Rhythmic Activity In Cells Of The Medial Septum And Diagonal Band Of Broca In Freely Moving Rats.

Charles King

Thesis submitted for the degree of Doctor of Philosophy

Department of Anatomy and Developmental Biology University College London February, 1997 ProQuest Number: 10106913

All rights reserved

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest.

ProQuest 10106913

Published by ProQuest LLC(2016). Copyright of the Dissertation is held by the Author.

All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. Microform Edition © ProQuest LLC.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 Der Ring des Nibelimgen R. Wagner Abstract

This thesis presents results from simultaneous, chronic recordings from the septum/diagonal band of Broca cells in freely moving rats. The rats were trained to run for a food reward in a feature-rich environment from end to end of a linear track, which was oriented in one of eight directions. The cells displayed similar activity profiles as those seen in anaesthetised rats, including; highly rhythmic bursting cells; cells with variable rhythmicity; and cells that displayed no rhythmicity. All rhythmic cells fired in phase with hippocampal theta rhythm. The preferential phases of the cells varied widely, but tended to correspond to two poles at the positive and negative peaks of the theta rhythm and were maintained over time.

The highly-rhythmic cells were strongly phase-locked to hippocampal theta rhythm, but often displayed rhythmic activity in the absence of theta, particularly before the start and after the end of a burst of theta in the hippocampus. These cells also varied in terms of burst reliability, with some cells demonstrating a significant number of burst failures during theta rhythm. Those cells that were most clearly phase-locked with the theta rhythm also showed an increase in frequency of burst firing corresponding to an increase in the rat's running speed. Cells were also found whose rhythmicity showed significant variation depending on the spatial orientation of the linear track and the direction of motion of the rat.

The findings are discussed with reference to the mechanisms of theta rhythm generation and current work concerning temporal coding in the nervous system, as opposed to coding by rate modulation.

This work was supported by the Wellcome Trust and the MRC (UK). Chapter One Introduction ...... 7 A. Temporal coding ...... 8 Binding in the visual cortex ...... 9 The information content of neural signals ...... 11 The detection of temporal coincidence ...... 13 Temporal variance in neural response rates ...... 21 Clock-referenced coding in the hippocampus ...... 27

B. The anatomy and connections of the medial septum/diagonal band of Broca ...... 34 Gross anatom y ...... 34 Afferent connections of the MS/DBB ...... 34 Efferent connections from the MS/DBB ...... 38

C. The classification of cells of the medial septum/diagonal band of Broca ...... 40 Functional categorisation ...... 40 Anatomical and Electrophysiological Categorisation ...... 50

D. The medial septum/diagonal band of Broca and spatial memory ...... 56

E. Non-spatial elements of MS/DBB function ...... 63

F. Theta rhythm ...... 67

G. The hippocampus: structure ...... 80 Gross anatom y ...... 80 Synaptic connections within the hippocampus ...... 84 Afferent connections to the hippocampus ...... 86 Efferent connections of the hippocampus ...... 91 Cortical Efferents ...... 91 Thalamic connections ...... 92 Subcortical connections ...... 93

H. The hippocampus: function ...... 94 Memory ...... 94 Cognitive mapping theory ...... 95 Working vs Reference memory ...... 100 Procedural vs Declarative memory ...... 103 Configurai Association Theory ...... 105 The comparator model ...... 108 Temporal continuity ...... 110 Spatial navigation ...... 113

I. Conclusion ...... 117 5

Chapter Two Methods ...... 119 Electrode Construction ...... 120

Surgery ...... 124 Implant into medial septum ...... 126 Implant into hippocampus ...... 128

Recording sessions ...... 132 Electrical recording apparatus ...... 132 Behavioural tasks ...... 135

Histology ...... 139

Data Analysis ...... 140

Chapter 3 Histology ...... 144

Chapter 4 Firing behaviour ...... 152 Results ...... 155 Discussion ...... 158

Chapter 5 Phase relationship with hippocampal theta rhythm 168 Methods ...... 170 R esults...... 171 Discussion ...... 182

Chapter 6 Determining cell type by action potential shape ...... 184 Results ...... 189 Discussion ...... 197

Chapter 7 Relationship between firing rate of MS/DBB cells and hippocampal theta 200 Methods ...... 201 Results ...... 203 Discussion ...... 206

Chapter 8 The behaviour of rhythmically bursting MS/DBB cells 211 Methods ...... 212 Results ...... 213 Burst frequency stability ...... 213 Correlations between burst frequency and running speed ...... 216 Discussion ...... 222 Chapter 9 Directional rhythmicity 226 Methods ...... 227 Results ...... 234 Discussion ...... 235

Chapter 10 Discussion: Models of septo-hippocampal function ...... 256 The production of theta rhythm ...... 260 Mechanisms of theta rhythm potentiation ...... 262 Mechanisms of theta rhythm generation ...... 265 Mechanisms of theta rhythm regulation ...... 270

Conclusion ...... 272

References ...... 274

Acknowledgements ...... 304 Chapter One

Introduction ______Page 8

One of the basic tasks of modem neuroscience is the continuing search to understand the method by which individual nerve cells encode the data which they transmit to other cells. Without such an understanding, our attempts to elucidate the computational mechanisms which underlie the function of the brain must be fruitless.

In this introduction I will, first, discuss the current research and opinions regarding neuronal codes, with special regard to recent evidence which suggests that the times at which single action potentials are fired conveys substantial information to the target neuron. I will subsequently focus on one particular example of temporal coding, that found in the hippocampus. This thesis is concerned with the behaviour of cells of the medial septum and diagonal band of Broca (MS/DBB), a group of cells responsible for sending rhythmic, clock­ like impulses to the hippocampus and which, therefore, play an important role in the coding mechanism used in this area.

A. Temporal coding

The classical method of neural codification is also the most simple: nerve cells express the presence of the stimuli to which they are responsive by increasing their rate of firing (Adrian, 1926, Adrian and Zotterman, 1926; Adrian, 1928;

Barlow, 1972), otherwise known as rate modulation. This has come to be accepted as the canonical method of data encoding by neurons, and is found in all parts of the central and peripheral nervous system.

It has been thought for some time, however that, under certain conditions, other coding methods may be used. The concept that the timing of neural activity ______Page 9

may play a role in the expression of information first appeared in the course of

investigations into neural responses to auditory phenomena (Wever, 1949), and

into the oscillations that occur in olfactory cortex when the animal inspires

(Adrian, 1950; Freeman, 1975). It was later found that the nature of this activity

varied corresponding to the odour presented and its significance to the animal,

and it was thought that the oscillations acted as a means of ensuring synchronous

activity among a large number of cells (Freeman, 1987), and this synchronisation

process was successfully modelled using a network of realistic neural

conductances (Wilson and Bower, 1989). In the visual system, as well,

statistical and information theoretical analyses of spike trains suggested the

presence of a temporal code (Eckhorn and Popel, 1974, 1975; Cattaneo et al.,

1981).

Binding in the visual cortex

It was not until the last decade, however, that it began to be suggested that temporal coding was present in other areas. Building on previous work that had shown the presence of oscillatory activity in field potential recordings over the visual areas (Lopes Da Silva et al., 1970; Rougeul et al., 1979), Charles Gray and

W olf Singer (1989) investigated areas 17 and 18 of the brains of kittens presented with a visual stimulus, and found a relationship between multi-unit activity and oscillations in the local field potential. Further studies have demonstrated correlations in unit activity between single units which are anatomically close to each other (Gray and Vianna Di Frisco, 1993), between different cortical columns separated by 7mm (Gray et al., 1992), between different areas of visual cortex (Engel et al., 1991b), and between different cortical hemispheres (Engel et al., 1991a). ______Page 10

It is thought that this synchronisation of neural activity provides a solution to

the "binding problem", in which the visual system must successfully segregate

the low-level features in the visual field into groups of elements which belong

to the same object. In this model (Engel et ah, 1992, Singer and Gray, 1995),

nerve cells in sensory areas represent the presence of a specific object by

means of a population code, in which the cells corresponding to the various

features which compose that object all fire synchronously (within 1 to 2ms).

This allows a cell that represents a specific feature to fire in response to the

presence of many different objects, while segregating its responses so that cells

upstream can distinguish between them.

This was elegantly demonstrated in a recent paper (Kreiter and Singer, 1996), in

which two cells from the M T (V5) visual cortex of macaque monkeys were

recorded simultaneously, while the animal was presented with a variety of bar

stimuli. The cells were sensitive to motion in different directions and their

receptive fields partially overlapped. A single bar moving at an angle midway

between the maximally-preferred directions of both cells caused the cells to

fire in a correlated fashion, and the spikes for the different cells were

synchronised at a millisecond level. W hen the two cells were stimulated by

different bars, each moving at the optimal direction for the corresponding cell,

however, there was a dramatic loss in the correlation between the two cells.

As Singer and Gray point out (1995), the presence of temporal coding adds a

powerful extra dimension to the range of data which can be expressed by individual nerve cells, allowing these cells to encode both the stimulus strength and the relation between that stimulus and other features in the sensory field.

Population coding offers a useful alternative to the problematic 'grandmother ______Page 11

cell' hypothesis, in which the presence of specific objects is represented by the

activity of correspondingly specific cells in the 'higher' areas of the cortex. This

method of object encoding also appears to coincide with the data available

concerning those sensory areas which are further removed from the primary

inputs, in which the neurons respond to complex clusters of features which

differ from cell to cell, but correspond to common objects (Gross et al., 1972;

Perret et al., 1987).

While the neurons which show these synchronous responses also demonstrate

activity oscillations in the gamma band (30-80Hz), these oscillations are

short-lived and do not exhibit a coherent overall pattern (Gray et al., 1992). It

appears that they serve to synchronise neural operation over only short time

periods, before decaying.

The information content of neural signals

This research has raised an important and controversial issue, namely: to what extent are individual neurons capable of producing and maintaining signals which have a temporal resolution on the order of a millisecond? It has been thought for a long time that single cell action potentials represent a stochastic response to the pooled inputs terminating on the neuron (Adrian, 1926), and that nerve cells merely acted as some form of 'leaky integrator' (Knight, 1972), summating inputs over a relatively long time-frame. This view has been challenged by theoretical work performed in the past few years.

Bialek and his co-workers have produced a model of sensory stimulus reconstruction of the spike trains produced by the H I neuron of the blowfly Page 12

Calliphora erythrocephela (Bialek at al., 1991; Bialek and Reike, 1992). They

recorded the action potentials output by this neuron when the relevant

photoreceptors were fed a pattern of rigid horizontal movements. The spike

timeline was then used to estimate the original signal using a set of nonlinear

filters of the form:

where t- is the time at which each spike arrives and the filters are chosen to minimise a measure of the error between the estimated and actual stimulus

patterns.

In the event, it was found that reconstructions using just the first, linear term of the equation produced results of sufficiently high quality, and the addition of further terms did not add significantly to the accuracy of the result. The reconstructed signal displayed very low noise levels, and, when integrated over a biologically relevant time period of 30ms, could discriminate movements with an amplitude of one tenth the distance between photoreceptors, illustrating that significant amounts of information can be transmitted using a very small number of spikes (the maximum firing rate of the HI neurons that were studied is

100-200Hz). Furthermore, it was found that 95% of the original information content of the signal could be reconstructed even when the spike timeline was subjected to a Gaussian timing jitter with a standard deviation of 2ms, indicating that, in this system, there may be a certain amount of leeway concerning the exact time of arrival of each spike. ______Page 13

The detection of temporal coincidence

Another important and influential piece of theoretical analysis was carried out

by Softky and Koch (1993). This work was based upon examination of

extracellularly recorded spike train data from the primary and extrastriate visual

cortices of monkeys. The authors derived a measure, the coefficient of variation

(Cy), of the variability in the times between each spike (inter-spike intervals)

that is adjusted to take into account variations in the overall firing rate of the neuron. This estimate ranges from zero to one, with low values indicating

highly regular output patterns, while a Cy of one indicates an essentially

random spike train with the characteristics of a Poisson process. They found

that the Cy of neurons in both areas was uniformly high, ranging from 0.4 to 1.2

(the presence of values greater than 1 is accounted for by the error inherent in their method of estimation of Cy).

The data firom real neurons was then compared to that produced by two types of neural model. The first was a simple 'integrate and fire' paradigm in which neural activity is approximated by that of a simple capacitor and associated voltage differential. When the voltage difference exceeds a pre-specified threshold condition, the 'neuron' fires and the capacitor discharges. The charge across the capacitor is incremented by a fixed amount with the arrival of each input. In this simple model, the 'neuron' merely fires after the occurrence of a sufficient number of inputs (N). The rate at which the inputs arrive is irrelevant. When this model is set up with realistic parameters for the threshold and input increment values (corresponding to a requirement for 40-50 inputs before the 'neuron' fires) it produces a highly regular output, as would be expected from the central limit theorem, since Cy becomes equal to INN. In ______Page 14

this case, the Cy produced by the model was 0.16. They found that Cy would

remain below 0.5 for all values of N above 3. This basic model was then refined

to incorporate both the presence of a refractory period, in which the 'neuron' is

incapable of firing for a certain period after the previous spike, and the presence

of 'leaks' across the cell membrane by placing a resistance in parallel with the

capacitor. Both additions resulted in a change of Cy dependent on the firing

rate. The incorporation of a refractory period was found to have little effect at

low firing rates, but caused the Cy to drop asymptotically as the mean time

between spikes (At) approached the refractory period. With the addition of a leak conductance to the model, Cy became dependent on the ratio of the

'membrane' time constant (x) and At. While the presence of a leak generally caused the Cy to attain higher values than in the condition without a leak, as it caused the system to "forget" previous spikes to a certain extent and thereby to act as a coincidence detector, a x value of 13 ms combined with an N of 51 produced a maximum Cy of only ~0.3 at low firing rates (mean time between spikes=30ms). An increase in the Cy to those levels found in the real data could only be achieved by reducing x to 0.2ms. This should be compared to data acquired by patch-clamping hippocampal cells, which demonstrate a membrane time constant of 23-107ms (Spruston and Johnston, 1992; Williams et al., 1994).

These data were acquired from cells in slices,, and it is possible that the actual time constant of the same cells in vivo is shorter because of the increased number of active membrane conductances in such a situation (Koch et al.,

1996). The membrane time constant values determined using sharp electrodes are much lower, presumably because of the large leak current caused by the membrane puncture intrinsic to this technique. The authors went on to test the effects of other variations in the model, including variation in the magnitude and time duration of the excitatory inputs and the presence of adaptation by the ______Page 15 addition of an I^pjp current, but none of these effects were capable of generating sufficiently high Cy values when the model was supplied with realistic parameters.

In another approach the nerve cells were modelled using compartmental techniques (Segev et al., 1989). This is a form of neural modelling in which the function of an individual nerve cell is approximated by a network of isopotential compartments, each containing a resistance in parallel with a capacitance to represent charge segregation across the cell membrane, as well as dynamic conductances which represent the effect of synaptic and voltage- or ligand-gated ionic membrane channels. This technique allows the fine structure of current flows through a neuron's dendritic tree to be modelled with an accuracy dependent only upon the basic data and the amount of computer power available.

A biologically realistic way of modifying the behaviour of such models is by changing the rise time and the peak conductance associated with each type of synapse, as well as by altering the position of the excitatory and inhibitory synapses. For the purposes of investigating output regularity, Softky and Koch constructed models using two different sets of parameters. One, the

"conventional" model, used parameters that agreed with those found in cortical neurons (Douglas et al., 1991); while the other, "barely plausible", model used parameters that were designed to maximise the cell's variability at the expense of low-level realism. For instance, the excitatory synapses were clustered at a point only 60|Lim away from the soma and had a maximal conductance of lOnS.

Only 19 of the "giant" EPSPs produced by these synapses were required to cause the cell to fire an action potential. Initially, they confined all the ______Page 16

voltage-gated conductances to the soma compartment, leaving a passive model of

the dendritic tree. Apart from the fast sodium and delayed-rectifier potassium

currents needed to form an action potential, they also included calcium, slow

sodium and the transient and after-hyperpolarising potassium currents.

Both the "conventional" and "barely plausible" variants produced output spike

trains with a Cy lower than that found in real nerve cells. As with the modified

versions of the leaky integrator model, the Cy varied with the firing rate, as

lower firing rates resulted in a higher Cy, although this effect rapidly saturated.

Only the "barely plausible" parameter set yielded results close to the real data,

and then only when the mean time between spikes was 10ms or more. Even so,

the Cy values that were yielded were in the region of 0.6 to 0.7, at the low end

of the values found in real neurons.

It was only when the authors added in active membrane conductances in the dendritic tree compartments that the Cy of the output spike trains mimicked those obtained from the monkey data. These conductances were pairs of strong voltage-gated depolarising and delayed-rectifier currents. In this model, every input was set up so that it would activate these conductances and fire a dendritic spike, but the cell was stripped of many of the voltage-dependent conductances at the soma, so that only the Hodgkin-Huxley sodium and potassium currents remained. High degrees of variability were only obtained when the maximal conductance of the delayed-rectifier current was 3.5 times that of its somatic counterpart. Due to the relatively high resistance between adjacent compartments in these thin branches of the distal dendritic tree, the individual synaptic areas were still effectively decoupled from each other despite the strong conductances involved. The strong dendritic repolarisation ______Page 17 current caused the model cell to "forget" about previous activity and acted to increase variability by changing the cell's response to input spikes from one of temporal integration to coincidence detection. It should, however, be noted that, while the conductances present in neural dendrites do include the fast sodium currents responsible for generating the action potentials found in this model, they also include the slower calcium conductances in sufficient number to prolong the overall length of most dendritic spikes (Benardo et al., 1982), negating the fast repolarisation necessary for a high Cy in this model.

Thus, an overall analysis of the results of these various modelling attempts yielded the hypothesis that there are four possible explanations for the variability seen in real neurons: an extremely strong outward leak current sufficient to drop the membrane time constant to around 0.2ms; the presence of extremely strong excitatory synapses capable of causing over 1 OmV of depolarisation per event; large numbers of all-or-none dendritic spike potentials coupled with strong repolarisation of these spikes; or the presence of synchronised, non-random behaviour in the inputs to the cell.

I have dealt with this paper in some length as it presents very strong theoretical support for the thesis of temporal coding and has engendered substantial debate.

A strong argument in opposition was proposed by Shadlen and Newsome

(1994). These authors compared three models of the functional output of a single nerve cell. Each model was constrained to produce an output firing rate which was roughly similar to the rates of each of its inputs, and was subject to a barrage of excitatory synaptic activity, as it was assumed that roughly 10% of each unit's

-3000 excitatory inputs would be active during a 30-50ms time period, leading ______Page 18

the authors to present each model with 300 inputs, each firing at approximately

lOOHz. The time of arrival of each spike was random.

The first, a simple integrate-and-fire model, produced, as expected from the central limit theorem, a highly regular output, as it simply waited for the arrival of 300 inputs before firing a spike. Secondly, they presented a coincidence-detector model which required that 35 inputs arrive within a certain time of each other (determined by the membrane time constant) in order to fire. In this case, they found that, owing to the high rate of excitatory input, the model produced far too high an output rate unless the time constant was set at 1ms or below, a figure which, as discussed above, is far below the physiological level.

Finally, a model was presented in which the large amount of excitation was counter-balanced by a comparable degree of inhibition originating from a smaller number (150) of inhibitory synapses, which provided strong inhibition.

The IPSP amplitude was twice that of the EPSP, leading to an approximate balance between excitatory and inhibitory influences. This was justified by the prevalence of inhibitory synapses close to the cell body, the higher firing rate of inhibitory cells and reports of frequent failures of excitatory inputs to produce an EPSP (see references in Shadlen and Newsome, 1994). This resulted in a model which followed a "random walk", in which the somatic membrane potential fluctuated in a random fashion, producing spike outputs whose inter-spike interval was similarly random. Thus, by balancing excitation and inhibition, the authors were able to reproduce the high degree of variance seen in cortical cell spike trains without incorporating any form of structure into the input patterns. ______Page 19

In reply, Softky (1995) recreated Shadlen and Newsome's model and analysed it

further by examining the cross-correlogram formed when each output spike was

correlated with the surrounding EPSPs and IPSPs. He showed that the "random

walk" model did indeed demonstrate a marked association between a typical

output spike and an increase (on the order of 60%) in the number of EPSPs

arriving in the previous millisecond as well as a 90% decrease in the number of

preceding IPSPs. It was suggested that the model was simply discovering the

inevitable coincidences within the input data. Furthermore, calculation of the

effective time constant in the face of the large amount of outward inhibitory

current yielded a value of 0.67ms, justifying the contention that integrative

events were occurring at a sub-millisecond timescale. As was noted by Softky, however, this model only includes fast sodium-spiking currents, and the

addition of the slower, calcium-dependent dendritic potentials may lead to a loss of temporal resolution. In conclusion, Softky presented the argument that temporal coding is attractive because of its efficiency, as a temporally-coded spike train can carry an order of magnitude more information than a rate code.

In a final comment, Shadlen and Newsome (1995) replied by presenting two sets of data from their random-walk model. In one, the set of excitatory and inhibitory inputs which had occurred in the two milliseconds prior to the production of a burst of two spikes was noted. A second set of input trains was then prepared, into which the exact configuration of inputs which had previously yielded a spike was inserted at three different places. In none of the three cases did this set cause a spike to be fired (this account is based on the figure the authors presented, as there is some discrepancy with the text in the body of the article). They contend that this provides evidence that the random-walk model disregards the exact temporal nature of its inputs, as the ______Page 20

same code was incapable of reliably firing an action potential. Although this is a

powerful rebuttal, this argument could benefit from further quantitative

analysis, and unfortunately, they do not provide data concerning the state of the

inhibitory synapses prior to the onset of each code insertion. Finally, the

Shadlen and Newsome model is based on a single compartment with no

segregation of the various input conductances. As Softky pointed out (Softky,

1995), nonlinear processes occurring in the fine branches of the dendritic tree

may, in conjunction with amplification by active dendritic membrane currents,

provide a basis for fine temporal resolution.

This topic has been further developed by Koch, Rapp and Segev (1996), in which the nature of the membrane time constant (t^) was discussed. This value

represents the time taken for the membrane voltage to settle back to its resting

potential after a charge is applied to it and can be expressed as where

equals the capacitance of the cell membrane and is the inverse of the

trans-membrane conductances that exist in parallel to this capacitance.

Historically, estimates of have increased dramatically from the early values of 2-4ms (Coombs et al., 1955) to present day values of 20-100ms (Spruston and

Johnston, 1992; Williams et al., 1994). As mentioned above, it is thought that the dramatic increase in found with modern perforated patch-clamp techniques is a result of errors introduced by the almost inevitable shunt conductance caused by disruption of the cell membrane when the cell is impaled with sharp intracellular electrodes. Koch and his co-workers note, however, that these high modern estimates still only reflect the effects of a charge injection at the soma. They present a model which demonstrates the different responses to a brief pulse of current injection in different parts of the dendritic tree. While such a pulse at the soma results in a long-lasting "EPSP", a similar event ______Page 21 occurring at the distal end of a basal dendrite causes a much briefer voltage change, as there exists a relatively low-resistance pathway to the soma, which shunts a large enough proportion of the charge to cause a relatively rapid return to the resting level. This implies that the effective time constant of current flows in the distal dendrites is much shorter than somatic measurements would lead us to believe, and supports the hypothesis that synaptic inputs may interact on a time scale shorter than the membrane time constant.

Temporal variance in neural response rates

It is significant that the bulk of the debate concerning temporal coding in the nervous system has focussed around the issue of neurons’ ability to respond to patterns of activity in the millisecond range. This thesis is central to the idea that nerve cells act as coincidence detectors, a function which is necessary if the claims of feature binding through synchrony are to be accepted. An alternative approach to temporal coding is found in the work of Richmond,

Optican and their co-workers.

These investigators have studied the strategies used in encoding the colour and form of objects in the visual scene. Their work has mostly been done in awake monkeys performing a fixation task. These are then presented with a range of visual stimuli in the form of Walsh patterns, an 8x8 array of binary pixels. The full set of 128 members contains all the possible variants of the binary elements.

In the first set of experiments discussing temporal coding (Richmond et al. 1987,

Optican and Richmond, 1987) they studied the neurons in the inferior temporal

(IT) area and demonstrated that these cells responded by producing an extended ______Page 22

train of spikes whose rate varied in a complex manner over time. Initial analyses

of the data using a bootstrap Monte Carlo technique suggested that these variations were not simply random. They went on to measure the information content of the signal using principal components analysis, modelling the neuron as an information channel transmitting messages about parameters of the stimuli.

While the results of such an analysis do not represent the actual manner in which the nervous system utilises the data coming from these neurons, it does reflect the amount and types of information present in the signal. Briefly, the method of analysis produces a set of orthogonal temporal waveforms which describe the covariance of the neuron’s data and are typically arranged in order of descending variance (Richmond and Optican, 1987). The mean value of the first principal component generally corresponds to the cell’s firing rate, and the following ones measure the information contained in temporal variations of that rate. Overall, this demonstrated that the information carried by the first three principal components was roughly double that presented by the spike count alone.

Subsequently, these authors demonstrated a similar effect in the primary visual cortex (Richmond et al., 1990, Richmond and Optican, 1990). As before, they showed that the bulk of the transmitted information was contained in the first three principal components and that these three components carried over twice the amount of information found in the spike rate alone.

They have also examined the activity of IT neurons when the animal is involved in a delayed non-matching to sample task in order to see whether the temporal coding of these cells displayed any memory effects (Eskander et al.,

1992a, 1992b). They used an information theoretic analysis (Optican and

Richmond, 1987) to compare the transmitted information present in different ______Page 23

input/output pairings, the various input codes being the different patterns that

were presented and the context (whether the pattern was a sample, matched the

previous one, or didn’t match), and the output codes were the first three

principal components or the spike count on its own. Analysis of the neurons’

response showed that they were transmitting data concerning both the patterns

and the context in which they occurred, with some cells showing a dramatic

change in their response to a particular test stimulus dependent on whether it

matched the previous one or not. Again, transmitted information was

significantly higher when the first three principal components were used

compared to the spike count alone. Further analysis examined the amount of information carried about the current and prior stimulus patterns. Although the output code based only on spike count transmitted no information about the previous patterns, one based on the first three principal components showed significantly more data about these than found in the firing of these cells in between trials. This suggested a segregation of the information being transmitted by these cells between the temporal and spike count codes, with temporal coding carrying 65% of the information about previous patterns and 70% of that concerning the behavioural context, while the spike count expressed 65% of the transmitted information about the current pattern.

One critique of the principal components methodology was raised by Tovée et al. (1993), who demonstrated that making a correction for differing response latency caused the information carried by the second and third principal components to drop dramatically. As the mean value of the first principal component is correlated with the overall mean firing rate, this implied that the temporal structure of the waveform did not add a great deal of information.

These authors also noted that much of the information about a stimulus was ______Page 24 transmitted over the first 50ms after the response latency. Despite the corrections, however, their calculations still showed that the information carried in the first three principal components was considerably higher than that transmitted in the firing rate alone. For instance, the information transmitted during the period between 100ms and 150ms after stimulus onset by the first three principal components was about 0.5 bits, whereas that transmitted by the mean firing rate alone was roughly 0.29 bits.

Principal component analysis is, of course, not the only method of measuring information transmitted in a temporal code. A recent set of experiments

(McClurkin and Optican, 1996, McClurkin et al., 1996) explored an alternative analytical method, and examined the role of temporal coding in binding together information about different aspects of the stimulus was examined. It has long been known that there are two major anatomical and processing pathways involved in visual perception, one leading to the parietal cortex and concerned with movement, the other leading to the inferior temporal cortex and dealing with colour and static form analysis (Zeki, 1980; Maunsell and van Essen, 1983).

On the basis of pure spike rate coding, some cells in the visual pathway leading to the inferior temporal cortex respond maximally to colour and some to monochrome patterns (Hubei and Livingston, 1987), and it has been hypothesised that there are two separate pathways leading from VI to V4.

In order to test this, McClurkin and Optican used a similar experimental paradigm to that described above, but they incorporated one of six colours into the Walsh patterns. The monkeys were trained on a pattern discrimination task, in which they were presented with a group of three stimuli, each composed of different colours and Walsh patterns, and one of which was located at the centre ______Page 25

of the receptive field of the cell under study, while the others were spaced

equidistant to the fixation point. A cue consisting of either a plain square of a

particular colour or a black and white Walsh pattern was then flashed on the

screen at the fixation point, and the monkey had to make a saccade towards the

peripheral stimulus that matched the cue, ie if a red square was flashed on, the monkey had to saccade towards the pattern coloured red. The authors then

analysed the information transmitted by the neuron using two possible output codes: the simple spike count over a 128ms epoch, and a vector formed by measuring the activity in each of 64 consecutive 2ms time periods. They showed that, while individual neurons in areas VI, V2 and V4 could be segregated into pattern- and colour-sensitive cells on the basis of their spike count alone, the amount of information transmitted about both parameters using the temporal code was nearly equal in the majority of cases. Further examination of the development of this information over time showed that, while the waveform code grew to carry considerably more information than the spike count, the latter’s information developed faster during the first 20ms of the response. They also showed that information about both pattern and colour developed more slowly in V4 than in VI or V2, although information transmission was still developing in VI and V2 when it started to be expressed by V4, suggesting the presence of feedback from V4.

In the second paper of the set (McClurkin et al., 1996), they went on to develop the multiplex filter theory first put forward by Optican and Richmond in 1987.

In this model, neurons contain a variety of filters sensitive to different aspects of the incoming stimuli. The results of these filters are then combined in a

(presumably) orthogonal manner and used to determine the temporal variation of the cell’s output spike rate. They showed that the waveforms representing ______Page 26

the various colours and patterns could be clustered into distinct groups, and

that neurons that represented a particular colour or pattern using similar

waveforms also tended to use similar waveforms for the representation of the

other patterns or colours. They also discovered that, when comparing the

representational waveforms found in the two monkeys used for the experiment,

those for three of the six patterns and five of the six colours were similar (in that

their differences were not statistically significant at the p<0.05 level).

Furthermore, they found that the waveforms for four of the six colours were

similar between all the cortical areas studied (VI,V2 and V4), whereas those for

the patterns all showed significant differences, suggesting that the temporal

codes for colour are fixed at a relatively low level, whereas those for pattern are

substantially altered with each successive relay.

Cortical cells in the visual sensory areas have a wide dynamic firing range, and can alter their output firing rate from close to 0 to over 300Hz. This provides

extensive scope for analogue data encoding at a fine scale as proposed by

Richmond and Optican (see above). In many parts of the brain, though, the dynamic range of the principal neurons is not nearly comparable.

These three sections have discussed lines of research which provide strong support for the presence of a temporal element to neuronal coding in the mammalian neocortex. In the following section I will introduce the evidence for a clock-based code in the hippocampus and, subsequently, discuss the mechanisms responsible for producing the oscillations against which the action potentials of the principal cells are aligned. ______Page 27

Clock-referenced coding- in the hippocampus

A good example lies in the pyramidal cells of the hippocampus, which, as will

be discussed, fire when the rat is in a particular part of the environment

(O’Keefe and Dostrovsky, 1971). These cells are virtually silent when the

animal is not in the place field, and fire at relatively low rates when it does

enter the field. Most cells show in-field firing rates of around 10-20Hz, which can, exceptionally, increase to 40-60Hz (Muller et al., 1987; Barnes et al., 1990).

While, when averaged out over many presentations, the place response of the cell does show a systematic rise in firing rate as the animal nears the centre of the field (O’Keefe, 1976; O'Keefe and Speakman, 1987), there is substantial variation in the firing rate for each individual transition of the place field.

Ensembles of hippocampal neurons have been shown to provide a highly accurate description of the animal’s position in space (Wilson and McNaughton,

1993), particularly when the output of large numbers of cells are taken into account (extrapolations of the data showed that a combination of the outputs of

380 cells integrated over 100ms would provide a positional accuracy of 1cm).

There is evidence that the change in firing rate varies substantially on individual runs, however. Fenton et al. (1995) compared the observed pattern of place cell firing on individual runs to that demonstrated by a Monte Carlo simulation repeated 100000 times. They found that the variation of individual spike trains from in vivo data was significantly higher than that found in the

Monte Carlo runs. The variation could not be explained by details of the path that was run, by drifts in the average firing rate, or by the grouping of spikes into bursts. ______Page 28

It has indeed been shown that the position information expressed by

hippocampal pyramidal cells is not only encoded through their firing rate. The

firing pattern of these cells also demonstrates a relationship with hippocampal

theta rhythm (O'Keefe and Recce, 1993). These cells fire bursts of action

potentials known as complex spikes, and the number of spikes in each burst

increases as the animal nears the centre of the cell's place field. It was found

that the phase of theta rhythm at which each spike fires also undergoes a

systematic change. When the animal enters the place field, the cell generally

fires at the positive peak of local theta rhythm. As it moves through the field, the firing phase precesses backwards, so that spikes fired further into the place

field occur at an earlier phase than the initial spikes. The total amount of phase shift varied from 100° to 355°, and could be fitted with a simple linear equation.

All the data was collected on a simple raised linear track, in which the rat was trained to run back and forth in pursuit of a food reward. The place fields showed highly directional properties, as 14 of the 15 cells analysed showed place fields only when the rat ran in one particular direction.

It was shown that this phase variation was far better correlated with the rat's position than with the time the rat had spent within the place field, alleviating the concern that the phase shift merely reflected an adaptive process within the cell. In fact, one run was reported in which the rat lingered in a particular part of the place field for several theta cycles while searching for food. When it began moving again, the phase at which the cell fired was the same as when it stopped, and subsequently showed the same phase shift as it had before, as if the delay had not occurred. Overall, the mean correlation between firing phase and location within the place field was far higher (0.66±0.17) than that between phase and time within the field (0.42±0.12). ______Page 29

This effect was confirmed by Skaggs et al. (1996). In this experiment, the animals were run on a triangular maze or in an open field. While the precession of firing phase could easily be demonstrated when the rat was confined to the linear environment of a raised track, they found that it was much less obvious in the open field. It was possible, however, to show the effect when the firing phases of the various spikes were segregated into two coarse groupings, each occupying half the complete cycle, and the authors discuss this with regard to the difficulty in determining the exact position of the animal in such an unbounded environment.

In addition, they noted several additional features of the phase precession.

While O'Keefe and Recce stated that the most simple characterisation of the phase correlation was as a simple linear relationship, Skaggs et al. argued that this may include more complex features, specifically:

1) The phase advance occasionally accelerated towards the end of the place field, causing the place/phase plot to curve downwards rather than remain straight.

2) Spikes fired later in a burst appeared to precess at a slower rate than the initial spikes, leading to the complex spike's phase plot broadening as the animal moves through, the place field. Close study of the four cells they present which showed this phenomenon suggests, however, that this effect may be caused by a loss of the phase/place correlate for these terminal spikes when the rat is about halfway through the field. Although the exact electrophysiological mechanisms by which the phase shift occurs are not yet known, it is tempting to speculate that these final spikes, firing on the tail of the burst's depolarising waveform, are, in part, triggered by a rhythmical process within the pyramidal cell, such as that described by Ylinen et al. (1995b). Such a rhythm may tend to act as an Page 30

attractor when the burst starts close enough that the attractive effects are not

negated by the short-term afterhyperpolarisation currents triggered by the

burst's spikes. A point in opposition to this explanation is that the position of

this putative attractor varies by approximately 90“ from the phase of maximal

pyramidal cell activity, whereas one would expect the two to coincide.

3) Those cells which were only weakly active demonstrated less phase

precession than those with a high in-field firing rate. It is possible, of course,

that a decreased firing rate reflects a decreased relevance in terms of coding the

animal's position.

4) Several cells showed a clustering of spikes in certain parts of the place field, yielding an increased density in the corresponding parts of the phase/position plot. As the clustering occurred in two dimensions, it is not clear whether this is due to the superimposition of a phase preference, or to some complex variation of the shape of the place field.

5) Finally, seven cells were found which appeared to undergo a double phase shift. Assuming that these data sets do not represent two cells with closely similar waveforms and place fields (a situation which is unlikely to occur with such frequency), it would seem that the phase advance occurs too rapidly to cover the full extent of the cells' place fields, leading to the firing phase being reset to the starting phase (in at least one example, the two shifts are clearly discontinuous) and then continuing to shift from there.

These differences are essentially minor, however, and the experiments of

Skaggs et al., largely serve to confirm O'Keefe and Recce's original finding.

While the environments in which the experiments were conducted differ, both involve linear movement along well defined paths, and produced place cells with a high degree of directionality (cf. McNaughton et al., 1983). The ______Page 31 importance with regard to temporal coding of directionality, or of the type of movement required is still not certain.

Modelling work is now underway to explain the functional and cellular mechanisms underlying this form of temporal coding. Tsodyks and co-workers

(Tsodyks et ah, 1996) used a network of simple integrate-and-fire units (both excitatory and inhibitory) and demonstrated that the phase shift could be replicated when these units were connected in an asymmetric 'synapses', with those connections arranged in the direction of motion being stronger than those in the opposite direction.

Another model, in which the phase shift is less explicitly encoded into the structure, is that of Jensen and Lisman (1996), which supposes that the 40Hz gamma rhythm carries individual memories and rides on top of the underlying theta wave, allowing the animal to store approximately 7 memories at a time

(Lisman and Idiart, 1995). During a learning mode enabled when the rat is in a new environment, the recurrent collaterals of CA3 are synaptically modified in a process involving NMDA channels. In this mode, the place cells code for the present or very recent past location of the rat. After having learnt the environment, the rat changes into recall mode, in which the place cells express predictions about the future location of the rat. They excite the cell corresponding to the next location along the track through NMDA-mediated synapses which allows a delay of approximately one gamma cycle per location.

As the rat moves into the successive locations, the stack of memories shifts with respect to the underlying theta rhythm, causing the observed phase shift. A merit of this model is that is contains clear predictions which are worthy of ______Page 32 testing, although the exact mechanisms involved in the state change between learning and recall modes are unclear.

In conclusion, this section has shown that there is a growing body of evidence suggesting that temporal coding is used in neuronal communication. While highly persuasive, the case for temporal coding is still far from complete. While different investigations have shown how temporal codes can be produced, transmitted and utilised, these three critical components have not yet been shown to coexist within a single system. For the temporal coding hypothesis to be fully accepted, it will be necessary to demonstrate that one set of cells produces a temporal code, that it is then transmitted to another cell group, and that the temporal qualities of the input have a significant effect on the activity of the target cells.

Further study of temporal coding will require an understanding of the underlying neural mechanisms of these systems. As hippocampal theta rhythm has been shown to be implicated in hippocampal coding strategies, this thesis is aimed at developing a better understanding of the processes responsible for the generation and control of this oscillatory process. As will be discussed in the following section, the area responsible for controlling hippocampal theta rhythm is the medial septum/diagonal band of Broca (MS/DBB), and the results of recordings from this area in freely moving rats will be described in the experimental section of this thesis. First, I will review the literature concerning the MS/DBB and hippocampal theta rhythm. Following that, I will review hippocampal anatomy and the details of the septo-hippocampal pathway. Finally, I will discuss the various current theories of hippocampal Page 33 function to give a wider perspective of the role of the septo-hippocampal system. ______Page 34

B. The anatomy and connections of the medial septum/diagonal band of

Broca

Gross anatomy

Figures 1.1 and 1.2 show the MS/DBB and its anatomical relationship to the hippocampus in the sagittal and coronal planes respectively. This structure is composed of two elements: the medial septum, which comprises a narrow

(~ 1.5mm) band of cells spanning the midline, and the band of Broca, a diagonal patch of cells which extends down from the medial septum bilaterally (Paxinos and Watson, 1986). This is further divided into the vertical and horizontal limbs. The medial septum and the vertical limb of the diagonal band of Broca share anatomical connections and physiological function, but the horizontal limb is related to the latter areas in name only, and it has been suggested that it would be more appropriate to term it the magnocellular preoptic nucleus

(Swanson, 1976; Meibach and Segal, 1977; Amaral and Kurz, 1985).

Afferent connections of the MS/DBB

The medial septum is richly innervated from a large number of brainstem areas, most notably the supramammillary nucleus (SUM), median raphe (MRN), pontine raphe, locus coeruleus, raphe obscurus, nucleus incertus, pars alpha of the central grey, nucleus of Darkschewitsch, lateral mesencephalic grey and the medial aspects of the pontis caudalis (Segal and Landis, 1974; Vertes, 1988).

Although examination of MS/DBB afferents using tritiated amino acid labelling showed a large input from the lateral septum (Swanson and Cowan, 1979), a recent study using PHAL techniques combined with staining for ChAT and Page 35

corpus callosum right cerebral hemisphere

cerebellum

septum anterior commisure fimbriafii temporal pole ventral hippocampal commisure hippocampus

Figure 1.1. The relationship between the hippocampus, the septum, and other major brain structures. The left hippocampus has been exposed by removing the overlying cortex and all forebrain structures apart from those at the midline. The hippocampus is shaded to give an impression of its three-dimensional structure. (After O'Keefe and Nadel, 1978) Page 36

dorsal third ventricle cerebral cortex corpus callosum

hippocampus fimbria

gyrus

medial t^septum

thalamus

vertical limb of the lateral diagonal band of Broca ventricle horizontal limb of the diagonal band of Broca

Figure 1.2. A coronal view of the rat brain showing the relationship between the medial septum/diagonal band of Broca and the hippocampus. The left half shows the major structures seen on a coronal section taken 3.6mm behind the bregma. The right half shows a shaded representation of the three-dimensional hippocampus as it would appear when separated from the surrounding tissue. The dotted line indicates the position and gross anatomy of the medial septum on a coronal section made 0.2mm in front of the bregma. Redrawn from Paxinos and Watson, 1986. ______Page 37 parvalbumin showed that this connection is, in fact very sparse, and mostly composed of parvalbumin-positive (presumed GABAergic) fibres. The main source of inputs which cause synchronised MS/DBB activity appears to be the nucleus pontis oralis (RPO), the central part of the pontine reticular formation.

Stimulation studies have shown that projections from this area separate into three systems, namely the medial longitudinal fasciculus (MLF), the ventrolateral-tegmental medial forebrain bundle (MFB) and the dorsolateral central tegmental tract (CTT) (Vertes, 1982). Fibres within the MFB may be separated into two groups: the dorsolateral group wherein stimulation results in hippocampal synchronisation; and the ventromedial portion along which are conveyed the desynchronising fibres from the MRN. Stimulation of the posterior hypothalamus (PH) also results in the inducement of hippocampal theta rhythm (Bland and Vanderwolf, 1972), and it has been shown that this synchronising effect is reliant on the intact functioning of the MS/DBB

(Smythe et al., 1991, Oddie et al., 1996).

The connection from the SUM to the MS/DBB is of particular interest, as it is thought that the SUM plays a role in determining the frequency of hippocampal theta rhythm and of the rhythmic bursting of MS/DBB cells (Kirk and

McNaughton, 1991, 1993). Lesions of the SUM, however, fail to eradicate hippocampal theta completely (Thinschmidt et al., 1995, McNaughton et al.,

1995), suggesting that other areas can also serve the function of transducing reticular activation into the appropriate frequency.

Inputs to the MS/DBB come largely from the medial part of the SUM (Vertes,

1988). The SUM also displays an extensive connection to all parts of the hippocampus, although this arises almost exclusively from the lateral areas of ______Page 38 the SUM (Wyss et al., 1979, Haglund et al. 1984, Vertes, 1992). Examination of this projection using PHAL tracers failed to reveal any postsynaptic targets which were positive for a variety of stains known to mark inhibitory neurons, suggesting that the SUM projection to the hippocampus targets pyramidal cells exclusively.

The MS/DBB also receives afferent connections from the hippocampus and subiculum (Alonso and Kohler, 1982). These connections arise from

GABAergic cells, as demonstrated by lesion studies combined with

PHAL-labelled cell tracing and immunoreactive staining for GABA (Shinoda et al., 1987, Toth and Freund, 1992, Toth et al., 1993), and they synapse onto inhibitory cells of the MS/DBB almost exclusively (Leranth and Frotscher,

1989, Toth et al., 1993). The axons from hippocampal cells project back to the

MS/DBB in a topographic manner, as those from the septal pole project to the dorsal and medial part of the medial septum and the anterior and dorsal parts of the diagonal band of Broca bilaterally, whereas those from the temporal pole project to the lateral and ventral parts of the medial septum and caudal and ventral areas of the vertical limb of the diagonal band of Broca. Thus, the topography of these afferents mirrors that of the efferent connections from the

MS/DBB to the hippocampus (Amaral and Kurz, 1985).

Efferent connections from the MS/DBB

The majority of efferent fibres from the MS/DBB project to the hippocampus, and these will be discussed in detail in conjunction with the review of hippocampal anatomy. ______Page 39

Apart from the hippocampus and entorhinal cortex, the MS/DBB projects to a range of other areas, although these projections are sparse (Naumann et ah,

1996). They include the lateral septum, the medial and lateral habenular nuclei and the anteromedial and paratenial nuclei of the thalamus (via the stria medullaris), the medial preoptic area and lateral hypothalamus, the pars posterior of the medial mammillary nucleus and the SUM (via the medial forebrain bundle) and thence on to the superior central nucleus, the ventral tegmental area and the dorsal raphe nuclei (Meibach and Segal, 1977, Swanson and Cohen, 1979, Naumann et al., 1996). There is no evidence, however, of collateral projections from single cells to both the cortex and brain stem (Semba et al., 1989). ______Page 40

C. The classification of ceils of the medial septum/diagonal band of Broca

Functional categorisation

Since the first discovery of rhythmic cells in the MS/DBB by Petsche and his co-workers (Petsche et ah, 1965), the heterogeneity of the cells in this area has been well documented. This heterogeneity covers the neurotransmitter released by the cells, their projection pattern and the patterns of firing behaviour. Naturally, most of the descriptions of MS/DBB cell activity have focussed on the rhythmically bursting cells that are phase locked to hippocampal theta rhythm, but many studies of MS/DBB functional have noted the presence of other cells. Below I shall outline the various observations made of the functional properties of MS/DBB cell firing, with special emphasis on the description of the different cell types that have been found.

Petsche et al. (1965) studied these cells in awake, curarised animals using extracellular recordings. They described three cell types:

•"Active burst" cells produced rhythmic bursts closely phase locked to hippocampal theta rhythm with an average rate of 800-1000 per minute (3 to 4 spikes per burst).

•"Passive burst" cells displayed a much lower rate of discharge (less than 500 per minute), and showed considerable variation in the phase of hippocampal theta at which each spike fired. Interestingly, these cells developed an improved phase lock upon stimulation of the reticular formation. The authors also noted that "they may occasionally be drawn into phase for varying periods of time by surrounding cells which are chiefly discharging in bursts." This ______Page 41 suggests that local interactions with nearby cells played an important role in these cells' behaviour.

•"Facultative burst" cells. These cells fired in what was described as a

"random" manner, with no clear relationship to theta rhythm. They could be induced to transform into active burst cells, however, upon the administration of reticular stimulation or parenteral eserine, indicating that their behaviour was dependent on basal forebrain cholinergic pathways.

This group subsequently (Gogolak et al., 1968) studied the phase relationships of rhythmically bursting MS/DBB cells in curarised rabbits. They found that summation of the phase histograms of multiple cells recorded in a single experiment (so that the theta phase reference was constant) demonstrated that the preferential phase of these cells tended to be similar.

Macadar et al. (1970) also examined these cells using extracellular recordings in awake, immobilised rats. They described only one type of septal cell, with the rhythmic bursting behaviour attributed to active burst cells by Petsche et al.

(1965).

Apostol and Creutzfeldt (1974) performed experiments on urethane anaesthetised rats and performed both intra- and extracellular recordings. They separated the cells they found into three groups:

Type A: These fired tonically, and the firing frequency appeared to vary in a

manner correlated with the "arousal rhythm" (which appears to be a

reference to hippocampal rhythmical activity). ______Page 42

Type B: These cells fired in an irregular pattern, and had no phase relationship

with theta rhythm. The authors noted, however, that discharges of these

cells appeared to be related to small increases in the EEG amplitude.

Type C: These cells fired in bursts which showed a constant phase relationship

with hippocampal theta rhythm. The authors noted that, while the phase

of any one particular cell remained constant, the phase relationship

differed between cells. They also stated that intracellular recordings of

these cells showed regular EPSPs which occurred at a constant theta

phase and acted as a trigger for burst firing, although the excitatory nature

of these membrane potential oscillations was not verified.

J.B. Ranch noted several types of cell in the medial septum of awake, unrestrained rats (Ranch, 1976). Most prominent were the rhythmically bursting "theta cells" which he divided into two groups: "continuous", composed of cells which fired rhythmically even in the absence of hippocampal theta rhythm; and "selective", consisting of cells which fired regular bursts only in the presence of hippocampal theta, and even then would lapse bach into irregularity at times (no behavioural specificity could be found for this variation in rhythmicity). A small number of cells showed theta characteristics along with "approach-orient" features, i.e. the cells tended to fire when the animal engaged in approaching or orienting actions, or the cell's firing became synchronised with movement during those actions. The final category,

"tight group" cells, did not fire rhythmically and did not have a specific behavioural correlate, although relations with grooming, drinking and "automatic sniffing" were described. ______Page 43

Wilson and his colleagues (Wilson et al., 1976), studied septal cells in immobilised cats. They separated MS/DBB cells into three simple groups: rhythmically bursting (39%), non-rhythmically bursting (24%) and non-bursting (37%). They noted that the rhythmically bursting group was not homogenous, comprising cells whose rhythmicity ranged from continual bursting even in the absence of hippocampal theta to cells which would fire phase locked bursts for a few cycles, then become desynchronised for a period of time before reestablishing a phase lock. Unfortunately, this variation was not systematically examined.

