Remodelling at the calyx of Held–MNTB synapse in mice developing with unilateral conductive hearing loss

Abstract

Structure and function of central synapses are profoundly influenced by experience during developmental sensitive periods. Sensory synapses, which are the indispensable interface for the developing brain to interact with its environment, are particularly plastic. In the auditory system, moderate forms of unilateral hearing loss during development are prevalent but the pre-and postsynaptic modifications that occur when hearing symmetry is perturbed are not well understood. We investigated this issue by performing experiments at the large calyx of Held synapse. Principal neurons of the medial nucleus of the trapezoid body (MNTB) are innervated by calyx of Held terminals that originate from the axons of globular bushy cells located in the contralateral ventral cochlear nucleus. We compared populations of synapses in the same animal that were either sound deprived (SD) or sound experienced (SE) after unilateral conductive hearing loss (CHL). Middle ear ossicles were removed 1 week prior to hearing onset (approx. postnatal day (P) 12) and morphological and electrophysiological approaches were applied to auditory brainstem slices taken from these mice at P17–19. Calyces in the SD and SE MNTB acquired their mature digitated morphology but these were structurally more complex than those in normal hearing mice. This was accompanied by bilateral decreases in initial EPSC amplitude and synaptic conductance despite the CHL being unilateral. During high-frequency stimulation, some SD synapses displayed short-term depression whereas others displayed short-term facilitation followed by slow depression similar to the heterogeneities observed in normal hearing mice. However SE synapses predominantly displayed short-term facilitation followed by slow depression which could be explained in part by the decrease in release probability. Furthermore, the excitability of principal cells in the SD MNTB had increased significantly. Despite these unilateral changes in short-term plasticity and excitability, heterogeneities in the spiking fidelity among the population of both SD and SE synapses showed similar continuums to those in normal hearing mice. Our study suggests that preservations in the heterogeneity in spiking fidelity via synaptic remodelling ensures symmetric functional stability which is probably important for retaining the capability to maximally code sound localization cues despite moderate asymmetries in hearing experience.

Introduction

Prolonged periods of sensory deprivation early in development have long-lasting maladaptive effects on the structure and function of the nervous system (Knudsen, 2004). In the auditory system, severe hearing loss early in development alters the synaptic connectivity of neurons throughout the central auditory neuraxis manifesting as abnormalities in axonal projections and branching (Kitzes et al. 1995; Russell & Moore, 1995; Hsieh & Cramer, 2006; Franklin et al. 2006, 2008), re-organization of midbrain and cortical maps (Harrison et al. 1991, 1998; Lomber et al. 2010; Pienkowski & Eggermont, 2011), changes in synaptic protein and receptor expression (Suneja et al. 1998; Hutson et al. 2007; Takesian et al. 2010; Zeng et al. 2012) and modifications in synaptic ultrastructure (Ryugo et al. 2005). Surprisingly though, even moderate forms of early hearing loss can alter protein synthesis, metabolism, receptor expression, connectivity and the response properties of auditory neurons (Moore et al. 1989; Suneja et al. 1998; Tucci et al. 1999; Hutson et al. 2007; Xu et al. 2007; Whiting et al. 2009; Popescu & Polley, 2010; Harrison & Negandhi, 2012; Takesian et al. 2012). Moreover, if the hearing loss is unilateral, these effects may occur bilaterally (Moore et al. 1989; Suneja et al. 1998; Tucci et al. 1999; Sumner et al. 2005; Popescu & Polley, 2010; Wang et al. 2011) and in some cases may be more pronounced than if the hearing loss occurred in both ears (Silverman & Clopton, 1977; Feng & Rogowski, 1980). Despite these findings, very little is known with regard to how moderate unilateral hearing loss impacts the morphology of auditory synapses and neurotransmission at individual synapses. Given the prevalence of moderate forms of hearing loss in children and its association with performance deficits related to sound localization, speech and language acquisition (Sanes & Woolley, 2011; Whitton & Polley, 2011), understanding how pre-and postsynaptic elements of auditory neurons are precisely affected by hearing asymmetries during development remains of central importance.

The calyx of Held, which is located in the auditory brainstem, is a glutamatergic nerve terminal that innervates a single principal neuron of the medial nucleus of the trapezoid body (MNTB; Borst & Soria van Hoeve, 2012). Owing to its large size and experimental tractability, the calyx of Held–MNTB synapse has served as a powerful model system to uncover key developmental features of synaptic transmission. Before hearing onset (∼P12 in rodents), the calyx of Held terminal resembles a spoon-shaped structure but subsequently transforms into a highly digitated/fenestrated structure about 1 week after hearing onset (Kandler & Friauf, 1993; Wimmer et al. 2006; Ford et al. 2009). This gross morphological change is accompanied by subsynaptic changes in the number of active zones (Taschenberger et al. 2002; Dondzillo et al. 2010), docked synaptic vesicles (Taschenberger et al. 2002), filamentous structures (Yang et al. 2010) and postsynaptic receptors (Hermida et al. 2010). In parallel, the calyx of Held–MNTB synapse transforms biophysically, which is reflected in the narrowing of presynaptic action potentials (Taschenberger & von Gersdorff, 2000; Yang & Wang, 2006), more efficient Ca²⁺-triggered neurotransmitter release (Fedchyshyn & Wang, 2005), speeding of AMPA receptor-mediated EPSCs (Taschenberger & von Gersdorff, 2000; Joshi et al. 2004) and downregulation of NMDA receptor-mediated EPSCs (Taschenberger & von Gersdorff, 2000; Futai et al. 2001; Joshi & Wang, 2002). This morphological and biophysical maturation during the first 3 weeks of postnatal development transforms this synapse to discharge at high rates and transmit with high fidelity (Borst & Soria van Hoeve, 2012), a feature that is critical for delivering well-timed inhibition to neighbouring brainstem nuclei needed for computing interaural timing and level differences (ITDs and ILDs) during sound localization (Grothe et al. 2010).

Studies have previously characterized synaptic development in experimental models of severe deafness but the effects of mild to moderate forms of unilateral deafness at the calyx of Held–MNTB synapse during early development are unknown. In the deafness (dn/dn) mutant mouse model of congenital deafness, synaptic transmission and connectivity at the calyx of Held–MNTB synapse appear to develop normally (Oleskevich et al. 2004; Youssoufian et al. 2005, 2008). Similarly in the deaf Ca_v1.3^−/− mouse mutant, presynaptic machinery and postsynaptic AMPA receptors are largely unaffected by the lack of neural activity generated at the cochlea during development (Erazo-Fischer et al. 2007). Although these and other forms of bilateral deafness alter some aspects of synaptic transmission (i.e. synaptic strength, down-regulation of NMDA receptors and intrinsic principal MNTB neuron characteristics), the effects are surprisingly not very dramatic (Futai et al. 2001; Leao et al. 2004, 2006; Erazo-Fischer et al. 2007; Hirtz et al. 2012). In contrast, the synaptic modifications caused by severe unilateral or asymmetrical hearing losses early in development are more profound. For example, following unilateral cochlear ablations and cochlear nucleus lesions, calyx terminals incorrectly innervate both the ipsi-and contralateral MNTB (Kitzes et al. 1995; Russell & Moore, 1995; Hsieh & Cramer, 2006). Although this does not occur if the ablations are performed at older ages such as P8/9, the normal development of calyx terminal morphology is altered (Ford et al. 2009). Moreover, the developmental refinement of principal MNTB neuron projections to other brainstem regions such as the lateral superior olive is eliminated (Sanes & Takacs, 1993) and the expression patterns of calcium-binding proteins in the MNTB (Hatano et al. 2009) are reduced following unilateral cochlear ablations. These studies, as well as others demonstrating more profound effects in unilateral versus bilateral forms of severe auditory deprivation (Silverman & Clopton, 1977; Feng & Rogowski, 1980) suggest that aural imbalances profoundly affect normal synaptic development. But whether the morphological and biophysical development typically observed at the calyx of Held–MNTB synapse is disrupted by moderate forms of unilateral deafness have not been examined.

