Sound frequency representation in cat auditory cortex

www.elsevier.com/locate/ynimg

NeuroImage 23 (2004) 1246–1255

Sound frequency representation in cat auditory cortex

Vassiliy Tsytsarev,1 Tadashi Yamazaki, Jerome Ribot, and Shigeru Tanaka*

Laboratory for Visual Neurocomputing, RIKEN Brain Science Institute, Wako, Saitama 351-0198, Japan

Received 23 December 2003; revised 24 May 2004; accepted 18 August 2004

Using the intrinsic signal optical recording technique, we reconstructed

the two-dimensional pattern of stimulus-evoked neuronal activities in

the auditory cortex of anesthetized and paralyzed cats. The average

magnitude of intrinsic signal in response to a pure tone stimulus

increased steadily as the sound pressure level increased. A detailed

analysis demonstrated that the evoked signals at early frames were

scaled by the sound pressure level, which in turn indicated the presence

of a minimum level of sound pressure beyond which stimulus-related

intrinsic signal can be generated. Intrinsic signals evoked significantly

by pure tone stimuli of different frequencies were localized and

arranged in an orderly manner in the middle ectosylvian gyrus, which

indicates that the primary auditory field (AI) is tonotopically

organized. The arrangement of optimal frequencies obtained from

optical recordings of the same auditory cortex, which were conducted

on different days, was highly reproducible. Furthermore, other

auditory fields surrounding AI in the recorded area were allocated

based on the observed tonotopicity. We also conducted unit recordings

on the cats used for optical recording with the same set of acoustic

stimuli. The gross feature of the arrangement of optimal frequencies

determined by unit recordings agreed with the tonotopic arrangement

determined by the optical recording, although the precise agreement

was not obtained.

D 2004 Elsevier Inc. All rights reserved.

Keywords: Optical recording; Intrinsic signal; Auditory cortex; Tonotopy;

Cat; In vivo

Introduction

Elucidating the structure of cortical maps is important to obtain

a better understanding of the neural substrate for information

representation and processing in the cerebral cortex. It is widely

accepted that the mammalian sensory cortices contain the maps

1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved.

doi:10.1016/j.neuroimage.2004.08.021

* Corresponding author. Laboratory for Visual Neurocomputing,

RIKEN Brain Science Institute, 2-1 Hirosawa, Wako, Saitama 351-0198,

Japan. Fax: +81 48 467 9685.

E-mail address: [email protected] (S. Tanaka).1 Current address: Human Brain Research Center, Kyoto University

School of Medicine, 54 Shogoin Kawahara-cho, Sakyo-ku, Kyoto 606-

8507, Japan.

Available online on ScienceDirect (www.sciencedirect.com.)

whose topography mirrors the physical arrangement of sensory

inputs in the body. The somatosensory cortex contains a map of the

body (Mountcastle, 1957), the visual cortex a topographically

arranged map of the visual space (Talbot, 1940), and the auditory

cortex a map of sound frequencies (Woolsey and Walzl, 1942). In

the last decade, the technique of optical recording of intrinsic

signals has developed rapidly, thus enabling the visualization of

various cortical maps (Bonhoeffer and Grinvald, 1991; Ts’o et al.,

1990). The intrinsic signal optical recording technique, which

detects small changes in the reflected light intensity caused by

neuronal activities in the brain tissue, has been proven to be useful

for visualizing particularly the tangential organization of orienta-

tion and ocular dominance columns in the visual cortex with a high

spatial resolution (Bonhoeffer and Grinvald, 1991; Ts’o et al.,

1990). Recent studies using optical recording of intrinsic signals

have revealed the presence of tonotopic maps in the auditory cortex

of guinea pigs (Bakin et al., 1996), rats (Bakin et al., 1996;

Tsytsarev and Tanaka, 2002), chinchillas (Harel et al., 2000), cats

(Dinse et al., 1997; Spitzer et al., 2001), and ferrets (Versnel et al.,

2002). Spitzer et al. (2001) have shown the tendency of optimal

frequency shift: low frequencies are represented caudally and high

frequencies rostrally in the cat primary auditory field (AI). Harel et

al. (2000) have visualized tonotopic maps not only in AI but also in

anterior auditory field (AAF) and the secondary auditory field

(AII) in the auditory cortex of chinchillas.

In the present study, we conducted intrinsic signal optical

recording on the cat auditory cortex to examine how far the

stimulus-related signal is reliable for the investigation of tonotopic

organization. It was found that the evoked signals were scaled by

the sound pressure level. The presence of the minimum sound

pressure level at approximately 20 dB SPL, beyond which stimulus-

related intrinsic signals are generated, is consistent with the average

minimum sound pressure level for neuronal spike generation in unit

recordings (Mendelson et al., 1997; Sutter and Schreiner, 1995). It

was demonstrated that optimal frequencies were systematically

arranged in the middle ectosylvian gyrus. The optimal frequency

maps were almost identical between two recordings in the same

auditory cortex conducted with a 4-day interval. The systematic

behavior of stimulus-related intrinsic signal such as the scaling

property and map reproducibility guarantee the reliability of our

optical recording. On the basis of the tonotopic representation

obtained from the optical recordings, we identified interareal

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–1255 1247

borders among the AI, AAF, AII, and the dorsoposterior auditory

field (DP). Furthermore, unit recording experiments were con-

ducted using two cats in which tonotopic maps were optically

reconstructed. The arrangement of optimal frequencies roughly

agreed with the tonotopic arrangement obtained by the optical

recording, even if the agreement was not so excellent.

