| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Department of Otolaryngology—Head and Neck Surgery, and 2W. M. Keck Center for Integrative Neuroscience, University of California, San Francisco, California 94143; and 3Laboratoire Intégration de données multimédia, Unité Propre de Recherche de l'Enseignement Supérieur Equipe d'Accueil 3192, Faculté de Médecine, 35043 Rennes Cedex, France
Submitted 9 January 2003; accepted in final form 26 September 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
; Reale and Imig 1980). The AI representation comprises tonotopically arrayed isofrequency domains resulting in an inhomogeneous, systematic spatial distribution of characteristic frequency (CF; i.e., the frequency of a tone burst to which a neuron is most sensitive).
In the adult cat, the spatial distributions of other response parameters are also nonhomogeneous, such that neighboring locations within the AI often exhibit similar tuning bandwidths (Schreiner and Mendelson 1990) and rate/level functions (Clarey et al. 1994; Imig et al. 1990; Phillips et al. 1985, 1994; Schreiner et al. 1992). Tuning bandwidths of local cell clusters in the central portion of the mature AI are sharper than bandwidths in flanking ventral and dorsal regions (Schreiner and Mendelson 1990). The spatial distribution of rate/level function shapes is also nonuniform, and sites exhibiting nonmonotonic rate/level functions often occur in spatially distinct patches (Clarey et al. 1994; Imig et al. 1990; Phillips et al. 1985, 1994; Schreiner et al. 1992). Regions of broad or narrow tuning, as well as regions of monotonic or nonmonotonic rate/level functions, are found within each isofrequency domain of the AI. While the algorithmic purpose of such an organization is still unclear, it represents a functional differentiation, similar to that found in the visual cortex, which is critical for its contribution to higher cortical processes.
To reach the adult configuration, cortical responses go through a period of development during which several response parameter values are modified. Previous studies have revealed that the statistical distribution of response properties of AI neurons changes from birth through an extended postnatal period (Brugge et al. 1988; Eggermont 1996): minimum response thresholds and latencies decrease with age, mean and maximum tuning bandwidths increase, the ability to follow repetitive stimuli improves, and spontaneous firing under anesthesia increases. Exponential time constants for these changes range from 5 to 50 days (Eggermont 1996). The changes in postnatal response reflect structural and functional development of cortical circuitry as well as peripheral auditory structures (Brugge et al. 1988; Eggermont 1996). For example, before postnatal day 10, minimum thresholds are high, typically >80 dB SPL. Thresholds rapidly decrease and reach adult values, near 5 dB SPL, by postnatal day 20. A similar change is observed in minimum thresholds in the cochlear nucleus (Brugge 1992), inferior colliculus (Blatchley and Brugge 1990), and auditory nerve (Kettner et al. 1985), indicating that the cortical threshold change is predominantly due to the threshold decrease in the periphery. In contrast, the average bandwidth of cortical tuning curves, which are initially narrow, increases to a nearly adult value only around postnatal day 40 (Eggermont 1996). During this period, tuning curve bandwidths in the inferior colliculus decrease by about half (Moore and Irvine 1979), suggesting that the increase in average bandwidth found in the cortex is due to development of central neuronal circuitry. Neural contributions to changing response properties include changes in myelination, specific membrane conductances, number and efficacy of synaptic contacts, synaptic transmission delays, nerve fiber conduction delay, and differential changes of relative excitatory and inhibitory input to neurons throughout the auditory pathway. Structural contributions include clearing of the middle ear, calcification of the ossicles, and changes in the stiffness and mass of the basilar membrane and organ of Corti (reviewed in Romand 1997 and Rubel et al. 1998).
Although temporal changes in statistical distributions of these response parameters have been studied, the development of the spatial distributions of response parameters in the immature AI remain uncharacterized. Earlier studies of postnatal development of these cortical response properties (Brugge et al. 1988; Eggermont 1996) have pooled data from unspecified regions of the AI, precluding conclusions regarding the spatial distributions of maturing responses. Two alternative hypotheses as to the time course of development of these spatial distributions exist. The initial distributions of features such as bandwidth in the young cortex may be spatially nonuniform and bandwidths increase globally during development. Alternatively, the distributions of bandwidth and rate/level functions in the young cortex may be more uniform initially, and bandwidths change nonuniformly in segregated regions during development. The current study examines the spatial organization of developing response properties by comparing extracellular multiunit responses to pure tone bursts recorded from entire AI isofrequency contours in kittens of different ages. Preliminary accounts of these results were previously published in abstract form (Bonham and Schreiner 1998; Bonham et al. 2001).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Fifteen cats of both genders, ranging in age from 2 to 16 wk were studied. Animals were initially anesthetized with an intramuscular injection of a mixture of ketamine hydrochloride (22 mg/kg) and acepromazine (0.11 mg/kg). After venous cannulation, sodium pentobarbital was administered and titrated to achieve an areflexive state for surgery. Lactated Ringer solution was delivered intravenously, as were the supplementary doses of sodium pentobarbital needed to maintain the areflexive state during recording. Heart and respiration rates, end-tidal CO2, and reflexes were monitored throughout surgery and recording to ensure that the anesthetic state was maintained. A tracheotomy was performed to maintain patency of the airway and to minimize breathing noise, and core temperature was held constant using a thermostatically controlled water blanket.
