In this study, 61 two- to three-month-old C57BL/6 J mice of both sexes (27 males and 34 females) were used. The mice were housed in groups of 3 to 5 animals per cage under standard laboratory conditions, including a 12 h light–dark cycle, room temperature of 23°C, and unrestricted access to food and water.

To induce tinnitus, mice were exposed for 1 hour to 1/3 octave narrowband noise centered at 10 kHz and with an intensity of 116 dB SPL. During exposure, mice were anesthetized with a mixture of ketamine (Calypsol, 35 mg/kg) and xylazine (Xylapan, 6 mg/kg) and placed on a heated pad (37°C) in a custom-made soundproof chamber. Exposure noise was created using a white noise generator RFT 03004, processed using a set of filters RFT 01013 and a custom-made amplifier, and presented by a DE700 speaker (B&C Speakers, Italy) connected to a horn. The sound pressure level calibration was performed using a Brüel & Kjaer 4,939 1/4″ microphone and a B&K 2231 noise meter (Brüel & Kjaer, Denmark).

Acoustic startle responses (ASRs) of mice were studied using the Acoustic Startle Response System (Habitest model E10-21, Coulbourn, Pennsylvania, United States) placed in a soundproof room. During testing, individual animals were confined in a small wire cage placed on a sensitive platform containing a load-cell transducer that converts pressure into voltages. The output electrical signals were acquired and processed using a TDT 3 system with an RP 2 real-time processor (Tucker Davis Technologies, Alachua, United States). Acoustic stimuli were generated by a TDT 3 system and presented through a 29TAF/W loudspeaker (SEAS, Norway) placed above the animal. The frequency response of the loudspeaker in the test rig was within ±9 dB at 1–32 kHz. The equipment was calibrated using a Brüel & Kjaer 4,939 1/4-inch microphone and a Brüel & Kjaer 2,231 sound meter (Brüel & Kjaer, Denmark). Stimulus generation, data collection and processing were controlled by Matlab software. Startle responses were analyzed within a 100 ms window beginning at stimulus onset and were measured as maximal peak-to-peak voltage amplitude.

Testing consisted of two identical sessions, conducted separately on two consecutive days. The aim of the first session was to adapt the animal to the test equipment and its results are not presented in this study. Each session began with a ten-minute habituation of the animal to the experimental setup, followed by 4 consecutive trains of 21 startle-eliciting stimuli at intervals randomly varying from 20 to 30 s. Intertrain intervals were 10 min, and responses to the first of 21 stimuli for each train were not included in further analysis. The sessions included three different types of behavioral paradigms (Figure 1A): (a) induction of ASR by the startle stimulus alone (white noise, 50 ms/110 dB SPL, rise/fall time 0.5 ms) without further modulation, (b) ASR inhibition by narrowband background noise (BGN, 1/3 octave band centered at 10 kHz, 65–85 dB SPL), and (c) inhibition of ASR by gap pre-pulse (50 ms, rise/fall time 3 ms, 100 ms time interval between onset of pre-pulse and startle stimulus) in the presence of BGN (GPIAS). The paradigm in (a) was applied during the first trains of the sessions, while the paradigms in (b) and (c) were applied during the second to fourth trains as 10 no gap and 10 gap trials presented in a random order in the presence of continuous BGN of 65, 75, 85 dB SPL intensity, different for each train. The raw ASR amplitude for each mouse was quantified as the average of 10 randomly selected voltage responses recorded during the first train of the session. ASR inhibition by BGN was quantified as follows:

where ASR_BGN and ASR_{no BGN} represent the average raw ASR amplitudes obtained from 10 replicate measurements for trials with no gap in the presence of BGN and from measurements obtained during the first train in the absence of BGN (see above). ASR inhibition by gap-prepulse was quantified as follows:

where ASR_gap and ASR_{no gap} represent the average raw ASR amplitudes, each obtained from 10 repeated measurements during the same stimulus train in the presence of BGN, with and without gap pre-pulse. The significance of the difference between ASR_gap and ASR_{no gap} was further tested using unpaired Student’s t-test for each mouse. Mice that showed non-significant differences between ASR_gap and ASR_{no gap} at 75 dB BGN 2 weeks after acoustic trauma were assigned to the tinnitus group (+T), while mice showing any significant differences were assigned to the non-tinnitus group (−T).