Assaf and Miller (1978) studied the effects of raphe stimulation on MS/DBB cell activity in urethane-anaesthetised rats. They classified the cells they discovered simply into rhythmically bursting (B-neurones) and irregular cells

(I-neurones). The I-neurones were inhibited by single pulse stimulation of the median raphe nucleus, whereas the B-neurones displayed no response.

Repetitive stimulation of the median raphe resulted in a loss of B-neurone rhythmicity, however, coupled with desynchronisation of the hippocampal

EEG. This effect could be prevented by pre-treating the animal with p-chlorophenylalanine in a dose sufficient to cause 60-80% depletion of serotonin levels in the forebrain, indicating that MS/DBB rhythmicity may be gated by serotonergic systems.

Gaztelu and Buno (1982) separated MS/DBB cells into three major categories, based on their observations from extracellular recordings in awake, immobilised animals. ______Page 44

Type 1: All type 1 cells demonstrated a rhythmic pattern in the

autocorrelogram (AC). Cells in this group differed in terms of their

burst-firing properties, though, and the group is further subdivided:

la: These cells fired long (1/4 to 1/2 of a theta cycle) bursts with high

intraburst rates (the mean firing rate within a burst was 140Hz). These

cells were most numerous (56%)

lb: The bursts in these cells were much shorter, and occasionally

consisted only of a single spike. Their rhythmicity, and their phase

relation with hippocampal theta rhythm, was more variable than for type

la cells, as demonstrated by a wider secondary peak in their inter-spike

interval (ISI) histograms and a lower degree of rhythmicity in the

cross-correlation of unit activity with hippocampal EEG. These cells

were infrequent (11%).

Ic: It was often difficult to identify visually the bursts from cells in this

group, as they tended to display rate modulation instead. While the AC

histogram showed clear rhythmicity, this was not as deeply modulated as

in the previous two groups. Furthermore, the secondary peak in the ISI

histogram denoting inter-burst frequency was occasionally lost

completely in the falling phase of the primary peak. These cells were

also found infrequently (6%).

Type 2: This group constituted 19% of the total, and was composed of cells

which did not manifest any rhythmic activity in their AC histograms,

but showed a clear phase relationship with the hippocampal EEG.

These units fired with either irregular bursts or with irregular single

spikes. Although such an observation appears surprising at first, such

cells have been found in several different experiments. The mechanisms

by which this arises are discussed further in chapter four. ______Page 45

Type 3: These cells (7% of total) showed no relationship with theta rhythm at

all, manifesting a "random” firing pattern without bursts, an irregular AC

histogram and no rhythmicity in the cross-correlation with hippocampal

EEG.

The distribution of the mean phases of the phase-related cells (types 1 and 2) did not shown any clear preference, although there was a weak clustering around the positive and negative poles of the hippocampal EEG wave.

The authors also stimulated at various sites (dorsal fornix, supramammillary nucleus, mesencephalic reticular formation and periaqueductal grey). They found that this stimulation resulted in a phase reset of both hippocampal theta rhythm and type la cells. Type 2 cells also showed a response to stimulation taking the form of either excitation or inhibition, but were not induced to fire periodic bursts, whereas type 3 cells showed no reaction at all.

Lamour and Dutar and their co-workers (Lamour et al., 1984; Dutar et al., 1986,

Dutar et al., 1987) classified MS/DBB cells projecting to the hippocampus of anaesthetised rats simply into two groups: bursting and non-bursting. The bursting cells comprised roughly half the population of cells recorded (45%,

43.4% and 48% in each of the respective reports). Sixteen percent of the cells classified as bursting actually spent some time in non-bursting behaviour as well, switching between the two behaviour types either spontaneously or following the iontophoretic application of drugs. They showed that bursting cells displayed a significantly higher mean rate of spontaneous activity and that the latency with which these cells could be driven by antidromic stimulation from the fimbria was significantly shorter than in non-bursting cells. They calculated mean conduction velocities of 3.6m/sec for bursting cells compared ______Page 46 to 2.9m/sec for non-bursting cells. Both cell groups were sensitive to the application of cholinergic drugs, with equal numbers from both groups showing excitatory responses to acetylcholine and carbachol (a small number of cells also showed inhibitory or biphasic responses).

Alonso et al. (1987) studied MS/DBB cells under urethane anaesthesia and classified their results in a manner based on that used by Gaztelu and Buno.

They also examined the relationship between these cells' firing rates and the presence of hippocampal theta rhythm. In only one case did they find a cell

(which belonged to type lb) which retained rhythmic activity in the absence of a rhythmic hippocampal EEG. In all other cases, the cells' rhythmicity and firing rate decayed with a decrease in the amplitude of hippocampal theta.

Brazhnik and Vinogradova (1988) studied awake restrained rats and examined the effects of pharmacological manipulations on MS/DBB cell activity. They divided the cells they recorded into four groups:

Type I: These "hyperstable" cells displayed continuous bursting activity at

theta frequencies. They were not affected by the administration of either

physostigmine or scopolamine. (4 cells)

Type II: These cells displayed pronounced, periodically modulated bursts at

theta frequency. (8 cells)

Type III: In these cells, a weak and intermittent theta-related rate modulation

was detected. (30 cells)

Type IV: These cells displayed no relationship with theta at all. (14 cells)

Physostigmine had no effect on some of the type II and III cells, but in others, its adminstration caused them to fire bursts in a much more stable manner.

Furthermore, they developed resistance to entrainment by rhythmic ______Page 47 stimulation of the medial forebrain bundle (MFB), with complete blockade in

66% of the cells studied. In the rest, MFB stimulation could sometimes cause a slight phase shift in the cells' bursts if the stimulation occurred close to the time of the cells' spontaneous discharge.

The administration of scopolamine had effects which were largely opposite to those produced by physostigmine. It caused a significant decrease in the presence, regularity and frequency of theta-related bursts in some of the type II and III cells, and enhanced the degree to which the cells would follow the frequency of MFB stimulation. It also eliminated "breakthrough" responses, in which a cell would tend to fire at its own spontaneous frequency as well as at the higher frequency imposed by MFB stimulation.

These results lead us to the conclusion that cholinergic systems in the MS/DBB play an important role in regulating the interaction between rhythmic processes intrinsic to the cells of the area (due either to properties of the cells themselves or to properties of the local synaptic networks in which the cells are connected) and those impinging on these cells from other areas.

Stewart and Fox (1989c) described three types of cells in the MS/DBB. Apart from rhythmically bursting cells, which formed 80% of the population, they recorded 15 "clock" cells (9.4%), which fired single spikes with a high rate. The only evidence of rhythmicity in these cells was the presence of a significant theta phase preference in three of these cells, although the depth of theta modulation (as expressed by the length of the mean resultant phase vector) was lower than seen in the rhythmically bursting group (0.16+/-0.14 compared to

0.43+/-0.16 respectively). They also described a group of 17 "irregular" cells

(10.6%), which fired at markedly lower rates than the other cells (8.1+/-6.5 Hz as opposed to 38.2+/-12.2Hz for "clock" cells and 26.1+/-14.9Hz for ______Page 48 rhythmically bursting cells) and whose spikes appeared to occur in a

"pseudo-random" fashion. Yet ten of these irregular cells displayed a significant phase lock with hippocampal EEG, and displayed relatively high levels of theta modulation (a mean resultant length of 0.31+/-0.15). The authors compared their results to those of Ranck (1976) and concluded that their "clock" cells were analogous to the "constant-firing" cells described by Ranck, as were Ranck's

"tight-group" and their "irregular" cell types. They also noted that, of 33 rhythmic cells tested in this manner, 11 were sensitive to the effects of systemically administered atropine, which caused them to lose their rhythmicity. Both atropine-sensitive and atropine-resistant cells could be driven antidromically by stimulation of the fimbria, indicating that both types projected to the hippocampus. They also noted that there was no systematic variation in terms of the theta phase at which the cells preferentially fired and this parameter was evenly distributed across all the cells which showed a phase lock.

Ford et al. (1989), working in urethane anaesthetised rats, studied the relation of rhythmic cells to various states of theta rhythm and divided the cells into two main groups, "theta-on" consisting of cells whose firing rate increased in the presence of theta rhythm and "theta-off" consisting of cells whose firing rate decreased with the onset of theta rhythm. The cells were further subdivided according to whether they were rhythmic or not (phasic or tonic firing) and their reaction to changes in theta frequency (linear cells showed an increase in firing rate with an increase in theta frequency, while nonlinear cells did not).

Theta-on cells were more numerous than theta-off cells (60 to 28) and phasic cells more common than tonic ones (55 to 33). Phasic theta-on cells were roughly evenly distributed between linear and nonlinear subtypes (16 and 21 ______Page 49 cells each, respectively). Also of note was the presence of 9 cells that maintained their rhythmicity even during large irregular activity (LIA) in the hippocampus, though this activity was less regular than that found in the presence of theta rhythm.

Barrenechea et al. (1995) performed intracellular recordings in animals that were anaesthetised with urethane, and described three types of cells:

Type A: These cells fired in rhythmic bursts which were phase locked to

hippocampal theta rhythm. Spontaneous rhythmic intracellular

oscillations were present, and these persisted even in the absence of

hippocampal theta. (8.9% of cells)

Type B: These cells were irregular in the absence of hippocampal theta

rhythm, but showed rhythmic membrane potential oscillations in its

presence. They tended to fire briefer bursts than seen in type A cells (2

to 4 spikes) and were also marked by the presence of a large but brief

postspike afterhyperpolarisation (amplitude 8.4+/-0.6mV, duration 4.8

+/-0.6ms). (17.9% of cells)

Type C: These cells were most numerous (73.2%) and produced slow spikes

with an afterdepolarisation. Some of them displayed intracellular

theta-frequency oscillations in the presence of hippocampal theta

rhythm.

Stimulation of the reticularis pontis oralis nucleus resulted in increased rhythmical activity in both type A and B cells, but not in those belonging to type C. Stimulation of the fornix produced an EPSP followed by an IPSP in all the type B and C cells so studied , but no synaptic response at all in type A cells. ______Page 50

While the bulk of the observations of MS/DBB cell activity have been focussed on the rhythmically bursting cells, it is clear that non-bursting cells make up a significant fraction of the total population. Moreover, four independent reports

(Petsche et ah, 1965; Wilson et al., 1976; Dutar et al., 1986; Brazhnik and

Vinogradova, 1988) showed cells which spontaneously changed mode from bursting to non-bursting independent of the state of hippocampal theta rhythm, suggesting that there exist different sub-populations of MS/DBB cells which are sensitive to differing modulatory effects. In chapter nine, I will present some evidence from this current study which shows that changes in the degree of rhythmicity of some MS/DBB cells can be correlated with an aspect of the animal’s behaviour.

Anatomical and Electrophysiological Categorisation

At around the same time that the rhythmic nature of MS/DBB cells was demonstrated, Shute and Lewis and their co-workers (1963, 1966) reported the presence of large numbers of cholinergic cells in the MS/DBB and a projection to the hippocampus along three paths: the fimbria, fornix and supracallosal striae. The detection of increased acetylcholinesterase (AChE) activity in the fimbria anterior to a transection confirmed that the MS/DBB projection contained a large cholinergic component (Lewis and Shute, 1967, Lewis et al.,

1967). This cholinergic projection has been confirmed using other techniques, including septal stimulation combined with an assay of hippocampal ACh

(Duder, 1975), selective cholinergic labelling (Bagnoli et al., 1981), and a combination of retrograde transport and AChE labelling (Mesulam et al., 1983a;

1983b). ______Page 51

As noted by Mesulam et al. (1977), there was evidence that the septo-hippocampal projection was not entirely cholinergic, however. The presence of non-cholinergic projection cells was confirmed by two independent double-labelling studies. One (Baisden et al., 1984) used retrograde transport of horseradish peroxidase (HRP) from the hippocampus and discovered that fewer than 50% of HRP-labelled MS/DBB cells also stained positive for AChE. The other study (Wainer et al., 1985) combined retrograde transport of an HRP-wheatgerm agglutinin complex (HRP-WGA) with staining for the activity of choline acetyltransferase (ChAT), a more specific cholinergic marker, in the MS/DBB. They also found that less than 50% of the cells were double-labelled, supporting the thesis that the septohippocampal projection contains a large non-cholinergic component.

The non-cholinergic projection neurons were subsequently identified as

GABAergic by a range of techniques. These included double-labelling studies using retrograde transport dyes and immunoreactive stains for glutamic acid decarboxylase (GAD - the enzyme responsible for G ABA synthesis) (Kohler et al., 1984) ; and injections of the anterograde tracer PHAL into the MS/DBB combined with immunological stains for G ABA in the hippocampus (Freund and Antal, 1988; Gulyas et al., 1990). They constitute roughly 30% of the septohippocampal projection (Kohler et al., 1984), and receive synapses from both GABAergic and cholinergic cells (Onténiente et al., 1987; Leranth and

Frotscher, 1989).

The cholinergic and GABAergic cell groups share many morphological characteristics (Brashear et al., 1986, Naumannn et al., 1992), both showing roughly the same size variation (the size of their elliptical cell bodies ranged ______Page 52 from 8-12|xm by 20-2 5^im to large cells of 10-15jim by 25-30p,m), although smaller GAD-positive cells were seen (with spherical cell bodies of around

12 pm). The cells of both groups are multipolar or fusiform and it was not possible to discriminate between the two in terms of their fine structural characteristics. Electron microscopy has supported these findings for

GABAergic cells (Onténiente et al., 1987). Their axons, however, appear to differ, the GABAergic projections being thicker and showing many en passant boutons and baskets of synapses while the cholinergic axons are thin and show varicosities (Nyakas et al., 1987; Gulyas et al., 1990).

The proportion of cholinergic and non-cholinergic projection neurons was also found to vary in a topographic manner. A study in which injections of

HRP-WGA were placed in different parts of the hippocampal formation

(Amaral and Kurz, 1985) and sections of the MS/DBB stained for ChAT immunoreactivity demonstrated that the medial half of the medial septum is predominantly non-cholinergic and projects to the septal, or rostral, pole of the hippocampus, whereas the lateral half of the medial septum contains far higher numbers of ChAT-positive neurons and projects most heavily to the temporal pole of the hippocampal formation. In contrast to the cholinergic neurons, staining for GAD demonstrated that GABAergic cells are located mainly in the lateral areas of the nucleus of the diagonal band (Brashear et al., 1986). Staining for parvalbumin, a protein shown to be a good marker for GABAergic activity

(Freund, 1989), has demonstrated that the midline and medial areas of the medial septum contain 95% of these cells (Kiss et al., 1990) and that they account for 20-47% of the total number of cells in the MS/DBB. ______Page 53

These two cell systems also differ in terms of the target lamina in the hippocampus, and this will be discussed later, in conjunction with the review of hippocampal anatomy.

Neurons expressing other neurotransmitters also project to the hippocampal formation. The most prominent are galanin and N-acetyl-aspartate-glutamate, which comprise roughly 22% and 26% of the projection neurons respectively

(Forloni et al., 1987, Senut et al., 1989, Senut et al., 1990). These two transmitters are also found to be co-localised with ACh in whole or in part

(Melander et al., 1985, Forloni et al., 1987). There is also evidence for septohippocampal projections involving luteinising hormone-releasing hormone, calcitonin gene-related peptide and enkephalin (Senut et al., 1989).

The functional importance of these projections is not clear at present.

The cholinergic and GABAergic cell populations also show certain key electrophysiological differences. The most prominent distinction between the two is the presence of a slow afterhyperpolarisation (sAHP) in cholinergic cells identified by AChE or ChAT staining as opposed to a fast afterhyperpolarisation (fAHP) in the rest of the cells (Griffith and Matthews,

1986; Markram and Segal, 1990; Gorelova and Reiner, 1996). Further study of these afterhyperpolarisations showed that both are mediated by a potassium conductance that is sensitive to cadmium, and are thus presumably activated by calcium at different latencies (Griffith, 1988).

These cells also differ consistently in terms of the length of their action potential, with sAHP cells displaying a markedly longer action potential duration (1.4ms in sAHP cells as opposed to 0.63ms in fAHP cells) with a ______Page 54 cadmium-sensitive component prolonging the falling phase (Griffith, 1988;

Markram and Segal, 1990). A study involving intra- and extracellular recordings from slice preparations has demonstrated that the sAHP of the cholinergic cells may be translated into a characteristic "hump" in the late phase of the impulse recorded in the extracellular milieu (Matthews and Lee, 1991).

Although Griffith (1988) found that both cells displayed similar passive electrical properties and were electrotonically compact (the cells’ electrotonic length ranged from 0.6 to 0.7 in both cases, implying relatively little attenuation of trans-membrane potential with distance), two other studies found distinct differences in mean input resistance (Segal, 1986, Alreja, 1996), and Alreja noted that presumed cholinergic cells with an sAHP had an input resistance of

103±14.3MQ, almost twice the 64±5.6MQ found in fAHP cells.

These two cell groups also show differences in terms of the rectification currents induced by hyperpolarisation. Both fAHP and sAHP cells show a clear inward rectification in response to hyperpolarising current pulses, but the time course and voltage dependency differs. fAHP cells show a slow, depolarising sag response, while sAHP cells demonstrate a very fast rectification which was sensitive to barium (Griffith and Matthews, 1986;

Griffith, 1988; Alreja, 1996). Markram and Segal (1990) also described a transient outward rectifier in sAHP cells that was activated by depolarisation from a hyperpolarised potential. A similar current was also seen in roughly a quarter of the non-cholinergic neurons they tested, though, indicating that this property is not exclusive to cholinergic cells.

The electrophysiological differences between different MS/DBB populations may be reflected in the variation in cell firing properties described earlier. ______Page 55

Griffith (1988) postulated that the sAHP cells recorded in vitro would display slow tonic firing activity in vivo, resulting in a widespread tonic cholinergic activation of the target areas. But application of serotonin to slices of septal tissue results in a loss of the sAHP together with transformation of the cholinergic cell firing characteristics from slow but regular firing to an irregular pattern displaying a high degree of frequency adaptation (Gorelova and Reiner,

1996). Changes in serotonin levels in the MS/DBB have been implicated in the transition of hippocampal EEG from a coherent to a disorganised state and in changes in MS/DBB cells from rhythmically bursting to irregular activity

(Assaf and Miller, 1978; Kasamo et al., 1994; Vertes et al., 1994; Kinney et al.

1996). This evidence suggests that sAHP neurons do in fact fire in a rhythmic, rather than tonic manner, and that this is necessary for the production of hippocampal theta rhythm. Serotonin does have other, inhibitory effects on

MS/DBB cells, however, which may be responsible for its suppressant effect on theta rhythm (Segal, 1986; Alreja, 1996). ______Page 56

D. The medial septum/diagonal band of Broca and spatial memory.

Given the crucial role the MS/DBB plays in the genesis of the hippocampal

theta rhythm and the extensive links between hippocampal function, spatial

navigation and memory, it is unsurprising that MS/DBB lesions should be

related to memory deficits.

The four testing procedures most commonly used are:

1) The radial-arm maze. This paradigm has several variants, and includes

elements such as the "working memory" task (Olton et al., 1979) in which

the animal has to visit each arm only once to collect all the food rewards

at the ends of the arms; and delayed match or non-match to sample tasks

in which the animal is allowed to visit a subset of the arms before a delay

after which it either has to pick the arms not visited (non-match) or pick

the arm previously visited from a presentation of two arms (match).

2) The Morris water maze (Morris, 1984). In this paradigm, the rat is

placed in a tank containing water made opaque with milk powder and has

to find a platform hidden just beneath the surface which allows it to

climb up out of the water. Commonly, the rat is placed into the tank at

one of four positions 90° apart around the edge of the pool in order to

ensure that it finds the platform by navigation rather than stereotyped

movements.

3) The T-maze alternation task. This simple task involves an initial

presentation or "information run" in which the animal finds food in one

arm of a T-maze, followed by a "choice run" in which the animal has to

pick the arm opposite to the one first presented. ______Page 57

4) The match- or non-match-to-position task. This consists of a box

equipped with levers, cue lights and a reward chute. Similar to the

above, the rat presses one cued lever during the sample phase and then

has to select either the same or a different lever when presented with

them both after a delay (which usually involves a task which forces the

rat to the centre of the box in order to avoid using simple position cues).

Many investigations of memory function have involved lesions of the fomix, a fibre tract that carries both a substantial proportion of the septohippocampal projection as well as projections from the hippocampal formation to the mammillary bodies (Allen and Hopkins, 1989) and the anterior thalamus and lateral septum (Swanson et al., 1987), implying that behavioural changes caused by lesions of this tract may reflect deficits in the operation of many different brain areas. Several studies have focused on impairments in spatial memory, though, often including a comparison of the different areas involved.

It was initially shown that lesioning the fornix, a major route for projections from the MS/DBB to the hippocampus, resulted in deficits in the animals' ability to navigate through space. Crush lesions of the fornix caused increased errors on tasks which included a spatial component: comparison of place and cue learning in fornix-lesioned animals demonstrated impaired spatial ability, but an improved score on cue learning (O'Keefe et al., 1975); while fornix lesions resulted in deficits on an “X”-maze task when the task required navigation using a set of external cues which were distributed around the room, rather than simply heading towards a clustered set of cues (O'Keefe and

Conway, 1980). ______Page 58

Lesions of the fimbria and fornix and of the medial septum were separately shown to have an effect on performance on a rewarded alternation task carried out on a raised T-maze, though the fimbria-fornix lesions had a more prolonged effect, perhaps due to recovery processes within the MS/DBB (Rawlins and

Olton, 1982). Suction lesions of the fimbria-fornix and the cingulate cortex also produced deficits on a delayed matching to position task (Dunnett, 1985), and comparison of fornix, anterior thalamic and mammillary body lesions by

Aggleton et al. (1991) demonstrated marked deficits on a delayed non-matching to position lever press task for the first two lesions only, with no effect caused by loss of the mammillary bodies.

Slow recovery was also reported by Crutcher et al. (1983) following electrolytic lesions of the MS/DBB or after ibotenic acid lesions of the MS/DBB or nucleus basalis magnocellularis (NBM) (Hepler et al., 1985). Crutcher and his co-workers were able to relate the extent of the recovery to the amount of damage to the cholinergic innervation of the hippocampus, finding a significant correlation between the amount of hippocampal acetylcholinesterase depletion and the number of errors made on a radial arm maze task 10 to 25 days after the lesion was made. The amount of delay between sample and test also plays a role on the severity of errors made after MS/DBB lesions, as shown by Kesner et al.

(1988), who found that small lesions of the MS/DBB cause significant errors on a spatial match-to-sample test on trials involving all but the most recent sample.

While proactive interference (an effect in which animals forget faster when the stimulus to be remembered is different to that which the animal responded to on a previous trial) is present in animals with MS/DBB lesions. Harper et al.

(1994) showed that its effect was no greater than in the control group and thus this mechanism cannot explain the memory loss. ______Page 59

Pharmacological manipulation of the MS/DBB has been widely used as it allows evaluation of the role of specific neurochemical systems within the MS/DBB.

Administration of drugs acting on opioid receptors has a significant effect on spatial memory, as tested by a 5-hour delayed non-matching to sample

(DNMTS) task in an 8-arm maze repeated in different rooms (Bostock et ah,

1988). Microinjection of the opioid receptor antagonist naloxone resulted in an improvement in performance at moderate, but not high doses. Conversely, administration of fi-endorphin, an opioid agonist, produced a doubling in the number of trials required to reach criterion. Interestingly, the same dose failed to impair working memory performance in a no-delay version of the same task.

Reversible inactivation experiments, in which microinjections of tetracaine in the region of the septum served to inhibit MS/DBB activity temporarily, demonstrated some significant deficits. Firstly, it was found that, while the place-specific firing of CA3 cells was reduced by around 50% and theta cells in

CAl showed markedly decreased activity and rhythmicity, the firing rate and place field specificity of CAl cells was almost completely unimpaired, although these cells did display some minor changes in the rhythmicity of their firing pattern (Mizumori et al., 1989). When rats were tested on an 8-arm maze working memory task, it was found that septal injections of tetracaine caused an increased number of errors. The animals also displayed a deficit in a non-spatial task, suggesting that septal inactivation produces more general behavioural effects as well (Mizumori et al., 1990b).

Anatomical and physiological studies have shown that GABAergic transmission plays an important role in modulation of activity in the MS/DBB (Segal, 1986;

Leranth and Frotscher, 1989). As a result several investigators have tested the ______Page 60

impact caused by alterations in the action of this neurotransmitter within the

MS/DBB, using a range of tests for spatial memory defects.

Microinjections of muscimol, a GABA^ receptor agonist caused a deficit in

spatial navigation in rats tested on the Morris water maze (Brioni et al., 1990),

shown by the lack of a bias towards swimming in the target quadrant on firee-

swim trials and persistently high mean escape latencies (even though both

muscimol and control groups displayed roughly similar mean escape latencies

in the initial trials). This effect was shown to be temporary, as the rats were

capable of performing the task normally when retrained later without muscimol

injections. Localised application of both muscimol (Chrobak et al., 1989) and

bicuculline, a GABAergic receptor antagonist (Chrobak and Napier, 1991) to the

MS/DBB produced increased errors on a DNMTS memory task in the 8-arm

maze. The latter has also been shown to have a dose-dependent effect, in that in

small doses it cause task-impairment at a 4 hour delay, but not after 1 hour,

whereas large doses affected performance over both delay periods (Chrobak and

Napier, 1992).

Agonists of the GABAg receptor also degrade spatial memory. Intraseptal

injections of baclofen caused a similar increase in errors on the radial-arm maze

task to that described above (Stackman and Walsh, 1994).

Differences between GABAergic and cholinergic activity within the MS/DBB were studied on a spatial alternation task using a T-maze (Givens and Olton,

1990). As before, intraseptal injections of tetracaine and muscimol were found

to impair choice accuracy, as was the infusion of scopolamine, a muscarinic

antagonist, into the MS/DBB. Carbachol, however, had no effect on accuracy. ______Page 61

though it did, together with scopolamine, produce an increase in the time

required to complete the task. Concurrent recording of the hippocampal EEG

revealed a significant decrease in the power of hippocampal theta rhythm upon

infusion of tetracaine, muscimol or scopolamine. The extent of this loss showed

a good correlation with the degradation in choice accuracy.

Microinjections of chlordiazepoxide, a benzodiazepine-receptor agonist which

potentiates the effect of G ABA, was shown to produce impairment in the

Morris water maze task only when it was administered to the medial septum

(McNamara and Skelton, 1993a, 1993 b). Although swim speeds were slower

after injection of chlordiazepoxide, spatial learning was measured by the

distance swum before reaching the platform rather than the escape latency. Cue

learning (measured by the distance swum when the escape platform was

visible) was not impaired, and subsequent testing a day later without drug

administration revealed that the effect had been reversed, as the rats showed

normal performance. The effect of chlordiazepoxide could be eliminated by

systemic administration of flumazenil, a benzodiazepine antagonist, although

microinjections of flumazenil localised to the septum were ineffective at

antagonising the effect of a systemic injection of chlordiazepoxide, indicating

that the amnestic effects of this drug are due to actions elsewhere in the brain as w ell.

In contrast to this. Decker et al. (1992), reported that radio frequency or

quisqualate lesions of the MS/DBB which resulted in significant loss of choline

acetyltransferase (ChAT) activity in the hippocampus caused increased errors on a radial maze task, but not in the Morris water maze. They did, however, show that radio frequency lesions increased the error rate in a spatial ______Page 62

discrimination task based on the water maze. In this latter task two identical

visible platforms were present in the pool, but one floated on the water. Thus

the rats had to learn to swim to the other platform in order to escape. It is

difficult to explain why they should have discovered such a distinction

between the radial and water maze tasks. It would seem that the authors

uncovered a functional difference between the stereotyped and highly directed

movements a rat makes on a radial arm maze (in which movement is only

possible in one of two directions along each arm) and the free, unbounded

movements which can be made in the open field of a Morris water maze. Even

so, this does not explain why radio frequency lesions caused an effect on the

spatial discrimination task. Furthermore, Miyamoto et al. (1987) performed a

similar experiment, comparing ibotenic acid lesions of the basal forebrain with

electrolytic lesions of the MS/DBB. While they related that basal forebrain

lesions produced a greater deficit in the water maze than lesions of the

MS/DBB, these latter did significantly increase escape times. Likewise, Hagan

et al. (1988) demonstrated a clear deficit in performance on the Morris water

maze after similar ibotenic acid MS/DBB lesions, whereas lesions of the NBM

failed to produce a deficit in spatial memory.

In summary, it is clear from these results that correct functioning of the

MS/DBB is vital for successful spatial navigation. There is evidence that this

area is involved in some other functions as well, as will be discussed in the next

section, but its importance for spatial tasks has been demonstrated unequivocally. This suggests that study of the firing characteristics of cells in

this area is best performed when the rat is engaged in activity with a spatial

component, which requires recording in an awake freely moving animal. ______Page 63

E. Non-spatial elements of MS/DBB function

Although the MS/DBB has primarily been studied with regard to its influence on memory and spatial navigation, there have been reports of other behavioural associations, although a coherent picture of these has yet to arise.

Lesions of the septal area have long been known to produce hyperactive and aggressive behaviour in rats. It has been well demonstrated, however, that these effects are related to damage to the lateral, rather than medial septum (see Albert and Chew, 1980 for a review); a finding reinforced by anatomical evidence that shows a link between the hypothalamic "attack area" (an area consisting of the intermediate hypothalamic area and the ventrolateral pole of the ventromedial hypothalamic nucleus) and the lateral, but not medial septum (Roeling et al.,

1994).

In what may have been the first chronic recording of MS/DBB cells in rats.

Morales et al. (1971) described the differing activities of MS/DBB cells during different states of alertness. As is the case with the hippocampal EEG

(Vanderwolf, 1969), coherent rhythmic activity was found during the restless awake state with exploratory behaviour and not when the animal remained quiet. In a corresponding fashion, highly rhythmic states were seen when the animal entered REM sleep, but the cells' activity (and the hippocampal EEG) became desynchronised on the change to slow wave sleep.

Sweeney et al. (1992) repeated this experiment and found a very similar pattern of activity, apart from the discovery of rhythmically-bursting cells that still showed rhythmicity during slow wave sleep. Interestingly, they also found that ______Page 64 these cells fired at a far higher frequency in paradoxical, or REM, sleep, and that in this state virtually all (94%) of the cells showed rhythmic firing behaviour, far more than the 64.2% of cells seen to be rhythmic during an awake, aroused state in which hippocampal theta was present.

Noxious peripheral stimuli have been shown to correlate well with the activity of MS/DBB units. Dutar et al. (1985) demonstrated, in urethane- or fluothane-anaesthetised rats, that painful stimuli such as pinching one of the animal's paws or immersing the tail in hot water, could excite (and, in a minority of cases, inhibit) cell activity. These responses could be elicited from application of stimuli to a large part of the body surface, with most units having a receptive field that including the tail and two hindlimbs or the entire body

(together, these comprised 83% of the units tested). Only in one third of the cases did the cell respond by developing rhythmicity, the rest demonstrated simple, tonic reactions. These reports may, in general, be interpreted as evidence for a relationship between general states of arousal and MS/DBB cell activity.

Studies of operant conditioning have suggested that the septal area plays a role in inhibiting responses. This has been observed in experiments using a differential reinforcement at low rates schedule, in which animals with septal lesions failed in a task which required them to wait 20s between lever-pressing responses in order to obtain a food reward (Burkett and Bunnell, 1966). This is controversial, and other experiments suggested that the increased response merely reflects increased thirst (Harvey and Hunt, 1965) or a more general increase in consummatory behaviour (Hothersall et al., 1970). Septal lesions also lead to elimination of the partial reinforcement extinction effect, in which. ______Page 65 in normal animals, partial reinforcement of a response leads to increased time to extinction when the reinforcement is withdrawn (Gray et al., 1972a, Henke,

1974), and this effect has also been seen after stimulation of the septum sufficient to disrupt the generation of theta rhythm (Gray et al, 1972b). Many of the lesion studies unfortunately included the lateral septum as well, making interpretation of the results difficult, although those of Gray et al., (1972a) were generally small and predominantly specific to the medial septum.

A recent investigation has examined MS/DBB activity in response to a non-spatial continuous discrimination task (Givens, 1996). This task presented the rat with a stimulus, either a light or a tone. If the stimulus was identical to that which had been presented previously, the rat had to press the 'match' lever; if different, the rat was required to press a second 'non-match' lever.

Correct responses yielded a reward. MS/DBB cells were found which responded by changes in firing rate to either the stimulus, the behavioural response or the reward. The majority of cells responded by increasing their firing rate, though a few showed inhibition. Eleven out of the eighteen cells that showed a reaction to the rats' behavioural response demonstrated significantly more activity when the rats' choice was accurate, and three of the five cells that responded to the stimulus did so by resetting the phase of theta rhythmic activity. All these effects were inhibited by a systemic dose of ethanol.

It is, at present, difficult to provide a coherent interpretation of these last results. Clearly, they demonstrate that MS/DBB activity is not simply related to the motor system or to the ascending systems mediating general arousal. They do suggest, though, that MS/DBB activity bears a specific relationship to behaviour that requires the use of memory, or is goal-oriented, as well as Page 66 demonstrating that some MS/DBB cells respond to external events by modifying their rhythmicity patterns. ______Page 67

F. Theta rhythm

While it has been demonstrated that theta rhythm is generated in the cingulate cortex (Holsheimer, 1982) and in the entorhinal cortex (Mitchell and Ranck,

1980), the bulk of the investigation into this phenomenon has concentrated on the hippocampus, where theta rhythm appears to have a major functional role.

A regular oscillation was first observed in the hippocampus of rabbits after nocioceptive or auditory stimulation (Jung and Kornmüller, 1938, cited in

Green and Arduini, 1954). The presence of hippocampal theta rhythm was later found to be dependent on the state of wakefulness of the animal and it could be induced by a variety of sensory stimuli, as well as by electrical stimulation at various sites in the brainstem (Green and Arduini, 1954).

First noted in passing by Yoshii and co-workers (1966) in recordings from dogs, recordings from the hippocampus of freely-moving rats demonstrated a clear relationship between theta activity and voluntary motion (Vanderwolf, 1969).

Vanderwolf described three patterns of hippocampal EEG: rhythmical slow activity (RSA) - theta rhythm; large irregular activity (LIA) and small amplitude irregular activity. Vanderwolf went on to make a distinction between

'automatic' activities such as shivering, grooming and chewing, which were accompanied by an irregular hippocampal EEG and voluntary activities such as walking or making head movements, which were associated with RSA. He also noted that RSA was prominent during paradoxical (REM) sleep and hypothesised that this was due to the increased cerebrospinal activity found in this state. ______Page 68

Further investigation of basic correlates of theta rhythm revealed a strong relationship between the core temperature of the animal and theta frequency, while temperature had little effect on the amplitude of theta rhythm (Whishaw and Vanderwolf, 1971). In contradiction to the findings of Komisaruk (1970), they found that, while the frequencies of sniffing and theta rhythm were often similar, there was no coherent phase relationship between the two.

The absence of mobility-related theta rhythm in the cat and the presence of theta rhythm in anaesthetised animals led to doubt concerning the relationship between theta rhythm and voluntary activity. Finally, it was proposed that two theta rhythm systems exist (Kramis et al., 1975, Vanderwolf et al., 1978). Type I theta was induced by voluntary motor activity, had a relatively high frequency

(7-12Hz in rats and rabbits) and was insensitive to systemic atropine, although it could be abolished by lesioning the entorhinal cortex (Vanderwolf and Leung,

1983). Type II theta could be observed at rest and under anaesthesia, was of lower frequency (4-7Hz in rats and rabbits) and could be eliminated by atropine. The exact behavioural significance of type II theta is debatable, and research into this area is hampered by extensive differences between species, as rabbits are reported to show extensive immobility-related theta, while this is uncommon in rats (Bland, 1986). Recently, the validity of this model of hippocampal rhythmical activity has come to be questioned, and it has been proposed that the variations in theta rhythm may be explained by one unified model of septo-hippocampal function (Lee et al., 1994).

The cellular systems responsible for generating the extracellular potential shifts seen in theta rhythm have been the subject of a considerable amount of investigation. It was discovered early on that the CAl area harboured a ______Page 69

significant generator of rhythmical current flows (Green et ah, 1960), and it was

hypothesised that theta rhythm spread out from this dorsal generator in a

spherical manner, allowing theta rhythm to be recorded in nearby tissues.

Further investigation of the phase and amplitude profiles of theta rhythm

within the hippocampus led to a revision of this theory with the proposal of a

second theta generator in the dentate gyrus (Winson, 1974). This was confirmed

in an independent study a year later (Bland et al., 1975). By slowly moving an

electrode down through the dorsal hippocampus (passing through CAl along the

axis of the pyramidal cells, across the hippocampal fissure and through both

blades of the dentate gyrus) and comparing the EEG with a reference electrode,

it was found that the amplitude of hippocampal theta was high at two locations:

the stratum oriens of CAl and the stratum moleculare of the dentate gyrus,

where it was maximal. The transition between these two points followed one of

two profiles: in the first, observed in anaesthetised and freely-moving animals

(Bland et al., 1975; Winson, 1976), theta amplitude diminished greatly just

below the stratum pyramidale of CAl (a "null" point), and it subsequently

underwent a rapid phase reversal; the second profile, found only in

freely-moving rats (Winson, 1974), did not go through a null point and the 180°

phase shift between stratum oriens and the dentate gyrus occurred slowly over

a distance of roughly 400pm through the stratum radiatum. The phase of theta

rhythm remained constant, however, when it was recorded on the surface of the

hippocampus even though the distance between reference and moving

electrode varied by up to 4mm (Bland et al., 1975). Another notable finding

from this study was the almost complete absence of extracellular theta rhythm in the CA3 area. ______Page 70

Further studies of the physiology of these generators led to the suggestion that their interaction was related to the two forms of theta rhythm seen in behavioural studies. Fenestra and Holsheimer (1979) produced a model which fit the profile of theta rhythm seen in anaesthetised animals. They proposed the presence of two coupled and approximately synchronous dipoles located in the dentate molecular layer and the stratum radiatum of CAl. This model only produced a rapid phase shift, though, and could not explain the more gradual shift seen in freely moving animals.

In a series of papers, Leung (1984a, 1984b) demonstrated that the degree of phase shift in the stratum radiatum varied with the animal's behaviour and could be altered by administering anaesthetic or cholinergic drugs. This led to the conclusion that both forms of theta rhythm were present during normal behaviour. The thesis that both types of theta coexist in different behavioural conditions gained support from later evidence that there is an atropine-resistant component of theta rhythm in urethane-anaesthetised rats (Stewart and Fox,

1989b).

Using the findings from his experiments, Leung constructed a model of theta rhythm generation in the freely moving rat (Leung, 1984c). In this, the electrical flows in the region of a CAl pyramidal cell were simulated by constructing a network of passive electrical cables (representing the soma and dendritic tree) together with active conductance changes (representing the synapses). Using this model, Leung found that the particular pattern of theta phase shift within the hippocampus could be reproduced by driving the cell with two rhythmic inputs shifted in phase from each other by about 30 to 90°, ______Page 71

an inhibitory input at the soma and an excitatory input on to the distal

dendrites.

Current source density analysis of electrical activity in the hippocampus

(Buzsaki et al., 1986) demonstrated a source of current in the pyramidal layer of

CAl accompanied by current sinks in the distal apical dendrites and the outer

third of the dentate molecular layer. About half a theta cycle later, a current

sink appeared in CAl stratum oriens, together with a deep sink in the middle

third of the dentate molecular layer. This was interpreted as supporting the

thesis that hippocampal theta rhythm is the result of two rhythmic inputs: one from the medial septum and one from the theta generator in the entorhinal cortex (Mitchell and Ranck, 1980) to the distal dendrites of CAl pyramidal cells. The current source density analysis was repeated using a more sophisticated technique which enabled reconstruction of the DC-coupled current flows in the dorsal hippocampus (Brankack et al., 1993). This revealed a prominent source at the level of the pyramidal cell bodies and the hilus, as well as a number of sinks in the stratum radiatum, the apical dendritic tree of CAl and the stratum moleculare of the dentate gyrus. In this paper it was proposed that, in addition to rhythmic excitatory inputs from the entorhinal cortex, the excitatory connections between cells in the 'tri-synaptic loop' (the dentate gyrus -> CA3 -> CAl pathway) also provide rhythmic inputs, although the principal cells of all these areas show very low firing rates when not induced to fire by behavioural circumstance (Fox and Ranck, 1981).

The primary cellular correlate of theta rhythm in the hippocampus are the theta cells. First described by Macadar et al. (1970), these cells were further investigated by Ranck (1973). They were found to fire single spikes (rather than ______Page 72

the complex spikes produced by pyramidal cells) phase locked to hippocampal

theta rhythm, and doubled their firing rate in the presence of theta rhythm. The definition of theta cells was later modified (Bland et al., 1983) to stipulate that they: 1) always fire single spikes; 2) are silent or fire irregularly during LIA; 3) are consistently rhythmic during both types of theta activity and fire phase locked to the negativity of local theta rhythm; and 4) change their firing frequency relative to the frequency of local theta. While there is evidence that theta cells are interneurons (Fox and Ranck, 1975, 1981), there is still doubt whether some principal cells also fulfil this function, especially in the dentate gyrus (Rose et al., 1983; Buzsaki et al., 1983).

The relationship between the firing of these cells and the phase of theta rhythm is still not entirely clear. While a number of studies found that theta cells discharge on or near the negative peak of local theta rhythm (Sinclair et al., 1982,

Bland et al., 1983, Bland et al., 1984), Fox et al. (1983, 1986) demonstrated that, while this is true for awake, walking rats, CAl theta cells fire on the positive peak of local theta when the animal is anaesthetised with urethane.

Furthermore, they found a small number of dentate granule cells that fired on the positive phase of their local (dentate) theta rhythm and on the negative phase under urethane. A separate experiment (Buzsaki et al., 1983) produced yet another set of results, showing both CAl and dentate gyrus theta cells which fired on the positive peak of local theta and did not change their phase under urethane anaesthesia. Unfortunately, as theta cells are relatively scarce, a full quantitative analysis of the phase variation of these cells has not yet been performed. ______Page 73

First noted by Fox and Ranck (1975) and later by Buzsaki et al. (1983), 'andtheta'

or 'theta-ofF cells were hilly investigated by Colom and his coworkers (Colom

et ah, 1987; Colom and Bland, 1987). They also found a class of cells which

increased their firing rate during theta rhythm, but did not fire in a rhythmic

manner. These were named tonic theta-on cells as opposed to the phasic

theta-on cells which fired in rhythmic bursts during theta. Theta-off cells

could likewise be divided into two groups: the tonic group, whose cells

remained silent during theta and fired irregularly during LIA; and the phasic

group, whose cells did not fire during higher frequencies of theta rhythm, but

began to fire rhythmically as the frequency of theta declined, and whose firing

rate was inversely related to the frequency of theta. The phasic theta-on cells were found to behave in a linear fashion, displaying an increase in firing rate

directly proportional to the frequency of theta. While these experiments were

conducted in urethane-anaesthetised rats, confirmation has been found for the

presence of theta-off cells in freely-moving animals (Mizumori et al., 1990a).

Further investigation of the properties of theta-off cells has suggested that they

are inhibited by septohippocampal afferents. Suppression of the MS/DBB by

procaine (Smythe et al., 1991) induced an increase in the firing rate of these cells which was not blocked by hypothalamic stimulation, which normally induced or enhanced theta rhythm and suppressed theta-off cells. Furthermore, intracellular recordings of theta-off cells demonstrated that they are subject to a hyperpolarising shift during the transition into theta rhythm, whereas theta-on cells showed a depolarising shift (Konopacki et al., 1992b).

Intracellular oscillations were first observed in hippocampal pyramidal cells by

Fujita and Sato (1964). They found a rhythmic membrane potential oscillation ______Page 74

(MPO) in cells from CAl, CA2 and CA4 that was synchronous with the theta

rhythm recorded extracellularly in the pyramidal layer. The hyperpolarised

phase of the intracellular MPO corresponded to the positive phase of the

extracellular rhythm. When they used KCl electrodes instead of the usual

K+-citrate, they found no change in the phase relationship between the intra-

and extracellular oscillations, leading to the conclusion that the MPO was

caused by rhythmic EPSPs.

The issue of whether intracellular MPOs are caused by rhythmic inhibitory or

excitatory potentials lies at the heart of attempts to determine the mechanisms underlying hippocampal theta rhythm, and it has received experimental attention from many different laboratories. Apart from Fujita and Sato (1964),

Nunez et al., (1987) also concluded that the MPOs were due to rhythmic EPSPs as they found that the amplitude of the oscillations increased as the cell was hyperpolarised and decreased with depolarisation.

On the other hand, several experiments have shown that IPSPs provide the driving force behind the intracellular MPOs. Leung and Yim (1986) demonstrated that the amplitude of the oscillations varied in direct correspondence with that of IPSPs generated by stimulation of the alveus.

When sufficient hyperpolarising current was passed to cause the IPSPs to invert, the MPO changed phase by 207° relative to extracellular theta (as measured in the stratum oriens). Soltész and Deschênes (1993) performed a similar experiment, but used ketamine-xylazine as an anaesthetic, rather than urethane. They also found that hyperpolarisation of the cell led to a change of

90-180° in the phase of the MPOs in both CAl and CA3 (although cells in CA3 showed greater variation in their phase/voltage relationships). Administration ______Page 75 of QX-314, a blocker of voltage-gated Na"^ channels, did not have any effect on the MPOs, while injection of chloride ions caused the oscillation phase to lose its dependency on voltage. Finally, the phase of the intracellular MPO was found to invert around the reversal potential for chloride ions in histologically identified basket cells as well as pyramidal cells (Ylinen et ah, 1995b).

An investigation of the changes in trans-membrane impedance during theta rhythm (Fox, 1989) demonstrated that the input resistance of pyramidal cells drops in phase with the negative peak of the intracellular oscillation seen when the cell is slightly depolarised. This provides additional support for the theory that rhythmic inhibition provides the major drive for intracellular MPOs. Fox did mention one example, however, in which he impaled a cell in the stratum radiatum (the location, along with the absence of action potentials or IPSPs suggested that he had impaled the trunk of the axial dendrite) and recorded

MPOs whose amplitude increased on hyperpolarisation and did not reverse on depolarisation, suggesting that they were driven by EPSPs. These oscillations were phase-shifted approximately 90“ to those found in the soma. Although only one such example was recorded, it is interesting that this closely matches the predictions made by Leung's 1984 model.

In general, apart from the two findings by Fujita and Sato (1964) and Nunez et al., (1987), these findings suggest that theta-related MPOs in both cell types are driven by the rhythmic influx of chloride ions through GABA^ synaptic channels. This provides physiological support for part of the Stewart and Fox

(1990) model of theta rhythm generation, in which GABAergic MS/DBB cells inhibit hippocampal interneurons, which in turn provide rhythmic inhibition in the pyramidal cells. ______Page 76

Further support for this model comes from a study of hippocampal cell firing rates during the transition into theta rhythm. It was found (Stewart, 1993) that interneurons are inhibited and pyramidal cells increase their firing rate during the one second period in which the EEG recording shows a change from a non-theta to a theta state. It was inferred from this that the initiation of theta rhythm involves increased activity in MS/DBB GABAergic cells, which causes inhibition of their target hippocampal interneurons and a resultant disinhibition of the pyramidal cells which receive connections from these neurons.

In contrast with the differences found for the firing phase of theta cells (as discussed above), most of the studies of pyramidal cell MPOs concur that the internal oscillation is approximately 180° out of phase with the local external

EEG. This was either measured directly by inserting an electrode into an appropriate position in the CAl layer or by establishing a stable reference electrode in the hippocampal fissure and subsequently comparing it with the

EEG in CAl close to the recorded cell.

While inactivation of the medial septum causes complete loss of hippocampal theta rhythm in otherwise untreated animals (Green and Arduini, 1954;

Mizumori et al., 1990b; Lawson and Bland, 1993), there is evidence that the hippocampus is capable of generating rhythmic activity in isolation from its inputs.This was first demonstrated by Konopacki and co-workers in the rat

(Konopacki et al., 1987a; Bland et al., 1988) and the cat (Konopacki et al., 1992a).

In these experiments, it was shown that bath application of carbachol onto slices of hippocampal tissue induced the ta-like oscillations which could be recorded from the stratum moleculare of the dentate gyrus. These began approximately 10 minutes after application of the carbachol, lasted over an hour and were blocked ______Page 77

by atropine but not D-tubocurarine. While carbachol-induced theta in the slice preparation shares many characteristics with that found in vivo, it does not display the same 180° phase shift between the CAl and dentate generators

(Konopacki et al., 1987b).

This activity was also shown to take place in area CA3 (MacVicar and Tse,

1989), even though CA3 does not produce theta rhythm in intact animals (Bland et al., 1975). Intracellular records from pairs of CA3 neurons after administration of carbachol showed synchronous MPOs with a frequency of

5-1 GHz. These oscillations persisted after complete removal of the dentate gyrus and CAl (to prevent volume conduction from these areas). The oscillations were, however, eradicated by tetrodotoxin and kynurenic acid

(excitatory synaptic antagonists), by inorganic Ca^"^-channel blockers and by raising the extracellular [Ca^+] above 7mM, all of which block the recurrent synaptic circuits within CA3. NMD A, GABA^ or GABAg blockers had no effect, though. It was concluded that these oscillations were mediated by recurrent synaptic excitation in a similar manner to that proposed by Traub and his co-workers (Traub et al., 1987a, 1987b).

