INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

One of the most common complaints of the aged listener is difficulty understanding speech in compromised listening situations, as occurs when background noise is present. This can occur even when there is only a mild peripheral hearing impairment. These clinical findings have motivated a research effort that focuses on the examination of neural correlates of the age-related hearing deficits suffered by nearly 20 million elderly listeners in the US (Gordon-Salant and Fitzgibbons 1993; Frisina and Frisina 1997; Snell and Frisina 2000). It is now generally accepted that there are at least two components to presbycusis, or the decline in hearing with age, a peripheral component involving cochlear degeneration and a central component involving declines in the auditory CNS.

Speech and other species-specific communication signals are characterized by rapid intensity perturbations, which carry important information-bearing parameters required for effective auditory perception. The temporal structure of these AM signals can be periodic or aperiodic. Periodic amplitude fluctuations, such as occur in vowels, can be modeled by using amplitude-modulated tones or noise stimuli, while aperiodic intensity modulations, which for example determine voice onset time, can be modeled by gap-detection paradigms. Previous psychoacoustic studies on gap detection provide a basis for the suggestion that more fundamental temporal acuity measures are linked to deficits in speech processing in the aged listener (Fitzgibbons and Gordon-Salant 1996; Moore et al. 1992; Snell and Frisina 2000).

AM is inherent in most species-specific vocalizations, including speech. Low-frequency AM can convey prosody, while AM carriers with modulation frequencies above 20 Hz convey a strong pitch sensation (Zwicker and Fastl 1990). Bregman et al. (1985) has also shown that AM signals embedded in cocktail party noise can convey sound object identification information. The fact that temporal cues are critically important for speech perception was reinforced by Shannon (1995), who found that speech signals, digitally processed to remove all spectral information, could still be correctly understood by young adult listeners fitted with cochlear implants. Since periodic fluctuations in sound intensity are a common acoustic signature of the speech envelope, it has been suggested that differences in AM during speech may be one important cue in discriminating differences between phonemic categories (Rosen 1992, Shannon et al. 1995). For example, Van Tasell and colleagues (1987) reported that in normalhearing listeners the information content of consonant/vowel syllables was significantly reduced when the speech envelope was altered in such a way as to remove temporal fluctuations above 20 Hz.

Recently, evidence has emerged suggesting that elderly listeners process AM signals differently than young adults (Snell and Frisina 2000). Age-related deficits in temporal acuity are found even when older listeners are closely matched to younger listeners with respect to audiometric hearing loss (Snell 1997; Snell and Frisina 2000) or when the gap stimulus is presented at equal sensation levels to the listeners in each age group (Schneider et al. 1994). These results highlight the important contribution of central temporal processing deficits to age-related speech perception disorders.

There has been exhaustive research describing the neural encoding of AM signals in the inferior colliculus (IC) of many mammals. It is now known that the IC manifests a transformation in the manner in which envelope periodicities are encoded in the spike trains of auditory neurons. Caudal to the IC, AM signals are primarily represented in the phase-locking of the ongoing spike train to each modulation cycle of the AM stimulus. This holds true for the auditory nerve (Javel 1980; Joris and Yin 1992; Moller 1976), cochlear nucleus (Backoff et al. 1999; Frisina et al. 1985, 1990; Rhode and Greenberg 1994), lateral superior olive (Batra et al. 1997a,b; Joris and Yin 1995), and medial superior olive (Grothe et al. 1997). In contrast, at the level of the mammalian IC, most units do not phase-lock to high AM frequencies (above 400 Hz), but display spike count changes which vary as a function of modulation depth and frequency (Langner and Schreiner 1988; Moller and Rees 1986; Rees and Moller 1987). In addition, rate coding has also been found to vary as a function of stimulus rise times (Barsz et al. 1998; Heil 1997a; Phillips 1988; Suga 1971). In guinea pig, Rees and Moller (1983) found that most IC units had best modulation frequencies for spike count (rBMFs), of <120 Hz, with band-pass (tuned) rate modulation transfer functions (rMTFs). Response increased monotonically with increases in modulation depth, typically saturating before modulation depths reached 100%. Modulation thresholds as small as 2% were measured in some units and in many units, rBMF remained constant with large changes in modulation depth.

The ability of IC neurons to encode temporal variations in sine-wave amplitude-modulated (SAM) stimuli was found to be intermediate between that of the auditory nerve and the cortex, with rBMFs ranging between 3 and 1000 Hz, and the majority of cells having rBMFs <400 Hz (Rees and Moller 1983, 1987). Finally, in 1988, Langner and Schreiner found that best modulation frequencies (BMF) were topographically organized in the IC and argued that this map of BMF would allow for early neural encoding of AM frequency, as might be used for pitch perception (Langner and Schreiner 1988).

There is also evidence which points to deficits in temporal processing in the aged auditory midbrain from both evoked potential and single-unit studies. Results from the early work by Willott and colleagues showed that single-unit responses to simple sounds from old CBA mice were mildly affected by the aging process (Willott et al. 1988). Boettcher et al. (1996) used a gap paradigm to elicit auditory brain stem responses in young and old gerbils and found that latencies assigned to midbrain components were prolonged in old gerbils, suggesting that the magnitude of the forward masking effect was greater. Importantly, these researchers simultaneously recorded from the auditory periphery and found no change in response latency with age. Walton et al. (1997) found that single-unit correlates of the gap detection paradigm were comparable to those obtained behaviorally in young adult CBA mice in that gap thresholds and their time courses of recovery reflected behavioral measures. Subsequently, they also reported that the majority of IC neurons from old CBA mice showed deficits in neural processing of gaps imbedded in noise carriers (Walton et al. 1998). This deficit was manifested as a shift in the frequency distribution of minimal gap thresholds where approximately 50% fewer "old" neurons had gap thresholds <3 ms, which is the normal behavioral threshold in the young adult mouse. In addition, nearly all phasic-type units displayed prolonged recovery times, as compared with units from young adults.

