Abstract
During developmental critical periods (CPs), early-life stress (ELS) induces cognitive deficits and alters neural circuitry in regions underlying learning, memory, and attention. Mechanisms underlying critical period plasticity are shared by sensory cortices and these higher neural regions, suggesting that sensory processing may also be vulnerable to ELS. In particular, the perception and auditory cortical (ACx) encoding of temporally-varying sounds both mature gradually, even into adolescence, providing an extended postnatal window of susceptibility. To examine the effects of ELS on temporal processing, we developed a model of ELS in the Mongolian gerbil, a well-established model for auditory processing. In both male and female animals, ELS induction impaired the behavioral detection of short gaps in sound, which are critical for speech perception. This was accompanied by reduced neural responses to gaps in auditory cortex, the auditory periphery, and auditory brainstem. ELS thus degrades the fidelity of sensory representations available to higher regions, and could contribute to well-known ELS-induced problems with cognition.
SIGNIFICANCE STATEMENT In children and animal models, early-life stress (ELS) leads to deficits in cognition, including problems with learning, memory, and attention. Such problems could arise in part from a low-fidelity representation of sensory information available to higher-level neural regions. Here, we demonstrate that ELS degrades sensory responses to rapid variations in sound at multiple levels of the auditory pathway, and concurrently impairs perception of these rapidly-varying sounds. As these sound variations are intrinsic to speech, ELS may thus pose a challenge to communication and cognition through impaired sensory encoding.
- auditory brainstem response
- auditory cortex
- development
- early-life stress
- gap detection
- temporal processing
Introduction
Sensory experience early in development is essential for the normal maturation of cortical circuits, allowing for accurate perception of the sensory world. Children who experience hearing loss (HL) are at risk for later problems with speech perception (Schönweiler et al., 1998; Psarommatis et al., 2001; Whitton and Polley, 2011). This is likely because of auditory deprivation during critical periods (CPs) for neural plasticity, where the brain undergoes increased susceptibility to alterations by sensory experience (Takesian and Hensch, 2013). Altered experience during these time windows (e g., by hearing loss or exposure to altered environmental soundscapes) causes perceptual impairments and changes the response properties of auditory cortical (ACx) neurons (Zhang et al., 2002; Han et al., 2007; Rosen et al., 2012; Gay et al., 2014; Park et al., 2015; Ihlefeld et al., 2016; Green et al., 2017; Yao and Sanes, 2018). In particular, the circuitry that encodes temporally-varying signals develops gradually postnatally, and is thus susceptible to auditory deprivation during the CP for ACx development (Sanes and Woolley, 2011; Mowery et al., 2015). This is important because the ability to detect rapid changes in sound (temporal processing) is intrinsic to deciphering our soundscape, including analyzing auditory scenes and understanding speech (Tallal, 2004).
Early-life stress (ELS) is well known to affect CP mechanisms identified in studies of sensory cortices, but ELS effects are described almost exclusively in nonsensory regions. ELS elicits long-term effects on anxiety, memory, learning, and cognition (Cameron et al., 2017). These behavioral changes have been intensely investigated in neural regions underlying those higher-level processes, including frontal cortex, hippocampus, and amygdala. In those regions, ELS affects the maturation of inhibitory neurons, perineuronal nets, brain-derived neurotrophic factor (BDNF), and dendritic morphology (Bath et al., 2013; Castillo-Gómez et al., 2017; Murthy et al., 2019; Page and Coutellier, 2019). Because these elements underlie sensory cortical plasticity during CPs (Hensch, 2005), ELS may also alter cortical sensory regions. Indeed, both motor and somatosensory cortices show reduced parvalbumin cell density following ELS (Manzano Nieves et al., 2020). Yet the effects of ELS on ACx neuronal properties, subcortical auditory regions, or behavioral sensitivity to sounds have not been examined.
The literature indicates that ELS may affect the auditory system. In rats, ELS reduced the magnitude of auditory evoked potentials (Ellenbroek et al., 2004; Yates et al., 2016). Children experiencing maltreatment, low socioeconomic status (SES), or low maternal education (proxies for developmental stress) have altered ACx gray matter volume and diminished auditory fiber tract integrity (Teicher et al., 2016), lower amplitude auditory brainstem responses (ABRs; Skoe et al., 2013), and poorer performance on speech perception tasks (Nittrouer and Burton, 2005). Although the underlying mechanisms may differ, stress induced during adulthood alters auditory perceptual sensitivity (Ashton et al., 2000; Hasson et al., 2013; Perez et al., 2013), neural morphology, and response properties in ACx and inferior colliculus (Bose et al., 2010; Dagnino-Subiabre et al., 2012; Ma et al., 2015). Thus, ELS may have widespread effects on the auditory pathway and even the auditory periphery; for example, a corticotropin-releasing factor signaling system is active in the cochlea (Vetter and Yee, 2018).
We examined whether ELS affects gap detection thresholds (GDTs) for short gaps in background sound. Gap detection represents a general measure of temporal acuity and requires an intact ACx (Green, 1971; Ison et al., 1991; Threlkeld et al., 2008). Because ACx circuitry that encodes gap detection develops gradually postnatally, we hypothesized that it may be affected by ELS, similarly to how this circuitry is affected by early auditory deprivation or hearing loss (Rosen et al., 2012; Mowery et al., 2015; Green et al., 2017). We induced ELS in Mongolian gerbils within a time window encompassing the CP for ACx maturation, using a combination of maternal separation and restraint. ELS gerbils showed poorer behavioral gap detection, and responses to short gaps in sound were reduced in ACx neurons and in the auditory nerve and brainstem. Because gap detection in children is predictive of later speech processing abilities (Benasich et al., 2006; Muluk et al., 2011), ELS-induced deficits in gap detection implicate ELS as a risk factor for persistent problems with speech processing.
Materials and Methods
Subjects and experimental design
All procedures relating to the maintenance and use of animals were approved by the Institutional Animal Care and Use Committee at Northeast Ohio Medical University. Mongolian gerbils (Meriones unguiculatus) are a well-established model for auditory processing. Gerbils from multiple litters were housed with littermates in a 12/12 h light/dark cycle. Control and ELS animals were from separate litters, to avoid any stress on Control animals or parents induced by removing siblings for ELS treatment. The numbers and the sex of animals used for each measurement are indicated in Results.
The experimental design is shown in Figure 1A. ELS animals experienced stress induction from postnatal day (P)9 to P24 as described below, while Control animals were left undisturbed in their home cages. Bodyweight was measured from a subset of animals on days of stress induction. Blood samples were collected from a subset of animals at P25 (the day following the last stress induction event) or P33 (immediately following one session of gap-PPI) to measure corticosterone levels. Auditory brainstem responses were collected from a subset of animals within an age range of P33–P40. Animals used for bodyweight, corticosterone, or ABRs did not contribute to any other measurements.
On P33, P36, and P39 (following 1 h of acclimation to the startle testing enclosure on P32), a subset of animals was behaviorally tested in 1-h acoustic startle gap-PPI sessions (the behavioral setup allows testing of eight animals simultaneously). Over the next ∼14 d (P40–P54), these animals underwent cortical recordings (one animal per day). Animals for cortical neurophysiology were chosen from each group without regard to behavioral performance. Each animal yielded a variable number of neurons for analysis, so recordings were conducted until a sufficient number was collected across groups (see Results).
