Download Corticofugal modulation of functional connectivity within the auditory

Journal of Neuroscience Methods 86 (1999) 161 – 178
Corticofugal modulation of functional connectivity within the auditory
thalamus of rat, guinea pig and cat revealed by cooling deactivation
Alessandro E.P. Villa a,*, Igor V. Tetko a,b, Pierre Dutoit a, Yves De Ribaupierre a,
François De Ribaupierre a
b
a
Institut de Physiologie, Faculté de Médecine, Uni6ersité de Lausanne, Rue du Bugnon 7, CH-1005 Lausanne, Switzerland
Department of Biomedical Applications, IBPC, Academy of Sciences of Ukraine, Murmanskaya 1, Kie6-660, 253660, Ukraine
Abstract
Microelectrode recordings were simultaneously performed at multiple sites in the medial geniculate body (MGB) of anesthetized
cats, rats and guinea pigs. We studied the effect of cortical deactivation on the association of neural activity within the thalamus
during spontaneous activity. The corticofugal influence was suppressed by temporary cooling of the auditory cortex. Pairs of spike
trains recorded from the same electrode were distinguished from cases where units were in MGB but recorded with different
electrodes. Time domain analyses included crosscorrelations and search for precise repetition of complex spatiotemporal firing
patterns of reverberating thalamic circuits. As a complementary approach we performed bispectral analyses of simultaneously
recorded local field potentials in order to uncover the frequency components of their power spectra which are non linearly
coupled. All results suggest that new functional neuronal circuits might appear at the thalamic level in the absence of input from
the cortex. The newly active intrathalamic connections would provide the necessary input to sustain the reverberating activity of
thalamic cell assemblies and generate low frequency non-linear interactions. The dynamic control exerted by the cortex over the
functional segregation of information processing carried out in the thalamus conforms with theoretical neural network studies and
with the functional selectivity-adaptive filtering theory of thalamic neuronal assemblies. Although this general conclusion remains
valid across species, specific differences are discussed in the frame of known differences of the microcircuitry elements. © 1999
Elsevier Science B.V. All rights reserved.
Keywords: Reversible inactivation; Auditory thalamus; Medial geniculate body; Auditory cortex; Single unit; Crosscorrelations;
Bispectral analysis; Spatiotemporal patterns
1. Introduction
The thalamus is the main source of input to the
sensory areas of the cerebral cortex of mammals. Reciprocal connections exist between these structures and
sensory information is processed in a complex way in
the thalamus, so it cannot be regarded as a mere relay
station conveying peripheral inputs to the cortex
(Jones, 1985; Steriade and Llinas, 1988). In all sensory
modalities multiple modules characterized by the same
basic connectivity are assumed to work in parallel and
include three main components: (1) dorsal thalamic
* Corresponding author. E-mail: [email protected]
neurons which receive the sensory input from the periphery (e.g. in the medial geniculate body, MGB, for
the auditory pathway or in the lateral geniculate nucleus, LGN, for the visual pathway); (2) cells of the
thalamic reticular nucleus (RE), a major component of
the ventral thalamus and (3) the cortical area receiving
the corresponding thalamic input. Corticothalamic
fibres make excitatory synaptic contacts on the distal
dendrites of thalamic principal cells, as well as on
inhibitory local circuit cells (Golgi Type II, inhibitory
interneurons) of the main thalamic sensory nuclei and
RE (Jones 1985). In addition, it is important to note
that RE sends inhibitory projections back to both
principal cells and local circuit cells in the main thalamic nuclei, although the presence of these local in-
0165-0270/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved.
PII S0165-0270(98)00164-2
162
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
hibitory neurons is species-specific and varies across the
different sensory thalamic subdivisions (Table 1).
The technique of reversible deactivation of the cerebral cortex (Payne et al., 1996) is a tool of importance
to study the cortical influence on thalamic activity. Of
the available deactivation techniques, reversible cooling
offers several advantages, in particular because steady
state deactivated conditions can be maintained over a
period of time long enough to allow the recording of
spike trains during several conditions of stimulation.
However, due to the complexity of the thalamocortical
circuitry and to the different experimental conditions,
there is still controversy about the effect of cortical
deactivation on thalamic information processing. Cooling deactivation of cat and rat primary visual cortex has
only limited and variable influences on LGN neurons
(Baker and Malpeli 1977; Geisert et al. 1981; Kayama et
al. 1984; McClurkin and Marrocco, 1984; Marrocco
and McClurkin, 1985), although a predominantly excitatory effect of corticofugal fibres in cat was clearly
observed in other cortical deactivation experiments
(Kalil and Chase, 1970; Singer, 1977; Tsumoto et al.,
1978). In rats, the effect of inactivation of the barrel
field cortex showed that the transmission of sensory
information through the ventral posterior medial nucleus of the thalamus (the primary pathway) occurs
independently of the state of the cortex. In contrast,
transmission through the rostral sector of the posterior
complex (the non-primary pathway) depends upon the
state of the cortex itself (Diamond et al., 1992). Cooling
deactivation of somatosensory areas SI and/or SII indicate that the corticofugal fibres exert mainly a facilitatory influence in rat (Yuan et al., 1986) as well as in cat
(Ghosh et al., 1994) somatosensory thalamus. In the
auditory pathway, an excitatory or facilitatory role of
the cortical input to MGB has been suggested in cats
Table 1
Species-specific and modality-specific relative frequency of local inhibitory neurons in the sensory dorsal thalamic nuclei
Thalamic nucleus
Rat
Guinea Pig
Cat
Medial geniculate body (auditory)
Ventrobasal complex (somesthetic)
Lateral geniculate nucleus (visual)
0a
0b
+++c
0d
++e
+++f
++g
++e
++h
Relative frequency: 0 =range 0–3%; ++= range 15–20%; +++ =
range 25–30%.
a
Winer and Larue, 1988.
b
Thompson et al., 1985.
c
Winer, 1985.
d
Harris and Hendrickson, 1987.
e
Spreafico et al., 1994.
f
Penny et al., 1983.
g
Ohara et al., 1983.
h
Montero, 1986.
(Orman and Humphrey, 1981; Villa, 1988; Villa et al.,
1991).
The role of corticothalamic modulation is clearly
prominent when stimulus-related activity is analyzed.
Results on the somatosensory system show that the
cortical areas may modulate the gain of transmission of
tactile signalling through the thalamus (Ghosh et al.,
1994) and visual-dependent complex balance of excitatory and inhibitory inputs exerted on the LGN principal
cells can induce correlated firing under control of cortical feedback (Sillito et al., 1994). Acoustically-evoked
activity in thalamic single units is often characterized by
a complex response pattern which can be modified in its
time course and in its components by cooling of the
auditory cortex (Villa et al., 1991). Corticofugal modulation could regulate the response properties of thalamic
units by modifying their firing rate and bandwidth
responsiveness to pure tones. This processing (called
‘functional selectivity-adaptive filtering’ theory) would
allow to selectively extract information from the incoming signals according to the cortical activity (Villa, 1988;
Villa et al., 1991; Tetko and Villa, 1997a).