In an extension of the slice experiments, Colom et al. (1991) lesioned the septal afferents in a urethane-anaesthetised animal and demonstrated that application of carbachol alone to the hippocampus failed to induce theta rhythm in this condition. The concomitant administration of bicuculline, however, led to a large, fast (15Hz) oscillation which they likened to theta rhythm as it showed a similar phase reversal in the stratum radiatum, although both the amplitude and shape of the recorded waveforms appear significantly different to natural theta oscillations. ______Page 78

These in vitro studies have demonstrated oscillations which are dependent on

excitatory mechanisms. Given the findings in intact animals in which IPSPs are

shown to drive theta rhythm, it is difficult to accept that the carbachol-induced

oscillations seen in slice preparations are an exact analogue of in vivo theta

rhythm. They may, however, be manifestations of some of the rhythmic

processes intrinsic to the hippocampus which provide the current flows

observed in theta rhythm, similar to those postulated by Garcia-Munoz et al.

(1993).

The mechanisms which determine the phase and frequency of theta were

illustrated by a recent experiment which demonstrated the role which

inhibitory interneurons may play in initiating and maintaining theta. Activation

of CAl interneurons by an intracellularly applied current pulse together with

simultaneous recording of pyramidal cells revealed that IPSPs can cause a phase

reset of rhythmic processes intrinsic to the pyramidal cell (Cobb et al., 1995).

Moreover, this phase reset was shown to affect multiple pyramidal cells at the

same time, presumably as a result of the broad arborisation of the interneuron's

axon field. This significant result demonstrates for the first time a cellular

mechanism which may underlie the widespread coherence of hippocampal

theta (Bland et al., 1975, Bullock et al., 1990).

Apart from the low-frequency theta rhythm, the other major form of EEG found

in the hippocampus is LIA (Vanderwolf, 1969). These irregular,

large-amplitude waves last for 50-100ms and are associated with quiet, stationary activities such as sleep, sitting quietly, eating or drinking. They are

often accompanied by a high-frequency (~200Hz) ripple in the EEG recorded

from the CAl pyramidal layer (O'Keefe and Nadel, 1978). The physiology of ______Page 79 these ripples has received increased attention over the past few years (Buzsaki et ah, 1992; Ylinen et al., 1995a), and it has been shown that pyramidal cells, although firing at a low rate, fire phase locked to the negative peaks of the ripples. Interneurons, however, increase their firing rate substantially during the ripples, and can sustain brief bursts at frequencies of 200Hz, leading to the hypothesis that this barrage of firing forms the basis for the fast field oscillations

(Ylinen et al., 1995a). It has been suggested that the lower-frequency

(40-100Hz) gamma oscillation supports information transfer from the entorhinal cortex to the CA3-CA1 region, while the high-frequency 200Hz oscillations are involved in enhancing the impact of the CAl-entorhinal cortex output system

(Chrobak and Buzsaki, 1996), a role which may be important in the consolidation of memory. ______Page 80

G. The hippocampus: structure

The hippocampus is the main efferent target of the MS/DBB and much of this

thesis will be concerned with the impact of these outputs. It is necessary,

therefore to describe in some detail the anatomical and physiological properties

of this area.

Gross anatomy

In rats, the hippocampus composes a substantial portion of the forebrain. It is a sausage-shaped structure which curves down from the dorsal part of the septum to end in the ventral aspect of the temporal lobe (figures 1.1 and 1.2). Running along the anterior aspect is the fimbria, which serves as the major pathway for communication with the contralateral hippocampus and brainstem structures.

The hippocampus proper is composed of two interlocking sheets of cells (figure

1.3): the dentate gyrus, containing granule cells and the Cornu Ammonis (CA) fields, which contain pyramidal cells (Lorente de No, 1934). These sheets contain the principal excitatory cells of the hippocampus packed into strata approximately 4 cells deep in the CA fields (Andersen, 1990), achieving a cell density of approximately 272.4xlOVmm^ within the principal cell layer (Aika et al., 1994). These are further broken down into the CAl and CA3 areas. CA2 represents the border area between CAl and CA3, but is not generally considered to comprise a separate functional unit, while CA4 merges with the area known as the dentate hilus. The subiculum has strong anatomical and functional relationships with the hippocampus and is included in what is known as the hippocampal formation. The CA fields are also divided into the Page 81

Entorhinal C

sub\c«'«"'

Figure 1.3. A sketch of the major regions of the hippocampus, including the principal synaptic connections. Based on the anatomy of a transverse section through the ventral hippocampus (Paxinos and Watson, 1986). ______Page 82 regio superior, the outer limb which contains the CAl cells, and regio inferior, the inner limb comprised of the CA3 cells.

As the hippocampus is dominated by two densely packed sheets of cells, its anatomical structure can easily be described in reference to these layers. The

CAl area is divided into four areas which correspond to parts of the structure of the pyramidal cells. The stratum oriens contains the basal dendrites; the stratum pyramidale is composed of the cell bodies; the stratum radiatum comprises the proximal trunks of the axial dendrites aligned roughly in parallel; and the stratum lacunosum-moleculare contains the distal tufts of the apical dendrites. CA3 is similarly divided into stratum oriens (basal dendrites), stratum pyramidale(cell bodies), stratum lucidum (apical dendrites receiving mossy fibre inputs), and stratum radiatum (the distal apical dendrites), and the dentate gyrus into stratum moleculare (apical dendrites), stratum granulosum

(cell bodies) and the hilus, which contains a mixture of the axon collaterals of the granule cells a variety of different types of interneurons, the most prominent of which are the mossy cells, which receive synaptic input from the dentate granule cells (Amaral, 1978, Ribak et al., 1985).

The structure of the principal cell types also varies between the three major areas of the hippocampus (Brown and Zador, 1990). In the dentate gyrus, the cell bodies are typically small (10pm), as are the dendrites (diameter l-2pm), which branch extensively and are covered in spines. The pyramidal cells of the

CA areas are, as implied by their name, conical, with a diameter of 20-40pm at the base and a height of 40-60pm. They have a thick (5-10pm diameter) apical dendrite which, in CAl, branches at its tip to form the stratum lacunosum- ______Page 83

moleculare. Several dendrites (3-6)im diameter) emerge from the basal aspect of

the cell body to form the stratum oriens.

Apart from the interneurons of the hilus mentioned above, a number of other

types of interneurons have been described, especially in the CAl field. These

include basket cells, whose dendritic trees span all the four layers, yet which

send axon collaterals only to the pyramidal cell layer (Schwartzkroin and

Kunkel, 1985, Sik et al., 1995). The chandelier cells have a dendritic tree

similar in distribution to that of the pyramidal cells, but synapse only on to the

initial segment of these cells (Somogyi et al., 1983, Li et al., 1992). The cell

bodies of the oriens/alveus interneurons lie on the border between the alveus

and the stratum oriens, their dendrites are concentrated in the alveus, with a

few extending to other layers, while their axonal tree ramifies extensively

throughout stratum oriens and pyramidale (Lacaille et al., 1987).

Lacunosum-moleculare interneurons have cell bodies situated in the

corresponding area of CAl, and their dendrites branch throughout stratum

lacunosum-moleculare, stratum radiatum, and stratum oriens, while their axon collaterals branch in stratum lacunosum-moleculare as well as crossing the hippocampal fissure into the stratum moleculare of the dentate gyrus (Lacaille and Schwartzkroin, 1988à). Another type of interneuron has been described which has a cell body and dendritic tree confined to the stratum oriens, but which sends axons to a compact but densely innervated region of the stratum lacunosum-moleculare (Sik et al., 1995).

The number of dentate granule cells in the rat has been estimated at 10*^ (Boss et al., 1985), while the number of pyramidal cells in CAl is roughly 250,000 and in

CA3 160,000 (Boss et al., 1987). Interneurons are far fewer than the principal ______Page 84 excitatory cells, and have been estimated to constitute between 7 and 11% of the total cell population (Woodson et ah, 1989; Aika et al., 1994).

Synaptic connections within the hippocampus

The dentate granule cells give rise to unmyelinated axons which course through the hilus to synapse on the apical dendrites of CA3 pyramidal cells. These axons have become known as "mossy fibres" due to the presence of large presynaptic expansions which appear roughly every 140jim (Brown and Zador,

1990). These are roughly 5)im in diameter and contact the target dendrites from roughly 17 to 190 pm from the cell body (Johnston and Brown, 1983). The large size of these synapses and their proximity to the cell body has excited considerable interest, although they do not function as "detonator" synapses as once thought (Brown and Johnston, 1983).

The Schaffer collaterals are branches of the CA3 axons which run through the

CAl field, making excitatory synapses en passant every 7pm on average (Lorente de No, 1934, Hjorth-Simonsen, 1973; Ishizuka et al., 1990). The projection pattern from CA3 to CAl is arranged in a topographic manner. The CA3 cells located near the dentate gyrus (proximal CA3) project most strongly to the CAl field of the septal portion of the hippocampus, whereas the more distal regions of CA3 (closer to CAl) project most heavily to the CAl field in the temporal part of the hippocampus. Proximal CA3 cells terminated more superficially

(closer to the hippocampal fissure) in the stratum radiatum of CAl, while the terminations of distal cells were deeper in stratum radiatum and extended to the stratum oriens. Examination of the projection's pattern in areas close to the same septo-temporal level as the injection site showed that proximal CA3 cells ______Page 85

tended to innervate the portions of CAl closer to the subiculum whereas distal

CA3 innervates the region of CAl closer to CA2. Finally, it was found in

general that the terminations from any specific part of CA3 spread more

superficially as they course towards the temporal end of the hippocampus, and

spread towards the deeper areas as they progress towards the septal end (Amaral

and W itter, 1989; Ishizuka et ah, 1990).

The interneurons of the hippocampus participate in a rich network of

interactions, which have not yet been fully determined. There is evidence that

lacunosum-moleculare interneurons provide feed-forward inhibition

exclusively, as they are excited by afferents to CAl and inhibit CAl pyramidal

cells (Lacaille and Schwartzkroin, 1988a, 1988b). Other interneurons take part

in feedback inhibition as well as inhibit other interneurons (Schwartzkroin and

Kunkel, 1985, Seress and Ribak, 1985, Lacaille et ah, 1987).

A large component of the inputs to hippocampal cells consists of commissural

fibres from the other hippocampus. These form several groups: fibres from

CA3, fibres from CAl, and fibres from cells of the hilus (Blackstad, 1956;

Laurberg and Sorensen, 1981). The precise mapping of the terminations of these

projections varies between species, but here I shall describe only those in the

rat. The projections from the hilus end in the inner third of the stratum moleculare of the contralateral dentate gyrus, and extend widely across the

septo-temporal extent, although the projection from the temporal pole appears to

be less intense than that from the septal pole. The fibres from CA3 end in the

stratum oriens and stratum radiatum of the contralateral CA3 and CAl, as well

as the portion of the subiculum closest to CAl. Those from CAl, which are ______Page 86 fewer in number than the projections from other areas, synapse on axons in the opposite CAl and subiculum (Van Groen and Wyss, 1988).

Finally, an important source of synaptic input in the CA3 area is from other

CA3 cells. Many of the axon collaterals of CA3 cells produce extensive local ramifications, synapsing onto other cells in the same area (e.g. Ishizuka et al.,

1990). This extensive local recurrent excitation is thought to be responsible for the CA3 area's high susceptibility to epileptic discharge (Johnston and Brown,

1983; Miles and Wong, 1986). Recordings from pairs of CA3 cells have found that 7 out of the 400 cells examined showed monosynaptic excitatory connections (Miles and Wong, 1986). While there is also some recurrent excitation in CAl, this is at least an order of magnitude smaller than that seen in

CA3, reflecting the relative absence of recurrent collaterals (Christian and

Dudeck, 1988). Figure 1.4 summarises the connections within the hippocampal formation.

Afferent connections to the hippocampus

The hippocampus receives inputs from two major sources: the entorhinal cortex

(EC), which is divided into medial and lateral areas, and the MD/DBB.The projections to the hippocampal formation and the entorhinal cortex form the primary output from the MS/DBB (Lewis et al., 1967; Meibach and Siegel,

1977; Swanson and Cowan, 1979; Mesulam et al., 1983a). These fibres pass via the supracallosal striae, dorsal fornix, fimbria and a diffuse ventral pathway

(Shute and Lewis, 1963; Lewis and Shute, 1967; Gage et al., 1983; Milner et al.,

1983), and while the majority project ipsilaterally, 15% of axons from the medial septum and 17% of those from the diagonal band cross to the contralateral ______Page 87 hippocampus (Peterson, 1989). These various routes carry axons from different parts of the MS/DBB: cells of the medial septum and the ventromedial part of the diagonal band project via the fimbria and fornix, while those of the dorsolateral diagonal band project via the supracallosal striae; no cells were found which projected via both routes (Peterson, 1994).

In the hippocampus, terminals from these projections are found in the subiculum, dentate hilus, the infra- and supragranular layers of the dentate gyrus, the whole of the stratum oriens of the Cornu Ammonis and the stratum lacunosum-moleculare of CAl (Meibach and Segal, 1977; Crutcher et al., 1981;

Chandler and Crutcher, 1983; Oka and Yoshida, 1985; Nyakas et al., 1987). The dorsal region of CAl receives input from the vDBB only, whereas both the MS and vDBB project to the ventral part (Nyakas et al., 1987).

The septo-hippocampal projection also displays a number of topographic properties related to the various cell types found in the MS/DBB, as revealed by PHAL labelling (Nyakas et al., 1986). In the dorsal hippocampus, input to the CAl and dentate gyrus fields comes almost exclusively from the vertical limb of the nucleus of the diagonal band of Broca (vDB), whereas these areas in the ventral hippocampus receive afferents from the medial septum as well as the vDB. The thick, type I axons which may include the GABAergic inputs are seen widely spread across the stratum oriens and moleculare of CAl, the dentate hilus and all areas of CA3, while the finer, varicose type II axons form distinctive laminar terminations, being heaviest in the subpyramidal area of

CAl and the supragranular field of the dentate gyrus as well as the middle third of the molecular layer. The medial septum and vDB also differ with respect to the paths taken by their projections, as retrograde dye study has shown that the ______Page 88 dorsolateral portion of the vDB projects to the hippocampus exclusively via the supracallosal striae, whereas the medial septum and the rest of the vDB utilises the fimbria/fornix route (Peterson, 1994).

Afferents from the entorhinal cortex are bundled into what is known as the perforant path (Cajal, 1901). This projection arises from layers II and III of the

EC and has been found to possess a number of topographical properties (Witter,

1989, 1993). First, the lateral part of both the medial and lateral EC projects preferentially to the septal portion of the hippocampus, whereas the medial EC projects to the temporal pole; this may have functional significance in that the lateral areas receive heavy inputs from the perirhinal cortex, whereas afferents to the medial parts arise predominantly from the limbic system and the periamygdaloid cortex.

Cells in layer II send projections almost exclusively to the dentate gyrus,

CA3/CA2 and subicular areas, while fibres from those in layer III all terminate in CAl and the subiculum. Furthermore, the projections to the dentate gyrus are arranged in such a way that cells of the lateral EC innervate the outer third of the stratum moleculare and the stratum lacunosum-moleculare of CAl, while those in the medial EC project to middle third of these layers. The layer III fibres innervating CAl and the subiculum display a more selective termination pattern. The projections from the lateral EC end in the distal area (close to the subiculum) of CAl and the proximal part (close to CAl) of the subiculum, whereas the medial area of the EC projects to proximal CAl and distal subiculum. Furthermore, the lateral portions of both the medial and lateral EC tend to project towards the septal pole, whereas the medial portion project towards the temporal end. Again, it has been suggested (Witter, 1993) that the Page 89

contralateral CAl contralateral CA3 I I I I I 1

CA3 pyramidal cells

mossy CAl fibres pyramidal cell dentate gyrus + subiculum

contralateral hilus Entorhinal cortex

Figure 1.4. A summary^ of the synaptic connections within the hippocampal formation and its links to the entorhinal cortex, subiculum and the contralateral hippocampus. The connections shown in bold comprise the classical 'tri-synaptic loop' (Andersen et ah, 1971) ______Page 90

difference between the diffuse innervation of all areas of CA3 and the dentate

gyrus by layer II and the selective activation of discrete portions of CAl and

the subiculum by layer III reflects differing information processing taking place in these areas.

The entorhinal cortex itself receives a wide array of inputs from the rest of the cortex, which, in the monkey, project in a convergent fashion from a variety of association areas (van Hoesen et ah, 1972; van Hoesen 1982) . It also receives a strong input from the perirhinal cortex (Kohler, 1986, Witter, 1989).

Apart from the input from the MS/DBB, which has been discussed previously, the hippocampus receives direct inputs from a number of brainstem structures.

Fibres from cells of the median raphe pass along the cingulum and fornix to terminate throughout the hippocampus, principally in the stratum lacunosum-moleculare of CAl and CA3 and in the dentate hilus (Segal, 1975,

Moore and Halaris, 1975). The locus coeruleus provides nor-adrenergic projections to the hippocampus. These enter through the retro-splenial cortex and terminate in a pattern similar to that of the serotonergic fibres, but are also found in close approximation to the mossy fibres of the stratum lucidum of CA3.

The supramammillary nucleus (SUM) also innervates the hippocampus (Wyss et al., 1979). These projections arise mostly in the lateral portion of the SUM and pass through the medial forebrain bundle to terminate in the stratum granulosum and stratum moleculare of the dentate gyrus (Vertes, 1992), where they synapse onto non-GABAergic cells (presumably the dentate granule cells)

(Magloczky et al., 1994). ______Page 91

Efferent connections of the hippocampus

•Cortical Efferents

The principal output of the CAl area is to the subiculum, which it innervates in a columnar fashion, in which the axons collaterals of individual CAl cells bifurcate many times in the subiculum and form a “slab” of synapses in a column of that area (Hjorth-Simonsen, 1973; Swanson and Cohen, 1977;

Swanson et al., 1978; Finch and Babb, 1980; Finch et ah, 1983; Tamamaki et ah,

1987; Tamamaki and Nojyo, 1990).

The subiculum (and, to a lesser extent, CAl) sends ipsilateral projections to the cingulate and retrosplenial cortex, the pre, post and parasubiculum, and the entorhinal cortex. In the entorhinal cortex the fibres terminate mostly in the deep layers (IV-VI), and predominantly in the medial entorhinal area (Swanson and Cowan, 1977; Swanson, 1981; Kohler, 1985). The EC acts as relay station, producing efferents which reach many parts of the olfactory, limbic and paralimbic cortex (Lopes da Silva et ah, 1990). In the rat, the EC subsequently projects to areas 20, 35, 36 and 41 in the temporal cortex, while in the monkey fibres from the EC terminate in much of the temporal lobe (Kosel et ah, 1982).

Swanson and Kohler (1986) reported widespread projections to the prefrontal, somatosensory, auditory and motor areas, although a recent observation suggests that these projections arise only from a small part of the EC adjacent to the perirhinal cortex (Witter, 1993). In the cat, however, EC projections are restricted to allocortical and olfactory areas, but it also sends axons to the perirhinal cortex, from which projections extend to widespread areas of the ______Page 92 cortex, particularly the sensory association areas (Witter and Groenewegen,

1986).

The efferent fibres have been shown to arise principally from layer IV, which is the site of termination of projections from CAl and the subiculum (Kosel et al., 1982; Swanson and Kohler, 1986). In contrast, the fibres that make up the perforant path arise from layers II and III (Witter, 1989), suggesting that the hippocampal inputs and outputs are served by different groups of cells, a laminar organisation not dissimilar to that found in the visual cortex (Zeki and

Shipp, 1988).

•Thalamic connections

The pre and postsubiculum have been shown to have reciprocal connections with the anterior and laterodorsal thalamic nuclei, in which afferent fibres terminate preferentially in layers I and III, whereas the efferents arise from layer VI (Shipley and Sorensen, 1975; Swanson and Cowan, 1977; Robertson and Kaitz, 1981; Kaitz and Robertson, 1981). The parasubiculum also projects to the anterior thalamic nuclei (Swanson and Cohen, 1977; Witter and

Groenewegen, 1986). Weak connections have been reported from the dorsal subiculum to the anterior and midline thalamus, but a more robust projection to the midline thalamus arises from the ventral subiculum (Namura et al., 1994) ______Page 93

•Subcortical connections

The subcortical efferents of the hippocampal formation are dominated by the projections to the septum, as outlined below. The GABAergic projections to the

MS/DBB have already been discussed in section B.

The Cornu Ammonis as well as the subiculum gives rise to a large projection to the lateral septum, which proceeds via the fimbria (Swanson and Cowan, 1977).

This projection is ipsilateral from CAl, but the projection from CA3 extends to the contralateral septum as well. These fibres terminate in a topographic manner, in that different areas along the septo-temporal axis of the CA fields innervate different parts of the lateral septum. These lateral septal cells go on to innervate the lateral and anterior hypothalamic regions, including the supramammillary nucleus and lateral mammillary region (Swanson and Cowan,

1979).

The dorsal subiculum and the pre and parasubiculum send projections via the descending column of the fornix to the mammillary bodies. The ventral subiculum on the other hand produces fibres which form the medial corticohypothalamic tract, and terminate in the bed nucleus of the stria terminalis, the nucleus accumbens, the medial and posterior olfactory nuclei, the taenia tecta and the infralimbic area (Swanson and Cowan, 1977). The dorsal and ventral portions of the subiculum are further distinguished by the segregated nature of cells in the dorsal subiculum, most of which project to only one of the three major output areas (lateral septum, mammillary bodies and anterior/midline thalamus), whereas many cells of the ventral subiculum project to more than one area (Namura et al., 1994). ______Page 94

H. The hippocampus; function

Memory

The importance of the hippocampus in memory was first demonstrated by

Scoville and Milner (1957). They described a patient, H.M., who suffered profound anterograde and retrograde memory loss after bilateral removal of the hippocampus and surrounding cortical tissue. The report of this patient's deficits had a widespread impact on the developing science of neuropsychology, as this provided concrete proof of the localisation of mental function, and the performance of H.M. in a variety of cognitive and memory tasks has been thoroughly explored (Milner et al., 1968; Corkin, 1984; Freed et al., 1987; Freed and Corkin, 1988).

The interest in defining the systems and mechanisms involved in memory have also led to the development of a number of animal models which allow the use of accurately targeted lesions. As this thesis is exclusively concerned with septohippocampal function in rats, and the discussion of memory and amnesia in humans involves many issues peculiar to the clinical paradigm, not least of which is the usual presence of brain damage in structures other than the hippocampus, I will restrict this discussion to experiments performed upon animals, and the theories designed to explain these findings.

Over the years, a number of different theories of memory have been put forward, and these various models will serve as a focus for the following discussion, with specific reference to their explanation of hippocampal ______Page 95 function. The theories which will be reviewed are (listed in chronological order):

• Cognitive mapping theory

• Working vs reference memory

• Declarative vs procedural memory

• Contextual memory and related theories

• Temporal buffering

Cognitive mapping theory

This was first proposed by O'Keefe and Nadel (1978), as an outgrowth on work regarding the role of the hippocampus in spatial navigation (O'Keefe and

Dostrovsky, 1971 - this particular feature of hippocampal function will be discussed at length in the following section). At the core of this model lies a distinction between the processing of information using "locale", or map based methods, and that using "taxon", or cue guidance methods.

The concept of a cognitive map was first proposed by Tolman (1948) to explain early experiments which showed that rats trained to find food by following a complex route were capable of using a direct short-cut when it was made available to them. At the time, however, the relationship between such mapping processes and neural function was undetermined, and this was to remain unclear for many years, despite a great deal of investigation into this field in the late 1940s and 1950s (see O'Keefe and Nadel, 1978 for a review).

O'Keefe and Nadel (1978) proposed that the animal constructed a "place hypothesis", in which it formed a representation of its location within a map of ______Page 96 its local space, and that neural processes within the hippocampus are critical for the formation of this map. In this context, a map was defined as a "set of ordered, connected places" (pg 93), in which the place representations were activated either by the simultaneous presence of sensory data from two or more external objects, or through the previous activation of another place representation combined with motor data which represented the path and distance through which the animal had moved. A key aspect of this representation is its flexibility, as it is not tied to the presence of any specific cue, and is capable of surviving intact if some of the external cues are moved or removed, as long as there are sufficient external data for the rat to locate itself within its internal map of the environment. The internal map is built afresh for each new environment, with the rat taking part in active exploration in order to sample the sensory data incident at a sufficient number of locations within that environment.

Another key element of this form of representation is that it is robust over time and is not substantively changed by repeated presentations of the same sensory data. Thus, a rat need only explore a particular location once in order to integrate it into its cognitive map. Further exploration may produce a finer grained map which contains more detail, but the basic configuration of external objects within the map is unchanged.This component of the theory accommodates both one-time learning (which will be discussed later) and the ability to differentiate between identical objects in different locations, or between environments which are similar. As repeated observation of any particular stimulus has no effect on the representation formed when it was initially observed, this representation is left intact when the same stimulus is discovered in a novel situation. ______Page 97

The taxon system, on the other hand, which encompasses all the major forms of neural processing within the rest of the central nervous system, is highly sensitive to repeated presentations of data. O'Keefe and Nadel argued that this system was dependent upon simple mechanisms of short- and long-term potentiation or depression and was responsible for a variety of repetition related phenomena;

• The priming effect, in which a single prior presentation of a stimulus can lower the threshold for perception of that stimulus, or bias the subject towards such a perception.

• The development of habituation, in which repeated presentations of the same stimulus lead to an increase of the threshold for or bias the subject away from perception of that stimulus.

• Over a longer time period, repetitive activations of a particular taxon representation (such as the reinforcement of a certain stimulus) lead to permanent changes in the threshold of that representation, usually lowering it, but occasionally raising it.

This variety of response was explained as being due to the large number of systems involved, each with differing latencies and decay rates, and sensitive to different aspects of the incoming sensory data.

O'Keefe and Nadel went on to propose that, while the taxon system is involved in the formation of habits and the learning of object properties and categories, the locale system was vital to integrate these objects together into a context.

They also stated that the hippocampus was but one memory area among many, but that it a part of the brain crucially involved in the formation of maps and the utilisation of the locale method to process sensory data. Furthermore, they argued that the hippocampus is vital for the long-term storage of the ______Page 98 representations involved in locale processing systems. In their view, ablation of the hippocampus forces the subject to adopt new strategies to solve problems which were based on taxon system processing techniques. While, in certain circumstances, these techniques may be useful in approximating the behaviour possible with an intact hippocampus, they fail to replicate it fully, a failure which can be catastrophic when features which are unique to the locale system are tested.

The one element of the cognitive map theory that has changed since 1978 is the contention that long-term memory is stored in the hippocampus. There is now a substantial body of evidence showing that, over time, memories which once depended on the hippocampus for their activation slowly lose this dependence

(presumably shifting to reside in the cortex), a process known as consolidation

(Zola-Morgan and Squire, 1990). While this was acknowledged (Nadel, 1991), it was pointed out that, in some species, the time span over which this occurs is measured in years, and that it has not yet been determined whether consolidation actually involves the disappearance of memories from the hippocampus.

A substantial amount of experimental evidence has confirmed the major aspects of O'Keefe and Nadel's spatial memory theory, the most striking example being that provided by O'Keefe and Speakman (1987). In this experiment, rats were trained to navigate a cross-maze in a cue-controlled environment. Recordings were made of the place activity of single hippocampal pyramidal cells, and in 33 out of the 55 units this was found to be determined by the configuration of cues within the environment. The behaviour of these cells was then investigated using spatial working memory trials, in which the rat was placed in the ______Page 99

environment and allowed to orient itself using the controlled cues. These cues were then removed and the rat then had to run to the correct arm and collect the reward. In this phase of the trial, hippocampal place cells were found to fire in the same location as they had when the cues were present.

While the cognitive map theory has been very successful at explaining spatial memory processes occurring in the hippocampus, it has become evident that the memory function of the hippocampus is not entirely spatial. Hippocampal lesions were ini tally thought to cause a deficit on non-spatial delayed-matching-to-sample test in monkeys (Zola-Morgan and Squire, 1986) as well as rats (Otto and Eichenbaum, 1992), although more recent work has shown that these deficits can be ascribed to concomitant lesions of the perirhinal and entorhinal cortices (Meunier et al., 1993). They also cause failure to learn negative patterning discrimination, a test in which the animal is trained to respond to the presence of both elements from a pair when each element is presented on its own, but not when they are presented together (Rudy and

Sutherland, 1989; Sutherland and Rudy, 1989). Finally, hippocampal damage causes a loss of long-term memory for olfactory stimulus associations (Bunsey and Eichenbaum, 1995).

There is also evidence from extracellular recordings that hippocampal cells participate in processes which are non-spatial. These include the findings from rabbits which show hippocampal cells which fire when the animal performs an eyeblink in response to a conditioned tone stimulus, but don't show any significant change in their firing rate when the animal is presented with an unpaired tone, or when the animal blinks in response to a puff of air (Berger et al., 1983). There is also a body of data showing that hippocampal cells fire in ______Page 100 association with behaviour in which sensory discrimination is required

(reviewed in Cohen and Eichenbaum, 1993, pg 117).

These findings, taken together with evidence from human studies of amnesia following pure hippocampal cell loss, in which the pattern of memory loss does not show a spatial character (Zola-Morgan et ah, 1986; Kartsounis et ah, 1995), have lead many to seek a more general explanation for the role the hippocampus plays in memory. These will be outlined below.

Working vs Reference memory

Olton and his colleagues (Olton et ah, 1979) have proposed that the hippocampus is vital for the operation of a process they call working memory, which is concerned with short-term information relevant to the current trial only. In contrast, reference memory deals with knowledge that is required over a longer term, is used in many trials within the same experiment, and is not entirely dependent on hippocampal function. In support of this distinction, and of the contention regarding the role played by the hippocampus, the authors presented two experiments.

In the first, rats were trained to retrieve food bait from the ends of four maze arms, each of which was decorated or marked in a unique manner (so that each arm offered different intramaze cues). One food pellet was placed at the end of each arm, and was not replenished, so that the rat was being rewarded to remember which arms it had already visited. After learning the task with the arms in static positions relative to the room, the arms were moved around relative to the room and to each other after every choice made by the rat (during ______Page 101

this period the rat was confined to the central platform and could not observe what was happening). This was designed to force the rat to learn to distinguish

between arms on the basis of their unique intramaze cues, rather than their

spatial position. While normal rats had some difficulty learning this task, they

did manage to reach a criterion of 3.8 out of 4 mean correct choices after about

50 trials. Rats which had their fimbria/fornix lesioned, however, were reported

to perform at chance levels throughout the 60 trials they were given, producing

an overall mean number of 2.8 correct choices out of 4 (pure chance activity would yield 2.5 out of 4 correct responses in this task). Unfortunately, a statistical analysis of these results was not conducted.

The second experiment involved a seventeen-arm maze, in which 8 of the arms were consistently baited once while the other 9 were not. The rat had to learn both which arms were the ones which were baited (different patterns of arm baiting were used, but this remained the same for each rat throughout the experiment), which was classed as a reference memory procedure, and which it had already visited during that particular trial (and thus no longer carried a food

reward), which was regarded as a procedure involving working memory.

Unlesioned rats learned this task well, developing a final rate of 97% correct responses. They chose an arm in the baited set 92% of the time. Following a fornix/fimbria lesion, however, rats completely failed to learn the working memory component of this task, whereas they slowly returned to criterion

levels on the reference memory component, taking about 30 trials to do so. In other words, whilst the rats learnt to dissociate between those arms which were consistently baited and those which were not, they had great difficulty remembering which arms they had already visited among the baited set on any one particular trial. ______Page 102

While this theory provides a useful account of some of the differences between different memory systems, it has been criticised on a number of points (see the commentary and response section of Olton et al., 1979, also summarised in

Cohen and Eichenbaum, 1993). Firstly, it was pointed out that the discrepancy between the working and reference memory tasks noted in the second experiment may simply be a manifestation of the difficulty with which the taxon or guidance system learns such tasks. It manages to acquire data which is repeatedly presented (the arms which are consistently unbaited), but the absence of such repetition in the working memory component causes it to fail through lack of sufficient exposure rather than as a result of a more encompassing deficit. It was also pointed out that the 30 trials required by normal rats to reach criterion in the 4 arm maze experiment is an order of magnitude higher than the 5 or so needed to reach the same criterion when the rat is allowed to use spatial cues, suggesting that working memory is of lower efficiency than memory utilising spatial elements and that, while the hippocampus may be vital for the operation of both, its role in spatial memory is of greater behavioural significance. While the concept of working memory describes activity in the eight-arm maze well, it fails to provide a reason why rats with hippocampal lesions prove incapable of navigating to the hidden platform in the Morris water maze task, even though its position is unchanged across trials and should thus be regarded as stored in reference memory (Morris et ai., 1982). ______Page 103

Procedural vs Declarative memory

The distinction between procedural and declarative memory was first proposed in the early 1980's (Cohen and Squire, 1980; NJ. Cohen's doctoral dissertation,

1981, as cited in Cohen and Eichenbaum, 1993). It rests on the concept that the hippocampus plays a unique role in the storage of memories concerning relations between distinct objects. The hippocampus receives highly processed input from the sensory areas, and participates in the formation of "declarative networks", which store the relationships between these inputs. These can occur at a range of different levels: relationships may be formed on the basis of low-level features such as size, colour, form or texture, or they may involve higher-order phenomena such as spatial or temporal associations and correlations between objects that may be either causal or accidental (Cohen and

Eichenbaum, 1993, chapter 3). The advantage of a scheme with such a wide-ranging functional basis is that it allows "representational flexibility and promiscuity" (pg 62) between the objects stored in the memory system, allowing them to be retrieved by a multitude of different processing systems in many different situations.

A crucial element of this idea is the property of "compositionality", in which the representation used by memory has access to representations of both the individual components of which the object, event or scene is composed, as well as a representation of the overall nature of the object. This is in distinction to configurai association theory (Sutherland and Rudy, 1989), in which the constituents of an object blend into a unified representation. ______Page 104

Procedural memory, on the other hand, does not involve the hippocampus, and

is concerned with memories which remain closely associated with the

processing systems with which they were originally associated. Procedural

memory processes are regarded as being common in areas such as the primary

visual system. Rather than being concerned with the storage of relational data,

these processes' function is the "tuning and biasing" (pg 79), the net effect of

which is to optimise the ability of the system with which the procedural

memory is associated.

Using this definition, it is argued that the notion of declarative memory

successfully accounts for a number of experimental results showing that

hippocampal activity is crucial for memory processes that do not involve any

spatial components, but instead rely on making discriminations between

olfactory cues (Eichenbaum et al., 1988; Cohen and Eichenbaum, 1993; Bunsey

and Eichenbaum, 1995, 1996). Cohen and Eichenbaum (1993) also demonstrate

how this thesis can explain certain peculiarities in the type of memory loss

suffered by H.M. and other amnesics, particularly their ability to learn to solve

the Tower of Hanoi problem without any explicit memory of having done so.

While the theory of declarative memory offers explanations for a wider range of

phenomena associated with hippocampal function than any other theory at

present, it is a matter of concern that here are very few constraints upon the way the putative underlying processes work, or even upon the definition of

"declarative" itself. Moreover, the extent to which declarative memory is

important for the successful completion of a task can vary dependent on subtle

variants of the experimental paradigm. As Cohen and Eichenbaum themselves

state, "different variants of nearly every paradigm can end up in either the

impaired' or the 'spared' category for the very same stimulus materials.... Thus, ______Page 105 the issue of 'replication' from study to study is more complicated than it has appeared." (Cohen and Eichenbaum, 1993, pp235-236). Although they go on to offer guidance on how to resolve the rather important issue of replication, it is worrying that the definitions of declarative memory used are still susceptible to subjective interpretation, rather than providing specific examples of how the theory could be proved or disproved. Although these are important objections, and the difficulties with replication are particularly damaging, the thesis of declarative vs procedural memory and its related variants (Sherry and Schacter,

1987; Squire, 1992) does offer interesting ideas concerning the shape which a future general theory of memory may take.

Configurai Association Theory

Sutherland and Rudy's proposal of configurai association theory (Sutherland and Rudy, 1989), elaborated earlier ideas that had been put forward concerning the role of the hippocampus in discriminating the different meanings cues may adopt in different situations (Hirsch, 1974, 1980; Winocur, 1980).

They posited that animals use two mechanisms to respond to stimuli. The

Simple Association System (SAS) provides the most basic response, in which there are simple links between stimuli and responses which can be either strengthened or weakened. The overall response of an animal to a set of stimuli involved in such relationships is the simple algebraic sum of the strengths of the elements composing that set. Tasks which require discrimination between different stimulus compounds will be settled by allowing behaviour to be regulated by the compound which has the highest stimulus strength. Although there may be common elements between the compounds, as long as the animal ______Page 106

can perceive at least one distinctive element that differentiates between them, it can use the SAS to determine different behaviours for the different stimulus

compounds.

The Configurai Association System (CAS), however, is based on a process in which the various elements of a compound stimulus are joined together into a single representation that uniquely identifies that specific compound. Once constructed, the representation as a whole can be triggered by the presence of a subset of the sensory stimuli which constitute it. Sutherland and Rudy initially identified the hippocampus as being critically involved in the operation of the

CAS.

In order to test this theory, they used a task which could only be solved by performing an exclusive-or on the input stimuli. In this situation, the correct response is to respond only when one of the stimuli is positive, but not when both are. The task requires the ability to form nonlinear associations, and cannot be solved simply by associating the presence of any particular cue with a response. An experiment was constructed in which the animal was rewarded when it responded to a light (L) or to a tone (T), but not to the presence of the light and tone together (LT) (Rudy and Sutherland, 1989). Half of the rats who had been trained to solve this task then received hippocampal lesions caused by multiple injections of kainic acid and colchicine. When subsequently tested, this group showed a significant and drastic impairment in its ability to solve the

LT portion of the test, despite performing as well as the controls on the tests which involved responses to single cues on their own. Furthermore, a group of naive lesioned rats which were newly trained on the task after surgery showed a similar pattern of impairment, again failing when presented with both stimuli ______Page 107

together, but showing themselves capable of learning the simple

stimulus-response associations on their own.

Sutherland and Rudy (1989) also demonstrated that the concept of configurai

associations can explain the processes at work in spatial memory tasks, where

the animal has to remember the specific configurations of cues at different locations in the environment. They also show how this theory can account for a number of other phenomena such as: latent inhibition, in which the rate at which a stimulus becomes conditioned is lowered if the animal has been repeatedly exposed to that stimulus without it being associated with the reinforcement; discrimination reversal, a state in which the animal is taught to associate one stimulus from a pair with a reward and is then required to learn the opposite stimulus-response pattern; and stimulus selection, in which the animal needs to learn not to respond to certain stimuli which are irrelevant.

In a recent paper (Rudy and Sutherland, 1995) the authors re-examined configurai association theory in the light of new evidence that showed that animals with hippocampal lesions were still able to learn configurai associations. They concluded that the intial formulation of the theory was wrong and that the critical locus of configurai processing lay outside the hippocampus (presumably in the neocortex), although they argued that the output of the hippocampus is used to enhance selectively certain cortical configurai units. In the revised model, this decreases the similarity between units representing the co-occurrence of cues and those that represent the individual cues themselves. Hippocampal output also leads to an increase in the rate at which the configurai units acquire their associative strength. ______Page 108

The comparator model

In contrast to the majority of theories of hippocampal function, Gray (1982a)

proposed that the septo-hippocampal system (SHS) plays little direct role in

memory, and that the amnestic effects of hippocampal damage are the result of

loss of the ability to perform stimulus/response analysis.

In this model, the SHS takes part in the “behavioural inhibition system” (BIS), which regulates the animal’s response to threatening or novel stimuli by causing it to generate the behaviours associated with anxiety. At a physiological level, the hippocampus is engaged in comparing actual sensory experiences with those which are expected. If these two don’t match, or if aversive conditions are predicted, the system takes control of the animal’s behaviour away from whatever other regions were responsible for it and causes the animal to perform the relevant actions.

On the basis, principally, of analysis of the anatomical circuits of the limbic system. Gray places the comparator in the subiculum. This is involved in two important processing loops: one that passes from the entorhinal cortex, through the various regions of the hippocampus and arrives at in the subiculum from

CAl; the other leaves the subiculum to pass to the mammillary bodies, anterior thalamus and cingulate cortex, which in turn project back to the subiculum.

It was theorised that the subiculum receives information of the actual sensory data directly from the entorhinal cortex, together with a prediction of these data

from the anterior thalamus and cingulate gyrus. The information passing from the entorhinal cortex through the hippocampus is used to determine whether Page 109 the detailed sensory data arriving directly from the entorhinal cortex is of importance or not. CAl therefore finally transmits a nonspecific “enabling” signal to the subiculum.

While the SHS does not store any memories itself, it can influence the storage of memories elsewhere in terms of its stimulus/response analysis, particularly via the projections to the temporal lobe. Theta rhythm, in this model, is used as a clocking device, maintaining the correct relationship of sensory data and predictions as they move through the various circuits that compose the system.

Although many of the assertions made by this model are quite radical in relation to other theories of hippocampal function. Gray points out that it is possible to reconcile many of the competing viewpoints. The central function of the hippocampal system as a comparator or hypothesis checker is distinctly similar to the role it performs in Olton’s working memory model (Olton et al., 1979). In its ability to compare multiple pieces of significant sensory data, this model also shares some elements with cognitive mapping theory, and it is interesting that both place and head direction cells may play a predictive function (Muller and

Kubie, 1989; Blair and Sharp, 1995), although Gray firmly located predictive functions in the cingulate cortex and anterior thalamus, rather than the hippocampus itself. The comparator model, together with other work on the pathophysiology of anxiety and schizophrenia, has recently been developed into a more general theory of conscious (Gray, 1995). ______Page 110

Temporal continuity

Rawlins (1985) proposed that the hippocampus acts as a temporal buffer for memory events. This theory was developed partly in response to the working theory hypothesis of Olton et al. (1979) discussed above. Although working memory is defined as being concerned with events that are varied within the same trial, as opposed to those that remain constant across trials, Rawlins suggested that the crucial function damaged by hippocampal lesions is the ability to integrate into memory events separated by more than a certain period of time.

In his model, memory is split into two primary systems, one for short-term and one for long-term requirements. Neither is located in the hippocampus.

Instead, the hippocampus acts as a temporal buffer or “high capacity intermediate-term memory store”, retaining events so that associations can be made between them when appropriate. One corollary of this is that hippocampal damage, rather than producing qualitative impairments (as predicted by most competing theories of hippocampal function), will result in quantifiable deficits as the delay between task-significant events increases.

In support of this, Rawlins described a number of experimental results which showed that hippocampal lesions had an effect on task performance regardless of whether the task was classified as using working or reference memory. Operant conditioning tasks that employ differential reinforcement dependent on the response rate (reinforcement requires there to be a pause between responses) are accepted as an example of a working memory task, and rats with hippocampal lesions generally perform poorly at such tasks as they over-respond. Rawlins ______Page 111 cited an experiment in which the delay between responses was incrementally increased (Boitano et ah, 1980). In this paradigm, the delay between responses was successively increased from 2s to 20s. At the 2s stage, all of the hippocampectomised rats could complete the task to the 50% criterion level, but this ability deteriorated markedly as the time interval increased. At 8s, 8 out of 11 rats reached criterion, but only 1 rat managed to do so at an interval of

14s. It was argued that this provided evidence that hippocampal damage impaired the animal’s ability to maintain memories at longer time intervals and that this effect was not dependent on the nature of the task.

Rawlins also presented an example of a reference memory task which showed impairment in rats with a hippocampal lesion (Rawlins et al., 1985).

Hippocampectomised and control rats were placed in a Y-maze in which one arm was always reinforced, whereas the other was only reinforced 25% of the time. All rats learned to choose the continually-reinforced (CR) arm at first.

Subsequently, a 10s delay was imposed between arrival in the CR arm and the administration of the reward. Subsequently, the hippocampectomised rats began to lose their preference for the CR arm, and by the seventh day their choice of arm had reached choice levels. This was taken to indicate that, as with the previous example, it was the time delay itself which caused the change in behaviour, rather than the type of task performed.

Although this model of hippocampal function is greatly dissimilar to most of the other major theories, Rawlins argued that many of their aspects can be explained in terms of the requirement for a temporal buffer. For instance, it was argued that spatial memory involves integration of multiple stimuli over time as the ______Page 112 animal explores. Similarly, the formation of contextual or configurai associations requires storage of sensory data over short periods of time.

While the temporal buffer theory provides a basis which can explain the major role that time plays in memory loss, it does not, on its own, explain the full range of deficits found after hippocampal damage. As Cohen and Eichenbaum pointed out (1993), there are several spatial tasks in which the impairment or maintenance of performance cannot be specifically related to the temporal relationships between cues and responses.

These last two theories of hippocampal function, the hippocampus as a comparator and the hippocampus as a temporal buffer, have complementary features. Gray and Rawlins (1986) have discussed the ways in which the two theories are mutually compatible, theorising that the SHS is required for two types of task: firstly for the processing of ‘emotional’ inputs relevant to the BIS, and secondly for the intermediate-term storage of information during any tasks which require such a bridging structure. These two functions were ascribed in the first case to the ascending monoaminergic inputs and the dentate-CA3 system, and in the second to the circulation of prediction information around the Papez circuit (Papez, 1937).

In conclusion, while many theories of hippocampal function exist, each capable of explaining a set of experimental findings, none is able to provide a satisfactory explanation for all the data available. While some theories offer highly structured frameworks which cover much, but not all, aspects of hippocampal function (such as spatial mapping theory), others (such as the theory of declarative memory) purposely seek to include as broad a range of phenomena ______Page 113 as possible, but suffer as a result from the problem of defining exactly the phenomena that they cover.

Spatial navigation

The place cell correlate of hippocampal principal cells is now well established

(O'Keefe and Dostrovsky, 1971, O'Keefe and Nadel, 1978, O'Keefe, 1979;

McNaughton et al., 1983; Muller et al., 1987; O'Keefe and Speakman, 1987).

These cells are located in CAl, CA3, the dentate gyrus and the medial entorhinal cortex (O'Keefe and Dostrovsky, 1971;Mitchell and Ranck, 1977;

Olton et al., 1978). The spatial specificity of place cells found in the medial entorhinal cortex is typically less clearly defined than that of cells in the CA fields, as well as exhibiting a poorer signal to noise ratio, and cells in CAl are, in turn, less spatially specific than those of CA3 (Barnes et al., 1990; Quirk et al.,

1992).

Place fields tend to adopt a circular or elliptical shape, in which the maximal firing rate (10-40Hz) occurs at the centre of the field and the rate progressively decreases moving outwards from this point (Muller et al., 1987). Outside the place field, most cells are functionally silent (a few spikes may be fired at long and random intervals). These place fields have been shown to be stable over a number of months (Best and Thompson, 1989; Thompson and Best, 1990).

Several experiments have demonstrated that the position of place fields can, in controlled circumstances, be linked to the positions of certain cues within the environment (O'Keefe and Conway, 1978, Muller and Kubie, 1987, O'Keefe and Speakman, 1987). Alteration of gross features (size, shape) of the environment can have a number of effects. It was found that simply scaling the ______Page 114 size of the environment led to place fields within that environment scaling as well, but that changing the shape of the environment from a cylinder to a square or vice-versa was associated either with the absence of a place field in one of the apparati (the most common occurrence), or the formation of a new and unrelated place field in the other environment (Muller and Kubie, 1987).

Recent examination of the determinants of place field location has shown that different place cells in the same animal may base their place field response on more than one spatial frame of reference, with a few cells using two reference frames in the same environment (Gothard et al., 1996). Systematic analysis has revealed that the edges of the environment (or the walls of the box in which the animal is confined) play an important role (O'Keefe and Burgess, 1996). It was demonstrated that the firing-rate map of a place cell can be modelled as the two-dimensional sum of a number of Gaussian tuning curves which are tied either to the absolute distance of the animal from a wall, or to the relative position of the animal between two opposite walls. In this model, each dimension can be controlled by multiple tuning curves, which may have different determinants, leading to the possibility that scaling the environment along one dimension may lead to the place field becoming stretched along that axis alone, or even breaking into two separate fields. The authors presented one cell which did indeed demonstrate such a behaviour.

Although it was initially suggested (Ranck, 1973; O'Keefe, 1976) that the behavioural significance of certain objects, especially food rewards, could have an effect on the cell's firing pattern, this has been shown not to have an impact on the place cell response. Specifically, a group of experiments have demonstrated that the position of the goal in a behavioural task does not appear ______Page 115 to have an effect (O'Keefe and Conway, 1978; O'Keefe, 1979; Speakman and

O'Keefe, 1990), although an alternative experiment did show that selective baiting of a particular portion of environment could change the firing map of a place cell (Breese et al., 1989). It is possible that this latter experiment was demonstrating the sensitivity of place cells to cue salience (Biegler and Morris,

1993).

It has been well established, though, that, in certain environments, the direction in which the animal is moving can play a major role in determining the firing rate of place cells: some cells fire equally whatever direction the rat moves through their place field, some show increased firing when the rat moves through the field in a preferred direction, and some fire only when the rat is in the place field and moving in correct direction (O'Keefe, 1976; McNaughton et al., 1983; O'Keefe and Recce, 1993). When the animal is placed in a open-field environment, however, the place fields that form are not directional (Muller et al., 1987; Bostock et al., 1988; Muller et al., 1994). The precise reason for this difference is still not understood, although it has been suggested that it is related to the nature of the task the animal is required to perform: those which involve a large amount of linear movement between places lead to the development of directional place fields, whereas movement patterns which cover all directions more-or-less equally lead to nondirectional place fields (Markus et al., 1995).