Early-life stress induction
ELS was induced postnatally by a combination of maternal separation and restraint from P9 to P24. Ten 2-h restraint sessions occurred either in the morning or late afternoon within this 16-d time window, to reduce predictability of the restraint session; unpredictability is widely known to be a strong contributor to stress-related behavioral changes (Willner, 1997; Mineur et al., 2006; Bath et al., 2017). During these sessions, the home cage was transported from the animal facility to the testing room, and pups were separated from their parents and littermates and individually isolated for 2 h. P9–P24 animals were placed into 50-ml centrifuge tubes. The tubes were perforated to allow for easy breathing, and were partly filled with a cork to prevent the animal from free movement. The tubes were placed either horizontally or vertically each in their own small, sound attenuated, anechoic booth, with the lights on. Animals that were too large for the centrifuge tubes (≥P20) were placed in a small restrainer, and then placed in the booth. The booths and small restrainers were those used for later behavioral acoustic startle testing to measure gap detection abilities. After each session, pups were returned to their home cages and then to the animal facility. ELS was induced in all pups of a litter. While this precludes using siblings as controls, it importantly prevents stress being unintentionally induced in control animals or parents when the home cage and pups are regularly disturbed and removed.
Corticosterone measurement
To measure corticosterone levels, blood was collected from the retroorbital vein in animals briefly anesthetized with isoflurane. Blood was centrifuged at 3000 rpm for 30 min, and plasma was separated and saved at −80°C. Plasma corticosterone levels were assessed by ELISA kits (Enzo Life Sciences). Plate readings were obtained with a SynergyTM 4 spectrophotometer (BioTek Instruments) at a wavelength of 405 nm. Raw optical density was read out between 570 and 590 nm and corrected with a blank control. The blood corticosteroid level was calculated by fitting the raw data with a four-parameter logistic curve.
Behavioral testing for gap detection
Gap detection abilities were assessed as inhibition of the reflexive acoustic startle response, where a detectable change that precedes a startling sound inhibits the startle [known as prepulse inhibition (PPI)]. The strength of inhibition corresponds with an animal's detection of the change. Here, the change was a silent gap in background noise (a variation on prepulse inhibition, gap-PPI). The procedure has been described previously (Green et al., 2016). Briefly, animals were placed inside a small acoustically transparent restrainer, on a force plate in a sound attenuated, anechoic booth, with the lights on. Two separate speakers in each booth presented either background noise at 50 dB SPL or a startling stimulus at 110 dB SPL (Kinder Scientific Inc.). The background noise was bandpassed from 2.5 to 20 kHz. We presented 190 trials in pseudorandom order. Of these, 46 trials were startle–only, with a startle stimulus of 20-ms broadband noise at 110 dB SPL, 1-ms rise/fall time. The remaining 144 were gap trials, where the startle stimulus was preceded (by 50 ms) by a silent gap in the noise background of either 2, 3, 5, 7, 10, 25, 50, or 125 ms, with 18 trials of each gap duration. The background noise preceding and following the gap was shaped with a 1-ms rise/fall time. At the beginning of each session, five startle-only trials were presented (not included in analysis) to habituate the responses to a steady-state level (Ison et al., 1973). Sessions lasted 1 h.
Behavioral data analysis
A gap detection threshold (GDT) was calculated for each animal at each session using custom MATLAB scripts (MATLAB, The MathWorks; D. Green and M. Rosen), as described previously (Longenecker et al., 2016; Green et al., 2017). First, the peak response magnitude to the startle stimulus was measured in the time window 20–50 ms after startle stimulus onset. Because the distribution of peak responses had a strong positive skew, a log10 transform was applied to generate a normal distribution of responses within each trial type. Then we determined the peak response threshold at which a reduction in startle was considered statistically significant. To do so, the median values of the peak log-transformed responses for startle only and each gap duration were plotted, and a cubic spline was fitted to this plot, creating a detection function. To find where that function crossed a detection criterion, the transformed startle-only values were bootstrapped (sampled with replacement 10,000 times) to generate a normal distribution, from which 95% confidence intervals were calculated. The lower confidence interval was the value where a reduction in startle indicated significant detection (Fechter et al., 1988). GDT was the shortest gap duration at which the fitted detection function crossed the lower confidence interval. Note that this approach calculates significant gap detection for each animal in comparison with their baseline startle-only response, which controls for any across-animal differences in their startle-only magnitude. This is important because startle-only magnitude can vary across animals. Group differences across sessions were assessed with mixed ANOVAs.
Gap-ABRs
Animals were anesthetized with ketamine (75 mg/kg) and xylazine (5 mg/kg) and presented with auditory stimuli [RZ6 Auditory Processor, BioSigRP software, Tucker Davis Technologies (TDT)]. For a subset of animals, behavioral GDTs were measured before ABR GDTs; these animals were anesthetized with dexdomitor during ABR recordings (0.2 mg/kg). Responses to individual stimuli were conducted using stainless steel needle electrodes inserted subdermally at the dorsal midline between the eyes (noninverting), posterior to the right pinna (inverting), and base of the tail (common ground), amplified (20×, TDT low-impedance RA4LI), bandpass filtered (0.3–3 kHz), and digitized (24.4 kHz, RZ5 BioAmp Processor, TDT). Auditory stimuli were presented at 40 dB SPL from a freefield speaker located 7 cm from the right ear. Stimuli were two bursts of 50-ms bandpass noise (2.5–20 kHz) with 1-ms rise/fall times for each burst. The bursts were separated by gaps of varying durations to match those used in behavioral gap detection testing: 2, 3, 5, 7, 10, 25, 50, or 125 ms. Burst pairs were repeated at intervals of 1 s. Responses were averaged over 200 presentations.
Each wave of the ABR was quantified separately. Gerbil wave i (equivalent to human wave I) is understood to arise from the distal portion of the auditory nerve (Achor and Starr, 1980; Boettcher, 2002). Gerbil waves ii and iii (equivalent to human wave III) arise primarily from the cochlear nucleus (Moller and Jannetta, 1982; Melcher et al., 1996). Gerbil wave iv (equivalent to human wave V) reflects activity from the cochlear nucleus, superior olivary complex, and inputs to the inferior colliculus (i.e., lateral lemniscus; Hashimoto et al., 1981; Moller and Jannetta, 1982, 1983). Amplitudes and latencies were measured for each wave, in response to both noise bursts. Peak amplitudes were measured as the voltage difference between each wave peak and the following trough, allowing calculation of the amplitude ratio: attenuation of the 2nd burst divided by the response to the 1st burst. Peak latencies were measured as the latency to each peak from the start of each noise burst. This allowed calculation of any latency shift: the difference in the response latency to the 2nd burst versus the response latency to the 1st burst. Detection threshold for each animal's gap-ABR was determined as the shortest gap duration with a visible response to the 2nd noise burst in any wave, although typically waves ii or iii were the last waves to disappear as gaps shortened. Group differences across gap durations were assessed with mixed ANOVAs.