To address this, we investigated the synaptic properties at the calyx of Held–MNTB synapse in brainstem slices taken from mice aged P17–19 that underwent middle ear ossicular removal 1 week before hearing onset to induce a unilateral conductive hearing loss (CHL). Owing to its neuroanatomical design and positioning within the auditory brainstem circuit, the calyx of Held–MNTB synapse offers a unique advantage to the examination of the effects of moderate unilateral hearing loss on synaptic development. Because principal neurons of the MNTB are innervated by calyx of Held terminals that originate from the axons of globular bushy cells (GBCs) located in the contralateral ventral cochlear nucleus (VCN; Borst & Soria van Hoeve, 2012), we can precisely identify the population of synapses that have been deprived of normal acoustic input. Thus, the main goals of the study were (1) to compare the synaptic properties of synapses located in the sound-deprived (SD) contralateral MNTB with those located in the intact sound-experienced (SE) ipsilateral MNTB, and (2) to determine whether the structural and functional continuum within the MNTB (Grande & Wang, 2011) is altered amongst the population of SD and SE synapses in mice developing with aural asymmetries. We found that following a unilateral CHL during early development, calyx terminal morphology increases in complexity in both the SD and SE MNTB. However, this bilateral shift in the structural continuum was not accompanied by a corresponding shift in the functional continuum as defined by the heterogeneities in principal cell spiking during high frequency stimulation. This occurred despite bilateral changes in various aspects of synaptic transmission, unilateral increases in principal cell excitability in the SD MNTB and unilateral changes in the form of short-term plasticity (STP) expressed in the SE MNTB. The preservation of heterogeneities in spiking fidelity via synaptic remodelling may ensure that the coding capability of sound localization cues is retained at the calyx of Held–MNTB synapse in the event of moderate asymmetrical disturbances in hearing experience during sensitive periods of development.

Methods

All experimental procedures were performed in CD–1 mice (Charles River Laboratories, Quebec, Canada) of either sex during the first 3 postnatal weeks in accordance with the Hospital for Sick Children Laboratory Animal Service review committee (under the guidelines of the Canadian Council on Animal Care). Data were collected in normal hearing (NH; n = 16) or surgical (SURG; n = 14) mice that underwent unilateral ear surgery (left ear only in all cases) at age P5–6 to produce a unilateral conductive hearing loss (CHL) approximately 1 week before hearing onset. In a small number of mice (n = 2) sham surgeries were also performed in the left ear. We also performed a separate set of control experiments in CD–1 mice (n = 5) included as Supporting information (Fig. S1, available online) to ensure that this mouse strain has normal hearing across all frequencies. These experiments were performed in mice aged 4–5 weeks. In all cases, mice were housed in the animal facility and thus exposed to ambient noise levels of the institutional vivarium until additional experiments were performed at P17–19.

Unilateral ear surgery

Mouse pups (P5–6) were anaesthetized using 2–3% isofluorane (maintained at 1%) delivered through a custom-made nose cone. After ensuring an absent withdrawal reflex, a small post-auricular incision was made behind the left ear to expose the ear canal with the aid of a dissecting microscope. To produce a CHL, the tympanic membrane was ruptured and ossicles (malleus and incus) were removed using fine forceps. Gelfoam was placed in the ear canal and the incision was closed using Vetbond (3M Company, St Paul, Minnesota, USA). Ketoprofen (5 mg kg^–1) and saline were given subcutaneously (s.c.) for analgesia and hydration, respectively, prior to gas anaesthesia removal. In sham experiments, all procedures were identical except for tympanic membrane rupture and ossicle removal. Pups recovered on a warm heating pad until mobility was restored and then returned to their litter.

Auditory brainstem responses (ABRs)

NH, SURG and sham mice aged P17–19, just prior to slice experiments, were given a ketamine (1.5 mg kg⁻¹) and xylazine (0.5 mg kg⁻¹) mix (s.c.) to induce anaesthesia. ABRs were recorded in response to acoustic click stimuli (31 μs duration) delivered using broadband (16 kHz) earphones (ER2, Etymotic Research, Elk Grove Village, IL, USA) inserted into the external auditory meatus. Click stimuli, which were presented with alternating polarity to minimize the contribution of the cochlear microphonic, were delivered at sound pressure level (SPL) from 0 to 80 dB in 10 dB increments. Needle electrodes in the standard vertex–mastoid configuration were used to record the ABR signals (Intelligent Hearing Systems, Miami, FL, USA) which were amplified (×1000), averaged (512 sweeps) and filtered (100–1500 Hz). Each stimulus was calibrated by connecting the output transducer into a small volume cavity, <0.3 ml, to simulate the small cavity of the mouse ear canal. This was probed with a half-inch condenser microphone (PCB Piezotronics, Model 377B02 Condenser Microphone, Depew, NY, USA) connected to an integrating sound level (Larson Davis, model 831, Depew, NY, USA) and to a sound spectrum analyser (SR760 FFT, Stanford Research Systems Sunnyvale, CA, USA). In wild-type CD–1 mice aged 4–5 weeks, ABRs were recorded following 8, 16 and 32 kHz tone stimulations delivered to the right and left ears using a high-frequency transducer (Intelligent Hearing Systems). All recordings were carried out in a sound attenuation room. The lowest sound level at which ABR peaks could be identified was considered to be ABR threshold. ABR waveforms were included in the analysis if the characteristic peaks I–V (see Fig. 1A) were clearly visible.

Figure 1.

A, sample ABRs from normal hearing (NH) and surgical (SURG) mice. Clicks presented in 10 dB increments from 0 to 80 dB SPL to each ear. ABR waveforms for NH mice are shown in black. ABR waveforms recorded from the impaired left ear of SURG mice shown in red and those from intact right side shown in blue. B, ABR thresholds recorded in the left and right ears of NH (n = 9), sham (green, n = 2) and SURG (n = 12) mice. Matching symbols between right and left ear correspond to the same animal (i.e. left filled circle and right filled circle etc.). Mean ABR thresholds are indicated. Because ABR thresholds in sham mice were not different versus NH, we included these data points in the mean values. C, ABR amplitude–intensity functions for waves I and IV (normalized to the amplitude at 80 dB SPL). Amplitudes were initially calculated by taking the difference between the negative peak and the preceding positive peak. Given the similarity in ABR threshold, data obtained from the right and left ears of NH and sham mice were combined. D, ABR waveforms from different mice (grey traces) were averaged (thick traces) at 70 and 40 dB. These averaged traces were normalized to peak I and overlaid to compare responses at 70 and 40 dB. *Significant difference.

