Department of Zoology, University of Oklahoma, Norman, Oklahoma 73019
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bastian, Joseph and Jerry Nguyenkim. Dendritic Modulation of Burst-Like Firing in Sensory Neurons. J. Neurophysiol. 85: 10-22, 2001. This report describes the variability of spontaneous firing characteristics of sensory neurons, electrosensory lateral line lobe (ELL) pyramidal cells, within the electrosensory lateral line lobe of weakly electric fish in vivo. We show that these cells' spontaneous firing frequency, measures of spike train regularity (interspike interval coefficient of variation), and the tendency of these cells to produce bursts of action potentials are correlated with the size of the cell's apical dendritic arbor. We also show that bursting behavior may be influenced or controlled by descending inputs from higher centers that provide excitatory and inhibitory inputs to the pyramidal cells' apical dendrites. Pyramidal cells were classified as "bursty" or "nonbursty" according to whether or not spike trains deviated significantly from the expected properties of random (Poisson) spike trains of the same average firing frequency, and, in the case of bursty cells, the maximum within-burst interspike interval characteristic of bursts was determined. Each cell's probability of producing bursts above the level expected for a Poisson spike train was determined and related to spontaneous firing frequency and dendritic morphology. Pyramidal cells with large apical dendritic arbors have lower rates of spontaneous activity and higher probabilities of producing bursts above the expected level, while cells with smaller apical dendrites fire at higher frequencies and are less bursty. The effect of blocking non-N-methyl-D-aspartate (non-NMDA) glutamatergic synaptic inputs to the apical dendrites of these cells, and to local inhibitory interneurons, significantly reduced the spontaneous occurrence of spike bursts and intracellular injection of hyperpolarizing current mimicked this effect. The results suggest that bursty firing of ELL pyramidal cells may be under descending control allowing activity in electrosensory feedback pathways to influence the firing properties of sensory neurons early in the processing hierarchy.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A commonly observed
characteristic of neuronal spontaneous activity as well as
stimulus-driven activity is that clusters of spikes separated by short
interspike intervals (ISIs) or "bursts" occur more often than
expected given the assumption that spikes are generated randomly and
independently as in a Poisson spike train. Results from studies of a
variety of neural systems indicate that bursts may be important
features of spike trains; bursts may ensure transmission at unreliable
synapses, they may be involved in the induction of synaptic plasticity,
they may signal important features of sensory stimuli, and relatively
synchronous bursts within populations of neurons may be involved in
"higher" aspects of sensory processing (see Lisman
1997 for review). Mechanisms underlying burst production can
include network properties that provide appropriate patterns of
synaptic input as well as intrinsic membrane properties that predispose
a neuron to produce bursts (see Connors and Gutnick 1990
for review). Additionally, both physiological and computational studies
show that morphological features of neurons can be linked to their
ability to produce bursty firing patterns. Among cortical pyramidal
cells, for example, apical dendritic morphology, is an important
determinant of a cell's firing pattern (Mainen and Sejnowski
1996
; Mason and Larkman 1990
).
Burst-like firing patterns may also be important characteristics of
neurons involved in the initial stages of sensory processing systems
where brief clusters of spikes may signal the presence of specific
stimulus features (Gabbiani and Metzner 1999). Recent in
vivo studies of the electrosensory system of South American weakly
electric fish showed that while electroreceptor afferents encode
detailed information about the amplitude and time course of
electrosensory stimuli (Nelson et al. 1997
;
Wessel et al. 1996
), electrosensory lateral line lobe
(ELL) pyramidal cells, which receive synaptic input from these
afferents, encode significantly less detailed information. Instead,
these neurons signal the occurrence of critical features of a stimulus
with short bursts of spikes (Gabbiani and Metzner 1999
;
Gabbiani et al. 1996
; Metzner et al. 1998
). In vitro studies of these same pyramidal neurons showed that characteristics of the cells' apical dendrites are critical for
burst production; bursts occur in response to summating depolarizing afterpotentials that result from longer-duration dendritic spikes depolarizing the soma (Lemon and Turner 2000
; Turner and
Maler 1999
; Turner et al. 1994
, 1996
).
This report describes the variability of spontaneous firing characteristics of ELL pyramidal cells in vivo and shows that the tendency of these cells to produce bursts of action potentials can be related to the size of the cell's apical dendritic arbor and that bursting behavior may be influenced or controlled by descending inputs from higher centers.
The weakly electric fish used in this study, Apteronotus
leptorhynchus, generates a quasi-sinusoidal electric organ
discharge (EOD) having a frequency ranging from about 600-1,000 Hz
depending on the individual. Electroreceptors scattered over the body
surface encode the amplitude (P-type receptors) and timing (T-type
receptors) of this discharge. Two broad categories of sensory stimuli
are thought to be encoded by this system. Electrolocation stimuli are
thought to consist of relatively localized changes in the amplitude and
timing of the discharge resulting from the presence of objects in the
animal's environment that have an impedance different from that of the
surrounding water. Electrocommunication stimuli also consist of changes
in EOD amplitude and timing, but these result from the interaction of
the EOD of an individual with that of conspecifics, and these are
typically spatially extensive influencing the activity of
electroreceptors over large regions of the body surface (see
Turner et al. 1999 for recent reviews of
electroreception). Electroreceptor afferents terminate within a
medullary nucleus, the ELL, and the P-type or amplitude encoding afferents provide excitatory synaptic input to basilar pyramidal cells
(E cells). A second category of pyramidal cells, nonbasilar or I cells,
are driven by inhibitory interneurons that receive excitatory afferent
input (Maler 1979
; Maler et al. 1981
;
Saunders and Bastian 1984
).
In addition to receiving electroreceptor afferent inputs, ELL pyramidal
cells receive, via elaborate apical dendrites, massive synaptic input
descending from higher centers (Maler et al. 1981; Sas and Maler 1983
, 1987
). One subdivision of the
descending pathways is thought to be involved in gain control
(Bastian 1986a
,b
; Bastian and Bratton
1990
; Nelson 1994
) while another may provide
positive feedback to accentuate important stimulus features
(Berman and Maler 1999
; Berman et al.
