Acute hyperbilirubinaemia induces presynaptic neurodegeneration at a central glutamatergic synapse

Authors


Corresponding author I. D. Forsythe: Neurotoxicity at the Synaptic Interface, MRC Toxicology Unit, University of Leicester, Leicester LE1 9HN, UK.  Email: idf@le.ac.uk

Abstract

There is a well-established link between hyperbilirubinaemia and hearing loss in paediatrics, but the cellular mechanisms have not been elucidated. Here we used the Gunn rat model of hyperbilirubinaemia to investigate bilirubin-induced hearing loss. In vivo auditory brainstem responses revealed that Gunn rats have severe auditory deficits within 18 h of exposure to high bilirubin levels. Using an in vitro preparation of the auditory brainstem from these rats, extracellular multi-electrode array recording from the medial nucleus of the trapezoid body (MNTB) showed longer latency and decreased amplitude of evoked field potentials following bilirubin exposure, suggestive of transmission failure at this synaptic relay. Whole-cell patch-clamp recordings confirmed that the electrophysiological properties of the postsynaptic MNTB neurons were unaffected by bilirubin, with no change in action potential waveforms or current–voltage relationships. However, stimulation of the trapezoid body was unable to elicit large calyceal EPSCs in MNTB neurons of hyperbilirubinaemic rats, indicative of damage at a presynaptic site. Multi-photon imaging of anterograde-labelled calyceal projections revealed axonal staining and presynaptic profiles around MNTB principal neuron somata. Following induction of hyperbilirubinaemia the giant synapses were largely destroyed. Electron microscopy confirmed loss of presynaptic calyceal terminals and supported the electrophysiological evidence for healthy postsynaptic neurons. MNTB neurons express high levels of neuronal nitric oxide synthase (nNOS). Nitric oxide has been implicated in mechanisms of bilirubin toxicity elsewhere in the brain, and antagonism of nNOS by 7-nitroindazole protected hearing during bilirubin exposure. We conclude that bilirubin-induced deafness is caused by degeneration of excitatory synaptic terminals in the auditory brainstem.

Abbreviations 
ABR

auditory brainstem response

aCSF

artificial cerebrospinal fluid

EM

electron microscopy

MEA

multi-electrode array

MNTB

medial nucleus of the trapezoid body

7-NI

7-nitroindazole

sulfa

sulfadimethoxine

nNOS

neuronal nitric oxide synthase

NO

nitric oxide

Introduction

Jaundice or hyperbilirubinaemia, as the name implies, is the result of elevated serum levels of bilirubin, a degradation product of haemoglobin. Bilirubin is conjugated in the liver so that this more water-soluble form can be excreted in the bile. Neonatal jaundice can be caused by failure to conjugate free bilirubin (due to enzyme deficiency in the liver), a high load following erythrocyte damage (haemolysis) or high turnover leading to bilirubin accumulation. Immaturity and compromised liver function means that jaundice is common in newborn babies but causes little lasting harm if monitored and treated rapidly; phototherapy is often sufficient to lower the concentration of free bilirubin. Severe cases of neonatal jaundice (particularly in premature infants) are associated with deafness (Gerrard, 1952; Rhee et al. 1999) and development of cognitive problems such as kernicterus (Shapiro et al. 2006) with symptoms including ataxia and epilepsy. Bilirubin causes the degeneration of Purkinje neurons in the cerebellum (Schutta & Johnson, 1969) and basal ganglia (Perlman et al. 1997) consistent with ataxia and a central site of action (Shapiro et al. 2006).

Our objective was to test the hypothesis that bilirubin contributes to deafness by causing neurodegeneration in the central auditory pathway, by examining synaptic transmission and neuronal function following bilirubin exposure. We have focused on the calyx of Held excitatory synapse with its target neuron in the medial nucleus of the trapezoid body (MNTB) (Schneggenburger & Forsythe, 2006). This glutamatergic relay synapse is part of the brainstem auditory pathway involved in sound localisation (Johnston et al. 2010) and both the synapse and its target have been well characterised.