These studies indicate an increasing interest in a
search for a general rule in understanding the modulatory mechanism exerted by cortical activity on thalamic
information processing, but due to modality- and species-specific features of the thalamocortical circuitry no
clear overview of the cortical feedback can be drawn so
far. The purpose of this study was to investigate the
corticofugal influence on MGB activity in three representative animal models: the cat, the guinea pig and the
rat. Reversible inactivation of the primary auditory
cortex by mean of a cooling chamber refrigerated by a
cool fluid is shown to be a valuable experimental set-up
for this study. We will demonstrate the importance of
applying complementary analytical methods of the electrophysiological signals for the study of correlated activity within the thalamus. Simultaneous recordings of cell
discharges in the auditory thalamus prior to, during and
after cooling deactivation allowed us to estimate to
which extent the thalamic functional connectivity is
controlled by the activity of the auditory cortex. Significant interactions between two thalamic cells were estimated by crosscorrelograms and comparisons between
cell pairs recorded from the same and from different
electrodes were performed. In addition, analysis of
crossbispectra was used to reveal nonlinear interactions
in the frequency domain from simultaneous recording of
local field potentials.
2. Materials and methods
2.1. Surgical and experimental procedures
The experiments were carried out in compliance with
Swiss guidelines for the care and use of laboratory
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
animals and after receiving governmental veterinary
approval. The animals were placed in a sound-attenuating room. All subjects were given atropine sulphate
intraperitoneally (0.08 mg/kg) immediately prior to
surgery as a prophylactic against respiratory distress.
All surgical wounds were infiltrated with the local
anesthetic Effocaine. In cats (n =7, weighing 1.5–2.9
kg) surgery was performed under deep anesthesia
(sodium pentobarbital, 40 mg/kg). After surgery the
cats were muscle relaxed (gallamine triethiodide, 10
mg/kg per h) and artificially ventilated with a mixture
of 80% N2O and 20% O2. The arterial pulse, rectal
temperature, and concentrations of CO2 and O2 in the
expired air were continuously monitored and the pupil
size periodically checked. The guinea pigs (Ca6ia porcellus GOHI strain, n =2, weighing 310 and 370 g)
were injected intraperitoneally with diazepam (8 mg/kg)
followed 20 min later by sodium pentobarbital (20
mg/kg). In rats (Long – Evans hooded rats, n= 2,
weighing 235 and 250 g) anesthesia was induced by an
i.m. injection (1.0 ml/kg body weight) of a 4:3 mixture
of ketamine and xylazine hydrochloride. This dose corresponded to 57 mg/kg of ketamine and 8 mg/kg of
xylazine. In guinea pigs and rats anesthesia was maintained during the whole recording session by supplementary intramuscular injections of approximately 0.3
ml/kg of the 4:3 mixture of ketamine and xylazine,
nearly every 90 min. The limb withdrawal reflex was
checked regularly in order to monitor the depth of
anesthesia. All animals were mounted in a stereotaxic
apparatus without earbars. Body temperature was
maintained between 37 and 39°C, by means of a heating pad. Openings for the thalamic microelectrodes and
for the cortical cooling probe were drilled through the
skull. The dura mater was removed from the cortical
area corresponding to the penetration of the thalamic
electrodes. Upon completion of the recording session
(lasting approximately 60 h in cats and 12 h in rats and
guinea pigs), an electrolytic lesion was placed at a
specific depth for each track, by passing a current of
about 8 mA for 10 s. The animals then received a
sub-lethal dose of sodium pentobarbital i.p. (180 mg/kg
body weight) and were perfused transcardially with
0.9% NaCl immediately followed fixative solution (4%
paraformaldehyde in phosphate buffer 0.1 M, pH 7.3).
The brains were removed after the perfusion, postfixed
and prepared for standard histological procedures.
More details on the surgical procedures are described
elsewhere (Villa et al., 1991, 1996).
2.2. Multisite electrophysiological recordings
Extracellular single unit recordings in the auditory
thalamus were made with glass-coated platinum-plated
tungsten microelectrodes having an impedance in the
range 0.5–2 MV measured at a frequency of 1 kHz
163
(Frederick Haer, Brunswick, Maine). The experiments
were performed with multielectrode devices constructed
at the Institute of Physiology of the University of
Lausanne (Fig. 1), which are now available commercially (Alpha Omega Engineering, Nazareth, Israel). In
these devices four microelectrodes are independently
controlled. According to the experimental constraints
and data acquisition capabilities it is possible to use
several multielectrode devices in the same set-up.
Briefly, the uninsulated end of each microelectrode is
kinked and inserted into a 24G (0.55 mm external
diameter, 48 mm long) stainless steel tube, the ‘drive
cannula’. The kink provides both mechanical fixation
and electrical contact for the electrodes. The drive
cannula slides within a fixed 20G (0.9 mm external
diameter, 50 mm long) guide tube, and is soldered to a
flexible stainless steel shaft (0.25 mm diameter) which is
in turn attached to a remotely controlled DC micromotor (4 V, 0.075 mA). The shaft is protected and insulated by a Teflon external sleeve. The electrodes extend
approximately 80 mm from the drive cannula and are
fed into stereotaxically oriented guide tubes. The microelectrodes can be advanced by steps or in continuous mode at a constant speed of 2 mm per 10 ms. A
magnetic incremental shaft encoder (model 30B20,
Minimotor SA, Agno, Switzerland) translates the rotational movement of the shaft into an electrical waveform. Two similar channels with a 90° phase difference
provide velocity and direction information. Digital output pulses for each channel (ten pulses in one complete
revolution) are generated by a comparator which
switches when the analog values are equal. The absolute advance and the relative depth to a zero point can
be read remotely and saved on a computer file. The
above description corresponds to the most recent
model, used on rat and guinea pigs experiments. Cat
experiments were performed with variants of this model
(especially in the length of the flexible shaft and the
measurement of electrode advance). Up to six microelectrodes were advanced independently in cat experiments but four electrodes were used in rat and
guinea pig thalamic recordings. The distance between
the stereotaxically oriented guides ranged between 360
and 600 mm. In most cases two microelectrodes were
inserted in the same guide.
One electrode aimed at the center of the rostro-caudal extension of MGB was advanced at first. Both the
activity recorded along the track and the first acoustically evoked activity were used to determine if the
region under recording corresponded to the electrophysiological activity of MGB. The other electrodes
were advanced in the auditory thalamus when the first
electrode track matched the required criteria and as
many as possible single units were isolated on all electrodes. If no acoustically evoked activity could be observed on one electrode, then the signals recorded from
164
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
Fig. 1. Schematic diagram of the microelectrode setup. Each electrode is connected to a moving stainless steel shaft independently driven by a
micromotor. One to three electrodes can fit into a stereotaxically oriented guide. The geometry of the electrode guide holding system is
independent from the electrode advance setup so that the whole system is adjustable to the targeted neuronal structures and to the subject’s brain.
that electrode were discarded from the dataset analyzed
in this study. The first recording session of an experiment started about 4 h after the beginning of surgery in
guinea pig and rat experiments and after 16 h in cat
experiments. Data were gathered in blocks of 80 –400 s.
