Disinhibition at Feedforward Inhibitory Synapses in Hippocampal Area CA1 Induces a Form of Long-Term Potentiation

by

John Oliver Heal Ormond

A thesis submitted in conformity with the requirements

for the degree of Doctor of Philosophy

Graduate Department of Cell and Systems Biology

University of Toronto

© Copyright by John Oliver Heal Ormond (2009)

Disinhibition at Feedforward Inhibitory Synapses in Hippocampal Area CA1 Induces a Form of Long-Term Potentiation

Doctor of Philosophy (2009). John Oliver Heal Ormond

Graduate Department of Cell and Systems Biology, University of Toronto.

Abstract

One of the central questions of neuroscience research has been how the cellular and molecular components of the brain give rise to complex behaviours. Three major breakthroughs from the past sixty years have made the study of learning and memory central to our understanding of how the brain works. First, the psychologist Donald Hebb proposed that information storage in the brain could occur through the strengthening of the connections between neurons if the strengthening were restricted to neurons that were co-active (Hebb, 1949). Second, Milner and

Scoville (1957) showed that the hippocampus is required for the acquisition of new long-term memories for consciously accessible, or declarative, information. Third, Bliss and Lømo (1973) demonstrated that the synapses between neurons in the dentate gyrus of the hippocampus could indeed be potentiated in an activity-dependent manner. Long-term potentiation (LTP) of the glutamatergic synapses in area CA1, the primary output of the hippocampus, has since become the leading model of due to its dependence on NMDA receptors

(NMDARs), required for spatial and temporal learning in intact animals, and its robust pathway specificity. Using whole-cell recording in hippocampal slices from adult rats, I find that the efficacy of synaptic transmission from

CA3 to CA1 can in fact be enhanced without the induction of classic LTP at the glutamatergic inputs. Taking care not to directly stimulate inhibitory fibers, I show that the induction of GABAergic plasticity at feedforward inhibitory inputs in CA1 results in the reduced shunting of excitatory currents, producing a long-term increase in the amplitude of Schaffer collateral-mediated postsynaptic potentials which is dependent on NMDAR activation and is pathway specific. The sharing of these fundamental properties with classic LTP suggests the possibility of a previously unrecognized target for therapeutic intervention in disorders linked to memory deficits, as well as a potentially overlooked site of LTP expression in other areas of the brain.

ii

Acknowledgements

Performing the experiments contained in this dissertation was a challenging and lengthy

undertaking, but was also a very rewarding experience thanks to all those who made

contributions, both big and small, along the way. First and foremost, I would like to thank my

supervisor and mentor of the past 4 and half years, Dr. Melanie Woodin. I came to the lab with

an incredibly vague idea of what I wanted to do, and Dr. Woodin initially provided me with the

training I needed to get started, and then gave me the freedom and resources to figure out what direction I wanted to go in. She waited patiently for publications to materialize, and when they finally did, she allowed me to participate directly in the submission process, from which I gained very valuable experience that I will no doubt be able to draw upon throughout my career.

For many thought provoking discussions (and for putting up with me), I thank my colleagues in the Woodin lab, Trevor Balena and Brooke Acton. Trevor and Brooke always had insightful ideas to offer during both formal meetings and informal discussions in the lab, and I will miss working alongside them. Additionally, I would like to thank all my colleagues in the

Peever, Buck, Lovejoy, Yeomans, and Stephenson labs who made the third floor of Ramsay

Wright a very pleasant and friendly place to work.

For guiding me through the bureaucratic maze of degree completion and graduation, I thank our graduate coordinator Ian Buglass.

Lastly, I would like to thank all members of my thesis committee, both for taking time from their very busy schedules, as well as for providing very helpful comments and suggestions, and for always asking very thoughtful and thought-provoking questions. The core members of the committee were Dr. Melanie Woodin, Dr. John Peever, and Dr. Min Zhuo; attending various iii

committee meetings in a number of different capacities were Dr. Les Buck and Dr. Richard

Stephenson; and joining us from the University of California at Los Angeles was the external examiner Dr. Dean Buonomano. Without their help, I would no doubt have had to live through another Toronto winter!

iv

Table of Contents

Abstract...... ii Acknowledgements...... iii Table of Contents...... v Table of Figures...... vii Table of Tables...... ix 1. General introduction to hippocampal physiology and synaptic plasticity...... 1 1.1 Opening remarks………………………………………………………...1 1.2 Fast and slow synaptic transmission……………………………………2 1.3 The excitatory synapse…………………………………………………..3 1.4 The inhibitory synapse…………………………………………………..5

1.5 Regulation of the driving force for GABAAR currents………...……..7 1.6 Circuitry of the hippocampus…………...……………………………...8 1.7 Long-term potentiation………………………………………………...10 1.8 Involvement of LTP in memory…………………...…………………..14 1.9 Interneurons in the hippocampus……………………………………..14 1.10 Types of interneurons………………………………………………….16 1.11 Feedforward inhibition………………………………………………...17 1.12 Reports of inhibitory plasticity……………………………………...…20 1.13 Hypotheses………………………………………………………………22

2. Disinhibition Mediates a Form of Hippocampal Long-Term Potentiation in Area CA1...20 2.1 Introduction: Feedforward inhibition in CA1...... 24

2.2 Results: Disinhibition increases the efficacy of excitatory transmission...... 27

2.3 Figures...... 32 2.4 Discussion: Disinhibition-mediated long-term potentiation...... 46

3. A Heterosynaptic Increase in GABAergic Conductance Maintains the Pathway Specificity of Disinhibition-Mediated LTP...... 51

v

3.1 Introduction: Glutamatergic LTP is pathway specific...... 52

3.2 Results: Heterosynaptic increases in GABAAR synaptic conductance confines LTP to the paired pathway...... 54

3.3 Figures...... 59

3.4 Discussion: Both classic and disinhibition-mediated LTP involve the interaction of synapse specific and cell wide plasticity...... 66

4. General discussion of the main findings: A role for disinhibition in memory...... 70 4.1 Major conclusions...... 70 4.2 Feedforward inhibition in vivo...... 72 4.3 Plasticity during theta/gamma and SWRs...... 79 4.4 Future experiments...... 82 4.5 Final remarks...... 86 5. Detailed Materials and Methods...... 87

5.1 Ethics approval...... 87 5.2 Slice preparation...... 87 5.3 Electrophysiology...... 88 5.4 Data analysis...... 90 5.5 Statistics...... 90

6. References...... 92

vi

Table of Figures

Figure 1. Circuitry of the hippocampus. 9

Figure 2. Feedforward inhibition in area CA1. 19

Figure 3. Feedforward inhibition reduces EPSP amplitude. 32

Figure 4. Feedforward inhibition overlaps the EPSP rising phase. 33

Figure 5. EGABA depolarization underlies a large component of mixed LTP. 34

Figure 6. LTP induction leads to simultaneous EGABA depolarization and LTP. 35

Figure 7. APV blocks pairing-induced GABAergic plasticity in single pathway experiments. 36

Figure 8. Depolarization of EGABA is sufficient for LTP expression. 37

Figure 9. Depolarization of EGABA can be induced by as few as 50 pairings. 38

Figure 10. The majority of recorded inhibition in mixed EPSP/IPSP recordings was feedforward. 39

Figure 11. Induction protocol. 40

Figure 12. Depolarization of EGABA contributes to LTP expression- representative recordings. 41

Figure 13. Pairing Schaffer collateral stimulation with postsynaptic spiking results in simultaneous EGABA depolarization and LTP expression. 42

Figure 14. Depolarization of EGABA is sufficient for LTP expression- representative recordings. 43

Figure 15. Disinhibition alone can increase the efficacy with which presynaptic CA3 pyramidal neurons excite postsynaptic CA1 pyramidal neurons. 44

vii

Figure 16. Electrophysiological identification of CA1 pyramidal neurons. 45

Figure 17. Depolarization of EGABA spreads to unpaired synapses, but inhibitory strength is maintained at those sites through increased GABAergic conductance. 59

Figure 18: Pairing-induced LTP is pathway-specific only when feedforward inhibition is intact. 60

Figure 19: Pairing-induced heterosynaptic increases in GABAergic conductance require Ca2+ entry through L-type calcium channels. 61

Figure 20: Increased GABAergic conductance at the unpaired pathway underlies the pathway specificity of mixed LTP. 62

Figure 21: The majority of recorded inhibition in mixed EPSP/IPSP recordings was feedforward. 64

Figure 22: Pathways in the 2-pathway experiments were independent. 65

viii

Table of Tables

Table 1. List of abbreviations...... x

ix

Table 1. List of abbreviations ______

µM micromolar ACSF artificial cerebrospinal fluid AIP autocamtide 2-related inhibitory peptide AMPA α-amino-3hydroxy-5-methyl-4-isoxazole propionic acid AMPAR α-amino-3hydroxy-5-methyl-4-isoxazole propionic acid receptor AP AP5/APV 2-amino-S phosphonopentanoate ATPase adenosine triphosphatase CA cornu ammonis Ca2+ calcium CaMKII calcium/calmodulin-dependent protein kinase II CB cannabinoid CCC cation chloride transporters CCK cholecystokinin Cl– chloride [Cl-] chloride concentration CNQX 6-cyano-7-nitroquinoxaline-2,3-dione CNS central nervous system DG dentate gyrus E-S epsp to spike EC entorhinal cortex – ECl equilibrium potential of Cl

EGABA equilibrium potential of GABAAR-mediated currents EPSPs excitatory postsynaptic potential eSWRs exploratory sharp wave associated ripples GABA γ-amino butyric acid

GABAAR GABAA receptor

GABABR GABAB receptor x

GABACR GABAC receptor GluR glutamate receptor subunit – HCO3 bicarbonate IPSC inhibitory postsynaptic current IPSP inhibitory postsynaptic potential KCC K-Cl cotransporter LFP local field potential LJP liquid junction potential LTD long-term depression LTP long-term potentiation 2+ Mg magnesium mGluR metabotropic glutamate receptor mM millimolar ms millisecond mV millivolt NKCC Na-K-2Cl cotransporter NMDA N-methyl-D-aspartic acid NMDAR N-methyl-D-aspartic acid receptor NR NMDAR subunit O-LM Oriens-Lacunosum moleculare PSD postsynaptic density PSP postsynaptic potential PV parvalbumin REM rapid eye movement RMP resting SPWs sharp waves SWRs sharp wave associated ripples ______

xi

1

1. General introduction to hippocampal physiology and synaptic

plasticity

1.1 Opening remarks

The study of long-term potentiation (LTP) in slices of hippocampal tissue from rodents has become a central component of research into the cellular and molecular mechanisms of memory formation in the mammalian brain. The origins of today’s LTP research can be traced back to Donald Hebb, who proposed that information storage in the brain could occur through the strengthening of the connections between co-active neurons (Hebb, 1949), and to Scoville and Milner, who showed from their study of the amnesic patient H. M. that the hippocampus is required for the acquisition of new long-term memories for consciously accessible, or declarative information (Scoville and Milner, 1957). In 1973, Bliss and colleagues demonstrated that the synapses between neurons in the dentate gyrus of the hippocampus could indeed be potentiated in an activity-dependent manner (Bliss and Lomo, 1973; Bliss and Gardner-Medwin, 1973), thus beginning an explosion of research into the various cellular and molecular aspects of synaptic plasticity, and by hopeful extension, memory. Research in the intervening years has focused on the potentiation of glutamatergic synaptic transmission as the cellular substrate of memory, and indeed, there is evidence linking this phenomenon to memory in intact animals (Morris et al.,

1986; Tsien et al., 1996a, b; McHugh et al., 1996). However, receiving much less attention has been inhibitory plasticity, owing in part to its more recent discovery, as well as uncertainty about how it might be involved in the encoding of memory. In this thesis, I present evidence that

GABAergic plasticity may in fact aid the encoding of memory in rodents in much the same way as glutamatergic plasticity, through synapse specific enhancements of glutamatergic 2

transmission. We begin with a summary of the hippocampal machinery, from the molecular

components to the circuitry, as well as the evidence implicating synaptic plasticity in memory.

1.2 Fast and slow synaptic transmission

Fast synaptic transmission refers to transmission involving presynaptically released

neurotransmitters binding to postsynaptic ionotropic receptors (reviewed by Greengard, 2001).

From neurotransmitter release to receptor binding and then receptor opening takes less than 1

millisecond (ms), and allows the entry or exit of ions for which the receptor is specific. There are

two types of fast transmission, excitatory, activating depolarizing currents which if large enough

lead to cell firing, and inhibitory, involving hyperpolarizing or shunting currents, which inhibit

the postsynaptic firing of action potentials (it should be noted here that shunting refers to

currents that may be depolarizing below action potential threshold, but who’s reversal potential,

described below, is hyperpolarized with respect to action potential threshold).

Slow synaptic transmission, on the other hand, refers to the binding of these same

neurotransmitters to g-protein coupled metabotropic receptors (reviewed by Greengard, 2001).

The receptors themselves do not form channels, but rather, through intracellular signaling

cascades, can open channels (as well as exert other functions, such as inhibiting neurotransmitter

release from the presynaptic terminal), but on a time-scale much too slow to lead to the

correlated firing thought to be important for N-methyl-D-aspartic acid (NMDA) receptor- dependent plasticity induction (discussed below). However, plasticity at certain synapses can rely on activation of metabotropic receptors (reviewed by Anwyl, 2009).

3

1.3 The excitatory synapse

The long-term potentiation originally described by Bliss and Lomo (1973) occurs at

excitatory synapses. These synapses consist of a presynaptic active zone releasing the

neurotransmitter glutamate, and a postsynaptic membrane, termed the postsynaptic density

(PSD), consisting of a combination of various glutamate receptors, including ionotropic AMPA

receptors (AMPARs), NMDA receptors (NMDARs), and kainate receptors, and metabotropic

glutamate receptors (mGluRs) (reviewed by Cattabeni, 1999; Dingledine et al., 1999).

AMPARs are composed of a combination of four different subunits, GluR-A to GluR-D

(or GluR1-4) and are permeable to potassium and sodium (reviewed by Sprengel, 2006). In the

adult brain, most excitatory neurons express GluR-B-containing AMPARs, although the

expression of Ca2+-permeable AMPARs without GluR-B have been described in glial cells and

GABAergic interneurons (Geiger et al., 1995; Muller et al., 1992). AMPAR channels typically

have rapid onset, offset, and desensitization kinetics (reviewed by Gouaux, 2004). The long-term

potentiation originally described by Bliss and Lomo (1973) is due primarily to increased

glutamate release leading to a greater opening of AMPARs, and thus greater postsynaptic

depolarization and action potential firing (Bliss et al., 1987).

Kainate receptors are tetrameric assemblies of subunits GluR5, GluR6, GluR7, KA1 and

KA2 that have a structure similar to those of the other ionotropic glutamate receptors (reviewed

by Isaac et al., 2004). Like AMPARs, they display rapid activation and desensitization

characteristics. However, unlike the other ionotropic glutamate receptors, kainate receptors are targeted to a variety of presynaptic and postsynaptic locations often within the same neuron. At these locations, they perform specific tasks related to the regulation of presynaptic function (of 4

glutamate transmission and transmission mediated by other transmitters, most notably GABA), postsynaptic neurotransmission, or regulation of membrane excitability (Frerking & Nicoll,

2000; Kullmann, 2001). Principal cells and interneurons in CA3 show high levels of expression, but in CA1, only interneurons show much expression.

NMDA receptors (NMDARs) are the most complex glutamate-gated ionotropic receptors

(reviewed by Köhr, 2006). The NMDARs require the co-agonist glycine to open, and are subject to a voltage-dependent block by physiological Mg2+ concentrations which is relieved upon membrane depolarization. They display high Ca2+ permeability, and owing to their voltage- dependent Mg2+ block, act as coincidence detectors, permitting ion flow only when both pre- and postsynaptic cells are excited. Seven NMDAR subunits have been identified: one NR1, four NR2

(A–D), and two NR3 (A, B) subunits. Most native NMDARs appear to function as heterotetrameric assemblies composed of two glycine-binding NR1 and two glutamate-binding

NR2 subunits. NMDARs are therefore NR2-containing receptors in which NR1/NR2 heterodimers appear to be the functional unit (Chen and Wyllie, 2006; Furukawa et al., 2005).

The mGluRs consist of a single subunit, with the N-terminal portion of the receptor containing the glutamate-binding site (O’Hara et al., 1993; Okamoto et al., 1998), and the intracellular domains physically interacting with G-proteins to initiate signal transduction events

(Pin et al., 1994; Gomeza et al., 1996). mGluRs have been implicated in many diverse functions of the mammalian CNS, including the mediation of slow excitatory and inhibitory responses, the regulation of calcium, potassium, and non-selective cation channels, the inhibition and facilitation of transmitter release, the induction of long-term potentiation and depression, the formation of various types of memory, the regulation of trafficking of ionotropic glutamate 5

receptors, the modification of N-methyl-D-aspartate (NMDA) receptor-mediated synaptic transmission and the regulation of neuronal development (reviewed by Ferraguti and Shigemoto,

2006).

1.4 The inhibitory synapse

Most inhibitory synapses in the brain involve transmission mediated primarily by the neurotransmitter GABA [inhibition in sensory and motor reflex circuits in the spinal cord, as well as at certain synapses in the brain stem, cerebellum and retina, is mediated by glycine

(reviewed by Webb and Lynch, 2007)]. The inhibitory synapse, like its excitatory counterpart, contains both ionotropic receptors mediating fast inhibition and metabotropic receptors mediating slow inhibition. The GABAA receptors (GABAARs) are the main ionotropic type, and

- - form channels permeable to chloride (Cl ), and to a lesser extent bicarbonate (HCO3 )(reviewed

by Michels and Moss, 2007). They are members of the nicotinicoid superfamily of ligand-gated ion channels, and are heteromeric pentamers composed of five subunits belonging to a variety of subfamilies. A total of 19 different subunits have been isolated: α1-6, β1-3, γ1-3, δ, ε, π, ρ1-3, and θ. The expression of GABAA receptor subtypes is spatially, regionally and developmentally

regulated, with individual subunits having distinct but overlapping expression patterns (Fritschy

and Mohler, 1995; Laurie et al., 1992). In addition to differential subunit expression throughout

brain regions, the GABAA receptor subunit composition varies between cell types and undergoes

differential subcellular targeting. The majority of GABAA receptor subtypes in the brain contain

diverse α and β subunit variants in combination with the γ or δ subunit. GABAA receptors that contain the γ subunit are usually located on synaptic sites, where they mediate the phasic currents underlying fast inhibitory transmission (Farrant and Nusser, 2005; reviewed by Mody and 6

Pearce, 2004; reviewed by Farrant and Kaila, 2007), whereas receptors containing the δ subunit

are typically located extrasynaptically, where they mediate the tonic GABAA currents caused by

persistent low concentrations of ambient GABA (Caraiscos et al., 2004; reviewed by Semyanov

et al., 2004). There is also a second type of ionotropic GABA receptor, the GABAC receptor

- (reviewed by Bormann, 2000). GABACRs, like GABAARs, form Cl permeable channels, but

they are exclusively located in the retina.

Slow inhibitory transmission is mediated by the metabotropic GABAB receptors

(GABABRs; reviewed by Huang, 2006; Bettler and Tiao, 2006). There are two receptor subunits,

GABAB1 and GABAB2 (Bettler et al., 2004), and it is now fairly well established that most

functional GABABRs in the brain are formed as GABAB1 and GABAB2 heterodimers (Mohler

and Fritschy, 1999). GABABRs inhibit adenylyl cyclase via the Gαi/o subunits of the activated G-

2+ + protein (Hill, 1985) and regulate Ca - and K -channels at pre- and postsynaptic sites via the Gβγ

subunits (Bowery et al., 2002; Calver et al., 2002; Bettler et al., 2004). Presynaptic GABABRs

are present on inhibitory and excitatory terminals where they function as auto- and

heteroreceptors, respectively, and their activation suppresses neurotransmitter release by

inhibition of voltage-sensitive Ca2+-channels (Mintz & Bean, 1993; Thompson et al., 1993;

Poncer et al., 1997). Postsynaptic GABABRs induce a slow inhibitory postsynaptic current

(sIPSC) by gating Kir3-type K+-channels, which hyperpolarizes the membrane and can shunt

excitatory currents (Lüscher et al., 1997). GABABRs have also been implicated in the

modulation of synaptic plasticity (Davies et al., 1991; Patenaude et al., 2003; Huang et al., 2005)

and heterosynaptic depression (Vogt & Nicoll, 1999). They can also inhibit population burst firing and inhibition of backpropagating action potentials (Zilberter et al., 1999; Leung &

Peloquin, 2006), indicating an additional extrasynaptic presence on axons and dendrites. 7

1.5 Regulation of the driving force for GABAAR currents

The polarity and magnitude of the postsynaptic current mediated by the opening of

- GABAARs (as well as glycine receptors) is due to the electrochemical Cl gradient, which is

under the control of cation chloride transporters (CCCs) (reviewed by Blaesse et al., 2009). The

CCC family in mammals consists of nine members falling into three categories based on

function: two members are Na-K-2Cl cotransporters (NKCCs; isoforms NKCC1 and NKCC2),

one is a Na-Cl cotransporter (NCC), four are K-Cl cotransporters (KCCs; isoforms KCC1-4),

and the physiological roles of the remaining two are unknown. The ubiquitous primary active

transporter, the Na+-K+ ATPase, generates plasmalemmal K+ and Na+ gradients that provide the main source of energy for all CCCs. Extrusion of Cl– is driven by the K+ gradient, and is mediated primarily by KCC2 in central neurons. Uptake of Cl– in most neurons is driven by

NKCC1 (Yamada et al., 2004), which exploits the plasmalemmal Na+ gradient as its energy

source for Na-K-2Cl cotransport. Because KCC2 is expressed mainly in the soma and dendrites

(Gulyas et al., 2001), and NKCC1 expression is low outside of the axon, the Cl- gradient is hyperpolarizing at most sites of GABAergic transmission. However, because GABAARs are also

– permeable to HCO3 , which owing to its important role in neuronal pH regulation is relatively concentrated intracellularly, the reversal potential of GABAA receptor-mediated responses

– (EGABA-A) is not identical to the equilibrium potential of Cl (ECl), but shows a deviation to more positive values (Kaila and Voipio, 1987). KCC2 expression is delayed in the hippocampus and neocortex, producing a developmental “switch” in the polarity of GABAAR currents from

depolarizing early in development to hyperpolarizing after the first two weeks postnatal (Rivera 8

et al., 1999). The depolarizing nature of GABAAR-mediated transmission early in postnatal life

is thought to play a role in generating activity important for the development of neural networks

prior to the maturation of sensory inputs (Ben-Ari, 2007).