In the present study, the encoding of AM noise carriers by IC neurons, in young adult and old CBA mice, were compared. Although many of the fundamental response features to SAM noise remained stable with age, we found an age-related alteration in the rate encoding of SAM, and in the upper range of SAM frequencies to which units from old CBA mice could accurately phase-lock.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experimental procedures

Young adult (2- to 4-months-old) and old (24- to 28-months-old) CBA/CaJ mice were obtained from the National Institute of Aging and Jackson Lab mouse colonies and kept in an isolated, noise-controlled vivarium on a 12-h light:dark cycle. Food and water were provided ad lib. Prior to each experiment, animals were lightly anesthetized with Metofane (methoxyflurane, Pittman-Moore) and the external auditory meatus was examined down to the tympanic membrane for blockage. Only animals found to have clear external canals were used. Animals were prepared under aseptic conditions according to the guidelines for recovery surgery approved by the University of Rochester's Committee on Animal Resources. Prior to surgery, mice were deeply anesthetized with Avertin; the skull was shaved, and the cranium was exposed by reflecting the scalp musculature. Subsequently, a 2% solution of lidocaine was applied locally. The area was cleaned and dried, and a small, threaded metal tube was attached to the skull surface using cyanoacrylate glue (Superglue) and dental acrylic. A sharpened tungsten wire served as the indifferent electrode and was implanted in the skull and secured with dental acrylic. Using stereotaxic coordinates, a small (0.5 mm) hole was made in the skull over the IC and then filled with bone wax, after which the animal was allowed to fully recover and then was returned to its cage.

On the experimental day, the animal was mildly tranquilized (Taractan 5-12 µg/g) and placed in a plastic restraint attached to a custom-built stereotaxic frame. Typically, old mice were administered one-third to one-half the dose required for young animals. The frame was located in the middle of a heated (27-30°C), double-walled, sound-treated room (IAC), lined with sound-absorbing foam (Sonex). The head was fixed to the frame by bolting the threaded tube to a rigid bar attached to the stereotaxic frame. Care was taken to ensure that the animal was as comfortable as possible to avoid undue distress and body movement.

Stimulus generation

Stimuli were generated using a digital signal-processing platform (Tucker-Davis Technologies AP2) running on a Pentium PC. Broadband noise was digitally generated (2-60 kHz), amplified, and fed to a high-frequency leaf tweeter (Panasonic THD 100) located on the horizontal plane 30° contralateral to the recording site. Sound calibration was performed by sampling the noise using a calibrated 1/4 in. condenser microphone (Bruel and Kjaer model 4135) placed at the location of the pinna and connected to a measuring amplifier (Bruel and Kjaer model 2610). The output of the measuring amplifier was fed to an A/D converter, and, via inverse-fast Fourier transform (FFT) signal processing, a speaker's transfer function was equalized to within ±2 dB from 2 to 60 kHz. Tone bursts, noise bursts, and AM stimuli were synthesized on-line. SAM noise bursts were synthesized by multiplying a wideband noise carrier by a sine-wave modulator with m = 1, thereby producing 100% amplitude-modulated noise. SAM stimuli were 100 ms in duration with rise/fall times that followed the SAM envelope. Unmodulated stimuli were also 100 ms in duration with 1-ms linear rise/fall times. Stimuli were typically presented at 65 dB SPL. Occasionally, noise bursts were adjusted to other intensities between 25 and 80 dB SPL to present stimuli near the unit's best noise intensity. The modulation frequency (MF) was varied from 10 to 800 Hz and presented 50 times at a rate of 4/s.

Measurement protocol

After a neuron was isolated, the following experimental protocol was followed: 1) audio-visual determination of best frequency (BF) and threshold; 2) measurement of rate-intensity functions; 3) measurement of spontaneous activity over a period of 5.62 or 11.25 s; and 4) presentation of AM series. Minimum threshold (MT) was defined as the lowest intensity at which an increase in activity above the spontaneous rate was just noticeable. The intensity of the noise carriers never exceeded 80 dB SPL. Tones for rate-intensity functions were generated digitally at the audiovisually determined BF. Rate-intensity functions were obtained by presenting 25 repetitions of 100-ms-long tones or noise bursts with 1-ms rise/fall times at the rate of 4/s and at intervals of 5 dB from about 10 dB below the audiovisually determined MT to 80 dB SPL. This protocol was followed until the unit was lost or the signal-to-noise ratio of the spike waveform fell below 3:1.

Data analysis

Spikes evoked by stimuli were time-stamped (10-µs accuracy) and displayed on-line in the form of poststimulus time histograms (PSTHs). An automated analysis program was used off-line to obtain both synchronization and rate measures from AM files which have been routinely used in our other studies of AM processing in the auditory nerve and central auditory system (Frisina et al. 1996; Lesser et al. 1990). Phase locking to the SAM stimulus periodicity was quantified by computing the synchronization index (SI) for each period histogram (Goldberg and Brown 1969). To eliminate onset effects, initial spike responses, which occurred within 10 to 20 ms of stimulus onset, were not used in SI calculations. SI is the normalized estimate of a neuron's tendency to discharge at a particular phase in the modulation cycle. SI values can range from 0 to 1. A value of 1 indicates that all spikes fall in one bin, and 0 indicates an even distribution of spikes throughout the modulation period. The significance of the SI measure was tested by use of the Rayleigh statistic and only significant SI values, at the 0.05 level, were included in the analysis.

