Auditory‐evoked potentials to changes in sound duration in urethane‐anaesthetized mice

Spectrotemporally complex sounds carry important information for acoustic communication. Among the important features of these sounds is the temporal duration. An event‐related potential called mismatch negativity indexes auditory change detection in humans. An analogous response (mismatch response) has been found to duration changes in speech sounds in rats but not yet in mice. We addressed whether mice show this response, and, if elicited, whether this response is functionally analogous to mismatch negativity or whether adaptation‐based models suffice to explain them. Auditory‐evoked potentials were epidurally recorded above the mice auditory cortex. The differential response to the changes in a repeated human speech sound /a/ was elicited 53–259 ms post‐change (oddball condition). The differential response was observable to the largest duration change (from 200 to 110 ms). Any smaller (from 200 to 120–180 ms at 10 ms steps) duration changes did elicit an observable response. The response to the largest duration change did not robustly differ in amplitude from the response to the change‐inducing sound presented without its repetitive background (equiprobable condition). The findings suggest that adaptation may suffice to explain responses to duration changes in spectrotemporally complex sounds in anaesthetized mice. The results pave way for development of a variety of murine models of acoustic communication.

Mismatch response has been recorded in anaesthetized guinea pigs and rats to changes in repetitive vowels (Kurkela, Lipponen, Hämäläinen, Näätänen, & Astikainen, 2016, syllables (Ahmed et al., 2011Kraus et al., 1994;McGee et al., 2001) and syllable patterns (Astikainen, Mällo, Ruusuvirta, & Näätänen, 2014). However, it has not been investigated whether the MMR is elicited to spectrotemporally complex sounds in mice. A mice model for MMR to such sounds would be particularly applicable for instance to determine the effect of language-related Foxp2 gene in change detection (Enard et al., 2009). Here we investigated detection of changes in vowel duration in anaesthetized mice. Epidural recordings of auditory-evoked potentials were obtained in anaesthetized mice exposed to changes in the duration of a sound (110,120,130,140,150,160,170 or 180 ms vs. 200 ms standard sound). Only decrements in sound durations were applied because increments in the sound durations would have elicited larger responses due to the higher stimulus energy delivered by deviants than standards. In addition to the oddball condition, we used an equiprobable control condition (Schröger & Wolff, 1996) to probe whether the MMR, if observed, reflected the detection of violations of standard sound durations (Näätänen et al., 2005) or the actual stimulus rate (May & Tiitinen, 2010).

| Animals
The experiments were approved by the Finnish National Animal Experiment Board (ESAVI/10646/04.10.07/2014), and they were carried out in accordance with the European Communities Council Directive (86/609/EEC) on the care and use of animals in experimental procedures. Eight male (n = 8) C57Bl/6J (RRID: MGI:5811150) mice from the Animal Center of UEF, Kuopio, Finland were used in the experiment (weight: 24.8 ± 2.0 g; age: 12.1 ± 1.4 weeks; mean ± SEM). The animals were housed in groups, with access to water and food ad libitum in a controlled environment (constant temperature of 22 ± 1°C, humidity of 50%-60%, lights on from 07.00 to 19.00 hr). At the end of the experiment, the anaesthetized animals were killed via cervical dislocation.

| Epidural recordings of auditoryevoked potentials
For the recordings of auditory-evoked potentials, the animals were first anaesthetized with isoflurane (an initial dose of 5% in 1 L/min of pressure air, followed by a maintenance dose of 1%-2% in 1 L/min of pressure air). The animal's head was fixed to a stereotactic device (Kopf, CA, USA) equipped with an external supporting arm that allowed the removal of an ear bar. Immediately, a single dose of urethane (Sigma-Aldrich, MO, USA; 7.5 g/100 ml solution) was administered intraperitoneally (1.2 g/kg), and the administration of isoflurane was gradually reduced over a 5-min period (Barth & Mody, 2011). The level of anaesthesia was controlled by regular testing of the pedal withdrawal reflex. If required, extra doses (0.1-0.2 ml) of urethane were administered.
Under full anaesthesia, the skin and muscle tissue over the skull were removed. For the reference electrode, a hole was drilled in the skull over the right side of the cerebellum, and a small insulin needle (BD Lo-Dose syringe; USA) was inserted in the cerebellum as the reference electrode (AP: −5.8 mm, ML: 2-3 mm and DV: 2 mm). A similar needle electrode inserted subcutaneously into the neck served as the ground electrode. Next, the upper half of the squamosal skull bone was removed to reveal the left primary auditory cortex. The dura was left intact (Stiebler, Neulist, Fichtel, & Ehret, 1997).
The tip of a Teflon-insulated silver wire (A-M Systems, WA, USA) with a diameter of 200 μm was placed on the surface of the dura above the auditory cortex (2.2-3.2 mm posterior and 4-4.5 mm ventral to bregma). Although the locations of the electrodes in the craniotomy were intended to capture primary auditory cortex activity, some activity from higher | 1913 LIPPONEN Et aL. sensory areas may also have been captured. The latter is due to differences between animals in the organization of the primary auditory cortex and the relatively large size of the tip of the electrode allowing signals to be conducted from adjacent areas. To reduce the variability in the electrode location, the location of the electrode was guided by online recorded auditoryevoked potentials in response to a stimulus of 200 ms, which was presented as the standard tone in the actual experiment.
A continuous electrocorticogram was first 10-fold amplified using a low-noise MPA8I pre-amplifier (MultiChannel Systems MCS GmbH, Germany). The signal was further fed to a filter amplifier (FA64I, filter: 1-5,000 Hz, MultiChannel Systems MCS GmbH, Germany). All the signals were digitized (USBME-64 System, MultiChannel Systems MCS GmbH, Germany) and recorded using McRack software (MultiChannel Systems MCS GmbH, Germany) at a 2,000-Hz sampling rate. Finally, all the signals were digitally band-pass filtered between 1 and 500 Hz (high-pass: low-pass: fourth-order Bessel).