A custom-designed fixture that left the ears unobstructed immobilized the head, and a hollow tube was inserted into the left ear canal. The right primary auditory cortex was exposed by craniotomy over the middle ectosylvan gyrus. The dura was incised and reflected to expose the auditory cortex to which heavy silicone oil was applied to prevent desiccation. A digital image of the exposed cortex, including a scale marker, was recorded and the location of each penetration was documented on this image.
Experiments were performed in a double-walled sound-attenuating chamber (IAC). Auditory stimuli were delivered through a headphone (STAX-54) enclosed in a small chamber that was connected by a tube into the left external acoustic meatus (Sokolich, U.S. Patent 4,251,686, 1981). The sound-delivery system was calibrated with a sound level meter (Brüel and Kjær 2209). The frequency response of the system was flat within 6 dB 
14 kHz, which encompassed frequency response areas (FRAs) of the majority of neurons studied.
Parylene-coated tungsten microelectrodes with impedance 1–2 M
were used for extracellular multiunit recordings. In some cases electrode tips were plated first with gold and then platinum black to reduce electrical noise; electrodes typically had impedances <100 k after plating. Electrodes were held in a micromanipulator and advanced using a hydraulic microdrive (500–1150 µm below the cortical surface) until bursts of spike activity that were phase-locked to the onset of repetitive tone or noise search stimuli were observed above the background noise. Electrode signals were amplified and band-pass filtered (0.3–3 kHz), and spike activity was discriminated from background noise using a window discriminator. Spike times after stimulus onset were recorded digitally with an accuracy of 0.1 ms.
To record at a given site, the electrode was initially inserted to the depth at which responses of neighboring sites had been recorded. If a response to tone bursts and/or broadband noise bursts at this depth was not observed, the electrode depth was increased and/or decreased to search for a response. If no response was found at any depth, the site was marked nonresponsive and used to determine the boundaries of the AI, especially in the younger animals. The approximate minimum stimulus intensity necessary to elicit a response (threshold) and the frequency of that stimulus [characteristic frequency (CF)] for each location were identified audiovisually. At each location, responses were recorded to 675 computer-generated tone bursts (50 ms duration, 3 ms rise/fall) bracketing the approximate CF and threshold. The tone bursts varied in frequency across 3–5 octaves in 45 logarithmically spaced increments, and varied in intensity across 70 dB in steps of 5 dB. Responses to the tone bursts, each presented once, were used to construct FRAs (Fig. 1). In cases where response latencies approached the duration of the tone stimuli, the tone duration was increased to verify that the responses were related to stimulus onset rather than offset.
|
). The 3- to 15-kHz frequency range was chosen for 2 reasons. First, visual identification of the location of each electrode penetration is required for the mapping procedure, and AI sites with CFs above 3 kHz are usually found on the surface of the middle ectosylvian gyrus. Second, the frequency response of the sound-delivery system began to drop off near 15 kHz. In some cases mapping was terminated before covering this region due to changes in the animal's physiological condition. All surgical and recording procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at UCSF. Analysis
The FRA recorded at each recording site was analyzed to determine CF, threshold, spike-count versus level (rate/level) function, and bandwidths and low-frequency and high-frequency band edges (20 and 40 dB above threshold, and expressed in octaves relative to CF). An additional measure of frequency selectivity, Q, is defined as the ratio of CF to linear bandwidth. High Q values correspond to high frequency-selectivity (narrow bandwidth), and a low Q value corresponds to a low frequency-selectivity (broad bandwidth).
The rate/level function at each site was constructed by averaging responses to 3 frequencies centered on the CF (spanning <1/3 octaves) at stimulus levels ranging over 70 dB (Fig. 2). The averaged rate/level functions were then fit using a least-squares-error algorithm to a 3-segment piecewise-linear continuous function. The first segment was constrained to have zero slope, and corresponded to subthreshold spontaneous activity. Slopes of the second and third segments were not constrained. The third segment, corresponding to response to the loudest sounds, was required to extend through a 
10-dB range of intensities for the recording site to be included in statistical analysis. The fit functions were normalized by the response rate at the intersection of the second and third line segments (the "transition point"). The degree of monotonicity was identified as the slope of the third segment. A response was designated as nonmonotonic if the third segment had a slope < -0.5%/dB. Sites with third segments that did not extend through a 10-dB range but whose second segments extended through 30dB were designated as monotonic.
|
Cortical maps
To compare spatial data from different animals, the CF gradient was aligned with the x-axis by fitting log (CF) to a plane using a linear least-squared-difference algorithm. The coordinate system was rotated to align the x-axis to the maximum slope of this plane. As a result, isofrequency contours were parallel to the y-axis and the x-axis was parallel to the posterior-anterior cortical axis. The origin of the x-axis was assigned to lie at the 4.0-kHz position of the plane.
Maps were created by constructing Voronoi polygons around the recording sites without smoothing or distortion of the original value distributions. Polygon boundaries in a Voronoi diagram are formed by the perpendicular bisectors of line segments connecting neighboring recording sites. Consequently, for each point in the interior of a Voronoi polygon, the closest recording site is at the center of the polygon. The response parameter value from a given recording site was assigned to the entire area of the corresponding Voronoi polygon and assigned a color code. Polygons corresponding to nonresponsive sites were left blank.