Methods

Preparation of animals

Experiments were conducted using four cats (cat ID: EH, EL,

EM, and EU) between 9 weeks and 12 months of age. The animals

(500–1500 g) were initially anesthetized with ketamine (15 mg/kg,

i.m.) and xylazine (2.5 mg/kg, i.m.). Intratracheal intubation was

performed and the animals were artificially ventilated with a 1:1

mixture of N2O/O2 containing 0.5–1.0% isoflurane. The electro-

cardiogram, end-tidal CO2, arterial oxygen saturation, and rectal

temperature were continuously monitored and maintained within

normal limits throughout the experiments. The animals were

paralyzed with pancuronium bromide (0.1 mg/kg/h, i.v.). The

initial surgery was performed preceding a series of optical

recording and electrophysiological recording. During the opera-

tion, a chamber made of dental cement was mounted on the skull

with anchoring stainless-steel screws, and the skull (about 8 � 12

mm2) located above the putative cortex was removed. A metal

holder for head fixation during the experiments was installed on

the cranium and fixed also with screws and dental cement. The

dura mater covering the recording area was removed. The interior

of the chamber was filled with 2% agar containing gentamicin

sulfate and dexamethasone sodium phosphate, and sealed with a

polyvinylidene chloride thin film to protect infection by virus and

bacteria. Finally, the chamber was covered with a plastic lid for the

protection of the cortex from mechanical damage. In optical

recording experiments, the lid, thin film, and agar were removed,

the chamber was filled again with fresh agar, and a cover glass was

placed on the agar surface before the solidification of the agar. In

the electrophysiological experiments, the cover glass was not used,

but a few drops of silicon oil were deposited on the agar surface to

prevent it from drying. At the end of the final session of recording,

the animals were sacrificed by an overdose injection with

nembutal. The animals were cared for in accordance with the

dGuiding Principles for the Care and Use of Animals in the Field of

Physiological ScienceT of the Japanese Physiological Society and

the dNIH Guide for the Care and Use of Laboratory AnimalsT.

Acoustic stimulation

Acoustic stimuli were generated digitally (SigGen and RPvds,

Tucker–Davis Technology, USA) and presented via earphones,

whose shape was designed for the external ears of cats. Acoustic

properties of the earphones were calibrated by a measuring

amplifier (ACO 6030, Japan) equipped with a capacitive micro-

phone (type 2124, ACO, Japan). Individual acoustic stimuli were

presented binaurally. Each stimulus was composed of five identical

pure tone peeps (10-ms cosine rise and fall, 100-ms duration, and

500-ms interpeep interval for intrinsic signal optical recording; 5-

ms cosine rise and fall, and 50-ms duration for unit recording). The

pure tones were of four frequencies: 5, 10, 15, and 20 kHz at a

sound pressure level of 40–60 dB SPL.

Intrinsic signal optical recording

The animal was firmly fixed by the headholder on a

vibroisolated table. Intrinsic signals were measured to visualize

functional maps around the AI area using standard techniques

(Bonhoeffer and Grinvald, 1996). The cortex was illuminated with

monochromatic light of 540-nm wavelength (Tsytsarev and

Tanaka, 2002; Versnel et al., 1998) using two fiber-optical light

guides to achieve uniform illumination of the cortical surface as

much as possible. The focal plane was adjusted at 500 Am below

the cortical surface using a tandem-lens microscope arrangement.

Images were obtained using a CCD camera (320 � 240 pixels) and

digitized using a differential video-enhancement system, Capos

(Tsunoda et al., 2001). Video images of the gyral surface between

the dorsal tips of the anterior ectosylvian and posterior ectosylvian

sulci were acquired in 15 frames; the duration of each frame was

660 ms. An optical imaging session began at least 3 h after the

induction of anesthesia and continued for 5–8 h. Acquisition of

intrinsic signals was synchronized with acoustic stimulus presen-

tation so that the stimulus onset was delayed by three frames (1980

ms) in each recording. The data acquisition period was followed by

a silent interval of 12 s. Optical recording experiments were

conducted twice at certain intervals for two cats examined.

In the data analysis, first of all, we subtracted the average signals

over the first three frames preceding the stimulus onset from the

recorded signals at each frame to extract stimulus-related intrinsic

signals. This procedure was applied to the recorded signals at each

trial. Next, we reconstructed intrinsic signals elicited in response to

each acoustic stimulus using the generalized indicator function

method, which can project the optical imaging data onto a subspace

of high signal-to-noise ratios (Yokoo et al., 2001).

Electrophysiological recording

Electrophysiological recording was performed on the two cats

(EM and EU) 5–10 days after the final recording of intrinsic

signals. A reference image of the cortical surface collected under

the illumination of light with 540-nm wavelength was used for

guiding electrode penetration and for the subsequent alignment of

optical and electrophysiological data. After removing the glass

cover, a tungsten microelectrode (about 3.0 MV impedance at 1

kHz) attached to a stepping motor microdrive (PC-5N, Narishige,

Japan) was introduced perpendicularly to the cortical surface. In

most cases, the recording depth ranged from 300 to 700 Am below

the cortical surface. For each recording site, the sound pressure

level was determined to be that at which a click tone could evoke

minimum spike activities. A single peep of pure tone at the sound

pressure level ranging from 40 to 50 dB SPL was presented to each

animal 50 times with 3-s interstimulus intervals. The same set of

pure tone stimuli was used as in optical imaging. Response signals

were amplified (�10000 gain), filtered using a 0.3–3.0 kHz band-

pass filter, and then fed to a window discriminator to construct

raster displays and peristimulus time histograms.