Anterograde tracing, image acquisition and analysis

The morphology of the calyx of Held in NH and SURG mice was determined as previously described (Grande & Wang, 2011). After decapitation using a custom-made guillotine, brainstem slices from mice were perfused with oxygenated aCSF at room temperature. Borosilicate glass microelectrodes, with a tip diameter of 10–15 μm, containing 5% biotinylated dextran amine (BDA; D1956, Invitrogen, Carlsbad, CA, USA) in 0.4 m KCl were introduced at the midline with the aid of a microscope and DIC monitor. Calyceal axons were labelled by 1–3 iontophoretic injections (5 μA, 7 s on/off duty cycle, 5 min each, A.M.P.I, Model Master-8, Jerusalem, Isreal). Slices were incubated for 90 min and fixed in 4% paraformaldehyde (PFA) in PBS at 4°C. Following PBS rinses, slices were incubated with streptavidin conjugated to Alexa 488 (Invitrogen, S11223) for 3 h at room temperature under gentle agitation to visualize BDA-labelled calyces under fluorescence. Slices were then mounted and coverslipped with Vectashield (Vector Laboratories, Burlingame, CA, USA). In a separate set of experiments, DiI crystals (Invitrogen) were placed in the left VCN of P17–19 SURG mice to ensure that ossicular removal at such early developmental ages did not result in aberrant calyx projections to the ipsilateral MNTB as shown to occur when cochlear ablations are performed (Kitzes et al. 1995; Russell & Moore, 1995). Brainstem blocks from SURG mice were placed in 4% PFA for 3 days and DiI crystals were then placed in left VCN. Blocks were then incubated for 4–5 weeks at 35°C to allow for DiI transport to contralateral MNTB. Slices (∼150 μm thick) were then collected and placed in PBS.

Confocal images of only those calyces well-labelled with BDA were acquired with a Zeiss laser scanning microscope (LSM 710) equipped with a 488 nm argon laser line for Alexa 488 excitation and ×63 (1.4 NA, oil) objectives. Z-stack images (0.5 μm steps) of individual calyces of Held were acquired and three-dimensional reconstructions were rendered using Volocity (Perkin Elmer, Waltham, MA, USA). Surface area and volume measurements were calculated using the region-of-interest function in Volocity. These measurements did not include the axon belonging to the calyx terminal. Visual inspection of 3–D-rendered calyces was used to determine the number of swellings on each calyx which are elliptical structures connected to the stalk via a thin neck (Grande & Wang, 2011). Low-power (×5; 0.12 NA) confocal images of DiI-labelled slices in PBS were taken using a 543 nm argon laser. Contrast and brightness were adjusted using Volocity.

Electrophysiology and data analysis

Following ABR recordings in NH, SURG and sham mice, transverse brainstem slices (∼250 μm thick) containing the MNTB were collected. After decapitation, brains were rapidly removed and immersed in oxygenated (95% O₂ and 5% CO₂) semifrozen artificial CSF (ACSF; in mm: 125 NaCl, 2.5 KCl, 10 glucose, 1.25 NaH₂PO₄, 2 sodium pyruvate, 3 myo-inositol, 0.5 ascorbic acid, 26 NaHCO₃, 1 MgCl₂ and 2 CaCl₂). Slices were collected using a vibratome (VT1000S, Leica, Nussloch, Germany) and incubated in a warm bath (35°C) for 1 h before experimentation. Before incubation, the right sides of the slices taken from SURG animals were notched. This allowed for easy identification of the right SD MNTB, which under normal hearing conditions would receive input from the left ear. Consequently the left MNTB, which we are referring to as the SE MNTB, corresponding to input received by the right ear, could also be easily identified.

Slices were transferred to a recording chamber continuously perfused with oxygenated ACSF at room temperature. ACSF was supplemented with bicuculline (10 μm) and strychnine (1 μm) to block inhibitory inputs. A bipolar platinum electrode, coupled to a Master 8 stimulator (A.M.P.I, Jerusalem), was placed near the midline to stimulate calyceal axons (2 × threshold) to calyx terminals. Using a dual-channel amplifier (Multiclamp 700 A, Molecular Devices, Union City, CA, USA) we obtained cell-attached and whole-cell voltage clamp recordings from principal neurons located in the SD and SE MNTBs of SURG mice and the MNTBs of NH and sham mice. Cell-attached voltage-clamp recordings (>1 GΩ resistance) were made at −60 mV holding potential during midline stimulation of calyceal axons (100–400 Hz trains, 200 ms) at room temperature. Under such conditions, we registered inward and outward compound currents due to action potentials (APs). Subsequently, membrane patches from the same cells were ruptured to allow whole-cell recordings of EPSCs. To record EPCSs during midline stimulation, patch electrodes with resistances of 2.5–3 MΩ containing (in mm): 97.5 potassium gluconate, 32.5 CsCl, 5 EGTA, 10 Hepes, 1 MgCl₂, 30 TEA and 3 lidocaine (lignocaine) N-ethyl bromide (pH 7.2) were used. Holding potential was maintained at −60 mV. Series resistance ranged between 4 and 8 MΩ and was compensated to 90%. To estimate current density, current–voltage (I–V) curves were generated by recording EPSCs −60 mV to +60 mV in 20 mV increments and normalized to capacitance (nA/pF). In a separate set of experiments, we examined principal MNTB cell excitability by performing whole-cell current-clamp recordings using electrodes with resistances of 2.5–3 MΩ containing (in mm): 97.5 potassium gluconate, 32.5 KCl, 0.5 EGTA, 40 Hepes, 1 MgCl₂ (pH 7.2). With resting membrane potential held at −70 mV, step current pulses (25 pA increments, 100 ms) were delivered to principal MNTB neurons to elicit APs (see Fig. 5).

Figure 5.

A, sample cell-attached voltage-clamp (VC) recordings from a principal MNTB neuron in a NH mouse during 100–400 Hz stimulation trains. Onset of spike failure indicated by arrowhead. Line plots depict spike failure rates in NH (a; n = 21), SD (b; n = 14) and SE (c; n = 13) synapses during 100–400 Hz stimulation. B, sample cell-attached and whole-cell recording from a principal MNTB neuron in a NH mouse. Average steady-state EPSC amplitude calculated from the last 10 EPSCs recorded during trains. C, correlation between postsynaptic spike failures during 300 Hz trains and steady-state EPSC amplitude in NH (n = 36), SD (n = 14) and SE (n = 13) synapses. Double-exponential line fits are plotted; R values is the correlation coefficient. D, sample traces of whole-cell current-clamp recording of APs in response to step-current injections. Holding potential was −70 mV and step-currents delivered in 25 pA increments. E, average number of spikes recorded following step-current injections in NH (n = 11), SD (n = 8) and SE (n = 8) synapses. Boltzmann line fits are shown.

Data were acquired online, digitized at 50 kHz, filtered at 4 kHz and analysed offline using the pCLAMP 10.3 software (Molecular Devices), Excel 2010 (Microsoft, Redmond, WA, USA), SigmaPlot 10.0 (SYSTAT, Chicago, IL, USA) and MiniAnalysis (Synaptosoft, Decatur, GA, USA). EPSC amplitudes were measured and normalized to the first EPSC. Multiple traces acquired from each cell at each frequency were averaged. The rate at which EPSC amplitude depressed during stimulation trains was fitted with a single exponential functional to obtain the time constant. Paired-pulse ratios (PPRs) were calculated using the amplitudes of the 1st and 2nd EPSC during 300 Hz stimulus trains (EPSC2/EPSC1 × 100). The size of the readily releasable pool (RRP) of synaptic vesicles was estimated from 300 Hz trains based on methods outlined by Wesseling & Lo (2002) which accounts for the variable rate at which synaptic vesicles are recruited back into the releasable pool during a depleting train (Grande & Wang, 2011). To estimate the probability of synaptic vesicle release (P_r), we divided the amplitude of the first EPSC from the 300 Hz trains by the size of the RRP.