1997
; Bratton and Bastian 1990
; Maler and
Mugnaini 1994
). Recent studies have also demonstrated robust synaptic plasticity at the pyramidal cells' apical dendrites
(Bastian 1999
; Bell et al. 1997
).
Studies in which single ELL pyramidal cells were intracellularly
labeled and reconstructed showed that several physiological characteristics of these neurons were strongly correlated with neuronal
morphology (Bastian and Courtright 1991). Apical
dendritic size is highly variable, and both spontaneous firing
frequency as well as rate of adaptation to changes in stimulus
amplitude are negatively correlated with the size of a cell's apical
dendritic arbor. This report extends these observations showing that
pyramidal cells are also highly variable in terms of their tendency to
produce bursts of action potentials, that apical dendritic size is
strongly correlated with a cell's tendency to burst, and that
pharmacological reduction of descending excitation to these dendrites
greatly reduces bursty firing.
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The South American weakly electric fish A. leptorhynchus was exclusively used in these studies. Animals were
housed in 50-gallon population tanks at 26-28°C and with water
conductivity ranging from 200 to 400 µS. Experiments were done in a
39 × 44 × 12-cm-deep experimental tank containing water
from the animal's home tank. Surgical techniques were the same as
previously described (Bastian 1996a,b
) and all
procedures were in accordance with the University of Oklahoma's animal
care and use guidelines.
Recording and stimulation
Extracellular recordings were made with metal-filled
micropipettes constructed as described by Frank and Becker
(1964). Intracellular recordings were made with borosilicate or
aluminosilacate sharp electrodes pulled with a Brown-Flaming P-87
pipette puller and filled with 3 M K-acetate. Initial electrode
impedances ranged from 150 to 200 M
and were beveled (K. T. Brown BV-10 bevelor) until resistances fell to between 60 and 100 M
.
For extracellular studies, electrodes were advanced with a Kopf 650 hydraulic micro-drive and signals with amplified with a WPI DAM50
preamplifier. For intracellular studies, electrodes were advanced with
a Burleigh piezoelectric microdrive and preamplified with a WPI 767 electrometer. Spike times and times of EOD zero-crossings, and during
intracellular recordings, membrane potential waveforms were acquired
with Cambridge Electronic Design 1401plus hardware and SpikeII
software. Spike and electric organ discharge (EOD) timing was measured
with a resolution of 0.1 ms, and analog waveforms were A/D converted at
a rate of 12.5 kHz. All subsequent data processing was done using
Matlab (The Mathworks, Natick, MA).
The electric organ of Apteronotus is composed of modified
motoneurons rather than muscle cells so the normal electric organ discharge remains intact during the neuromuscular blockade used in
these experiments. The pyramidal cell spontaneous firing properties described therefore refer to activity in the presence of the normal baseline receptor afferent activity, which is very constant in the
absence of modulations of the EOD amplitude. Electronically produced
stepwise increases or decreases of the EOD, typically 1 mV/cm in
amplitude and 300 ms in duration, were applied between electrodes
straddling the fish and used as search stimuli. Stereotyped responses
to this stimulus enabled categorization of cells as either basilar
pyramidal cells (E cells) or nonbasilar pyramidal cells (I cells). The
former respond to increased EOD amplitude with short-latency increases
in spike frequency while the latter respond to increased EOD amplitude
with reductions in firing frequency (Saunders and Bastian
1984).
Data analysis
Pyramidal cells were divided into subgroups, bursty or
nonbursty, depending on whether or not autocorrelograms of spontaneous activity deviated significantly from that expected assuming Poisson spike trains (Abeles 1982). Autocorrelograms, 1-ms
binwidth and 200-ms duration, were produced from records of spontaneous
activity typically containing 1,500-2,000 spike times. For a few very
low frequency cells, only 500 spike times were used. Spike trains were
initially displayed as plots of instantaneous frequency versus time and
only cells showing stable firing frequencies were studied. Expected
correlogram bin contents (y) was determined as
![]() |
Given an expected bin contents, y, the probability of
finding m spikes within a bin is
![]() |
![]() |
Pharmacological techniques and current injection
Micropressure ejection techniques were used to apply the
non-NMDA glutamate antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) to local regions of the ELL molecular layer containing the
apical dendrites of a recorded cell. Multibarrel pipettes were pulled
to a fine tip and broken back to a total tip diameter of about 10 µm.
Typically two barrels were filled with a 1 mM solution of disodium
CNQX, two barrels were filled with 1 mM glutamate, and a fifth control
barrel contained distilled water. After a well- isolated single-unit
extracellular recording or a stable intracellular recording was
established, the pressure pipette was slowly advanced into an
appropriate region of the ELL molecular layer while periodically
ejecting "puffs" of glutamate. Typically ejection duration ranged
from 50 to 100 ms, and ejection pressure was usually 40 psi. As
described earlier (Bastian 1993), proximity to the
apical dendrite of the recorded cell was indicated by short-latency increases in firing rate following glutamate ejection. Following correct placement, CNQX was delivered as a series of pulses (e.g., 100-ms puffs at 0.5 Hz for 20 s), and this treatment typically resulted in tonic alterations in pyramidal activity lasting
approximately 5 min. In experiments where the effects of tonic
hyperpolarization on pyramidal cell activity were studied, negative
current injection, typically
0.5 nA for 5 min, was applied via the
bridge circuit of the electrometer. Unless indicated otherwise, sample
means are given ±1 SE.
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Spontaneous activity
Spontaneous firing frequencies of 126 pyramidal cells were
recorded from the lateral and centrolateral segments of the ELL of 30 fish in which the animals' neurogenic electric organ discharge was
intact. Recording site within the lateral and centrolateral segments
were estimated from surface landmarks and recording depth, and no
obvious differences among cells from these regions were seen. Cells
were categorized as either E or I type, excited or inhibited by
increasing electric organ discharge amplitude, respectively, and the
distribution of spontaneous firing frequencies is shown in Fig.