Results

Dependence on sound pressure level

Intrinsic signals evoked by a pure tone stimulus of 10 kHz at

four different sound pressure levels (40, 50, 55, and 60 dB SPL)

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–12551248

were recorded using cat EH. Fig. 1a shows the spatial patterns of

the signals at the 5th frame for the four sound pressure levels. We

observed steady increase of the magnitude of intrinsic signal

evoked by the pure tone stimulus with the increase of the sound

pressure level. Comparison between the stimulations at 55 and 60

dB SPL showed similar basic structure of active domains. This

indicates that the locations of the cores of activated domains did

not change, whereas the magnitudes of the signals differ at

different sound pressure levels.

Particularly, inside the cortical region delimited by the dotted

square, we examined at different sound pressure levels the number

of pixels whose intrinsic signal is stronger than a given control

threshold T. These numbers N plotted for different sound pressure

levels S decreased monotonically as control threshold T increased,

as shown in Fig. 1b. When the sound pressure level increased, the

plotted curve extended to larger T. Next, we plotted the control

thresholds TN that gave N = 500, 1000, 2000, 3000, and 4000 for

different sound pressure levels S = 40, 50, 55, and 60 dB SPL, as

shown in Fig. 1c. The TN for each N depended linearly on the

sound pressure levels, and all the regression lines crossed the

horizontal axis at approximately Smin = 21.5 dB SPL. This analysis

reveals that the relation TN = a(S � Smin) holds, which indicates

that only the slope a depends on N. Inversely, N is uniquely

determined by a. This means that plotting N against T/(S � Smin)

for different sound pressure levels should yield a single curve

irrespective of S, as shown in Fig. 1d. Therefore, intrinsic signals

have a scaling property at least in a local domain exhibiting

significant responses and monotonic behavior for the sound

pressure level. This scaling property suggests that when conducting

optical recording at a single sound pressure level, we can obtain a

sufficient amount of information on activity patterns at least inside

monotonic domains (Sutter and Schreiner, 1995). Accordingly, in

Fig. 1. Dependence of stimulus-related intrinsic signals in response to 15-kHz aco

maps for different sound pressure levels, 40, 50, 55, and 60 dB SPL. The region in

as the sound pressure level increased. (b) Number of pixels that elicited stimulu

against the signal strength for four sound pressure levels. (c) The signal strength b

level. (d) Figure b was replotted against the scaled control threshold.

the present study, the sound pressure level was set at 60 dB SPL for

all other optical recordings.

From the scaling property, we obtained another important

feature: the presence of a minimum sound pressure level around

Smin = 21.5 dB SPL, beyond which stimulus-related intrinsic signal

was elicited. Even if we really changed the sound pressure level

from 0 to 40 dB SPL, we could not have seen clearly the onset of

signal response, because of unreliably weak stimulus-related

intrinsic signal below 40 dB SPL. Instead, utilizing the scaling

property obtained from the reliable signals between 40 and 60 dB

SPL, we could successfully determine the minimum sound

pressure level by extrapolating the regression lines of TN vs. S

beyond the range that we used for recording. Thus determined

minimum sound pressure level is close to the average threshold of

the sound pressure level for the generation of spike activities in the

unit recordings (Mendelson et al., 1997; Sutter and Schreiner,

1995). The existence of the common minimum sound pressure

level implies that the intrinsic signals at very early frames are

correlated with neuronal spike activities.

Intrinsic signals in response to different frequencies

Fig. 3 shows the temporal evolution of the spatial pattern of

intrinsic signals evoked by pure tone stimuli of four frequencies (5,

10, 15, and 20 kHz) at six frames (4th, 6th, 8th, 10th, 12th, and

14th frames), where the data were acquired in the first optical

recording of cat EL. For all the stimulus frequencies, localized

activities increased steadily from the 4th to 10th frames and then

decreased. The activated domains appeared as spots of irregular

shape and not necessarily as strips that might be expected from

isofrequency bands obtained by the conventional electrophysio-

logical mapping. It is likely that the evoked signals expand and

ustic frequency on the sound pressure level for cat EH. (a) Single-condition

side the dotted square shows that stimulus-related intrinsic signals increased

s-related intrinsic signals stronger than a given signal strength was plotted

eyond which N pixels were activated was plotted against the sound pressure

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–1255 1249

shrink without changing the locations of the cores of activated

domains.

Around the 10th frame at which the most significant signals

were observed, the activated domains for 10 and 15 kHz

stimulations were fragmented and widely scattered while 5 and

20 kHz stimulations yielded sharply localized active domains. The

signals elicited in response to 15 kHz were weaker than those

elicited in response to the other frequencies. This is regarded to be

due to the occlusion of the signals by the thick blood vessel

running in the dorsoventral direction.

Each stimulus frequency activated two major domains located at

the rostral and caudal parts, which are indicated by the arrowheads in

Fig. 2. The distance between the two arrowheads is maximal for 5-

kHz stimulation and diminishes as the frequency increases. For 20-

kHz stimulation, the two arrowheads are located very closely. This

suggests that there is a mirror symmetric arrangement of optimal

frequencies. The activated caudal domains revealed a progressive

shift in a caudal-to-rostral direction as the stimulus frequency

increased from 5 to 20 kHz, and the activated rostral domains

changed their locations from the rostral to the caudal part. Compared

Fig. 2. Single-condition maps in response to stimulus frequencies of 5, 10, 15, and 2

The arrowheads indicate the locations of significantly activated domains.

with the previous electrophysiological observations of the character-

istic frequency representation in the cat auditory cortex, the caudal

tonotopic representation exists in AI and the rostral representation in

AAF. The direction of the progressive shift in the probable AI agrees

with the direction of gradient of characteristic frequency arrange-

ments determined by the unit recordings in cat AI (Ehret, 1997). The

location of the thus identified AI based on the tonotopic representa-

tion is consistent with that of the anatomically identified AI, which

extends across the middle ectosylvian gyrus and the banks of the

dorsal tip of the posterior ectosylvian sulcus (Reale and Imig, 1980).