Postsynaptic spike failures during stimulation trains were identified by a clear absence of AP current. Spike failures were expressed as a percentage of the number of observed APs/the number of expected APs at the particular frequency × 100. Raster plots of postsynaptic cell-attached recordings were generated representing the occurrence of an AP. Stimulation artifacts from raw traces during stimulus trains were removed for clarity. Differences in excitability of principal MNTB cells were based on the number of APs occurring within a 100 ms as a function of step current amplitude. Statistical significance was assessed using one-way and two-way ANOVAs, Holm–Sidak post hoc tests and Kolmogorov–Smirnov (KS) tests with a P value cut-off of P < 0.05 (Systat, SigmaPlot 12). Data are expressed as the mean ± SEM (standard error of the mean) from a population of n synapses.

Results

Prior to performing morphological and electrophysiological experiments in brainstem slices, we first recorded click-evoked ABRs in mice aged P17–19 to confirm that removal of middle ear ossicles induced a unilateral CHL. ABRs in NH mice displayed the characteristic peaks (I–V) with detectable ABRs ranging from ∼5 to 20 dB SPL in both the right and left ears (Fig. 1A and B). ABRs in sham mice were also detectable in the same range as those in NH mice (Fig. 1B). When clicks were presented to the intact right ear of SURG mice, ABR thresholds were similar to those in NH mice. However, when presented to the impaired ear of SURG mice, ABR thresholds were higher compared to the right and left ears of NH mice and the intact ear of SURG mice, as reflected in the mean values (F_(3,42) = 85.9; P = 5.9⁻¹⁸; Fig. 1B). This significant difference reflected an approximate 50 dB shift in the average ABR threshold compared to normal hearing mice and is typical of a CHL (Larsen & Liberman, 2010; Qin et al. 2010). The summed sound-evoked activity in NH and SURG mice was also assessed by plotting the mean ABR amplitude versus intensity function (Fig. 1C). For clarity, we combined the data obtained from the right and left ears of NH and the sham mice, given the ABR thresholds were the same. We found that for both wave I and IV, the amplitudes increased monotonically as a function of sound intensity from 30 to 80 dB SPL. Two-way ANOVAs comparing the amplitude–intensity functions did not reveal significant interactions between the NH, sham and intact right ears for wave I (F_(10,148) = 0.34; P = 0.96) and wave IV (F_(10,136) = 0.62; P = 0.79). Because ABRs from the impaired ear of SURG mice could not be recorded similarly from 30 to 80 dB SPL, we made comparisons using one-way ANOVAs at 60 and 70 dB SPL. At 70 dB SPL, no differences were apparent for waves I and IV. However, at 60 dB, the amplitude of both waves had significantly decreased (wave I: F_(3,32) = 6.69; P = 0.001; wave IV: F_(3,27) = 6.02; P = 0.003) indicative of a much steeper rate of decline for ABR amplitudes recorded from the impaired ear (Fig. 1C). Differences were also apparent when ABR waveform amplitudes were averaged, normalized to peak I and overlaid (Fig. 1D). At 70 and 40 dB SPL, peaks I–V were clearly distinguishable for NH mice and the intact auditory pathway of SURG mice. However, for the impaired left ear of SURG mice, the ABR peaks were not discernible at 70 dB SPL and largely absent at 40 dB SPL. Moreover for intact auditory pathways, the averaged ABRs were ∼2.5-fold larger compared to NH mice at both 70 and 40 dB SPL. These findings confirm that unilateral removal of middle ear ossicles led to a CHL and asymmetries in the activation patterns of auditory brainstem neurons between the intact and impaired pathways.

Previous studies report that CD–1 mice experience early progressive hearing loss (Shone et al. 1991; Le Calvez et al. 1998), which may confound the interpretation of our results from the surgically induced CHL. To ensure that CD–1 mice have normal hearing within the age range used in this study (<3 weeks), we performed additional control experiments in older CD–1 mice (n = 5) aged 4–5 weeks. Following tone stimulation, we found that ABRs were detectable at all frequencies tested (8, 16 and 32 kHz) with thresholds ranging from 20 to 40 dB SPL (Supporting information Fig. S1). No differences were observed for the mean values across the frequencies tested (F_(2,27) = 2.1; P = 0.15). These findings are similar to the study by Li & Verkman (2001) which also found no evidence of high-frequency hearing loss in CD–1 mice at 4–5 weeks of age. Based on this we believe that the onset of high-frequency hearing loss in CD–1 mice due to an underlying genetic basis does not occur within the age range used in this study.

Unilateral CHL increases the morphological complexity of calyx terminals bilaterally

During the first 3 weeks of postnatal development, the calyx of Held undergoes well-described changes in morphology such that it initially resembles a spoon-shaped structure but subsequently transforms into its mature phenotype as a highly digitated and fenestrated structure after hearing onset (Kandler & Friauf, 1993; Wimmer et al. 2006; Ford et al. 2009). We previously showed that the structure of mature calyx terminals is highly heterogeneous within the MNTB and forms a structural continuum such that calyx morphologies range from simple to complex. Simple calyces are characterized mainly by digit-like (stalks) structures with few swellings whereas complex calyces have many swellings emanating from the stalks (Grande & Wang, 2011). Thus, we asked whether unilateral CHL alters (1) the developmental transition of calyx morphology from spoon to digit-like structure and (2) the structural continuum formed by the heterogeneous calyx morphologies. To confirm that unilateral removal of middle ear ossicles did not result in gross re-wiring of the calyx of Held–MNTB circuit as occurs when unilateral cochlear ablations are performed at early postnatal ages (Kitzes et al. 1995; Russell & Moore, 1995) we placed DiI crystals in the VCN ipsilateral to the manipulated left ear (Fig. 2A). DiI-labelled calyces were found exclusively in the contralateral MNTB (Fig. 2B) verifying that the unilateral CHL did not aberrantly innervate ipsilateral principal MNTB neurons.

Figure 2.

A, schematic diagram of auditory brainstem. Middle ear ossicle removal in left ear reduces the auditory input to the contralateral MNTB (red). We refer to this MNTB as the sound-deprived (SD) MNTB. The opposing MNTB (blue) we refer to as the sound-experienced (SE) MNTB. BDA injections were placed at the midline and DiI crystals were placed in the VCN ipsilateral to the surgical ear. B, low-power photomicrographs (merged) of DiI-labelled calyces innervating the appropriate contralateral MNTB (dashed region) in a mouse developing with unilateral CHL. C, sample 3–D renderings of calyces of Held (P17–19) taken from normal hearing (NH) mice and those located in the SD and SE MNTBs of surgical mice. Examples of swellings emanating from the stalks of calyx terminals are indicated. D–F, average surface area, volume and number of swellings from calyx terminals in NH mice (n = 45) and those located in the SD (n = 32) and SE (n = 28) MNTB. Individual raw data points are shown adjacent to bars. The corresponding cumulative distributions are shown below. *Significantly different from NH mice. **Significantly different from SE calyces. Scale bar in B, 500 μm; scale bar in C, 5 μm and applies to all.

We quantified calyx morphology by injecting BDA at the midline of transverse brainstem slices (Fig. 2A). These injections led to well-labelled calyces in both the SD and SE MNTB within the slice that could be reconstructed for quantitative analysis (Fig. 2C; Wimmer et al. 2006; Ford et al. 2009; Grande & Wang, 2011). For both NH and SURG mice, all calyces acquired their mature phenotype. Regardless of their location in the SD or SE MNTB, calyces were characterized by stalks and numerous swellings demonstrating that the developmental transition from spoon to digit-like structure is unaffected by hearing asymmetries caused by unilateral CHL.