We have used the Gunn rat model of hyperbilirubinaemia (Gunn, 1938; Uziel et al. 1983; Shapiro & Nakamura, 2001); this spontaneous mutation was isolated from a Wistar colony and first recognised by their yellow coat colour by Charles Gunn. Homozygous (jj) Gunn rats are distinguished from their heterozygous litter mates (Jj) by this yellow discolouration. Homozygous Gunn rats lack the liver enzyme uridine-diphosphoglucuronyl-tranferase, thus rendering them unable to efficiently conjugate and excrete bilirubin. This results in jaundice within 48 h after birth and the rats stay jaundiced for the rest of their lives. Hyperbilirubinaemia is induced by administration of a sulfonamide, which displaces bilirubin from serum albumin binding sites, raising free bilirubin levels sufficiently to allow entry into the brain (Schutta & Johnson, 1969; Rose & Wisniewski, 1979), producing hearing loss (Shapiro, 1988) and ataxia in homozygous Gunn rats. The site of action of bilirubin in the auditory system is unknown. A peripheral site of action is unlikely, since cochlea hair cells were unaffected (Uziel et al. 1983; Rhee et al. 1999; Shaia et al. 2005). Anatomical post-mortem studies of humans suffering from severe neonatal jaundice report no pathological lesions in either the cochlea or dorsal root ganglia but extensive destruction of neurons in the cochlear nucleus (Dublin, 1951). Sound-evoked auditory brainstem responses (ABRs) are commonly used to test for hearing loss and hyperbilirubinaemia reduces ABRs in human neonates (Chisin et al. 1979) and homozygous Gunn rats (Shapiro, 1988). Functional hearing tests in homozygous Gunn rats suggest that the failure of the ABR is associated with a reduced size of the cochlear nuclei and the MNTB (Conlee & Shapiro, 1991) but the underlying cellular defect has not been identified.

Our results show reduced ABR amplitudes indicating hearing loss after 18 h hyperbilirubinaemia. Intra- and extracellular electrophysiological recordings demonstrate that this was accompanied by failure of synaptic transmission at the calyx of Held/MNTB synapse. Multi-photon imaging and electron microscopy reveal substantial degeneration of the presynaptic terminal, while the postsynaptic MNTB neurons remain largely unaffected, confirming a presynaptic site of action for bilirubin damage.

Methods

Ethical approval

All experiments were performed at the MRC Toxicology Unit in Leicester (UK), in accordance with the UK Animals (Scientific Procedures) Act 1986 and under the approval of the local ethical committee of the University of Leicester. All experiments reported in this study comply with the policies and regulations of The Journal of Physiology (Drummond, 2009).

A total of 73 Wistar and Gunn rats (Harlan UK, strain HsdBlu:GUNN-UDPGTj) aged from postnatal day 14 to 20 (P14–20) were used because blood bilirubin levels naturally peak in homozygous Gunn rats at between 2 and 3 weeks of life (Schutta & Johnson, 1969). Animals were bred in-house and homozygous (jj) Gunn rats were obtained by mating jj male with heterozygous (Jj) female Gunn rats (Takagishi & Yamamura, 1994).

Induction of hyperbilirubinaemia

Separate studies from Wistar, Jj or jj Gunn rats showed similar electrophysiological properties under whole-cell patch and multi-electrode array (MEA) recording and were pooled for later experiments as the control group (control). The intraperitoneal (i.p.) administration of the long-lasting sulfonamide sulfadimethoxine (sulfa; Rose & Wisniewski, 1979; Shapiro, 1988) (200 mg kg−1; Sigma) was used to trigger hyperbilirubinaemia in jj-Gunn rats. The experiments were conducted 18 h after this treatment (treated group). Prior to the animals being killed and 18 h following sulfa treatment, 77% of jj-Gunn rats showed motor deficits or ataxia. All treated jj-Gunn rats that had motor deficits also showed compromised synaptic transmission or degenerated/absent calyces. Wild-type Wistar rats that received identical 18 h sulfa treatment were asymptomatic, showing no motor deficits, ataxia, no compromised ABRs (Fig. 4C) or impaired synaptic transmission in the MNTB. Similar results from Jj-Gunn rats have been reported (Rose & Wisniewski, 1979; Shapiro, 1988).

Figure 4.