1A. Spontaneous firing rates
for this sample of E and I cells were not different, averaging 18.32 ± 1.34, n = 62 and 17.72 ± 1.03, n = 64, respectively, and these values are similar to those seen in earlier
studies of this and the related fish Eigenmannia virescens
(Bastian 1986a; Metzner et al. 1998
).
|
The coefficient of variation (CV) of each
pyramidal cell's record of ISIs was calculated as an initial estimate
of spike train variability and, as described for Eigenmannia
(Metzner et al. 1998), CVs were
highly variable ranging from 0.45 to 2.24. The CVs were also negatively correlated with these
cells' spontaneous firing frequency as shown in Fig. 1B.
The CV is expected to be reduced with increasing
spike frequency since firing must become more regular as the ISI
approaches the cell's refractory period; however, effects of
refractory period alone cannot account for the range of
CVs seen in this sample of pyramidal cells.
Given a cell with an absolute refractory period of 5 ms, a spontaneous firing rate of 5 spikes/s, and a CV of 1.5, increasing firing rate to 50 spikes/s is expected to decrease
CV by about 25% (Gabbiani and Koch
1998
, equation 9.12). The reduction in
CVs seen across pyramidal cells having a similar
range of firing frequencies was approximately 50% or twice as large as
predicted given refractory effects alone.
Unlike most weakly electric fish, the neurogenic electric organ
discharge in Apteronotus remains intact under neuromuscular blockade, hence the "spontaneous activity" referred to herein is
activity in the presence of the normal discharge. Electroreceptor afferents' time of firing is strongly phase-coupled to the EOD waveform (Bastian 1981a; Hagiwara et al.
1965
; Hopkins 1976
) while the ELL pyramidal
cells generally show a weaker phase relationship to the EOD cycle
(Bastian 1981b
). Phase histograms of firing times within
the period of the EOD cycle were produced for a subset of the cells
studied. Phase coupling was measured by computing the mean vector
length for each histogram (Batschelete 1981
); this
statistic ranges from zero for histograms of activity unrelated to the
EOD period to 1.0 for perfectly synchronized firing. The strength of
the phase coupling to the EOD varied significantly among pyramidal
cells, and phase histograms of spontaneous activity showing stronger
(mean vector = 0.42) and nonsignificant (mean vector = 0.06)
phase relationships to the EOD are shown in Fig. 2, A and B,
respectively. The degree of phase coupling to the EOD was also found to
be highly correlated with the cells' spontaneous firing frequency
(Fig. 2C). Cells with higher rates of spontaneous activity
showed stronger phase coupling. A similar correlation was seen for both
E and I cells although the highest frequency cells and those showing
the strongest phase coupling were E cells (Fig. 2,
). The
higher-frequency pyramidal cells' stronger phase coupling to the EOD
contributes to the lower ISI variability seen for these cells since
spikes preferentially occur during a restricted phase of the highly
regular discharge waveform.
|
In a previous study in which individual ELL pyramidal cells were
intracellularly labeled with either horseradish peroxidase or Lucifer
yellow, it was found that spontaneous firing rate and other
physiological properties were significantly correlated with cellular
morphology (Bastian and Courtright 1991). In particular, the size of both basilar and nonbasilar pyramidal cells' apical dendritic trees was found to be negatively correlated with spontaneous rate. Examples of reconstructed pyramidal cells that span the range of variation seen are shown in Fig.
3, A-C. The spike trains from
17 morphologically described cells of this earlier study were
reexamined, and Fig. 3D shows the relationship between the summed length of all apical dendritic branches and each cell's spontaneous firing rate and spike train CV.
Apical dendritic lengths are significantly correlated with both
spontaneous rate and ISI coefficient of variation (r's
0.81 and 0.74, respectively). Hence the pyramidal cells with smaller
apical dendritic arborizations fire at relatively high frequencies and
their spike trains are more regular. Additionally, the higher frequency
cells' firing patterns shows relatively strong phase coupling to the
EOD cycle, indicating that receptor afferent inputs predominantly drive
these cells. Conversely, the lack of phase coupling between the firing times of the low-frequency pyramidal cells and the EOD waveform suggests that receptor afferent activity is less effective in driving
these cells, at least on an EOD cycle by cycle basis. Instead
descending inputs that terminate in the ELL molecular layers and
provide synaptic input to the extensive apical dendrites of
lower-frequency pyramidal cells may play a larger role in driving these
cells, and the more irregular character of the low frequency cells'
spontaneous activity (high CV) may be a
consequence of the biophysical properties of the more extensive apical
dendrites, a result of patterned inputs to these dendrites or a
combination of both.
|
Burst-like firing patterns
ISI histograms (ISIHs) of three pyramidal cells are shown in Fig.
4, A1-C1. Spike trains from a
subset of the pyramidal cells studied had simple exponential ISI
distributions as expected for a Poisson spike train (Fig.
4A1); however, the ISIHs of most cells consisted of two
phases: an initial peak at short ISIs followed by a longer tail (Fig.
4, B1 and C1) as described for pyramidal cells of
the related fish Eigenmannia virescens (Metzner et al. 1998). The latter ISIH pattern is typical for cells having a
tendency to produce short high-frequency "bursts" or clusters of
action potentials.
|
A variety of techniques have been used to determine whether or not to
classify a cell as bursty and to define the ISI size considered to be
characteristic of spikes within a burst. These range from identifying
features such as the first trough or inflection point following the
peak of ISI distributions or autocorrelation functions (Metzner
et al. 1998; Turner et al. 1996
) to statistical methods based on the occurrence probability of short ISI sequences (Legendy and Salcman 1985
). A method described by
Abeles (1982)
was adopted for identifying bursty
pyramidal cells and for determining the ISIs characteristic of bursts.
Autocorrelograms were constructed, and the expected bin contents based
on the cell's average firing rate as well as the +99.9% confidence
limit for the expected bin contents (
and - - -, respectively, Fig.
4, A2-C2) were determined. Cells showing initial peaks in
the autocorrelogram exceeding the confidence interval were categorized
as bursty.