Auditory brainstem responses (ABRs) were obtained from ketamine/xylazine anesthetized mice using methods described previously (Willott, 2006; Rybalko et al., 2019). During recording, mice were placed on a heated pad in a soundproof anechoic room. A TDT System 3 Real-Time Processor RP 2.1 (Tucker Davis Technologies, FL, United States) was used to generate sound stimuli that were transmitted to the animal via a PMA 720 acoustic system (Denon, Japan) and a loudspeaker system consisting of a 140-15D ribbon tweeter (RAAL advanced loudspeakers, Serbia) and an Alpha 6A speaker (Eminence, United States) under free-field conditions. The frequency response of the system was flat (±3 dB) in the range from 1 to 50 kHz. Calibration of the equipment was performed using a Brüel & Kjaer 4,939 1/4-inch microphone and a B&K 2231 sound level meter (Brüel & Kjaer, Denmark).

ABRs were elicited by clicks of 0.1 ms duration and a series of tones of 2, 4, 8, 16, and 32 kHz and 5 ms duration, including 1 ms onset and offset ramps with a rate of 15/s. The level of both types of stimuli was set between 100 and 0 dB SPL and gradually decreased in steps of 5 or 10 dB. Evoked brainstem potentials were recorded using subcutaneous needle electrodes placed on the vertex (active electrode) and in the region of the neck muscles (reference electrode). Recorded signals were band-pass filtered (300 to 3,000 Hz) and averaged from 500 traces elicited by repetitive stimulation (TDT BioSig software). The ABR threshold was the minimum tone or click intensity that elicited a visually detectable ABR within the expected time window. The amplitude of ABR waves I and IV was measured from the negative peak to the positive peak in responses evoked by click stimuli. The amplitudes of ABR waves I and IV obtained at varying click levels were used to construct ABR growth functions. To determine the slope of the function, for each mouse, the region of the function starting 10 dB above threshold was fitted using linear regression. Absolute latencies were measured from the time point at which the sound stimulus reached the ear to the time of the positive peak of each wave.

The datasets were analyzed using GraphPad Prism 9.5. All values are presented as the mean ± SD, the significance level is set to be <0.05. Two-sample Student’s t-tests and two-way ANOVA followed by Sidak’s multiple comparison test (SMCT) were used to assess differences in means, unless otherwise stated. The normality of the data was assessed using D’Agostino and Pearson’s test and the presence of outliers was checked using the ROUT method (Q = 0.05%). Correlations between variables were assessed using Pearson and Spearman correlation coefficients.

To induce tinnitus, we exposed anesthetized mice (n = 61) to 1/3 octave narrow-band noise, centered at 10 kHz with intensity of 116 dB SPL for 1 hour (Rybalko et al., 2019). The presence of tinnitus was assessed by inhibition of ASR with a gap pre-pulse in BGN (Turner et al., 2006) (Figure 1A). The intensity of BGN was first set to 75 dB SPL which is well audible even after hearing loss induced by acoustic trauma (see later). Two weeks after exposure, mice showed a significant decrease in the magnitude of ASR inhibition by gap pre-pulse (GPIAS) from 49.6 ± 8.9% to 20.8 ± 23.4% (p<0.001, paired Student’s t-test) (Figure 1B). This was consistent with the presence of noise-induced tinnitus, which masked gap pre-pulse, thereby reducing its inhibitory effect on ASR. The simultaneous increase in the coefficient of variation of GPIAS values from 18.0 to 112.6% suggests heterogeneity in the effect of exposure on the development of tinnitus. To distinguish between mice with high and low severity of tinnitus symptoms, we divided the entire group of animals into those that exhibited either non-significant or any significant inhibition of ASR by the gap embedded in 75 dB BGN after exposure (Figure 1B). Significance of inhibition in each mouse was assessed by statistical comparison of the differences between mean raw ASRs calculated from repeatedly stimulated responses with and without a gap pre-pulse (Kraus et al., 2011) (see Methods). These two distinct groups are hereafter referred to as tinnitus (+T) and non-tinnitus (−T) (Figure 1C). Of note, pre- and post-exposure GPIAS values were not significantly different in males and females (before exposure: p = 0.961; after exposure: p = 0.979; SMCT, data from 27 males and 34 females) and in subsequent analyses, data from mice of both sexes were pooled together.