1.6 Circuitry of the hippocampus

The cortex that forms the hippocampal formation is composed of three layers (reviewed by van Strien et al., 2009; Figure 1). The first layer is comprised of a mixture of afferent and efferent fibres and interneurons. In the dentate gyrus, this layer is called the hilus, whereas in the cornu ammonis (CA) regions it is referred to as the stratum oriens. Next is the cell body layer, called the granule layer in the dentate, and stratum pyramidale (or simply the pyramidal cell layer) in the CA regions and the subiculum. In addition to densely packed principal cells, this layer also contains interneurons. On the other side of the cell layer is stratum moleculare (the molecular layer), which in the CA regions is subdivided into two or three sublayers: the stratum lucidum, which receives input from the DG, and is located only in CA3; the stratum radiatum, comprising the apical dendrites of the neurons located in the stratum pyramidale, located in all

CA regions; and, most superficially, the stratum lacunosum-moleculare, comprising the apical tufts of the apical dendrites, also located in all CA regions.

9

Figure 1. Circuitry of the hippocampus.

The perforant path forms the main input to the hippocampus, synapsing with granule cells of the dentate as well as with CA1 pyramidal neurons. The granule cells excite the pyramidal neurons of area CA3, which in turn, excite the pyramidal neurons of area CA1, which form the primary output of the hippocampus. The mossy fibers are the axons emerging from the granule cells; the Schaffer collaterals are the axons emerging from CA3 and synapsing in CA1. Adapted from Neves et al., 2008. 10

Bliss and Lomo originally demonstrated LTP at the perforant path synapses from the

entorhinal cortex to the dentate gyrus of the hippocampus. The dentate gyrus is considered the

primary input to the hippocampus (reviewed by van Strien et al., 2009). The circuitry of the

hippocampus, from the main input in the dentate gyrus to the primary output in CA1, is

commonly referred to as the trisynaptic pathway. This circuit includes the perforant pathway

projection to the dentate gyrus, the subsequent mossy fiber projection to field CA3, and finally

the Schaffer collaterals targeting neurons in CA1. However, the concept of the trisynaptic

pathway is oversimplified. While inputs do enter the molecular layer of dentate gyrus through the perforant path from entorhinal cortex (EC) layer II, EC layer II also sends axons directly to stratum lacunosum-moleculare of area CA3. Additionally, EC layer III sends direct projections to stratum lacunosum-moleculare of area CA1 and the molecular layer of the subiculum.

1.7 Long-term potentiation

The LTP originally described by Bliss and colleagues in the dentate gyrus was induced

with high frequency stimulation to the perforant path in rabbits, but it was later replicated in

hippocampal slices from rodents (Schwartzkroin and Wester, 1975; Alger and Teyler, 1976), which is the most widely used preparation for its study. LTP has since been observed at all the glutamatergic synapses in the hippocampus and has been extensively characterized using electrophysiological, biochemical and molecular techniques (reviewed in Neves et al., 2008).

Research into LTP has revealed its three major properties: cooperativity, associativity, and input specificity (reviewed by Bliss and Collingridge, 1993). Cooperativity refers to the existence of an intensity threshold for the production of LTP, such that tetanic activation of a relatively small number of afferents will have no potentiating effect. The threshold for LTP induction is the 11

stimulation strength required to elicit a population spike (McNaughton et al., 1978), indicating a

requirement for postsynaptic spiking. Associativity refers to the ability of a weakly tetanized

input to express LTP when concurrently activated with a strong tetanized input (McNaughton et

al., 1978; Levy and Steward, 1979). It is this property which provides a cellular analogue of

classical conditioning, and which was predicted by Hebb’s model as the synaptic strengthening

of the connection between cell A and B, when cell A is one of the cell’s involved in driving cell

B to fire (Hebb, 1949). In reality, associativity and cooperativity are simply two names for the

same property, the requirement for postsynaptic spiking. Input specificity refers to the restriction

of LTP to those inputs activated at the time of tetanization (Andersen et al., 1977; Lynch et al.,

1977), showing that it is a local phenomenon and not a generalized increase in neuronal

excitability.

Low frequency stimulation of hippocampal afferents mainly activates postsynaptic

receptors of the AMPA/Kainate type, since pharmacological blockade with 6-cyano-7- nitroquinoxaline-2,3-dione (CNQX) prevents postsynaptic responses (Davies et al., 1989).

Tetanus-induced LTP additionally involves temporary activation of NMDA receptors during high frequency stimulation which allows loading of the postsynaptic compartment with Ca2+.

NMDA receptors are cellular coincidence detectors and require simultaneous binding of glutamate and depolarization in order to remove the voltage-dependent Mg2+ block from the

channel (Ascher and Nowak, 1988). The increase in postsynaptic Ca2+ is critical for the

activation of further intracellular processes underlying LTP expression. Blockade of NMDA

receptors using 2-amino-S phosphonopentanoate (AP5) consequently blocks the induction of

LTP in CA1 in vitro (Collingridge et al., 1983: Harris et al., 1984) and in the dentate gyrus in

vivo (Morris et al., 1986; Riedel et al., 1995) 12

The expression of LTP requires that the molecules activated during the induction must give rise to a persistent molecular modification of the synapse (reviewed by Malinow et al,

2000). The study of the molecular mechanisms has given rise to a persistent controversy: the localization of the synaptic modifications underlying LTP. A postsynaptic mechanism for LTP was proposed on the basis of observations that LTP selectively increases the postsynaptic current that is attributable to AMPA receptor channels (Muller et al., 1988; Kauer et al., 1988); and that

LTP can increase the sensitivity to exogenously delivered AMPA (Davies et al., 1989). This view was seriously challenged a few years later by studies showing that the frequency of failures in synaptic transmission (a classic measure of presynaptic function) changes during LTP (eg.

Malinow and Tsien, 1990; Kullman and Nicoll, 1992; Larkman et al., 1992). It has been argued that presynaptic failure, which at other synapses is attributable to failure in the release of neurotransmitter (del Castillo and Katz, 1954; Redman, 1990), is actually due to postsynaptic mechanisms at glutamatergic synapses in the cortex (eg. Liao et al., 1995; Wu et al., 1996).

Specifically, transmitter release that acts only on NMDA receptors will be recorded as a failure if the postsynaptic membrane is at resting levels due to the Mg2+ block. Thus, synaptic failures are not necessarily due to a failure of transmitter release, but can be indicative of an absence of

AMPARs. Postsynaptic responses that are mediated solely by activation of NMDA receptors

(that do not produce a response when the cell is at resting potential) have been termed ‘silent synapses’. Silent synapses have now been documented in virtually every glutamatergic synapse where NMDA-receptor dependent LTP occurs, including hippocampal area CA1 (Liao et al.,

1995; Isaac et al., 1995) and the dentate gyrus (Min et al., 1998).

The postsynaptic expression of LTP is thought to occur mainly through post-translational modification of AMPARs which increase their conductance (Benke et al., 1998; Barria et al., 13

1997), and through the insertion of additional AMPARs, which has been directly visualized by

investigation of fluorescent protein-tagged AMPA receptors in hippocampal slices (Shi et al.,

1999). Supporting the postsynaptic membrane as the site of LTP expression, several independent

groups (eg. Mainen et al., 1998; Manabe and Nicoll, 1994) have found no increase in transmitter

release following LTP. However, some results in slice preparations are also not supportive of

postsynaptic modifications (Stevens and Wang, 1994; Bolshakov and Siegelbaum, 1995). While

proponents of the postsynaptic theory argue that results against their theory are not universally

obtained (Malinow et al., 2000), the fact that even still some labs are reporting that LTP is due to

presynaptic mechanisms (Enoki et al., 2009) indicates that this argument cuts both ways.

Recently, it was shown that LTP in slices from adult rats did not depend at all on postsynaptic

changes, but rather was due almost completely to changes in the probability of neurotransmitter

release (Enoki et al., 2009), with the authors reporting no change in the amplitude of quantal

EPSPs. Why then have so many other studies found a dominant role for postsynaptic expression

mechanisms? The authors of this particular study put forth the idea that postsynaptic expression

of LTP is a developmental phenomenon relying on the presence of silent synapses (Enoki et al.,

2009). Silent synapses are rare by 3 weeks of age in rats (Durand et al., 1996) highlighting the

problematic use of slices from juvenile animals by the majority of slice electrophysiologists

studying LTP. Nevertheless, this may not be the full story, given evidence that increased

AMPAR insertion does parallel learning in adult rats (Whitlock et al., 2006) and mice (Matsuo et

al., 2008).

14

1.8 Involvement of LTP in memory

Early support for the idea of a link between LTP and spatial learning came from studies exploring the NMDAR-dependence of memory (reviewed by Morris and Frey, 1997) beginning with the demonstration by Morris et al. (1986) that chronic intraventricular infusion of the selective NMDA antagonist APV results in impairment of spatial learning in the open-field water maze. Importantly, this inhibition of spatial learning occurred at extracellular concentrations of

APV comparable to those that block LTP in vitro (Davis et al., 1992). Furthermore, APV has no effect on performance when applied after learning (Morris, 1989). Confirmation has come from experiments with mice containing site-specific deletion of the NR1 subunit from CA1 pyramidal cells which results in a blockade of LTP at the Schaffer collateral input to CA1, an impairment of spatial learning, and changes in the size and specificity of CA1 place-cell fields (Tsien et al.,

1996a, b; McHugh et al., 1996).

1.9 Interneurons in the hippocampus

In contrast to the more homogeneous principal cell population, interneurons display a striking diversity in their morphological appearances, biophysical properties, patterns of gene/protein expression, and connectivity (reviewed by Buzsáki et al., 2004). GABAergic interneurons in the hippocampus proper comprise 15%–20% of the neuronal population, approximately a third of which innervate the perisomatic region of pyramidal neurons (Freund and Buzsáki, 1996; reviewed in Buzsáki et al., 2007). The axon length of interneurons is two to four times less than CA3 pyramidal cells (Li et al., 1994), although this varies considerably amongst interneuron subtypes, with perisomatic targeting basket and chandelier cells having the shortest axons (40–55 mm; 9,000-12,000 boutons), and dendrite targeting interneurons having 15

the longest (80–220 mm; 16,000-80,000 boutons) (Li et al., 1992; Sik et al., 1994; Sik et al.,

1995; Sik et al., 1997). Dendritic arbors of interneurons are also shorter than those of the principal cells. A single CA1 pyramidal cell has approximately 12 mm in total dendrites,

receiving approximately 30,000 excitatory inputs (Megias et al., 2001), whereas a typical

interneuron has 4 mm of dendrites and between 5,000 and 17,000 excitatory inputs (Gulyás

et al., 1999). Basket cells receive the largest number of excitatory inputs (16,000) and a similar

ratio of inhibitory and excitatory inputs (6% versus 94%) as pyramidal cells (Megias et al., 2001;

Gulyás et al., 1999]). Other interneurons have a lower ratio of excitatory inputs (70%–80%;

Gulyás et al., 1999).

The resting membrane potential of interneurons is considerably less negative, and thus closer to spike threshold than that of the principal cells (Fricker and Miles, 2000; reviewed by

Markram et al., 2004) making interneurons respond more effectively to inputs. Contributing to this effect, the glutamatergic terminals on interneuron dendrites are generally larger than on pyramidal cells (Acsády et al., 1998; Gulyás et al., 1993). This leads to a rapid rise of depolarization, larger amplitude EPSPs, and less frequent failures, making synaptic transmission more effective onto interneurons than principal cells (Miles, 1990; Gulyás et al., 1993; Jonas et al., 2004; Kraushaar and Jonas, 2000; Losonczy et al., 2004), though considerable variability among the various cortical interneuron classes does exist (reviewed by Markram et al., 2004).

This combination of factors leads to higher frequency spiking in interneurons for the same amount of glutamate released onto the postsynaptic membrane, allowing them to control the complex network operations performed by the principal neurons (Shadlen and Newsome, 1998; 16

Shu et al., 2003; Swadlow, 2003) despite the relatively low numbers of inhibitory synapses on

their targets (Megias et al., 2001).

1.10 Types of interneurons

Interneurons have been classified on the basis of firing patterns, molecular expression

profiles, and their innervations of distinct subcellular domains of pyramidal cells (reviewed by

Klausberger and Somgyi, 2008). In CA1 alone, they are at least 21 classes of interneurons, with the most well-studied being the axo-axonic cells, basket cells, bistratified cells, Oriens-

Lacunosum moleculare (O-LM) cells, and Ivy cells. Fast-spiking axo-axonic cells (or 'chandelier'

cells) project exclusively to the axon initial segments of pyramidal neurons, where they form

characteristic clusters of presynaptic specializations that are orientated perpendicular to the

pyramidal cell. O-LM cells have their cell bodies and dendrites in the stratum oriens and project

their axons to the stratum lacunosum-moleculare, where they innervate the apical dendritic tuft coaligned with the direct entorhinal input. Neurogliaform cells also target the apical dendritic tuft, and provide a slower form of GABAA receptor–mediated inhibition (Hardie and Pearce,

2006; Szabadics et al., 2007), as well as GABAB receptor–mediated inhibition (Tamás et al.,

2003; Price et al., 2005). The recently discovered Ivy cells innervate more proximal pyramidal

cell dendrites aligned with the CA3 input. Basket cells innervate primarily the cell bodies, as well

as the proximal dendrites. Two subtypes of basket cell are distinguished by their expression of

either cholecystokinin (CCK) or parvalbumin (PV) and by their firing properties. CCK-positive

basket cells also express endocannabinoid CB1 receptors at their terminals. Bistratified cells

innervate the basal and oblique dendrites coaligned with the CA3 glutamatergic input. In

addition, there are a number of GABAergic projection neurons, projecting both between areas of 17

the hippocampus as well as between other brain areas and the hippocampus, which likely prime

and reset activity in specific target areas (Klausberger and Somgyi, 2008).

The compartmentalized structure of pyramidal cells allows spatially segregated activities

at the same time, but also allows for distinct activities to occur temporally segregated as well,

owing to the differential timing of the various inputs. Differences in the short-term plasticity of

glutamatergic synapses onto distinct interneurons (Ali et al., 1998a,b; Biro et al., 2005;

Silberberg et al., 2004) further contribute to a temporally distinct and spatially distributed recurrent inhibition in perisomatic or dendritic domains of pyramidal cells (Pouille and

Scanziani, 2004). In the somatosensory cortex, the firing frequency of perisomatic- and dendrite- targeting interneurons may differentially entrain the output of postsynaptic pyramidal cells

(Tamás et al., 2004). The dissection of cellular properties in vitro has provided stimulating

possibilities of how distinct types of interneuron might act. It remains a challenge to explain how

these concepts relate to the information flow when the neurons are embedded in ongoing network

activity.

1.11 Feedforward inhibition

Recent advances revealing a spatiotemporal division of labor in cortical circuits are perhaps best exemplified by our current understanding of feedforward inhibition in area CA1 of the hippocampus. When recording from CA1 pyramidals, evoking Schaffer collateral-mediated

transmission (i.e. transmission from presynaptic CA3 pyramidals) produces an EPSP/IPSP

sequence (Pouille and Scanziani, 2001; Figure 2). Both the glutamatergic and GABAergic

component are blocked by AMPAR antagonists indicating that the IPSP is disynaptic and

requires the firing of the glutamatergic Schaffer collaterals. The delay between the EPSP and 18

IPSP, caused by the extra round of synaptic transmission required for the disynaptic IPSP, is

only 1.9 ms. This brief delay underlies the ability of feedforward inhibition to limit the

integration of temporally separate excitatory inputs (Pouille and Scanziani, 2001). Feedforward inhibition is mediated by the soma targeting PV-positive basket cells, sometimes termed fast-

spiking basket cells for the non-adapting trains of action potentials they fire in response to step

depolarization, in contrast to the CCK-positive basket cells which show strong spike-frequency

adaptation (Freund and Buzsaki, 1996). These cells receive stronger and more persistent

excitation than the CCK-positive basket cells, which they integrate over very narrow time windows, thereby faithfully reporting the timing of ongoing hippocampal activity in area CA3.

Interestingly, PV-positive basket cells receive inputs not only from the Schaffer collaterals, but also from the direct entorhinal inputs via the perforant path and feedback connections from CA1 pyramidals, highlighting their importance in providing somatic inhibition time-locked to excitatory transmission regardless of the source (Gulyás et al., 1999).

19

Figure 2. Feedforward inhibition in area CA1.

When CA3 pyramidal neurons fire, they excite both CA1 pyramidals as well as feedforward interneurons, which inhibit CA1 pyramidals. Levels of CA3 firing which produce only subthreshold EPSPs in CA1 pyramidals, easily drive action firing in the feedforward interneurons. As a result, CA3 firing generates an EPSP-IPSP sequence (or mixed PSP) in CA1 pyramidal neurons.

20

1.12 Reports of inhibitory plasticity

While synaptic plasticity in the hippocampus has been studied mainly at glutamatergic

synapses, a number of reports have demonstrated the expression of activity-dependent

modification at GABAergic synapses as well. Most of these studies have demonstrated that

tetanic stimulation induces changes in inhibitory synaptic conductance, both increases and

decreases, analogous to LTP and long-term depression (LTD) at glutamatergic synapse. A potent

LTD of GABAergic synapses (commonly referred to as disinhibition) in CA1 can be induced with strong tetanic stimulation (used to induce epileptic activity in vitro) to the Schaffer collaterals (Stelzer et al., 1987). A briefer tetanic stimulation applied to the Schaffer collaterals, identical to that frequently used to induce glutamatergic LTP has also been demonstrated to induce disinhibition onto CA1 pyramidals (Patenaude and Lacaille, 2000; Lu et al., 2003;

Chevaleyre and Castillo, 2003). In another study, tetanic stimulation of the Schaffer collaterals was shown to induce disinhibition at inhibitory synapses on the dendrites of CA1 pyramidals, but potentiation at synapses onto the somatic compartment (Stelzer et al., 1994). Interestingly, a more physiological theta-burst stimulation, representing activity patterns thought to be important for memory formation in vivo, was found to induce a potentiation of GABAergic transmission under conditions in which tetanic stimulation had induced disinhibition (Patenaude and Lacaille,

2000). Tetanically-induced LTP specifically at feedback inhibitory connections in CA1 has also been reported (Grunze et al., 1996). Long term potentiation of GABAergic transmission has also been demonstrated in the CA3 region of slices from neonatal rats (Caillard et al., 1999a). Lastly, low frequency pairing of pre- and postsynaptic action potential firing has been demonstrated to reduce the driving force for GABAAR-mediated currents through a reduction in KCC2 activity 21

leading to intracellular Cl- accumulation, with no change in synaptic conductance observed

(Woodin et al., 2004).

The induction and expression mechanisms for these various forms of hippocampal

inhibitory plasticity vary widely. A number of these studies showed that like glutamatergic

plasticity, inhibitory plasticity also depends on activation of NMDAR (Stelzer et al., 1994; Lu et

al., 2000; Caillard et al., 1999b; Grunze et al., 1996). In some cases, however, mGluR activation,

rather than NMDAR activation, was required (Patenaude and Lacaille, 2000; Chevaleyre and

Castillo, 2003). The locus of plasticity was only investigated in a handful of these studies.

Measuring the paired-pulse ratio before and after plasticity induction indicated a postsynaptic

locus of expression in one study (Patenaude and Lacaille, 2000), whereas another study found

that disinhibition was the result of reduced transmitter release from the presynaptic terminal due

to activation of cannabinoid receptors by endocannabinoids released by the postsynaptic

pyramidal neuron (Chevaleyre and Castillo, 2003). The change in the driving force for

GABAAR-mediated currents was of course postsynaptic (Woodin et al., 2004).

Does inhibitory plasticity play a role, along the lines of glutamatergic LTP, in memory

formation? A handful of the above studies (Stelzer et al., 1994; Lu et al., 2000; Chevaleyre and

Castillo, 2003) have described effects of general (i.e. not localized to a specific circuit) inhibitory

plasticity on excitatory transmission indicating either direct or indirect involvement in LTP,

further detailed in Chapter 2. However, the circuitry of CA1 suggests that those inhibitory

synapses mediating feeforward inhibition should be active with the appropriate timing relative to

Schaffer collateral-mediated excitation to have an impact on its efficacy. Plasticity at these

specific connections has not previously been described. 22

1.13 Hypotheses

With multiple reports of GABAergic plasticity in the hippocampus, we thought it would be important to determine whether GABAergic plasticity plays a significant role in the hippocampus’ ability to mediate the storage of information. While technical limitations make a

direct examination of the effect of disinhibition on memory in vivo unfeasible at present, the link between CA1 LTP and memory has been fairly well established in rodents. As such, a demonstration of inhibitory plasticity’s involvement in CA1 LTP could provide an important basis for continued research into a role for GABAergic plasticity in memory. Based on the tight temporal relationship between Schaffer collateral-mediated glutamatergic transmission and disynaptic feedforward inhibitory transmission, we hypothesized that changes in the strength of feedforward inhibitory transmission, like the more commonly studied changes in glutamatergic transmission, would be able to regulate the efficacy of transmission between areas CA3 and

CA1. Our experiments, performed in slices from adult rats, show that disinhibition in the feedforward circuitry can indeed contribute to LTP expression. Furthermore, based on our suspicion that disinhibition is involved in memory, we hypothesized that disinhibition would share classic glutamatergic LTP’s pathway specificity, a crucial aspect of LTP’s usefulness as a mnemonic device (Dunwiddie and Lynch, 1978). 23

2. Disinhibition Mediates a Form of Hippocampal Long-Term Potentiation in Area CA1 Abstract

The hippocampus plays a central role in memory formation in the mammalian brain. Its ability to encode information is thought to depend on the plasticity of synaptic connections between neurons. In the pyramidal neurons constituting the primary hippocampal output to the cortex, located in area CA1, firing of presynaptic CA3 pyramidal neurons produces monosynaptic excitatory postsynaptic potentials (EPSPs) followed rapidly by feedforward (disynaptic) inhibitory postsynaptic potentials (IPSPs). Long-term potentiation (LTP) of the monosynaptic glutamatergic inputs has become the leading model of synaptic plasticity, in part due to its dependence on NMDA receptors (NMDARs), required for spatial and temporal learning in intact animals. Using whole-cell recording in hippocampal slices from adult rats, we find that the efficacy of synaptic transmission from CA3 to CA1 can be enhanced without the induction of classic LTP at the glutamatergic inputs. Taking care not to directly stimulate inhibitory fibers, we show that the induction of GABAergic plasticity at feedforward inhibitory inputs results in the reduced shunting of excitatory currents, producing a long-term increase in the amplitude of Schaffer collateral-mediated postsynaptic potentials. Like classic LTP, disinhibition-mediated LTP requires NMDAR activation, suggesting a role in types of learning and memory attributed primarily to the former and raising the possibility of a previously unrecognized target for therapeutic intervention in disorders linked to memory deficits, as well as a potentially overlooked site of LTP expression in other areas of the brain.