Recordings during a total of 300 – 1000 s without external stimulation, referred to as spontaneous activity,
were collected. A set of auditory stimuli was used to
characterize the response properties to pure tones,
white noise bursts and complex sounds (Villa et al.,
1991). The results relative to the acoustically evoked
activity are not discussed in this study. A full protocol
of activity characterization was performed prior to
cortical inactivation. This protocol lasted on average 45
min. During cooling of the primary auditory cortex, a
simplified protocol was performed, lasting approximately 30 min. After cooling the same simplified protocol was performed in order to check the recovery of the
activity. Generally, no injection of drugs was done
during this period. If a sign of distress was noticed and
a supplementary dose of anesthetic was needed the
recording was paused and resumed at least 15 min after
the injection of the drug. We assume that these recording conditions corresponded to a steady level of
anesthesia.
In cat experiments up to two distinct single units
could be recorded from the same microelectrode using
an analog template matching spike sorter according to
a technique described elsewhere (Ivarsson et al., 1988;
Villa, 1990). In the other experiments, a commercially
available digital template-matching spike sorter (MSD,
Alpha Omega Engineering, Nazareth, Israel), allowing
sorting up to three distinct single units per channel, was
used in addition to or instead of the analog multi-unit
analyzer. The overall number of simultaneously
recorded single units was up to eight in cat and up to 15
in rat and guinea pig experiments. The spike firing
times of each unit were given by interrupts generated by
a digital acquisition board driven by an external clock
with accuracy of 1 ms and stored digitally for off-line
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
analyses. Dot rasters of the spike trains were displayed
in real-time. The local field potential (LFP) from each
electrode was measured by bandpass filtering the neural
signal at 7–100 Hz (MCP-8000, Alpha Omega Engineering, Nazareth, Israel) with a notch filter at a 50 Hz
cut-off frequency (-34 db attenuation) and a bandwidth9 2 Hz. The analog LFPs were sampled with 2
ms resolution at 12-bit precision (National Instruments
NB-MIO-16H-9 I/O board connected to a NBDMA2800 board) and stored in files for off-line analyses on a Apple Macintosh IIfx microcomputer (running
version 7.1 of MacOS).
2.3. Cooling technique
Separate skull-holes were drilled for the thalamic
microelectrodes and for the cortical cooling probe.
However, according to the size of the probes and to the
rostro-caudal distance of the planes corresponding to
auditory thalamic and cortical structures the two holes
sometimes formed a contiguous opening. The cooling
probes were placed above the cortical area referred to
as primary auditory cortex, stereotaxically determined
following the lambda and bregma landmarks for rats
(Zilles and Wree 1985; Paxinos and Watson, 1986) and
guinea pigs (Redies et al., 1989) and following the
suprasylvian, anterior ectosylvian and posterior ectosylvian sulci in cats (Reale and Imig, 1980). The cooling
probe was fixed to the skull with dental cement. A
different model of probe was used for each species, but
all consisted of an aluminium cylindrical core, refrigerated by a circulating mixture of chilled ethanol and
ice-cold water, in tight contact with the dura mater.
Fig. 2 illustrates the three set-ups used in this study.
The surface of the probe in contact with the cat’s brain
was circular (diameter 10 mm), whereas with guinea
pig’s (9× 6 mm) and rat’s brain (5×4 mm) the surface
was rectangular. In addition, in the rat probe the
surface was slightly curved. In the cat probe the flux of
the refrigerant liquid was maintained at 100 ml/min,
whereas the flux was 12 ml/min in the other probes, by
means of a DC peristaltic pump. This difference in the
flux rate was due to a specific design of tubular sleeves
around the aluminium core of the cat probe which
resulted in a larger section of the circulating fluid. The
tubing was mounted in such a way that the refrigerant
flowed from the chilled reservoir to the cooling probe,
then from the probe to the peristaltic pump entrance
and back to the reservoir.
The temperature of the probe-end proximal to the
cerebral cortex was monitored continuously by a thermocouple and the degree of reversible inactivation of
cortical activity was assessed by means of microelectrodes (at least one) monitoring the auditory evoked
responses in the cortex. Cortical rewarming was
achieved by interrupting circulation of the cooling fluid.
165
The cooling and rewarming kinetics were observed to
be similar in all experiments. The steady-state was
reached in 15–20 min and a detailed analysis of the
kinetics as a function of depth in the cortex was studied
for the cat probe (Villa et al., 1991). During cooling the
temperature of the probe was near 5°C and the temperature of the deep cortical layers in the range 16–22°C,
low enough to inactivate cortical synapses (Brooks,
1983). Although portions of surrounding cortical fields
would have probably been affected by a lateral spread
of cooling or by a partial cooling if the probe was
slightly overlapping these areas, we will refer to the
cooling as primary auditory cortex inactivation
throughout this article.
2.4. Data analysis
Spike trains were analyzed with the help of time
series renewal density histograms according to Abeles
(1982). Using this technique all histograms were scaled
in rate units (spikes/s) and smoothed after convolution
with a moving Gaussian-shaped bin of 10 ms width. In
particular, the functional interaction between single
units was assessed by computing the crosscorrelograms.
Peaks in the crosscorrelograms were identified as crossings of the 99% confidence limits, calculated assuming
that a Poisson distribution underlay the spike train
discharges. Peaks spanning time zero (bilateral peaks)
were interpreted as synchronous firing of the pair of
units due to a shared input, referred to as ‘common
input’, CI. Peaks near to but on one side only of time
zero (unilateral peaks) were interpreted as one cell
increasing the probability of firing of the other cell,
referred to as ‘direct excitation’, Direct. Unilateral
troughs, bilateral troughs, peaks far from the zero delay
and more complex features of the correlograms were
observed in only a few cases and will not be discussed
further in the present report.
However, crosscorrelograms provide only a rough
average picture of the temporal interactions between
spike trains and their value is almost limited to the
quasi-synchronous domain (near time zero delay). They
cannot, for instance, detect recurrent activity in a cell
assembly because complex spatio temporal patterns of
firing are generated. On the other hand, statistically
significant repeated appearance of an identical firing
pattern indicates that the corresponding single units
have been repeatedly engaged in some kind of information processing carried out by a single cell assembly.
Therefore the existence of an excess, over chance expectancy, of spatiotemporal patterns of firing reveals the
presence of cell assembly information processing. The
estimation of the number of precise patterns of discharges that occur by chance in records of multiple
single unit spike trains was performed by combinatorial
and probabilistic combinatorial methods (PCM) (Tetko
166
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
Fig. 2.
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
and Villa, 1997b). The confidence limits estimated by
these methods are in good agreement with different sets
of simulated test data as well as with the ‘ad-hoc’
method (Abeles and Gerstein, 1988) but in non-stationary data PCM provided better estimations than the
ad-hoc method. A jitter for searching patterns with a
precision of a few milliseconds and burst filtering procedures were also introduced (Tetko and Villa, 1997b),
because of the tendency of spike trains to increase
bursting activity during cortical inactivation (Villa et
al., 1991).
