| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Physiology, University College London, Gower Street, London, WC1E 6BT, United Kingdom
Submitted 11 August 2003; accepted in final form 29 September 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
), including those of the auditory system. The classic "primary-like" response of auditory nerve fibers is a welldescribed example of adaptation in the peripheral auditory pathway (Westerman and Smith 1984; Yates et al. 1985). Adaptation in the visual system appears to function as a gain control mechanism adjusting neural responses to maintain their responsiveness over a wide range of intensities in both the retina (Laughlin 1989) and the visual cortex (Ohzawa et al. 1985). However, specific roles for adaptation in auditory processing are yet to be established. One possible role for adaptation in the auditory system lies in determining the sensitivity of auditory neurons to the stimulus context. The context in which a sound is heard is likely to influence the way in which the auditory nervous system processes that sound, and a number of recent studies have begun to examine how sensitivity to stimulus context, including stimulus history, emerges in the ascending auditory pathway (Malone and Semple 2001; Malone et al. 2002; Sanes et al. 1998; Spitzer and Semple 1991,1998; Ulanovsky et al. 2003). One of the paradigms used in these studies is interaural phase modulation (IPM). IPM is a binaural stimulus in which the interaural phase disparity (IPD) is modulated at different rates, to different depths, or around different values of interaural phase, to produce the percept of auditory motion. Many neurons in the inferior colliculus (IC), the major auditory nucleus in the midbrain, are sensitive to the context in which IPDs are presented as part of the IPM stimulus. Their responses may be enhanced or suppressed depending on the direction, speed, or center-locus of the virtual motion cues (McAlpine et al. 2000; Spitzer and Semple 1991, 1995, 1998). In contrast, responses in the medial superior olive (MSO) appear to reflect the instantaneous interaural disparity of a stimulus (Spitzer and Semple 1995; Yin and Chan 1990), suggesting that the main source for the creation of sensitivity to dynamic binaural cues in the IC lies within the IC itself (Borisyuk et al. 2002).
Subsequent investigations have shown IC neurons to be sensitive to the context in which acoustic cues other than IPDs, such as interaural level differences (Sanes et al. 1998) or frequency glides (Malone and Semple 2001), are presented, and that the influence of dynamic context on neural responses is even more pronounced in auditory cortex than in the IC (Malone et al. 2002). Together, these data suggest a hierarchy of context sensitivity in the ascending auditory nervous system, with MSO neurons showing the least sensitivity (Spitzer and Semple 1995, 1998), cortical neurons showing the most sensitivity (Malone et al. 2002), and IC neurons showing intermediate sensitivity (Malone and Semple 2001; McAlpine et al. 2000; Spitzer and Semple 1991, 1995, 1998).
Although the phenomenology of stimulus-context dependency in the auditory pathway is increasingly becoming established, the mechanisms that contribute to this sensitivity remain to be determined. In terms of dynamic interaural cues, it was originally suggested that binaurally sensitive inhibition, possibly derived from the dorsal nucleus of the lateral lemniscus, was responsible for the appearance of context sensitivity in the IC (Spitzer and Semple 1995, 1998). However, McAlpine et al. (2000) argued that an adaptation process residing above the level of binaural integration was sufficient to account for the data. In support of this, McAlpine and Palmer (2002) demonstrated that blocking GABA-ergic inhibition in IC neurons enhanced differential sensitivity to dynamic cues of IPM, whereas adding tonic inhibition reduced such sensitivity. The explanation posited for this behavior was that increases in the discharge rate that accompanied blockade of inhibition moved neurons closer to an adapting state, whereas decreases in discharge rate that accompanied addition of tonic inhibition moved neurons away from the adapting state. Recent modeling studies explain the responses of IC neurons to a range of static and dynamic interaural timing cues using a combination of ITD-sensitive and ITD-insensitive GABA-ergic inhibition, postinhibitory rebound, as well as an adaptation mechanism intrinsic to the IC in their calculations (Cai et al. 1998a,b).