24

2.1 Introduction: Feedforward inhibition in CA1

Plasticity of synaptic connections between neurons in the hippocampus is thought to play

a central role in learning and memory. Synaptic plasticity can be induced by patterned electrical

stimulation at a number of synapses in the hippocampus, including the excitatory synapses of the

trisynaptic pathway, as well as at certain excitatory onto interneuron synapses, and inhibitory

onto pyramidal neuron synapses [see Nelson and Turrigiano (2008) for a review]. LTP of CA3-

CA1 glutamatergic transmission has become the leading model of synaptic plasticity, in part

because of its dependence on NMDAR activation (Harris et al., 1984), which provides a

mechanism for associating pre- and postsynaptic action potential firing, and which is also required in vivo for hippocampal-dependent spatial and temporal learning (Morris et al., 1986;

Tsien et al., 1996; Huerta et al., 2000).

An analysis of the literature on feedforward inhibition in CA1 suggests that plasticity at

inhibitory synapses might also be able to play a role in regulating the efficacy of CA3-CA1 transmission. When presynaptic CA3 pyramidals fire, the EPSP recorded in CA1 is followed in less than 2 ms by a disynaptic IPSP (Pouille and Scanziani, 2001) originating from basket cells targeting the somatic compartment (Glickfeld and Scanziani, 2006). This delay between EPSP and IPSP is only half as long as the rise time of unitary EPSPs evoked by single cell firing in

CA3 (Sayer et al. 1990). Furthermore, feedforward inhibition has been shown to overlap with the rising phase of the EPSP in hippocampal slices from guinea pigs (Karnup and Stelzer, 1999).

Thus, feedforward inhibition should reduce EPSP amplitude recorded at the soma, as

demonstrated for unitary EPSPs between pairs of CA3 neurons (Miles, 1990). It follows that

disinhibition, if expressed at feedforward synapses, would reduce the shunting of excitatory 25

currents, leading to an increase in EPSP amplitude, and thereby augmenting the efficacy of CA3-

CA1 excitatory transmission without the need for classic LTP at glutamatergic synapses.

A number of studies have, in fact, reported the expression of activity-dependent reductions in strength at inhibitory synapses in area CA1 (Stelzer et al., 1994; Lu et al., 2000; Chevaleyre and Castillo, 2003; Woodin et al., 2003). Two of these studies showed that disinhibition, requiring either NMDAR (Lu et al. 2000) or mGluR (Chevaleyre and Castillo, 2003) activation during induction, can cause EPSP-spike (E-S) potentiation, an increase in postsynaptic excitability normally co-expressed with glutamatergic LTP. A third study showed that a reduction in inhibitory strength at monosynaptic GABAergic synapses (i.e. synapses directly activated by the stimulating electrode) induced by tetanic stimulation contributed to LTP expression through reduced shunting of excitatory currents (Stelzer et al., 1994). While these studies provide compelling evidence of inhibitory plasticity’s ability to modulate the efficacy of synaptic transmission between areas CA3 and CA1, a number of factors suggest that the disinhibition demonstrated in these studies did not occur at the feedforward inhibitory synapses, and would therefore be unlikely to mediate the reported effects under physiological conditions.

All three studies applied extracellular stimulation within 200 µm of the postsynaptic neuron, too close to avoid the recruitment of unphysiological monosynaptic inhibition [see Marder and

Buonomano (2004) for a detailed discussion; in fact, Karnup and Stelzer (1999) show that monosynaptic inhibition can still be elicited from as far away as 1 mm with relatively moderate levels of stimulation]. Because monosynaptic inhibition is not delayed relative to the EPSP, and is comprised of all synapses within range of the stimulation electrode (located in the dendritic layer in the cited studies), it has a greatly enhanced, albeit artificially, ability to shunt excitatory 26

currents. Thus, using disinhibition at sites of monosynaptic transmission to probe the ability of

GABAergic plasticity to influence excitatory transmission should lead to an overestimation of its effect, and has the potential to produce false positives due to the involvement of inhibitory synapses not normally co-active with the Schaffer collateral synapses. For instance, Stelzer et al.

(1994) observed disinhibition at monosynaptic connections in the dendrites, but actually reported a potentiation of inhibition at the soma, indicating that feedforward inhibition was strengthened and therefore not contributing to LTP expression. Likewise, Chevaleyre and Castillo (2003) reported that the disinhibition underlying E-S potentiation was due to activation of cannabinoid receptors on presynaptic interneurons, yet it has been reported that the basket cells mediating feedforward inhibition are cannabinoid-insensitive (Glickfeld and Scanziani, 2006), suggesting the recruitment of cannabinoid-sensitive interneurons was monosynaptic. Thus, whether disinhibition can occur at sites of feedforward inhibition and whether such disinhibition would contribute to LTP expression remains unknown.

Being careful not to directly stimulate inhibitory fibers (see Figure 3b, Figure 10) while making whole-cell recordings in hippocampal slices from 2 month old rats, we investigated the role of feedforward inhibition in regulating Schaffer collateral-mediated excitatory transmission.

We found that feedforward inhibition does indeed reduce EPSP amplitude, and that disinhibition at feedforward synapses, expressed as a depolarization of the reversal potential for GABAAR-

mediated currents (Woodin et al., 2003) contributes to the increase in EPSP amplitude seen

during LTP expression.

27

2.2 Results: Disinhibition increases the efficacy of excitatory transmission

To investigate the effect of feedforward inhibition on the amplitude of Schaffer collateral- mediated EPSPs, we stimulated the Schaffer collaterals while manipulating the driving force for

GABAergic currents by either hyperpolarizing or depolarizing the postsynaptic pyramidal cell with current injection through the whole-cell pipette. Unlike previous studies examining the interaction between feedforward inhibition and excitation, the effect of disynaptic inhibition here was not enhanced by manipulating the chloride gradient with a low [Cl-] intracellular recording

solution (Pouille and Scanziani, 2001; Bissière et al., 2003). Rather, we initially performed

control recordings with gramicidin perforated patch to maintain the intracellular chloride

concentration, and then in subsequent whole-cell recordings, determined the intracellular [Cl-] needed to reproduce the EGABA measured with gramicidin (10 mM; see Methods and Materials).

The Schaffer collaterals were stimulated close to their site of origin in CA3, in order to avoid directly stimulating local inhibitory fibers in CA1. We found that membrane depolarization, which increases the driving force for Cl- in the inward direction, had a strongly depressive effect

on EPSP amplitude (Figure 3a). This depression of EPSP amplitude was completely blocked by

the GABAA receptor (GABAAR) antagonist GABAzine (6 µM). In separate recordings,

perfusion of the AMPA receptor (AMPAR) antagonist CNQX (10 µM; Figure 3b; Figure 10)

largely abolished the postsynaptic response; the absence of an isolated IPSP after CNQX

perfusion indicated that inhibition was activated disynaptically (i.e being driven by glutamatergic

transmission) rather than directly by the stimulating electrode (CNQX was applied at the end of

all subsequent EPSP/IPSP recordings). Taken together, these results show that disynaptic

inhibition elicited by Schaffer collateral-mediated glutamatergic transmission decreases the 28

amplitude of EPSPs recorded at the soma. To confirm that the onset of feedforward inhibition is

indeed rapid enough to reduce EPSP amplitude, we directly measured the EPSP/feedforward

IPSP delay, previously found to be 1.9 ms in hippocampal slices from 1 month old Wistar rats

recorded at 33°C (Pouille and Scanziani, 2001). To this end, we made paired recordings between

synaptically connected feedforward interneurons and pyramidal cells. Feedforward interneurons

were identified electrophysiologically; they fired APs in response to relatively moderate levels of

Schaffer collateral stimulation (which never elicited pyramidal cell firing), and when made to

spike with intracellular current injection, evoked unitary IPSPs in neighbouring pyramidal cells.

To calculate the EPSP/IPSP delay, the delay between the onset of the Schaffer collateral-evoked pyramidal cell EPSP and the interneuron AP (Figure 4a) was added to the delay between the intracellularly-evoked interneuron AP and the resulting pyramidal unitary IPSP (Figure 4b). This

gave an EPSP/IPSP onset delay of 1.3 ± .3 ms (n = 10). To determine whether the EPSP/IPSP delay is indeed brief enough to reduce EPSP amplitude, we next determined EPSP rise time.

Pharmacologically isolated Schaffer collateral-evoked EPSPs were recorded (Figure 4c). No correlation was observed between 90% amplitude (range = 0.5 - 4.5 mV) and rise time to 90% amplitude [adj. r2 = 0, p = 0.623, n = 43 (4 cells)]. The mean rise time to 90% amplitude was 6.1

± .3 ms, leading us to conclude that feedforward inhibition arrives in time to reduce EPSP

amplitude. PSPs evoked in the absence of blockers, being comprised of both EPSPs and

feedforward IPSPs, are referred to as mixed PSPs henceforth.

We next examined how feedforward inhibition affects the magnitude of pairing-induced

LTP. Pairing Schaffer collateral stimulation with postsynaptic spiking (300 pairings at 5 HZ;

Experimental Procedures; Figure 11) produced a robust increase in mixed PSP amplitude (Figure 29

5a, Figure 12a). However, when feedforward inhibition was blocked with gabazine, LTP was

reduced (Figure 5b, Figure 12b). The mere presence of feedforward inhibition would not be

expected to enhance LTP, and has in fact been reported to instead diminish its magnitude

(Meredith et al. 2003). We therefore suspected that a reduction in inhibitory strength induced by

pairing was responsible for the enhanced LTP. To determine if this might be the case, we

isolated inhibition pharmacologically, and paired extracellular stimulation close to the cell body

layer [the site of feedforward synapses (Glickfeld and Scanziani, 2006)] with postsynaptic spiking. This induced a depolarization of EGABA (Figure 5c, Figure 12c), as previously reported in slices from juvenile rat (Woodin et al., 2003). Importantly, this also confirmed the previously reported observation that whole-cell recording does not interfere with the expression of pairing-

induced GABAergic plasticity (Woodin et al., 2003), which results from intracellular Cl-

accumulation due to a Ca2+ dependent down-regulation of the K+/Cl- cotransporter KCC2. No long-term changes in relative synaptic conductance were observed (Figure 5c, p = 0.173). These results suggest that pairing-induced weakening of the GABAergic driving force increases the magnitude of LTP.

We next sought a direct confirmation that EGABA depolarizes during LTP expression. To

this end, we induced and monitored LTP with Schaffer collateral stimulation, while

simultaneously monitoring EGABA by evoking APs in a synaptically connected feedforward

interneuron. Indeed, both EGABA depolarization and LTP were expressed after pairing (Figure 6,

Figure 13). Thus, pairing Schaffer collateral stimulation with pyramidal cell spiking induces a

weakening of the driving force for GABAergic currents at feedforward synapses that contributes

to LTP expression. 30

As a final confirmation of our results, we asked whether EGABA depolarization alone would

be sufficient to produce LTP. Our strategy was to block glutamatergic LTP pharmacologically

with a drug that would leave pairing-induced inhibitory plasticity unaffected. We first tested the

NMDAR blocker APV (25 µM), but it completely blocked both pairing-induced GABAergic

plasticity and LTP (Figure 7; data not shown). We next tested inhibition of calcium/calmodulin-

dependent protein kinase II (CaMKII), a Ca2+-dependent kinase required for classic LTP

induction (Malinow et al., 1988). Applying the CaMKII inhibitor autocamtide 2-related

inhibitory peptide (AIP; 5 µM; Ishida et al., 1995) intracellularly through the patch pipette had

no effect on inhibitory plasticity (Figure 8A, Figure 14A; change in conductance p = 0.078), but

abolished glutamatergic LTP when inhibition was blocked with gabazine (p = 0.537, Figure 8b,

Figure 14b). When both inhibition and excitation were left intact, LTP induced by pairing was

only slightly reduced by AIP (Figure 8c, Figure 14c), as expected from the loss of the

glutamatergic LTP component (Figure 5b). Thus, we conclude that disinhibition is sufficient to

produce LTP of Schaffer collateral-mediated excitatory transmission.

Given the view that the hippocampal synaptic plasticity responsible for memory formation

is thought to require only short periods of action potential firing (Mehta et al., 2000), we wanted

to know whether EGABA depolarization could be induced by a briefer induction protocol. Leaving

the frequency of stimulation unchanged, we examined the ability of 10, 25, or 50 pre- and postsynaptic pairings to induce plasticity (Figure 9). Neither 10 (p = 0.74) nor 25 (p = 0.139) pairings produced significant plasticity (no significant difference from control in Figure 4), but

50 pairings (elapsed time of 10 sec) produced a depolarization of EGABA to a level 31

indistinguishable from that induced by 300 pairings [p < 0.001 compared to ctrl (Figure 4) and

10 pairings, p = 0.005 compared to 25 pairings]. 32

2.3 Figures

Figure 3. Feedforward inhibition reduces EPSP amplitude. a, left, PSP amplitude vs. Vm for one cell before (black) and after (open) gabazine (left). Insets: Sample traces before (1) and after (2) drug application. right, The slope of the PSP vs. Vm graph before and after gabazine, normalized to the before gabazine slope (n= 4). * denotes statistical significance from control (* p < 0.05). b, PSP amplitude vs. Vm for one cell before (black) and after (open) CNQX (right). Insets: Sample traces before (1) and after (2) drug application. Membrane potential denoted to left of sample traces. 33

Figure 4. Feedforward inhibition overlaps the EPSP rising phase. a, b sample traces demonstrating: a, the delay between the evoked pyramidal PSP and interneuron AP, b, the delay between the interneuron AP and unitary IPSP (same cell as left). Inset: interneuron firing in response to current injection; c, pharmacologically isolated EPSP with feedforward IPSP delay superimposed. Membrane potential denoted to left of sample traces where relevant. 34

Figure 5. EGABA depolarization underlies a large component of mixed LTP. a, Mixed LTP (pairing n=6, ctrl n=5), b, glutamatergic LTP (pairing n=7, ctrl n=5) , c, GABAergic plasticity (pairing n=6, ctrl n=5, same cells left and right). Insets: Sample traces (a, c) or average of 10 traces (b) before (1) and after (2) paired activity. Black circles = pairing (denoted by arrow), open circles = ctrl. *denotes statistical significance from control (* p<0.05; ** <0.001) from that time point on. Membrane potential denoted to left of sample traces where relevant. 35

Figure 6. LTP induction leads to simultaneous EGABA depolarization and LTP. left, Simultaneous GABAergic plasticity and mixed LTP induced by pairing Schaffer collateral stimulation close to CA3 with postsynaptic spiking (n=7). Data presented as the mean change at 40 min. right, sample traces from one cell. *denotes statistical significance (* p<0.05). Membrane potential denoted to left of sample traces. 36

Figure 7. APV blocks pairing-induced GABAergic plasticity in single pathway experiments.

left, Change in EGABA induced by paired activity in the absence of APV (black circles, n= 4) or in its presence (white circles, n= 5) plotted against time. right, Change in relative GABAergic conductance in the same cells. Black circles represent cells receiving paired activity in the absence of APV. Open circles represent cells receiving paired activity in APV. * denotes statistical significance from control (* p < 0.05) from that time point on. 37

Figure 8. Depolarization of EGABA is sufficient for LTP expression. a, GABAergic plasticity (pairing n=7, ctrl n=5, same cells left and right). b, Glutamatergic LTP (pairing n=5, ctrl n=4). C, Mixed LTP (pairing n=6, ctrl n=5). Insets: Sample traces (a, c) or average of 10 traces (b) before (1) and after (2) paired activity. CaMKII inhibitor AIP present in all recordings. Black circles = pairing (denoted by arrow), open circles = ctrl. * denotes statistical significance from control (* p<0.05; ** <0.001). Membrane potential denoted to left of sample traces where relevant. 38

Figure 9. Depolarization of EGABA can be induced by as few as 50 pairings.

GABAergic plasticity induced by 10 pairings (diamond, n = 7), 25 pairings (square, n = 6), or 50 pairings (triangle, n = 5). ** denotes statistical significance from control (** <0.001; for statistical significance between groups see text). 39

Figure 10. The majority of recorded inhibition in mixed EPSP/IPSP recordings was feedforward.

The slope of the PSP vs. Vm graph after CNQX application divided by the slope before application for all experiments (see Figure 3b for example). Experimental groups: 1, mixed LTP (Figure 5a); 2, mixed control (Figure 5a); 3, mixed LTP (paired recordings, Figure 6); 4, mixed LTP with AIP (Figure 8c); 5, mixed control with AIP (Fig. 6c). 40

Figure 11. Induction protocol.

Sample trace of plasticity induction, showing beginning of current injection followed by presynaptic stimulus (visible from stimulus artifact) and then postsynaptic spiking. Repeated 300 times at 5 Hz. Scale bars: 20 mV, 2ms.

41

Figure 12. Depolarization of EGABA contributes to LTP expression- representative recordings. a, Change in mixed PSP amplitude induced by paired activity. b, Change in EPSP amplitude induced by paired activity. c, Change in EGABA induced by paired activity.

42

Figure 13. Pairing Schaffer collateral stimulation with postsynaptic spiking results in simultaneous EGABA depolarization and LTP expression.

a, Change in EGABA induced by paired activity. b, Change in mixed PSP amplitude induced by paired activity (same cells as (a). * denotes statistical significance from baseline (* p < 0.05) from that time point on.

43

Figure 14. Depolarization of EGABA is sufficient for LTP expression- representative recordings.

a, Change in EGABA induced by paired activity with CaMKII activity inhibited. b, Change in EPSP amplitude induced by paired activity with CaMKII activity inhibited. c, Change in mixed PSP amplitude induced by paired activity with CaMKII activity inhibited. 44

Figure 15. Disinhibition alone can increase the efficacy with which presynaptic CA3 pyramidal neurons excite postsynaptic CA1 pyramidal neurons. a, Classic LTP is expressed mainly as an increase in AMPAR insertion at the postsynaptic side of the Schaffer collateral synapses onto CA1 pyramidal neurons. Much of the excitatory current generated by Schaffer collateral transmission is shunted by temporally overlapping feedforward transmission, such that the depolarization measured at the soma is smaller than that generated at the site of excitatory transmission in the dendrites. b, Disinhibition at the feedforward connections can also increase the efficacy with which presynaptic CA3 pyramidals excite their postsynaptic CA1 pyramidal targets. Increased intracellular [Cl-] reduces the driving force for GABAergic currents, thereby reducing the shunt of excitatory current. 45

Figure 16. Electrophysiological identification of CA1 pyramidal neurons.

Recorded neurons were identified by the presence of an after-depolarization following AP firing, as well as spike accommodation (inset). Scale bars. Scale bars inset: 40 mV, 250 ms.

46

2.4 Discussion: Disinhibition-mediated long-term potentiation

To our knowledge, this is the first report that long-term weakening of feedforward connections can augment the efficacy of Schaffer collateral-mediated glutamatergic transmission

(see Figure 15 for model). We took great care to make sure that the inhibition being elicited was done in a way that was physiological so as to avoid overestimating the impact of inhibition on

EPSP amplitude. By activating inhibitory inputs disynaptically rather than directly, we have maintained the physiological ratio of, and delay between, excitation and inhibition, and avoided the recruitment of interneuron types not involved in feedforward inhibition. Furthermore, under our whole-cell recording conditions, the driving force for GABAAR-mediated currents closely

matched the driving force we measured with gramicidin perforated patch recording (see

Materials and Methods), which leaves the intracellular [Cl-] unperturbed. Lastly, we used slices

from adult rats in all experiments. We believe the use of slices from adults is crucial for studying

the cellular mechanisms underlying learning and memory, as ongoing neural development

complicates the interpretation of data from younger animals. Highlighting this fact, we

previously found that EGABA, measured with gramicidin perforated-patch, hyperpolarizes a further 15 mV between 3 and 7 weeks of age (Ormond and Woodin, 2006 abstract) indicating

that the developmental changes involved in strengthening inhibition extend into adolescence in

the rat. As such, we think our study provides strong evidence of a role for disinhibition in LTP

beyond the end of development, when synaptic plasticity is thought to primarily subserve the

long-term storage of information.

Our results show that disinhibition-mediated LTP is virtually indistinguishable from classic

LTP when recorded in whole-cell mode from the soma. However, it is important to note that 47

classic glutamatergic LTP results in the potentiation of EPSPs at their site of initiation in the dendrites. Thus, the site of classic LTP expression confers it with a unique role in modulating the dendritic integration of excitatory inputs. For example, non-linear summation in small dendritic branches (Polsky et al., 2004) could cause LTP at a given synapse to be amplified when other synapses in the same branch are co-active. Once EPSPs spread to the soma, though, we find feedforward inhibition blunts their impact, in most cases causing a complete loss of depolarization before action potential threshold is reached (see Figure 3a, Figure 5a). Thus, as demonstrated here, modulation of inhibitory strength is a powerful mechanism for regulating the efficacy with which EPSPs are able to depolarize the soma.

As mentioned in the introduction, a role for disinhibition in E-S potentiation, the increase in postsynaptic excitability which normally accompanies classic LTP, has been previously reported (Lu et al., 2000; Chevaleyre and Castillo, 2003). While we did not directly investigate postsynaptic excitability here, our data are consistent with a role for disinhibition, but also indicate that E-S potentiation needn’t refer explicitly to a change in the postsynaptic neuron’s excitability. E-S potentiation has most often been studied with extracellular recording, and observed as a disproportionate increase in the population spike at the cell body layer relative to the potentiation of the field potential (which provides a measure of the EPSP recorded in the dendritic layer) (Bliss and Gardner-Medwin, 1973; Bliss and Lomo, 1973; Andersen et al.,

1980). The disproportionate change in the population spike underlies the belief that a change in intrinsic excitability is responsible for this phenomenon. While one study implicating

GABAergic plasticity did find evidence that this was the case (the AP threshold was reduced after LTP induction; Lu et al., 2000), a change in excitability is not required to explain 48

extracellularly recorded E-S potentiation. Feeforward interneurons synapse primarily onto the

soma (Glickfeld and Scanziani, 2006), so the contribution of disinhibition at those sites to LTP

would be largely confined to that compartment. Thus, disinhibition should increase the amount

of LTP expressed at the soma relative to the dendrites, which extracellularly, would be recorded

as an increase in the population spike relative to the dendritic field potential. Our data therefore

suggests that E-S potentiation may be simply due to an additional LTP component restricted to

the soma.