Calyx terminals from SURG mice appeared to display heterogeneous morphologies comparable to NH mice in that some were more structurally complex than others in line with our previous study (Grande & Wang, 2011). This was well reflected in the wide range of surface areas and volumes measured (Fig. 2D and E). In NH mice, calyx surface area ranged from ∼800 to 2800 μm² and in the SD and SE MNTB the values ranged from ∼1000 to 2600 μm² and ∼1000 to 2300 μm², respectively. The mean surface area did not differ between any of the groups (NH 1660.8 ± 78.4 μm²; SD 1760.7 ± 62.9 μm²; SE 1619.5 ± 55.0 μm²; F_(2,102) = 0.91; P = 0.41) nor did their cumulative distribution profiles (Fig. 2D; KS test NH vs. SD P = 0.15; NH vs. SE P = 0.39; SD vs. SE P = 0.36). For calyx volume, the values ranged from ∼300 to 1500 μm³ in NH mice and in the SD and SE MNTB the values ranged from ∼300 to 1400 μm³ and ∼500 to 1200 μm³, respectively (Fig. 2E). Similar to calyx surface area, the mean calyx volume did not differ between any of the groups (NH 679.1 ± 40.9 μm³; SURG SD 752.7 ± 39.6 μm³; SE 633.8 ± 33.1 μm³; F_(2,102) = 1.96; P = 0.14) nor did the cumulative distribution profiles (KS test NH vs. SD P = 0.10; NH vs. SE P = 0.32; SD vs. SE P = 0.10). The number of swellings found on calyces, which is a measure of structural complexity, was significantly different between groups (F_(2,102) = 21.3; P = 1.8⁻⁸; Fig. 2F). Calyces in NH mice had an average of ∼13 swellings, which was significantly lower than the average number found on calyces in the SD (P = 1.9⁻⁸) and SE (P = 0.007) MNTB, which had an average of ∼26 and ∼19 swellings, respectively. Moreover, the higher incidence of swellings in SURG mice was more pronounced for SD versus SE calyces (P = 0.001). This bilateral increase in morphological complexity was well reflected in the rightward shifts of the cumulative distribution plots for swelling numbers found on SD and SE calyces versus NH calyces (KS test NH vs. SD P = 0.00002; NH vs. SE P = 0.009; SD vs. SE P = 0.03). These results suggest that asymmetries in hearing caused by unilateral CHL during development shifts the structural continuum within the MNTB such that calyx morphologies amongst the population are more complex.

Short-term plasticity is altered in SE but not SD synapses

Given the known functional changes at the calyx of Held–MNTB synapse during development (Borst & Soria van Hoeve, 2012), we examined whether synaptic responses are altered in mice developing with unilateral CHL. We performed whole-cell recordings in principal MNTB neurons during axonal stimulation of the presynaptic bundle to measure evoked EPSCs. Figure 3A shows sample EPSC recordings taken from NH mice and those from the SD and SE MNTB of SURG mice. In sham mice, EPSC amplitudes, rise times and decay times were not different compared to NH mice. On average, EPSC amplitudes in the left (n = 10) and right (n = 9) MNTB of sham mice were 9.0 ± 0.8 and 8.5 ± 0.5 nA, respectively, whereas in NH mice it was 9.5 ± 0.4 nA (F_(2,38) = 1.15; P = 0.32). EPSC rise times in the left and right MNTB of sham mice averaged 0.22 ± 0.004 and 0.22 ± 0.003 ms, respectively, whereas in NH mice it was 0.22 ± 0.002 ms (F_(2,38) = 1.03; P = 0.37). Similarly, decay times in the left and right MNTB of sham mice averaged 0.98 ± 0.04 and 1.12 ± 0.07 ms, respectively, whereas in NH mice it was 1.13 ± 0.04 ms (F_(2,38) = 2.27; P = 0.11). Based on these findings, and the fact that ABR thresholds did not differ between NH and sham mice, we pooled the data for the remainder of the electrophysiological analysis.

Figure 3.

A, sample EPSC traces recorded from principal MNTB neurons in NH and SURG mice (P17–19) during midline axon stimulation (whole-cell voltage clamp, holding potential −60 mV). Mean EPSC amplitude, rise time and decay constant compared for principal MNTB cells in NH mice (black) and those in the SD (red) and SE (blue) MNTB of SURG mice. Data from sham and NH mice was pooled (see text for explanation). Raw data points are indicated adjacent to bars. B, sample EPSCs recorded at different holding potentials (−60 to +60 mV, 20 mV increments). Peak amplitudes at various holding potentials were measured, normalized to capacitance, averaged and plotted for current–voltage (I–V) relationships. In A, * indicates different from NH. In B, * indicates difference between NH and SE, and ** indicates difference between NH and SD.

In SURG mice, we observed bilateral decreases in EPSC amplitude compared to NH synapses (F_(2,77) = 22.8; P = 1.7⁻⁸; Fig. 3A). EPSCs in SD (n = 21) and SE (n = 18) synapses, which were not significantly different (P = 0.2), had an average amplitude of 6.7 ± 0.43 and 5.9 ± 0.44 nA, respectively, compared to NH synapses (n = 41, 9.2 ± 0.36 nA). For EPSC rise time, we observed bilateral increases in SURG versus NH mice (F_(2,77) = 13.9; P = 7.3⁻⁶; Fig. 3A). Rise times in SD (0.236 ± 0.004 ms) and SE (0.24 ± 0.004 ms) were not significantly different (P = 0.48) but were slower compared to NH synapses (0.22 ± 0.002 ms). However, EPSC decay times did not differ across groups (F_(2,77) = 1.3; P = 0.27; Fig. 3A). To determine whether synaptic conductance was altered, we measured evoked EPSCs at various holding potentials and constructed I–V curves (Fig. 3B). When normalized to capacitance, a two-way ANOVA revealed reductions in synaptic conductance in SD and SE synapses versus NH synapses (F_(12,511) = 7.02; P = 0.0001) although no differences occurred between SD and SE across all holding potentials (F_(2,6) = 0.9; P = 0.38). This bilateral decrease between SURG and NH mice is most apparent at −60 mV (NH vs. SD, P = 0.0001; NH vs. SE, P = 0.002), −40 mV (NH vs. SD, P = 0.04; NH vs. SE, P = 0.003) and +60 mV (NH vs. SD, P = 0.01; NH vs. SE, P = 0.0004) holding potentials as determined by post hoc tests. Thus, similar to the changes in calyx morphology, we observed bilateral effects on function in SURG mice despite a unilateral CHL.