As mentioned before, the activated domains in the probable AI

appeared as spots but not elongated domains, particularly for the 5-,

10-, and 15-kHz representations. Since they were located along the

posterior ectosylvian sulcus, most of the domains were possibly

extended into the dorsal bank of the sulcus. Therefore, we might be

able to observe only parts of the optimal frequency domains,

resulting in spots. On the other hand, optimal frequency domains in a

rostral part seemed to be elongated for the 15- and 20-kHz

representations or scattered for the 10-kHz representation in the

dorsoventral direction.

0 kHz at 60 dB SPL sound pressure level are shown in six frames in cat EL.

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–12551250

Time courses of intrinsic signals

Fig. 3 depicts the time courses of the number of pixels activated

stronger than given signal strengths for individual stimulus

frequencies inside the ROI. The signal strengths for thresholding

were given as one, two, and three times the standard deviation of

the evoked signals for each stimulus frequency as shown in Figs.

Fig. 3. Time courses of the number of pixels at which stimulus-related

intrinsic signals were stronger than three different thresholds for the four

stimulus frequencies. The thresholds of intrinsic signals in a, b, and c were

given by SD, 2SD, and 3SD of signals over the ROI for the first recording

(SD: standard deviation). The timings of stimulation of pure tone peeps are

indicated by the gray short vertical line segments.

4a,b, and c, respectively. In Fig. 3a, evoked signals emerged at the

4th frame immediately after the stimulus onset. Even after the

stimulus offset in the middle of the 7th frame, the signal strengths

still built up steadily to reach maxima around the 10th to 11th

frames (2.6–3.3 s after the stimulus offset), and then the signal

strengths decreased. For higher signal strengths for thresholding,

however, the number of pixels tended to decrease and be more

sharply localized around the 10th frame. For 10- and 15-kHz

frequency stimulation, the number of activated pixels markedly

decreased as the signal strength for thresholding increased. This is

consistent with the observation that the activated domains were

rather scattered for 10- and 15-kHz stimulation, while the activated

domains were closely localized for 5- and 20-kHz stimulation.

Organization of optimal frequency domains

When we define a stimulus frequency that evoked the

maximum strength of intrinsic signal to be an optimal frequency

at each pixel, we can draw the optimal frequency maps as

illustrated in Fig. 4 according to the color code shown below the

figure. The map in Fig. 4a was obtained from the first optical

recording of cat EL. We can find the above-mentioned progressive

shift of optimal frequency domains as the frequency changes from

low to high.

Such a progressive shift can also be seen in the matrix of

correlation coefficients shown in Table 1, where the correlation

coefficient c( f1, f2) between two frequencies f1 and f2 is defined by

the normalized overlap integral of stimulus-related intrinsic signal

strengths:

c f1; f2ð Þ ¼RROI

s xY; f1ð Þs xY; f2ð ÞdxYffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiRROI

s xY; f1ð Þ½ �2dxYq ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiR

ROIs xY; f2ð Þ½ �2dxY

q ð1Þ

Here, s xY; f1ð Þ represents the strength of stimulus-related intrinsic

signal at cortical position xY in response to frequency f1. The

diagonal elements of the correlation matrix are all unity because

they are the normalized overlaps between the identical stimulus

frequencies. As shown in Table 1, the correlation coefficients

decreased with the difference between two frequencies. This

indicates that domains activated by two frequencies were farther

separated as the difference between the frequencies increased. This

supports the visually observed progressive shift of the activated

domains with the stimulus frequency. Thus, part of the auditory

cortex of cat EL is tonotopically organized.

Map reproducibility in intrinsic signal optical recording

To confirm that we could successfully reconstruct tonotopic

maps by the optical recording, we examined the reproducibility of

the maps in the same cat between different recordings conducted

with a certain interexperimental interval. To this end, we compared

optimal frequency maps between the first and second recordings

conducted with a 4-day interval using cat EL. The optimal

frequency map shown in Fig. 4b was obtained from the second

optical recording conducted with a low magnification. The arrange-

ments of optimal frequencies were almost the same between the two

recordings (Figs. 4a,b). Fig. 4c is a montage composed of Figs.

4a,b: The map in Fig. 4a was superimposed on the map in Fig. 4b so

that the blood vessel images between the two figures coincided at

the border of the ROI for the first recording. The continuity in the

Fig. 4. Optimal frequency maps in a and b obtained from the first and second recordings of cat EL with a 5-day interexperimental interval, respectively. (c) Map

a was superimposed in map b so that the images of blood vessels were continuously connected at the edges of the ROI for the first recording.

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–1255 1251

optimal frequency representation seen in Fig. 4c supports the

reproducibility of the map in different optical recordings.

The coefficients of cross-correlations of intrinsic signals

between the two recordings were estimated using the normalized

overlap integrals between the two recordings:

ccross f1; f2ð Þ¼RROI

s1 xY; f1ð Þs2 xY; f2ð ÞdxYffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiRROI

s1 xY; f1ð Þ½ �2dxYq ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiR

ROIs2 xY; f2ð Þ½ �2dxY

q : ð2Þ

Table 2 shows the symmetrized coefficients of cross-correlations

between the first and second recordings; that is, [ccross( f1, f2) +

ccross( f2, f1)]/2. The correlation coefficients in the diagonal

elements are about 0.5 and evidently larger than those in the off-

diagonal elements, whereas the coefficients decrease with the

difference between pairs of frequencies. The large values of the

diagonal elements and small values of the off-diagonal elements

substantiate the reliability of the recorded signals in response to

pure tone stimuli as well as biological plausibility of the tonotopic

map.