Surgical preparation
Gerbils were prepared for auditory cortical recordings as previously described (Mattingly et al., 2018). Briefly, animals were anesthetized with isoflurane and held in a stereotaxic apparatus. A small headpost was positioned along the midline and secured with dental acrylic, and a silver ground wire was implanted into the posterior contralateral skull. A craniotomy was made over the left temporal cortex caudal to the bregma suture and the dura was left intact. A thin well of dental acrylic was built along the perimeter of the craniotomy, the cortical surface was covered with silicone oil, and the craniotomized area was covered with a disposable cap of silicone elastomer (ImageLB-28, Matrics Inc.). The entire skull was covered with dental acrylic to form a headcap.
Neurophysiological recordings
On the day of recording, animals were anesthetized with urethane (1.3 g/kg, administered in two doses over 1.5 h) and placed in a soundproof chamber (Industrial Acoustics Company) on a heating pad. The head was stabilized using the headpost, the silicone elastomer cap removed, and the dura was covered with saline during recording to maintain moisture. The dura was nicked and platinum-plated tungsten electrodes (1.5–2.5 MΩ; MicroProbe) were advanced to isolate neurons in primary ACx based on response characteristics (reliable, short-latency, nonadapting responses to tones). Electrical signals from the brain were amplified (250×, TDT RA16PA Medusa preamplifier), filtered (0.25–10 kHz), and digitized (24.4 kHz, TDT RZ5 BioAmp Processor). The TDT equipment was controlled by custom software written in MATLAB and TDT RPvdsEx programming environments (TytoLogy by S.J. Shanbhag). Units were isolated by spike amplitude. Spikes were detected offline (Plexon Offline Sorter), and sorted based on spike shape and principal component analysis.
Acoustic stimulation for neurophysiology
TDT equipment (RZ6 Auditory Processor) was used to deliver auditory stimuli and record neural responses using custom software written in MATLAB and TDT RPvdsEx (TytoLogy by S. J. Shanbhag; modified by M. J. Rosen). Auditory stimuli over a frequency range of 200 Hz to 35 kHz were calibrated (custom MATLAB scripts, S. J. Shanbhag). Calibration data were collected using a ¼ inch microphone [Brüel and Kjær (B&K) model 4939], a preamplifier (B&K 2670) and a conditioning amplifier (B&K Nexus model 2690). Stimuli were delivered through a freefield speaker positioned 25 cm in front of the animal.
Each ACx unit's response to tones was assessed by presenting 200-ms tone pips (1-s intertrial intervals, 5-ms cosine-ramped rise/fall). First, the frequency range over which the neuron was responsive was obtained with an iso-intensity function at 60 dB SPL to determine the best frequency (BF) of the unit (the frequency eliciting the highest firing rate). This was followed by a rate-level function (RLF) at BF, measured at increments of 10 dB SPL, for 15 trials. Threshold was visually determined as 5 dB below the lowest level with a clear increase in firing above lower levels. A gap detection function was obtained at the BF of the unit and 30 dB above threshold, with 20 trials at each gap duration, presented in random order. Gap detection stimuli consisted of two consecutive tone bursts lasting a total of 400 ms (with 5 ms cosine-ramped rise/fall). The bursts were separated by gaps of varying durations matching those presented behaviorally (0, 1, 2, 3, 5, 7, 10, 15, 25, 50, or 125 ms with a 0.5-ms cosine-ramped rise/fall) inserted between the two bursts at 200 ms after the onset of the first burst. The 0-ms gap stimulus was the control against which gap detection was measured, and thus did not contain rise/fall ramps at 200 ms, but was instead a continuous 400-ms tone burst. This control was chosen to mimic the behavioral stimuli, where detection of gaps in background noise were calculated in comparison with continuity in the background noise. After collecting the gap detection tone function, the unit was presented with 200-ms bursts of noise at 50 dB SPL, bandpassed from 2.5 to 20 kHz, to match the background noise used for the behavioral testing. If the unit showed a clear response to the noise, a noise gap detection function was obtained with 20 trials at each gap duration, presented in random order.
Neural data analysis
Rate-level functions
Firing rates to tones were calculated over a time window equal to the stimulus duration (200 ms). Threshold, dynamic range, and monotonicity were determined from the RLF. Threshold was defined as the dB SPL level at which there was at least a 35% increase in firing rate, stepping up from one level to the next; threshold firing rates were calculated at this sound level. Dynamic range was defined as the range between the sound levels where each cell responded at 10% and 90% of its maximum firing rate, calculated by interpolation. A monotonicity index (MI) was calculated by dividing the firing rate at the maximum presented sound intensity (80 dB SPL) by the maximum firing rate within the RLF (Moore and Wehr, 2013). The MI ranges from 1 (no intensity tuning) to 0 (strong intensity tuning). All data were analyzed with custom MATLAB scripts (M. J. Rosen). Group differences were assessed with Kruskal–Wallis (KW) nonparametric ANOVAs.
Gap detection
Responses to each gap duration were measured from poststimulus time histograms (PSTHs) with 5-ms bins calculated across the 20 trials. Only units with onset responses to the first burst of the gap detection stimuli were used for gap detection analysis. A valid onset response (within 100 ms following the first burst) was determined as spiking 20% of the time within a single bin across trials (Eggermont, 2000). A valid gap response [within a region of interest (ROI) of 100 ms following gap offset (equivalent to second burst onset)] was determined based on the bin with maximal firing in this window. We evaluated the response in a time window following gap offset [rather than gap onset (equivalent to offset of the first burst)] because (1) most units did not have clear offset responses, and (2) cortical responses at gap offset have been directly linked to behavioral gap detection (Weible et al., 2014). To determine a significant response within the ROI, the response to each gap was compared with the no-gap stimulus response using a bootstrapping procedure. In response to each stimulus, 20 response trials were drawn with replacement and used to create a PSTH with 5-ms bins. The peak firing rate (within the largest bin) was noted. We bootstrapped 500 times to repeat this process, to calculate the peak firing rate mean and 95% CI for each stimulus. Any gap duration where the mean firing in the peak bin within the ROI exceeded the 95% CI from the no-gap condition was considered a significant response. The shortest gap with a significant response was considered threshold. If a significant response to the second burst was absent for all gap durations, a GDT of 135 ms was assigned [since only cells with valid onset responses were analyzed, a gap longer than the longest presented (125 ms) would necessarily elicit a response]. First spike latency (FSL) and FSL jitter were measured directly from spike timing rather than from binned PSTHs, and were based on the onset response to the first burst. All data were analyzed with custom MATLAB scripts (M. J. Rosen). Group differences were assessed with Kruskal–Wallis nonparametric ANOVAs.