A further assessment of the degree of coordinated
activity within a localized population of neurons was
achieved by applying third order frequency domain
analysis to LFP recordings. As second order cumulant
statistics decompose the electrophysiological signal into
a linear combination of mutually uncorrelated frequency components they can be applied only to stationary Gaussian signals. Conversely, third order cumulant
statistics, i.e. bicorrelation, bispectrum and bicoherence,
are based on skewness so that they can help to detect if
the signal deviates from normality, i.e. is non-Gaussian
(Brillinger, 1965; Huber et al., 1971; Nikias and Raghuveer, 1987). In addition, third order cumulants retain
the phase relationship between the signal components
and thus can detect if some of them are non linearly
coupled. In brain electrical activity, such non-linearities
occur when two frequency components are in harmonic
resonance or if two brain electrical waves interact and
generate a new energy component at a frequency which
is the sum of the frequency of its primary components.
Such phase coupling can be detected by third order
spectral (bispectral) analysis because the phase of the
generated components is also the sum of the phase of
its primary components (Nikias and Raghuveer, 1987).
Bispectra and cross-bispectra were estimated by using
the Fourier method (Matlab v4.2c, The MathWorks,
Natick, MA). In order to detect significant peaks we
used bicoherence (bic) and crossbicoherence (xbic) (i.e.
normalized indexes of bispectrum and crossbispectrum). The comparison between distributions of the
corresponding frequencies of non-linear interaction in
the bispectra can reveal, to some extent, if the information processing tends to be limited locally (high frequencies will predominate, ‘local processing’) or if it
167
tends to spread within a wider thalamic area (low
frequencies will predominate, ‘distributed processing’).
3. Results
The data presented here refer only to the sample of
well isolated single units maintained throughout the
experimental protocol and when at least two spike
trains were recorded simultaneously (n = 193 in cats,
n =37 in guinea pigs, n= 48 in rats). Pairs of spike
trains were differentiated if they were recorded from the
same electrode (SE) or from different electrodes (DE) in
the crosscorrelation analysis.
The significant crosscorrelograms were classified as
‘common input’ (CI) if a symmetrical peak near time
zero was observed, thus suggesting a synchronization
between the times of firing of the single units, or as
‘Direct’, if an asymmetrical unilateral peak was observed. All other types of significant correlogram features pooled together accounted for 0–5%, according
to species, of the significant number of crosscorrelograms. Therefore, for statistical reasons and for the
sake of comparing the experiments performed on different animal models we limit the present study to the two
major classes CI and Direct. The qualitative effect of
reversible deactivation of the primary cortex on the
synchronization of the spike trains was determined by
counting the correlogram features that were maintained
across all experimental conditions (‘stable’), those that
went away during cooling but reappeared during the
recovery period (‘disappearing’) and those that were
only observed during cooling (‘appearing’). In several
cases the correlogram type was not modified and it was
classified in the ‘stable’ class but an increase (Fig. 3a) or
a decrease (Fig. 3b) in strength could be observed. A
preliminary analysis indicated that the strength of the
interactions was increased twice as often during cooling
as decreased. The assessment of the quantitative
changes can be done by computing the asynchronous
gain (Abeles, 1991), but these results are not presented
here and will not be discussed further. In other cases
during cortical inactivation an additional significant
feature of the correlogram appeared. Fig. 3c illustrates
one such case with the appearance of a Direct interac-
Fig. 2. Schematic representation of the anatomical location of the regions of interest and of cooling devices used in cat (a), rat (b) and guinea
pig (c). The left panels represent a view of the left hemisphere of the cerebral cortex for each species. The scales are different. The right panels
show two frontal sections, for each species, representative of the rostro-caudal levels (interaural coordinates are indicated near the corresponding
sections) of the auditory cortex and auditory thalamus. The most rostral section illustrates also the cooling probe fixed to the skull with dental
cement (in gray). See text for more details on the cooling probes. Arrows indicate the direction of the refrigerant flow. The thermocouple (T°C)
used to monitor the probe temperature and the microelectrode (melectrode) used to monitor the auditory evoked potentials are also indicated. The
scale bar on the right panels corresponds to 5 mm. Abbreviations: A, anterior auditory cortex in guinea pig; AI, primary auditory cortex in cat;
AII, secondary auditory cortex in cat; aes, anterior ectosylvian sulcus; Hi, hippocampus; LGN, lateral geniculate nucleus; LT, lateral nucleus of
the thalamus; MGB, medial geniculate body; OT, optic tract; pes, posterior ectosylvian sulcus; ps, pseudosylvian sulcus; R, red nucleus; rf, rhinal
fissure; SC, superior colliculus, SN, substantia nigra; ss, suprasylvian sulcus; sy, sylvian sulcus; Te1, temporal area 1 (rat primary auditory cortex);
Te3, temporal area 3 (rat non-primary auditory cortex).
168
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
Fig. 3.
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
tion during cooling without a simultaneous modification of the CI. On the opposite, Fig. 3d illustrates an
increase in synchronization between two cells induced
by deactivation of the primary cortex, even if one cell
was steadily exciting the other cell. Superimposed features in the same graph were observed in barely 5% of
the overall number of significant crosscorrelograms. In
these cases two separate counts, one for each feature,
are reported in the distribution of crosscorrelogram
types and the total number of cell pairs is increased by
one for all such cases. Because of the limited number of
these observations, no particular procedure was applied
to correct these numbers.
Table 2 summarizes the results of the qualitative
count of significant correlograms across species and in
all sampled groups. During control and recovery conditions the vast majority of the correlograms in the SE
pairs showed a significant pattern of correlation (56%
in cats, 56% in guinea pigs and 77% in rats). These
correlograms were predominantly of CI type, but the
proportion of Direct versus CI was higher in guinea
pigs and rats (3/22, and 5/32, respectively) than in cats
(2/67). In most cases, different electrodes were in different subdivisions of MGB. Preliminary analysis could
not differentiate statistically the relative frequency of
significant crosscorrelograms from DE recordings in the
same or different subdivision. Therefore all data
recorded from DE were pooled together. Only a minority of significant correlograms (8% in cats, 8% in guinea
pigs and 6% in rats) occurred in crosscorrelations between DE.
The picture of a relatively similar distribution across
species does not extend to the analysis of the corticofugal influence (Table 2). In rats and guinea pigs the
number of ‘disappearing’ CI correlograms during cooling, in both SE and DE samples, was always larger
than the number of ‘stable’ CI. Conversely, the reversible deactivation of cat’s primary auditory cortex
tended to leave the majority of CI unaltered. However,
in cat’s data a remarkable difference exists between
pairs recorded from the same electrode or from different electrodes. In the SE group we observed as many
‘appearing’ and ‘disappearing’ CI, but in the DE group
the cortical deactivation tended to produce a larger
169
number of synchronizations. The corticofugal effect on
the Direct type of correlogram was uniform across all
species and was always characterized by a majority of
‘appearing’ correlations. This tendency was much
stronger between cells recorded from the same electrode. A general overview of the corticofugal modulation can be drawn by plotting the percentage of
‘disappearing’ versus the percentage of ‘appearing’ correlograms with respect to the significant correlograms
observed in the control and recovery conditions (Fig.
4). In summary, reversible cortical deactivation tended
to increase the number of apparent direct interactions
between thalamic neurons at both local (SE) and areal
(DE) levels. Simultaneously, the number of shared inputs tended to decrease in rat and guinea pig but it
remained moderately unaffected at the local level in cat
thalamus.