| Stimuli
The speech sound /a/ was used in the experiment. The speech sound was originally recorded at a sampling rate of 44.1 kHz using a female native Finnish speaker. The sound was then digitally edited using SoundForge software (SoundForge 10, Sony Corporation, Japan) to ensure it had a constant duration of 200 ms. The 200-ms speech sound was digitally shortened in 10-ms steps, and a 5-ms fade-out was added in each sound in Soundforge Pro 10.0 (MAGIX Software GmbH, Germany). This resulted in 10 speech sounds (200,190,180,170,160,150,140,130,120 and 110 ms in duration), which were used in the experiment (Figure 1).
The sounds were presented via an active loudspeaker system (Studiopro 3; N-audio, CA, USA), and the presentation of each stimulus was controlled by E-prime 2.0 software (Psychology Software Tools, PA, USA). The stimuli were presented using the passive part of the loudspeaker system directed towards the right ear of the animal at a distance of 20 cm, with a sound pressure level of 70 dB, as measured using a C-weighted sound level metre (Sound level metre Type 2240; Brüel & Kjaer Sound & Vibration, Denmark).
Two different types of stimulus conditions were presented to each animal: an equiprobable condition and an oddball condition ( Figure 2). In one stimulus block, the equiprobable condition was applied. Under this condition, all 10 speech sounds were presented with the same probability (p = 0.10) in random order at an inter-stimulus interval of 400 ms. In the oddball condition, nine stimulus blocks were applied. Under this condition, the durations of the deviant sounds were 110, 120, 130, 140, 150, 160, 170, 180 and 190 ms. Due to a technical failure, the oddball stimulus block with the 190-ms deviant tone is not reported. In each oddball condition, the 200-ms speech sound served as the repeatedly presented standard sound (p = 0.90), and one of the nine shorter sounds served as the deviant sound (p = 0.10). When delivering the sounds, the deviant sounds were interspersed with the standard sounds in a random order, but there were always at least two standard sounds between two deviant sounds. There were 1,000 stimuli in each stimulus block, and these were presented in a counterbalanced order between the subjects.

| Data analysis
The data analyses were performed offline using Brain Vision Analyzer (Brain Products, Gilching, Germany), GraphPad Prism 5.03 (GraphPad Software), Excel 2016 (Microsoft Office), Rstudio v. 1.1.456 and Matlab R2017b. The data were segmented (from −50 to 400 ms from stimulus onset) separately for the responses to the deviant and standard sounds immediately preceding the deviant sound in the oddball condition and for the responses to each tone in the equiprobable condition. The data segments were baseline corrected against the mean of the signal during a 50-ms time window prior to sound onset. The data segments were averaged for each animal separately for the different deviant and standard tones preceding the deviant tones and for each tone presented in the equiprobable condition.