Spike counts refer to the total number of spikes recorded during the specified time window, usually equated to the stimulus duration. Spike count distributions include units where the SI was <0.3 and where the SI was nonsignificant (P > 0.05), as tested by the Rayleigh statistic. Meaningful comparisons of spike counts require that they reflect the same time interval. All total stimulus data are based on 5 s (50 repetitions each 100-ms duration) of stimulus presentation, unless otherwise noted, and reflect unit response to the onset and remaining portion of the stimulus. Spike responses to stimulus offset have been excluded. Spike counts for individual SAM cycles are based on unit responses related to the time interval, or period, of that cycle. For example, for a 100-Hz SAM stimulus, the first-cycle spike count includes all spikes occurring in the first 10 ms of a unit's driven response; the second cycle spike count includes all spikes occurring in the next 10 ms of the response, etc. Second and subsequent phasic-unit spike bursts are elicited in response to the SAM while first cycle spikes also reflect responses to stimulus onset. Spikes per stimulus values have been calculated by dividing the number of spikes elicited in 50 stimulus presentations by 50 and, consequently, reflect the probability of discharge to a single stimulus presentation. Spikes per second values in Fig. 9, A and B, have been calculated by dividing the measured spike count by the time interval over which it was measured. Spike counts and SI values were used to provide a quantitative description of the ability of a neuron to encode the envelope of the periodic waveform either by the strength of response (number of evoked spikes) or by phase-locking to envelope periodicities (SI value).

Histological verification

In addition to close inspection of specific response types according to area, horseradish peroxidase (HRP) microinjections were used to calibrate locations of recordings within the IC. HRP (10% Sigma type XII in 0.5 M KCl, 0.05 M Tris buffer, pH 7.3) was iontophoretically injected (electrode positive) using 1.5 µA constant DC for 15-20 min, into the center of the area of the IC in which recordings were made (Meininger et al. 1986; Walton et al. 1997, 1998). Animals were returned to their cage and perfused transcardially 24 h later with heparinized saline and fixed with glutaraldehyde/paraformaldehyde. Three serial sets of coronal sections were cut at 60 µm. Two sets were processed with tetramethylbenzidine (TMB), and one was counterstained with safranin-O. The third set was reacted with diaminobenzidine (DAB) and counterstained with cresyl violet. The centers of the injection sites were 500-990 µm in diameter and were confined to the central nucleus of the IC. These procedures were similar to our previous reports of structure-function HRP mapping studies in the IC of unanesthetized mammals (Frisina et al. 1989; Frisina and Frisina 1997).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Effect of age on fundamental response properties

Recordings were obtained from 87 single units from a total of 10 young adult mice and 58 single units from 9 old mice. To decrease the heterogeneity of the AM sample, only units with phasic responses to both tone and noise stimuli were included in the analysis, reducing the sample to 65 and 40 units from young adult and old mice or, 75 and 69% of typed units, respectively. BFs ranged from 5.3 to 54.1 kHz in units from young adult mice and from 4.3 to 46.2 kHz for old mice. Spontaneous activity ranged from 0 to 36 spikes/s in young adult units, and 0 to 25 spikes/s in old. There was no significant, age-related difference in the frequency distribution or means of spontaneous activity (Student's independent t-test, P = 0.2 and 0.49, respectively).

Thresholds were derived from the tone and noise rate-level functions. In young adult mice, single-unit thresholds ranged from 0 to 70 dB SPL for tones and 5 to 78 dB SPL for noise, with respective means of 18.2 and 35.3 dB SPL. In old mice, single-unit thresholds ranged from 5 to 65 dB SPL for tones and 5 to 80 dB SPL for noise, with respective means of 37.7 and 52.8 dB SPL. Minimum thresholds for both tone and noise stimuli were significantly elevated (P < 0.001) in old mice by 19.5 and 17.5 dB, respectively. These data are in agreement with previous studies reporting age-related threshold shifts of 20-30 dB across frequency for CBA mice and rats (Finlayson and Caspary 1993; Li and Borg 1993; Palombi and Caspary 1996a; Willott 1986).

Stimulus intensity and absolute (pa/s) rate of change

Given the age difference in noise thresholds noted above, the mean dB SPL of the noise stimuli applied to the old unit sample was greater than that applied to the young adult. However, because maximum stimulus intensity was limited to 80 dB SPL, the stimulus intensity above noise threshold (dB SL) applied to the old unit sample was less than that of the young adult. Mean stimulus levels for young adult and old distributions were 65.6 dB SPL, 30.4 dB SL and 71.0 dB SPL, 18.3 dB SL, respectively. Student's independent t-tests show P < 0.001 for both dB SPL and dB SL distributions. Median young adult values were 65.0 dB SPL, 30.0 dB SL with old medians of 75.0 dB SPL and 19.5 dB SL. Therefore, the mean and median stimulus levels, applied to the old unit sample, were greater by 5.4 and 10.0 dB SPL, respectively, but less by 12.1 and 20.5 dB SL, respectively, than those applied to the young adult unit sample. The absolute (SPL) intensity difference, noted above, results in a greater stimulus rate of change in the old sample. Mean stimulus rate of change, in pascals per second, is 1.86 times greater in the old than the young adult sample, while median rate of change is 3.16 times greater.

Age-related changes in rate encoding of SAM noise carriers

Many neurons from both young adult and old mice displayed strong phase locking and rate tuning to 100% SAM noise stimuli. This is illustrated in Fig. 1, where PSTHs are shown for rate-tuned neurons from a young adult (top) and an old (bottom) CBA mouse. Modulation transfer functions (MTFs), for spike count (rMTFs), are shown at the bottom right of both panels. Each unit displays an onset temporal discharge pattern to the unmodulated carrier. Both units have strong, rate-tuned responses with the young adult unit (top) peaking at an SAM rate, or MF, of 400 Hz, and the old unit (bottom) peaking at an MF of about 50 Hz. Note that the PSTHs illustrate three of the main findings. Units from older mice generally produce greater spike counts and greater phase-locked responses to multiple modulation cycles. Young adult mice exhibit stronger responses at higher MFs.

