Current Concepts in Cochlear Ribbon Synapse Formation

Abstract

In mammals, hair cells and spiral ganglion neurons (SGNs) in the cochlea together are sophisticated “sensorineural” structures that transduce auditory information from the outside world into the brain. Hair cells and SGNs are joined by glutamatergic ribbon-type synapses composed of a molecular machinery rivaling in complexity the mechanoelectric transduction components found at the apical side of the hair cell. The cochlear hair cell ribbon synapse has received much attention lately because of recent and important findings related to its damage (sometimes termed “synaptopathy”) as a result of noise overexposure. During development, ribbon synapses between type I SGNs and inner hair cells form in the time window between birth and hearing onset and is a process coordinated with type I SGN myelination, spontaneous activity, synaptic pruning, and innervation by efferents. In this review, we highlight new findings regarding the diversity of type I SGNs and inner hair cell synapses, and the molecular mechanisms of selective hair cell targeting. Also discussed are insights into cell adhesion molecules and protein constituents of the ribbon synapse, and how these factors participate in ribbon synapse formation. We also note interesting new insights into the morphological development of type II SGNs, and the potential for cochlear macrophages as important players in protecting SGNs. We also address recent studies demonstrating that the structural and physiological profiles of the type I SGNs do not reach full maturity until weeks after hearing onset, suggesting a protracted development that is likely modulated by activity.

1.0. INTRODUCTION

1.1. Sound Detection and the Molecular Composition of the Ribbon Synapse

1.1.1. Basic Events in Sound Detection by the Cochlea

Sound perception for mammals begins with detection of sound by the auditory sense end organ known as the cochlea. During hearing, sound waves are collected by the outer ear and transmitted through the tympanic membrane and the ossicular chain to the cochlea in the inner ear. As the sound waves travel through the cochlear spiral, a series of events converts this mechanical energy into electrochemical signals perceivable by our nervous system. The spiral structure of the cochlea is divided into three chambers – the scala vestibuli, scala media, and scala tympani (see Figure 1A). Scala media and scala tympani are separated by the basilar membrane, which is rigid and narrow at the base of the cochlear spiral and becomes progressively more flexible and wider towards the apex. This peculiar architecture of the basilar membrane results in higher frequency sounds vibrating the basilar membrane at the base and lower frequency sounds vibrating the basilar membrane at the apex and sets the first level of tonotopic specificity within the cochlea. Recent reviews and book chapters elaborate on biomechanics and mechanoelectrical transduction underlying this process and highlight important findings (Goodyear & Richardson 2018; Hudspeth 2014; Mann & Kelley 2011).

Figure 1. Summary of cochlear hair cell afferent innervation and inner hair cell ribbon synapses.

(A) P0 mouse cochlear cross section with Tuj1 immunostaining (green) to reveal the spiral ganglion neurons (SGNs) and efferent fibers. Actin (phalloidin; magenta) was used to reveal the anatomy of the cochlea. sv, scala vestibuli; st, scala tympani; pa, peripheral axon; ca, central axon. (B) An E16.5 cochlear cross-section immunostained with Tuj1 antibodies to show the orientation of the nerve fibers within the organ of Corti, along with the inner and outer hair cells (IHCs and OHCs). (C) A representative whole-mount view of the cochlea at E17.5. Sox2 immunostaining reveals the organ of Corti (oC), and is also apparent in the spiral ganglion, but at lower levels in this image. (D) Cartoon schematic of the different classes of nerve fibers that innervate the organ of Corti. The high-magnification view of the inner hair cell shows the arrangement of ribbon bodies and postsynaptic densities of different sizes. TM, tectorial membrane; LOC, Lateral Olivary Complex; MOC, Medial Olivary Complex; SGN, spiral ganglion neuron; SR, spontaneous rate; m, modiolar; p, pillar; c, cuticular; h, habenular. Parts of this schematic were published previously (Zhang & Coate 2017) and reproduced here with permission from Elsevier. (E-G) Whole-mount immunostaining of the P12 cochlea using anti-Ctbp2 (red) and anti-Shank (green) antibodies reveals the pre- and postsynaptic components of the inner hair cell ribbon synapses. Scale bar in G: A = approximately 300 μm; B = 25 μm; C = 150 μm, E-G = 8 μm.

1.1.2. The Organ of Corti

The organ of Corti rests on top of the basilar membrane and contains specialized sensory epithelial cells called hair cells, which are organized into one row of inner hair cells and three rows of outer hair cells (Figure 1B). Classic studies on the development of the cochlear sensory domain have demonstrated the proneural transcription factor Atoh1 (red staining in 1B) is essential for hair cell development (Bermingham et al. 1999; Woods et al. 2004). The hair cells and supporting cells comprise the organ of Corti (“oC” in C; Sox2 staining; Figure 1C) where mechano-electric transduction begins. Figures 1A and 1C show cross-sectional views of the cochlea with Tuj1 immunostaining, which illuminates the spiral ganglion neuron cell bodies, their peripheral axons (“pa” in 1A and 1C; a.k.a “dendrites”) projecting toward the hair cells, and their central axons (“ca” in 1A and 1C) extending toward the brainstem. Hair cells are characterized by the presence of mechanosensory hair bundles at the apical surface of the cell that contain ion channels that open or close depending on the degree of deflection of hair bundles (Fettiplace 2017). In mammals, hair bundles are deflected through shearing forces against the gelatinous tectorial membrane, which sits on top of hair cells and is anchored by interdental cells, an arrangement that allows it to vibrate in tandem with the vibrations in the basilar membrane (Goodyear & Richardson 2018).

1.1.3. Introduction to Spiral Ganglion Neurons

Spiral ganglion neurons (SGNs) connect hair cells in the cochlea to the cochlear nucleus in the brainstem and serve as the afferent arm of the peripheral auditory pathway (Nayagam et al. 2011; Yu & Goodrich 2014). The majority of SGNs (~95%) are type I SGNs that form ribbon-type synapses (see section 1.1.4) with inner hair cells. In the cochlea, the ribbon synapse is where glutamate is released from hair cells onto SGNs as a result of sound input. As illustrated in Figure 1D, each SGN forms only a single ribbon synapse with one inner hair cell, whereas each inner hair cell forms ribbon synapses with multiple SGNs (Meyer et al. 2009). The minority 5% of SGNs, the type IIs, form ribbon synapses with outer hair cells, and each type II SGN synapses onto multiple outer hair cells via en passant connections after turning towards the base of the cochlea (Weisz et al. 2012). Both type I and type II SGNs are excited by glutamate (Glowatzki & Fuchs 2002; Weisz et al. 2009), although it has also been shown that type IIs are able to respond to adenosine triphosphate (ATP) released after hair cell ablation (Liu et al. 2015). The focus of this review is on the development of type I SGN/inner hair cell ribbon synapses. Much of this review focuses on studies where mouse was used as a model system. Unless otherwise noted, the staging nomenclature (“E” for embryonic day and “P” for postnatal day) refers to the staging in mouse. Many of the topics addressed here were also discussed in a previous review (Bulankina & Moser 2012). Aspects of type II SGN/outer hair cell development and function were also reviewed recently (Zhang & Coate 2017). The axons of olivocochlear efferent neurons are also observed in the cochlea (Figure 1D and these cells are also labeled by Tuj1 antibodies in 1A-C); the development and function of these fascinating cells was also reviewed recently (Frank & Goodrich 2018).