Autocorrelograms of approximately 20% of the cells studied (12 E cells and 13 I cells) were either flat, as in the case of Fig. 4A2, or had small initial peaks that did not exceed the threshold confidence interval. These cells typically had simple exponential ISI distributions, had relatively high rates of spontaneous activity, and their average CV was 0.84. These cells were categorized as "nonbursty." The majority of cells (80%) were categorized as bursty and the maximum intraburst interval (IBImax) of spike doublets, or longer bursts, was taken as the time at which the falling phase of the autocorrelogram's initial peak crossed the confidence limit (arrow of Fig. 4B2). In cases where the correlogram contained multiple peaks that exceeded the threshold, as in Fig. 4C2, the time of the minimum following the first peak was taken as IBImax. The value of IBImax for each cell was then used to identify bursts of from 2 through 10 spikes in the records of spontaneous activity.
The values of IBImax for 50 E cells and 51 I cells averaged 15.98 ± 0.58 and 15.29 ± 0.71 ms, respectively, and the values of IBImax were negatively correlated with the cells' spontaneous firing frequencies (Fig. 5A). After identifying the IBImax for a given cell, burst probability was determined as the ratio of the total number of bursts, each burst considered as a single event, to the total number of events (burst events plus single spikes) in the sample. Burst probabilities for E and I cells averaged 0.25 ± 0.014 and 0.22 ± 0.012, respectively, and were not significantly correlated with pyramidal cell spontaneous firing frequency (Fig. 5B).
|
Pyramidal cells were categorized as bursty based on significant
deviations from the properties of a Poisson spike train; that is,
bursty cells are, by definition, those that produce an excess of spike
clusters over those expected from a Poisson spike train. The
probabilities shown in Fig. 5B are overestimates of the
"excess" bursts since some number ISIs shorter than
IBImax would be expected to occur even in a
Poisson spike train. Therefore an alternative measure of burst
probability designed to show the proportion of a cell's total bursts
above or in excess of those expected was also calculated. This was done
by removing the fraction of bursts expected given a Poisson spike train
of an average firing frequency predicted from the tail region of the
ISIH: first, data from the tail of the ISIH were fit with a single
exponential, which was extrapolated to its intersection with the start
of the initial peak of the ISIH. An example is shown by the smooth line
of Fig. 5C. The peak of the histogram above this
extrapolation, including bins through 2 times
IBImax, was then replaced with values predicted by the exponential fit to the tail. The solid line shown in the inset
of Fig. 5C shows the altered initial phase of the resulting histogram, and the dotted line indicates values that were replaced. The
mean spike frequencies (f) determined from histograms with the initial peaks removed in this manner were used as the basis for
predictions of the number of bursts expected due to an underlying Poisson process. The probability of n events
p(n) in a given time period, T, is
given as:
![]() |
The resulting corrected probabilities represent a cell's tendency to
fire bursts in excess of those accounted for by an underlying Poisson
process. These "non-Poisson bursts," expressed as a percentage of
the total burst probability, are plotted in Fig. 5D, and
this measure is strongly correlated with the cells' spontaneous firing rates (r = 0.73, P < 0.001). The
majority of bursts seen in low-frequency spike trains are non-Poisson,
but for higher-frequency cells, approximately 50% of the bursts seen
are expected given a Poisson spike train. Given the strong negative
correlation between spontaneous firing frequency and apical dendritic
size, this analysis indicates that cells with larger apical dendrites
produce more bursts in excess of those expected.
Intraburst spike interval distributions
To determine if characteristic patterns of ISIs occur within
bursts, successive intervals within bursts of six spikes were analyzed
for 64 pyramidal cells. Ten such bursts were produced by the cell of
Fig. 6A within a 3-min record,
and the successive ISIs within each burst are shown by . The means
of these intervals are shown by
. No systematic pattern of ISIs was
seen for this cell except for the relatively long interval at the end
of the burst. Mean ISIs within bursts containing six spikes from a
sample of 10 cells are shown by
of Fig. 6B. The second
through fifth intervals are expressed as a percentage of the first, and
no systematic pattern of successive intervals within bursts was seen
for these cells. The
shows the grand mean of within-burst intervals
normalized in this manner for the 64 cells analyzed, and no
statistically significant trend in interval length was found.
|
A second analysis was performed on all bursts containing three or more
spikes. The differences between successive intervals within each burst
were determined as described by Turner et al. (1996),
and histograms of these interval differences were produced. Figure
6C shows the distribution of intraburst interval (IBI) differences for the cell of Fig. 6A. That the interval
differences are approximately normally distributed with a mean (0.035 ms) not significantly different from zero indicates that there are no
simple trends in interval lengths within bursts. Successive intervals
neither increase nor decrease systematically; rather, for this cell,
intervals vary randomly about a mean length of approximately 16 ms. The
means of IBI difference distributions, like that of Fig. 6C,
were determined for 99 cells and are summarized in the histogram of
Fig. 6D. The mean of this distribution (
0.016 ms) is also
not different from zero. Although, in in vitro preparations, significant numbers of pyramidal cells showed trends or nonrandom patterns of IBI differences (Turner et al. 1996
), in
vivo, no consistent patterning of IBIs is seen.
Blockade of descending excitation reduces burst-like behavior
Previous in vivo studies showed that localized blockade of
excitatory inputs to pyramidal cell apical dendrites can be achieved via micropressure ejection of glutamate antagonists within the ELL
dorsal and ventral molecular layers (Bastian 1993). This
technique was used to determine if alterations of synaptic inputs to
pyramidal cell apical dendrites influenced bursting behavior. As in
previous studies, a multibarrel pressure pipette was positioned in
close proximity to the apical dendrite of a pyramidal cell being
recorded from. As the pressure pipette was advanced through the
molecular layer, brief puffs of glutamate were periodically ejected,
and short-latency responses to the glutamate were taken to indicate proximity of the pipette to the recorded cell. Previous pressure ejection experiments coupled with extracellular labeling of ejection site and intracellular labeling of the recorded cell verified that
short-latency excitatory responses occurred when the pipette was very
close to or within the cell's dendritic arbor (Bastian 1993
). Slightly longer-latency inhibitory responses were
typically seen when the pressure pipette was further outside the
dendritic arbor where glutamate preferentially activated inhibitory
interneurons. Following recording of 200 s of spontaneous firing,
the non-NMDA antagonist CNQX was ejected. Typically 50-ms puffs were
delivered at a rate of one per 2 s, resulting in a reduction in
the cell's spontaneous firing rate and a decrease in the cell's
tendency to produce spike bursts. The reduction of spontaneous activity is not due to CNQX blockade of glutamatergic receptor afferent input to
pyramidal cells. As shown earlier (Bastian 1993
) and repeated in this study (data not shown), responses to changes in EOD
amplitude are increased rather than decreased by CNQX application to
the molecular layers.