The relationship between the firing-rate map and the animal's movement over time has also been investigated (Muller and Kubie, 1989). The quality of a firing-rate map was assessed using a number of parameters: firing area, the number of pixel locations at which the cell's firing rate was greater than zero; patchiness, the number of connected areas which all have a firing rate within ______Page 116 the same range; and spatial coherence, a first-order spatial correlation which estimates the degree to which the firing rate at any location is determined by that at neighbouring locations. Using these three measures, Muller and Kubie found that all three measures demonstrated a place field of optimal quality if the spikes were shifted an average of 119ms forward in time relative to the position of the rat. This was taken to indicate that these place fields were best interpreted as predicting the position of the rat in the near future. Different cells showed different amounts of shift, though, and it was proposed that these different shifts represented different possible routes through the environment allowing the hippocampus to calculate simultaneously different solutions to the navigational task with which it was presented.

As initially stated (O'Keefe, 1976, O'Keefe and Nadel, 1978), the position of place fields could be determined by both sensory cues and internal estimates derived from monitoring the direction and distance the rat had moved ("dead reckoning" or path integration). The balance between these two has been the subject of frequent investigation, but it still not fully resolved. It has been suggested that sensory data commands a preeminent place in determining place-cell firing and that the activation of a place cell depends on the presence of a specific configuration of sensory inputs, which were learned during exploration (McNaughton et ah, 1989). There is, however, substantial evidence that individual cues within the environment can be changed without affecting the animal's overall navigation ability (O'Keefe and Speakman, 1987; Markus et ah, 1994). At present, investigation of the relative contributions of sensory and internal cues is complicated by experimental difficulties in separating the two elements (Knierim et ah, 1995). ______Page 117

I. Conclusion

In this introduction, I have presented the evidence for the importance that the timing of a cell’s action potentials plays in passing information to it target neurons. In the hippocampus in particular, it has been shown that the phase of theta rhythm against which each spike fires may convey more detailed information about the position of the animal within the cell’s place field.

The function of the hippocampus is still a matter of much debate, with many different theoretical interpretations of the results of lesions and cellular recordings under different behavioural circumstances. The presence of a temporal code has only been shown in the place cell paradigm, however, which is why this thesis is concerned in depth with that particular theory of septohippocampal function. Theta rhythm, likewise, has been shown to have major correlates with the state of movement of the animal (particularly in rats), another fact suggesting that any temporal code based on the theta rhythm will be particularly applicable to the role the hippocampus plays in spatial navigation, and this consideration has guided the focus of this thesis.

Hippocampal theta rhythm is directly dependent on the function of the

MS/DBB. Further understanding of the temporal coding mechanisms used by the hippocampus will depend on developing a model of the way in which theta rhythm is generated and controlled. With this in mind, this thesis has addressed the question of describing the firing behaviours of these cells and the way in which these vary during movements in different speeds or directions which may affect the rhythmic information transmitted from these cells to the hippocampus. ______Page 118

There have been extensive recordings of MS/DBB cell activity in anaesthetised or restrained animals, as reviewed here. Although some of these studies were concerned only with separating MS/DBB cells into rhythmic and non­ rhythmic groups, several demonstrated additional differences, particularly with regard to a detailed analysis of their firing behaviour and the relationship between firing rate and hippocampal theta. As spatial navigation is particularly dependent on the animal’s behaviour in an awake and unrestrained state, it was felt necessary to ascertain whether these distinctions persisted or were modified in this condition.

The rest of this thesis describes a series of experiments in which extracellular recordings were made of single cells in the MS/DBB of freely-moving rats performing a stereotyped running task. The questions to be asked were: whether the firing characteristics of these cells differed from those described in non-moving or anaesthetised animals; whether it was possible to distinguish between the two major cell types on the basis of the extracellular record; and whether the firing properties of the cell, such as burst frequency or rhythmicity, showed any variation with the spatial behaviour of the animal. Chapter Two

Methodology ______Page 120

Electrode Construction

Microelectrodes were constructed from O.OOOSin diameter Rediohm(tm) 800 wire supplied by H.P.Reid (P.O.Box 352440, Palm Coast, Florida 32135-2440,

USA), coated with polyamide insulation. The wire was wound on a loom so that two bundles were constructed, each containing five or six wires. With the wires held under mild tension, the centre of the bundles was lightly coated in cyanoacrylate adhesive and then heated using a hand-held gas torch. This caused the insulation to flow slightly and bundles of wires to adhere together.

The ends of the bundles were then splayed so that the wires were separated.

The ends were heated vigorously over a bunsen burner so that the polyamide insulation was burnt off. This was done with care to ensure that the insulation remained intact between individual wires making up the bundle. After this, the wires were cut off the loom, and each bundle was cut in half, producing four electrode bundles.

The electrode bundles were loaded into a microdrive apparatus as depicted in figure 2.1. Two electrode bundles were loaded, each having four active electrodes (the extra wires added in the construction stage were necessary to allow for kinks or breakages as the wires were wound onto the microdrive's connection posts). This apparatus consists of a stationary part composed of two metal posts joined at both ends. One of the posts bears a piece of tubing carrying a screw-thread. The dimensions of these two tubes are chosen so that the second tube may turn freely on the first without wobbling. Small sections of heatshrink tubing are fitted over the screw-thread and over the second post and heated. The tubing over the screw thread becomes moulded to the thread in the ______Page 121 process and forms a 'nut'. Dental cement is then applied around both these pieces of tubing and built up to form the moveable part of the microdrive. The plug and connection wires are mounted onto this at the back (next to the unthreaded post). The connection wires are brought forward, fastened to the moveable section with dental cement, and stripped of insulation to form posts and a short length of 23g stainless-steel cannula is mounted on the front between these posts using dental cement. A piece of 19g tubing is cut sufficiently long to cover the length of cannula protruding from its cement mounting and is slid over this. The microelectrode bundles are mounted by sliding them through the cannula from the top, leaving a predetermined length protruding from the far side. The bundles were arranged so that they were spaced 3 00-500pm in the vertical dimension and had as small a horizontal separation as possible. Care was taken to ensure that the electrode bundles remained straight. The bare ends of the microelectrode wires protruding from the top of the cannula are wound around the connection posts and then painted with electrically-conductive paint.

The ends of the electrodes are plated with a 0.2M gold solution sufficient to bring the electrode's impedance in saline down below 500kQ. Typically, currents on the order of 50nA for 30 seconds were used. The impedance was checked by immersing the end of the electrode in a bath of saline and using a signal generator producing a lOkHz sine wave to pass current through a IMD resistor in series with the electrode. The voltage across the resistor was measured, then the electrode was removed from the circuit and replaced by a resistor of known value. Different resistors were inserted until a voltage reading was obtained close to that found with the electrode in place. ______Page 122

The microdrive/electrode assembly was then sealed using non-conductive nail varnish. Many thin coats were applied, and the assembly was allowed to dry under a 60 watt bulb between each coat. This served to prevent the inclusion of gas pockets in the drying varnish.

mm "0mm 9 pin connector rTmnTI solder

pressure 3.5mm spring (Mg) 7mm crew turner dig)

heat main screw shrink (Mg)' sleeve électrode dental wire 9.5mm acrylic connecting 3mm wire bearing surfaces collar

shoulder solder guide 2mm for collar joint tube (23g) ^25mm protective sleeve (19g) elecrodes protrude from guide tube

Figure 2.1. The microdrive used to hold recording microelectrodes in the septum, including an expanded view showing the construction of the main elements. The main screw was cut with a precise pitch so that one revolution lowered the electrode 200p.m. In practice, movements were made in increments of l/8th of a turn, or 25pm. Adapted from an original drawing by A. Ainsworth.

EEG electrodes were constructed from 75pm diameter stainless-steel wire coated with PTEE insulation (Advent Research Alaterials Ltd., Hales worth,

Suffolk, England IP 19 8DD). A single loop of wire was twisted (approx 10 turns ______Page 123 per cm) while held under tension with a weight. Cyanoacrylate adhesive was then applied to a section of the wire to ensure they remained together. The wires were separated at one end of the bundle and the insulation was stripped with a sharp knife under a microscope. The other end was cut so that the ends were separated longitudinally by 1mm. The stripped ends were wrapped around gold connection pins, treated with phosphoric acid or soldering flux and soldered using low-melting-point 2% silver solder (R.S. no 561-094). The connection pins were then encased in enough dental cement to provide a mechanical support for further handling. The ends of the electrodes were plated with gold in a similar fashion to the microelectrodes described above.

Typically, impedances well below 50kQ were obtained in this fashion. ______Page 124

Surgery

The animals used were adult male Hooded Lister rats, supplied by OLAC and weighed between 250 and 400g.

The animals were removed from the animal house and placed in individual cages in a holding room close to the operating theatre on the day prior to surgery. After surgery, the animals were kept in this room until they were sacrificed at the end of the experiment. The only other environment they were exposed to was that of the laboratory.

The animals were deprived of food for at least 12 hours prior to surgery, but allowed to drink. They were anaesthetised by placing them in a vetinary anaesthetic enclosure which was fed 3 % isofluorane in a mixture of oxygen and the ambient air. W hen the animal showed signs of deep sleep (loss of whisker movement combined with slow, deep breathing), the chamber was flushed with pure oxygen and the animal removed. The head and two areas either side of the chest were then shaved using electric clippers. During this, the animal might begin to regain consciousness, at which point further anaesthetic consisting of

3% isofluorane (Abbott) in a 33-66% O^/N^O mix was administered by nose mask.

After shaving, further anaesthetic was given by nose mask until a satisfactorily deep level of anaesthesia was obtained (particularly: failure of the animal to respond to pinching of a rear paw). The animal was then fitted with ear bars, the points of which were located in the animal's external auditory meatus ______Page 125 bilaterally, and this assembly was mounted in the stereotaxic frame (Baltimore

Instrument Co., Baltimore, MD, USA).

At this point, EGG electrodes treated with conductive jelly were placed over the shaved areas on either side of the animal's chest. A rectal thermometer was inserted and the animal was given an intra-perineal injection of 10ml of 0.9%

NaCl to provide fluid replacement during the operation. The animal's body was placed on a bed of artificial fur and an incandescent lamp was positioned over the body to provide heat. The EGG electrodes were connected to an oscilloscope, a pulse counter and an alarm. The heart rate was monitored and used as an index of overall level of anaesthesia. Typically, after induction with

3% isofluorane, the percentage was turned down to 2% for the first hour and then gradually reduced during the remainder of the operation. Towards the end, the animals could be maintained in deep sleep with levels as low as 0.5%, and it was important not to administer excessive isofluorane in order to avoid a prolonged recovery phase after the operation.

After ensuring that the animal's head was level, a longitudinal incision was made down the centre through both skin and the scalp aponeurosis. The aponeurosis was bluntly dissected back from the skull and five retraction sutures were inserted through the skin and aponeurosis, one on either side at the front, one on either side at the back, and one at the rear end of the incision.

These were pulled taut and fastened to the surrounding apparatus. Bleeding from superficial vessels was then controlled. When the skull had been exposed, the attitude of the rat's head was adjusted so that the upper surface of the skull was parallel to the base plate of the stereotaxic frame. ______Page 126

Using a straight edge and a bone scribe, a line was drawn in the skull perpendicular to the sagittal suture and passing through the bregma in order to preserve the bregmatic reference point after the holes were drilled. Likewise, lines were inscribed on the skull surface so that a point 4mm posterior to the bregma and 2.5mm lateral to the midline on the left would easily be found later.

Five holes for screws were drilled using a 2.3mm trephine dental drill. Two holes were drilled over the occipital bone and two over the frontal, one on each side in both cases. A further hole was drilled approximately in the middle of the right parietal bone. Slightly larger holes for the electrodes were then drilled, one over the bregma (the implant site for the medial septum) and one over the 4mmAP/2.5mmML point on the left parietal bone (the implant point for the hippocampal EEG electrode). After each hole was drilled, it was covered with a small piece of cotton wool soaked in saline to prevent desiccation. Stainless steel screws, fashioned by cutting 6BA screws in a tightened die so their diameter was 2.6mm, were screwed into the four occipital and frontal holes. One screw, with a ground lead attached, was screwed into the right parietal hole.

Implant into medial septum

The dura mater on the left side of the sagittal sinus was split using the tip of a

30ga needle. This usually resulted in bleeding from tributaries running into the sinus, and bleeding was controlled using haemostatic sponge. This bleeding could, at times, be excessive, but the fluid pre-load given at the start of surgery served to prevent serious consequences. Using a pair of curved, blunt forceps, the sagittal sinus was gently displaced to the right. Provided that the split in the ______Page 127 dura mater was long enough, the sinus could usually be moved sufficiently to grant free access to the midline.

The microdrive assembly was then mounted in a micro-manipulator attached to the stereotaxic frame by means of a acrylic band which slid over the plug cover on the microdrive and was held in place with a small screw. The microdrive was oriented so that its legs would not foul on the skull screws, but would be close enough to them so that a strong cement bridge could be formed later. The microdrive was supported on a piece of thick but flexible wire, which was twisted so that the vertical posts of the drive, which would form the axis along which the electrode would be advanced, were perpendicular to the upper surface of the skull of the rat. The drive assembly was then positioned with the electrode tip over the implantation hole and lowered so that the tip touched the brain surface at the bregma (using the line inscribed earlier on the skull as an antero-posterior(AP) reference). During this, care was taken to ensure that the tip of the electrode did not touch the skull, which might disturb the plating on the tip. A reading was taken of the AP position of the micro-manipulator and the assembly was raised a few millimetres. The assembly was then moved 0.5mm anterior to its prior location and the assembly was again lowered so the electrode tip was touching brain surface. A reading was taken of the micro­ manipulator's vertical position, and the assembly was them lowered so that the electrode tip was advanced to a depth of 4.5mm below brain surface. The movement of the electrode was monitored continuously using an operating microscope to ensure that it advanced without meeting any impediment (which would be manifest by bending of the electrode). This left the electrode tip approximately 1.5mm above the medial septum. ______Page 128

The sleeve around the electrode cannula was then lowered so that its lower surface rested on the brain surface. In this position, the sleeve covered the length of exposed electrode between the bottom of the cannula and the brain surface. The sleeve was also positioned so that it held the sagittal sinus away from the electrode wire. The exposed brain around the sleeve was loosely packed with haemostatic sponge and acrylic dental cement (Simplex Rapid:

Associated Dental Products, Swindon) was applied around the sleeve and the two frontal screws so that the sleeve would be held in position when the cement dried (care was taken to keep the top of the sleeve free of cement so that the cannula would be free to move within the sleeve). At this time, a cement bridge was formed from the feet of the microdrive to the surrounding screws. If the distance between foot and electrode was more than about 5mm a short piece of metal tubing or a pin was used to reinforce the bridge. Once the cement had dried, the microdrive assembly was released from the micro-manipulator arm.

Implant into hippocampus

The hippocampal EEG electrode was attached to the micro-manipulator by means of a crocodile clip grasping the acrylic pedestal holding the connector pins. As before, this was mounted on a short piece of flexible wire which was twisted so that the electrode was perpendicular to the upper surface of the skull. The dura mater over the hippocampal implant site was stripped away using a 30ga needle.

The electrode was positioned over the hole at the point where the lines drawn on the skull at 4mm posterior to bregma and 2,5mm lateral to the midline crossed. The micro-manipulator was lowered so that the tip of the deep ______Page 129 electrode was touching the brain surface and a note was made of the micro- manipulator's vertical position. The electrode was then lowered while being monitored through the microscope to a depth of 3 mm (as the electrodes were spaced 1mm apart vertically, this placed the deep electrode 3mm deep in the dentate gyrus and the superficial electrode 2mm deep in CAl). The hole around the electrode was loosely packed with haemostatic sponge and dental cement was applied to the electrode wires, running over the skull surface to the occipital screws. When the dental cement had dried, the electrode was removed from the micro-manipulator.

The pedestal holding the connection pins for the EEG electrode was then cemented to a suitable place on the skull (the upper surface of which was now covered in dental cement). Dental cement was also used to attach a short length of 3 or 4mm nylon screw, which was used as an attachment point for the headstage, and short lengths of metal tubing were cemented in where needed to protect the electrodes from scratching by the rat.

The edge of the acrylic covering the skull was then tidied up. Sharp points were removed with scissors and sharp edges were blunted by further application of the cement. The surrounding skin was cleaned of blood and any cement which may have landed on it. A small amount of neomycin/bacitracin antibiotic powder (Cicatrin: Wellcome) was then sprinkled under the skin around the edges of the cement platform and the retraction sutures were removed. The skin was arranged so that it lay over the edges of the cement and was pulled moderately taut, with any loose skin being gathered at the front of the incision. This was then sutured with 3/0 silk using one or two 'mattress' sutures to guard against inversion of the skin edges. This arrangement ensured ______Page 130 that the skin was held firmly and smoothly in place around the sides and rear of the cement platform, which both decreases the amount of irritation caused to the rat and ensures that the rat cannot cause serious harm by scratching at the area.

During the operation the animal’s degree of anaesthesia was monitored by observation of the heart rate and the presence of any reactions to pain. The need for anaesthetic agents typically declined as the operation progressed and towards the end the rats’ anaesthesia was usually maintained on a 1 % concentration of isofluorane. Reduction of the dose of anaesthetic in this manner also ensured that recovery times were not unduly long.

After the anaesthetic gases were turned off the rat was removed from the stereotaxic frame. When the rat began to manifest voluntary movements it was given an intramuscular injection of 0.15ml of Temgesic (Reckitt and Coleman), equivalent to 45pg of buprenorphine to provide post-operative pain relief. The rat was then placed on a sheet of nylon mesh suspended above the floor of the recovery cage and given access to water. Nylon mesh was used so that, should the rat develop the stereotyped ingestion behaviour often seen after administration of buprenorphine it would not fill its stomach with sawdust or other material, which can lead to alimentary obstruction.

When the rat showed signs of full recovery (typically after 12-18 hours) it was transferred to the long-term holding cage. Only one rat was kept in each cage in order to prevent the rats damaging each other's head implants. The rats were weighed and inspected daily and fed so that their weight gain was restricted to around lOg per week until they reached 400g when weight gain decreased to a ______Page 131 rate of 2-5g per week. This was generally sufficient to ensure motivation for the behavioural requirements of the experiment. The holding room environment was maintained on a 12-12 light/dark cycle which was scheduled such that the dark phase (when rats’ activity is maximal) occurred during the day. ______Page 132

Recording sessions

After approximately one week, recording began. First, the microelectrodes positioned above the medial septum had to be lowered into the relevant area.

This was performed over the course of a week, moving the electrodes 300-

400pm in 50pm increments over 3-4 hours each day. When the medial septum was reached, movement was made in 25pm increments with a longer time between movements (approximately 30 to 45 minutes). It was often found that cells only appeared if a substantial (12 hours or more) period was left after moving the electrodes. This may have been caused by tissue drag slowing the effective advance of the electrodes when advancing into the deeper brain regions. When the electrodes were still superficial (not far advanced from the

4,5mm implant point), new cells would often become apparent within 5 minutes of moving the electrodes. These cells were in the cortex (based on the calculated position of the electrodes), and they are larger and, possibly, more robust than the small cells in the medial septum.

It was far more difficult to find sufficiently strong unit signals in the medial septum compared to the hippocampal or cortical areas. But, when a set of units was found, they could often be held for appreciable periods of time, and it was not unusual to be able to record from the same unit for five days or more.

Electrical recording apparatus

A nine-pin socket mated with the plug on the microdrive. A bundle of wires approx 2cm long issued from the socket and led to 2 quad-input unity-gain FET operational amplifiers (surface-mount packages were used to cut down on ______Page 133 weight). A separate ground wire led from the ground screw in the rat's skull to the headstage amplifier assembly. The buffered signals then passed along lengths of flexible hearing-aid wire to the main amplifiers, which were situated in a rack close to the recording area (see diagram 2.3 for the layout of the laboratory).

The amplifiers were used in differential mode, so that the signal from an electrode in one bundle was subtracted from a corresponding electrode in the other bundle (the polarity of the subtraction was usually arranged so that the action potentials being observed had an initial positive deflection). As the tips of the electrode bundles were separated by about 300pm, this served to remove common-mode noise while preserving the signals caused by action potentials close to one or other of the electrodes. Thus the eight electrodes produced four signal channels. As each electrode occupied a unique position relative to the source of the action potentials, the size and shape of the resultant waveform would be different on each channel, thus providing a unique signature for each source of action potentials. This allowed separation of multiple units in those circumstances when many sources were active at the same time (Recce and

O'Keefe 1991). Several examples showing the different wave shapes recorded are given in figures 5.2, 5.3 and 6.2.

The main amplifiers band-pass filtered the signal between 800Hz and lOkHz and applied a variable gain, which was typically set at between 30000-45000 in order to produce a signal with a 5V pk-to-pk range. When a recording was to be made, the signals were taken directly from the output of the amplifiers to a 12- bit five-channel multiplexed analog to digital converter (ADC) (Pentland

Systems MPV952) under the control of a microprocessor (Force CPU 20-XB), ______Page 134 all mounted in a VME rack along with hard-disk and serial port controllers. The

ADC was capable of producing a sample every 4ps, thus each of the five channels had an effective sample interval of 20jis, or a sampling rate of 50kHz.

During recording a control program on the host microprocessor monitored the input continuously. When the input level passed a threshold (this was fixed before recording started and was almost always 2.5V) a set of 50 samples (1ms at the 50kHz sampling rate) was recorded to the hard disk, starting ten intervals before the sample that crossed threshold. This provided sufficient data to characterise the action potential waveform.

At the same time, the position of the rat was continuously monitored by means of a small light fixed to the headstage assembly and a video camera placed above the recording area. The output of this camera was fed to a tracking computer

(HVS Image VPl 12) which scanned the video image from the top left corner and encoded the first point brighter than threshold as an X/Y position with 8 bit resolution. Experiments were conducted in semi-darkness to ensure that the light on the rat's head was the brightest object in the image, and the threshold was set accordingly. One position was recorded for each video frame (running at the standard PAL frame rate of 50Hz).

The EEG signal from the rat (referenced to the right parietal ground screw) was also buffered by buffer amplifiers in the headstage before passing along hearing- aid wires to the main amplifier bank, where it was amplified 400 times. The signal was then low-pass filtered at 125Hz with a 50Hz notch filter to remove power-supply noise and amplified further so that it has a lOV peak-to-peak amplitude. This signal was then presented to the fifth channel of the ADC. The ______Page 135 signal was sampled at 250Hz and stored along with the position data (five EEG samples per position) on the hard disk.

At the end of a recording period, the data was downloaded from the VME rack to a network of Sun computers where it was analysed using a custom-built software package. Initially, spurious data caused by noise or chewing artifact were removed, and then the spikes were grouped according to differences in size and shape on each of the four channels. Further analysis was then performed according to the dictates of the experiment and will be described where relevant.

Behavioural tasks

The rat was trained to run along a 4.5cm wide wooden track raised approximately 30cm above floor level. Bait was placed at either end of the track and the rat was trained to run to one end of the track, eat the bait and then turn round, run to the bait at the other end and repeat the process, as demonstrated in figure 2.2. Rats generally learnt this task with little difficulty, and usually only required two or three training sessions of 3 hours each which was done a week after surgery. Figure 2.3 shows an overhead view of the laboratory displaying the track on its turntable and its position relative to the equipment and other environmental cues in the room.

In those experiments where the animal's direction of motion was varied, the linear track was rotated relative to a fixed axis on the floor of the laboratory.

Initially a line drawn on the floor was used for reference. Later, in the majority of the experiments, the track was placed on a turntable which allowed it to be Page 136

suspension points m video camera

data/power cables microdrive and headstage amplifiers

lamp

/ ■ c~a %

track

food reward

Figure 2.2. A side-on view of the rat running on the track showing the disposition of the data collection apparatus. ______Page 137 rotated to predetermined orientations. When recording a full set of 16 directions across 360 degrees, 8 orientations were used at 22,5 degree intervals

(each orientation yielded two motion directions), which were recorded in random sequence. No change was made to the surface of the track between orientations, and thus smell cues on the track surface remained in place.

Each recording lasted two minutes. The rat would be picked up from the holding platform and placed at the centre of the track. When the rat began to run towards one of the two ends, the recording was started. The experimenter stood approximately 0.5m from the middle of the track and his position relative to the track was invariant of the track's orientation. Brief movements were made to

replenish the bait at the ends of the track, otherwise the experimenter stayed still. Typically, two recording sessions would be run consecutively, after which the rat would be returned to the holding platform and allowed to rest.

Virtually all the rats demonstrated a tendency to turn in one direction only, although the preferred direction of turn varied among animals. This caused the connecting wires running from the headstage gradually to become twisted and the rat had to be rotated in the opposite direction on the holding platform after each set of recordings. This was done immediately on returning the rat to the holding platform and the rat was allowed to recover on the platform for several minutes afterwards in order to alleviate any disorientation.

The rat was also allowed to drink at regular periods. A complete set of 16 recordings generally took around one and a half hours, after which the rat was returned to its home cage Page 138

shelves carrying window old equipment

radiator

baited track on turntable

holding box mm •J.WS- m # amplifiers and computers H i door

Fi^re 7. An overhead view of the laboratory. The track on which the rat ran was baited at both ends and mounted on a turntable. This was oriented at random in one of eight directions (alternatives shown in light grey) spaced at 22.5° intervals. The laboratory contained a large number of surrounding environmental cues which were constantly visible. A busy road lies outside the window, which provided a set of auditory cues. ______Page 139

H istology

After it was decided that no further material could be collected from the animal

(usually when the electrode had been lowered to a depth of 7,5-8mm), the rat was left without moving the electrode for a period (typically 1 to 2 weeks) and then culled with an intramuscular injection of 0.7ml/kg body weight of

Lethobarb (Duphar Vetinary, Southampton). When deep anaesthesia was obtained, the animal's abdomen was opened, the diaphragm cut in the midline and the mediastinum exposed by cutting along the left costo-chondral joint lines. A 19g needle was inserted into the left ventricle and held in place with a clamp. The descending aorta was clamped and the inferior vena cava cut to allow exsanguination. The rat was then perfused transcardially with 0.9% saline for 5 minutes and then with a 4% formaldehyde/phosphate buffer solution for

30 minutes.

After perfusion, the head was removed, the skull exposed and cut away with small rongeurs and the brain was dissected free. Care was taken to ensure that the electrodes were disturbed as little as possible in this process and were removed vertically. The brain was then fixed in 4% formaldehyde for a week and slices were mounted to examine for electrode tracks. In all cases, there was sufficient gliosis or local tissue damage around the electrode track for it to be found and charted. ______Page 140

Data Analysis

Data from the recording computer were downloaded onto a network of Sun workstations for analysis using custom-written software. The initial phase consisted of eliminating noise, most commonly caused by myogenic potentials from chewing, grooming and scratching, from the recording.

As the four-channel waveform of each data capture could be viewed on­ screen, noise removal was easily performed by eliminating those data samples whose wave shape did not conform to a biphasic pattern on at least three of the four channels. Most data sets contained waveforms which clearly clustered in this pattern. In a small number of cases and particularly when the spike was small, however, no reliable distinction could be made between an action potential and noise. In these instances there was a continuum of wave shapes varying between those which may be regarded as a spike and those which were clearly noise. Such data sets were not used for further analysis and were discarded.

At the same time as removing noise, the waveforms were grouped into clusters corresponding to different cells. This could be done with ease as the wave shapes of different cells were highly dissimilar. The initial distinction was usually made regard to the size of the waveform on each of the four channels, but cells often differed in terms of the wave's shape as well.

Although the wave's size could vary, especially during a burst, the basic shape of the wave usually remained static under such circumstances. ______Page 141

Multiple cells were recorded simultaneously in approximately half (29/61)

of the recording sessions. The maximum number of cells recorded at one

time was 5. Figure 2.4 shows the interface used to separate cells and an

example of the wave representations corresponding to spikes from four cells

recorded simultaneously. On one occasion, five cells could be separated out

from one set of data, although this was unusual, with recordings of two or

three simultaneous cells being more typical.

The position data returned by the video tracking system was converted into velocity data after having been passed through a boxcar filter to smooth it.

The position data would occasionally be lost at the acquisition stage when

the lamp on the rat's head was occluded by the signal cables and these missing data points were interpolated from the surrounding data. Once smoothed and interpolated, the position data were used to calculate the velocity and direction of motion.

Initial analysis consisted of calculating the cell's autocorrelation histogram

(ACH) and the phase of hippocampal theta rhythm at which each spike fired. The phase was determined by fitting the EEG data from the hippocampus to a sequence of sinusoids. For each cycle, the data set was

compared to a set of half-period inverted sine waves varying in frequency

from 11.1 to 20.3Hz in steps of 0.4Flz. The goodness of fit was measured by

determining the sum of the squares of the differences between the data and

each model. A number of successive start points (corresponding to the maximum acceptable half period) were used. The start position and period "COREL) f^MPwrSpcj .MATCH ) J+VWBAR ) & AND j % l OR -W NClR|^Thresh ) Undo) illllliU b t J Std Summary 4#* Spk phasJ Assist) , : 8 ox car j# . Clear) ‘-^Dsplay ); rVelf vWdens)

DirCut) filter) \ G d at»)’^ f Strip ) ' Input)> Jolnj % l ^ l b ) % "eXcl ) Zoorn ) 4- 0 5-0 >5-0 tt-0 [267]:c(1,0) @(168.128); ^ Loct) ' M agnf) Nxtmap ^ ^O utput) Recr) Spikes ' purstsj ' , Velocs ') '"-DSTRBjT« tr0,pos77.dt39.dst15808. pxy(168,128) ptxy(130.102) 017+018 il- 0 ol-O in tas-0 .0 eem-0 sp«ag-2.5 1-0 2-0 3=0 4-0 5-0 >5-0 tt=0 5[602]: c(1.0) @(224.135); params +J delta 'W4 )% W 5 ) p p k s 8128 cntx 114 r vail •«' tr0.pos173,dt94.dst35524, pxy(224.135) crate 0.00 cnty 132 ptxy(13.225) V 20__ 0 —T - J ^■%4|2 _ il-0 ol=0 int#s-0.0 mem-0 spmag-2.5 1-0 2=0 3=0 bksize mad va 3 4.n s=n >s=n rr.n J Axsamp: W1 Axsamp: W2

Figure 2.4.1 he interface screens of the software program used to analyse the data. The top window carries the controls for the various analysis functions. The bottom left window shows the position associated with each spike in the sub-window on die top left. The other three sub-windows show the maximum minus the minimum voltage of the spike waveform on each of the four channels (the axes are labelled here for clarity). In this case, the data contained four separate cells. The bottom right window shows the first 100 spikes from one of the cells. ______Page 143 of the wave model achieving the best goodness of fit was stored and used to calculate the starting point for the next cycle.

In many cases further analysis was performed. These techniques will be described in the later chapters where relevant. Chapter Three

Histology ______Page 145

Provided that the microelectrodes were constructed to a high enough standard, specifically that the electrodes emerging from the guide tube were parallel to the shaft of the main screw, location of the electrodes in the MS/DBB proved relatively easy. It was found, though, that the major factor affecting the success of a recording was the size of wire and the technique used in constructing the electrode bundles.

Figure 3.1 shows an implant in which 25pm platinum wires were used (in this case, the electrodes were inserted at a 10° angle to avoid the sagittal sinus). The tetrode bundle was constructed by twisting four wires together and then cutting them with scissors. As platinum is a relatively soft metal, it retained the twist well. Although the bundle has clearly passed through the medial septum and into the vertical limb of the diagonal band of Broca, no cells were recorded from this implant. This was also true of all the other MS/DBB implants which used this technique.These electrodes were used initially as they provide good recordings when used in the hippocampus (O’Keefe and Recce, 1993).

From examination of the histology, it is suspected that this was due to two factors: the amount of cell damage caused simply by the physical size of the

(relatively) large electrodes; and to the growth of glial tissue into the interstices of the bundle's spiral (clearly visible in this example), which may have caused additional damage when the electrode was moved. It is possible that such ingrowth could have occurred overnight between recording sessions. This would only affect cells at the level of or above the electrode tips, although it is possible that it caused disruption to the blood supply of the deeper cells. Paçe 146

cerebral cortex

m 0 X corpus callosum

Vf-./. - > lateral Iventricles J - K h iWS electrod àtS'sr.Æ ^s' » é ^ "A'-f f #W#a# track ,'wu % r fk.'Tvsx^W -^‘' medial I® septum

anterior commissure

Figure 3.1. A coronal section showing an early implant into the medial septum. In this case, the tetrode was constructed from four 25pm teflon-insulated platinum wires twisted together. No cells were successfully recorded from this implant, even though it clearly passed through the medial septum and into the vertical limb of the diagonal band of Broca. This may have been due either to the large electrodes used in this case or to ingrowth into the interstices of the electrode spiral, which is clearly visible in this sample and may have caused tearing of the surrounding tissue as the electrode was moved. Calibration bar: 1mm. ______Page 147

Figures 3.2 and 3.3 demonstrate MS/DBB implants using untwisted 12pm wire which yielded successful recordings. As can be seen from figure 3.2, the amount of tissue damage caused by these electrodes was much smaller (the tissue damage seen in figure 3.3 was almost certainly caused during extraction of the electrode, as the tissue fragments curve upwards), furthermore, there is no evidence of tissue ingrowth into the electrode bundle. Figure 3.3 shows three successive coronal sections taken at 80pm which show the electrode path. This usually passed through the medial septum, into the vertical limb of the diagonal band of Broca, and ended in the horizontal limb (recordings were terminated when no further rhythmic cells were found on moving the electrode). Page 148

corpus callosum

tricles

medial septum"

vertical ILy^ diagonal

3^

anterior commissure

Figure 3.2. A demonstration of an electrode track (arrows) passing through the medial septum. In this case (as with all the implants yielding successful recordings) a non­ twisted bundle of 12 pm electrodes was used. The amount of tissue damage is noticeably lower than in the previous example. Animal: r480. Calibration bar: 1mm. Page 149

A- A)'#

lateral ventricles

of Broc medial septum m #

anterior commissure

Figure 3.3. Three successive sections (anterior to posterior) showing the passage of an electrode (arrows) through the medial septum and into the diagonal band of Broca. The configuration of the tissue fragments within the electrode track suggests that the majority of the damage observed was caused during extraction of the electrode. Animal: r484. Calibration bars: 1mm ______Page 150

Two examples of hippocampal implants are shown in figure 3.4. As can be seen,

the tip of the deep electrode was successfully located in the stratum moleculare

or granulosum of the dentate gyrus. As the superficial electrode terminated

1mm above that and was in line along the axis of the penetration, it could be

accepted that the upper electrode tip was placed in the CAl field. This was

confirmed by ensuring that the theta oscillations on the deep electrode were

180° out of phase with those from the superficial one. Page 151

t >.

%

dentater gyrusv

1

Figure 3.4. Two examples from different animals of the position of the EEG electrodes in the hippocampus, showing the deep electrode track (arrows) terminating in the dentate gyrus. The superficial electrode was always 1mm above along the line of the track, and thus positioned in the stratum pyramidale of CAl. This was confirmed by observing a 180° phase difference between the EEG recorded from the two electrodes. All the EEG recordings used the deep electrode. Calibration bars: 1mm Chapter Four

Firing behaviour ______Page 153

Various methods have been used to classify the different types of firing

behaviour displayed by MS/DBB cells, although apart from one example

(Ranch, 1976), they have all been concerned with cells recorded in

anaesthetised or restrained animals. The various analyses of MS/DBB cell

behaviour in anaesthetised animals have identified several differences between the cells on the basis of their raw firing behaviour during theta rhythm.

As discussed in the introduction, however, there are several clear differences

between the theta rhythm present in anaesthetised and that present in awake animals. In the anaesthetised state, theta rhythm has a lower frequency than in

freely-moving animals, displays a different pattern of phase shift on moving the recording electrode down through the laminae of the hippocampus, and is abolished by anti-muscarinic drugs (Winson, 1974; Kramis et al., 1975; Bland et al., 1975; Winson, 1976; Vanderwolf et al, 1978).

Furthermore, there is evidence showing that anaesthetics affect MS/DBB cell firing behaviour. Administration of urethane led to a decrease in the burst frequency of rhythmically bursting cells, and a drop in the number of spikes fired in each burst (Stumpf et al., 1962). Systemic injection of pentobarbital into restrained rabbits caused the frequency of rhythmically bursting cells to lower, and led to the elimination of theta bursts in cells with unstable rhythmicity

(Brazhnik and Vinogradova, 1986). The number of rhythmically bursting cells in the MS/DBB of restrained rats increased significantly over the unanaesthetised state when urethane was used as anaesthetic, however, and the cells were found to exhibit a higher variability of discharge rate in the unanaesthetised condition (Lee et al., 1991). This was confirmed in a later study on restrained rats (Sweeney et al., 1992), which also found that the ______Page 154

percentage of rhythmically bursting cells was far higher during paradoxical

sleep (94%) than during wakefulness with hippocampal theta rhythm (64.2%)

and particularly low during slow wave sleep (8%).

These data indicate that the rhythmic behaviour of MS/DBB cells varies

significantly with the introduction of anaesthesia. As the majority of studies of

the MS/DBB have been conducted in anaesthetised animals, this raises the

question of the extent to which results from these studies can be applied to animals in the awake and freely moving state.

In the introduction, I gave details of the numerous different classification schemes which have been used to distinguish the firing behaviours of different

MS/DBB cells. The scheme which provides the most comprehensive and detailed account is the one first proposed by Gaztelu and Buno (1982) and subsequently revised by Alonso and his co-workers (1987), and this was used to compare results from anaesthetised animals with those obtained in this study in the awake, freely moving rat. This method assesses the cells' firing behaviour on the basis of three criteria: the presence of rhythmicity in the autocorrelation plot; the presence and nature of a secondary peak in the inter-spike interval

(1ST) histogram; and whether the cell fires preferentially at a particular phase of theta rhythm.

Three types were described: type 1 cells are rhythmic and phase-locked to theta rhythm; type 2 cells do not show rhythmicity on their autocorrelation plots, yet are still significantly phase-locked to theta (the presence of a phase- lock was gauged using the Rayleigh test (Mardia, 1972) at the p< 0.001 level) ; type 3 cells are also non-rhythmic, and their firing has no relationship with ______Page 155

theta rhythm. Type 1 is farther subdivided into three groups: type la cells fire

in relatively long bursts and have a small secondary peak in the ISI histogram;

type lb cells fire much shorter bursts, often only firing one spike per theta

cycle, and this is reflected in a more pronounced secondary ISI peak and a

smaller autocorrelation pedestal (the height of the autocorrelation histogram

between the peaks); type Ic cells have a high background firing rate, on which

theta rhythmicity is imposed as a form of rate modulation, leading to a large

autocorrelation pedestal and a broadening of the primary ISI peak which often

totally obscures any secondary peak which might be present.

R esults

Out of implants in 14 rats, 73 cells were recorded. Of these cells, 15 were

recorded with the linear track in a single orientation only, the rest were

recorded over multiple orientations as detailed in chapter two. Aspects of the

behaviour of cells which varied with the direction in which the rat ran will be

discussed in chapter nine. The majority of cells, however, did not display

significant differences with direction and are considered as a group in this and

the following chapters.

It was found that MS/DBB cells in the awake, freely moving state could easily

be separated into groups using the criteria outlined above. Examples of the five

different types of cells are given in figures 4.I-4.5. The numbers of cells of the

various types found and their firing rates are given in Table 4.1. Most of the

groups could be separated with ease as the distinctions between them were

clear. The differences between cells belonging to type la and lb were, in some

cases, less obvious though. As the original definitions of the classification do not Page 156 state rigid numerical criteria for differentiating between the two groups, it was not felt appropriate to introduce such criteria here. In only two cases, however, was there doubt concerning the class to which the cell should be allocated.

The classification of the majority of cells remained stable across the recording period. There were two cases, however, in which a rhythmic (type Ic) cell was recorded on two different days, and the cell’s record was arrhythmic (type 2) on the second day (the cells both maintained the same preferential phase). In both these cases, the firing rate was dramatically lower on the second day (around

2Hz), which would account for the loss of rhythmicity as such a low rate is insufficient to produce an autocorrelogram which reflects rhythmicity at the frequency of theta rhythm. A small number of cells varied their rhythmicity during the course of a recording session. These cells will be described in full in chapter nine.

Type Number of Cells Firing Rate (Hz±S.D.)

la 8 30.3±15.4

lb 15 21.6+22.5

Ic 11 23.7±18.8

2 24 10.2±13.5

3 15 21.8±23.7

Table 4.1 Cells of the MS/DBB classified according to the method of Gaztelu and Buno (1982), together with the average firing rates of the different classes. All Type 1 cells fired spikes phase-locked to theta rhythm and demonstrated rhythmicity in their auto-correlation histograms (ACH). Tvq^e la cells fired in ______Page 157 long bursts (25-50% of a theta cycle) with high intraburst rates. Type lb cells fired much shorter bursts, often consisting of a single spike. Both these cells demonstrated a clear secondary peak in their interspike interval histogram (ISIH) corresponding to the burst frequency. The spikes of Type Ic cells were difficult to separate into bursts, and the secondary ISIH peak was very small or non-existent. Type 2 cells demonstrated a phase lock with theta rhythm, but showed no rhythmicity in their ACH. Type 3 cells showed no relationship with theta rhythm at all.

All type of cells showed large firing rate standard deviations, indicating a poor match between the firing rate and the firing behaviour of the cell. The standard deviation is particularly high for type 3 cells. This can be explained by the substantial skew that existed in the firing rate distributions, which are shown in figure 4.6.

The rhythmic activity of the cells was linked to the animal’s mobility, which was predictable given the correlation between hippocampal theta rhythm and voluntary movement (Vanderwolf, 1969). Figure 4.7 shows representative examples of the changes made to the autocorrelation function of a rhythmically bursting (type lb) cell and a type Ic cell when those spikes fired when the animal was moving faster than 0.05m/sec were removed. In both cases, the rhythmicity of the cell drops dramatically, leaving just a small residue which may be related to movement made when the animal had passed outside the field of view of the overhead camera (in such cases, the position of the rat was interpolated between the points of exit and re-entry into the field of view, and the velocity was zero). ______Page 158

D iscu ssion

Table 4.2 summarises the differences in number of rhythmically bursting cells between the two experiments performed in awake, freely moving rats and those performed in urethane anaesthetised animals in which the quantities of different cells were measured. The number of rhythmically bursting cells is significantly larger under urethane anaesthesia (%- test, p<0.001), confirming the earlier results described above.

Awake, freely moving animals

This study 23/73 (31.5%)

Ranck, 1976 18/35 (51.4%)

Urethane anaesthetised animals

Dutar et al., 1986 136/313 (43.4%)

Ford et al., 1989 55/88 (62.5%)

Stewart and Fox, 1989 128/160 (78%) Table 4.2 A comparison of the numbers of rhythmically bursting MS/DBB neurons recorded under urethane anaesthesia and in awake, freely moving animals. The percentage of rhythmically bursting cells was significantly higher in urethane anaesthetised animals (%^=13.02, d.f.=l, p<0.001)

At first, it would appear anomalous that type 2 cells are phase-locked to theta rhythm yet do not have a rhythmic autocorrelogram. This may be explained, however, by the low sensitivity of the autocorrelogram to rhythmic processes that are highly discontinuous (ie, the cell fires phase-locked to theta, but the Page 159 Type 1 a

W^VVVHW'A

Raw data samples

)o 10000 .. ik

-502 -376 -251 -125 0 125 251 376 502 -502 -376 -251 -125 0 125 251 376 502 m sec m sec Autocorrelation histograms

5000 5000 - 4500 , 4500 . 4000 T 4000 - 3500 . 3500 i 3000 : 3000 2500 . 2500 2000 . 2000 - 500 - 0 0 0 1 10 0 0 - 500 .

100 125 150 175 125 150 175 IlliCl. iiiacv. Inter-spike interval histograms

400 , 350 ,

350 -

300 ;

250 r

0 30 60 90 120 150 180 210 240 270 300 330 0 30 60 90 120 150 180 210 240 270 300 330 Phase of hippocampal dentate theta (degrees) kJIPhase of hippocampal dentate theta (degrees) Firing phase of each spike

Figure 4.1. Two cells belonging to type la from different animals. These cells were phase-locked to hippocampal theta and fired in long, rhythmic bursts which lasted a substantial portion of the theta cycle. As a result, the secondary peak in the inter­ spike interval distribution is small. Calibration bars: 150msec; left, 962pV; right, 329pV Page 160 Type 1 b

I I I I

V\/VVVlAAA/^A^ Raw data samples

4500 - 3000 T 4000 - 2500 ; 3500 . 3000 . 2000 i 2500 1500 - 2000

376 502 502 -376 376 502 Autocorrelation histograms

1000 . 10 0 0 - 900 - 900 ! 800 . 700 . 700 . 600 - 600 - 500 . 400 i 400 - 300 - 300 i

200 - 100 1

Inter-spike interval histograms

140 .

30 60 90 120 150 180 210 240 270 300 330 0 30 60 90 120 150 180 210 240 270 300 330 Phase of hippocampal dentate theta (degrees) Phase of hippocampal dentate theta (degrees) Firing phase of each spike Figure 4.2. Two cells from different animals which were classified as type lb. These cells fire short, rhythmic bursts phase-locked to hippocampal theta rhythm. The short burst/gap ratio causes a pronounced secondary peak in the in ter-spike interval histogram. Calibration bars: 150msec; right, 962pV; left, 379|lV. Page 161 Type 1 c

a III H iiini I

Raw data samples

4000 - 20000 - 18000 - 3500 - 15000 - 14000 2500 12000 2000 10000 8000 6000 4000 2000 0 25 251 376 502 -502 -376 -251 -125 0 125 251 376 502 m sec mse' Autocorrelation histograms

1000 - 4500 • 4000 . 800 . 3500 - 700 - 3000 - 600 : 2500 ^ 500 . 2000 -

300 - 200 .

25 50 75 100 125 150 175 0 25 50 75 100 125 150 175 msec Inter-spike interval histograms

350 350 -

300 ;

30 60 90 120 ISO 180 210 240 270 300 330 30 60 90 120 150 180 210 240 270 300 330 Phase of hippocampal dentate theta (degrees) Phase of hippocampal dentate theta (degrees) Firing phase of each spike

Figure 4.3. Two type Ic ceils from (different animals. These cells are phase-lockecd to hippocampal theta rhythm, but the rhythmicity is manifest as rate mo(dulation rather than rhythmic burst firing. Calibration bars: 150msec; right, 962pV; left, 357pV. ______Page 162 intervals between phase-locked bursts are too long to be visible above the accumulated noise and timing jitter).

Type 2 cells have been described in the hippocampus (Garcia-Austt et al.,

1977; Garcia-Sànchez et al., 1978) and in the MS/DBB (Petsche et al., 1962;

Gaztelu and Buno, 1982, Alonso et al., 1987; Stewart and Fox, 1989), although in far smaller numbers than found in this study. Of the four previous reports of type 2 cells in the MS/DBB, three quantify the number of cells showing this behaviour (Table 4.3). It is noticeable that there are a significantly higher number of type 2 cells in the freely-moving animal {y} test, p<0.001).

Petsche et al. Gaztelu and Stewart and Fox, This study

1962 Buno, 1982 1989

5/113 (4.3%) 18/96 (19%) 13/160 (8.1%) 24/73 (32.9%)

Table 4.3 A summary of the numbers of type 2 cells (non-rhythmic cells which are significantly phase-locked to hippocampal theta rhythm) in urethane- anaesthetised animals (first three columns) and the awake, freely moving rat

(this study: final column). The number of type 2 cells is significantly increased in this study compared to the others (%^=28.06, d.f.=l, p<0.001)

Stewart and Fox (1989) noted that many of the type 2 (“irregular”) cells they recorded could be activated orthodromically by stimulation of the fimbria/fornix. They suggested that the phase-relationship of these cells may be the result of back-projections from hippocampal complex-spike (pyramidal) cells. Although current anatomical evidence, as detailed in the introduction. ______Page 163

indicates that the hippocampo-septal projections arise instead from GABAergic

interneurons, this does not negate the possibility that rhythmic inputs from the

hippocampus are responsible for modulating the activity of these cells.

In chapter six, I will present evidence arising from further analysis of the wave

shape data which suggests that type 2 cells belong to a distinct functional group

of MS/DBB cells. Initially, it is necessary to consider the theta phase

relationships of the different types of cells. Page 164 Type 2

I I I I M i l l I I g III 1 1 1 1 II 1 1 II III II I I 1

Raw data samples

120 - 700 -

600 . I I

300

200 100

-502 -376 -251 -125 0 125 251 376 502 -502 -376 -251 -125 0 125 251 376 502 m sec m sec Autocorrelation histograms

180 - 160 - 140 -

100 . I

m sec m se c Inter-spike interval histograms

0 30 60 90 120 150 180 210 240 270 300 330 30 60 90 120 150 180 210 240 270 300 330 Phase of hippocampal dentate theta (degrees) Phase of hippocampal dentate theta (degrees) Firing phase of each spike Figure 4.4. Two ceils from different animals which were classified as type 2. These cells do not display any notable rhythmicity in their autocorrelation histograms, and do not fire in bursts, as shown by the inter-spike intervals, yet they are significantly phase-locked to hippocampal theta rhythm. Calibration bars: left, 300msec, 694pV; right, 150msec, 379pV. Page 165 Type 3

!w

Raw data samples

180 T 1200 - 160 - 1000 - jj. 140 - 120 100

-502 -376 -251 -125 125 251 376 502 -502 -376 -251 -125 125 251 376 502 Autocorrelation histograms

250 ,

40 i

A ijW 50 75 100 125 150 175 0 25 50 75 100 125 150 175 m se c m se c Inter-spike interval histograms

0 30 60 90 120 150 180 210 240 270 300 330 30 60 90 120 150 180 210 240 270 300 330 Phase of hippocampal dentate theta (degrees) Phase of hippocampal dentate theta (degrees) Firing phase of each spike

Figure 4.5. Two examples from different animals of cells that were classified as type 3. These cells did not display any rhythmicity at all, firing in an essentially random fashion. Calibration bars: 150 msec; left, 313p.V; right, 625pV. Page 166 Type 1 a

3 9 IS 21 27 33 39 45 51 57 63 69 75 81 37 firing rate (Hz) Type 1 b Type 1 c 12 -

9 15 21 27 33 39 45 51 57 63 69 75 81 87 9 15 21 27 33 39 45 51 57 63 69 75 81 87 firing rate (Hz) firing rate (Hz)

Type 2 Type 3

3 9 15 21 27 33 39 45 51 57 63 69 75 81 87 3 9 15 21 27 33 39 45 51I 57 63 69 75 81 87 I firing rate (Hz) firing rate (Hz)

Figure 4.6. The firing rate distributions of cells belonging to the various categories of firing behaviour. These are highly skewed and do not conform to a normal distribution. As the average firing rate could change from day to day, the graphs plot the average rate for each day a cell was recorded. Three of the type 3 cells were recorded very early in the experiment and their firing rate could not be calculated because of data loss. Page 16',

Rhythmically bursting cell at all speeds

-502 -376 -251 -125 0 125 251 376 502 msec 7000 -

6000 . Rhythmically 5000 - bursting cell 4000 - at low speeds3000 . 2000 only

-125 0 125 502 msec

4000 3500 .

