potential by up to about 10 mV, but in terms of rates of flow of ions, these currents are far smaller than channel currents It is possible that these pumps may contribute to extremely slow frequency changes in the extracellular field, including so-called DC shifts, small step changes in the electric field that have been seen in certain brain states including epileptic seizures—this remains an open question It is quite clear, however, that any rhythms higher than about 1 Hz are too fast to

be attributed to pumps, and that all such faster rhythms must therefore arise from flow through ion channels For the purpose of this chapter, therefore,

when referring to cell membrane currents, we will use the term conductance

as a synonym for ion channel Patch clamp studies can derive the unitary conductance of a single ion channel, and so the total conductance is simply the product of this unitary conductance and the number of channels that are open Since the total conductance is a nice intuitive term, reflecting the total number of channels open, it is used in preference to its reciprocal, the resistance, when talking about membrane currents

Figure 1.1: local circuits induced by transmembrane currents A: Changes in

intra-cellular membrane potential (upper panel), and in the membrane conductance of

Na + and K + during an action potential B: as the action potential propagates, the

highly focal changes in transmembrane conductance create large movements of

Na into, and K out of the cell Currents moving into cells are referred to as current

sinks (i.e., relative to the extracellular space), and those moving out of cells and into

the extracellular space are called current sources Charge redistribution away from

these sinks and sources can be measured by electrodes as fluxes in the electric field

C: schematic showing how the electric field drops with distance down a cable a

steady state holding potential drops off with a space constant, l = (d·Rm/4·Ri) 0.5 The

space constants for oscillating currents are shorter (the drop off is faster), because

the capacitance absorbs current proportional to the rate of change of the potential

The dissipation of electric field in the extracellular space is likely to be similar,

al-though the complexities of the circuit are far greater than for intracellular currents,

and so we do not have analytical solutions for the extracellular space d: local

elec-trical circuits may also be created through synaptic conductances E: linear iV

rela-tionships follow ohm’s law, where the gradient is the conductance (reciprocal of the

resistance), and (F) the intersection of the abscissa is called the reversal potential

Persistent currents are also termed “passive conductances”—synaptic currents that

are ohmic are not strictly speaking “passive” because they are gated currents, by

neurotransmitter binding in contrast, other currents show large voltage-dependent

changes in conductance These are referred to as “active conductances.” all gated

currents, including both neurotransmitter-gated and voltage-gated currents, can be

very rapidly activated, thereby generating large currents and rapid fluxes in the

ex-tracellular field.

juxtaposition of the firing neurons, and their positions relative to the ing electrodes, influence the recorded signal Consider, for instance, the case when two closely apposed neurons are firing repeatedly at high frequency (>250 Hz) but 1 to 2 milliseconds out of phase with each other Locally, this induces small currents reverberating in the extracellular space between the cells, as Na+ ions go in and K+ ions come out of the cells These rever-berating extracellular currents may be recorded, but they are small and very local, and thus their visibility is very dependent on the exact positioning of the recording and reference electrode, and the impedance of the electrodes Now consider a larger cluster of neurons, say 10 cells firing at more typical rates for pyramidal cells, at 50 Hz, but again with some millisecond jitter As before, if the recording electrode is located within the cluster, then one may

record-be able to discern the direction and amplitude of small, local extracellular currents relating to particular cells (relative to the distant location of the ref-erence electrode), and this is the principle behind spike-sorting algorithms When the recording electrode is outside the cluster of neurons, the jitter in firing results in some destructive interference, and there may be little current flow at the actual electrode If, however, firing becomes more synchronous, then the currents become far more visible at that electrode The visibility of the currents, therefore, requires synchrony at the timescale of the transmem-brane currents We will return to this point shortly when considering the much longer duration synaptic currents

A second factor determining the visibility of currents is their amplitude Hodgkin and Huxley based their model on recordings of the squid giant axon, the largest axon in the natural world, but much subsequent work has shown that the essential features of the Hodgkin-Huxley model are also replicated in the far smaller neurons in mammalian brains Regarding how these currents appear in the EEG, however, in this instance, size matters The essential details are exactly analogous to changes in concentration

A drop of ink added to a thimbleful of water will cause a big change in color, but the color change might be imperceptible when a drop is added

to a bucketful With neuronal membrane potential, we are dealing with charging up membranes; so the structure of interest is a surface area, not a volume, but the principle is exactly the same For large structures, like a cell body, or the squid giant axon, a large charge flow is needed to substantially change the membrane potential Conversely, a large membrane potential change implies a large transmembrane current But for tiny structures, like most mammalian axons, the currents involved are, relatively, very small Furthermore, many mammalian axons are myelinated (although many lo-cally connecting grey matter axons are not), and so still smaller currents

Action Potentials

The elucidation of the currents flowing during an action potential, by Alan

Hodgkin and Andrew Huxley, was a seminal moment in the history of

neuroscience (2–5) (Fig 1.1) They described how, as a neuron is

depolar-ized from rest, there is a point at which voltage-dependent Na+ channels

are suddenly opened There is then a surge in the inward current, creating

a positive feedback cycle, which is only terminated by the equally sudden

inactivation of the Na+ channels Inactivation occurs very quickly, after

about a millisecond, and coincides roughly with the slightly slower

open-ing of voltage-dependent K+ channels This combination of closing of Na+

channels and opening of K+ channels rapidly brings the membrane

poten-tial back down to resting membrane potenpoten-tial (or even overshoots briefly)

Restoration of the resting membrane potential results in shutting of the

voltage-dependent K+ channels

The density and kinetics of the ion channels (particularly the K+

chan-nels) vary in different neuronal classes, allowing some cells to fire at far

higher rates than others Firing patterns are also influenced by other slower

currents, which shape the tail end of action potentials, leading to after-

hyperpolarizations and after-depolarizations The major currents

underly-ing an action potential, though, last in the region of just 1 to 3 milliseconds

Consequently, action potentials are very high-frequency signals, and this

affects how far they travel through electrical fields with significant

ca-pacitance (Fig 1.1) Electrophysiologists refer to the “transfer function”

(6), which is the drop in electrical field going from point A to point B

A steady holding potential at point A drops off rapidly when recorded

at increasing distances Most theoretical work on transfer functions has

considered the flow of membrane potential changes intracellularly; for

in-stance, examining how a synaptic potential drops from the dendritic spine

to where it is usually recorded at the soma Extracellular transfer functions

are considered to be analogous, except that the mathematical model is far

more complex, with multiple parallel routes through the network and a

less clearly defined capacitive element, and an analytical solution, even if

it were possible, has not been worked out Empirically though, it is clear

that high-frequency voltages such as action potentials dissipate rapidly in

the extracellular space In contrast, slower electrical fields, such as those

associated with synaptic currents, influence the electric field over a larger

volume of the brain

The very transient nature of action potential currents is also an

impor-tant issue when there are multiple neurons firing The spatial and temporal

According to these equations, the driving force for any transmembrane current is the difference between the reversal potential and the membrane potential When a receptor opens on a dendritic spine, small ion movements cause big changes in membrane potential (the ink and thimble analogy men-tioned earlier), and so the membrane potential will shift rapidly toward the reversal potential of the receptor As a result, the driving force drops sharply

in these small structures, and in turn, the current also drops quickly In trast, receptors on larger structures, such as the dendritic trunk or soma, do not shift the membrane potential toward the reversal potential as easily; so the driving force persists and the currents also remain high

con-A second important feature of synaptic receptor currents is that, relative

to action potential currents, they are long lasting; most last 5 to 20 seconds, while others last hundreds of milliseconds The best characterized

milli-of these long-lasting synaptic currents are N-methyl-d-aspartate (NMDA) plateau potentials (7) NMDA receptors require glutamate binding to open, but like Na+ channels are also voltage-gated, and can thus also pro-duce regenerative currents This is a kind of action potential, but unlike our previous description of the Hodgkin-Huxley model, NMDA receptors are inactivated far more slowly than Na channels and thus induce a self-sustaining depolarization This “plateau potential” may also be boosted by other voltage-dependent Ca2+ currents, because these too are inactivated slowly, and hence also contribute to the self-sustaining depolarizing current over tens of milliseconds A further important feature of NMDA receptors

is their modulation by other extracellular ions and molecules Mg2+ ions are critical for the voltage- dependency (8); in addition various amino acids, including glycine and serine, bind at extracellular sites and modulate the NMDA opening properties (9) Thus, changes in the ionic and molecular constitution of cerebrospinal fluid may influence the likelihood of such pla-teau potentials occurring and also their duration

The protracted nature of these postsynaptic currents means that they will summate strongly in the extracellular field, even when the initial, causative, presynaptic action potentials are not absolutely synchronous The time scales of these currents are simply far longer than those of action potentials Furthermore, there is one obvious, and commonly occurring, situation when the postsynaptic currents are highly synchronized: when they arise from an action potential in a single axon, releasing vesicles at multiple varicosities

We shall return to this point later, in our discussion on basket cells

To summarize the section so far, postsynaptic currents are larger and ger lasting and have lower-frequency components than those underlying ac-tion potentials, and consequently are more visible in extracellular recordings

lon-are involved per unit length of axon, because the current only flows at

particular hotspots, the nodes of Ranvier Consequently, as it propagates

through the axonal tree, an action potential does not constitute a large

“sink” of current away from the extracellular space Moreover, since action

potentials propagate rapidly down the axon, the current sink is

distrib-uted over a spatially extensive area very quickly In contrast, a rather more

visible current in the EEG arises from the postsynaptic consequence of

axonal firing

Synaptic Currents

Neurotransmitters released from axonal terminals and varicosities cause

postsynaptic receptors to open on dendritic spines, on the dendrites

them-selves, and also on the soma and axon hillock The same size argument as

outlined earlier, for axons, also applies for dendritic spines In contrast, the

dendritic trunks may be relatively large (>5 μm diameter, compared to

ap-proximately 1 μm or smaller for most axons and dendritic spines)

Theoreti-cal Theoreti-calculations show that these size considerations impact on the currents

flowing through receptors at these different locations The currents flowing

are dictated by Ohm’s law

Turning this equation round, we get

R

This may be rewritten more appropriately for transmembrane currents, in

terms of the membrane driving forces and the conductance, g, which is the

reciprocal of resistance (g = 1/R) Thus, the current (note that the

conven-tion is to plot inward currents as negative) is expressed as

where Em is the membrane potential, and Erev is the reversal potential for the

ion in question, which is set by the concentrations inside and outside the cell,

as given by the Nernst equation

Erev α In[ion]out (4)