Despite the tendency of +T mice to have higher GPIAS values than −T mice before exposure, the differences did not reach statistical significance (Figure 1C). Accordingly, the correlation between pre- and post-exposure GPIAS values in the whole mouse population was not significant (Spearman r = −0.233, p = 0.071, n = 61). Furthermore, mice of both groups showed similar raw ASR amplitudes in the absence or presence of 75 dB BGN after exposure (for −T and + T mice, ASR_{no BGN} were 0.631 ± 0.244 V and 0.695 ± 0.353 V, p = 0.603, and ASR_{no gap} were 0.399 ± 0.167 V and 0.368 ± 0.247 V, p = 0.881, two-way RM ANOVA with SMCT, n = 24 and 37), and no significant correlation was observed between these amplitudes and postexposure GPIAS at 75 dB BGN (ASR_{no BGN}: Spearman r = 0.009, p = 0.945, ASR_{no gap}: Spearman r = 0.188, p = 0.147, n = 61). Finally, a 10 dB SPL shift in BGN level up or down did not alter the significance of differences in GPIAS between the two groups (Figure 1D). These observations indicate that noise induced GPIAS failure in +T mice did not depend on the baseline value of this index before exposure or on its parameters such as post-exposure ASR amplitude and BGN intensity. Thus, our results support that the +T and −T groups differ in the presence or intensity of tinnitus and are consistent with the previously described heterogeneous sensitivity of mice to noise-induced tinnitus (Li et al., 2015; Fabrizio-Stover et al., 2022).

In these experiments, we compared the auditory function in mice from the +T and −T groups before noise exposure. We first analyzed bilateral ABRs evoked by pure tones and click stimuli (Figure 2). The thresholds of ABRs showed the lowest values in the frequency range of 8 to 16 kHz, consistent with previously reported values (Zheng et al., 1999), and were not significantly different in +T and −T mice (Figure 2B). On the other hand, ABR wave amplitudes showed a slight increase at higher sound levels in +T mice (Figures 2C,D), resulting in a steeper slope of their input–output functions (Figure 2E). The increase was manifested by the amplitudes of both early and late ABR waves (I, IV), so that their ratio (IV/I), a measure of neural gain (Mohrle et al., 2019), was similar in both animal groups (Figure 2F). Peak latencies were not significantly different in +T and −T mice (Figures 2G,H). Thus, our findings show that mice that significantly vary in their susceptibility to noise-induced tinnitus do not show substantial differences in their ABRs.

