Introduction

Many features of cochlear activity are directly regulated through innervation from the olivocochlear system, which comprises the medial olivocochlear component (MOC) originating from medial nuclei of the superior olivary complex and the lateral olivocochlear component (LOC) originating from the lateral superior olive (Guinan 2011). The MOC system projects onto outer hair cells (OHCs; see, for review, Wersinger and Fuchs 2011) and the LOC component projects onto primary auditory neurons dendrites beneath the inner hair cells (IHCs). Whereas MOC terminals likely use acetylcholine as a primary neurotransmitter and to some extent γ−amino-butyric acid (GABA), LOC efferent terminals express dopamine (DA), acetylcholine (ACh), γ−amino-butyric acid (GABA), enkephalins, dynorphins and calcitonin gene-related peptide (CGRP) (Eybalin 1993; Puel 1995; Sewell 2011). During the two last decades, significant advances in the understanding of the olivo-cochlear efferent function have been made in developmental, normal and pathological conditions, demonstrating the substantial contribution of the olivo-cochlear efferents to auditory physiology.

Efferent inhibition and ascending auditory pathway mapping

In rodents, the efferent innervation of IHCs occurs after birth and persists until the end of the second postnatal week (Katz et al. 2004; Roux et al. 2011). During this stage, the vast majority of efferent terminals are found on IHCs as well as on afferent fibers contacting IHCs (Pujol et al. 1978). Patch-clamp recordings have unambiguously demonstrated the cholinergic nature of the neurotransmitters released by the efferent fibers onto the IHCs (Glowatzki and Fuchs 2000). ACh secretion from the efferent terminals activates the α9α10 nicotinic receptors of the IHCs. Calcium influx through the α9α10 receptor opens, in turn, the small-conductance Ca2+ activated K+ channels (SK2), thereby hyperpolarizing the developing IHCs (Glowatzki and Fuchs 2000; Marcotti et al. 2004; Kong et al. 2008; Katz et al. 2011; Wersinger and Fuchs 2011; Johnson et al. 2013a). Recent work has highlighted the molecular machinery involved in ACh release by the efferent terminals. ACh secretion depends on the P/Q and N type calcium channels and is negatively regulated by the tight coupling between the L-type calcium channel and the big-conductance Ca2+-activated K+ current (BK type; Zorrilla de San Martín et al. 2010). Interestingly, GABA is also released by the efferents and inhibits the ACh secretion evoked by P/Q-type calcium channels through GABAB autoreceptors signaling (Wedemeyer et al. 2013). Finally, the retrograde signaling by nitric oxide (NO), which is produced by the hair cells in a calcium-dependent manner, results in the enhancement of ACh secretion (Kong et al. 2013). Taken together, the cholinergic input can be regulated in an autonomous and dependent fashion.

Developing hair cells fire action potentials in a burst-like fashion (Johnson et al. 2011; Sendin et al. 2014; Fig. 1a, b) with each action potential triggering glutamate exocytosis onto the afferent fibers (Beutner and Moser 2001; Marcotti et al. 2003). This bursting activity has been suggested to consolidate and refine the tonotopic map along the ascending auditory pathways (Tritsch et al. 2010). The cholinergic efferent feedback inhibits the spiking pattern of the developing IHCs to coordinate peripheral and central development (Glowatzki and Fuchs 2000; Johnson et al. 2011; Sendin et al. 2014; Fig. 1c, d). This hypothesis, however, has been called into question because efferent fibers are severed during the experiments in vitro. Indeed, the loss of the α9 or α10 subunits in the mouse does not change auditory thresholds (Vetter et al. 1999, 2007; May et al. 2002), although the α9 subunit or the SK2 channel are required for the proper maturation of synaptic machinery in IHCs (Johnson et al. 2007, 2013a, b). However, mice lacking the α9 subunit show aberrant wiring of auditory pathway circuitry, i.e., impairment of the medial nucleus of the trapezoid body–lateral superior olive connections (Clause et al. 2014). These results demonstrate, therefore, that the cholinergic system in the developing cochlea contributes to the organization of the ascending structures. At the onset of hearing, cholinergic control of the IHCs ceases and the medial efferent neurons project onto the mature OHCs (for review, see Simmons 2002). The α9α10 receptors can be directly modulated by opioids (Lioudyno et al. 2002) expressed in both LOC and MOC efferents (Eybalin 1993). Because the cholinergic inhibition mechanisms are highly similar between developing IHCs and adult OHCs, it has been proposed that the efferent fibers, which transiently connect to the immature IHCs, belong to the medial olivocochlear neuron pool (Simmons 2002; Simmons et al. 2011).

