Download Experience-dependent corticofugal adjustment

Proc. Natl. Acad. Sci. USA
Vol. 95, pp. 12663–12670, October 1998
Neurobiology
This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences
elected on April 28, 1998.
Experience-dependent corticofugal adjustment of midbrain
frequency map in bat auditory system
(descending systemyhearingylearningyplasticityytonotopic map)
ENQUAN GAO
AND
NOBUO SUGA*
Department of Biology, Washington University, One Brookings Drive, St. Louis, MO 63130
Contributed by Nobuo Suga, August 24, 1998
However, the neural mechanisms underlying these changes
have mostly remained unexplored.
Recent studies indicate that cortical neurons of the mustached bat (Pteronotus parnellii) mediate, via corticofugal
projection, a highly focused positive feedback to subcortical
neurons ‘‘matched’’ in tuning to a particular acoustic parameter and a widespread lateral inhibition to ‘‘unmatched’’
subcortical neurons. This cortical feedback results in the
adjustment and improvement of subcortical signal processing
(i.e., the adjustment and improvement of the cortical neurons’
own input). This function, named egocentric selection, enhances the neural representation of frequently occurring signals in the central auditory system (8–10). Egocentric selection
shifts the BFs of collicular neurons not only toward the BF of
electrically stimulated cortical neurons but also toward the
frequency of a repetitively delivered acoustic stimulus (tone
burst), resulting in local reorganization of the frequency map
in the IC (11). It is likely that the changes observed in
subcortical auditory nuclei after conditioning are partially, if
not totally, mediated by the corticofugal system. Therefore, we
hypothesized that the corticofugal system adjusts and improves
subcortical information processing according to auditory experience based on associative learning.
To test our hypothesis, we delivered an acoustic stimulus
paired with a mild electric stimulus of the leg of the big brown
bat (Eptesicus fuscus). This classical conditioning paradigm
permitted the animal to learn that the acoustic stimulus
predicted the electrical stimulation and to make behavioral
and neuronal adjustments based on the learned importance of
the acoustic stimulus. We found that the frequency-response
curves of neurons in the central nucleus of the IC shift toward
the frequency of the acoustic stimulus, that the AC is necessary
to evoke the shift, and that the IC sustains the shift for some
time once it is evoked, without further input from the AC.
We studied the IC rather than the MGB for the following
reasons. The central nucleus of the IC receives ascending
inputs from 10 or more lower-order auditory nuclei and
projects to the MGB and, to a lesser extent, to other divisions
of the IC and the superior colliculus. In addition, the central
nucleus of the IC receives the corticofugal projections and
projects to the lower-order auditory nuclei. The central nucleus of the IC is an important site where ascending auditory
signals are processed, and the processing is modulated by
descending corticofugal fibers (12). Despite this complexity of
ABSTRACT
Recent studies of corticofugal modulation of
auditory information processing indicate that cortical neurons mediate both a highly focused positive feedback to
subcortical neurons ‘‘matched’’ in tuning to a particular
acoustic parameter and a widespread lateral inhibition to
‘‘unmatched’’ subcortical neurons. This cortical function for
the adjustment and improvement of subcortical information
processing is called egocentric selection. Egocentric selection
enhances the neural representation of frequently occurring
signals in the central auditory system. For our present studies
performed with the big brown bat (Eptesicus fuscus), we
hypothesized that egocentric selection adjusts the frequency
map of the inferior colliculus (IC) according to auditory
experience based on associative learning. To test this hypothesis, we delivered acoustic stimuli paired with electric leg
stimulation to the bat, because such paired stimuli allowed the
animal to learn that the acoustic stimulus was behaviorally
important and to make behavioral and neural adjustments
based on the acquired importance of the acoustic stimulus. We
found that acoustic stimulation alone evokes a change in the
frequency map of the IC; that this change in the IC becomes
greater when the acoustic stimulation is made behaviorally
relevant by pairing it with electrical stimulation; that the
collicular change is mediated by the corticofugal system; and
that the IC itself can sustain the change evoked by the
corticofugal system for some time. Our data support the
hypothesis.
The response properties of neurons in the subcortical auditory
nuclei, as well as in the auditory cortex (AC) can be changed
by associating an acoustic stimulus (conditioning stimulus: CS)
with an electric foot stimulus (unconditioned stimulus). After
conditioning, the response of neurons in the AC (1–3) and the
medial geniculate body (MGB) of the adult guinea pig (4, 5)
increases to the frequency of a CS tone, but decreases to other
frequencies, including the original best frequencies (BFs) of
the neurons. These changes result in a shift in BF toward the
frequency of the CS tone. The BF shift lasts at least 8 weeks
in the AC (3), at least 1 hr in the dorsal division of the MGB
(4), and less than 1 hr in the ventral division of the MGB (5).
In the rat, the response of neurons in the inferior colliculus
(IC) increases to frequencies used for a discrimination task (6),
and 2-deoxyglucose uptake in the IC increases in the iso-BF
band representing the frequency of the CS tone (7). All of
these studies indicate that information processing in the central auditory system can be modified by auditory experience.
Abbreviations: AC, auditory cortex; AI, primary AC; ASr, repetitive
acoustic stimulus; ASt, a train of acoustic stimuli; BF, best frequency;
BFc, BF in the control condition; BFs, BF in the shifted condition; ESl,
electric shock to the leg; IC, inferior colliculus; MGB, medial geniculate body; PST, poststimulus time; SI, primary somatosensory cortex.
*To whom reprint requests should be addressed. e-mail: suga@
biodec.wustl.edu.
© 1998 by The National Academy of Sciences 0027-8424y98y9512663-8$2.00y0
PNAS is available online at www.pnas.org.