[ion]in

Active Conductances

In addition to numerous synaptic receptors, the dendrites and somata have

many other conductances The most relevant for EEG are so-called active,

or non-Ohmic, conductances These change their conductivity with changes

in membrane voltage, unlike Ohmic conductances, which have a linear I-V

relationship Hodgkin-Huxley channels and NMDA receptors are all

exam-ples of active conductances, but there are many others besides Of particular

importance are the voltage-dependent Ca2+ channels, which in addition to

their role in gating neurotransmitter release at synaptic terminals, are also

found in the dendritic tree and soma where they underlie bursting

behav-ior in many neuronal classes, including certain classes of pyramidal cells in

cortex and hippocampus In these pyramidal cells, there are great

concentra-tions of voltage-dependent Ca2+ channels at the tuft of the apical dendrite,

which may play a role in certain rhythmic activity and in distorting, and

especially boosting, synaptic inputs Since these are concentrated within the

upper cortical layers, they are examples of highly localized current sinks

An important feature of these active conductances is that they can give

rise to sudden and large changes in local transmembrane currents near their

threshold voltage However, synchronization of these currents in populations

of neurons, which is likely to be a prerequisite for such currents being visible

in EEG recordings, requires interactions between neurons The most

obvi-ous way this happens is through chemical or electrical synapses, although

recent work has provided experimental evidence also for so-called “ ephaptic”

interactions Ephaptic interactions refer to how very small fluctuations in the

local field potential caused by one cell’s firing, passively entrain the activity

of other neurons (10–13) (Fig 1.2) Several groups have now shown that

elec-trical fields, of comparable amplitude to those actually recorded in the brain

(2 to 4 mV/mm), can modulate firing patterns (13) Interestingly, the effect

appears to be proportional to the rate of change of the field, so higher

fre-quency oscillations entrain more powerfully than lower frequencies, although

as we discussed earlier, lower frequencies propagate further in the brain

Thus, even though these local ephaptic interactions are tiny, the entrainment

may still grow to synchronize many neurons, in the same way that fireflies

may coordinate their flashes, or theater audiences come to clap in synchrony

Gap-Junction Coupling

Electrical synapses, through gap junctions, have been described in a number

of cortical populations The best evidence for such connections is between

Figure 1.2: entrainment of neuronal bursting induced by small fluctuations in

the extracellular field (adapted from fröhlich f, McCormick Da endogenous

electric fields may guide neocortical network activity Neuron 2010;67:129–143)

spontaneous bursting activity in a brain slice was modulated by application of small extracellular voltages of comparable size to those recorded in vivo The switch between up and down states is rapidly entrained to whatever extracellular rhythm is being imposed, as shown by the close correlation between the period

of imposed field and the period of the entrained activity Demonstrations like this provide a proof of principle that ephaptic effects are real, albeit very small indeed the intracellular effect is likely to be no larger than the summed activation

of less than about three to five excitatory synapses.

members of the same inhibitory neuronal classes Thus, parvalbumin- expressing basket cells are connected through gap-junction syncytia, as are somatostatin positive neurons, but it appears that the two classes of neuron do not connect with each other, and that the syncytia are thus entirely separate There are theoretical arguments and some experimental data also suggesting that pyramidal cells may be interconnected through gap junctions located between axons, creating what is referred to as an axonal plexus It has been suggested that, in certain conditions, activity may spread and escalate through this plexus, building up eventually to a full ictal event This model of ictal initiation derives from a rather artifi-cial, in vitro epilepsy model in which all synaptic activity is prevented by removing Ca2+ ions from the extracellular medium In contrast to synaptic

there is continued debate regarding to what extent these data merely show the potential of glia, rather than what really happens in vivo Our under-standing of these phenomena continues to evolve, and will soon clarify what glia do in both physiological and pathologic brain states.

A more established function for glial syncytial networks is the tion of local inhomogeneities of extracellular ions After intense bursts of neuronal activity, extracellular K+ rises and Ca2+ drops markedly Follow-ing an epileptic burst, extracellular K+ may easily exceed 10 mM Steven Kuffler, in the 1960s, suggested that glial networks may redistribute K+ ions

redistribu-in these conditions Recently, strong support for this has been provided by the demonstration that glucose imbalances in parts of cortical networks cause increased coupling of glia, and thereby also enhanced the spread of

a fluorescent glucose analog (17) If the same process occurs with charged ionic species such as K+ ions, then focal epileptic activity would trigger po-tentially a large current sink, with characteristically slow kinetics, and the current circuits would match the dimensions of the extent of changes in extracellular K+ This needs further research, but the theoretical argument suggests that such low-frequency currents, reflecting their slow kinetics, ap-pear in the EEG as a change in very low-frequency (or DC) power, and may thus provide a useful marker of the location of intense epileptiform discharges

ThE AnATOmICAl ORGAnIzATIOn

OF CORTICAl CuRREnTS hippocampal Anatomy

The hippocampus in humans is tucked away from accessible sites for ment of EEG electrodes, or even subdural electrodes As such, activity in the hippocampus is largely invisible in EEG or ECoG recordings, although

place-it can be recorded readily using depth electrodes Hippocampal recordings both in vivo and in vitro have contributed hugely to our understanding of synaptic physiology and network oscillations The usefulness of the hip-pocampus for these researches is that its anatomical organization is re-markably laminar Pyramidal somata are packed densely in the pyramidal layer, with dendritic trees that project almost perpendicular to that layer

on both sides The excitatory drive on to pyramidal cells is arranged pendicular to the dendritic trees, and parallel with the pyramidal cell layer

per-In short, one could not design a structure more suited to the study of sociative learning, allowing the simple independent stimulation of parallel

as-function, gap-junction coupling appears enhanced in zero Ca2+ medium,

and sharp electrode recordings of neurons commonly show “spikelets,”

which are presumed to be retrogradely propagating action potentials that

invade the soma, but which fail to induce a full action potential there

be-cause of the impedance mismatch going from the small axon to the large

soma The trigger for the back-propagating action potential has been

sug-gested to be a gap-junction connection with another active axon

An important point to make is that gap-junction-mediated currents

be-tween cells are likely to be virtually invisible to extracellular recordings

This is because gap-junction currents flow directly from one cell to the

next, without creating a local sink or source of current in the extracellular

compartment They may be visible indirectly, however, by synchronizing

firing between cells (14,15) As such, gap-junction coupling may lead to

summation of transmembrane currents in such synchronized cells Two

other points are worth making First, gap junctions may extend the

elec-trotonic length of neurons: Quite simply, the cellular “unit” is extended

into the adjacent cell, creating a syncytium, and so charge entry into one

cell is then dissipated and eventually returned to the extracellular space

over a larger area In this way, the effective circuit is increased in size,

although in doing so, the “return” part of the circuit, away from the

ini-tial current, carries a lower density current because it is distributed over a

larger surface area The second is that while the conventional arrangement

is for hemichannels to be juxtaposed with hemichannels of adjacent cells,

thus making a full gap junction, there are cases when the hemichannel

appears to exist in isolation It is possible that such hemichannels may be

induced by epileptic activity (16), and cause further trauma to the

partici-pating neurons, while also allowing currents that may be visible in EEG

recordings

nonneuronal Currents

Finally, we come to currents flowing through glial networks Glia are also

highly connected through gap junctions, and this facilitates the spread of

waves of Ca2+ transients from cell to cell The cytosolic Ca2+ surges come

primarily from intracellular stores, making them unlike those in neurons,

which primarily reflect transmembrane currents As such, glial Ca2+

tran-sients are not likely to have a direct impact on EEG signals They may do

so indirectly though, by releasing neuroactive substances, which cause local

neuronal discharges There is good evidence from many laboratories

sup-porting this view, although much of this derives from in vitro work, and

at very high frequencies (>200 Hz) Furthermore, while current thinking is that all fast-spiking basket cells express parvalbumin, it is also the case that other cell classes express parvalbumin too, and even that parvalbumin ex-pression may be altered by activity levels Thus, all fast-spiking basket cells may be parvalbumin positive, but not all parvalbumin positive neurons are fast-spiking basket cells We make these points mainly to emphasize the fact that interneuronal classification is a minefield for the unwary, and that many specialists, quite rightly, get agitated when others are sloppy with their clas-sification As ever, the devil is in the details, and we would encourage read-ers to consult more specialist texts on this topic, of which there are many (19,20) We will limit ourselves here to some general, but important, points that are particularly relevant to epilepsy and to the nature of EEG signals.There are many notable features of these remarkable cells; the most im-portant though regard their output Basket cells project extremely densely

on to the local pyramidal population Indeed, each basket cell appears to synapse on to every single pyramidal cell locally, and furthermore, makes on average at least 10 synapses per connection These synapses are all clustered densely around the proximal dendrites and the somata of the pyramidal cells The synapses show a high transmitter release probability, and the post-synaptic receptors have a high conductance Thus a single action potential

in a basket cell may trigger a very large summed conductance in a very cal region of cortex We make a very important distinction here between conductance and current, because as Ohm’s law states, the current depends

fo-on both the driving force and the resistance Thus there may be a large cfo-on-ductance, but if the membrane potential is close to the GABAergic reversal potential, the driving force is small At rest, EGABA (typically about −60 mV) lies close to the membrane potential, and under these conditions, the large basket cell-induced, postsynaptic current is small

con-In one very important situation, however, the driving force is large, and therefore so is the GABAergic current This occurs when there is concur-rent glutamatergic drive and basket cell firing The glutamatergic input drives the pyramidal cell to a more depolarized state, thus shifting the membrane potential away from the GABAergic reversal potential and creating a large driving force The inhibitory postsynaptic currents (IPSCs) are thus large, and so constitute a powerful hyperpolarizing drive They also have a second inhibitory action, which is referred to as shunting inhibition: They make the cell far more “leaky” and thus more current is required to depolarize them This is particularly relevant for pyramidal cells, which have a very striking arrangement of synapses on their dendritic trees: All their excitatory inputs are located on distal dendrites, with nothing on the most proximal dendrites

inputs onto a single cell by placing the stimulating electrodes at different

depths along the somato-dendritic axis Such stimuli, and indeed any

bar-rage of presynaptic inputs, induce a focal synaptic current with respect to

this same axis, and conversely, one may record current sinks along that axis

and associate them with particular presynaptic inputs Further, one may

progressively move a recording electrode along the somato-dendritic axis,

and record an inversion of the sign of the recorded field, and thus ascertain

the location on the dendrites of the transmembrane current The direction

of the sign inversion may then indicate whether the transmembrane current

represents a current sink (positive charge entering the cell, and thus

“leav-ing” the extracellular space) or a source (negative charge entering the cell

[e.g., a hyperpolarizing GABAergic current] or positive charge leaving the

cell [late action potential K+ currents]) This approach to dissecting out the

location of synaptic currents and neuronal firing is referred to as “current

source density analysis” (18)