1.1.4. The Molecular Composition of the Ribbon Synapse

Ribbon synapses differ greatly from “conventional” synapses in terms of their structure, function, and molecular composition (see Safieddine et al. 2012 for a summary of differences between CNS and ribbon synapses). In terms of the molecular constituents of the inner hair cell ribbon synapse, detailed summaries of the known proteins facilitating pre- and postsynaptic function have been published recently (Pangrsic et al. 2018; Reijntjes & Pyott 2016; Wichmann 2015), and these lists are likely incomplete given the postsynaptic density of excitatory synapses in the CNS was estimated to contain upwards of 620 distinct proteins (Collins et al. 2006). Visually resolving the molecular components of the ribbon synapse using stimulated emission depletion (STED) microscopy was also discussed recently (Rutherford 2015). In section 3.3 here, we describe how some of these factors are needed, at least in part, for ribbon synapse formation. The dominating structure of the ribbon synapse is the electron-dense round or spherical presynaptic ribbon body (Sobkowicz et al. 1982) decorated by synaptic vesicles and contains scaffolding proteins like RIBEYE (Khimich et al. 2005; Schmitz et al. 2000), Bassoon (Khimich et al. 2005), and Piccolino (Regus-Leidig et al. 2013, 2014). Interestingly, RIBEYE is transcribed from the same gene as the transcriptional regulator C-terminal binding protein 2 (Schmitz et al. 2000). Ctbp2 and Ribeye each have their unique promoters and N-terminal regions, but they share an identical C-terminal domain. Thus, antibodies that bind to Ctbp2 also bind to RIBEYE, as illustrated by the immunostained ribbon bodies in Figure 1D. Other presynaptic proteins include Ca_v1.3 isoforms (Vincent et al. 2017) for voltage-gated Ca²⁺ entry (Frank et al. 2009), Vglut3 for the uptake of glutamate (Seal et al. 2008), Otoferlin for vesicle fusion (Michalski et al. 2017) and vesicle resupply (Pangšrič et al. 2010), and several others (Wichmann 2015). The main constituents of the post-synaptic ribbon synapse immediately adjacent to the site of glutamate release are the scaffolding proteins PSD-95 (Davies et al. 2001), Shank1 (Braude et al. 2015), Homer (Becker et al. 2018), and a large variety of glutamate receptors (Reijntjes & Pyott 2016; Sebe et al. 2017). Shank1 immunostaining adjacent to Ctbp2 is shown in Figures 1F and G. In terms of glutamate receptor subtypes expressed by SGNs, the current wisdom is that α-amino-3-hydroxy-5-methyl-isoxazolepropionic acid (AMPA)-type glutamate receptors are primarily associated with afferent excitability because of classic studies of excitatory post-synaptic currents (EPSCs) in the afferent bouton (Glowatzki & Fuchs 2002). It has now become customary to illuminate the afferent postsynaptic densities with antibodies against GluA2/3, and GluA puncta juxtaposing Ctbp2 puncta are generally regarded as “functional” ribbon synapses (Liberman & Liberman 2016; Meyer et al. 2009; Nemzou et al. 2006). N-methyl-D-aspartate (NMDA) receptors, kainate receptors, and metabotropic glutamate receptors have all been reported as being expressed by type I SGNs, but their function in afferent transmission appears to be minor, related to nuances of afferent excitability, or controversial (Reijntjes & Pyott 2016). It is also possible the non-AMPA glutamate receptors participate in various aspects of SGN development or survival, as is the case for GluN1 (Zhang-Hooks et al. 2016).

1.2. Synaptic Specificity Between Type I SGNs and Inner Hair Cells

What are the molecular factors that control synaptic specificity between SGNs and hair cells? Compared to other systems such as the visual system, where research over the past two decades has unveiled numerous cellular and molecular mechanisms related to synaptic specificity (Sanes & Yamagata 2009; Zhang et al. 2017), our knowledge toward answering this question is quite limited. During the formation of a typical excitatory synapse, scaffolding proteins, synaptic adhesion molecules, glutamate receptors, proteins associated with synaptic vesicle release, and synaptic vesicles themselves all need to be assembled in a timely and coordinated manner (see McAllister 2007; Scheiffele 2003a for reviews on cellular and molecular mechanisms of synapse formation). In terms of the inner hair cell ribbon synapse, we really only have an organ-level understanding of this process: we understand how axon guidance factors help guide the SGNs toward either the inner or outer hair cells (see section 2), we know neurotrophins are necessary for SGN survival and maintaining ribbon contacts after noise damage (Green et al. 2012; Wan et al. 2014), and we know Mafb (a transcription factor) ensures appropriate ribbon synapse terminal differentiation (Appler & Lu 2013; Yu et al. 2013). As described in sections 3.2 and 3.3, a small handful of factors that may facilitate ribbon synapse formation have been identified, but unfortunately, we are not yet close to understanding how these factors and others may act at the cellular or subcellular level.