12663
12664
Neurobiology: Gao and Suga
connections, the central nucleus of the IC has a single, simple
cochleotopic frequency map (13).
MATERIALS AND METHODS
Preparation. Experiments were performed with 26 adult big
brown bats, E. fuscus, whose auditory system is basically the
same as that of other mammalian species (13–15). Procedures
for animal preparation, acoustic stimulation, and recording of
action potentials have been previously described (11, 16). The
protocol for this research was approved by the animal studies
committee of Washington University. Under neuroleptanalgesia (Innovar 4.0 mgykg body weight), a 15-mm-long metal
post was glued on the dorsal surface of the bat’s skull.
Experiments for recording auditory responses from single or
multiple neurons began 3–4 days after the surgery. The
unanesthetized bat was placed in a styrofoam restraint suspended by an elastic band at the center of a soundproof room
maintained at a temperature of about 31°C. The head was
immobilized by fixing the metal post glued on the skull onto a
metal rod with set screws, and it was adjusted to face directly
at a loudspeaker located 74 cm away.
Acoustic andyor Electric Stimuli. For the measurement of
the frequency-tuning or frequency-response curves of collicular neurons, acoustic stimuli (20-ms tone bursts with a 0.5-ms
rise-decay time) were delivered to the bat at a rate of 5ys. Their
frequency and amplitude were manually varied. The frequency-tuning curve of single or multiple neurons first was measured audio-visually. Then, the tone bursts were computercontrolled. The sharpness of the manually measured tuning
curve was used to determine the step size (0.2–0.5 kHz) of a
computer-controlled frequency scan: the sharper the tuning
curve, the smaller the step. The frequency scan usually consisted of 21 200-ms time blocks or 34 150-ms time blocks when
a tuning curve was particularly wide. In each scan, a single tone
burst was delivered at the beginning of each block, and the
frequency of the tone burst was shifted in 20 or 33 steps across
the BF of the neuron. The amplitude of the tone bursts in the
scan was always set at 10 dB above the minimum threshold of
the neuron, so as to easily detect a shift in BF. An identical
frequency scan was delivered 50 or 100 times to obtain an array
of poststimulus time (PST) histograms as a function of frequency. BFs and frequency-response curves were obtained by
counting the total numbers of impulses discharged to 50 or 100
identical tone bursts.
To evoke changes in the BFs of collicular neurons, i.e., in the
frequency map of the IC, two types of acoustic stimuli were
delivered. We named them repetitive acoustic stimuli (ASr)
and a train of acoustic stimuli (ASt) for convenience. In ASr
and ASt, short tone bursts were delivered at rates chosen to be
consistent with the acoustic behavior of the bat; the echolocating big brown bat emits short orientation sounds at '10ys
in a search phase, 30–40ys at the middle of an approach phase,
and 100–200ys in a terminal phase (17, 18). In ASr, the tone
bursts at a given frequency between 21 and 55 kHz were 50 dB
SPL (decibels in sound pressure level re. 20 mPa) and 20 ms
long. They were delivered at a rate of 10ys for 30 min In ASt,
the tone bursts at a given frequency also between 21 and 55
kHz were 50 dB SPL and 10 ms-long. The rate of the tone
bursts within ASt was 33ys. ASt was 1.0 s long and was delivered
every 30 s for 30 min (60 times in total).
The electric shock to the bat’s leg (ESl) was a 50-ms
monophasic electric pulse and was delivered every 30 s for 30
min (60 times in total). ESl was either delivered alone or in a
pair with ASt. When paired, it was delivered 1.0 s after or
before ASt. Therefore, ASt 1 ESl was 1.0-s ASt 1 1.0-s gap 1
50-ms ESl, and ES1 1 ASt was 50-ms ES1 1 1.0-s gap 1 1.0-s
ASt. One of these paired stimuli was delivered every 30 s for
30 min. The intensity of ESl was just above the threshold
(0.15–0.57 mA) for eliciting a just-noticeable leg flexion, which
Proc. Natl. Acad. Sci. USA 95 (1998)
was monitored with a strain gauge. ASt and ESl are a conditioning and an unconditioned stimulus, respectively.
There were four steps in our experiments: (step 1) to
confirm the shift in BF, i.e., the shift in the frequency map
observed in the central nucleus of the IC by Yan and Suga (11),
we delivered ASr in the same way as in that research. The BFs
of single or multiple IC neurons were measured before and
after 30-min delivery of ASr. (Step 2) To examine whether
auditory experience based on associative learning evokes a
larger BF shift than that observed in step 1, the BFs of IC
neurons were measured before and after 30-min delivery of
ASt 1 ESl or ES1 1 ASt. They also were measured before and
after 30-min delivery of ASt or ESl. (Step 3) To examine
whether BF shifts in the IC evoked by ASt 1 ESl are mediated
by the corticofugal system, the corticofugal system was inactivated during ASt 1 ESl or immediately after collicular BF
shifts were evoked by ASt 1 ESl by applying 0.4 mg of muscimol
(an agonist of an inhibitory synaptic transmitter, g-aminobutyric acid) to the primary AC (AI) 20 min before or 30–50 min
after ASt 1 ESl. The method used to apply muscimol to the AC
was the same as that in a previous study (19). (Step 4) To
examine whether the primary somatosensory cortex (SI) influences the collicular BF shifts evoked by ASt 1 ESl, the SI
was inactivated during ASt 1 ESl by applying 0.4 mg of
muscimol to the SI 20 min before ASt 1 ESl. The AI and SI,
respectively, were localized by identifying the frequency (15)
and somatotopic (20) maps, and by recording neural responses
to acoustic or touch stimuli before muscimol application.