Ideal observer analysis
An ideal observer classifier model was used to determine how well the population of neural units could discriminate between gap and no-gap trials. By definition, an ideal unbiased observer maximizes hits and minimizes false alarm rates to perform a given task optimally (Geisler, 2011). We compared the ideal observer's ability to detect gaps of varying durations. We trained a support vector machine with 85% of the neural data, and ran a 10-fold cross-validation test of classifier performance with the remaining 15% of the neural data, separately for Control and ELS groups. The classifier was trained and tested with trial-by-trial firing rates from 100-ms time windows following gap offset. The classifier, implemented in MATLAB 2015a, used custom scripts (M. J. Rosen, Y. E. Cohen, and A. Ihlefeld) that trained a binary support vector machine classifier (fitcsvm in MATLAB) and tested the performance on the remainder of the data by computing a loss estimate using cross validation and misclassification rate (crossval and kfoldLoss in MATLAB). Equal numbers of neurons were used across the treatment groups, limited by the group with the fewest units that had valid onset responses to the first burst (104 units for trials with tone carriers at BF; 94 units for trials with bandpass noise carriers). The cross-validated classifier model was run 100 times to generate mean performance with SEM error bars.
Data availability
The data are available from the corresponding author on reasonable request.
Results
Establishing a model of early-life stress in the Mongolian gerbil
While early-life stress is frequently studied in rodent models, to date there is not an established model for ELS in gerbils. The Mongolian gerbil is a well-established model of auditory processing, and is being actively investigated to understand the development of auditory perception and attention. Adding ELS to the developmental repertoire establishes a model that will help elucidate interactive effects of stress with other developmental issues, such as hearing loss. Here, we induced early-life stress through a combination of intermittent, unpredictable maternal separation and restraint over the developmental window P9–P24 (Fig. 1A,B). This time window encompasses the CP for the maturation of several auditory cortical properties, including evoked firing rates (Mowery et al., 2015). Furthermore, auditory deprivation during this time window is known to induce perceptual deficits (Caras and Sanes, 2015; Green et al., 2017). Thus, induction of stress over this time window has the potential to affect development of the auditory cortex. To assess the effectiveness of the stress induction, we measured three established physiological markers of early-life stress: blood corticosterone levels, bodyweight gain over development, and response to a startling stimulus (Buynitsky and Mostofsky, 2009; van Bodegom et al., 2017; Demaestri et al., 2020).
Experimental design and early-life stress induction method. A, ELS was induced by intermittent maternal separation and restraint as depicted, placing the restrained animals in isolated enclosures for 10 2-h sessions at unpredictable times over the age range P9–P24. B, The experimental design timeline depicts measures of auditory processing of gaps in ongoing sound (gap-PPI for behavioral sensitivity, gap-ABR for peripheral and brainstem sensitivity, and ACx recordings for cortical sensitivity), along with measures verifying successful stress induction (bodyweight, blood corticosterone, and startle-only amplitude collected during gap-PPI sessions).
Stress enhances the activity of the hypothalamus-pituitary-adrenal (HPA) axis, resulting in increased secretion of corticosteroids from the adrenal cortex of the adrenal gland (van Bodegom et al., 2017). In rodents including gerbils, plasma corticosterone (CORT) is the main glucocorticoid involved in regulating stress responses, and is thus often used as a biomarker for stress. Chronic or developmental stress can either increase or decrease CORT levels, depending on the details of stress induction (van Bodegom et al., 2017). To establish whether our ELS induction protocol affected baseline levels of this stress hormone, we measured corticosterone levels in blood samples collected from animals at P25, the day following the last stress induction event (Fig. 1A). To determine whether ELS induction affected CORT levels following an acute stressor, we collected blood samples at P33 in a separate group of animals, immediately following a session of acoustic startle gap-PPI measurements. CORT levels were significantly lower in ELS than Control animals, both at baseline and following this acute stressor [two-way ANOVAs, main effect of Treatment: Baseline F(1,35) = 28.625, p < 0.001, η2 = 0.450, n = 20 Control, n = 19 ELS; Acute F(1,34) = 4.317, p = 0.045, η2 = 0.113, n = 18 Control, n = 20 ELS; Fig. 2A). Because stress effects can differ across sex, we tested whether ELS affected CORT in a sex-specific manner. There were no main effects of sex for either baseline or acute measurements. In alignment with this, at baseline, pairwise comparisons indicated lower CORT levels in ELS versus controls for both males and females (Males (M): F(1,35) = 14.801, p < 0.001, η2 = 0.297, CTR M: 11.0 ± 1.8, ELS M: 1.8 ± 0.1, n = 9 CTR, n = 12 ELS; Females (F): F(1,35) = 13.824, p < 0.001, η2 = 0.283, CTR F: 12.9 ± 2.4, ELS F: 4.2 ± 1.0, n = 9 CTR, n = 8 ELS). For the acute stressor, CORT was lower in ELS animals only for males (Males: F(1,34) = 7.838, p = 0.008, η2 = 0.187, CTR M: 26.3 ± 3.5, ELS M: 16.7 ± 2.0; CTR F: 17.8 ± 2.5, ELS F: 15.4 ± 3.2). Additionally for the acute stressor, there was a sex difference in CORT levels for control but not ELS animals (F(1,34) = 4.681, p = 0.038, η2 = 0.121). These data demonstrate that the P9–P24 ELS induction protocol affects the HPA axis in Mongolian gerbils, with a greater sex effect in males than females following an acute stressor.
Successful induction of early-life stress in the gerbil, without a nutritional deficit. A, Blood serum corticosterone levels were reduced by ELS in two behavioral situations: baseline (P25, 1 d after the completion of ELS induction) and following an acute stressor (P33, immediately after first experiencing a gap-PPI acoustic startle test session). B, Bodyweight gain over postnatal development was not affected by ELS induction. C, Startle reactivity was reduced in ELS animals. ***p ≤ 0.001, **p ≤ 0.01, *p < 0.05; error bars show SEM; pale lines or small circles show individual animals.
When rodents are raised with ELS, their body weights may be reduced compared with Control animals, although this varies depending on how ELS is induced (Demaestri et al., 2020). We weighed ELS gerbils after each restraint session (at postnatal days 12, 13, 16, 18, 19, 21, 22, and 24), and weighed Control animals at the same ages. Body weight gain did not differ across this developmental period [repeated measures ANOVA: F(1,24) = 1.159, p = 0.29, n = 10 Control (three males), n = 16 ELS (9 males); Fig. 2B]. This reduces the likelihood of differential nutritional consequences, or of any potential confound arising from comparing animals of different body weights in a startle paradigm.
Startle reactivity measured by the acoustic startle response, which involves a simple subcortical circuit, can be modified by hormones of the HPA axis (Lee et al., 1994; Lee and Davis, 1997; Grillon et al., 2006), and is often used as an assessment tool for stress and anxiety (Kevin and Jennifer, 2013; Poli and Angrilli, 2015). To determine whether our ELS induction protocol altered startle reactivity, we measured the amplitude of startle-only responses (normalized by each animal's mass) during behavioral testing for gap detection using gap-PPI. The normalized amplitudes of startle responses in trials that did not contain gaps were compared between Control and ELS groups, for each of three sessions of gap-PPI testing. A mixed ANOVA with a between-subject factor of treatment and a within-subject factor of session revealed a significant effect of treatment, with lower startle-only amplitudes in ELS animals (F(47,1) = 7.734, p = 0.008, η2 = 0.141, n = 26 Controls, n = 23 ELS; Fig. 2C). This altered startle reactivity is consistent with an effective induction of stress in our model.