In terms type I SGNs and inner hair cells, it is logical to think distinct molecular mechanisms might be in place to facilitate their connectivity to ensure the development of intensity or frequency coding. With respect to intensity coding, it is known that the type I SGNs are heterogeneous in terms of their morphology and rates of spontaneous discharge (Liberman 1982; Liberman & Oliver 1984), although more recent work suggests the spiking characteristics of an individual type I peripheral fiber may relate mostly to presynaptic release properties (Wu et al. 2016). For additional information on SGN spiking characteristics and spontaneous activity, see (Heil & Peterson 2017). As shown in Figure 1, type I SGNs with high spontaneous discharge rates (“SR”) and low thresholds are relatively thick and tend to terminate onto the abneural or “pillar cell” side of the inner hair cell (Liberman 1982; Liberman & Oliver 1984). Conversely, type I SGNs with low-SR and high thresholds tend to be relatively thin and terminate onto the neural or “modiolar” side of the inner hair cell (Liberman 1982; Liberman & Oliver 1984). The micrographs in Figures 2C and D illustrate SGN fibers contacting the modiolar and pillar sides of the inner hair cell. To complicate matters, there is also a third “medium-SR” class of type I SGNs (Liberman 1978), and there are clear differences in pre- and postsynaptic synapse volumes along the long (“habenular-cuticular”) axis of the inner hair cell (Liberman & Liberman 2016). More recently, three groundbreaking single cell RNA-sequencing studies have demonstrated three distinguishable subdivisions of SGNs (Petitpré et al. 2018; Shrestha et al. 2018; Sun et al. 2018). Each subdivision can be defined by varying levels of mRNAs encoding neurotransmitter receptors, factors related to synaptic transmission (e.g. channels), transcription factors, adhesion molecules, and cytoskeletal proteins. In all likelihood, the three classes of type I SGNs – the 1a, 1b, and 1c classes (Shrestha et al. 2018; Sun et al. 2018) – correspond to the high, medium, and low-SR populations, but this issue remains to be addressed experimentally. Interestingly, Petitpré et al. performed physiological studies of type I SGNs in vitro and found that their “type Ib” population (named “type Ic” in the other two reports) could be further subdivided into two groups based on how they spiked during sustained depolarization. Half of these neurons showed only a single spike, while the remaining half spiked multiple times. Functionally, the heterogeneity in firing characteristics and morphologies of type I SGNs most likely underlies the ability of the cochlea to communicate a very large dynamic range of sound intensities to the brain (Liberman 2017). The stereotyped spatial positioning of the type I afferent endings and their ribbon synapses around the base of the inner hair cell complements, and may result from, measurable differences in the voltage dependence of Ca²⁺ influx at glutamate release sites along the modiolar-pillar axis (Ohn et al. 2016). However, the cellular and molecular mechanisms underlying the interesting pattern of inner hair cell innervation (shown in figures 1D and 2E) are overall not well understood.

Figure 2. Highlights of Ribbon Synapse Development in the Mouse Cochlea.

(A) A whole-mount confocal image of a P4 mouse cochlea showing sparse numbers of genetically labeled SGNs. The cochlea came from a mouse carrying Neurog1^CreERT2 (Koundakjian et al. 2007) and R26R-tdTomato and its mother was not exposed to tamoxifen (so “background” labeling is shown). The hair cells are immunostained with anti-myosin VI antibodies (blue) and tdTomato was immunolabeled using anti-dsRed antibodies (white). The arrowheads point to small branches from the type I SGNs not yet refined at this stage. IHC, inner hair cell; OHC, outer hair cell. (B) Similar preparation as in A, but at P10. The micrograph illustrates how the type I SGNs show unramified bouton endings at this stage. (C and D) High magnification images from the boxed regions in B. 3D reconstructions were generated (using Imaris software) from the original confocal z-stacks and then rotated to show SGN fibers contacting the “pillar” or “modiolar” side of the inner hair cell. C is an image in the XY plane that was rotated slightly toward the Z axis. D is an image that was rotated 90° to show the YZ plane. Scale bar in D: A and B = 25 μm; C and D = 8 μm. (E) A schematic of the developmental time line of ribbon synapse formation. See section 3.1 for additional details. Around P0, the SGNs show elaborate branching and the ribbon synapses (red) are small and immature. The small branches refine through P10 (~hearing onset). By P30, the ribbon bodies and their apposed postsynaptic densities (blue) have fully matured, and the three type I subdivisions (1a, 1b, and 1c) are apparent. m, modiolar; p, pillar; c, cuticular; h, habenular. (F) Approximations of the timing of different important events in SGN development with respect to the time points illustrated in E.

In terms of frequency (tonotopic) differences, there are several morphological and physiological differences notable in terms of the ribbon synapses joining type I SGNs and inner hair cells. First, the number of type I SGN contact points per inner hair cell is by no means uniform along the tonotopic axis: the greatest concentration of ribbon synapses is found in the middle of the cochlea (~8–32 kHz) where hearing is most sensitive (Meyer & Moser 2010). Whether this pattern manifests during development as a result of preferred targeting to the middle region of the cochlea by the SGNs, or as a result of variable levels of pruning, or some combination of both is not at all clear. Second, it was shown in gerbil cochleae that ribbon bodies vary in that ribbons in high frequency regions have an ellipsoid appearance while ribbons in low frequency regions are spherical (Johnson et al. 2008). Third, the inner hair cell presynaptic architecture varies between the apex and base, particularly with respect to spatial differences between Ca²⁺ channels (i.e. Ca_v1.3) and the Ca²⁺ sensor (i.e. Otoferlin; Roux et al. 2006). Whereas Ca²⁺ “nanodomains” are found where the opening of a single Ca²⁺ channel is sufficient for the release of a single neighboring glutamate vesicle, Ca²⁺ “microdomains” are found where the opening of multiple Ca²⁺ channels is required for the release of distant vesicles (Kim et al. 2013). In physiological studies by Johnson et al., the differential modes of actions of the Ca²⁺ chelating agents BAPTA and EGTA were leveraged to determine that Ca²⁺ nanodomains are most prominent in inner hair cells of the cochlear apex, whereas Ca²⁺ microdomains feature prominently in the cochlear base (Johnson et al. 2017). While these studies reveal fascinating differences in modes of frequency-dependent exocytosis, the mechanisms that control the development of different presynaptic Ca²⁺ channel/Ca²⁺ sensor architectures between the apex and base remain unknown. Fourth, in terms of SGN morphology, the SGNs show comparatively larger cell bodies and axonal diameters in high frequency regions compared to low frequency regions (Echteler & Nofsinger 2000; Liberman & Oliver 1984; Nadol Jr. et al. 1990), but the extent to which these features correlate with the morphology of the ribbon bodies is unclear. Fifth, in vitro studies have shown the SGNs have distinguishable firing kinetics based on their tonotopic location: SGNs found at high frequency regions show a comparatively faster firing rate compared to SGNs found in low frequency regions (Adamson et al. 2002). Thus, the pre- and postsynaptic ribbon structures in the cochlea are heterogeneous in terms of morphology and physiology in the context of both frequency and intensity coding and distinct synaptogenic and/or differentiation mechanisms logically could underlie these differences. In addition, neuronal activity is known to strongly influence synapse maturation and refinement in many contexts as a result of transcriptional changes (Flavell & Greenberg 2008), thus many of the differences (as described above) may also manifest as a result of auditory activity and changes in gene expression. In strong support of this idea, SGNs lacking normal excitation because of mutations in genes encoding hair cell mechanotranduction proteins (Sun et al. 2018) or loss of Vglut3 (Shrestha et al. 2018; Sun et al. 2018), show reduced diversification.

