Dysfunction of specific auditory fibers impacts cortical oscillations, driving an autism phenotype despite near‐normal hearing

Autism spectrum disorder is discussed in the context of altered neural oscillations and imbalanced cortical excitation–inhibition of cortical origin. We studied here whether developmental changes in peripheral auditory processing, while preserving basic hearing function, lead to altered cortical oscillations. Local field potentials (LFPs) were recorded from auditory, visual, and prefrontal cortices and the hippocampus of BdnfPax2 KO mice. These mice develop an autism‐like behavioral phenotype through deletion of BDNF in Pax2+ interneuron precursors, affecting lower brainstem functions, but not frontal brain regions directly. Evoked LFP responses to behaviorally relevant auditory stimuli were weaker in the auditory cortex of BdnfPax2 KOs, connected to maturation deficits of high‐spontaneous rate auditory nerve fibers. This was correlated with enhanced spontaneous and induced LFP power, excitation–inhibition imbalance, and dendritic spine immaturity, mirroring autistic phenotypes. Thus, impairments in peripheral high‐spontaneous rate fibers alter spike synchrony and subsequently cortical processing relevant for normal communication and behavior.


| INTRODUCTION
Autism describes a constellation of early-appearing social communication deficits. 1Though it is thought to have a genetic basis, with Fragile X Syndrome being the leading inherited cause, a majority of cases are idiopathic. 24][5] This becomes evident through reduced frontal electroencephalographic (EEG) power across all brain oscillation frequencies, a characteristic neural marker in autistic infants. 6,7[10][11] Therefore, the pathological features of autism are generally discussed in the context of reduced cortical stimulusdriven PV-interneuron (IN) function. 3,5,124][15] Some studies discuss E/I imbalance in neuropsychiatric disorders, including autism, in the context of impairments of PV-IN progenitors migrating to the telencephalon. 167,18 As cortical PV cells determine the sequential nature of critical periods and related maturation of sensory modalities, 5,19 both theories focus on dysfunctional PV cells in the cortex as a causal factor of autism.
We previously described 20 an autistic-like phenotype in mice in which brain derived neurotrophic factor (BDNF) was deleted in Pax2-lineage descendants in hindbrain regions (Bdnf Pax2 KO).This phenotype included deficits in learning, social behavior, and anxiety. 20The deletion, however, left numbers of cortical and hippocampal PV-INs and BDNF levels in frontal brain regions unaffected. 20,21s BDNF drives synaptogenesis of the cortical PV-IN network, 22,23 normal BDNF levels in frontal brain regions of Bdnf Pax2 KO mice exclude a contribution of cortical BDNF to the autism phenotype.Bdnf Pax2 KO mice also exhibited normal basic hearing thresholds, but lower amplitude and delayed supra-threshold auditory nerve responses, linked to a reduced dynamic range, elevated spontaneous firing rates (SRs), delayed first-spike latency, and impaired inhibitory strength of high-frequency sidebands in neurons of the dorsal cochlear nucleus (DCN) 20 and the inferior colliculus (IC). 21In addition, despite normal numbers of PV-INs, Bdnf Pax2 KO mice showed reduced dendritic PV staining in the auditory cortex (AC) and hippocampus (HC), which correlated with deficits in learning and social behavior and with increased anxiety, altogether displaying an autistic-like phenotype. 20ubpallium-derived GABAergic IN precursors, which have been discussed in the context of autism, 16,24,25 are positive for the transcription factor Pax6 26 and migrate to the telencephalon.8][29] This might suggest deficits relating to the autistic phenotype in Bdnf Pax2 KO mice may originate in the brainstem rather than in the frontal cortical brain regions.
Thus, our overarching goal was to determine whether the autistic phenotype observed in Bdnf Pax2 KO mice could be caused primarily by a peripheral deficit in the lower auditory brainstem regions.We investigated this by addressing the following points: (i) The first aim was to examine whether Bdnf Pax2 KO mice display typical changes of brain oscillations or other signs of E/I imbalance that are commonly seen in autism spectrum disorders. 6,7ii) The second aim was to explore whether the source of these cortical changes in Bdnf Pax2 KO mice can be attributed to a peripheral origin.We aimed to identify the source of altered cortical brain activity in the periphery of Bdnf Pax2 KO mice with normal basic hearing function.
Addressing (i), we analyzed cortical local field potentials (LFPs) in the AC, prefrontal cortex (PFC), visual cortex (V1), and HC in response to auditory stimuli of behaviorally relevant frequencies and intensity levels, amplitude-modulated auditory steady-state responses (ASSRs), and stimulusevoked (phase-coherent), induced (phase-incoherent), and spontaneous (resting-state) brain oscillations. 30As we present auditory stimuli, V1 was used as a control region which should not directly respond to sound stimuli but may still be affected by intracortical feedback.Additionally, markers of E/I imbalance, such as vesicular glutamate transporter 1 (vGlut1) and PV, were analyzed in addition to spine morphology upon Golgi's staining.
Addressing (ii), we analyzed compound action potential (CAP) thresholds and peristimulus time responses (PSTRs) in the same frequency ranges presented during LFP recordings.Both approaches provide information about the spike synchronization within auditory fibers and might identify relations between peripheral and cortical responses.
A connection was observed between deficits in synchronized response behavior of high-SR auditory nerve fiber (ANF) processing and reduced evoked and enhanced induced EEG power in Bdnf Pax2 KO mice.

| Animals
The care and use of mice and the experimental protocol were reviewed and approved by the University of Tübingen, Veterinary Care Unit, and by the Animal Care and Ethics Committee of the Regional Board of the Federal State Government of Baden-Württemberg, Germany, and followed the guidelines of the European Union Directive 2010/63/EU for animal experiments.Bdnf Pax2 KO and control mice were obtained by crossing the Pax2 Cre mouse line 31 with the Bdnf flox mouse line. 32Both lines were obtained from the Mutant Mouse Regional Research Center, MMRRC. 33For all experiments, adult mice between ten weeks and six months old of either sex were used.

| Tissue preparation
Tissue preparation was carried out as previously described in detail. 34In brief, brains were dissected and fixed in 2% paraformaldehyde for 48 h and then stored in 1% paraformaldehyde at 4°C until sectioned.Brains were sectioned at 40-60 μm on a VT1000S Leica vibratome (Leica, Wetzlar, Germany).Sections were stored at −20°C in cryoprotectant (150 g of sucrose in 200 mL 1× PBS and 150 mL ethylene glycol, volume adjusted to 500 mL with 1× PBS) until use.

| Immunohistochemistry
Immunohistochemistry was carried out as described in detail previously. 35Sections were selected according to the mouse brain atlas 36 focusing on hippocampal regions between −1.7 and −2.45 bregma.In brief, slices were washed in PBS (pH 7.2) followed by a permeabilization and blocking step using 3% BSA in PBS containing 0.2% Triton-X 100.Primary antibodies against PV (rabbit, diluted 1:2000, Abcam, Cambridge, UK) and vGlut1 (guinea pig, diluted 1:1000, Synaptic Systems, Göttingen, Germany) were diluted in 1.5% BSA in PBS containing 0.1% Triton-X 100.Slices were incubated with primary antibodies overnight at 4°C.After removing the antibody solution, slices were washed in PBS and primary antibodies were detected using secondary antibodies Cy3 (anti-rabbit, 1:1500, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and Alexa 488 (anti-guinea pig, 1:500, Molecular Probes, Eugene, OR, USA) diluted in 1.5% BSA in PBS containing 0.1% Triton-X 100.After 1h incubation with secondary antibodies at room temperature, slices were washed again in PBS and finally embedded using Vectashield mounting medium with DAPI.
Stained samples of 10 control mice and 12 Bdnf Pax2 KO mice were viewed using an upgraded Olympus BX61 microscope (EVIDENT Europe GmbH, Hamburg, Germany).Images were acquired using a Hamamatsu ORCA-Flash4.0LT PLUS monochrome camera (Hamamatsu Photonics K.K., Herrsching, Germany) and analyzed with cellSens Dimension software (OSIS GmbH, Münster, Germany).To increase spatial resolution, slices were imaged over a distance of ∼15 μm within an image stack along the z-axis (z-stack), followed by 3-dimensional deconvolution using cellSens Dimension's built-in algorithm.

