Dynamic representation of spectral edges in guinea pig primary auditory cortex

Abstract

The central representation of a given acoustic motif is thought to be strongly context dependent, i.e., to rely on the spectrotemporal past and present of the acoustic mixture in which it is embedded. The present study investigated the cortical representation of spectral edges (i.e., where stimulus energy changes abruptly over frequency) and its dependence on stimulus duration and depth of the spectral contrast in guinea pig. We devised a stimulus ensemble composed of random tone pips with or without an attenuated frequency band (AFB) of variable depth. Additionally, the multitone ensemble with AFB was interleaved with periods of silence or with multitone ensembles without AFB. We have shown that the representation of the frequencies near but outside the AFB is greatly enhanced, whereas the representation of frequencies near and inside the AFB is strongly suppressed. These cortical changes depend on the depth of the AFB: although they are maximal for the largest depth of the AFB, they are also statistically significant for depths as small as 10 dB. Finally, the cortical changes are quick, occurring within a few seconds of stimulus ensemble presentation with AFB, and are very labile, disappearing within a few seconds after the presentation without AFB. Overall, this study demonstrates that the representation of spectral edges is dynamically enhanced in the auditory centers. These central changes may have important functional implications, particularly in noisy environments where they could contribute to preserving the central representation of spectral edges.

an important question in auditory neuroscience addresses the understanding of how acoustic stimuli are represented in the central auditory system (Escabí and Read 2005; Nelken 2004; Oswald et al. 2006; Reyes 2011; Sutter 2005; Winer et al. 2005). This process is particularly challenging because sounds in the natural environment (such as conspecific vocalizations) have complex spectra that change rapidly with time and are often embedded in background noise (Palmer and Shamma 2004; Shamma and Micheyl 2010; Singh and Theunissen 2003; Theunissen and Elie 2014). Although we are far from a complete understanding of how complex sounds are represented in the central auditory system, it appears that the representation of many sound features is context dependent, i.e., it depends on the spectrotemporal past and present of the acoustic mixture (Blake and Merzenich 2002; Brosch et al. 1999; Brosch and Scheich 2008; Brosch and Schreiner 2000; Catz and Noreña 2013; Gourévitch et al. 2009; Noreña et al. 2008).

The sensitivity of central auditory neurons to the spectrotemporal history of the acoustic environment has traditionally been examined using simple stimulus paradigms such as two-tone sequences (Brosch et al. 1999; Brosch and Scheich 2008; Brosch and Schreiner 1997, 2000; Kadia and Wang 2003; Rhode and Greenberg 1994; Shamma and Symmes 1985). Namely, a “target” tone presented at a neuron's best frequency (BF) is preceded by a “modulating” tone at variable frequency, sound pressure level (SPL), and interstimulus interval (typically between 0 and a few hundred milliseconds). Such studies showed that neural responses to the target tone could be strongly suppressed, or in some cases enhanced, by the prior presentation of the modulating tone; that is, they could be strongly context dependent. Other studies derived neural properties, such as spectrotemporal receptive fields (STRFs), from a variety of more complex stimuli, such as white noise, spectrotemporally modulated (“rippled”) noise, ensembles of random tone bursts, and conspecific vocalizations (Blake and Merzenich, 2002; Catz and Noreña 2013; deCharms et al. 1998; Depireux et al. 2001; Eggermont et al. 1983; Escabí and Read 2005; Gourévitch et al., 2009; Noreña et al. 2006, 2008; Rabinowitz et al. 2011, 2012; Theunissen et al. 2001; Theunissen and Elie 2014; Valentine and Eggermont 2004). It was shown that although the shapes of STRFs are relatively insensitive to some stimulus characteristics (i.e., SPL; Pienkowski and Eggermont 2010; Valentine and Eggermont 2004), they are sensitive to others (i.e., stimulus spectrotemporal density; Blake and Merzenich 2002; Noreña et al. 2008).

Animal vocalizations, including human speech, contain sharp spectral peaks and transitions (i.e., vowel formants) that differentiate them from each other and from the noise background (Assmann and Summerfield 2004; Baer et al. 1993; Moore and Glasberg 1983; Palmer and Shamma 2004; Summerfield and Assmann 1989). We recently investigated the auditory cortical representation of spectral edges or contours, and how this representation varies with their physical properties (i.e., width, depth, and slope) (Catz and Noreña 2013). To investigate the respective representation of frequencies near to and remote from the spectral edges, we devised a stimulus ensemble composed of random tone bursts (between 500 Hz and 32 kHz, ⅛ octave step) with or without an attenuated frequency band (AFB), within which tone levels are reduced to create edge contrasts. It was shown that the representation of tone frequencies adjacent to the AFB is enhanced, whereas the representation of frequencies in the AFB, especially near the spectral edges, is reduced. These cortical changes are very sensitive to the characteristics of the spectral notch. These results are broadly consistent with the general idea that the representation of acoustic features is context dependent and suggest that some mechanisms act to enhance the central auditory representation of spectral edges (Gourévitch et al. 2009; Shamma et al. 2011; Von Békésy 1967, 1969).

The present study is aimed at extending the results of our previous work. First, we further investigated the effects of spectral notch depth to determine whether edge enhancement could be seen for spectral contrasts as small as 10 dB. Second, the edge enhancements described in our previous study (Catz and Noreña 2013) were observed after a 180-s presentation of the AFB stimulus. This is relatively fast but still too slow to dynamically shape the representation of real acoustic environments, which can vary on much smaller (fractions of a second) timescales (Singh and Theunissen 2003). We thus investigated whether edge enhancement also occurs during a much shorter (2–10 s) presentation of the AFB stimulus. Third, it was unknown whether edge enhancement persisted for some time after the presentation of the AFB stimulus. This question was addressed by interleaving 10-s AFB stimulus presentations with 10-s periods of the control multitone stimulus (without AFB). Fourth, the edge enhancement reported in our previous study was observed in anesthetized animals. To assess the effects of anesthesia on the neural changes at spectral edges, we carried out a subset of experiments in a few awake animals.