1.3. Cochlear Ribbon Synapse Loss After Noise Exposure

The cochlear ribbon synapse has recently captured broad attention because it is the focal point of recent revelations concerning age-related and noise-induced hearing loss in humans (Liberman 2017). Seminal findings by Kujawa and Liberman (Kujawa & Liberman 2009) showed that moderate environmental noise causing temporary hearing loss (with no loss of hair cells) leads to reduced numbers of ribbon synapses followed by a slow, progressive loss of SGNs. Interestingly, the SGNs with low rates of spontaneous discharge (low-SR; and presumably the thinner ones) are particularly susceptible to these effects (Furman et al. 2013). In correlation with this, the 1c subclass of type I SGNs, which may represent the low-SR population, is preferentially eliminated in aging mice (Shrestha et al. 2018). The primary cause of noise-induced synapse loss can likely be attributed to the SGN peripheral termini being overexposed to glutamate. This idea of “glutamate excitotoxicity” is supported by previous reports showing that both moderately loud noises and glutamate analogs can cause the SGN termini to swell and become damaged, and that these detrimental effects can be blocked by glutamate antagonists (Pujol et al. 1985, 1993; Robertson 1983). In addition, evidence suggests glutamate overexposure at the ribbon synapse due to the activity of calcium permeable AMPA receptors leads to toxic levels of cytosolic calcium within the afferent terminal (Sebe et al. 2017), a concept supported by many studies in the CNS (Sattler & Tymianski 2001). Thus, drugs to mitigate glutamate excitotoxicity in SGNs are currently in development, as are neurotrophin-based therapies, because brain-derived neurotrophic factor (BDNF) and neurotrophin-3 (NT3) have been shown to support the survival of SGNs and ribbon synapses in the context of both cochlear development and regeneration (Green et al. 2012; Wan et al. 2014; Wang & Green 2011; Wise et al. 2005). Despite the recent enthusiasm for ribbon synapses from this standpoint of hearing loss, we still know very little about the mechanisms controlling their assembly during development, and what mechanisms may be needed for their re-assembly in the context of damage.

2.0. Early steps of hair cell targeting by SGNs

2.1. Mechanisms of selective hair cell targeting – classical axon guidance mechanisms within the sensory domain

The expression of many axon guidance factors has been reported in the inner ear and these appear to be poised to affect SGN development (Coate & Kelley 2013). Previously, investigators had examined several guidance factors and their impact on targeting neurites to vestibular organs (Battisti et al. 2014; Gu et al. 2002; Tessarollo et al. 2004). More recently, the molecules influencing how radially-projecting SGNs selectively innervate different hair cell regions within the cochlear sensory domain have been identified (Table 1).

Table 1.

Proteins that facilitate ribbon synapse formation

Protein name(s) or pathway	Known or proposed function(s)	Reference(s)
Activity in general	Type I SGN differentiation	(Barclay et al. 2016; Shrestha et al. 2018; Sun et al. 2018)
Ephrin-A5 and EphA4	Restricts type I SGNs from the OHC region	(Defourny et al. 2013)
Sema3F and Nrp2	Restricts type I SGNs from the OHC region	(Coate et al. 2015)
Wnt-9a	Upregulates factors that attract auditory afferents in chick basilar papilla	(Munnamalai et al. 2017)
Vangl2, Frzd3, Celsr1	Controls type II SGN turning toward the cochlear base	(Ghimire et al. 2018)
Rspo2	Controls type II SGN turning toward the cochlear base	(Mulvaney et al. 2013)
Prox1	Controls type II SGN turning toward the cochlear base	(Fritzsch et al. 2010)
Thryoid hormone signaling	Pruning of ribbon synapses in the cochlea	(Sendin et al. 2007; Sundaresan et al. 2016)
NrCAM	May regulate ribbon synapse maturation	(Harley et al. 2018)
Np65	Ribbon synapse formation	(Carrott et al. 2016)
Bassoon	Assembly of presynaptic ribbon synapse constituents	(Buran et al. 2010; Frank et al. 2010; Jing et al. 2013; Khimich et al. 2005)
Shank1	Ribbon synapse maturation	(Braude et al. 2015)
Ribeye	Assembly of presynaptic ribbon synapse constituents, vesicle recruitment, synapse maturation	(Becker et al. 2018; Jean et al. 2018; Lv et al. 2016; Sheets et al. 2011, 2017)
GluN1	SGN spontaneous activity and survival	(Zhang-Hooks et al. 2016)

SGN, spiral ganglion neuron; OHC, outer hair cell

Classic axon guidance factors that influence SGN radial innervation include Ephrin-A5, a membrane-bound ligand, and one of its receptors, EphA4. During periods of innervation and refinement in the organ of Corti, EphA4 is expressed by type I SGNs while Ephrin-A5 is expressed by the outer hair cells as well as a subset of type I SGNs (Defourny et al. 2013). Interestingly, Ephrin-A5 is shown to be enriched in type II SGNs starting at P3 (Petitpré et al. 2018). To investigate the impact of EphA4/Ephrin-A5 signaling on radial innervation, Defourny and colleagues examined Efna5^−/− cochleae and found they exhibited increased numbers of peripherin-negative (type I) fibers and decreased numbers of peripherin-positive (type II) fibers projecting to the outer hair cells at P14 (Defourny et al. 2013). Complementing this, Efna5^−/− cochleae, compared with wild type controls, showed fewer presynaptic inner hair cell ribbon bodies and proportionally higher outer hair cell ribbon bodies. This radial shift of afferent innervation to the outer hair cells is maintained in three-month-old Efna5^−/− mice and was accompanied by reduced peak 1 amplitudes in response to subthreshold click stimuli in ABRs (Defourny et al. 2013). Similar effects were observed in the ABRs of Epha4^−/− mice (Miko et al. 2008), further supporting that EphA4 is acting as the receptor for Ephrin-A5. These data support a model in which EphA4 on the plasma membrane of the growth cones of type I SGNs are repelled by Ephrin-A5 present on the surface of outer hair cells. This model is further supported by growth cone collapse assays in which the size of growth cones in type I SGNs was reduced when cultured in the presence of Ephrin-A5, while type II SGN growth cones were unchanged relative to controls (Defourny et al. 2013).

