| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Auditory Science Laboratory, Department of Otolaryngology and Brain and Behaviour Division, The Hospital for Sick Children, Toronto, Ontario; and Department of Physiology and the Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada
Submitted 1 June 2004; accepted in final form 20 August 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
). Analogous "topographic" maps of receptor/effector space are common to other sensory/motor systems.
The construction of topographic maps is directed, at least initially, by (genetically expressed) protein factors that are matched between axonal growth cones and their intended target locations (Sperry 1963; reviews: Meyer 1998; Rubel and Fritzsch 2002). Early in development, topographic projections appear (relatively) diffuse, i.e., axonal arbours from closely spaced sources may innervate well-separated targets. Synaptic connections that are out of place are subsequently eliminated, and in this, it is spike activity that appears to play the crucial role (e.g., Meyer 1983; Schmidt and Eisele 1985; reviews: Fawcett and O'Leary 1985; Kaas 2000). As first conjectured by Hebb (1949), the spike trains of erroneous inputs will more likely be uncorrelated with respect to the majority of (correct) inputs, and so will lose in the competition for trophic molecules required for synapse consolidation (review: Fritzsch et al. 2004).
An ongoing debate is whether the early developmental refinement of diffuse topographic projections is driven by spontaneous, or externally stimulated, spike activity (or both). We emphasize "early developmental" as it has long been known that sensory deprivation (or augmentation) can lead to the rewiring of established topographic maps (e.g., Harrison et al. 1993; Kaas et al. 1990; Merzenich et al. 1984). Indirect evidence for a possible role of spontaneous activity was furnished by Naegele et al. (1988), who found that the refinement of diffuse retinotopic projections to visual cortex (in the hamster) appeared completed prior to initial eye opening, and by Horton and Hocking (1996), who showed that the maturation of ocular dominance columns (in the monkey) also precedes vision-onset. Direct evidence for the importance of waves of correlated spontaneous activity in the optic nerve (first described by Maffei and Galli-Resta 1990; Meister et al. 1991) for orderly retinotopic map development was recently provided (in the knockout mouse) by Grubb et al. (2003) and McLaughlin et al. (2003). Rhythmic spontaneous activity has also been observed in the developing auditory system (Lippe 1994, 1995), but its role in tonotopic map refinement has not yet been established.
In the auditory system, even the developmental refinement of initially diffuse tonotopic maps has only been shown within the last few years (Leake et al. 2002; Zhang et al. 2001). In primary auditory cortex of the rat, it is completed well after the onset of hearing at P12 (12th postnatal day). During the 2- to 3-wk refinement period, the (sound) frequency tuning of rat cortical neurons also becomes much narrower (Zhang et al. 2001), although it may yet broaden in later development; this broadening has been shown in the cat (Bonham et al. 2004; Eggermont 1996). It has also recently been shown that both tonotopic map refinement and the sharpening of frequency tuning can be blocked by rearing the rat in continuous, moderate-level broadband noise (Chang and Merzenich 2003). Still, it remains unclear whether the noise only masks the external acoustic environment, or perhaps more importantly, whether it also disrupts the (purported) rhythmic spontaneous activity bursts.
In this paper, we report that well-ordered tonotopic maps and sharply tuned neurons are found in auditory cortex of the near-newborn (P3) chinchilla (laniger). Thus their refinement would seem to be completed in utero, where both external and maternal sounds reach the cochlea by bone conduction (Sohmer et al. 2001) and are considerably attenuated in the process (Gerhardt et al. 1992; review: Sohmer and Freeman 2001). On the other hand, the complexity of (single-unit) receptive fields continues to increase postnatally, and this may reflect the emergence of selectivity for fragments of species-specific vocalizations (including human speech).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Harrison et al. 1996b). We recorded tone-evoked spike activity of auditory cortical neurons from normal-hearing, ketamine-anesthetized chinchillas in four age groups: P3, P15, P30, and adult. Standard extracellular one-electrode methods were used. Spike sorting was followed by the calculation of (single-unit) driven firing rates as functions of tone frequency, level, and time after stimulus-onset. Parameters such as best frequencies (BFs) and tuning widths were determined, and receptive field complexity and tonotopic map disorder were quantified. Animal preparation
All procedures were approved and carried out within the regulations of the local animal care committee. Animals were given atropine sulfate (0.04 mg/kg body mass), primarily to reduce bronchial secretion. They were anesthetized with ketamine hydrochloride (15 mg/kg) and xylazine hydrochloride (2.5 mg/kg). One-half doses of ketamine and xylazine were administered every 1–2 h, as required to maintain sedation for the duration of data collection, after which the animals were killed. Surgery consisted of tracheotomy and intubation, followed by craniotomy over the left inferior temporal cortex; the dura mater was removed for a clear view of the vasculature.