Behavioral gap detection is impaired by ELS
Having established that our developmental manipulation effectively induces early-life stress, we measured behavioral auditory sensitivity to temporally-varying sounds that require auditory cortex: brief gaps in an ongoing noise (Ison et al., 1991; Kelly et al., 1996; Syka et al., 2002; Threlkeld et al., 2008). We chose to use a reflexive, preattentive measure of gap detection: gap-PPI of the acoustic startle response. PPI, a measure of detection rather than perception, is not influenced by attentional factors provided the interval between stimulus and startle is ≤60 ms (Li et al., 2009), but see (Cope et al., 2022). This allows measurement of auditory processing without the potentially confounding contribution of cognition or attention, both of which can be affected by stress. Furthermore, significant gap detection for each animal was calculated in comparison with their baseline startle-only response, controlling for any across-animal differences in startle magnitude.
The day before testing, animals were acclimated to the testing enclosure for 1 h. Testing sessions were conducted at the same time each day, on P33, P36, and P39, and GDTs were calculated for each session (animals tested for GDTs were not used for body weight or CORT measures; Fig. 1A). A mixed ANOVA with a between-subject factor of treatment and a within-subjects factor of test session revealed significantly poorer (higher) GDTs for ELS animals, collapsed across sessions (F(1,47) = 7.105, p = 0.011, partial η2 = 0.131; n = 26 Control, n = 23 ELS; Fig. 3A). There was also greater variability across the ELS animals (Levene's test F(1,35) = 7.59, p < 0.0001). Post hoc tests with Sidak corrections for multiple comparisons showed that ELS GDTs were worse than Controls in sessions 2 and 3 (p = 0.005 and p = 0.014), although not in session 1 (p = 0.56). Even when examining the best performance for each animal, ELS animals had poorer and more variable thresholds than Controls (Kruskal–Wallis: χ2(1,47) = 4.34, p = 0.037; Levene's test: F(1,47) = 9.94, p = 0.002; Fig. 3B). Beyond examining just threshold, the response ratio comparing gap to no-gap startle amplitudes depicts sensitivity across gap durations, with smaller amplitude indicating better detection. The Controls had smaller amplitude ratios across gap durations than ELS (main effect of treatment: F(1,47) = 232.7, p < 0.001, partial η2 = 0.83; Fig. 3B), with group differences at longer gap durations for session 1 (50 and 100 ms, p = 0.026 and p = 0.046), but shorter gap durations for sessions 2 and 3 (7, 10, and 25 ms: session 2, p = 0.026, p = 0.009, p = 0.002; session 3, p = 0.029, p < 0.001, p = 0.023).
Behavioral gap detection was impaired by early-life stress. A, Across three sessions of gap-PPI, ELS animals had poorer gap detection thresholds, concurrent with greater variability across ELS than Control animals. B, The best threshold across all sessions was worse for ELS animals. C, Gap to no-gap response ratios were higher for ELS animals, indicating poorer detection overall. Controls showed greater improvement based on response ratio across sessions than ELS. D, For gap detection thresholds, there was a trend for Controls to improve more across gap-PPI sessions than ELS animals. E, Males and females performed equivalently to one another within each treatment group. ***p ≤ 0.001, **p ≤ 0.01, *p < 0.05, †p < 0.1; ◊: unequal variance, p < 0.005; in boxplots, box edges are 25th and 75th percentiles, with whiskers extending to the most extreme data points excluding outliers; error bars show SEM; pale lines or small circles show individual animals.
Because ELS impaired gap detection, we evaluated whether gap detection performance was correlated with behavioral measures of stress that differed by treatment: corticosterone levels and startle reactivity (Fig. 2A,C). There were no significant correlations between CORT levels (Fig. 2A) and behavioral gap detection thresholds within any group (the animals used for CORT measurements were only tested for one gap-PPI session; data not shown). This was the case when grouped by treatment (ELS and CTR), or by sex and treatment (CTR females, ELS females, CTR males, and ELS males). Similarly, we correlated best behavioral gap detection thresholds (Fig. 3B) with startle-only magnitude (from Fig. 2). There were no significant correlations.
Testing across multiple days allowed us to assess improvement of gap detection, which may indicate learning. Both groups of animals improved their GDTs across sessions (sessions 1 vs 3: CTR: p < 0.0001, ELS: p = 0.009). While not significant, there was a trend for CTR animals to improve thresholds across sessions more than ELS animals (sessions 3 vs 1: t(47) = 1.3717, p = 0.088; Fig. 3A,D). Consistent with this trend, response amplitudes across all gap durations improved more for Control than ELS animals (comparing sessions 1 and 3, t(326) = 2.45, p = 0.015).
Because ELS often has differential effects on males and females, and because CORT levels were more affected by an acute stressor in male gerbils, we applied a mixed ANOVA with between-subject factors of sex and treatment and a within-subject factor of test session. There was no significant effect of sex on GDTs, indicating that ELS affected gap detection similarly in males and females (ELS n = 10 females, 13 males; Control n = 15 females, 11 males; Fig. 3E).
Neural gap detection in the auditory periphery and brainstem is impaired by ELS
The detection of brief gaps in sound (<100 ms) using the gap-PPI measure is known to require an intact auditory cortex (Ison et al., 1991; Kelly et al., 1996; Syka et al., 2002; Threlkeld et al., 2008), and gap detection is worsened by manipulation of neurons in ACx (Weible et al., 2014). Yet disruption of signal encoding earlier in the auditory pathway could also impact gap detection. We therefore measured the auditory brainstem response to gaps of varying durations, matching those used in the behavioral experiments (Fig. 4B). Amplitudes and latencies of the responses to two sequential noise bursts separated by varying gaps were measured for ABR waves i, ii, iii, and iv (Fig. 4A). Gap-ABR GDTs (defined as the shortest gap where any response was visible to the 2nd burst) were higher in ELS animals [one-way ANOVA between CTR (n = 20, 6 males) and ELS (n = 19, 10 males): F(1,38) = 9.882, p = 0.003, η2 = 0.211; Fig. 4C]. There was also greater variability of GDTs across the ELS animals (Levene's test: F(1,37) = 12.86, p = 0.001). This may have been because of a floor effect for the CTR animals, as 70% and 95% of Controls had visible responses at 2 and 3 ms, respectively, compared with 26% and 53% of ELS animals.
Peripheral and brainstem sensitivity to gaps was reduced by early-life stress. A, Example ABR waveform from gerbil, showing waves i, ii, iii, iv, and v. B, Example ABR response to two 50-ms noise bursts (gray shaded regions) separated by gaps of varying durations. C, Higher ABR gap detection thresholds for ELS animals, based on the shortest gap duration with a visible response to the 2nd noise burst. Individual data points are slightly jittered on the y-axis for visibility. D, For ELS animals, higher gap-ABR GDTs correlated with higher behavioral GDTs; there was no correlation for CTR animals, potentially because of a floor effect for CTR gap-ABR GDTs. E, ELS did not alter latency shift between the two noise bursts except at the shortest gap duration for wave ii. F, ELS reduced the gap-ABR amplitudes to the 2nd burst normalized by the 1st burst, for waves i, ii, and iii. G, ELS reduced the amplitude to the 2nd burst significantly for wave i and with a trend for wave ii, but increased the amplitude to the 1st burst for wave iv. Pale lines depict individual animal responses to the 2nd burst for waves i, ii, and iii, and to the 1st burst for wave iv. ***p ≤ 0.001, **p ≤ 0.01, *p < 0.05, †p < 0.1; in boxplots, box edges are 25th and 75th percentiles, with whiskers extending to the most extreme data points excluding outliers; pale lines show individual animals.
