Talking back: development of the olivocochlear efferent system

Olivocochlear efferents develop together with typical motor neurons

The earliest stages of IEE development are intertwined with those of FBMNs. In all vertebrate species examined thus far—including mice, chickens, and zebrafish—IEEs and FBMNs originate together in the progenitor domain of visceral motor neurons (pMNv) in the ventricular zone, just medial to the floor plate of r4^{47,48,51,52,71,82}. This NKX2-2+ domain, analogous to p3 in the spinal cord, produces branchial and visceral motor neurons, while somatic motor neurons are generated from a more dorsal PAX6+ progenitor domain^83,84. The overall identity of r4 is determined by the homeobox transcription factor HOXB1,whose expression is restricted to r4 by E8.5 in mice^{57,59,61,85–88}. Loss of Hoxb1 has major consequences for r4, including loss of both IEEs and the facial motor nucleus, where FBMN cell bodies reside^89–91. However, ISL1-positive motor neurons are still present, suggesting that Hoxb1 is not necessary for motor neuron development per se, but instead directs a subset of motor neuron precursors to assume “r4” identities^71,89.

IEEs, like FBMNs and other branchial motor neurons, also express a collection of transcription factors that together assign a motor neuron identity. These genes include Isl1^92,93, Phox2a and Phox2b^94–96, Nkx genes (most critically Nkx2-2, Nkx2-9, Nkx6-1, and Nkx6-2)^{59,62,97–100}, and Tbx20^101–103. Expression of these transcription factors is restricted to motor neuron progenitors as part of the broader SHH signalling network, with pMNv forming just ventral to the pMN domain, which produces somatic motor neurons^63,83.

Motor neuron progenitors are subsequently diversified to produce subtypes of motor neurons by integrating patterning information from the anterior-posterior and dorsal-ventral axes. For IEEs and FBMNs, this means assuming a branchiovisceral fate within r4. The branchiovisceral fate is controlled by the combined action of NKX2-2 and NKX2-9 in progenitors within the pMNv domain¹⁰⁰. A key effector is the homeodomain protein PHOX2B, which promotes expression of Ascl1 and production of branchial and visceral motor neurons while inhibiting an alternative serotonergic fate^{95,96,104–107}. Thus, Phox2b knock-out animals lack all branchial and visceral cranial motor nuclei¹⁰⁴. In these animals, NKX2-2+ progenitors are specified normally but progress atypically to the next stage of development¹⁰⁴. This general PHOX2B-driven MNv program is customized in r4 through the combined action of HOXB1 and NKX family members. HOXB1 is required for Phox2b expression⁹⁰ and can directly bind to regulatory sequences in the Phox2b locus that are necessary and sufficient for expression in ventral r4, where the IEEs and FBMNs develop¹⁰⁸. Moreover, NKX2-2 synergizes with HOXB1 and other co-factors, such that expression of Phox2b is a direct readout of the intersection between the A/P (i.e. HOXB1) and D/V (i.e. NKX2) patterning systems. NKX6 proteins reinforce this system by maintaining expression of Hoxb1 in r4⁹⁵. Hence, in the absence of NKX6 protein activity, Hoxb1 expression decreases prematurely and very few IEEs are produced⁹⁵. As a result of this complex spatiotemporal network, MNv progenitors in r4 produce only branchial and visceral MNs, including both IEEs and FBMNs.

IEEs also depend on unique genetic programs to differentiate them from the FBMNs developing simultaneously in pMNv. In chicks and mammals—but not zebrafish^109,110—IEEs are distinguished from FBMNs early on by the expression of zinc-finger transcription factors GATA2 and GATA3^111–114. Although both GATA2 and GATA3 are expressed in the pMNv progenitor domain, analysis of Gata3 mutants indicates that GATA3 is not expressed by postmitotic FBMNs^96,111,112. In mice lacking either GATA2 or GATA3, efferent neurons project aberrantly and fail to cross the floorplate, with axons instead joining the VII^th nerve^111,112,114. In r4, Gata3 expression is downstream of Gata2^111,114. Thus, a GATA2/3 network may drive a binary IEE versus FBMN fate decision, similar to the role of GATA factors in the immune system¹¹⁵. This network, in turn, is downstream of both Hoxb1 and Phox2b. Thus, misexpression of Hoxb1 in r2 induces ectopic expression of both Gata2 and Gata3, as well as aberrant midline crossings, suggesting an induction of IEEs^86,111. PHOX2B may serve as an intermediary for this effect, as HOXB1 can induce Phox2b¹⁰⁸, and PHOX2B in turn drives expression of Gata3 in the ventral region of r4⁹⁶. Indeed, IEEs in Phox2b knock-out mice fail to differentiate normally, exhibit reduced levels of GATA3, and occasionally mis-specify as serotonin progenitors⁹⁶. Phox2a, on the other hand, is not required for IEE development⁹⁶. Thus, IEE differentiation depends on a general HOXB1/PHOX2B network, with further specificity arising through the action of GATA factors.

