Mass Potentials Recorded at the Round Window Enable the Detection of Low Spontaneous Rate Fibers in Gerbil Auditory Nerve

Abstract

Auditory nerve fibers (ANFs) transmit acoustic information from the sensory hair cells to the cochlear nuclei. In experimental and clinical audiology, probing the whole ANF population remains a difficult task, as the ANFs differ greatly in their threshold and onset response to sound. Thus, low spontaneous rate (SR) fibers, which have rather higher thresholds, delay and larger jitter in their first spike latency are not detectable in the far-field compound action potential of the auditory nerve. Here, we developed a new protocol of acoustic stimulation together with electrophysiological signal processing to track the steady state activity of ANFs. Mass potentials at the round window were recorded in response to repetitive 300-ms bursts of 1/3 octave band noise centered on a frequency probe. Analysis was assessed during the last 200-ms of the response to capture the steady-state response of ANFs. To eliminate the microphonic component reflecting the sensory cells activity, repetitive pairs of sounds of opposite polarities were used. The spectral analysis was calculated on the average of two consecutive responses, and the neural gain was calculated by dividing point-by-point the spectrum to sound over unstimulated condition. In response to low-sound-level stimulation, neural gain predominated in the low-frequency cochlear regions, while a second component of responses centered on higher cochlear frequency regions appeared beyond 30 dB SPL. At 60 dB SPL, neural gain showed a bimodal shape, with a notch near 5.6 kHz. In addition to correlate with the functional mapping of ANFs along the tonotopic axis, the deletion of low-SR fibers leads to a reduction in the high-frequency response, where the low-SR fibers are preferentially located. Thus, mass potentials at the round window may provide a useful tool to probe the SR-based distribution of ANFs in humans and in other species in which direct single-unit recordings are difficult to achieve or not feasible.

Fig 1. Extraction of a neural index from the round-window response.

(A, B) Round-window responses (bottom, s) were evoked by repetitive 300-ms bursts of third-octave band noise (top, e) centered at a probe frequency (50 dB SPL and 4 kHz in this example). To reduce the cochlear microphonic: i) two consecutive stimuli (that form a pair) were presented in opposite polarity and ii) all the pairs (n = 50) differed from each other (by changing the seed of the pseudorandom stream). (C) Calculation of halved sum (red) and halved difference (green) within each pair of response using black and blue traces shown in B. (D-O) Spectral analysis of round-window response before (D-I) and after 10 μM TTX application (J-O) into the round window niche. The grey trace is the power spectral density of the signal recorded at the round window in the absence of sound stimulation (unstimulated activity). Black and blue traces (D, E) are calculated from responses with even-numbered (black, first element of each pair) or odd-numbered rank (blue, second element of each pair). Note the mixture of a neural component centered around 900 Hz and the microphonic centered at 4 kHz. To segregate neural and microphonic components, the spectral analysis was calculated after the half summation (red, F) or the half difference (green, G) of traces within each pair of response (see equations in inset). The neural and microphonic sound-evoked activities (red and green traces in H and I) were derived from the traces shown in F and G respectively (ratio between colored and grey traces). Note the complete disappearance of neural activity after TTX application, leaving the cochlear microphonic unaffected (J-O). The area under the red curve in panel H was used as an index of neural sound-evoked activity (34.5 dB×kHz in this example).

Fig 2. Neural subcomponents in round window responses.

Spectral analysis of the halved sum responses across frequency (1 to 16 kHz) and intensity (0 to 60 dB SPL), before (1^st row, power spectral density, A-E, 2^nd row, gain of activity, F-J) and after 10 μM TTX application (1^st row, power spectral density, K-O, 2^nd row, gain of activity, P-T). PSD recorded in the absence of sound stimulation are shown in black (unstimulated activity). PSD obtained in response to level of stimulation from 0 to 60 dB SPL in 10 dB increments are shown in red. The spectral subcomponent (1), which reflects the steady-state firing of fibers, decreases in frequency for probe frequency below 8 kHz (C-E and H-J). The subcomponent (2) set at twice the stimulus frequency is a second harmonic resulting of half summation.

Fig 3. Low-pass modulation transfer functions account for spectrum shift behaviour.

A. Protocol used to record unit contribution at the round window (from [29,30,33]). Spontaneous action potentials (black trace) recorded with an electrode in the auditory nerve were used as trigger pulses to average the corresponding action potentials recorded with a gross electrode at the round window (blue trace). After more than 10,000 averages, a biphasic waveform of 0.3 μV amplitude, 1 ms second duration, was obtained. Red fit: f(t) = A×((cos(2πf₁t)+1)×sin(2πf₂t) with A = 0.31 ± 0.04 μV, f₁ = 997 ± 20 Hz, f₂ = 920 ± 33 Hz, R² = 0.95 ± 0.007, 59 fibers, >10,000 averaging per fiber. B. Parameters of the unit contribution (n = 59 ANFs). C. Amplitude of the spectrum density function (ASD) of the unit contribution (adequate zero padding was applied to improve the frequency resolution of the spectral estimate). The peak in the ASD is around 1 kHz. Inset: Unit contribution model used to estimate the PSD. D. Location of the spectral component as a function of the probe frequency for experimental (black curve, 10 gerbils) and simulated data (red curve) at 60 dB SPL. Note the spectrum shift for a probe frequency below 10 kHz. Inset: 3-dB cut-off frequency of modulation transfer functions as a function of the fiber CF (f(CF) = 2000×(1-exp(-CF/9000)) with CF in Hz). E. Low-pass modulation transfer functions obtained from C and D (8^th order Butterworth filters, 0 dB in band pass and cut-off frequency at 3-dB) for fibers with CF ranging from 1 to 32 kHz in 1 octave steps. Note that cut-off frequency is positively correlated with the CF of the fibers.