Another classic axon guidance factor shown to impact radial innervation in the cochlea is the secreted ligand Semaphorin-3F (Sema3F) and its receptor, Neuropilin-2 (Nrp2). Sema3F is expressed in the lateral compartment of the organ of Corti, whereas Nrp2 localizes to all the SGNs (type I and type II) during periods of hair cell innervation. To test the hypothesis that Sema3F in the lateral compartment inhibits type I projections to the outer hair cells, Coate and colleagues examined peripheral afferent fibers in Nrp2^−/−, Nrp2^+/−, and Sema3f^−/− mice. Compared to wildtype, Nrp^−/−, Nrp^+/−, and Sema3f^−/− mice (in some cases mutant mice were crossed with Neurog1^CreERT2; R26R^tdTom to visualize a subset of SGNs) showed ectopic SGN projections to the outer hair cells by P0. Interestingly, and in contrast with the Ephrin-A5/EphA4 work mentioned above, both Nrp2^+/− and Sema3f^−/− cochleae showed no significant differences in numbers of ribbon bodies, ABRs, or numbers of type II projections (Coate et al. 2015). The authors speculated that early innervation errors could be corrected by synaptic pruning events, which normally occur after birth. Given that EphA4 and Ephrin-A5 have been shown to inhibit type I innervation of the outer hair cells and that those effects are maintained at older ages (Defourny et al. 2013), one possibility is that these other molecules are able to correct the innervation errors in the Nrp2^+/− and Sema3f^−/− mice. These data support a model whereby Nrp2-expressing type I SGNs are repelled from the Sema3F-expressing lateral compartment and then preferentially innervate inner hair cells. Interestingly, type II SGNs also express Nrp2, yet they still project into the lateral compartment. Given how Nrps form receptor complexes with various co-receptors and the composition of these complexes can impact the resulting intracellular signaling events and function (Zhou et al. 2008), one possible explanation is that an Nrp2 co-receptor is expressed by type I SGNs, but not type II SGNs. While one candidate co-receptor, neuronal cell adhesion molecule (NrCAM), has already been ruled out (Harley et al. 2018), other co-receptors such as Plexin (Plxn)A1 and PlxnA3 are indeed expressed by the SGNs (Coate et al. 2015; Katayama et al. 2013; Murakami et al. 2001) and remain candidate co-receptors. PlxnA3 protein was shown to be enriched in type I SGNs (Coate et al. 2015), but functional studies remain to be performed.

While much of the work in radial patterning of auditory afferents has been done in mouse models, another example of a molecular regulator of radial patterning across the sensory domain can be found in the chicken. Wnt9a is a secreted ligand and its transcripts are expressed on the neural edge of the basilar papilla (homologous to the organ of Corti) from E4.5 to E12 (Sienknecht & Fekete 2009), the time period of afferent innervation and synaptogenesis. After Munnamalai and colleagues overexpressed Wnt9a using a retrovirus, they observed, compared with controls, increased numbers of hair cells receiving afferent innervation as early as E6 and as late as E18. Wnt9a overexpression also resulted in an increase in the number of hair cells assuming a tall hair cell (homologous to inner hair cell) fate at the expense of the short hair cell (homologous to outer hair cell) fate (Takasaka & Smith 1971; Tanaka & Smith 1978). Wnt9a-overexpressing hair cells on both the neural and abneural sides had increased numbers of ribbon bodies relative to the controls (Munnamalai et al. 2017). In this case, Wnt9a is not likely acting directly as an axon guidance factor, as statoacoustic ganglion neurites have been previously shown to be unresponsive to Wnt9a and other Wnts (Fantetti 2011; Fantetti et al. 2011). One possibility is Wnt9a acts upstream of an axon guidance cue, which can influence radial innervation by upregulating cues attractive to afferents on the neural side or downregulating cues repulsive to afferents on the abneural side, such as Semas or Eph/ephrins in the mouse. RNA deep sequencing of Wnt9a-overexpressing and control basilar papillae indicate both Sema3D and Nrp2 are decreased in the presence of ectopic Wnt9a. These changes in gene expression, coupled with an overall increase in innervation density, are consistent with Sema3D serving as a repulsive guidance cue in the chicken cochlea. However, the expression pattern of the Sema3D transcripts (relatively uniform across the radial axis of the basilar papilla) marks a difference from the mouse Sema3F pattern (Munnamalai et al. 2017; Scott et al. 2018).

2.2. The turning of type II SGNs

While several guidance molecules (summarized above) have been identified that restrict a subset of afferents (type I SGNs in mouse) to the medial side of the organ of Corti, less is known about the molecular cues that regulate the projections of type II afferents. Whereas type I SGNs project to the inner hair cells, type II SGNs must project past the tunnel of Corti to the lateral compartment and do so starting around E15.5 in mouse. At E16.5, the type II SGNs make a 90^o turn and begin projecting toward the base of the cochlea (Figure 2A). As these fibers continue to project toward the base, they track along the basal surface of a row of outer hair cells and send branches to outer hair cells to synapse with them. One type II SGN fiber can form synapses with at least 10–15 outer hair cells (Berglund & Ryugo 1987; Koundakjian et al. 2007) and it has been estimated the initiation of an action potential in a type II SGN requires transmitter release by at least six outer hair cells (Weisz et al. 2014).

Recently, it was discovered by Ghimire and colleagues that planar cell polarity (PCP) proteins are involved in this curious turning of the type II SGNs. Vangl2 and other PCP proteins (Fzd3, Celsr1 and Celsr3) are expressed by SGNs, supporting cells, and hair cells and high-resolution imaging indicates Vangl2 localizes to the basolateral walls of the supporting cells. Vangl2^−/−, Frzd3^−/−, and Celsr1^−/− mice all showed type II SGN turning errors relative to littermate controls (Table 1). Through an elegant series of conditional knockout models, the authors demonstrated Vangl2 operates non-autonomously: loss of Vangl2 in supporting cells causes turning errors, whereas loss of Vangl2 in SGNs does not. Based on these results, the authors suggested two alternative models. In a “direct” model, Frzd3-expressing growth cones encounter Vangl2 on the basolateral wall of the inner pillar cell, which stabilizes that side of the growth cone and steers it toward the base. In an “indirect” model, the authors suggest classic axon guidance mechanisms (repulsion, attraction, or a combination of the two) could act downstream of Vangl2 and channel axon growth toward the base (Ghimire et al. 2018). We note here that, given how stereociliary bundle orientation is maintained in the rows of OHC where type II turning is disrupted in Celsr1^−/− and Vangl2^−/− cochleae (Duncan et al. 2017; Yin et al. 2012), the type II SGN turning errors are likely not due to a secondary effect from changes in hair cell polarity. In addition, we note that factors like Rspo2 (Mulvaney et al. 2013) and Prox1 (Fritzsch et al. 2010) were previously implicated in type II SGN turning, and it will be important to determine how PCP signaling acts in the context of these factors.

3.0. SGN development after birth

3.1. Synapse formation during the early neonatal period before hearing onset: a menagerie of cellular and molecular signaling events.

Several studies have shown that, in mice, type I and II SGNs arrive at their inner and outer hair cell synaptic locations by P0 (Coate et al. 2015; Druckenbrod & Goodrich 2015; Koundakjian et al. 2007), so it seems reasonable to estimate P0 as the time when ribbon synapse formation begins. At P0, presynaptic ribbon bodies are visible, but glutamate receptor staining appears diffuse (Huang et al. 2012). From P0 through hearing onset (Mikaelian & Ruben 1965; Polley et al. 2013; Shnerson & Willott 1979), glutamate receptor immunolabeling becomes progressively more punctate in appearance, as the receptors coalesce into discrete patches that juxtapose the ribbon bodies (Huang et al. 2012). This period between P0 and the onset of hearing represents a highly dynamic period of type I SGN development with at least five distinguishable events (see the following two paragraphs) occurring simultaneously and may all affect ribbon synapse formation. And to complicate matters, any molecular factors (e.g. cell adhesion molecules) facilitating ribbon synapse assembly may also be necessary during any of these other five events, and experimentally determining cell-type specific effects can often pose significant challenges.