The segregation of IEEs and FBMNs is reflected in their migration to distinct nuclei

Shortly after undergoing their final mitotic division, both FBMN and IEE cell bodies begin to migrate away from their shared point of origin (Fig. 2). In mice, these movements are initiated at about E10 and conclude around E15.5^{48,89,116–119}. During this time, IEEs and FBMNs separate from each other, as FBMNs travel caudally into their final position in r5 and r6¹²⁰. The IEEs populate r4-5, and are subsequently subdivided into the vestibular efferent neurons (VENs), which migrate dorsally, and the OCNs, which migrate ventrally and eventually give rise to MOCs (residing mainly in VNTB) and LOCs (located in and around LSO)^48,96. Importantly, FBMNs and IEEs extend axons together from their common origin and into the periphery as early as E10.5, so the trajectory of axons within and out of the brainstem reflects their migratory history⁴⁷.

Whereas FBMNs project ipsilaterally to innervate muscles on the same side of the body, many IEEs project to the contralateral cochlea. In particular, MOCs tend to project contralaterally and LOCs tend to project ipsilaterally, although the proportions of ipsilateral and contralateral projections vary across species²⁵. In mouse, about 66% of MOCs (but only 1% of LOCs), project contralaterally, with some neurons even projecting to both ears^25,121,122. The contralateral projections, which form the olivocochlear bundle (OCB), appear as early as E10^111,112, and the OCB is fully formed by E13.5⁴⁸, thereby overlapping substantially with the period of IEE migration. Indeed, in chickens, contralateral projections are achieved by a translocation of the IEE cell bodies across the floor plate, thereby leaving a crossed projection in their wake^51,52. In mice, the means of midline crossing are less clear, as extensive cell body translocations have not been detected^47,48,123. However, neurons positive for the IEE marker GATA3 have been observed in the midline of E10.5 mice, suggesting that at least some IEE soma may move from one side to the other^96,112. In mammals, the formation of a contralateral projection may involve the axon guidance receptor EPHB2. In mice, EPHB2 is expressed by IEEs, FBMNs, and the floor plate¹²⁴. In the absence of EphB2, IEE axons meander aberrantly across the midline before achieving anatomically normal projections, albeit several days later than wild-type controls¹²⁴.

The mechanisms that entice IEEs across the midline —and prevent FBMNs from following— may also involve the Slit-Robo pathway. In Slit1, Slit2 double knock-out mice, the total number of fibers crossing the midline in r4 is decreased by about half compared to wild-type controls, with some axons turning and running longitudinally along the floor plate rather than crossing directly¹²⁵. In contrast, Robo2 knock-out animals show a marked increase in the number of projections crossing the floor plate, as well as aberrant migration of ISL1-positive cell bodies onto the floor plate^125,126. These findings suggest that ROBO2 may ordinarily provide a repulsive cue that keeps FBMNs away from the midline. One can speculate that a similar mechanism might be at play in guiding the ipsilateral projections of IEEs.

Some IEE migratory decisions appear to depend on transcriptional networks active in all motor neurons (Fig. 3). For instance, both IEEs and FMBNs require the pan-motor neuron marker ISL1 for proper differentiation. For FBMNs, ISL1 is crucial for their caudal migration, as reflected by an accumulation of FBMNs in r4-5 rather than r6 in mice with reduced ISL1 levels¹²⁷. Although no major IEE migratory phenotypes were described, the IEEs developed truncated and disorganized projections out of the brainstem, with the few remaining axons barely reaching the ear by E11.5. Both types of defects may be due to altered expression of Slit2, as Slit2 levels are decreased in Isl1-deficient motor neurons¹²⁷.

Like ISL1, TBX20 is expressed by both IEEs and FBMNs, but has distinct effects on the migration of each population¹⁰¹. Caudal migration of FBMNs is controlled by the planar cell polarity (PCP) pathway^{110,128–140}, and Tbx20 mutant mice exhibit defects in caudal FBMN migration as well as a downregulation of PCP genes in FBMN somata¹⁰¹. Tbx20 expression is also required for midline crossing of IEEs, but it is unclear whether this phenotype is also due to changes in PCP gene expression¹⁰¹. Nevertheless, IEEs exhibit aberrant midline crossing in mice lacking the PCP component PRICKLE1, with reduced contralateral projections and some axons turning to wander along the floorplate longitudinally¹⁴⁰. In zebrafish, the caudal migration of octavolateralis efferents is also stalled in animals lacking the PCP proteins Scrib or Celsr2^136,137. More work is needed to understand what role—if any—the PCP pathway plays in the migration of mammalian IEEs. Thus, despite a common reliance on TBX20 and possibly PCP genes, there appear to be significant differences in how these genes act in IEEs and FBMNs.