To identify spatial clustering of rate/level function shapes, the monotonicity classification at each recording site was compared with each of its nearest neighbors. If monotonic (M) or nonmonotonic (N) sites are clustered, then neighbors should have similar rate/level functions. A Monte Carlo method was used to determine whether the proportion of similar neighboring rate/level functions was statistically significant. The classification values were shuffled and randomly distributed over the recording sites, and the matching pairs counted. The shuffle/count procedure was repeated 10,000 times to derive a baseline distribution of the number of matching pairs. The true number of matching pairs was considered to be statistically significant if 500 or fewer (P < 0.05) of the shuffled counts had a larger number of matching pairs.
Determination of ventral and dorsal boundaries of the AI
The ventral transition from the AI to the second auditory cortical area AII is not sharp (Rose 1949; Schreiner and Cynader 1984). A distinguishing physiological feature of the AI is its smooth gradient of tonotopy—a feature that is not shared by the AII. For 12 of 15 animals, recordings covered the area from locations near this ventral breakdown of the tonotopic gradient dorsally to the SSS. In some animals, tonotopy also broke down dorsally. Nontonotopic recording sites near the SSS were possibly in the dorsoposterior area (DP; Reale and Imig 1980).
To ensure that recording sites were within the boundaries of AI, an iterative selection process was used. CF values between 3.0 and 15.0 kHz were first plotted against x-axis location. A linear regression was performed for CF against the x-axis location. Sample points whose CFs lay more than 2 SDs from this regression line were rejected as outliers and the regression was repeated. This process of regression/rejection converged after a few (11) iterations in every case, thus identifying the strictly tonotopically organized portion of AI. [In this narrow frequency range, the linear regression of CF against X-location retained a larger number of points than a similar regression of log(CF) against X-location.] This process accommodated any local curvature of the isofrequency contours. The rostral AI boundary with the anterior auditory field (AAF) normally is identified by a reversal in the CF gradient. For some animals in this study mapping did not extend to include the gradient reversal; however, the limit of mapping was rostral to the region of cortex with 13-kHz CFs for every animal. Thirteen sites in one animal (k-128) were rejected as unlikely to be in the AI before applying the iterative procedure based on their anterior location and large deviation of CF from the posterior-anterior CF gradient.
Age grouping and CF selection
To determine age-related changes in measured parameters, data from animals of similar ages were pooled into 4 approximately logarithmically spaced age groups (G1: <P17; G2: P17–P30; G3: P31–P70; G4, >P70). The specific ages of animals in each group are presented in Table 1. In the youngest animals, it is possible that the time since conception is a better determinant of auditory system development than postnatal age. The age since conception for each animal is provided in Table 1 (queens were generally mated on 2 or more successive days, resulting in a range of age for most subjects).
|
; Evans and Whitfield 1964). Potential bias attributed to correlation of CF with any parameter was avoided by choosing subsets of recording sites with equivalent CF distributions for statistical comparison between age groups (Kolmogorov–Smirnov two-sample goodness-of-fit test). These subsets (Fig. 3) included all AI sites with CFs in the octave centered at 7.5 kHz (5.3–10.6 kHz). Nonparametric measures were used for all statistical comparisons. The Wilcoxon two-sample rank-sum (RS-test) was used to identify differences in distributions of spectral parameters, and the Fisher exact test (FE-test) was used to identify differences in distribution medians of spectral parameters.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Extracellular multiunit recordings were made at 1,086 locations from the surface of the right middle ectosylvan gyrus of 15 animals ranging in age from P14 through P110. Data presented here are from 855 sites that fulfilled the criteria (see METHODS) for inclusion in the study. CFs ranged from 3 to 15 kHz, and maps constructed from these recordings contained between 19 and 195 points. Recording depths and character of responses were consistent with those of cortical layer(s) IIIb/IV. A frequency gradient with low-CF posterior and high-CF anterior was present in all animals; the magnitude of this gradient decreased as a function of age, so that the cortical region encompassing CFs between 3 and 15 kHz increased in size along its posterior-anterior dimension. In 12 animals the responsive area of the AI was evenly sampled from its most dorsal border ventrally through a region with narrow bandwidths. The median bandwidth of samples pooled over the entire ventral-dorsal extent of the AI increased as a function of age. In most (12 of 15) animals a centrally located cluster of narrowly tuned sites was apparent. Based on the ventral-dorsal location of this cluster, the AI was segregated into 3 partitions and responses in each partition were compared. Age-dependent changes of bandwidth in the 3 partitions were different. The proportion of recorded sites that had nonmonotonic response/intensity functions decreased as a function of age.