The arrangement of inhibitory circuitry in the hippocampus is similarly

ordered (19,20) Thus, anatomists have been able to classify many

differ-ent types of hippocampal interneuron, characterizing them according to

their stereotyped axonal and dendritic morphology and thus their drivers

and their outputs, the location of their somata, the expression of

par-ticular proteins (parvalbumin, somatostatin, VIP, calretinin, etc.), their

firing patterns, and many other cellular attributes A major research goal

in recent years has been to characterize how these different interneuronal

groups participate in different network oscillations, or even sculpt these

rhythms

Basket Cells

Special mention needs to be made of one particular inhibitory cell class,

that which targets the somata of pyramidal cells The interneurons wrap

their axons around the pyramidal somata, which thus appear like baskets

cupping the somata, hence the derivation of Cajal’s term, the basket cell

(his Golgi stains typically only labelled small number of cells, and so he saw

many examples when the presynaptic axonal basket was seen without the

contents of the basket, the pyramidal soma, being visible—it looks even

more basket-like when the basket appears empty!) These cells also go by

other names, most commonly fast-spiking interneurons, and parvalbumin

positive neurons, but these are not, strictly speaking, synonyms It is clear,

as the classification of cortical interneurons becomes more sophisticated,

that there are at least two populations of basket cells, only one of which fires

pyramidal cells are receiving essentially the same IPSCs, they are all ited synchronously, but then all “bounce back” to fire in synchrony too Essentially, the same mechanism explains how a depolarizing agent such as kainate, applied continuously to a cortical slice, sets up a gamma rhythm

inhib-In short, it creates a powerful depolarizing drive to most of the network, which is then overridden by firing of basket cells, but then the depolarizing influence kicks in again to create a pacemaker current The exact rhythm

is essentially dictated by the rhythm of the basket cell firing, which is also coordinated by gap-junction coupling as described earlier

There is ongoing debate about whether this mechanism is the same for all gamma rhythms Gamma activity is considered to extend from about 30 to

150 Hz, but this may be further subdivided into “low” and “high” gamma It

is also relevant that in neocortical brain slices, certain pharmacological nipulations may induce concurrent, but separate, gamma frequency rhythms

ma-in the supragranular and ma-infragranular layers (24) It remama-ins to be resolved what distinguishes the two rhythms in this model It is notable, though, that the predominant laminar frequencies are the same as the firing rates of bas-ket cells in those same layers, suggesting that the two frequencies arise sim-ply because the different laminar basket cells vary in their sensitivity to the pharmacological agent, kainate What this work does show, however, is that

a single cortical column need not follow the same rhythm throughout the cortical depth, and that the separation of rhythms may reflect the external drive Gamma activity in vivo may show increased variations in the instan-taneous peak frequency, which is thought to reflect the fluctuations in the amplitude of the population IPSC: a large IPSC, perhaps unsurprisingly, delays the rebound firing, and thus extends the period of the oscillation

neocortical Anatomy and Thalamic Connections

Most of the same cell classes have been identified in neocortex, as have been described in hippocampus One difference is that in neocortex, pyramidal somata are distributed across almost the entire cortical depth, from layers 2

to 6 It is possible that this arrangement reduces the potential for ephaptic influences in neocortex, but the truth is we do not know why this evolution-arily more recent cortical structure (“neo”-cortex) is arranged differently from archeocortex (hippocampus) A consequence is that the neocortical anatomy is not as crystalline as in hippocampus, but it seems reasonable to presume that the same basic principles elucidated using hippocampal prepa-rations will also apply in neocortex Because the somato-dendritic axes of the dominant cell type (pyramidal cells constitute about 80% to 85% of the

and soma, where instead, a huge density of inhibitory synapses is located

This distribution of synaptic input means that these proximal inhibitory

syn-apses can veto virtually any level of excitation This is clearly predicted on

theoretical grounds, and has been shown experimentally for a comparable

arrangement of synapses in crayfish (21), and more recently, in pyramidal

cells themselves, in a simple model of epileptiform propagation (22) In this

model, ictal discharges are induced by removing Mg2+ ions from the bathing

medium, thereby increasing excitation while preserving (at least initially)

in-hibition The cortical territories immediately ahead of the ictal wavefront are

bombarded by excitatory barrages arising from the wavefront itself, but this is

not necessarily translated into postsynaptic firing because of the power of the

inhibitory restraint on these neurons Thus, there arises a kind of “ictal

pen-umbra” around territories that have been recruited to a seizure, in which there

may be an extreme discrepancy between the level of synaptic currents (both

excitatory and inhibitory) and the level of neuronal firing This is a very

im-portant scenario in human seizures as well, which we will return to later

An important factor in the coordination of this restraining inhibitory

burst is that gap junctions interconnect these interneurons Gap-junction

coupling has the effect of synchronizing firing in connected cells (14,15)

Consequently, the high-frequency discharges of basket cells during these

bursts of activity may be synchronized with submillisecond precision across

an extended syncytium of cells The pattern of inhibition during this

re-straint is of high frequency (100 to 300 Hz) and large amplitude IPSCs are

experienced by all pyramidal cells in a local area Studies of these types

of discharges in the magnesium washout model described earlier showed

that these currents are highly visible in the local field potential Although

results from an in vitro model should be regarded only as proof of principle

rather than conclusive evidence of cause, this may be one of the

mecha-nisms that can give rise to high-frequency oscillations in clinical recordings,

a potentially important epileptic biomarker We will address this topic in

more detail in a later section

In addition to a possible role in high-frequency oscillations, basket cells

have a central role in gamma rhythm activity Again, fundamental to this

rhythm is the power of basket cells in inhibiting pyramidal firing, which

allows them to entrain pyramidal activity extremely well This entrainment

has been shown compellingly by recording the initial desynchronized firing

of pyramidal cells from a baseline depolarizing state, and then imposing a

gamma frequency basket cell inhibition on the pyramidal cells (23) Each

volley of IPSCs shut the pyramids down transiently, but this is followed by

a narrow window between IPSCs when the pyramid can fire Since all local

acronym, HCN stands for “hyperpolarization-activated, cyclic- nucleotide modulated”, which already tells us much about these conductances) (27) These channels carry maximal current at relatively hyperpolarized levels (about −80 mV), dropping down to almost nothing when cells are depo-larized to about −30 mV The key effect of this activation curve is that the currents act to resist changes from a moderately hyperpolarized state Take the situation where a cell is stable at rest at −65 mV At this membrane po-tential, there will be a subpopulation of these channels opened, which are providing a tonic depolarizing current that will be contributing to the steady state at that membrane potential If the cell experiences a hyperpolarizing drive though, some of these HCN channels will close, thus reducing the depolarizing drive, and the membrane potential heads back toward −65 mV Remarkably, exactly the same effect happens if a hyperpolarizing drive

is given, because this will cause HCN channels to open, thus providing a stronger depolarizing current, and again causing the membrane potential

to head back toward −65 mV The currents have quite slow kinetics ever, opening or closing over a time course of about 10 to 15 milliseconds Thus, if these cells are given a sharp inhibitory synaptic drive, arising for instance in the reticular nucleus and mediated by both GABAA and GABABreceptors, the HCN channels will then create a pacemaker current from this inhibitory trough If the inhibition is sufficient to de-inactivate the low- threshold voltage-gated Ca2+ channels (i.e., membrane repolarization following closure and inactivation of the Ca2+ channel changes its confor-

how-mation to a closed, but activatable, state), the Ih pacemaker current can then reactivate these channels, and the neurons fire a burst of action potentials This reactivates the reticular nucleus, which reinhibits the thalamus, and we have the potential for a self-perpetuating rhythm

This pattern of bursting provoked by voltage-gated Ca2+ channel vation, the latency of synaptic delays between the reticular and thalamic

acti-nuclei, and the slope of the Ih-driven pacemaker drive a rhythmic 7 to 14 Hz oscillation in the cortex, referred to as sleep spindles (28) These occur every

3 to 10 seconds in the early stages of sleep, and are most prominent in the frontal and midline regions A related mechanism is responsible for the spike and wave discharges (SWDs) that are a hallmark of idiopathic generalized epilepsy syndromes These differ from spindles in several respects They are slower (3 to 6 Hz), larger in amplitude, have a prominent spike component preceding the slow wave, are recorded throughout the cortical mantle, but may be either frontally or posteriorly predominant The most notable dif-ference, though, is that unlike sleep spindles, SWDs may also occur during wakefulness, when they may be clinically manifest as absence seizures

total neuronal population) are not all aligned, though, neocortical current

sinks and sources may not be so cleanly recorded as in hippocampal tissue

An important feature of neocortical circuits is the pattern of recurrent

connections with thalamus Thalamic nuclei not only provide the major

ex-ternal drive to cortical circuits, but also receive back from cortex a large

descending input, primarily from layer 6 pyramidal cells This corticofugal

pathway sends collateral branches also to the reticular nucleus, a group of

inhibitory neurons that then project on to the same set of thalamic neurons

as the direct cortical pathway targets Cortical discharges therefore not only

create a direct excitation of thalamic neurons, but also cause a disynaptic

inhibition mediated through the reticular nucleus

A further key feature of this thalamocortical circuit is that both

tha-lamic and reticular neurons have a peculiar arrangement of low-threshold

activated Ca2+ channels The two nuclei express slightly different forms of

voltage-gated Ca2+ channel: Both nuclei express the Cav33.3 isoform, but the

reticular nucleus also expresses Cav2.3, which is activated at higher threshold

than those found in thalamocortical neurons, and this may be relevant to

the subtle distinction between pathologic and physiologic thalamocortical

bursting (25), which we will discuss shortly The presence of these channels

means that thalamic neurons discharge in two different ways If they are

activated from a relatively hyperpolarized state (below about −65 mV), then

the depolarization opens both Na+ and Ca2+ channels The latter inactivate

far slower than Na+ channels, and consequently, cause secondary firing; in

other words, the combined activation of both Na+ and Ca2+ channels causes

a burst of action potentials This is the dominant pattern in sleep and also

underlies the spike and wave pattern of idiopathic generalized epilepsies In

the waking state, thalamic neurons at rest are in a more depolarized state, at

about −60 mV, and in this state, the Ca2+ channels are already inactivated

A further, transient depolarization then only activates Na+ currents, and

the cells fire just a single action potential This slightly paradoxical state of

affairs, whereby neurons burst more intensely if driven from a more

hyper-polarized state, is thought to underlie two different patterns of information

transfer in the thalamus Bursts of activity may provide some kind of

wake-up call, but the normal waking mode, in which thalamic neurons only fire

single action potentials may more faithfully transfer sensory information to

the cortex

There are three other notable membrane currents in thalamic neurons,

which may influence the likelihood of rhythmic bursting (26,27): Ih, GABAB

and tonic GABAA currents Ih is the hyperpolarization-activated, nonspecific

cation current (Erev = −30 mV), encoded by HCN1 and HCN2 genes (the

(“blocking”) with eye opening were described by Berger in 1929 (1) ilarly, the mu rhythm, with components in the alpha and beta frequency bands, is related to sensorimotor processing in the frontal and parietal lobes (38) As we have discussed, spindles and rhythmic slow waves are distinctive features of the sleep states, whereas during wakefulness, polymorphic and rhythmic, local or diffuse slow waves are commonly associated with cerebral dysfunction There are also many highly characteristic patterns associated with particular seizure phenotypes, and we could list many other examples While these long-recognized clinical associations are reason enough to study oscillations, it is increasingly clear that the very nature of brain function is tied to oscillations (39,40) We will not understand the brain until we under-stand how and why groups of neurons behave this way.