Next, we assessed hearing function in all mice based on behavioral measures. As noted above, the +T and −T groups showed similar pre-exposure GPIAS values, and our measurements also showed that pre-exposure raw ASR amplitudes in the absence of BGN did not differ between these groups (−T vs. + T: 0.764 ± 0.222 V vs. 0.847 ± 0.272 V, p = 0.269, two-way RM ANOVA with SMCT, n = 24 vs. 37). Consistent with previous reports (Gerrard and Ison, 1990), we further observed inhibition of ASR in the presence of BGN, with the degree of inhibition dependent on BGN level (Figure 3A). Interestingly, while ASR inhibition in +T and −T mice was similarly low or high at BGN levels of 65 dB or 85 dB SPL, respectively, at 75 dB SPL BGN elicited significantly greater ASR inhibition, resulting in lower raw ASR amplitudes (ASR_BGN) in +T mice than in −T mice (−T vs. + T: 0.642 ± 0.187 V vs. 0.445 ± 0.156 V, p < 0.01, two-way RM ANOVA with SMCT, n = 24 vs. 37). Accordingly, pre-exposure relative ASR amplitudes in the presence of 75 dB BGN were significantly lower in +T than in −T mice (Figure 3A) and significantly correlated with post-exposure GPIAS values in all mice (Spearman’s r = 0.641, p < 0.001, n = 61). In contrast, the correlation between GPIAS and ASR inhibition by BGN at 65 dB or 85 dB was not significant (Spearman’s r = 0.048 or 0.091, p = 0.714 or 0.488, n = 61). We add that relative ASR amplitudes at 75 dB BGN were found to be similar in males and females before noise exposure (p = 0.781, unpaired Student’s t-test, n = 27 and 34). After exposure, ASR inhibition in the presence of 75 dB BGN decreased in −T mice from 86.0 ± 16.9% to 65.3 ± 51.2% while it remained unchanged in +T mice (p = 0.004 vs. 0.927, mixed-effects model with Tukey’s multiple comparisons test) (not shown). However, the difference between +T and −T groups remained significant after exposure (p = 0.003, two-way ANOVA with SMCT). Thus, the two groups of mice showed differential ASR sensitivity to background sounds. The results also suggest that measuring ASR inhibition in the presence of BGN could represent a simple test of animal susceptibility to noise-induced tinnitus.

When using such a test, it would be essential to know the probability of tinnitus in individual animals with a given BGN-induced ASR inhibition. Therefore, we further investigated the association between the degree of ASR suppression and the presence of tinnitus in mice using regression analysis. Figure 3B shows a logistic regression curve that models the probability of tinnitus in +T and −T mice as a function of a continuous variable, the predictor, which is the relative amplitude of ASR in the presence of BGN. Based on this analysis, mice in which BGN at 75 dB suppressed ASR amplitudes to less than 60% of baseline could be expected to have tinnitus with a probability greater than 80% after acoustic trauma. To estimate the predictive ability of the logistic regression, the receiver operating characteristic (ROC) curve was constructed (Figure 3C). The data suggest that prediction of the presence of tinnitus based on ASR suppression provides highly reliable results because the shape of the ROC curve approximates right-angle lines and the area under the curve (AUC) is equal to 0.94. In comparison, the AUC of an ideal classifier is equal to 1 and a prediction equal to chance has an AUC of 0.5 (Figure 3C).

In the last part of the experiments, we compared auditory function in +T and −T mice after noise exposure. Both groups showed similar increases in ABR thresholds of about 10 to 30 dB, with the largest shift observed in the 8–16 kHz band, which roughly corresponds to the frequency of narrowband noise used for traumatic overexposure (Figures 4A,B). Furthermore, we observed a decrease in the maximum amplitude of both early and late ABR waves by more than half (p < 0.001, mixed-effects model with SMCT) and a corresponding decrease in the slope of the input–output functions (p < 0.001 for wave I in both mouse groups and p < 0.01 or p < 0.001 for wave IV in −T or +T mice, respectively; mixed-effects model with SMCT) to values that were not significantly different in +T and −T mice (Figures 4C–E). The IV/I ratio increased after exposure for ABRs elicited by moderate-intensity stimulation in +T mice (Figure 4F), whereas it did not change in −T mice across the entire range of stimulation levels (mixed-effects model with SMCT; not shown). The latencies of wave I and wave IV peaks were prolonged after exposure in both mouse groups (p < 0.001, mixed-effects model), with the change significantly greater in +T mice (Figures 4G,H). These observations indicate noise-induced hearing loss caused by acoustic trauma in both groups of mice, and the differences between +T and −T mice are consistent with previously published audiometric correlates of noise-induced tinnitus.