Stimulus delivery and calibration
Acoustic stimuli were produced by Intelligent Hearing Systems' (Miami, FL) high-frequency transducer and conducted through a short tube to the right (contralateral) ear canal, which was sealed with a soft rubber-tip probe. The stimulus-generation systems had been calibrated in the course of pilot recording sessions, as follows. In each age group, an intact outer ear was excised from a fresh cadaver; the eardrum was removed, and a one-quarter-inch microphone (7017, AcoPacific, Belmont, CA) fixed in its place. The microphone was connected to a spectrum analyzer (SR760, Stanford Research Systems, Sunnyvale, CA) and calibrated at 1 kHz, 94 dB SPL (AcoPacific 511E SPL Calibrator). The stimulus probe was inserted into the excised ear, and the entire set of tone frequencies and intensities used in the experiments was calibrated.
Auditory brain stem response thresholds
Auditory brain stem responses (ABRs) to noise and tone bursts (0.5–16 kHz) were measured on the right mastoid prior to surgery using Intelligent Hearing Systems' SmartEP instrument, which was calibrated as outlined above. ABR thresholds were determined to 5-dB accuracy and were consistent between animals within an age group (<10 dB from the group mean for noise and all tones). Measurements were repeated after surgery and intermittently throughout the recording session; an experiment was terminated if thresholds rose to >10–15 dB above preoperative levels.
Hardware and software for extracellular recording
Recording was done on a vibration-isolation table in a soundproof booth. A remotely controlled microdrive held a high-impedance (3–4 M
) tungsten-tip microelectrode that was advanced into auditory cortex perpendicular to the surface to depths of 400–1,000 µm. The electrode signal was amplified, band-filtered (0.4–4 kHz), and sent to an oscilloscope and audio monitor; it was digitized at 25 kHz and saved for analysis. Stimulus generation and extracellular potential recording was controlled by Tucker Davis Technologies (Gainesville, FL) hardware (AP2 DSP card and System II components) and software (SigGen32 3.6 and BrainWare 7.3).
Experiment protocol
Stimuli for cortical recording were 50-ms tone pips (including 5 ms ON and OFF ramps) and were presented at the rate of 2/s. The extracellular potential was recorded during the stimulus presentation and the preceding 50 ms (to determine spontaneous activity levels). As the electrode was slowly advanced through the cortex, a complex of 1-, 4-, and 16-kHz tones (each at 70 dB SPL) was presented to search for responding units. The frequency response area of the typically multi-unit activity was roughly determined. A final set of tones at 25 and 50 dB SPL, which spanned the response area in one-quarter-octave steps (9–29 tones/SPL), was presented (1 at a time in random order) for 40 trials. These procedures were repeated in as many different positions in auditory cortex as an animal's hearing status allowed.
Spike sorting
Two voltage triggers were used to detect prospective spikes on-line, at ±3
(SD) relative to the µ = 0 (0 mean) electrode noise. If both triggers were crossed within a span of 1 ms, 40 voltage samples (total duration of 1.6 ms) around the crossings were saved for off-line analysis along with the time of first crossing, tone frequency, and level. At each electrode position, the set of all such prospective spikes was imported into Mathematica ver4.1 for sorting. Also imported were "clean" samples of each "distinct" spike waveform (clean, as opposed to samples containing a superposition of several spikes), which served as templates for sorting. Distinct waveforms could typically be found in separate clusters on a scatter plot of spike peak amplitude versus peak duration. Each template was obtained by aligning a large number of (clean, distinct) waveforms in time, averaging, and excising the trace on either side of the main peak.
Next, all prospective spikes at a given electrode position were compared with the set of templates, as follows in brief. Each template was swept across a spike voltage record, one sampling interval at a time, and the mean square difference (error) was calculated at each step. The lowest difference (reflecting the optimal alignment of spike and template) was normalized by the peak amplitude of the template, since spike waveform variance was observed to increase roughly as its amplitude. If now the normalized root mean square difference of the candidate spike and its best template was within 2 of the template mean, the spike and template were deemed to match. Following a match, the template was subtracted from the spike voltage record, and the procedure was repeated with the remaining templates until no more matches could be made.
While not state-of-the-art (cf. Sahani 1999; review: Lewicki 1998), our variant on the long-standing "template-matching" approach was found to give reliable sorting results (many examples were checked "by eye"), particularly compared with "cluster-cutting" alone. Its most significant shortcoming was an incapacity to deal with superposed waveforms; consequently, many of these went unclassified. Nevertheless, we are confident that well-isolated single-unit responses were obtained in most cases (again, at the expense of missing some spikes) and that the sorting algorithm did not bias the comparison of results between age groups.