Sim-Interest in brain oscillations has grown immensely in recent years, vated largely by two pieces of research that suggested particular functional significance to oscillations The first was a series of recordings made by John O’Keefe and Colleagues using extracellular electrodes implanted into the entorhinal cortex of rats that were trained to run along simple mazes and tracks (41) They could isolate the activity of single cells (“unit” recordings, detected and sorted from the high-pass filtered signal as we described ear-lier), and also the local field potential (low-frequency bandpass recordings), all done with the same electrode They found that as the rat moved along a simple track, cells consistently fired when the rat was at certain points along the track That is to say, individual cells represented particular places along the track, and these cells were duly termed “place cells.” The other notable finding was that the firing showed a very special relationship with the local field potential, which had a predominant theta rhythm (5 to 8 Hz): As the rat moved through the place field for a particular neuron, the cell fired at a fractionally earlier time point during each oscillation of the theta rhythm Another way of putting this is that the cell’s firing was at a very slightly higher frequency than the field oscillation Its firing appeared to “progress” forward

moti-as the animal moved forward, and this gave rise to the term “phmoti-ase sion” for this phenomenon Thus the position of the rat was represented not only by activity of particular neurons, but also by the rhythm of their firing.The second study was of recordings in cat visual cortex made by Singer and Colleagues, of the responses of groups of neurons to presentations of bars of light (42) What Singer and his team did was identify neurons whose receptive fields were aligned, such that they could be stimulated by flashes

progres-of light, either independently with two short bars or together, by a single tended long bar of light They found that both approaches increased the fir-ing of the pair of neurons, but when they were activated together by a single

ex-Sleep spindles and SWDs also appear to differ in the relative involvement

of the reticular and thalamic neurons In vivo recordings in cats (29) and

the GAERS rats (“Generalized Absence Epilepsy Rats of Strasbourg”) (30)

suggest that the dominant bursting in SWDs occurs in the reticular nucleus

neurons, possibly secondary to some initial event in cortex (31) The burst of

reticular firing creates a flurry of inhibitory synaptic potentials in thalamic

neurons, which consequently fire only single spikes or are silent, unlike their

behavior during sleep spindles (29,32) The failure of the burst of inhibitory

postsynaptic potentials to deinactivate the Ca2+ channels in the thalamic

neurons in these recordings has been attributed to a pathologically high level

of tonic inhibitory currents (these “tonic” currents pass through GABAA

channels that are persistently open—this may happen because there remain

low levels of GABA in the CSF, and certain receptor subtypes, notably those

that include δ-subunits, do not inactivate), which act to clamp the membrane

potential above the level needed for deinactivation of the voltage-gated Ca2+

channels (33) Intense cortical pyramidal and interneuron firing is also

ob-served during the spike component of the pathologic discharges, which is

not present in spindles (34) The two rhythms may be linked by the

phenom-enon of augmenting potentials in thalamocortical circuits; stimulation

any-where in the circuit at 10 Hz leads to incremental potentials, and ultimately

to spike and wave discharges very similar to those in absence seizures (35)

Thus, both these rhythms involve burst firing in the thalamic and reticular

nucleus, but differ in the relative involvement of the two nuclei and cortex

The rhythms may be influenced by neuromodulation (36), such as

choliner-gic projections from the brainstem and basal forebrain, which act by

mod-ulating the various active conductances, and in particular Ih (26) In both

cases however, it is easy to see how the intensity of these rhythms, cycling

between the thalamus and the reticular nucleus, may disrupt the flow of

sen-sory information through this relay, during both sleep and absence seizures

We have provided just a brief outline of these fascinating thalamic behaviors,

which is covered in far greater detail in more specialist monographs (35,37)

OSCIllATIOnS

One of the most important tasks performed by electroencephalographers

interpreting clinical EEG studies is the identification of visually prominent

rhythmic patterns, or oscillations These oscillations often provide clinically

useful markers of both normal and pathologic activity The first

oscilla-tion noted in human EEG was the 8 to 12 Hz posterior dominant (alpha)

rhythm The presence of this rhythm during wakefulness and its attenuation

Figure 1.3: structure and interaction of brain oscillations A: amplitude spectrum of eeG recorded from a

lateral frontal subdural grid electrode during wakefulness, with frequency ranges shown as color-coded bars

Note the near linear decrease in amplitude with frequency shown on a natural logarithm scale B: schematic

of hierarchical, phase-amplitude coupling in eeG The top trace illustrates the typical observation: oscillations recorded in the brain are normally complex mixtures of components at different frequencies The traces below illustrate the individual oscillatory components in the delta, theta and gamma band that comprise the com- posite waveform Gamma oscillatory amplitude varies with the phase of the underlying theta oscillation, and

theta oscillatory amplitude varies with the phase of the underlying delta oscillation C: hierarchical coupling

in human microelectrode data recorded from lateral temporal neocortex during an auditory task (Besle and schevon, unpublished) a modulation index (56) was computed for phases between 0.1 and 20 hz, and for both amplitudes between 1 and 250 hz and the amplitude of multiunit activity (Mua, above), filtered 500 to 5

khz and rectified significant phase-amplitude coupling was found between delta phase and Mua (A), gamma

(B), and theta (C) amplitudes, and between alpha phase and both Mua (D) and high gamma amplitude (E)

color version is available online.

time (s)

Comp 1.5 Hz

7 Hz

40 Hz 0

B

long bar of light, that is to say they were responding to the same object, their

firing became synchronized at gamma frequencies The implication was that

gamma synchronization was how the brain represents “unity”—that activity

in these disparate neurons were reflecting one and the same stimulus This

has proved to be a very influential observation, triggering many subsequent

studies, but is limited in that it is just a correlation; what we need now is

to be able to modulate these rhythms, either to create or destroy the

bind-ing oscillation, so that we can really demonstrate causality between gamma

rhythms and the so-called “binding problem.” With the development of

optogenetic techniques, we may now have the tools to test Singer’s able hypothesis

remark-The Structure of EEG

Cerebral signals recorded as EEG contain superimposed oscillations in a wide range of frequencies (Fig 1.3) In humans, recorded oscillations range from 0.05 to 600 Hz, divided into fixed frequency bands At the high end

of the spectrum, above approximately 150 to 200 Hz, oscillations begin to

excellent tool for detecting brain rhythms with fine temporal resolution, but its usefulness for locating the source of these rhythms is often limited We will revisit this point later on, in the context of seizure recordings.

In previous sections, we described the electrical currents induced by firing neurons, and gave examples of EEG rhythms that are generated by bursting cells in the thalamus We now shift our focus to the EEG oscillations them-selves Earlier, we discussed several mechanisms by which neuronal firing can be synchronized and regulated to generate rhythmic extracellular field potentials, such as shunting inhibition from basket cells onto perisomatic pyramidal neurons Conversely, slow oscillations themselves may influence the firing probability of a neuron through their effect on endogenous electri-cal fields, as described previously in the discussion of ephaptic interactions The best described of these slow oscillations at the cellular and network levels are transitions from a relatively hyperpolarized membrane potential generally below −70 mV, when there is little firing activity, to a more depo-larized state around −60 mV and far higher rates of firing These two states are conventionally referred to as “down” and “up” states, respectively Such activity is seen in many anesthetized states, and also during sleep, and be-cause many neurons switch in unison, the transitions between up and down states are readily seen in EEG recordings as a delta rhythm (roughly 0.5

to 2 Hz) The synchrony appears far more global in the anesthetized state, whereas during sleep, evidence suggests that the transitions propagate across the cortex as a wave, and are only synchronized locally, over a few square centimeters

The transition from down to up state can be triggered by thalamic puts It is also likely to occur spontaneously within cortical networks, aris-ing from activity reverberating within the extensive recurrent connections The greatly increased firing during the up state is modulated broadly in the gamma frequencies (20 to 80 Hz) (54), which is evident both in the statistics

in-of population firing and also in the pattern in-of synaptic currents recorded in individual cells This is an excellent example of higher frequency rhythms being nested within lower-frequency oscillations Again, this rhythm is likely

to arise simply from the pattern of recurrent connections with their acteristic synaptic latencies and kinetics What terminates these up states is less clear, although obvious possibilities include a progressive synaptic de-pression, the kinetics of NMDA receptors and the consequent time course

char-of NMDA plateau potentials, increasing levels char-of Ca2+, which in turn raise

K+ conductance by opening Ca2+-gated K+ channels, and changes in other intrinsic neuronal conductances

The clearest examples of spontaneous EEG oscillations with up and down states are the physiological rhythms that occur during sleep, as we discussed

overlap with multiunit activity, or the extracellular signatures of action

po-tentials, rather than synaptic currents Signal amplitude is much higher in

the lowest frequency bands, and decreases with frequency in a steep “1/f”

curve This inverse relationship of signal amplitude to frequency is

char-acteristic of many biological signals A similar inverse relationship also

ex-ists between the spatial extent of oscillations and frequency; as alluded to

earlier, higher frequency oscillations are restricted to small cortical volumes,

while low frequencies are more broadly distributed (39)

EEG recorded from the scalp is typically limited to frequencies under

about 30 Hz, due to attenuation of the recorded signals by the skull and

in-tervening layers of tissue (43), as well as interference from extracerebral

elec-trical sources such as muscle activity Higher frequencies are preferentially

screened out due to the smaller amplitude of the high-frequency signals, and

due to their limited spatial extent, which causes signal to be lost due to the

effects of averaging across the listening sphere of scalp electrodes (44) For

example, interictal epileptiform discharges are rarely detected in scalp EEG

if they encompass less than 6 cm2 of cortex (45) It is not surprising,

there-fore, that EEG recorded from electrodes implanted onto the cortical surface