Protection from bilirubin-induced hearing loss by the nNOS inhibitor, 7-nitroindazole (7-NI)
A, ABRs from a jj-Gunn rat in response to a 94 dB ‘click’ stimulus before (black) and after (grey) induction of hyperbilirubinaemia (18 h after sulfadimethoxine, 200 mg kg−1). B, ABRs from a jj-Gunn rat in response to a 94 dB ‘click’ stimulus before (black) and after (grey) induction of hyperbilirubinaemia (18 h after sulfadimethoxine, 200 mg kg−1, and 7-NI, 150 mg kg−1). C, summary plot of ABR thresholds in response to a ‘click’ stimulus. Wild-type Wistar rats and the jj-Gunn rats have similar auditory thresholds under control conditions. The Wistar rats suffered no auditory deficit after 18 h sulfadimethoxine, whereas jj-Gunn rats show significantly elevated thresholds after sulfa treatment. 7-NI protected the hearing from this threshold elevation. Box plots show mean ± upper and lower threshold range; ANOVA, *P < 0.05. The number of animals tested is indicated in the respective box.

Auditory brainstem responses (ABRs)

Homozygous Gunn rats and Wistar rats were anaesthetised with an i.p. injection of Hypnorm/midazolam (2.7 ml kg−1, VetaPharma Ltd) and recordings were made before (control) and 18 h after (treated) administration of sulfadimethoxine (200 mg kg−1, Sigma) using a Medelec Sapphire 2A amplifier. Data were sampled at 16 kHz and stored on a personal computer. ABRs were evoked using pure-tone (12, 24, 30 kHz) and ‘click’ stimuli delivered via a reverse-driven microphone (B&K 4134) over a range of intensities (24–94 dB) using a digital attenuator (Tucker Davies Technologies, USA). For some experiments an nNOS antagonist, 7-nitroindazole (7-NI, Sigma) was suspended in peanut oil and administered i.p. at a concentration of 150 mg kg−1, 30 min prior to the i.p. injection of sulfadimethoxine.

Patch clamp

Brain slices were prepared from animals killed by decapitation. Whole-cell patch recordings were made from visually identified rat MNTB neurons in acute brain slices (200 μm thick) of the auditory brainstem as described previously (Steinert et al. 2008). Experiments were performed at 36 ± 1°C using a feedback-controlled Peltier device (manufactured by University of Leicester Mechanical and Electrical Workshop). Synaptic stimulation (using a DS2A isolated stimulator (Digitimer, Welwyn Garden City, UK; 1–10 V, 0.1–0.2 ms) was delivered via a bipolar platinum electrode placed at the midline across the trapezoid body nerve fibres. Patch pipettes were pulled from glass capillaries (GC150F-7.5, o.d. 1.5 mm, Harvard Apparatus, Edenbridge, UK) and had resistances of 2–5 MΩ when filled with the pipette solution (see below). Series resistances were between 4 and 9 MΩ (series resistance compensation and prediction were around 70%). Data were recorded using a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA). Stimulation, data acquisition and analysis were performed using pCLAMP 9 and Clampfit 10.2 (Molecular Devices). Average data are presented as mean ±s.e.m.

Calcium imaging

Acute brainstem slices (200 μm) were prepared as for patch clamp experiments. Slices were loaded with 5 μm of the acetoxy methyl ester form of Fura-2 (Fura-2 AM; Molecular Probes, Eugene, OR, USA, dissolved in DMSO with 5% pluronic acid) for 10 min, then maintained for 30 min in artificial cerebrospinal fluid (aCSF) for de-esterification of the AM-dye. Fura-2 fluorescence (excitation at 340 and 380 nm) was recorded and Metafluor imaging software (Series 7, Molecular Devices) was used to display the fluorescence images (emission > 505 nm).

Extracellular multi-electrode array

Acute brainstem slices (300 μm thick) were prepared as described above. One slice was placed on top of a multi-electrode array dish (MED-P515A, 150 μm electrode spacing) and placed in a recording chamber (MED64, Alpha MED Scientific Inc., Japan), which was adapted for perfusion and Peltier-driven temperature control (manufactured by University of Leicester Mechanical and Electrical Workshop). Stimulation with a biphasic pulse (±100 μA, 160 μs) to one or two electrodes at the midline was used to evoke field potentials in the superior olivary complex. Evoked MNTB field potentials (Haustein et al. 2008) were recorded from between 4 and 10 underlying electrodes. Data were recorded at a sampling rate of 20 kHz with 12 bit resolution. Extracellular recordings were performed at 31 ± 1°C using Mobius software (Alpha MED Scientific Inc.) for stimulation, recording and data analysis.