Different auditory fields

As found in Fig. 4c, different remote locations were activated by

a common acoustic frequency. Particularly, there were optimal

frequency domains for 15-kHz stimulation (green domains) both

rostrally and caudally to the optimal frequency domain for 20-kHz

stimulation (blue domain). This reveals the mirror symmetric

representation of acoustic frequencies, as pointed out in the

previous subsection. It has been reported that isofrequency domains

in AI and AAF appeared in a mirror-symmetric fashion (Reale and

Imig, 1980). Therefore, it is expected that the interareal border

between AI and AAF is located along the middle of the 20-kHz

optimal frequency domain parallel to the dorsoventral axis. There

was an ambiguous tonotopic representation in the region caudal to

the 5-kHz domain. According to the previous observations of

Table 1

Correlation matrix of tonotopic representation in the first recording, which

was obtained from the calculation based on Eq. (1) given in the text

5 kHz 10 kHz 15 kHz 20 kHz

5 kHz 1.0000 0.2576 0.1188 0.0497

10 kHz 1.0000 0.2477 0.1435

15 kHz 1.0000 0.6046

20 kHz 1.0000

characteristic frequencies in several auditory fields (Reale and Imig,

1980), the larger caudal region exhibiting ambiguous tonotopicity

may correspond to the dorsoposterior auditory field (DP).

We attempted to allocate several different auditory fields in the

two cats examined, which are shown in Fig. 5a for cat EL and Fig.

5b for EM. Fig. 5b again reveals that optimal frequency domains

are elongated or scattered along the dorsoventral axis. It has also

been revealed that the characteristic frequencies are inconsistent

across the interareal border between AI and DP (Reale and Imig,

1980). Therefore, in Fig. 5b for cat EM, the interareal border

between AI and DP was determined by the discontinuity in

frequency representation, across which the 5-kHz representation

(red) abruptly changed to the 15-kHz representation (green). As

seen in Figs. 5a,b, the location of interareal borders determined by

optical recording varies among individual cats. We could not detect

any significantly intense signals in the dorsal part of the probable

AI in cat EL. The area in which significantly intense signals could

not be detected may correspond to AII because neurons in AII are

broadly tuned and have relatively higher thresholds in the sound

pressure level (Schreiner and Cynader, 1984). In cat EM, we

accidentally removed the skull and dura mater covering the

suprasylvian gyrus (area 7). We could detect significant signals

in response to pure tone stimuli even in this area, which is regarded

to be outside the proper auditory fields (Fig. 5b). This unexpected

observation of intrinsic signals in response to pure tone stimuli

may corroborate the idea that area 7 works as an association cortex

integrating polysensory information (Thompson et al., 1963a,b).

Comparison with electrophysiological recordings

Extracellular unit recordings were also performed on two cats

(cats EM and EU) before the off-line analysis of recorded intrinsic

signals was conducted. Examples of electrophysiological responses

obtained from eight recording sites in cat EU are illustrated in Fig. 6.

Table 2

Cross-correlation matrix of tonotopic representations between the first and

second recordings, which was obtained from the calculation based on Eq.

(2) and the symmetrization procedure described in the text

5 kHz 10 kHz 15 kHz 20 kHz

5 kHz 0.5019 0.3702 0.1271 0.0354

10 kHz 0.5823 0.3548 0.1641

15 kHz 0.4616 0.4805

20 kHz 0.4854

Fig. 5. Hypothetical interareal borders are indicated by the dotted curves superimposed on the optical frequency maps for cat EL in a and for cat EM in b. The

sulci were indicated by solid curves. AI: primary auditory field; AAF: anterior auditory field; DP: dorsoposterior auditory field; SSS: suprasylvian sulcus; AES:

anterior ectosylvian sulcus; PES: posterior ectosylvian sulcus.

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–12551252

Spike activities were recorded by level triggering only at recording

sites where spikes of a unit could be clearly distinguished visually

using an oscilloscope. In practice, most of the recordings were

obtained at depths between 300 and 700 Am. First, the minimum

Fig. 6. Peristimulus time histogram of units recorded in the auditory cortex of cat

indicated by the hatched band in gray. Most of the units exhibited a single optimal

for h. Two units showed double optimal frequencies: 5 kHz and 10 kHz for c; an

sound pressure level for a click tone stimulus, beyond which spike

activities were elicited, was determined at each recording site. For

all subsequent recordings of responses to pure tone stimuli, we used

sound pressure levels higher than thus determined minimum sound

EU. The range of fluctuations in spontaneous spike activity of each unit is

frequency: 5 kHz for a and b; 10 kHz for e; 15 kHz for f and g; and 20 kHz

d 10 kHz and 15 kHz for d.

Fig. 7. Locations of recorded units superimposed on the images of domains activated by stimulus frequencies in intrinsic signal optical recordings of cats EU

and EM. The symbols indicating the recording sites specify the optimal frequency determined by unit recording. The letters in a correspond to the letters

indicating the peristimulus time histograms in the Fig. 6. The regions delineated by the colored loops show the domains activated more strongly than the

standard deviation of signals over the ROI in intrinsic signal optical recordings.