| Toluidine blue staining
Electrode position was confirmed using toluidine blue staining, using 1% toluidine blue in acetate buffer (pH 3.9).Slices were washed for five minutes with PBS and then submerged in the toluidine blue solution for 30 seconds before being washed two times with double-distilled water.

| Co-localization of mRNA and protein in brain sections
Sections were selected according to the mouse brain atlas 36 between −1.7 and −2.45 bregma.mRNA and protein were co-localized on free-floating brain sections as previously described 37 to enable the direct correlation of changes in expression patterns of excitatory and inhibitory markers that are mainly visible on either mRNA (activity-regulated cytoskeletal protein (Arc)/BDNF) or protein (PV) level.Sections were incubated overnight with BDNF or Arc riboprobes as previously described. 20For protein detection, sections were incubated overnight at 4°C with a primary antibody against PV (anti-rabbit, 1:500, ab11427, Abcam, United Kingdom), followed by incubation with the secondary antibody (biotinylated goat anti-rabbit, BA-1000, Vector Laboratories).
All samples were viewed using the upgraded Olympus BX61 microscope (EVIDENT Europe GmbH, Hamburg, Germany).Images were acquired using a DP 71 brightfield camera (EVIDENT Europe GmbH).

| Electrocochleographic recordings
Electrical potentials of ANFs were examined in anesthetized nine control mice and ten Bdnf Pax2 KO mice by electrocochleography.The mice were anesthetized as described above, and then, 20-40 μL Xylocaine 2% (AstraZeneca, Wedel, Germany) was applied subcutaneously at sites of surgical incisions.Mice were laid on a prewarmed resting pad (37°C).The bony auditory bulla was exposed by cutting the skin behind the ear and carefully moving muscles, nerves, and connective tissues aside.A small hole (0.6 mm diameter) was drilled into the bulla in order to access the round-window niche of the cochlea.A silver wire electrode insulated with varnish and silicone that had a small silver bead at the tip was placed within the niche.The skin above the ear was closed, and the mouse was placed in the sound-attenuating booth in front of a loudspeaker for recording.CAP threshold responses from the auditory nerves were measured by stimulation with short tone pips (3ms duration including 1ms on-and off-ramp cos 2 -shaped, 32-96 repetitions with stimulus interval 16 ms and alternating polarity) presented with 5 dB 12 incremental steps from 0 to 100 between 2 and 34 kHz.Electrical potentials were amplified (80 dB) and filtered between 0.2 and 5 kHz before being sampled at 20 kHz A/D rate, averaged, and saved to file.Thresholds were determined from individual ears from averaged waveform responses as the lowest sound pressure level (SPL), resulting in a signal visually distinguishable from noise.
For the CAP latency, electrical responses were recorded for 100 μs click stimuli of 0 to 100 dB SPL.Responses were amplified, filtered (DC, 50 kHz low pass), sampled at 100 kHz A/D rate, and averaged for 64 repetitions (interstimulus interval (ISI) 50 ms).For CAP input-output analysis, the averaged waveform was manually inspected for the first negative amplitude deflection after stimulus onset.
PSTRs were measured and analyzed as described. 38The PSTR stimulus, a 200 ms pseudo-randomized noise burst with a 2.5 ms ramp, was presented with alternated polarity and 100 iterations.The noise had a center frequency of 5.6, 8, 11.3, 16, or 22.6 kHz and a bandwidth of 1/3 of an octave.The intensity ranged from 10 to 80 dB SPL in 10 dB steps.The ISI was 420 ms, and the recording window was 410 ms long.The signal was hardware filtered (bandwidth with 1 kHz high pass and 30 kHz low pass, and the signal was amplified by factor 5000.For analysis, the two electrical recordings within each polarity pair were averaged for isolation of the neurophonic signal.The signal was digitally bandpass filtered with 300-1200 Hz, rectified, and finally smoothed (window of 1 ms).For further analysis, the maximum amplitude within the first 30 ms after stimulus onset (peak) and the average of the amplitude within the last 30 ms before stimulus offset (plateau) were calculated and used for determination of the peak/plateau ratio.

| Cortical local field potentials
Eight control mice and nine Bdnf Pax2 KO mice were anesthetized (see above); the fur in the surgical area was removed, and the skin was disinfected with Octenisept® (Schülke & Mayr GmbH, Norderstedt, Germany).20-40 μL Xylocaine 2% (AstraZeneca, Wedel, Germany) was applied subcutaneously at sites of surgical incisions for local anesthesia, and the mice were laid on a prewarmed resting pad (37°C).Breathing was monitored and supported by oxygen supply.The skin was cut ~12 mm with a scalpel, and connective tissue was removed from the surface of the skull.A 0.6 mm hole was drilled in the skull 1 mm posterior to lambda.A handmade silver wire electrode (0.125 mm) insulated by varnish and silicone and ending in a small silver bead was placed within the niche and glued with Histoacryl (B.Braun, Melsungen, Germany) was used as a reference electrode.
Two electrodes similar to the previous one were placed on the cortex's surface after drilling holes above the right PFC (0.2 mm lateral and 2 mm anterior to bregma) and the left V1 (2.5 mm lateral and 2 mm posterior to bregma).For local field potentials of the right AC (4 mm lateral, 3 mm posterior, and 2 mm ventral to bregma) and the left HC (1.5 mm lateral, 2 mm posterior, and 1.3 mm ventral to bregma), a 1 mm drill was used and the 0.125 mm silver wires were inserted and glued into 27G cannulas used as guiding shafts.Again, the electrodes were glued with Histoacryl to the edges of the drilling hole.All coordinates for the electrodes were chosen according to the mouse brain atlas, 36 and the electrode positions were verified ex vivo in brain slices stained either using toluidine blue or the staining protocol used for double detection of mRNA and protein 37 (Figure S1).All electrodes were connected to an active head stage (LabRat AC16LR, Tucker Davis Technologies, Alachua, FL, USA).The signal was transferred to a programmable gain amplifier (PGA16 Rev.B, Multichannel Systems MCS, Reutlingen, Germany).The PGA, supporting frequency bands between 5 Hz to 5 kHz, was used for 5000x amplification.The PGA was connected to a multi-I/O measurement card (NI PCIe-6321, National Instruments, Austin, TX, USA) housed in a personal computer.

| Resting-state activity
The ~17.3 min of resting-state recording were subdivided into ~1000 epochs with a duration of 1 s.The Fourier transform of each epoch was computed using the FFT function of Matlab (version R2021b).Afterward, the absolute value of the FFT of all epochs was averaged.The final step was to compute the genotype group averages.

| Auditory-evoked potentials
For measurements of auditory-evoked potentials, puretone frequency stimuli were presented at either 5.6, 11.3, or 22.7 kHz with an intensity of ~90 dB SPL.To compare the evoked signals with baseline activity, another recording without a stimulus was performed.The repetition rate for each stimulus condition was 2000, the stimulus length was 100 ms, and the recording interval was 507 ms.
For all cortical measurements of ASSR, a carrier of 11.3 kHz was chosen.For the I/O function, the carrier was modulated by frequencies of either 10, 40, or 128 Hz at 100% modulation depth.For each modulation frequency, the stimulus was presented at intensities from 10 to 90 dB SPL in 10 dB steps.For the modulation transfer function, the carrier was modulated by 5, 10, 16, 20, 32, 40, 64, 80,  128, 160, 256, 320, 512, 640, 1024, 1280, and 2048 Hz at 100% modulation depth.The stimulus intensity was 90 dB SPL.The repetition rate for each ASSR stimulus condition was 64, the stimulus length was 1000 ms, and the recording interval was 1207 ms.