Multi-photon live imaging

Calyces in the MNTB were anterogradely labelled by injection of dextran–tetramethyl-rhodamine (MW 3000; Invitrogen) in vitro into the ventral cochlear nucleus using methods adapted from Burger et al. (2005). Dextran–rhodamine (5% in sterile PBS) was pressure-injected and electroporation via a bipolar tungsten electrode connected to a BTX ECM 830 electroporator (Harvard Apparatus; 1 × 120 V pulse for 130 ms, then 60 × 50 ms 50 V pulses) was employed to induce uptake of the dye into the bushy cells that give rise to the calyces of Held. Slices were then kept in the dark at room temperature in a maintenance chamber for at least 4 h before imaging. Images and z-stacks from the living slice were taken using a Zeiss LSM510 scanning microscope equipped with a Mai-Tai Deep-See multi-photon laser (Spectra-Physics). Images were analysed and the surface area of calyces was calculated from z-stacks using Volocity 5.3 software (Improvision, UK). The calyces were visually identified by their shape. Great care was applied to only measure the calyces individually and to avoid adjacent or overlapping structures, such as neighbouring calyces or axons.

Electron microscopy

Rats were anaesthetised with isofluorane (3–4%) and then perfusion-fixed with 2% glutaraldehyde + 2% paraformaldehyde in 0.1 m sodium cacodylate buffer (final pH 7.4). Slices (500 μm thick) were prepared using a vibrating blade microtome (Leica Microsystems, Milton Keynes, UK). These slices were post-fixed in 1% osmium tetroxide + 1% potassium ferrocyanide, stained en bloc with 5% uranyl acetate and embedded in epoxy resin (TAAB Laboratories Equipment Ltd, Aldermaston, UK). Semi-thin (1 μm) sections were stained with toluidine blue and examined to select areas for ultramicrotomy. Ultrathin sections (70 nm) were stained with lead citrate and examined in a JEOL 100-CXII electron microscope (JEOL (UK) Ltd, Welwyn Garden City, UK) equipped with a ‘Megaview III’ digital camera (Olympus Soft Imaging Solutions GmbH, Münster, Germany). Analysis of electron microscopy (EM) images was done double-blinded using ImageJ 1.42 (NIH) or Axiovision 4.8 software (Zeiss). Presynaptic profiles were identified as structures enclosed by a membrane in contact with the cell body, showing a postsynaptic density and containing synaptic vesicles. Neurons containing the principal cell nucleus were selected for this analysis.

Solutions

An artificial cerebrospinal fluid (aCSF) was used for slice incubation, maintenance after slicing and perfusion during recordings (in mm): NaCl 125, KCl 2.5, NaHCO3 26, glucose 10, NaH2PO4 1.25, sodium pyruvate 2, myo-inositol 3, CaCl2 2, MgCl2 1, ascorbic acid 0.5; pH was 7.4 when gassed with 95% O2, 5% CO2. Osmolarity was around 310 mosmol l−1. A low-sodium aCSF was used during preparation of slices, with a composition as above for aCSF, except that NaCl was replaced by sucrose 250 mm, and CaCl2 and MgCl2 were changed to 0.1 mm and 4 mm, respectively. The pipette solution for whole-cell recordings contained (in mm): potassium gluconate 97.5; KCl 32.5, Hepes 10, EGTA 5, MgCl2 1; pH was adjusted to pH 7.3 with KOH and osmolarity to 290 mosmol l−1 with sucrose.

Numbers and statistics

The number (n) for statistical analysis depended on the experimental design (i.e. animals used in vivo; the number of cells where definable or the number of independent electrodes for MEA recordings). A summary of the animals used is as follows:

  • 1ABR. 8 jj-Gunn rats (control), 7 rats (sulfa-treated), 2 rats (saline), 4 Wistar rats (sulfa), 7 jj-Gunn rats (7-NI); each animal served as its own control (ABR only).
  • 2MEA. 52 electrodes/6 slices/3 rats (control), 55 electrodes/9 slices/4 rats (treated).
  • 3Synaptic stimulation and voltage clamp. 15 cells/7 rats (control), 9 cells/5 rats (treated).
  • 4Voltage clamp. 9 cells/3 rats (control), 18 cells/3 rats (treated).
  • 5Current clamp. 26 cells/10 rats (control), 18 cells/7 rats (treated).
  • 6Multi-photon imaging. For box-plot summary: 20 calyces/3 rats (control), 13 calyceal remnants/2 rats (treated).
  • 7Electron microscopy. 24 cells/3 rats (control), 18 cells/2 rats (treated).