Shortly before the IEEs and FBMNs have fully separated, IEEs begin to further diversify and separate from each other, with VENs migrating dorsally, MOC neurons moving ventrally, and LOC neurons settling in a more ventrolateral position⁴⁸. In mice, this migration happens between E12.5 and E14.5, with VENs and OCNs separating first, followed by a separation of MOC and LOC neurons⁴⁸. This process requires the transcription factor ASCL1 (formerly referred to as MASH1), as the VEN nucleus does not form in Ascl1 mutants⁹⁶. ASCL1 is also expressed in OCNs and FBMNs^96,104,141. However, although Ascl1 deletion delays the caudal migration of FBMNs, it has no effect on the movement of OCNs^96,141. Unfortunately, the identity of additional cell-type specific transcription factors that might direct subsets of OCNs to their final destinations remains unclear.

The final separation of VENs from OCNs relies in part on a ventral migration of OCNs, with both OCNs and FBMNs undergoing similar migrations away from the ventricular zone to arrive close to the ventral pial surface of the hindbrain^117,119 (Fig. 2). This process of radial migration depends on the Reln gene in both OCNs and FBMNs, although LOC neurons only appeared to be partially affected in Reln mutant mice, suggesting that additional mechanisms may be at work¹¹⁷. In FBMNs, Reln interacts with Dab1 and Cdk5. It is unknown whether similar mechanisms mediate its effects in OCNs^117,142 (see Table 1 for a summary of genes involved in OCN specification, migration, and guidance).

Efferents grow along existing nerves to find their synaptic targets

As they are migrating and settling in their final locations, OCNs begin to extend axons out of the brainstem and into the cochlea, ultimately terminating on post-synaptic targets in the organ of Corti (Fig. 4). These initial guidance events occur after the VIII^th nerve has begun to form, with SGN axons reaching the brainstem by E11.5⁷⁹, days before the IEEs have scattered to their distinct locations across the brainstem. The formation of efferent projections coincides with the formation of connections within the auditory brainstem, with bushy cell neurons in the CNC beginning to extend axons towards the midline around E13 and reaching the contralateral SOC by E17¹⁴⁸.

Consistent with the close relationship with the FBMNs, OCN axons initially follow along the facial genu in the brainstem but bypass the VII^th nerve and instead exit to join the nascent VIII^th nerve⁴⁸. Almost nothing is known about the mechanisms that direct OCN axons to grow into the VIII^th nerve, rather than continuing to follow the VII^th. One possibility is that the inner ear provides an attractive signal that could attract any motor neuron axons, but that OCN axons are in the best location to respond to the signal, an idea that has been supported by the behaviour of motor neuron axons presented with an ectopic otic vesicle^149,150. However, the inner ear is not necessary for growth into the periphery, as OCN axons mis-route into the VII^th nerve if the VIII^th nerve is absent¹⁵¹. IEEs also join the VII^th nerve inappropriately in Gata3 mutant mice, but more work is needed to understand whether this effect reflects a role for GATA3 in IEEs specifically, since GATA3 is also expressed in the developing cochlea and SGNs^112,145.

Upon reaching the VIII^th nerve exit point, OCN axons are thought to grow along a scaffold of SGN processes into the sensory epithelium of the inner ear^152,153. In support of this hypothesis, a wide array of mutations affecting SGN development and targeting also affect efferent innervation of the cochlea, with the phenotype of efferent neurons often matching the aberrant projections of SGNs^{145,146,151,154–156}. Lateral line efferents in zebrafish appear to rely on a similar mechanism to reach their targets¹⁵⁷. However, no molecular cues mediating fasciculation between afferents and efferents have been identified to date. In other peripheral nerves with both afferent and efferent components, this process depends on axon-axon interactions mediated by Eph signaling^158,159, but analogous roles for Eph receptors in the auditory system have not yet been described¹⁶⁰. Even more mysterious is the subsequent targeting of OHCs by MOC axons and of SGN terminals by LOC axons. However, it is thought that MOC outgrowth toward the OHCs may not rely on SGN scaffolding, as MOCs and Type II SGNs follow different trajectories between the IHC and OHC regions (schematized in Fig. 1), and MOCs still reach the OHCs in BDNF mutant mice that lack afferent innervation of OHCs^161,162.