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–1255 1253

pressure level. In Fig. 6, the stimulus-related spike responses can be

identified by the average spike number per second beyond the range

of fluctuation in spontaneous spike activity, which is indicated by

the gray band. Most of the units showed significant responses to

single optimal frequencies, although two units exhibited double but

adjacent optimality in spike activities, as shown in Figs. 6c (5 and 10

kHz) and d (10 and 15 kHz).

The cortical sites of unit recordings were plotted, superimposing

them on the maps of significant activation of intrinsic signals, as

shown in Fig. 7, where the activated domains are marked by loops

of different colors indicating the stimulus frequencies. The loops

were drawn along the contour at the standard deviation of stimulus-

related intrinsic signals. A larger portion of units responding

strongly to given pure tone frequencies were located inside the

domains activated by the same frequency in optical recording:

62.5% of recorded units (5 units out of 8) for cat EU (Fig. 7a) and

66.6% (6 units out of 9) for cat EM (Fig. 7b) were consistent in

optimal frequency with the optical recordings. Therefore, though

the gradient behavior of optimal frequency representation was well

preserved in the two recording methods, precise matching of

optimal frequency representation was not obtained.

Discussion

In the present study, monochromatic light of 540-nm wave-

length was used for cortical illumination. Intrinsic signal at 540-

nm wavelength of the light is often thought to reflect mostly

cortical blood volume (CBV) changes in the parenchymal and

nonparenchymal tissue, while the more conventional 610-nm

wavelength for optical imaging probes into the deoxygenation

content of the parenchyma. Whether the signal based on CBV can

resolve cortical structure at columnar resolution remains highly

controversial. We have tried to visualize tonotopic maps in rats and

cats using the light of 630- and 700-nm wavelengths, both of

which are usually used for imaging orientation and ocular

dominance maps in the cat visual cortex in our laboratory (Ohki

et al., 2000; Tani et al., 2003). However, we could obtain only

unreliably weak stimulus-related signal. We had no way other than

to use the light of shorter wavelengths for auditory cortex imaging.

In the optical imaging of the ferret auditory cortex conducted by

Versnel et al. (2002), which is one of the most successful imaging

of tonotopic maps, green light (546 nm) illumination was adopted.

They mentioned that the usage of the green light significantly

improved both signal strength and signal-to-noise ratio. Harel et al.

(2000) reconstructed very impressive tonotopic maps in chinchilla

auditory cortex using the light of 540-nm wavelength same as

ours. Several other laboratories attempting intrinsic signal imaging

of auditory cortex also adopted green light illumination (Dinse et

al., 1997; Spitzer et al., 2001).

Recently, Grinvald et al. (2001) have shown that light scattering

signal in response to electrical stimulation was measured at 540-nm

wavelength of the light in a blood-free hippocampal slice

preparation. They also reported that the change of the reflected

light intensity related to neuronal activity was up to 6% of the total

reflected light intensity for 540 nm, while it was about 0.1–0.2%

for 605 nm. It is currently accepted that although intrinsic signal

has different components that originate in different signal sources,

resultant functional maps at different wavelengths of the light are

very similar. Therefore, it appears that all of these components can

be used for functional mapping (Grinvald et al., 2001). Never-

theless, our optically reconstructed tonotopic maps were not as

clear as optically reconstructed functional maps in the cat visual

cortex. This probably came from the difference in the spatial

resolutions of the stimulus-related signals at 540-nm wavelength

and longer wavelengths.

In our intrinsic signal recordings, different localized activities in

single condition maps were observed in response to different

frequencies of pure tone stimuli (Fig. 2). As the sound pressure

level increased, activated domains expanded with the increase in

the strength of evoked signals (Fig. 1). This is probably due to the

observation that a pure tone stimulus at higher sound pressure

levels can activate not only neurons for which a stimulus frequency

corresponds to the characteristic frequency but also other neurons.

That is, as the sound pressure level increases, the number of

activated neurons increases. As previously reported (Mendelson et

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–12551254

al., 1997; Sutter and Schreiner, 1995), the single-unit threshold for

evoking spike responses was approximately 20 dB SPL, below the

sound pressure levels for the reliable induction of intrinsic signals.

Therefore, the suprathreshold stimulation in intrinsic signal optical

recording may activate a larger cortical territory beyond an iso-

frequency band determined by the characteristic frequency-based

electrophysiological mapping. On the other hand, the spatial

profiles of stimulus-related intrinsic signals for 55 and 60 dB

SPL were evidently similar. This indicates that the location of a

core region in the activated domain does not change so much

whereas the activated domain expands.

To confirm the reliability of our optical recordings for the

reconstruction of tonotopic maps, we carried out optical recordings

twice from the same cortex with a 4-day interexperimental interval

and compared tonotopic maps between the two recordings. We

could find excellent reproducibility even though the interexper-

imental interval was as long as 4 days. Furthermore, we carried out

unit recordings from the auditory cortex, in which optical

recording had been conducted. We compared the locations of

units exhibiting optimal frequencies with the tonotopic maps

obtained from the optical recording, and found that the optimal

frequencies were roughly consistent on the whole between the two

measurements.

It has been reported that neurons tuned to particular sound

pressure levels were observed in AI of an anesthetized cat (Brugge

et al., 1969; Phillips and Irvine, 1981). Sutter and Schreiner (1995)

have found a cluster of neurons exhibiting nonmonotonic behavior.