The specific statistical test is noted in the text or the respective figure legend, which was performed using Sigma Plot 11 (Systat Software Inc.). Statistical significance was accepted if P < 0.05.

Results

Control untreated Gunn rats aged P16–P18 showed normal in vivo ABR responses, but following 18 h of hyperbilirubinaemia (Fig. 1A) the characteristic ABR waveforms were severely disrupted, consistent with damage to the central auditory system (Shapiro, 1988). Wave III of the ABR, which is thought to reflect activity in the superior olivary complex (Melcher et al. 1996) was significantly reduced in amplitude in all treated jj-Gunn rats (n= 7, Fig. 1A and B). Two animals which received saline instead of sulfadimethoxine showed no change in ABR response (data not shown). Extracellular MEA recordings from brainstem slices in vitro (Fig. 1C) showed a characteristic presynaptic conducted waveform on stimulation of the trapezoid body (C1) and a postsynaptic component (C2) (Haustein et al. 2008). In treated rats the C2 component was absent or vastly reduced (control −186.4 ± 18.7 μV, n= 52; vs. treated −32.1 ± 3.2 μV, n= 55; P < 0.001, Mann–Whitney-Rank-Sum test) as shown by the C2/C1 ratio for both conditions (Fig. 1D; P < 0.001). The MEA data also revealed an increased synaptic delay, from 0.46 ± 0.01 ms in control (n= 52) rising to 1.12 ± 0.03 ms in animals with hyperbilirubinaemia (treated, n= 55, P < 0.001, Mann–Whitney-Rank-Sum test). In order to confirm the link between the ABR and MEA data, we made brain slices from three treated jj-Gunn rats. Each showed reduced ABRs in vivo, and the postsynaptic C2 component in vitro was either absent or reduced in amplitude and exhibited increased synaptic delay. Clearly compromised ABRs in hyperbilirubinaemia reflected impaired synaptic transmission in the MNTB.

Figure 1.

Jaundice causes hearing loss through failure of transmission in the auditory brainstem
A, in vivo auditory brainstem responses (ABRs) exhibit well-defined waves (I–IV) in response to 30 kHz pure-tone at 94 dB (black trace, control, n= 8); hyperbilirubinaemia in the jj-Gunn rat (grey trace, treated, n= 7) caused severe suppression of waves I–III within 18 h. B, bar graph shows that the mean amplitudes of ABR waves I–III are significantly reduced after 18 h hyperbilirubinaemia. t test, *P < 0.05, **P < 0.01, ***P < 0.001. C, extracellular field potentials from the MNTB in vitro show a presynaptic (C1) and postsynaptic (C2) component evoked by stimulation of the trapezoid axons (control, black trace); C1 is unaffected while C2 is delayed and suppressed by hyperbilirubinaemia (grey trace; s, stimulation artefact). D, bar graph shows mean reduction of C2 amplitude in treated animals. All values are mean ±s.e.m.; ANOVA, ***P < 0.001.

We performed whole-cell patch recording from MNTB principal neurons in the same in vitro preparation. Treated jj-Gunn rats showed no large amplitude evoked EPSCs, with mean amplitudes of 1.5 ± 0.5 nA (n= 9) compared with control EPSCs of 9.6 ± 1.2 nA (n= 9, at −60 mV). At higher stimulus intensities (>6 V, Fig. 2A) non-calyceal EPSCs were observed (Hamann et al. 2003) in treated Gunn rats. The input/output plot showed control calyceal EPSCs as low threshold and large magnitude events (Fig. 2A) while responses from treated animals were of high threshold and low magnitude (Fig. 2A). No difference was observed in excitability or action potential waveform between MNTB neurons from control and treated animals (Fig. 2B) nor were inward sodium currents (control 13 ± 1 nA, n= 9; vs. treated 12 ± 1 nA, n= 18; at −100 mV) or outward potassium currents affected (Fig. 2C). These results show that 18 h elevated bilirubin causes a rapid failure of synaptic transmission without affecting postsynaptic excitability.

Figure 2.