Recording and Evaluation of Neural Responses. To assess
the BF shifts of collicular neurons (i.e., the change in the
frequency map of the central nucleus of the IC) evoked by
acoustic andyor electric stimulation, multi-unit mapping and
single-unit recording experiments were performed before and
after the stimulation. In the multi-unit mapping experiments,
the BFs of multiple IC neurons (2–3 neuron clusters) first were
recorded with a tungsten-wire microelectrode ('7-mm tip
diameter) at about 100-mm intervals along a dorsoventral
electrode penetration. These recordings yielded a BF-depth
curve in the control condition (i.e., the condition before the
30-min delivery of acoustic andyor electric stimuli). We then
delivered either acoustic stimuli (ASr or ASt), electric stimuli
(ESl), or both for 30 min. Within a half-minute thereafter, we
started to remeasure the BF-depth curve. The curve was
repeatedly measured while the electrode was first withdrawn
near to the surface of the IC while remaining within the IC, and
then reinserted. This procedure continued until the BF-depth
curve returned to its control values (11). It took about 30 min
to obtain a single BF-depth curve. Each multi-unit mapping
experiment produced many data points indicating BF shifts.
In the single-unit experiments, responses of single neurons
to tone bursts were recorded, and their BFs and frequencyresponse curves were measured before and after the stimulations described above. A recording electrode was not moved
until all of the measurements on a given single neuron were
completed. Each single-unit experiment produced just one
datum indicating a BF shift.
The criteria used to define shift in the frequency-response
curve (or BF) of a single neuron by acoustic andyor electrical
stimulation (9, 11) are summarized below. If a shifted frequency-response curve did not recover by more than 50%
before the disappearance of action potentials of the neuron,
the data were excluded from our analysis. In a stable, longlasting recording, all curves shifted by the stimulation recovered by more than 50%. This recovery itself helps to prove that
the shift was significant. Furthermore, the shift was considered
to be significant only if the shift in BF was accompanied by a
shift in the same direction as that of the BF shift of more than
75% of the data points in the frequency-response curve. When
a BF shift was small and its significance was not obvious, a
weighted average frequency (i.e., BF) was calculated for the
Neurobiology: Gao and Suga
summed response to five consecutive frequency scans. Then
the mean and SD of these weighted averages were computed,
and a two-tailed, paired t test was used to determine whether
or not the weighted average frequencies (BFs) obtained for
control and stimulus conditions were significantly different for
P , 0.01. The t test also was used to test the significance of
difference between the data sets obtained before and after the
acoustic andyor electrical stimulation.
RESULTS
BF Shifts in the IC Evoked by Acoustic Stimulation Paired
with Electric Leg-Stimulation. In dorsoventral electrode penetrations through the IC (Fig. 1A, arrows), BFs of neurons
systematically increased, as previously reported by Casseday
and Covey (13). The BF-depth curve remained highly consistent over multiple identical penetrations. That is, the BF
measured at the same location stayed unchanged. However,
when ASr was delivered for 30 min, the BF-depth curve (i.e.,
BFs) was lowered toward the ASr frequency, but only at
locations tuned to frequencies within 10 kHz above the ASr
frequency (Fig. 1C, E), as previously shown by Yan and Suga
(11) (Fig. 1C, short dashed line). For example, BF shift toward
an ASr of 20 kHz occurred only in neurons with a BF between
20 and 30 kHz, whereas BF shift toward an ASr of 35 kHz
occurred only in neurons with a BF between 35 and 45 kHz.
FIG. 1. Shift in the BFs of neurons in the IC evoked by acoustic
stimuli (AS) or AS paired with electric stimuli (ES). A and B show the
left IC (frontal section) and the left cerebral cortex (dorsolateral
surface view), respectively. The numbers and lines in A indicate iso-BF
contour lines. The arrows in A indicate dorsoventral electrode penetrations. AI, primary auditory cortex; CBL, cerebellum; LL, lateral
lemniscus; m.c.a., middle cerebral artery; PAG: periaqueductal gray;
SC, superior colliculus. A and B are based on Dear et al. (15) and
Casseday and Covey (13), respectively. (C) The amount of shift in BF
as a function of the difference between the BFs of collicular neurons
and the frequency of AS. ASr, delivered repetitively for 30 min. ASt,
delivered as a 1.0-s train. ESa, focal electric stimulation of the AI. Each
symbol and vertical bar indicate a mean and a SD, respectively. ASt
alone evoked no BF shift, so that the data points are plotted only for
1–8 kHz differences between IC BF and AS frequency. N, the numbers
of BF-depth curves used for averaging. The frequency of ASt ranged
from 21 to 55 kHz (25.33 6 7.84 kHz, n 5 15). The short and long
dashed curves were obtained by Yan and Suga (11).
Proc. Natl. Acad. Sci. USA 95 (1998)
12665
The mean maximum shift was 0.43 6 0.22 kHz, which occurred
at 5.0 kHz above the ASr frequency.
The behavioral response to an electrical leg stimulation
(ESl) was leg flexion and body movement. When ASt followed
by ESl was delivered for 30 min, the animal showed the
behavioral response to every ESl, and in addition, it showed the
response to ASt in the last 10 min of the 30-min period. The
animal apparently was conditioned, i.e., it learned the relationship between ASt and ESl. When ASt alone was delivered
to the animal for 30 min, 2–3 hr after ASt 1 ESl, it did not
evoke the behavioral response. This lack of behavioral response was probably because of the extinction of the conditioned response by repetitive delivery of tone bursts, which
were used for the measurement of frequency-response curves
during the 2- to 3-hr period.