Driven firing rate
Raster plots were generated for each sorted single-unit, as shown in Fig. 1A. On occasion, a stimulus trial was contaminated by rapid bursts of nonstimulated activity, possibly an injury discharge due to the microelectrode; such trials were removed from the record by an automated algorithm. For each tone frequency and level, spikes from uncontaminated trials were assigned to 0.1-ms-wide bins (much shorter than the refractory period so that each contained at most 1 spike), and were convolved with a Gaussian probability density function (µ = bin center, = 3 ms) to smooth spike occurrence times. The 40 (or so) trials were averaged, bin by bin, and the result was converted to units of spikes per second (Hz). The average spontaneous rate (obtained from the 50 ms preceding stimulus-onset, and the 1st 5 ms following it—this being less than the minimum signal transit time from ear canal to cortex) was subtracted from each bin after stimulus-onset to give the driven firing rate. Figure 1B shows contour plots of driven firing rate versus tone frequency and time after stimulus-onset (derived from the raster plots of Fig. 1A). The P value, or confidence level that the driven firing rate was in fact significantly higher than the spontaneous rate, was computed for each bin (Student's t-test).
|
At every frequency, driven firing rates were averaged over the 50-ms tone duration (except the 1st 5 ms). A response was considered to be present if the global maximum of the mean driven firing rate versus tone frequency function was significantly greater than the spontaneous rate (throughout this section, significant will denote P < 0.05). Most firing rate versus tone frequency functions also had local maxima; to establish that these constituted distinct peaks in the response area, it was required that the rate at the local max be significantly higher than that at the adjacent local min (closest to the global max) and also be greater than the spontaneous rate (Fig. 2). The tuning width of each distinct peak was defined as the span of the frequency interval (in octave units) in which the firing rate does not differ significantly from that at the peak and remains above the spontaneous rate. The BF of the peak was simply the median of this interval. It should be noted that the more customary definition of tuning width is the interval in which the firing rate is greater than some arbitrary fraction of the maximum (usually 50 or 25%). The approach introduced here is equivalent, but using a significance criterion normalizes for fluctuations in the spontaneous rate.
|
The driven firing rate contour plots were classified by an automated procedure into four categories: simple, complex spectral, complex temporal, and complex spectral and temporal (see Fig. 12 for examples). Complex spectral units have more than one distinct peak in the firing rate versus tone frequency function (following the definition of distinct given in the preceding section). Complex temporal units have firing rates that are not statistically different in at least two adjacent frequencies, but with significantly different onset delays at those frequencies. Complex spectral and temporal units satisfy both the above criteria, whereas simple units satisfy neither. Simple units most closely resemble the responses of primary auditory neurons, which do not (normally) show multi-peaked rate-frequency curves or appreciable latency differences between (similar frequency) tones that evoke similar firing rates. Complex units, in contrast, are a reflection of spatial-temporal integration across separate tonotopic channels, either via interneuron connections within cortical areas or brain stem nuclei or via projections between them.
|
For each animal, BFs were displayed over a photograph of primary (AI) and/or secondary (AII) auditory cortex. Figure 3 is an example of AI from an adult chinchilla, with BFs determined at 50 dB SPL. Dots mark locations from which significant auditory responses were recorded (at 50 dB SPL); x indicates no response. A central core area (white dots) could be differentiated from a surrounding belt (black dots), which is not tonotopically organized. Each number, or pair/triplet in parentheses, gives the BF(s) of a single unit (color coded by octave). In complex spectral units, if one of the distinct peaks in firing rate was significantly greater than the other(s), and/or had a significantly shorter onset latency, it almost always fit better into the prevailing tonotopic arrangement (this was the case when just 1 of these conditions was met; rarely were the 2 in opposition). This was considered the dominant or true BF. In Fig. 3, dominant BFs are denoted by underscores, whereas the remaining peaks go unlisted and are not considered in the computation of map disorder. Only when the distinct peaks could not be distinguished on the basis of firing rates and onset latencies are all BFs listed in parentheses and given equal status in the computation.
|
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The number of animals in each age group and the number of single-units isolated from the core (AI and AII) and belt auditory cortex is given in Table 1. The total is 1,077 neurons from 19 animals (average >50 neurons/animal but rather variable), with 80% of the units from the core area. Here we report results only for the core units. They could generally be reliably differentiated from the surrounding belt based on differences in best frequency (belt lacks significant tonotopic organization), tuning width, and onset latency (significantly sharper and shorter, respectively, in core units; P < 0.001, Student's t-test); (the few) borderline cases were labeled belt.