METHODS

Animal preparation.

The care and use of animals in this study were approved by the Animal Care Committee of Bouches du Rhones, France (#A 13-504). A total of 18 guinea pigs weighing between 400 and 800 g were used for this study. Animals were deeply anesthetized with the administration of 50 mg/kg ketamine hydrochloride (Imalgene 1000) and 3 mg/kg of xylazine (Rompun; 2%), injected intramuscularly; 0.1 ml of Atropine methyl nitrate was also administered to reduce mucus secretion in the throat. Throughout the experiment, anesthesia was maintained with half the dose of ketamine and xylazine administrated every hour. The tissue overlying the frontal lobe was opened, four small screws were fixed to the top of the skull, and dental cement was used to fix a large screw. This screw provided a mechanically stable point of attachment to a metal head post. The tissue, skull, and dura overlying the primary auditory cortex were removed over a surface of roughly 1 cm² (see Grimsley et al. 2012 for precise location). The body temperature was maintained at around 37°C with a thermostatically controlled heating blanket. After the experiment, a lethal dose of pentobarbital sodium was administered.

Three additional animals were implanted chronically with microelectrodes to investigate whether cortical changes reported in anesthetized animals were also present in the awake preparation. In these animals, a cylindrical plastic chamber (9-mm diameter and 7-mm height) was fixed on the skull with dental cement. The center of the chamber was positioned roughly on the center of the craniotomy. The microelectrodes were then advanced perpendicularly to the cortex through the chamber and inserted into the primary auditory cortex. The microelectrodes were fixed to the chamber with a silicone sealant. The animals were allowed to recover for 1 wk before cortical recordings began. The animals' heads were fixed using the same method described above for anesthetized preparation.

Acoustic stimulation.

Stimuli were generated in MATLAB and transferred to an RP2.1-based sound delivery system (Tucker Davis Technologies). Acoustic stimuli were presented in a sound booth room from a headphone (Sennheiser HD595) placed 10 cm in front of the ear contralateral to the cortex where the recordings were carried out. The amplitude of each tone pip was adjusted to the transfer function of the sound delivery system so that they were presented at the desired level in decibel sound pressure level (dB SPL).

STRFs were obtained from 180-s multitone pip stimuli (Blake and Merzenich 2002; Catz and Noreña 2013; deCharms et al. 1998; Gourévitch et al. 2009; Noreña et al. 2008; Valentine and Eggermont 2004). Tone pips (49 frequencies, 8 frequencies per octave covering 6 octaves between 500 Hz and 32 kHz) were presented randomly over time (independent Poisson process for each frequency with a rate of 2 Hz and a 50-ms dead time designed to prevent tones of the same frequency from overlapping in time). However, tone pips of different frequencies could occasionally overlap in time. The envelope of the tone pips is given by γ(t) = (t/4)²e^−t/4 with t in milliseconds (the stimulus duration is 50 ms, and the maximum amplitude is reached at 8 ms). The average rate of tone pip presentation was around 16 Hz/octave (considering the number of tone frequencies present per octave, along with the average presentation rate of each). This pip rate per octave of the stimulus was chosen as a trade-off between spectrotemporal density and neural adaptation (to preserve cortical activity) (Blake and Merzenich 2002; Noreña et al. 2008). Control STRFs were obtained from multitone stimuli with tone pips presented at 70 (ctrl-70), 60 (ctrl-60), 50 (ctrl-50), and 40 dB SPL (ctrl-40) (Fig. 1A).

Fig. 1.

Schematic representation of the acoustic stimuli used in the present study. A: time onsets (black dots) and long-term spectrum of control (ctrl; left) and attenuated frequency band (AFB) stimuli (right). B: time onsets (black dots) of tone pips for stimulus sequences alternating control stimulus and silence (ctrl-sil), AFB stimulus and silence (AFB-sil), or control stimulus and AFB stimulus (ctrl-AFB).

A first experiment was carried out to study the effects of spectral contrast depth on the representation of acoustic spectral edges. Multitone pip stimuli (as described above) with an AFB were used: all pure tones were presented at 70 dB, except those corresponding to the frequency band of the AFB where pure tones were omitted (AFB 70-0) or presented at 40 (AFB 70-40), 50 (AFB 70-50), or 60 dB SPL (AFB 70-60) (Fig. 1A). The frequencies within a half octave above or below the AFB were called the edge-out frequencies. On the other hand, the frequencies within a half octave above or below the lower or upper edge, respectively, were called the edge-in frequencies. In this experiment, the width of the AFB was fixed to 1 octave. The center frequency of the AFB was set as follows: the BF for each cortical site was derived from the control stimulus (ctrl-70). The center frequency of the AFB was then set to the BF of a given cortical site. Cortical responses were obtained for all stimulus conditions (different depths) for that specific center frequency of the AFB. Once a set of recordings was completed, another set of recordings was carried out with a different AFB stimulus (centered on the BF of another cortical site). One notes that because we recorded from many cortical sites simultaneously, the BFs could correspond to the center frequency of the AFB, an edge frequency of the AFB, or a remote frequency from the AFB.