We previously showed that the form of short-term plasticity (STP) expressed at the calyx of Held–MNTB synapse is largely dictated by the morphology of calyx terminals independent of tonotopicity (Grande & Wang, 2011). Those with simple morphologies typically display strong short-term depression (STD) whereas those that are more complex display short-term facilitation (STF) followed by slow depression. Given our observation that a unilateral CHL led to a bilateral increase in the morphological complexity of calyx terminals, we predicted that both SD and SE synapses would predominantly display STF followed by slow depression. To test this we recorded evoked EPSCs during high-frequency train stimulation (100–300 Hz; Fig. 4A). At all frequencies tested, NH synapses (n = 41) displayed heterogeneous forms of STP similar to our previous observations (Grande & Wang, 2011). This is reflected in the variability of the line fits for EPSC amplitude (Fig. 4B) showing that in some cells there is an immediate and continuous decline after the onset of train stimulation (i.e. STD). In other cells, there is an initial short-term facilitation (STF) followed by a slow depression. The initial STF is apparent from the earliest time points whereby the line fits for the normalized EPSC amplitudes begin above the dashed line (Fig. 4B). Surprisingly, however, the heterogeneity in STP observed in SD synapses (n = 21) was similar to that of NH synapses. This is supported by the similarity in the variability of the EPSC amplitude line fits and their average (Fig. 4B) and the absence of significant differences found in the rates of synaptic depression following a two-way ANOVA post hoc test (at 100 Hz, 54.4 ± 3.5 versus 54.2 ± 2.4 ms, P = 0.79; at 200 Hz, 32.7 ± 2.4 versus 29.9 ± 1.1 ms, P = 0.44; at 300 Hz, 22.7 ± 1.5 versus 20.9 ± 0.7 ms, P = 0.56; Fig. 4C). In contrast, the average steady-state EPSC amplitude was higher in SE synapses (n = 16; Fig. 4B) and their rates of synaptic depression were slower (F_(2,225) = 199.7; P = 0.0001) at 100 and 200 Hz; Fig. 4C). At 100 Hz, the rate of synaptic depression in SE synapses was 68.6 ± 4.1 ms and statistically different from NH (P = 0.0001) and SD (P = 0.0001) synapses. At 200 Hz, the rate of synaptic depression was 38.9 ± 2.1 ms and different from NH synapses (P = 0.024) but at 300 Hz (26.1 ± 1.1 ms) no differences were observed compared to NH and SD synapses. This predominance of facilitation in SE versus NH and SD synapses was also reflected in the PPR (F_(2,75) = 9.2; P = 0.0003; Fig. 4D). Despite our observation that STP was only altered in SE synapses, there was a bilateral decrease in the RRP size in SURG mice (F_(2,75) = 8.5; P = 0.0005; Fig. 4E). On average, the RRP size in SD and SE synapses was 50.3 ± 4.0 and 54.9 ± 3.8 nA, respectively (P = 0.41), compared to that in NH mice which was 68.8 ± 2.9 (vs. SD, P = 0.0004; vs. SE P = 0.01). This was not paralleled by bilateral changes in P_r as we only observed a decrease in the average P_r for SE synapses (0.11 ± 0.006; F_(2,75) = 5.24; P = 0.007) relative to SD (0.15 ± 0.01; P = 0.003) and NH (0.14 ± 0.006; P = 0.003) synapses (Fig. 4F). Thus despite a bilateral increase in the morphological complexity of calyx terminals, only SE synapses had a corresponding functional shift in STP towards facilitating synapses.

Figure 4.

A, sample EPSC traces recorded during 200 Hz stimulation. B, exponential line fits to EPSC amplitudes during 100, 200 and 300 Hz trains recorded from NH, SD and SE synapses. Line fits are normalized to the initial EPSC amplitude (EPSC₁). Lowest plots display the averaged EPSC amplitude. C, average decay constant of EPSC depression in principal MNTB neurons calculated from the exponential fits obtained during 100, 200 and 300 Hz stimulation trains. Raw data points are adjacent to bars. D–F, paired-pulse ratios (PPRs), estimated readily releasable pool (RRP) size and probability of release (P_r) calculated for NH, SD and SE synapses. *Different from NH synapses (P < 0.01). **Different from SD synapses.

Heterogeneities in spiking fidelity are not altered in mice developing with unilateral CHL

During high-frequency stimulation, the calyx of Held–MNTB synapse displays heterogeneities in spiking fidelity (Hermann et al. 2007; Mc Laughlin et al. 2008; Englitz et al. 2009; Lorteije et al. 2009; Grande & Wang, 2011; Lorteije & Borst, 2011). The spiking heterogeneity between different MNTB neurons, which manifests as differences in postsynaptic spike failures, is uncovered during 300 Hz stimulation in standard slice conditions and is strongly correlated to the steady-state EPSC amplitude (Grande & Wang, 2011). Spike failures are most frequent in cells with small steady-state EPSC amplitudes and correspond to principal MNTB neurons innervated by calyces with simple morphologies that display strong STD. In contrast, few postsynaptic spike failures occur in principal MNTB neurons innervated by calyces with complex morphologies which display STF followed by slow depression. These cells typically display large steady-state EPSC amplitudes. To determine whether the relationship between spike failures and heterogeneities in STP and consequently steady-state EPSC amplitude is altered in mice developing with unilateral CHL, we performed cell-attached and whole-cell voltage-clamp recordings in principal MNTB neurons in SURG mice. Figure 5A shows a sample cell-attached recording of compound currents attributed to AP firing of a principal MNTB neuron during stimulation trains at various frequencies. In this example from a NH synapse, no spike failures were observed below 200 Hz (Fig. 5Aa), a small number were observed at 300 Hz (∼8%) but at 400 Hz the incidence increased substantially (∼40%). At 300 Hz, heterogeneities in spike failures became apparent, ranging from 0 to 45%, and at 400 Hz, ranging from 32 to 57%, replicating our previous findings (Grande & Wang, 2011). Surprisingly, the results were similar in SURG mice. For principal neurons in the SD and SE MNTB (Fig. 5Ab and c), no failures were observed at 100 and 200 Hz. At 300 Hz, the spike failure rate ranged from 0 to 52% in SD synapses and from 0 to 39% in SE synapses. At this stimulation frequency, no differences were found in the spike failure rate distributions (KS test NH vs. SD P = 0.12; NH vs. SE P = 0.07; SD vs. SE P = 0.47). At 400 Hz, the spike failure rate ranged from 30 to 60% in SD synapses and from 33 to 58% in the SE synapses. Similar to the spike failure rate at 300 Hz, no difference were found in the spike failure rate distributions at 400 Hz (KS test NH vs. SD P = 0.38; NH vs. SE P = 0.52; SD vs. SE P = 0.73). Thus, the range in spike failures typically observed in populations of calyx of Held–MNTB synapses during high frequency stimulation is unaffected in mice developing with aural asymmetries despite our observed bilateral changes in calyx morphology as well as the unilateral changes in STP for SE synapses.

Following cell-attached recordings, we switched to whole-cell mode in the same cell to determine the steady-state EPSC amplitude. Figure 5B shows sample cell-attached and whole-cell recordings from the same cell during 300 Hz trains displaying spike failures and a steady-state amplitude close to 1 nA. For both NH and SURG mice, we plotted the average steady-state amplitude versus spike failures. For NH synapses, the percentage of spike failures decreased exponentially as steady-state amplitude increased (Fig. 5C). This is expected given synaptic input ≥1200 pA rarely leads to many spike failures. For SE synapses (Fig. 5C), the results were similar although the exponential decrease in spike failures as steady-state EPSC amplitude increased occurred more slowly compared to NH synapses. However, for SD synapses (Fig. 5C), the relationship between spike failures and steady-state EPSCs shifted leftward. Despite these slight differences, in all cases the onset of failures gradually occurred at longer intervals and the number of failures decreased as the steady-state EPSC amplitude increased (Supporting information Fig. S2). Thus, the overall bilateral stability of the functional continuum is preserved in mice despite disruptions in the balance of hearing.