|
ABR and cortical neuron thresholds
ABR thresholds to tonal stimuli are shown averaged by age group in Fig. 5. In the mid-frequencies (2–4 kHz), differences in ABR sensitivity between groups are small (<5 dB). Relative to the young animals, adult chinchillas are more sensitive at lower frequencies but less sensitive at higher frequencies (5–10 dB in each case). In fact, the P3 group was slightly but significantly more sensitive than the adult in an overall average across all frequencies (P < 0.001, ANOVA).
|
). Unlike these species, the newborn chinchilla appears to have a fully mature cochlea (Harrison et al. 1996a), a middle ear free of loose mesenchyme or other fetal tissue (Hsu et al. 2000), and well-developed relay synapses (see next two sections). The differences in ABR sensitivity observed between chinchilla age groups may simply be due to the growth of the middle ear, progressively shifting its transfer function to lower frequencies (Hsu et al. 2000, 2001). We did not directly measure the response thresholds of cortical neurons (in the interest of recording time). Nevertheless, of the neurons responsive to stimulation at 50 dB SPL, similar fractions in each age group were also responsive at 25 dB SPL (from Table 1, the numbers are as follows: P3, 44%; P15, 46%; P30, 40%; adult, 38%). This would not be the case had there been an improvement in cortical unit sensitivity with age in the chinchilla, as observed in the more altricious species. In fact, these fractions correspond to the trends in ABR thresholds very closely. For example, the fraction of units responsive at 25 dB SPL was lowest in the adults, which had the highest average ABR thresholds. Nevertheless, a higher fraction of low-frequency neurons responded at 25 dB SPL in the adult compared with the P3 group, just as the low-frequency adult ABR threshold was lower. The effects of these relatively small but significant differences in ABR and cortical audiograms on the comparisons of results between age groups will be taken up in DISCUSSION.
Mean cortical driven firing rate
The maturity of the auditory periphery and of relay synapses in the near-newborn chinchilla is shown by the vigorous firing rates observed in P3 cortical neurons. Figure 6 shows mean driven firing rates (averaged over the 50-ms stimulus duration) to best frequency stimuli at 25 (light bars) and 50 dB SPL (dark bars). To reiterate, firing rates were averaged over the entire sample of neurons from core auditory cortex (AI and AII) by age group. At both SPLs, firing rates were somewhat higher at P3 and P15 than at P30 and adulthood (P < 0.01, ANOVA); perhaps this merely reflects the slightly lower average sensitivity of the younger animals. Again, the chinchilla is contrasted with the more altricious species (listed above), in which adult firing rates (at a given SPL) are typically much higher than those in the young (review: Sanes and Walsh 1998).
|
More evidence that relay synapses in the chinchilla auditory pathway are well developed by birth is the observation at P3 that cortical response times to repeated presentations of the BF tone show relatively little variability. Figure 7A gives an example of a P3 single-unit response to 40 presentations of a 50 dB SPL BF tone. The mean first spike response time is 32.1 ms (stimulus-onset is at 50 ms) and the SD is 2.1 ms, representative of the lowest values for found at any age. Figure 7B shows that there was in fact no significant change in the average value for between age groups (P > 0.05, ANOVA).
|
Although we observed no significant age-related differences in the variability of cortical response times, there were differences in mean response times. Figure 8 shows that the mean latency to response-onset at the BF was significantly lower at P30 (at 50 dB SPL only) and adulthood than at P3 and P15 (P < 0.005, ANOVA). Here latency is defined not by the first spike time, but by the time of the first bin in which the driven firing rate becomes significant (with respect to the spontaneous rate) at the 95% level. Given that both the auditory periphery and central relay synapses appear mature in the newborn chinchilla [as evidenced by histology (Harrison et al. 1996a), response thresholds, firing rates, variability in response times], it is likely that the shorter response times in the older groups are due to the growth of axon diameters and myelin sheaths, both of which result in an increased spike conduction speed.
|
A main finding of this work is that the tonotopic organization of chinchilla core auditory cortex is fully established in the near-newborn. Figure 9 shows maps from representative animals in each age group, with BFs determined at 50 dB SPL. Every animal studied (irrespective of age) showed a well-ordered, low-to-high frequency gradient roughly along the anterior-posterior axis. To a good approximation, tonotopic maps obtained at 25 dB SPL were simply shifted anteriorly with respect to those at 50 dB SPL; this follows from the fact that the BF of an auditory nerve fiber is higher at 25 than 50 dB SPL. No systematic variation in the (estimated) dimensions of the core area was found with age.
|
; Crandall and Caviness 1984; Zhang et al. 2001).