Fig. 1
figure 1

Spontaneous action potentials and inhibitory post-synaptic potentials in developing inner hair cell. a Patch-clamp recording at physiological temperature (top) of the spontaneous spiking activity from an apical P6 inner hair cell in current-clamp configuration (bottom). Bursts of action potential (AP, stars) and inhibitory post-synaptic potentials (IPSPs, arrowheads) populate the activity of the developing hair cell. b Train of AP at higher temporal resolution, corresponding to the green star in (a). c Train of IPSPs, corresponding to the red arrowhead in (a). d Higher temporal resolution of a single IPSP, corresponding to the black arrowhead in (c). Please note that each IPSP consists in a small depolarization, elicited by the α9α10 activation, followed by the membrane hyperpolarization, triggered by the SK2 potassium current activation

Full size image

Efferents control cochlear amplification in the adult cochlea

Akin to developing IHCs, cholinergic release from MOC terminals inhibits adult OHCs through the interplay of α9α10 nicotinic receptors and SK2 or BK potassium channels in OHCs located in apical or basal regions of the cochlea, respectively (Fuchs and Murrow 1992a; Oliver et al. 2000; Wersinger et al. 2010). In addition, a calcium-induced calcium release (CICR) mechanism has been suggested to contribute to the fast inhibitory effect of the acetylcholine input. One such CICR mechanism has also been identified in the developing IHCs and adult OHCs, involving the near-membrane postsynaptic cistern facing the efferent terminals, as calcium store (Lioudyno et al. 2004; Fuchs 2014). In addition, GABAB metabotropic receptors may act on type II afferents to regulate OHC activity through reciprocal synapses (Thiers et al. 2002, 2008; Maison et al. 2009). It is thus tempting to propose that the medial efferent olivocochlear bundle controls cochlear amplification through hyperpolarization of OHCs (Murugasu and Russell 1996). Indeed, MOC efferents constitute a sound-evoked negative feedback loop (Guinan 2011). This feedback suppresses the normal contribution of OHCs to sound (Wiederhold and Kiang 1970). Therefore, it has been proposed that MOC efferent feedback provides a powerful means to improve selective attention (Oatman and Anderson 1977; Puel et al. 1988; Scharf et al. 1994), signal detection in a noisy environment (Winslow and Sachs 1987), or ear protection from acoustic injury (Rajan 1988; Maison and Liberman 2000). This latter view is now well established though the use of olivo-cochlear bundle activation, de-efferentation or its genetic manipulation: (1) electrical stimulation of the efferent pathway reduces temporary hearing loss from simultaneous acoustic overexposure (Rajan 1988; Reiter and Liberman 1995), (2) surgical de-efferented animals are more vulnerable to noise exposure (Kujawa and Liberman 1997) and (3) overexpression of α9 subunit reduced noise-induced acoustic injury (Maison et al. 2002) as well as point-mutated α9 subunit, which prolongs cochlear inhibition (Taranda et al. 2009). Beyond reducing noise trauma protection, MOC removal exacerbates the loss of IHC ribbon synapse induced by moderate sound exposure (Maison et al. 2013) or associated with aging (Liberman et al. 2014). These results suggest that fine control of OHC activity is mandatory for the preservation of the synaptic structure of the cochlea.

Control of auditory nerve fiber excitability in the adult cochlea

Most studies from which LOC function has been inferred rely on olivocochlear bundle lesions. In this framework, completely de-efferented cochleas showed very little change in threshold levels but a very large reduction in the spontaneous discharge rate of the auditory nerve fibers (Liberman 1990; Zheng et al. 1999). However, these studies rely on a section of the entire olivocochlear bundle, including both the LOC and MOC (medial olivo-cochlear) efferents, making interpretation difficult (Fig. 2a). Stereotaxic injection of the neurotoxin melittin has been used as an alternative method to achieve selective damage of lateral superior olive (LSO) neurons (Le Prell et al. 2003). The unilateral destruction of the LSO did not affect thresholds but rather increased the amplitude of the ipsilateral neural response together with a reduction of the contralateral neural response amplitude (Darrow et al. 2006). Taken together, this study suggests that the LOC bundle may be involved in binaural balance to achieve spatial sound localization, although protection against sound trauma has also been proposed (Darrow et al. 2007). However, the chemical ablation of the LSO i) may not result in a complete loss of the efferent neurons, ii) may provoke MOC loss in the case of neurotoxin spread (Liberman et al. 2014) and iii) does not discriminate between efferents that may employ different neurotransmitters (Fig. 2a and b).