In a second experiment, we investigated whether the cortical changes produced by multitone pip stimuli presented for 180 s could also be induced by a shorter presentation (10 s). Three stimuli were constructed to address this question. A control stimulus consisted of an alternation of 10-s presentation of ctrl-70 and 10-s silence (ctrl-sil) (Fig. 1B). One test stimulus consisted of an alternation between 10-s AFB 70-0 stimulus and 10-s silence (AFB-sil). This test stimulus was specifically designed to address the cortical changes during the 10-s presentation of the stimulus with AFB. The other test stimulus consisted of an alternation of 10-s AFB 70-0 stimulus and ctrl-70 (ctrl-AFB). This test stimulus was specifically designed to address whether the cortical changes produced by the stimulus with AFB produce some persistent effects, i.e., during the following presentation of the ctrl-70 stimulus. In this experiment, two widths of the AFB (½ or 1 octave) were investigated.

Multiunit activity and local field potential recording procedure.

Each set of recordings was obtained with 1 array of 16 electrodes (Alpha-Omega, Nazareth, Israel) arranged in an 8 × 2 pattern with 0.25-mm electrode separation within the long row and 0.5-mm separation between columns. The electrodes had impedances between 0.8 and 1.4 MΩ. The array was manually advanced into the primary auditory cortex by using a Narishige microdrive (Wallace et al. 2000). The signals were then amplified 10,000 times with filter cutoff frequencies set at 2 Hz and 5 kHz. The amplified signals were processed by a TDT-System 3 multichannel data acquisition system. Multiunit activity (MUA) was sampled at 24,414 Hz and was extracted from the 300-Hz high-pass filtered signal. Local field potentials (LFPs) were sampled at 1,061 Hz and were extracted from the 300-Hz low-pass filtered signal. In this way, we were able to record spikes and LFPs simultaneously.

At an initial stage of the experiments, a “search procedure” was used that consisted of recording cortical activity induced by clicks, noise bursts, and tone pips (from 500 Hz to 32 kHz, ⅛ octave step). This procedure provided a rough estimate of the tonotopy and the amplitude of LFPs. Electrodes were placed at a depth where the (negative) amplitude of stimulus-induced LFPs was near maximal (region of the border between layer III and IV; Szymanski et al. 2011).

Data analysis.

All results were computed using custom MATLAB routines. MUA or “spike events” were detected by using an amplitude threshold on the high-pass filtered data. The median was calculated on the negative values of the filtered signal; the threshold was then set to six times the median (Quiroga et al. 2004).

The methodology for computing STRFs was similar to that used in previous studies (Catz and Noreña 2013; Noreña et al. 2008; Valentine and Eggermont 2004). Briefly, STRFs for MUA were determined by constructing poststimulus time histograms (PSTHs) with time bins of 1 ms for each tone pip frequency. In other words, spikes falling in the averaging time window (starting at the stimulus onset and lasting 100 ms) are counted. Because the average interstimulus interval in the stimulus ensemble (∼10 ms) is smaller than the averaging time window, a spike can be counted in the PSTH of several pip frequencies. STRFs for LFPs were obtained by using a similar procedure, except that the LFP waveforms (0–100 ms after stimulus onset) were averaged for each tone pip frequency. The buildup and breakdown of the neural changes produced by AFB stimulus were investigated by alternating 10-s AFB 70-0 stimulus with control stimulus or silence. STRFs were obtained by averaging PSTHs or LFP waveforms over the time periods of interest (either control or AFB). Moreover, the 10-s stimulation periods (AFB or control) were divided into three time periods, i.e., 0–2.5, 2.5–5, and 5–10 s, and STRFs were derived from each of these time windows. This more detailed analysis of the neural changes produced by AFB stimulus was made possible by the fact that nine samples of each multitone ensemble were available to derive the STRFs.

The maximal MUA response (or the minimal LFP amplitude) within the 10- to 30-ms time window after stimulus onset and over all frequencies was obtained from the ctrl-70 STRF. All STRFs (including those obtained from the ctrl-70 condition) were then normalized by dividing the mean neural activity by this single value. This normalization was aimed at minimizing the firing rate variability across recording sites. By definition, the maximum neural activity for the ctrl-70 condition was 1 (at the best frequency). One notes that values >1 are frequently observed in the AFB conditions (i.e., at the edge frequencies of the AFB); this indicates that the maximum of absolute firing rate in the AFB conditions is larger than the maximum of absolute firing rate in the ctrl-70 condition. This normalized mean neural activity is the dependent variable displayed in the STRFs.

To compare the STRFs obtained from control and AFB stimuli (and for display purposes), the differences between their frequency profiles were computed (Catz and Noreña 2013). The frequency profiles were obtained from the normalized STRFs by taking the maximum neural activity within a time window of 10–30 ms after stimulus onset for each tone pip frequency. For the frequencies outside the AFB, which were presented at 70 dB, the responses were compared with the corresponding frequencies obtained from the ctrl-70 condition. For the frequencies inside the AFB, which were presented at 40, 50, or 60 dB, the responses were compared with the corresponding frequencies obtained from the ctrl-40, ctrl-50, or ctrl-60 condition, respectively.

Statistics.

The main purpose of this study was to investigate whether the cortical responses at or near the acoustic spectral edges were modified compared with the neural responses induced by control stimuli (without spectral edges). Because neural responses were not normally distributed in all conditions and frequencies tested, we chose to use nonparametric tests. The Wilcoxon signed-rank test was used to compare the same recording sites over different conditions (AFB condition vs. control condition, or AFB conditions between them). The threshold for the significance value was Bonferroni corrected. The differences between AFB stimuli and the control condition were compared at all frequencies over 4 octaves (±2 octaves on either side of the AFB center, 33 frequencies overall). A difference was considered significant if P < 0.05/33 = 0.0015 at one or more frequencies. The statistical analysis was performed for MUA and LFP data separately. The Mann-Whitney test was used to compare independent samples, i.e., neural responses obtained at different cortical sites (comparison of neural responses for sites with BF corresponding to the upper edge of the AFB and those with BF corresponding to the lower edge of the AFB).