|
Tuning widths of the dominant peaks in firing rate versus tone frequency functions were calculated as described in METHODS (Best frequency and tuning). The results, averaged by age group, are presented in Fig. 11. Tuning widths were much broader at 50 (dark bars) than 25 dB SPL (light bars) in each group (P < 0.001, Student's t-test), as is observed in the auditory nerve due to saturating nonlinearities in cochlear mechanics and neural transduction. A conservative statement is that tuning widths were no less sharp by P3 than in the adult at either SPL. Because of the high variability observed between neurons (see Fig. 16), the apparent broadening of tuning widths in the adult group was only of modest statistical significance, with P values in the 0.05–0.1 range (ANOVA). However, it has been reported in the cat that cortical neurons are in fact more broadly tuned in adults than in the young (Bonham et al. 2004; Eggermont 1996). Again, sharp frequency tuning in the near-newborn chinchilla is in contrast to the very broad tuning seen in the rat and mouse for some weeks after birth (Chang and Merzenich 2003; Ehret and Romand 1992; Zhang et al. 2001).
|
|
Although tonotopic projections to chinchilla auditory cortex are well ordered and sharply tuned by P3, the fraction of complex cortical neurons (as defined in Receptive field complexity) was observed to increase steadily beyond P30. This is shown in Fig. 12, which shows examples of firing rate contour plots at 50 dB SPL from each age group (column). Exact numbers are given in Fig. 13; the developmental increase in both the fraction of complex spectral units (bottom 2 bars) and complex temporal units (top 2 bars), is evident. We note that response complexity depends on the sound pressure level; in every age group, the fraction of complex responses increased from 25 to 50 dB SPL. The firing rates of neurons in our sample were almost always higher at 50 than at 25 dB SPL, so the chance that an excitatory convergent input is effective, giving rise to a complex response, increases with SPL. Finally, we found that the average frequency separation of distinct peaks in complex spectral units was significantly larger in adult chinchillas than in the three young groups (P < 0.001, ANOVA), implying that the additional input connections in adults arrive from more distant iso-frequency channels.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; review: Sohmer and Freeman 2001). These findings in the chinchilla distinguish it from more altricious laboratory rodents, such as the rat, which remains deaf for several weeks after birth and shows very poor tonotopic organization and frequency tuning for 2–3 wk after the onset of hearing (Zhang et al. 2001). While the chinchilla cochlea and tonotopic pathway appear fully developed by birth, the establishment of connections between cortical and/or subcortical iso-frequency bands continues postnatally, as indicated by the observed increase in receptive field complexity with age. Methodological considerations
SAMPLING BIAS. The results presented in this paper consist mostly of averaged data from all core auditory cortical neurons sampled in a given age group, with only stimulus sound pressure level held constant. The results, however, depend on several parameters in addition to the SPL. For example, response latency and tuning width both also depend on a unit's best frequency, threshold sensitivity, and depth in the cortex. It's convenient when the distributions of the latter parameters are similar between age groups; otherwise, they may confound relationships in the data that have been attributed to developmental processes.
The large majority of units sampled in each age group were from cortical layer IV. Layer IV is 
400–500 µm below the cortical surface in young animals and 600–700 µm in adults. It can also be identified during recording as the source of the strongest multi-unit activity, because it contains the highest density of principal cells (receiving direct thalamic input). In this study, the distribution of cortical depths was not significantly different between age groups (after linearizing for the increasing depth of layer IV with age).
There were, however, differences in response threshold (Fig. 5), and these were likely largely responsible for differences in the distributions of best frequencies sampled between groups, shown in histogram form in Fig. 14. Light bars group BFs determined at 25 dB SPL and dark bars at 50 dB SPL. In the youngest animals, BFs are skewed toward the higher frequencies. Their distribution becomes progressively more uniform by adulthood, reflecting the changes in hearing sensitivity with age. The higher the threshold for a given sound frequency, the smaller the chance of detecting responses at that frequency, particularly at the low-moderate SPLs used in this study. Since the newborn chinchilla is more sensitive at high frequencies yet less sensitive at low frequencies (relative to the adult), a larger fraction of P3 neurons was in the high-frequency range. This is important, because, for instance, response latency decreases with (increasing) BF, as shown in Fig. 15 (at 50 dB SPL). (The decrease is consistent with the shorter distances traveled by higher frequency waves on the basilar membrane.) The reported decrease in response latency with age (Fig. 8) should therefore be adjusted for the bias toward higher BFs in younger animals, which effectively deflates the average latency value. We find that normalizing response latencies by frequency slightly increases the differences between groups, so that the reported decrease in latency with age becomes more significant. Figure 16 shows that tuning width also decreases with the BF in each age group. (Lower frequencies are mapped to longer basilar membrane segments than higher frequencies.) When the normalization is made, the increase in average tuning width in adults over the young animals becomes slightly less significant than reported in Fig. 11. The conclusion that tuning widths by P3 are as sharp as at any age does not change.
|
|
.) The average duration of sedation per dose did not vary with age (doses were normalized by animal weight). As an additional check, the time between doses was divided into three equal intervals. Tuning widths were slightly but significantly sharper in the first compared with the last interval (stronger compared with weaker anesthesia effect), and the fraction of complex units was slightly lower (again, in the 1st interval compared with last). However, these differences appeared no bigger in the young animals than in the adults, indirect evidence that the effects of anesthesia were not more severe in the younger animals.