A similar nonmonotonic behavior has also been observed by Dinse

et al. (1997) by using intrinsic signal optical recordings of the cat

auditory cortex. Regarding the dependence of intrinsic signals on

the sound pressure level, we rather analyzed signals in the

monotonic domain to show the scaling property and the presence

of a minimum sound pressure level for the generation of stimulus-

related intrinsic signals. The minimum sound pressure level

obtained from this analysis was about 21.5 dB SPL, which is

close to the average threshold of sound pressure level for the

generation of neuronal spikes (Mendelson et al., 1997). This

implies that stimulus-related intrinsic signals are correlated with

neuronal spike activities.

The tonotopic arrangement was almost consistent with the

previously reported electrophysiological mapping of the character-

istic frequencies in the cat auditory cortex. We observed, however,

that the active domains in the single condition maps often

appeared as spots of irregular shape and not as strips elongated

in the dorso-ventral direction. In addition, the agreement of

optimal frequency arrangements between optical and unit record-

ings was not excellent. Spitzer et al. (2001) pointed out a

discrepancy in the tonotopic representation in cat AI between

characteristic frequency-based tonotopic maps and activated

domains in intrinsic signal optical recordings. Intrinsic signal

optical recordings of the rat AI also revealed the presence of spot-

like activated domains (Bakin et al., 1996), whereas in the

electrophysiological mapping, narrow iso-frequency bands were

observed (Ehret, 1997). They suggested that the major source of

the discrepancy is the involvement of subthreshold neuronal

activity in stimulus-related intrinsic signals. In our analysis of the

dependence of intrinsic signals on the sound pressure level, we

found that intrinsic signals were correlated with suprathreshold

neuronal activities at least in early frames of recordings. Therefore,

the discrepancy between unit recordings and intrinsic signal

optical recordings may not simply be accounted for by the

contribution of subthreshold neuronal activity to the stimulus-

related intrinsic signals. The discrepancy may be attributable to

variability in the dependence of neuronal activities on sound

intensity. Different neurons possess different thresholds of sound

pressure levels for spike generation. Even at the same sound

intensity, neurons with lower thresholds may respond vigorously

to stimuli and elicit strong intrinsic signals, while neurons with

higher thresholds may show sluggish responses and elicit weak

intrinsic signals. In addition, some neuronal population has a

nonmonotonic property of response to the sound intensity. Such

variability in the sound intensity dependence of neuronal activities

is likely to produce nonuniform signal distribution, which may

form spot-like activated domains.

Based on the acoustic frequency representation obtained from

optical recording, we attempted to allocate different auditory fields

in the two cats examined, which are shown in Fig. 5. It is likely

that we can allocate not only AI but also other auditory fields such

as AAF, AII, and DP. The anterior auditory field was identified by

the mirror symmetric representation of the optimal frequencies, and

AII and DP were identified by the absence of tonotopic

organization and discontinuous changes in optimal frequency

representation beyond AI and AAF. Particularly, in AI, the high-

frequency domains were likely to be in the rostral part and the low-

frequency domains in the caudal part. This consistency in

representation may guarantee that our intrinsic signal optical

recording could work well for the visualization of a tonotopic

map in the cat auditory cortex.

We found that there were disadvantages of the intrinsic signal

optical recording from the cat auditory cortex. The optically

reconstructed tonotopic maps were not necessarily as clear as

orientation maps reconstructed in the same recording technique

from the cat visual cortex (Bonhoeffer and Grinvald, 1991; Ohki

et al., 2000), due to worse spatial resolution at 540-nm

wavelength of the light. We could see only rough agreement

in the location of optimal acoustic frequencies between the unit

recording and the optical recording. Nevertheless, the optical

recording still has an advantage in that we can visualize wide-

field tonotopic organization over the ectosylvian gyrus only by

several hours recording, even if the tonotopic map may not be as

precise as that reconstructed by unit recording. Therefore,

intrinsic signal optical recording can be safely used to obtain

rough tonotopic organization and broad allocation of different

auditory fields as a reference map for electrode penetration in

the subsequent electrophysiological recording, which may reduce

the time for searching optimal frequencies in unit recording

experiments.

Conclusion

The intrinsic signal optical recording technique was applied to

the visualization of acoustic frequency representation in the cat

auditory cortex. The tonotopic arrangements were visualized in the

middle ectosylvian gyrus, in which AI and AAF were identified

based on the mirror symmetric representation of acoustic frequen-

cies. The surrounding auditory fields such as AII and DP were also

distinguished by the absence of tonotopicity. The reproducibility of

optically reconstructed tonotopic maps indicates that the optical

recording technique is a reliable method for the visualization of

tonotopic organization and the allocation of different auditory

fields in the cat auditory cortex.

V. Tsytsarev et al. / NeuroImage 23 (2004) 1246–1255 1255

Acknowledgments

We thank Dr. Kazuyuki Imamura (Lab. for Visual Neuro-

computing, RIKEN BSI) and Dr. Hisayuki Ojima (Lab. for Cortical

Organization and Systematics, RIKEN BSI) for valuable advice

concerning the electrophysiological part of experiments, Dr. Dmitri

Sibarov (St. Petersburg State University, Russia) for help in data

analysis, and Mr. Yuji Akimoto and Katsuya Ozawa (Lab. for

Visual Neurocomputing, RIKEN BSI) for technical support in all

our experiments.

References

Bakin, J.S., Kwon, M.C., Masino, S.A., Weinberger, N.M., Frostig, R.D.,

1996. Suprathreshold auditory cortex activation visualized by intrinsic

optical imaging. Cereb. Cortex 6, 120–130.

Bonhoeffer, T., Grinvald, A., 1991. Iso-orientation domains in cat

visual cortex are arranged in pinwheel-like patterns. Nature 353,

429–431.