As the 1st stage in assessing the specific contribution of IC adaptation mechanisms to the coding of dynamic stimulus cues, we have developed a novel stimulus paradigm that enables adaptation characteristics of neurons at or above the level of binaural integration to be measured free of the influences of adapting monaural components in the pathway. The stimulus consists of a 3,000-ms binaural tone, the first 1,000 ms of which is presented at a neuron's least favorable (worst) IPD. During this time, neural components in the auditory pathway that are insensitive to interaural phase cues will be responding, and adapting to their steady-state level. After 1,000 ms, the interaural phase of the stimulus is rapidly switched to the neuron's most favorable (best) IPD, allowing the time course of neural adaptation to be assessed free from the influences of adapting monaural components. This is possible because neurons sensitive to IPDs are essentially "gated OFF" by the worst IPD epoch, becoming active only when "gated ON" on stepping to the best IPD. IPD is the only stimulus parameter of which we are aware whose influence on neural activity is delayed until such a late stage in the auditory pathway. Stepping stimulus parameters such as sound frequency and level (including interaural level) in this manner would influence the output of one or both cochleae and their target neurons, removing the possibility of examining binaural adaptation free from the influence of changing/adapting monaural components. Our data indicate this to be an important consideration when only the binaural configuration of a stimulus is altered. The adaptation profiles obtained by this method most probably reflect mechanisms intrinsic to the IC neurons themselves (Borisyuk et al. 2002), and not the properties of their input neurons in the MSO, given that MSO neurons show little evidence of the differential sensitivity to dynamic cues (Spitzer and Semple 1998). Although only interaural phase cues were used to assess adaptation in the IC, the data have more general application because any stimulus parameter that evokes activity in IC will also evoke some form of adapting response to it. These data will thus be useful in constructing neural models that seek to determine the contribution of different auditory centers to analysis of the auditory scene. Elements of these data were previously published in abstract form (Ingham and McAlpine 2001).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
All experiments were carried out in accordance with the guidelines of the UK Home Office, under the control of the Animals (Scientific Procedures) Act 1986. Pigmented guinea pigs (Cavia porcella) were anesthetized with urethane (1 g/kg; 25% solution) in preparation for surgical procedures and experimental recordings. Analgesia was induced, and supplemented as required, using 0.1 ml intramuscular injections of Hypnorm (fentanyl citrate/fluanisone). Local anesthesia at surgical sites was achieved using subcutaneous injections of 2% lignocaine. A tracheal cannula was inserted, and body temperature was maintained at 37°C using a thermostatically controlled heating blanket and rectal probe. Animals were mounted in a modified stereotaxic frame situated inside a sound-attenuating booth (IAC, Winchester, UK). Hollow ear speculae allowed the insertion of custommade ear phones and probe-tube microphones to form a sealed pressure-field sound delivery system. Pressure equalization of the middle ear was achieved by sealing high acoustic-impedance cannulae into each bulla. A craniotomy was performed, extending 2–3 mm rostral and caudal of the interaural axis, and 1–4 mm lateral from the midline on the right side. The dura overlying the cortex was removed, allowing microelectrode access through the cortex to the right inferior colliculus, and the cranium sealed with 2% agar.
Single-neuron recordings
Neural recordings were made using parylene-coated tungsten microelectrodes (1–5 M
impedance), mounted on a piezo-electric stepper motor, and advanced stereotaxically through the inferior colliculus in a dorsoventral fashion. Electrical activity from the microelectrode was filtered (300 Hz –3 kHz) and amplified (variable gain) using an AC differential amplifier and a PC1 spike conditioner (Tucker Davis Technologies, Alachua, FL). Single neurons were isolated using variable frequency and intensity diotic tones. Single spikes were discriminated from background noise using an SD1 spike discriminator (TDT) and time stamped to 1-µs accuracy using an ET1 event timer (TDT), linked to the computer. Single-neuron isolation was confirmed by the consistency of the discriminated spike waveform displayed on a digital oscilloscope.
Stimulus presentation and data analysis
Acoustic stimuli were produced and presented under computer control. Digitally generated dichotic stimuli (TDT, AP2 digital signal processor; at 100- or 48-kHz sampling rate) were converted to analog signals (TDT, DA3-2). The signals were filtered (TDT, FT6; fc = 40 kHz), and attenuated (TDT, PA4), before amplification and delivered to the ears by RadioShack 40-1377 (Fort Worth, TX), or Beyerdynamic DT-48 (Burgess Hill, UK) loudspeaker units fitted with brass tube attachments sealed into the hollow ear speculae supporting the animal. The sound field inside the sealed system was sampled using Knowles FG3452 microphones by a probe tube inserted to within a few millimeters of each tympanic membrane. Probe microphones were previously calibrated against a Bruel and Kjaer Type 4136 
-in. microphone, and the sound systems for each ear were flat to within ±5 dB from 50 to 2,000 Hz and matched to within ±5 dB for this range.