(e.g., subdurally) is far better than scalp electrodes at detecting gamma and

high gamma oscillations (up to 150 to 200 Hz) Frequencies above 200 Hz

are dominated by action potentials, or the extracellular signatures of firing

neurons To detect action potentials, which are most prominent in

neocorti-cal layers 3 to 5 about 1 mm below the cortineocorti-cal surface (46), we need to be yet

more invasive Action potentials can be recorded by tiny, high-impedance

electrodes inserted into the brain parenchyma, such as microwires added to

the end of a clinical depth array (47,48), or the microelectrodes built into the

“Utah” array (9–1) Magnetoencephalography, which records from sensors

on the scalp but uses magnetic signals that are less prone to attenuation by

skull and scalp tissue (52), is therefore more sensitive to high frequencies

than is scalp EEG, but is subject to a similar spatial averaging effect

The Relationship between Oscillations and Cellular Activity

In the 1930s, Bishop proposed that EEG is the direct result of rhythmic

fluctuations in neuronal activity (53) Specifically, postsynaptic potentials,

generated from neuronal ensembles firing in synchrony, induce extracellular

potential fluctuations that summate to create waves that are the substrate of

EEG signals These waves may appear at sites far removed from the location

of the presynaptic neurons, due to rapid and wide distribution of synaptic

potentials down axonal pathways This is a fundamental property that is

important to keep in mind when interpreting clinical studies: EEG is an

that would otherwise cause unwanted interference High excitability ods, that is “up” states, are more likely to occur when an input stimulus is expected, leading to improved accuracy and response time, while cortical response is reduced during “down” states (57,61).

peri-At the top of the oscillatory hierarchy, it appears that gamma activity, especially high gamma (>80 Hz), provides a reliable index of colocated pop-ulation firing (62,63) (Fig 1.3) In addition to the work already described, there is evidence from related fields to support this view Increases in in-duced gamma power (>40 Hz) during motor and language tasks in patients implanted with subdural grid electrodes has been found at sites positive for the same functions identified by electrocortical stimulation mapping (64,65) Increased high gamma power is also positively correlated with high values

of fMRI (BOLD) signals (66), which are presumed to reflect task-related cortical activation This tight relationship has proven to be advantageous for applications such as creating control signals from EEG for a brain- computer interface (67), and it will also be relevant in the discussion of high-frequency oscillations in the next section

CEllulAR BASIS OF EPIlEPTIFORm ACTIvITy neural Activity during Epileptiform discharges

The primary pathologic EEG findings in epilepsy are transient, high- amplitude deflections in the local field potential, typically on the order of hundreds of microvolts and seen over several square centimeters of cortex

At the peak of the deflection, there is a burst of intense, hypersynchronous multiunit activity, which is why epileptiform discharges appear “sharp.” Following this sharp peak, there is an after-going slow wave or amplitude attenuation, with little or no multiunit activity These may accompany a clin-ical seizure, or occur during the quiescent interictal period between seizures Interictal discharges most often occur as isolated events, while discharges that are part of an ongoing seizure generally appear in rhythmic trains The intracellular signature of both ictal and interictal discharges is thought to be the paroxysmal depolarizing shift (PDS), first described in the early 1960s

in an in vitro penicillin (68) and in vivo freeze-induced lesion (68) seizure models Since these early descriptions of the PDS, it has been demonstrated

in many animal models of both acute and chronic seizures (for detailed view, see de Curtis and Avanzini, 2001 [69]) Indeed, the PDS appears to be

re-a defining fere-ature of seizures in these models

The most notable feature of the PDS is that the cell receives such a strong excitation that following each action potential, the cell remains in a relatively

previously, and also pathologic slow waves The delta rhythms of pathologic

focal or diffuse slow activity, as opposed to the slow waves that characterize

stage 3 sleep, appear to be generated from the cerebral cortex, and their

pres-ence in the waking EEG is usually ascribed to underlying white matter

abnor-mality or cortical deafferentation Recordings from large deafferented cortical

slabs taken from cats revealed spontaneously synchronized slow rhythms, in

which neurons were hyperpolarized during positive peaks of the field

poten-tial, and depolarized during negative peaks, at which time bursts of action

potentials were seen The EPSPs (excitatory postsynaptic potential)

propa-gated at speeds of 10 to 100 mm/second, to recruit nearby neurons (55) Thus,

rhythmic oscillations with up and down states may have either a thalamic or

cortical origin

hierarchical Phase-Amplitude Coupling

Oscillations recorded by EEG can be arranged in a nested structure, in

which the amplitude envelope of a higher frequency rhythm modulates (or

is modulated by) the phase of a lower-frequency rhythm, as described

ear-lier for gamma activity nested within the delta-waves (Fig 1.3) An example

of such cross-frequency coupling that is often seen in clinical recordings

is the so-called “sinusoidal alpha rhythm,” or a spontaneous modulation

of the posterior dominant (alpha) rhythm amplitude that appears to vary

in a very slow (<1 Hz) rhythm The amplitude envelope of high gamma

(80 to 150 Hz) activity has been shown to be maximal during the troughs

of theta or alpha oscillations at many recording sites in human

intracra-nial EEG (56) Thus, there are good experimental examples of gamma

having a phase-amplitude relationship with almost all lower-frequency

rhythms Taking this concept one step further, a three-tiered hierarchy has

been described in recordings from macaque sensory cortices in which delta

phase modulates theta amplitude, theta phase modulates gamma (30 to

50 Hz) amplitude, and gamma phase modulates stimulus-averaged

multi-unit activity (57,58)

Entrainment of delta rhythms by external stimuli has also been

demon-strated in human intracranial EEG While these stimuli can be simple

sen-sory inputs (59), it is also possible to demonstrate entrainment following a

nested theta and delta cadence, by stimuli as complex as spoken language

Notably, experiments combining multiple stimuli, for example a “cocktail

party” with two simultaneous conversation streams, showed that the

en-trainment follows the conversation to which the listener is paying attention

(60), entrainment has been postulated to be an effective mechanism for

fine-tuning attention to follow behaviorally relevant stimuli, and ignore stimuli

Figure 1.4: Burst firing correlated with field oscillations in normal and pathologic tions (panels A and B adapted from Timofeev i, steriade M Neocortical seizures: initia-

condi-tion, development and cessation Neuroscience 2004;123:299–336) A: slow wave sleep in

anesthetized cat, with simultaneous intracellular recordings from a neuron in cortical area

4 and thalamocortical (TC) neuron in the ventrolateral (Vl) nucleus, together with surface and depth eeG from area 4 The depth-positive (upward) eeG waves are associated with hyperpolarization of cortical and thalamic cells, whereas the depth-negativities are asso- ciated with cortical depolarization and action potentials, followed by a rebound spike-

burst (expanded view, arrow) in the TC neuron B: Paroxysmal depolarizing shifts during

a spontaneous neocortical seizure in anesthetized cat eeG is shown in the top trace, with fast ripples (80 to 300 hz) in the second trace The fast ripple correlates with intracellularly

recorded spike bursts C: human in vivo microelectrode recording of interictal epileptiform

discharges (schevon, Columbia univ unpublished) fast activity can be seen over the

nega-tive eeG peaks (top), correlating with multiunit activity (300 to 3 khz, middle), and detected action potentials (bottom raster) Note the disorganized appearance of the firing bursts

accompanying the discharge peaks, compared to the spontaneous firing at other times The change in action potential shape, combined with likely distortion and/or contamination

of the signal from the filtering process, interferes with unit detection and greatly reduces its sensitivity.

depolarized state, and consequently, there is incomplete deinactivation of

the population of Na+ channels With many Na+ channels not being

“acti-vatable,” subsequent action potentials are smaller, and also broader,

reflect-ing a lower K+ conductance too (Figs 1.4 and 1.5) As a result, the burst

of firing, superimposed on the negative peak of the large depolarizing tential, is characterized by a progressive change in the shape of the action potentials, and ultimately, somatic recordings often show a depolarizing blockade of firing Theoretic studies show that even when the soma is in a

po-FIGuRE 1.5: Gamma rhythms are dictated by fast-spiking interneurons A: Cellular firing patterns during gamma rhythms recorded in two different brain slice preparations Panel Ai (adapted from shu Y, hasenstaub

a, McCormick Da Turning on and off recurrent balanced cortical activity Nature 2003;423:288–293) shows an

up state in a ferret brain slice, induced and terminated by white matter stimulation Panel Aii (adapted from

ainsworth M, lee s, Cunningham Mo, et al Dual gamma rhythm generators control interlaminar synchrony in

auditory cortex J Neurosci 2011;31:17040–17051) shows persistent gamma rhythms induced in auditory cortex

by bathing the brain slice in kainate Note the strong depolarizing drive reflected in the “pacemaker” voltages,

which follow each iPsC B: firing patterns recorded during gamma rhythms in vivo, in ferret prefrontal cortex

(adapted from hasenstaub a, shu Y, haider B, et al inhibitory postsynaptic potentials carry synchronized

frequency information in active cortical networks Neuron 2005;47:423–435) The fast-spiking interneurons fire

most cycles, in contrast to regular spiking cells (presumptive pyramidal cells), which fire only intermittently There is a marked dip in firing probability of the pyramidal cells shortly after the firing of the fast-spiking in- terneurons, consistent with the model of gamma that has been derived primarily from brain slice models, and

which is schematized in (C) The schematic shows a pattern of large rhythmic iPsCs onto pyramidal cells, which

then have a narrow window of opportunity to fire between these iPsCs if the frequency of iPsCs increases, however, this window progressively shuts, and the network is suppressed locally in this case, an extrinsic excit- atory drive, such as arising from an adjacent cortical territory, may still provide a strong depolarization, thereby shifting e m away from e GaBa and creating a driving force for current through these large oscillatory inhibitory conductances We predict that this “relatively pure” inhibitory oscillation may have a very distinctive spatial distribution reflecting an inhibitory surround mechanism, although this remains to be tested experimentally

Bi The fast-spiking interneurons Bii There is a marked dip.

field potential deflection Several different synchronization mechanisms have been proposed, including the release of neuroactive substances from glia (5–7), short-range recurrent excitation, release of feedforward inhibition (78), ephaptic interactions resulting from the effect of the large extracellular currents on membrane potential (79,80), electrical coupling between pyra-midal cells via gap junctions (81,82), and low-calcium field bursts (10,83).Despite the universal association of the PDS with animal epileptic dis-charges and its clear extracellular signature, it has rarely been reported in

in vivo recordings of seizures from microelectrodes in humans (for obvious reasons, these invariably are extracellular recordings) Recent results have finally begun to shed light on the reasons for this apparent conundrum The development of microelectrode arrays approved for use in humans has al-lowed us to link multiunit activity (action potentials from many neurons) recorded over a small patch of cortex to standardly interpreted, clinically relevant EEG events—an unprecedented level of detail It is now appar-ent that the extent of cortex that displays hypersynchronous burst firing is often smaller than has hitherto been understood from conventional EEG recordings This is such an important point that is worth considering in more detail

lessons from microelectrode Arrays:

Ictal discharges and the “Ictal Penumbra”

Epileptiform events that correspond closely to the extracellular signature of the PDS have been detected in a small number of patients during seizures (84) and interictal discharges (49,85) recorded with microelectrode arrays The low-frequency signals recorded from the microelectrodes correspond very closely to those recorded from the overlying subdural grid electrodes Critically, though, the multielectrode arrays could also record action poten-tials, and so for the first time, we have a tool that allows a direct comparison between local firing and the EEG over a spatially extended territory These multielectrode recordings showed unequivocally that there can be a large discrepancy between the low-frequency (EEG) signal, and the local level of neuronal firing (51), and that this is a routine occurrence at particular loca-tions during the extreme conditions of an electrographic seizure In short, a large EEG signal does not necessarily equate with local firing This happens where there are large synaptic currents that are not being translated into postsynaptic firing

The large amplitude, rhythmic oscillations that are traditionally used to fine the onset and progression of a seizure were visible in all electrodes in the

de-depolarizing block, there may still be firing down the axon, away from the

excitatory drive that is located on the dendrites Thus an apparent somatic

depolarizing block should not be taken to mean that there is no axonal

out-put from these cells This change in action potential shape distinguishes the

pathologic PDS from the physiological “up” state, in which likewise there

is a sustained, and often rhythmically repeated, depolarization, with bursts

of action potentials superimposed, but the individual action potentials all

have the same shape The primary difference, however, is probably simply

the intensity of the excitatory synaptic drive The PDS then terminates with

a period of hyperpolarization and suppressed neural firing, resulting in

at-tenuation of local field potentials This after-hyperpolarization may

contrib-ute to the periodicity of epileptiform discharges during seizures and other

situations, such as periodic discharges in acutely an injured brain

How PDSs are initiated remains unclear At least part of the difficulty

discerning the mechanism is that there are many possible ways this might

happen It is appropriate to distinguish PDSs that occur during a full

ic-tal event, or the afterdischarges during the clonic phase of a seizure or

fol-lowing electrical stimulation, from those that are interictal or at the very

start of a seizure The full ictal PDSs and afterdischarges are likely to arise

simply from the coordinated synaptic drive from other neurons Computer

modeling work has suggested that the transition from sustained, tonic firing

to the clonic bursting behavior may arise from the interaction between

fir-ing rates and synchrony, combined with slowed firfir-ing under strong synaptic

drive (70) Once the seizure is established, there are likely to be multiple

“re- entrant” loops (similar to the origin of certain pathologic heartbeats)

Newly recruited territories, which are still at the tonic phase of firing, may

also act as a pacemaker for territories that have already entered the clonic

phase This is the case in brain slices, when propagation is essentially

one-dimensional, thereby allowing the visualization of activity propagating both

forward and backward from the moving ictal wavefront (71)

Regarding PDSs that arise de novo (interictal events, or those that

initi-ate a seizure), early theories focused on the concept of “epileptic neurons.”

This concept developed from studies showing intrinsically generated

syn-chronized bursting in multiple types of neuronal populations in neocortical

layers IV–V and the CA3 layer of hippocampus (72,73), and that such

burst-ing can entrain the occurrence of PDSs, albeit in a disinhibited brain slice,

which generated PDSs spontaneously (74) The intensity of the depolarizing

drive, which underlies the PDS, however, suggests that its development

re-quires synchronized neural activity in the local network, hence its

associa-tion with large populaassocia-tion bursts and the characteristically high-amplitude,

multielectrode array, and also simultaneously over an extended region

sam-pled by the overlying subdural grid Volume conduction, or spread of

cur-rent in a surrounding field by passive electrical properties of the extracellular

medium, has often been advocated as an explanation for the large field of

these discharges However, the fractional delays and directional asymmetry

(49,86,87) instead indicate that they reflect the rapid axonal distribution of

postsynaptic currents Similarly rapidly spreading field potential fluxes have

also been demonstrated using current source density analysis in laminar

mi-croelectrode recordings of animal (88) and human (89,90) epileptiform

dis-charges, and in seizure recordings from cats using a dense subdural grid (91)

Figure 1.6: Microelectrode recordings of animal and human seizures A: intracellular recording of an epileptiform event in

the 0 Mg 2+ mouse model, showing the transitions between the pre-recruitment state, with sporadic firing, the tonic phase with sustained firing, and the post-recruitment, clonic phase with burst firing Note the decrease and variability in action poten- tial amplitude in the latter two phases (Trevelyan, Newcastle univ

unpublished) B: Clinical seizure recorded with a 4 × 4 mm

mi-croelectrode array, filtered into “ueeG” (gray) and Mua (black)

data streams Two channels 3 mm apart are shown The same phase transitions seen in the mouse model are evident in Mua, but have little effect on the predominant ictal rhythm that arises from postrecruitment clonic bursting Note the irregular pre- recruitment firing in the bottom channel, and the post-recruit- ment synchronization of burst firing and the ictal alpha-range

rhythm C: Clinical seizure shown in its entirety, recorded from

a site that is never recruited Note the irregular multiunit firing

and lack of correspondence with the low-frequency potentials d:

advance of the ictal wavefront during the clinical seizure shown in panel B, as shown by a striking transient increase in multiunit fir- ing rate The recruited area, or “seizure focus” is to the left of the ictal wavefront, and the pre-recruited area, or “ictal penumbra,”

is to the right The progress of the wavefront, moving at a rate of 0.8 mm/second, is largely hidden when the view is restricted to 2

to 50 hz amplitude (84).

A critical observation was that these rapid reverberations of low- frequency signals (<50 Hz) did not reflect the actual recruitment of neurons to the sei-zure Thus, in contrast to the rapidly moving field potentials, a slow-moving (<1 mm/second) “ictal wavefront” was apparent in the multiunit firing re-corded with the multielectrode arrays (Fig 1.6) The wavefront gradually in-corporated the entire array over a 5- to 10-second period Individual electrodes showed an abrupt start of sustained, intense firing, approximately 30 times greater than seen during the interictal period This tonic firing phase lasted about 1 to 2 seconds, before transitioning into a repeated bursting pattern that was phase-locked to the dominant theta or alpha frequency rhythm (clonic

phase) These recordings thus definitively show that the PDS and its context in

laboratory seizure models is also a feature of human seizures; however, there

are additional implications that are highly relevant to ictal EEG interpretation

Equally informative was the activity pattern at sites ahead of the ictal

wavefront, or in areas that it never reached (Fig 1.6) During the seizure,

activity outside the recruited territory was slightly raised above baseline

(less than fivefold), although many electrodes were almost quiescent

Fur-thermore, this low-level firing was not phase-locked to the low-frequency

rhythm: these neurons were clearly not dancing to the tune of the seizure

Remarkably, this was true even in recording sites located within the seizure

onset zone, as it is standardly defined Thus, there are relatively large

corti-cal areas that display an extreme discrepancy between the large amplitude

EEG signals and the low level of firing locally We suggest the term “ictal

penumbra” for this region of cortex adjacent to the recruited territories (84)

The penumbra may be fluid, shifting as the ictal wavefront progresses

(Fig 1.6) Interestingly, some territories in the penumbra, despite being

bombarded with ictal synaptic barrages throughout, were never actually

re-cruited to the seizure Thus we can make a distinction between areas that are

pre-recruitment and those that are never recruited, but both exist within the

ictal penumbra When repeated seizures were analyzed, the patterns of firing

during recruitment were highly stereotyped, whereas firing in the penumbra

was very variable from seizure to seizure (84) In contrast, the EEG tends to

be stereotyped throughout the penumbra, indicating that this spatial

wide-spread stereotypy at low frequencies is explained by the presynaptic neural

activity within the seizure focus, and not that in the penumbra In summary,

the recruited territories and the ictal penumbra show extreme differences

in the level of neuronal firing, but may appear very similar in the EEG

Indeed, the seizure onset zone, as it is usually defined, appears to include

both regions

Surround Inhibition

Considerable support for this view of an ictal focus and penumbra comes from

animal studies, suggesting that the penumbra region reflects surround

inhibi-tory effects The very first intracellular recordings of cortical neurons made in

intact animals, by Powell and Mountcastle, showed that a focal activation of

cortex induces large inhibitory currents in surrounding territories A decade

later, in 1967, Prince and Wilder showed similar lateral inhibition during

interictal bursts in rat hippocampus (92) They observed that pyramidal cells

adjacent to the focus were bombarded with inhibitory postsynaptic currents

during interictal discharges Other researchers have confirmed and extended these observations (3–5) Optical imaging of a 4-aminopyridine seizure model

in vivo in the rat (96) revealed a slowly expanding area of vasodilation rounded by a narrow band of vasoconstriction, propagating at a speed similar

sur-to that of the ictal wavefront observed in humans (Fig 1.7)

The most detailed view of these inhibitory effects, though, comes from studies of epileptiform propagation in brain slices (22,97) This simple preparation allows a level of control and accessibility that is impossible to achieve in human recordings, and thus presents a far more detailed view of ictal events Epileptiform activity was induced in these studies by remov-ing Mg2+ ions from the extracellular medium, thereby boosting excitatory neurotransmission by reducing the voltage-dependent blockade of NMDA receptors, while preserving inhibition During seizure-like events, a very large excitatory drive extends forward, to sites well ahead of the ictal wave-front (Fig 1.8) This induces an even more powerful feedforward inhibition, mediated through local interneurons, which project onto local pyramidal cells This disynaptic inhibition beats the monosynaptic excitation because the interneurons are recruited very rapidly, and because their axons target the somata of pyramidal cells, whereas the monosynaptic excitatory drive is onto the more distal dendrites of the pyramids Indeed, patch clamp stud-ies suggest that the threshold for firing an action potential may be exceeded 20-fold during this period of restraint (22) Moreover, the low level of firing ahead of the wavefront is not well correlated in different neurons, in exactly the same pattern as was noted in the penumbra of human seizures These experiments therefore provide a cogent explanation of the ictal penumbra as

a region in which neural firing is held off by feedforward inhibition from the active focus—an intrinsic defense against seizures

EEG markers of Ictal Territories:

high-frequency Oscillations

One of the most active areas of investigation in epilepsy neurophysiology

in recent years has been focused on potential biomarkers of epileptic brain areas that can be identified from clinical EEG recordings Historically, the most commonly used method of validating these markers has been the quality of seizure control post-surgery This approach, however, is greatly limited, as it is not possible to know whether the resection was larger than the minimum volume required to control seizures, or whether the length of the follow-up period is adequate Now, the new microelectrode technologies offer a more immediate means of assessing whether territories have been

Figure 1.7: high-frequency oscillations are associated with epileptic pathology A: high-frequency tions recorded through a depth electrode close to the apparent epileptic focus in a human (98) B: similar

oscilla-oscillations can be recorded in brain slices in which synaptic neurotransmission is prevented by removing Ca 2+

ions from the bathing medium in this case, activity is presumed to spread through gap junctions, and support for this comes from the observation that gap-junction blockers reversibly suppress the high-frequency oscilla-

tions (118) C: intracellular recordings (bottom trace) in this model, often show evidence of “spikelets,” which

are like mini-action potentials These are generally taken as evidence of gap-junction coupling, in which a full action potential in one cell induces the spikelet in the coupled, recorded cell without triggering a full action

potential d: The spread of activity in the non-synaptic (0 Ca2+ ) model has been explained in terms of ephaptic interactions, which are likely to be extremely localized, and also by percolation of activity through axons in-

terconnected via gap junctions E: The axonal location of the gap junctions in this model is predicated by the

need to explain very-fast-rising spikelets, which in this case are thought to be back-propagating axonal action potentials which fail to induce a somatic action potential because of an impedance mismatch.