2000 Type 1 c cell at all speeds

376 502 msec

800 I Type Ic cell at

low speeds only400

200

0 -502 -376 -251 376 502 msec

Figure 4.7, An example of the dependence of rhythmicity on the speed at which the animal runs. For each pair of autocorrelation histograms, the top graph shows the autocorrelation of all the spikes fired by a cell during the recording session, and the lower graph illustrates the autocorrelation of the same cell after all those spikes fired when the animal was moving faster than 0.05m/sec were removed. The residual rhythmicity in the rhythmically-bursting cell (some residual is also just visible in the plot of the type Ic cell) is due to movement at the ends of the track, outside the view of the overhead camera. Chapter Five

Phase relationship with hippocampal

theta rhythm ______Page 169

The current model of septo-hippocampal function regards the rhythmically-

bursting cells of the MS/DBB as “pacemaker cells” (Stewart and Fox, 1990)

which rhythmically inhibit GABAergic interneurons in the hippocampus,

which in turn produce rhythmic IPSPs in the pyramidal cells. Although there

is good evidence that normal MS/DBB activity is vital for the production of

hippocampal theta rhythm (Bland and Bland, 1986; Mizumori et ah, 1989;

Lawson and Bland, 1993), this model is unable to explain the wide disparity in

the preferential firing phase of these cells in anaesthetised animals (Gaztelu and

Buno, 1982;Stewart and Fox, 1989c).

This disparity is reflected in measurements of the preferential firing phase of

“theta cells” in the hippocampus (Bland et ah, 1980; Sinclair et al, 1982;

Buzsâki et ah, 1983; Fox et ah, 1986). Taken as a whole, these experiments show

that the preferential phase of hippocampal theta cells can vary substantially (see

especially Sinclair et ah, 1982), and the mean preferential phase varies from

study to study and between anaesthetised and unanaesthetised preparations.

Although it may be argued that at least part of this variance is due to error in

measuring the phase of theta rhythm itself, rhythmic hippocampal cells have

been recorded simultaneously and found to fire at different phases (Garcia-

Austt et ah, 1977).

If MS/DBB cells were pacing hippocampal theta rhythm directly, it would be

expected that they would exhibit little variation in their preferential firing

phase. This is clearly not the case in anaesthetised preparations, but it is

necessary to test whether these cells’ behaviour becomes more coherent in the

awake, freely moving state. Also, the acute preparations used previously

measure preferential phase only over a short time period. It is important to Page 170 discover whether the preferential phase of rhythmic MS/DBB cells varies significantly over time or remains a stable property.

M ethods

The mean theta phase (0) at which a cell fires was determined by finding the angle of the vector formed when all the individual phase vectors (one for each spike) were added: S if- C <:(), 6= atan c + 7T 'I s ' else i f S <0, 0= atan c + 2 k Is else 6= atan c

where 0^ is the theta phase of each spike,

Significance was tested at the p<0.001 level by using the property that

2NR^ is distributed as where R is the mean length of the resultant vector formed by addition of all the phase vectors in the data set:

The standard deviation of the preferential phase was evaluated using the formula given by Mardia (1972): 5o=V-21og,f R ) ______Page 171

R esults

The preferential phase of the MS/DBB cells recorded in this series of

experiments show a similar pattern of variation to that demonstrated in

anaesthetised animals (Gaztelu and Buno, 1982; figure 5.1), and little difference

can be attributed to the firing behaviour of the cells. As noted by Gaztelu and

Buno, it is possible that the phase distribution is grouped around two poles, at

90° and 270°, although such grouping is manifestly quite poor.

The preferential phase of firing of each spike belonging to a single cell may be

affected by the burst characteristics of the cell: cells which fire long bursts will

inevitably fire many spikes off the mean preferential phase. This is borne out

by examining the standard deviation of the preferential phase (figure 5.2).

The phase deviation clearly varies according to the type of firing behaviour,

with type lb cells showing the smallest deviation. Even so, the phase deviation

of a single cell is not sufficient to explain the variation seen in figure 5.1.

The average mean vector lengths for the different types of cell were:

Type la: 0.32±0.10

Type lb: 0.40±0.14-

Type Ic: 0.15±0.072

Type 2: 0.12±0.094

If types la and lb (the rhythmically bursting cells) and type Ic and 2 (non­

bursting) were grouped together, the averages of the mean vectors were:

Rhythmically bursting cells: 0.38±0.13

Non-bursting cells: 0.13±0.087 ______Page 172

The average mean vector of rhythmically bursting cells is comparable to the

0.43±0.16 found by Stewart and Fox (1989c) in urethane anaesthetised animals.

The phasic modulation of non-bursting cells in this study was significantly lower than that of the bursting cells (t test, p«0.0001), and lower than the

0.16±0.14 (n=3) and 0.31±0.15 (n=10) reported by Stewart and Fox for the two groups of non-rhythmic phase-locked cells that they described.

The preferential phase was evaluated against theta rhythm measured at the dentate gyrus. As the phase of hippocampal theta rhythm varies according to the depth of penetration of the electrode (Leung, 1984a,b), it may be argued that the variation in preferential phase is partly accountable to differences in the placement of the hippocampal EEG electrodes, although, as has been demonstrated, this was verified histologically. In many cases, however, multiple cells were recorded at the same time, and they could be differentiated on the basis of the four waveforms recovered from the recording tetrode. In nine cases a pair of cells, in ten cases three cells and in four cases four cells were recorded simultaneously. Page 173 Type 1 a Type 1 b

2

1 -

0 — 10 50 90 130 170 210 250LI 290 330 10 50 90 130 170 210 250 290 330 preferential phaseJ (degrees) preferential phase (degrees)

Type 1 c Type 2

2 - 3 -

2 -

1 1 ^ 10 50 90 130n 170 210 250 il 290 330 10 50 90 130 170 210 250 290 330 preferential phase (degrees) preferential phase (degrees)

All rhythmic cells

90 180 270 360

10 50 90 130 170 210 250 290 330 preferential phase (degrees)

Figure 5.1The mean theta phase of rhythmic MS/DBB cells grouped according to their firing behaviour (see chapter 4). The diagram at bottom right displays how the phase was measured with reference totheta rhythm in the dentate gyrus. As found by Gaztelu and Buno (1982), the preferential phase is widely distributed across the full range, although it is possible that there is some grouping around two poles at 90° and 270°. Page 174

Cell Type ■la Qlb 12^

■ Ic 02

Phase deviation (degrees)

Figure 5.2. The standard phase deviations of all cells which showed a relationship with hippocampal theta rhythm, grouped according to the cells' firing behaviour. ______Page 175

Figures 5.3-5.5 give examples of three different groups of cells which were recorded simultaneously, and in which the preferential phase varied widely.

As shown by the tetrode waveforms accompanying the phase plots, these cells could be separated with a high degree of reliability. The average difference between the mean phases of cells recorded simultaneously was 81.4°±56.5°, with a range from 3.02° to 175.6°. In cases where more than two phase-locked cells were recorded, the phase difference for each possible pairing within the group was calculated. No clustering of preferential phase was found along individual electrode tracks, either.

The preferential theta phases of different cells recorded simultaneously have been reported in only one previous study (Stewart and Fox, 1989c). In this case, on eleven occasions two different cells were recorded simultaneously. They found an average difference in the preferential phase of 60.8°±42.7°, similar to that demonstrated here.

These data demonstrate convincingly that the preferential phase of MS/DBB cells is not subject to any coherent organisation. It may be argued that phase disparity in the MS/DBB serves to maintain a degree of coherence in the resultant theta rhythm across the hippocampus (Bullock et al, 1990) by compensating for conduction delays. This hypothesis does not stand up to further analysis, though, as the septo-temporal length of the rat hippocampus is on the order of 1cm (Paxinos and Watson, 1986), yielding a time delay of around

20ms in unmyelinated fibres (at a propagation rate of 0.5m/sec). This is about l/6th of a theta cycle and is thus insufficient to account for the wide spread of preferential phase which has been observed. Furthermore, the projection from the MS/DBB to the hippocampus is, in part at least, topographic (Amaral and ______Page 176

Kurz, 1985;Nyakas et al, 1987), while the data presented here show that

preferential phase diversity is also seen in neighbouring cells, and it is thus

unlikely that MS/DBB topography reflects segregation on the basis of phase.

No previous study has measured the preferential phase of MS/DBB cells over

more than a few hours, as they have all been conducted in acute preparations.

The chronic implantations used in this experiment, however, allowed

individual cells to be monitored over many days. Twenty cells were recorded

over more than one day, and they had a similar preferential phase on each day of

recording. Figure 5.6 demonstrates cells which were recorded over five or

more days without substantial change in the preferential phase. Figure 5.7 plots

the standard error of the mean phase of all cells which were recorded on two or

more days, which gives an indication of the degree of variation in phase

between different recordings.

Not only do theta-related MS/DBB cells display preferential phases that fill the

full range, but these preferences are stable, indicating that the variance in

preferential phase cannot be explained as the result of phase drift over time. It is possible that not all of the cells recorded here project directly to the hippocampus as there is evidence for internal projections between MS/DBB

cells (Grimwood et al., 1995; Alreja, 1996; Yang et al., 1996). But, even so, it

must seriously be considered that the septo-hippocampal projections carry

rhythmic activity synchronised to widely differing phases of the theta rhythm which is ultimately produced. Page 177

350 - 300 -

300 - 250 ,

250 - 200 200 150 - 150 100 100 i

50 -

0 30 60 90 120 150 180 210 240 270 300 330 0 30 60 90 120 150 180 210 240 270 300 330 phase (degrees) phase (degrees)

350 - 30 -

300 - 25 -

20 .

15 -

10

0 30 60 90 120 150 180 210 240 270 300 330 0 30 60 90 120 150 180 210 240 270 300 330 phase (degrees) phase (degrees)

Figure 5.3. Four MS/DBB cells recorcded simultaneously. For each cell, the wave forms recorded from the four electrodes of the tetrode bundle are shown above a histogram of the phase of each spike relative to hippocampal theta recorded in the dentate gyrus. The spike waveforms of the different cells could be clearly differentiated, and the four resulting groups of spikes demonstrate preferential firing phases at approximately 90° intervals. Calibration bars: 1msec, 100p,V. Page 178

450 -

400 -

350 .

300 :

250 -

200

I 50 100

0 30 60 90 120 150 180 210 240 270 300 330 0 30 60 90 120 150 180 210 240 270 300 330 phase (degrees) phase (degrees)

350 -

300 .

250 :

200 :

150 -

30 60 90 120 150 180 210 240 270 300 330 phase (degrees)

Fig^ire 5.4. A further example of three ceils recorded simultaneously which fire at different phases of hippocampal theta rhythm (measured at the dentate gyrus). The mean wave forms of each cell are shown above a histogram plotting the theta phase at which each spike belonging to that cell fired. Calibration bars: 1msec, 100|lV. Page 179

200 , 40 - 180 . 160 . 140 - 120 .

100 -

0 30 60 90 120 150 180 210 240 270 300 330 0 30 60 90 120 ISO 180 210 240 270 300 330 phase (degrees) phase (degrees)

300 -

250

200 ,

1 50 100

0 30 60 90 120 150 180 210 240 270 300 330 phase (degrees)

Figure 5.5. Three cells recorded simultaneously which fired preferentially on different phases of theta rhythm. For each cell, the four tetrode wave forms are shown above a histogram of the phase of dentate theta rhythm at which each spike from that cell fired. Although two of the cells have a similar preferential phase, the third (top right) differs markedly. Calibration bars: 1msec, lOOpV. Page 180

360 - 330 - 300 - 270 - cuO) 240 - OJ 3 210 -

(LI 180 - to (T3 1 50 - CL 1 20 90 60 - 30 - 0 ------1 3 4 5 6 7 Day 360 - 330 ^ 300 +

CU 180 to n3 150 - 120 90 1 60 30 t 0 -

Day

360 -T 330 300 I cu 270 I " (V O) 240 '

Figure 5.6. The mean firing phase (+std. dev.) of three cells which were recorded on a number of occasions over five or more days. In no case did the mean phase display any significant variation. Page 181

5 -

IL 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 Standard error of mean phase (degrees)

Figure 5.7. The standard error(across daily records) of the mean phase plotted for all cells which were recorded on two or more days. ______Page 182

D iscu ssion

Given the pattern of innervation of the CA fields by MS/DBB afferents

(Nyakas et al, 1987), it would seem reasonable to assume that each theta cell receives input from more than one inhibitory MS/DBB cell. As there is a high degree of variation in the preferential phase of these cells, it is unlikely that a single theta cell would receive multiple inputs which were all in phase.

In order to explain the data presented here, it will be necessary to modify current models of septohippocampal function, as none can account adequately for the multiplicity of preferential phases. Stewart and Fox (1989c, 1990), suggested that septal inputs may be responsible for “gating” hippocampal theta rhythm rather than directly pacing it, with the septum regulating oscillatory activity intrinsic to the hippocampus. They do not advance any explanation for the variety of preferential phase, though. Lee et al. (1994), produce a model of the MS/DBB in which they assume that the cholinergic and GABAergic cell populations exhibit “population phase-locking”, with cells in each group firing synchronously and out of phase with cells in the other group. This model is clearly not supported by the experimental data presented here.

The need for a new model of septohippocampal function, and the production of hippocampal theta rhythm, is apparent. Despite the wide diversity in the phase at which MS/DBB cell fire, all reports of hippocampal theta cells show that the phase variation of these cells is considerably lower (Garcfa-Austt et al., 1977;

Bland et al., 1980; Sinclair et al., 1982; Buzsâki and Eidelberg, 1983; Buzsâki et al., 1983; Fox et al., 1986). This suggests, therefore, that the mechanisms by which the final phase of firing of the hippocampal theta cell is determined are ______Page 183 situated within the theta cell itself, and probably depend on integration of or competition between the disparate phases of the cell’s inputs. The inputs competing for determination of the interneuron output may have substantially different phases. Data presented later in this thesis has a bearing on this issue, and it will be discussed further in the final chapter. Chapter Six

Determining cell type by action

potential shape ______Page 185

As the MS/DBB contains two major cell types (cholinergic and GABAergic), which exert markedly different effects on their target cells, it is a matter of interest as to whether the two types can be distinguished by the waveforms recorded in the extracellular compartment. It is of especial interest as to whether classifying MS/DBB cells using this method yields differences in bursting behaviour or the phase of theta at which the cells fire.

Intracellular studies of GABAergic and cholinergic MS/DBB cells have revealed a number of differences, most notably the presence of a slow afferhyperpolarisation (AHP) in cholinergic and a fast AHP in GABAergic cells

(Griffith and Matthews, 1986; Markram and Segal, 1986; Griffith, 1988;

Markram and Segal, 1990; Gorelova and Reiner, 1996). Comparison of intra- and extracellular recordings from MS/DBB neurons in a slice preparation

(Matthews and Lee, 1991) has demonstrated that the slow repolarisation phase in cholinergic cells causes a "hump" in the action potential's derivative with respect to time. Extracellularly-recorded potentials from cells which had similar firing characteristics (slow versus fast rhythmic firing) and a similar response to apamin (a blocker of the calcium-activated potassium channel responsible of the long afterhyperpolarisartion) also show a very similar "hump"

(figure 6.1).

Matthews and Lee also distinguished MS/DBB cells on the basis of their action potential (AP) length: those cells with a “hump” also had an AP length of 1.8-

5.0ms measured at the baseline (these experiments were conducted at a temperature of 32°C, thus slowing down physiological processes), whereas those without had an AP length of 0.6-2.4ms. Cells with a “hump” also fired at a Paçe 186

S-AHPCELL F-AHP CELL

30«V Imi INTRA

-60mV

dV/dt

EXTRA

Figure 6.1. The action potential shapes of MS/DBB cells with slow (s-AHP) and fast (f-AHP) afterhyperpolarisations respectively. (A) The intracellular record of an action potential triggered by a 3 ms threshold depolarising pulse from a holding potential of -60mV (B) The first derivative of the action potential trace, displaying the "hump" produced by the slow afterhyperpolarisation. (C) Extracellular records from slow rhythmic firing (left) and fast rhythmic firing (right) MS/DBB neurons, which display a close resemblance to the derivatives of the s-AFIP and f-AHP cells respectively. Taken from Matthews and Lee (1991) Page 187

"Cholinergic"

GABAergic"

Figure 6.2. Examples of mean extracellular waveforms from putative cholinergic cells (top left group) which display a "hump" in the repolarisation phase of their action potential (see text), and putative GABAergic cells (bottom right group) which do not have such a hump. Each record comprises four waveforms, one from each electrode in the tetrode bundle, representing the average of several thousand action potentials recorded over a 32 minute period. Calibration bars: 1msec, lOOpV Page 188

Figure 6.3. Examples of two cells which could not easily be classified as G ABA or cholinergic by subjective examination of their wave forms. Each record comprises four waveforms, one from each electrode in the tetrode bundle, representing the average of several thousand action potentials recorded over a 32 minute period. Calibration bars: 1msec, lOOpV. ______Page 189 low rate, typically 5 to 6 Hz with a maximum of 20Hz. Non-”hump” cells fired at rates of 15 to lOOHz.

Results

A similar pattern of action potential shapes was found in this study. Figure 6.2 shows cells which can clearly be classified into the "hump" (putative cholinergic) and "no-hump" (putative GABAergic) classes. Unfortunately, many of the cells produced waveforms that could not easily be assigned to one or the other group, as shown in figure 6.3. Given this I attempted to find an objective method of classifying the waveforms.

The first method involved simply measuring the length of the action potential.

As the computerised waveform sampling system used only stores the first millisecond, the full waveform including the return to resting potential was not available for analysis. The length was determined, therefore, by finding the point in the repolarisation phase at which the largest of the four waveforms from the tetrode closed to within a certain value of the final sample. This value was set at l/50th of the full range. This value was selected from several other candidates as the one yielding the best spread of values. Even so, it did not yield good separation, as shown in figure 6.4. Figure 6.5 plots action potential length against the mean preferential phase of the corresponding cell, and shows no clear relationship.

An alternative, and more successful method involved computing the power spectrum of the extracellular waveform. As those waves which included a

"hump" would necessarily incorporate higher frequencies, it was thought that Page 190

trigger pointI

threshold 1 /SQth of full range

Action potential length

16 14

12

0.325 0.375 0.425 0.475 0.525 0.575 0.625 0.675 0.725 0.775 0.825 0.875 0.925 length (msec)

Figure 6.4. Attempts to separate cells on the basis of the length of their action potential did not yield fniithil results. At the top is a diagram demonstrating the method used to determine action potential length. As the length was measured from the start of the record, which was always 0.2msec before the wave crossed the threshold to trigger the start of recording, this method is potentially sensitive to variations in the height of the action potential. The histogram below plots the distribution of action potential lengths produced by this method. As can be seen, it is difficult to separate the cells into two classes by this criterion. Page 191

X 350 - X ^ X ♦Type la XX X ♦ oType 1b 300 - ^ □ □ A ^ypelc xTypeZ □ a % 2 ^ A ° 01 200 - ♦ X « n □

1 50 - X

100- □Ate ^ □ ♦ 4 X X AX 50 - ^ ^

X 0 0 0.2 0.4 0.6 0.8 Action Potential Length (msec)

Figure 6.5. Action potential length as measured by the method elaborated in figure 6.4 plotted against the preferential phase of the respective cell. This method yields no clear distinction in terms of either firing behaviour of the cell or its preferential phase. Page 192

o Ln Ln Ln O Ln O IS. Is. - Î5 m O 00 m "4- r\iLn njO njm Frequency (Hz) putative GABAergic cells

c3 ro 15

$ o Q_

O Ln Ln Ln O Ln Ln Ln o Ln o IS. L n r s r\i 00 oo Ü S - fSI (SI Frequency (Hz) putative cholinergic cells

Figure 6.6. The power spectra of the wave terms of two example cells which were putatively identified as GABAergic (top) and cholinergic (bottom). The power of a cell's wave form was measured at 16 ordinates between OHz and the critical sampling frequency of 25kHz. Calibration bars: 1msec, lOOpV. ______Page 193 this would show up on the power spectrum as a secondary peak distinct from the lower frequency main peak corresponding to the basic spike. As figure 6.6 shows, cells with a hump in their repolarisation phase instead produced a distinct broadening of the primary peak. The difference between the two spectral patterns was sufficiently large, however, to allow the cells to be split into two groups. This was done by examining the proportional difference between the values of the second and third ordinates. Those in which the third ordinate was more than 3/4 of the second were assigned to the putative s-

AHP/cholinergic group, and the rest to the putative f-AHP/GABAergic group.

Figure 6.7 demonstrates that this method yielded a fairly clear separation between the two types of cells. As some cells were recorded over more than one day, and the grouping of records from different days was performed by comparison of action potential wave forms, this graph uses the ordinate ratio of each day's record to avoid inadvertent corruption of the data. 34 cells (53 records) were assigned to the cholinergic group, and 36 cells (59 records) to the

GABAergic group. The average number of daily records per cell is very similar between the two groups (1.56 versus 1.64 respectively), suggesting that this did not introduce any notable bias. In only two cases did records assigned to the same cell fall into different categories. In these cases, the classification of the cell was decided by visual inspection of the wave form.

As figure 6.8 shows, this classification did produce two groups with a significantly different firing behaviour distribution (%^=14.3, d.f=4, p<0.01).

One third (12/36) of the putative GABAergic cells belonged to type lb (firing rhythmic, phase-locked bursts of one or a few spikes), whereas 44% (15/34) of Page 194

0 0 1 5

putative GABAergic “O cells c ou to

O)to 3 fC > c O) Ol g O l _Q

03

Figure 6.7. A plot of the ratio between the values of the third and second ordinates of the wave form power spectrum. As the hand-fitted curves show, it is plausible that these measurements reflect two separate underlying distributions, which allow the cells to be divided into two groups at a criterion ratio of 0.75. Page 195

□ CABAn

lb ic Cell Type Figure 6.8. GABA and cholinergic cells, sorted by examination of their action potential wave form spectrum, manifested significant differences in terms of their firing behaviour (%2=I4.3, d.f.=4, p<0.01). The cell type refers to the classification of Gaztelu and Buno (1982), as detailed in the previous chapter and describes the cells' relationship with hippocampal theta rhythm. Type 1 cells fire rhythmic bursts, type 2 cells are significantly phase-locked to theta rhythm, but fail to show rhythmicity in their autocorrelogram, and type 3 cells are non-rhythmic.

"ACh"

5 9 13 17 21 25 29 33 37 41 45 49 53 57 61

"GABA"

5 9 13 17 21 25 ■ 29 33 Jl 37 41 45 49iL 53 57 61L

Figure 6.9. The distributions of the mean firing rates of daily cell records. The cells were assigned to the cholinergic or GABAergic groups by examination of the spectrum of their action potential wave forms. Page 196

4 - Putative cholinergic cells

3 ^

2

1

10 30 50 70 90 n o 130 150 170 190 210 230 250 270 290 310 330 350 phase (degrees)

Putative GABAergic cells

10 30 50 70 90 110 130 150 170 190 210 230 250 270 290 310 330 350 phase (degrees)

Figure 6.10. The preferential phases of MS/DBB cells after they had been classified as cholinergic or GABAergic by examination of the wave form spectrum. There is no evidence of any systematic distribution. ______Page 197

the putative cholinergic cells belonged to type 2 (displaying a significant theta

phase lock without rhythmic bursting or noticeable rate modulation). There was no difference, however, in the distribution of average firing rates between

the two types of cells (figure 6.9). Neither did either of the two cell groups

display any overall preferential phase (figure 6.10).

D iscussion

It is important to note that the reliability of the classification presented here is not entirely certain. In separating the cells on the basis of a “hump” in the repolarisation phase of the action potential, this method is indirectly determining the presence of a slow AHP in the recorded cell. This, in and of itself, does not, of course, guarantee the type of neurotransmitter used by the cell’s projections, although the current evidence (as discussed previously) does suggest that this criterion can be used to make broad distinctions. It should be noted as well that Matthews and Lee themselves found a degree of uncertainty in classifying cells on the basis of their intracellular observations.While many of the cells that were found could easily be assigned to one of the two groups,

42% of the fast-AHP cells had action potential lengths which overlapped with those of the slow-AHP cells, and some fast-AHP cells failed to fire any faster than the slow-AHP cells.

Despite these caveats, however, the data presented here provide good evidence for a fundamental difference in the firing behaviours of GABAergic and cholinergic cells, namely that GABAergic cells tend to fire rhythmic bursts

(type lb), whereas cholinergic cells tend to non-rhythmic firing which still shows some significant (albeit lesser) theta phase modulation (type 2). As ______Page 198 demonstrated previously (chapter four), there is a significantly higher number of type 2 cells in awake, freely moving animals compared to the anaesthetised condition. This suggests that one of the differences between MS/DBB cell function in the anaesthetised and awake states is a higher degree of phasic modulation of cholinergic cells in the latter case. At present, it is still not possible to say whether this modulation arises from brainstem inputs, from changes in rhythmic interactions within the MS/DBB, or from the inhibitory backprojeétions from the hippocampus.

These data contrast with the conclusions reached by Stewart and Fox (1989a), who discovered that systemic injection of atropine caused one group of rhythmically bursting MS/DBB cells to fire in an incoherent, non-rhythmical manner, while a second group was unaffected. The authors assumed that the atropine-sensitive cells were cholinergic and the atropine-resistant ones

GABAergic, although there is anatomical evidence that GABAergic cells receive cholinergic synapses (Léranth and Frotscher, 1989), and that nearly all

GABAergic and cholinergic cells carry muscarinic cholinergic receptors (van der Zee and Luiten, 1994). As the data presented here show that the majority of rhythmically bursting (type la and lb) cells are probably GABAergic, it is more likely that the atropine-sensitive cells recorded by Stewart and Fox were a subgroup of the GABAergic population.

Another important implication of the low rhythmicity of cholinergic cells is that it contradicts the model of Stewart and Fox (1990) in which cholinergic septo-hippocampal projections cause fast, rhythmic excitation of hippocampal interneurons. The non-rhythmic, type 2 firing behaviour of MS/DBB cholinergic cells, together with physiological evidence for the slow onset of ______Page 199 cholinergic effects (Benardo and Prince, 1982) makes it unlikely that cholinergic projections are important in relaying rhythmic impulses. Although direct intracellular recordings of an interneuron’s response to cholinergics have not been performed, it has been shown that application of carbachol causes an increase in the number of spontaneous IPSPs recorded in pyramidal cells in a hippocampal slice preparation. This result implies that cholinergics do indeed excite hippocampal interneurons. This excitation has an onset latency of several seconds, however, which is far too slow for rhythmic transmission at the 8-9Hz theta rhythm frequency found in rats. It must be concluded, therefore, that cholinergic MS/DBB cells do not directly pace hippocampal cells to produce theta rhythm, but that their effect is exerted over a longer timeframe. Chapter Seven

Relationship between the firing

rate of MS/DBB cells and

hippocampal theta ______Page 201

Ford, Colom and Bland (Ford et ai., 1989) studied the relationship between the

firing rates of MS/DBB cells and the presence and frequency of hippocampal

theta rhythm. They found that it showed a similar pattern to that found in the

hippocampus (Colom and Bland, 1987), with cells that increased ("theta-on") and

decreased ("theta-off) their firing rate with the onset of theta rhythm in the

hippocampus. The method used by Ford and co-workers to determine whether

theta rhythm or LIA was present in the hippocampus was not fully specified,

and this was presumably performed by subjective visual inspection.

M ethods

In order to provide an objective assessment of any relationship between the

cell's firing rate and hippocampal theta rhythm, the data was segregated into a series of overlapping windows. In each window the power spectrum of the

EEG signal was estimated over 128 ordinates (approximately 1 per hertz) and the average firing rate was calculated. The amplitude of hippocampal theta rhythm was estimated by finding the ordinate with the highest power in the theta band (7 to 11 Hz) and adding the values of the neighbouring two ordinates.

This ensured that the power of theta rhythm was accurately reflected whether the frequency of theta rhythm fell exactly on an ordinate or not. The length of the window was dictated by the minimum number of data samples need to produce a reliable estimate (3 x number of ordinates). As the EEG was sampled once every 4.096 milliseconds (yielding a Nyquist frequency of 122.07Hz), the

384 samples required produced a window length of 1.573 seconds. The centres of each window were spaced 0.492 seconds (120 data samples) apart. Although these windows are relatively long they were found to produce results which Page 202

a lb le 2 3

theta-on theta-off

Figure 7.1. "Theta-on" and "Theta-off cells of the MS/DBB, grouped according to their firing behaviour (see chapter three). ______Page 203 correlated well with visual inspection of the EEG. Although shorter windows would have provided finer temporal resolution, this would be at the expense of resolution in the frequency domain, and the estimation of theta power would be more prone to corruption due to leakage from low frequencies.

Results

Of the 73 cells, 12 produced linear theta power/firing rate correlations which were significant at p<0.01, and a further 24 had significant correlations at p<0.05. Figure 7.1 shows the positive ("theta-on") and negative ("theta-off') correlations sorted by the firing behaviour of the cell. There was a definite bias in favour of cells with a significant phase relationship with theta rhythm (non type 3 cells), suggesting that the "tonic" theta-on and theta-off cells described by

Ford et al. (1989) were in fact mostly type 2 cells.

Figure 7.2 shows several data samples of a cell which showed a positive correlation between theta power and firing rate. The figure shows four separate examples of the cell’s firing behaviour. In each set, the top line plots the time of occurrence of each spike. Below this is the graph of the cell’s firing rate and below that a graph of the theta power. In both these cases, the parameter (firing rate or theta power) was calculated over a window centred, on the middle of the bar. The length of the bar represents the time between each window, which is approximately one third the length of the window itself. At the bottom of each diagram is the actual EEC waveform recorded from the hippocampus, which shows that the measurements of theta power correlated very well with the fluctuations in the degree of extracellular oscillation. Figure 7.3 shows a plot of the correlation between the two parameters, which exhibit a clear linear ______Page 204

B i Biflt c n ] y y i"iLULiLL.jj ii ------m ------1 . i ii ü u iL-U D üium üi

rjniTi fljiiiBBE : ] c. :\\ ii m n iiima l i iBiar:ii— ni~iTDniniiffi3mTTim^xBE3im

TTorni I n I II lamuLiuiui Branini— □ iD' : ai l u z m :

T a iB~BI B BBOniD IP i IIBB'^ 131JICMII II I"'3 I IIBII P nL J-lJI.U lir'’"!'], I,''U U 'V riJ .llIlLL[Cflll]J n iTlEr

Figure 7.2. Four examples of raw data from one cell demonstrating a positive correlation between the cell's firing rate and the power of hippocampal theta rhythm. Each example consists of four parts (from top to bottom): the spike timeline showing when each spike fired (because of the high firing rate, one mark may represent several spikes); the firing rate averaged over a period roughly 6 times the length of the bar (the data windows overlapped considerably); the power of theta rhythm measured over the same period; and the raw EEG record from the dentate gyrus. Calibration bars: Is; 50Hz; 329pV. Page 205

V i

i* <

20 30 40 50 60 80 90 firing rate (Hz)

Figure 7.3. Theta power plotted against firing rate for the cell illustrated in figure 7.2. This set of data contains each data window from 32 minutes of recording while the rat ran along the linear track. ______Page 206 relationship (r=0.60). Figures 7.4 and 7.5 contain similar data for a cell which had a negative correlation (r=-0.53). Cells belonging to the theta-on and theta-off groups showed no appreciable difference in terms of cell type as determined using the technique outlined in chapter 5: 12 "GABAergic" and 7 "cholinergic" cells showed a positive correlation, while 8 "GABAergic" and 9 "cholinergic" cells displayed a negative correlation. The two groups did not differ significantly in terms of average firing rates, either: theta-on cells had an average rate of 18.38±15.12Hz, while the average rate of theta-off cells was

24.13±22.80Hz.

D iscussion

The theta-on and theta-off cells recorded in this study constitute a much smaller proportion of the total population than did those reported by Ford et al.

(1989). They found that only 15 of the 103 cells recorded could not be classified as theta on/off by their classification. There are two possible explanations for this discrepancy. Firstly, the lower proportion found in this study may be related to the nature of the preparation, indicating that the firing rate of

MS/DBB neurons is more robust and has less tendency to fluctuate with theta rhythm in the awake, behaving rat. Secondly, the discrepancy may be ascribed to differences in the manner in which the theta relations were determined.

Apart from looking for a change in firing rate between the theta and non-theta states. Ford et al. also classified a cell as theta-on or off if its firing rate increased or decreased in conjunction with an increase in the frequency of hippocampal theta rhythm. This latter criterion was not used in this study. This was in part because the hippocampal EEG frequency produced during running remained stable at 8.5 to 9Hz throughout the experiments, whereas the EEG recorded Page 207

7 'ii: 0 1 III— r m — i- m — 11 iii no ] s in i i laflniifli: iii iii— m— rr-rr

i m — r iiflnnnonm ]i i i a iimoiiB i i iiini i iciniai \— i— mr

T i — n — ri------: k ;i: :i i: ]— iiDisfliia i i i i T------1—I— r 1—r

1— n—mrr— n — n —i—rmrmmnrTT T " ! — FTTT— r r n n i — i— r

Figure 7.4. Four examples of raw data from one cell demonstrating a negative correlation between the cell's firing rate and the power of hippocampal theta rhythm. Each example consists of four parts (from top to bottom): the spike timeline, showing when each spike fired; the firing rate averaged over a period roughly 6 times the length of the bar (the data windows overlapped considerably); the power of theta rhythm measured over the same period; and the raw EEG record from the dentate gyrus. Calibration bars: Is; lOHz; 367 |liV Paçe 208 .! • # c 3 e # 2 iû • ü IT3 • # a !: O$ CL (T3 # I e I • •

# • e 5 20 firing rate (Hz)

Figure 7.5. Theta power plotted against firing rate for the cell illustrated in figure 7.4. This set of data contains each data window from 32 minutes of recording while the rat ran along the linear track. The discontinuities in the firing rate values are due to the limitation on temporal resolution caused by the length of the data windows. ______Page 209 from urethane-anaesthetised rats in the case of Ford et al. showed substantial fluctuations, from 3 to 4.8Hz, a variation of 60%. Also, Ford et al. claim a precision of well below 0.1 Hz in measuring the frequency of theta rhythm. As can be seen from the data samples presented here, even the best EEG recordings produce a wave form which varies from cycle to cycle, which introduces a notable degree of uncertainty in any estimate of the oscillation frequency. For these reasons, the second criterion used by Ford et al. was not used. The impact of this on the number of theta-on/off cells detected may be judged from the comment by Ford et al. that “Cells in the MS/vDBB often had smaller differences in mean discharge rates between 0 and LIA [Large Irregular

Activity] states, compared to hippocampal cells and were thus more often classified as 0-on or 0-off by the criteria of linearity”, i.e. by observing an increase in firing rate correlated with an increase in theta frequency. They do not give any numerical indication of the number of cells that would be excluded if this criterion were not used, however.

The presence of theta-on cells can easily be explained as concordant with the increase in hippocampal acetylcholine release during voluntary movement

(Day et al., 1991; Mizuno et al., 1991), which is related to hippocampal theta rhythm (Vanderwolf, 1969). It is also in agreement with the disinhibition of hippocampal pyramidal cells found during the transition into theta rhythm

(Stewart, 1993), which may be the result of increased activity in GABAergic septal afferents onto hippocampal interneurons. The function of theta-off cells is less clear, however. It is not known whether these cells project out of the

MS/DBB (which could be tested in an acute preparation by recording from the cells while stimulating the fimbria/fornix and observing the presence of any ______Page 210 antidromic activation), or whether they are simply part of the internal network of MS/DBB cells responsible for controlling the theta-on projection cells.

It is clear that further study of this type of cells is necessary in order to understand their role in MS/DBB function. The median raphe produces serotonergic projections to both the MS/DBB and the hippocampus (Moore and

Halaris, 1975; Vertes, 1982) and these are known to inhibit hippocampal theta rhythm (Vertes et al., 1994). The serotonergic inputs to the MS/DBB have also been shown to cause the firing of rhythmic cells in this area to become desynchronised (Assaf and Miller, 1978; Kinney et ah, 1996). Furthermore, recent intracellular recordings in brain slices have shown that serotonin directly excites a certain subgroup of GABAergic MS/DBB cells (Alreja, 1996).

It is tempting to speculate that these cells are the same as the theta-off cells recorded here. This would need to be confirmed by combining the chronic in vivo techniques used here with pharmacological manipulations of the MS/DBB. Chapter Eight

Behaviour of rhythmically bursting

MS/DBB cells ______Page 212

Ever since the first electrophysiological recordings from the MS/DBB, the most notable feature of cells in this area has been their tendency to fire in rhythmic bursts. As noted in chapter four, the proportion of rhythmically bursting cells is significantly decreased in awake, freely moving rats, whereas that of the poorly rhythmic type 2 cells is increased. In this chapter I will explore two further issues regarding bursting cells of the MS/DBB, namely the stability of their rhythmicity and whether they display correlations with behaviour.

M ethods

Of the 73 recorded cells, 23 fired in a rhythmically- bursting manner (types la and lb), producing groups of spikes which could be clearly separated. The spikes were gathered into bursts on the basis of their inter-spike interval. A burst was defined as a cluster in which none of the spikes are separated by more than a certain inter-spike interval. This was varied between 25 and 40ms and the value which produced the sharpest burst frequency peak was used. This method was adopted as it offered the most simple and uncomplicated definition of a burst as a clearly defined and isolated group of spikes. This definition provides a clear distinction between cells which fire bursts (types la and lb) and those which display rhythmic rate modulation (type Ic).

The mean intraburst firing rate was calculated together with the skew coefficient , given by

1 ‘ X - X

where % is the mean and a the standard deviation) . ______Page 213

The time of occurrence of each burst was determined by finding the mean time at which each spike within the burst occurred. Using these markers, the burst frequency was defined as the inverse of the period between the current burst and the next in sequence. An alternative method, in which the frequency was calculated as the average of the inverse of the previous and following periods, was rejected as it produced peaks at anomalous frequencies when the cell missed cycles.

Results

The average intraburst firing rate was 138.2±38.9 Hz. The firing rate plots showed a substantial degree of positive skew, as shown by figure 8.1, and the overall skew coefficient was 2.33±1.1.

All of the rhythmically bursting cells were phase-locked to theta rhythm and thus produced bursts at the same frequency as hippocampal theta (figure 8.2).

As can be seen, the burst frequency peak of these cells is clearly visible. When it was attempted to isolate bursts in type Ic cells (which manifested rate modulation rather than discrete bursts) the resulting frequency distribution was much broader, though still centred around theta frequency (figure 8.3).

Burst frequency stability

Fourteen of the cells (61% of the all the rhythmically bursting units) frequently failed to fire on each cycle. As figure 8.4 shows, this was manifested by the formation of a secondary peak at 4Hz (representing one dropped cycle) and, in some circumstances, additional peaks are visible at lower frequencies (2.66Hz Page 214

.ml—i.j ■_ SO 200 250 300 350 400 50 100 ISO 200 250 300 350 400 frequency (Hz) frequency (Hz)

140

120 100

80

60

U i k t a . l l L i . . L . . . l l M . l L j U l .j |.j u |L l l j .l l ( _ » l _ . Lm_l 1 0 50 100 ISO 200 250 300 350 400 0 150 200 250 300 350 400 frequency (Hz) frequency (Hz)

Figure 8.1. Four examples of tJie distribution of intra-burst firing rates in rhythmically bursting MS/DBB cells. The skewed distribution leads to a mean rate which is higher than the median, (y-axis: number of cycles) Page 215

300

250 -

.200 _ _0J 50 -

100 -

50 -

0 0 2 3 4 5 6 7 8 9 10 11 12 13 14 15 frequency (Hz)

Figure 8.2. Rhythmic cells in which bursts could be clearly distinguished (type lb) produced sharp frequency peaks corresponding to hippocampal theta rhythm.

120 ^

00

80 cu Ü 60 u>. 40 i

20

0 0 1 2 3 4 5 6 7 8 9 10 1112 13 14 15 frequency (Hz)

Figure 8.3. The burst frequency plot from a rate modulated (type Ic) cell, displaying a far broader peak centered on the frequency of hippocampal theta rhythm. ______Page 216 and 2Hz, representing 2 and 3 dropped cycles respectively). The height of the secondary peaks varied widely, reflecting differences in the prevalence of dropped cycles, as shown in figure 8.5, but all were clearly visible. Eight of these cells were classified as GABAergic and six as cholinergic.

These cells’ frequent failure to fire on every cycle has serious implications for the pacemaker theory of MS/DBB function. If these cells exerted direct control of their target neurons through rhythmic inhibition, the missed cycles seen here would also be present at the hippocampal level. This will be discussed in greater depth in the final chapter.

Correlations between burst frequency and running speed

Fifteen of the rhythmically-bursting cells (65%) also demonstrated a significant

(p<0.05) correlation between burst frequency and the running speed of the rat

(figure 8.6), and 6 of those cells were recorded on two separate days, displaying significant correlations on both occasions. Eleven of the cells were classified as

GABAergic and the remaining 4 as cholinergic, although this distinction probably simply reflects the greater number of GABAergic rhythmically bursting cells. The average slope using linear regression was 0.87+0.2

Hz/m/sec and the average y-intercept was 8.4±0.13Hz. Although the method of delineating bursts inevitably yielded a number of outliers, especially at low frequencies, the robust maximum likelihood estimator (a technique which is designed to lessen the effect of outlying data points: Huber, 1981) gave very similar results: a slope of 0.94±0.2Hz/m/sec and a y-intercept of 8.3±0.17Hz. ______Page 217

It has been demonstrated that the frequency of hippocampal EEG can be influenced by the body temperature of the animal (Whishaw and Vanderwolf,

1971), presumably by altering the kinetic rates of the metabolic processes underlying the generation of theta rhythm. In order to test whether the brain temperature of the rat altered significantly while it ran on the linear track, a fast-response (10ms) thermocouple was implanted in the cortex above the

MS/DBB after having been stiffened and insulated with epoxy glue. This was connected (using identical metals to avoid the formation of spurious thermoelectric effects) to a specialised thermocouple amplifier (Analogue

Devices AD595AQ) which also provided thermal compensation and was mounted on the rat’s head in a similar fashion to the microelectrode headstage.

The signal from this thermocouple/amplifier combination was factory-set at

10mV/°C with a gain error of ±1.5%, this was checked before implantation and found to be accurate (within the limits of the mercury thermometer used). The output from the amplifier was filtered at O-lOOHz with a 50Hz notch filter and amplified before being fed to the digitiser usually used for EEG recording. The animal’s position was also monitored using the methods described in chapter 2.

Figure 8.7 shows the results gained by measuring temperature and speed at the same time while the rat ran on the track. As can be seen, the rat tended to run in short bursts, which were too short to build up temperature through muscular activity. While the thermocouple worked properly before and after implantation, it is possible that some local circumstance prevented it from measuring the rat’s body temperature accurately, and this would have best been tested by artificially raising the body temperature and monitoring the thermocouple output. Unfortunately, our laboratory is not licensed to perform such a procedure, and this further test was not possible. Page 218

100 -

kUj4_k 12 13 14 15 10 11 12 13 14 15 frequency (Hz) frequency (Hz)

140 - 2 0 0 -

100 -

6 7 S 9 10 11 12 13 14 15 2 3 4 5 6 7 8 9 10 12 13 14 15 frequency (Hz) frequency (Hz)

140 -

120 -

2 3 4 5 6 7 8 9 10 11 12 13 14 15 3 4 5 6 7 8 9 10 11 12 13 14 15 frequency (Hz) frequency (Hz)

Figure 8.4. The burst frequency plots of rhythmic cells which frequently skippe(d cycles of theta rhythm. The seconcdary, tertiary and quaternary peaks reflect 2,3 or 4 dropped cycles and are marked with arrowheads. Page 219

0 3 4 5 6 7 8 9 10 11 12 13 14 15 frequency (Hz)

350 -

300 -

250 -

200 -

150 ^

100 i

50 -

0 0 2 3 4 5 6 7 8 9 10 1112 13 14 15 frequency (Hz)

Figure 8.5. Two examples of the least obvious secondary frequency peaks in rhythmic cells caused by skipped theta cycles. The secondary peaks are marked by arrowheads. Page 220

14

12

• • • •

0 — 0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 running speed (m/sec)

• t

Î

0.1 0.2 0.3 0.4 0.5 0.5 0.7 0.8 0.9 running speed (m/sec)

♦ - •

• * • ; 0.2 0.4 0.6 0.8 1.2 running speed (m/sec)

Figure 8.6. Three examples of the relationship between the burst frequency of rhythmically-bursting cells and the running speed of the rat. Superimposed over the data is the linear regression obtained using a robust estimator (see text). Page 221

running speed core body temperature

-^w-

w— v«V

Figure 8.7. Six examples of the rat's running speed plotted above relative change in core body temperature measured using a rapid response thermocouple implanted in the cortex above the MS/DBB. Although some small variation in core temperature is seen during brisk bouts of running, the changes are too small to account for the variation in burst frequency with running speed. Calibration bars: Isec; 0.5m/sec; 0.5°C. ______Page 222

D iscu ssion

A clear behavioural relationship between theta rhythm in the rat and voluntary movement has been well established (Vanderwolf 1969). Together with the involvement of the hippocampus in spatial navigation (O'Keefe and Nadel

1978), this has prompted interest in studying the interaction between various parameters of theta rhythm and the rat's movement through the environment.

The linear relationship between the frequency of rhythmical bursting activity in the MS/DBB and the speed at which the rat runs corresponds with earlier studies which showed similar relationships between the frequency of hippocampal theta rhythm and speed of motion. The first evidence to suggest such a correlation was reported by Vanderwolf (1969). Rats were trained in a shock-avoidance task which involved jumping out of a box with an electrified floor. The animals soon began to jump out spontaneously, and developed a stereotyped response to being placed in the box. Before jumping out there would be a brief moment of stillness, as the rat tensed its muscles. Theta rhythm was present throughout, and showed a distinct increase in frequency

(from 6-7Hz to 8-12Hz), beginning 10-15 cycles before the jump took place. In this case, the increase in frequency occurred before the actual movement took place, which may be related to the instantaneous, ballistic nature of the rats' jump.

Whishaw and Vanderwolf (1973) and Morris and coworkers (1976) both described an increase in theta rhythm frequency prior to jumping in the same shock-avoidance task. It was further shown that the frequency change increasedwith an increase in the height of the jump (Whishaw and Vanderwolf ______Page 223

1973) and that this increase could not be replicated by attaching weights to the rat to increase the amount of force generated in the jump (Morris et al 1976). It was concluded that the velocity or distance travelled was the important factor in the relationship. It should be noted that, though they found an increase in

EEG frequency before jumping, Whishaw and Vanderwolf (1973) found no correlation between EEG frequency and speed of motion while the rats were running at a steady speed in a treadmill, a result which is at odds with other evidence outlined below.

Further support for a correlation between theta frequency and speed of onset of motion is given by Bland and Vanderwolf (1972). They found a linear correlation between the frequency of hippocampal theta rhythm in the first second after stimulation and the running speed of the rat. Theta rhythm, and corresponding motor behaviour, was induced in the rats in this study by stimulation of the dorso-medial posterior hypothalamus. The authors found that the frequency of theta rhythm induced under these circumstances was also dependent on the voltage used to stimulate, with high voltages causing a high onset frequency. The frequency would decline over roughly 30 seconds, though, to a stable level of 8Hz for all levels of stimulation

McFarlane and coworkers (1975) examined theta activity in rats running on a treadmill at various speeds. They measured theta rhythm by passing the raw

EEG through two band-pass filters, one set at 6-8Hz, the other at 10-12Hz and then using a Schmidt trigger to count the number of waveforms present in the filtered outputs. EEG records from the hippocampus showed a dramatic increase in wave counts in the 6-8Hz compared to the 10-12Hz band as the treadmill speed was increased. They also present example waveforms of 10.3 ______Page 224 seconds each which show the records obtained with the treadmill running at various speeds. At the three higher speeds, in which synchronous activity is

continuous, the number of waveforms increases from 70 (at8 cm/sec) to 76 (at

15cm/sec) and 78 (at 22cm/sec).

Arnolds and coworkers (1979a, 1979b), studying hippocampal EEG in dogs walking on a treadmill, found a direct correlation between muscle activity in the dogs' legs, hippocampal theta rhythm frequency, and the resultant velocity of the dogs. Furthermore, passive movement of the dogs across the same space resulted in no changes to theta frequency, demonstrating the necessity for active involvement of the motor system.