In general, responses in the youngest animals were less robust (smaller in amplitude and less reliable) than responses in older animals and often required substantial effort to identify and isolate from background activity. Action potentials in younger animals were substantially smaller compared with background than in adults, and response latencies were substantially longer (to be reported). In 2 of the youngest animals responses were recorded only from a restricted central region along the ventral-dorsal axis, and all other sites were unresponsive. It is possible that some percentage of the responses recorded using the extracellular electrodes arose from afferent MGB fibers, rather than from cortical layer IIIb/IV cells, and that the proportion of recorded responses of this origin was larger in younger animals. The median response magnitude increased monotonically from group G1 through group G4 (G1, 15.2; G2, 16.4; G3, 18.8; and G4, 22.9). However, pairwise comparisons (rank sum) between successive age groups indicated no significant difference in the response magnitude between groups G1 and G2 or between groups G2 and G3. The distribution of the response magnitude differed significantly (P < 0.01) between groups G3 and G4. Mean spontaneous rate for these multiunit recordings was essentially independent of age (Fig. 4).
|
A correlation of optimally responsive recording depth with age was found for animals 24 days of age or less. The equation of the regression line was 510 µm + 14.7 µm/day (r2 = 0.31, F = 162.7, P << 0.001). For animals older than 24 days, there was no significant correlation of recording depth with age. It is likely that the correlation of depth with age in the younger animals reflects a smaller amount of cortical tissue superficial to layer IIIb/IV neurons. Differentiation of superficial cortical layers is incomplete until 3 to 4 wk after birth, at least in visual cortex (Shatz and Luskin 1986), which may lag development of the auditory cortex by 3 to 4 days (Payne 1992).
Frequency gradient
The cortical frequency gradient was determined from linear regression of CF versus x-axis (Fig. 5) after rejection of outlying points (see METHODS). Examination of CF maps (Fig. 6) showed that most (148–161 of 231) sites rejected by the fitting procedure were on the periphery of the target recording area, dorsally including sites probably located in DP and ventrally including sites in the AII. The slopes of the CF regression lines decreased as a function of age reflecting an overall increase in size of the cortex consistent with increase in the overall size of the brain during this period (Larsen 1984). The initial value of the cortical frequency gradient was 3.5 kHz/mm at P14 and 2.5 kHz/mm at P100 (Fig. 7).
|
|
|
GLOBAL DISTRIBUTIONS. In general, FRAs were V-shaped at all tested ages. Representative FRA examples from each age group are shown in Fig. 8. Bandwidth at 20 and 40 dB above minimum threshold (Fig. 9) increased with age. Median bandwidth at G2 was significantly larger than that at G1. Median bandwidth at G3 was larger than that at G2, but the difference was only significant 40 dB above threshold. There was no difference in bandwidth between G3 and G4.
|
|
These pooled data suggest that spectral properties of AI neurons change during the first 2 postnatal mo (before the maximum age of group G3), but that no further changes occur over the next 2 mo. However, when analyzed regionally, clear differences are seen during this later developmental period.
REGIONAL DISTRIBUTIONS. The pooled analysis of AI described above presupposes a homogeneous population of spectral response properties throughout AI. However, studies of the isofrequency domain in adult AI have identified spatially localized regions having dissimilar distributions of bandwidth and Q (= CF/bandwidth) (Heil et al. 1992; Schreiner and Mendelson 1990; Schreiner and Sutter 1992) with one consistent central region of high-Q that is flanked by regions of lower-Q.
To determine whether there are developmental differences in these regions, AI was subdivided into 3 partitions based on measurements of Q and bandwidth measured 30 dB above minimum threshold. In groups G2, G3, and G4, a central region of high-Q was clearly evident. In some cases, a small number of more dorsally located recording sites had higher Q than that of their near neighbors when measured at 20 and/or 40 dB above the minimum threshold. This observation is consistent with previous studies reporting a second region of high-Q found dorsally to the central high-Q region in some adult cats (Heil et al. 1992; Read et al. 2001; Schreiner and Mendelson 1990). However, in animals younger than P24, it was not possible to unambiguously identify more than one region of high-Q. Therefore for all age groups AI was subdivided into 3 partitions based solely on an estimate of the location of the more apparent, ventrally located high-Q region. The identifi-cation of this region was based on 3 criteria: 1) inclusion of the predominant peak in a smoothed function describing Q30 versus ventral-dorsal position, 2) inclusion of a maximal number of high-Q recording sites, and 3) inclusion of a minimal number of low-Q recording sites. The 2-mm wide anterior-posterior swath that most effectively satisfied these criteria was chosen as central-AI. Recording sites dorsal to this swath were labeled as dorsal-AI, and those sites ventral to the swath were labeled as ventral-AI (Fig. 10). The CFs between these partitions were not significantly different (Wilcoxon rank-sum, P > 0.05). To quantify the effectiveness of this method in partitioning AI, distributions of bandwidth in the central partition for each age group were compared with the ventral and dorsal partitions (Fig. 9). In each case, the median bandwidth was smaller and the distribution of bandwidth values was narrower in the central partition than in either the ventral or dorsal partitions though there was some overlap in these distributions. For age groups G2, G3, and G4, the median bandwidth in the central partition was significantly different from the distribution in both the ventral and dorsal partitions (RS-test, P < 0.05). However, the division of AI into 3 partitions was unsuccessful for the youngest group, G1 (RS-test, not significant) and all measurements from the youngest group were pooled for further analysis and considered to be from a single population. Thus during the postnatal period, spatially and functionally differentiated responses emerge from what is initially a relatively homogeneous response population.
|
Comparison of all parameters in the youngest age group with those in the ventral partition in G2 indicated that frequency tuning was significantly narrower in the pooled G1 group. Median bandwidth was larger in the G2 recordings, low-frequency band edge was lower, and high-frequency band edge was higher.