Efferents reach the mammalian sensory epithelium during a period of circuit refinement

Efferents sit in a key node for auditory processing, both receiving sound information through the afferent pathway in the brainstem and then transmitting signals back to the afferents through synaptic contacts in the cochlea. It is therefore not surprising that this final pattern of connectivity depends on activity-dependent mechanisms that operate after the basic axon tracts have formed and synaptogenesis has begun, which occurs right around birth in mice¹⁶³. Indeed, the first two weeks of postnatal development in mice are a transformative epoch for the inner ear. Both SGNs and auditory brainstem neurons undergo extensive remodeling during this period to adopt distinct functional properties and converge on their final targets^3,164. As in other developing neural circuits, spontaneous activity plays a major role in the maturation and refinement of auditory circuitry¹⁶⁵. Before the onset of hearing (~P10 in mice¹⁶⁶), IHCs produce spontaneous calcium spikes, which propagate throughout the auditory system^167–175. In parallel, neurons in the auditory brainstem segregate into ever-more refined tonotopic arrangements, a process that is thought to depend on spontaneous activity from the inner ear¹⁷⁶.

During development, MOC axons are poised to influence the pattern of activity emanating from the cochlea, though the precise functions of this early cholinergic signaling remain unclear. In mice, tracing studies reveal that MOC neurons are the first to arrive in the sensory epithelium, appearing in the IHC region as early as E16.5 and in the OHC region around E18.5⁴⁸. Experiments in rats, hamsters, and gerbils identified a similar pattern, with dye-labeled MOCs first appearing near IHCs by P0, and extending to OHCs between P2 and P5, depending on the species^177–180. This time course also aligns with the appearance of choline acetyltransferase (ChAT) and acetylcholinesterase (AChE)-positive terminals in the cochlea. Both enzymes, which regulate acetylcholine synthesis and degradation, have been identified in the inner ear at birth or shortly thereafter in mice, rats, and hamsters^181–183. Consistent with dye-labeling studies, ChAT staining is first visible in the region of IHCs and can be seen around OHCs several days later¹⁸³. Like other aspects of cochlear development, OCNs arrive first in the base, and development progresses toward the apex¹⁷⁷.

Collectively, these findings suggest the existence of a “developmental waiting period,” in which MOC neurons transiently innervate the IHC region before synapsing onto their final targets, the OHCs^152,179. During this waiting period, OCNs also send projections to Kölliker’s organ^48,178, a transient structure believed to induce spontaneous activity in IHCs before the onset of hearing¹⁶⁷. It is not known whether OCNs form synapses with cells in this structure; however, subunits of acetylcholine receptors are expressed in Kölliker’s organ, raising the possibility of functional contacts^184,185. While the purpose of these temporary connections is unknown, transient innervations do influence the maturation of other developing circuits, such as thalamocortical neurons, which form short-lived connections in the subplate that prepare them for subsequent formation of synapses in layer four of the cortex^186–188.

MOC neurons affect prehearing synaptic activity

During this early postnatal waiting period, MOC terminals form synaptic contacts with IHCs, which transiently express SK2 channels and AChR subunits that mediate post-synaptic responses (Fig. 5)^189,190. Although little is known about how the MOC-to-IHC synapse forms, this transient connection does appear to be functional. Electrophysiological recordings from IHCs of rats during this time identified the presence of inhibitory postsynaptic currents (IPSCs) that disappeared with the application of the AChR antagonists strychnine and alpha-bungarotoxin, indicating that MOC neurons (the only cholinergic cells presynaptic to IHCs) inhibit IHCs^189,190. Similarly, both efferent stimulation and exogenous application of acetylcholine in cochlear explants reduce IHC firing rates, whereas blockade of cholinergic signaling with strychnine increases IHC firing rates^190–193. IHCs remain responsive to acetylcholine through at least the second postnatal week in both mice and rats^194–196. In rats, IHCs cease to respond to acetylcholine sometime between P16 and P21^194,195. IHCs in mice remain responsive to acetylcholine slightly longer than in rats, but they also stop responding to acetylcholine by the sixth postnatal week¹⁹⁶. By adulthood, the IHC membrane no longer responds to OCN stimulation^196,197.