Once isolated, a neuron's characteristic frequency (CF) was estimated audiovisually, and confirmed by generating a frequency-versus-level response area, extending 2 octaves above and 4 octaves below the estimated CF, using diotic 50-ms tones covering a range of 60–100 dB, presented at 5 Hz in pseudo-random order. Binaural sensitivity to interaural phase disparity (IPD) cues was assessed using binaural beats of 3-s duration, with 1-Hz difference in the carrier frequency presented to the 2 ears. The initial and final 500-ms periods of the response were omitted from analysis, and the middle 2,000 ms were averaged and plotted on-line with respect to the IPD of the stimulus to generate period histograms. Using the methodology of Goldberg and Brown (1969), the vector strength (R) of the response was calculated from the period histogram. The average best phase was calculated as a vector average of the response magnitudes at each point in the cyclic IPD phase histograms. These circular distributions were assessed using the Rayleigh test for the significance of the mean best phase. Where the Rayleigh coefficient < 13.816, the neuron was considered as not significantly modulated by the IPD of the binaural beat, and such units were excluded from further analyses. Best and worst IPD were determined from the IPD period histograms. Because many period histograms were roughly symmetric, the best IPD equated to the mean best phase described above. However, for some neurons, where the IPD-sensitive response produced asymmetric functions, the best IPD was taken to be that evoking the peak response. The worst IPD was generally determined as 0.5 cycles removed from the best phase, or where the phase response dropped to zero at less-favorable IPDs. In other cases, for example, where the plot was asymmetric, or where there was a baseline response, the worst IPD was taken as the IPD evoking the lowest spike activity.
Best and worst IPD values were used to generate IPD-step stimuli designed to determine the time course of adaptation and recovery from adaptation. Digital sound files 3,000 ms in duration ("wav" format, 48-kHz sample rate), were generated using a custom script in Matlab. Two digital sound files were constructed, one for presentation to either ear. Parameters such as carrier frequency, best IPD, worst IPD, and the rate at which the IPD was stepped between these values were under user control. The total phase shift at each ear was minimized for each phase step by advancing the phase by half the required amount (the difference between best and worst IPD in this case) in one ear, and regressing the phase by the same amount in the other ear. This "IPD-step" stimulus is illustrated in Fig. 1, A–E, along with a stylized neural response (Fig. 1, F and G). After stimulus onset at worst IPD, neurons usually responded with a train of discharges that decayed exponentially over time from a maximum (Omax) to reach a steady state (Oss) by the end of the first 1,000 ms of the stimulus. At this point, the interaural phase was stepped rapidly from worst IPD to best IPD, and the neuron responded with a new maximal discharge rate (Bmax), which once more decayed (Bd) over the 300-ms best-IPD epoch to reach a new steady-state level (Bss) above that of Oss. At the end of the best-IPD epoch, the IPD was stepped back to the worst IPD and the neuron responded with a reduced discharge rate, equal to that of the initial steady-state level, Oss. After some predetermined time [interstep interval (ISI)], the IPD was again stepped from worst IPD to best IPD for a further 300 ms. The response during this 2nd best-IPD epoch depends on the duration of the interval between the 1st and 2nd IPD steps. For longer intervals, a similar response pattern was produced to that seen during the 1st best-IPD epoch; that is, the discharge rate was initially high (Rmax) and decayed over time to reach a steady state (Rss). For shorter intervals, the adapting region of the response (Rd) was reduced, such that the initial peak of the recovery response (Rmax) was lower (see Fig. 1G, open circles), or not apparent, followed by a reduction in discharge rate back to the same plateau level seen for the 1st best-IPD epoch (Bss). Thus it was not always possible to fit the response to the 2nd step to best IPD with an exponential function.
|
on), monaural adaptation (mon), and binaural adaptation (bin) were extracted (see Eq. 1)
 | (1) |
is the time constant of adaptation (ms). For onset and binaural adaptation profiles, where responses to multiple stimuli with different recovery intervals were recorded, equivalent data bins from each response were averaged before being smoothed and subject to curve fitting (this was not possible for the monaural adaptation because the response to only the longest recovery time was recorded). The curve fits were subject to statistical analysis using ANOVA. Neurons whose ANOVA test results produced values of P > 0.05 were classified as "nonadapting"; the exponential decay function did not produce a significant fit to the physiological data. Curves fitted to the data for the various adaptation profiles were normalized to enable comparison of the curves and rates of decay for individual neurons. The extent to which a neuron's response adapted was assessed by calculating an adaptation ratio, defined as the ratio of the discharge rate at steady state to that at the peak of the adapting profile (e.g., binaural adaptation ratio = Bss/Bmax).