When ASt 1 ESl was delivered for 30 min, the shift in the
BF-depth curve became approximately three times larger than
that produced by ASr alone, and the shift occurred within 15
kHz above the ASt frequency (Fig. 1C, F) and persisted 2–3 hr,
even though the total number of tone bursts in 60 ASts was only
1,980, instead of 18,000 tone bursts in a single ASr, and each
tone burst in ASt was only 10 ms, instead of 20 ms in ASr. The
mean maximum shift for ASt 1 ES1 was 21.1 6 0.44 kHz,
which occurred at 7.0 kHz above the ASt frequency. Unlike ASt
1ES1, ES1 1 ASt delivered for 30 min did not evoke BF shift
of collicular neurons (0.01 6 0.2 kHz, n 5 11, P . 0.5). Neither
ASt nor ESl delivered alone for 30 min evoked a BF shift (Fig.
1C, Œ). That is, BF shift was 20.01 6 0.18 kHz (n 5 15, P .
0.5) for ASt alone and 0.02 6 0.23 kHz (n 5 15, P.0.5) for ES1
alone. Each mean 6 SD of the above was calculated for
collicular neurons with a BF of 6.0 kHz above ASt frequency,
where the largest BF shift was expected (see Fig. 1C). Because
ASr alone evoked a BF shift, ASt also should have evoked a BF
shift, but it was perhaps too small to be detected.
A BF shift always was associated with frequency-dependent
changes in response magnitude. Before stimulation by ASt, the
single IC neuron in Fig. 2A was tuned to 29.5 kHz (E in Fig.
2 Ac). When a 25.0-kHz tone burst was delivered in the ASt 1
ESl paradigm for 30 min, the neuron’s response to a 29.5-kHz
tone burst decreased by 88%, whereas its response to a
26.5-kHz tone burst increased by 300% (2 in Fig. 2 A a and b).
In other words, the BF of this neuron shifted downward (Fig.
2 Ac, F). About 1 hr after ASt 1 ESl, the frequency-response
curve started to recover (Fig. 2 Ac, Œ). The complete recovery
of the curve occurred 125 min after ASt 1 ESl (dashed curve
in Fig. 2 Ac). The PST histograms in Fig. 2 A a and b show the
responses at the BFs in the control condition (BFc) and in the
‘‘shifted’’ condition (BFs) in four different situations corresponding to the four curves in Fig. 2 Ac.
IC neurons with the same BF as the ASt frequency showed
no change in frequency-response curve or response magnitude
for ASt 1 ESl (Fig. 2B). All collicular neurons with BF lower
than or more than 15 kHz higher than the ASt frequency also
showed no change in frequency-response curve or response
magnitude.
Because the decrease in response at BFc, the increase in
response at BFs, and the BF shift differed quantitatively from
neuron to neuron, all the data obtained from 46 single
collicular neurons are presented in Fig. 3. Thirty of 32 neurons
with BF between the ASt frequency and 14 kHz above it
showed changes in response magnitude (decrease at BFc and
increase at BFs) and downward shift in BF toward the ASt
frequency. The amount of decrease in response at BFc was
about two times larger than the amount of increase in response
at BFs (Fig. 3A, E vs. F). Four neurons with a BF that was the
same as the ASt frequency showed no BF shift, but two of them
showed a small amount of increase in response only at the BFc
(Fig. 3, ‚). Three of four neurons with a BF between the ASt
frequency and 2.5 kHz below it showed a decrease in response
to BFc and an increase in response to BFs. Their BFs were
12666
Neurobiology: Gao and Suga
FIG. 2. Changes or lack of changes in the responses and frequencyresponse curves of two single collicular neurons evoked by ASt 1 ESl.
The responses and frequency-response curves were obtained with tone
bursts fixed at 10 dB above the minimum threshold of a given neuron.
(A) The frequency of ASt was 25.0 kHz, and the BF of the neuron was
29.5 kHz. (B) The frequency of ASt was 44.0 kHz, which was the same
as the BF of the neuron. The data obtained before, immediately after,
and 90 and 125 min (A) or 45 and 95 min (B) after ASt 1 ESl are
represented by open circles, filled circles, filled triangles and dashed
lines, respectively. BFc and BFs are indicated by arrows. The PST
histograms in A a and b show the changes in response at the BFc and
BFs evoked by ASt 1 ESl. The PST histograms in Ba show no change
in response at the BFc, which was not shifted by ASt 1 ESl. In these
PST histograms, 1–4, respectively, correspond to the four conditions
for the frequency-response curves: control, ASt 1 ESl, 90 or 45 min
after and 125 or 95 min after ASt 1 ESl.
shifted either away or toward the ASt frequency. These
changes were small, but statistically significant. As observed
with multi-unit recording, the maximum BF shift was observed
for single neurons with a BF that was 5–7 kHz higher than the
ASt frequency. These single-unit data showed no significant
difference from the multi-unit data (P 5 0.50). The largest BF
shift observed was 3.0 kHz (Fig. 3B). Because BF shifts of
almost all collicular neurons were downward toward the ASt
frequency, the shifts resulted in an over-representation of the
frequency of ASt and a related under-representation of frequencies between 8 and 13 kHz above the ASt frequency.
Effect of Cortical Inactivation on BF Shifts in the IC.
Because focal electrical stimulation of the AC evokes BF shifts
(long dashed line in Fig. 1C) (11) similar to those evoked by
ASt 1 ESl, we hypothesized that the changes evoked by ASt 1
ESl were mediated by the corticofugal system. To test this
hypothesis, cortical inactivation experiments were performed
with muscimol, which is a potent agonist of g-aminobutyric
acid (GABA), an inhibitory synaptic transmitter.