PURE TONES.
Pure tone stimuli allow for a rapid assessment of the frequency tuning of auditory neurons (when obtaining detailed tonotopic maps is a priority), but are less useful for predicting responses to complex inputs. Since a physical stimulus can be approximated to any desired degree of accuracy by a finite sum of suitably chosen sinusoids, the output of a linear system to any given input can be determined by summing the separate outputs to each component sinusoid. However, at best, (auditory cortical) neurons may exhibit only roughly linear behavior and over a limited range of stimulus parameters at that; a reliable description of neuronal responses thus requires a nonlinear approach, e.g., Volterra-Wiener methods (Marmarelis and Marmarelis 1978). Nevertheless, we speculate that the observed increase in tonal receptive field complexity in the developing chinchilla cortex reflects the emergence of selectivity for fragments of complex sounds, such as the frequency (formant) stacks (complex spectral units) and modulations (complex temporal units), characteristic of animal vocalizations (e.g., Sen et al. 2001). Abeles and Goldstein (1972) noted long ago that auditory cortical neurons responsive to two (separate) frequency bands typically fired at greater rates if the two tones (1 from each band) were presented simultaneously rather than individually. More recently, deCharms et al. (1998) showed that a more systematic but still linear estimate of cortical single-unit selectivity (the 1st-order Wiener kernel) was generally reliable, at least in the ballpark sense. Thus a stimulus whose spectrogram was matched to the neuron's "spectrotemporal receptive field" typically produced firing rates in excess of those evoked by pure tones, noise, and nonmatched complex stimuli (of equivalent intensity).
Development of tonotopic projections to, and lateral connections in, auditory cortex
This study provides indirect evidence that intrinsic mechanisms (under genetic control) may be sufficient to establish precisely ordered (tonotopic) maps of the cochlea in the auditory pathway. More specifically, it is now generally accepted that protein factors guide developing (relay) axons to appropriate target locations, where considerable axonal arborization leads to synapse formation over a relatively large area. This initially diffuse projection may be refined during a period of rhythmic spontaneous bursting (probably of cochlear origin), because synchronized activity in groups of nearby primary neurons provides the (Hebbian) drive for the elimination of misplaced connections. The evidence provided complements that of earlier studies in vision: retinotopic projections (Naegele et al. 1988) and ocular dominance columns (Horton and Hocking 1996) are well developed before the eyes first open. We found that tonotopic projections to auditory cortex were fully mature in all P3 chinchillas (recording times actually ranged from about P40 h to P90 h). It therefore appears that the (purported) refinement of diffuse projections occurs in utero, where, as elaborated below, sound reaching the fetal cochlea is considerably attenuated.
Measurements of cochlear microphonics (electric potentials measured at the round wind that are generated by outer hair cell activity) in (near-term) fetal lambs and again in the same lambs after birth (using the same implanted recording electrodes) showed that the threshold of hearing in utero was elevated by about 20 dB at 250 Hz, 35 dB at 500 Hz, 40 dB at 1 kHz, and 45 dB at 2 kHz (Gerhardt et al. 1992), continuing to rise with frequency. Most of the attenuation appears to be due to the immobile fetal middle ear (filled with amniotic fluid and loose tissue), which leaves conduction through skull as the only path to the cochlea (Sohmer et al. 2001). This means that maternal sounds (e.g., heartbeat, movement) are also attenuated, to the point, in fact, of being inaudible at all but the lowest frequencies. Of course, sufficiently loud external sounds will be audible to the fetus, but there are reasons to think that these have limited (if any) impact on tonotopic map refinement. First, broadband noise at (the total power of) 70 dB SPL may also fail to mask the loudest environmental sounds, yet sufficed to prevent map refinement in the rat (Chang and Merzenich 2003). Second, sound attenuation in utero increases steadily with frequency, yet in the near-newborn chinchilla, we found no evidence for an increase in the degree of map disorder at the higher iso-frequency bands (data not presented).
