Sound frequency representation in cat auditory cortex Vassiliy Tsytsarev, 1 Tadashi Yamazaki, Je ´ro ˆme 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 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 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.) www.elsevier.com/locate/ynimg NeuroImage 23 (2004) 1246 – 1255
10
Embed
Sound frequency representation in cat auditory cortex
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
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
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
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
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