When 0.4 mg of muscimol was applied to the large region of
the primary AC (Fig. 1B) containing matched neurons, which
had the same BF as that of given collicular neurons, the
magnitude of responses of the collicular neurons to tone bursts
Proc. Natl. Acad. Sci. USA 95 (1998)
FIG. 3. Changes in response magnitude (A) and BF (B) of 46 single
collicular neurons as a function of difference in their BFs and the ASt
frequency. (A) Almost all open circles show reduced responses to a
tone burst at the original BF in the control condition (BFc), and all
filled circles show increased responses to a tone burst at the BF shifted
by ASt 1 ESl (BFs). (B) The filled circles indicate that almost all
neurons with a BF higher than the ASt frequency shifted their BFs
downward toward the ASt frequency. The dashed curve in B shows the
BF shift observed in the multi-unit mapping experiments shown in Fig.
1C. The triangles indicate the data obtained from neurons whose BFs
were the same as the ASt frequency. The vertical lines on the right
indicate 6 one SD for variations in response magnitude or BF in the
control condition.
decreased by 29–51% (38 6 8.1%, n 5 8) over 2–3 hr, but their
BFs showed no change (0.00 6 0.21 kHz, n 5 8). Fig. 4A shows
the mean time course of the change in response magnitude
evoked by muscimol in eight neurons. The large effect of
cortical inactivation on collicular responses indicates that the
corticofugal system amplifies collicular responses to single
tone bursts 1.6 times on the average. Similar data to the above
also has been obtained in the mustached bat, in which cortical
inactivation by muscimol reduced collicular responses to single
tone bursts by 34 6 7.8%. That is, corticofugal amplification
was 1.5 times (19).
The mean time courses of changes in response magnitude
and BF evoked by ASt 1 ESl in 15 collicular neurons are shown
in Fig. 4B. The maximum reductions in response (10 6 7.2%)
and BF (1.1 6 0.44 kHz) were obtained by the end of ASt 1
ESl. Fig. 4 A and B is plotted to be used as the reference for
evaluating the data presented in Fig. 4 C and D.
To examine the role of the AC in the collicular changes
evoked by ASt 1 ESl, muscimol was applied to the primary AC,
and then, 20 min later, ASt 1 ESl was delivered to the bat. That
is, ASt 1 ESl was delivered during cortical inactivation. As
expected, the collicular response to ASt was reduced significantly (39 6 9.3%, n 5 8). Interestingly, however, neither a
further reduction in response nor a BF shift was evoked in the
IC by ASt 1 ESl (Fig. 4C). When ASt 1 ESl was first delivered
to the bat, and then 30–50 min later, muscimol was applied to
the primary AC, muscimol evoked an additional reduction in
response (32.1 6 8.1%, n 5 8) to that evoked by ASt 1 ESl
(7.9 6 6.2%, n 5 8), but affected little both the amount (1.3 6
Neurobiology: Gao and Suga
Proc. Natl. Acad. Sci. USA 95 (1998)
12667
FIG. 4. Mean time courses of changes in response magnitude (F) and BF (E) observed in collicular neurons. (A and B) The effect of 0.4 mg
of muscimol applied to the AI and the effect of ASt 1 ESl on 8 and 15 collicular neurons, respectively. Muscimol evoked a reduction of response,
but not a BF shift. ASt 1 ESl evoked a reduction in both response and BF. A response magnitude was measured at the BF in the control or shifted
condition shown by E in the lower graph. (C) Muscimol first was applied to the AI, and then, during AI inactivation, ASt 1 ESl were delivered
to the bat. These stimuli failed to evoke changes in the IC. (D) ASt 1 ESl were delivered to the bat, and, after the changes in the IC were evoked
by these stimuli, muscimol was applied to the AI. The BF shift evoked by ASt 1 ESl stayed unaffected. The arrows indicate the time when muscimol
or ASt 1 ESl was delivered. (C and D) Eight collicular neurons were studied. Each filled or opened circle and vertical bar indicates a mean and
a SD, respectively.
0.32 kHz, n 5 8) and the time course of the BF shift evoked
by ASt 1 ESl (Fig. 4D).
The results shown in Fig. 4 are substantiated with PST
histograms and frequency-response curves of single collicular
neurons in Fig. 5. The neuron in Fig. 5A was tuned to 21.5 kHz
(BFc). Application of muscimol to the primary AC reduced the
response of this neuron at its BFc by 36%, but evoked no BF
shift. The effect of muscimol lasted for nearly 4 hr. Fig. 5B
shows the data obtained from a collicular neuron tuned to 25.0
kHz. When muscimol was applied to the primary AC and then,
20 min later (during muscimol inactivation), ASt 1 ESl was
delivered to the bat, the response of this neuron was affected
by muscimol, but not by ASt 1 ESl. Fig. 5C shows the data
obtained from a collicular neuron when ASt 1 ESl was
delivered to the bat and then, 30 min later (during the change
evoked by ASt 1 ESl), muscimol was applied to the primary
AC. ASt 1 ESl shifted the frequency-response curve of this
neuron toward the ASt frequency along the frequency axis. The
BFs was 56.75 kHz instead of 57.75 kHz (BFc). The subsequent
application of muscimol reduced the curve toward the baseline
without changing BFs. Complete recovery from the effects of
ASt 1 ESl and muscimol occurred 250 min after ASt 1 ESl.
Because ESl excites the somatosensory system, muscimol
was applied to the SI (Fig. 1B), and then, 20 min later (during
muscimol inactivation), ASt 1 ESl was delivered to the bat.
Unlike muscimol applied to the primary AC, muscimol applied
to the SI did not reduce the auditory responses of the eight
collicular neurons studied at all, but did abolish both the
reduction in auditory response and BF shift of collicular
neurons, which otherwise would be evoked by ASt 1 ESl (Fig.