Fig 4. Probing the auditory nerve using round-window response in control and ouabain-poisoned cochleae.

(A, B) Iso-level neural index in control (A, n = 10 gerbils) and ouabain-poisoned cochleae (B, n = 9 gerbils). The frequency probe varied from 1 to 32 kHz in 1/3 octave steps and sound level from 10 (red curve) to 60 dB SPL (green curve) in 10 dB steps. Note that below 2 kHz (light-coloured area), the neural index amplitude decreases because of the phase locked response of fibers and their cancelation by opposite polarities. (C-F) Amplitude versus intensity functions of the neural index at 2 (C), 4 (D), 8 (E), and 16 kHz (F), in artificial perilymph control (black) and ouabain-poisoned cochleae (red). Inset: examples of neural sound-evoked response in control (black) and ouabain-poisoned cochleae (red) as shown in Fig 1G. (G, H) Ouabain-poisoned versus control correlations from low- (< 5.6 kHz, G) and high-frequency probe (> 5.6 kHz, H) derived from panels A and B. The data obtained below 2 kHz were excluded. The coordinate of each small dot corresponds to the neural index amplitude in control (x-coordinate) and ouabain condition (y-coordinate) measured for the same frequency and sound level. Large dots represent the average of small dots pooled per sound level (from 10 dB SPL in red to 60 dB SPL in green). The black line is the invariant model y = x, simulating an absence of drug effect. Red curves are lowest-order polynomial fits to the data (G, y = 0.96×x—0.06, r² = 0.96; H: y = -0.002×x²+0.8×x+0.18, r² = 0.97). Data are expressed as the mean ± SEM. X and Y error bars display the mean ± SEM of data shown in x and y axis, respectively; * p<0.05, two-way ANOVA test followed by post hoc Tukey’s test.

Fig 5. Compound action potential of the auditory nerve in control and ouabain-poisoned cochleae.

CAP amplitude-intensity functions in response to 2 (A), 4 (B), 8 (G), and 16 kHz (H) tone bursts, in artificial perilymph control (black, n = 10) and ouabain-poisoned cochleae (red, n = 9). Inset: Example of CAP in control (black) and ouabain-poisoned cochleae (red). CAP amplitude was measured between N₁ (the first negative wave) and P₁ (the subsequent positive wave). Data are mean ± SEM. No statistical difference was found between control and ouabain-perfused animals. p > 0.05, two-way ANOVA.

Fig 6. Synapse counts in control and ouabain-poisoned cochleae.

(A, B) Confocal microscopy of immunolabeled CtBP2 (green) and GluA2 (red) from the 16-kHz encoding region in artificial perilymph control (A) and ouabain-poisoned cochleae (B). Top panel: enlarged view of inner-hair-cell innervation (6 IHCs; n indicates the nucleus of IHCs). Middle panel: z-projection of the white square shown above (4 μm × 4 μm), showing CtBP2 and GluA2 immunolabeling alone or together (merged). Bottom panel: Three-dimensional (3D) views of the white square shown above (4 μm × 4 μm × 4 μm). Note the presence of an orphan ribbons in ouabain-poisoned condition (B, 12 ± 2% in the basal end (>5.6 kHz) against 1 ± 0.6% in the apical end (<5.6 kHz). (C) Number of synapses per IHC along the gerbil tonotopic axis [24] in control and ouabain-poisoned cochleae (black, control, 5 cochleae, 324 IHCs, 5790 synapses; red, ouabain, 5 cochleae; 344 IHCs, 5494 synapses). Each dot represents the average over 6 consecutive IHCs [14]. Black and red curves are fits using the sum of two Gaussian models (control, black, f(x) = 22.6×exp(-((x-30)/35.2)²) + 14.2×exp(-((x-78.6)/24.1)²), r² = 0.92; ouabain, red, f(x) = 22.9×exp(-((x-30)/33.4)²) + 9.7×exp(-((x-75.2)/23.6)²), r² = 0.88, with x the position from the apex in percent). Inset: Estimates of the number of synapses per cochlea calculated from IHC and synapse counts. (black, control: 19,659 synapses/cochlea; red, ouabain: 17,568 synapses/cochlea). (D) Number of synapses per IHC pooled per octave band, in control (black) and ouabain-poisoned (red) cochleae. Numerical values indicate the number of IHCs for which the number of synapses was assessed. Data were expressed as the mean ± SEM, P<0.05, P<0.01, two-way ANOVA test followed by post hoc Tukey’s test.