Figure 7, A, 1 and 2, and B, 1 and 2, shows the interval histograms and autocorrelograms of a bursty basilar pyramidal cell prior to and during the application of CNQX, respectively. Spontaneous firing rate was reduced from 10.0 to 5.1 spikes/s, and this reduction in firing frequency was mainly associated with a loss of spikes separated by short intervals. This results in a large reduction in the early peak of the ISIH and reduction in the height of the initial peak in the autocorrelogram. The inset of Fig. 7B1 shows the initial phases of the interval histograms acquired before and after recovery from CNQX (thin lines) superimposed on that taken during application of the antagonist (thick line). Burst probability for this cell was initially 0.35 given an IBImax of 23 ms. In the presence of CNQX, this was reduced to 0.14 and IBImax was increased to 26 ms. Cells typically recovered within 3 min of the cessation of CNQX application, and this cell's spontaneous rate returned to 10.2 spikes/s while burst probability recovered to 0.39 with an IBImax of 24 ms (Fig. 7C, 1 and 2).
|
The effects of CNQX application on the spontaneous firing patterns of 24 pyramidal cells (11 basilar and 13 nonbasilar) are summarized in Fig. 8. There was no significant difference in the behavior of these cell types so the data were pooled. On average, both total burst probability and non-Poisson burst probability were significantly reduced during CNQX application (Fig. 8, black and dark gray bars, P < 0.005, t-tests), and these measures returned toward their initial values following termination of CNQX application (Fig. 8, light gray bars). However, only 13 of the 24 cells were recorded through complete recovery. The coefficients of variation were also reduced by CNQX treatment in most cells (16 of 24); however, the average Cv of all cells during CNQX (1.04 ± 0.05) was not significantly less than that prior to treatment (1.09 ± 0.04). Although the reduction in burst probability during CNQX injection averaged approximately 45% for all cells studied, seven of these were only minimally affected by this treatment (burst probabilities changed by less than ±5%). It is not known if the lack of significant CNQX effects in these cells is a result of failure to deliver effective CNQX doses or if the firing characteristics of these cells are truly insensitive to this treatment.
|
The reductions in spontaneous firing frequency seen with CNQX application (compare Fig. 8, black and dark gray hatched bars) were not due to simply shifting the ISI distributions to longer values. Instead, as shown by Fig. 7B1, inset, this antagonist preferentially reduced the probability of short ISIs, and this loss of the higher-frequency spike bursts must contribute to the observed reduction in spontaneous rate. For comparison with the CNQX effects on firing rate, the bursts were artificially removed from the pre-CNQX spike trains and replaced with single spikes. The average frequency of these modified spike trains, which contained no bursts, is comparable to the frequency seen in the presence of CNQX (Fig. 8, hatched white bar). Although this manipulation of pre-CNQX spike trains indicates that loss of burst spikes alone could account for the reductions in firing rate seen with CNQX treatment, the application of CNQX also resulted in increased numbers of long (more than 300 ms) ISIs; this also contributes to the lower average firing rates seen.
Hyperpolarizing current injection reduces burst-like behavior
Possible mechanisms by which CNQX blockade of pyramidal cell
apical dendritic inputs could reduce spike bursts include elimination of patterned synaptic inputs that directly evoke pyramidal cell bursting as well as reduction of tonic excitatory inputs resulting in
hyperpolarization and reduced probability that a "burst-threshold" is exceeded. To determine if moderate hyperpolarization is sufficient to reduce bursts, intracellular recordings of spontaneous activity before, during, and after hyperpolarizing current injection were compared. Figure 9A1 shows two
typical 1-s segments of intracellularly recorded basilar pyramidal cell
spontaneous activity prior to hyperpolarizing current injection.
Spontaneous firing rate was initially 15.3 spikes/s and the
CV was 1.24. Constant current injection, 0.5 nA, initially silenced pyramidal cells but typically within 30 s spontaneous firing resumed at a lower but stable rate. Figure 9A2 shows 1-s epochs of spontaneous activity after
the cell adopted a new steady-state firing frequency. This cell was hyperpolarized by 1.5 mV, which reduced spontaneous firing frequency to
8.6 spikes/s and CV to 1.07, and this
small hyperpolarization reduced burst probability from an initial
valued of 0.24 to 0.04. Interval histograms of this cell's spontaneous
firing before and during current injection are shown in Fig.
9A3 by the thin and thick lines, respectively, and, as shown
for CNQX blockade of descending excitation, hyperpolarization
preferentially reduced the initial peak corresponding to the shortest
ISIs.
|
A pressure ejection pipette was also in place while recording from this cell so intracellular activity was monitored before and during CNQX blockade of inputs to the cell's apical dendrite. Comparison of the records (Fig. 9B, 1 and 2) shows that CNQX blockade resulted in similar changes in pyramidal cell spontaneous activity and bursting as did hyperpolarizing current injection. The brackets above the spike records show bursts identified based on IBImax of 12 and 13 ms for the data of Fig. 9, A and B, respectively. Spontaneous firing rate was reduced from 15.5 to 8.4 spikes/s, CV was reduced from 1.3 to 0.74, and burst probability was reduced from 0.27 to 0.03. The interval histograms of Fig. 9B3 summarize this cell's firing before and during the CNQX application (thin and thick lines, respectively), and the dashed line shows that the effects of CNQX were dose dependent; 50% of the original dose was applied in this case.