Patch-clamp recording reveals that principal neurons exhibit normal excitability, but the giant calyceal EPSC is absent following hyperbilirubinaemia
A, under voltage clamp MNTB neurons normally receive a large calyceal EPSC (black trace) but only small slow EPSCs are present following hyperbilirubinaemia (grey trace). A plot of EPSC amplitude against threshold stimulus intensity shows that control calyceal EPSCs (black diamonds) were low threshold and large amplitude (grey, highlighted area), whereas the EPSCs from treated animals (grey squares) were small amplitude and high threshold. B, postsynaptic MNTB neuron excitability was unchanged by hyperbilirubinaemia, as indicated by the similar action potential (AP) waveforms: although AP halfwidth was similar, there was a small increase in AP amplitude in treated animals. C, control (black) and treated (grey) principal neurons show near identical current–voltage relationships. Insets show example current traces from one cell, with voltage commands plotted below. All values are mean ±s.e.m.; Mann–Whitney-Rank-Sum test; n.s., not significant; *P < 0.05.

The electrophysiological data indicate a presynaptic failure of synaptic transmission, so we performed imaging experiments at the light and EM level to test for presynaptic degeneration. The presynaptic axons and calyces of Held were anterogradely labelled with dextran–rhodamine and live imaging performed using multi-photon microscopy, in vitro. Control rats showed well-labelled calyceal structures in the MNTB (Fig. 3A) but 18 h after induction of hyperbilirubinaemia no healthy calyces were observed (Fig. 3B). The surface area (2334 ± 301 μm2, n= 20) of healthy calyces was consistent with previous reports (Sätzler et al. 2002), while calyceal remnants from treated animals (1117 ± 318 μm2, n= 13, P > 0.001; Mann–Whitney-Rank-Sum test) were significantly smaller and there were considerable numbers of isolated dye-containing structures, indicating calyceal breakdown (Fig. 3B, inset). These remnants were excluded from this analysis, and so the above mean areas underestimate the total degree of synaptic breakdown. Intensely stained axons were present in both conditions (Fig. 3A and B inset), consistent with intact axons and functional dye transport.

Figure 3.

Hyperbilirubinaemia causes degeneration of the calyceal presynaptic terminal
A, multi-photon imaging of live calyces (labelled with dextran–rhodamine, MW 3000) from a control rat (P14, jj-Gunn, untreated) showing axons and 3 calyceal synaptic structures. Inset: comparison of the calyx surface areas between control and treated animals shows a significantly reduced surface area in the degenerating calyces. Mann–Whitney-Rank-Sum test, ***P < 0.001. B, 18 h following hyperbilirubinaemia, no healthy calyces were observed, but considerable labelled debris was present; inset: an example of a rare calyx remnant with marked degenerative morphology. Scale bars in A, B and inset: 20 μm. C, electron microscopy confirms the presence of the calyceal presynaptic terminal, showing a control MNTB neuron, surrounded by presynaptic terminal profiles (yellow). Mitochondria, ER and myelinated axon profiles are clearly visible. D, hyperbilirubinaemia reduced the number of presynaptic profiles (yellow), but the principal neuron and internal organelles appeared otherwise normal. E and F, MNTB neuron perimeter was unchanged, but the length and number of presynaptic profiles were significantly reduced by hyperbilirubinaemia, confirming the hypothesis of presynaptic degeneration. t test; n.s., not significant; **P < 0.01. Scale bars in C and D: 2 μm.

Electron-micrographs of the MNTB confirmed the presynaptic pathology; controls showed the target principal neuron surrounded by multiple large presynaptic calyceal profiles containing vesicles and mitochondria (yellow overlay, Fig. 3C). In treated animals the MNTB neuron looked healthy, but closer examination showed reduction in the number of presynaptic profiles (control 11 ± 1, n= 24; vs. treated 7 ± 1, n= 18; P > 0.006; Fig. 3D and F) and their length of apposition to the target neuron (control 23 ± 2 μm, vs. treated 14 ± 2 μm; P > 0.006; Fig. 3E). Principal neuron perimeters were unaffected by bilirubin treatment (control 68 ± 2 μm, vs. treated 74 ± 3 μm; not significant Fig. 3E). Mitochondria and endoplasmic reticulum (ER) showed no pathology, consistent with minimal postsynaptic excitotoxicity. An uncompromised, healthy postsynaptic neuron was also confirmed by Ca2+ imaging which showed no elevation of resting [Ca2+]i in MNTB neurons between control (110 ± 7 nm, n= 71) and treated (111 ± 10 nm, n= 46) Gunn rats.