Fig. 2
figure 2

Modulation of the synaptic transmission by acetylcholine, dopamine and GABA. a Schematic representation of afferent and efferent olivocochlear innervation of the organ of Corti in adult cochlea (modified from Puel 1995). Inner hair cells are innervated by type I afferent auditory fibers, which approximate about 90–95 % of the total afferent fiber population of the auditory nerve. Outer hair cells (OHCs) are innervated by the 5 % remaining afferent fibers of the auditory nerve (type II). Both type I and type II project to the cochlear nuclei. Two types of efferents innervate the cochlea: the efferents projecting underneath the inner hair cells (IHCs) and those connecting directly the OHCs. These efferents originate from different areas in the superior olivary complex in the brainstem and run through the vestibular nerve. The efferents connecting the afferent type I fibers originate in the lateral superior olive (LSO). The OHCs efferent neurons are located in the ventral nuclei of the trapezoid body (VTB). Consequently, most of the authors use the terms lateral and medial efferent systems to designate the efferents below the IHC and those of the OHCs. The lateral efferents, which represent about 50~65 % of the olivocochlear bundle fibers, are unmyelinated and project towards the ipsilateral cochlea. The medial efferents are myelinated and reach the OHCs via the crossed and uncrossed components of the olivocochlear bundle. The crossed component predominates and forms about 70–75 % of the medial efferent innervation. b–b” Cryostat sections through the second turn of a guinea pig cochlea showing immunoreactivities to the vesicular acetylcholine transporter (b, VAChT, green), glutamic acid decarboxylase (b’, GAD, green) and tyrosine hydroxylase (b”, TH, green) in lateral efferent fibers. For VACht and GAD immunostaining, IHC is outlined with dashed lines. For TH immunostaining, IHC is labelled with anti-calbindin antibody (blue). IHC Inner hair cell; ISB inner spiral bundle. Scale bar 5 μm. c, d Spontaneous activity of a single afferent fiber modulated by LOC neurotransmitters. c Perfusion with artificial perilymph (AP) containing 100 μM ACh + 10 μM strychnine (α9α10 antagonist) increases the spiking rate of the auditory nerve fiber. Subsequent perfusion of AP containing 1 mM dopamine (DA) has the opposite effect as decreasing the discharge rate. d Perfusion with artificial perilymph (AP) containing 1 mM GABA transiently reduces the discharge rate of the afferent fibers. Spiking rate was fitted using sigmoidal (c, red line) or transient exponential recovery models (d, red line)

Full size image

To circumvent these issues, pharmacological manipulation of the LOC efferent neurotransmitters has been used, notwithstanding the difficulty of cochlear perfusion in vivo. Perilymphatic perfusion of dopamine (DA) has been shown to reduce the neural response, i.e., the compound action potential (CAP) that reflects the synchronous activity of auditory nerve fibers to sound stimulation (Ruel et al. 2001). Single-unit recordings show, in addition, that exogenous DA applied in the cochlea reduces spontaneous firing of action potentials (Fig. 2c) and increases the threshold of the auditory nerve fibers (Ruel et al. 2001; Garrett et al. 2011). Consistent with this effect, the selective DA transporter (DAT) inhibitors abolish the spontaneous and sound-evoked activity of auditory nerve fibers as they increase the endogenous level of extracellular DA (Ruel et al. 2006). To know whether DA is released in a tonic fashion, the effect of dopamine receptor antagonists on spontaneous activity was examined. In this case, auditory nerve fibers increased their spontaneous firing with a reduction of the auditory threshold (Ruel et al. 2001; Garrett et al. 2011). Altogether, these data suggest that a tonic release of dopamine operates in the cochlea, which can therefore control the behavior of the afferent fibers. Because fibers with different spiking rates and auditory thresholds populate the auditory nerve, it is conceivable that the tonic inhibition dictates, or at least modulates, the phenotypes of these fibers. In addition, morphological observation shows that a fraction of the IHC–auditory fiber synapses are swollen after DA antagonists perfusion (Ruel et al. 2001). The pool of auditory fibers showing swollen terminals after the DA inhibition removal may correspond with the high spontaneous rate fibers. This may explain why DA antagonists increase the firing rate of the lowest spontaneous rate fibers but completely suppress high spontaneous rate fibers (Ruel et al. 2001). Dopamine release may adjust the activity of the auditory fibers and prevent the excitotoxic side effect of glutamate at the first auditory synapse. Accordingly, dopamine receptor (D4 and D5) knock-out mice have been shown to be more susceptible to noise trauma than wild-type (Maison et al. 2012).