RESULTS

Effects of contrast depth on the representation of spectral edges.

For this experiment, 139 cortical sites were recorded from 8 animals. Figure 2 shows a representative example of MUA and LFP responses obtained for the different control and AFB conditions. In this example, the BF of the STRF derived from the LFPs and MUA was near 2,000 Hz and the center frequency of the AFB corresponded to the BF. This example shows the two main results of this study. First, the responses at the edge-out frequencies were increased in all AFB conditions compared with the responses obtained in control conditions. Remarkably, there was a clear and sharp increase of neural responses at the edge-out frequencies even for the smallest depths tested (10 and 20 dB). Second, the responses within the notch frequency band were suppressed compared with the respective control conditions.

Fig. 2.

Neural tuning of individual recordings obtained from a selected example at a given location in the primary auditory cortex. Each column corresponds to a stimulus condition (indicated at top). 1st and 2nd rows: spectrotemporal receptive fields obtained from control stimuli (multitone ensemble without AFB) for multiunit activity (MUA) and local field potentials (LFPs), respectively. 3rd and 4th rows: spectrotemporal receptive fields obtained from multitone ensemble with AFB for MUA and LFPs, respectively. Horizontal dotted lines represent the edge frequencies of the AFB. Neural activity in the spectrotemporal receptive field is represented by a color continuum from blue (minimum values) to red (maximum values). 5th and 6th rows: frequency profiles obtained by taking the maximum firing rate of MUA or the minimal amplitude of LFPs in the 10- to 30-ms time window, respectively. The red and black lines represent the neural responses for the stimulus ensemble with AFB and without AFB, respectively. Control conditions at 60, 50, and 40 dB are represented with a blue line. Vertical dotted lines represent the edge frequencies of the AFB. Neural responses are greatly enhanced at the edges of the AFB and decreased within the AFB.

Figure 3 shows the medians of the effects on neural activity of AFB stimuli, relative to the control stimuli, for three positions of BF relative to the AFB center: when the AFB is centered on BF (at ±⅛ octave, n = 82 sites; 2nd column), when the lower edge of the AFB is centered on BF (at ±⅛ octave, n = 25 sites; 1st column), or when the upper edge of the AFB is centered on BF (at ±⅛ octave, n = 32 sites; 3rd column). The effects of the AFB stimuli relative to the control stimuli were tested statistically for both MUA and LFPs (see methods). In all relevant comparisons, the results were identical for both MUA and LFPs. Instead of reporting the identical effects for MUA and LFPs, we reported the results for MUA only. The increase in neural responses at the edge-out frequencies of the AFB were statistically significant in all AFB conditions and for the three positions of AFB center relative to the BF (P < 0.0015). We also compared the AFB 70-60 condition with all other AFB conditions (3 comparisons) at the first edge-out frequencies of the AFB (for the 3 positions of AFB center relative to the BF). Neural responses at edge-out frequencies were significantly larger for large spectral contrasts (AFB 70-40 and AFB 70-0) compared with small spectral contrast (AFB 70-60) (P < 0.017, after Bonferroni correction). The decrease in neural responses at the edge-in frequencies of the AFB were statistically significant in all AFB conditions and positions of the AFB center relative to the BF (P < 0.0015). We compared the neural responses at the lower edge-in frequency with those at the upper edge-in frequency. Consistent with our previous study (Catz and Noreña 2013), we found that the neural suppression of responses at the lower edge-in frequency was significantly larger compared with the suppression at the upper edge-in frequency (P < 0.05).

Fig. 3.

Medians of the difference between the AFB conditions and the control conditions for 3 positions of the AFB relative to the neural best frequency (BF), as a function of frequency. Each column represents a position of the AFB relative to the neural BF. 1st and 2nd rows: the average MUA and LFP, respectively. 1st column: average data for neurons with BF corresponding to the lower edge of the AFB. 2nd column: average data for neurons with BF corresponding to the AFB center. 3rd column: average data for neurons with BF corresponding to the upper edge of the AFB. Vertical dotted lines represent the edge frequencies of the notch. Circles indicate statistically significant differences between control and AFB conditions for MUA (results are identical for LFPs and are not shown). Neural responses are enhanced near and outside the notch, whereas they are reduced near and within the notch.

The percentages of recording sites showing at least 20% increase or decrease of neural activity for the three groups of neural responses (whether BF corresponded to the center of the AFB or to the lower or upper edge of the AFB) are shown in Fig. 4. The percentages of the recording sites showing an increase of neural responses at edge-out frequencies ranged between nearly 100% for the AFB 70-00 condition and 50% for the AFB 70-60 condition. The percentages of the recording sites showing a decrease of neural responses at the edge-in frequencies ranged between 90% for the AFB 70-00 condition and 40% for the AFB 70-60 condition. These results suggest that the cortical changes induced by the AFB stimuli are very systematic for all the depths tested in this study, although fewer cortical sites are affected by spectral contrast with small depths. Remarkably, the percentage of sites showing a decrease of neural responses at edge-in frequencies was larger when BF corresponded to the lower edge of the AFB (50–90% of the cortical sites) than when BF corresponded to the upper edge of the AFB (30–70% of the cortical sites). This result is also consistent with our previous study (Catz and Noreña 2013).

Fig. 4.