Figure 10 summarizes the effects of
hyperpolarization on the spontaneous activity of 14 pyramidal cells (7 basilar, 7 nonbasilar). Current magnitude was typically 0.5 nA, and
the average hyperpolarization that resulted was 3.25 mV. As with CNQX
treatment, overall burst probability as well as non-Poisson burst
probabilities were significantly decreased by hyperpolarization as was
spontaneous firing rate. The effects of removing bursts of from 2 through 10 spikes from spike trains recorded at normal membrane
potential and replacing them with single events was also examined. As
seen with CNQX treatment, the resulting reduction in firing frequency
due to artificial burst removal closely matched that due to
hyperpolarization (Fig. 10, dark gray and clear hatched bars).
Following hyperpolarization burst probabilities recovered but, unlike
the recovery from CNQX where probabilities and spike rates remained
somewhat depressed, following hyperpolarization both burst
probabilities and firing rates were larger than initially, although
these increases were not statistically significant. The result that
hyperpolarization mimicked the effects of CNQX treatment is compatible
with the idea that the CNQX blocked tonic excitation and altered a
threshold for burst generation. However, this result does not rule out
the possibility that CNQX treatment also eliminates patterned dendritic inputs. The intracellular recordings from cells that were also treated
with CNQX did not reveal any consistent effects of this drug on resting
potential; hyperpolarization was not systematically observed in
response to this treatment. Although it is possible that the slow onset
of CNQX effects along with small changes in impalement quality and
other artifactual sources membrane potential drift precluded clear-cut
indications of CNQX-caused hyperpolarization, the lack of changes in
resting potential may indicate that patterned inputs are also blocked
by this treatment.
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The results of this and of previous studies (Bastian
1993) show that there is considerable variability among ELL
pyramidal cells in terms of their physiology and morphology as well as
in the distribution of their neurotransmitter receptor types
(Bottai et al. 1997
, 1998
; Dunn et al.
1999
) and in the presence of various intracellular signaling
molecules (Berman and Maler 1999
; Berman et al.
1995
; Zupanc et al. 1992
). The variability of
pyramidal cell apical dendrite morphology is particularly striking
(Fig. 3, A-C), and the strong correlation between
spontaneous firing rate and dendritic structure (Fig. 3D)
enables us to predict the morphological characteristics of cells
studied with extracellular recording techniques. Cells with the
smallest apical dendrites (deep basilar pyramidal cells) also lie
deeper within the ELL laminae (Bastian and Courtright
1991
) and exhibited the highest firing rates and the lowest
probabilities of bursts in excess of those expected based on a Poisson
spike train, and their spontaneous activity showed the strongest phase
coupling to the electric organ discharge. This indicates that these
cells are principally driven by receptor afferent inputs whose activity
is also strongly phase-coupled to the EOD. Conversely, the
lowest-frequency cells have the largest apical dendrites, they are
found most superficially within the pyramidal cell lamina, and their
activity shows little to no phase relation to the EOD waveform. These
cells' activity is likely to be more strongly influenced by apical
dendritic inputs; their spontaneous activity is more irregular (high
CV), and these cells show the highest
probability of producing unexpected spike bursts. The correlation
between bursting and extensive apical dendrites has been previously
demonstrated in mammalian cortical pyramidal cells in vitro
(Mason and Larkman 1990
) and modeling studies also showed that apical dendritic size alone, without differential distributions of ion channels across morphological categories, was
correlated with a cell's ability to produce spike bursts
(Mainen and Sejnowski 1996
).
The tendency of various ELL pyramidal cell types to produce bursts may not be constant; instead, as sensory processing requirements change, bursty behavior may be modulated optimizing the performance of these cells for the task at hand. The large reductions in burst probability seen following blockade of descending excitation to the pyramidal cell apical dendrites and ELL inhibitory interneurons supports the idea that bursty firing may be under descending control.
Sources of apical dendritic inputs
The pyramidal cell apical dendrites receive input from two
distinct sources. Distal regions of the dendrites receive glutamatergic inputs, via AMPA as well as NMDA receptors, from typical cerebellar parallel fibers (Berman et al. 1997) and enormous
numbers of these granule cell axons comprise the ELL dorsal molecular
layer (DML) (Maler 1979
; Maler et al.
1981
). The granule cell bodies are found in a region
superficial to the DML termed the posterior eminentia granularis (EGp),
and the granule cells receive descending electrosensory inputs as well
as proprioceptive information and, possibly corollary discharges of
motor commands (Sas and Maler 1987
). The descending electrosensory inputs to the EGp originate in the rhombencephalic nucleus praeminentialis (nP), and since the nP also receives ascending electrosensory inputs from the ELL pyramidal cells, the nP to EGp to
DML circuit comprises a feedback loop allowing descending electrosensory as well as nonelectrosensory signals to influence pyramidal cell activity. The nP efferents carrying the descending electrosensory information are tonically active and track longer-term changes in EOD amplitude, and this feedback pathway is thought to be
involved in pyramidal cell gain control (Bastian
1986a
,b
; Bastian and Bratton 1990
; Nelson
1994
; Shumway and Maler 1989
). In addition, this
pathway participates in mechanisms that adaptively cancel reafferent
patterns of sensory input that arise, for example, as consequences of
the animals' body movements (Bastian 1995
, 1996a
,b
, 1998a
,b
,
1999
).
In addition to providing excitatory input directly to pyramidal cell
apical dendrites, the DML parallel fibers excite molecular layer
inhibitory interneurons including the DML stellate cells. The finding
that large CNQX injections result in decreased spontaneous activity
indicates that the direct excitatory drive due to parallel fiber
activity outweighs the disynaptic inhibition mediate via DML inhibitory
interneurons in the absence of changes in electrosensory input.