Recovery from adaptation was assessed in 2 ways. Neural responses to the 1st best-IPD epoch were analyzed as an average of the presentations made with each different interval period. Responses evoked during the 2nd epoch at best IPD were fitted with exponential decay functions, as outlined above (Eq. 1) and described by a 2nd adaptation time constant (bin2). Because responses to the 2nd step at best IPD were based on an average of fewer stimulus presentations compared with the 1st step to best IPD, they were more variable, resulting in fewer neurons producing sufficient data for detailed analysis of the recovery from adaptation. Recovery time constants (rec) were determined for 37 of the above neurons. Here, peak discharge rates from the function fitted to the 2nd step to best IPD were plotted as a function of the interval between the 2 best-IPD epochs (see Fig. 1G). With longer recovery intervals, the peak discharge rate of the 2nd step to best IPD increased toward that of the 1st step to best IPD. These data were fitted with exponential rise-to-maximum curves (see Eq. 2). The steady-state discharge rate evoked by the 2nd step to best IPD was similar, for all ISIs, to the discharge rate evoked during the 1st step to best IPD
 | (2) |
is the time constant of recovery rec (ms). |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Single-neuron responses to IPD steps
The IPD-step stimulus was designed to adapt out the influences of monaural components of the ascending auditory pathway up to the level of the IC, so that the time course of adaptation when only the binaural configuration of a sound is changed could be assessed. This method requires neurons to be IPD sensitive. The top row of Fig. 2 shows responses for 4 IPD-sensitive IC neurons to 3 s of a 1-Hz binaural beat stimulus. In each case, the discharge rate (black histogram bars) was modulated with each cycle of the beat (gray lines), indicating neurons to be IPD sensitive. These data, plotted in the 2nd row of Fig. 2 in the form of period histograms, were used to determine a neuron's best IPD [the IPD that evoked the highest discharge rate (filled circle in each period histogram)] and worst IPD [the IPD that evoked the lowest discharge rate (open circle in each period histogram)]. The remaining rows of Fig. 2 show the responses of each neuron to the IPD-step stimulus. Each neuron responded to the onset at worst IPD with an initially high discharge rate that adapted to a much lower rate (to zero in some cases; e.g., Fig. 2, column B) during the subsequent 1,000 ms. At this point, the IPD was stepped rapidly from worst IPD to best IPD, and held at best IPD for 300 ms, before being stepped rapidly back to worst IPD. This initial step to best IPD evoked a 2nd, initially high, response that adapted to a lower steady-state firing rate over the 300-ms duration of the step. This steady-state response was higher than the steady-state firing rate during the first 1,000 ms of the stimulus, even for those neurons for which the firing rate evoked during the first 1,000 ms remained relatively high after stimulus onset (e.g., Fig. 2, column D). After the step back to worst IPD, firing rates fell back to, and sometimes below, firing rates evoked before the step to best IPD.
|
Measuring spike-frequency adaptation after a step to best IPD
To assess the time course of adaptation after the step to best IPD, single exponential decay functions (Eq. 1) were fitted to the first 300-ms period of the PSTH over which the interaural phase was stepped from worst to best IPD (adjusted for neural latency, see METHODS). The majority of neurons (111/120, 92.5%) showed an adapting profile to this 1st step (Fig. 3, A and B) that was well described by a single exponential decay function (ANOVA, P < 0.0001). Fitted functions to the raw (Fig. 3A) or normalized (Fig. 3B) data indicate mean and median adaptation time constants for this step to best IPD (bin) of 61.4 and 46.9 ms, respectively. Time constants were not normally distributed [Kolmogorov–Smirnov (K-S); P < 0.0001] but, rather, were skewed toward lower values of bin (Fig. 4A). Using boxplot analysis, 4 neurons (dashed lines in Fig. 3B and open bars in Fig. 4A) were classified as outliers. Removing these outliers from the analysis reduced the mean bin from 61.4 to 52.9 ms and produced a dramatic reduction in the variance of the population data (see Table 1). These outliers were excluded from further analysis. Responses of the remaining 9/120 (7.5%) neurons to the initial step to best IPD were not well described by exponential functions (ANOVA, P > 0.05). These neurons tended to show a "build-up" response to the step to best IPD, or a stepped response that did not adapt (data not shown).
|
|
|
Recovery from adaptation
A neuron's time course of recovery from adaptation was assessed from the response to the 2nd step to best IPD (see columns in Fig. 2). The magnitude of the initial response to this 2nd step was highly dependent on the duration of the interval between the 2 steps, and by fitting exponential rise-to-max functions to the peak discharge rates over the range of ISIs (Fig. 5; and see Fig. 1G), we were able to determine the rate at which neurons recovered from the effects of adaptation to the 1st step. Of the 48 neurons for which sufficient data points were available, 37 showed a significant exponential fit (ANOVA, P < 0.1), and recovery time constants (rec) derived from these fits (Fig. 6, A and B) indicate that recovery from binaural adaptation was a much slower process than adaptation itself (mean rec = 225.5 ± 210.6 ms; see distribution of rec in Fig. 6C). Relative to bin, rec was more variable, and there was no correlation between bin and rec for individual neurons.