A correlation between velocity and firing rate in hippocampal principal cells has also been noted (McNaughton et al 1983, Wiener et al 1989), and

McNaughton and coworkers found such a relationship in theta cells as well. It is difficult to integrate these results with those showing a frequency/velocity correlation, though, as an increase in firing rate of theta cells could be achieved by increasing the number of spikes per burst, rather than increasing the number of bursts per second and thus the frequency of the theta rhythm that is produced. It is possible, however, that an increased theta frequency results in an increase in the number of cholinergic action potentials arriving in the hippocampus, and thus an increase in the amount of acetylcholine released.

There is evidence that cholinergic MS/DBB cells are rhythmic (Stewart and

Fox 1989c), and nearly half of the rhythmic cells in the MS/DBB increase their firing rate linearly with an increase in the frequency of theta rhythm (Ford et al

1989). Cholinergic MS/DBB cells synapse onto both pyramidal and inhibitory cells in the hippocampus (Frotscher and Lérânth 1985), and acetylcholine has a ______Page 225 noted excitatory effect on both pyramidal (Benardo and Prince 1982, Cole and

Nicoll 1984) and inhibitory (Pitler and Alger 1992) cells.

Recently, a more formal investigation of the relationship between speed and the frequency of hippocampal EEC has been performed (Recce 1994). These data demonstrated that changes in frequency occurred earlier than the corresponding changes in speed, thus, the correlation between EEG frequency and speed improved when the speed data's time scale was shifted by about 15%, resulting in an overall correlation coefficient of 0.57. The correlation could be fit by a

linear function with a slope of 2,3Hz/(m/sec) and an intercept of 8 .5Hz. The predictive effect of changes in EEG frequency is similar to the data reported in

Vanderwolf (1969) and Arnolds and coworkers (1979b), which may be due to the inevitable delay between initiation of faster motor activity and the resultant increase in speed. Chapter Nine

Directional rhythmicity ______Page 227

As noted in the introduction to this thesis, the place cells of the hippocampus display considerable directionality in their response. Furthermore, a group of cells have been found in the dorsal presubiculum which are highly responsive to the direction in which the animal’s head is pointed (these cells will be reviewed later in this chapter). It was felt to be of interest, therefore, to test whether MS/DBB cells showed any change in activity with the direction in which the rats ran.

M ethods

To assess the effect of the direction of motion of the rat on the activity of

MS/DBB cells, the linear track was placed on a simple turntable, as shown in figure 2.3. The linear track was then rotated relative to a fixed axis drawn on the floor of the laboratory. When recording a full set of 16 directions across 360° (at

22,5° intervals), 8 orientations were used (each orientation yielded two motion directions as the rat ran back and forth), which were recorded in random sequence. No change was made to the surface of the track between orientations, and thus smell cues on the track surface remained in place, although the track was cleaned from day to day.

In other areas of the brain, the intracellular processes underlying rhythmic action-potential generation have been shown to be continuous, sinusoidal or saw-tooth functions driven by interactions between ionic flows across differing membrane channels (Jahnsen and Llinas 1984, Ylinen et al. 1995b). In analysing rhythmicity in data recorded from extracellular studies we are faced with the task of reconstructing an analogue of the intracellular trans-membrane potential

(TMP) fluctuations. Yet the action-potential record is merely a time-series ______Page 228 indicating the points at which the TMP crossed the threshold for activation of the sodium channels in the cell's axon hillock, and deprives us of the higher- frequency data which represent the actual shape of the TMP fluctuation.

In terms of characterising the cell's overall rhythmic behaviour, and particularly with regard to discovering its effect on cells downstream, it is useful to regard the spike sequence as a set of discrete samples from a function

(which will be referred to as the transmission function) which can be characterised by Fourier analysis. This reconstructed function is an altered version of the original TMP fluctuation, distorted by the non-linear nature of action-potential generation and its limited bandwidth, but it represents the generative source of TMP fluctuations in the receiving cell (subject to non­ linear interactions at the synaptic junction), and is thus worthy of analysis.

Using conventional techniques, the power spectrum of a spike timeline may be estimated directly by representing each spike as a Dirac 0-function (French and

Holden, 1971; Lange and Hartline, 1979; Schild, 1982). After filtering and sampling, the resultant continuous waveform may be used to compute the power spectrum by conventional Fourier techniques.

Unfortunately, this method assumes that the spike train itself is continuous.

The method used to separate the spikes according to the animal's direction of motion leads to large gaps in the data at irregular intervals (due either to the animal running in the opposite direction or to the animal stopping and moving its head to look at objects off the track), and this causes the power spectrum to contain large and unpredictable low frequency components. The power from these readily leaks into the theta band and makes it extremely difficult to ______Page 229 compare power spectra from different trials, as the low frequency leakage can dominate the theta band in some cases.

These problems are largely obviated if, instead, the rhythmicity analysis is based on the cell’s autocorrelogram (Perkel et al, 1967). Furthermore, this averages out the rhythmicity of those cells which tend to miss cycles in a random fashion. As has been shown, some cells skip cycles in this way relatively frequently, a behaviour which would cause direct measurements of the power spectrum to show power at a sub-harmonic of theta rhythm (for example, if the theta rhythm frequency were 8Hz, one skip would produce a

4Hz cycle, while two skips would produce a 2.67Hz cycle). The powerspectrum ol such cells was computed using the direct methods described above, and it was found that the subharmonic frequencies led to a broad hump covering the range from twoto ten Hz rather than a sharp peak at the frequency of hippocampal theta.

The autocorrelation of a typical data set is shown in fig 9.1.

Figure 9.1. The autocorrelation function (over +/- 1200ms) of a directionally rhythmic cell while the rat is running in the cell's preferred direction.

As can clearly be seen, the autocorrelation converts underlying rhythmicity into a sine wave whose amplitude decreases as time increases (this decrease is caused by the accumulation of stochastic error in the rhythmicity of the transmission function as discussed in Perkel et al (1967)). Such a function can Page 230 then be analysed in terms of its power spectrum density, producing the periodogram shown in figure 9.2.

4,50i-

4 00;

3 5Q-

3.00

2.50

2.001

1.50(-

1.00

0.50-

o.oo| •UlllrUl 5&P0 F re q u e n c y Figure 9.2. The power spectrum of the autocorrelation function shown in figure 9.1. Frequencies below 2.5Hz have had their values set to zero.

There are several points of caution which need to be respected in the execution of these various transformations, especially when differing data sets are to be compared with each other.

The autocorrelation function is given by N Corr(j) = Ylf(i)f(i +jX

where/(i) is one if a spike occurs at time i and zero otherwise, and the time value j ranges over -s to s, producing a histogram with 2s data points, representing the time window over which the autocorrelogram is calculated. To determine the power spectrum, the data is segmented into k segments of 2m points each, where m equals the number of frequencies between zero and

(the Nyquist frequency, which is the highest frequency the sampled data can represent) at which the power spectrum is to be estimated (Press et al, 1988, ______Page 231

chapter 12). After being passed through a Parzen windowing function (to

minimise power leakage from neighbouring frequency bins), the K data

segments are passed through a fast Fourier transformation to produce

periodograms which are then averaged. In order to achieve the smallest spectral

variance per data point, the data are overlapped, as described in Press et al

(1988), producing a requirement of (k+l)m data points.

Each bin in the final periodogram will correspond to the power at (pym)Hz.

The time-resolution of the data collection system used in the experiments is

fixed at a sampling interval of O.lps, giving a maximum possible of 5kHz. This is far beyond the requirements for analysis of the cells in the area studied, most

of which have a rhythmicity in the region of 4-lOHz. Also, since the bins in the

resultant periodogram will be spread evenly from 0 to F^, it is necessary to restrict the power spectrum to the low frequencies in order to provide sufficient resolution at the frequencies of interest. This is performed by resampling the data during the construction of the autocorrelogram. This

requires calculation of the bin size and the number of points over which the

autocorrelogram will be summed.

The bin size of the resulting autocorrelogram is equal to the granularity which with the value y in the equation given above is varied. This is determined by

the required sampling interval (A), which is given by A=1/(2F^). The number of points is then calculated as follows. The sampling interval is equal to the time Page 232

over which the autocorrelogram is summed (s), divided by the number of samples required:

A = ----- seconds (k+ \),m and we can therefore determine s as

_ (k+ 1 )m seconds.

Since each data point in the autocorrelogram occurs at 0.0001 sec, the number of points over which we need to correlate (n) is

(k+ 1 )m 2 .f ,. 0.0001

The periodogram estimator used will provide the mean squared amplitude of the function at each of m frequencies between 0 and Fc. The sum of these values will give the mean squared amplitude of the autocorrelogram. If periodograms from two different autocorrelograms are to be compared, however, we must compensate for their differing amplitudes, as these are directly related to the firing rate of the cell over the time the data was collected. Thus, to prevent analysis of the cell's rhythmicity being confounded by the cell's firing rate, the autocorrelograms are normalised such that the value of each bin is divided by the total value of all bins in that autocorrelogram, ie; N

Corr(j)= ^ n N

j = - n i

This ensures that the total power of all the autocorrelograms is the same. ______Page 233

Due to the nature of the autocorrelogram as a 'signal' (none of the data points has a negative amplitude), the final periodogram contains substantial DC and very low frequency components which are irrelevant for our analysis. The periodograms are thus analysed from IHz upwards.

In practice, the procedure followed for determining the presence of directional rhythmicity is as follows:

1) 8 data sets are recorded from the rat running on a track set at 22,5° increments through a 180° arc (since the rat runs in both directions along the track, each data set provides two directions for the final analysis). The directions are recorded in random order.

2 ) The data are filtered to remove noise and isolate the cell's spikes as described

in chapter 2 , and the spikes were segregated on the basis of the animal's direction of motion at the time the spike fired. The track on which the animals were running was so narrow as to allow motion in only one of two directions, which facilitated this process. Spikes which fired when the animal was not moving in one of the two axial directions were deleted, as were spikes which fired when the animal’s speed was less than 0.05 m/sec. This effectively eliminated spikes fired when the animal was not moving in the corresponding direction along the track. As the runs in the two directions were by necessity interleaved, any effects due to fatigue, satiety or changes in the external environment (such as changes in the volume of noise coming from the traffic outside) were balanced evenly between the two directions of motion. Note that each direction set only contains spikes recorded with the track set to that direction. Cells whose overall firing rate was less than 5Hz were not considered for further analysis. ______Page 234

3) The rhythmicity of each direction set is then analysed by the method described above. The periodogram for each direction is scanned for a peak in the theta rhythm band (7-12Hz). The significance of a peak was then tested using the Kolmogorov-Smirnov test at the p<0.01 level.

Results

Four cells were found whose rhythmicity varied depending on the direction of

motion of the rat (Figures 9.3-9. 6 ). These cells fired rhythmically at the frequency of theta rhythm when the animal moved in the preferred direction, but the firing behaviour lost coherence when running in other directions. The power spectrum variance of the directional cells was significantly different

from that of the other, non-directional cells (F^^ 3 3 3 =3.09, p< 0 .0 0 1 ). In three of

these cells (r374d2, r432dl, r480dl; figures 9.4-9. 6 ) the variation in rhythmicity demonstrated a single, significant peak (Kuiper test, p<0.05). A fourth showed directionality, but failed to pass criterion, possibly due to poor behaviour

(r374dl, fig 9.3). On one occasion, two cells were recorded simultaneously, one of which maintained a relatively constant rhythmicity across the full range of directions while the other demonstrated a marked change in rhythmicity as the rat's direction of motion changed, with a sharp peak at 135° (Fig 9.7).

It is possible that the rhythmicity of MS/DBB cells is linked to the rat's running speed or the firing rate of the cell being observed. To test for this,

ANOVA statistics were computed to test for significant variation in mean velocity or firing rate. This hypothesis was strongly rejected in all cases

(p>0.5). Spearman ranked cross correlations between the power of theta ______Page 235 rhythmicity and the velocity parameters or average firing rate failed to reveal

any correlation with a significance below the p< 0 . 1 level.

The cells with directional rhythmicity did not form an homogenous group in terms of their firing rate (r374dl: 6.0Hz; r374d2: 6.2Hz; r432dl: 29.7Hz; r480dl: 59.5Hz) or in terms of their classification. Using the classification system outlined above , two cells (r374dl, r374d2) were classified as type lb, one cell (r432dl) as type la and one (r480dl) as typelc. Two cells were classified as GABAergic by their action potential shape, and the other two as cholinergic. Three of the four cells showed a weak but significant (p<0.05) negative correlation between their firing rate and the power of hippocampal theta rhythm (r=-0.1691, -0.1207, -0.1350).

Discussion

The change in rhythmicity found here has parallels with results from anaesthetised or restrained animals. As discussed in the introduction, there are four reports from different laboratories (Petsche et al., 1965; Wilson et al., 1976;

Dutar et al., 1986; Brazhnik and Vinogradova, 1988) of MS/DBB cells which changed their rhythmicity independent of the state of hippocampal theta rhythm. These data, together with the results presented here, suggest that modulation of the rhythmicity of a subgroup of MS/DBB cells may play a part in the overall function of this area.

Before discussing this in further depth, I will review the head direction cells mentioned earlier. As we are examining theta rhythm, which is strongly Page 236 I II II a i I I

7 - 6 -

I 6 -

•500 -400 -300 -200 -100 0 100 200 300 400 500 -500 -400 -300 -200 -100 0 100 200 300 400 500 lÉiiiÉiiiiimsec msec

^ g 0.15 c n 1

: 5 0 05

45 90 135 ISO 225 270 Direction of motion (degrees)

Figure 9.3. Variation in MS/DBB cell rhythmicity with direction of motion. At the head of the page are examples of the raw spike timeline as the rat ran through identical portions of the track. At the top, the rat is running in the preferred direction, and the cell fires in phase with hippocampal theta rhythm. Below, the rat is running the other way, and the cell's firing is disorganised. Beneath this are the autocorrelogram plots from spikes fired when the rat ran in the preferred (left) and null (right) directions. At the bottom is the plot of correlogram rhythmicity versus direction. The grey bars represent rhythmicity values which failed to be significant at the p<0.()l level. Calibration bars: 150msec; 260pV Page 237

I I II I I I I

-500 -400 300 -200 300 400 500 -500 -400 -300 -200 300 400 msec msec

-Û 0.05

45 90 135 1 80 225 270 31 5 Direction of motion (degrees)

Figure 9.4. Variation in MS/DBB cell rhythmicity with direction of motion. At the head of the page are examples of the raw spike timeline as the rat ran through identical portions of the track. At the top, the rat is running in the preferred direction, and the cell fires in phase with hippocampal theta rhythm. Below, the rat is running the other way, and the cell's firing is disorganised. Beneath this are the autocorrelogram plots from spikes fired when the rat ran in the preferred (left) and null (right) directions. At the bottom is the plot of correlogram rhythmicity versus direction. The grey bars represent rhythmicity values which failed to be significant at the p<0.()l level. Calibration bars: 150msec; 260|lV Page 238 1 II1 I I 1 II III 1 III III I I I I 1 mill nil II nil 1 II II1 11 V/ViA r - V wvA/v v V\a v I I I II II 11 1 II III 1III IIII1 1 1 Mini Mill III Hill 1 1 II II ^VvV\AAAV'/l,vuV \A V

150 ,

■500 -400 ■300 -200 -too 0 too 200 300 400 500 -500 -400 -300 -200 -100 0 100 200 300 400 500 msec msec

> Ü 0.35 E 0.3 'c % 3 0.25 _c 0.2 - TD C s 0.15 - _o 0.1 - rtJ 0.05 - 0» JZ 0 ■ 45 90 135 180 225 270 315 Direction of motion (degrees)

Figure 9.5. Variation in MS/DBB cell rhythmicity with direction of motion. At the head of the page are examples of the raw spike timeline as the rat ran through identical portions of the track. At the top, the rat is running in the preferred direction, and the cell fires in phase with hippocampal theta rhythm. Below, the rat is running the other way, and the cell's firing is disorganised. Beneath this are the autocorrelogram plots from spikes fired when the rat ran in the preferred (left) and null (right) directions. At the bottom is the plot of correlogram rhythmicity versus direction. Calibration bars: 150msec; 625)iV Page 239 III mill m il II mil iiiiiii iiii iiiiii iiiii iiii iiiii iiiiiiiiiiiiiiiiiiii

■500 -400 •300 -200 -100 0 100 200 300 400 500 -500 -400 -300 -200 -100 0 100 200 300 400 500 msec msec

0 45 90 135 180 225 270 315 Direction of motion (degrees) Figure 9.6. Variation in MS/DBB cell rhythmicity with direction of motion. At the head of the page are examples of the raw spike timeline as the rat ran through identical portions of the track. At the top, the rat is running in the preferred direction, and the cell fires in phase with hippocampal theta rhythm. Below, the rat is running the other way, and the cell's firing is disorganised. Beneath this are the autocorrelogram plots from spikes fired when the rat ran in the preferred (left) and null (right) directions. At the bottom is the plot of correlogram rhythmicity versus direction. The grey bars represent rhythmicity values which failed to be significant at the p<0.01 level. Calibration bars: 150msec; 417|lV Page 240

Cell 1 Cell 2

1 35 degrees

150 200 250 300 350 400 450 200 250 300 350 4 00 500 m s e c m s e c 31 5 degrees

150 200 250 300 350 400 450 100 150 200 250 300 350 400 450 500 m s e c m s e c

0.35 - 0.35 0.3 i 0.25 - 0.25 0.2 r 0.15 0.1 0.05

45 90 135 180 225 270 315 0 45 90 135 180 225 270 315 Direction of motion (degrees) Direction of motion (degrees)

Figure 9.7. Two MS/DBB cells recorded simultaneously. Cell 1 fired rhythmically when the rat was travelling at 135°, but not in the opposite direction. Cell 2 fired rhythmically in both directions. From the top: the four-channel spike waveform of each cell (Calibration bars: 1msec, lOOpV); autocorrelogram when travelling at 135° autocorrelogram when travelling at 315°; plot of normalised theta power (arbitrary units) against direction of motion - the grey columns represent rhythmicity values in which there was not a significant (p< 0 .0 1 ) amount of power in the theta band, which may be related to poor behaviour as the rat tired. ______Page 241 correlated with active movement, this study has concentrated on direction of motion rather than head direction. When running in a particular direction, however, the rat's head is usually oriented along its direction of motion, so it is relevant to consider the results found in this study in relation to the literature concerning cells responsive to head direction.

The directional properties of cells located in the dorsal presubiculum, which is also known as the postsubiculum (PoS), were first described by Ranck (1984,

1985) and were subsequently studied in greater detail by Taube and his co-workers (Taube et al., 1990a,b). These cells increase their firing rate markedly when the animal’s head is pointing in a specific direction. Cells tuned to head direction have since been found in the anterior thalamic nuclei (ATN)

(Taube 1995); the lateral dorsal thalamic nuclei (LDN) (Mizumori and

Williams, 1993); and the retrosplenial and medial prestriate cortex (Chen et al

1994, McNaughton et al, 1991). Of these areas, the cells in the PoS and ATN appear to have very similar discharge properties (Taube 1995). The cells of the other two areas have not been examined closely enough to ascertain whether or not they share this similarity.

Postsubicular head direction cells show highly specific tuning (Taube 1990a).

The preferred directions of these cells were evenly distributed across the full range of the compass. The average extent over which firing rate was elevated above the baseline was 83.4°, with a mean signal to noise ratio of 83.1:1. The background firing rate was less than 1 Hz, while the mean peak firing rate was

35.7Hz. The change in firing rate with direction could be modelled as a triangular function in which the limbs on either side of the peak were equal in length and angle (an isosceles triangle). There was weak evidence that the ______Page 242 animal's location modulated the cells' activity, but this was of borderline significance. It was also found that the head direction cells maintained their directional specific firing when the animal was restrained and its head rotated by the experimenter, although peak firing rates did decrease in this situation.

Head direction cells in the ATN, however, ceased firing completely under these conditions (Taube, 1995).

It was found that the direction of the firing rate peak could be manipulated by altering the salient visual cues in the environment. The cells' preferred direction was shown to rotate to follow a white cue card in an otherwise featureless grey cylindrical environment (Taube et al, 1990b). On two occasions, two head direction cells were recorded at the same time. In both instances, both cells rotated their preferred directions by the same amount.

Evidence suggests that the directional tuning of these cells is primarily determined by sensory input rather than 'dead reckoning' mechanisms

(vestibular inputs or a motor efference copy) (Goodridge and Taube 1995;

Taube and Burton, 1995). This appears in contrast to the situation in the hippocampus, in which both sensory and internal cues play important roles, as discussed in chapter one. This primacy is so strong that head direction cells have been shown to rotate their preferred direction even when the sensory cue is moved in the presence of the animal (Taube et al 1990b). While sensory inputs take precedence, there is evidence that rats use direction cues arising from internal, or "idiothetic " sources when they are exploring novel environments, and that this allows them to maintain a constant global orientation reference over different environments (Taube and Burton, 1995). This is supported by evidence that head direction cells maintain their preferred ______Page 243 direction for several minutes in darkness, subsequently showing a slow rotational drift (Mizumori and Williams, 1993).

Although is has been generally assumed that head direction cells represent a static direction in space, there is now evidence that those in the anterior dorsal thalamus (ADN) have a predictive role (Blair and Sharp, 1995). It was found that the direction at which the cells' firing rates peaked shifted to the left during clockwise head turns, and to the right during counterclockwise turns. This implies that, rather than representing the current direction in which the animal's head is pointed, these cells may be coding for the direction in which the head will be pointed in the near future. A set of averaged tuning curves was constructed using different amounts of temporal shift, and these were compared with the overall average tuning curve, showing that ADN cells yielded the closest match when the curves were shifted ~37ms into the future. This effect was not found in cells of the postsubiculum. As noted above, there is evidence that hippocampal place cells may also be interpreted as predicting the future of the animal, although in this case the time advance is much longer, on the order of 120ms (Muller and Kubie, 1989).

It was first noted in the investigation by Taube et al (1990b) that head direction cells all appear to be linked to one another, in that a rotation in the preferred direction of one cell would be matched by a similar rotation in another cell recorded at the same time. This was confirmed by the observations of Knierim and his colleagues (Knierim et al., 1995), who recorded both head direction and place cells at the same time. Rotating a prominent cue card within the environment caused rotation of both the preferred direction of the head ______Page 244 direction ceils and the place field of the hippocampal cells. The two rotations

were closely coupled, displaying a rotational correlation coefficient of 0 .8 6 .

Not surprisingly, head direction cells have been found to form an essential component of the hippocampal formation spatial navigation system. Taube,

Kesslak and Cotman (Taube et al., 1992) demonstrated that lesions of the postsubiculum caused by either electrolysis or NMDA excitotoxin led to significantly impaired performance on both the radial and water maze tasks.

Animals with both types of lesions showed some improvement over time (this was more noticeable in the water maze task), but performance remained well below that of the controls. In contrast, non-spatial tasks, such as swimming to a prominently labelled platform or a classically-conditioned taste aversion task, were unimpaired.

Given the strong and robust directional response of the PoS cells discussed above, it is important to discover whether anatomical links exist which might provide a route for the transmission of directional information between these cells and the MS/DBB. Below I review the current anatomical literature relating to these areas and propose two possible pathways which may be responsible for this link.

Many of the anatomical areas involved in head direction share reciprocal connections or common input/output sites. The ATN and LDN have reciprocal projections to the PoS (Niimi 1978, Thompson and Robertson,

1987a,b, van Groen and Wyss, 1990a,b, 1995, Shibata, 1993), which also projects to the retrosplenial cortex (van Groen and Wyss 1990b). The anterior thalamic nuclei and dorsal presubiculum also connect to the mammillary nuclei, a ______Page 245 relationship which will be described in more detail below. In the light of the anatomical data, the area most intimately connected with the MS/DBB complex is the postsubiculum, which will be the focus of the following discussion

An issue of some importance is the identity of the postsubiculum itself. Taube and his co-workers repeatedly refer to it as the dorsal presubiculum (Ranck

1985, Taube 1995, Goodridge and Taube 1995), although van Groen and Wyss

(1990a, 1990b) state that they found clear distinctions between the areas they named the pre- and postsubiculum. Specifically, in the postsubiculum there were cell islands in layer II and dark staining for AChE in layer I, the layer III cells were arranged in rows parallel to the pial surface and, in Vogt-silver stained material, there was a dense fibre plexus in layer la and few fibres in layer II. van Groen and Wyss situate the postsubiculum immediately adjacent to the retrosplenial cortex, however, and this corresponds with the areas described as dorsal presubiculum in the studies quoted above. Also, a comparison of figure 2 of Taube et al 1990 with figure 12 of Shibata's paper demonstrating the connection between the dorsal presubiculum and the pars posterior of the medial mammillary nuclei (MP) (Shibata 1989) suggests that the two authors are indeed referring to the same area.

At present, there are two pathways which might serve as a channel for the passage of directional data between the PoS and MS/DBB: a basal complex involving the mammillary bodies and supramammillary nucleus; and a cortical route involving the entorhinal cortex and back-projections from the hippocampus to the septum. ______Page 246 a) mammillary connections:

The PoS projects to the lateral posterior thalamus (Swanson and Cowan, 1977), areas 17 and 18 (Vogt and Miller, 1983), layer I of cingulate and retrosplenial cortex (Swanson and Cowan 1977, Vogt and Miller 1983) and the mammillary nuclear complex, specifically, the lateral mammillary nucleus (LMN) and area

MP (Donovan and Wyss, 1983, Shibata 1989, Namur a et al, 1994).

Area MP subsequently projects to the supramammillary nucleus (SUM)

(Gonzalo-Ruiz et al, 1992), an area lying above the mammillary bodies which projects strongly to the medial septum and nucleus of the diagonal band (Vertes

1992), and is thought to contain the mechanisms which translate activity ascending from the pontine reticular formation (RPO) into rhythmic impulses which are relayed to the MS/DBB complex (Kirk and McNaughton, 1991) and govern the presence and frequency of hippocampal theta rhythm. Both the mammillary bodies and, to a lesser extent, the SUM, project to the anterior thalamic nuclei (Seki and Zyo 1984, Sikes and Vogt 1987, Vertes 1992). Both the

SUM and the MS/DBB complex project to the PoS (van Groen and Wyss

1990b), and the medial septum projects back to the LMN and MP areas

(Meibach and Siegel 1977, Swanson and Cowan 1979, Gonzalo-Ruiz et al 1992).

Both the anterior dorsal (AD) and ventral (AV) thalamic nuclei receive projections from the medial mammillary nucleus and the LMN projects to AD

(Shibata 1992).

Thus the mammillary nuclei (specifically the lateral nucleus and the pars posterior of the medial nuclei) appear to be an area in which directional data

(through their connections with the PoS and ATN) come into contact with the rhythmic impulses which set the frequency of MS/DBB oscillation (through Page 247 the MP-SUM projections), the pathway being PoS->MP->SUM->MS/DBB.

The connections of this area are summarised in figure 9.8.

Functionally, the presence of oscillations in the theta range in cells of the medial mammillary nucleus (MMN) is well established (Komisaruk 1970,

Mignard et al 1987, Alonso and Llinas 1992), furthermore, cells in the MMN have been found to fire bursts in phase with hippocampal theta rhythm (Kocsis et al 1993, Kocsis and Vertes 1994, Bland et al 1995). It is also interesting to note that these cells did not show any significant difference with respect to discharge rates between the theta and non-theta states (Bland et al, 1995), a property they share with the directionally-rhythmic cells described in this paper.

It must, however, be noted that inactivation of the medial septum by procaine has been shown to abolish phasic discharge in the MMN (Kirk et al 1993).

Although this does not negate the possibility that this area plays a role in conveying directional data to the MS/DBB, it does raise implications concerning the manner in which these data become incorporated into the input streams feeding the MS/DBB, particularly that directional data may affect the area's output by modulating the feedback streams from higher areas (Kirk et al.

1994, Kocsis and Vertes 1994, Bland et al 1995). If this route does indeed carry directional data, the hypothesis implies that directionally-responsive cells should be found in the SUM, an area which has been little studied for behavioural correlates. Page 248

Figure 9.8. The anatomical connections between those areas containing cells tuned to head direction (AD, AV and PrS), the mammillary and supramammillary nuclei, and the medial septum. A putative path along which head direction information may be channelled is indicated by the paths shown in bold. The stroke type used to indicate a connection does not reflect the strength of that connection. Abbreviations: AD: antero-dorsal thalamic nucleus; AV: antero-ventral thalamic nucleus; LMN: lateral mammillary nucleus; LS: lateral septum; ML: medial mammillary nucleus, pars lateralis; MM: medial mammillary nucleus, pars medialis; Vln: medial mammillary nucleus, pars medianus; MP: medial mammillary nucleus, pars posterior; MS: medial septum/diagonal band of Broca complex; PrS: dorsal presubiculum (postsubiculum); Sub: subiculum; SuiVI: supramammillary nucleus; TM: tuberomammillary nucleus. (Meibach and Segal, 1977; Thompson and Robertson, 1987a, 1987b; Shibata, 1989, 1992; Gonzalo Ruiz et al., 1992; Léranth et al., 1992; Vertes 1992) ______Page 249 b) entorhinal connections:

The PoS produces a strong projection to the lateral entorhinal cortex and perirhinal cortex (van Groen and Wyss 1990b). van Groen and Wyss noted a distinction between these projections, which they describe terminating in the deep layers (IV-VI) of the lateral entorhinal cortex, and those coming from the presubiculum, which they found to terminate in the superficial layers (II-III) of the medial entorhinal cortex. It is notable that the superficial layers are the primary sites for projections to the hippocampal formation (Steward and

Scoville 1976, Witter et al 1989), while the deep layers are the primary sites for the termination of afferents from the hippocampus and subiculum (Hjorth-

Simonsen 1971, Sorensen and Shipley 1979). This laminar segregation is not absolute, although, as deep layers also provide some efferents while afferent terminations are found in the superficial layers (van Groen et al 1986, Witter et al 1989, Witter 1993). In the light of this and the uncertainty concerning the localisation of the head direction cells in the pre- or postsubicular areas as defined by van Groen and Wyss (1990b), it seems reasonable to accept that efferents from the head direction cells may influence the perforant path projections back to the hippocampus.

The perforant path contains a significant projection to the CA areas of the hippocampal formation (Witter et al 1988) which terminates in the stratum lacunosum-moleculare area of CAl. This projection is segregated, with projections from the lateral entorhinal area terminating on the outer one third of the stratum lacunosum-moleculare, while the medial area projects to the middle one third of this layer (Witter 1989, Witter 1993). ______Page 250

Functionally, the direct effects (ie, separate from activation of the dentate gyrus-CA3-CAl loop) of stimulation of the entorhinal cortex are mixed. While some authors have shown monosynaptic excitatory events (Segal 1972, Fox and

Ranch 1981, Yeckel and Berger 1990), there is strong evidence that the primary effect on the CA areas is to activate interneurons (Kehl and McLennan 1985,

Soltész and Deschênes 1993, Empson and Heinemann 1994). Wliile the exact role of the direct perforant path-CAl link is open to debate (Buzsaki et al 1995), it is clear that entorhinal inputs activate GABAergic cells in the CA areas.

The presence of GABAergic projections from the hippocampal CA areas onto both cholinergic and GABAergic septohippocampal projection cells in the

MS/DBB complex has been well established (Alonso and Kohler 1982, Toth et al 1993). Furthermore, studies examining calbindin D^gj^-containing nonpyramidal neurons have shown not only that they project to the MS/DBB, but also that many of these cells in the CAl area have long vertical dendrites running up to the distal stratum radiatum and stratum lacunosum-moleculare fields, or have dendrites running horizontally within stratum lacunosum- moleculare (Toth and Freund 1992) (it must be pointed out, however, that, while calbindin-positive cells were found to project to the MS/DBB, such cells were all confined to the stratum oriens in this study). Thus GABAergic hippocampo-septal projection neurons may have dendrites in the termination field of the direct perforant path inputs to CAl, which are known to synapse on inhibitory cells.

This provides a second possible pathway for the passage of directional data from the PoS to the MS/DBB: PoS -> entorhinal cortex -> inhibitory cells in CAl ->

MS/DBB (summarised in figure 9,9). ______Page 251

Clearly, both of these pathways are putative, as there is as yet no evidence in either pathway that the cells receiving inputs from the preceding area do indeed project to the next link in the chain. This is an area worthy of further study, and it would be interesting to study the intermediate areas for evidence of directional modulation of cell firing.

As the hippocampal formation is the main target of MS/DBB efferents, we need to consider the effect of directionally-selective rhythmicity on hippocampal pyramidal cells. The firing of hippocampal pyramidal cells defines the animal's location in two ways. The rate of firing partitions oE the place field from the rest of the environment (in that the firing rate of the cell shows a clear distinction between when the animal is in the place field and when it is elsewhere) (O’Keefe, 1979). And the phase of firing of each burst relative to the theta rhythm locates the animal within the place field (O'Keefe and Recce,

1993; Skaggs et al, 1995). This dependency may serve to explain the finding that

MS/DBB blockade results in a dramatic impairment in spatial tasks despite little change in the place-specific firing of hippocampal CAl cells (Mizumori et al, 1989).

Two possible mechanisms exist whereby directionally-modulated changes in the rhythmicity of MS/DBB cells would have an effect on the rhythmic firing of hippocampal cells:

a) The induction of rhythmic behaviour in a neuron not sufficiently inhibited by the rhythmic inhibition from non-varying inputs. Theoretical studies of the interaction between inhibitory inputs (Douglas and Martin 1990) have shown Page 252

osCTbf

pyramrdgg^ll

perforant entorbim- V '‘^ i ^ # - - ï l pa^h

dèrîtàte y y r U S

»> V ^ .A ^ ^ ^ friedial septum/ W & j^onal band of

Figure 9.9. A possible connection route between the postsubiculum and the MS/DBB via the entorhinal cortex and GABAergic cells of CAl. (Alonso and Kohler, 1982; Witter et ah, 1988; van Groen and Wyss, 1990b; Soltész and Deschênes, 1993; Toth and Freund, 1992; Toth et ah, 1993; Witter 1993; Empson and Heinemann, 1994) ______Page 253 that inputs on the soma or dendrite can exert an effect synergistic with inputs to the axon initial segment, allowing the shunting of excitatory currents that would otherwise cause a high firing rate. As GABAa synapses are effective for periods of time as short as 4ms (Hille 1992), it is possible that many inhibitory inputs must be active at the same time in order to shunt sufficient current to control the cell. Thus, variations in the rhythmicity of one of the cells may effectively turn off theta rhythm in the target interneuron.

b) It is possible that cells showing variable rhythmicity act by altering the phase of theta rhythm expressed in the target interneurons. The presence of additional inhibition either just before or just after the onset of the primary rhythmic input may be sufficient to shift the phase of the resultant rhythm in the target cell. Thus, variable rhythmicity may serve to fine-tune the local phase of theta rhythm. Although it is not certain whether GABAergic septo­ hippocampal neurons synapse onto the whole population of hippocampal interneurons or only onto a particular class, it is known that chandelier cells at least have an effect on a large number of pyramidal cells (Li et al 1992) and that only three of them (approximately) contact any one particular pyramidal cell

(Somogyi et al 1983). Thus, single interneurons may govern theta rhythm over a relatively large area of the hippocampus, accounting for the high degree of local coherence found (Bullock et al 1990). This also implies that changes made to the phase of theta rhythm in a small number of interneurons may have effects across a large number of pyramidal cells, possibly in a manner similar to the competitive interactions which have been used in artificial neural networks

(Rumelhart and Zipser, 1986). ______Page 254

To understand the way neural systems may use oscillatory mechanisms, it is necessary to recognise the difference between the theta frequency phase- encoding method described here and the gamma frequency spike-correlatdon method postulated to take place in the visual cortex (Gray and Singer 1989,

Engel et al. 1991) or the spatial amplitude patterns found in olfactory cortex

(Freeman and Skarda 1985). It is notable that both the latter feature oscillations in the gamma band (40-100Hz) and that the oscillations involved are far less regular on recording than the theta rhythm waves found in the hippocampus. As the hippocampus also displays gamma oscillations (Bragin et al 1995), it is possible that multiple coding mechanisms are being superimposed, possibly combining the tasks of pattern completion and spatial navigation (Buzsaki and

Chrobak 1995).

For the theta rhythm to act in the clock-like manner described, it is necessary for the same clock to be present at both the input and output stages (or, at the very least, these two clocks must be related in a predictable manner), so that encoding and decoding may be performed coherently. Although at first this may seem a potential failing, the entorhinal cortex, which serves as the major efferent stage for CAl and the subiculum (van Groen and Wyss 1990) has been found to generate theta rhythm independently (Mitchell and Ranck 1980), and this function is reliant on a functioning MS/DBB (Dickson et al 1995, Jeffery et al

1995). The phase and wave shape of entorhinal theta is also closely related to that found in CAl (Mitchell and Ranck 1980). It is not clear, though, to what extent theta rhythm generation in the entorhinal cortex interacts with that in

CAl, as lesions of the entorhinal cortex can cause changes in the phase and coherence of theta rhythm in CAl (Bragin et al 1995, Buzsaki et al 1995). ______Page 255

At present the purpose that theta phase-encoding might serve is not entirely clear. It may be necessary for efficient information transfer (either increasing the robustness or the information capacity of the channel) or it may be involved in a more subtle form of data processing. Whatever the actual nature of the mechanisms used to encode and decode data in such a system, it is clear that the phase and frequency of the theta rhythm clock plays a vital role in the overall process.

In this light, it is interesting to see that the behaviour of the animal, such as running speed or direction of motion, may alter the firing behaviour of sub­ groups of MS/DBB neurons. Such mechanisms, which alter the clock used to encode data passing out of the hippocampus, may be responsible for determining the context in which neuronal data is used and transmitted to other areas. If so, such context-dependent coding may play an important part in the flexibility of hippocampal function and may have important implications with regard to the role played by the hippocampus in memory. Chapter Ten

Discussion:

Models of septo-hippocampal

function ______Page 257

This thesis challenges the conventional view of septo-hippocampal function, in which rhythmically bursting MS/DBB cells fire in phase, directly entraining oscillations in hippocampal interneurons and pyramidal cells by both rhythmic inhibition (GABAergic projections) and excitation (cholinergic projections)

(Stewart and Fox, 1990; Lee et al., 1994). As an alternative, I offer the following model;

1) MS/DBB cells do not form an homogenous group, even within the

GABAergic and cholinergic types. Rather, as evidenced by the presence of theta-on and theta-off cells within both types, these cells form various subgroups. It is not clear at present whether all these sub-groups project from the MS/DBB, or whether some are committed to local interactions within it.

2) Cholinergic MS/DBB cells are only poorly rhythmic, their rhythmicity possibly reflecting interactions within the internal network of the MS/DBB.

They serve the purpose of altering the function of various hippocampal trans­ membrane channels to allow the production of theta rhythm.

3) GABAergic MS/DBB cells determine the frequency and phase of theta rhythm, but often skip cycles. Hippocampal theta rhythm remains regular despite this behaviour as the hippocampus contains rhythmic generators which are capable of maintaining theta rhythm oscillations over short periods of time.

4) These cells do not fire at one coherent phase, and the final phase of theta rhythm is determined by an interaction between many different cells, in a winner-take-all style.

5) The behaviour of the animal determines different aspects of rhythmicity in different MS/DBB cells. A large subset of rhythmically bursting cells vary the frequency of their bursts with the speed at which the animal runs. Some

MS/DBB cells modify the extent of their rhythmicity based on the direction in which the animal is moving. This variable rhythmicity is one mechanism by ______Page 258 which cells firing at different phases may be given differing weights in determining the actual phase of hippocampal theta rhythm.

Overall, cells in the medial septum/diagonal band of Broca (MS/DBB) display very similar firing characteristics in the awake, behaving state as they do in anaesthetised or restrained preparations. The rhythmic firing can be classified into the same five classes reported by Alonso et al. (1987). The proportion of cells which show phasic modulation without firing rhythmic bursts (type 2) is, however, significantly higher in freely moving animals than in the anaesthetised state. As in the anaesthetised animals studied by Ford et al. (1989) both positive and negative correlations have been found between the firing rate of MS/DBB cells and the presence of hippocampal theta rhythm, although the number of such cells is lower in the awake state.

In addition, it has been shown that, in a manner consistent with the findings of

Matthews and Lee (1991), it is possible to separate MS/DBB cells on the basis of correlates between their extracellular wave shape and cellular physiology.

This is, of course, subject to the caveat that such a simple criterion is bound to produce inaccuracies in the classification of cells and, as noted by Matthews and Lee, there is some overlap between the two groups. Even so, they show interesting differences in terms of firing behaviour, with the bulk of the putative GABAergic cells firing in a rhythmically bursting manner while the putative cholinergic cells are grouped in type 2: non-rhythmic cells which still show a significant phase-lock to hippocampal theta.

An important conclusion to arise from the data reported in this study is that the pacemaker hypothesis of septo-hippocampal function needs to be revised. In ______Page 259 current models of the generation of theta rhythm, hippocampal interneurons

("theta cells") play a vital role (Stewart and Fox, 1990). These cells are the targets of inhibitory MS/DBB cells (Freund and Antal, 1988) and their own inhibitory synapses onto hippocampal pyramidal cells may form the phasic current source observed in the stratum pyramidale of CAl (Brankcack et al,

1993). While there is ample evidence that unimpaired functioning of the

MS/DBB is necessary for the production of hippocampal theta rhythm (as discussed in chapter one), it is now impossible to accept a simple model in which the rhythmic firing of MS/DBB cells directly generates the hippocampal oscillations that are recorded as theta rhythm.

Firstly, this study has shown that rhythmically-bursting MS/DBB cells frequently fail to fire on a regular basis, skipping up to four cycles in between bursts. If these cells are directly responsible for generating theta rhythm in the hippocampus, then it would be expected that such skipped cycles would lead to substantial and random fluctuations in the power of theta rhythm. Yet such fluctuations are seen in neither extracellular or intracellular (Ylinen et al.,

1995b) recordings of theta rhythm. This suggests that, while cells of the

MS/DBB may determine the presence and frequency of theta rhythm, they are not directly responsible for its actual generation. Thus, the Stewart and Fox model, while correct in its basic anatomical details, needs to be revised with respect to the functions of the various elements which comprise the septo- hippocampal system.

Secondly, it has been conclusively shown that the cells of the MS/DBB have no overall preferential phase, whether they are grouped by firing behaviour or putative cell type (both GABAergic and cholinergic). Although the uniform ______Page 260 phase distribution of MS/DBB cells has been conclusively demonstrated in anaesthetised animals (Gaztelu and Buno, 1982, Stewart and Fox, 1989c), it was not known whether this changed to a more coherent pattern when the animal was allowed to move freely. As the data provided in this study shows, this hypothesis can be rejected. While there is considerable variation between cells, each cell maintains its preferential phase over time, indicating that this is a stable aspect of the cells' function.

In order to interpret these data, I will first present such a revised model of septo-hippocampal function, based on the separation of roles into those elements that potentiate theta rhythm, and those that control it. This model will be used to explain the behavioural findings of this study.

The production of theta rhythm

Oscillatory activity can be induced in the isolated hippocampus through simple stimulation or pharmacological manipulations (Konopacki et al, 1987; Cobb et al,

1995), indicating that rhythmic inputs from the MS/DBB are not necessary for the hippocampus to generate oscillatory activity. Such oscillations have also been induced in vivo in the deafferented hippocampus by application of both carbachol and bicuculline (Colom et al., 1991), although the EEC rhythm produced by this method differs from hippocampal theta in both frequency and waveform.

WTiilst there is some evidence that the cholinergic inputs may be rhythmic

(Stewart and Fox, 1989), pharmacological studies have shown that cholinergic effects on hippocampal pyramidal cells have a time constant on the order of ______Page 261 several seconds (Benardo and Prince, 1982, Pitler and Alger, 1992), which is too long to allow them to induce BHz rhythmic activity directly. Furthermore, theta rhythm can be produced (albeit at a lower amplitude) even after severe lesions of the MS/DBB cholinergic cells (Lee et al, 1994).

These data suggest that any rhythmic activity in cholinergic MS/DBB cells is incidental to their primary role, which is to potentiate the generation of hippocampal theta by inducing a variety of mechanisms (discussed below) which may form the basis for the oscillation system intrinsic to the hippocampus. These oscillations are then regulated by the action of MS/DBB

GABAergic cells on inhibitory interneurons (theta cells) in the hippocampus

(Freund and Antal, 1988). The presence of highly rhythmic MS/DBB cells which regularly fail to fire on each cycle suggests that these cells do not directly gate the activity of their targets, as one would then expect to observe similar irregularities in the output of hippocampal theta cells, in contradiction to the high degree of regularity that has been observed (Ylinen et al, 1995b).

The first step in a discussion of the hippocampal theta rhythm must be to decide what, exactly, this phenomenon consists of. The primary way in which theta rhythm is observed is by extracellular electrodes, and this has been the focus of much of the research. On the other hand, modern slice technology is revealing the intracellular correlates of theta rhythm and is providing us with an understanding of the fundamental mechanics of the neurons involved in its generation. If theta rhythm does play an important part in the regulation of firing behaviour in the hippocampus, then its actions must be exerted at an intracellular level. It seems logical that this is the arena in which we must seek to define theta rhythm and its function. ______Page 262

In considering the systems involved with the production of theta rhythm, it is useful to divide the analysis into three parts:

a) Potentiation: the mechanisms which are necessary to enable generation to occur. These consist primarily of modulatory systems which alter the resting potential and activate or block ionic membrane channels.

b) Generation: the mechanisms which produce the actual membrane potential differences which are measured as theta rhythm. Possible sources for this which will be discussed are synaptic sites and intrinsic membrane channels.

c) Regulation: the mechanisms which control the frequency and phase of theta rhythm. These are distinct from the generation mechanisms and, as will be shown, either have their origin in, or are relayed through the MS/DBB

A.) Mechanisms of theta rhythm potentiation

As mentioned previously, it is well established that theta rhythm can be induced in the dentate gyrus and CA3 without external rhythmic inputs by bathing slices of hippocampal tissue in carbachol, a cholinergic-receptor agonist, and that this action is antagonised by atropine, a muscarinic-receptor antagonist

(Konopacki et al., 1987, 1992; MacVicar and Tse, 19890). This reveals two things: firstly, that the hippocampus has intrinsic mechanisms which can produce rhythmic oscillations, and secondly, that cholinergic stimulation is sufficient to turn these mechanisms on. Page 263

Acetylcholine (ACh) has a broad range of effects on the hippocampal slice preparation. Nevertheless, we can group the effects of ACh into two broad categories:

1) Effects on synaptic transmission at non-cholinergic sites.

The physiological importance of this in vivo is hard to determine as it would require significant leakage of ACh from the synaptic cleft. Excitatory transmission has been reported to show both inhibition and enhancement in the presence of cholinergic agonists (Krnjevic and Ropert, 1982; Herreras et al.,

1988). And while inhibitory synaptic currents are diminished in the presence of cholinergic agonists, the frequency of IPSPs is increased (Pitler and Alger,

1992). Cholinergic agents have also been shown to augment long-term potentiation (LTP), the process in which synaptic transmission is enhanced by prior activity in the pre- and post-synaptic cells at the Schaffer collateral-CAl locus (Blitzer et al., 1990; Huerta and Lisman, 1993), although they depress

LTP in the CA3 area (Williams and Johnston, 1988).

2) Effects on intrinsic membrane conductances.

ACh has effects on a variety of trans-membrane ionic channels, the net effect generally being to cause stimulation of the target cell . Although relatively little is known of interneuron membrane conductances, pyramidal cell conductances have been studied in depth. Cholinergic agonists affect many of the channels present in hippocampal pyramidal cells. They cause mild depolarisation due to blockade of a potassium conductance open at rest (Madison et al., 1987; Benson et al., 1988). They also cause blockade of and I^yp, two conductances involved in accommodation to prolonged depolarisation and the after­ hyperpolarisation response (Benardo and Prince, 1982; Halliwell and Adams,

1982). Cholinergics also inhibit the transient channel responsible for altering ______Page 264 the time of onset of action potentials (Nakajima et ah, 1986). The various subtypes of calcium ion channels present in hippocampal cells are also affected by the administration of cholinergic agonists. Carbachol causes an increase in T channel activity, the low-threshold Ca^^ channel responsible for low- frequency rhythmic bursts in the thalamus Qahnsen and Llinas, 1984). It has no effect on the N channel and suppresses the high-conductance L channel

(Gahwiler and Brown, 1987; Fisher and Johnston, 1990).

A key feature to be noted concerning these effects is that they all have a relatively long latency, typically on the order of over a second (Benardo and

Prince, 1982; Cole and Nicoll, 1984; Stewart et al., 1992). A few neocortical neurons have been shown to respond to acetylcholine with a much shorter latency, in one case producing an action potential only 25ms after the application of ACh (McCormack and Prince, 1986), although such cells were extremely rare (8/255) and the average latency was on the order of Is. Clearly, then, these latencies are too long to transmit a rhythm in the theta range, even though, as demonstrated in this thesis, there is good evidence that some cholinergic cells are both phase-locked to theta rhythm and fire in a rhythmically-bursting manner.

It is now possible to destroy basal forebrain cholinergic neurons selectively using the immunotoxin IgG-192 saporin. This causes a dramatic reduction in the power of hippocampal theta rhythm, but fails to eradicate it completely (Lee et al., 1994), indicating that cholinergic inputs augment the amplitude of hippocampal theta oscillations rather than generate the oscillations themselves. ______Page 265

Consonant with this, Stewart et al. (1992) demonstrated that blockade of cholinergic effects in the hippocampus of urethane-anaesthetised animals led to a decrease in the firing rate of theta cells, though they remained phase-locked to extracellular theta. It also caused the elimination of the burst mode of firing of complex-spike cells (presumed to be pyramidal cells). It seems, therefore, that cholinergic transmission along the septo-hippocampal pathway plays a role in altering intracellular and/or synaptic conditions such that theta oscillations may be generated, without directly transmitting the theta rhythm itself.