Further indirect evidence of the corticofugal control
on thalamic cell assemblies synchronization is provided
by the analysis of complex spatiotemporal patterns of
spikes. The number of spikes necessary for a fair statistical evaluation of the significance of the patterns could
not be met in all datasets reported here. This is due not
only to an insufficient time of recording in certain
groups of spike trains, but also to the effect of massive
decrease in selected single units firing rate induced by
cortical cooling, so that only a limited number of spike
trains recorded simultaneously could match the requirements to be analyzed during all experimental conditions. A detailed analysis of the spatiotemporal patterns
requires also a systematic evaluation of the complexity
of the pattern (i.e. the number of spikes forming the
pattern, the number of exact repetitions of the pattern
within the record, the number of distinct single units
contributing to the pattern and the time interval structure of the significant patterns). This goal is far beyond
the scopes of the present paper and is discussed in
detail elsewhere (Villa et al., 1998).
A preliminary analysis showed that cortical deactivation can disrupt precise timed activity within the thalamus in a reversible way. Fig. 5 shows an example of a
pattern formed by three spikes belonging to two single
units recorded from two different electrodes in the rat
auditory thalamus. The pattern is formed by cell 7
Fig. 3. Examples of corticofugal modulation on thalamic crosscorrelations. Abscissa: time (ms). Ordinate: rate (spikes/s). The ordinate corresponds
to the instantaneous firing rate of one cell (follower) at variable delays before (negative times on the abscissa) and after (positive times) the firing
of the other cell (trigger) of a neuronal pair. (a) The main feature of these crosscorrelograms, between a cell pair recorded from one electrode,
is an asymmetrical peak on one side of time zero, referred to as Direct interaction. Cortical deactivation increased the hump; this is the
predominant effect observed across species. (b) The effect of cortical cooling is the opposite of that shown in (a). The Direct interaction almost
disappeared in absence of cortical activity. This cell pair was recorded from one electrode. (c) Complex pattern of correlation showing two salient
features. The first feature, identified as a nearly symmetrical peak spanning time zero is referred to as a shared input, CI. This feature remained
unaltered by cortical deactivation. The second feature was observed only during cooling and was identified as the appearance of a Direct
interaction. The cell pair was recorded from one electrode. (c) This example shows a Direct interaction on top of a shared input, two features that
were observed during all recording conditions. Note that during cortical cooling the CI feature is increased without a change in the Direct
interaction. The increase of CI during cortical deactivation was a peculiar observation of cell pairs recorded in the feline thalamus from different
electrodes.
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
170
firing, then after 350 ms by a discharge of cell 11 and
then after 21 ms cell 11 firing again. This exact pattern
was used as template for searching approximate matching patterns with jitter of9 2 ms. Note that several
spikes may occur in between the spikes belonging to the
pattern without affecting the precise intervals separating the occurrences of the cell discharges. In this example, the significant pattern was observed nine times in
400 s of spontaneous activity recorded during the control condition. No patterns were observed in 300 s of
recording during cooling but the very same precise
pattern was observed five times in 300 s during the
recovery period. Note that between the first and the last
occurrence of the pattern more than 90 min had passed.
Another example (Fig. 6) illustrates that significant
patterns observed during cortical cooling might be completely absent prior to and after cortical deactivation.
This case shows a pattern of three spikes repeating nine
times (with jitter of 92ms) starting by a spike of cell 8,
followed by a spike of cell 5 180 ms later, then after 280
ms by a spike of cell 6. Note that only cells 5 and 6
were recorded from the same electrode. This second
example also illustrates the tendency observed in firing
patterns across species during cortical cooling to exhibit
a higher complexity than patterns observed during control and recovery conditions.
The last results described here are based on higher
order spectral analysis performed on rat and guinea pig
thalamic LFPs. In approximately half of the recordings
(43–54% according to species) the energy of the power
spectrum density was decreased during cortical cooling,
whereas only a small fraction (3–9%) of the energies
were increased. During the control condition the LFPs
showed a high degree of synchronicity from a simple
visual inspection of the traces but during cooling the
synchronicity becomes even more prominent and spindles can be observed more frequently (Fig. 7a). This
observation is confirmed by the analysis of phase-coupled non-linear interactions revealed by the bispectral
analysis. In the example of Fig. 7 virtually no peaks of
bicoherence were observed between LFPs recorded
from channels 1 and 2 during control and recovery
conditions. In contrast, a number of peaks were observed during cooling, bimodally distributed along the
frequency axis (Fig. 7b). The bispectral analysis of all
rat results pooled together revealed a corticofugal effect
on intrathalamic interactions of paramount importance
(Fig. 8). We analyzed the distributions of phase-
Table 2
Summary of the distribution of significant crosscorrelations during spontaneous activity across species when the cell pair was recorded from the
same electrode and from different electrodes
Crosscorrelogram type
Rat (n= 2)
Guinea pig (n = 2)
Cat (n =7)
Control and recovery
During cooling
Control and recovery
During cooling
Control and recovery
During cooling
Same electrode (SE)
CI
Stable
Disappearing
Appearing
15
17
0
15
0
0
9
13
0
9
0
2
55
12
0
55
0
11
Direct
Stable
Disappearing
Appearing
4
1
0
4
0
7
3
0
0
3
0
4
2
0
0
2
0
4
Total number of pairs
48
45
121
Different electrodes
(DE)
CI
Stable
Disappearing
Appearing
3
7
0
3
0
3
0
4
0
0
0
1
50
6
0
50
0
24
Direct
Stable
Disappearing
Appearing
0
2
0
0
0
5
3
3
0
3
0
4
5
4
0
5
0
14
Total number of pairs
216
121
The numbers highlighted in boldface refer to the prevailing tendency of corticofugal modulation discussed in the text.
Cl= common input type of crosscorrelogram. Direct = Direct type of crosscorrelogram.
801
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
171
Fig. 4. Crosscorrelation analysis of pairs recorded from the same electrode (left panel) and pairs recorded from different electrodes (right panel).
The scattergrams for different samples display the percent of ‘disappearing’ versus ‘appearing’ correlograms. The origin of the sample is indicated
as follows: R, rat; G, guinea pig and C, cat. The type of significant correlogram feature is indicated as CI (shared input) and d (Direct interaction).
See text for more details.
coupled frequencies corresponding to the first 200
peaks of bicoherence, thus indicating the most prominent non-linear interactions. The mean level of bicoherence of these peaks was equal to 0.55 during cortical
deactivation but it was below 0.3 during the other
recording conditions. In the graphs of Fig. 8 bispectral
and cross-bispectral results were pooled together because no particular differences were found between
these two samples. In the control and recovery conditions, the distribution of phase-coupled frequencies was
flat and ranged between 20 and 100 Hz (Fig. 8). Conversely, a clear-cut bimodal distribution of these frequencies was observed during cooling. The low
frequency values were well clustered and ranged between 7 and 32 Hz, centered near 22 Hz, whereas the
high frequency values ranged from 80 to 100 Hz. The
bispectral analysis of guinea pig recordings provided
similar results. The distribution of phase-coupled frequencies was flat during control and recovery, but it
was bimodal during cortical cooling. The only difference with the corresponding cluster in rat results is that
in the guinea pig the distribution of low frequencies
ranged between 20 and 60 Hz and showed a larger
kurtosis. The non-linear interactions occurring at low
frequencies suggest that resonance might be due to a
spread of activity over a wider area within the thalamus
with respect to the spread of activity occurring during
control and recovery conditions.