| Golgi-Cox staining
Brains of 10-to 12-week-old mice were dissected and stained according to the manufacturer's guidelines using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA).In brief, dissected brains were immersed in A:B solution for 10 days at room temperature in the dark.Brains were transferred to solution C and kept in the dark for three more days.Afterward, coronal vibratome sections were made (VT1000S, Leica, Wetzlar, Germany) at a thickness of 150 μm and mounted on gelatin-coated microscope slides (FD NeuroTechnologies).The sections dried overnight at room temperature in the dark and were stained according to the manufacturer's guidelines.For embedding, Eukitt Quick-hardening mounting medium (Sigma-Aldrich, Darmstadt, Germany) was used.
Individual pyramidal neurons of the hippocampal CA1 region were included for analysis if (i) completely impregnated and (ii) the integrity of the dendritic branches was preserved.The first dendritic branch from the main apical shaft was analyzed if uninterrupted by other dendrites and at least 20 μm long.Z-stack images were acquired using an upgraded Olympus BX61 microscope (EVIDENT Europe GmbH, Hamburg, Germany) with 60× magnification.On average, four to eight neurons were included per animal for each brain region.

| Statistics and numbers
All statistical information, including the statistical tests and post hoc tests used, the exact value of n, what n represents and the precision documentation of statistical outcome, can be found in Table S1.In brief, data were tested for normal distribution (Shapiro-Wilk normality test, α = 0.05).Differences between the means were compared for statistical significance by unpaired twotailed Student's t-test (parametric) and Mann-Whitney U test for parametric and non-parametric data, respectively.Multivariate sample mean differences were tested by permutation test and 2-way ANOVA or MANOVA for non-parametric and parametric data, respectively.The Bonferroni-Holm method was used to correct for type I error.

| Electrocochleographic recordings
Thresholds were determined from individual ears from averaged waveform responses as the lowest SPL resulting in a signal visually distinguishable from noise.For CAP I/O analysis, the averaged waveform was visually inspected for the first negative amplitude deflection after stimulus onset.The latency of the CAP was registered for each stimulus intensity for each individual ear, and the resulting growth function was averaged and presented as the mean and SEM.PSTR peak/plateau ratios were averaged for all animals within the group and presented as mean and SEM.

| Cortical local field potential analysis
Field potential analysis comprised resampling with the eeglab anti-aliasing filter cutoff function (rate 3000 Hz, cutoff 0.8, band width 0.2), artifact detection and rejection by a variance-amplitude criterion, rejecting the 5% trials with the highest variance, segmentation and transformation of the first 100 of 120 ms stimulus duration (ISI 387 ms) into frequency space by FFT (rectangular window), compute evoked responses be simple time domain averaging, peak-to-peak (min/max within predefined time interval) analysis of voltage values, separating evoked and induced oscillations by phase-coherent/phase-incoherent averaging, and computation of peak-to-peak (min/max within predefined time interval).The frequency bands were analyzed for the following bands: alpha (8-15 Hz, including 5-7 Hz theta band), beta (15-25 Hz), low gamma (25-35 Hz), mid gamma (35-65 Hz), and high gamma (65-125 Hz); results were tabulated and statistically evaluated by permutation analysis confining data permutation to individual setups, and non-parametric signed-rank testing across genotypes for individual frequency bands and stimulus conditions (4.66, 5.66, 6.66, 10.3, 11.3, 12.3,  21.7, 22.7, and 23.7 kHz).For the statistical analysis, the 9 stimulus frequencies were grouped into low (4.66, 5.66, and 6.66 kHz), mid- (10.3, 11.3, and 12.3), and high (21.7,22.7, a 23.7 kHz) frequencies.The peak-to-peak amplitude stimulus frequency regions were analyzed similarly to the ASSR via permutation, while ensuring that only measurements from the same setup were permuted.

| Dendritic spine analysis
Spine analysis and classification were conducted as previously described, 39 using the Reconstruct software (http:// synap ses.clm.utexas.ed).Briefly, all protrusions that were in direct contact with a given segment of a dendrite were counted as spines.Segments of at least 30 μm long were analyzed for each dendrite.Spine density was calculated as the number of protrusions per μm.The length and the width of spines were then transferred to a spreadsheet. 39 3.5 | Fluorescence analysis of brain immunohistochemistry Z-stacks were analyzed using the free software ImageJ (NIH, Bethesda, MD, USA). Fr each z-stack, the three channels (PV, vGlut1, and DAPI) were split and analyzed separately as a maximum intensity projection over the zaxis.A binary mask was created using the default parameters, and then, a ROI (270 × 270 pixels for 1024 × 1024 images and 540 × 540 pixels for 2048 × 2048 images) was placed on the dendrites of the pyramidal neurons in each single-channel picture.Afterward, the average fluorescence intensity within the ROI was calculated.For analysis, the data were normalized between groups in order to account for variability in staining.For each individual animal (ten control mice and 12 Bdnf Pax2 KO mice), one slice was processed, imaged, and between 3 and 5 pictures were analyzed, the average of which was then used for further statistical analyses.

| Bdnf Pax2 KO mice exhibit reduced auditory-evoked local field potentials
To determine whether Bdnf Pax2 KO mice exhibit cortical activity changes with characteristics of an autism phenotype, sound-evoked LFPs were recorded in adult Bdnf Pax2 KO and control mice in response to pure-tone stimuli between 4 and 23 kHz at 90 dB SPL from the AC, HC, PFC, and V1 (Figure 1; Figure S1; for statistics, see Table S1).
When evoked LFP amplitudes were compared between the two groups in response to 5, 11, and 22 kHz stimuli, it became evident that control mice exhibited a strong event-related potential (ERP) in the AC in response to 11 and 22 kHz auditory stimuli and a much weaker one in response to 5 kHz stimuli, whereas Bdnf Pax2 KO mice had weak ERPs in the AC and the other brain regions in response to all frequencies (Figure 1A).No significant amplitude differences were seen in the PFC, HC, or V1 (Figure 1A).
To further classify evoked LFP responses over the hearing range, LFP amplitudes were plotted against stimulation frequencies (Figure 1B).Bdnf Pax2 KO mice exhibited significantly lower evoked response amplitudes in the AC in comparison with controls, particularly in the best hearing range (i.e., 10-12 kHz) up to the extended frequency range (i.e., 21-23 kHz), most critical for communication and behavior 40,41 (Figure 1B; Table S1).In PFC, HC, and V1, no significant difference was detected for LFP amplitudes in response to any presented frequency (Figure 1B; Table S1).
Taken together, a gradual increase in sound-evoked LFP amplitudes in response to acoustic stimuli in the range of best hearing frequencies in mice (> 10 kHz) is seen in the AC, but not in PFC, HC, or V1.In Bdnf Pax2 KO mice, the sound-evoked LFPs in response to >10 kHz were significantly reduced in the AC, suggesting, when  S1.
considering principles of evoked LFP, 42 that the Bdnf Pax2 KO mice display a profound impairment of input and output intracortical network activity.