5D).
These cortical inactivation experiments with muscimol indicate the following important facts. (i) The auditory corticofugal system evokes changes in both the auditory response
and the frequency map of the IC, when the animal is exposed
to an acoustic stimulus. (ii) The corticofugal system evokes
larger changes in the IC when the acoustic stimulus becomes
behaviorally relevant to the animal. (iii) The changes in the IC
occur predominantly in neurons tuned to frequencies slightly
higher than the frequency of the acoustic stimulus, resulting in
an augmented representation of that sound in the IC. (iv) Once
the changes in the IC are evoked by the corticofugal system, the
changes persist for some time in the IC without further
corticofugal input. In other words, the IC itself shows plasticity. (v) Subcortical interactions between the ascending auditory and somatosensory systems are not involved in the
changes in the IC evoked by the paired acoustic and somatosensory stimuli. (vi) The SI is necessary for the association of
somatosensory information with auditory information (see
Discussion).
DISCUSSION
Corticofugal System and Plasticity in Central Sensory
Systems. The auditory, visual, and somatosensory systems
have cochleotopic (frequency), retinotopic, and somatotopic
12668
Neurobiology: Gao and Suga
Proc. Natl. Acad. Sci. USA 95 (1998)
FIG. 5. Effects of cortical inactivation on auditory responses and frequency-response curves of four collicular neurons. (A) The effects of 0.4
mg of muscimol applied to the AI on a single collicular neuron. (B) Muscimol was applied to the AI immediately before ASt 1 ESl. That is, ASt
1 ESl were delivered to the bat during the AI inactivation. (C) Muscimol was applied to the AI immediately after the collicular changes were evoked
by ASt 1 ESl. (D) Muscimol was applied to the SI immediately before ASt1 ESl. That is, ASt1 ESl were delivered to the bat during the SI
inactivation. The frequency-response curves and PST histograms were obtained with tone bursts fixed at 10 dB above the minimum threshold of
a given neuron. The PST histograms show the change or lack of change in the response magnitude at the BFc or BFs condition. 1–4, respectively,
correspond to the four conditions in which the frequency-response curves were obtained (see Fig. 2 legend).
maps in their central neural pathways, respectively. These
topographic maps are modified by deprivation, injury and
experience, even in adult animals (21–26). Such plasticity has
been explained by changes in divergent and convergent projections of the ascending sensory system. The contribution of
the massive corticofugal system to the plasticity of sensory
systems has been given little consideration, although there has
been a great deal of data demonstrating the corticofugal
modulation of the responses of subcortical neurons to sensory
stimuli.
In the visual system, studies have shown excitatory and
inhibitory effects of the corticofugal projection on matched
and unmatched thalamic neurons, respectively (27). The contribution of the corticofugal system to visual signal processing
has been demonstrated in many experiments (e.g., refs. 28 and
29). However, modification of the retinotopic map (i.e., shift
in receptive field) in the thalamus via the corticofugal system
has not yet been shown.
In the somatosensory system, modification of the somatotopic map because of somatosensory experience is evident in
the thalamus as well as in the cortex (30, 31). A treatment of
the SI for several months with an NMDA receptor agonist
induces a large change in the somatotopic map in the thalamus
(32). This intriguing finding most likely indicates that the
corticofugal system can modulate the subcortical somatotopic
map.
In the auditory system, modification of the frequency and
echo-delay map in the IC andyor MGB by the corticofugal
system has been demonstrated in bats (8, 9, 11). In the present
experiment, we have further demonstrated that the corticofu-
gal system evokes large plastic changes in the IC during
classical conditioning.
Relationships of Our Findings to Learning, Memory, and
Attention. In our present studies, a classical conditioning
paradigm was used to evoke conditioned behavioral response
to ASt, so that the animal learned the relationship between ASt
and ESl and anticipated ESl following ASt. Because ASt 1 ESl
was repeatedly delivered at a rate of 2ymin for 30 min, the
animal presumably paid attention to ASt once ASt 1 ESl began
to be delivered to the animal. Changes in the BFs and auditory
responses of collicular neurons (hereafter simply collicular
changes) were not evoked by ASt alone, ESl alone, or ES1 1
ASt, but by ASt 1 ESl. Therefore, we conclude that the
collicular changes are the result of auditory experience based
on associative learning and, perhaps, attention. In the mustached bat, it has been demonstrated that egocentric selection
adjusts subcortical signal processing in the same way in both
the frequency (9) and time domains (8). Therefore, it may also
be concluded that the corticofugal modulation described
above occurs not only for a frequency-domain analysis but also
for a time-domain analysis.
In our present studies, we made ASt behaviorally relevant by
pairing it with ESl, and found that ASt 1 ESl evoked collicular
changes that were larger than those evoked by ASr. Therefore,
we may conclude that corticofugal adjustment and improvement of subcortical signal processing is larger for behaviorally
relevant sounds than for behaviorally irrelevant sounds.
As reviewed in the Introduction, neurons in the MGB and IC
of the guinea pig andyor the rat change their responses to
acoustic stimuli according to the frequency of a conditioned
Proc. Natl. Acad. Sci. USA 95 (1998)
Neurobiology: Gao and Suga
acoustic stimulus. These changes are similar to those observed
in the present studies. Therefore, it is likely that the subcortical
changes observed in the guinea pig andyor rat are evoked by
the corticofugal system.