Percentage of sites showing an increased response (1st and 2nd rows: MUA and LFPs, respectively) or a decreased response by at least 20% (3rd and 4th rows: MUA and LFPs, respectively). Otherwise configuration is same as in Fig. 3.

Time course of the edge enhancement.

All the changes reported in the previous section were induced by the 180-s presentation of the AFB stimulus. Although these changes were relatively rapid, it is unknown whether these changes occur over a shorter time period. Moreover, it is unclear whether these changes can persist after the end of the AFB stimulus presentation. We designed specific stimuli for answering these questions (see methods and Fig. 1). For this experiment, 145 cortical sites were recorded from 10 animals. Figure 5 shows an example at a given cortical site for two widths of the AFB (½ and 1 octave, respectively) and for which the AFB was centered on the BF (near 10 kHz). First, neural responses were globally enhanced in the ctrl-sil condition compared with the ctrl-70 condition (Fig. 5, 1st and 2nd columns). Second, AFB stimuli presented in alternation with silence or control stimulus produced a sharp neural enhancement at both edges of the AFB (Fig. 5, 3rd and 4th columns and 6th and 7th columns). On the other hand, the STRF derived from the control condition presented in alternation with the AFB stimulus was unchanged compared with the control STRF (Fig. 5, 5th and 8th columns). These results indicate that 10-s presentation of the AFB stimulus is sufficient to produce a neural enhancement at the spectral edges of acoustic stimuli and that this effect is short lived, since neural enhancement has disappeared within the 10 s of ctrl-70 stimulus presentation (in the ctrl-AFB condition).

Fig. 5.

Neural changes produced by 10-s AFB stimulus alternated with 10-s silence or control stimulus. Stimulus condition and time period (in parentheses) are indicated at top of columns. 1st and 2nd rows: spectrotemporal receptive fields obtained from MUA and LFPs, respectively. 3rd and 4th rows: frequency profiles of neural responses for MUA and LFPs, respectively. AFB bandwidth (notch width, Nw; indicated at top) was 0.5 octave in the 3rd–5th columns and 1 octave in the 6th–8th columns. The black and red lines represent the neural responses for the stimulus ensemble without AFB and with AFB, respectively. The black lines from the 1st column (for MUA and LFPs) are replicated in the 2nd–8th columns for comparison. Neural responses are greatly enhanced at the edges of the notch and decreased within the notch. The neural changes at the spectral edges are fast, i.e., they occur within 10 s of the stimulus with AFB presentation (3rd–4th columns and 6th–7th columns), and labile, i.e., they disappear within the following 10 s of the stimulus ensemble without AFB presentation (5th and 8th columns).

Neural recordings were grouped according to the position of BF relative to the AFB center: BF at the lower edge of the AFB (Fig. 6, left; n = 41 and 35 sites for the 1 and ½ octave-width AFB, respectively), BF at the center of the AFB (Fig. 6, middle; n = 72 and 65 sites for the 1 and ½ octave-width AFB, respectively), and BF at the upper edge of the AFB (Fig. 6, right; n = 32 and 28 sites for 1 and ½ octave-width AFB, respectively). The neural responses obtained from the ctrl-sil and ctrl-AFB conditions (during stimulation with ctrl stimulus) were significantly increased at frequencies of the AFB (and at adjacent frequencies in the ctrl-sil condition) compared with the ctrl-70 condition (P < 0.0015). More interestingly, neural responses obtained from AFB stimuli in the ctrl-AFB and AFB-sil conditions were significantly increased at the edge-out frequencies of the AFB (P < 0.0015). Finally, we tested whether the edge enhancement reported during the presentation of the AFB stimulus in the ctrl-AFB condition (AFB period, Fig. 1) was also present during the presentation of the control stimulation (ctrl period, Fig. 1): we did not find any significant increase of neural responses at spectral edges (P > 0.0015). In summary, these results indicated that the edge enhancement occurs within 10 s of acoustic stimulation and disappears very shortly after the presentation of the AFB stimulus (Fig. 6).

Fig. 6.

Medians of the difference between the various stimulus conditions and the control condition (ctrl-70) for 3 positions of the AFB relative to the neural BF, as a function of frequency and for the different stimulus conditions. Each column represents a position of the AFB relative to the neural BF (see Fig. 3). 1st and 2nd rows: results for MUA at the 0.5- and 1-octave notch width (Nw), respectively. 3rd and 4th rows: results for LFPs at the 0.5- and 1-octave notch width, respectively. Vertical dotted lines represent the edge frequencies of the AFB. Circles indicate statistically significant differences for MUA (results are identical for LFPs and are not shown). Neural responses at both edges of the AFB are largely enhanced, whereas they are reduced within the AFB.

The buildup and breakdown of neural changes during and after stimulation with AFB stimulus, respectively, were further investigated with a finer temporal resolution, i.e., within the following time windows: 0–2.5, 2.5–5, and 5–10 s (see methods). Our results indicate that the neural enhancement at the edge-out frequencies builds up and breaks down within a few seconds (0–2.5 s) of stimulation with AFB stimulus and the consecutive ctrl stimulus, respectively (P < 0.0015; Fig. 7).

Fig. 7.

Buildup and breakdown of neural changes produced by the AFB stimulus alternated with control stimulus or silence. Stimulus condition and time period (in parentheses) are indicated at top of each panel. The 10-s stimulation periods (ctrl or AFB) are divided into 3 time windows, i.e., 0–2.5, 2.5–5, and 5–10 s. 1st and 2nd rows: the neural changes for stimulus sequence with AFB of 1- or 0.5-octave bandwidth, respectively. The neural enhancement at the edge-out frequencies build up and break down within a few seconds (0–2.5 s) of stimulation.