Why is this synapse so susceptible to bilirubin toxicity? Recent reports (Brito et al. 2010) have implicated nitric oxide (NO) in bilirubin toxicity. Previously we have shown that the MNTB expresses high levels of neuronal nitric oxide synthase (nNOS) (Steinert et al. 2008), so we reasoned that an nNOS antagonist, 7-nitroindazole (7-NI) might protect hearing in this Gunn rat model. Administration of 7-NI during bilirubin exposure provided significant protection from loss of auditory function (Fig. 4). This was quantified by using the threshold to a ‘click’ sound stimulus. Wistar and Gunn rats had similar thresholds of around 40 dB under control conditions but 18 h after induction of hyperbilirubinaemia the jj-Gunn rats’ thresholds were raised to 80 dB (Fig. 4C); this threshold elevation was substantially prevented when the jj-Gunn rats received the nNOS antagonist 7-NI. These results suggest that the mechanism of bilirubin toxicity involves pathological changes in the nitrergic signalling of the MNTB.

Discussion

ABR measurements reveal substantial hearing loss in Gunn rats within 18 h of exposure to high bilirubin. Extracellular multi-electrode array recordings showed impaired synaptic transmission through the MNTB in vitro. Whole-cell patch-clamp recordings from MNTB neurons in hyperbilirubinaemic rats confirmed that their electrophysiological properties were essentially unchanged from control animals. However, stimulation of the trapezoid body was unable to elicit large amplitude calyceal EPSCs in MNTB neurons of hyperbilirubinaemic Gunn rats. Multi-photon imaging of anterogradely labelled calyceal terminals revealed dramatic degeneration of the presynaptic calyx, supporting a neurodegenerative mechanism. Electron microscopy confirmed the loss of presynaptic terminals and healthy postsynaptic neurons. The protection from hearing loss by an nNOS antagonist suggests involvement of nitric oxide signalling in this presynaptic toxicity. We conclude that degeneration underlies the synaptic failure in the MNTB, resulting in reduced ABR amplitudes in an acute model of hyperbilirubinaemia.

Following 18 h hyperbilirubinaemia (induced by sulfa administration) we observed decreased amplitudes or complete absence of ABR waves I–III. These ABR data are in accordance with findings in homozygous Gunn rats by Shapiro (1988) where declining amplitude and increased waveform latency occur between 1 and 8 h after induction of hyperbilirubinaemia. Acute in vitro slices obtained from animals which had undergone in vivo ABR measurements confirmed a common defect. Field potentials measured from treated Gunn rats showed prolonged synaptic delays and decreased amplitudes of the postsynaptic field potential (C2). The MNTB contributes to the generation of waveform III in ABR recordings (Melcher et al. 1996); we can therefore directly link the impaired synaptic transmission measured with MEAs to the reduced waveform III amplitude indicative of hearing loss. The MEA C2 synaptic latency matched that of in vivo recordings (0.46 ± 0.12 ms) (Kopp-Scheinpflug et al. 2003). Interestingly, the prolonged synaptic delay measured here in treated jj-Gunn rats (1.12 ± 0.03 ms) is similar to that reported for the endbulb of Held/ventral cochlear nucleus of treated jj-Gunn rats (1.15 ± 0.23 ms Zhang et al. 1989), consistent with an early presynaptic compromise causing increased synaptic delay. MEA data did not provide sufficient information on the source of transmission failure in the MNTB so we performed patch-clamp recordings on single MNTB neurons to address this question. Whole-cell patch-clamp recordings from treated MNTB neurons were unaffected by bilirubin and the postsynaptic neuron did not degenerate in this acute model of hyperbilirubinaemia. This argument was strengthened by measurement of resting [Ca2+]i levels which were within normal ranges. However, low-threshold, large amplitude calyceal EPSCs could not be evoked in any of the MNTB neurons from treated Gunn rats. In contrast, neurons from control rats commonly showed calyceal EPSCs. Bilirubin has no direct effect on postsynaptic ionotropic glutamate receptors (Warr et al. 2000), so together these data suggest a presynaptic rather than postsynaptic site of failure in synaptic transmission.