For a subset of animals, before ABR recordings, behavioral GDTs were measured and averaged across the three sessions, to assess any correlation between behavioral and ABR GDTs. Control animals did not show any correlation, potentially because of a floor effect for ABR GDTs [CTR (n = 10): Spearman's correlation: ρ = −0.18, p = 0.62; Fig. 4D]. Yet for ELS animals, poorer behavioral GDTs were positively correlated with poorer ABR GDTs [ELS (n = 12): ρ = 0.62, p = 0.031]. This suggests that in ELS animals, a response following the gap does not occur reliably at the level of the auditory nerve, and thus may not provide high-fidelity signal input to auditory cortex, which is required for the detection of short gaps.
As expected, for both Control and ELS animals and all ABR waves, the response to the 2nd noise burst occurred at a delayed latency compared with the response to the 1st burst, with larger delays following shorter gaps (Fig. 4E). Similarly, the response amplitude to the 2nd burst was reduced compared with that to the 1st burst, particularly for shorter gaps (Fig. 4F,G). The latency shift did not differ between CTR and ELS animals for ABR wave i, iii, or iv. For wave ii, ELS animals had a smaller latency shift that was confined to the 2-ms gap duration, although only 21% of the animals showed visible responses and contributed at this gap duration (univariate mixed ANOVA, with independent variables of treatment and gap durations: wave ii, F(1,261) = 7.348, p = 0.007, partial η2 = 0.027; post hoc at 2 ms, p < 0.001; Fig. 4E). However, the amplitude ratio of the two responses (effectively normalizing the 2nd burst response for each animal) was smaller for the ELS animals for waves i, ii, and iii (wave I, F(1,261) = 10.910, p = 0.001, partial η2 = 0.04; wave ii, F(1,261) = 7.117, p = 0.008, partial η2 = 0.027; wave iii, F(1,261) = 4.544, p = 0.034, partial η2 = 0.018; Fig. 4F). For wave i and a trend for wave ii, this effect was driven by a reduced response to the 2nd burst (wave i: F(1,261) = 9.324, p = 0.002, partial η2 = 0.034; wave ii: F(1,261) = 2.765, p = 0.098, partial η2 = 0.010; Fig. 4G, left two panels). In contrast, for wave iv only, there was a decreased response in ELS animals to the 1st burst (F(1,295) = 4.409, p = 0.037, partial η2 = 0.015; Fig. 4G, right panel).
Neural gap detection in the auditory cortex is impaired by ELS
Although gap-PPI does not involve attention, the detection of short gaps (<100 ms) requires an intact auditory cortex (Ison et al., 1991; Kelly et al., 1996; Syka et al., 2002; Threlkeld et al., 2008). Thus, the behavioral deficit in the ELS animals suggests that cortical sensitivity to gaps may have been impaired by ELS induced during the CP for ACx maturation. To test this hypothesis, we recorded single-unit and multiunit activity [104 CTR from nine animals (two males) and 108 ELS units from 12 animals (four males)] from primary ACx with tungsten microelectrodes during the week following the last behavioral session, when animals were P40–P49. We conducted these recordings in urethane-anesthetized animals, to minimize the impact of top-down influences from higher-level regions, as we were interested in bottom-up auditory encoding (Mashour, 2014; Cai et al., 2016). Gap-detection functions were collected with carriers of either tones at each unit's best frequency, or bandpass noise matching that used in the behavioral testing, and GDTs were calculated for each unit. Figure 5A–D depicts four different gap-responsive units, to represent the variability of response types and to show different GDTs. Gap detection was measured based on firing immediately following gap offset (i.e., the response to the 2nd sound burst).
Early-life stress reduced cortical sensitivity to gaps, with a population cortical model predicting poorer behavioral sensitivity. A–D, PSTHs (with overlaid rasters) from four units in response to gaps of varying durations. Shaded regions indicate sound presentation, and blue-dotted boxes indicate the time windows used to evaluate responses, with significant gap-evoked responses indicated by thicker lines. The shortest gap with a significant response is the threshold. Panels A, B, and D are from ELS animals; panel C is from a Control animal. E, GDTs measured from gap detection functions with tone carriers at the BF of each unit were higher for ELS units. Cells with GDTs > 125 showed no response following the longest gap presented, despite having clear sound-evoked responses. F, Compared with Controls, the distribution of GDTs was shifted to the right across ELS units responding to tones, with fewer units sensitive to short gaps. G, Differences in firing rate between the 1st and 2nd burst were larger in Controls, particularly at shorter gap durations. H, A population-based ideal observer model trained with tone-evoked firing rates to gaps of varying durations showed greater error (a higher misclassification rate) for short gaps in ELS animals. I, Gap detection functions with a noise carrier matching that used in behavioral testing produced higher overall GDTs in both groups than with tone carriers. Noise-evoked GDTs were not significantly worse in ELS animals. J, The distribution of noise-evoked GDTs was similar in CTR and ELS animals. K, Differences in firing rate between the 1st and 2nd burst. L, Despite nonsignificant GDTs, an ideal observer model trained with noise-evoked firing showed higher misclassification across nearly all gap durations for ELS animals. ***p ≤ 0.001, **p < 0.01, *p < 0.05; in boxplots, box edges are 25th and 75th percentiles, with whiskers extending to the most extreme data points excluding outliers; shaded regions around FR indicate SEM; error bars show SEM; small pale circles show individual animals.
For tone carriers, cortical GDTs were higher for ELS units than Controls (Kruskal–Wallis ANOVA: χ2(1,210) = 16.8, p < 0.0001; CTR n = 104 units, 9 animals, ELS n = 108 units, 12 animals; Fig. 5E). Comparing the distribution of neural GDTs shows a clear shift for ELS animals, with more cells having higher GDTs and fewer cells with sensitivity to short gaps (Fig. 5F). Across gap durations, firing following the gap (i.e., in response to the second tone burst) was reduced by ELS, both when normalized by the response to the first tone burst (subtracting firing rates of burst 2 from burst 1 in each cell, at each gap duration; Fig. 5G; χ2(1,2451) = 11.1, p < 0.001), and also in response to each tone burst (burst 1: χ2(1,2451) = 47.2, p < 0.0001, burst 2 χ2(1,2451) = 111.6, p < 0.0001). Gap detection functions collected with a noise carrier matching that used in the behavioral testing (n = 101 Control and 94 ELS units with noise responsivity) showed the same pattern, although GDTs were higher in both groups than with a tone carrier, with many more units insensitive to even 125 ms, the longest gap presented (χ2(1,193) = 5.8, p = 0.016; Fig. 5I,J). Across gap durations, firing was reduced in response to each noise burst (burst 1: χ2(1,2429) = 30.3, p < 0.0001, burst 2 χ2(1,2429) = 155.1, p < 0.0001), although normalized firing was not significantly reduced (χ2(1,2429) = 1.2, p = 0.27; Fig. 5K). Despite the general correspondence between cortical and behavioral GDTs across treatment groups, there was no significant correlation between cortical and behavioral GDTs for either group, using either tone or noise carriers (data not shown).