4. Discussion
Thalamocortical interactions are based on rich interconnections between the components of this network
and understanding of information processing undergoing in the thalamus necessarily rests on the investigation of corticofugal modulation. Reversible cooling
offers the opportunity to investigate the role of cortex
by examining thalamic network activity in its absence.
However, a firm grounding in the anatomy of connections is essential for adequate interpretations of the
effects of deactivating cortical input on thalamic activity. These effects represent the sum of the effects on the
composite pathway linking the thalamus and the cerebral cortex mediated directly via primary projections or
indirectly via the thalamic reticular nucleus and corticotectal projections. Furthermore, there are features of
the thalamocortical circuitry which are modality- and
species-specific demonstrating the advisability of performing such experiments in several species and repeating them for each sensory modality. In addition, the
presence of multiple local circuits within the thalamus
suggests that the use of simultaneous multisite recording will be of extreme importance. Our study represents
the first attempt to apply such a global approach to the
study of the auditory thalamocortical pathway. We
have also demonstrated the complementary information
on functional local connectivity that can be obtained by
combining single unit and local field potential analyses.
In addition, this is the first report of the effect of
reversible cooling of the primary auditory cortex on
thalamic activity in rat and guinea pig.
The general results of crosscorrelation analyses
across species indicate that cortical cooling induced the
appearance of Direct interaction type correlograms,
especially between pairs recorded from the same electrode. This could be interpreted as the activation of
direct excitatory links between thalamic neurons. Such
172
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
fast plasticity of the thalamic circuitry can hardly be
viewed as ‘classical’ neural compensation in terms of
‘hard wiring’ remodeling. On the contrary, these results
indicate that existing, but possibly pre-synaptically inhibited intrathalamic circuits become functionally significant in the absence of primary auditory cortex
activity (Fig. 9). This pre-synaptic inhibition might be
effected under normal conditions by auditory cortical
activation of the corresponding sector of RE, projecting
to MGB. Additional evidence in support of this interpretation is given by the effect of cortical cooling on CI
interaction type correlograms. In rats and guinea pigs
cortical deactivation induced disappearance of CI to a
large extent, especially between cells recorded from
different electrodes. This can be explained as absence of
shared input from the auditory cortex and diminished
activity of shared input from RE. In contrast, while the
number of CI was stable in cat SE data, a remarkable
increase in CI was observed between cells recorded
from different electrodes. This apparent contradiction
could be resolved by considering the different local
circuitry in cat’s thalamus with respect to rat and
guinea pig. As well as shared excitation, shared inhibition tend to synchronize the firing times of a pair of
neurons and produce similar CI features on the crosscorrelograms (Abeles, 1991; Hill and Villa, 1997). In
cats the local inhibitory neurons represent approximately 25% of the cells but they are virtually absent in
the other species studied here (Table 1). Thus, it is
possible that, in the cat, local MGB inhibitory neurons
are usually inhibited by cortically activated RE projections, while during cortical deactivation local inhibitory
neurons might provide an additional common source of
input.
In addition, in the absence of input from the cortex,
the newly active intrathalamic direct connections would
provide the necessary input to sustain reverberating
activity within thalamic cell assemblies (Fig. 9). Indeed,
the data reported in the current study confirm previous
findings (Villa and Abeles, 1990) about the existence of
precise spatiotemporal patterns of spikes in the auditory thalamus. The reported patterns are extremely
Fig. 5. Example of a precise spatiotemporal pattern of spikes detected nine times prior to and five times after recovery from cooling of the auditory
cortex, but absent during cortical deactivation. The firing pattern was formed by three spikes, starting by spike 7, then after 359 9 2 ms spike 11
and then after 21 92 ms spike 11 again. The spike occurrences corresponding to the pattern are indicated by thick ticks in the raster displays (right
panels) aligned by displaying the first spike in the pattern at time 0. Note that spikes 7 and 11 were recorded from different electrodes. The
analyzed records corresponded to 400 s of spontaneous activity during control condition, 300 s during cortical cooling and 300 s in the recovery
condition. Each block of 100 s is represented by a thick mark along the time axis of the experimental protocol and the dots indicate the start event
of the pattern.
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
173
Fig. 6. Example of a precise spatiotemporal pattern of spikes detected nine times exclusively during cortical deactivation. The firing pattern was
formed by three spikes, starting by spike 8, then after 180 92 ms spike 5 and then after 28092 ms spike 6. The spike occurrences corresponding
to the pattern are indicated by thick ticks in the raster displays (right panels) aligned by displaying the first spike in the pattern at time 0. Note
that only spikes 5 and 6 were recorded from the same electrode. The analyzed records corresponded to 300 s of spontaneous activity during
control condition, 500 s during cortical cooling and 300 s in the recovery condition. Each block of 100 s is represented by a thick mark along
the time axis of the experimental protocol and the dots indicate the start event of the pattern.
unlikely to have occured by chance (p B 0.001) assuming Poisson distribution of firing times (Abeles and
Gerstein, 1988). The existence of such patterns is predicted by the theory of synfire chains, based on chains
of divergent/convergent links (Abeles, 1991). Interestingly, recent simulation studies suggest that time-structured neural assemblies may be sustained by the
cortico-thalamo-cortical loop, which provides a high-security link from one active node of the chain to nodes
activated at a later time in the sequence (Miller, 1996).
Higher order precise spatiotemporal patterns of activity
within the auditory thalamus are only moderately affected by cooling deactivation, generally showing a
tendency to become more complex during deactivation
of the primary auditory cortex (present results and Villa
and Abeles, 1990). Due to the low probability of pattern occurrence and to the relatively small size of our
sample (in rat and guinea pig) it is not possible to
determine a level of significance of this tendency but
this result goes also in the direction of increased intrathalamic functional connectivity during cortical
deactivation.
In contrast to spike train data that correspond to
single neuron activity, the local field potential represents the average activity of several hundreds or thousands of neurons. Therefore, each neuron generates a
fraction of the local field potential, but some degree of
synchronization among the neurons is required to generate the slow (up to 100 Hz) electric waves seen. Power
spectrum density analysis provides an estimate of the
energy of local field potentials and determines the frequency components corresponding to the synchronized
activity. In thalamic electrodes the power spectrum
density was decreased during cortical cooling in approximately half of the recordings. This effect is probably be due to the deactivation, or reduction in activity,
of a generator (presumably the inactivated cortex) that
normally spreads a common pattern of waveform over
large thalamic areas. However, this result does not
provide any information on the functional interactions
within the thalamus and bispectral analyses (third order
cumulants) were calculated to address this question.
Bicoherence is a normalization method which compares
174
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
the actual bispectrum with a zero phase bispectrum, i.e.
a bispectrum with the highest degree of phase coupling.