| Bdnf Pax2 KO mice exhibit reduced evoked and induced ASSRs and increased spontaneous oscillations
To further determine whether the source of the reduced sound-evoked LFP responses in Bdnf Pax2 KO mice stems from cortical or subcortical nuclei, we analyzed the signal envelope following responses to amplitude-modulated stimuli (ASSRs).ASSRs reflect the summation of phaselocked activity from multiple generators of neural stimuli within the auditory system, including the cochlea, auditory nerve, IC, and AC. 43The contributions of ASSR sources in response to lower (10-40 Hz) modulation rates are dominated by cortical components, 43 while ASSRs in response to higher (>100 Hz) modulation rates are dominated by subcortical components. 43odulation frequencies between 5 and 2048 Hz were used at an 11.3 kHz carrier frequency.ASSR amplitudes were depicted for 100% modulation depth at 90 dB SPL and plotted against the frequency oscillation bands (Figure 2; Figure S2).Evoked ASSR activity reflects bottom-up sensory transmission primarily by driving inputs. 43This evoked ASSR activity was distinguished from induced ASSR activities, which are not strictly locked to stimulus onset and rather represent internal dynamics of cortical networks related to higher cognitive functions. 30,43s shown for the AC (Figure 2A), PFC (Figure 2D), HC (Figure S2A), and V1 (Figure S2D), evoked and induced ASSR amplitudes were significantly reduced in Bdnf Pax2 KO mice in response to 10-, 40-, and 128 Hz-modulated tones and remained significantly diminished in comparison with control mice up to 320 Hz modulation frequency (see dashed lines in Figure 2A,D; Figure S2A,D).S1.
These reduced evoked and induced ASSR amplitudes in AC, PFC, V1, and HC suggest an impaired summation of phase-locked activity from subcortical and cortical regions in Bdnf Pax2 KO mice that have impacted cortical networks related to the AC.
To better explain the reduced ASSR amplitudes in Bdnf Pax2 KO mice, we analyzed the dependence of the ASSR amplitude on stimulus intensity.This could shed light on deficits in a specific type of ANF.The high-spontaneous rate (high-SR) low-threshold ANFs dominate the encoding of the envelope of amplitude-modulated tones at low sound intensities in the best frequency range and contribute at higher sound intensities through activation of sidebands of neurons with characteristic frequencies outside of best frequency range. 44SSR amplitudes were plotted against increasing sound intensities from 10 to 100 dB SPL and specifically analyzed for 10-, 40-, or 128 Hz-modulated tones (Figure 2B,E; Figure S2B,E).In the AC, control mice exhibited a characteristic increase in response amplitude with increasing stimulus intensity (>50 dB SPL) for 10-and 40 Hz-modulated tones.This effect was significantly less pronounced in Bdnf Pax2 KO mice (Figure 2B; Table S1).
In both control and Bdnf Pax2 KO mice, ASSR amplitudes in the PFC, HC, and V1 showed a progression identical to the respective AC response, particularly to the higher SPL range.This points to a coherence between 10 and 40 Hz activity within these brain regions in response to higher intensity sound (Figure 2B,E; Figure S2B,E).The induced (non-phase-locked) responses were slightly higher in Bdnf Pax2 KO mice compared to controls specifically for low sound intensities in response to 40 Hz-modulated tones in the AC (Figure 2C; Table S1) and PFC (Figure 2F; Table S1) and in response to 10 Hz-modulated tones in the PFC (Figure 2F).This effect was not present in the HC or V1 (Figure S2C,F; Table S1).
In summary, Bdnf Pax2 KO mice exhibit diminished evoked ASSR amplitudes at higher sound intensities, best explained through missing high-SR ANF-driven activation outside of neurons' characteristic frequency.Furthermore, we find slightly elevated induced ASSR amplitudes at lower sound intensities in Bdnf Pax2 KO mice, best explained through missing high-SR ANFdriven activation in neurons' best frequency range.Overall, this characteristic response profile suggests that Bdnf Pax2 KO mice exhibit disturbed high-SR ANF function, preventing them from properly contributing to the ASSR amplitude.
If the cause of reduced sound-evoked cortical LFPs and diminished evoked ASSRs in Bdnf Pax2 KO mice is of subcortical origin, we would expect spontaneous LFPs (independent of stimuli) to be elevated in Bdnf Pax2 KO mice, as intracortical network activity is modulated indirectly through elevated subcortical baseline activity.
To elucidate this, spontaneous cortical activity was measured in the AC, PFC, HC, and V1 in Bdnf Pax2 KO mice and controls (Figure 3).In the AC, the spontaneous activity of Bdnf Pax2 KO mice was significantly higher than that of controls in the alpha (5-15 Hz), beta (15-25 Hz), lowgamma (25-35 Hz), and high-gamma (35-65 Hz) bands (Figure 3A; Table S1).Interestingly, the spontaneous LFP power was also elevated, although less prominently, in the PFC in the alpha, beta, and low-gamma bands (Figure 3B; Table S1) and in the V1 in the beta and low-gamma bands (Figure 3D; Table S1).No significant differences were found in the HC (Figure 3C; Table S1).
Conclusively, diminished evoked ASSR at higher sound intensities, elevated induced ASSR at lower sound intensities, and elevated LFPs at rest in Bdnf Pax2 KO mice all suggest that the spontaneous ongoing activity at the input site of intracortical networks in the AC has changed.This may influence the operation point for feed-forward synchronization and self-organizing intracortical induced synchronization, both required for proper evoked and induced brain oscillations.

| Bdnf Pax2 KO mice exhibit reduced evoked and enhanced induced alpha, beta, and gamma oscillations in cortical regions
As autism is suggested to be a "critical period" disorder, 3,4 we hypothesized that elevated spontaneous LFP activity in Bdnf Pax2 KO mice is associated with a cortical inhibitory component that failed to mature during the critical period after hearing onset.This could result from underdeveloped near-threshold auditory processing as concluded from the aforementioned results.
If this were the case, we would expect a particularly profound impact on the alpha and beta neuro-oscillatory bands, as they are generated earlier than the higher frequency gamma bands in humans and rodents, coinciding with the differential onset of sensory modalities driven by the establishment of PV-IN networks. 5,10e therefore tested specific frequency-band power differences in response to sound frequencies increasing from 4 to 23 kHz in Bdnf Pax2 KO and control mice.Evoked and induced LFP responses to increasing stimulus frequencies are distinguished and shown for each neuro-oscillatory frequency band separately in the AC (Figure 4), PFC, HC, and V1 (Figure S3) of Bdnf Pax2 KO mice and controls.In comparison with controls, Bdnf Pax2 KO mice exhibited significantly lower evoked LFP responses in the alpha, beta, and low-gamma bands, particularly in response to higher sound frequencies (>6 kHz) (Figure 4A-C; Table S1).This effect was less prominent in the mid-gamma band (Figure 4D; Table S1) and not present in the high-gamma band (Figure 4E; Table S1).This indicates that soundevoked EEG power is diminished across a wide range of frequency bands in Bdnf Pax2 KO mice.
Although the amplitude sound-evoked LFP was significantly lower only in the AC of Bdnf Pax2 KO mice (Figure 1), when the evoked LFP was divided into specific oscillatory frequency bands, differences in other brain regions became apparent.In the PFC, HC, and V1, Bdnf Pax2 KO mice exhibited significantly lower evoked LFPs compared to controls mainly in alpha and beta bands and particular for stimulation frequencies above 6 kHz (Figure S3; Table S1).
Induced LFPs were also examined across different oscillatory frequency bands and overall stimulus frequencies.In the course of this analysis, Bdnf Pax2 KO mice exhibited higher induced LFPs in comparison with controls particularly in the AC, PFC, and V1 in the beta band (Figure 4; Figure S3; Table S1).
This indicates that the diminished sound-evoked EEG power across <65 Hz frequency bands in the AC and in alpha-and beta-frequency bands in the PFC, HC, and V1 of Bdnf Pax2 KO mice is at least partially accompanied by an enhancement of induced LFP power in the AC, PFC, and V1 in the beta band.Here, an association with the increased ongoing activity of induced ASSR responses to low-SPL stimuli in the AC and PFC becomes apparent.In both cases, elevated spontaneous feed-forward synchronization may have influenced self-organizing intracortical induced synchronization.