View larger version (20K): [in this window] [in a new window]   Fig. 1. Poststimulus time histograms (PSTHs) showing the response of 1 unit from a young adult mouse (3.5 months, BF = 16.1 kHz, top), and 1 unit from an old mouse (29 months, BF = 21.2 kHz, bottom), to sinusoidally amplitude-modulated (SAM) noise bursts having modulation frequencies (MF) of 20, 60, 80, 100, and 400 Hz. The solid bar beneath the absicissa denotes stimulus timing and duration. The complete rate modulation transfer function (rMTF), spike count plotted as a function of modulation frequency, is shown in the bottom right of each panel. Both units were classified as band pass with rBMFs of 400 Hz for the young adult unit and between 40 and 60 Hz for the old unit. In the young adult unit, a maximum synchronization index (SI) of 0.99 occurred at 20 Hz and remained above 0.9 through 60 Hz, while, in the old unit, a maximum SI of 0.95 occurred at 60 Hz and remained above 0.9 through 80 Hz. Both units were classified as on-sustained with minimum tone thresholds of 32 and 61 dB SPL for the young adult and old mouse units, respectively.

The shape of each unit's rMTF was analyzed with respect to its filter properties, resulting in the categorization as a low-pass (LP), band-pass (BP), high-pass (HP), all-pass (AP), or band reject (BR) unit. Figure 2A shows representative examples of four filter-shape categories, from four different units (2 young adult and 2 old), and Fig. 2B shows the young adult and old unit type distributions. Classification based on the filter characteristics of the rMTF shows that the distributions are similar in both sample groups (no statistically significant differences) with trends toward a lower percentage of LP and HP and a higher percentage of BP units in the young adult mice. The distribution of the MF eliciting the greatest spike count (rBMF) is shown in Fig. 2C. Young adult and old rBMF distributions are primarily BP, with most peaks in the 40-80 Hz MF range. The young adult distribution extends to higher MFs than the old, with 36.0% of units having rBMFs >100 Hz as compared with only 12.5% of the old unit sample. This difference is statistically significant (² = 8.4, P < 0.005).

View larger version (22K): [in this window] [in a new window]   Fig. 2. Characterization of age-related effects on rate encoding of envelope periodicities. A: examples of 4 different unit MTFs showing the classification scheme used to group rMTF filter shapes into 5 categories. Units were classified as low pass (LP), here with the maximum SI of 0.889; band pass (BP), here with the maximum SI of 0.939; high pass (HP), here with a maximum SI of 0.946; all pass (AP), here with a maximum SI of 0.703; or band reject (BR), example not shown. B: the proportion of young adult (open bars) and old (filled bars) units of each rMTF filter shape. C: the proportion of units from young adult (open bars) and old (closed bars) mice with rBMFs at each of the 9 MFs tested.

Age-related changes in phase-locking of envelope periodicities

Figure 3 summarizes the abilities of young and old units to phase lock to SAM noise carriers, as measured by the SI. MF distributions are shown for the minimum percentage of units with SI >0.3 and SI <0.3 in A (young adult), B (old), and, for the MF eliciting the highest, significant SI (sBMF) in C. SI distributions are shown at rBMF (D) and at sBMF (E). For MFs from 10 to 100 Hz, the percentage of young adult and old units with SI >0.3 are 78.2 ± 6.3 and 80.0 ± 5.0%, respectively, and, with SI <0.3 are 10.2 ± 2.1 and 7.5 ± 3.7%, respectively. These percentages drop significantly at MFs of 200 Hz and above. SI is uncertain (P > 0.05) in 10.2 ± 3.8% of young adult and 11.3 ± 3.7% of old units 100 Hz MF and increases rapidly at higher MFs. The distributions for sBMF are LP, with the proportion of units decreasing abruptly at MFs above 10 Hz in young adult and more gradually in old units.

View larger version (40K): [in this window] [in a new window]   Fig. 3. Characterization of age-related effects on phase locking of envelope periodicities of 65 units from young adult and 40 units from old CBA mice. A: stacked bar graph showing the proportion of units having SI >0.3 (open bar) and <0.3 (grayed bar) for units from young adult mice, as a function of MF. For example, at an MF of 20 Hz, just over 80% of the units had SIs >0.3, while ~10% had SIs <0.3. B: same as A, but for units from old mice, and filled bars show proportion of units with SI >0.3. C: the proportion of units from young adult (open bars) and old (filled bars) mice with the BMF for maximum SI (sBMF) at each of the 9 MFs tested. D: the proportion of units from young adult (open bars) and old (filled bars) animals, having SIs between 0 and 0.3, 0.3 and 0.5, 0.5 and 0.7, and 0.7 and 1.0, at the BMF for rate (rBMF). E: same as D except the SI was taken at the sBMF. All SIs were significant as tested by the Rayleigh coefficient.

Differences between rBMF and sBMF distributions

Comparing Fig. 2C and Fig. 3C, young adult rBMF and sBMF distributions are significantly different ( < 0.001), while those of the old are not ( = 0.065). In Fig. 3D, the age-related difference in the percentage of units with SI >0.7 at rBMF is not significant (² = 3.4, 0.050 < P < 0.075). While there are minimal age-related differences in SI distributions at sBMF, as shown in E, note the large percentage of units where SI exceeds 0.7 at one or more SAM rates. A includes data from 65 young adult and 40 old units. In C and D, data from two young adult units were not included because of nonsignificance (P > 0.05) at sBMF. In C, SI nonsignificance at rBMF reduced data by 21.5% to 54 young adult units and by 5.0% to 37 old units.