Multi-photon imaging of labelled calyces of Held in acute brain slices from treated rats revealed degenerated or completely absent calyces. Axons were clearly labelled in both conditions indicating functional transport of dextran–tetramethyl-rhodamine from the bushy cell bodies of the anteroventral cochlear nucleus to the calyces of Held. While calyces in control animals had the same morphology as previously described (Forsythe, 1994), treated animals showed little intact calyceal structure and a large amount of stained debris, where calyces would have been expected. These remnants are probably parts of disintegrated calyces. While other authors have described a decrease in cell size of MNTB neurons (Conlee & Shapiro, 1991) after 4–5 days of sulfa treatment, and a reduction in the expression of the calcium-binding protein parvalbumin (Spencer et al. 2002) after 3 days of sulfa treatment, no previous studies have investigated or commented on the structural integrity of the presynaptic terminals. Shaia and colleagues reported degenerative changes in spiral ganglion neurons of jj-Gunn rats after 3 days of sulfa treatment (Shaia et al. 2005) although cochlear hair cells were not affected.

There is EM evidence from sulfa-treated Gunn rats that both acute lesions as well as chronic degeneration occurs in the cerebellar Purkinje cells in this model of hyperbilirubinaemia (Rose & Wisniewski, 1979). At 18 h the acute effects of bilirubin observed in this study caused no measurable change in MNTB neurones. Darkening of the remaining MNTB presynaptic profiles in treated Gunn rats matches observations in the cochlear nuclei of Gunn rats (Jew & Williams, 1977; Jew & Sandquist, 1979), although other hallmarks of neurodegeneration such as distorted mitochondria, accumulation of mitochondrial glycogen and abnormally increased extracellular space were not observed in the MNTB. In contrast, postsynaptic disease mechanisms seem more prevalent in the cerebellum of jj-Gunn rats which has long been known to be affected in hyperbilirubinaemia (Schutta & Johnson, 1969; Rose & Wisniewski, 1979). The sensitivity of the excitatory synapses in the auditory brainstem to bilirubin (Shapiro & Nakamura, 2001) probably explains the association of neonatal jaundice with deafness, and suggests that other cognitive and motor deficits associated with jaundice have similar underlying mechanisms, e.g. in the basal ganglia, cerebellum and hippocampus (Shapiro et al. 2006).

Human neonates suffering from bilirubin encephalopathy/kernicterus also show light microscopic evidence of bilirubin-induced neurodegeneration with lesions in the cochlear nuclei, with pyknosis and disintegration (Gerrard, 1952). The cochleae, 8th nerve and spiral ganglion neurons did not show abnormalities. In an autopsy study (Dublin, 1951) reduced numbers of fibres in the trapezoid body were reported, implying fewer synapses, consistent with the present data from the Gunn rat model. In a recent autopsy case study, Perlman and colleagues also report necrosis in the hippocampus (Perlman et al. 1997) following kernicterus. These studies emphasise the particular vulnerability of the auditory system and basal ganglia and highlight the need for more research in animal models (such as the Gunn rat) to unravel the underlying causes of bilirubin neurotoxicity and mitigate the long-term detrimental effects of bilirubin-induced neurodegeneration.

Previous reports have suggested that injection of albumin causes some reversal of bilirubin toxicity, presumably by binding free bilirubin (Shapiro, 1993). In terms of an underlying mechanism, it has been suggested that antagonists for NMDA receptors (McDonald et al. 1998) or nNOS (Brito et al. 2010) can provide protection from bilirubin toxicity. The MNTB has high levels of nNOS (Steinert et al. 2008) and we show here that an nNOS antagonist provides substantial protection of hearing from bilirubin-induced toxicity. Further work is required to identify the mechanism by which bilirubin causes NO production, and to establish the downstream actions of nitric oxide and mechanisms of toxicity (Steinert et al. 2010). We conclude that hyperbilirubinaemia causes degeneration of excitatory synaptic terminals in the auditory brainstem and this is associated with activation of neuronal nitric oxide synthase. Further investigation is required to test the therapeutic and wider implications of this result to other areas of the nervous system.

Appendix

Author contributions

M.D.H. conducted the experiments and analysis. D.J.R. provided expert assistance and helped with analysis of multi-photon imaging and EM data. J.R.S. conducted the calcium-imaging experiments. N.P. performed additional ABR recordings. D.D. conducted the EM imaging. I.D.F. developed the experimental design, interpreted data and supervised the project. M.D.H. and I.D.F. wrote the manuscript. All authors provided discussion, comments and approved the manuscript.

Acknowledgements

We would like to thank Judy McWilliam and Tim Smith for their help in preparing samples for electron microscopy, Robert Fern for advice on CNS pathology and Brian Robertson for drawing our attention to this fascinating problem. This work was funded by the MRC and by a PhD Scholarship to M.D.H. from Deafness Research UK.

Ancillary