In the central partition from G2 through G4, changes in frequency tuning were restricted to intensities 40 dB above minimum threshold. There was no change in distributions of bandwidth, low-frequency band edge, or high-frequency band edge measured 20 dB above minimum threshold, but there were significant changes in distributions of 2 parameters measured at the higher stimulus intensity. The distribution of bandwidth measured 40 dB above threshold differed in groups G2 and G3 accompanied by a significant difference in the distribution of the high-frequency band edge. There were no later changes of either bandwidth or high-frequency band edge. The moderate change in bandwidth with age in this partition is predominately a result of changing high-frequency band edge.
Comparison of the spectral response parameters from the central partition of AI with the pooled distribution from the youngest age group G1 identified initial differences in the high-frequency band edge. Measurements 20 and 40 dB above minimum threshold indicated that the median high-frequency band edge was higher in the central partition of the G2 group than in the pooled population of sites in the G1 group. However, the increase in high-frequency band edge was not suffi-ciently large that it created a significant change in bandwidth in either intensity range.
In the dorsal partition from G2 through G4, changes in tuning were also restricted to changes in the high-frequency band edge. There was no significant change in bandwidth 20 dB above threshold. At the higher stimulus intensity, the distribution of bandwidths slowly changed, becoming significantly different only in the comparison of G2 with G4. However, the median bandwidths of these 2 groups were not significantly different. The distribution of low-frequency band edges did not show a significant change for either range of stimulus intensities. The distribution of high-frequency band edges changed from G2 to G4 in both stimulus intensity ranges accompanied by significant increases in the median high-frequency band edges.
Comparison of the distribution of response parameters of the G1 pool with those of the dorsal partition indicated significant differences in all parameters in both stimulus intensity ranges. The median values were consistent with broader tuning in the dorsal partition than in the pool of G1 responses. Bandwidth was larger, low-frequency band edges were lower, and high-frequency band edges were higher in the dorsal partition of each older group than in the pooled G1 group.
Monotonicity
As mentioned earlier, the slope of the third segment of the rate/level function was examined as a measure of monotonicity. The mean slope of this third segment increased with age, although in all age groups most responses were nonmonotonic. In the youngest (G1) group, 86% of cortical sites were nonmonotonic, whereas in the oldest (G4) group, only 61% of the sites were nonmonotonic. An R x C test of association using the G-test indicates that the percentage of nonmonotonic responses was not independent of the age group (G = 15.4 with Williams adjustment, P < 0.01). After restricting the sample to include only those sites for which the slope of the third segment could be identified (see METHODS), 90% of the sites in the youngest (G1) group were nonmonotonic with a median nonmonotonicity of -1.8%/dB, and 64% of the recording sites in the oldest (G4) group were nonmonotonic with a median nonmonotonicity of -0.5%/dB. The distribution of the monotonicity index for each age group is shown in Fig. 11.
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
, 1992; Rubel et al. 1998), and the development of cortical function necessarily reflects the changes that are occurring at subcortical as well as cortical levels. Consequently, interpretation of the results presented here is aided by consideration of anatomical and physiological evidence from maturing cortical and subcortical structures. Interconnections between the auditory cortex and subcortical auditory structures (Winer 1992) are already present at birth (Cant 1998), but the majority of these pathways, as well as local and remote cortico-cortical connections, have not reached their mature form. In fact, cellular differentiation in the cortical subplate and the resultant formation of cortical layers have not yet been completed by that time (Payne 1992). In the first 3 postnatal mo that are the focus of the current study, there is a dramatic transient increase of synapses after the completion of neuron migration (at wk 3), which peaks after several weeks and then declines to levels found in the mature brain by 10–15 wk after birth (reported for visual cortex: Cragg 1975; Winfield 1983), as well as a large decrease in the total number of trans-callosal fibers (Berbel and Innocenti 1988). Additionally, interneuronal communication pathways, including neurotransmitter (e.g., GABA: Gao et al. 1999; Owens and Kriegstein 2002) and neuromodulator (e.g., serotonin: Vu and Tork 1992) pathways, change through the first 2 postnatal mo. The current report focuses on 3 aspects of functional organization that have been shown to have spatial gradients or clustering in adult animals: frequency, receptive field bandwidth, and monotonicity of rate/level functions.
Different aspects of cortical architecture and cortico-cortical interconnection likely mature through an extended postnatal period, but development has been studied for only a few of these. For example, previous studies have identified that development of the binaural processing columns found in the cat AI extends through the 3rd postnatal mo (Brugge 1985; Feng and Brugge 1983; Imig and Adrian 1977; Imig and Brugge 1978; Moore 1985; Payne 1992).
In this series of experiments, short tone bursts were used to investigate age-dependent changes in frequency response areas, identified by changes in frequency tuning and in rate/level functions, of the primary auditory cortex. The main findings of the study can be summarized as follows. 1) Responses of AI sites are tonotopically arranged by postnatal day 14, the earliest age included in the study. 2) The cortical frequency gradient is steeper at postnatal day 14 than at 3.5 mo of age. 3) The majority of sites in the AI from postnatal day 14 through 3.5 mo exhibit nonmonotonic rate/level functions, although the proportion with nonmonotonic rate/level functions decreases with age. 4) For a fixed CF range, the median response bandwidth in the AI increases with age. 5) Large increases in response bandwidth occur in ventral and dorsal regions of the AI, but response bandwidths remain narrow in what is to become the central narrowly tuned region found in the adult AI.