B.) Mechanisms of theta rhythm generation

Ever since its discovery (Green and Arduini, 1954), theta rhythm has been defined as a sinusoidal extracellular signal. The introduction of intracellular recording, though, has allowed the exploration of the intracellular correlates of theta rhythm, as described in the introduction. It is, presumably, at this site

(within the pyramidal cells) that theta rhythm has its effect.

Given that the intra- and extracellular theta oscillations must be the result of current flow through pyramidal cell transmembrane channels, we now have two alternatives for the site of generation of the electrical potentials which constitute theta rhythm. The first is at inhibitory or excitatory synapses, in which case the rhythmicity of the current flows would be imposed by neural elements external to the pyramidal cell. The second is via a subset of the voltage-gated channels of the pyramidal cell, many different varieties of which are found in the cell membrane of pyramidal neurons (Brown et al., 1990). In this second case, the rhythmicity would be the result of interactions between elements internal to the pyramidal cell. ______Page 266

1) The potentials are generated at synapses of either pyramidal or inhibitory cells

This hypothesis is attractive in light of the rhythmic synaptic input known to come from the medial septum. Laminar analysis of the field potentials, current sources and phase shift of theta rhythm in CAl has been interpreted as the product of two rhythmic current sources, one inhibitory and located in the stratum oriens, the other excitatory, shifted by ~45° and located in the distal dendrites of stratum lacunosum-moleculare (Leung 1984). More recent current source density analysis has revealed a more complex picture

(Buzsaki et al., 1986; Brankack et al., 1993) with multiple sources and sinks dominated by a current source in the CAl stratum pyramidale (which may correspond to the site of inhibitory synapses from interneurons) and a current sink in the hippocampal fissure, which the authors ascribed to excitatory inputs from the lateral entorhinal cortex.

While there are excitatory cells of the Cornu Ammonis and dentate gyrus which discharge rhythmically, their average firing rate of 1-2 Hz would seem too low to sustain oscillations at theta frequencies without careful temporal interleaving

(although they can fire at up to 60Hz when the animal is in the cell’s place field)

(Buzsaki et ah, 1983; Fox et ah, 1986; Barnes et ah, 1990). There are rhythmic inputs from entorhinal cortex to all areas of the hippocampal formation (Stewart et ah, 1992), which are far more likely candidates for a source of excitatory drive, particularly as the entorhinal projections terminate in the stratum lacunosum-moleculare (Witter et ah, 1988). ______Page 267

It should be noted, however, that the two reports of the effects of entorhinal

cortex lesions on theta rhythm have produced contradictory results:

Vanderwolf and Leung (1983) reported a loss of theta power in the region of

dorsal CAl and normal power at the hippocampal fissure, whereas the major

effect of entorhinal lesions in the study of Ylinen et al., (1995b) was the

elimination of the power peak at the fissure. Also, there is still some debate as to

the degree to which entorhinal projections to CAl excite inhibitory or

excitatory neurons in the stratum lacunosum-moleculare area (Yeckel and

Berger, 1990; Empson and Heinemann, 1994; Levy et al., 1995; Soltész, 1995).

Current evidence gives more support to the thesis that the primary generative

sites for extracellular theta rhythm potentials are the currents occurring at

inhibitory synapses as multiple studies have shown that intracellular

membrane potential oscillations (MPOs) are related to rhythmic IPSPs (Leung

and Yim, 1986; Soltész and Deschênes, 1993; Ylinen et al. 1995b). Synaptic

connections from the inhibitory inputs from the MS/DBB have been found

exclusively on inhibitory interneurons in the hippocampus (Freund and Antal,

1988; Stewart and Fox, 1990). These hippocampal interneurons then synapse on

hippocampal pyramidal cells (HPCs) (Lacaille et al., 1987). The rhythmic

inhibition of the interneurons thus causes rhythmic disinhibition of the HPCs, which may manifest as theta rhythm.

In this arrangement there are, of course, two inhibitory synaptic events per

theta cycle, which are out of phase by roughly 180°. Furthermore, both of these

events will appear as current sources in the extracellular milieu, yet only one

current source has been recorded in the stratum pyramidale (Brankack, 1993).

This may be explained if the inhibitory synapses arising from MS/DBB inputs ______Page 268 are markedly fewer than the inhibitory synapses onto HPCs, and thus their out- of-phase signal is swamped by the larger current source provided by the latter.

This is an acceptable proposition, given that hippocampal interneurons synapse with a large number of HPCs (axo-axonic chandelier cells contact approximately

1200 cells, making 2-15 synapses with each (Li et ah, 1992), while basket cells connect to 500-2500 post-synaptic cells (Sik et al., 1995; Miles et al., 1996)), yet may be controlled by a much smaller number of inhibitory synapses from the

MS/DBB.

There is still, however, a piece of evidence which does not fit with this hypothesis. Ylinen et al (1995b) found that intracellular theta-related MPOs and accompanying bursts of action potentials were far more regular and robust in interneurons than those found in HPCs. Furthermore, they found that these oscillations and action potential bursts would persist when extracellular theta lapsed into transient irregularity. As these interneurons are, therefore, producing rhythmic bursts at a time when corresponding extracellular theta oscillations are not present, it would appear that an extra factor, apart from the rhythmic operation of inhibitory synapses from the hippocampal interneurons, is necessary to generate the large extracellular theta potentials.

3) The potentials are generated by active membrane currents in the pyramidal cells and/or interneurons.

As discussed previously, hippocampal tissue can, when bathed in carbachol, create oscillations similar to theta rhythm without the need for external synaptic input (Konopacki et al., 1987). It was not clear for many years. ______Page 269 however, whether cholinergic activation was primarily responsible for this effect or whether it was due to the secondary activation of other processes.

Recent intracellular recordings from CAl pyramidal cells have shown that these cells can spontaneously produce rhythmic oscillations at theta frequencies when presented with suitable stimuli. In one case, when bathed in a Ca^'^-free solution containing tetraethylammonium and Co^"^ to block the

and Ca-^ channels, both sustained (200ms) depolarisation and a brief (10ms) pulse caused the cell to fire a spike followed by a depolarised plateau potential, upon which theta-frequency oscillations were superimposed (Garcia-Munoz et al., 1993) (it should be noted that a mild but sustained depolarisation is one of the primary effects of cholinergic agonists). These oscillations, and the plateau potential, were found to be sensitive to tetrodotoxin, and therefore probably dependent on a persistent Na'^ current similar to the one shown to exist in hippocampal pyramidal cells (French et ah, 1990). An oscillatory process dependent on such a persistent Na"^ channel is known to be present in the thalamus (Limas et ah, 1984).

In a separate experiment, it was found that, after minimal stimulation of an interneuron to produce a single IPSP in a synaptically-connected pyramidal cell, the pyramidal cell exhibited a rhythmic rebound depolarisation at theta frequency which was sustained for two to three cycles (Cobb et ah, 1995).

These data suggest that at least some of the rhythmic current flows seen in laminar analysis (Brankack et ah, 1993) may be the result of rhythmic oscillations in non-synaptic trans-membrane ion channels. ______Page 270

C) Mechanisms of theta rhythm regulation

The two prior sections have involved the systems concerned with setting up the hippocampus and then generating the current flows recorded as extracellular theta rhythm. As discussed in the introduction, however, there is evidence that theta rhythm acts as a clock providing the basis for a temporal code of the neuronal responses. In light of this, the most important factors to consider are those which control the frequency and phase of theta rhythm, as these are the elements that have an impact on the role it plays in the coding system.

As I have shown in chapter 6, although cholinergic cells do display rhythmic properties, the majority of them are poorly rhythmic, and it is, thus, unlikely that the cholinergic projections are responsible for controlling theta rhythm on a cycle-by-cycle basis. Instead, I would argue that the function of cholinergic release is, as discussed in section A, to modify the operation of various ligand- and voltage-gated channels in order to allow them to produce theta rhythm oscillations. The GABAergic cells then provide the rhythmic stimulus that leads directly to the generation of these oscillations as well as determining their phase and frequency.

As noted earlier, however, the inputs from the GABAergic cells frequently fail to fire on every burst. As the pyramidal cells, at least, are capable of independently maintaining oscillations over two to three cycles (Garcia-Muhoz et al., 1993; Cobb et al., 1995), it may be that regular pacing inputs from the

MS/DBB are not necessary. Thus the role of the GABAergic MS/DBB inputs is more strictly that of a regulator, correcting the frequency and phase of the theta ______Page 271 rhythm oscillations in the pyramidal cells from time to time, rather than directly driving them.

This leads to the question of whether the frequency and phase of theta rhythm carry any particular behavioural significance. As discussed in the introduction, there is good evidence that the hippocampus employs a form of temporal coding in which the spikes fired at different phases of theta represent different subsections of the cell’s place field. In the work presented here, there is evidence of correlation between the frequency of burst firing and the animal’s speed, and, in a small number of cells, a correlation between the degree of rhythmicity of the cell and the direction in which the direction in which the animal is running. The means by which this may affect the rhythmic oscillation of hippocampal cells was discussed in detail in chapter nine. ______Page 272

C onclusion

Further experiments are necessary to produce a comprehensive model of septohippocampal function. I have suggested that hippocampal interneurons are the site at which the varying rhythmic inputs from the MS/DBB are integrated and the final phase of hippocampal theta is determined. This needs to be tested by intracellular examination of inhibitory synaptic integration in interneurons in vivo during theta states. It is also important to gain a better understanding of the functional relations between the different cell populations of the MS/DBB.

As shown here, the cholinergic population in the freely moving rat exhibits a firing behaviour significantly different to that found in GABAergic cells. These data, founded on an analysis of extracellular wave forms, need to be confirmed, however. This may be performed by analysing the behaviour of MS/DBB cells after the cholinergic populations has been eliminated by the immunotoxin 192-

IgG saporin (Lee et al., 1994).

This thesis has shown that the current models of septo-hippocampal function need to be revised. Rhythmic activity in the septum is too incoherent and unstable to drive hippocampal theta rhythm directly. Instead, this thesis has shown that the MS/DBB provides a controlling influence which is modulated by aspects of behaviour such as the speed at which the rat runs and the direction in which it is heading. While the behaviour of MS/DBB cells in the awake, freely moving animal shows the same general patterns as seen in anaesthetised preparations, behavioural correlates must be considered in subsequent investigation of MS/DBB function. ______Page 273

These mechanisms may allow the MS/DBB to set the context in which the information from hippocampal pyramidal cells is coded and decoded. This implies that the temporal coding mechanism used by the hippocampus uses the theta rhythm clock as a means of adaptively formatting the codes. This would, presumably, allow the positional responses of hippocampal pyramidal neurons to be robust against changes in the animal’s behaviour. Further understanding of this will, however, need to be founded on additional study of the decoding of hippocampal representations. References ______Page 275

E.D. Adrian, The impulses produced by sensory nerve endings. Part 1. J. Physiol. 61:49-72, 1926

E.D. Adrian, The Basis of Sensation. Christophers, London, 1928

E.D. Adrian, The electrical activity of the mammalian olfactory bulb. Electroenceph. Clin. Neurophysiol. 2:377-388, 1950

E.D. Adrian and Y. Zotterman, the impulses produced bv sensory nerve endings. Part 2. J, Physiol. 61:151-171, 1926

J.P. Aggleton, A.B. Keith and A. Sahgal, Both fornix and anterior thalamic, hot not mammillary, lesions disrupt delayed non-matching-to-position memory in rats. Behav. Brain res. 44:151-161, 1991

Y. Aika, J.Q. Ren, K. Kosaka and T. Kosaka, Quantitative analysis of GABA-like immunoreactive and parvalbumin-containing neurons in the CAl region of the rat hippocampus using a stereological method, the disector. Exp. Brain Res. 99:267-276, 1994

G.V. Allen and D.A. Hopkins, Mammillary body in the rat: Topography and synaptology of projections from the subicular complex, prefrontal cortex, and midbrain tegmentum. J. Comp. Neurol. 28611-336, 1989

A. Alonso and C. Kohler, Evidence for separate projections of hippocampal pyramidal and non-pyramidal neurons to different parts of the septum in the rat brain. Neurosci. Lett. 31:209-214, 1982

A. Alonso and R.R. Llinas, Electrophysiology of the mammillary complex in vitro. II Medial mammillary neurons. J. Neurophysiol 68:1321-1331, 1992

A. Alonso, J.M. Gaztelu, W. Buho and E. Garcia-Austt, Cross-correlation analysis of septohippocampal neurons during theta-rhythm. Brain Res. 413:135-146, 1987

M. Alreja, Excitatory actions of serotonin on GABAergic neurons of the medial septum and diagonal band of Broca. Synapse 22:15-27, 1996.

D.G. Amaral, A Golgi study of cell types in the hilar region of the hippocampus in the rat. J. Comp. Neurol. 182:851-914, 1978

D.G. Amaral and J. Kurz, An analysis of the origins of the cholinergic and noncholinergic septal projections to the hippocampal formation of the rat. J. Comp. Neurol. 240:37-59, 1985

D.G. Amaral and M.P. Witter, The three-dimensional organization of the hippocampal formation: A review of anatomical data. Neurosci. 31:571-591, 1989

P. Andersen, Synaptic integration in hippocampal CAl pyramids. Prog. Brain res. 83:215-222,1990

P. Andersen, V.P. Bliss and K.K. Skrede, Lamellar organisation of hippocampal excitatory pathways. Exp. Brain Res. 13:222-238, 1971

G. Apostol and O.D. Creutzfeldt, Crosscorrelation between the activity of septal units and hippocampal EEG during arousal. Brain Res. 67:65-75, 1974 ______Page 276

D.E.A.T. Arnolds, F.H. Lopes da Silva, J.W. Aitink and A. Kamp, Hippocampal EEG and behaviour in dog. I. Hippocampal EEG correlates of gross motor behaviour. Electroenceph. Clin. Neurophysiol. 46:552-570, 1979a

D.E.A.T. Arnolds, F.H. Lopes da Silva, J.W. Aitink and A. Kamp, Hippocampal EEG and behaviour in dog. II. Hippocampal EEG correlates with elementary motor acts. Electroenceph. Clin. Neurophysiol. 46:571-580, 1979b

S.Y. Assaf and JJ. Miller, The role of a raphe serotonin system in the control of septal unit activity and hippocampal desynchronisation. Neuroscience 3:539-550, 1978

I.P. Bagnoli, A. Beaudet, M. Stella and M. Cuenod, Selective retrograde labelling of cholinergic neurons with (3H)choline. J. Neurosci. 7:691-695, 1981

R.H. Baisden, M.L. Woodruff and D.B. Hoover, Cholinergic and non-cholinergic septo-hippocampal projections: A double-label horseradish peroxidase- acetylcholinesterase study in the rabbit. Brain Research 290:146-151, 1984

C.A. Barnes, B.L. McNaughton, S.J.Y. Mizumori, B.W. Leonard and L.-H. Lin, Comparison of spatial and temporal characteristics of neuronal activity in sequential stages of hippocampal processing. Prog. Brain Res. 83:287-300, 1990

C. Barrenechea, M. Pedemonte, A. Nunez and E. Garcia-Austt, In vivo intracellular recordings of medial septal and diagonal band of Broca neurons: relationships with theta rhythm. Exp. Brain Res. 103:31-40, 1995

H.B. Barlow, Single units and sensation: a neuron doctrine for perceptual psychology. Perception 1:371-394, 1972

L.S. Benardo and D.A. Prince, Cholinergic excitation of mammalian hippocampal pyramidal cells. Brain Res. 249:315-331, 1982

L.S. Benardo, L.M. Masukawa and D.A. Prince, Electrophysiology of isolated hippocampal pyramidal dendrites. J. Neurosci. 11:1614-1622, 1982

D.M. Benson, R.D. Blitzer and E.M. Landau, An analysis of the depolarisation produced in guinea-pig hippocampus by cholinergic receptor stimulation. J. Physiol. 404:479-496, 1988

T.W. Berger, P.C. Rinaldi, D.J. Weisz and R.F. Thompson, Single-unit analysis of different hippocampal cell types during classical conditioning of rabbit nictitating membrane response. J. Neurophysiol. 50:1197-1219, 1983

P.J. Best and L.T. Thompson, Persistence, reticence, and opportunism of place-held activity in hippocampal neurons. Psychobiology 17:236-246, 1989

W. Bialek and F. Reike, Reliability and information-transmission in spiking neurons. Trends Neurosci. 15:428-434, 1992

W. Bialek, F. Reike, R.R.D. van Steveninck and D. Warland, Reading a neural code. Science 252:1854-1857, 1991

R. Biegler and R.G.M. Morris, Landmark stability is a prerequisite for spatial but not discrimination learning. Nature, 361:631-633, 1993

T.W. Blackstad, Commissural connections of the hippocampal region in the rat, with special reference to their mode of termination. J.Comp.Neurol. 105:417-537, 1956 ______Page 277

H.T, Blair and P.E. Sharp, Anticipatory head direction signals in anterior thalamus: Evidence for a thalamocortical circuit that integrates angular head motion to compute head direction. J. neurosci. 15:6260-6270, 1995

B.H. Bland, The physiology and pharmacology of hippocampal formation theta rhythms. Prog. Neurobiol. 26:1-54, 1986

B.H. Bland and C.H. Vanderwolf, Diencephalic and hippocampal mechanisms of motor activity in the rat: Effects of posterior hypothalamic stimulation on behaviour and hippocampal slow wave activity. Brain Res. 43:67-88, 1972

B.H. Bland, P. Andersen and T. Gaines, Two generators of hippocampal theta activity in rabbits. Brain Res. 94:199-218, 1975

B.H. Bland, P. Andersen, T. Gaines and O. Sveen, Automated analysis of rhythmicity of physiologically-identified hippocampal formation neurons. Exp. Brain Res. 38:205- 219, 1980

B.H. Bland, M.G. Seto and C.J. Rowntree, The relation of multiple hippocampal theta cell discharge rates to slow wave theta frequency. Physiol. Behav. 31:111-117, 1983

B.H. Bland, M.G. Seto, B.R. Sinclair and S.M. Fraser, The pharmacology of hippocampal theta cells: Evidence that the sensory processing correlate is cholinergic. Brain Res. 299:121-131, 1984

B.H. Bland, L.V. Colom, J. Konopacki and S. Roth, Intracellular records of carbachol- induced theta rhythm in hippocampal slices. Brain Res. 447:364-368, 1988

B.H. Bland, J. Konopacki, I.J. Kirk, S.O. Oddie and C.T. Dickson: Discharge patterns of hippocampal theta-related cells in the caudal diencephalon of the urethan- anaesthetised rat. J. Neurophysiol. 74:322-333, 1995

S.K. Bland and B.H. Bland, Medial septal modulation of hippocampal theta cell discharges. Brain Res. 375:102-116, 1986

R.D. Blitzer; O. Gil and E.M. Landau, Cholinergic stimulation enhances long-term potentiation in the CAl region of the hippocampus. Neurosci. Lett. 119:207-210, 1990

J.J. Boitano,C.J.P. Dokla, P. Mulinski, S. Misikonis and T. Kaluzynski, Effects of hippocampectomy in an incremental-step DRL paradigm. Physiol. And Behav. 25:273- 278, 1980

B.D. Boss, G.M. Peterson and W.M. Cowan, On the numbers of neurons in the dentate gyrus of the rat. Brain Res. 338:144-150, 1985

B.D. Boss, K. Turlejski, B.B. Stanfield and W.M. Cowan, On the numbers of neurons in fields CAl and CA3 of the hippocampus of Sprague-Dawley and Wistar rats. Brain Res. 406:280-287, 1987

E. Bostock, J. taube and R.U. Muller, The effects of head orientation on the firing of hippocampal place cells. Soc. Neurosci. Abs. 13:127, 1988

A. Bragin, G. Jando, Z. Hetke, K. Wise and G. Buzsaki, Gamma (40-100Hz) oscillation in the Hippocampus of the behaving rat. J. Neurosci. 15:47-60, 1995

J. Brankack, M. Stewart and S.E Fox, Current source density analysis of the hippocampal theta rhythm: associated sustained potentials and candidate synaptic generators. Brain Res. 615: 310-327, 1993 ______Page 278

H.R. Brashear, L. Zaborszky and L. Heimer, Distribution of GABAergic and cholinergic neurons in the rat diagonal band. Neuroscience 17:439-451, 1986

E.S. Brazhnik and O.S. Vinogradova, Control of the neuronal rhythmic bursts in the septal pacemaker of theta-rhythm: Effects of anaesthetic and anticholinergic drugs. Brain Res. 380:94-106, 1986

E.S. Brazhnik and O.S. Vinogradova, Modulation of the afferent input to the septal neurons by cholinergic drugs. Brain Research 451:1-12, 1988

C.R. Breese, R.E. Hampson and S.A. Deadwyler, Hippocampal place cells: Stereotypy and plasticity. J. Neurosci. 9:1097-1111, 1989

J.D. Brioni, M.W. Decker, L.P. Gamboa, I. Izquierdo and J.L. McCaugh, Muscimol injections in the medial septum impair spatial learning. Brain Res. 522:227-234, 1990

D.A. Brown, B.H. Gahwiler, W.H. Griffith and J.V. Halliwell, Membrane currents in hippocampal neurons, Prog. Brain Res. 83:141-160, 1990

T.H. Brown and D. Johnston, Voltage-clamp analysis of mossy fiber synaptic input to hippocampal neurons. J. Neurophysiol. 50:487-507, 1983

T.H. Brown and A.M. Zador, Hippocampus. In G.M. Shepherd (ed) The Synaptic Organization of the Brain, Oxford University Press, New York, 1990

T.H. Bullock, G. Buzsaki and M.G. McClune, Coherence of compound field potentials reveals discontinuities in the CAl-subiculum of the hippocampus in freely- moving rats. Neuroscience 38:609-619, 1990

W. Buno, J.L. Garcia-Sanchez and E. Garcia-Austt, Reset of Hippocampal rhythmical activities by afferent stimulation. Brain Res. Bull. 3: 21-28, 1978

M. Bunsey and H. Eichenbaum, Selective damage to the hippocampal region blocks long-term retention of a natural and nonspatial stimulus-stimulus association. Hippocampus 5:546-556, 1995

M. Bunsey and H. Eichenbaum, Conservation of hippocampal memory function in rats and humans. Nature, 379:255-257, 1996

E.E. Burkett and B.N. Bunnell, Septal lesions and the retention of DRL performance in the rat. J. Comp. Physiol. Psych. 62:468-471, 1966

G. Buzsaki and J.J. Chrobak, Temporal structure in spatially organized neuronal ensembles: a role for interneuronal networks. Curr. Opin. Neurobiol. 5:504-510, 1995

G. Buzsaki and E. Eidelberg, Phase relations of hippocampal projections cells and interneurons to theta activity in the anaesthetized rat. Brain Res. 266:334-339, 1983

G. Buzsaki, J. Czopf, I Kondakor and I. Kellényi, 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, 1986

G. Buzsaki, Z. Horvath, R. Urioste, J. Hetke and K. Wise, High-frequency network oscillation in the hippocampus. Science 256:1025-1027, 1992

G. Buzsaki. L.S. Leung and C.H. Vanderwolf. Cellular bases of hippocampal EEG in the behaving rat. Brain Res. Rev. 6:139-171, 1983 ______Page 279

G. Buzsaki, M, Penttonen, A. Bragin, Z. Nadasy and J.J. Chrobak, Possible Physiological role of the perforant path-CAl projection. Hippocampus 5:141-146, 1995

S. Ramon Y. Cajal, Studies on the human cerebral cortex IV: Structure of the olfactory cerebral cortex of man and mammals. Trabajos del Laboratorio de Investigaciones Biolôgicas de la Universidad de Madrid, 1901-1902. Trans. J. DeFelipe and E.G. Jones, Cajal on the Cereb7-al Cortex Oxford University Press, New York, 1988

A. Cattaneo, L. Maffei and C. Morrone, Patterns in the discharge of simple and complex visual cortical cells. Proc. R. Soc. Lond. B. 212:279-297, 1981

J.P. Chandler and K.A. Crutcher, The septohippocampal projection in the rat: an electron microscopic horseradish peroxidase study. Neuroscience 10:685-696, 1983

L.L. Chen, L.-H. Lin, E.J. Green, C.A. Barnes and B.L. McNaughton, Head- direction cells in the rat posterior cortex. I. Anatomical distribution and behavioural modulation. Exp. Brain Res. 101:8-23, 1994

E.P. Christian and E.E. Dudeck, Electrophysiological evidence from glutamate microapplications for local excitatory circuits in the CAl area of rat hippocampal slices. J. Neurophysiol. 59:110-123, 1988

J.J. Chrobak and T.C. Napier, Intraseptal administration of bicuculline produces working memory impairments in the rat. Behav. Neur. biol. 55:247-254, 1991

J.J. Chrobak and T.C. Napier, Antagonism of GABAergic transmission within the septum disrupts working/episodic memory in the rat. Neuroscience 47:833-841, 1992

J.J. Chrobak and G. Buzsaki, High-frequency oscillation in the output networks of the hippocampal-entorhinal axis of the freely-behaving rat. J, Neurosci. 16:3056-3066, 1996

J.J. Chrobak, R.W. Stackman and T.J. walsh, Intraseptal administration of muscimol produces dose-dependent memory impairments in the rat. Behav. Neur. Biol. 52:357- 369, 1989

S.R. Cobb, E.H. Buhl, K. Halasy, O. Paulsen and P. Somogyi, Synchronisation of neuronal activity in hippocampus by individual GABAergic interneurons. Nature, 378:75-78, 1995

N.J. Cohen and H. Eichenbaum, Memory, amnesia, and the hippocampal system. MIT Press, Cambridge, Mass., 1993

N.J. Cohen and L.R. Squire, Preserved learning and retention of pattern analysing skill in amnesia: Dissociation of knowing how and knowing that. Science 210:207-209, 1980

A.E. Cole and R.A.Nicoll, The pharmacology of cholinergic excitatory responses in hippocampal pyramidal cells. Brain Res. 305:283-290, 1984

L.V. Colom and B.H. Bland, State-dependent spike train dynamics of hippocampal formation neurons: evidence for theta-on and theta-off cells. Brain Res. 422:277-286, 1987

L.V. Colom, R.D. Ford and B.H. Bland, Hippocampal formation neurons code the level of activation of the cholinergic septohippocampal pathway. Brain Res. 410:12-20, 1987 ______Page 280

L.V. Colom, S. Nassif-Caudareila, C.T. Dickson, J.W. Smythe and B.H. Bland, In vivo intrahippocampal microinfusion of carbachol and bicuculline induces theta-like oscillations in the septally deafferented hippocampus. Hippocampus 1:381-390, 1991

J.S. Coombs, J.C. Eccles and P. Fatt, the electrical properties of the motorneurone membrane. J. Physiol. 130:396-413, 1955

S. Corkin, Lasting consequences of bilateral medial temporal lobectomy: Clinical course and experimental findings in H.M.. Seminars in Neurology 4:249-259, 1984

K.A. Crutcher, R. Madison and J.N. Davis, A study of the rat septohippocampal pathway using anterograde transport of horseradish-peroxidase. Neuroscience 6:1961- 1973,1981

K.A. Crutcher, K.P. Kesner and J.M. Novak, Medial septal lesions, radial arm maze performance, and sympathetic sprouting - A study of recovery of function. Brain Res. 262:91-98, 1983

J. Day, G. Damsma and H.C. Fibiger, Cholinergic activity in the rat hippocampus, cortex and striatum correlates with locomotor activity: an in vivo microdialysis study. Pharmacol. Biochem. Behav. 38:723-729, 1991

M.W. Decker, R.J. Radek, M.J. Majchrzak and D.J. Anderson, Differential effects of medial septal lesions on spatial memory tasks. Psychobiology 20:9-17, 1992

M.K. Donovan and J.M. Wyss, Evidence for some colateralization between cortical and diencephalic efferent axons of the rat subicular cortex. Brain Res. 259:181-192, 1983

R.J. Douglas and K.A.C. Martin, Control of neuronal output by inhibition at the axon initial segment. Neural Comp. 2:283-292, 1990

R. Douglas, K. Martin and D. Whitteridge, An intracellular analysis of the visual responses of neurones in the cat visual cortex. J. Physiol. 440:656-696, 1991

J.D. Duder, The effect of septal nuclei stimulation on the release of acetylcholine from the rabbit hippocampus. Brain res. 83:123-133, 1975

S.B. Dunnett, Comparative effects of cholinergic drugs and lesions of nucleus basalis or fimbria-fornix on delayed matching in rats. Psychopharm. 87:357-363, 1985

P. Dutar, Y. Lamour and A. Jobert, Activation of identified septo-hippocampal neurons by noxious peripheral stimulation. Brain res. 328:15-21, 1985

P. Dutar, Y. Lamour, O. Rascol and A. Jobert, Septo-hippocampal neurons in the rat - Further study of their physiological and pharmacological properties. Brain res. 365:325-334,1986

P. Dutar, O. Rascol, A. Jobert and Y. Lamour, Modulation of the excitability of septohippocampal terminal in the rat: Relation to neuronal discharge rate, brain Res. 418:98-110,1987

R. Eckhorn and B. Popel, Rigorous and extended application of information theory to the afferent visual system of the cat. I. Basic concepts. Kybernetik 16:191-200, 1974

R. Eckhorn and B. Popel, Rigorous and extended application of information theory to the afferent visual system of the cat. II Experimental results. Biol. Cybern. 17:7-17, 1975 ______Page 281

H. Eichenbaum, A.Fagan, P. Matthews and NJ. Cohen, Hippocampal system dysfunction and odour discrimination learning in rats: impairment or facilitation depending on representational demands. Behav. Neurosci. 102:331-339, 1988

R.M. Empson and U. Heinemann, Excitatory perforant path connections to CAl predominantly activate inhibition in the rat hippocampal-entorhinal slice in vitro. J. Physiol. 476:70P, 1994

A.K. Engel, P Konig, C.M. Gray and W. Singer, Synchronization of oscillatory neuronal responses between striate and extrastriate visual cortical areas of the cat. Proc. Natl. Acad. Sci. USA, 88:6048-6052, 1991

A.K. Engel, P. Konig, A.K. Kreiter and W. Singer, Interhemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science 252:1177-1179, 1991

A.K. Engel, P. Konig, A.K. Kreiter, T.B. Schillen and W. Singer, Temporal coding in the visual cortex: new vistas on integration in the nervous system. TINS 15:218-226, 1992

E.N. Eskander, B.J. Richmond and L.M. Optican, Role of inferior temporal neurons in visual memory I. Temporal encoding of information about visual images, recalled images, and behavioral context. J. Neurophysiol. 68:1277-1295, 1992a

E.N. Eskander, L.M. Optican and B.J. Richmond, Role of inferior temporal neurons in visual memory II. Multiplying temporal waveforms related to vision and memory. J. Neurophysiol. 68:1296-1306, 1992b

B.W.A. Feenstra and J. Holsheimer, Dipole-like neuronal sources of theta rhythm in dorsal hippocampus, dentate gyrus and cingulate cortex, electroenceph. Clin. Neurophysiol. 47:532-538, 1979

A.A. Fenton, J.L. Kubie and R.U. Muller, The firing of hippocampal place cells is either noisy or signals more than just position. Soc. Neur. Abs. 21:944, 1995

D.M. Finch and T.L. Babb, Neurophysiology of the caudally directed hippocampal efferent system in the rat: projections to the subicular cortex. Brain Res. 197:11-26, 1980

D.M. Finch, N.L. Nowlin and T.L. Babb, Demonstration of axonal projections of neurons in the rat hippocampus and subiculum by intracellular injection of HRP. Brain Res. 271:201-216, 1983

R. Fisher and D. Johnston, Differential modulation of single voltage-gated calcium channels by cholinergic and adrenergic agonists in adult hippocampal neurons. J. Neurophysiol. 64:1291-1302, 1990

R.D. Ford, L.V. Colom and B.H. Bland, The classification of medial septum-diagonal band cells as 0-on or 0-off in relation to hippocampal EEG states. Brain Res. 493:269- 282,1989

G. Forloni, R. Grzanna, R.D. Blakely and J.T. Coyle, Co-localization of N-acetyl- aspartyl-glutamate in central cholinergic, noradrenergic and serotonergic neurons. Synapse 1:455-460, 1987

T.C. Foster, C.A. Castro and B.L. McNaughton, Spatial selectivity of rat hippocampal neurons is dependent on preparedness for movement. Science 244:1580-1582, 1989 ______Page 282

S.E. Fox, Membrane potentials and impedance changes in hippocampal pyramidal cells during theta rhythm. Exp. Brain Res. 77:282-294, 1989

S.E. Fox and J.B. Ranck, Localization and anatomical identification of theta and complex spike cells in dorsal hippocampal formation of rats. Exp. Neurol. 49:299-313, 1975

S.E. Fox and J.B. Ranck, Electrophysiological characteristics of hippocampal complex- spike cells and theta cells. Exp Brain Res. 41:399-410, 1981

S.E. Fox and E. Brazhnik, Action potentials and relations to theta rhythm of septohippocampal neurons in vivo. Soc. Neurosci. Abs. 21:1205, 1995

S.E. Fox, S. Wolfson and J.B. Ranck, Investigating the mechanisms of hippocampal theta rhythms: Approaches and progress. In W. Seifert (ed) Neurobiology of the Hippocampus Academic Press, New York, 1983

S.E. Fox, S. Wolfson and J.B. Ranck, Hippocampal theta rhythm and the firing of neurons in walking and urethane anaesthetized rats. Exp. Brain Res. 62:495-508, 1986

D.M. Freed and S. Corkin, Rate of forgetting in H.M.: 6-month recognition. Behav. Neurosci. 102:823-827, 1988

D.M. Freed, S. Corkin and N.J. Cohen, Forgetting in H.M.: A second look. Neuropsychologica 25:461-471, 1987

W.J. Freeman, Mass Action in the Nervous System. Academic Press, New York, 1975

W.J. Freeman and C.A. Skarda, Spatial EEG patterns, non-linear dynamics and perception: the neo-Sherringtonian view. Brain Res. Rev. 10:147-175, 1985

W.J. Freeman and B.W. von Dijk, Spatial patterns of visual cortical fast EEG during conditioned reflex in a rhesus monkey. Brain Res. 422:267-276, 1987

A.S. French and A.V. Holden, Alias-free sampling of neuronal spike trains. Kybernetik 8:165-171, 1971

C.R. French, P. Sah, K.J. Buckett and P.W. Gage, A voltage-dependent persistent sodium current in mammalian hippocampal neurons. J. Gen. Physiol. 95:1139-1157, 1990

T.F. Freund, GABAergic septohippocampal neurons contain parvalbumin. Brain Res. 478:375-381, 1989

T.F. Freund and M. Antal, GABA-containing neurons in the septum control inhibitory interneurons in the hippocampus. Nature 336:170-173, 1988

M. Frotscher and C. Lérânth, Cholinergic innervation of the rat hippocampus as revealed by choline acetyltransferase immunocytochemistry: A combined light and electron microscopic study. J. Comp. Neurol. 239:237-246, 1985

Y. Fujita and T. Sato, Intracellular records from hippocampal pyramidal cells in rabbit during theta rhythm activity. J. Neurophysiol. 27:1011-1026, 1964

W.A. Fuller, Introduction to Statistical Time Series. John Wiley and Sons, New York, 1976 ______Page 283

F.H. Gage, A. Bjorklund and U. Stenevi, Reinnervation of the partially deafferented hippocampus by compensatory collateral sprouting from spared cholinergic and adrenergic afferents. Brain Res. 268:27-37, 1983

B.H. Gahwiler and D.A. Brown, Muscarine affects calcium-currents in rat hippocampal pyramidal cells in vitro. Neurosci. Lett. 76:301-306, 1987

E. Garcia-Austt, W. Buno and J.L. Garcia-Sanchez, Analysis of unit pairs in the hippocampus. Acta Neurol. Latinoam. 23:141-153, 1977

A. Garcia-Muhoz, L.C. Barrio and W. Buho, Membrane potential oscillations in CAl hippocampal pyramidal neurons in vitro: intrinsic rhythms and fluctuations entrained by sinusoidal injected current. Exp. Brain Res. 97:325-333, 1993

J.L. Garcia-Sanchez, W. Buho, J. Fuentes and E. Garcia-Austt, Non-rhythmic hippocampal units, theta rhythm and afferent stimulation. Brain Res. Bull. 3:213-219, 1978

R.P.A. Gaykema, J. van der Kuil, L.B. Hersh and P.G.M. Luiten, Patterns of direct projections from the hippocampus to the medial septum-diagonal band complex: Anterograde tracing with Phaseohis Vulgaris leucoagglutinin combined with immunochemistry of choline acetyltransferase. Neuroscience 43:349-360, 1991

J.M. Gaztelu and W. Buho, Septo-hippocampal relationships during EEG theta rhythm. Electroenceph. Clin. Neurophysiol. 54:375-387, 1982.

B. Givens, behavioural correlates of single units in the medial septal area: the effect of ethanol. Neuroscience 71:417-427, 1996

B.S Givens and D.S. Olton, Cholinergic and GABAergic modulation of medial septal area: effect on working memory. Behav. Neurosci. 104:849-855, 1990

G. Gogolak, Ch. Stumpf, H. Petsche and J. Sterc. The firing pattern of septal neurons and the form of the hippocampal theta wave. Brain Res. 7:201-207, 1968

A. Gonzalo-Ruiz, A.Alonso, J.M. Sanz and R.R. Llinas, Afferent projections to the mammillary complex of the rat, with special reference to those from surrounding hypothalamic regions. J. Comp. Neurol. 321:277-299, 1992

J.P. Goodridge and J.S. Taube, Preferential use of the landmark navigational system by head direction cells in rats. Behav. Neurosci. 109:49-61, 1995

N. Gorelova and P.B. Reiner, Role of the afterhyperpolarization in control of discharge properties of septal cholinergic neurons in v itro .]. Neurophysiol. 75:695-706, 1996

K.M. Gothard, W.E. Skaggs, K.M. Moore and B.L. McNaughton, Binding of hippocampal CAl neural activity to multiple reference frames in a landmark-based navigation task. J. Neurosci. 16:823-835, 1996

C.M. Gray, Synchronous oscillations in neuronal systems: Mechanisms and functions. J. Comp. Neurosci. 1:11-38, 1994

C.M. Gray and W. Singer, Stimulus-specific neuronal oscillations in orientation columns of cat visual cortex. Proc. Natl. Acad. Sci. USA 86:1698-1702, 1989

C.M. Gray and G. Viana Di Prisco, Properties of stimulus-dependent rhythmic activity of visual cortical neurons in the alert cat. Soc. Neurosci. Abs. 19:359.8, 1993 ______Page 284

C.M. Gray, A.K. Engel, P. Konig and W. Singer, Synchronization of oscillatory neuronal responses in cat striate cortex: Temporal properties. Vis. Neurosci. 8:337- 347, 1992

J.A. Gray, The Neuropsychology of Anxiety: An enquiry into the functions of the septo­ hippocampal system. Oxford University Press, Oxford, 1982a

J.A. Gray, Precis of The Neuropsychology of Anxiety: An enquiry into the fimctions of the septo-hippocampal system. Behav. Brain Sci. 5:469-534, 1982 b

J.A. Gray, The contents of consciousness: A neuropsychological conjecture. Behav. Brain Sci. 18:659-722, 1995

J.A. Gray and J.N.P. Rawlins, Comparator and buffer memory: An attempt to integrate two models of hippocampal functions. In R.L. Isaacson and K.H. Pribram (Eds), The hippocampus Vol pp 4 15 9-2 02, Plenum Press, New York, 1986

J.A. Gray, L. Quintao and M.T. Araujo-Silva, the partial reinforcement extinction effect in rats with medial septal lesions. Physiol. Behav. 8:491-496, 1972a

J.A. Gray, M.T. Araujo-Silva and L. Quintao, Resistance to extinction after partial reinforcement training with blocking of the hippocampal theta rhythm by septal stimulation. J. Comp. Physiol. Psych. 8:497-502, 1972b

J.D. Green and A.A. Arduini, Hippocampal electrical activity in arousal. J. Neurophysiol. 17:533-557, 1954

J.D. Green, D.S. Maxwell, W.J. Schindler and C. Stumpf, Rabbit EEG "Theta" rhythm: Its anatomical source and relation to activity in single neurons. J. Neurophysiol. 23:403-420, 1960

W.H. Griffith, Membrane properties of cell types within guinea pig basal forebrain nuclei in v itr o .f Neurophysiol. 59:1590-1612, 1988

W.H. Griffith and R.T. Matthews, Electrophysiology of AChE-positive neurons in basal forebrain slices. Neurosci. Lett. 71:169-174, 1986

P. Grim wood, H.-W.Yang, Z. Henderson and K.Appenteng, Intrinsic circuitry in the rat medial septum/diagonal band complex. Soc Neurosci. Abs. 21:430, 1995

C.G. Gross, E.C. Rocha-Miranda and D.B. Bender, Visual properties of neurons in infero-temporal cortex of the macaque. J. Neurophysiol. 35:96-111, 1972.

A.I. Gulyas, T.J. Gores and T.F. Freund, Innervation of different peptide-containing neurons in the hippocampus by GABA-ergic septal afferents. Neuroscience 37:31-44, 1990

A.I. Gulyas, L. Seress, K Toth, M. Antal and T.F. Freund, Septal GABAergic neurons innervate inhibitory interneurons in the hippocampus of the macaque monkey. Neuroscience 41:381-390, 1991

B. Gustafsson, M. Galvan, P. Grafe and H.A.Wigstrom, Transient outward current in mammalian central neurone blocked by 4-aminopyridine. Nature 299:252-254, 1982

J.J. Hagan, J.D. Salamone, J. Simpson, S.D. Iversen and R.G.M Morris, Place navigation in rats is impaired by lesions of the medial septum and diagonal band but not nucleus basalis magnocellularis. Behav. Brain Res. 27:9-20, 1988 ______Page 285

L. Haglund, L.W. Swanson and C, Kohler, The projection of the supramammillary nucleus to the hippocampal formation: An immunohistochemical and anterograde transport study with the lectin PHA-L in the rat. J. Comp, Neurol. 229:171-185, 1984

J.V. Halliwell and P.R. Adams, Voltage-clamp analysis of muscarinic excitation in hippocampal neurons. Brain Res. 250:71-92, 1982

D.N. Harper, A.P. McLean and J.C. Dalrymple-Alford, Forgetting in rats following medial septum or mammillary body damage. Behav. Neurosci. 108:691-702, 1994

J.A. Harvey and H.F. Hunt, Effect of septal lesions on thirst in the rat and operant responding for water reward. J. Comp. Physiol. Psych. 59:49-56, 1965

P.G. Henke, Persistence of runway performance after septal lesions in rats. J. Comp. Physiol. Psych. 86:760-767, 1974

D.J. Hepler, D.S. Olton, G.L. Wenk and J.T. Coyle, Lesions in nucleus basalis magnocellularis and medial septal area of rats produce qualitatively similar memory impairments. J. Neurosci. 5:866-873, 1985

O. Herreras, J.M. Solis, A.S. Herranz, R. Martin del Rio and J.C. Lerma, Sensory modulation of hippocampal transmission. II. Evidence for a cholinergic locus of inhibition in the Schaffer-CAl synapse. Brain Res. 461:303-313, 1988

B. Hille, Ionic channels of excitable membranes. 2nd ed. Sinauer Associates, Massachussetts, 1992

R. Hirsh, The hippocampus and contextual retrieval of information from memory: A theory. Behav. Biol, 12:421-444, 1974

R. Hirsh, The hippocampus, conditional operations and cognition. Physiol. Psych. 8:175-182, 1980

A. Hjorth-Simonsen, Hippocampal efferents to the ipsilateral entorhinal area: An experimental study in the rat. J. Comp. Neurol. 142:417-438, 1971

A. Hjorth-Simonsen, Some intrinsic connections of the hippocampus in the rat: An experimental analysis. J. Comp. Neurol. 147-145-162, 1973

J. Holsheimer, Generation of theta activity (RSA) in the cingulate cortex of the rat. Exp. Brain Res. 47:309-312, 1982

J.J. Hopfield, Pattern recognition computation using action potential timing for stimulus representation. Nature 376:33-36, 1995

D. Hothersall, D.A. Johnson and A. Collen, Fixed-ration responding following septal lesions in the rat. J. Comp. Physiol. Psych. 73:470-476, 1970

D.A Hubei and M.S. Livingstone, Segregation of form, colour and stereopsis in primate area 18. J. Neurosci. 7:3378-3415, 1987

P.J. Huber, Robust Statistics. J.Wiley and Sons, New York, 1981

P.T. Huerta and J.E. Lisman, Heightened synaptic plasticity of hippocampal CAl neurons during a cholinergically induced rhythmic state. Nature 364:723-725, 1993

N. Ishizuka, J. Weber and D.G. Amaral, Organization of intrahippocampal projections originating from CA3 pyramidal cells in the rat. J. Comp. Neurol. 295:580-623, 1990 ______Page 286

H. Jahnsen and R. LLinas, Ionic basis for the electroresponsiveness and oscillatory properties of guinea-pig thalamic neurones in vitro .]. Physiol. 349:227-247, 1984

K.J. Jeffery, J.G. Donnett and J. O'Keefe, Medial septal control of theta-correlated unit firing in the entorhinal cortex of awake rats. NeuroReport 6:2166-2170, 1995

D. Johnston and T.H. Brown, Giant synaptic potential hypothesis for epileptiform activity. Science 211:294-297, 1981

D. Johnston and T.H. Brown, Interpretation of voltage-clamp measurements in hippocampal neurons. J. Neurophysiol. 50:464-486, 1983

R. Jung and A.E. Kornmiiller, Eine Methodik der Abkitung localisierter Potentialschweankungen aus subcorticalen Hirngebieten. Arch Psychiat. Nervenkr. 109:1-30, 1938

S.S. Kaitz and R.T. Robertson, Thalamic connections with limbic cortex. II Corticothalamic projections. J. Comp. Neurol. 195:527-545, 1981

L.D. Kartsounis, P. Rudge and J.M. Stevens, Bilateral lesions of CAl and CA2 fields of the hippocampus are sufficient to cause a severe amnesic syndrome in humans. J. Neurol. Neurosurg. Psych. 59:95-98, 1995

K. Kasamo, K. Tada, N. Ueda, T. Kojima, M. Kogure and K. Ishikawa, Effects of several 5-HT(lA) agonists on hippocampal rhythmical slow activity in unanaesthetized rats. Neuropharmacol. 33:905-914, 1994

S.J. Kehl and H. McLennan, An electrophysiological characterization of inhibitions and postsynaptic potentials in rat hippocampal CA3 neurons in vitro. Exp. Brain Res. 60:299-308, 1985

R.P. Kesner, K. Crutcher and D R. Beers, Serial position curves for item (spatial location) information: Role of the dorsal hippocampal formation and medial septum. Brain Res. 454:219-226, 1988

C. King and M. Recce, Simultaneous recording of multiple cells from the medial septum of freely moving rats. Soc. Neurosci. Abs. 21:1439, 1995

G.G. Kinney, B. Kocsis and R.P. Vertes, Medial septal unit firing characteristics following injections of 8-OH-DPAT into the median raphe nucleus. Brain Res. 708:116-122, 1996.