The leftward shift in the relationship between spike failures and steady-state EPSCs for SD synapses suggests an increased intrinsic excitability of principal neurons of the SD MNTB. We tested this possibility by delivering step current pulses in principal MNTB neurons of NH and SURG mice (Fig. 5D). In all cases, the number of spikes generated by principal MNTB neurons in response to injected current increased then reached a plateau indicating the maximum saturation point. For each cell, the average responses were then fitted with a Boltzmann function (Fig. 5E). We first calculated the half-maximum activation point and the maximum slope from the line fits and found no differences between NH, SE and SD synapses for both parameters. On average, the half-maximum activation point for SD and SE synapses was 178.3 ± 18.1 and 223.9 ± 26.5 nA, respectively, versus that found in NH synapses which was 171.5 ± 17.9 nA (F_(2,24) = 1.8; P = 0.18). The average maximum slope for SD, SE and NH synapses was 0.09 ± 0.01, 0.08 ± 0.01 and 0.07 ± 0.02 spikes nA^–1, respectively (F_(2,24) = 0.41; P = 0.66). However, as shown in Fig. 5E, the average maximum number of spikes was significantly higher in SD (17.5 ± 1.6) versus SE (10.9 ± 1.6) and NH (8.5 ± 1.4) synapses (F_(2,24) = 8.96; P = 0.001). This suggests that unilateral upregulation in the excitability of SD neurons compensates for the downregulation in synaptic drive (i.e. steady-state EPSCs) to maintain the stability of neurotransmission bilaterally.

Discussion

Little is known about the modifications in synaptic transmission that occur in auditory neurons when hearing symmetry is disrupted by moderate forms of unilateral hearing loss (with minimal perturbation to the cochlea) during development. Principal neurons of the MNTB are innervated by calyx of Held terminals that originate from the axons of GBCs located in the contralateral VCN (Borst & Soria van Hoeve, 2012). Because of this anatomical feature, our unilateral CHL approach allowed us to isolate and examine populations of synapses that were sound deprived (i.e. in the MNTB contralateral to the impaired ear) and those of the opposing MNTB where hearing experience remained intact. Our study shows that compared to NH mice, calyx of Held synapses located in both the SD and SE MNTB of SURG mice (1) are structurally more complex, (2) have smaller synaptic currents and RRP size, and (3) display an increase in excitability. Despite the observed changes in synaptic structure and function in mice developing with unilateral CHL, the continuum of spiking fidelity across the population of principal MNTB cells remained bilaterally symmetric.

In P17–19 mice that underwent middle ear ossicular removal 1 week before hearing onset, calyx terminals in both MNTBs had properly crossed the midline without ipsilateral sprouting and innervated contralateral MNTB neurons. These terminals acquired their mature morphological phenotype suggesting that the developmental transition from the immature spoon-shape to mature digitated structure was unaffected by unilateral CHL. At mature terminals, we did not observe any differences in surface area and volume of SD and SE calyces versus those in NH mice, which is similar to reports in the deafness (dn/dn) mouse (Youssoufian et al. 2008), a model of severe bilateral congenital deafness. However, when we quantified calyx morphology based on the number of swellings, a novel measure of structural complexity we previously employed to reveal heterogeneity amongst calyces (Grande & Wang, 2011), we found a much higher incidence of calyces with a larger number of swellings in mice developing with unilateral CHL. This increase in structural complexity occurred bilaterally and was more pronounced in SD versus SE calyces. Presently, it is unclear whether the morphological complexity of calyx terminals we observed (as defined by the number of swellings) also occurs in more severe forms of unilateral or bilateral deafness. In spite of this, our study shows that asymmetries in hearing experience caused by a moderate unilateral CHL elicit a shift in the structural continuum such that the population of calyx terminals within the MNTB are more complex compared to NH animals.

Given our previous findings demonstrating a strong correlation between calyx morphology, the form of STP expressed and incidence of postsynaptic spike failures (Grande & Wang, 2011), we had predicted that the bilateral shift in the structural continuum to more complex calyces would have resulted in a corresponding shift in the functional continuum such that more synapses would express STF followed by slow depression and thus more synapses would demonstrate few postsynaptic spike failures. Surprisingly, we found that the incidence of postsynaptic spike failures amongst the population of SD and SE synapses was not different from NH synapses during high-frequency trains. Thus, the functional continuum as characterized by the heterogeneities in spike failures was bilaterally preserved in unilaterally hearing mice. This occurred despite observing bilateral decreases in EPSC amplitude, synaptic conductance and the size of the RRP. Moreover, for the SE MNTB, we observed a unilateral decrease in P_r, increase in PPR and a shift in STP to a higher incidence of synapses expressing STF followed by slow depression. Previous studies show deafness has little effect on the synaptic properties at the calyx of Held–MNTB synapse but more striking changes at the upstream giant synapse, the endbulb of Held in young mice (i.e. ≤P12; Oleskevich et al. 2004; Youssoufian et al. 2005). In another study examining the synaptic characteristics at the calyx of Held–MNTB synapse in deaf mice (Ca_v1.3^−/−), Erazo-Fischer et al. (2007) report numerous changes including an increase in P_r and a higher incidence of STD. Although our results contrast with these findings, the Ca_v1.3^−/− and deafness mice are models of severe forms of bilateral congenital deafness which lack any form of sound-evoked cochlea-driven activity. Thus it is possible that different forms and degrees of severity of hearing loss elicit different modes of remodelling at auditory synapses. We postulate that in response to moderate imbalances in hearing symmetry caused by unilateral CHL during sensitive periods of development, different forms of synaptic remodelling in each MNTB may ensure that heterogeneities in spiking fidelity, and hence diversity of response properties, remain bilaterally stable. This idea needs to be tested in future experiments.

Heterogeneity of action potential failures

In our slice preparation, heterogeneity in principal MNTB spike failures emerges during high frequency stimulation. These heterogeneities, which are reflected in temporal onset and rate of AP failures during 200 ms axonal stimulation trains, become apparent at 300 Hz or at 400 Hz when in vivo-like conditions are mimicked (Grande & Wang, 2011). AP failures observed during in vitro experiments are probably associated with the physiological phenotypes of MNTB neurons as several in vivo studies demonstrate AP failures in principal MNTB neurons during acoustic stimulation. For example, during tone repetitions, Lorteije & Borst (2011) show that AP failures emerge at frequencies above 200 Hz stimulation and show large variability in spike failure rate at 400 Hz, in line with our previous in vivo-like experiments. In response to low-frequency sinusoidal acoustic stimulation, Kopp-Scheinpflug et al. (2008) show that principal MNTB neurons can reliably phase-lock. But as stimulation frequency increases, the units’ discharge skips stimulus cycles reducing the degree of entrainment and the ability to respond to each cycle. Importantly, however, APs are still phase-locked allowing principal cells to follow high-frequency acoustic input despite limits imposed by the refractory period. AP failures may be relevant for coding sound duration. In response to low-frequency acoustic stimulation at 600 Hz for 100 ms, principal MNTB cells generate APs faithfully during early cycles of the sine wave (Kopp-Scheinpflug et al. 2011) but the probability of generating APs in later cycles drastically declines, consistent with our observations in some cells (Fig. 5). These lines of evidence support the view that AP failures at the calyx of Held–MNTB synapse normally occur during acoustic processing.