Occasionally, we could get single-unit activity over an entire set of stimulus conditions. This provides an opportunity to apprehend whether the central changes observed for MUA are comparable to the putative central changes for single-unit activity. Figure 8 shows an example of single-unit activity for different stimulus conditions. This example shows a very clear increase of neural activity at the edge-out frequencies within 10 s of stimulus presentation. This result suggests that the effects observed at the level of MUA are likely also present at the level of single-unit activity.

Fig. 8.

Cortical responses evoked by stimulus ensembles with or without AFB for individual recordings where a single-unit activity over all the stimulus conditions could be obtained. Stimulus condition and time period (in parentheses) are indicated at top of each panel. 1st row: individual (and averaged; red line) waveforms of the single-unit activity. 2nd row: spectrotemporal receptive fields in all stimulus conditions. 3rd row: frequency profiles of cortical responses in all stimulus conditions. One observes a clear response enhancement at the edge-out frequencies, in particular on the upper edge of the spectral notch.

Effects of anesthesia.

Three animals were implanted chronically with microelectrodes to investigate the effects of anesthesia on the spectral edge enhancement (see methods). Figure 9 shows the STRFs obtained simultaneously from five recording sites in an awake animal (channels 2, 4, 6, 13, and 14). The STRF is unchanged when it is remote or adjacent from the AFB (Fig. 9, 1st and 3rd columns, respectively). On the other hand, for STRFs with BF close to the AFB center, the neural responses at the spectral edges of the AFB are greatly enhanced (Fig. 9, 2nd, 4th, and 5th columns). Remarkably, the neural enhancement at the lower spectral edge was particularly strong for the recording site shown in the second column of Fig. 9 (because there was no neural response at these frequencies produced by the control condition). Figure 10 shows the medians of the effects on neural activity of AFB stimuli, relative to the control stimuli, for three positions of BF relative to the AFB center. The number of recording sites was 15, 30, and 19 for recordings with BF at the lower edge, center, or upper edge of the AFB, respectively. The percentages of recording sites showing at least 20% neural increase are shown in Fig. 10 (bottom row). Because MUA and LFP have been documented above, and since they show very similar results in awake animals compared with anesthetized animals, LFPs are not shown. The neural responses at the spectral edges of the AFB stimuli were significantly enhanced (P < 0.0015, except when BF is at the lower edge of the AFB, where statistical significance does not survive the Bonferroni correction: P = 0.004).

Fig. 9.

Effects of anesthesia on the neural enhancement produced by AFB stimulus. Five individual examples of cortical responses evoked by the stimulus ensemble without (1st and 2nd rows: MUA and LFPs, respectively) or with AFB (3rd and 4th rows: MUA and LFPs, respectively) obtained simultaneously in an awake animal. 5th and 6th rows: frequency profiles of cortical responses for MUA and LFPs, respectively. One observes a clear response enhancement at the edge-out frequencies, suggesting that anesthesia did not play a major role in our results.

Fig. 10.

Medians of the cortical changes induced by the stimulus ensemble with AFB on awake animals. Each column corresponds to a position of the cortical BF relative to the AFB (see Figs. 3 and 6). 1st row: medians of the difference between the AFB conditions and the control conditions. Circles indicate statistically significant differences between control and AFB conditions. 2nd row: percentage of sites showing an increased response by at least 20%.

Finally, the time course of the spectral edge enhancement was also investigated in two awake animals. Figure 11 shows an example of a cortical site where the BF corresponded to the AFB center. One notes the very clear enhancement of neural responses at the spectral edges of the AFB in the AFB-sil and ctrl-AFB conditions. On the other hand, there was no enhancement at spectral edges in the ctrl-AFB condition (ctrl period). The results on this example are representative of the few other examples obtained in the two animals. Altogether, these results suggest that anesthesia plays only a minor role, if any, in the cortical changes reported in this study and our previous one (Catz and Noreña 2013).

Fig. 11.

Time course of the neural changes produced by AFB stimulus with 1-octave-wide spectral notch in an awake animal. Each column corresponds to a stimulus condition and period (indicated at top of panels). 1st and 2nd rows: spectrotemporal receptive fields obtained from MUA and LFPs, respectively. 3rd and 4th rows: frequency profiles of neural responses for MUA and LFPs, respectively. The black and red lines represent the neural responses for the stimulus ensemble without AFB and with AFB, respectively. The enhancement of neural activity at the spectral edges is fast, i.e., it occurs within the 10-s presentation of the stimulus ensemble with AFB (3rd and 4th columns), and labile, i.e., it disappears within the following 10-s presentation of the stimulus ensemble without AFB (5th column). These results are comparable to those reported in anesthetized animals.

DISCUSSION

In a previous study, we showed that the cortical representation of tone pip frequencies near spectral edges is greatly “distorted”: the representation of frequencies outside a spectral notch is greatly enhanced, whereas the representation of frequencies inside a spectral notch is reduced. These cortical changes were exquisitely sensitive to the characteristics of the spectral notch, including the width of the notch and the slope of the spectral edges (Catz and Noreña 2013). The present study extends these initial findings. We found that the neural enhancement at spectral edges are larger for deeper spectral notches but that it is statistically significant for depths as small as 10 dB. Moreover, the cortical changes at the spectral edges of the notch are rapid and labile, since they build up and break down within 2.5-s of the multitone stimulus presentation with or without AFB, respectively. These results imply that the neural mechanisms (discussed below) involved in these central changes are necessarily rapid, labile, and highly sensitive to small changes in the physical characteristic (depth, slope, and width) of the spectral notches.