I.J. Kirk and N. McNaughton, Supramammillary cell firing and hippocampal rhythmical slow activity. Neuroreport 2:723-725, 1991

I.J. Kirk and N. McNaughton, Mapping the differential effects of procaine on frequency and amplitude of reticularly elicited hippocampal rhythmical slow activity. Hippocampus 3:517-525, 1993

J. Kiss, A.J. Patel and T.F. Freund, Distribution of septohippocampal neurons containing parvalbumin or choline acetyltransferase in the rat brain. J. Comp. Neurol. 298:362-372, 1990

J.J. Knierim, H.S. Kudrimoti and B.L. McNaughton, Place cells, head direction cells, and the learning of landmark stability. J. Neurosci. 15:1648-1659, 1995

B. Knight, Dynamics of encoding in a population of neurons. J. Gen. Physiol. 59:734- 766, 1972 ______Page 287

C. Koch, M. Rapp and I. Segev, A brief history of time (constants). Cerebral Cortex 6:93-101, 1996

B. Kocsis and R.P. Vertes, Characterization of neurons in the supramammillary nucleus and mammillary body that discharge rhythmically with the hippocampal theta rhythm in the rat J Neurosci 14:7040-7052, 1994

B. Kocsis, R.P. Vertes and J.S. Thinscmidt, Discharge properties of neurons of the supramammillary nucleus and mammillary body:relationship to hippocampal theta rhythm Soc Neurosci Abstr. 19:1442, 1993

C. Kohler, Intrinsic projections of the retrohippocampal region in the rat brain. I. The subicular complex. J. Comp. Neurol. 236:504-522, 1985

C. Kohler, Intrinsic connections of the retrohippocampal region in the rat brain. II. The medial entorhinal area. J. Comp. Neurol. 246:149-169, 1986

C. Kohler, V. Chan-Palay and Jang-Yen Wu, Septal neurons containing glutamic acid decarboxylase immunoreactivity project to the hippocampal region in the rat brain. Anat. Embryol. 169:41-44, 1984

B.R. Komisaruk, Synchrony between limbic system theta activity and rhythmical behaviour in rats. J. Comp. Physiol. Psychol. 70:482-492, 1970

J. Konopacki, M.B. Maclver, B.H.Bland and S.H. Roth, Carbachol-induced EEG 'theta' activity in hippocampal brain slices. Brain Res., 405:196-198, 1987a

J. Konopacki, B.H. Bland and S.H. Roth, Phase shifting of CAi and dentate EEG theta rhythms in hippocampal formation slices. Brain Res. 417:399-402, 1987b

J. Konopacki, H. Golebiewski and B. Eckersdorf, Carbachol-induced rhythmic slow activity (theta) in cat hippocampal formation slices. Brain res. 578:13-16, 1992a

J. Konopacki, B.H. Bland, L.V. Colom and S.D. Oddie, In vivo intracellular correlates of hippocampal formation theta-on and theta-off cells. Brain Res. 586:247-255, 1992b

K.C. Kosel, G.W. van Hoesen and D.L. Rosene, Non-hippocampal cortical projections from the entorhinal cortex in the rat and rhesus monkey. Brain Res. 244:201-213, 1982

R.C. Kramis, C.H. Vanderwolf and B.H. Bland, Two types of hippocampal rhythmical slow activity in both the rabbit and the rat: Relations to behaviour and effects of atropine, diethyl ester, urethane and pentobarbital. Exp. Neurol. 49:58-85, 1975

A.K. Kreiter and W. Singer, Stimulus-dependent synchronization of neuronal responses in the visual cortex of the awake macaque monkey. J. Neurosci. 16:2381- 2396,1996

K. Krnjevic and N. Ropert, Electrophysiological and pharmacological characteristics of facilitation of hippocampal population spikes by stimulation of the medial septum. Neuroscience 7:2165-2183, 1982

J.-C. Lacaille and P.A. Schwartzkroin, Stratum lacunosum-moleculare interneurons of hippocampal CAl region. I Intracellular response characteristics, synaptic responses, and morphology. J. Neurosci. 8:1400-1410, 1988a ______Page 288

J.-C. Lacaille and P.A. Schwartzkroin, Stratum lacunosum-moleculare interneurons of hippocampal CAl region. II Intrasomatic and intradendritic recordings of local circuit synaptic interactions. J. Neurosci. 8:1411-1424, 1988b

J.-C. Lacaille, A.L. Mueller, D.D. Kunkel and P.A. Schwartzkroin, Local circuit interactions between oriens/alveus interneurons and CAl pyramidal cells in hippocampal slices: Electrophysiology and morphology. J. Neurosci. 7:1979-1993, 1987

Y. Lamour, P. Dutar and A. Jobert, Septo-hippocampal and other medial septum- diagonal band neurons:Electrophysiological and pharmacological properties. Brain Res. 309:227-239, 1984

G.D. Lange and P.H. Hartline, Fourier analysis of spike train data. Biol. Cybern. 34:31-34,1979

S. Laurberg and K.E. Sorensen, Associational and commissural collaterals of neurons in the hippocampal formation (hilus fasciae dentatae and subfield CA3). Brain Res. 212:287-300,1981

V.H. Lawson and B.H. Bland, The role of the septohippocampal pathway in the regulation of hippocampal field activity and behaviour: Analysis by the intraseptal microinfusion of carbachol, atropine and procaine. Exp. Neurol. 120:132-144, 1993

B.H. Lee, Y. Lamour and M.H. Bassant, lontophoretic study of medial septal neurons in the unanaesthetized rat. Neurosci. Lett. 128:29-32, 1991

M.G. Lee, J.J. Chrobak, A. Sik, R.G. Wiley and G. Buzsaki, Hippocampal theta activity following selective lesion of the septal cholinergic system. Neuroscience 62:1033-1047,1994

C. Lérânth and M. Frotscher, Cholinergic innervation of hippocampal GAD- and somatostatin-immunoreactive commissural neurons. J. Comp. Neurol. 261:33-47, 1987

C. Lérânth and M. Frotscher, Organisation of the septal region in the rat brain: cholinergic-GABAergic interconnections and the termination of hippocampo-septal fibres. J. Comp. Neurol. 289:304-314, 1989

C. Lérânth, T. Deller and G Buzsâki, Intraseptal connections redefined: lack of a lateral septum to medial septum path. Brain Res. 583:1-11, 1992

J. Lerma and E Garcia-Austt, Hippocampal theta rhythm during paradoxical sleep. Effects of afferent stimuli and phase relationships with phasic events. Electroenceph. Clin. Neurophysiol. 60:46-54, 1985

L.-W.S. Leung, Pharmacology of theta phase shifts in the hippocampal CAl region of freely moving rats. Electroenceph. Clin. Neurophysiol. 58:457-466, 1984a

L.-W.S. Leung, Theta rhythm during REM sleep and waking: Correlations between power, phase and frequency. Electroenceph, Clin. Neurophysiol. 58:553-564, 1984b

L.-W.S. Leung, Model of gradual phase shift of theta rhythm in the rat. J. Neurophysiol. 52:1051-1065, 1984c

L.-W.S. Leung and C.Y. Yim, Intracellular records of theta rhythm in hippocampal CAl cells of the rat. Brain Res. 367:323-327, 1986 ______Page 289

W.B. Levy, C.M. Colbert and N.L. Desmond, Another network model bites the dust: Entorhinal inputs are no more than weakly excitatory in the hippocampal CAl region. Hippocampus 5:137-140, 1995

P.R. Lewis and C.C.D. Shute, The cholinergic limbic system: Projections to hippocampal formation, medial cortex, nuclei of the ascending cholinergic reticular system, and the subfornical organ and supra-optic crest. Brain 90:521-539, 1967

P.R. Lewis, C.C.D. Shute and A. Silver, Confirmation from choline acetylase analyses of a massive cholinergic innervation to the rat hippocampus. J. Physiol. 191:215-224, 1967

X.-G. Li, P. Somogyi, J.M. Tepper and G. Buzsaki, Axonal and dendritic arborization of an intracellularly labelled chandelier cell in the CAl region of rat hippocampus. Exp. Brain Res. 90:519-525, 1992

R. Linke, H. Schwegler and M Boldyreva, Cholinergic and GABAergic septo­ hippocampal projection neurons in mice: a retrograde tracing study combined with double immunocytochemistry for choline acetyltransferase and parvalbumin. Brain Res., 653:73-80, 1994.

J.E. Lisman and M.A.P. Idiart, Storage of 7±2 short-term memories in oscillatory subcycles. Science 267:1512-1515, 1995

R. Lorente de No, Studies on the structure of the cerebral cortex. II. Continuation of the study of the ammonic system. J. Psychol. Neurol. 46:113-177, 1934

F.H. Lopes da Silva, A. van Rotterdam, W. Storm van Leeuwen and A.M. Tielen, Dynamic characteristic of visual evoked potentials in the dog. II. Beta frequency selectivity in evoked potentials and background activity. Electroenceph. Clin. Neurophysiol. 29:260-268, 1970

F.H. Lopes da Silva, M.P. Witter, P.H. Boejinga and A.H.M Lohman, Anatomical organization and physiology of the limbic cortex. Physiol. Rev. 70:453-511, 1990

O. Macadar, J.A. Roig, J.M. Monti and R. Budelli, The functional relationship between septal and hippocampal unit activity and hippocampal theta rhythm. Physiol. Behav. 5:1443-1449, 1970

B.A. MacVicar and F.W.Y. Tse, Local neuronal circuitry underlying cholinergic rhythmical slow activity in CA3 area of rat hippocampal slices. J. Physiol. 417:197-212, 1989

D.V. Madison, B. Lancaster and R.A. Nicoll, Voltage clamp analysis of cholinergic action in the hippocampus. J. Neuroscience 7:733-741, 1987

Z. Magloczky, L. Ascady and T.F. Freund, Principal cells are the postsynaptic targets of supramammillary afferents in the hippocampus of the rat. Hippocampus, 4:322-334, 1994

K.V. Mardia, Statistics of Directional Data. Academic Press, London, 1972

H. Markram and M. Segal, Electrophysiological characteristics of cholinergic and non­ cholinergic neurons in the rat medial septum-diagonal band complex. Brain Res. 513:171-174, 1990 ______Page 290

E.J. Markus, C.A. Barnes, B.L. McNaughton, V.L. Gladen and W.E. Skaggs, Spatial information content and reliability of hippocampal CAl neurons: Effects of visual input, hippocampus 4:410-421, 1994

E.J. Markus, Y.-L. Qin, B. Leonard, W.E. Skaggs, B.L. McNaughton and C.A. Barnes, Interactions between location and task affect the spatial and directional firing of hippocampal neurons. J. Neurosci. 15:7079-7094, 1995

R.T. Matthews and W.L. Lee, A comparison of extracellular and intracellular recordings from medial septum/diagonal band neurons in vitro. Neuroscience 42:451- 462,1991

J.H.R. Maunsell and D.C. van Essen, Functional properties of neurons in the middle temporal visual area of the macaque monkey. I. Selectivity for stimulus direction, speed, and orientation. J. Neurophysiol. 49:1127-1147, 1983

J.W. McClurkin and L.M. Optican, Primate striate and prestriate cortical neurons during discrimination I. Simultaneous temporal encoding of information about color and pattern. J. Neurophysiol. 75:481-495, 1996

J.W. McClurkin, J.A. Zarbock and L.M. Optican, Primate striate and prestriate cortical neurons during discrimination II. Separable temporal codes for color and pattern. J. Neurophysiol. 75:496-507, 1996

D.A. McCormik and D.A. Prince, Mechanisms of action of acetylcholine in the guinea-pig cerebral cortex in vitro.]. Physiol. 375:169-194, 1986

W.L. McFarlane, H. Teitelbaum and E.K. Hedges, Relationship between hippocampal theta activity and running speed in the rat. J. Comp. Physiol. Psychol. 88:324-328, 1975

R.K. McNamara and R.W. Skelton, Effects of intracranial infusions of chlordiazepoxide on spatial learning in the Morris water maze. I. Neuroanatomical specificity. Behav. Brain Res. 59:175-191, 1993a

R.K. McNamara and R.W. Skelton, Effects of intracranial infusions of chlordiazepoxide on spatial learning in the Morris water maze. II. Neuropharmacological specificity. Behav. Brain Res. 59:193-204, 1993b

B.L.McNaughton, C.A. Barnes and J. O'Keefe, The contributions of position, direction, and velocity to single unit activity in the hippocampus of freely-moving rats. Exp. Brain Res. 52:41-49, 1983

B.L. McNaughton, B. Leonard and L. Chen, Cortico-hippocampal interactions and cognitive mapping: A hypothesis based on reintegration of the parietal and inferotemporal pathways fro visual processing. Psychology 17:236-246, 1989

B.L. McNaughton, L.Chen and E.J. Markus, "Dead reckoning", landmark learning, and the sense of direction: a neurophysiological and computational hypothesis. J. Cognit. Neurosci. 3:189-202, 1991

N. McNaughton, B. Logan, K.S. Panickar, I.J. Kirk, W.-X. Pan, N.T. Brown and A. Heenan, Contribution of synapses in the medial supramammillary nucleus to the frequency of hippocampal theta rhythm in freely moving rats. Hippocampus 5:534-545, 1995. ______Page291

R.C. Meibach and A. Siegel, Efferent connections of the septal area in the rat: An analysis utilizing retrograde and anterograde transport methods. Brain Res. 119:1-20, 1977

T. Melander, W.A. Staines, T. Hokfelt, A Rokaeus, F. Eckstein, P.M. Salvaterra and B.H. Wainer, Galanin-like immunoreactivity in cholinergic neurons of the septum- basal forebrain complex projecting to the hippocampus of the rat. Brain Res. 360:130- 138, 1985

M.M. Mesulam, G. van Hoesen and D.G. Rosene, Substantia innominata, septal area and nuclei of the diagonal band in rhesus monkey: Organization of efferents and their acetylcholinesterase histochemistry. Soc. Neurosci. Abs. 3:202, 1977

M.M. Mesulam, E.J. Mufson, A.I. Levey and B.H. Wainer, Cholinergic innervation of the cortex by the basal forebrain: cytochemistry and cortical connections of the septal area, diagonal band nuclei, nucleus basalis (substantia innominata), and hypothalamus in the rhesus monkey. J. Comp. Neurol. 214:170-197, 1983a

M.M. Mesulam, E.J. Mufson, B.H. Wainer and A.I. Levey, Central cholinergic pathways in the rat: An overview based on an alternative nomenclature (Chl-Ch6). Neuroscience 10:1185-1201, 1983b

M. Meunier, J. Bachelavier, M. Mishkin and E.A. Murray, Effects on visual recognition of combined and separate ablations of the entorhinal and perirhinal cortex in rhesus monkeys. J. Neurosci. 13:5418-5432, 1983

M. Mignard, D. Bentzinger, N. Bender and M. Gabriel, Type I and type II theta-like unit activity in structures of the Papez circuit during differential avoidance conditioning in rabbits. Soc Neurosci. Abstr. 13:305, 1987

R. Miles and R.K.S. Wong, Excitatory synaptic interactions between CA3 neurones in the guinea-pig hippocampus. J. Physiol. 373:397-418, 1986

R. Miles, K. Toth, A.L Gulyas, N. Hajos and T.F. Freund, Differences between somatic and dendritic inhibition in the hippocampus. Neuron 16:815-823, 1996

B. Milner, S. Corkin and H.L. Teuber, Further analysis of the hippocampal amnesic syndrome: 14-year followup study of H.M.. Neuropsychologica 6:215-234, 1968

T.A. Milner, R. Loy and D.G. Amaral, An anatomical study of the development of the septo-hippocampal projection in the rat. Dev. Brain Res. 8:343-371, 1983

S.J. Mitchell and J.B. Ranck, Firing patterns and behavioural correlates of neurones in entorhinal cortex of freely-moving rats. Soc. Neurosci. Abs. 3:202, 1977

S.J. Mitchell and J.B. Ranck, Generation of theta rhythm in medial entorhinal cortex of freely moving rats. Brain Res. 189:49-66, 1980

S.J. Mitchell, J.N.P. Rawlins, O. Steward and D.S. Olton, Medial septal area lesions disrupt 0 rhythm and cholinergic staining in medial entorhinal cortex and produce impaired radial arm maze behaviour in rats. J. Neurosci. 2:292-302, 1982

M. Miyamoto, J. Kato, S. Narumi and A. Nagaoka, Characteristics of memory impairment following lesioning of the basal forebrain and medial septal nucleus in rats. Brain Res. 419:19-31, 1987 ______Page 292

S.J.Y. Mizumori and J.D. Williams, Directionally selective mnemonic properties of neurons in the lateral dorsal nucleus of the thalamus of rats. J. Neurosci. 13:4015-4028, 1993

S.J.Y. Mizumori, B.L. McNaughton, C.A. Barnes and K.B. Fox, Preserved spatial coding in hippocampal CAl pyramidal cells during reversible suppression of CA3c output: Evidence for pattern completion in hippocampus. J. Neurosci. 9:3915-3928, 1989

S.J.Y. Mizumori, C.A. Barnes and B.L. McNaughton, Behavioural correlates of theta- on and theta-off cells recorded from the hippocampal formation of mature young and aged rats. Exp. Brain Res. 80:365-373, 1990a

S.J.Y. Mizumori, G.M. Perez, M.C. Alvarado, C.A. Barnes and B.L. McNaughton, Reversible inactivation of the medial septum differentially affects two forms of learning in rats. Brain Res. 528:12-20, 1990b

T. Mizuno, Y. Endo, J. Arita and F. Kimura, Acetylcholine release in the rat hippocampus as measured by the microdialysis method correlates with motor activity and exhibits a diurnal variation. Neuroscience 44:607-612, 1991

R.Y. Moore and A.E. Halaris, Hippocampal innervation by serotonin neurons of the midbrain raphe in the rat. J. Comp. Neurol. 164:171-184, 1975

F.R. Morales, J.A. Roig, J.M. Monti, O. Macadar and R. Budelli, Septal unit activity and hippocampal EEG during the sleep-wakefulness cycle of the rat. Physiol. Behav. 6:563-567, 1971

R.G.M. Morris, Developments of a water-maze procedure for studying spatial learning in the rat. J. Neurosci. Meth. 11:47-60, 1984

R.G.M. Morris, A.H. Black and J. O’Keefe, Hippocampal EEG during a ballistic movement. Neurosci. Lett. 3:102, 1976

R.G.M. Morris, P. Garrud, J.N.P. Rawlins and J. O'Keefe, Place navigation impaired in rats with hippocampal lesions. Nature 297:681-683, 1982

R.U. Muller and J.L. Kubie, The effects of changes in the environment on the spatial firing of hippocampal complex-spike cells. J. Neurosci. 7:1951-1968, 1987

R.U. Muller and J.L. Kubie, The firing of hippocampal place cells predicts the future position of freely moving rats. J. Neurosci. 9:4101-4110, 1989

R.U. Muller, J.L. Kubie and J.B Ranck, Spatial firing patterns of hippocampal complex-spike cells in a fixed environment. J. Neurosci. 7:1935-1950, 1987a

R.U. Muller, E. Bostock, J.S. Taube and J.L. Kubie, On the directional firing properties of hippocampal place cells. J. Neurosci. 14:7235-7251, 1994

L. Nadel, the hippocampus and space revisited. Hippocampus 1:221-229, 1991

Y. Nakajima, S. Nakajima, R.J. Leonard and K. Yamaluchi, Acetylcholine raises excitability by inhibiting the fast transient potassium current in cultured hippocampal neurons. Proc. Natl. Acad. Sci. USA 83:3022-3026, 1986

S. Namura, M. Takada, H. Kikuchi and N. Mizuno, Topographical organization of subicular neurons projecting to subcortical regions. Brain Res. Bull. 35:221-231, 1994 ______Page 293

T. Naumann, T. Deller and M. Frotscher, Multiple projections are unlikely to account for the survival of rat medial septal neurons after axotomy. Neurosci. Lett. 211:117- 120, 1996

K. Niimi, M. Niimi and Y. Okada, Thalamic afferents to the limbic cortex in the cat studied with the method of retrograde axonal transport of horseradish peroxidase. Brain Res. 145:225-238, 1978

A. Nuhez, E. Garcia-Austt and VV. Buho, Intracellular 9-rhythm in identified hippocampal pyramids. Brain Res. 416:289-300, 1987

C. Nyakas, P.G.M. Luiten, D.G. Spencer and J. Traber, Detailed projection patterns of septal and diagonal band efferents to the hippocampus in the rat witdi emphasis on innervation of CAl and dentate gyrus. Brain Res. Bull. 18:533-545, 1987

S.D. Oddie, W. Stefanek, I.J. Kirk and B.H. Bland, Intraseptal procaine abolishes hypothalamic stimulation-induced wheel-running and hippocampal theta field activity in rats. J. Neurosci. 16:1948-1956, 1996

J. O'Keefe, Place units in the hippocampus of the freelv moving rat. Exp. Neurol. 51:78-109, 1976

J. O'Keefe, A review of the hippocampal place cells. Prog. Neurobiol. 13:419-439, 1979

J. O'Keefe and J. Dostrovsky, The hippocampus as a spatial map. preliminary evidence from unit activity in the freely moving rat. Brain Res. 34:171-175, 1971

J. O'Keefe and D.H. Conway, hippocampal place units in the freely moving rat: Why they fire where they fire. Exp. Brain. Res. 31:573-590, 1978

J. O'Keefe and L. Nadel, The Hippocampus as a Cognitive Map. Clarendon Press, Oxford, 1978

J. O'Keefe and A. Speakman, Single unit activity in the rat hippocampus during a spatial memory task. Exp. Brain Res. 68:1-27, 1987

J. O'Keefe and M. Recce, Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 3:317-330, 1993

J. O'Keefe and N. Burgess, Geometric determinants of the place fields of hippocampal neurones. Nature 381:425-428, 1996

J. O'Keefe, L. Nadel, S. Keightley and D. Kill, Fornix lesions selectively abolish place learning in the rat. Exp. Neurol. 48:152-166, 1975

H. Oka and K. Yoshida, Septohippocampal connections to field CAl of the rat identified with field potential analysis and retrograde labelling by horseradish peroxidase. Neurosci. Lett. 58:19-24, 1985

D.S. Olton, M. Branch and P. Best, Spatial correlates of hippocampal unit activity. Expl. Neurol. 58:387-409, 1978

D.S. Olton, J.T. Becker and G.E. Handlemann, Hippocampus, space, and memory. Behav. Brain Sci. 2:313-365, 1979 ______Page 294

B. Onteniente, M. Geffard, G. Campistron and A. Galas, An uitrastructurai study of GABA-immunoreactive neurons and terminals in the septum of the rat. J. Neurosci. 7:48-54

L.M. Optican and B.J. Richmond, Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. III. Information theoretic analysis. J. Neurophysiol. 57:162-178, 1987

T. Otto and H. Eichenbaum, Dissociable roles of the hippocampus and orbitofrontal cortexin an odor-guided delayed non-match to sample task. Behav. Neurosci. 105:111- 119, 1992

J.W. Papez, A proposed mechanism of emotion. Arch. Neurol. Psychiat. 38:725-743, 1937

G. Paxinos and C. Watson, The Rat Brain in Stereotaxic Coordinates. 2nd Ed., Academic Press, London , 1986

D.L Perret, A.J. Mis tin and A.J. Chitty, Visual neurones responsive to faces. Trends Neurosci. 10:358-364, 1987

G.M. Peterson, A quantitative analysis of the crossed septohippocampal projection in the rat brain. Anat. Embryol. 180:421-425, 1989

G.M. Peterson, Differential projections to the hippocampus by neurons of the medial septum and vertical limb of the diagonal band. Brain Res. 646:129-134, 1994

H. Petsche, Ch. Stumpf and G. Gogolak, The significance of the rabbit's septum as a relay station between the midbrain and the hippocampus. I. The control of the hippocampus arousal activity by the septum cells. Electroenceph. Clin. Neurophysiol. 14:202-211, 1962

H. Petsche, G. Gogolak and P.A. van Zwieten, Rhythmicity of septal cell discharges at various levels of reticular excitation. Electroenceph. Clin. Neurophysiol. 19:25-33, 1965

T.A. Pitler and B.E. Alger, Cholinergic excitation of GABAergic interneurons in the rat hippocampal slice. J. Physiol. 450:127-142, 1992

W.H. Press, S.A. Teukolsky, W.T. Vetterling and B.P. Flannery, Numerical Recipes in C. 2nd Ed., Cambridge University Press, Cambridge, 1992

G.J. Quirk, R.U. Muller, J.L. Kubie and J.B. ranck, the positional firing properties of medial entorhinal neurons: Description and comparison with hippocampal place cells. J. Neurosci. 12:1945-1963, 1992

J.B. Ranck, Studies on single neurons in dorsal hippocampal formation and septum in unrestrained rats. Expl. Neurol. 41:461-555, 1973

J.B. Ranck, Behavioural Correlates and firing repertoires of neurons in septal nuclei in unrestrained rats. In: J.E. DeFrance (ed), The Septal Nuclei. Plenum Press, New York, 1976

J.B. Ranck, Head direction cells in the deep cell layer of dorsal presubiculum in freely moving rats. Soc. Neurosci. Abs 10:599, 1984 ______Page 295

J.B. Ranck, Head direction cells in the deep cell layer of dorsal presubiculum in freely moving rats. In: Electrical activity of the Archicortex, Akadémiai Kiado, Budapest, 1985

J.N.P. Rawlins, Associations across time: The hippocampus as a temporary memory store. Behav. Brain Res. 8:479-496, 1985

J.N.P. Rawlins and D.S. Olton, the septo-hippocampal system and cognitive mapping. Behav. Brain Res. 5:331-358, 1982

J.N.P. Rawlins, J. Feldon and J.A. Gray, Septo-hippocampal connections and the hippocampal theta rhythm. Exp. Brain Res. 37:49-63, 1979

J.N.P. Rawlins, J. Feldon and S. Butt, The effects of delaying reward on choice preference in rats with hippocampal or selective septal lesions. Behav. Brain Res. 15:191-203, 1985

M. Recce, The representation of space in the rat hippocampus as revealed using new computer-based methods. PhD. thesis. University College London, 1994

M. Recce and J. O'Keefe, The tetrode: an improved technique for multi-unit extracellular recording. Soc Neurosci Abs 15:1250, 1991

G.E. Ribak, L. Seress and D.G. Amaral, The development, ultrastructure and synaptic connections of the mossy cells of the dentate gyrus. J. Neurocytol. 12:577-597, 1985

B.J. Richmond and L.M. Optican, Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. II. Quantification of response waveform. J. Neurophysiol. 57:147-161, 1987

B.J. Richmond, L.M. Optican, M. Podell and H. Spitzer, Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. I. Response characteristics. J. Neurophysiol. 57:132-146, 1987

B.J. Richmond, L.M. Optican and H. Spitzer, Temporal encoding of two-dimensional patterns by single units in primary visual cortex. I. Stimulus-response relations. J. Neurophysiol. 64:351-369, 1990

B.J. Richmond and L.M. Optican, Temporal encoding of two-dimensional patterns by single units in primate primary visual cortex. II. Information transmission. J. Neurophysiol. 64:370-380, 1990

R.T. Robertson and S.S. Kaitz, Thalamic connections with limbic cortex. I. Thalamocortical projections. J. Comp. Neurol. 195:501-5525, 1981

T.A.P. Roeling, J.G. Veening, M.R. Kruk, J.P.W. Peters, M.Ê.J. Vermellis and R. Nieuwenhuys, Efferent connections of the hypothalamic 'aggression area' in the rat. Neuroscience. 59:1001-1024, 1994

G. Rose, D. Diamond and G.S. Lynch, Dentate granule cells in the rat hippocampal formation have the behavioural characteristics of theta neurons. Brain Res. 266:29-37, 1983

A. Rougeul, J.J. Bouyer, L. Dedet O. Debray, Fast somato-parietal rhythms during combined focal attention and immobility in baboon and squirrel monkey. Electroenceph. Clin. Neurophysiol. 46:310-319, 1979 ______Page 296

J.W. Rudy and R.J. Sutherland, The hippocampal formation is necessary for rats to learn and remember configurai discriminations. Behav. Brain. Res. 34:97-109, 1989

J.W. Rudy and R.J. Sutherland, Configurai association theory and the hippocampal formation: An appraisal and reconfiguration. Hippocampus 5:375-389, 1995

D.E. Rumelhart and D. Zipser, Feature discovery by competitive learning. In: D.E. Rumelhart and J.L. McClelland (eds.) Parallel Distributed Processing: Explorations in the Microstnicture of Cognition MIT Press, Cambridge, Mass., 1986

D. Schild, An efficient method for the fourier transform of a neuronal spike train. Int. J. Neurosci. 17:179-182, 1982

P.A. Schwartzkroin and D.D. Kunkel, Morphology of identified interneurons in the CAl regions of guinea pig hippocampus. J. Comp. Neurol. 232, 205-218, 1985

W.B. Scoville and B. Milner, Loss of recent memory after bilateral hippocampal lesions. J. Neurol. Neurosurg. Psych. 20:11-21, 1957

M. Segal, Hippocampal unit responses to perforant path stimulation. Exp. Neurol. 35:541-546, 1972

M. Segal, Physiological and pharmacological evidence for a serotonergic projection to the hippocampus. Brain Res. 94:115-131, 1975

M. Segal, Properties of rat medial septal neurones recorded in vitro. J. Physiol. 379:309-330, 1986

M. Segal and S.C. Landis, Afferents to the septal area of the rat studied with the method of retrograde axonal transport of horseradish peroxidase. Brain Res. 82:263-268, 1974

I. Segev, J.W. Fleshman and R.E. Burke, Compartmental models of complex neurons. In: C. Koch and I. Segev (eds) Methods in Neuronal Modelling MIT Press, Cambridge, Mass, 1989

M. Seki and K. Zyo, Anterior thalamic afferents from the mammillary body and the limbic cortex in the rat. J. Comp. Neurol. 229:242-256, 1984

K. Semba, P.B. Reiner, E.G. McGeer and H.C. Fibiger, Brainstem projecting neurons in the rat basal forebrain: Neurochemical, topographical, and physiological distinctions from cortically projecting cholinergic neurons. Brain Res. Bull. 22:501-509, 1989

M.C. Senut, D. Menetrey and Y. Lamour, Cholinergic and peptidergic projections from the medial septum and the nucleus of the diagonal band of Broca to dorsal hippocampus, cingulate cortex and olfactory bulb: A combined wheatgerm agglutinin- apohorseradish peroxidase-gold immunohistochemical study. Neuroscience 30:385- 403, 1989

M.C. Senut, F. de Bilbao and Y. Lamour, Medial septal neurons containing N-acetyl- aspartyl-glutamate-like immunoreactivity project to the hippocampal formation in the rat. Neurosci. Lett. 113:12-16, 1990

L. Seress and C.E. Ribak, A combined Golgi-electron microscopic study of non- pyramidal neurons in the CAl area of the hippocampus. J. Neurocytol. 14:717-730, 1985 ______Page 297

M.N. Shadlen and W.T. Newsome, Noise, neural codes and cortical organization. Curr. Opin. Neurobiol. 4:569-579, 1994

M.N. Shadlen and W.T. Newsome, Is there a signal in the noise? Curr. Opin. Neurobiol. 5:248-250, 1995

P. Sharp, Computer-simulation of hippocampal place cells. Psychobiology 19:103-115, 1991

D.F. Sherry and D.L. Schacter, The evolution of multiple memory systems. Psych. Rev. 94:439-454, 1987

H. Shibata, Descending projections to the mammillary nuclei in the rat, as studied by retrograde and anterograde transport of wheat germ agglutinin-horseradish peroxidase. J. Comp. Neurol. 285:436-452, 1989

H. Shibata, Topographic organization of subcortical projections to the anterior thalamic nuclei in the rat. J. Comp. Neurol. 323:117-127, 1992

H. Shibata, Direct projections from the anterior thalamic nuclei to the retrohippocampal region in the rat. J. Comp. Neurol. 337:431-445, 1993

K. Shinoda, M. Tohyama and Y. Shiotani, Hippocampal gamma-aminobutyric acid (GABA)-containing neuron system in the rat: A study using a double-labelling method that combines retrograde tracing and immunocytochemistry. Brain Res. 409:181-186, 1987

M.T. Shipley and K.E. Sorensen, On the laminar organisation of the anterior thalamus projections to the presubiculum in the guinea pig. Brain Res., 86:473-477, 1975

C.C.D. Shute and P.R. Lewis, Cholinesterase-containing systems of the brain of the rat. Nature 199:1160-1164, 1963

C.C.D. Shute and P.R. Lewis, Electron microscopy of cholinergic terminals and acetylcholinesterase-containing neurons in the hippocampal formation of the rat. Zeit. Zellforsch. 69:334-343, 1966

W. Singer and C.M. Gray, Visual feature integration and the temporal correlation hypothesis. Ann. Rev. Neurosci. 18:555-586, 1995

A. Sik, M. Penttonen, A. Ylinen and G. Buzsaki, Hippocampal CAl interneurons: An in vivo intracellular labelling study. J. Neurosci. 15:6651-6665, 1995

R.W. Sikes and B.A. Vogt, Afferent connections of anterior thalamus in rats: Sources and association with muscarinic acetylcholine receptors. J. Comp. Neurol. 246:312- 342, 1987

B.R. Sinclair, G.S. Seto and B.H. Bland, Cells in CAl and dentate layers of hippocampal formation: Relations to slow wave activity and motor behaviour in freely moving rabbit. J. Neurophysiol 48:1214-1225, 1982

W.E. Skaggs, J.J. Knierim and B.L. McNaughton, Hippocampal theta and the spatial firing of dentate gyrus granule cells. Soc Neurosci. Abs. 21:940, 1995

W.E. Skaggs, B.L. McNaughton, M.A. Wilson and C.A. Barnes, Theta-phase precession in hippocampal neuronal populations and the compression of temporal sequences. Hippocampus 6:149-172, 1996 ______Page 298

J.W. Smythe, B.R. Christie, L.V. Colom, V.H. Lawson and B.H. Bland, Hippocampal 9 field activity and 0-ON/0-OFF cell discharges are controlled by an ascending hypothalamo-septal pathway. J. Neurosci. 11:2241-2248, 1991

J.W. Smythe, L.V. Colom and B.H. Bland, The extrinsic modulation of hippocampal theta depends on the coactivation of cholinergic and GABAergic medial septal inputs. Neurosci. Biobehav. Rev. 16:289-308, 1992

W.R. Softky, Simple codes versus efficient codes. Curr. Opin. Neurobiol. 5:239-247, 1995

W.R. Softky and C. Koch, The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs. J. Neurosci. 13:334-350, 1993

I. Soltész, Brief history of cortico-hippocampal time with a special reference to the direct entorhinal input to CAl. Hippocampus 5:120-124, 1995

I. Soltész and M. Deschênes, Low- and high-frequency membrane potential oscillations during theta activity in CAl and CA3 pyramidal neurons of the rat hippocampus under ketamine-xylazine anaesthesia. J. Neurophysiol. 70:97-116, 1993

P. Somogyi, M.G. Nunzi, A. Gorio and A.D. Smith, A new type of specific interneuron in the monkey hippocampus forming synapses exclusively with the axon initial segments of pyramidal cells. Brain Res. 259:137-142, 1983

K.E. Sorensen and M.T. Shipley, Projections from the subiculum to the deep layers of the ipsilateral presubicular and entorhinal cortices in the guinea pig. J. Comp. Neurol. 188:313-334,1979

A. Speakman and J. O'Keefe, Hippocampal complex spike cells do not change their place fields if the goal is moved within a cue controlled environment. Eur. J. Neurosci. 2:544-555, 1990

N. Spruston and D. Johnston Perforated patch-clamp analysis of the passive membrane properties of three classes of hippocampal neurons. J. Neurophysiol. 67:508-529, 1992

L.R. Squire, memory and the hippocampus: A synthesis from findings with rats, monkeys and humans. Psych. Rev. 99:195-231, 1992

R.W. Stackman and T.J. Walsh, Baclofen produces dose-related working memory impairments after intraseptal injection. Behav. Neur. Biol. 61:181-185, 1994

O. Steward and S. Scoville, Cells of origin of entorhinal cortical afferents to the hippocampus and fascia dentata of the rat. J. Comp. Neurol. 169:347-370, 1976

M. Stewart, Disinhibition of hippocampal pyramidal cells during the transition into theta rhythm. Exp. Brain Res. 93:1-5, 1993

M. Stewart and S.E. Fox, Two populations of rhythmically bursting neurons in rat medial septum are revealed by atropine. J. Neurophysiol. 61:982-993, 1989a

M. Stewart and S.E. Fox, Detection of an atropine-resistant component of the hippocampal theta rhythm in urethane-anesthetized rats. Brain Res. 500:55-60, 1989b

M. Stewart and S.E. Fox, Firing relations of medial septal neurons to the hippocampal theta rhythm in urethane anaesthetised rats. Exp. Brain Res. 77:507-516, 1989c ______Page 299

M. Stewart and S.E. Fox, Do septal neurons pace the hippocampal theta rhythm? TINS 13:163-168, 1990

M. Stewart, Y. Luo and S.E. Fox, Effects of atropine on hippocampal theta cells and complex-spike cells. Brain Res. 591:122-128, 1992a

M. Stewart, G.J. Quirk, M. Barry and S.E. Fox, Firing relations of medial entorhinal neurons to the hippocampal theta rhythm in urethan anesthetized and walking rats. Exp. Brain Res. 90:21-28, 1992b

J.E. Storm, Potassium currents in hippocampal pyramidal cells. Progress in Brain Research 83:161-187, 1990

Ch. Stumpf, H. Petsche and G. Gogolak, The significance of the rabbit's septum as a relay station between the midbrain and the hippocampus. II. The differential influence of drugs upon both the septal cell firing pattern and the hippocampus theta activity. Electroenceph. Clin. Neurophysiol. 14:212-219, 1962

R.J. Sutherland and J.W. Rudy, Configurai association theory: The role of the hippocampal formation in learning, memory and amnesia. Psychobiology 17:129-144, 1989

L.W. Swanson, An autoradiographic study of the efferent connections of the preoptic region in the rat. J. Comp. Neurol. 167:227-256, 1976

L.W. Swanson, A direct projection from Ammon’s horn to prefrontal cortex in the rat. Brain Res. 217:150-154, 1981

L.W. Swanson and W.M. Cohen, An autoradiographic study of the efferent connections of the hippocampal formation in the rat. J. Comp. Neurol. 172:49-84, 1977

L.W. Swanson and W.M. Cohen, The connections of the septal region in the rat. J. Comp. Neurol. 186:621-656, 1979

L.W. Swanson and C. Kohler, Anatomical evidence for direct projections from the entorhinal area to the entire cortical mantle in the rat. J. Neurosci. 6:3010-3023, 1986

L.W. Swanson, C. Kohler and A. Bjorklund, The limbic region. I: The septohippocampal system. In A. Bjorklund, T. Hokfelt and L.W. Swanson (eds) Handbook of Chemical Neuroanatomy. 5; Integrated Vol Systems of the CNS, Part Elsevier,I. Amsterdam, 1987.

L.W. Swanson, J.M. Wyss and W.M. Cowan, An autoradiographic study of the organization of intrahippocampal association pathways in the rat. J. Comp. Neurol. 181:681-716, 1978

J.E. Sweeney, Y. Lamour and M.H. Bassant, Arousal-dependent properties of medial septal neurons in the unanaesthetized rat. Neurosci. 48:353-362, 1992

N. Tamamaki and Y. Nojyo, Disposition of the slab-like modules formed by axon branches originating from single CAl pyramidal neurons in the rat hippocampus. J. Comp. Neurol. 291:509-519, 1990

N. Tamamaki, K. Abe and Y. Nojyo, Columnar organization in the subiculum formed by axon branches originating from single CAi pyramidal neurons in the rat hippocampus. Brain Res. 412:156-160, 1987 ______Page 300

J.S. Taube, Head direction cells recorded in the anterior thalamic nuclei of freely moving rats. J. Neurosci. 15:70-86, 1995

J.S. Taube and H.L. Burton, Head direction cell activity monitored in a novel environment and during a cue conflict situation. J. Neurophysiol. 74:1953-1971, 1995

J.S. Taube, R.U. Muller and J.B. Ranch, Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. J. Neurosci. 10:420-435, 1990a

J.S. Taube, R.U. Muller and J.B. Ranch, Head-direction cells recorded form the postsubiculum in freely-moving rats. II. Effects of environmental manipulations. J. Neurosci. 10:436-447, 1990b

J.S. Taube, J.P. Kesslah and C.W. Cotman, Lesions of the rat postsubiculum impair performance on spatial tashs. Behav. Neur. Biol. 57:131-143, 1992

J.S. Thinschmidt, G.G. Kinney and B. Kocsis, The supramammillary nucleus: Is it necessary for the mediation of hippocampal theta rhythm? Neuroscience 67:301-312, 1995

L.T. Thompson and P.J. Best, Long-term stability of the place-held activity of single units recorded from the dorsal hippocampus of freely behaving rats. Brain Res. 509:299-308,1990

S.M. Thompson and R.T. Robertson, Organisation of subcortical pathways for sensory projections to the limbic cortex. I. Subcortical projections to the medial limbic cortex in the rat. J. Comp. Neurol. 265:175-188, 1987a

S.M. Thompson and R.T. Robertson, Organisation of subcortical pathways for sensory projections to the limbic cortex. II. Afferent projections to the thalamic lateral dorsal nucleus in the rat. J. Comp. Neurol. 265:189-202, 1987b

E.C. Tolman, Cognitive maps in rats and men. Psychol. Rev. 55:189-208, 1948

K. Toth and T.F. Freund, Calbindin D 28k“Containing nonpyramidal cells in the rat hippocampus: their immunoreactivity for GABA and projection to the medial septum. Neuroscience 49:793-805, 1992

K. Toth, Z. Borhegyi and T.F. Freund, Postsynaptic targets of GABAergic hippocampal neurons in the medial septum-diagonal band of Broca complex. J. Neurosci. 13:3712-3724, 1993

M.J. Tovée, E.T. Rolls, A. Treves and R.P. Beilis, Information encoding and the responses of single neurons in the primate temporal visual cortex. J. Neurophysio. 70:640-654, 1993

R.D. Traub, R. Miles and R.K.S. Wong, Models of synchronised hippocampal bursts in the presence on inhibition. I. Single population events. J. Neurophysiol. 58:739- 751,1987a

R.D. Traub, R. Miles, R.K.S. Wong, L.S. Shulman and J.H. Schneiderman, Models of synchronised hippocampal bursts in the presence on inhibition. II. Ongoing spontaneous population events. J. Neurophysiol. 58:752-764, 1987b

M.V. Tsodylcs, W.E. Skaggs, T.J. Sejnowski and B.L. McNaughton, Population dynamics and theta rhythm phase precession of hippocampal place cell firing: A spiking neuron model. Hippocampus 6:271-280, 1996 ______Page 301

U. Ungerstedt, Stereotaxic mapping of the monoamine pathways in the rat brain. Acta physiol. Scand. Suppl. 367:1-48, 1971

E.A. van der Zee and P.G.M. Luiten, Cholinergic and Gabaergic neurons in the rat medial septum express muscarinic acetylcholine receptors. Brain Res. 652: 263-272, 1994

T. van Groen and J.M. Wyss, Species differences in hippocampal commissural connections: studies in rat, guinea pig, rabbit and cat. J. Comp. Neurol. 267:322-334, 1988

T. van Groen and J.M. Wyss, The connections of presubiculum and parasubiculum in the rat. Brain Res. 518:227-243, 1990a

T. van Groen and J.M. Wyss, The postsubicular cortex in the rat: characterization of the fourth region of the subicular cortex and its connections. Brain Res. 529:165-177, 1990b

T. van Groen and J.M. Wyss, Projections from the anterodorsal and anteroventral nucleus of the thalamus to the limbic cortex in the rat. J. Comp. Neurol. 358:584-604, 1995

T. van Groen, F.J. van Haren, M.P. Witter and H.J. Groenewegen, The organization of the reciprocal connections between the subiculum and the entorhinal cortex in the cat: I. A Neuroanatomical tracing study. J. Comp. Neurol. 250:485-497, 1986

G.W. van Hoesen, The parahippocampal gyrus. New observations regarding its cortical connections in the monkey. TINS 5:345-350, 1982

G.W. van Hoesen, D.N. Pandya and N. Butters, Cortical afferents to the entorhinal cortex of the rhesus monkey. Science 175:1471-1473, 1972

C.H. Vanderwolf, Hippocampal electrical activity and voluntary movement in the rat. Electroenceph. Clin. Neurophysiol. 26:407-418, 1969

C.H. Vanderwolf and L.-W.S. Leung, Hippocampal rhythmical slow activity: A brief history and the effects of entorhinal lesions and phencyclidine. In: G.W. Seifert (ed) Neurobiology of the Hippocampus Academic Press, New York, 1983

C.H. Vanderwolf, R. Kramis and T.E Robinson, Hippocampal electrical activity during waking behaviour and sleep: analyses using centrally acting drugs. In: Functions of the septo-Hippocampal System. CIBA Foundation Symposium Vol 58 Churchill, London pp 199-226, 1978

R.P. Vertes, Brain stem generation of the hippocampal. EEG. Prog Neurobiol. 19:159- 186, 1982.

R.P. Vertes, Brainstem afferents to the basal forebrain in the rat. Neuroscience 24:907- 935, 1988.

R.P. Vertes, PHA-L analysis of projections from the supramammillary nucleus in the rat. J. Comp. Neurol. 326:595-622, 1992

R.P. Vertes, G.G. Kinney, B. Kocsis and W.J. Fortin, Pharmacological suppression of the median raphe nucleus with serotonin ia agonists, 8-OH-DPAT and buspirone, produces hippocampal theta rhythm in the rat. Neuroscience 60:441-451, 1994 ______Page 302

B.A. Vogt and M.W. Miller, Cortical connections between rat cingulate cortex and visual, motor, and postsubicular cortices. J. Comp. Neurol. 216:192-210, 1983

B.H. Wainer, A.I. Levey, D.B. Rye, M.M. Mesulam and E.J. Mufson, Cholinergic and non-cholinergic septohippocampal pathways. Neurosci. Lett. 54:45-52, 1985

E.G. Wever, Theory of Hearing, Wiley, New York, 1949

I.Q. Whishaw and C.H. Vanderwolf, Hippocampal EEG and behaviour: Effects of variation in body temperature and relation of EEG to vibrissae movement, swimming and shivering. Physiol. Behav, 6:391-397, 1971

I.Q. Whishaw and C.H. Vanderwolf, Hippocampal EEG and behaviour: Changes in amplitude and frequency of RSA (theta rhythm) associated with spontaneous and learned movement patterns in rats and cats. Behav, Biol. 8:461-484, 1973

S.I. Wiener, C.A. Paul and H. Eichenbaum, Spatial and behavioural correlates of hippocampal neuronal activity. J. Neurosci. 9:2737-2763, 1989

S. Williams and D. Johnston, Muscarinic depression of long-term potentiation in CA3 hippocampal neurons. Science 242:84-87, 1988

S. Williams, D.D. Samulack, C. Beaulieu and J.C. Lacaille, membrane properties and synaptic responses of interneurons located near the stratum lacunosum- moleculare/radiatum border of area CAl in whole-cell recordings from rat hippocampal slices. J. Neurophysiol. 71:2217-2235, 1994

C.L. Wilson, B.C. Motter and B.D. Lindsley, Influences of hypothalamic stimulation upon septal and hippocampal electrical activity in the cat. Brain Res. 107:55-68, 1976

M.A. Wilson and J.M. Bower, The simulation of large-scale neural networks. In: C. Koch and I. Segev (eds) Methods in Neuronal Modelling MIT Press, Cambridge, Mass, 1989

M.A. Wilson and B.L. McNaughton, Dynamics of the hippocampal ensemble code for space. Science 261:1055-1058, 1993

J. Winson, Patterns of hippocampal 0 rhythm in the freely moving rat. Electroenceph. Clin. Neurophysiol. 36:291-301, 1974

J. Winson, Hippocampal theta rhythm. II. Depth profiles in the freely moving rabbit. Brain Res. 103:71-79, 1976

G. Winocur, The hippocampus and cue utilization. Physiol. Psych. 8:280-288, 1980

M.P. Witter, Connectivity of the rat hippocampus. In: The Hippocampus - New Vistas pp53-69 ed V. Chan-Palay and C. Kohler, A.R. Liss Inc., New York, 1989

M.P. Witter, Organization of the entorhinal-hippocampal system: A review of current anatomical data. Hippocampus 38:33-44, 1993

M.P. Witter and H.J. Groenewegen, Connections of the parahippocampal cortex in the cat. III. Cortical and thalamic efferents. J. Comp. Neurol. 252:1-31, 1986

M.P. Witter, A.W. Griffioen, B. Jorritsma-Byham and J.L.M. Krijnen, Entorhinal projections to the hippocampal CAl region in the rat: an underestimated pathway. Neurosci. Lett. 85:193-198, 1988 ______Page 303

M.P. Witter, G.W. van Hoesen and D.G. Amaral, Topographical organization of the entorhinal projection to the dentate gyms of the monkey. J. Neurosci, 9:216-228, 1989

W. Woodson, L, Nitecka and Y. Ben-Ari, organization of the GABAergic system in the rat hippocampal formation: A quantitative immunocytochemical study. J. Comp. Neurol. 280:254-271, 1989

J.M. Wyss, L.W. Swanson and W.M. Cowan, A study of subcortical afferents to the hippocampal formation in the rat. Neuroscience 4:463-476, 1979

H.W. Yang, P. Grimwood, Z. Henderson and K. Appenteng, Neuron types identified in the brain slice of the rat medial septal/diagonal band complex. Brain Res. Assoc. Abstr. 12:83, 1995

M.F. Yeckel and T.W. Berger, Feedforward excitation of the hippocampus by afferents from the entorhinal cortex: redefinition of the role of the trisvnaptic pathway. Proc. Natl. Acad. Sci. USA 87:5832-5836, 1990

A. Ylinen, A. Bragin, Z. Nadasdy, G. Jando, I Szabo, A. Sik and G. Buzsaki, Sharp wave-associated high-frequency oscillation (200 Hz) in the intact hippocampus: Network and intracellular mechanisms. J. Neurosci. 15:30-46, 1995a

A. Ylinen, I Soltész, A. Bragin, M. Penttonen and G. Buzsaki, Intracellular correlates of hippocampal theta rhythm in identified pyramidal cells, granule cells, and basket cells. Hippocampus 5:78-90, 1995b

N. Yoshii, M. Shimokochi, K. Miyamoto and M. Ito, Studies on the neural basis of behaviour by continuous frequency analysis of EEG. Prog Brain Res. 21a:217-250, 1966

S.M. Zeki, The representation of colours in the cerebral cortex. Nature 284:412-418, 1980

S. Zeki and S. Shipp, The functional logic of cortical connections. Nature 335:311-317, 1988

S. Zola-Morgan and L.R. Squire, Memory impairment in monkeys following lesions limited to the hippocampus. Behav. Neurosci. 100:155-160, 1986

S. Zola-Morgan and L.R. Squire, The primate hippocampal formation: Evidence for a time-limited role in memory storage. Science 250:288-290, 1990

S. Zola-Morgan and L.R. Squire, Neuroanatomy of memory, Ann. Rev. Neurosci. 16:547-563,1993

S. Zola-Morgan, L.R. Squire and D. Amaral, Human amnesia and the medial temporal region: Enduring memory impairment following a bilateral lesion limited to the CAl field of the hippocampus. J. Neurosci. 6:2950-2967, 1986 ______Page 304 Acknowledgements

I would like to thank my supervisor, Michael Recce and Professor John O’Keefe for their invaluable assistance, and Hajime Hirase and Ken Harris for their useful suggestions. Alan Ainsworth, Clive Parker and Jim Donnett provided indispensable technical help.

The author was supported while working on this thesis by a grant from the Wellcome Trust. Professor O’Keefe’s laboratory, in which the research was conducted, is funded by a programme grant from the Medical Research Council.