To determine how well the combined information across the population of neural units could discriminate between gap and no-gap trials, and whether this population discrimination performance predicted behavioral differences across the groups, we trained and tested an ideal observer classifier model using the firing rates following gap offset, i.e., in a time window following the 2nd sound burst. Figure 5G,K depict the means of those firing rates, normalized to the response following the 1st sound burst. Cortical responses at gap offset have been shown to directly influence gap detection measured by gap-PPI (Weible et al., 2014). Thus, a model trained with these responses should reflect behavioral GDTs, showing poorer detection at shorter gaps. Furthermore, that model should show poorer performance when trained with ELS neurons.
We compared the ideal observer's ability to detect gaps of varying durations for both groups, with either tone or noise carriers (Fig. 5H,L). Consistent with the behavior, for both Controls and ELS the model shows greater error distinguishing short gap responses from no-gap responses. Comparing treatments, for tone carriers the model further shows greater error detecting short duration gaps (<5 ms) in the ELS group than in Controls. For noise carriers, greater ELS error occurs for both short and long gaps. The model is thus consistent with the poorer behavioral gap detection seen in ELS animals.
ELS reduces the magnitude of auditory cortical responses
Manipulating auditory experience during the developmental period over which ELS was induced here is known to affect basic response properties of auditory cortical neurons (Zhang et al., 2002; Rosen et al., 2012; Green et al., 2017). We thus assessed whether ELS during that period altered cortical responses to tones. First-spike latencies (FSL) were measured from responses to the first burst of sound in gap detection functions acquired with tone carriers at each unit's BF. Neither FSL nor FSL jitter were altered by ELS (Fig. 6C,D). However, the firing rate evoked by the first sound burst was reduced in ELS animals, as was the spontaneous activity in the 100 ms preceding the first burst (KW ANOVAs, spontaneous activity: χ2(1,221) = 19.11, p < 0.0001; sound-evoked activity: χ2(1,221) = 3.94, p = 0.047; Fig. 6E,F). This effect extended to measures acquired from rate level functions acquired at BF. ELS animals had lower firing rates at sound-evoked threshold, and those thresholds were higher for ELS animals (threshold FR: χ2(1,199) = 7.85, p = 0.005; threshold dB: χ2(1,199) = 8.66, p = 0.003; Fig. 6G,H). The dynamic range was not affected by ELS, but there were fewer nonmonotonic units in ELS animals, indicated by a higher monotonicity index (monotonicity index: χ2(1,199) = 6.61, p = 0.010; Fig. 6I,J).
Sound-evoked and spontaneous cortical firing, but not first-spike latency, were reduced by early-life stress. Examples of rate-level functions from (A) Control and (B) ELS animals are shown for cells from the 99th, 35th, 25th, and 15th percentiles of the monotonicity index distributions for each treatment group. Gray dashed lines indicate threshold FR and threshold dB SPL; purple shaded regions show dynamic range; insets are PSTHs of responses at 60 dB SPL, where gray shaded regions indicate tone presentation, and axes are firing rate over time. Neither (C) FSL nor (D) FSL jitter were affected by ELS. In contrast, ELS reduced (E) evoked firing, (F) spontaneous firing, and (G) firing at threshold, along with causing (H) higher sound-evoked thresholds. I, Dynamic range was unaffected by ELS, but (J) monotonicity was increased by ELS. ***p ≤ 0.001, **p ≤ 0.01, *p < 0.05; in boxplots, box edges are 25th and 75th percentiles, with whiskers extending to the most extreme data points excluding outliers; small pale circles show individual animals.
Discussion
To examine the effects of ELS on temporal processing, we evaluated gap detection, a common measure of temporal acuity important for perceiving communication calls and speech. We developed a model of ELS in the Mongolian gerbil, verified by shifts in corticosterone and responses to startling stimuli. ELS impaired perception of short gaps in sounds, and reduced the improvement in gap detection across behavioral sessions, suggesting a possible learning deficit. Consistent with the necessity of ACx for short gap detection, behavioral deficits coincided with poorer ACx gap detection thresholds and fewer neurons sensitive to short gaps. Further, ABR responses to gaps were reduced in the auditory nerve and brainstem. In ELS animals, behavioral and ABR gap detection thresholds were correlated. This suggests that the behavioral deficits may arise at least in part from an unreliable response to short gaps in the auditory nerve, providing a poor-fidelity signal to ACx, a region required for the detection of short gaps. These effects reveal a temporal processing deficit at multiple levels of the auditory pathway, reducing the fidelity of information available to higher-level regions necessary for speech perception and cognition. This impaired sensory encoding may contribute to known ELS-related problems with attention and learning, and may be a risk factor for auditory processing disorder or specific language impairment.
Neural loci of ELS effects on auditory gap detection
While there were deficits in temporal encoding at multiple levels of the auditory system, ELS-induced changes in ACx likely contributed to altered perception, because ACx combines inputs from the ascending auditory pathway and is necessary for gap detection. As hypothesized based on the critical period elements commonly affected by both ELS and sensory deprivation, ELS has effects similar to those caused by hearing loss (HL) in gerbil during this window for ACx maturation: poorer behavioral and auditory cortical gap detection (Green et al., 2017). Because gap perception relies on activity of ACx inhibitory neurons (Weible et al., 2014), deficits from both HL and ELS may arise from altered ACx inhibitory function. The ELS-induced reduction in nonmonotonic ACx cells (Fig. 6J) also suggests inhibitory changes, as nonmonotonicity in ACx is shaped by local inhibition (Wu et al., 2006). Similarly, altered inhibitory function in prefrontal cortex, hippocampus, and amygdala arising from ELS coincides with emotional and cognitive problems (Bath et al., 2016; Castillo-Gómez et al., 2017; Murthy et al., 2019; Page and Coutellier, 2019; Manzano Nieves et al., 2020). ELS causes early maturation of perineuronal nets, reduction in inhibitory neurons, and changes in BDNF levels in frontal cortex and amygdala (Cameron et al., 2017). These elements are involved in CP regulation, including in ACx (Hensch, 2005; Do et al., 2015). Thus, the locus and details of ACx changes underlying ELS-induced and HL-induced perceptual deficits may arise from common mechanisms involving CP perturbation. An additional link is that both ELS and HL reduce ACx excitability (Fig. 6E,F; Green et al., 2017). Since levels of cortical activity are tightly linked with mechanisms of plasticity, the overall reduction in neural activity by both manipulations indicates a common feature linking CP dysregulation.