|
|
Under natural listening conditions it is likely that a neuron is subject to continuously varying states of adaptation and recovery from adaptation, mirroring the time-varying nature of most natural sounds. To assess a neuron's adaptation during different adaptation states, exponential decays were fitted to the response profiles recorded to the 2nd step to best IPD and a 2nd binaural adaptation time constant (bin2) was calculated (see Fig. 5 and Fig. 1G) in a similar manner to the step to best IPD. Adaptation time constants measured for the 2nd step to best IPD were similar to those measured for the 1st step to best IPD for individual neurons. For recovery intervals for which sufficient data were available, bin2 was generally in accordance with bin. Figure 7 compares bin2 with bin for ISIs of 1,000 ms (7A), 300 ms (7B), 100 ms (7C), and 30 ms (7D). Linear regression fits to the data for each interval, and for the pooled (Fig. 7E) and average (7F) data indicate a clear relationship between the 2 adaptation time constants. These observations were confirmed by plotting the averaged bin2 against its recovery interval (Fig. 7G). There was no systematic change in adaptation time constant recorded across the range of recovery intervals tested. Thus whenever IC neurons adapt, they tend to adapt at the same rate irrespective of their current state of adaptation.
|
The tones presented to each ear in the IPD-step stimulus contain a rapid phase shift during the transition from worst to best IPD. Even with monaural presentation, there is an audible "blip" when this transition occurs (4 times in the 3-s IPD-step stimulus), a percept that likely arises as a result of the simultaneous phase shift across many peripheral auditory filters—the rapid phase shift is equivalent to a FM. Because we record the activity of single neurons, presumably reflecting the output of very few auditory filters, this "blip" is unlikely to influence greatly, or at all, neural discharge rates. However, to discount any influence of the monaural phase shifts per se on the response of individual neurons to the IPD-step stimulus, we recorded responses to monaural and diotic stimulation that provide for the same rapid phase transitions as the dichotic stimuli, but not for the change in interaural phase cues.
PSTHs for two representative neurons, in response to contralateral monaural stimulation (Fig. 8, A and B) or diotic stimulation in which the phase transition in each ear was in the same direction (Fig. 8, C and D), indicate no influence of the phase transition on evoked discharge rates (cf. with large increases in discharge rate to dichotic transitions in Fig. 2). Analysis of the discharge rate during the 5- to 10-ms period of the phase transition for 23 neurons (monaural or diotic) revealed little or no influence of these phase shifts on evoked discharge rates (Fig. 8, E and F). No systematic increase or decrease in activity could be attributed to the time of the phase transition. This finding is important because it demonstrates that IC neurons are insensitive to the rapid monaural phase shifts that generate the steps between worst and best IPD, and that any changes in discharge rate can be attributed to the gating of activity through the binaural coincidence detectors in the MSO that project to these neurons.
|
The rationale for the IPD-step stimulus is that it enables adaptation time constants to be measured above the level of binaural integration free from the influence of adaptation mechanism at levels lower than that of binaural integration. This is possible because holding the stimulus at worst IPD effectively gates off the output of binaural neurons in the MSO, evoking little or no activity in the IPD-sensitive IC neurons to which they project. When the stimulus is stepped to best IPD after 1,000-ms components of the binaural pathway below binaural integration in the MSO, and any monaural inputs to the IC, should be fully adapted. Any subsequent adaptation to changes in the binaural configuration therefore reflects adaptation of binaurally sensitive components only. Note that although binaural adaptation could reflect processing in the MSO, which is simply projected to the IC, indirect evidence, such as the relative insensitivity of MSO neurons compared with IC neurons to different directions of motion, for example, argues against this (see DISCUSSION). Here, we compare responses at stimulus onset to those at the step to best IPD.
Most neurons showed adapting profiles to the onset of monaural stimulation, with discharge rates decaying from a peak rate to a lower, steady-state rate. In some cases, monaural steady-state discharge rates were comparable to those at worst IPD (Fig. 2, columns A and B), although, consistent with previous reports (e.g., Kuwada and Yin 1983; McAlpine et al. 1996; Yin and Kuwada 1983), discharge rates evoked by monaural stimulation could be higher (Fig. 2, column C) or lower (Fig. 2, column D) than discharge rates evoked at worst IPD. Similar responses were also observed using diotic stimulus presentations, where the left waveform was presented simultaneously to both ears (data not shown).
To assess adaptation at stimulus onset, responses for the IPD-step stimulus, and responses to monaural presentation of the left-channel wav file, were fitted with single-exponential decay functions (Eq. 1) for the first 300 ms after stimulus onset. Decay functions fitted to the onset at worst IPD (on) are shown in Fig. 3, C and D (n = 96), and gave mean and median adaptation time constants of 48.1 and 32.0 ms, respectively. Decay functions fitted to the monaural onset (mon) are shown in Fig. 3, E and F (n = 52) and gave mean and median time constants of 51.7 and 34.5 ms, respectively. Neurons with time constants above 133 ms for the worst IPD at onset condition (n = 7, dashed lines in Fig. 4D) and neurons with time constants above 136 ms (n = 6, dashed lines in Fig. 3F) for the monaural onset condition were classified by boxplot analysis as outliers, and were excluded from subsequent analyses.