| Bdnf Pax2 KOs exhibit deficits in synchronized responses and high-SR ANF processing
Thus far, all results would be in line with an immaturity of high-SR ANFs in Bdnf Pax2 KO mice.High-SR ANFs develop during the critical time period after hearing onset, nearly independent of basic hearing function. 45,46They define the perceptual hearing thresholds at behaviorally relevant frequency regions 47,48 and contribute to level-dependent response properties. 49High-SR ANFs have a high synchronicity and phase-locking capacity 44 and thus contribute to both ASSRs and the CAP amplitude of the auditory nerve. 50,51hus, as an indicator for ANF functionality, the CAP signals were recorded at the round window of  S1.S1.
anesthetized Bdnf Pax2 KO mice and controls (Figure 5A).CAP thresholds showed significant differences between Bdnf Pax2 KO and control mice at frequencies below 11 kHz (Figure 5B; Table S1); these differences mirror the differences in auditory brainstem response thresholds reported in previous studies. 33Interestingly, when analyzed in a threshold-normalized manner, the CAP amplitudes in response to click stimuli were lower in Bdnf Pax2 KO mice at all sound intensities (Figure 5C, left; Table S1), while the latency was unaffected (Figure 5C, right; Table S1).
When amplitude growth functions were measured for different frequencies ranging from 2 to 32 kHz, Bdnf Pax2 KO mice displayed lower CAP amplitudes at frequencies between 8 and 32 Hz (Figure 5F-J; Table S1).Addressing frequency gradients, CAP amplitudes of all measured frequencies were compared at 20 dB above threshold (Figure 5K).Again, CAP amplitudes of Bdnf Pax2 KO mice were significantly lower in response to stimuli at and above the best hearing frequency range (Figure 5K; Table S1).Notably, stimuli of this same frequency range also elicited lower evoked cortical brain activity in Bdnf Pax2 KO mice (Figure 1, Figure 4).Similar effects were observed when comparing CAP amplitudes of all frequencies at 30 (Figure 5L; Table S1) and 40 dB above threshold (Figure 5M; Table S1).
The significantly reduced CAP amplitudes around the best frequency ranges in Bdnf Pax2 KO mice may suggest an inability to properly phase lock and synchronize summed spike activity.As a result, when action potentials are out of phase, this desynchronization destructively adds to the CAP amplitude, which is defined through the summation of all synchronized action potentials.
High-SR ANFs, rather than low-spontaneous rate (low-SR) high-threshold ANFs, were previously shown to be the primary contributor to phase locking and synchronization of auditory nerve activity. 50The proportion of high-SR ANFs to low-SR ANFs can be evaluated by measuring PSTRs.Noise-band stimuli in alternating polarity were presented, and the electrical signal was recorded at the round window; then, the responses were averaged in pairs. 38Using a bandpass filter, the neurophonic response was isolated, then rectified, and averaged across all repetitions (Figure 6A).The PSTR has an envelope shape (Figure 6B) similar to the peristimulus time histograms derived from single-fiber recordings of ANFs.The envelope consists of a synchronized response at the onset of stimulation followed by an alternating current component arising from a phase-locked activity of ANFs to stimulus envelope fluctuations. 38,52The peak-to-plateau ratio, which reflects the proportion of high-to low-SR ANFs, was calculated for all traces and then compared in a threshold-dependent manner.
Bdnf Pax2 KO mice and controls did not differ in their peak-to-plateau ratio for low-to mid-frequency stimuli (i.e., below 11.3 kHz) (Figure 6C-E; Table S1).At higher frequencies (i.e., 16 and 22 kHz), however, Bdnf Pax2 KO mice had a significantly smaller peak-to-plateau ratio in comparison with controls (Figure 6F,G; Table S1).When the peak-to-plateau ratio was compared at all frequencies at 30 dB above threshold, an overall significantly reduced ratio was found in Bdnf Pax2 KO mice (Figure 6H; Table S1).
Conclusively, the lower CAP amplitude along with the lower peak-to-plateau ratio observed predominantly in high frequencies in Bdnf Pax2 KO mice point to a deficit in high-SR ANF function that leads to failed maturation of precise spike synchronization.This high-SR ANF deficit alone may explain the stark reduction in evoked cortical LFPs in the AC, linked to reduced evoked power in the alpha, beta, and low-gamma bands, typical for an autismlike phenotype.This would lead us to expect in Bdnf Pax2 KO also structural changes in cortical and associated regions linked to E/I imbalance, comparable to what has been observed in autistic phenotypes.

| BDNF Pax2 KO mice exhibit impaired spine maturation and PV-IN levels
Reduced cortical inhibition, leading to E/I imbalance in neuropsychiatric diseases, was associated with changes in GABAergic proteins and spine maturation patterns, for example, in the HC (e.g., [53][54][55] ).
Consistent with the notion that cortical brain activity changes in Bdnf Pax2 KO mice are a consequence of altered driving force stemming from the periphery of the cochlea rather than resulting directly from cortical network changes, we want to highlight the previous finding of similar cell counts of PV-INs and similar BDNF mRNA levels in the AC and HC in Bdnf Pax2 KO mice and controls (Figure S4A-C, with the permission of Front Mol Neurosci 20 ).However, PV-IN labeling in the neuropil domain in layer III/IV of the AC, the stratum radiatum, and the stratum lucidum of hippocampal CA1/CA3 regions of Bdnf Pax2 KO mice showed reduced PV-IN levels.This coincided with significantly higher Arc mRNA levels in the same regions (Figure S4D,E; with the permission of Front Mol Neurosci 20 ).
In the present study, the structural correlate of enhanced cortical and hippocampal excitability in Bdnf Pax2 KO mice was validated by measuring the expression of PV protein and vGlut1 protein as markers of inhibitory and excitatory synapses, respectively, in combination with spine maturity, as described. 56Images in Figure 7 were taken of the stratum radiatum (Figure 7A), a hippocampal region where the Schaffer's collaterals contact the dendrites of CA1 pyramidal neurons.Bdnf Pax2 KO mice exhibited significantly lower PV fluorescence levels (Figure 7 B,C; Table S1) and significantly higher vGlut1 fluorescence levels in comparison with controls (Figure 7B,C; Table S1).
Golgi's staining was performed on hippocampal brain slices in order to visualize dendritic spines (Figure 7D).Bdnf Pax2 KO mice exhibited higher spine density, defined as the number of protrusions per μm  S1.
(Figure 7E).In addition, the length-to-width ratio, reflecting different spine morphology, was higher in Bdnf Pax2 KO mice, indicating longer, thinner spines (immature spines) in comparison with controls (Figure 7F-H).Overall, the decreased PV and increased vGlut1 levels along with an immature spine morphology in the hippocampus of Bdnf Pax2 KO mice, which otherwise show normal PV-IN numbers, strengthen the assumption that Bdnf Pax2 KO mice exhibit autism-like structural changes in the HC.
In conclusion, Bdnf Pax2 KO mice exhibit lower CAP amplitudes and a lower peak-to-plateau ratio of the PSTR predominately in response to higher frequency stimuli.This suggests that high-SR ANF processing did not properly develop in Bdnf Pax2 KO, possibly due to reduced inhibitory shaping, leading to poor spike synchronization (Graphic Abstract i, controls: left, Bdnf Pax2 KO: right).Under these conditions, the lack of driving force for tonic inhibition in the ascending pathway would hamper the maturation of inhibitory strength in ascending auditory neurons. 20,21his prevents proper spike synchronization in Bdnf Pax2 KO mice, resulting in coarser receptive fields and poorer temporal coding (Graphic Abstract ii).The subsequently elevated spontaneous feed-forward synchronization through the thalamocortical input (Graphic Abstract iii, input MGB) lowers the operation point for signal-to-noise ratio of stimulus-driven pyramidal neurons and stimulusdriven PV-IN inhibition.As a result, enhanced excitability disrupts the internal dynamics of cortical networks, lowering evoked stimulus-driven oscillatory activity and proper network-centered homeostasis for cortical circuit function of coherently activated brain regions (Graphic Abstract iii, output).

| DISCUSSION
The cause of altered E/I balance and cortical EEG power changes in autism spectrum disorder, if not hereditary, has remained unclear.Through cortical LFP recordings and their spectral analyses, as well as through CAP and PSTR measurements, to analyze the spike synchronization of ANFs, we discovered that a failed maturation of high-SR ANF processing prevents proper establishment of feed-forward inhibitory networks and homeostasis of intracortical networks after hearing onset, providing a new rationale to explain the development of an autistic phenotype.