Since both patterned environmental input and patterned spontaneous input may drive topographic map refinement, perhaps different combinations of the two mechanisms are favored in different species (e.g., environmental input in the rat, spontaneous input in the chinchilla). Alternatively, it may prove that spontaneous activity is the more common (and efficient) mechanism. Patterned spontaneous activity in sensory neurons seems to exist only during short time windows in early development (Demas et al. 2003; Lippe 1994), after which its disappearance may be permanent. This might explain the necessity of patterned environmental input for the refinement of regenerating retinotopic maps (following optic nerve injury) in mature nonmammals (e.g., Schmidt and Eisele 1985). Furthermore, patterned spontaneous activity can be disrupted by external stimuli (e.g., noise; Lippe 1994). Perhaps such disruption occurs in the course of continuous noise stimulation of the developing rat (Chang and Merzenich 2003), and this, rather than the masking of the patterned acoustic environment, is the (initial) reason map organization remains poor.
While spontaneous activity (along with chemoaffinity) may suffice for the establishment of sharply tuned, orderly representations of sensory epithelia in the primary cortices, there is no doubt that neuronal receptive fields (and neuronal assemblies) are subsequently shaped by the external environment (review: Buonomano and Merzenich 1998). To take an example from the auditory system, Recanzone et al. (1993) showed that (extensive) practice on frequency discrimination tasks caused receptive fields in the adult monkey cortex to narrow, while practice with frequency modulations caused them to broaden. Thus it's likely that the somewhat wider tuning and greater map disorder observed in the adult chinchilla are a consequence of correlated activity stimulated by broadband species-specific calls (and not, for instance, a reflection of more diffuse tonotopic projections). The acoustic environment of the developing chinchilla includes a considerable repertoire of such calls. We suggest that the increasing numbers of complex units with postnatal age reflect specialization for the detection of the predominant spectral-temporal features of chinchilla vocalizations (cf. Sen et al. 2001).
The proposed roles of spontaneous and externally stimulated activity in auditory system development should be further tested by rearing animals in an attenuated acoustic environment. Such an experiment in the rat would leave any patterned spontaneous activity unaffected, and it is interesting if this would allow map refinement to proceed unimpeded despite reduced environmental drive. A similar experiment in the chinchilla might confirm the importance of this external drive for the emergence of complex cortical receptive fields.
Unlike more common laboratory rodents (e.g., rat, mouse, gerbil), the chinchilla has a long gestation period (for its size) and begins to hear in utero. The human fetus likewise begins to respond to (loud) external sound (by increasing its heart rate, movement) as early as the start of the third trimester. At birth, human ABR thresholds are close to their adult values (e.g., Sininger and Abdala 1996; Stuart et al. 1993), as is the case in the chinchilla. It also appears that behavioral sound frequency resolution in humans (as indicated by masked critical bandwidths, psychophysical tuning curves, etc.) is mature <6 mo after birth (review: Werner and Marean 1996). It's therefore tempting to use the chinchilla developmental timeline as a model for the human auditory system, with sharply tuned, orderly cortical tonotopic maps at birth continuing to be bridged by intracortical (and/or subcortical) circuits early in postnatal life. Attenuated levels of speech sound stimulation in infants with chronic otitis media (middle ear infection) are correlated with deficits in language and literacy skills that can persist well into childhood, despite the recovery of normal audiometric thresholds (e.g., Brandes and Ehdinger 1981; Luotonen et al. 1996). The neurobiological bases for these deficits, however, remain largely unknown. Our work suggests that inadequate stimulation in early infancy may not affect the sensorineural machinery required for precise frequency resolution, but may impair the formation of neural connections that integrate information across iso-frequency bands to extract complex spectral-temporal features for more robust speech perception.
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: M. Pienkowski, Rm. 3005, Elizabeth McMaster Bldg., The Hospital for Sick Children, 555 University Ave., Toronto, Ontario M5G 1X8, Canada (E-mail: martin.pienkowski{at}utoronto.ca)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Bonham BH, Cheung SW, Godey B, and Schreiner CE. Spatial organization of frequency response areas and rate/level functions in the developing AI. J Neurophysiol 91: 841–854, 2004.
Brandes PJ and Ehdinger DM. The effects of early middle ear pathology on auditory perception and academic achievement. J Speech Hear Disord 46: 301–307, 1981.
Buonomano DV and Merzenich MM. Cortical plasticity: from synapses to maps. Annu Rev Neurosci 21: 149–186, 1998.[CrossRef][ISI][Medline]
Chang EF and Merzenich MM. Environmental noise retards auditory cortical development. Science 300: 498–502, 2003.
Crandall JE and Caviness VS. Thalamocortical connections in newborn mice. J Comp Neurol 228: 542–556, 1984.[CrossRef][ISI][Medline]
deCharms RC, Blake DT, and Merzenich MM. Optimizing sound features for cortical neurons. Science 280: 1439–1443, 1998.