Spike-count response distributions to fifty 100-ms stimulus presentations

Neurons from old mice exhibited higher driven counts to both unmodulated and SAM noise stimuli than neurons from young adult mice. The box plots in Fig. 4 compare spike count distributions under four noise carrier conditions: 1) unmodulated; 2) SAM at 10 Hz; 3) SAM at rBMF; and 4) SAM at sBMF. Median counts for the unmodulated and 10-Hz SAM stimuli (A and B) in the old population are 1.66 and 1.77 times greater than young adult medians, respectively, with the difference at 10 Hz MF significant ( = 0.025). At 10 Hz, the old population, 75th percentile spike count, is noticeably greater than the young adult (255 and 140, respectively). However, the median spike counts to both unmodulated and 10-Hz SAM stimuli are comparable in the young adult (65 vs. 64), as well as in the old (106 vs. 103) sample. Median spike counts at rBMF and sBMF (C and D) are greater in the old units by 1.63 and 2.29 times those in young adult units, respectively, and the 75th percentile, old sBMF count, is 1.52 times the young adult.

View larger version (30K): [in this window] [in a new window]   Fig. 4. Box plots showing median values (horizontal lines), and 25th and 75th confidence limits are represented by the length of the box, of young adult and old sample spike count distributions, for 4 stimulus conditions: (A) unmodulated; (B) 10 Hz SAM; (C) rBMF SAM; and (D) sBMF SAM. The vertical lines represent the 5th and 95th confidence limits of the distributions.

Given the difference in the unmodulated spike count between neurons from young adult and old mice, an important question is whether the SAM response might be predicted by the unmodulated response. Figure 5 shows scatterplots of each unit's response to an unmodulated stimulus plotted against its response to rBMF (A and B) and 10-Hz SAM stimuli (C and D). As the total stimulus duration is 100 ms, the 10-Hz SAM stimulus represents a single sine-wave cycle signal with 50 ms rise and fall times. The main observation is that the slopes of the linear regressions for spike count response to rBMF and 10-Hz SAM stimuli are significantly different for the young adult units (1.38 vs. 0.61) and nearly identical for the old units (0.82 vs. 0.83). Consequently, in old units the change in activity is the same for 10-Hz and rBMF stimuli, while in young adult units, the change is approximately twice as great for rBMF than the 10-Hz stimuli. This difference may be viewed as a depression in spike coding at rBMF in units from old, as compared with young adult, mice. Regression results for the 10-Hz SAM data (C and D) yield similar explained variation (r²) for young adult and old data (0.65 vs. 0.60) and similar spike counts at the y-intercept (30 vs. 38.5). Regressions for rBMF response in both populations yield lower r² values (0.47, young adult; 0.35, old), and greater spike counts at an unmodulated response of 0 (95, young adult; 140, old).

View larger version (26K): [in this window] [in a new window]   Fig. 5. Scatterplots showing the relationship between driven spike count to the unmodulated carrier and the driven spike count at the rBMF (top panels), and at an MF of 10 Hz (bottom panels), for units from young adult (A and C) and old (B and D) CBA mice. Linear regressions were computed for each distribution. Predicted, driven spike count response to SAM stimuli at rBMF are as follows: spike count at rBMF = 94.4 + 1.38 (unmodulated spike count) for young adult units (A); and spike count at rBMF = 141.4 + 0.82 (unmodulated spike count) for old units (B). Predicted total spike count response to SAM stimuli at 10 Hz MF are as follows: spike count at 10 Hz = 29.7 + 0.67 (unmodulated spike count) for young adult units (C); spike count at 10 Hz = 38.7 + 0.82 (unmodulated spike count) for old units (D).

Total spike count distributions, for neurons from young adult and old mice, to SAM stimuli with MFs 200 Hz, are shown in the box plots of Fig. 6A. The median spike counts for old neurons are significantly greater than the young adult at all MFs, except 200 Hz (Mann-Whitney U test, P < 0.05). Old, median total counts are more than twice that of young adult total counts for MFs from 20 to 100 Hz and >1.5 times for unmodulated and 10-Hz stimuli.

View larger version (24K): [in this window] [in a new window]   Fig. 6. Three different presentations comparing rate encoding of SAM noise stimuli, as a function of modulation frequency, for units from young adult and old mice are shown. A: box plots showing the distribution of the total number of spikes to 50 stimulus presentations, at each MF, for young adult (open bars) and old (filled bars) samples. The dark horizontal line in each bar represents the median value. Total counts were significantly different between the young adult and the old samples (Mann-Whitney U test, P < 0.05) for MFs from 20 to 100 Hz. B: median, probability of discharge values (spikes/stimulus), plotted as a function of MF, for the young adult () and old () units. C: same as B except that the median values have been normalized to their respective, maximum values, at an MF of 60 Hz.

Rees and Moller (1983) have shown that spike responses of IC units to SAM stimuli increase with intensity over the first 20-30 dB SL, but begin to saturate and, in some units, decrease markedly above this level. To test if young adult population responses were depressed by inclusion of units operating in a saturated or nonmonotonic region, young adult spike count statistics were recalculated using only the 35 units with stimulus level of 30 dB SL or less. Response medians were nominally lower than those of the 65-unit population for all stimuli except at 200 Hz MF, where it increased by 42%. Mean response to the unmodulated stimulus increased by 8.6%, while mean responses to all other stimuli decreased by 0.0 to 14.0% of the relevant, 65-unit population value. Therefore, with the possible exception of 200 Hz MF, the lower spike counts elicited in the young adult (compared with the old) sample do not appear to be a result of their greater stimulus levels.