First, and as shown in Figures 2E and F, the type I SGNs refine their branching patterns down to a single bouton ending, a process commencing around E15.5 (Koundakjian et al. 2007) and ends around hearing onset (Figures 1 and 2). This process of branching refinement likely accounts, in part, for the observed changes in glutamate receptor immunostaining profiles described above. Second, the type I SGNs begin to fire glutamate-dependent action potentials due to prehearing spontaneous activity that originates in cochlear supporting cells (Tritsch et al. 2010a; Wang et al. 2015; Zhang-Hooks et al. 2016). Spontaneous activity in the sensory epithelium and SGNs commences at least as early as P3 with the onset of TMEM16a expression in cells of the greater epithelial ridge (Wang et al. 2015). TMEM16a is a Ca²⁺-activated Cl⁻ channel necessary for supporting cell fluid secretion, which stimulates spontaneous activity in hair cells (Wang et al. 2015). Third, the type I SGN peripheral processes (the segment outside of the cochlear epithelium) become myelinated and sodium and potassium channels appear at heminodes and nodes (Kim & Rutherford 2016), a process appearing to commence around P5 (Kim & Rutherford 2016). With respect to branching refinement or spontaneous activity, it is not yet clear whether deficiencies in either of these events alter ribbon synapse formation or pruning. With respect to myelination, Schwann cells are known to be necessary for SGN migration during development because of defects observed in neural crest-specific Sox10 conditional knockout mice (Mao et al. 2014), but it is not clear whether the timing of myelination or node formation affects any aspect of the ribbon itself.

Fourth, the type I SGNs undergo a phase of synaptic pruning where about 50% of the ribbon synapses are shed (Huang et al. 2012; Sundaresan et al. 2016), an interesting process shown to be dependent on thyroid hormone signaling. The evidence for this came first from studies Pax8 knockout mice (Sendin et al. 2007) then later from Pit^dw mice (Sundaresan et al. 2016), both of which are mouse models of hypothyroidism. In each case, the normal dramatic loss of inner hair cell ribbon synapses between P5 and P14 failed to occur (Sendin et al. 2007; Sundaresan et al. 2016). Importantly, thyroid hormone treatment was able to restore normal calcium currents in Pax8-deficient inner hair cells (Sendin et al. 2007) and ribbon synapse pruning in cochleae of Pit^dw mice (Sundaresan et al. 2016). While much work needs to be done to determine possible mechanisms, thyroid hormone remains one of the only known mediators of ribbon synapse pruning in the ear. We note here that ribbon synapse pruning appears to coincide with SGN apoptosis (Echteler et al. 2005), but a direct link between these two events has not yet been demonstrated experimentally. Thus, it remains unclear whether the loss of ribbon synapses during this phase is a result of pruning extraneous secondary branches coming off the SGN peripheral process, SGN apoptosis, or some combination of both. And as a fifth event, the type I SGN peripheral processes become innervated by lateral olivocochlear (LOC) efferent axons (Bergeron et al. 2005), which protect against sound-evoked damage (Darrow et al. 2007) and may also regulate auditory input (Guinan 2017). We note here that the SGNs must also compete for synaptic space on inner hair cells with medial olivocochlear (MOC) efferents, which make transient contacts with inner hair cells before innervating outer hair cells (Frank & Goodrich 2018). Studies have shown efferent innervation starting as early as E18.5 in mouse cochleae (Bruce et al. 1997) and soon after birth in rat cochleae (Knipper et al. 1995; Simmons 2002). The extent to which afferent and efferent synapses on the SGN peripheral terminals form independently or in some coordinated manner is not well understood.

3.2. Cell Adhesion Molecules in Ribbon Synapse Formation

Many published works from studies in the CNS (McAllister 2007; Scheiffele 2003b) have delineated many mechanisms by which cell adhesion molecules (CAMs) establish and maintain contacts between pre- and postsynaptic cells, and later refine those connections. Arguably, the area of CAMs in synapse formation still represents a significant gulf in knowledge between the auditory system and the CNS, but recently several CAMs have been implicated in ribbon synapse formation in the ear (Table 1).

Neuronal cell adhesion molecule (NrCAM) is a member of the immense immunoglobulin (Ig) family of cell adhesion molecules and was recently shown to play a possible role in ribbon synapse formation (Harley et al. 2018). Since NrCAM localizes to the cell membranes of postnatal hair cells, supporting cells, and SGNs from embryonic through postnatal stages, the authors predicted NrCAM homophilic interactions may help facilitate ribbon synapse development in some way. Consistent with this, at P10, Nrcam^−/− mice showed increased numbers of presynaptic ribbons in both inner and outer hair cells compared to wild type. These ribbon bodies were smaller in size compared to WT, suggesting they are less mature; although, it was unclear from the work whether each of these ribbon bodies were associated with an afferent. This reduction in mature ribbon bodies in Nrcam^−/− mice appears to recover over time: when analyzed at later ages (P47–48), ribbon bodies from Nrcam^−/− inner hair cells were indistinguishable from controls. This apparent recovery in the Nrcam nulls was supported by comparatively normal ABRs and normal organ of Corti morphology. Nevertheless, the early increase in apparently immature ribbon bodies suggests NrCAM may promote afferent pruning or inhibit ribbon synapse formation (Harley et al. 2018) in some way.

Another Ig-CAM, Neuroplastin-65 (Np65), was also recently shown to impact ribbon synapse formation in the mammalian inner ear (Carrott et al. 2016), but at what seems like a later-shifted phase compared to NrCAM. At P16, Np65 localizes to the cuticular plate of both outer and inner hair cells and in the basolateral region of hair cells. Np65 mutant mice at P16 showed no differences in the number of presynaptic ribbons and postsynaptic densities; however, the mutant mice showed an increased number of ribbon bodies misaligned with post-synaptic densities, suggesting Np65 is involved in their pairing. Further analyses of neurotransmitter release from presynaptic ribbons indicated the readily-releasable pool was similar in both strains, but the secondary releasable pool was reduced in the Np65 mutants. Additionally, auditory brainstem response measurements revealed that Np65 mutants have elevated thresholds at P16 relative to the wildtype mice (Carrott et al. 2016), so the defect in pre- and postsynaptic pairing was sufficient to cause hearing deficits.