FIGuRE 1.8: activity patterns in the ictal penumbra, and

fol-lowing recruitment to an epileptiform event This schematic

is based on imaging and electrophysiological recordings in

brain slice preparations in which epileptiform events are

in-duced by removing Mg 2+ ions from the bathing medium

The territories ahead of the ictal wavefront are characterized

by very low pyramidal firing despite cells receiving

patho-logically high levels of glutamatergic barrages projecting

forward from the wavefront The restraint of these neurons

appears to be achieved through very intense feedforward

inhibition, because interneurons ahead of the wavefront

may fire at extremely high rates (>300 hz).

recruited or are within the ictal penumbra, which we hope will resolve some

outstanding issues in this field, exploring feasibility of various measures,

and establishing their clinical utility

The “DC shift,” or transient fluctuation in very low-frequency (<0.5 Hz)

power, is perhaps indicative of glial shunting of high extracellular K+ local

to seizure discharges (98,99), and has been recorded in clinical EEG (100)

Another potential biomarker that has received a great deal of attention is

high-frequency oscillations, or transient increased signal amplitude above

80 Hz These were first recorded from microwire bundles in mesial

tem-poral structures in humans (1–3) and in animal models of seizures (4–7)

Most high-frequency oscillations occur with epileptiform discharges,

although they may sometimes be independent Their typical duration is

100 milliseconds or less These may be subdivided into ripples, in which the

predominant frequencies are 80 to 200 Hz, and fast ripples, which contain

frequency components higher than 200 Hz and may extend up to 600 Hz

A hotly debated topic is to what extent these very high-frequency tions reflect the firing rates of individual cells Fast ripples oscillate at fre-quencies well above the apparent maximal firing rate of most pyramidal cells The maximal rate, though, is usually assessed in response to square pulse current injections, and it is possible that patterned inputs, including strong transient inhibition, may allow faster repolarization and thus higher firing rates briefly Furthermore, there are other cell classes in neocortex that can fire at close to these rates, including both fast-spiking interneurons (basket cells) and fast-rhythmic-bursting neurons, also termed “chattering” neurons (125) It is also relevant that these spiking properties may change with experimental conditions (126,127) It is possible then that pro-epileptic states are those that favor high-frequency burst firing.

oscilla-Another possibility is that fast ripples emerge from lower-frequency terns due to locally desynchronized pyramidal cell populations, based on data from an in vitro model (128) and supported by in vivo data from a kainic acid rat model (124) (Fig 1.9) Analyses of these oscillations sug-gest that a sudden doubling of frequency can arise, possibly reflecting out-of-phase firing An intriguing possibility is that this may reflect another well- documented pathologic mechanism in epileptic foci, that of increased intracellular chloride Increases in intracellular chloride cause a positive shift

pat-in the GABAergic reversal potential, thus compromispat-ing neuronal pat-tion, which has been hypothesized to be the primary pathology in several epilepsy syndromes, including postanoxic injury and pharmacoresistant temporal lobe epilepsy (129) This may be especially true for neonatal sei-zures, which can display a paradoxical response to anticonvulsant drugs that act on GABAA receptors, due to different expression of chloride transport channels compared to the adult brain In tissue resected during epilepsy sur-gery, some cells have abnormally low levels of the potassium-chloride co-transporter, KCC2, the main mechanism for extruding chloride in cortical neurons; notably, these cells fire excessively during interictal events (130) In-tense activation of GABAergic channels can overwhelm the chloride clear-ance mechanism even without underlying deficits in KCC2 levels (131–133) Also in simple in vitro models of epilepsy, chloride-sensitive dyes show a pronounced shift toward higher chloride in all neurons with each successive full ictal event (134)

inhibi-The way these shifts in chloride translate to out-of-phase firing at high frequencies is secondary to the feedforward, high-frequency, GABAergic barrages In experimental seizures in cats, if cells are artificially loaded with chloride through the recording electrode, those cells are recruited dispropor-tionately early to the seizure, with intense bursting (135) This is consistent

to that obtained with larger, traditionally defined, resection boundaries

(114) Fast ripples recorded interictally from neocortex may be normal

findings, but may demonstrate broad spatial distribution when they are

pathologic (115) This can explain their detection from surface electrodes

(114), although reduced frequencies may result from spatial averaging and

the filtering effects of the electrode Thus, caution must be exercised when

using large, clinical electrodes to record high-frequency oscillations, as the

frequency ranges of these different types can blur together, and their

gen-eration mechanisms and pathologic significance can be very different

The source of high-frequency oscillations remains controversial; implicit

in this statement is that there are likely to be multiple sources

Hippocam-pal ripples (i.e., the fast components of sharp-wave complexes, generally at

<200 Hz) correlate with population spike bursts in normal and kainic acid

treated rats (107,116,117), but neocortical ripples in epilepsy patients are

of-ten seen without evidence of prominent neural firing in extracellular

record-ings (115) They may be generated by summated inhibitory postsynaptic

potentials (118), reflecting a feedforward inhibition by perisomatic

interneu-rons that restrain pyramidal cell firing (97) This mechanism is thus likely

to give a distinctive spatial pattern of high-frequency oscillations, reflecting

the delineation of active territories brought about by this intense inhibition

A nonsynaptic mechanism has also been proposed based on neural network

modeling (119,120) of the zero-Ca2+ brain slice model (121), reflecting axonal

and somatic action potentials reverberating through a gap-junction coupled

network (Fig 1.7) The model is critically dependent on the degree of

cou-pling between neurons, and increased frequencies of oscillations up into the

fast ripple range (>200 Hz) have been suggested to occur from pathologically

high levels of coupling This model can display escalating activity, leading

ulti-mately to seizure-like behavior, and thus cleverly combines the apparent

asso-ciation of very high-frequency oscillations with ictogenesis It is important to

realize, however, that the mechanism of escalating activity in this model is quite

similar to that also proposed for ephaptic escalation, except that the ephaptic

interactions are likely to be highly localized, whereas gap-junction coupling

may be on distributed axonal processes and thus be over far larger distances

Fast ripples, generally, appear to be a by-product of the burst of

pyra-midal cell firing at the negative peaks of epileptiform discharges, since the

duration of typical action potential waveforms matches the periodicity of

signals in the 200 to 600 Hz range (107,115,122–124) The duration of

ac-tion potentials during a PDS may be increased by the inactivaac-tion of

po-tassium channels due to high levels of excitation, thus bringing the action

potential waveforms into the fast ripple range (72)

display almost instantaneous transitions from complete quiescence to extremely high firing rates in certain conditions These synaptically connected cells were recorded in

a brain slice, bathed in 0 Mg 2+ artificial cerebrospinal fluid (Trevelyan, unpublished) in normal conditions, this inhibitory discharge creates a powerful restraint on local

pyra-midal activity, but in epileptic tissue, a similar burst of firing may give rise to out-of-phase firing (B) in populations of cells with different levels of intracellular chloride (125) C: artificially loading chloride into cells, via the recording electrode, causes them to fire very intensely at times when there is a large, high-frequency inhibitory synaptic drive

(135) Notably, the action potentials in these chloride-loaded cells appear to be at almost 180° phase shift from the firing of cells with low intracellular chloride.

23

AJT is currently a Senior Lecturer at Newcastle University AJT currently holds a fellowship from Epilepsy Research UK CAS is supported by NIH NINDS K08 NS48871

REFEREnCES

1 Berger H Über das Elektroencephalogramm des Menschen [On the

electroencephalo-gram of man] Arch Psychiatr Nervenkr 1929;87:527–570.

2 Hodgkin AL, Huxley AF Currents carried by sodium and potassium ions through the

membrane of the giant axon of Loligo J Physiol 1952;116:449–472.

3 Hodgkin AL, Huxley AF, Katz B Measurement of current-voltage relations in the

mem-brane of the giant axon of Loligo J Physiol 1952;116:424–448.

4 Hodgkin AL, Huxley AF The components of membrane conductance in the giant axon

of Loligo J Physiol 1952;116:473–496.

5 Hodgkin AL, Huxley AF The dual effect of membrane potential on sodium

conduc-tance in the giant axon of Loligo J Physiol 1952;116:497–506.

6 Jack JB, Noble D, Tsien RW Electric current flow in excitable cells New York, NY:

Oxford University Press, 1983:534.

7 Schiller J, Major G, Koester HJ, et al NMDA spikes in basal dendrites of cortical

pyra-midal neurons Nature 2000;404:285–289.

8 Herron CE, Lester RA, Coan EJ, et al Frequency-dependent involvement of NMDA

receptors in the hippocampus: a novel synaptic mechanism Nature 1986;322:265–268.

9 Kleckner NW, Dingledine R Requirement for glycine in activation of NMDA-receptors

expressed in Xenopus oocytes Science 1988;241:835–837.

10 Jefferys JG Nonsynaptic modulation of neuronal activity in the brain: electric currents

and extracellular ions Physiol Rev 1995;75:689–723.

11 Bikson M, Inoue M, Akiyama H, et al Effects of uniform extracellular DC electric fields

on excitability in rat hippocampal slices in vitro J Physiol 2004;557:175–190.

12 Anastassiou CA, Perin R, Markram H, et al Ephaptic coupling of cortical neurons Nat Neurosci 2011;14:217–223.

13 Fröhlich F, McCormick DA Endogenous electric fields may guide neocortical network

activity Neuron 2010;67:129–143.

14 Galarreta M, Hestrin S A network of fast-spiking cells in the neocortex connected by

electrical synapses Nature 1999;402:72–75.

15 Gibson JR, Beierlein M, Connors BW Two networks of electrically coupled inhibitory

neurons in neocortex Nature 1999;402:75–79.

16 Takehara-Nishiuchi K, McNaughton BL Spontaneous changes of neocortical code for

associative memory during consolidation Science 2008;322:960–963.