Median spike response per stimulus presentation

Figure 6B plots median spike count per stimulus presentation and highlights the greater magnitude and variability with MF of old unit median spike count. For all stimuli, median spike response per stimulus presentation is 3.6 ± 42% in the old and 1.7 ± 23% in the young adult sample. The median response is greatest in both young adult and old units at an MF of 60 Hz. In Fig. 6C, median responses are normalized to the 60-Hz MF response. Both young adult and old responses vary similarly at MFs from 10 to 100 Hz. At MFs of 200 Hz, the old median response is 37.5% lower, and the young adult response is approximately equal to their respective 60-Hz levels.

Onset versus steady-state encoding of SAM in young and old samples

In Fig. 7, median spike counts are divided into two count windows: the total number of spikes elicited by the first SAM cycle of the stimulus, which includes the onset response, referred to as the transient response (A); and the response to the remaining SAM cycles of the stimulus, which we refer to as the periodic response (B). The total number of spikes evoked by the unmodulated stimulus, and the 10-Hz SAM stimulus, were considered transient responses. In all the phasic units included in this analysis, the transient response to stimulus onset was an abrupt burst of spike activity, lasting for only a fraction of the first SAM cycle. Median probability of discharge of the transient response, plotted against the MF, is shown in Fig. 7A. Note that the probability of discharge in units from old mice are significantly greater than that of the young adult sample from 20 to 100 Hz MF (Mann-Whitney U test, P < 0.05). In both young adult and old units, the transient spikes/stimulus count decreases as the MF increases and is explained quite well by the logarithm of the MF. Regression fits for young adult and old median transient responses, as a function of the logarithm of the MF, shown in A, have explained variations (r²) of 0.954 and 0.908, respectively. The median probability of discharge for the periodic response (B) is also greater in the old population and significantly so at MFs of 20, 40, and 80 Hz (Mann-Whitney U test, P < 0.05). Here, however, the probability of discharge for the old sample is two to three times greater than the young sample up to 100 Hz and then drops to the young adult level at 200 Hz. Interestingly, both young adult and old functions peak in the MF region of 60 to 80 Hz, with the young adult response remaining relatively constant up to an MF of 200 Hz.

View larger version (21K): [in this window] [in a new window]   Fig. 7. Probability of discharge decreases systematically as a function of modulation frequency for the first SAM cycle, or transient response (A), but not for the remaining SAM cycles, or periodic response (B). A: median spikes/stimulus response plotted as a function of MF for the young adult () and old () samples, to the first SAM cycle of the stimulus, denoted as the transient response, and then fit with negative log functions. The r2 for the young adult fit was 0.91 and for the old fit was 0.95. B: same as A, but for the periodic response, and without regression lines.

In Fig. 7 we showed that the number of spikes measured by the transient and periodic responses to SAM noise varied in both young adult and old IC neurons as a function of MF. To directly compare the effects of aging on the transient and periodic responses, we computed a spike ratio, normalized to the median old data set. In Fig. 8A, the spike ratios of old median to young adult median transient and periodic responses are plotted as functions of MF. A value of 1.0 signifies that the young and old sample medians were the same at a given MF. The transient response ratio remains constant at 1.62 ± 10% for both unmodulated and 10- to 200-Hz SAM stimuli. The periodic response ratio decreases linearly with increasing MF with an r² = 0.900. 

View larger version (21K): [in this window] [in a new window]   Fig. 8. A: spike ratios were calculated by normalizing the young adult median values, at each MF, to the corresponding old median. A value of 1.0 indicates the medians were equal and, for example, a value of 2.0 would indicate that the old median value was twice that of the young. The spike ratios are plotted as a function of MF, for the transient component (), and for the periodic component () of the SAM response. B: spikes/stimulus response to each SAM cycle of the stimulus, illustrating the difference between the median young adult (open symbols) and old (filled symbols) responses to each period of modulation, from 40 Hz (, old; , young) to 200 Hz (, old; , young) MF.

The age-related difference in the number of spikes elicited by the first cycle of modulation, compared with the last cycle, also varies as a function of MF and is illustrated in Fig. 8B. For both young adult and old samples, Fig. 8B shows median spikes per stimulus, evoked by each SAM cycle, for three of the nine MFs. Note that SAM cycle duration varies with MF. The spike response to the 40-Hz stimulus was measured in four 25-ms periods, while responses to 80- and 200-Hz stimuli were measured in eight 12.5-ms and twenty 5-ms periods, respectively. After the first SAM cycle, there is a systematic decrease with increasing MF in the degree to which age affects the spike response to each subsequent cycle of modulation. The old sample, however, displays a greater median spike response, when compared with the young adult, at all MFs except for 200 Hz where the response, after the first (transient) cycle, is nearly equal. This data show that, up to 100 Hz MF, the larger median periodic responses noted previously in old units also apply on a modulation cycle-by-cycle basis. Consequently, old units are discharging at a significantly greater rate than young adult units throughout the entire stimulus response interval.

Figure 9A shows the transient response from Fig. 7A re-computed as an average spike rate and plotted versus MF. The transient response duration was always limited to one SAM cycle, or modulation period (MP), of the stimulus. Consequently, the percentage of total stimulus response time represented by the transient response varied from 100% (100 ms) at 10 Hz MF, to 5% (5 ms) at 200 Hz MF. In Fig. 9A, average transient response rates are re-computed by dividing the transient responses, from Fig. 7A, by the corresponding MP. Both young adult and old responses increase linearly with MF (r² = 0.98). Since the transient response consists of a brief spike burst, lasting only a fraction of the MP, the actual median transient response rate is not constant over the interval, but peaks at a value greater than the average shown in A. Periodic response measurement duration also varies with MF and is equal to 100 ms less the transient response duration (one MP). The average periodic response rate is calculated in B by dividing the median periodic response, from Fig. 7B, by the time duration over which the response is measured. The periodic response rate remains relatively constant with MF in the young adult sample but decreases linearly (r² = 0.87) with MF in the old sample. Consequently, the periodic response rate is independent of MF in the young sample and decreases by almost 66% as MF goes from 80 to 200 Hz in the old sample. This compares favorably to the cycle-by-cycle analysis shown in Fig. 8B, which illustrates that, after the first SAM cycle, the difference in cycle-by-cycle median response, between young adult and old samples, decreases with MF, becoming equal at 200 Hz.