However, CNQX treatment has the opposite effect on pyramidal cell
responses to electrosensory stimulation; it increases both the
magnitude and duration of responses to EOD increases (Bastian 1993). The seemingly paradoxical finding that a cell's
spontaneous activity is reduced by glutamatergic blockade while
responses to electrosensory stimuli are enhanced was recently suggested by Berman and Maler (1999)
to be due to differences in
the dendritic architecture of the pyramidal and stellate cells. The
large pyramidal cell apical dendrites receive excitatory inputs from
very large numbers of parallel fibers, and, in the absence of
electrosensory stimulation, assumed spatially diffuse activity across
large populations of parallel fibers is expected to provide significant
excitatory drive to pyramidal cells. Hence blockade of this diffuse
excitation reduces spontaneous activity. Dorsal molecular layer
stellate cells, with far smaller dendritic arborizations, are thought
to be minimally driven by this spatially diffuse activity. In response to electrosensory stimulation, however, spatially restricted
populations or "beams" of parallel fibers are proposed to be
activated that could have a proportionally larger effect on the
stellate cells. Hence stimulus-driven DML stellate cells are capable of
inhibiting pyramidal cells, and release from this inhibition occurs
when their parallel fiber inputs are blocked.
More proximal regions of pyramidal cell apical dendrites receive
synaptic input from a directly descending electrosensory feedback
pathway also originating in the nP (Sas and Maler 1987). This pathway terminates in the ELL ventral molecular layer (VML) and
provides both glutamatergic and GABAergic inputs (Maler and Mugnaini 1994
; Wang and Maler 1994
). The
excitatory portion of this pathway shows reciprocal topographic mapping
between the ELL and nP; that is, subsets of ELL pyramidal cells project
to stellate cells of the contralateral nP and the axons of these stellate cells provide descending excitation to the same population of
ELL pyramidal cells (Sas and Maler 1987
). It has been
proposed that the positive feedback loop resulting from these
reciprocal excitatory connections functions as a "sensory searchlight
mechanism" that highlights the representation of important
electrosensory stimuli (Berman and Maler 1998b
, 1999
;
Bratton and Bastian 1990
; Maler and Mugnaini
1994
).
The differences in pyramidal cells' apical dendrites suggest that
those that have the lower spontaneous firing rates, the weaker phase
coupling to the EOD waveform, and the most irregular patterns of
spontaneous activity (large CV and high
proportion of non-Poisson bursts) may also be most sensitive to changes
in dorsal molecular layer activity simply because their more extensive apical dendrites are expected to receive input from larger numbers of
DML parallel fibers. Those with reduced dendritic arborizations are
expected to be less influenced by DML inputs and may have more static
physiological properties. Other properties of the pyramidal cells also
suggest that they comprise a functionally heterogeneous population. The
more superficial cells with extensive dendritic arbors also show more
rapidly adapting responses to long-duration electrosensory stimuli
(Bastian and Courtright 1991). These cells have high
concentrations of both inositol trisphosphate and ryanodine receptors
(Berman et al. 1995
; Zupanc et al. 1992
) as well as high concentrations of the NR2B NMDA receptor subunit (R. J. Dunn and L. Maler, personal communication). That the most superficial pyramidal cells also have the largest intracellular Ca2+ stores while those with the smallest
dendrites have far less (Berman and Maler 1999
) and that
the pyramidal cell plasticity is Ca2+ dependent
(Bastian 1998b
) leads to the prediction that the
superficial pyramidal cells should also show more robust synaptic
plasticity. Hence bursting behavior may not only be modulated by
changes in DML afferent activity, but may also change with alterations
in the strength of these apical dendritic synapses.
Mechanisms of burst generation and functional implications
The mechanism of burst generation in ELL pyramidal cells has been
extensively studied in-vitro (Lemon and Turner 2000;
Turner and Maler 1999
; Turner et al. 1994
,
1996
). Bursts arise as a consequence of prolonged somatic
depolarizations, depolarizing afterdepolarizations (DAPs), that follow
somatic spikes. The somatic spikes can trigger broader dendritic
Na+ spikes that backpropagate within the apical
dendritic tree. The DAPs result from electrotonic conduction of the
longer-duration dendritic Na+ spikes back into
the soma, and if the DAP rises above threshold, additional somatic
spikes are initiated and the process repeats producing spike bursts.
The bursts are terminated when the intraburst spike interval falls
below a critical value and dendritic spike failure occurs due to
refractoriness (Lemon and Turner 2000
).
Neuronal membrane properties including ion channel types and densities
as well as morphological features are known to be correlates of both
spontaneous and driven activity patterns (Connors and Gutnick
1990), and dendritic morphology seems to be a particularly important factor in determining bursting characteristics. A recent modeling study (Mainen and Sejnowski 1996
) showed
that activity patterns ranging from regularly firing to intrinsically
bursting could be demonstrated in neurons having constant ion channel
distributions and differing only in dendritic structure. As originally
described by Turner et al. (1994)
a delayed
depolarization, or DAP, resulting from activation of voltage-gated
dendritic conductances, is critical for the generation of bursts, and
as seen in this study, bursting was positively correlated with the size
of a neuron's dendritic arbor.
In addition to showing that burst probability is related to cellular
morphology, our results suggest mechanisms by which dendritic inputs
may modulate a cell's firing pattern. In vitro studies showed that the
pattern of ELL pyramidal cell activity evoked by depolarizing current
injection could shift from a tonic to a bursty pattern contingent on
the degree to which the cell was depolarized (Lemon and Turner
2000). In the present study, blockade of direct excitatory and
disynaptic inhibitory inputs to the apical dendrites reduced bursting
and hyperpolarizing current injection, resulting in membrane potential
changes as small as
1 to
2 mV, mimicked this result. Hence
modulation of descending inputs could bias pyramidal cells in a way
that determines whether responses to electrosensory stimuli consist of
bursts, enabling the cells to act as feature detectors (Gabbiani
and Metzner 1999
; Gabbiani et al. 1996
;
Metzner et al. 1998
) or consist of more tonic changes in
firing rate that are better suited to encoding detailed information about a stimulus.
Bursts seen in vivo and in vitro also showed some important
differences. In vitro burst durations were much longer and bursts contained many more spikes than seen in vivo. In the former case, the
average numbers of spikes/burst ranged from 8 to 61 depending on the
ELL subdivision or map recorded from (Turner et al.
1996), while in vivo the average was between 2 and 3 spikes/burst (Gabbiani et al. 1996
; and this study).