Mechanisms of the dynamic changes in spectral contrast representations.

The rapid changes of cortical responses observed in the present study preclude the involvement of mechanisms with slower time constants such as long-term potentiation or depression, synaptic scaling, or intrinsic plasticity (Buonomano and Merzenich 1998; Desai et al. 1999; Grubb and Burrone 2010; Kuba et al. 2010; Turrigiano 2008; Turrigiano et al. 1998; Watt and Desai 2010). Instead, they are likely the results of one or a combination of relatively fast mechanisms occurring on the order of seconds or less.

A first mechanism that comes to mind is synaptic inhibition (including “lateral” inhibition), because it has long been suggested to play a critical role in emphasizing the representation of edges and contours in audition and vision (Carterette et al. 1969; Catz and Noreña, 2013; Hartline et al. 1956; Hartline and Ratliff 1957; Ratliff and Hartline 1959; Ratliff et al. 1967; Shamma and Symmes 1985; Von Békésy 1967, 1969). Synaptic inhibition has been proposed to be involved in the neural changes produced by two-tone sequences and after hearing loss (Calford 2002; Calford et al. 1993; Rajan, 2001; Rhode and Greenberg 1994; Shamma and Symmes 1985; Wang et al. 2002). Although synaptic inhibition may contribute to shape the tuning of cortical neurons, its effects are, however, limited in frequency range (cortical excitation and inhibition are approximatively co-tuned) and time (i.e., up to 100 ms after the stimulus onset) (Tan et al. 2004; Wehr and Zador 2003, 2005). One notes, however, that the respective tuning of excitation and inhibition may be different at the subcortical than at the cortical level. In particular, lateral (or side band) inhibition may play a major role in type “O” and “I” patterns of responses in the cochlear nucleus and inferior colliculus (Ramachandran et al. 1999; Ropp et al. 2014). It is conceivable that part of the findings reported here result from neural changes produced at the subcortical level (see below). Disinhibition, i.e., inhibitory neurons inhibiting other inhibitory neurons (Pi et al. 2013), might also play a role in our results, producing a response enhancement at the edge-out frequencies. Disinhibition has been postulated as a possible mechanism accounting for the response enhancement in two-tone sequences (Brosch et al. 1999).

Another mechanism that may be involved in the dynamic representational changes reported in this study is short-term synaptic plasticity, including synaptic depression and facilitation. Synaptic transmission is a dynamic and context-dependent process that can contribute to shape the representation of rapidly evolving acoustic features (Abbott and Regehr 2004; Brenowitz et al. 1998; Fortune and Rose 2001, 2002; Oswald et al. 2006; Reyes 2011; Zucker and Regehr 2002). Synaptic depression is usually attributed to depletion of some readily releasable vesicle pools due to repeated stimulations at a rate above the vesicle turnover rate. Synaptic depression can occur in the central auditory system but also as early as the synapses between inner hair cells and cochlear fibers (Goutman and Glowatzki 2007; Zilany et al. 2009). In general, synapses with high-probability release are subjected to synaptic depression, and those with low-probability release can demonstrate synaptic facilitation (Regehr 2012; Zucker and Regehr 2002). Other postsynaptic mechanisms can contribute to synaptic depression, such as receptor desensitization (Jones and Westbrook 1996; Otis et al. 1996).

Interestingly, inhibition and short-term synaptic plasticity may interact in a complex way to shape the spectrotemporal properties of cortical neurons. At the end-bulb glutamatergic synapse, for example, it has been shown that whereas GABA_B presynaptic inhibition reduces the initial synaptic current evoked during a stimulation train, it can also greatly reduce the synaptic depression produced over time by the train (Brenowitz et al. 1998). By reducing the probability of neurotransmitter release, presynaptic inhibition produces less receptor desensitization and therefore causes a net increase of the synaptic transmission during a prolonged stimulation (Brenowitz and Trussell 2001).

Any discussion relative to the mechanisms involved in our results is necessarily speculative. At this stage, we suggest that complex interactions between inhibition and short-term synaptic plasticity can account for the neural changes we observe at the spectral edges of an acoustic stimulus. Another important question relative to the mechanisms enhancing spectral contrasts is whether they operate at the cortical level or are inherited from lower levels of the auditory pathway. In broad agreement with our previous study (Catz and Noreña 2013), the similar pattern of responses for MUA and LFPs suggests that the enhancement of spectral contrasts observed in the cortex is largely inherited from lower levels. Finally, our study also shows that anesthesia plays only a modest role, if any, in our results. This corroborates with an earlier study showing that anesthesia marginally affects poststimulation suppression and facilitation produced by tone sequences (Brosch and Scheich 2008).

Functional implications.

Given the demonstration in the present study that the representation of spectral contrasts is rapidly enhanced in the auditory cortex for spectral depth as small as 10 dB, one can wonder whether these neural changes have functional implications. Importantly, the spectral profile of the stimuli used in the present study shows a clear contrast only when they are time-averaged over a few hundreds of milliseconds. This stimulus design (overlapping but asynchronous tone pips) was used to assess the representation of each tone pip frequency for different spectral profiles. One can therefore wonder whether the results obtained from this particular stimulus can apply to other (more ecological) stimuli. Many natural sounds, such as vocalizations, usually last a few hundreds of milliseconds or more and often consist of a succession of similar acoustic motifs or syllables (Aubin and Jouventin 2002; Huetz et al. 2009). Guinea pig vocalizations, for instance, though they may vary over time, usually maintain their harmonic structure over the course of the sequence (Berryman 1976). The enhancement observed in the present study, implying temporal integration over a few hundreds of milliseconds, could therefore well apply to some natural temporal sequences.