Tonotopic gradient
The present observations of tonotopy in the developing AI are in general agreement with results of previous studies in adult and developing animals and confirm the tonotopic arrangement of the AI at P14. Tonotopy of multiunit recordings has been reported in the adult cat in numerous studies (e.g., Merzenich et al. 1975; Reale and Imig 1980; Schreiner and Mendelson 1990). In developing animals, the cortical frequency gradient of local field potentials has been used to identify the AI in kittens as young as P10 (Eggermont et al. 1993), in spite of the fact that auditory nerve thresholds are still quite elevated (Walsh and McGee 1987). However thresholds at all levels in the auditory system drop dramatically at about this age (Brugge et al. 1988).
The cortical gradient of frequency was found to rapidly decrease with age from an initial value of 3.5 kHz/mm at P14 to 2.5 kHz/mm in the 3rd postnatal wk. This gradient decrease could be explained by a number of mechanisms, such as change in cochlear place-code during development (Rubsamen and Lippe 1998) or reorganization of the place-code at one or more subcortical stations (Romand 1997). However, the most parsimonious explanation is simply that the cortex increases in volume during development. From P10 through P50, the brain of the cat doubles in weight, and it increases by an additional 50% by adulthood (Larsen 1984). Corresponding increases in linear dimensions are consistent with the decrease in the tonotopic gradient observed in this study. Previous estimates of the adult cortical frequency gradient range from 3 to 7 kHz/mm (Eggermont and Komiya 2000; Harrison et al. 1991; Merzenich et al. 1975; Rajan et al. 1993; Reale and Imig 1980; Stanton and Harrison 1996). The cortical frequency gradients reported for the oldest animals in the present study are on the low end of this distribution.
Frequency tuning
Changes in tuning bandwidth of AI neurons as a function of age are not uniform over the ventral to dorsal extent of the AI. Eggermont (1996) reported an increase during development of bandwidth measured 20 dB above threshold for AI neurons with an associated time constant of 25 days. That time course for neurons pooled from unspecified locations in the AI is consistent with the changes in the present sample of multiunit recordings pooled from all of the AI, whose bandwidth distribution did not change significantly after the G3 period (P35–P58). Eggermont's study noted that the observed change in bandwidth was predominantly attributed to broader tuning in a subset of about 30% of the recorded neurons. The current results suggest that broadening of tuning does not occur uniformly across the ventral to dorsal extent of the AI. Instead, during maturation, AI develops from an initial state, of homogeneously narrow tuning, to an adult state characterized by segregated regions of narrow or broad tuning (Heil et al. 1992; Schreiner and Mendelson 1990; Schreiner and Sutter 1992).
The developmental progression of cortical FRA bandwidth organization may relate to 1) changes in projection from the MGB and 2) changes in inhibitory shaping. Studies to correlate physiologically defined clusters of broader and narrower bandwidths with anatomically distinct regions of the AI have only recently begun (Read et al. 2001). Cortical tuning does not simply reflect tuning at precortical levels, although the range of bandwidths in the source of thalamic afferents does cover nearly the range of bandwidths recorded in the adult AI (Calford et al. 1983). The majority of thalamic input to the AI arises from the ventral division of the medial geniculate body (MGBv). The MGBv has been reported to contain a gradient of bandwidths in which caudally located neurons tend to be more broadly tuned than rostrally located neurons (Rodrigues-Dagaeff et al. 1989). Projections from the MGBv to the AI do appear to be topographically arrayed in the dimension of bandwidth (as well as the dimension of frequency). That is, ventral injections into the AI tend to label cells caudally positioned within the MGBv and dorsal injections tend to label rostrally positioned cells (Brandner and Redies 1990). However, if cortical tuning were merely a result of topographic projection from the MGBv, one would expect that tuning in the dorsal region of the AI would tend to be sharper than tuning in the ventral region, which is not the case. On the other hand, both broadly and narrowly tuned responses are found in all regions of the MGBv regardless of their rostral-caudal position (Brandner and Redies 1990). More recently it has been shown that thalamic afferents to narrowly tuned and broadly tuned cortical neurons may have tuning that is broad or narrow (Miller 2001). Consequently, the clustered topography of bandwidth in the adult AI appears not to be merely a reflection of clustering of thalamic afferents with different bandwidths. Developmental changes in the cortical bandwidth organization may thus reflect a progressive cortical pruning or selection mechanism of thalamic inputs. These changes may well be governed by cortical plasticity mechanisms that are influenced by the statistics of the acoustic environment of the animal. The increase in more broadly tuned neurons would suggest an increased influence of the presence or behavioral significance of broadband sound structure (Zhang et al. 2002, 2001).
A role of local inhibitory properties in defining spectral integration and bandwidth characteristics is highly likely. Based on anatomical observations, Prieto and colleagues (1994) have suggested that nonuniform concentrations of GABA-ergic neurons in the AI may be responsible in part for differences in frequency tuning. They found within cortical layers III and IV a moderate but significantly higher density of GABA-ergic cells in a centrally located region compared with that in ventrally located and dorsally located regions. Additional support for inhibitory effects on AI tuning comes from a study of iontophoretic application of the GABAA antagonist bicuculline (BIC) (Wang et al. 2000) that reported an increase in bandwidth in 28 of 36 AI neurons in chinchilla.