The BF shifts observed in the AC of the guinea pig by
Weinberger et al. (3) are also similar to those evoked by the
corticofugal system, except for the duration of the shift. In our
present studies, we did not measure the cortical BF shifts
evoked by ASt 1 ESl. However, it has been found that BF shifts
basically the same as those reported in the present paper are
evoked in both the IC (11) and the AC of the big brown bat by
focal electric stimulation of the AC, and that the amount of BF
shift is larger for the AC than for the IC (S. A. Chowdhury and
N.S., unpublished work). It also has been found that in the
mustached bat, cortical inactivation with muscimol reduces the
responses of IC and MGB neurons to single tone bursts by 34 6
7.8% and 60 6 16%, respectively (19). This finding certainly
indicates that the AC is influenced more than the IC by the
corticofugal feedback loop. Therefore, the BF shifts observed
in the AC of the guinea pig (3) presumably are influenced by
the feedback loop consisting of the ascending and corticofugal
(descending) systems.
In our current studies, the collicular changes are retained for
up to 3 hr after the animal was exposed to ASt 1 ESl. During
this 3-hr period, we repeatedly delivered tone bursts of different frequencies to measure frequency-response curves.
These tone bursts may have shortened the duration of the
changes, because of a phenomenon called extinction. The
duration of the collicular changes observed probably would last
longer than 3 hr if we could minimize the effect of extinction.
It is also quite possible that the collicular changes would last
longer if ASt 1 ESl were delivered for longer than 30 min. We
also found that the collicular changes, once evoked by ASt 1
ESl, last for some time without a corticofugal input. This
finding indicates that the IC itself has plasticity related to
learning and memory.
Neural Pathways for Collicular Plasticity. In our current
studies, inactivation of either the AI or SI during ASt 1 ESl
abolished the collicular changes, which otherwise would be
evoked by ASt 1 ESl. On the other hand, once the collicular
changes were evoked, inactivation of the AI did not abolish
them. These data indicate that the plasticity of the subcortical
sensory nuclei depends at least partly on neural mechanisms
that have not been considered before. The mechanisms, we
propose, are as follows. The AC first shows plastic changes.
Then, subcortical auditory nuclei are changed by the corticofugal system. The subcortical changes are maintained for
some time without descending cortical input. The AC is
influenced by the subcortical changes, because the ascending
and descending systems form a positive feedback loop. The
changes mediated by this feedback loop are stabilized by an
inhibitory mechanism yet to be explored. Suga et al. (33)
hypothesized that the thalamic reticular nucleus is involved in
this inhibitory mechanism.
The SI, not subcortical multisensory nuclei, plays an essential role in sending somatosensory information to the AC.
There is no direct cortico-cortical connection between the SI
and the AC. How is somatosensory information associated
with auditory information? The amygdala of the limbic system
and the cholinergic forebrain nuclei may be involved.
During learning of a behavioral task, cortical acetylcholine
levels increase (34). Blockage of cortical cholinergic receptors
(35) or lesion of a cholinergic nucleus (nucleus basalis) in the
forebrain (36) prevents learning of a behavioral task. Lesions
of the nucleus basalis prevent reorganization of the SI (37),
whereas electrical stimulation of the nucleus paired with
acoustic stimulation induces long-lasting plastic changes in the
AC (38, 39). AS 1 ES results in increased response in both the
nucleus basalis and the AC in rats (40). The nucleus basalis
responds to stimuli associated with behavior (41). Cortical
12669
sensory areas as well as thalamic sensory nuclei send signals to
the lateral nucleus of the amygdala, which interacts with the
amygdala’s central nucleus interfaced with the motor system
(42).
Based on these findings, LeDoux and Muller (42) proposed
a model to explain how the brain forms memories about
unpleasant experiences. This model contains a pathway from
the sensory thalamus to the amygdala, then to the nucleus
basalis, and then to the primary sensory cortex. This pathway,
as well as the projection from the amygdala to the primary
sensory cortex, modifies neural activity in the primary sensory
cortex. They speculated that this pathway mediated cortical
arousal and attention.
Weinberger (43) proposed a model similar to the model
above to explain cortical frequency-tuning plasticity in the
learning of a classical conditioning paradigm. An auditory
signal (tone burst, conditioning stimulus) is sent to the primary
AC through the ventral division of the medial geniculate body
(MGBv) and also is sent to the magnocellular division of the
MGB (MGBm) and the posterior intralaminar complex (PIN),
both in the thalamus. A somatosensory signal (electric footshock, unconditioned stimulus) also is sent to the MGBm and
PIN, where the somatosensory and auditory signals are associated with each other. The associated signal is sent up to the
AI to modulate cortical frequency tuning. This associated
signal also is sent to the amygdala, which, in turn, presumably
projects to the nucleus basalis. Then, the nucleus basalis
increases cortical acetylcholine levels, so that cortical modulation based on the signal from the MGBm and PIN is
augmented.
Based on our findings and what has been known of the
amygdala and the nucleus basalis, we propose the following
model, which is somewhat different from Weinberger’s model.
ASt and ESl excite the AI and SI, respectively. These sensory
cortices send signals to the amygdala through the association
cortex. When associative learning takes place in the amygdala,
that is, when an animal is conditioned by ASt 1 ESl, the
cholinergic nucleus in the forebrain is excited by ASt 1 ESl
through the amygdala. Then, the acetylcholine level in the
cortex increases and the ASt-related changes in the AC, which
are highly specific to acoustic stimuli, are augmented. These
changes are transmitted down to the subcortical auditory
nuclei by the corticofugal system. As a result, processing of
behaviorally relevant acoustic stimuli is adjusted and improved. Cortical changes, which are highly specific to acoustic
stimuli, are based on egocentric selection mediated by the
corticofugal (descending) system that forms a feedback loop
with the ascending system.