Moreover, ecologically relevant stimuli are often composed of frequency-specific information such as spectral edges (Assmann and Summerfield 2004; Moore and Glasberg 1983; Palmer and Shamma 2004). It seems that the auditory system relies heavily on these spectral cues. For instance, the detection of spectral peaks caused by vocal tract resonances (formants) can play a major role in speech recognition (Assmann and Nearey 1987; Darwin 1984; Henry et al. 2005). These acoustic features may also help differentiate relevant acoustic features from background noise (Darwin 1984; Roberts and Moore 1990). This hypothesis that spectral contrasts are important for speech recognition in noise has been further corroborated by psychoacoustic studies. These studies showed an improvement of intelligibility in noise for processed speech with enhanced spectral contrasts (increase of peak-to-valley ratio) (Alexander et al. 2011; Baer et al. 1993; Simpson et al. 1990). The perceptual effect of spectral enhancement was equivalent to a speech-to-noise ratio improvement of 4.2 dB (Baer et al. 1993). The increase and decrease of edge-out and edge-in frequency representation, respectively, may contribute to emphasize the representation of spectral edges and/or to prevent the loss of the spectral information degraded in a noisy environment (Rhode and Greenberg 1994).

The neural enhancement at the spectral edges may also account for “Mach bands” in hearing, i.e., perceptual enhancement at the spectral edges of acoustic stimuli (Carterette et al. 1969; Houtgast 1972). In particular, this may account for the pitch induced by noise bands at their spectral edges (Bilsen 1977; Small and Daniloff 1967). Indeed, low- and high-pass noises have been shown to produce a pitch up to 10 kHz, which is correlated with the cutoff frequency (Bilsen 1977; Small and Daniloff 1967). One notes that the high cutoff frequency of the noise (>10 kHz), which can produce a pitch, precludes a mechanism based on a temporal enhancement, because no temporal information is available above 3.5–5 kHz (Palmer and Russell 1986; Rose et al. 1967). The enhancement at spectral edges could also account for the dominant role played in pitch perception by the lowest and highest partials of a harmonic complex, especially when the low-numbered (resolved) partials are removed from the complex (Dai 2000; Moore and Gockel 2011).

Stimulation with notched stimuli (filtered noise or harmonic complex) is known to produce interesting auditory phenomena. The Zwicker tone is a tonal illusory percept produced immediately after the noise presentation. The pitch of this sensation lies within the frequency range of the notch (Noreña et al. 2000, 2002; Parra and Pearlmutter 2007; Wiegrebe et al. 1996; Zwicker 1964). Additionally, auditory enhancement can be produced immediately after the presentation of a broadband signal. A deleted frequency component from a harmonic complex perceptually “pops out” when it is reinserted. The detection thresholds for the component that was deleted are also lowered (Byrne et al. 2011; Summerfield and Assmann 1989; Viemeister and Bacon 1982; Wiegrebe et al. 1996). It is unknown whether the stimuli used in the present study, which are composed of tone pips with asynchronous onsets, can produce Zwicker tone and/or auditory enhancement immediately after the presentation of the AFB stimuli. Our results suggest, however, that these two auditory phenomena are preceded (during stimulation with the notched stimulus) by a reduction of neural activity within the frequency range of the spectral notch. One speculates that this neural suppression might be followed by a rebound of activity immediately after the AFB stimulus presentation, which could account for the Zwicker tone and/or auditory enhancement [see Fig. 7, ctrl-AFB (ctrl) condition with 1-octave AFB, and Noreña and Eggermont (2003)].

Artificial hearing loss, acoustic environment, and long-term central changes.

We and others have previously suggested that notched stimuli, such as our stimulus ensemble with AFB or notched noise, mimics the discontinuity in sensory inputs over frequency produced by sharp hearing loss (Catz and Noreña 2013; Noreña et al. 2000; Okamoto et al. 2007; Pantev et al. 1999). In this context, the stimulus ensemble used in the present study can be interpreted as being an equivalent of those producing an artificial scotoma in vision (Das and Gilbert 1995; DeAngelis et al. 1995; Kapadia et al. 1994; Parks and Corballis 2012; Pettet and Gilbert 1992; Ramachandran and Gregory 1991; Weil et al. 2007, 2008). The neural enhancement at the edge-frequencies of the spectral notch is reminiscent of the long-term central changes triggered by cochlear hearing loss, i.e., neural enhancement or even unmasking at the edge-frequency of hearing loss (Calford et al. 1993; Noreña and Eggermont 2005; Robertson and Irvine 1989). In case of an extensive exposure (for a few weeks) to an acoustic environment with spectral notch, it can be speculated that the short-term changes reported in this study may be gradually converted into long-term changes (Noreña et al. 2006; Pienkowski and Eggermont 2010; Pienkowski et al. 2013). Our results further suggest that chronic exposure to an acoustic environment with a spectral contrast as small as 10 dB and presented at a moderate level (∼70 dB SPL) may be sufficient to produce chronic central changes. These chronic central changes, including changes in the pattern of spontaneous firing (Noreña et al. 2006), may have functional consequences, such as tinnitus and hyperacusis (Noreña 2011; Noreña and Farley 2013).

GRANTS

This work was supported by L'Agence Nationale de la Recherche Grant ANR-2010-JCJC-1409-1.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

N.M. and A.J.N. performed experiments; N.M. and A.J.N. analyzed data; N.M. and A.J.N. prepared figures; N.M. and A.J.N. drafted manuscript; N.M. and A.J.N. edited and revised manuscript; N.M. and A.J.N. approved final version of manuscript; A.J.N. conception and design of research; A.J.N. interpreted results of experiments.