| Statistical analysis
Paired timepoint-by-timepoint t tests were applied to compare the averaged response amplitudes. The averaged response amplitudes of the responses to the standard and deviant sounds for each oddball condition were first compared. If the timepoint-by-timepoint comparison values reached statistical significance (p < 0.05) for at least 20 consecutive sample points (i.e. over a 10-ms time span), differential responses to the oddball deviant sounds were considered to exist and not to be accounted for by random signal fluctuations (Guthrie & Buchwald, 1991). See below for data-driven confirmation of this. Next, if a significant difference was observed between the standard and deviant responses in the oddball condition, the oddball deviant response and control-deviant response (i.e. response to the same sound presented in the equiprobable condition) were compared. Due to technical problems, the recording of the equiprobable control condition in one animal could not be recorded. Therefore, the comparisons of the responses in the oddball and control conditions were based on those of seven animals. Two-tailed t tests were used to compare the standard and deviant responses in the oddball condition because we were not able to predict the direction of the potential differential response. The latter was not possible because the location of the reference electrode and anaesthesia in experimental animal models can affect the polarity of the differential response (Harms et al., 2016). The t tests comparing deviant and control-deviant responses were one-tailed because we specifically tested, whether the responses were larger for the deviant than for the control-deviant sounds. Cohen's d is reported for the effect size.
In addition, to provide data-driven estimate of the probability of finding a random consecutive sequence of p-values below 0.05 of the same length as that found in the present data, randomization procedure was applied. First, the labels of the standard and deviant conditions were assigned randomly for each animal and a timepoint-by-timepoint twotailed t test was calculated. This was carried out 56 times (the maximum number of permutations for eight animals). Second, the maximum number of consecutive time points with p < 0.05 was extracted for each permutation. Third, these values were sorted into ascending order and the original number of time points was compared against this vector providing an estimate of the probability for randomly finding equally long significant time window.
To investigate whether the differential response was significant in individual animals, we performed paired timepoint-by-timepoint t tests for single-trial data. For each animal, the amplitude values for the standard and deviant responses in each trial were compared. Similarly, the responses of each animal to the deviant sounds and control sounds were compared whenever difference was found between deviant and standard responses. Again, at least 20 consecutive F I G U R E 2 Stimulus series used in the experiment. The tones were presented in the oddball condition (a) and in the equiprobable condition (b). In the oddball condition, a deviant sound (p = 0.1) was interspersed with a repetitive standard sound (p = 0.9). In the equiprobable condition, control stimuli of 10 different frequencies were presented with the same probability (p = 0.1) sample points we expected to significant for a robust effect. Similarly, to the group-level analysis, two-tailed tests were applied for the standard versus deviant response comparisons and one-tailed tests for deviant versus control-deviant response comparisons. Individual-level of analysis was conducted for those deviant sound durations that showed a significant differential response in oddball condition at the group-level analyses.

| RESULTS
Overall, the sounds evoked a prominent response, with positive polarity peaking approximately 35 ms after the stimulus onset (Figure 3). A visual inspection revealed a larger response of positive polarity to the deviant sound in comparison to the standard sound (200 ms) in the oddball condition, especially to the shortest (110 ms) deviant sound (Figure 3).
At group-level, the timepoint-by-timepoint t tests revealed a statistically significant difference in the amplitude values in response to the standard and 110-ms deviant sounds 53-259 ms post-change (all tests, p < 0.05). The effect size was largest 117 ms post-change (Cohen's d = 1.56). The randomization procedure used to validate this finding showed that the probability of finding 206 ms long time window by chance is below 0.017. Deviant sounds of longer durations than 110 ms, and therefore higher similarity to standard sounds, did not elicit responses different in amplitude from those to standard sounds as analysed with the timepoint-by-timepoint t tests (Figure 3).

F I G U R E 3 Grand-averaged responses
recorded from the dura above the auditory cortex. (a) Evoked potentials for the deviant (110, 120, 130, 140,150,160, 170 and 180 ms) and to the 200-ms standard sound (n = 8). The sounds were presented in the oddball condition. (b) Evoked potentials to the 110-ms deviant and 200-ms standard stimulus in the oddball condition (n = 8). This was the only deviant condition where responses between the standard and deviant sounds were significantly different in amplitude. The point-by-point t tests to response amplitudes showed that these responses differed (p < 0.05) at latency of 163-369 ms (corresponding to 53-259 ms after the change onset). The grey bar refers to 95% confidence interval (CI) for the deviant-standard difference from the t test without zero. (c) Evoked potentials to the 110-ms deviant sound (oddball condition, n = 8) and corresponding control sound (equiprobable condition, n = 7). Stimulus onset at 0 ms. The waveforms are presented as mean (a) and mean ± SEM (b and c). [Colour figure can be viewed at wileyonlinelibrary.com] Next, the responses to the 110-ms deviant sound in the oddball condition were compared in amplitude to those to the corresponding control-deviant sound in the equiprobable condition. There was no difference in the response amplitudes elicited by the deviant and control-deviant sound. The effect size was largest 227 ms post-change (p = 0.06, Cohen's d = 0.8). Figure 3 shows the p-values for the whole response time.
Single trial analysis showed that responses to the 110-ms deviant sound and those to 200-ms standard sound in the oddball condition differed in five of eight animals. The same analysis revealed that responses to the 110-ms deviant sound and 110-ms control-deviant sound differed only in one of seven animals (Table 1 and Figure 4).