Cochlear function is still immature during our stress induction, suggesting that ELS could induce cochlear plasticity (Puel and Uziel, 1987; Arjmand et al., 1988; Mills and Rubel, 1996; Abdala and Keefe, 2012). Developmental conductive HL does not affect the cochlea in terms of frequency tuning (Caras and Sanes, 2015; Ye et al., 2021) or temporal processing (Yao and Sanes, 2018). Yet ELS impaired auditory nerve gap detection (Fig. 4F,G, wave i). This could arise from direct effects of stress on elements of the cochlea. Acute stress induction protects against cochlear noise damage (Yoshida et al., 1999; Wang and Liberman, 2002), via glucocorticoid receptor activation within the cochlea (Jin et al., 2009), and a cochlear system equivalent to the HPA axis (Basappa et al., 2012). Indeed even in adult animals, circulating levels of corticosterone correlate with wave I amplitude and inner hair cell synapse number (Singer et al., 2018). This suggests that changes in corticosteroid circulation could alter peripheral sound processing (Knipper et al., 2015). Here, wave i amplitude to only the 2nd noise burst was reduced, indicating problems with recovery from adaptation rather than with auditory sensitivity, perhaps implicating involvement of vesicular release from inner hair cells.
Alternatively, stress may modulate efferent feedback from the olivocochlear system, which normally refines inner hair cell sensitivity and frequency selectivity, and protects against HL (Fuchs and Lauer, 2018). This efferent system is susceptible to top-down effects, as it can improve selective attention to stimuli by modifying cochlear sensitivity (Delano et al., 2007; Gehmacher et al., 2022). Further, behaviorally-relevant plasticity has been demonstrated at ABR wave i (Rotondo and Bieszczad, 2020), and may indicate a role of olivocochlear feedback in perceptual learning (de Boer and Thornton, 2008; Irving et al., 2011; Fuchs and Lauer, 2018). Thus, ELS-induced changes in the brainstem, centrally, or directly on the cochlea could alter peripheral function.
Given that ELS resulted in both central and peripheral effects, it is tempting to speculate about the relative contributions of peripheral and central gap detection deficits to perception. The impaired gap detection seen here in ACx neurons may be inherited from the poorer gap detection in the auditory nerve. However, ABR wave iv (reflecting midbrain activity) showed a recovery of gap detection encoding compared with auditory nerve and cochlear nucleus (Fig. 4F,G). This may indicate central compensation, which is known to occur following HL and aging (Caspary et al., 2008; Sanes and Kotak, 2011). The shapes of the response profiles across gap durations can be compared across behavioral, peripheral, and cortical responses, but yield an imperfect match. Further, within-animal measures of behavioral and ACx GDTs were not correlated for either tone or noise carriers. This is presumably because of the variability of the gap-PPI measure, the numbers of units recorded per animal, and the nonsimultaneous measures, and can be addressed by future recordings conducted during gap-PPI sessions.
Relations to known effects of stress
Our ELS manipulation decreased baseline and acute stress-induced corticosterone levels when measured in juvenile (P25) or adolescent (P33) gerbils, respectively. While ELS frequently increases CORT levels, effects of neonatal stress on HPA function vary greatly with type of stress, the developmental timing of ELS induction, and the age of assessment, and these variations can yield reduced CORT levels (van Bodegom et al., 2017). A decreased CORT response may arise from early disruption of the HPA axis, causing increased engagement of negative feedback. Because ELS is typically induced earlier in development than in our model (which is timed to begin just before hearing onset), and because ELS is rarely studied in gerbils, there are no other studies that mimic our ELS model, making it difficult to predict the effects on the HPA axis. Both hyperreactivity and hyporeactivity of the HPA axis are detrimental to many aspects of normal development, so it is expected that perturbation in either direction could negatively impact auditory function. The auditory temporal processing deficits in our model provide the opportunity to evaluate the influence of reduced CORT on the development of sensory neural circuits.
We did not evaluate whether adult stress similarly affected gap detection, because restraint will inevitably induce different amounts of stress in preweaned versus adult animals because of their relative nursing and mobility differences. However, we previously showed that when animals were restrained as preweanlings versus adults and tested two months later, only the younger animals had gap detection deficits (Green et al., 2016). Further, the reduced ACx responses to pure tones in ELS animals (Fig. 6E–G) contrast with adult-stressed rodents, where a glucocorticoid agonist applied to ACx enhanced tone-evoked responses and broadened tuning (Lei et al., 2014), and with human adults, where corticosteroids or stress enhanced auditory evoked potentials and heightened auditory perceptual sensitivity (Ashton et al., 2000; Hasson et al., 2013). Yet our reduced ACx responses are consistent with reduced ACx evoked potential magnitudes in developmentally-stressed rats (Ellenbroek et al., 2004). Thus, while adult activation of the HPA axis likely affects ACx, the underlying mechanisms may differ.
While it is not surprising that altered HPA activity affects central auditory function, studies have consistently interpreted stress-induced auditory deficits in terms of altered sensory gating, a preattentional mechanism that filters irrelevant information by reducing responses to successive sounds (White and Yee, 1997; Rosburg et al., 2009). Experimentally, gating is a smaller P50 event-related potential to the 2nd (vs 1st) of two sounds separated by ∼500 ms. Impaired gating is a biomarker for psychiatric disorders (Patterson et al., 2008; Javanbakht et al., 2011), and gating is altered by adult and early-life stress (White and Yee, 1997; Ellenbroek et al., 2004; Maxwell et al., 2006; Cromwell and Atchley, 2015; Ma et al., 2015; Yates et al., 2016). While sensory gating involves ACx (Josef-Golubic, 2020), gap detection operates on a much faster timescale: detection of gaps >100 ms does not require ACx, and thus sensory gating for a 500 ms interval is likely probing a different phenomenon. In rats, ELS-impaired gating is driven by a smaller P50 response to the first of two sounds (Ellenbroek et al., 2004), consistent with our result of reduced ACx firing to simple tones (Fig. 6C). Separately but similarly, the behavioral phenomenon of “sensorimotor gating” (the reduction of a startle response by a preceding pulse) is also altered in stress and psychiatric disorders (Light and Braff, 1999; Ellenbroek et al., 2004; Huggenberger et al., 2013). However, sensorimotor gating is distinct from the gap-PPI used here because it measures the response to a single sound, using prepulse inhibition (PPI) of acoustic startle. PPI does not require ACx, as it is intact with ACx inactivation (Threlkeld et al., 2008).
One question is whether our effects are a result of top-down influences from regions known to be altered by ELS. For example, the amygdala has indirect effects on ACx via neuromodulatory inputs (Wenstrup et al., 2020), which may partially contribute to the reduced gap-PPI. Importantly, our neural recordings were conducted under anesthesia, minimizing the impact of top-down influences from higher-level regions (Mashour, 2014; Cai et al., 2016). Further, we used a short (50 ms) interval between stimulus and startle in our gap-PPI to ensure that it was preattentive (Li et al., 2009), reducing the impact of higher-level regions.
In conclusion, the behavioral deficits seen here may arise from plasticity at both the level of ACx and earlier in the auditory pathway. Critical period mechanisms can now be examined to identify elements that are altered by early-life stress, and whether those differ from elements altered by disruptions in early auditory experience, such as developmental hearing loss.
Footnotes
This work was supported by the National Institute on Deafness and Other Communication Disorders Grant R01 DC013314 (to M.J.R.). We thank Dr. J. J. Huyck and Dr. L. Coutellier for comments on an earlier version of the manuscript.
The authors declare no competing financial interests.
- Correspondence should be addressed to Merri J. Rosen at mrosen{at}neomed.edu