In addition to the IgCAMs discussed above, Thrombospondins (TSPs), secreted extracellular matrix glycoproteins, have also been shown to impact synapse formation (McAllister 2007). Recently, Mendus and colleagues examined the role of TSPs in synapse formation in the inner ear using TSP1 and TSP2 knockout mice (Mendus et al. 2014). At P29, TSP2^−/− and TSP1/2^−/− (double knockout) mice have significantly reduced numbers of ribbon bodies; however, the number of Shank1 puncta per HC is unchanged. In addition, there was a significant reduction in juxtaposition of the ribbon bodies with Shank1 puncta in all three knockout mice (TSP1^−/−, TSP2^−/−, TSP1/2^−/−), suggesting fewer functional synapses per inner hair cell. TSP2^−/− and TSP1/2^−/− mice showed elevated ABR thresholds relative to the controls at P15 and older ages, and TSP1/2^−/− mice showed significantly reduced wave 1 amplitudes and prolonged latencies (Mendus et al. 2014). However, it remains unclear exactly how TSP1 and 2 function in the context of ribbon synapse development, and this same issue applies to NrCAM and Np65. Future studies to examine cell type-specific roles of these factors, as well as inter- and intracellular signaling mechanisms will be needed to decipher their mechanistic function during ribbon synapse formation or maturation.

3.3. The role of ribbon synapse constituents in ribbon synapse formation.

In addition to the small number of CAMs described above, there are several reports demonstrating at least partial roles of the ribbon synapse scaffolding proteins in the formation of the ribbon synapse (Table 1). Perhaps the most dramatic effects have been reported from Bassoon mutants, where loss of Bassoon in hair cells leads to a spectrum of defects. In terms of auditory acuity, Bassoon mutants show elevated ABR thresholds along with significantly reduced wave 1 amplitudes (Buran et al. 2010). Bassoon mutants also show reduced spontaneous and sound-evoked firing rates (Buran et al. 2010), reduced inner hair cell fast exocytosis (Khimich et al. 2005), and smaller EPSC amplitudes (Jing et al. 2013). Morphologically, Bassoon mutants show reduced numbers of juxtaposed presynaptic ribbon bodies and postsynaptic densities, reduced numbers of synaptic vesicles, reduced numbers of Ca_v1.3 clusters around the ribbon bodies (Frank et al. 2010), and defective ribbon anchoring (Jing et al. 2013). With respect to the loss of calcium channel clustering and reduced alignment of pre- and postsynaptic sites, the Bassoon mutants show similarities to mice lacking the α₂δ2 subunit of Ca_v1.3 (Fell et al. 2016). Although not as pronounced, Shank1 mutants also show postsynaptic densities with smaller volumes (Braude et al. 2015), suggesting Shank1 may facilitate synapse maturation.

So, it would seem the scaffolding proteins of the ribbon synapse, in general, serve some function in ribbon synapse formation. But, what about Ribeye itself? One of the first inquiries into Ribeye came in a 2011 study of zebrafish lateral line hair cells (Sheets et al. 2011). In this study, Morpholino-mediated ribeye a and ribeye b knockdown (zebrafish have two copies), led to reduced hair cell afferent innervation and reduced numbers of MAGUK-positive postsynaptic densities (Sheets et al. 2011). Interestingly, some ribeye a/b double morphants had residual numbers of Ribeye-positive puncta and these typically had juxtaposed postsynaptic densities, but in many cases, these pre- and postsynaptic puncta were misaligned. Overall, these data supported the idea that Ribeye levels somehow dictated the postsynaptic densities within the afferent fibers and possibly helped recruit or maintain their termini (Sheets et al. 2011). Contrary to this notion, zebrafish overexpressing ribeye b have enlarged ribbons (and presumably more Ribeye per ribbon), and these mutants do not show changes in morphology or relative numbers of the postsynaptic densities (Sheets et al. 2017). Recently, two separate studies described the effects of genetic loss of Ribeye (removal of the A-domain of the Ctbp2 gene) in mouse in the context of inner hair cell afferent synapse formation and function (Becker et al. 2018; Jean et al. 2018) and, overall, the synapse formation phenotypes were not nearly as pronounced compared to Bassoon mutants (above). Interestingly, Ribeye knockouts show only moderate hearing impairment and weakened, but not absent, hair cell exocytosis (Becker et al. 2018; Jean et al. 2018). Interestingly, the inner hair cells from Ribeye knockout mice showed a normal presence and overlap of the presynaptic marker Bassoon and the postsynaptic marker Homer. By electron tomography, the Ribeye knockout hair cells showed a comparatively smaller presynaptic density with reduced numbers of synaptic vesicles (Becker et al. 2018), suggesting Ribeye is necessary for vesicle recruitment or maintenance. Ribeye knockout hair cells were also reported to have larger postsynaptic densities (Jean et al. 2018). In one of these studies (Jean et al. 2018), the Ribeye knockout hair cells were shown to have smaller and more numerous Ca_v1.3-positive structures compared to wild type hair cells. In this case, the loss of Ribeye disrupted the normal singular and larger Ca_v1.3-positive cluster at each active zone, but Ca_v1.3-positive puncta were still present. Interestingly, zebrafish ribeye a/b double morphants showed a complete loss of Ca_v1.3-positive clusters (Sheets et al. 2011), suggesting Ribeye is necessary for Ca_v1.3 recruitment. More recently, studies of zebrafish ribeye a and b double homozygous mutants, where Ribeye protein was dramatically reduced, showed some curious differences: these mutants showed reduced electron density within the presynaptic ribbon, but the presence of “ghost” ribbons still maintained synaptic vesicles (Lv et al. 2016). So, there appears to be some moderate discrepancies in Ribeye-deficient hair cells between rodents and fish, or it could be that phenotypic differences relate to biological differences between rodent cochlear hair cells and fish lateral line hair cells.