It shows the degree of phase coupling between fre-
quency components of one or more signals (Sigl and
Chamoun, 1994). During deactivation of the auditory
cortex a remarkable increase (80–90 times) in the num-
Fig. 7. Analysis of local field potentials (LFPs) recorded from a pair of electrodes in the rat auditory thalamus. (a) Traces of simultaneously
recorded signals during 3 s are displayed in the central panels for the three recording conditions. Power spectrum densities (PSD) of the
corresponding channels are displayed on the side panels. The PSDs were computed over 200 s. Note the effect of notch filtering centered on 50
Hz. (b) Bicoherence analysis within- and cross-channel of the same dataset illustrated on (a). The bicoherence peaks (represented by dots) of phase
coupled frequencies f1 and f2 are plotted for frequency ranges up to 100 Hz at a resolution of 0.5 Hz. The triangular shape of this plot is due
to the symmetry properties of bispectral analysis and redundant values are not plotted. Several peaks were revealed during cortical cooling but
virtually no peaks (n B5) were observed during control and recovery conditions.
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
175
Fig. 8. Results of bispectral analysis for rat data. The 200 highest bicoherence values from pooled data are plotted during the three recording
conditions. The left panels show the density histogram (in percent) of the phase coupled frequencies ( f1 +f2). The right panels show the
bicoherence plot (see Fig. 7b for legend). Note the bimodal distribution of phase coupled frequencies during cooling, particularly characterized
by the appearance of low frequency components.
ber of significant bicoherence peaks was observed in
within- as well in cross-channel analyses, thus indicating a general increase in interacting thalamic subpopulations. In particular, the majority of phase couplings
occurred at low frequencies (7 – 32 Hz) that were absent
during control and recovery conditions. Thus, the degree of coupling between frequency components of
local field potentials is varied according to changes in
cortical activity.
Resonances occur as a result of interference between
waves traveling in different directions that combine to
form standing wave patterns, called ‘eigenfunctions’
(Nunez, 1995a). This phenomenon requires that each
neural mass has a relatively large number of synapses
so that signals are transferred between two locations in
a time roughly proportional to the distance between
locations (Nunez, 1995b). In one-dimensional standing
waves, a spatial frequency corresponds to each resonant
frequency. In multidimensional waves likely to be generated in a complex structure such as the thalamus, the
concept of spatial frequency is not precise but the
general idea that high spatial frequencies correspond to
high temporal frequencies is preserved (Nunez, 1995a).
Thus, the non-linear interactions occurring at low frequencies suggest that resonance might be due to a
spread of activity over a wider area within the thalamus
176
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
with respect to the spread of activity occurring during
control and recovery conditions (as an analogy, in the
violin body low tones occur with a spatial pattern
consisting of a few large areas having equal signs of
deflection at any fixed time; high tones correspond to
patterns with many such small areas).
In summary, we have presented several lines of evidence that corticothalamic projections may control the
extent to which information is processed locally in the
MGB. This general conclusion is valid for rat, guinea
pig and cat, but the presence of local inhibitory neurons
in cat MGB can account for some peculiar findings of
corticofugal modulation of feline thalamic activity. In
normal conditions, selective activation of auditory cortical assemblies would initiate a feedback influence to
the thalamus inducing local processing in MGB, which
in turn would enhance the effect of thalamocortical
input (Fig. 9). This fits well with the functional selectiv-
ity-adaptive filtering theory of thalamic function (Villa,
1988; Villa et al., 1991; Tetko and Villa, 1997a), where
the gain of filtering is indirectly controlled by the
thalamic reticular nucleus (Villa, 1990). In this model, a
particular pattern of cortical activity (‘template feature’) could be fed to the thalamus and compared to the
pattern elicited by an external sensory stimulus (‘test
feature’). Consequently, the increased coupling of cortical and thalamic activity would amplify the effectiveness of a particular feature of the external sensory input
allowing its detection and binding to higher cognitive
processing. Conversely, during cortical deactivation the
absence of cortical input would induce an intrathalamic
spread of activity, thus characterizing a distributed
processing state in which the thalamus speaks primarily
with itself.
In conclusion, it is evident that the combination of a
reliable and reversible deactivation technique (Payne et
Fig. 9. Diagram of the effect of corticofugal modulation on thalamic processing. Multisite simultaneous recording of spike trains is illustrated by
two microelectrodes inserted at two different locations of MGB. In this virtual example three single units (Nos. 1, 2, 3) are detected from the first
electrode and two single units are detected from the second electrode (Nos. 4, 5). Single units from the same electrode are sorted according to their
waveshapes (shown on the oscilloscope traces) and the rasters of the corresponding spike trains are displayed. During normal conditions (upper
panel) cortico-thalamo-cortical feedback together with projections to medial geniculate body (MGB) from the thalamic reticular nucleus (RE)
induces a segregation of information processes within the thalamus (‘local processing’ mode). During cortical deactivation (lower panel) cortical
inputs to MGB and to RE are absent so that segregation is released to some extent and increased effectiveness of intrathalamic connections can
sustain complex spatiotemporal activity in thalamic cell assemblies (‘distributed processing mode’).
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
al., 1996) with multisite electrophysiological recordings
and complex analytical methods can provide data with
far reaching implications for theories of thalamic function and more generally for cerebral network
interactions.
Acknowledgements
The authors wish to thank A. Audergon, V.M. Bajo
Lorenzana, M. Capt, D. Carretta, A. Celletti, C.
Eriksson, C. Haeberli, M. Jadé, F. Rodriguez Nodal,
E.M. Rouiller, G.M. Simm, A. Singy and P. Zurita for
taking part in some experimental sessions reported here
and for their technical assistance. We are indebted to
the critical reading of Brian Hyland. This research was
partially supported by Swiss SNF 31-37723.93 and
OFES 93.0241 grants and by the European Union
CHRX-CT93-0269 grant.
References
Abeles M. Quantification, smoothing, and confidence limits for single unit histogram. J Neurosci Methods 1982;5:315–25.
Abeles M. Corticonics: Neural Circuits of the Cerebral Cortex.
Cambridge: Cambridge University Press, 1991.
Abeles M, Gerstein GL. Detecting spatiotemporal firing patterns
among simultaneously recorded single neurones. J Neurophysiol
1988;60:909 – 24.
Baker FH, Malpeli JG. Effects of cryogenic blockade of visual
cortex on the responses of lateral geniculate neurons in the
monkey. Exp Brain Res 1977;29:433–44.
Brillinger DR. An introduction to polyspectra. Ann Math Statist
1965;36:1351 – 74.
Brooks VB. Study of brain function by local, reversible cooling. Rev
Physiol, Biochem Pharmacol 1983;95:1–109.
Diamond ME, Armstrong-James M, Budway MJ, Ebner FF. Somatic sensory responses in the rostral sector of the posterior
group (POm) and in the ventral posterior medial nucleus (VPM)
of the rat thalamus: dependence on the barrel field cortex. J
Comp Neurol 1992;319:66–84.
Geisert EE, Langsetmo A, Spear PD. Influence of the cortico-geniculate pathway on response properties of cat lateral geniculate
neurons. Brain Res 1981;208:409–15.
Ghosh S, Murray GM, Turman AB, Rowe MJ. Corticothalamic
influences on transmission of tactile information in the ventroposterolateral thalamus of the cat: effect of reversible inactivation of somatosensory cortical areas I and II. Exp Brain Res
1994;100:276 – 86.