View larger version (18K): [in this window] [in a new window]   Fig. 9. Median spike count data have been re-computed as an average spike firing rate. A: transient spike rates increase with increasing MF for both young adult () and old () samples, and both are fit well by a linear regression (r2 = 0.98 for young and r2 = 0.98 for old). B: the median periodic spike rate of the old sample decreases linearly with increasing MF (r2 = 0.87), while the young adult spike rate remains constant.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A long-range objective of our research is to examine neural processing of complex sounds and to uncover the biological correlates of presbycusis using both simple and complex acoustic signals. The present study was prompted by previous results that showed age-related alteration in neural correlates of gap detection by IC neurons (Walton et al. 1998). The present report identifies differences in response, between young adult and old IC phasic neurons, to 100% SAM broadband noise stimuli, over an MF range of 10-800 Hz.

It is well known that non-age-related variation in single-unit response can result from differences between sample groups, including location within the IC, response type (tonic/phasic), binaural factors, and rMTF distribution. Young adult and old unit samples characterized here are closely matched in all, non-stimulus-dependent dimensions, except for trends in rMTF response type and distribution. To statistically compare these two data sets, we used conservative, nonparametric techniques, as both distributions were skewed (Sokol and Rohlf 1995). Results from the current study indicate subtle, yet consistent, age-related alteration in neural encoding of amplitude-modulated noise. We found that aging disrupts the upper range of rate encoding of SAM, and that overall, there is an age-related degradation in the cycle-by-cycle coding of SAM.

Multiple measures indicate an age-related effect on rate encoding of SAM signals

We observed several important, age-related changes in the encoding of SAM noise based on spike count measures. First, we observed that both the total and the transient responses to SAM noise stimuli, with MFs from 10 to 100 Hz, were significantly greater in units from old CBA mice. The increased responsiveness of old units is consistent with findings of some studies that have found increased spontaneous rates (Willott et al. 1988). Unlike the median transient response, there is a marked, age-related difference in the variation of median periodic response with MF as shown by the linear decrease in the ratio of median old periodic response to median young adult periodic response with increasing MF (see Fig. 8A). This decrease results from the fact that the young adult, median periodic response rate remains relatively constant with increases in MF, while the old median response rate decreases markedly, as shown in Fig. 9B. It is interesting that the largest difference between the young and old samples is in the range of MFs where most units also have their rBMF. Possible causes of this behavior include a differential aging of the neural mechanism controlling the periodic response and the transient response and a decreased ability of old units to periodically discharge in the same cycle-by-cycle manner as young adult units at higher MFs, due to their greater transient response.

Palombi et al. (2001), in the only other study of age-related alteration in neural coding of SAM signals in the IC, found that discharge rate was stable with age at the maximum rBMF for tone carriers. They observed that the maximum discharge rates were 19.4 spikes/s for units from young rats as compared with 16.9 spikes/s for old. Conversion of median values reported here at rBMF to spikes/s rates results in 28 spikes/s for the young adult sample, as compared with 50 spikes/s for the old sample. Methodological differences between the two studies, such as differences in anesthetic, SAM carrier and level, may have contributed to the differences in the observed results. We used wideband noise as the SAM carrier in the current study to permit future comparison with single-unit correlates of gap detection using noise markers. The use of noise carriers also removes the contribution of differences in spectral integration between units from consideration. ON-BF versus OFF-BF age-related changes in the interplay between excitation and inhibition also may contribute to the observed differences in rate coding of SAM tones and noise in the two studies.

Age-related decrease in coding high-frequency envelope periodicities

The selectivity to certain modulation frequencies as encoded by variation in discharge rate or phase-locking by neurons in the auditory system is typically represented by the MTF. With the following exception, we found similar distributions of MTF types for both rate and synchronization measures of SAM encoding in young adult and old samples. There was a significant, age-related difference in sample distributions for rBMF. rBMFs above 100 Hz were more prevalent in the young adult sample than in the old sample, and the 75th quartile upper limit decreased from 200 Hz in the young adult to 70 Hz in the old sample. Palombi et al. (2001) did not characterize the actual rBMF of each unit, but did find that the filter characteristics between young and old samples were comparable for tonal carriers (Fig. 3).

Age-related differences in rMTF type may be related to carrier intensity

Several reports have shown that when carrier intensity is increased from near threshold to 30-50 dB above MT, the MTF filter characteristics of IC neurons can change from low pass to band pass (Langner and Schreiner 1988; Rees and Moller 1987). The lower percentage of LP and higher percentage of BP units reported here in the young adult sample, as compared with the old sample, are consistent with the observation that the carrier intensity was on average greater in the young adult sample. A similar transition in receptive fields, from LP to BP filtering characteristics with increasing stimulus level, has been documented in the visual system. Retinal receptive field organization changes with adaptation to increasing light intensity (Atick and Redlick 1990). Here, the change has been linked to minimization of retinal coding redundancy in response to the increase in signal-to-noise ratio resulting from increased light intensity (Atick 1992). It is possible that a stimulus level dependent change in auditory processing of SAM stimuli, at the IC, might reflect a similar, stimulus signal-to-noise ratio-dependent, phenomena. However, as the difference in the median SLs for carrier intensity between the young and old samples was on the order of 10 dB, and all carrier intensities were 65 dB SPL or greater, we believe that the aging effects we observed are not strongly influenced by differences in presentation level.