The response patterns of principal MNTB cells to acoustic stimuli vary considerably. Some units display responses such as phase-locking (PhL), primary-like (PL) and primary-like with notch (PLn) based on peristimulus time histograms (PSTHs) which reflect spike discharge patterns during acoustic stimulation (Smith et al. 1998). PSTHs following acoustic stimulation describe two response components of principal MNTB cells: the onset and sustained components (Kopp-Scheinpflug et al. 2011). The onset component (i.e. the first cycle in the sine wave) is very reliable regardless of the response type (Smith et al. 1998) and provides accurate timing onset information to downstream nuclei (Grothe et al. 2010). The ability to generate APs with such high probability, which is evident in the first peak of PSTHs, can be explained by the magnitude of synaptic current generated at the calyx terminal. As many labs have shown, the synaptic current delivered to the principal MNTB cell by the calyx terminal upon initial stimulation is many times larger than needed to generate an AP. Our data show that this large safety factor occurs regardless of calyx morphology, forms of STP or other heterogeneities, and ensures that all principal MNTB cells deliver precise onset timing information. The sustained component is much more heterogeneous and is used to categorize cells as PhL, PL or PLn. During the sustained component the incidence of AP failures increases due to the lower probability of generating APs in later cycles compared to the first (Kopp-Scheinpflug et al. 2011). As a consequence, heterogeneities in AP failures diversify the response patterns of principal MNTB cells evident in PSTHs following prolonged acoustic stimulation. Incidentally, this sustained component is relevant for coding complex features of sound (Joris & Yin, 1998; Tollin & Yin, 2005; Tolnai et al. 2008). The variability of AP failures during sustained stimulation as seen in our present and previous study (Grande & Wang, 2011) will result in heterogeneity in the representation of sound in the spike trains of principal MNTB neurons. Such heterogeneity may contribute to the coding of complex features of sound such as harmonic structure, spectrotemporal fluctuations and rapid amplitude modulations (Skoe & Kraus, 2010).

Heterogeneity, synaptic remodelling and sensory experience

The functional continuum as characterized by heterogeneities in principal MNTB spiking was bilaterally maintained in mice developing with mild-to-moderate unilateral hearing loss. This occurred despite differential synaptic remodelling in the population of SD and SE synapses. For example, the presynaptic release parameters were relatively unchanged in SD synapses whereas in SE synapses we observed increases in the PPR and reductions in P_r probably contributing to the shift in STP to facilitating-type synapses (Fig. 4). With respect to postsynaptic changes, the most striking effect was the increase in excitability for principal cells located only in the SD MNTB (Fig. 5). This change seemed to stabilize the heterogeneities of the population SD principal MNTB cells such that they remained functionally similar to those of the SE MNTB with respect to spike failures. Increased excitability following sound deprivation also occurs in other regions of the auditory pathways such as in A1 cortical neurons. This is thought to reflect a feedback mechanism to maintain a normal operative state (Kotak et al. 2005).

The precise mechanisms that ensure that principle MNTB cells remain functionally diverse bilaterally despite unilateral hearing loss remain unknown. The answer may reside in the neurons positioned upstream of the MNTB, and bilateral cross-talk. For example it is well known that neurons of the cochlear nucleus also display significant functional and morphological heterogeneities (Rhode, 2008). Thus an important question to be answered is how the heterogeneity of GBCs relates to the heterogeneity of the calyx of Held–MNTB synapse. In other words, who are the upstream synaptic partners of complex versus simple structured calyces of Held? By determining this through future anatomical experiments, we can more precisely determine whether the heterogeneity observed at the calyx terminal arises entirely de novo at the level of the MNTB, at the level of the cochlear nucleus or perhaps even more upstream. This may help us understand the mechanisms that drive the diversity observed at the calyx of Held synapses and more importantly, the mechanisms by which diversity is maintained under conditions of altered sensory experience.

Following a period of atypical sensory experience during development, some sensory systems, unlike the visual system, have the capacity to recover. In the olfactory system for example, odour discrimination and related processing can be sustained owing to the regenerative capacity of olfactory neurons (Lledo & Saghatelyan, 2005), appropriate learning-related reinforcement (Kato et al. 2012; Wu et al. 2012) and the diversity of glomeruli (Padmanabhan & Urban, 2010; Angelo & Margrie, 2011). In a recent study performed in mice expressing a single odorant receptor, Angelo et al. (2012) demonstrated that this form of olfactory deprivation eliminated the diverse cellular features of mitral cells thus severely reducing their capacity to code odour-related information. This study strongly suggests that maintaining heterogeneities within an identified class of neurons after sensory experience is disrupted is critical for reducing redundancies and ensuring that the capacity to code sensory information remains maximal. In both adult animals and humans, binaural disruptions in spatial cues via unilateral CHL leads to poor sound localization performance in azimuth but through daily training tasks, performance can improve (Kacelnik et al. 2006; Van Wanrooij & Van Opstal, 2007; Kumpik et al. 2010; Irving & Moore, 2011; Irving et al. 2011). For animals developing with a unilateral CHL during infancy, their ability to locate sound during spatial hearing tasks can reach comparable levels to normal hearing controls (King et al. 2000). Moreover in deaf children, cochlear implants improve sound localization performance probably due to their ability to use binaural cues such as interaural level differences (Gordon et al. 2012). These studies strongly suggest that auditory brainstem circuits remain remarkably stable and ready to reinstate their functions in the event of disrupted hearing experience. Our findings lead us to conclude that asymmetric sensory experience can significantly impact morphological and functional remodelling at the synaptic level within auditory brainstem but the overall bilateral functional continuum even under adverse conditions is preserved.

Key points

• Severe forms of hearing loss during development are well known to profoundly alter synaptic structure and function through the auditory neuraxis; however, the pre-and postsynaptic modifications that occur during development following moderate forms of unilateral hearing loss remain unknown.

• We performed anatomical and electrophysiological experiments at the calyx of Held synapse in slices taken from mature mice developing with a unilateral conductive hearing loss and compared populations of synapses in the same animal that were either sound deprived or sound experienced.

• Compared to normal hearing mice, we found that calyx of Held synapses that were both sound deprived and sound experienced as a result of a unilateral conductive hearing loss were (1) structurally more complex, (2) had smaller synaptic currents and a readily releasable pool size, and (3) were more excitable.

• Despite these changes in structure and function, spiking fidelity within the populations maintained a continuum as in normal hearing mice such that heterogeneities remained bilaterally symmetric.

• Preservations in the heterogeneity in spiking fidelity via synaptic remodelling may ensure that a functional stability important for the ability to code sound localization cues is maintained despite asymmetries in hearing experience during development.

Glossary

ABR

auditory brainstem response

AP

action potential

BDA

biotinylated dextran amine

CHL

conductive hearing loss

GBC

globular bushy cell

LSO

lateral superior olive

MNTB

medial nucleus of the trapezoid body

MSO

medial superior olive

NH

normal hearing

P

postnatal day

PhL

phase-locking

PL

primary-like

PLn

primary-like with notch

PPR

paired-pulse ratio

P_r

probability of release

PSTH

peristimulus time histograms

RRP

readily releasable pool

SD

sound deprived

SE

sound experienced

SPL

sound pressure level

STD

short-term depression

STF

short-term facilitation

STP

short-term plasticity

SURG

surgical

VCN

ventral cochlear nucleus

Additional information

Competing interests

The authors declare no competing interests.

Author contributions

G.G. and L.-Y.W. conceived and designed the research. G.G. performed the electrophysiological and anatomical experiments. Experiments were performed in the labs of L.Y.W. and R.V.H. at the Hospital for Sick Children in Toronto, Canada. G.G. and J.N. performed the auditory experiments. G.G. performed all the data analysis and drafted the manuscript. G.G., J.N., R.V.H. and L.-Y.W. edited and revised the manuscript. All authors approved the final version.

Funding

This work was supported by individual operating grants from the Canadian Institutes of Health Research (CIHR; MOP-77610 and MOP-114990) and Canada Research Chair (to L.-Y.W.), a CIHR team grant (to R.V.H. and L.-Y.W.) as well as a Restracomp Postdoctoral Fellowship from the SickKids Research Institute (G.G.).

Supporting Information

Supplementary Material