| DISCUSSION
In this study, we investigated whether the MMR analogous to MMN in humans, is elicited in anaesthetized mice in response to changes in the durations of spectrotemporally complex sounds (a human speech sound /a/). We studied not only whether such changes elicited differential responses to rare sounds in comparison to frequently presented sounds but also whether these differential responses reflected the detection of sound duration changes rather than the lower presentation rate of the deviant sound relative to that of the standard sound (May & Tiitinen, 2010). To shed light on the aforementioned issue, we used an equiprobable condition (Harms et al., 2016) as a control condition for the oddball condition.
In the oddball condition, we found a differential response only to the shortest (110-ms) deviant sound interspersed with the 200-ms standard sound (45% duration decrease). As shown by the analysis of the individual-level data, five of eight animals showed a significant differential response to the 110-ms deviant sound. Consistently, freely moving mice have been found to generate a differential response to a 50% reduction (from 100 to 50 ms) in the duration of a repeated sinusoidal sound (Umbricht et al., 2005). In another study (Roger, Hasbroucq, Rabat, Vidal, & Burle, 2009), freely moving rats showed a differential response to a 33% (100-ms deviant vs. 150-ms standard) but not a 16% change in the duration of a sinusoidal sound. In contrast, robust MMN has been detected in humans, even to 10% change in sound duration (100 ms vs. 110 ms, Jaramillo, Paavilainen, & Näätänen, 2000). Thus, the neurophysiological detection of sound duration changes in rodents seems not to be as advanced as it is in humans.
The current experiments were conducted in anaesthetized animals, and it is not known how anaesthesia affects representation of sound duration in spectrotemporarily complex sounds. However, previous studies applied in urethane-anaesthetized rodents have shown that MMR is elicited to changes in vowels (Kurkela et al., 2018), syllables (Ahmed et al., 2011) and syllable patterns (Astikainen et al., 2014). Future studies should investigate MMR to complex sounds in awake and anaesthetized animals to reveal possible effect of anaesthesia on neural change detection.
We also investigated whether the differential response reflected the detection of regularity violations (i.e. a true MMR) or merely the different presentation rate of the stimuli. At the group-level analysis, we found no difference between the responses to the deviant and control sound (110 ms sounds). When single trial analysis was conducted for each animal separately, we detected a significant difference in the responses to the 110-ms deviant and corresponding control sound in only one of the seven animals. This pattern of results suggests that the differential response in the oddball condition could be explained by the different presentation rates of the standard and deviant stimuli (May & Tiitinen, 2010). This finding is in contrast with results in humans showing larger responses to deviant than control sounds (Jabobsen & Schröger, 2003). While the human data suggest that duration changes are detected as violation in regularity (Jabobsen & Schröger, 2003), our data suggest that in mice the detection of duration deviations could be processed as physical feature encoding.
Differential responses to deviants were reasonably long in duration considering that no comparison mechanism could be postulated to precede these responses. However, at the auditory cortical level, such slow responses are expected. Rodent auditory pathway has been under extensive studies related to spectrotemporally complex sounds. By applying dynamically changing stimuli in which temporal modulations were defined by the speed and direction of the amplitude peaks' change, it has been found that temporal response has been faster in thalamic than cortical cells. Namely, the thalamic and cortical neural populations seem to be simultaneously active for most of their response durations, but cortical neurons seem to have longer response latencies than the thalamic neurons (Miller, Escabí, Read, & Schreiner, 2002). In summary, the results show that mice can detect changes in the durations of spectrotemporally complex sounds if the difference in the durations of these sounds is large (at least 45%). The neural detection of rare sounds can be based on different presentation rates of rare and frequent sounds, leading to different levels of neural refractoriness or adaptation in the neural populations responding to these sounds. F I G U R E 4 Difference waves for the oddball (a, deviant minus standard) and control (b, oddball-deviant minus controldeviant) responses in individual animals and group mean (mean) for the 110-ms deviant condition. Statistical analysis revealed that five (out of eight) animals showed significant difference in the oddball condition between the standard and deviant responses, and that only one (out of seven) animals showed significant difference between the deviant and control-deviant responses. Stimulus onset at time 0 ms, and change onset at 110 ms (dotted line) (c)