The lack of robust glutamatergic LTP in our experiments was beneficial in that it allowed

us to more easily identify the contribution that disinhibition makes to LTP in the mixed

EPSP/IPSP condition. Nevertheless, some might find this curious given the number of

publications showing that paired pre- and postsynaptic activity, when separated with a positive

(pre before post) spike-timing delay, produces robust glutamatergic LTP in CA1 pyramidal cells

(eg. Magee and Johnston, 1997). However, a number of publications have instead shown that

such pairing induces either moderate LTP (eg. Nishiyama et al. 2000), weak LTP (eg. Hardie and

Spruston, 2009), or even long-term depression (LTD; Christie and Johnston, 1996). There are a

number of factors to consider when evaluating the effectiveness of a given induction protocol to

induce glutamatergic LTP in addition to the pre-/postsynaptic delay: these include the number of

pre- and postsynaptic spikes per pairing, the frequency of pairing, and the number of pairings.

For example, while it is well known that LTD can be induced by switching the order of pre- and

postsynaptic pairing so that the postsynaptic firing occurs first (Bi and Poo, 1998; Nishiyama et

al., 2000), it can also be induced with pre- before post pairing if the number of pairings is greatly

increased (Christie et al., 1996). This can be partly understood due to the fact that that both LTP 49

and LTD require calcium influx through NMDARs [reviewed by Bear and Malenka (1994)].

Recent work has shed more light on this issue by showing that the processes underlying LTP and

LTD can occur simultaneously, and that the resulting plasticity is due to the combinatory effects of these processes (Wittenberg and Wang, 2006). Given that LTP is normally induced by 50-100 pre- before post- pairings, and LTD requires approximately 900 such pairings, it is not surprising that our induction protocol consisting of 300 pairings, occupying the middle ground, induced only a weak glutamatergic LTP [though Hardie and Spruston, (2009) reported a similar level of

LTP induced with fewer pairings].

Our demonstration here that depolarization of EGABA can be induced by both relatively low

(50) and relatively high (300) numbers of pairings suggests that GABAergic plasticity may in some cases be co-expressed with glutamatergic LTP, and in other cases expressed in its absence.

Pairing-induced LTP is hypothesized to modulate various aspects of place cell firing in the hippocampus, and possible examples of both low and high numbers of pairings can be found in the literature. For example, the experience-dependent expansion of place fields is induced by just a few passes through a given field (Mehta et al., 2000); this level of activity would be expected to induce both glutamatergic and GABAergic plasticity. Requiring more activity is the reactivation during sleep of neurons which co-fired during waking exploration (O’Neill et al.,

2008); this study reported that reactivation was most robust in those neurons which fired the most number of action potentials, across an examined range of 40-280, during exploration. As discussed above, increasing the number of pairings across this range would actually decrease the magnitude of glutamatergic LTP, suggesting that this phenomenon might depend solely on 50

GABAergic plasticity, and might in fact be disrupted by the glutamatergic LTP induced with

fewer pairings.

The glutamatergic LTP hypothesis of memory has been strengthened by demonstrations of

LTP-like changes accompanying memory formation in vivo (Gruart et al., 2006; Whitlock et al.,

2006). Does any such evidence exist for pairing-induced GABAergic plasticity? In fact, EGABA

depolarization has been demonstrated in CA1 pyramidal neurons of the dorsal hippocampus in

slices cut from animals that had completed spatial memory acquisition in a water maze task

(Gusev and Alkon, 2001). In that study, no parallel changes in excitatory transmission were

observed, suggesting that GABAergic plasticity alone might underlie some forms of memory.

This possibility is supported by our demonstration here that disinhibition at feedforward

connections is sufficient for LTP expression. Furthermore, the NMDAR-dependence of pairing-

induced EGABA depolarization suggests that it may be involved in forms of memory attributed until now solely to classic glutamatergic LTP (Morris et al, 1986; Tsien et al., 1996; Huerta et

al., 2000). 51

3. A Heterosynaptic Increase in GABAergic Conductance Maintains the Pathway Specificity of Disinhibition- Mediated LTP Abstract

The hippocampus plays a central role in memory formation in the mammalian brain. Its ability to encode information is thought to depend largely on the plasticity of synaptic connections between neurons. In the pyramidal neurons constituting the primary hippocampal output to the cortex, located in area CA1, firing of presynaptic CA3 pyramidal neurons produces monosynaptic excitatory postsynaptic potentials (EPSPs) followed rapidly by feedforward (disynaptic) inhibitory postsynaptic potentials (IPSPs). We previously reported that a

depolarization of the reversal potential for GABAA receptor-mediated currents at the feedforward inhibitory synapses induced by paired pre- and postsynaptic action potential firing can produce a long lasting increase in the amplitude of Schaffer collateral evoked postsynaptic potentials, brought about by the reduced shunting of excitatory currents. In the present study, we investigate the pathway specificity of this disinhibition-mediated LTP. We find that the underlying depolarization of EGABA is not restricted to the paired pathway, with EGABA depolarizing to the same extent at an unpaired control pathway. However, the overall strength of GABAergic transmission is maintained at the unpaired pathway by a heterosynaptic increase in conductance induced by the pairing protocol. As a result, the LTP resulting from pairing-induced disinhibition remains confined to the paired pathway. Combined with our previous demonstration that pairing- induced GABAergic plasticity is NMDA receptor-dependent in slices from adult rats, these results demonstrate that disinhibition-mediated LTP shares the same fundamental properties that have made glutamatergic LTP a leading model of synaptic plasticity, suggesting the possibility of a previously unrecognized target for therapeutic intervention in disorders linked to memory deficits, as well as a potentially overlooked site of LTP expression in other areas of the brain. 52

3.1 Introduction: Glutamatergic LTP is pathway specific

Plasticity of synaptic connections between neurons in the hippocampus is thought to play a central role in learning and memory (Hebb, 1949). Synaptic plasticity can be induced by patterned electrical stimulation at a number of synapses in the hippocampus, including the excitatory synapses of the trisynaptic pathway, as well as at certain excitatory onto interneuron synapses, and inhibitory onto pyramidal neuron synapses (reviewed by Nelson and Turrigiano,

2008). LTP of CA3-CA1 glutamatergic transmission has become the leading model of synaptic plasticity, in part because of its dependence on NMDAR activation (Harris et al., 1984), which provides a mechanism for associating pre- and postsynaptic action potential firing, and which is also required in vivo for hippocampal-dependent spatial and temporal learning (Morris et al.,

1986; Tsien et al., 1996; Huerta et al. 2000).

We recently demonstrated that disinhibition expressed at feedforward inhibitory synapses onto pyramidal cells in area CA1 reduces the shunting of excitatory currents, leading to an increase in the amplitude of Schaffer collateral-mediated postsynaptic potentials which we termed disinhibition-mediated LTP (Ormond and Woodin, submitted). This form of LTP, induced with paired pre- and postsynaptic action potential firing and expressed through a reduction in potassium/chloride co-transporter KCC2 activity and subsequent intracellular chloride accumulation (Woodin et al., 2003), is dependent on NMDA receptor activation in slices from adult rats (Ormond and Woodin, submitted), suggesting that it too might play a role in hippocampal-dependent learning and memory. 53

Synaptically-based neurophysiological theories of learning postulate that there are changes resulting from neural activity which are long-lasting and confined to specific sets of synapses

(Hebb, 1949; Marr, 1969; Albus, 1971). The expression of glutamatergic LTP is, generally speaking, input specific, meaning that LTP is restricted to the pathway at which it was induced

(Andersen et al, 1977), although it does spread to other synapses within ~70 um of the site of induction (Engert and Bonhoeffer, 1997). There have been a number of reports that pathway specificity also involves a heterosynaptic depression in the strength of glutamatergic transmission, further differentiating the gain of test and control pathways (Abraham and

Goddard, 1983; Scanziani et al., 1996; Lynch et al 1977).

We wondered whether pairing-induced GABAergic plasticity and the resulting disinhibition-mediated LTP might also display pathway specificity. We found that the depolarization of EGABA induced by pairing postsynaptic spiking with extracellular stimulation in

S. pyramidale, the main site of feedforward inhibitory synapses, was not confined to the paired

pathway. However, the overall strength of inhibition was maintained at control pathways by a

heterosynaptic increase in synaptic conductance induced by the pairing protocol. We found that

this heterosynaptic potentiation of GABAA receptor-mediated transmission prevents the spread

of disinhibition-mediated LTP to control pathways. 54

3.2 Results: Heterosynaptic increases in GABAAR synaptic conductance confines LTP to the paired pathway

To investigate the pathway specificity of pairing-induced GABAergic plasticity, we paired extracellular stimulation in S. pyramidale, the site of most feedforward inhibitory synapses onto CA1 pyramidal neurons (Pouille and Scanziani, 2001), with postsynaptic spiking evoked with intracellular current injection through the whole-cell recording pipet (300 pairings at

5 Hz; the postsynaptic cell typically fired two APs per pairing within 5 ms of the preceding presynaptic stimulation; Woodin et al., 2003; Ormond and Woodin, submitted). We monitored a second unpaired control pathway with extracellular stimulation also in S. Pyramidale, but on the opposite side of the recorded neuron [Figure 17a inset; separate pathways were confirmed by the lack of short-term plasticity between them (Figure 22, methods)]. Pairing induced a depolarization of EGABA at the paired pathway (Figure 17a left) as previously reported (Woodin et al., 2003; Ormond and Woodin, submitted). At the unpaired pathway, EGABA depolarized to a similar magnitude and with a similar time course (Figure 17a left), indicating that this form of synaptic plasticity is not pathway specific. However, while synaptic conductance was not significantly changed at the paired pathway, as previously reported in single pathway experiments (Ormond and Woodin, submitted), synaptic conductance was significantly increased at the unpaired pathway (Figure 17a right). This increase in synaptic conductance made IPSPs more depolarizing when evoked while holding the membrane potential below EGABA, but more hyperpolarizing when holding the membrane potential above EGABA (Figure 17b). Thus, at control pathways, the effect of EGABA depolarization on inhibitory strength is negated as the cell approaches action potential threshold, restricting the disinhibition to the paired pathway. 55

Having shown previously that feedforward inhibition reduces PSP amplitude recorded at

the soma, we expected that the increased GABAergic conductance at the unpaired pathway

would help to restrict LTP of Schaffer collateral-mediate PSPs to the paired pathway. When LTP

was recorded with GABAergic inhibition blocked pharmacologically with gabazine (6 µm), LTP

was weak and not long-lasting, and spread to the unpaired pathway (Figure 18a). We previously

showed that that this pairing protocol induces only weak glutamatergic LTP in single pathway

experiments (Ormond and Woodin, submitted), likely due to the large number of pairings which

lead to an overlapping NMDA receptor dependent long-term depression (LTD) which negates

the potentiation (Christie et al., 1996; Wittenberg and Wang, 2006). To maintain the

independence of the paired and control pathway, we used weaker extracellular stimulation than

in our previous single pathway study, which likely contributed to the greater transience of

glutamatergic LTP observed here. The spread to the unpaired pathway is somewhat surprising,

though pairing induced LTP has been shown previously to spread to nearby synapses (Engert and

Bonhoeffer, 1997) and to nearby cells (Schuman and Madison, 1994). It is possible that at higher stimulation intensities, or with a different number of pairings, LTP would have shown pathway specificity. In contrast to LTP with inhibition blocked, when inhibition was left intact, the LTP induced at the paired pathway was significant and lasted for the duration of the recording, indicating that inhibitory plasticity was responsible for the increase in PSP amplitude (Figure

18b). Furthermore, at the unpaired control pathway, no significant change in PSP amplitude was observed, indicating that the increased GABAergic conductance does indeed oppose the effect of

EGABA depolarization, restricting LTP expression to the paired pathway. 56

We wanted to verify this effect of increased GABAergic conductance by examining it in

isolation from glutamatergic LTP and EGABA depolarization. We had previously used such a

strategy to determine the effect of EGABA depolarization on PSP amplitude by blocking glutamatergic LTP with an inhibitor of CaMKII (AIP) which spared the GABAergic plasticity

(Chapter 2; Ormond and Woodin, submitted). For the present study, we wanted to find an induction protocol or pharmacological blocker that would prevent glutamatergic LTP and EGABA

depolarization, but that would not block the increase in GABAergic conductance induced at the

unpaired pathway. We first tested the Ca2+-dependence of the conductance increase by applying

the calcium chelator BAPTA through the patch pipette. This completely blocked both the

depolarization of EGABA (Figure 19a left; no significant difference relative to baseline, paired

pathway p = 0.613 at 45 min, unpaired pathway p = 0.590) and the increased GABAergic

conductance at the unpaired pathway (Figure 19a right; conductance showed a trend towards a

decrease relative to baseline at 45 min, p = 0.112), confirming the Ca2+-dependence of both forms of plasticity. We next tested the ability of an induction protocol consisting of postsynaptic spiking alone (i.e. no presynaptic stimulation) to induce plasticity. This protocol failed to induce

EGABA depolarization (Figure 19b left; both pathways showed an insignificant trend towards

hyperpolarization: paired pathway p = 0.087, unpaired pathway p = 0.149 at 45 min), and

prevented the increase in synaptic conductance at the unpaired pathway (Figure 19b right; no

significant change relative to baseline, p = 0.286 at 45 min). The L-type calcium channel blocker

nimodipine has previously been reported to block the depolarization of EGABA induced by pairing

in hippocampal cell cultures (Woodin et al, 2003). With nimodipine in the bath, pairing did

produce a slight EGABA depolarization at both pathways (Figure 19c left) that was neither

significantly different from baseline (paired pathway p = 0.162 at 50 min, unpaired pathway p = 57

0.137 at 45 min) nor from the non-drug treated cells in Figure 17a (paired pathway p = 0.252 at

50 min, unpaired pathway p = 0.083 at 45 min), suggesting that EGABA depolarization in

hippocampal slices from adults depends only partially on Ca2+ entry through L-type Ca2+

channels. Nimodipine did, however, completely block the increase in GABAergic conductance at

the unpaired pathway (Figure 19c right; no significant change relative to baseline, p = 0.551 at

45 min; statistically significant difference relative to non-drug treated cells from Figure 17a right, p = 0.023), indicating that L-type calcium channel opening is required for at least some aspects of GABAergic plasticity in the adult (Figure 19c right). Lastly, we tested the ability of the NMDAR antagonist APV (25 µm) to prevent the increase in GABAergic conductance. We previously found that APV blocked the pairing-induced depolarization of EGABA in slices from adult rats (Ormond and Woodin, submitted). This was confirmed here at both the paired pathway

(Figure 20a left, p = 0.169) and the unpaired pathway (p = 0.326). The change in conductance at

the unpaired pathway was slightly reduced, but was still statistically significant (Figure 20a

right) suggesting that calcium entry through L-type calcium channels is sufficient for this form of

plasticity.

Given its ability to block EGABA depolarization as well as glutamatergic LTP (Harris et al.

1984), we chose to use APV to isolate the effect of increased GABAergic conductance at the

unpaired pathway on Schaffer collateral-mediated PSPs. We first verified that APV blocked isolated glutamatergic LTP (i.e. with gabazine in the bath). No significant potentiation was seen at either the paired or unpaired pathway (Figure 20b), suggesting that the transient potentiation observed in LTP recordings without APV (Figure 18b) was simply weakly induced conventional

NMDAR-dependent LTP. With the increased GABAergic conductance at the unpaired pathway 58

being the only plasticity spared by APV in our experiments, we expected that pairing under conditions of intact glutamatergic and feedforward GABAergic transmission would lead to a depression of PSP amplitude at the unpaired pathway, with little change at the paired pathway.

Surprisingly, however, PSPs recorded at the resting membrane potential (measured at the outset of the recording; see Methods), there was no significant change seen at either pathway (Figure

20c top left). However, as the membrane was depolarized toward AP threshold, the effect at the unpaired pathway became apparent. Specifically, when we measured PSP amplitude from the membrane potential at which PSPs had reversed polarity prior to plasticity induction (i.e. the PSP reversal potential, at which excitation and inhibition cancelled each other out completely), PSPs became hyperpolarizing, indicating that the strength of inhibition was considerably increased relative to excitation (Figure 20c right). The reason for this discrepancy in the effect of increased

GABAergic conductance at the two membrane potentials is a result of the increasing driving force for GABAergic currents as the membrane is depolarized. As can be seen in the example recording (Figure 20c bottom left), the effect of increased GABAergic conductance is enhanced as the membrane potential is increased. These recordings confirm that increased synaptic conductance at GABAergic synapses of the unpaired pathway underlies the pathway specificity of disinhibition-mediated LTP. 59

3.3 Figures

Figure 17. Depolarization of EGABA spreads to unpaired synapses, but inhibitory strength is maintained at those sites through increased GABAergic conductance.

a, Left: Change in EGABA at the paired and unpaired pathways (n= 6). Right: Change in relative GABAergic conductance at the paired and unpaired pathways for the same cells. Black circles represent paired pathway. Open circles represent unpaired pathway. * denotes statistical significance from control pathway (* p < 0.05) from that time point on. † denotes statistical significance from baseline († p < 0.05) from that time point on. Paired activity denoted by arrow. b, Left: IPSP amplitude plotted against Vm for the unpaired pathway before (black triangles) and after (open triangles) pairing for one cell from a. Right: Example traces from the same cell before (1) and after (2) pairing. Scale bars: 1 mV, 10 ms. 60

Figure 18: Pairing-induced LTP is pathway-specific only when feedforward inhibition is intact. a, Change in EPSP amplitude at the paired and unpaired pathways (n= 6). Black circles represent paired pathway. Open circles represent unpaired pathway. b, Change in mixed PSP amplitude (calculated from RMP) at the paired and unpaired pathways (n= 6). Black circles represent paired pathway. Open circles represent unpaired pathway. * denotes statistical significance from control (* p < 0.05) from that time point on. † denotes statistical significance from baseline († p < 0.05) from that time point on. § denotes statistical significance from baseline (§ p < 0.05) for that time point only. Paired activity denoted by arrow. 61

Figure 19: Pairing-induced heterosynaptic increases in GABAergic conductance require Ca2+ entry through L-type calcium channels.

a, Left: Change in EGABA at the paired and unpaired pathways with BAPTA applied intracellularly through the patch pipet (n= 4). Right: Change in relative GABAergic conductance at the paired and unpaired pathways for the same cells. b, Left: Change in EGABA at the paired and unpaired pathways induced by a plasticity induction protocol consisting of postsynaptic spiking alone (i.e. no presynaptic stimulation; n= 7). Right: Change in relative GABAergic conductance at the paired and unpaired pathways for the same cells. c, Left: Change in EGABA at the paired and unpaired pathways induced in the presence of bath applied nimodipine (nimodipine applied for the duration of the recording; n= 5). Right: Change in relative GABAergic conductance at the paired and unpaired pathways for the same cells. Black circles represent paired pathway. Open circles represent unpaired pathway. 62

Figure 20: Increased GABAergic conductance at the unpaired pathway underlies the pathway specificity of mixed LTP.

a, Left: Change in EGABA at the paired and unpaired pathways (n= 6) with NMDARs antagonized. Right: Change in relative GABAergic conductance at the paired and unpaired pathways for the same cells. b, Change in EPSP amplitude at the paired and unpaired pathways (n= 8) with NMDARs antagonized. c, Left: Change in mixed PSP amplitude (calculated from RMP) at the paired and unpaired pathways (n= 6) with NMDARs antagonized. Right: Change in mixed PSP amplitude (calculated from pre-induction 63

reversal potential) at the paired and unpaired pathways in the same cells at left. d, PSP amplitude plotted against Vm for the unpaired pathway before (black triangles) and after (open triangles) pairing for one cell from C). Inset: Example traces from the same cell before (1) and after (2) pairing. Black circles represent paired pathway. Open circles represent unpaired pathway. * denotes statistical significance from control (* p < 0.05) from that time point on. † denotes statistical significance from baseline († p < 0.05) from that time point on. Paired activity denoted by arrow. Scale bars: 1 mV, 10 ms. 64

Supplementary Figure 21: The majority of recorded inhibition in mixed EPSP/IPSP recordings was feedforward.

The degree to which inhibition affected PSP amplitude was determined from the slope of the PSP vs. Vm graph (Vm has no effect on EPSP amplitude within the measure range when inhibition is blocked pharmacologically; Ormond and Woodin, submitted). The slope was then recalculated after CNQX was washed in to block glutamatergic transmission. The slope after CNQX divided by the slope before its application thus gave an indication of the ratio of disynaptic to monosynaptic inhibition in these mixed excitatory/inhibitory recordings. Experimental groups: 1, two-pathway mixed LTP paired pathway (Figure 18a); 2: 2-pathway mixed LTP unpaired pathway (Figure 18a), 3: 2-pathway mixed LTP with APV paired pathway (Figure 20c), 4: 2-pathway mixed LTP with APV unpaired pathway (Figure 20c).

65

Supplementary Figure 22: Pathways in the 2-pathway experiments were independent.

Top right: Short-term plasticity within pathways (i) and between pathways (ii) for all 2 pathway groups [a value of 1 indicates no short-term plasticity; 1: 2-pathway GABA STDP (Figure 17a), 2: 2-pathway GABA STDP w/ intracellular BAPTA (Figure 19a), 3: Postsynaptic spiking only GABA STDP (Figure 19b), 4: 2-pathway GABA STDP in nimodipine (Figure 19c), 5: 2- pathway GABA STDP in APV (Figure 20a), 6: 2-pathway glutamatergic LTP (Figure 18b), 7: 2- pathway glutamatergic LTP in APV (Figure 20b), 8: 2-pathway mixed LTP (Figure 18a), 9: 2- pathway mixed LTP in APV (Figure 20c). Inset: Sample recording demonstrating the protocol used. Scale bars: 50 pA, 25 ms.