Other factors affecting spike response

Several studies have shown that SAM stimulus level above threshold can also affect spike count response (Langner and Schreiner 1988; Rees and Moller 1987). Typically, spike counts increase as stimulus level increases from 0 to 30 dB SL, remaining relatively constant or declining at higher stimulus levels. As mentioned previously, the mean and median spike counts for the 35-unit, young adult sample, with stimulus levels of 30 dB SL or below, were less than those of the 65-unit sample, except at 200 Hz where the median was 42% greater. Consequently, with stimulus levels within the monotonic spike response region for both sample groups, the old sample spike response levels remained significantly greater than those of the young adult sample at MFs 100 Hz.

Effects of rise time on probability of discharge

The effects of stimulus rise/fall time and rate of change (Pascals/s) on spike count are interdependent and highly nonlinear. Over certain ranges, reduced rise time or increased rate of change of tone stimuli have been shown to increase spike response in single neurons within the auditory cortex of cats (Heil 1997a). Stimulus rise/fall times used here are identical in young adult and old samples. Average rise/fall times decrease linearly from 50 to 2.5 ms as MF increases from 10 to 200 Hz. The variation in young adult and old sample spike counts with MF, as shown in Figs. 6, 7, and 8B, do not exhibit the increase with MF indicative of a rise/fall time affect. Comparison of spike response to unmodulated and 10-Hz SAM stimuli provides a further indication of the lack of rise/fall time influence on spike count in our data. As shown in Fig. 6B, median responses to these two stimuli are almost identical within each age sample, even though the rise/fall time of the unmodulated stimulus is 1 ms and that of the 10-Hz SAM stimulus is 50 ms.

Co-varying stimulus factors

The processing of SAM sounds in the IC is governed by temporal interplay of excitation and inhibition and by stimulus changes that co-vary with the MF. Such factors may have an effect on observed spike counts (Heil 1997b; Phillips 1988). The following additional temporal factors which vary with MF in the stimuli used here, and which may influence spike count, include: the time duration of the transient response (MP); the time duration of the periodic response (100 - MP); and, the number of SAM cycles in the periodic response (0.1 × MF  1). The relationship between individual unit spike response to unmodulated and 10-Hz SAM stimuli in young adult and old samples, as established by the linear regression correlation coefficients of 0.62 and 0.82 (Fig. 5, C and D), respectively, may be attributable to the 50 times slower average rise time of the 10-Hz SAM stimulus compared with that of the unmodulated stimulus (50.0 ms vs. 1.0 ms).

The difference in median noise thresholds between young and old samples results in a threefold increase in Pascal per second rate of change of the carrier envelope in the old sample, and this might account for the larger correlation (0.82) between the 10-Hz SAM and unmodulated stimuli spike response in the old, compared with the young adult (0.62), sample group. Consequently, the difference between the responses observed in individual, young adult and old units, when stimulated with unmodulated versus 10 Hz SAM stimuli, may result from their age-related threshold difference.

Age-related loss of inhibition and neural correlates of age-related changes in temporal processing

Auditory information from lower binaural centers converge onto target neurons in the IC in a complex spatial and temporal interplay of excitation and inhibition (Covey et al. 1996; Palombi and Caspary 1996a; Pollak and Park 1993). Moreover, GABAergic circuits are known to play a central role in shaping response properties of auditory midbrain neurons to AM (Burger and Pollack 1998; Palombi and Caspary 1996b; Yang and Pollak 1997). The data of Yang and Pollack (1997) support the hypothesis that acute blockade of GABAergic circuits can lead to changes in encoding of AM in dorsal nucleus of the lateral lemniscus (DNLL) neurons. Blockade of -aminobutyric acid-A (GABA_A) receptors with bicuculline resulted in an increase in the periodic response to AM. Data presented in Fig. 14 of Yang and Pollack (1997) show increased cycle-by-cycle spike counts on the order of 1.5 to 2.2 times for both unmodulated and 50 to 200 Hz tone SAM carriers. This compares favorably to our data showing greater median total spike counts (1.3-2.5 times) for units from old mice as compared with young adult mice.

To a large extent, the ability of IC neurons to follow, and even sharpen, neural encoding of rapidly changing acoustic signals must play a role in the perception of AM (Eggermont 2001; Langner and Schreiner 1988). It is plausible that the age-related changes in IC neurochemistry could result in changes in overall gain. This hypothesis is supported by data from Caspary and colleagues, which have demonstrated age-related alterations in inhibitory neurotransmitter synthesis and receptor properties in the rodent IC (Caspary et al. 1995; Gutierrez et al. 1994; Milbrandt et al. 1994, 1997; Raza et al. 1994). They have shown, in Fisher 344 rats, an age-related decline of over 30% in both GABA_A and GABA_B immunoreactive neurons. Furthermore, these changes appear to be compensated, in part, by changes in receptor subunit expression which result in increased ionic flux of chloride in the aged IC (Caspary et al. 1999). Compensatory changes are again supported by the data of Helfert et al. (1999), which reports an age-related decline in GABA-positive and -negative synaptic terminals, suggesting a proportional decrease in both inhibitory (GABA) and excitatory (glutamate) inputs to principle cells in the IC of aged rats. Age-related changes in sensitivity at the auditory periphery are certainly responsible for the decline of audibility, but the deficits in the neural encoding of static and dynamic signals may lead to a better understanding of the age-related deficits in temporal acuity observed in human listeners.