The inhibitory control of afferent fibers activity is not restricted to DA release by the LOC terminals. For example, micro-iontophoretic application of GABA reduces the afferent spiking rate increase evoked by glutamate (Arnold et al. 1998). Also, single unit recordings show a clear reduction of the spontaneous rate following GABA perfusion into the cochlear fluids (Fig. 2d). Because the GABA-induced reduction does not persist throughout the time course of application, i.e., the response elicited by GABA shows a strong desensitization, it contradicts a tonic effect of GABA in the cochlea. It should also be stressed that deletion of GABA receptors sub-unit β3 leads to a loss of afferent innervation (Maison et al. 2006), suggesting a role for GABA in synaptic maintenance.

In addition to inhibitory input, auditory nerve fibers undergo excitatory modulation. Indeed, excitatory action of LOC efferents is supported by the fact that micro-iontophoretic injections of ACh in the IHC synaptic area increase the spontaneous and glutamate-induced firing activity of radial afferent fibers in a dose-dependent manner (Felix and Ehrenberger 1992; Arnold et al. 1998). However, the very low spontaneous rate firing of fibers recorded in the vicinity of the afferent synapses suggested that micro-iontophoresis may have damaged fibers. To probe the cholinergic component of the LOC, ACh was applied by intracochlear perfusion. In order to block the effect of ACh on the MOC efferents, ACh was perfused together with strychnine, a potent blocker of the α9α10 receptors (Fuchs and Murrow 1992b; Kujawa et al. 1994; Elgoyhen et al. 1994, 2001). ACh/strychnine perfusion increased the spontaneous and sound driven activities of single auditory nerve fibers, suggesting an excitatory role for the cholinergic LOC efferents (Fig. 2c).

In addition to classical neurotransmitters, dynorphin may modulate the neural output of the cochlea as the CAP amplitude is depressed following diffusion of the kappa-opioid receptor agonist (−) pentazocine across the round window membrane (Le Prell et al. 2014). Finally, CGRP application increases the spike rate and decreases the mechanically driven response of primary afferents in the lateral line (Bailey and Sewell 2000a, b). Conversely, targeted gene deletion of CGRP in mice decreased suprathreshold neural responses, without the corresponding changes in DPOAE amplitudes that would be expected if OHCs were affected (Maison et al. 2003).

Altogether, these results show that LOCs can cause opposing effects on afferent fiber activity. One working hypothesis is that the LOCs adjust the sensitivity of the auditory nerve fibers according to the level of environmental sound. Thus, the ability of the auditory nerve to encode sound level in a noisy environment could be driven by LOC neurotransmitter release onto the afferent fibers, as when, for example, DA modifies auditory thresholds (Ruel et al. 2001).

Pathophysiological consequences

After environmental injury (excitotoxic injury, noise trauma, ototoxic drugs) or during aging, synaptic rearrangements occur in the cochlea. In all these cases, afferent synaptic contacts are lost and efferent neurons appear to form synapses with IHCs, at least transiently. Major challenges remain in determining the mechanisms that drive these changes and to understand what functional consequences they may have.

The IHC-afferent synapse is vulnerable to glutamate excitotoxicity and acoustic trauma, with potentially adverse consequences for long-term spiral ganglion neuron survival. In our hands, the excitotoxicity caused by intracochlear perfusion of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) induced a massive swelling of radial dendrites, which resulted in a disruption of membranes and a loss of cytoplasmic content (Puel et al. 1991, 1994). Structural rearrangements were clearly seen 1 day after excitotoxic injury (Fig. 3a, b). The basal pole of the IHC was surrounded by neurites, which appeared relatively normal at low magnification. However, upon closer examination, thin filopodia appeared to make elongated contact with the IHC membrane (covering up to 20–30 μm), with multiple dense thickenings, mimicking what has been described early in development (Pujol et al. 1978). At higher magnification, both afferent boutons (auditory nerve dendrites) and efferents could be recognized in direct contact with the IHC membrane (Fig. 3b). Five days after excitotoxic exposure, the pattern of IHC innervation seemed, at least qualitatively, almost normal, correlating with the full recovery of the compound action potential (CAP) (Puel et al. 1995; Ladrech et al. 2003; Fig. 3c). Normal synaptic differentiation between IHCs and auditory nerve endings was more frequently observed, together with efferents connecting to the afferent dendrites. Finally, the number of spiral ganglion neurons was not significantly different in AMPA-treated cochleas (24,293 ± 976) versus the non-treated contralateral cochleas (23,598 ± 852) when counted 1 month after exposure (Puel 1995).