66

3.4 Discussion: Both classic and disinhibition-mediated LTP involve the interaction of synapse specific and cell wide plasticity

To our knowledge, this is the first report of a pathway-specific enhancement of excitatory

transmission by inhibitory plasticity. In our LTP experiments, we took great care to make sure

that the inhibition being elicited was done in a way that was physiological so as to avoid

overestimating the impact of inhibition on EPSP amplitude. Inhibition was activated

disynaptically in these recordings, thereby maintaining the EPSP/IPSP delay and preventing the

recruitment of additional non-feedforward interneurons, and was verified as such at the end of

experiments. Furthermore, under our recording conditions, the driving force for GABAAR-

mediated currents closely matched what we have previously recorded with gramicidin perforated

patch recording (Ormond and Woodin, submitted); this is particularly important, as a number of

studies examining feedforward inhibition increased the driving force in whole-cell experiments

by lowering intracellular Cl- to unphysiological levels, thereby artificially enhancing the effect of

inhibition on excitatory transmission (Pouille and Scanziani, 2001) and LTP induction (Bissiere

et al., 2003). Lastly, we used slices from adult rats in all experiments. We believe the use of

slices from adults is crucial for studying the cellular mechanisms underlying learning and

memory, as ongoing neural development complicates the interpretation of data from younger

animals. Highlighting this fact, gramicidin perforated-patch recording in our previous study showed that EGABA hyperpolarizes a further 15 mV between 3 and 7 weeks of age, indicating that

the developmental changes involved in strengthening inhibition last much longer than commonly

assumed on the basis of KCC2 expression (Rivera et al., 1999). As such, we think our study

provides strong evidence of a role for disinhibition in pathway-specific strengthening of CA1- 67

CA3 transmission beyond the end of development, when these processes are thought to be

primarily involved in the long-term storage of information.

Together with our previous results showing that disinhibition-mediated LTP is NMDAR

dependent (Ormond and Woodin, submitted), the present results show this form of LTP shares

the three main features of classic glutamatergic LTP: pathway-specificity, cooperativity and associativity [these last two properties are due to the NMDAR dependence of LTP, and though they have frequently been treated as separate in the literature, they are in fact the same property

(McNaughton et al., 1978; McNaughton, 2003)]. The mechanism underlying this pathway

specificity is unusual, though, in that it does not actually result from a confinement of the cellular

changes responsible for the potentiation, but instead requires a separate form of plasticity to be

expressed at control pathways. There are, nevertheless, some parallels with classic LTP as well.

Heterosynaptic plasticity does in fact accompany classic LTP expression, where it takes the form

of a depression of glutamatergic transmission, enhancing the contrast between test and control

pathways (Lynch et al., 1977; Scanziani et al.,. 1996). Furthermore, in a reversal of the situation

for disinhibition-mediated LTP, the heterosynaptic depression accompanying classic LTP is

apparently a cell wide-phenomenon which is counteracted by LTP expression at the test pathway

(Scanziani et al., 1996). Lastly, the heterosynaptic GABAergic plasticity reported here not only

maintains inhibitory strength but actually strengthens it as the membrane potential is depolarized

towards AP threshold (Figure 17b), thereby enhancing the contrast between pathways.

While our data suggests that the pairing-induced depolarization of EGABA is a cell-wide

process, it is entirely possible that the depolarization of EGABA was confined to the soma as we

primarily stimulated somatic inhibitory synapses. Such a compartmentalized Cl- regulation has 68

been demonstrated. In pyramidal cells of the cerebral cortex, for instance, GABAergic inhibition

is strongly depolarizing at the axon-initial segment, but not in the soma or dendrites, due to the

absence of KCC2 expression (Szabadics et al., 2006) and the presence of the Na–K–2Cl

cotransporter NKCC1 (Khirug et al., 2008) in the axon. In contrast, EGABA is hyperpolarizing at

the soma, and becomes progressively more negative with distance into the dendrites (Khirug et

al, 2008). In preliminary experiments, pairing-induced EGABA depolarization did appear confined

to the paired pathway when the unpaired pathway was located near the distal dendrites in S.

radiatum rather than in the cell body layer (unpublished results). Future work will be needed to

verify this compartmentalization, and to determine its significance.

Given that disinhibition-mediated LTP shares the major features of classic glutamatergic

LTP, it seems reasonable to speculate that it may be involved in memory. Unfortunately,

investigating the importance of this form of plasticity in memory will be a difficult task given the

lack of specific pharmacological agents for blocking changes in KCC2 activity. It should be

noted, however, that the NMDAR blockers used in behavioural experiments linking classic LTP

to memory are not particularly specific either, also blocking the disinhibition-mediated LTP

described here, as well as other forms of GABAergic plasticity (Stelzer et al., 1994; Lu et al.,

2000; Caillard et al., 1999; Grunze et al., 1996; Ormond and Woodin, submitted). A number of molecules that specifically inhibit KCC2 were recently identified in a high-throughput screening study based on 234,000 compounds (Delpire et al., 2009); hopefully, some of these will become available for behavioural testing.

A recent study (Carvalho and Buonomano, 2009) showed that glutamatergic and

GABAergic plasticity can have very different effects on the input/output functions of CA1 69

pyramidal neurons. Plasticity at glutamatergic synapses was demonstrated to shift the threshold

(i.e. the EPSP slope required to reach AP threshold) without affecting its gain, while inhibitory plasticity was shown to change both the threshold and the gain (i.e. the rate of change the function). Balanced changes in excitatory and inhibitory synaptic strength therefore allowed for changes to the gain of the function while maintaining a constant threshold. However, changes in

EGABA were not examined, and it is therefore uncertain how such changes might affect the I/O function. Determining how simultaneous changes to synaptic conductance and EGABA affect the

I/O function may be key to determining if there are computational benefits in expressing LTP through cell or soma-wide changes in EGABA combined with heterosynaptic changes in

GABAergic conductance, as observed here. 70

4. General discussion of the main findings: A role for disinhibition in memory

4.1 Major conclusions

The findings presented in this thesis show that disinhibition at sites of feedforward

transmission in the CA1 region of rat hippocampal slices produces a long-term potentiation of

CA3-CA1 transmission that shares with classic CA1 glutamatergic LTP the properties of

cooperativity, associativity, and pathway specificity. As discussed in Chapter 1, the first two

properties are actually the same and relate to classic LTP’s dependence on NMDAR activation

during plasticity induction, which requires a threshold depolarization to relieve the Mg2+ block

and allow the influx of Ca2+. This requires cooperativity amongst inputs to produce sufficient

depolarization. The inputs are thereby associated with one another through their simultaneous

NMDAR currents and subsequent co-expression of LTP. Because disinhibition-mediated LTP, like pairing-induced glutamatergic LTP (Markram et al., 1997; Magee and Johnston, 1997), requires both the firing of postsynaptic action potentials and requires NMDAR activation for its induction, it clearly displays the associativity/cooperativity of classic LTP. In contrast, the third property of classic LTP, pathway specificity, is expressed quite differently during disinhibition- mediated LTP. Whereas the cellular and molecular changes underlying classic and pairing- induced glutamatergic LTP, namely the increased conductance of existing AMPARs and the recruitment of additional AMPARs to the postsynaptic membrane (but see Chapter 1 for a discussion of presynaptic sites of LTP expression), are restricted to the synapses active during the induction, the depolarization of EGABA responsible for disinhibition is not similarly confined.

Rather, a heterosynaptic increase in GABAergic synaptic conductance maintains the strength of

inhibition in the face of a reduced driving force for GABAAR-mediated currents at control 71

pathways, thereby restricting disinhibition-mediated LTP to the paired pathway. However, there

are also some similarities in the pathway-specificity of both forms of LTP. Heterosynaptic plasticity does in fact accompany classic LTP expression, where it takes the form of a depression of glutamatergic transmission, enhancing the contrast between test and control pathways (Lynch et al., 1977; Scanziani et al.,. 1996). Furthermore, in a reversal of the situation for disinhibition-

mediated LTP, the heterosynaptic depression accompanying classic LTP is apparently a cell

wide-phenomenon which is counteracted by LTP expression at the test pathway (Scanziani et al.,

1996). Lastly, the heterosynaptic GABAergic plasticity reported here not only maintains

inhibitory strength but actually strengthens it as the membrane potential is depolarized towards

AP threshold, producing a depression of Schaffer collateral-mediated transmission at control

pathways similar to that accompanying classic LTP.

Our findings are supported by three previous reports showing that disinhibition can

enhance the efficacy of CA3-CA1 glutamatergic transmission, either through a contribution to

LTP expression (Stelzer et al., 1994) or as the mechanism underlying E-S potentiation (Lu et al.,

2000; Chevaleyre and Castillo, 2003), the increase in postsynaptic excitability normally co-

expressed with classic LTP. Our results extend these previous findings in four important ways.

First, we have shown that disinhibition produces LTP under conditions where the activation of

inhibitory synapses is restricted to the feedforward connections, and where this activation is

solely disynaptic, being activated by the firing of presynaptic CA3 pyramidal cells. This is in

contrast to the previous studies, which examined the effect of disinhibition without restricting the

activation of inhibitory synapses to those synapses mediating feedforward/ disynaptic inhibition.

Because there is no monosynaptic inhibitory component in our mixed excitatory/inhibitory 72

recordings, the amount of inhibition elicited is less than in the previous studies, and it is delayed with respect to the excitation. Thus, we show that even when inhibition is restricted to the feedforward connections, and the amount of inhibition is reduced to more physiological levels and delayed with respect to excitation, disinhibition can still exert a powerful influence on the efficacy of excitatory transmission. Second, our demonstration of the pathway specificity of disinhibition-mediated LTP, a necessary property for synaptic plasticity-dependent storage of

information (Hebb, 1949), is in fact the first such demonstration for increases in the efficacy of

excitatory transmission caused by disinhibition. Third, the disinhibition-mediated LTP described here is induced with physiologically realistic low-frequency pairing, as opposed to tetanic presynaptic stimulation. This is particularly significant given that a report indicating that previous demonstrations of disinhibition in CA1 were the result of unphysiological tetanic stimulation, and that a more realistic theta-burst stimulation instead produced inhibitory potentiation (Patenaude et al., 2003). Why we don’t see the changes in synaptic conductance at the paired pathway described in these previous studies is not entirely clear, but likely results from the more restricted activation of inhibitory synapses in our study. It would be interesting, and worthwhile, to investigate pairing-induced plasticity at other inhibitory synapses, particularly those along the pyramidal neuron dendrites which are the synapses most likely investigated in the previous studies based on positioning of the stimulating electrode.

4.2 Feedforward inhibition in vivo

Indications that interneurons might contribute differentially to the temporal coordination

of pyramidal cells has come largely from in vitro recordings in brain slices of rats. In particular,

the demonstration that feedforward inhibition follows Schaffer-collateral mediated excitation by 73

less than 2 ms in the hippocampal slice (Pouille and Scanziani, 2001) led us to hypothesize that

disinhibition at feedforward connections might be involved in LTP expression. However, an

involvement of the disinhibition described here in memory formation requires that the neurons

mediating feedforward inhibition in the slice, the PV-positive basket cells, also mediate feedforward inhibition in vivo. Is this the case?

To answer this question, it is important to note that in slice, there is little to no intrinsic or spontaneous neural activity, whereas in vivo, the hippocampus, and in fact the whole cortex, displays highly coordinated neural activity as demonstrated by network oscillations (Soltesz

Deschenes, 1993). There are three main types of network oscillation: theta oscillations (4 to 10

Hz), gamma oscillations (30 to 80 Hz), and sharp-wave associated ripples (100 to 200 Hz)

(reviewed by Klausberger and Somogyi, 2008). Theta oscillation and gamma oscillations typically occur simultaneously, and are observed during exploratory behavior (i.e. spatial navigation), memory tasks, and during rapid-eye-movement sleep. Sharp-wave associated ripples occur during resting, consummatory behavior, and slow-wave sleep, and are thought to support

offline replay and consolidation of previous experiences (Foster and Wilson, 2006; Diba and

Buzsaki, 2007). Thus, asking whether the timing of PV-positive basket cell activity in vivo is

consistent with observations in the slice is too simplistic. Instead, we must ask how it is

temporally coordinated during each of the different hippocampal states.

One approach to understanding the timing with which the various subtypes of

interneurons neurons fire in vivo has been to record from identified interneurons while

simultaneously recording the local field potential (LFP), allowing a determination of the phase

of firing with respect to extracellular oscillations. Unfortunately, this technique, which relies on 74

juxtacellular labeling of the recorded neuron and post-hoc identification with histological methods, requires for technical reasons that the animal be anaesthetized, which fundamentally alters the theta rhythm as demonstrated by sensitivity to muscarinic blockers which do not block theta in awake animals (Kramis et al., 1975). Using this method, it has been shown that during theta oscillations, PV-positive basket cells fire almost completely out of phase with CA1 pyramidals (Klausberger et al., 2003). In the cited study, the difference in the phase of firing of

PV-positive basket cells and CA1 pyramidals corresponded to a delay in actual time of approximately 76 ms, far too long for the generated fast IPSP to interact with the EPSP responsible for the pyramidal cell firing. Importantly, this timing is indeed consistent with data from awake animals and is likely due to the fact that the firing of CA1 pyramidals during theta doesn’t reflect firing of excitatory inputs from CA3. Instead, the majority of CA1 firing may be due to the rhythmic excitation of the distal dendrites by the entorhinal afferents, which are believed to play the most important role in the generation of extracellular theta (Holsheimer et al., 1982; Buzsáki et al., 1983, Leung, 1984a, Leung 1984b and Brankack et al., 1993). Data from behaving animals showing that CA3 pyramidal cells fire out of phase with CA1 discharge

(Csicsvari and Buzsaki 1999) instead suggest that the primary role of CA3 may be to help entrain

CA1 to the theta rhythm through rhythmic feedforward inhibition. While only a small percentage of CA3 pyramidal cells are active during a given theta cycle [supported by the small magnitude of the sink in CA1 stratum radiatum (Buzsáki et al., 1986; Brankack et al. 1993)], the strong coupling between CA3 pyramidals and feedforward interneurons in CA1 suggests that significant inhibition should occur at the soma (Buzsáki, 2002). The absence of a large current source in the cell body layer does call this into question, though. However, this may be explained by the fact that during theta, most pyramidal cells are silent, and thus there is not a large driving 75

- force for GABAAR-mediated Cl currents, which have an equilibrium potential close to rest

(Harris et al., 2000). Nevertheless, if CA3’s sole role during theta is entrainment of the CA1

pyramidal population, it is hard to imagine what role pathway specific depolarization of EGABA at

the feedforward connections might have to play.

The phenomenon of phase precession, however, indicates that CA3 pyramidals and their

feedforward interneurons in CA1 may in fact play a central role in the encoding of specific

information during the theta rhythm. The hippocampus in rodents has primarily been studied as it

relates to the encoding spatial memory, one of its central functions. Many principal neurons in

the hippocampus are place cells, meaning that they fire primarily at certain locations in the

physical environment (O’Keefe and Dostrovsky, 1971; O’Keefe, 1976). Theta-phase precession

refers to a change in place cell spike timing as the animal crosses the spatially restricted firing

field of the neuron, with the cell firing at progressively earlier phases of the extracellular theta

rhythm until it exits the field (O’Keefe and Recce, 1993; Skaggs et al., 1996). Until recently,

phase precession was thought to occur exclusively in the hippocampus, where it has been

observed in all subregions (O’Keefe and Recce, 1993; Skaggs et al., 1996). However, it has now

been observed in the grid cells of entorhinal cortex layer II, but importantly, not by the grid cells

of entorhinal cortex layer III which constitute the direct entorhinal pathway to CA1 (Hafting et

al., 2008). Thus, while the direct entorhinal pathway may indeed mediate most of the excitatory

drive to CA1 during theta, the hippocampus inherits phase precession from the entorhinal inputs

to the trisynaptic pathway, indicating that the CA3 inputs to CA1 convey specific information

relating to the exact spatial position of the animal. Disinhibition in CA1 would therefore be well

positioned to exert memory encoding functions during spatial navigation. 76

Gamma oscillations are also prevalent in the hippocampal formation, where they have

been proposed to assist in encoding and retrieval of memory traces (Bragin et al., 1995; Lisman

and Idiart, 1995; Chrobak and Buzsáki, 1998; Hasselmo et al., 1996). Although gamma patterns can sometimes occur on their own (Buzsáki et al., 1983; Leung, 1998; Fisahn et al., 1998), their power correlates with the power of theta oscillations, indicating an important relationship

between the two (Buzsaki 2003). The major inputs responsible for gamma waves in the CA1 area

are basically the same as those which generate theta (Bragin et al., 1995; Leung, 1998; Csicsvari

et al., 1999; Wang and Buzsáki, 1996; Penttonen et al., 1998; Fisahn et al., 1998; Charpak et al.,

1995). The increased discharge of the interneuron population and the peak power of gamma

oscillation occur on the same phase of theta (Buzsaki et al., 2003) indicating that interneuron

activity is critical for the generation of physiological gamma patterns (Whittington et al., 1995;

Wang and Buzsáki, 1998). Furthermore, the gamma phase-locked discharge of CA3 pyramidal

cells is responsible for gamma oscillations in the CA1 region as well, evident in part by the

monosynaptic activation of activation of CA1 interneurons by CA3 pyramidal neurons (Csicsvari

et al., 2003). Under average conditions, feedforward inhibition appears to curtail the excitatory

drive of CA1 pyramidal cells, as it is out of phase with the discharge of CA1 pyramidals

(Csicsvari et al., 2003), suggesting that CA3 is not the primary source of excitatory drive for

CA1 pyramidals. However, the observed increase in gamma coherence between CA3 and CA1

during retrieval suggests that the synchrony between CA3 and CA1 unit firing increases at short

(monosynaptic) and gamma time scales during the retrieval process. This would suggest that

feedforward inhibition might, under “normal” conditions be primarily involved in generation of

the rhythm/entrainment, but during retrieval, gate the Schaffer collateral mediated transmission 77

which may become more important for CA1 output, analogous to the situation described above

during the simultaneous theta oscillation.

In the absence of theta oscillations (e.g., waking immobility, slow wave sleep),

irregularly occurring sharp waves (SPWs) are the most dominant network pattern observed in

CA1 stratum radiatum (Buzsaki et al., 1983). CA3 plays a particularly prominent role in the

generation of SPWs, as the release of subcortical neuromodulators is substantially reduced, and,

as a consequence, the spread of activity in the CA3 collateral system is disinhibited (Hasselmo et

al., 1995). Furthermore, during SPWs, neurons in layers II–III of the entorhinal cortex are

relatively silent (Chrobak and Buzsaki, 1994), in contrast to theta/gamma, highlighting the

importance of CA3 in driving CA1. Both CA3 and CA1 pyramidal cells fire synchronously

during sharp waves (Buzsaki, 1986; Csicsvari et al., 2000), and the strong excitatory input from

the CA3 region triggers short-lived fast (140–200 Hz) oscillatory patterns [sharp wave associated

ripples (SWRs)] in the CA1 region (O'Keefe and Nadel, 1978; Suzuki and Smith, 1988; Buzsaki

et al., 1992; Ylinen et al., 1995; Draguhn et al., 1998; Csicsvari et al., 1999a; Csicsvari et al.,

1999b; Traub and Bibbig, 2000) due to recruitment of CA1 basket and chandelier cells (Buzsaki

et al., 1992; Ylinen et al., 1995). Because CA3-CA1 neuronal activity during sleep occurs frequently in synchronous bursts associated with SWRs, it has been suggested that reactivation of neuronal assemblies is confined mostly to SWRs. Furthermore, it has been hypothesized that the key pattern for consolidation of learned information is SWRs, not sleep per se (Buzsaki,

1989; Buzsaki, 1996; Nadasdy et al., 1999; Kudrimoti et al., 1999). Indeed, there are indications that organized ensembles of CA3 and CA1 cells fire in SWRs (Nadasdy et al., 1999, Kudrimoti et al., 1999; Csicsvari et al., 2000; Lee and Wilson, 2002; Lee and Wilson, 2004), some correlated with waking neuronal patterns (Kudrimoti et al., 1999). We conclude that disinhibition 78

at feedforward inhibitory synapses in CA1 would be well placed to exert effects on encoding and retrieval during each of the three main hippocampal rhythms.

4.3 Plasticity during theta/gamma and SWRs

The hippocampus is thought only to temporarily stores memory traces of waking experience, which are subsequently reactivated and transferred back to other cortical areas for permanent storage (Squire and Zola-Morgan, 1991; Buzsaki, 1989). The initial memory storage and reactivation/consolidation phases are believed to occur in separate behavioral states; labile memories are formed during waking exploration when the theta and gamma rhythms are prominent, while reactivation/consolidation occurs primarily during immobility, consummatory behaviors, and sleep when sharp waves occur (Buzsaki, 1989; Wilson and McNaughton, 1994).

Although indirect, several observations suggest the possible involvement of theta oscillations in synaptic plasticity. A number of in vitro and in vivo studies have reported that induction of long-term potentiation (LTP) is optimal when the time interval between stimuli is approximately 200 ms (corresponding to a 5 Hz theta rhythm; Larson and Lynch, 1986;

Greenstein et al. 1988). Furthermore, single tetanic trains (also called bursts) can induce LTP when delivered at the positive peak of theta measured in stratum radiatum (corresponding to the highest probability of pyramidal cell discharge; Fox et al., 1986; Buzsáki et al., 1983), both in carbachol-treated slices (carbochol is used to induces thet in slices) and anesthetized rats (Huerta and Lisman, 1996; Höschler et al., 1997). Conversely, stimulation of CA1 afferents at the theta trough can result in depotentiation of a previously potentiated synapse (Pavlides et al., 1988;

Huerta and Lisman, 1996; Höschler et al., 1997). Theta-associated somatic hyperpolarization 79

may provide silent periods necessary for the occurrence of complex spike bursts (Harris et al.,

2001), a condition necessary for synaptic potentiation (Paulsen and Sejnowski, 2000). Thus, theta oscillation may provide a mechanism for bringing together in time afferent-induced depolarization of pyramidal cell dendrites and dendritic invasion of fast spikes, key elements for the induction of synaptic plasticity. As a result, the repeated pairing of dendritic depolarization by the entorhinal input and the trisynaptically activated recurrent and Schaffer collaterals could lead to modification of the intrahippocampal associational pathways, allowing them to give rise to endogenous population patterns in the absence of the entorhinal inputs during subsequent non- theta activity (Buzsaki, 2002). Two lines of evidence suggest that the induction of GABAergic plasticity described here should be facilitated by the theta rhythm. The first is that it can be induced by pairing at 5 Hz, within the frequency range of theta, and that this appears to be the ideal frequency for its induction (Woodin et al., 2003; M. Woodin personal communication).

Secondly, it is dependent on NMDAR activation, which in vivo would likely require the maximal dendritic depolarization brought about by theta oscillations. Importantly, the pathway specificity of disinhibition-mediated LTP would mean that the theta-associated somatic inhibition required for the occurrence of complex spike bursts would be unaffected, preventing LTP already expressed at some synapses from disrupting induction at other synapses.