There was substantial variability in the peak discharge rate evoked at stimulus onset for both the monaural and onset at worst-IPD conditions. However, the steady-state firing rates to the monaural onset were less variable than the steady-state firing rates to the worst-IPD onset. Visual inspection indicates that neurons adapted more rapidly at stimulus onset, for both the worst-IPD onset and monaural onset conditions, than at the step to best IPD (cf. Fig. 4, B, D, and F), and this is confirmed by the distribution of time constants in Fig. 9, A and B, which also indicate a significant skewing of the distributions for mon and on (K-S, P < 0.0001). However, this was less pronounced for adaptation at the step to best IPD (Fig. 9, A and B; open bars), with relatively more neurons showing longer adaptation time constants than for onset at worst IPD (Fig. 9A; closed bars) or monaural onset (Fig. 9B; closed bars).
|
mon and the onset at worst IPD on were significantly correlated (Fig. 9E; P < 0.001). However, there was no appreciable correlation between measurements of the recovery time constant rec, and mon or on for the same neurons (r2 < 0.01). For the majority of neurons, responses at onset at worst IPD and monaural onset were more strongly adapted than at the step to best IPD. The steady-state rate was just 12 ± 11% (Fig. 10A; filled bars) and 18 ± 13% (Fig. 10B; filled bars), respectively, of the initial discharge rate evoked at stimulus onset, compared with 42 ± 18% for the step to best IPD (Fig. 10, A and B; open bars, reproduced from Fig. 4B). This trend was also evident for individual neurons (Fig. 10, C–E). Where comparison of the adaptation ratio for monaural and onset worst IPD was possible (n = 49), adaptation to the onset worst-IPD stimulus was usually more pronounced [i.e., had a lower adaptation ratio (Fig. 10E; 35/49 cases)] than adaptation at monaural onset. The remaining 14 neurons showed equal, or more pronounced, adaptation to the onset of the monaural stimulus (Fig. 10E; data points on or below the line of equivalence). Just 5 of 96 neurons demonstrated more pronounced adaptation to the 1st step to best IPD compared with the onset at worst IPD (Fig. 10D), and just 2 of 51 neurons demonstrated more pronounced adaptation to the 1st step to best IPD compared with the monaural onset (Fig. 10C); that is, the depth of adaptation was normally more pronounced at stimulus onset than at the step to best IPD after 1,000 ms of a worst-IPD stimulus.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Distinguishing central adaptation from peripheral adaptation
No previous study of which we are aware has been able to distinguish between peripheral (in this context, monaural) adaptation and binaural adaptation time constants, and the ability to do so here is an important outcome of the current study. Using an adaptor-probe stimulus, Finlayson and Adam (1997) found superior olive neurons to have a binaural adaptation time constant of about 20 ms, considerably shorter than the 51.7 ms observed here, and for adaptation to monaural stimulation to be similar to that for binaural stimulation. Recovery from adaptation was also faster (about 100 ms) than in the current study (mean = 226 ms) and less variable. It is tempting to suggest possible reasons for these differences, including the fact that we have been able to exclude adapting monaural influences using the IPD-step stimulus, whereas this and previous studies of binaural adaptation have not been able to do so. The rapid adaptation in auditory nerve–fiber responses (Westerman and Smith 1987; Yates et al. 1985), and the rapid recovery from adaptation (Yates et al. 1983), suggests that the time course of adaptation in the peripheral nerve fibers might dominate the time course of adaptation in higher centers, unless it is somehow filtered out by neurons at subsequent stages. In the current study, all of the neurons were low-frequency neurons sensitive to IPDs. The binaural brain stem neurons from which they receive their input must, therefore have phase-locked inputs by specialized pathways that retain, even enhance, temporal sensitivity. Nevertheless, the responses of low-frequency auditory nerve fibers also adapt, albeit the rapid component of their adaptation is less pronounced (slower) than for high-frequency fibers (Westerman and Smith 1985). If the combined effect of hair cell, auditory nerve, and brain stem adaptation is transmitted to brain stem nuclei, MSO neurons will receive a diminishing input during the initial 1 s of our IPD-switch stimulus, and will likely show adaptation at stimulus onset. This raises the question, aside from adaptation at stimulus onset brought about by adapting inputs to the MSO: Are the ongoing responses of MSO neurons themselves subject to adaptation? Recordings from the MSO in response to IPM stimuli indicate little or no influence of rate, direction, or depth of the phase modulation (Spitzer and Semple 1995). This suggests that MSO neurons might not be subject to intrinsic adaptation processes, once adaptation of monaural inputs has been accounted for.