Spike intervals within bursts recorded from cells in vitro were
typically shorter, ranging from 3 to 13 ms (Lemon and Turner
2000
) compared with 7 to 25 ms in vivo, and this may indicate
differences in the temporal characteristics of the DAPs under in vitro
and in vivo conditions. In vitro burst ISIs also often show serial
patterning; alternating long and short intervals as well as serially
decreasing intervals are seen and, in vitro, the last ISI of a burst is
typically the shortest since refractory effects terminate in vitro
bursts (Lemon and Turner 2000
; Turner and Maler
1999
; Turner et al. 1996
). No consistent intraburst spike interval patterning was seen in vivo, and the last
ISIs within bursts were not systematically different from any other
intraburst ISI. The significantly shorter burst durations seen in vivo
plus the absence of indications that short ISIs lead to burst
termination suggests that processes other than refractory effects
terminate bursts in vivo.
One possibility is that inhibitory interactions within the ELL
terminate bursts and detailed analyses of inhibition in the ELL have
recently appeared (Berman and Maler 1998a-c
, 1999
).
Dorsal molecular layer parallel fibers excite pyramidal cells as well as inhibitory interneurons, DML stellate cells, and ventral molecular layer neurons, and both of these provide GABAA
inhibition to pyramidal cells (Berman and Maler 1998c
).
Stellate inhibition primarily reduces the amplitude of individual
parallel fiber excitatory postsynaptic potential (EPSPs), while VML
neuron inhibition is of longer duration and may modulate persistent
Na+ channel currents (Berman and Maler
1998c
, 1999
). Given the importance of the persistent
Na+ current in burst generation, VML cell
inhibition is especially attractive as a candidate mechanism for burst
termination in vivo. Appropriately timed excitatory parallel fiber
inputs to pyramidal cells could initially augment the somatic
spike-evoked depolarizing afterpotentials initiating bursts, and the
disynaptically evoked inhibition could repolarize the cell terminating
the burst. In addition, to terminating individual bursts, the results
of the hyperpolarizing current injection experiments suggest that
inhibitory inputs to the proximal apical dendrites and somatic regions
could also modulate overall burst probability by participating in the control of a cell's membrane potential.
In addition to the possibility that DML parallel fibers modulate
pyramidal cell burst probability in a tonic fashion by altering the
cell's membrane potential and therefore the probability that a burst
threshold will be exceeded, it is also possible that excitatory ventral
molecular layer inputs trigger individual bursts. Nucleus praeminentialis stellate cells project bilaterally to the ELL ventral
molecular layers and provide powerful excitatory inputs to the initial
segments of the pyramidal cells' apical dendrites (Sas and
Maler 1983, 1987
). These cells have very low rates of spontaneous activity, but they respond with high-frequency bursts of
activity to certain patterns of electrosensory stimuli (Bratton and Bastian 1990
), and stellate cell evoked EPSPs studied in
vivo and in vitro show extreme frequency-dependent posttetanic
potentiation (Bastian 1996b
; Berman and Maler
1998b
, 1999
; Wang and Maler 1997
, 1998
).
The reciprocally topographic relationships between ELL pyramidal cells
and the nP stellate cells and the nP stellate cells' burst-like
responses to appropriate electrosensory stimuli and the highly
facilitating pyramidal cell EPSPs evoked by the descending stellate
activity all support the idea that this circuit behaves as a positive
feedback loop that greatly amplifies responses to certain patterns of
electrosensory input perhaps by evoking pyramidal cell bursts.
Electrosensory information serves at least two separate categories of
behavior. Electrocommunication behaviors involve an animal's detection
of conspecifics' discharges followed by the generation responses such
as the jamming avoidance response (Heiligenberg 1991) or
chirps (Bullock 1969
; Larimer and Macdonald
1968
). The stimuli received during electrocommunication consist
of the sum of an animal's own discharge plus that of a conspecific.
This summation produces a beat waveform consisting of cyclic patterns of amplitude and relative phase modulations (Heiligenberg
1991
). These EOD modulations are spatially extensive and
influence pyramidal cell receptive field centers and antagonistic
surrounds simultaneously. Previous studies demonstrated that pyramidal
cell spike bursts encode the occurrence of EOD amplitude increases and
decreases more reliably than either electroreceptor afferent spikes or
pyramidal cell single spikes (Gabbiani and Metzner 1999
;
Metzner et al. 1998
), and accurate representation of the
timing of these stimulus features is critical for the control of these
electrocommunication behaviors (Heiligenberg 1991
).
Electrolocation stimuli also generate EOD amplitude and phase
modulations, but unlike electrocommunication signals, these AMs are
spatially localized and typically have a movement component. Hence
electrolocation stimuli may affect subdivisions of pyramidal cell
receptive fields sequentially. As in the case of electrocommunication stimuli, increases or decreases in EOD amplitude comprise important electrolocation stimulus features and reliably encoding these as bursts
of spikes produced by a subset of the somatotopically mapped pyramidal
cell population could, for example, indicate the current position of a
moving object. However, it also seems likely that detailed information
about the time course of the AM waveform may be important for
electrolocation and cells with high firing rates are better suited for
encoding this information (Metzner et al. 1998).
Although the highest spontaneous firing rates seen for pyramidal cells,
40-50 spikes/s, are still well below that typical for receptor
afferents (150-400 spikes/s), additional studies focusing on these
nonbursty pyramidal cells are needed to determine their potential
for encoding detailed information about the spatially localized EOD
modulations that occur during electrolocation.
![]() |
ACKNOWLEDGMENTS |
---|
We thank Drs. L. Maler and R. Turner for helpful discussions.
This work was supported by National Institute of Neurological Disorders and Stroke Grant NS-12337 to J. Bastian.
Present address of J. Nguyenkim: Division of Biology and Biomedical Sciences, Washington University, 660 S. Euclid Ave., St. Louis, MO 63110.
![]() |
FOOTNOTES |
---|
Address for reprint requests: J. Bastian, Dept. of Zoology, University of Oklahoma, 730 Van Vleet Oval, Norman, OK 73019 (E-mail: jbastian{at}ou.edu).
Received 20 June 2000; accepted in final form 14 September 2000.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|