The hypothesis that the reactivation-consolidation process occurs during sleep in the hippocampus is supported by several studies showing that neuronal activity patterns observed during waking exploration recur in subsequent sleep. Cells that fire together during exploration also do so in sleep (Kudrimoti et al., 1999; Hirase et al., 2001), and cells with overlapping place fields show correlated firing during sleep (Wilson and McNaughton, 1994). Finally, it has been 80

demonstrated that sequential firing of place cells recurs during sleep, but only after they have

been repeatedly activated in the same order during behavior, as on linear tracks or circular mazes

(Skaggs and McNaughton, 1996b; Louie and Wilson, 2001; Lee and Wilson, 2002; Lee and

Wilson, 2004). Because CA3-CA1 neuronal activity during sleep occurs frequently in synchronous bursts associated with SWRs, it has been suggested that reactivation of neuronal assemblies is confined mostly to SWRs. Furthermore, it has been hypothesized that consolidation of learned information specifically requires SWRs, rather than simply sleep in general (Buzsaki,

1989; Buzsaki, 1996; Nadasdy et al., 1999; Kudrimoti et al., 1999). Indeed, there are indications that organized ensembles of CA3 and CA1 cells fire in SWRs (Nadasdy et al., 1999; Kudrimoti et al., 1999; Csicsvari et al., 2000; Lee and Wilson, 2002; Lee and Wilson, 2004), some correlated with waking neuronal patterns (Kudrimoti et al., 1999). Reactivation during ripples has been proposed to drive consolidation of previously formed memories through LTP induced in the cortex by the high frequency firing of presynaptic cells in CA1 (the main hippocampal output) (Buzsaki, 1989; Sutherland and McNaughton, 2000). Indeed, spiking during ripples has a temporal structure similar to that used to induce LTP (Buzsaki, 1986a) and ripple-like stimulation is effective in inducing plasticity in the deep layers of the entorhinal cortex, the major target of CA1 outputs (Yun et al., 2002). The emerging picture of LTP’s involvement in reactivation and consolidation is that prior LTP in CA3 and CA1, induced during theta, is likely required for the reactivation of cell assemblies in the hippocampus, which then induce LTP in cortical areas during the ripple events themselves. A specific role for disinhibition-mediated LTP in these processes is described below. 81

The involvement of glutamatergic LTP in a specific cellular component of spatial

memory has recently been described. When rats run repeatedly down a track, place fields, which

are initially symmetrical in shape, become negatively skewed, such that firing is low when they

enter the field, increases to a maximum, and then rapidly drops off (Mehta et al., 2000). This

rapid change in the shape of place fields is NMDAR-dependent indicating that it relies on the

induction of LTP. Modeling showed that this phenomenon likely results from the asymmetrical

shape of the spike-timing relationship for glutamatergic plasticity, i.e. presynaptic firing before

postsynaptic firing leads to LTP, becoming stronger the closer in time the pairing of pre- and

postsynaptic spikes are, and then switching immediately to depression as soon as the order of

firing is reversed. In contrast, pairing-induced GABAergic plasticity has a symmetrical spike- timing window, with depolarization of the reversal potential being induced so long as pre- and postsynaptic spiking occur within a time window of 20 ms, regardless of the order (Woodin et al., 2003). Given that the asymmetrical nature of the glutamatergic spike-timing curve can apparently make itself evident in the asymmetrization of place fields, it seems reasonable to suggest that GABAergic plasticity, if actually involved in memory, might reveal itself through phenomena that do not display asymmetry, or do not display a dependence on the order, but only relative timing of events. In fact, such observations have recently been made in studies demonstrating the occurrence of sharp wave ripples during waking exploration (termed eSWRs).

In one of the studies (O’Neill et al, 2006), cells which tended to fire together during theta showed a tendency to fire together during both eSWRs and the more typical SWRs occurring during immobility and sleep, similar to the replay during sleep of neuronal activity patterns observed during waking exploration discussed above. Importantly, cell pairs that did not show sequential firing bias during theta still showed reactivation of their co-firing patterns, indicating 82

that glutamatergic LTP was not responsible for the replay. The requirement for plasticity, but the lack of a requirement for temporal order between spiking neurons points to the involvement of pairing-induced GABAergic plasticity. In the other study (Sheng and Frank, 2008), pairs of CA1 cells representing overlapping locations in a novel environment were found to be initially more coactive and more precisely coordinated than are cells representing overlapping familiar locations specifically during brief, high-frequency events analogous to the eSWRs reported by

O’Neill et al. (2006). Similarly, across the population of cell pairs, no consistent bias in spike ordering was observed, indicating a lack of involvement by glutamatergic STDP. Again, the apparent involvement of an LTP-like mechanism, but the lack of a requirement for temporal order between spiking neurons suggests the involvement of pairing-induced GABAergic plasticity.

4.4 Future experiments

Our examination of the literature, particularly on memory phenomena in vivo, points to a number of avenues for further investigation of pairing-induced GABAergic plasticity. Firstly, there are a number of experiments to be done in slice. When we initially began the project, in addition to the pairing protocol described, we also used the uncorrelated pairing protocol described in Woodin et al. (2003) as a control. Surprisingly, this protocol, in preliminary experiments, appeared to hyperpolarize EGABA. While the importance of a counterbalancing plasticity is clear from theories of glutamatergic LTP/LTD (Sejnowski, 1977), the significance of the timing used to induce this hyperpolarization escaped us at the time. In fact, this timing, which can be described as anti-theta because of the perfectly out of phase firing of the pre- and postsynaptic firing (like our correlated pairing protocol, uncorrelated pairing was repeated at 83

theta frequency), has previously been implicated in memory extinction. Place cells for familiar

locations reverse their phase during rapid eye movement (REM) sleep, whereas place cells for novel locations, presumably not yet consolidated in the cortex, do not (Poe et al., 2000). As described above, reversing the phase of postsynaptic firing during ongoing theta can lead to LTD induction, leading to the hypothesis that the phase reversal observed during REM might erase memories already consolidated in the cortex, thereby making room for new memory traces in the hippocampus. Based on our observations, uncorrelated pairing-induced GABAergic plasticity might be able to subserve such a role. Unfortunately, our early experiments were performed in

APV as we did not realize that pairing-induced GABAergic plasticity in the adult would require

NMDAR activation. Thus, these experiments should be repeated in the absence of NMDAR blockers, and might form the starting point for a new project investigating the ability of

GABAergic plasticity to mediate bidirectional changes in the strength of Schaffer collateral- mediated transmission.

The original observation of GABAergic STDP was made in hippocampal culture. Unlike the slice, where particular regions of the hippocampus and classes of principal and inhibitory neurons can be targeted, in the culture, there is a heterogeneous mix of neurons which are generally only distinguished on the basis of the transmitter they release (i.e. GABA or glutamate). Thus, the demonstration of GABAergic plasticity in culture indicates that

GABAergic STDP is not restricted to a particular synapse, such as the feedforward interneuron/

CA1 pyramidal synapse examined here, but is rather a general phenomenon existing at most hippocampal inhibitory synapses, both onto excitatory and inhibitory cells. Furthermore, LTP is thought to be important at a number of other hippocampal synapses, perhaps most notably at the recurrent connections in CA3 and at the direct entorhinal connections in CA1. The strong 84

feedforward inhibition in CA3, which can be elicited by the spiking of a single CA3 pyramidal

neuron (Miles, 1990), makes this an obvious place to examine for the presence of disinhibition-

mediated LTP. Furthermore, the studies showing reactivation of cell pairs during exploratory

ripples (O’Neill et al, 2006; Sheng and Frank, 2008) point to the involvement of LTP within the

CA3. This is because the highly coordinated activity observed in CA1 requires either the

recorded cells or their inputs to drive or be driven by recurrent connections with their

neighbours, which occur only in CA3. Thus, the firing of a CA3 pyramidal (or a group of

pyramidals) will lead to both the firing of a CA1 pyramidal and to the firing of a neighbouring

CA3 pyramidal, which will in turn lead to the firing of another CA1 pyramidal, evident in multi-

unit recording as the coordinated firing of the two CA1 pyramidals. With an asymmetric spike-

timing plasticity relationship, LTP would lead to the learning of sequences, because the sequential firing of CA3 neurons would reinforce the sequence through LTP in the sequence direction and LTD in the “anti-sequence” direction. However, the observations by O’Neill et al.

(2006) and Sheng and Frank (2008) instead point to a symmetrical spike-timing/plasticity

relationship in CA3. Given that this relationship is indeed symmetrical for pairing-induced

GABAergic plasticity, it seems plausible that it could be involved in the replay observed in these

studies. Thus, a demonstration of pairing-induced GABAergic plasticity in CA3 following a

symmetrical spike-timing/plasticity rule would greatly strengthen our hypothesis that it is

involved in memory formation. In vivo, a demonstration that these phenomena require NMDAR

activation would further implicate the GABAergic plasticity described here. Obviously, this

would require technically challenging multi-unit recording from behaving animals, but as such

techniques become more accessible, more labs may be able to perform such studies. A perhaps

more realistic goal is to implicate GABAergic plasticity in memory through a simpler combined 85

behavioural/pharmacological approach. This would be aided by specific blockers or activators of

KCC2 activity. Unfortunately, none are presently available, but a number of molecules that specifically inhibit KCC2 were recently identified in a high-throughput screening study based on

234,000 compounds (Delpire et al., 2009); hopefully, some of these will become available.

Additionally, there is a hypomorphic KCC2 deficient mouse [Vilen et al., 2001; the full knock-

out dies shortly after birth (Woo et al., 2002)] that does demonstrate some learning deficits along

with a number of other nervous system defects (Tornberg et al., 2005). A genetic manipulation

restricted to the hippocampus, or even a particular subregion, would be ideal, and technologies

do exist to restrict gene knock-out both spatially and temporally (Tonegawa et al., 2003);

hopefully, these technologies will eventually be used to investigate the importance of

GABAergic plasticity.

Lastly, future slice experiments should take into account the different patterns of activity

and the differential involvement of neuromodulators underlying the various brain states. For

example, the high level of acetylcholine present during active waking (Hasselmo, 1999) when the theta and gamma rhythms dominate is thought to set the appropriate dynamics for encoding new information in the hippocampus. Given that our data suggests the induction of GABAergic plasticity may take place during theta oscillation, it might be interesting to investigate whether acetylcholine can facilitate its induction. While bath applying neuromodulators might seem unphysiological, their absence from the slice might be equally unphysiological. The ideal strategy would be to perform experiments such as those described in this thesis under a range of conditions reflecting the various brain states observed in vivo.

86

4.5 Final remarks

We have shown that disinhibition at the feedforward connections in CA1 can lead to a

strengthening of Schaffer collateral-mediated excitatory transmission. This disinhibition- mediated LTP is similar to classic glutamatergic LTP in many respects, suggesting that it too may play an important role in memory. By creating a closer approximation of the in vivo conditions, both in terms of neuromodulators present as well as the different activity patterns, and by investigating plasticity at other important hippocampal synapses, future slice experiments will hopefully be able to strengthen the link between pairing-induced GABAergic plasticity and memory. The advent of pharmacological agents to disrupt or enhance GABAergic plasticity, which we eagerly await, will in all likelihood allow for the confirmation of an important role for

GABAergic plasticity in memory, and may explain certain features that cannot be accounted for by glutamatergic plasticity alone. 87

5. Materials and Methods

5.1 Ethics Approval

All rats were maintained on a 12 h light/dark cycle with food and water provided ad

libitum. The Animal Studies Committee at the University of Toronto approved all experimental

protocols.

5.2 Brain slice preparation

400 µm hippocampal slices were prepared from 50-75 day old (except where noted) male

Sprague Dawley rats anaesthetized with a mixture of ketamine and xylazine, and perfused

through the ascending aorta with chilled modified artificial cerebrospinal fluid (ACSF). Modified

ACSF was composed of 180 mM sucrose, 25 mM sodium bicarbonate, 25 mM glucose, 2.5 mM

KCl, 1.25 mM sodium phosphate, 2 mM MgCl2, 1 mM CaCl2, 0.4 mM sodium ascorbate, and 3

mM sodium pyruvate, and saturated with 95% O2/5% CO2 (pH 7.4, osmolarity ~305 mOsm).

After cardiac perfusion, the brain was quickly removed, and hemispheres were separated and

placed into the chilled solution for another 30 seconds. Hippocampi were then partially isolated

by removing the cerebellum and all cortex except that directly overlying the hippocampus, to

avoid damaging area CA1 while increasing the hippocampal surface area contacting oxygenated

solution. The hippocampi were mounted vertically on an agar block and 400 µm slices cut with a

Vibratome 1000 plus. Slices recovered in 35-37oC ACSF composed of 125 mM NaCl, 25 mM

glucose, 25 mM sodium bicarbonate, 2.5 mM KCl, 1.25 mM sodium phosphate, 1mM MgCl2,

and 2 mM CaCl2 and saturated with 95% O2/5% CO2 (pH 7.4, osmolarity ~305 mOsm) for 1

hour. 88

5.3 Electrophysiology

Whole-cell recordings were made in oxygenated ACSF at 35-37oC from CA1 pyramidal

cells, and in some experiments, presynaptically connected feedforward interneurons. Pyramidal

cells were identified by the presence of an after-depolarization following action potential firing,

as well as action potential accommodation during prolonged AP trains (Figure 16). Feedforward

interneurons were recorded in the pyramidal cell layer. Generally, we targeted cells that had

much larger, and more irregularly shaped, cell bodies than the pyramidal cells. Their identity was

confirmed electrophysiologically; they were excited to threshold by relatively low levels of

Schaffer collateral stimulation (blocked by CNQX), and inhibited CA1 pyramidal neurons.

Intracellular current injection always produced APs with large after-hyperpolarization (see

Figure 4B, inset), and which were non-accommodating.

Whole-cell recording pipettes were pulled from thin-walled borosilicate (World Precision

Industries, TW-150F) with a Sutter Instruments P-87 to resistances of 5-8 MOhms. Pipettes were filled with a solution consisting of 130 mM potassium gluconate, 10 mM KCl, 10 mM HEPES,

0.2 mM EGTA, 4 mM ATP, 0.3 mM GTP, 10 mM phosphocreatine (pH 7.25, osmolarity 275-

285 mOsm). IPSPs recorded with this intracellular solution reversed at -88.3 ± 1.6 mV (LJP corrected; -74.3 mV uncorrected; n = 12). This was nearly identical to the reversal recorded with gramicidin perforated patch (-88.5 ± 1.2 mV LJP corrected; -87 uncorrected; n = 6), and there

was no statistical difference between the two groups (p = 0.673). All membrane potential values

in the text and figures are uncorrected for the liquid junction potential. Signals were amplified

using an Axon Instruments Multiclamp 700b and digitized using an Axon Instruments Digidata

1322a. The bridge was balanced upon going whole-cell, and then monitored and adjusted as 89

necessary throughout the duration of recording. Extracellular stimulation was applied through a

whole-cell recording pipette containing a silver chlorided wire and filled with ACSF. The

stimulus was generated by an AMPI ISO-Flex stimulus isolator triggered by an AMPI Master 8 stimulator, and was 100 µsec in duration. The baseline recording frequency was 0.0333 Hz.

Plasticity was induced by pairing extracellular stimulation with simultaneous current injection (1 nA for 10 ms) at 5 Hz (300 pairings; Figure 11). APV (Sigma), CNQX (Sigma) and gabazine

(Tocris) were applied through bath perfusion, while AIP (Sigma) was applied through the patch pipette diluted in internal recording solution. CNQX (Sigma) was used to block glutamatergic transmission in all recordings of isolated IPSPs. Gabazine (Tocris) was used to block

GABAergic transmission in all recordings of isolated EPSPs.

To determine the EPSP/feedforward IPSP delay, the delay between PSP onset in the pyramidal neuron and AP firing in the interneuron (Figure 4A) was added to the delay between the interneuron AP and the postsynaptic unitary IPSP (Figure 4B).

To verify that the recorded inhibition was feedforward and not due to direct stimulation of inhibitory fibers, CNQX was perfused into the bath at the end of all mixed recordings. CNQX reduced the mean slope of the PSP vs. membrane potential relationship for all recordings by 89%

(Figure 10 displays this data for each individual experimental group). Because glutamatergic

EPSPs make no significant contribution to the slope at the membrane potentials examined

(Figure 3A), the reduction in slope after CNQX can be attributed to reduced inhibition.

In 2-pathway experiments, the independence of the two pathways was demonstrated by

the confinement of short-term plasticity within pathways (Figure 22). 4 pulses at 20 Hz were

applied to one pathway, followed by a pulse to other pathway. The protocol was then repeated 90

with 4 pulses to the second pathway followed by 1 to the first. For each pathway, the ratio of the fourth to the first pulse was taken as the measure of short-term plasticity with the pathway, while the ratio of the single fifth pulse to the first pulse in the other recording gave the short-term plasticity between pathways.

5.4 Data Analysis

Data was acquired using Axon Instruments Clampex 9 software, and analyzed using

Axon Instruments Clampfit and Microsoft Excel. In all recordings, intracellular current steps were applied simultaneously with extracellular stimulation in sequences of 10, from the most negative to the most positive, so that PSP amplitude versus membrane potential graphs could be constructed in Microsoft Excel for each 5 min segment of the recording. For GABAergic recordings, the shift in the PSP amplitude versus membrane potential graphs’ x-intercept gave the change in EGABA, while the change in slope provided the relative change in synaptic conductance. For mixed EPSP/IPSP recordings, the graphs were used to determine the average

PSP amplitude from resting membrane potential (RMP; measured at the outset of each recording) for each 5 min segment. As EPSP amplitude showed no voltage dependence

(glutamatergic recordings), it was averaged for each 5 min segment to aid comparison with the other experimental groups.

5.5 Statistics

Results are expressed as mean ± s.e.m. All statistical tests were performed in Sigmastat.

Significance was determined using either a two-way repeated measures ANOVA with post-hoc

Tukey test (Figures 5, 7, 8, 9, 17a right, 18a, 20a right, 20c; P values reflect the results of the 91

post-hoc Tukey test) or a paired Student’s t-test (Figure 3A, Figure 6, 17a left, 18b, 19 a, b, c,

20a left, 20b) with significance level of P < 0.05. For all multiple comparisons in which statistical significance is reported, ANOVA values were significant to 0.05 or smaller. EPSP

90% amplitude and rise time correlation was determined with linear regression. 92

6. References

Abraham, W. C., Goddard, G. V. (1983). Asymmetric relationships between homosynaptic long-term potentiation and heterosynaptic long-term depression. Nature 305, 717-9.

Acsády, L., Kamondi, A., Sik, A., Freund, T., and Buzsáki, G. (1998). GABAergic cells are the major postsynaptic targets of mossy fibers in the rat hippocampus. J Neurosci 18, 3386– 3403.

Albus, J. S. (1971). A theory of cerebellar function. Mathematical Biosciences 28, 167–171.

Alger B. E., Teyler, T. J. (1976). Long-term and short-term plasticity in the CA1, CA3, and dentate regions of the rat hippocampal slice. Brain Res 110, 463-80.

*Ali, A. B., Thomson, A. M. (1998). J Physiol 507, 185-.

*Ali, A. B., Deuchars, J., Pawelzik, H., Thomson, A. M. (1998). J Physiol 507, 201

Andersen P, Sundberg SH, Sveen O, Wigström H (1977). Specific long-lasting potentiation of synaptic transmission in hippocampal slices. Nature 266, 736-7.

*Anwy1, R. (2009). Metabotropic glutamate receptor-dependent long-term potentiation. Neuropharmacology

Ascher, P., Nowak, L. (1988). The role of divalent cations in the N-methyl-D-aspartate responses of mouse central neurones in culture. J Physiol 399, 247-66.

Barria, A., Muller, D., Derkach, V., Griffith, L.C., and Soderling, T.R. (1997). Regulatory phosphorylation of AMPA-type glutamate receptors by CaM-KII during long-term potentiation. Science 276, 2042–2045.

Bear, M.F., Malenka, R.C. (1994). Synaptic plasticity: LTP and LTD. Curr Opin Neurobiol 4, 389-399.

Ben-Ari, Y., Gaiarsa, J.L., Tyzio, R., and Khazipov R. (2007). GABA: a pioneer transmitter that excites immature neurons and generates primitive oscillations, Physiol Rev 87, 1215–1284. 93

Benke, T. A., Luthi A., Isaac J. T. and Collingridge G. L. (1998). Modulation of AMPA receptor unitary conductance by synaptic activity. Nature 393, 793–797.

Bettler, B., Kaupmann, K., Mosbacher J. and Gassmann M. (2004). Molecular structure and physiological functions of GABAB receptors. Physiol Rev 84, 835–867.

Bettler B., and Tiao, J. Y. (2006). Molecular diversity, trafficking and subcellular localization of GABAB receptors. Pharmacol Ther 110, 533-543.

Biro, A. A., Holderith, N. B., and Nusser, Z. (2005). Quantal size is independent of the release probability at hippocampal excitatory synapses. J Neurosci 25, 223-232

Bi, G. Q., Poo, M.M. (1998). Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type. J Neurosci 18, 10464- 10472.

Bissière, S., Humeau, Y., and Lüthi, A. (2003). Dopamine gates LTP induction in lateral amygdala by suppressing feedforward inhibition. Nat Neurosci 6, 587-592

Blaesse, P., Airaksinen, M. S., Rivera, C., Kaila, K. (2009). Cation-chloride cotransporters and neuronal function. Neuron 61, 820-838.

Bliss, T.V., Collingridge, G.L. (1993). A synaptic model of memory: long-term potentiation in the hippocampus. Nature 361, 31-39.

Bliss, T.V. and Lømo, T. (1973). Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol 232, 331–356.

Bliss, T.V and Gardner-Medwin, A.R. (1973). Long-lasting potentiation of synaptic transmission in the dentate area of the unanaestetized rabbit following stimulation of the perforant path. J Physiol 232, 357-74. 94

Bliss, T.V., Errington, M.L., Laroche, S. and Lynch, M.A. (1987). Increase in K+-stimulated, Ca2+-dependent release of [3H]glutamate from rat dentate gyrus three days after induction of long-term potentiation. Neuroscience Letters 83, 107–112.

Bolshakov, V.Y. and Siegelbaum, S.A. (1995). Regulation of hippocampal transmitter release during development and long-term potentiation. Science 269, 1730–1734.

Bormann, J. (2000). The 'ABC' of GABA receptors. Trends Pharmacol Sci 21, 16-9.