Harris RM, Hendrickson AE. Local circuit neurons in the rat ventrobasal thalamus — a GABA immunocytochemical study. Neuroscience 1987;21:229–36.
Hill SL, Villa AEP. Dynamic transitions in global network activity
influenced by the balance of excitation and inhibition. Network
1997;8:165 – 84.
Huber PJ, Kleiner B, Gasser T, Dumermuth G. Statistical methods
for investigating phase relations in stationary stochastic process.
IEEE Trans Audio Electroacoust 1971:AU-19; 78–86.
Ivarsson C, deRibaupierre Y, deRibaupierre F. Influence of auditory localization cues on neuronal activity in the auditory thalamus of the cat. J Neurophysiol 1988;59:586–606.
177
Jones EG. The Thalamus. New York: Plenum Press, 1985.
Kalil RE, Chase R. Corticofugal influence on activity of lateral
geniculate neurons in the cat. J Neurophysiol 1970;33:459–74.
Kayama Y, Shosaku A, Doty RW. Cryogenic blockade of the visual
cortico-thalamic projection in the rat. Exp Brain Res
1984;54:157 – 65.
Marrocco RT, McClurkin JW. Evidence for spatial structure in the
cortical input to the monkey lateral geniculate nucleus. Exp
Brain Res 1985;59:50 – 6.
McClurkin JW, Marrocco RT. Visual cortical input alters spatial
tuning in monkey lateral geniculate nucleus cells. J Physiol
1984;348:135 – 52.
Miller R. Cortico-thalamic interplay and the security of operation of
neural assemblies and temporal chains in the cerebral cortex.
Biol Cybern 1996;75:263 – 75.
Montero VM. The interneuronal nature of GABAergic neurons in
the lateral geniculate nucleus of the rhesus monkey: a combined
HRP and GABA-immunocytochemical study. Exp Brain Res
1986;64:615 – 22.
Nikias CL, Raghuveer MR. Bispectrum estimation: a digital signal
processing framework. Proc IEEE 1987;75:869 – 91.
Nunez PL. Mind, Brain, and Electroencephalography. In: Nunez
PL, editor. Neocortical Dynamics and Human EEG Rhythms.
New York: Oxford University Press, 1995a:133 – 94.
Nunez PL. Global contributions to EEG dynamics. In: Nunez PL,
editor. Neocortical Dynamics and Human EEG Rhythms. New
York: Oxford University Press, 1995b:475 – 533.
Ohara PT, Lieberman AR, Hunt SP, Wu JY. Neural elements
containing glutamic acid decarboxylase (GAD) in the dorsal
lateral geniculate nucleus of the rat; immunohistochemical studies by light and electron microscopy. Neuroscience 1983;8:189–
211.
Orman SS, Humphrey GL. Effects of changes in cortical arousal
and of auditory cortex cooling on neuronal activity in the medial
geniculate body. Exp Brain Res 1981;42:475 – 82.
Paxinos G, Watson C. The Rat Brain in Stereotaxic Coordinates,
2nd edition. New York: Academic Press, 1986.
Payne BR, Lomber SG, Villa AEP, Bullier J. Reversible deactivation of cerebral network components. Trends Neurosci
1996;19:535 – 42.
Penny GR, Fitzpatrick D, Schmechel DE, Diamond IT. Glutamic
acid decarboxylase-immunoreactive neurons and horseradish peroxidase-labeled projection neurons in the ventral posterior nucleus of the cat and Galago senegalensis. J Neurosci
1983;3:1868 – 87.
Reale RA, Imig TJ. Tonotopic organization of auditory cortex in
the cat. J Comp Neurol 1980;192:265 – 91.
Redies H, Sieben U, Creutzfeldt OD. Functional subdivisions in the
auditory cortex of the guinea pig. J Comp Neurol 1989;282:473–
88.
Sigl JC, Chamoun NG. An introduction to bispectral analysis for
the electroencephalogram. J Clin Monit 1994;10:392 – 404.
Sillito AM, Jones HE, Gerstein GL, West DC. Feature-linked synchronization of thalamic relay cell firing induced feedback from
the visual cortex. Nature 1994;369:479 – 82.
Singer W. Control of thalamic transmission by corticofugal and
ascending reticular pathways in the visual system. Physiol Rev
1977;57:386 – 420.
Spreafico R, Frassoni C, Arcelli P, DeBiasi S. GABAergic interneurons in the somatosensory thalamus of the guinea-pig: a light
and
ultrastructural
immunocytochemical.
Neuroscience
1994;59:961 – 73.
Steriade M, Llinas RR. The functional states of the thalamus and
the associated neuronal interplay. Physiol Rev 1988;68:649–742.
Tetko IV, Villa AEP. Efficient partition of learning datasets for
neural network training. Neural Networks 1997;8:1361 –74.
178
A.E.P. Villa et al. / Journal of Neuroscience Methods 86 (1999) 161–178
Tetko IV, Villa AEP. Fast combinatorial methods for estimation of
complex temporal patterns of spikes. Biol Cybern 1997;76:397 –
407.
Thompson GC, Cortez AM, Lam DM. Localization of GABA
immunoreactivity in the auditory brainstem of guinea pigs. Brain
Res 1985;339:119 – 22.
Tsumoto T, Creutzfeldt OD, Legéndy CR. Functional organization of
the corticofugal system from visual cortex to lateral geniculate
nucleus in the cat (with an appendix on geniculo-cortical monosynaptic connections). Exp Brain Res 1978;32:345–64.
Villa AEP. Influence de l’ecorce cérébrale sur l’activité spontanée et
évoquée du thalamus auditif du chat. Lausanne: Presse Imprivite,
1988.
Villa AEP. Functional differentiation within the auditory part of the
reticular nucleus of the cat. Brain Res Rev 1990;15:25–40.
Villa AEP, Abeles M. Evidence for spatio-temporal firing patterns
within the auditory thalamus of the cat. Brain Res 1990;509:325 – 7.
Villa AEP, Rouiller EM, Simm GM, Zurita P, deRibaupierre Y,
deRibaupierre F. Corticofugal modulation of the information
.
processing in the auditory thalamus of the cat. Exp Brain Res
1991;86:506 – 17.
Villa AEP, Bajo VM, Vantini G. Nerve Growth Factor (NGF)
modulates information processing in the auditory thalamus. Brain
Res Bull 1996;39:139 – 47.
Villa AEP, Tetko IV, Celletti A. Corticofugal modulation of complex
spatiotemporal firing patterns in the rat auditory thalamus, 1998
(in prep).
Winer JA. The medial geniculate body of the cat. Adv Anat Embryol
Cell Biol 1985;86:1 – 97.
Winer JA, Larue DT. Anatomy of glutamic acid decarboxylase
immunoreactive neurons and axons in the rat medial geniculate
body. J Comp Neurol 1988;278:47 – 68.
Yuan B, Morrow TJ, Casey KL. Corticofugal influences of S1 cortex
on ventrobasal thalamic neurons in the awake rat. Neuroscience
1986;8:3611 – 7.
Zilles K, Wree A. Cortex: areal and laminar structure. In: Paxinos G,
editor. The Rat Nervous System, vol. 1. New York: Academic
Press, 1985:375 – 416.