| Reduced evoked and elevated spontaneous local field potentials in Bdnf Pax2 KO mice
Control mice exhibited a significant increase in LFP amplitude in the AC in response to sound stimuli higher than 6 kHz, which was significantly diminished in Bdnf Pax2 KO mice (Figure 1).This frequency is within the range in which the hearing of mice is the most sensitive and is considered crucial for communication and behavior. 40,41eficits in responses to this frequency range therefore suggest that Bdnf Pax2 KO mice are severely limited in the processing and integration of behaviorally relevant acoustic information close to threshold.An increased sensitivity to the band of frequencies best for communication gradually developed during evolution to improve communication, mainly through an improved ability to hear higher frequencies. 57,58While a failure to process sounds in this best frequency range may have only a minimal effect on the survival and evolutionary fitness of non-lingual species such as mice, a similar deficit in humans would have dramatic consequences for behavior, particularly for language acquisition and social learning, 59 notably associated with autism spectrum disorder. 3,17 link between autism and deficits in processing sounds in the best frequency range becomes particularly clear when considering the altered cortical LFP response in Bdnf Pax2 KO mice.Overall and across oscillatory frequency bands, the most profound differences between Bdnf Pax2 KO mice and controls were seen in response to stimuli higher than 6 kHz.Moreover, Bdnf Pax2 KO mice exhibited differences in evoked and spontaneous LFP responses in partially overlapping frequency bands in distinct brain regions.These were in part also accompanied by altered induced LFP responses.The reduced evoked LFP responses, higher spontaneous LFP power, and partially elevated induced LFP responses observed in Bdnf Pax2 KO mice may share a common origin in an elevated ongoing spontaneous activity in the auditory pathway occurring after failed maturation of high-SR ANFs (Graphic Abstract i).This hypothesis is supported through the following observations: Reduced proportion of high-SR ANFs in Bdnf Pax2 KO mice.(A) Data acquisition was performed by presenting noise-band stimuli in alternating polarity.The round-window recordings of both polarities were averaged in pairs and then bandpass filtered to isolate the neurophonic response.Within a step of rectification, the envelope was calculated and finally averaged for all repetitions.(B) Example trace of a peristimulus time response with a stimulus ranging from 10 to 210 ms during a 400 ms recording window, resulting in a typical slope with a peak and a plateau.(C) The threshold-normalized peak-to-plateau ratio was similar for Bdnf Pax2 KO mice and controls at 1/3 octave noise bands with a center at 5.7 kHz, (D) 8 kHz, and (E) 11.3 kHz.(F) The peak-to-plateau ratio was reduced in Bdnf Pax2 KO mice at 16 kHz and at (G) 22.6 kHz.(H) The peak-to-plateau ratio was reduced in Bdnf Pax2 KO mice when compared for all frequencies at 30 dB above threshold.Mean ± SEM. *p < .05,**p < .01,***p < .001,ns p ≥ .05.Detailed statistical information can be found in Table S1.S1.
(i) Reduced LFP power in Bdnf Pax2 KO mice: The elevated spontaneous LFP activity in Bdnf Pax2 KO mice is comparable to EEG recordings from individuals with Fragile X Syndrome, the leading inherited cause of autism. 60][62] In Bdnf Pax2 KO mice, these findings could be linked to an elevated excitatory baseline activity at the thalamocortical input (Graphic Abstract iii).Typically, in a feed-forward mode of function, LFP power in layer IV is defined by thalamocortical input activity, which transmits the input activity of layer IV to infragranular layer V-VI (output) by a large and early current sink into the supragranular layers I-III. 42,635][66][67] The neurons contributing to output activity are extensively shaped by intracortical neural circuits, 68,69 resulting in significantly sharper frequency tuning of neurons that is constructed by output activity rather than input activity. 63An elevated baseline activity within the ascending auditory pathway that is unable to drive feed-forward inhibition after hearing onset in intracortical (supragranular or infragranular layers) output networks would result in elevated excitatory network responses of output activity.It would thereby provide the explanation for elevated spontaneous LFP power in the AC and coherently activated brain regions in Bdnf Pax2 KO mice (Figure 3).It would be of interest if individuals suffering from Fragile X Syndrome, or other mouse models for autism spectrum disorder show similar deficits of inhibitory strength along the auditory pathway.
(ii) Reduced ASSRs in Bdnf Pax2 KO mice: The altered ASSR is perhaps the strongest indicator of an association between the diminished peripheral input of high-SR ANFs and the abnormal intracortical response pattern and autism-like phenotype in Bdnf Pax2 KO.The ASSR reflects the entrainment of neural activities elicited by periodic auditory stimulation that are sinusoidally modulated in amplitude and frequency.This measurement was used to evaluate sensory and cognitive functions in the central nervous system in numerous clinical studies. 70mportantly, ASSRs reveal the synchronous discharge of auditory neurons phase-locked to the modulation frequency of the stimulus [71][72][73] and thereby critically depend on proper phase-locking capacity along the auditory pathway.The power (magnitude) and phase-locking ability (phase consistency across trials) measured with ASSRs is dependent upon the summation of phase-locked activity from multiple neural generators within the auditory system, including the cochlea, auditory nerve, IC, and AC. 43It is likely that the phase-locked activity of all these neural generators depends on the proper maturation of high-SR ANFs, as this ANF type mainly contributes to the synchronized spike responses within the dynamic range of the auditory nerve response, while low-SR ANFs have limited influence on synchronization. 50Immaturity of high-SR ANFs would lower synchronization of action potentials to fire sufficiently in phase, as was observed in the profoundly diminished CAP amplitudes in Bdnf Pax2 KO mice (Figure 5).This would further hamper synchronized spike rates at stimulus onset when the frequency exceeds 11.3 kHz, as observed through the lower PSTR peak-toplateau ratios in Bdnf Pax2 KO mice (Figure 6).Moreover, high-SR ANFs are not only predicted to have significant effects on level-dependent response properties 74,75 but they have also been suggested to be primarily responsible for determining ASSR amplitudes at higher SPLs by enabling the activation of frequencies outside of the characteristic frequencies. 44High-SR ANFs additionally define the perception thresholds at any given characteristic frequency 47,48,51 and thereby, when functionally impaired, would elevate the detection threshold of ASSR signals at low SPLs (Figure 2C,F; Figure S2C,F).Considering this dual function of high-SR ANFs, their impairment in Bdnf Pax2 KO mice would therefore best explain the severe deficits observed in the ASSR amplitudes in response to stimuli of higher SPLs in the AC and other regions (Figure 2B; Figure S2B,E).The lower evoked and induced ASSRs, observed not only in the AC of Bdnf Pax2 KO mice but also in the PFC, V1, and HC, may suggest a coherence between these brain regions.Coherence between brain regions activated by amplitude-modulated stimuli has previously been shown for the AC and PFC 76,77 and for the AC and V1. 78,79This process is driven by feedback from neurons in AC layer V/VI to the dorsal medial geniculate body and the mediodorsal thalamus, which feeds into the PFC and HC. 80,81In addition, projections from the AC to the V1 have been described. 82n analogy, reduced ASSRs and diminished temporal integration and binding of sensory functions across sensory modalities, including multisensory temporal acuity, have been reported in autism spectrum disorder. 3,83,84hese findings support the hypothesis that observed ASSR network changes in the AC and coherent brain regions may be correlates of the autism-like phenotype in Bdnf Pax2 KO mice.