Although a proportion of the adapting response to stimulus onset at worst IPD may have been a consequence of the activation of monaural pathways to the IC, it is likely that this response also reflects activity that "bleeds through" the binaural cross-correlator at stimulus onset. This is evident even in the responses to binaural beats in the current study (e.g., top row of Fig. 2, A–D), where there is a large response at stimulus onset, corresponding to zero IPD, that is not evoked during the subsequent cycles of the beat.
Consequences for binaural processing
Spitzer and Semple (1998) noted that primary binaural neurons in the superior olivary complex—neurons with phase-locked monaural inputs—were considerably less influenced by the differences in the rate, direction or depth of the IPM stimulus than were neurons in the IC, leading them to conclude that a hierarchy of binaural responses existed, with sensitivity to apparent motion cues emerging between the site of primary binaural integration in the brain stem and primary auditory cortex (Malone and Semple 2001; Malone et al. 2002). Subsequent investigation positing a mechanism of adaptation-of-excitation as being responsible for the differential sensitivity of IC neurons to acoustic motion cues (Ingham et al. 2001; McAlpine and Palmer 2002; McAlpine et al. 2000). This suggests the existence of a hierarchy of adaptation-of-excitation, with presumably wider significance for auditory processing than simply low-frequency acoustic-motion cues. Several recent studies (Borisyuk et al. 2002; Cai et al. 1998a,b) used arbitrarily determined adaptation time constants in an attempt to explain the differential sensitivity of IC neurons to auditorymotion cues. The measure of binaural adaptation we have obtained could be used in models of IC processing to determine the contribution of adaptation mechanisms in the IC to any time-varying stimulus, whether binaural or monaural.
Adaptation and sensitivity to stimulus context
The original observation by Spitzer and Semple (1991) of noncontiguous responses to the dynamic IPD cues of interaural phase modulation suggested that instantaneous IPD was not the only factor that influenced the responses of neurons, but that stimulus history also had an influence on discharge rates. Although McAlpine et al. (2000) argued that response history rather than stimulus history was the critical determinant in rendering IC neurons sensitive to different directions or depths of the motion cues in IPM, stimulus history and response history are intimately linked, given that the only means by which a stimulus feature, such as its context, can be encoded in the brain is by the pattern of action potentials generated. In this sense, stimulus history is, necessarily, response history.
What possible role might spike-frequency adaptation perform in auditory processing? A number of studies, most notably in the visual system, have suggested a role for adaptation-of-excitation in scaling neural output to take account of, for example, stimulus variance (Brenner et al. 2000; Fairhall et al. 2001). It is also well described that complex cells of the visual cortex adapt to the local contrast (Carandini and Ferster 1997; Laughlin 1989; Ohzawa et al. 1982), the effect being to position a neuron's dynamic range of discharge rates over the relevant range of contrasts. Cells in the lateral geniculate and simple cells in the visual cortex are sensitive only to absolute contrast (Ohzawa et al. 1982). If mechanisms underlying binaural adaptation recorded in the IC are inherent to the IC and not a reflection of potential adaptation process occurring in the MSO, it suggests the emergence of context sensitivity at the level of the auditory midbrain driven by adaptation. Although the role of this context sensitivity remains to be determined, it appears to influence sensitivity to dynamic time and level cues, as well as frequency modulations, suggesting it has some general function in providing sensitivity to stimulus context, and may render neurons more useful in allowing them to adjust their output to take account of local stimulus conditions.
Is adaptation in the IC additive or multiplicative?
One outstanding question concerns the form of spike-frequency adaptation in the IC, whether it is multiplicative or additive (i.e., whether neurons adapt their response by a constant factor or by a constant number of action potentials). Additive and multiplicative processes are implicated throughout the brain in neural gain control, the scaling of discharge rate to match the properties of the excitatory input to the dynamic range of neuronal firing (e.g., Mitchell and Silver 2003). Further, auditory tasks that are accomplished by comparison of activity between populations of neurons, as has been recently suggested for sound localization (McAlpine et al. 2001), would require neurons to adapt at the same rate independent of their current adaptation state, a feature of multiplicative adaptation. Although the IPD-step stimulus was not designed to test whether adaptation in the IC is additive or multiplicative, we have preliminary evidence that IC neurons adapt multiplicatively (Ingham et al. 2002). In addition, adaptation time constants appear to be independent of the current state of adaptation (see Fig. 7), consistent with a multiplicative process. Finally, the population adaptation profile is better described by a proportional reduction in spike rate than by a subtractive reduction in spike rate. Figure 11A shows the distribution of unadapted discharge rates at onset to the 1st step to best IPD. The adaptation profiles in Fig. 11B lead to the distribution of final, steady-state discharge rates in Fig. 11C. The mean onset and steady-state responses were used to calculate the expected population response if spike-frequency adaptation adapts on average by a constant proportion (Fig. 11D) or by a constant number of spikes (Fig. 11E). The observed steady-state distribution (Fig. 11C) better reflects adaptation by a constant proportion, more consistent with a multiplicative process, than adaptation by a constant number of spikes, more consistent with an additive process. Although this is not conclusive evidence for multiplicative adaptation, it argues that the population of IC neurons always adapts proportionally.