Bowery, N.G. Bettler, B., Froestl, W., Gallagher, J.P., Marshall, F., Raiteri, M., Bonner, T.I. and Enna, S. J. (2002). International union of pharmacology: XXXIII. Mammalian γ- aminobutyric acid B receptors: structure and function. Pharmacol Rev 54, 247–264.

Bragin, A., Jandó, G., Nadasdy, Z., Hetke, J., Wise, K. and Buzsáki, G. (1995). Gamma (40– 100 Hz) oscillation in the hippocampus of the behaving rat. J Neurosci 15, 47–60.

Brankack, J., Stewart, M. and Fox, S.E. (1993). Current source density analysis of the hippocampal theta rhythm: associated sustained potentials and candidate synaptic generators. Brain Res 615, 310–327.

Buzsáki, G., Leung, L.W. and Vanderwolf, C.H. (1983). Cellular bases of hippocampal EEG in the behaving rat. Brain Res 287, 139–171.

Buzsaki, G. (1986) Hippocampal sharp waves: their origin and significance. Brain Res 398, 242–252.

Buzsáki, G., Czopf, J., Kondakor, I. and Kellenyi, L. (1986). Laminar distribution of hippocampal rhythmic slow activity (RSA) in the behaving rat: Current source density analysis, effects of urethane and atropine. Brain Res 365, 125–137.

Buzsaki, G. (1989) Two-stage model of memory trace formation: a role for “noisy” brain states. Neuroscience 31, 551–570.

Buzsaki, G., Horvath, Z., Urioste, R., Hetke, J., and Wise, K. (1992). High-frequency network oscillation in the hippocampus. Science 256, 1025–1027. 95

Buzsaki, G. (1996). The hippocampo-neocortical dialogue. Cereb Cortex 6, 81–92.

Buzsáki, G. (2002). Theta oscillations in the hippocampus. Neuron 33, 325-340.

Buzsáki, G., Buhl, D. L. Harris, K. D. Csicsvari, J. Czéh B. and Morozov A. (2003). Hippocampal network patterns of activity in the mouse. Neuroscience 116, 201-211.

Buzsáki, G., Geisler, C., Henze, D. A., Wang, X. J. (2004). Circuit complexity and axon wiring economy of cortical interneurons. Trends Neurosci 27, 186-93.

Buzsáki, G., Kaila, K., Raichle, M. (2007). Inhibition and brain work. Neuron 56, 771-783.

Caillard, O., Ben-Ari, Y., Gaiarsa, J. L. (1999). Long-term potentiation of GABAergic synaptic transmission in neonatal rat hippocampus. J Physiol 518, 109-19.

Caillard, O., Ben-Ari, Y., Gaiarsa, J. L. (1999). Mechanisms of induction and expression of long-term depression at GABAergic synapses in the neonatal rat hippocampus. J Neurosci 19, 7568-77.

Calver, A. R., Davies, C. H. and Pangalos, M. N. (2002). GABAB receptors: from monogamy to promiscuity. Neurosignals 11, 299–314.

Caraiscos, V. B., Elliott, E. M., You-Ten, K. E., Cheng, V. Y., Belelli, D., Newell, J. G., Jackson, M. F., Lambert, J. J., Rosahl, T. W., Wafford, K. A. (2004). Tonic inhibition in mouse hippocampal CA1 pyramidal neurons is mediated by alpha5 subunit-containing gamma- aminobutyric acid type A receptors. Proc Natl Acad Sci USA 101, 3662–3667.

Carvalho, T. P., Buonomano, D. V. (2009). Differential effects of excitatory and inhibitory plasticity on synaptically driven neuronal input-output functions. Neuron 61, 774-85.

Cattabeni, F., Gardoni, F., Di Luca, M. (1999). Pathophysiological implications of the structural organization of the excitatory synapse. Eur J Pharmacol 375, 339-347.

Charpak, S., Pare, D. and Llinás, R. (1995). The entorhinal cortex entrains fast CA1 hippocampal oscillations in the anaesthetized guinea-pig: role of the monosynaptic component of the perforant path. Eur J Neurosci 7, 1548–1557. 96

Chen, P. E., Wyllie, D. J. (2006). Pharmacological insights obtained from structure-function studies of ionotropic glutamate receptors. Br J Pharmacol 147, 839–853.

Cheng, S., Frank, L. M. (2008). New experiences enhance coordinated neural activity in the hippocampus. Neuron 57, 303-313.

Chevaleyre V., Castillo P. E. (2003). Heterosynaptic LTD of hippocampal GABAergic synapses: a novel role of endocannabinoids in regulating excitability. Neuron 38, 461-472.

Christie, B. R., Magee, J. C., Johnston, D. (1996). The role of dendritic action potentials and Ca2+ influx in the induction of homosynaptic long-term depression in hippocampal CA1 pyramidal neurons. Learn Mem 3, 160-169.

Chrobak, J. J. and Buzsaki, G. (1994). Selective activation of deep layer (V-VI) retrohippocampal cortical neurons during hippocampal sharp waves in the behaving rat. J. Neurosci 14, 6160–6170.

Chrobak, J. J. and Buzsáki, G. (1998). Gamma oscillations in the entorhinal cortex of the freely behaving rat. J. Neurosci 18, 388–398.

Collingridge, G. L., Kehl, S. J., McLennan, H. (1983). The antagonism of amino acid-induced excitations of rat hippocampal CA1 neurones in vitro. J Physiol 334, 19-31.

Csicsvari, J., Hirase, H., Czurko, A., Mamiya, A. and Buzsáki, G. (1999). Oscillatory coupling of hippocampal pyramidal cells and interneurons in the behaving rat. J Neurosci 19, 274–287.

Csicsvari, J., Hirase, H., Czurko, A., Mamiya, A. and Buzsáki, G. (1999). Fast network oscillations in the hippocampal CA1 region of the behaving rat. J. Neurosci 19, RC20.

Csicsvari, J., Hirase, H., Mamiya, A. and Buzsáki, G. (2000). Ensemble patterns of hippocampal CA3–CA1 neurons during sharp wave-associated population events. Neuron 28, 585–594.

Csicsvari, J., Jamieson, B., Wise, K. D., Buzsáki, G. (2003). Mechanisms of gamma oscillations in the hippocampus of the behaving rat. Neuron 37, 311-322. 97

Davies, S. N., Lester, R. A., Reymann, K. G., Collingridge, G. L. (1989). Temporally distinct pre- and post-synaptic mechanisms maintain long-term potentiation. Nature 338, 500-503.

Davies, C. H., Starkey, S. J., Pozza, M. F., and Collingridge, G. L. (1991). GABA autoreceptors regulate the induction of LTP. Nature 349, 609–611.

Davis, S., Butcher, S. P., and Morris, R. G. (1992). The NMDA receptor antagonist D-2- amino-5-phosphonopentanoate (D-AP5) impairs spatial learning and LTP in vivo at intracerebral concentrations comparable to those that block LTP in vitro. J Neurosci 12, 21–34.

Delpire, E., Days, E., Lewis, M., Mi, D., Kim, K., Lindsley, C. W., and Weaver, C. D. (2009). Small molecule screen identifies inhibitors for the neuronal K-Cl cotransporter KCC2. Proc. Natl. Acad. Sci. USA 106, 5383-5388. del Castillo, J. and Katz, B. (1954). Quantal components of end-plate potential. J Physiol 124, 570–573.

Diba, K. and Buzsaki, G. (2007). Forward and reverse hippocampal place-cell sequences during ripples. Nat Neurosci 10, 1241-1242.

Dingledine, R., Borges, K., Bowie, D., and Traynelis, S. F. (1999). The glutamate receptor ion channels. Pharmacol Rev 51, 7-61.

Draguhn, A., Traub, R. D., Schmitz, D., and Jefferys, J. G. (1998). Electrical coupling underlies high-frequency oscillations in the hippocampus in vitro. Nature 394 pp. 189–192.

Dunwiddie, T. and Lynch, G. (1978). Long-term potentiation and depression of synaptic responses in the rat hippocampus: localization and frequency dependency. J Physiol 276, 353-67.

Durand, G. M., Kovalchuk, Y., and Konnerth, A. (1996). Long-term potentiation and functional synapse induction in developing hippocampus. Nature 381, 71–75.

Engert, F. and Bonhoeffer, T. (1997). Synapse specificity of long-term potentiation breaks down at short distances. Nature 388, 279-284. 98

Enoki, R., Hu, Y. L., Hamilton, D., and Fine, A. (2009). Expression of long-term plasticity at individual synapses in hippocampus is graded, bidirectional, and mainly presynaptic: optical quantal analysis. Neuron 62, 242-53.

Farrant, M. and Nusser, Z. (2005). Variations on an inhibitory theme: phasic and tonic activation of GABA(A) receptors. Nat Rev Neurosci 6, 215–229.

Farrant, M. and Kaila, K. (2007). The cellular, molecular and ionic basis of GABA(A) receptor signalling. Prog. Brain Res 160, 59–87.

Ferraguti, F. and Shigemoto, R. (2006). Metabotropic glutamate receptors. Cell Tissue Res 326, 483-504.

Fisahn, A., Pike, F. G., Buhl, E. H., and Paulsen, O. (1998). Cholinergic induction of network oscillations at 40 Hz in the hippocampus in vitro. Nature 394, 186–189.

Foster, D. J. and Wilson, M. A. (2006). Reverse replay of behavioural sequences in hippocampal place cells during the awake state. Nature 440, 680-683.

Fox, S. E., Wolfson, S. and Ranck Jr., J. B. (1986). Hippocampal theta rhythm and the firing of neurons in walking and urethane anesthetized rats. Exp. Brain Res 62, 495–508.

Frerking, M. and Nicoll, R.A. (2000). Synaptic kainate receptors, Curr Opin Neurobiol 10, 342–351.

Freund, T.F. and Buzsaki, G. (1996). Interneurons of the hippocampus. Hippocampus 6, 347– 470.

Fricker, D. and Miles, R. (2000). Differences between EPSP—spike transduction in pyramidal and inhibitory cells of the hippocampus. Neuron 28, 559–569.

Fritschy, J. M. and Mohler, H. (1995). GABAA-receptor heterogeneity in the adult rat brain: differential regional and cellular distribution of seven major subunits. J Comp Neurol 359, 154– 194. 99

Furukawa H, Singh SK, Mancusso R, Gouaux E (2005) Subunit arrangement and function in NMDA receptors. Nature 438:185–192

Geiger JR, Melcher T, Koh DS, Sakmann B, Seeburg PH, Jonas P, Monyer H (1995) Relative abundance of subunit mRNAs determines gating and Ca2+ permeability of AMPA receptors in principal neurons and interneurons in rat CNS. Neuron 15:193–204

Glickfeld LL, Scanziani M. Distinct timing in the activity of cannabinoid-sensitive and cannabinoid-insensitive basket cells. Nat Neurosci 9: 807-815, 2006.

Gomeza J, Joly C, Kuhn R, Knopfel T, Bockaert J, Pin JP (1996) The second intracellular loop of metabotropic glutamate receptor 1 cooperates with the other intracellular domains to control coupling to G-proteins. J Biol Chem 271:2199–2205

Gouaux E (2004) Structure and function of AMPA receptors. J Physiol (Lond) 554:249–253

Greengard P. The neurobiology of slow synaptic transmission. Science. 2001 Nov 2;294(5544):1024-30. Review.

Greenstein, Y.J., Pavlides, C. and Winson, J., 1988. Long-term potentiation in the dentate gyrus is preferentially induced at theta periodicity. Brain Res. 438, pp. 331–334.

Gruart A, Muñoz MD, Delgado-García JM. Involvement of the CA3-CA1 synapse in the acquisition of associative learning in behaving mice. J Neurosci 26: 1077-1087, 2006.

Grunze HC, Rainnie DG, Hasselmo ME, Barkai E, Hearn EF, McCarley RW, Greene RW. NMDA-dependent modulation of CA1 local circuit inhibition. J Neurosci. 1996 Mar 15;16(6):2034-43.

A.I. Gulyás, R. Miles, A. Sik, K. Toth, N. Tamamaki and T.F. Freund, Hippocampal pyramidal cells excite inhibitory neurons through a single release site, Nature 366 (1993), pp. 683–687.

A.I. Gulyás, M. Megias, Z. Emri and T.F. Freund, Total number and ratio of excitatory and inhibitory synapses converging onto single interneurons of different types in the CA1 area of the rat hippocampus, J. Neurosci. 19 (1999), pp. 10082–10097. 100

Gulyas, A. I., Sik, A., Payne, J. A., Kaila, K., and Freund, T. F. (2001). The KCI cotransporter, KCC2, is highly expressed in the vicinity of excitatory synapses in the rat hippocampus. Eur J Neurosci 13, 2205–2217.

Gusev, P. A. and Alkon, D. L. (2001). Intracellular correlates of spatial memory acquisition in hippocampal slices: long-term disinhibition of CA1 pyramidal cells. J Neurophysiol 86, 881-899.

Hafting, T., Fyhn, M., Bonnevie, T., Moser, M. B. and Moser, E. I. (2008). Hippocampus- independent phase precession in entorhinal grid cells. Nature 453, 1248-52.

Hardie, J. B. and Pearce, R. A. (2006). Active and passive membrane properties and intrinsic kinetics shape synaptic inhibition in hippocampal CA1 pyramidal neurons. J Neurosci 26, 8559- 8569.

Hardie J and Spruston N. (2009). Synaptic depolarization is more effective than back- propagating action potentials during induction of associative long-term potentiation in hippocampal pyramidal neurons. J Neurosci 29, 3233-3241.

Harris, E. W., Ganong, A. H., Cotman, C. W. (1984). Long-term potentiation in the hippocampus involves activation of N-methyl-D-aspartate receptors. Brain Res 323, 132-7.

Harris, K., Henze, D. A., Hirase, H., Csicsvari, J., and Buzsáki, G. (2000). The accuracy of tetrode spike separation as determined by simultaneous intracellular and extracellular measurements. J Neurophysiol 84, 401–414.

Harris, K., Hirase, H., Leinekugel, X., Henze, D.A., and Buzsáki, G. (2001). Temporal interaction between single spikes and complex spike bursts in hippocampal pyramidal cells. Neuron 32, 141–149.

Hasselmo, M. E., Schnell, E., and Barkai, E. (1995). Dynamics of learning and recall at excitatory recurrent synapses and cholinergic modulation in rat hippocampal region CA3. J Neurosci 15, 5249–5262. 101

Hasselmo, M. E., Wyble, B. P., and Wallenstein, G. V. (1996). Encoding and retrieval of episodic memories: role of cholinergic and GABAergic modulation in the hippocampus. Hippocampus 6, 693–708.

Hasselmo, M.E. (1999). Neuromodulation: acetylcholine and memory consolidation. Trends Cogn Sci 3, 351–359.

Hebb, D.O. (1949). The Organization of Behavior. Wiley, New York.

Hill, D.R. (1985). GABAB receptor modulation of adenylate cyclase activity in rat brain slices. Br J Pharmacol 84, 249–257.

Hirase, H., Leinekugel, X., Czurko, A., Csicsvari, J., and Buzsaki, G. (2001). Firing rates of hippocampal neurons are preserved during subsequent sleep episodes and modified by novel awake experience. Proc Natl Acad Sci USA 98, 9386–9390.

Holsheimer, J., Boer, J., Lopes da Silva, F.H., and van Rotterdam, A. (1982). The double dipole model of theta rhythm generation: simulation of laminar field potential profiles in dorsal hippocampus of the rat. Brain Res 235, 31–50.

Höschler, C., Anwyl, R., and Rowan, M.J. (1997). Stimulation on the positive phase of hippocampal theta rhythm induces long-term potentiation that can be depotentiated by stimulation on the negative phase in area CA1 in vivo. J Neurosci 17, 6470–6477.

Huang, Z. J. (2006). GABAB receptor isoforms caught in action at the scene. Neuron 50, 521-4.

Huang, C. S., Shi, S. H., Ule, J., Ruggiu, M., Barker, L. A., Darnell, R. B., and Jan, Y. N., and Jan, L. Y. (2005). Common molecular pathways mediate long-term potentiation of synaptic excitation and slow synaptic inhibition. Cell 123, 105–118.

Huerta, P. T. and Lisman, J. E. (1996). Bidirectional synaptic plasticity induced by a single burst during cholinergic theta oscillation in CA1 in vitro. Neuron 15, 1053–1063.

Huerta, P. T., Sun, L. D., Wilson, M. A., and Tonegawa, S. (2000). Formation of temporal memory requires NMDA receptors within CA1 pyramidal neurons. Neuron 25, 473-480. 102

Isaac, J. T., Nicoll, R. A. and Malenka, R. C. (1995). Evidence for silent synapses: implications for the expression of LTP. Neuron 15, 427–434.

Isaac, J. T., Mellor, J., Hurtado, D., Roche, K. W. (2004). Kainate receptor trafficking: physiological roles and molecular mechanisms. Pharmacol Ther 104, 163-72.

Ishida, A., Kameshita, I., Okuno, S., Kitani, T., and Fujisawa, H. (1995). A novel highly specific and potent inhibitor of calmodulin-dependent protein kinase II. Biochem Biophys Res Commun 212, 806-812.

Kaila, K. and Voipio, J. (1987). Postsynaptic fall in intracellular pH induced by GABA- activated bicarbonate conductance. Nature 330, 163–165.

Karnup, S. and Stelzer, A. (1999). Temporal overlap of excitatory and inhibitory afferent input in guinea-pig CA1 pyramidal cells. J Physiol 516, 485-504.

Kauer, J.A., Malenka, R. C., and Nicoll, R. A. (1988). A persistent postsynaptic modification mediates long-term potentiation in the hippocampus. Neuron 1, 911–917.

Khirug, S., Yamada, J., Afzalov, R., Voipio, J., Khiroug, L., and Kaila, K. (2008). GABAergic Depolarization of the Axon Initial Segment in Cortical Principal Neurons Is Caused by the Na-K-2Cl Cotransporter NKCC1. J Neurosci 28, 4635-4639.

Klausberger, T. and Somogyi, P. (2008). Neuronal diversity and temporal dynamics: the unity of hippocampal circuit operations. Science 321, 53-7.

Köhr, G. (2006). NMDA receptor function: subunit composition versus spatial distribution. Cell Tissue Res 326, 439-46.

Kramis, R., Vanderwolf, C. H., and Bland, B. H. (1975). Two types of hippocampal rhythmical slow activity in both the rabbit and the rat: relations to behavior and effects of atropine, diethyl ether, urethane, and pentobarbital. Exp Neurol 49, 58–85.

Kraushaar, U. and Jonas, P. (2000). Efficacy and stability of quantal GABA release at a hippocampal interneuron-principal neuron synapse. J Neurosci 20, 5594–5607. 103

Kudrimoti, H. S., Barnes, C. A., and McNaughton, B. L. (1999). Reactivation of hippocampal cell assemblies: effects of behavioral state, experience, and EEG dynamics. J Neurosci. 19, 4090–4101.

Kullmann, D. M. and Nicoll, R. A. (1992). Long-term potentiation is associated with increases in quantal content and quantal amplitude. Nature 357, 240-244.

Kullmann, D. M. (2001). Presynaptic kainate receptors in the hippocampus: slowly emerging from obscurity. Neuron 32, 561–564.

Larkman, A., Hannay, T., Stratford, K., and Jack, J. (1992). Presynaptic release probability influences the locus of long-term potentiation. Nature 360, 70-73.

Larson, J. and Lynch, G. (1986). Induction of synaptic potentiation in hippocampus by patterned stimulation two events. Science 232, 985–988.

Laurie, D. J., Wisden, W., Seeburg, P. H. (1992). The distribution of thirteen GABAA receptor subunit mRNAs in the rat brain. III. Embryonic and postnatal development. J Neurosci 12, 4151–4172.

Lee, A. K. and Wilson, M. A. (2002). Memory of sequential experience in the hippocampus during slow wave sleep. Neuron 36, 1183–1194.

Lee, A. K. and Wilson, M. A. (2004). A combinatorial method for analyzing sequential firing patterns involving an arbitrary number of neurons based on relative time order. J Neurophysiol 92, 2555–2573.

Leung, L. S. (1984). Theta rhythm during REM sleep and waking: correlations between power, phase and frequency. Electroencephalogr Clin Neurophysiol 58, 553–564.

Leung, L. W. (1984). Model of gradual phase shift of theta rhythm in the rat. J Neurophysiol 52, 1051–1065.

Leung, L. S. (1998). Generation of theta and gamma rhythms in the hippocampus, Neurosci Biobehav Rev 22, 275–290. 104

Leung, L. S. and Peloquin, P. (2006). GABAB receptors inhibit backpropagating dendritic spikes in hippocampal CA1 pyramidal cells in vivo. Hippocampus 16, 388-407.

Levy, W. B., Steward O. (1979). Synapses as associative memory elements in the hippocampal formation. Brain Res 175, 233-45.

Li, X. G., Somogyi, P., Tepper J. M., and Buzsáki, G. (1992). Axonal and dendritic arborization of an intracellularly labeled chandelier cell in the CA1 region of rat hippocampus, Exp Brain Res 90, 519–525.

Li, X. G., Somogyi, P., Ylinen, A., and Buzsáki, G. (1994). The hippocampal CA3 network: an in vivo intracellular labeling study. J Comp Neurol 339, 181–208.

Liao, D. Hessler N.A., and Malinow, R. (1995). Activation of postsynaptically silent synapses during pairing-induced LTP in CA1 region of hippocampal slice. Nature 375, 400–404.

Lisman, J. E. and Idiart, M. A. (1995). Storage of 7 +/− 2 short-term memories in oscillatory subcycles. Science 267, 1512–1515.

Losonczy, A., Biro, A.A., and Nusser, Z. (2004). Persistently active cannabinoid receptors mute a subpopulation of hippocampal interneurons. Proc Natl Acad Sci USA 101, 1362–1367.

Louie, K. and Wilson, M.A. (2001). Temporally structured replay of awake hippocampal ensemble activity during rapid eye movement sleep. Neuron 29, 145–156.

Lu, Y. M., Mansuy, I. M., Kandel, E. R., Roder, J. (2000). Calcineurin-mediated LTD of GABAergic inhibition underlies the increased excitability of CA1 neurons associated with LTP. Neuron 26, 197-205.

Lüscher, C., Jan, L. Y., Stoffel, M., Malenka, R. C., and Nicoll, R. A. (1997). G protein- coupled inwardly rectifying K+ channels (GIRKs) mediate postsynaptic but not presynaptic transmitter actions in hippocampal neurons. Neuron 19, 687–695.