The increase in the bandwidth of cortical neurons during the postnatal period is particularly interesting, as bandwidths in the central nucleus of the IC, the main collicular relay in the pathway to the AI, are narrower in adults than in animals <4 wk of age (Moore and Irvine 1979). This suggests that the anatomical locus of frequency tuning change is either in the thalamus or the cortex. Changes in tuning-curve properties of thalamic neurons during development have not been studied.
The cause of developmental tuning changes may be related to developmental changes in the efficacy of GABA-ergic synapses and the cortical distribution of GABA-ergic neurons. In adult cortical neurons, a GABAA-receptor–mediated increase in membrane conductance normally causes hyperpolarization of the membrane potential, and thereby suppresses generation of action potentials. However, in very young rodents (<3 wk), a maturational difference in the intracellular chloride ion concentration causes similar GABAA-mediated conductance increases to depolarize the membrane potential. This depolarization may result in evoking of membrane action potentials (Owens and Kriegstein 2002). The earliest changes in frequency tuning reported here may in part be related to this global change in synaptic efficacy through a mechanism of disinhibition acting in the thalamus or the cortex. Both early and later changes in tuning may in part be related to differences in the cortical distribution of GABA-ergic neurons. In ferret auditory cortex the density of GABA-ergic neurons was initially high at birth and declined around P20 (corresponding approximately to P0 in the cat). The proportion of GABA-ergic neurons exhibited a transitory peak at around P60 (P40 in cats) and then decreased (Gao et al. 1999). This global time course of inhibitory potency matches the changes seen in this study. The transitory developmental peak in the proportion of GABA-ergic neurons may coincide with the transient increase in bandwidth in the ventral partition of the AI observed in the current study. This interpretation would also suggest a spatially nonuniform decline of GABA-ergic influences during development.
The current results suggest that during development, different regions of the AI develop differently. In the central region, bandwidths change only modestly during the first 3 mo of life. In contrast, there is a dramatic change in the distribution of bandwidths in areas ventral and dorsal to this region. At the present time, it is not possible to conclusively determine whether the responsive neurons in younger animals constitute the majority of responsive neurons in older animals; the same neurons may grow to respond to a broader range of stimuli or, alternatively, the initially responsive neurons may have been supplemented (or perhaps even replaced) by a new population of neurons with broader tuning. Despite the possibility that a multiunit recording that included a narrowly tuned neuron together with a broadly tuned neuron would likely appear broadly tuned, the results present compelling evidence that sites with broad bandwidths are underrepresented in the juvenile cortex when compared with the adult cortex.
The most significant tuning changes occurred during the G1–G2 period, although distributions of bandwidth and band edges continued to change through the G3–G4 transition. Moreover, the observed changes in frequency tuning were not uniform throughout the AI. A schematic diagram illustrating the observed changes in each region is shown in Fig. 13. Broadly tuned sites occur infrequently in the young cortex. Coordinated activity of thalamic afferents over a broad range of CFs, such as that evoked by a broadband acoustic stimulus, may be absent while thresholds are high. Subsequently, ventral and dorsal regions with broadly tuned bandwidths appear coincidentally as acoustic thresholds decline to adult levels. This contrasts with the cortical frequency gradient, which appears to be present before acoustically generated activity, and suggests that some intrinsic developmental factor is responsible for tonotopic organization. The delay in development of ventral and dorsal broadly tuned sites might suggest that acoustic experience is necessary for development of cortical circuitry appropriate for integrating sounds over a broad spectrum.
|
). Moreover, two observations suggest that the decreases in OFF-CF thresholds occur independently of this global decrease. First, OFF-CF thresholds decrease over a longer time than the global decrease. Second, OFF-CF thresholds decrease at site-specific locations. They decrease in the dorsal and ventral regions, but not in the central region of the AI. For these reasons, we have interpreted our experimental observations as changes in bandwidth rather than as changes in OFF-CF thresholds. Rate/level functions
A majority of multiunit sites in the juvenile AI exhibit nonmonotonic rate/level functions, and the proportion of nonmonotonic responses decreases with age. Earlier studies reporting a proportion of nonmonotonic responses in the AI used two basic types of methods for identification of monotonicity. The first type compares the response rate 10–20 dB above threshold or highest response rate with the response rate at a high stimulus intensity (e.g., Phillips et al. 1994; Sutter and Schreiner 1995). The second method identifies monotonicity based on the slope of the rate/level function at intensities above the "transition point" (Schreiner et al. 1992). The latter method was most appropriate for the current study because presenting stimuli at intensities substantially above the high thresholds of the youngest animals would have exceeded the maximum output of the speakers. Schreiner and colleagues (1992) reported that 60.5% of the multiunit responses they recorded were nonmonotonic—a proportion similar to that identified here for the older age groups. Proportions of nonmonotonic units reported in other studies of single-unit (e.g., Calford and Semple 1995; Clarey et al. 1994; Phillips et al. 1994; Sutter and Schreiner 1995; Wang et al. 2000) and multiunit (e.g., Clarey et al. 1994; Heil et al. 1994; Schreiner et al. 1992; Sutter and Schreiner 1995) responses ranged broadly from about one third to two thirds. Eggermont (1996) reported that about half of the recorded single-unit responses in the AI of juvenile cats (P9–P300) were nonmonotonic, without indication that the proportion of nonmonotonic responses was related to age. Differences between that study and the current one may be attributable to differences in the method used to identify nonmonotonic responses or may relate to differences inherent in single-unit and multiunit recording techniques. In the adult cat, monotonic and nonmonotonic response types are known to be clustered (Clarey et al. 1994; Imig et al. 1990; Phillips et al. 1985, 1994; Schreiner et al. 1992). However, consistent clustering of monotonic and nonmonotonic sites was not apparent for most animals in the current study.