|
The term "adaptation" appears to be used to describe a range of phenomenological observations, and does not provide much insight into any specific mechanisms that might account for neuronal response properties. Potential mechanisms include intrinsic membrane properties involving small-conductance calcium-activated potassium, or SK, channels (Cordoba-Rodriguez et al. 1999; Tang et al. 1997; Wagner et al. 2001; Wikström and El Manira 1998). SK channels have been localized within the mammalian IC using in situ hybridization (Stocker and Pedarzani 2000). Prolonged depolarization of the cell membrane—such as that seen during high discharge rates—results in the slow activation of an inward potassium current (Xia et al. 1998), which builds up over time producing a significant hyperpolarizing effect on a neuron's resting membrane potential. Such a mechanism is supported by in vitro studies indicating that 
25% of intracellularly recorded responses of IC neurons show adapting spike patterns in response to depolarizing current injection (Li et al. 1998; Peruzzi et al. 2000; Sivaramakrishnan and Oliver 2001). An alternative mechanism to account for the adaptation profiles we observe involves neuronal networks, either local or from other midbrain regions, feeding inhibitory input onto IC neurons (Burger and Pollak 2001; Chen et al. 1999; Kelly and Li 1997). However, recordings from a single IC neuron indicate that adaptation is still evident even when GABA-ergic inhibition is blocked using bicuculline methiodide, which did not abolish the neuron's adapting response profile (Fig. 12). This suggests that GABA-ergic inhibition does not underlie the adaptation process observed in the IC, consistent with recent data indicating that blocking GABA-ergic inhibition does not abolish sensitivity to the dynamic interaural phase cues of IPM (McAlpine and Palmer 2002). Such sensitivity has been suggested to arise from a process of spike-frequency adaptation (McAlpine et al. 2000). Interpretation of this particular result should be treated cautiously, however, because bicuculline methiodide is also known to alter directly SK channel currents (e.g., Tonkovic-Capin et al. 2001). A third possibility is the involvement of fast synaptic plasticity, such as that observed in the barrel cortex (Chung et al. 2002). However, this fast synaptic plasticity appears to have a slower time course than the adapting response profiles we observe, although this does not necessarily rule it out as a potential mechanism.
|
), suggesting their responses do not adapt to any great extent. MSO neurons possess certain cellular specializations for precise coincidence detection such as AMPA receptors (reviewed in Trussell 1997), ideal for the rapid and secure transmission of high-rate, phase-locked inputs critical to their primary role in binaural coincidence detection. Once binaural integration is complete, however, such precision may no longer be required. Thus any change in sensitivity to dynamic interaural cues may reflect a transition from a purely "cues"-based encoding of interaural disparities in the MSO to a more "context"-based encoding of interaural disparities in the IC. One could conclude that spike-frequency adaptation in the IC is brought about by intrinsic membrane properties of IC neurons themselves, perhaps by specific ion channels, rather that by the action of networks of inhibitory connections or binaurally adapting inputs from the MSO. |
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Present address: N. J. Ingham, Centre for the Neural Basis of Hearing, The Physiological Laboratory, Downing St., Cambridge, CB2 3EG, UK.
GRANTS
This work was supported by an MRC Career Establishment Grant awarded to D. McAlpine.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: D. McAlpine, Department of Physiology, University College London, Gower Street, London, WC1E 6BT, UK (E-mail: d.mcalpine{at}ucl.ac.uk).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Borisyuk A, Semple MN, and Rinzel J. Adaptation and inhibition underlie responses to time-varying interaural phase cues in a model of inferior colliculus neurons. J Neurophysiol 88: 2134–2146, 2002.
Brenner N, Bialek W, and de Ruyter VS. Adaptive rescaling maximizes information transmission. Neuron 26: 695–702, 2000.[CrossRef][ISI][Medline]
Burger RM and Pollak GD. Reversible inactivation of the dorsal nucleus of the lateral lemniscus reveals its role in the processing of multiple sound sources in the inferior colliculus of bats. J Neurosci 21: 4830–4843, 2001.
Cai H, Carney LH, and Colburn HS. A model for binaural response properties of inferior colliculus neurons. I. A model with inte