We thank Drs. S. Dear, L. S. Green, M. Konishi, W. E. O’Neill, and
N. M. Weinberger, Mr. N. Laleman, and Ms. J. F. Linden for their
comments on the manuscript. This work has been supported by a
research grant from the National Institute on Deafness and Other
Communicative Disorders (DC 00175).
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Diamond, D. M. & Weinberger, N. M. (1986) Brain Res. 372,
357–360.
Bakin, J. S. & Weiberger, N. M. (1990) Brain Res. 536, 271–286.
Weinberger, N. M., Javid, R. & Lepan, B. (1993) Proc. Natl.
Acad. Sci. USA 90, 2394–2398.
Edeline, J. M. & Weinberger, N. M. (1991) Behav. Neurosci. 105,
154–175.
Edeline, J. M. & Weinberger, N. M. (1991) Behav. Neurosci. 5,
618–639.
Disterhoft, J. F. & Stuart, D. K. (1977) Brain Res. 130, 315–333.
Gonzalez-Lima, F. & Agudo, J. (1990) NeuroReport 1, 161–164.
Yan, J. & Suga, N. (1996) Science 273, 1100–1103.
Zhang, Y., Suga, N. & Yan, J. (1997) Nature (London) 387,
900–903.
Suga, N., Yan, J. & Zhang, Y. (1997) Trends Cognit. Sci. 1, 13–20.
Yan, W. & Suga, N. (1998) Nat. Neurosci. 1, 54–58.
12670
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
Proc. Natl. Acad. Sci. USA 95 (1998)
Neurobiology: Gao and Suga
Oliver, D. L. & Huerta, M. F. (1992) in The Mammalian Auditory
Pathway: Neuroanatomy, eds. Webster, D. B., Popper, A. N. &
Fay, R. R. (Springer, New York), pp. 168–195.
Casseday, J. H. & Covey, E. (1992) J. Comp. Physiol. 319, 34–50.
Jen, P. H., Sun, X. D. & Lin, P. J. (1989) J. Comp. Physiol. A 165,
1–14.
Dear, S. P., Simmons, J. A. & Fritz, J. (1993) Nature (London)
364, 620–623.
Suga, N., O,Neill, W. E., Kujirai, K. & Manabe, T. (1983)
J. Neurophysiol. 49, 1573–1626.
Griffin, D. R. (1962) Symp. Zool. Soc. London 7, 61–72.
Simmons, J. A., Fenton, M. B. & O’Farrel, M. J. (1979) Science
203, 16–21.
Zhang, Y. & Suga, N. (1997) J. Neurophysiol. 78, 3489–3492.
Krubitzer, L. A. & Calford, M. B. (1992) J. Comp. Neurol. 317,
1–30.
Kaas, J. H., Krubitzer, L. A., Chino, Y. M., Langston, A. L.,
Polley, H. & Blair, N. (1990) Science 248, 229–231.
Irvine, D. R. F. & Rajan, R. (1996) Clin. Exp. Pharmacol. Physiol.
23, 939–947.
Merzenich, M. M., Nelson, R. J., Stryker, M. P., Cynader, M. S.,
Schoppmann, A. & Zook, J. M. (1984) J. Comp. Neurol. 224,
591–605.
Recanzone, G. H., Schreiner, C. E. & Merzenich, M. M. (1993)
J. Neurosci. 13, 87–103.
Snyder, R. L., Rebscher, S. J., Cao, K. L., Leake, P. A. & Kelly,
K. (1990) Hear. Res. 50, 7–33.
Weinberger, N. M. (1995) Annu. Rev. Neurosci. 18, 129–158.
Tsumoto, T., Creutzfeldt, O. D. & Legendy, C. R. (1978) Exp.
Brain Res. 32, 345–364.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
Sillito, A. M., Cudeiro, J. & Murphy, P. C. (1993) Exp. Brain Res.
93, 6–16.
Sillito, A. M., Jones, H. E., Gerstein, G. L. & West, D. C. (1994)
Nature (London) 369, 479–482.
Pollin, B. & Albe-Fessard, P. (1979) Brain Res. 173, 431–449.
Garraghty, P. E. & Kaas, J. H. (1991) NeuroReport 2, 747–750.
Ergenzinger, E. R., Glasier, M. M., Hahm, J. O. & Pons, T. P.
(1998) Nat. Neurosci. 1, 226–229.
Suga, N., Butman, J. N., Teng, H., Yan, J. & Olsen, J. F. (1995)
in Active Hearing, eds. Flock, A., Ottoson, D. & Ulfendahl, M.
(Elsevier, London), pp. 13–30.
Orsetti, M., Casamenti, F. & Pepeu, G. (1996) Brain Res. 724,
89–96.
Leanza, G., Muir, J., Nilsson, O. G., Wiley, R. G., Dunnet, S. B.
& Bjorklund, A. (1996) Eur. J. Neurosci. 8, 1535–1544.
Butt, A. E. & Hodge, G. K. (1995) Behav. Neurosci. 109, 699–713.
Webster, H. H., Hanisc, U., Dykes, R. W. & Biesold, D. (1991)
Somatosens. Mot. Res. 8, 327–346.
Bjordahl, T. S., Dimyan, M. A. & Weinberger, N. M. (1998)
Behav. Neurosci. 112, 467–479.
Kilgard, M. P. & Merzenich, M. M. (1998) Science 279, 1714–
1718.
Maho, C., Hars, B., Edeline J. M. & Hennevin, E. (1995)
Psychobiology 23, 10–25.
Richardson, R. T. & DeLong, M. R. (1991) Adv. Exp. Med. Biol.
295, 233–252.
LeDoux, J. E. & Muller, J. (1997) Philos. Trans. R. Soc. London
B 352, 1719–1726.
Weinberger, N. M. (1998) Neurobiol. Learning Memory 69, in
press.