17 Rouach N, Koulakoff A, Abudara V, et al Astroglial metabolic networks sustain

hip-pocampal synaptic transmission Science 2008;322:1551–1555.

18 Mitzdorf U Current source-density method and application in cat cerebral cortex:

inves-tigation of evoked potentials and EEG phenomena Physiol Rev 1985;65:37–100.

19 Freund TF, Buzsaki G Interneurons of the hippocampus Hippocampus 1996;6:347–470.

20 Somogyi P, Klausberger T Defined types of cortical interneurone structure space and

spike timing in the hippocampus J Physiol 2005;562:9–26.

21 Vu ET, Krasne FB Evidence for a computational distinction between proximal and distal

neuronal inhibition Science 1992;255:1710–1712.

22 Trevelyan A, Sussillo D, Watson B, et al Modular propagation of epileptiform activity:

evidence for an inhibitory veto in neocortex J Neurosci 2006;26:12447–12455.

with other descriptions of the feedforward inhibition ahead of the ictal

wavefront Chloride loading also occurs naturally though, as noted above

COnCluSIOn

In order to make good use of EEG to identify and localize both normal

and pathologic brain activity, it is important to understand what features

of these recordings reflect the neuronal activity that drives the extracellular

signals While action potentials do not generally contribute directly to the

EEG, they may in unusual circumstances generate signals in the high gamma

band—such as the highly synchronized, intense burst firing that

accompa-nies the clonic phase of a seizure Further, the summated postsynaptic

po-tentials that result from the synchronized firing of neuronal populations can

be detected at a distance, in much the same manner in which a radio receiver

picks up a tower broadcast

In this chapter, we have focused particularly on focal seizure localization,

since this is one of the most important clinical applications of EEG—and

is possibly the facet of EEG that is the least understood In vitro and in

vivo studies provide a coherent view of how seizures spread, but must be

reconciled with observations from clinical experience EEG recordings of

clinical seizures often appear to spread very rapidly, follow anatomical and

functional connections (136), and may involve disparate cortical sites Our

previous discussion adds another dimension to the story: that is, the

appar-ent large-scale pattern of seizure spread in humans may be in part explained

by broadcasting of postsynaptic currents from relatively small seizure foci

These results allow us to envision new ways of mapping the trajectories

of seizures through the brain, which can be translated to relatively low-

resolution techniques

A similar understanding can be applied to the study of high-frequency

os-cillations Much work still needs to be done to link the clinical studies to those

aimed at uncovering their neural mechanisms The abrupt boundaries of

epileptic activity that have been defined using animal models, where intense

firing in the epileptic focus contrasts sharply with restrained firing in the

sur-rounding “penumbra,” cannot be readily appreciated in the raw EEG

Limit-ing the view to high frequencies makes it possible to distLimit-inguish these areas

The recording technology that has recently become available, and that

continues to be developed at an ever-increasing pace, promises a more

nu-anced reading of EEG signals, relating them directly to particular patterns

of cortical activity The concepts that are now being developed may usher in

a complete change in how we define and understand seizures

23 Cobb SR, Buhl EH, Halasy K, et al Synchronization of neuronal activity in

hippocam-pus by individual GABAergic interneurons Nature 1995;378:75–78.

24 Ainsworth M, Lee S, Cunningham MO, et al Dual gamma rhythm generators control

interlaminar synchrony in auditory cortex J Neurosci 2011;31:17040–17051.

25 Zaman T, Lee K, Park C, et al Cav2.3 channels are critical for oscillatory burst

dis-charges in the reticular thalamus and absence epilepsy Neuron 2011;70:95–108.

26 McCormick DA, Pape HC Properties of a hyperpolarization-activated cation

cur-rent and its role in rhythmic oscillation in thalamic relay neurones J Physiol 1990;431:

291–318.

27 Robinson RB, Siegelbaum SA Hyperpolarization-activated cation currents: from

mol-ecules to physiological function Annu Rev Physiol 2003;65:453–480.

28 Steriade M, McCormick DA, Sejnowski TJ Thalamocortical oscillations in the sleeping

and aroused brain Science 1993;262:679–685.

29 Steriade M, Contreras D Relations between cortical and thalamic cellular events during

transition from sleep patterns to paroxysmal activity J Neurosci 1995;15:623–642.

30 Slaght SJ, Leresche N, Deniau J-M, et al Activity of thalamic reticular neurons

dur-ing spontaneous genetically determined spike and wave discharges J Neurosci 2002;22:

2323–2334.

31 Neckelmann D, Amzica F, Steriade M Spike-wave complexes and fast components of

cortically generated seizures, III: synchronizing mechanisms J Neurophysiol 1998;80:

1480–1494.

32 Contreras D, Steriade M Cellular basis of EEG slow rhythms: a study of dynamic

corti-cothalamic relationships J Neurosci 1995;15:604–622.

33 Cope DW, Di Giovanni G, Fyson SJ, et al Enhanced tonic GABAA inhibition in typical

absence epilepsy Nat Med 2009;15:1392–1398.

34 Kandel A, Buzsaki G Cellular-synaptic generation of sleep spindles, spike-and-wave

dis-charges, and evoked thalamocortical responses in the neocortex of the rat J Neurosci

1997;17:6783–6797.

35 Steriade M Neuronal substrates of sleep and epilepsy, 1st ed Cambridge, England:

Cambridge University Press, 2003:536.

36 Steriade M, Gloor P, Llinás RR, et al Report of IFCN Committee on Basic

Mecha-nisms Basic mechanisms of cerebral rhythmic activities Electroencephalogr Clin

Neuro-physiol 1990;76:481–508.

37 Sherman SM, Guillery RW Exploring the thalamus and its role in cortical function, 2nd

ed Cambridge, MA: MIT Press; 2009:512.

38 Pineda JA The functional significance of mu rhythms: translating “seeing” and

“hear-ing” into “doing.” Brain Res Brain Res Rev 2005;50:57–68.

39 Buzsaki G, Draguhn A Neuronal oscillations in cortical networks Science 2004;304:

1926–1929.

40 Varela F, Lachaux J, Rodriguez E, et al The brainweb: phase synchronization and

large-scale integration Nat Rev Neurosci 2001;2:229–239.

41 O’Keefe J, Recce ML Phase relationship between hippocampal place units and the EEG

theta rhythm Hippocampus 1993;3:317–330.

42 Gray CM, Singer W Stimulus-specific neuronal oscillations in orientation columns of cat

visual cortex Proc Natl Acad Sci U S A 1989;86:1698–1702.

43 Grieve PG, Emerson RG, Fifer WP, et al Spatial correlation of the infant and adult

electroencephalogram Clin Neurophysiol 2003;114:1594–1608.

44 Pfurtscheller G, Cooper R Frequency dependence of the transmission of the EEG from

cortex to scalp Electroencephalogr Clin Neurophysiol 1975;38:93–96.

45 Tao JX, Ray A, Hawes-Ebersole S, et al Intracranial EEG substrates of scalp EEG

inter-ictal spikes Epilepsia 2005;46:669–676.

46 Schroeder CE Defining the neural bases of visual selective attention: conceptual and

empirical issues Int J Neurosci 1995;80:65–78.

47 Jacobs J, Kahana M, Ekstrom A, et al Brain oscillations control timing of single-neuron

activity in humans J Neurosci 2007;27:3839–3844.

48 Cerf M, Thiruvengadam N, Mormann F, et al On-line, voluntary control of human

temporal lobe neurons Nature 2010;467:1104–1108.

49 Schevon CA, Goodman RR, McKhann G, et al Propagation of epileptiform activity on

a submillimeter scale J Clin Neurophysiol 2010;27:406–411.

50 Hochberg L, Serruya M, Friehs G, et al Neuronal ensemble control of prosthetic devices

by a human with tetraplegia Nature 2006;442:164–171.

51 Truccolo W, Donoghue JA, Hochberg LR, et al Single-neuron dynamics in human focal

epilepsy Nat Neurosci 2011;14:635–641.

52 Cappell J, Schevon C, Emerson RG Magnetoencephalography in epilepsy: tailoring

interpretation and making inferences Curr Neurol Neurosci Rep 2006;6:327–331.

53 Bishop G Cyclic changes in excitability of the optic pathway of the rabbit Am J Physiol

1932;103:213–224.

54 Le Van Quyen M, Staba R, Bragin A, et al Large-scale microelectrode recordings of

high-frequency gamma oscillations in human cortex during sleep J Neurosci 2010;30:

7770–7782.

55 Timofeev I, Grenier F, Bazhenov M, et al Origin of slow cortical oscillations in

deaf-ferented cortical slabs Cereb Cortex 2000;10:1185–1199.

56 Canolty RT, Edwards E, Dalal SS, et al High gamma power is phase-locked to theta

oscillations in human neocortex Science 2006;313:1626–1628.

57 Lakatos P, Karmos G, Mehta AD, et al Entrainment of neuronal oscillations as a

mech-anism of attentional selection Science 2008;320:110–113.

58 Lakatos P, Shah AS, Knuth KH, et al An oscillatory hierarchy controlling neuronal

excitability and stimulus processing in the auditory cortex J Neurophysiol 2005;94:

1904–1911.

59 Besle J, Schevon CA, Mehta AD, et al Tuning of the human neocortex to the temporal

dynamics of attended events J Neurosci 2011;31:3176–3185.

60 Zion Golumbic EM, Poeppel D, Schroeder CE Temporal context in speech

process-ing and attentional stream selection: a behavioral and neural perspective Brain Lang

2012;122:151–161.

61 Schroeder C, Lakatos P The gamma oscillation: master or slave? Brain Topogr 2009;22:

24–26.

62 Ray S, Maunsell JHR Different origins of gamma rhythm and high-gamma

activ-ity in macaque visual cortex PLoS Biol 2011;9:e1000610 doi:10.1371/journal.pbio

1000610.

63 Haegens S, Nácher V, Luna R, et al α-Oscillations in the monkey sensorimotor network

influence discrimination performance by rhythmical inhibition of neuronal spiking Proc Natl Acad Sci U S A 2011;108:19377–19382.

64 Crone N, Hao L, Hart J, et al Electrocorticographic gamma activity during word

pro-duction in spoken and sign language Neurology 2001;57:2045–2053.

65 Miller KJ, Zanos S, Fetz EE, et al Decoupling the cortical power spectrum reveals

real-time representation of individual finger movements in humans J Neurosci 2009;29:

68 Goldensohn ES, Purpura DP Intracellular potentials of cortical neurons during focal

epileptogenic discharges Science 1963;139:840–842.

69 de Curtis M, Avanzini G Interictal spikes in focal epileptogenesis Prog Neurobiol 2001;

63:541–567.