(iii) Diminished brain oscillations in Bdnf Pax2 KO mice: Evoked oscillations are stimulus-locked activities and reflect bottom-up sensory transmission by primarily driving inputs. 30,85Elevated baseline activity, as here hypothesized to occur in Bdnf Pax2 KO mice, may potentially lower the operation point for stimulus-driven inputs on pyramidal neurons.This imbalance could emerge when tonic inhibition of PV-IN on pyramidal neurons is not properly established during the critical period, as shown previously in subcortical auditory brainstem regions in Bdnf Pax2 KO mice. 20Subsequent to failed maturation of synchronized power to maintain sustained inhibition in the periphery, as here suggested to occur in Bdnf Pax2 KO mice, cortical pyramidal neurons might be unable to build up sufficient functional surround inhibition and generate sufficient feed-forward synchronization in intracortical networks, which would lead to weaker stimulus-evoked oscillations, as also hypothesized in previous studies. 86,87In addition, reduced stimulus-evoked intracortical inhibition would further contribute to changes in induced oscillations.Under these conditions, it is thus plausible that the internal dynamics of cortical networks which are critical for establishing and maintaining oscillatory activity and higher cognitive functions 84,88,89 would be impaired in Bdnf Pax2 KO mice.Fast PV-IN activity in intracortical networks contributes to the generation of oscillatory brain activity during the critical period of sensory experience with slow waves developing earlier than faster ones, 5,9,90-92 explaining the selective effect on EEG power in the alpha and beta bands of Bdnf Pax2 KO mice.Particular alpha oscillations are suggested to be associated with temporal integration of the different sensory modalities that mature in cascades at different time points, driven by PV-IN networks. 3,5As in turn the proper integration of all sensory modalities through brain oscillation, driven by PV-IN networks, is a prerequisite for attentional control of speech processing and verbal fluency in humans, 7 diminished alpha band oscillations in Bdnf Pax2 KO mice would be one of the most critical links to their cognitive deficits.In line with these observations, reduced evoked brain activity across alpha, beta, and low-gamma oscillations is a characteristic feature of the early-onset "critical period" disorders, such as autism, [3][4][5][6][7] as opposed to late-onset psychiatric disorders, such as schizophrenia, which are characterized primarily by reduced evoked gamma oscillations. 93ntil now, the source of enhanced spontaneous and reduced evoked broadband EEG power in autism, if not hereditary, has remained unknown.The time period in which autism symptoms start to manifest (three to six months of age) is the same period in which children exhibit a dramatic shift in their EEG power spectrum, resulting from changes in conduction speed, synapse number and strength, and cellular composition and maturity. 5uring this time, EEG waveforms are assumed to integrate neuronal activity driven by the individual sensory modalities. 9,90It has long been hypothesized that if sensory input was integrated improperly in higher brain regions regarding its modality, it would result in deficits in complex executive behavior, such as language in autism. 3hile improper sensory integration could occur because of sensory deprivation, autism spectrum disorder in children is typically associated with normal basic sensory function, 17,59,94 which would exclude sensory deprivation as a main cause of altered cortical brain oscillations.
These data provide a new understanding of autism through an immature synchronization of peripheral neuronal processes.This peripheral immaturity would, despite the development of basic hearing function, prevent the maturation of intracortical networks that allow the processing of higher frequencies critical for communication.
(iv) E/I imbalance in Bdnf Pax2 KO mice: Numerous studies in both humans 4,95 and animals 89 emphasize E/I imbalance in frontal brain regions as a characteristic feature of autism.E/I imbalance has been suggested to involve aberrant neural connectivity linked to a reduction of brain volume in the frontal cortex and HC linked with a loss of synapses rather than neurons themselves. 96,97ndeed, psychiatric disorders have been associated with reduced dendritic outgrowths of PV-INs and with more immature dendritic spines on pyramidal cells in postmortem studies 96,97 and in mouse models. 98In line with this, Bdnf Pax2 KO mice exhibit normal numbers of PV-INs and normal BDNF mRNA/protein levels in the AC and HC, but immature (low) levels of PV in dendrites that persist also into adulthood (Figure S4). 20,21As shown in the present study, Bdnf Pax2 KO mice exhibit reduced PV-IN staining and enhanced vGlut1 levels in hippocampal stratum radiatum (Figure 7), confirming E/I imbalance to occur on synaptic levels.
The normal PV-IN number in the cortex of Bdnf Pax2 KO mice suggests that the source of the structural and functional E/I imbalance in the cortex and HC of Bdnf Pax2 KO mice is neither a general impairment in the migration of GABAergic INs from the subpallium to telencephalon 14,27 nor a lack of cortical BDNF in pyramidal neurons, required for the experience-driven synaptogenesis of fastspiking PV-INs and pyramidal neurons. 22,23,99he Pax2+ GABAergic IN precursor cells, suggested to be impaired in Bdnf Pax2 KO mice, derive from the third ventricle and migrate to hindbrain regions and the auditory periphery, including ANFs. 27,29It was previously shown that BDNF deletion in Pax2-lineage descendants resulted in reduced sustained tonic inhibitory strength of the high-frequency sidebands of sharply tuned projecting neurons in the DCN and IC, particularly in response to stimuli with frequencies higher than 10 kHz. 20,21,33The fact that Bdnf Pax2 KO mice exhibit deficits in higher frequency responses in (i) CAP amplitudes and PSTR peak-to-plateau ratios (Figures 5 and 6), (ii) high-frequency sideband inhibition in the DCN and IC, 20,21 and (iii) cortical LFP responses (Figure 1; Figure 4) underscores a relationship between these processes.
We hypothesize that the mechanism behind the influence of BDNF deletion in Pax2-lineage descendants on high-SR fiber processing is an impaired maturation of inhibitory lateral olivocochlear (LOC) efferents.LOC efferents modulate high-SR ANF responses through axo-dendritic tonic inhibition.When this inhibition is impaired, spontaneous firing rates of ANFs are elevated, while response amplitudes of particularly high-SR ANFs are reduced. 100Hypothesizing that in Bdnf Pax2 KO mice, the shaping of auditory nerve responses remains immature, we would expect a reduced activation of dendrites contacting DCN neurons, leading to longer membrane time constraints that would consequently widen the temporal window for specific high-frequency sensoryevoked information. 101This accounts for the observed elevated thresholds, enlarged bandwidth, and reduced high-frequency sideband inhibition accompanied by elevated firing rates observed in the DCN 20 and IC of Bdnf Pax2 KO mice. 21Dysfunctional inhibitory shaping of ANFs following impaired function of Pax2+ precursor cellsas postulated here-may also provide the connection to reduced GABA function in autism: During the critical period-the suspected onset of autism 4,5,18 -GABA receptor function switches from excitatory to inhibitory in all sensory systems. 102,1035][106][107] High-SR ANFs were previously suggested to drive this switch through activity-driven BDNF upregulation. 45,108In support of this proposed mechanism, auditory nerve transection that includes high-SR ANFs leads to a decline of KCC2 and a subsequent reemergence of depolarizing GABAergic signaling. 109urthermore, a deafferentation of ANFs after acoustic trauma that, from its percentage, includes high-SR ANFs triggers a reduction of activity-driven BDNF expression levels in the auditory brainstem. 34An imbalance of excitatory/inhibitory effects of GABA would also explain elevated epileptiform activity observed in children with autism spectrum disorders and reduced GABAergic activity in the brain in humans 4,95 and mouse models. 15n conclusion, diminished high-SR ANF processing in Bdnf Pax2 KO mice would primarily influence synchronized spike responses, 110 evoked population synchrony, 38,50 and diminished tonic inhibition in ascending circuits, thereby diminishing the network-centered homeostasis of the overall cortical circuit and ultimately resulting in an autistic phenotype.This is expected to affect the corticofugal feedback loops which, as postulated by others, 64,111,112 control perception and attention.Based on the current findings, these should be reconsidered in the context of the influence of LOC activation o high-SR ANFs as also postulated previously. 111 the light of this new perspective, various observations made in the context of autism require further examination: (i) Disturbed feed-forward and intracortical synchronization, as observed in Bdnf Pax2 KO mice, 45 may provide a rationale for predictive coding deficits observed in autism spectrum disorder. 113,114(ii) Deficits in the visual orientation tuning of the autistic Fmr1 KO mice model linked to decreased PV-IN activity in V1 19 can be considered within the context of a thus far unknown source in the retina that may provide improved driving force for central inhibition in the visual cortex.(iii) In the context of the present findings, deficits in high-SR ANF function can be reconsidered as a source of immediate unmasking of cortical neurons after peripheral damage without causing a change in topographic map plasticity. 87Through such an event, the increased risk for cognitive deficits observed with hearing loss 116,117 or listening difficulties in children with clinically normal hearing 17,59,[118][119][120][121] may be explained.(iv) Any processes of either hereditary or acquired origin that might disrupt proper migration of GABA-IN precursor neurons to brainstem regions, including intraventricular or periventricular hemorrhage, or hypoxic ischemic encephalopathies, linked with neonatal bleeding during the critical developmental period of children, 122 should be newly reconsidered as a possible cause of autism.