3.4. An Extended Phase of Auditory Maturation: hearing onset to P30

Given that hearing onset can be approximated to as early as postnatal day 10 in mouse (Polley et al. 2013), one might reasonably suspect the type I SGNs and inner hair cell ribbon synapses reach full maturation around this time or shortly thereafter. However, the results from several recent studies suggest a protracted period of maturation for these structures that has some previously unrecognized features and extends well past the point of hearing onset. Using immunostaining and confocal microscopy, Liberman and Liberman (Liberman & Liberman 2016) constructed a timeline of the development of inner hair cell pre- and postsynaptic ribbon synapse volumes at eight frequency locations along the tonotopic axis with the goal of determining the time point when the ribbon synapse profile mirrors that of the adult. Around the time of birth, the synapse profiles were similar to the adult profile: synapse volumes on the abneural/pillar cell side of the inner hair cell were greater compared to the volume sizes on the neural/modiolar side. However, shortly after the onset of hearing (from P14 to P18) this pattern broke down – and even reversed – and synapse volumes on the abneural/pillar cell side of the inner hair cell were comparatively smaller. Starting around P21, the pattern reverses course again, and by P28 the positions of the ribbons around the inner hair cell based on volume mirrored what is seen in adult cochleae. Similar findings were made recently in rat cochleae (Kalluri & Monges-Hernandez 2017). During this extended period (P7 to P28), the GluA2-positive postsynaptic puncta showed relatively little variation (Liberman & Liberman 2016), which suggests that these dramatic configuration changes may be limited to the presynaptic region. However, GluA2 is calcium-impermeable, and it is possible that the calcium-permeable AMPA receptors (GluA1, 3 and/or 4; Man 2011) show age-dependent changes in size or distribution around the inner hair cell. From the standpoint of postsynaptic physiology, it was recently shown in rat cochleae that the diversification of spontaneous firing rates directly follows hearing onset (Wu et al. 2016), a process preceding changes in spike train characteristics (Wu et al. 2016). During this period of time in mouse, hearing becomes more sensitive (as measured by auditory brainstem response thresholds) (Song et al. 2006), so perhaps synapse morphology and afferent firing characteristics also might be going through a small-scale evolution. Since these events coincide with the onset of activity, it is natural to wonder to what extent they depend on activity. Unfortunately, in vivo experimental approaches to address this issue are quite difficult for at least two different reasons. First, experiments to ablate all activity in the ear are difficult because of the well-characterized period of prehearing spontaneous glutamate release from hair cells onto the SGNs afferents (starting around P5) resulting from ATP signaling by supporting cells (Tritsch et al. 2010b; Tritsch & Bergles 2010; Wang et al. 2015; Zhang-Hooks et al. 2016). And second, as shown by a Vglut3^−/− mouse model, the complete loss of glutamate signaling between inner hair cells and type I SGNs leads to about a 50% reduction in SGN survival compared with controls (Seal et al. 2008). In addition, we now know SGNs from Vglut3^−/− maintain excitability due to potassium coming from the cochlea (Babola et al. 2018). Recently, though, mice with “reduced sensory stimulation” (after conductive hearing loss achieved by surgery) were found to have enlarged presynaptic ribbon bodies, and enlarged Shank1-, GluA2/3-, and GluA4-positive postsynaptic patches (Barclay et al. 2016). While these studies did not indicate whether this enlargement effect was specific to any region of the inner hair cell (i.e. pillar cell side vs. modiolar side), it does provide a general link between activity and ribbon synapse morphology. Overall, the maturation of the peripheral afferent structures in the cochlea is a process that continues on well past synaptic pruning and hearing onset and future studies will reveal the extent to which different aspects of this maturation period are dependent on activity.

4.0. Do immune cells serve a function in ribbon synapse development?

Over the past few decades, research in the CNS has shown remarkably diverse roles for immune cells, namely microglia, in the formation and remodeling of synapses. As part of their well-known role in phagocytosis, microglia engulf and remove dying cells, but they’ve also been shown to remove weakly active synapses and do so in response factors such as CX3CL1 (fractalkine) and the complement cascade proteins C1q and C3 (Schafer & Stevens 2015; Wu et al. 2015). Recently, macrophages have captured the attention of several investigators in the context of cochlear development, regeneration, and injury response and it is intriguing to consider the possibility that macrophages may act in ribbon synapse development in a manner akin to their microglia counterparts in the CNS. In support of this idea, Iba-1-positive macrophages were recently shown in the mouse cochlea throughout early postnatal development and in the vicinity of the ribbon synapses (Dong et al. 2018), which puts them in the right place at the right time. We also note that we have observed Iba-1 positive cells (presumably macrophages) in the cochlea as early as E15.5 (not shown). Interestingly, macrophages were shown to be present among SGNs and within the organ of Corti, and in each of these locations macrophage numbers were highest between P1 and P10 (Dong et al. 2018), which corresponds to the initial period of ribbon synapse formation and pruning (Huang et al. 2012; Nemzou et al. 2006). While it is intriguing to consider macrophages as somehow facilitating ribbon synapse development, it is also possible macrophages help remove SGNs undergoing apoptosis during this time, as apoptosis among SGNs during early postnatal stages has been observed in developing gerbil cochleae (Echteler et al. 2005). In the context of the injured cochlea, macrophages are known to enter the cochlea in response noise damage (Hirose et al. 2005) and several recent studies have revealed important roles for macrophages and fractalkine signaling in the context of SGN protection. Neurons commonly express the fractalkine ligand (a.k.a CX₃CL1), which is a chemotactic cytokine that stimulates microglia and macrophages through the receptor CX₃CR1 (Limatola & Ransohoff 2014). In 2015, Kaur and colleagues showed SGNs express CX₃CL1 and genetically ablating hair cells leads to elevated numbers of macrophages within the spiral ganglion and the organ of Corti (Kaur et al. 2015). Genetic loss of CX₃CR1 led to reduced numbers of macrophages in both of these compartments and, interestingly, poor survival of the SGNs. The same group followed up these findings by showing that SGNs in CX₃CR1^−/− mice are similarly lost after either acoustic damage or aminoglycoside exposure (Kaur et al. 2018b). Importantly, CX₃CR1 appears to also potentiate the recovery of functional ribbon synapses after an injury to the cochlea conferring only a temporary threshold shift (Kaur et al. 2018a; conference proceedings). Although the mechanisms remain to be determined, it appears macrophages and fractalkine signaling protect SGNs and may also help facilitate ribbon synapse recovery in the context of damage. If these processes recapitulate some aspects of developmental connectivity between SGNs and hair cells, then it is reasonable to suspect macrophages may also facilitate the formation or pruning of cochlear ribbon synapses during development.

5.0. Conclusions and Perspectives

In our view, the field has made strides in the area of ribbon synapse formation in the cochlea, but significant gaps still remain to be filled. In particular, the molecular regulation of ribbon synapse formation is quite poorly understood, and we only know a small handful of factors (like CAMs) involved in the process (see sections 3.2 and 3.3). These factors could potentially be valuable targets for those who endeavor to rewire the cochlea in cases of damage – whether it be through gene therapy, cell replacement strategies, or through drug-based treatments of synapse loss (section 1.3). In addition, the extent to which synaptic specificity between different subdivisions (apex vs. base) or classes (1a, 1b, or 1c) of type I SGNs and inner hair cells is determined by classic differentiation or guidance mechanisms, along with activity, remains to be determined. The mature inner hair cell shows a fascinating presynaptic landscape divisible along its short and long axes in terms of ribbon synapse volumes (Liberman & Liberman 2016) and how these differences are achieved will be of great interest to the field. These issues will certainly be more easily addressed given the newly-available SGN transcriptome data (Petitpré et al. 2018; Shrestha et al. 2018; Sun et al. 2018) and ongoing improvements in experimental mouse technology. In addition, it will be important to determine how other, less well characterized cells, like immune cells, blood vessels, and glia, may affect ribbon synapse formation.