Fig. 3
figure 3

Synaptic re-arrangement after AMPA-induced excitotoxicity. a The compound action potential (CAP), which reflects the synchronous activation of the auditory fibers, is completely abolished following intra-cochlear 200 μM AMPA perfusion and recovers during the next days (top). Schematic representation of the innervation changes during excitotoxic injury in the cochlea. Swelling of the afferent terminals caused by AMPA application accounts for the loss of CAP. During the repair of the afferent fibers terminals, efferent terminals connect in a transient fashion the hair cells (bottom). b–b” Electron microcopy examination shows a synaptic body (arrow) facing a neurite terminal (a) and the efferent (e) making axo-dentritic contact in normal condition (b, scale bar 0.5 μm). Ten-minute perfusion of AMPA provoked a massive swelling of radial dendrites, which ends in a disruption of membranes and a loss of cytoplasmic content (b’, scale bar 1 μm). One day later, efferents contact the hair cell (b”, scale bar 1 μm). Efferents show a very high density of vesicles, sometimes with postsynaptic cisterns in the hair cell (red arrowheads). c CAP amplitude (N1–P1) plots against the intensity of the sound-stimulation (probed at 8 kHz, 80 dB SPL). Note that the CAP amplitude recovers completely within 5 days after the excitotoxic injury, according to (c). (a) and (c) are adapted from Ladrech et al. (2003)

Full size image

Efferents are therefore able to synapse onto the hair cell during excitotoxic injury until the auditory nerve fiber terminals repair. The nature of the efferent fibers remains undetermined (LOC vs. MOC), although the loss of the afferent dendrite, as target of the LOC, may drive the LOC terminals to contact the nearby IHC. It is not known whether this efferent projection to the presynaptic hair cell is functional or whether this efferent connection is involved in afferent neuron survival. Worthy of note is that noise exposure also shows a change in innervation patterns (i.e., efferent neurons reinnervate the IHCs) during synaptic repair (Puel et al. 1998). Similar features of innervation have also been observed with aminoglycoside intoxication (Lenoir et al. 1999; Ladrech and Lenoir 2002). It might be conjectured that the projection of efferent neurons onto the hair cells favors the activity-dependent stabilization of afferent neurons by regulating hair cell excitability.

If efferent re-innervation somehow sustains the hair cell to enable spiral ganglion neuron survival, the death of hair cells would be expected to lead to spiral ganglion death. However, it has been found that spiral ganglion neurons can survive in the absence of IHCs (Zilberstein et al. 2012). In this latter case, however, it should be noted that efferent neurons still contact the auditory nerve fibers so that they may provide an input required for afferent fiber survival (Zilberstein et al. 2012). Finally, it has been found that efferent neurons re-innervate inner hair cells during the aging process (Lauer et al. 2012), perhaps to slow down the degenerative loss of the auditory nerve fibers due to multiple injuries throughout the life span.

Conclusions

The activity of the cochlea is under the regulation of efferent terminals that originate from the brainstem. During the two last decades, new roles for the efferent systems have emerged. Beside their contribution to the circuitry organization and activity, the olivo-cochlear bundle protects the cochlear structures. MOC innervation is required for auditory pathway development and to prevent the loss of synaptic structure in IHCs via the inhibition of cochlear amplifier in noise- and age-related hearing loss. In parallel, LOC innervation governs the excitability of the auditory nerve fibers to prevent glutamate-induced excitotoxicity and to promote synaptic repair. Others have suggested that LOC efferents are involved in binaural balance to achieve spatial sound localization. Based on the resistance of low-spontaneous rate fibers to masking noise (Costalupes et al. 1984), the larger number of efferent synapses connected onto the auditory nerve fibers of the modiolar side of the IHCs (i.e., the lowest spontaneous rate fibers) may also explain the better discrimination in background noise. In other words, LOC efferent may adjust the sensitivity of the fibers to improve the signal -to -noise ratio. Finally, Yin et al. (2014) demonstrated a role for the efferents in maintaining functional heterogeneity in the cochlear nerve. Although a vast amount of data on functional roles for these two efferent synapses in the regulation of cochlear development, maintenance and pathophysiology has been identified, we are still far from a clear and comprehensive portrait. A better understanding of cochlear efferent function will not only increase our knowledge of sound coding but also improve our understanding of the cellular changes that take place after cochlear trauma.