Demas J, Eglen SJ, and Wong RO. Developmental loss of synchronous spontaneous activity in the mouse retina is independent of visual experience. J Neurosci 23: 2851–2860, 2003.
Eggermont JJ. Differential maturation rates for response parameters in cat primary auditory cortex. Auditory Neurosci 2: 309–327, 1996.
Eglen SJ, Demas J, and Wong RO. Mapping by waves. Patterned spontaneous activity regulates retinotopic map refinement. Neuron 40: 1053–1055, 2003.[CrossRef][ISI][Medline]
Ehret G and Romand R. Development of tone response thresholds, latencies and tuning in the mouse inferior colliculus. Dev Brain Res 67: 317–26, 1992.[CrossRef][Medline]
Fawcett JW and O'Leary DD. The role of electrical activity in the formation of topographic maps in the nervous system. Trends Neurosci 8: 201–206, 1985.
Fritzsch B, Tessarollo L, Coppola E, and Reichardt LF. Neurotrophins in the ear: their roles in sensory neuron survival and fiber guidance. Prog Brain Res 146: 265–78, 2004.[ISI][Medline]
Gerhardt KJ, Otto R, Abrams RM, Colle JJ, Burchfield DJ, and Peters AJ. Cochlear microphonics recorded from fetal and newborn sheep. Am J Otolaryngol 13: 226–233, 1992.[CrossRef][ISI][Medline]
Grubb MS, Rossi FM, Changeux JP, and Thompson ID. Abnormal functional organization in the dorsal lateral geniculate nucleus of mice lacking the beta 2 subunit of the nicotinic acetylcholine receptor. Neuron 40: 1161–1172, 2003.[CrossRef][ISI][Medline]
Harel N, Mori N, Sawada S, Mount RJ, and Harrison RV. Three distinct auditory areas of cortex (AI, AII, and AAF) defined by optical imaging of intrinsic signals. Neuroimage 11: 302–312, 2000.[CrossRef][ISI][Medline]
Harrison RV, Cullen JR, Takeno S, and Mount RJ. The neonatal chinchilla cochlea: morphological and functional study. Scan Microsc 10: 889–894, 1996a.
Harrison RV, Kakigi A, Hirakawa H, Harel N, and Mount RJ. Tonotopic mapping in auditory cortex of the chinchilla. Hear Res 100: 157–163, 1996b.[CrossRef][ISI][Medline]
Harrison RV, Stanton SG, Ibrahim D, Nagasawa A, and Mount RJ. Neonatal cochlear hearing loss results in developmental abnormalities of the central auditory pathways. Acta Otolaryngol 113: 296–302, 1993.[Medline]
Hebb DO. The Organisation of Behaviour. New York: Wiley, 1949.
Horton JC and Hocking DR. An adult-like pattern of ocular dominance columns in striate cortex of newborn monkeys prior to visual experience. J Neurosci 16: 1791–1807, 1996.
Hsu GS, Margolis RH, and Schachern PA. Development of the middle ear in neonatal chinchillas. I. Birth to 14 days. Acta Otolaryngol 120: 922–932, 2000.[CrossRef][Medline]
Hsu RW, Margolis RH, Schachern PA, and Javel E. The development of the middle ear in neonatal chinchillas. II. Two weeks to adulthood. Acta Otolaryngol 121: 679–688, 2001.[CrossRef][Medline]
Kaas JH. Organizing principles of sensory representations. Novartis Found Symp Ser 228: 188–198, 2000.
Kaas JH, Krubitzer LA, Chino YM, Langston AL, Polley EH, and Blair N. Reorganization of retinotopic cortical maps in adult mammals after lesions of the retina. Science 248: 229–231, 1990.
Leake PA, Snyder RL, and Hradek GT. Postnatal refinement of auditory nerve projections to the cochlear nucleus in cats. J Comp Neurol 448: 6–27, 2002.[CrossRef][ISI][Medline]
Lewicki MS. A review of methods for spike sorting: the detection and classification of neural action potentials. Network Comput Neural Syst 9: 53–78, 1998.[CrossRef]
Lippe WR. Rhythmic spontaneous activity in the developing avian auditory system. J Neurosci 14: 1486–1495, 1994.[Abstract]
Lippe WR. Relationship between frequency of spontaneous bursting and tonotopic position in the developing avian auditory system. Brain Res 703: 205–213, 1995.[CrossRef][ISI][Medline]
Luotonen M, Uhari M, Aitola L, Lukkaroinen AM, Luotonen J, Uhari M, and Korkeamaki RL. Recurrent otitis media during infancy and linguistic skills at the age of nine years. Pediatr Infect Dis J 15: 854–858, 1996.[CrossRef][ISI][Medlin