Bonhoeffer, T., Grinvald, A., 1996. Optical imaging based on intrinsic

signals: the methodology. In: Toga, A.W., Mazziotta, J.C. (Eds.), Brain

Mapping: The Methods. Academic Press, San Diego, p. 55.

Brugge, J.F., Dubrovsky, N.A., Aitkin, L.M., Anderson, D.J., 1969.

Sensitivity of single neurons in auditory cortex of cat to binaural tonal

stimulation; effects of varying interaural time and intensity. J. Neuro-

physiol. 32, 1005–1024.

Dinse, H., Godde, B., Hilger, T., Reuter, G., Cords, S.M., Lenarz, T., von

Seelen, W., 1997. Optical imaging of cat auditory cortex: cochleotopic

selectivity evoked by acute electrical stimulation of a multichannel

cochlear implant. Eur. J. Neurosci. 9, 113–119.

Ehret, G., 1997. The auditory cortex. J. Comp. Physiol 181, 547–557.

Grinvald, A., Shoham, D., Shmuel, A., Glaser, D., Vanzetta, I., Shtoyer-

man, E., Slovin, H., Sterkin, A., Wijnbergen, C., Hildesheim, R., Arieli,

A., 2001. In-vivo optical imaging of cortical architecture and dynamics.

In: Windhorst, U., Johansson, H. (Eds.), Modern Techniques in

Neuroscience Research. Springer.

Harel, N., Mori, N., Sawada, S., Mount, R.J., Harrison, R.V., 2000. Three

distinct auditory areas of cortex (AI, AII and AAF) defined by optical

imaging of intrinsic signals. NeuroImage 11, 302–312.

Ohki, K., Matsuda, Y., Ajima, A., Kim, D.-S., Tanaka, S., 2000.

Arrangement of orientation pinwheel centers around area 17/18

transition zone in cat visual cortex. Cereb. Cortex 10, 593–601.

Mountcastle, V.B., 1957. Modality and topographic properties of

single neurons of cat’s somatic sensory cortex. J. Neurophysiol. 20,

408–434.

Mendelson, J.R., Schreiner, C.E., Sutter, M.L., 1997. Functional

topography of cat auditory cortex: response latencies. J. Comp. Physiol.,

A 181, 615–633.

Phillips, D.P., Irvine, D.R.F., 1981. Responses of single neurons in

physiologically defined primary auditory cortex (AI) of the cat:

frequency tuning and response to intensity. J. Neurophysiol. 45, 48–58.

Reale, R., Imig, T., 1980. Tonotopic organization in auditory cortex of the

cat. J. Comp. Neurol. 192, 265–291.

Schreiner, C.E., Cynader, M.S., 1984. Basic functional organization of

second auditory cortical field (AII) of the cat. J. Neurophysiol. 51,

1284–1305.

Spitzer, M.W., Calford, M.B., Clarey, J.C., Pettigrew, J.D., Roe, A.W.,

2001. Spontaneous and stimulus-evoked intrinsic signals in primary

auditory cortex of the cat. J. Neurophysiol. 85, 1283–1298.

Sutter, M.L., Schreiner, C.E., 1995. Topography of intensity tuning in cat

primary auditory cortex: Single-neuron versus multiple-neuron record-

ings. J. Neurophysiol. 73, 190–204.

Talbot, S.A., 1940. Arrangement of visual field on cat’s cortex. Am. J.

Physiol. 129, 477–478.

Tani, T., Yokoi, I., Ito, M., Tanaka, S., Komatsu, H., 2003. Functional

organization of the cat visual cortex in relation to the representation of a

uniform surface. J. Neurophysiol. 89, 1112–1125.

Thompson, R., Johnson, R., Hoopes, J., 1963a. Organization of auditory,

somatic sensory, and visual projections to association fields of cerebral

cortex in the cat. J. Neurophysiol. 26, 343–364.

Thompson, R., Smith, H., Bliss, D., 1963b. Auditory, somatic sensory, and

visual response interactions and interrelations in association and

primary cortical fields of the cat. J. Neurophysiol. 26, 365–378.

Tsunoda, K., Yamane, Y., Nishizaki, M., Tanifuji, M., 2001. Complex

objects are represented in macaque inferotemporal cortex by the

combination of feature columns. Nat. Neurosci. 4, 832–838.

Ts’o, D.Y., Frostig, R.D., Lieke, E., Grinvald, A., 1990. Functional

organization of primate visual cortex revealed by high-resolution

optical imaging. Science 249, 417–420.

Tsytsarev, V., Tanaka, S., 2002. Intrinsic optical signal from rat primary

auditory cortex in response to sound stimuli presented to contralateral,

ipsilateral and bilateral ears. NeuroReport 13, 1661–1665.

Versnel, H., Mossop, J.E., Moore, D.R., 1998. Optical imaging of intrinsic

signals in ferret auditory cortex. Br. J. Audiol. 32, 93–94.

Versnel, H., Mossop, J.E., Mrsic-Flogel, T.D., Ahmed, B., Moore, D.R.,

2002. Optical imaging of intrinsic signals in ferret auditory cortex: Res-

ponses to narrow-band sound stimuli. J. Neurophysiol. 88, 1545–1558.

Woolsey, C.N., Walzl, E.M., 1942. Topical projection of nerve fibers from

local regions of the cochlea to the cerebral cortex of the cat. Bull. Johns

Hopkins Hosp. 71, 315–344.

Yokoo, T., Knight, B.W., Sirovich, L., 2001. An optimization approach

to signal extraction from noisy multivariate data. NeuroImage 14,

1309–1326.