Lynch, G. S., Dunwiddie, T., Gribkoff, V. (1977). Heterosynaptic depression: a postsynaptic correlate of long-term potentiation. Nature 266, 737-9. 105

Jonas, P., Bischofberger, J., Fricker, D., and Miles, R. (2004). Fast in, fast out—temporal and spatial signal processing in hippocampal interneurons. Trends Neurosci 27, 30–40.

Klausberger, T., Magill, P. J., Márton, L. F., Roberts, J. D., Cobden, P. M., Buzsáki, G., Somogyi, P. (2003). Brain-state- and cell-type-specific firing of hippocampal interneurons in vivo. Nature 421, 844-848.

Klausberger, T. and Somogyi, P. (2008). Neuronal Diversity and Temporal Dynamics: The Unity of Hippocampal Circuit Operations. Science 321 , 53 - 57

Kraushaar, U. and Jonas, P. (2000). Efficacy and stability of quantal GABA release at a hippocampal interneuron-principal neuron synapse. J Neurosci 20, 5594–5607.

McHugh, T. J., Blum, K. I., Tsien, J. Z., Tonegawa, S., Wilson, M. A. (1996). Impaired hippocampal representation of space in CA1-specific NMDAR1 knockout mice. Cell 87, 1339– 1349.

McNaughton, B. L., Douglas R. M., and Goddard, G. V. (1978). Synaptic enhancement in fascia dentata: cooperativity among coactive afferents. Brain Res 157, 277–293.

McNaughton, B. L. Long-term potentiation, cooperativity and Hebb's cell assemblies: a personal history. (2003). Philos Trans R Soc Lond B Biol Sci 358, 629-634.

Magee, J. C. and Johnston, D. (1997). A synaptically controlled, associative signal for Hebbian plasticity in hippocampal neurons. Science 275, 209–213.

Mainen, Z. F., Jia, Z., Roder, J., and Malinow, R. (1998). Use-dependent AMPA receptor block in mice lacking GluR2 suggests postsynaptic site for LTP expression. Nat Neurosci 1, 579–586.

Malinow, R., Madison, D. V., Tsien, R. W. (1988). Persistent protein kinase activity underlying long-term potentiation. Nature 335, 820-824.

Malinow R. and Tsien, R. W. (1990). Presynaptic enhancement shown by whole-cell recordings of long-term potentiation in hippocampal slices. Nature 346, 177–180. 106

Malinow, R., Mainen, Z. F., Hayashi, Y. (2000). LTP mechanisms: from silence to four-lane traffic. Curr Opin Neurobiol 10, 352-357.

Manabe, T. and Nicoll, R. A. (1994). Long-term potentiation: evidence against an increase in transmitter release probability in the CA1 region of the hippocampus. Science 265, 1888–1892.

Marder, C. P., Buonomano, D. V. (2004). Timing and balance of inhibition enhance the effect of long-term potentiation on cell firing. J Neurosci 24, 8873-8884.

Markram, H., Lubke, J., Frotscher, M., and Sakmann, B. (1997). Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs. Science 275, 213–215.

Markram, H., Toledo-Rodriguez, M., Wang, Y., Gupta, A., Silberberg, G., and Wu, C. (2004). Interneurons of the neocortical inhibitory system. Nat Rev Neurosci 5, 793–807.

Marr D. (1969). A theory of cerebellar cortex. J Physiol 202, 437-70.

Matsuo, N., Reijmers, L., and Mayford, M. (2008). Spine-type-specific recruitment of newly synthesized AMPA receptors with learning. Science 319, 1104–1107.

Megias, M., Emri, Z., Freund, T. F., and Gulyás, A. I. (2001). Total number and distribution of inhibitory and excitatory synapses on hippocampal CA1 pyramidal cells. Neuroscience 102, 527–540.

Mehta, M. R., Quirk, M. C., Wilson, M. A. (2000). Experience-dependent asymmetric shape of hippocampal receptive fields. Neuron 25, 707-715.

Meredith, R. M., Floyer-Lea, A. M., Paulsen, O. (2003). Maturation of long-term potentiation induction rules in rodent hippocampus: role of GABAergic inhibition. J Neurosci 23, 11142- 11146.

Michels, G. and Moss, S. J. (2007). GABAA receptors: properties and trafficking. Crit Rev Biochem Mol Biol 42, 3-14.

Miles R. (1990). Synaptic excitation of inhibitory cells by single CA3 hippocampal pyramidal cells of the guinea-pig in vitro. J Physiol 428, 61–77. 107

Miles, R. (1990). Variation in strength of inhibitory synapses in the CA3 region of guinea-pig hippocampus in vitro. J Physiol 431, 659–676.

Miles, R., Tóth, K., Gulyás, A. I., Hájos, N., and Freund, T. F. (1996). Differences between Somatic and Dendritic Inhibition in the Hippocampus. Neuron 16, 815-823.

Min, M. Y., Asztely, F., Kokaia, M., and Kullmann, D. M. (1998). Long-term potentiation and dual-component quantal signaling in the dentate gyrus. Proc Natl Acad Sci USA 95, 4702–4707.

Mintz, I. M. and Bean, B. P. (1993). GABAB receptor inhibition of P-type Ca2+ channels in central neurons. Neuron 10, 889–898.

Mody, I. and Pearce, R.A. (2004). Diversity of inhibitory neurotransmission through GABA(A) receptors. Trends Neurosci 27, 569–575.

Mohler, H. and Fritschy, J. M. (1999). GABAB receptors make it to the top as dimmers. Trends Pharmacol Sci 20, 87–89.

Morris, R. G., Anderson, E., Lynch, G. S., and Baudry, M. (1986). Selective impairment of learning and blockade of long-term potentiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature 319, 774-6.

Morris, R. G. (1989). Synaptic plasticity and learning: selective impairment of learning rats and blockade of long-term potentiation in vivo by the N-methyl-D-aspartate receptor antagonist AP5. J Neurosci 9, 3040–3057.

Morris, R. G. and Frey, U. (1997). Hippocampal synaptic plasticity: role in spatial learning or the automatic recording of attended experience? Philos Trans R Soc Lond B Biol Sci 352, 1489- 503.

Muller, D., Joly, M., and Lynch, G. (1988). Contributions of quisqualate and NMDA receptors to the induction and expression of LTP. Science 242, 1694–1697.

Muller, T., Moller, T., Berger, T., Schnitzer, J., Kettenmann, H. (1992). Calcium entry through kainate receptors and resulting potassium-channel blockade in Bergmann glial cells. Science 256, 1563–1566. 108

Nadasdy, Z., Hirase, H., Czurko, A., Csicsvari, J., and Buzsaki, G. (1999). Replay and time compression of recurring spike sequences in the hippocampus. J Neurosci 19, 9497–9507.

Nelson, S. B., Turrigiano, G. G. (2008). Strength through diversity. Neuron 60, 477-482.

Neves, G., Cooke, S. F., and Bliss, T. V. P. (2008). Synaptic plasticity, memory and the hippocampus: a neural network approach to causality. Nature Reviews Neuroscience 9, 65-75.

Nishiyama, M., Hong, K., Mikoshiba, K., Poo, M. M., Kato, K. (2000).Calcium stores regulate the polarity and input specificity of synaptic modification. Nature 408, 584-588.

O’Hara, P. J., Sheppard, P. O., Thogersen, H., Venezia, D., Haldeman, B. A., McGrane, V., Houamed, K. M., Thomsen, C., Gilbert, T. L., Mulvihill, E. R. (1993). The ligand-binding domain in metabotropic glutamate receptors is related to bacterial periplasmic binding proteins. Neuron 11, 41–52

O'Keefe, J. and Dostrovsky, J. (1971). The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res 34, 171-5.

O'Keefe, J. (1976). Place units in the hippocampus of the freely moving rat. Exp Neurol 51, 78- 109.

O'Keefe, J. and Nadel, L. (1978). Hippocampus as a Cognitive Map. Clarindon, Oxford.

O'Keefe, J. & Recce, M. L. (1993). Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 3, 317–330.

O'Neill, J., Senior, T., Csicsvari, J. (2006). Place-selective firing of CA1 pyramidal cells during sharp wave/ripple network patterns in exploratory behavior. Neuron 49, 143-55.

O'Neill, J., Senior, T. J., Allen, K., Huxter, J. R., Csicsvari, J. (2008). Reactivation of experience-dependent cell assembly patterns in the hippocampus. Nat Neurosci 11, 209-215.

Okamoto, T., Sekiyama, N., Otsu, M., Shimada, Y., Sato, A., Nakanishi, S., and Jingami, H. (1998). Expression and purification of the extracellular ligand binding region of metabotropic glutamate receptor subtype 1. J Biol Chem 273, 13089–13096. 109

Ormond, J. and Woodin, M. A. (2006). Bidirectional spike-timing plasticity of GABAA receptor-mediated inhibition in the hippocampus. Society for Neuroscience Annual Meeting Abstract 133.3/C59.

Patenaude, C., Chapman, C. A., Bertrand, S., Congar, P., and Lacaille, J. C. (2003). GABAB receptor- and metabotropic glutamate receptor-dependent cooperative long-term potentiation of rat hippocampal GABAA synaptic transmission. J Physiol 553, 155–167.

Paulsen, O. and Sejnowski, T. J. 2000. Natural patterns of activity and long-term synaptic plasticity. Curr Opin Neurobiol 10, 172–179.

Pavlides, C. Greenstein, Y.J. Grudman M. and Winson, J. (1988). Long-term potentiation in the dentate gyrus is induced preferentially on the positive phase of theta-rhythm. Brain Res 439, 383–387.

Penttonen, M., Kamondi, A., Acsady, L. and Buzsáki, G. (1998). Gamma frequency oscillation in the hippocampus of the rat: intracellular analysis in vivo. Eur J Neurosci 10, 718– 728.

Pin, J. P., Joly, C., Heinemann, S. F., and Bockaert, J. (1994). Domains involved in the specificity of G protein activation in phospholipase C-coupled metabotropic glutamate receptors. EMBO J 13, 342–348.

Poe, G. R., Nitz, D. A., McNaughton, B. L., and Barnes, C. A. (2000). Experience-dependent phase-reversal of hippocampal neuron firing during REM sleep. Brain Res 855, 176-80.

Polsky, A., Mel, B. W., and Schiller, J. (2004). Computational subunits in thin dendrites of pyramidal cells. Nat Neurosci 7, 621-627.

Poncer, J. C., McKinney, R.A., Gahwiler, B.H., and Thompson, S.M. (1997). Either N- or P- type calcium channels mediate GABA release at distinct hippocampal inhibitory synapses. Neuron 18, 463–472. 110

Pouille, F. and Scanziani, M. (2001). Enforcement of temporal fidelity in pyramidal cells by somatic feed-forward inhibition. Science 293, 1159-1163.

Pouille, F. and Scanziani, M. (2004). Routing of spike series by dynamic circuits in the hippocampus. Nature 429, 717-723.

Price, C. J., Cauli, B., Kovacs, E. R., Kulik, A., Lambolez, B., Shigemoto, R., Capogna, M. (2005). Neurogliaform Neurons Form a Novel Inhibitory Network in the Hippocampal CA1 Area. J Neurosci 25, 6775-6786.

Redman S. (1990). Quantal analysis of synaptic potentials in neurons of the central nervous system. Physiol Rev 70, 165–198.

Riedel, G., Casabona, G., and Reymann, K. G. (1995). Inhibition of long-term potentiation in the dentate gyrus of freely moving rats by the metabotropic glutamate receptor antagonist MCPG. J Neurosci 15, 87-98.

Rivera, C., Voipio, J., Payne, J. A., Ruusuvuori, E., Lahtinen, H., Lamsa, K., Pirvola, U., Saarma M., and Kaila, K. (1999). The K+/Cl- co-transporter KCC2 renders GABA hyperpolarizing during neuronal maturation. Nature 397, 251–255.

Rivera, C., Li, H., Thomas-Crusells, J., Lahtinen, H., Viitanen, T., Nanobashvili, A., Kokaia, Z., Airaksinenm M. S., Voipio, J., Kaila, K., and Saarma, M. (2002). BDNF- induced TrkB activation down-regulates the K+-Cl- cotransporter KCC2 and impairs neuronal Cl- extrusion. J Cell Biol 159, 747–752.

Sayer, R. J., Friedlander, M. J., and Redman, S. J. (1990). The time course and amplitude of EPSPs evoked at synapses between pairs of CA3/CA1 neurons in the hippocampal slice. J Neurosci 10: 826-36.

Scanziani, M., Malenka, R. C., and Nicoll, R. A. (1996). Role of intercellular interactions in heterosynaptic long-term depression. Nature 380, 446-50.

Schuman, E. M. and Madison, D. V. (1994). Locally distributed synaptic potentiation in the hippocampus. Science 263, 532-536. 111

Scoville, W. B. and Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. J Neurol Neurosurg Psychiatry 20, 11-21.

Schwartzkroin, P. A. and Wester, K. (1975). Long-lasting facilitation of a synaptic potential following tetanization in the in vitro hippocampal slice. Brain Res 89, 107-119.

Sejnowski, T. J. (1977). Statistical constraints on synaptic plasticity. J Theor Biol 69, 385-389.

Semyanov, A., Walker, M. C., Kullmann, D. M., and Silver, R. A. (2004). Tonically active GABA A receptors: modulating gain and maintaining the tone. Trends Neurosci 27, 262–269.

Shadlen, M. N. and Newsome, W. T. (1998). The variable discharge of cortical neurons: implications for connectivity, computation, and information coding. J Neurosci 18, 3870–3896.

Shi, S. H., Hayashi, Y. Petralia, R. Zaman, S. Wenthold, R. Svoboda K. and Malinow, R. (1999). Rapid spine delivery and redistribution of AMPA receptors after synaptic NMDA receptor activation. Science 284, 1811–1816.

Shu, Y., Hasenstaub, A., and McCormick, D. A. (2003). Turning on and off recurrent balanced cortical activity. Nature 423, 288–293.

Sik, A., Ylinen, A., Penttonen, M., and Buzsáki, G. (1994). Inhibitory CA1–CA3-hilar region feedback in the hippocampus. Science 265, 1722–1724.

Sik, A., Penttonen, M., Ylinen, A. and Buzsáki, G. (1995). Hippocampal CA1 interneurons: an in vivo intracellular labeling study. J Neurosci 15, 6651–6665.

Sik, A., Penttonen, M., and Buzsáki, G. (1997). Interneurons in the hippocampal dentate gyrus: an in vivo intracellular study. Eur J Neurosci 9, 573–588.

Silberberg, G., Wu, C., and Markram, H. (2004). Synaptic dynamics control the timing of neuronal excitation in the activated neocortical microcircuit. J Physiol 556, 19-27.

Skaggs, W. E., McNaughton, B. L., Wilson, M. A., and Barnes, C. A. (1996). Theta phase precession in hippocampal neuronal populations and the compression of temporal sequences. Hippocampus 6, 149–172. 112

Skaggs, W. E. and McNaughton B. L. (1996). Replay of neuronal firing sequences in rat hippocampus during sleep following spatial experience. Science 271, 1870–1873.

Soltesz, I. and Deschenes, M. (1993). Low- and high-frequency membrane potential oscillations during theta activity in CA1 and CA3 pyramidal neurons of the rat hippocampus under ketamine- xylazine anesthesia. J Neurophysiol 70, 97-116.

Squire, L. R. and Zola-Morgan, S. (1991). The medial temporal lobe memory system, Science 253, 1380–1386.

Stelzer, A., Slater, N. T., and Bruggencate, G. (1987). Activation of NMDA receptors blocks GABAergic inhibition in an in vitro model of epilepsy. Nature 326, 698-701.

Stelzer, A., Simon, G., Kovacs, G., and Rai, R. (1994). Synaptic disinhibition during maintenance of long-term potentiation in the CA1 hippocampal subfield. Proc Natl Acad Sci U S A 91, 3058-62.

Stevens, C. F. and Wang, Y. (1994). Changes in reliability of synaptic function as a mechanism for plasticity. Nature 371, 704–707.

Stricker, C., Field, A. C., and Redman S. J. (1996). Changes in quantal parameters of EPSCs in rat CA1 neurones in vitro after the induction of long-term potentiation. J Physiol 490, 443- 454.

Sutherland, G. R. and McNaughton, B. (2000). Memory trace reactivation in hippocampal and neocortical neuronal ensembles. Curr Opin Neurobiol 10, 180–186.

Suzuki, S. S. and Smith, G. K. (1988). Spontaneous EEG spikes in the normal hippocampus. II. Relations to synchronous burst discharges. Electroencephalogr Clin Neurophysiol 69, 532–540.

Swadlow, H. A. (1988). Efferent neurons and suspected interneurons in binocular visual cortex of the awake rabbit: receptive fields and binocular properties. J Neurophysiol 59,1162–1187.

Szabadics, J., Varga, C., Molnar, G., Olah, S., Barzo, P., and Tamas, G. (2006). Excitatory effect of GABAergic axo-axonic cells in cortical microcircuits. Science 311, 233–235. 113

Szabadics J., Tamas G., Soltesz I. (2007). Different transmitter transients underlie presynaptic cell type specificity of GABAA,slow and GABAA,fast. Proc Natl Acad Sci U S A 102, 14831- 14836.

Tamas, G., Buhl, E. H., and Somogyi, P. (1997). Fast IPSPs elicited via multiple synaptic release sites by different types of GABAergic neurone in the cat visual cortex. J Physiol 500, 715–738.

Tamás, G., Lörincz, A., Simon, A., Szabadics, S. (2003). Identified sources and targets of slow inhibition in the neocortex. Science 299, 1902-1905.

Tamas, G., Szabadics, J., Lorincz, A., Somogyi, P. (2004). Input and frequency-specific entrainment of postsynaptic firing by IPSPs of perisomatic or dendritic origin. Eur J Neurosci 20, 2681-2690.

Thompson, S. M., Capogna, M., and Scanziani, M. (1993). Presynaptic inhibition in the hippocampus. Trends Neurosci 16, 222–227.

Tonegawa, S., Nakazawa, K., and Wilson, M. A. (2003). Genetic neuroscience of mammalian learning and memory. Philos Trans R Soc Lond B Biol Sci. 358, 787-95.

Tornberg, J., Voikar, V., Savilahti, H., Rauvala, H., and Airaksinen, M. S. (2005). Behavioural phenotypes of hypomorphic KCC2-deficient mice. Eur J Neurosci 21, 1327-1337.

Traub, R. D. and Bibbig, A. (2000). A model of high-frequency ripples in the hippocampus based on synaptic coupling plus axon-axon gap junctions between pyramidal neurons. J Neurosci 20, 2086–2093.

Tsien, J. Z., Chen, D. F., Gerber, D., Tom, C., Mercer, E. H., Anderson, D. J., Mayford, M., Kandel, E. R., Tonegawa, S. (1996). Subregion- and cell type-restricted gene knockout in mouse brain. Cell 87, 1317–1326.

Tsien, J. Z., Huerta, P. T., and Tonegawa, S. (1996). The essential role of hippocampal CA1 NMDA receptor-dependent synaptic plasticity in spatial memory. Cell 87, 1327–1338. 114

van Strien, N. M., Cappaert, N. L., and Witter, M. P. (2009). The anatomy of memory: an interactive overview of the parahippocampal-hippocampal network. Nat Rev Neurosci 10, 272- 282.

Vilen, H., Eerikäinen, S., Tornberg, J., Airaksinen, M. S., and Savilahti, H. (2001). Construction of gene-targeting vectors: a rapid Mu in vitro DNA transposition-based strategy generating null, potentially hypomorphic, and conditional alleles. Transgenic Res 10, 69-80.

Vogt, K. E. and Nicoll R. A. (1999). Glutamate and gamma-aminobutyric acid mediate a heterosynaptic depression at mossy fiber synapses in the hippocampus. Proc Natl Acad Sci U S A 96, 1118–1122.

Wang, X. J. and Buzsáki, G. (1996). Gamma oscillation by synaptic inhibition in a hippocampal interneuronal network model. J Neurosci 16, 6402–6413.

Webb, T. I. and Lynch, J. W. (2007). Molecular pharmacology of the glycine receptor chloride channel. Curr Pharm Des 13, 2350-2367.

Whitlock, J. R., Heynen, A. J., Shuler, M. G. and Bear, M. F. (2006). Learning induces long- term potentiation in the hippocampus. Science 313, 1093–1097.

Whittington, M. A., Traub, R. D., and Jefferys, J. G. (1995). Synchronized oscillations in interneuron networks driven by metabotropic glutamate receptor activation. Nature 373, 612– 615.

Wilson, M. A. and McNaughton, B. L. (1994). Reactivation of hippocampal ensemble memories during sleep. Science 265, 676–679.

Wittenberg, G. M. and Wang, S. S. (2006). Malleability of spike-timing-dependent plasticity at the CA3-CA1 synapse. J Neurosci 26, 6610-6617.

Woo, N. S., Lu, J., England, R., McClellan, R., Dufour, S., Mount, D. B., Deutch, A. Y., Lovinger, D. M., and Delpire, E. (2002). Hyperexcitability and epilepsy associated with disruption of the mouse neuronal-specific K-Cl cotransporter gene. Hippocampus 12, 258-268. 115

Woodin, M. A., Ganguly, K., and Poo, M. M. (2003). Coincident pre- and postsynaptic activity modifies GABAergic synapses by postsynaptic changes in Cl- transporter activity. Neuron 39, 807-820.

Wu, G., Malinow, R., and Cline, H. T. (1996). Maturation of a central glutamatergic synapse. Science 274, 972–976.

Yamada, J., Okabe, A., Toyoda, H., Kilb, W., Luhmann, H. J., and Fukuda, A. (2004). Cl- uptake promoting depolarizing GABA actions in immature rat neocortical neurones is mediated by NKCC1. J Physiol 557, 829–841.

Ylinen, A., Bragin, A., Nadasdy, Z., Jando, G., Szabo, I., Sik, A., and Buzsaki, G. (1995). Sharp wave-associated high-frequency oscillation (200 Hz) in the intact hippocampus: network and intracellular mechanisms. J Neurosci 15, 30–46.

Yun, S. H., Mook-Jung, I., and Jung, M. W. (2002). Variation in effective stimulus patterns for induction of long-term potentiation across different layers of rat entorhinal cortex. J Neurosci 22, RC214.

Zilberter, Y., Kaiser, K.M., and Sakmann, B. (1999). Dendritic GABA release depresses excitatory transmission between layer 2/3 pyramidal and bitufted neurons in rat neocortex. Neuron 24, 979–988.