| Limitations of the study
Though it is important to acknowledge that neurooscillatory frequency bands in rodents may not directly correspond to those in the primate brain, 85 there is meanwhile evidence that brain oscillations, including slower ones, can indicate compatible neural processes in different species, including rodents. 10 A clear connection between the role of high-SR processing, KCC2, and BDNF has not been brought and likely cannot be provided due to the enormous complexity of the integrated networks, which exhibit modality-specificity and high variability in temporal and spatial dynamics due to the cascading development of sensory modalities during critical developmental periods. 5,102,103This includes the predicted need of proper tonic GABAergic activity for spine maturation that cannot be proven beyond doubt due to temporal and spatial heterogeneity of circuits involved. 123The relationship between lower inhibitory GABAergic processing in Bdnf Pax2 KO mice and an

F
I G U R E 1 Reduced LFP amplitudes in response to auditory stimuli of best frequencies in Bdnf Pax2 KO mice.(A) Traces of event-related potentials averaged from evoked responses of 8 control and 9 Bdnf Pax2 KO mice for three different stimulus frequencies: 5 kHz (left panels), 11 kHz (middle panels), and 22 kHz (right panels) for the auditory cortex (1st row), prefrontal cortex (2nd row), hippocampus (3rd row), and visual cortex (bottom row).Controls: black curves; Bdnf Pax2 KO: red curves.(B) Group data of peak-to-peak amplitude of single ERPs for controls (black) and Bdnf Pax2 KOs (red) based on the first 50 ms of the ERP (LFP amplitude) in response to stimuli presented in groups of frequencies ranging from ca. 4-6, 10-12, and 21-23 kHz (abscissa).Gray arrows in (A) indicate the LFP.Data in (B) present mean ± SEM. *p < .05,ns p ≥ .05.Detailed statistical information can be found in Table

F
I G U R E 2 Averaged evoked and induced ASSR response amplitude in cortical regions of control and Bdnf Pax2 KO mice.Distribution of evoked and induced response amplitude measured at different modulation frequencies (f m ) in the (A) auditory cortex and (D) prefrontal cortex (controls: black circles; Bdnf Pax2 KO: red triangles).Distribution of (B,E) evoked and (C,F) induced response amplitudes in response to increasing SPL of the stimuli for the (B,C) auditory cortex and (E,F) prefrontal cortex.Left panels in (B,C,E,F) show 10 Hz f m , middle panels 40 Hz f m , and right panels 128 Hz f m .Mean ± SEM of 8 control and 9 Bdnf Pax2 KO mice.*p < .05,**p < .01,***p < .001.Detailed statistical information can be found in Table

F I G U R E 3
Elevated spontaneous LFP power in cortical regions of Bdnf Pax2 KO mice.Power spectra for LFP in frequency oscillation bands alpha (5-15 Hz), beta (15-25 Hz), low gamma (25-35 Hz), mid-gamma (35-65 Hz), and high gamma (65-125 Hz), separated by vertical lines.The spontaneous LFP power was significantly higher in cortical regions of Bdnf Pax2 KO mice (red lines) as compared to controls (black lines) (A, auditory cortex, B, prefrontal cortex, and D, visual cortex).(C) In the hippocampus, the higher spontaneous LFP power does not reach statistical significance in the analyzed frequency bands.The 50 Hz powerline interference has been removed (gap).Spectral power decreases below 10 Hz due to bandpass filtering (>10 Hz-5 kHz with 60 dB attenuation per decade).Controls: black curves; Bdnf Pax2 KO: red curves.Mean ± 95% confidence range of 8 controls and mean of 9 Bdnf Pax2 KO mice.*p < .05,**p < .01,ns p ≥ .05.Detailed statistical information can be found in Table

F I G U R E 4
Reduced evoked and elevated induced LFP power at alpha, beta, and low-gamma bands in the AC of Bdnf Pax2 KO mice.Evoked (left panels) and induced (right panels) oscillation power (mean LFP in μV 2 ) in response to low-(4-6 kHz), middle-(10-12 kHz), and high-frequency stimuli (21-23 kHz).The effect of stimulus frequencies on (A) alpha oscillations (5-15 Hz), (B) beta oscillations (15-25 Hz), (C) low-gamma oscillations (25-35 Hz), (D) mid-gamma oscillations (35-65 Hz), and (E) high-gamma oscillations (65-125 Hz) were compared in the respective frequency group between control and Bdnf Pax2 KO mice.Evoked oscillations in Bdnf Pax2 KO mice (red triangles) were smaller in the lower oscillatory bands (A-D, left), black empty circles), most consistently at stimulus frequencies around 11 kHz.Induced oscillations (right) were higher in Bdnf Pax2 KO mice in comparison with controls, though the difference reached statistical significance only for the beta oscillatory band (B, right).Controls: black empty circles; Bdnf Pax2 KO: red triangles.Mean ± SEM of 8 control and 9 Bdnf Pax2 KO mice.*p < .05,**p < .01,ns p ≥ .05,and (*) p < .05uncorrected for multiple testing.Detailed statistical information can be found in Table

F I G U R E 5
Reduced compound action potentials (CAP) in Bdnf Pax2 KO mice.(A) Schematic drawing of electrode placement for electrocochleography at the round window.(B) Click-and pure-tone frequency stimulus-evoked CAP thresholds.(C) Click-evoked threshold-normalized CAP input/output growth-function amplitude (left panel) and latency (right panel).(D) Threshold-normalized CAP input/output growth-function amplitude elicited with pure-tone stimuli at 2 kHz, (E) 5.7 kHz, (F) 8 kHz, (G) 11.3 kHz, (H) 16 kHz, (I) 22.6 kHz, and (J) 32 kHz.(K) Comparison of CAP signals for all frequencies at a similar intensity of 20 dB above threshold, (L) 30 dB above threshold, and (M) 40 dB above threshold.(B) Left panel = individual ears (single dots).(B-M, right panel) = mean ± SEM.Red highlights indicate significance in post hoc comparison.ns p ≥ .05.Detailed statistical information be found in Table

F I G U R E 7
Protein analysis of parvalbumin (PV) and vesicular glutamate transporter 1 (vGlut1) and spine analysis in the CA1 region of the hippocampus.(A) Schematic drawing of a coronal section of the hippocampus; DG = dentate gyrus; SP = stratum pyramidale; SR = stratum radiatum.The red box indicates the area analyzed.(B) Quantification of PV and vGlut1 fluorescence in the dendrites of CA1 pyramidal neurons.Bdnf Pax2 KO mice had lower PV fluorescence intensity and higher vGlut1 fluorescence intensity than controls.(C) Representative images of controls (top) and Bdnf Pax2 KO mice (bottom).(D) Overview picture of the hippocampus after Golgi-Cox staining.(E) Spine analysis of pyramidal neurons in CA1.Spine density is indicated by the number of protrusions per micrometer.Lengthto-width ratio (LWR) reflects differences in spine maturation, with higher LWR indicating less mature spines.Bdnf Pax2 KO mice had significantly higher spine density and LWR than controls, indicating longer, thinner spines (immature spines).(G) Higher magnification of Golgi-Cox-stained CA1 pyramidal neurons of controls (upper panels) and Bdnf Pax2 KO mice (lower panels).(H) Proportion of different spine types.Control mice had a higher percentage of mature mushroom-type spines and a lower percentage of immature filopodia-type spines than Bdnf Pax2 KO mice.For C: PV, red; vGlut1, green.DAPI nuclear stain, blue.For B, E, F: *p < .05,**p < .01,****p < .001.Detailed statistical information can be found in Table