Given the evidence that spontaneous activity originating in the cochlea ultimately propagates along the entire developing auditory neuraxis, it is tantalizing to suppose that efferent inhibition of IHCs during early postnatal stages might also shape auditory circuits in the CNS. Several studies have attempted to address this question by investigating auditory development in mice lacking the nAChR subunits α9 or α10, which are required for OCNs to form functional cholinergic synapses on hair cells^198,199. Indeed, tonotopic refinement of brainstem circuitry is disrupted in nAChRα9 knockout mice, at least in the LSO¹⁷⁶. However, the primary defect appears to be a failure to prune synapses after the onset of hearing, suggesting a role independent of spontaneous activity. The IHCs also appear to have defects in vesicle fusion at ribbon synapses, indicating a possible role for early efferent innervation on hair cell maturation²⁰⁰, although other properties of IHCs seem to be normal in mice lacking cholinergic OCN innervation^200,201.

Following this period of temporary MOC-IHC contacts, MOC axons gradually transition into the OHC region, concomitant with the arrival of LOC neurons in the sensory epithelium during the first postnatal week, although distinguishing LOCs from MOCs by morphology at this stage remains a challenge. In rats, OHCs begin responding to acetylcholine at P6, shortly after the onset of SK2 expression^202,203. In the cat and rat, MOCs first form transient, axo-dendritic synapses on Type II SGNs before synapsing directly onto OHCs^163,204,205. At least some of the early contacts onto OHCs appear to be branched projections from MOCs innervating the IHC region^180,206 (Fig. 5). The transition from IHCs to OHCs may involve BDNF, as MOC innervation of the OHC region is delayed in BDNF mutant mice¹⁶². However, the factors that ultimately guide MOCs toward the OHC region have yet to be identified.

The arrival of MOC neurons in the OHC region also coincides with a loss of Type II afferent synapses in several species and the onset of OHC electromotility (the physical contraction of OHCs in response to sound)^205,207,208. These findings led early researchers to speculate that the arrival of MOC neurons influences the final maturation of OHCs, including the formation of synapses with Type II SGNs^209,210. In support of this idea, in Hoxb1 mutant mice, MOC neurons fail to develop and their absence is accompanied by a gradual degeneration of OHCs, and hence progressive hearing loss⁷¹. However, other essential auditory nuclei are also disrupted in these mutants, and the loss of OHCs is limited to the apical turn, leaving the specific contributions of MOCs unclear. Additionally, OHCs still become electromotile in the absence of functional MOC inputs^211,212. Other research suggests that MOCs are not required for normal synaptic connectivity between Type II SGNs and OHCs: in cats that have had all efferent connections to the cochlea severed neonatally—before MOC neurons typically reach the OHCs—SGN synapses are still pruned to a number similar to that seen in wild-type animals²¹³. Whether the removal of MOC contacts with IHCs makes room for the newly arrived LOC axons remains unknown.

Does the arrival of LOC neurons affect SGN subtype determination?

During this period of efferent rearrangement, SGNs acquire their own unique traits, raising the possibility that these two aspects of circuit maturation are causally linked. Type I SGNs are generally classified into three subtypes, based on their spontaneous firing rates and thresholds²¹⁴. In rats, diversification of firing rates occurs primarily between the second and third postnatal weeks, around the time OCNs are finalizing their own connections²¹⁵. Moreover, LOCs preferentially innervate the subclass of SGNs with the lowest spontaneous firing rate and highest firing threshold, raising the possibility that SGN subtype differentiation may involve cues from LOC neurons^216,217.

This hypothesis was tested directly in kittens by severing the OCB shortly after birth and measuring the firing rates of single auditory nerve fibers. Although firing rates overall were decreased in these animals, all three SGN subtypes remained present, as assessed by recording from fibers in the VIII^th nerve²¹⁸. LOC neurons do, however, seem to play a role in both establishing and maintaining morphological properties of the SGN-IHC synapse^219,220. Detailed anatomical studies suggest that low spontaneous firing rate (low-SR) SGNs form synapses on the modiolar side of the IHC (i.e. closer to the ganglion) whereas high spontaneous firing rate (high-SR) SGNs synapse instead on the opposite side (i.e. the pillar side). SGN axon segregation is accompanied by striking differences in synaptic morphology: low-SR SGNs receive signals through small post-synaptic boutons apposed to large pre-synaptic ribbons, whereas high-SR SGNs develop larger boutons opposite to small ribbons²²¹. This distribution of synapses emerges gradually over the first month of life, with the mature pattern apparent right around the time that OCN axons exhibit fully mature electrophysiological properties. When the OCB was severed in adult mice, the gradient of synaptic morphology was disrupted²¹⁹, raising the possibility that LOC axons influence SGN firing rates and hence synaptic morphology. Unfortunately, since no genetic tools are available to manipulate LOC or MOC neurons prior to birth, the full extent of OCN influence on auditory maturation remains unclear.