Because rate/level functions of auditory nerve fibers are monotonic, the presence of nonmonotonic rate/level functions in the cortex indicates that there are cortical or subcortical inhibitory factors influencing cortical responses (Phillips et al. 1985). One possible cortical source of inhibition is the population of GABA-ergic neurons that are distributed throughout the AI and are implicated in the determination of frequency tuning. The proportion of GABA-ergic neurons declines during postnatal development (Gao et al. 1999); however, it is unlikely that this decrease alone is the source of the observed decrease in the proportion of nonmonotonic responses, given that iontophoretic application of the GABAA antagonist BIC is usually ineffective in converting nonmonotonic cortical responses to monotonic responses (Wang et al. 2000). Thus it appears that other mechanisms, such as precortical inhibition, must play a significant role in determining the rate/level properties of cortical neurons.
Methodological considerations
Three aspects of the method used to collect data in the current experiment deserve brief mention. These are the stimulus calibration, the effects of anesthesia on the cortical response properties, and the use of the extracellular multiunit recording technique. The frequency response of the acoustic system was flat ex situ. It is possible that age-related differences in the dimensions of the ear canal might produce differences in the acoustic load on the speaker, although any such differences did not affect measurements of frequency response within the frequency range reported here. Factors that would have affected frequency response measurements constitute resonant peaks or notches in acoustic transmission and a resultant peak or notch in stimulus intensity. An observable consequence of such a resonance would be a region of high (or low) threshold at that resonant frequency in all FRAs measured from an individual animal (e.g., FRAs with CFs at that frequency would have high minimum thresholds, and FRAs with other CFs would have a "notch" or "shoulder" at that frequency). Rather than this, we observed age-related differences in frequency tuning that were independent of CF: e.g., narrowly tuned FRAs were observed for all CFs in the youngest age group. There were no apparent effects of age-specific differences in the acoustic load that would affect the findings reported in this manuscript.
The groups of juvenile animals studied here represented stages of development over a period when all physiological systems are changing rapidly. It is possible that there are differential effects of anesthesia in young and adult animals that affect AI responses, but are unrelated to developmental functional changes. Effects of barbiturate anesthesia on auditory cortical responses in adult animals were previously described by other investigators (Gaese and Ostwald 2001; Goldstein and Abeles 1975; Goldstein et al. 1959; Phillips et al. 1985; Volkov et al. 1985; Zurita et al. 1994).
The frequency responses of the multiunit recordings reported here might differ from their constituent single-unit frequency responses (Schreiner and Mendelson 1990; South and Weinberger 1995). In the adult cat, a gradient of broad frequency tuning of individual neurons in the dorsal region of the AI coincides with a similar gradient of broad frequency tuning of neuron clusters. In contrast, there is only a weak gradient of frequency tuning of individual neurons in the ventral region of the AI, even though neuron clusters do exhibit a strong gradient. The reduced gradient for single units in the ventral portion of the AI results from increasing local scatter of CFs resulting in broad multiunit frequency tuning (Schreiner and Sutter 1992). The rate/level functions of the reported multiunit responses may also differ from rate/level functions of their constituent single units.
A high degree of nonmonotonicity in a multiunit response would be most likely to arise from similarly high nonmonotonicities of its single-unit constituents (Heil et al. 1994). It should be noted, however, that a single strongly responding nonmonotonic constituent might be sufficient to mask other weakly responding monotonic constituent single units in a multiunit response. It is possible that the constitutive population of recorded multiunit clusters changes during development, with new subpopulations beginning to respond at different times, or with other subpopulations that initially responded dropping out. Alternatively, the response properties of individual single units within multiunit clusters may change during development, either in the range of stimuli to which they respond, or in the manner in which the single units respond. Spatial distributions of single-unit bandwidths and nonmonotonicities are beyond the scope of the current report.
This study has focused on functional changes during the period of intense neural remodeling that occurs immediately after the auditory nervous system has first become responsive to acoustic stimulation. Uniformity of bandwidth and lack of spatial differentiation of responses in the youngest age group suggest that the responses recorded in those animals were from a relatively homogeneous population. Significant differences in response parameter distributions between spatially segregated clusters of neurons in older animals, identified in this and previous studies, suggest either that the initial homogeneous population of responding neurons has become functionally differentiated during the course of development or that the responses that are recorded in older animals are from neurons that are initially unresponsive in the youngest animals.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
, +, and x were classified as monotonic; curves indicated by 
and 
were classified as nonmonotonic.
