Axon guidance cues in auditory development
The innervation of the cochlear sensory epithelium is intricately organized, allowing the tonotopy established by the auditory hair cells to be maintained along the ascending auditory pathways. These auditory projections are patterned by several gene families that regulate neurite attraction and repulsion, known as axon guidance cues. In this review, the roles of various axon guidance molecules, including fibroblast growth factor, ephs, semaphorins, netrins and slits, are examined in light of their known contribution to auditory development. Additionally, morphogens are discussed in the context of their recently described influence on axonal pathfinding in other sensory systems. The elucidation of these various mechanisms may guide the development of therapies aimed at maximizing the connectivity of auditory neurons in the context of congenital or acquired sensorineural hearing loss, especially as pertains to cochlear implants. Further afield, improved understanding of the molecular processes which regulate innervation of the organ of Corti during normal development may prove useful in connecting regenerated hair cells to the central nervous system. Anat Rec Part A, 2006. © 2006 Wiley-Liss, Inc.
The identification of molecules that guide spiral ganglion projections to their central and peripheral targets is critical in furthering our understanding of the fundamental principles behind the establishment of topographic connections in the auditory system. Such signaling cues are in the forefront of genetic therapies being proposed to reestablish the connectivity of auditory neurons in the context of congenital or acquired sensorineural hearing loss. For example, guidance factors have the potential for enhancing the interface between afferent dendrites and the electrode array of a cochlear implant. Additionally, the past few years have seen remarkable progress in the area of hair cell regeneration using both stem cell and gene therapy approaches (Li et al.,2003; Izumikawa et al.,2005). Improved understanding of the molecular processes that regulate innervation of the organ of Corti during normal development will be essential in developing strategies to connect regenerated hair cells to the central nervous system. In light of such rapid progress in this field, the goal of this review is to serve as a state-of-the-art reference regarding molecular guidance of innervation in the auditory periphery.
In recent years, several gene families that regulate neurite attraction and repulsion have been described in developing sensory systems including the netrins (Kennedy et al.,1994; Serafini et al.,1994), semaphorins (Kolodkin et al.,1992; Luo et al.,1993), ephs and ephrins (Cheng et al.,1995; Drescher et al.,1995) and the slit/robo signaling family (Seeger et al.,1993; Kidd et al.,1998).
Already several gene families have already been implicated in guiding auditory innervation, including the neurotrophins BDNF and NT-3 (Ernfors et al.,1992; Pirvola et al.,1992; Ylikoski et al.,1993; Aletsee et al.,2001), fibroblast growth factors (FGFs) (Dazert et al.,1998; Hossain and Morest,2000), ephrins (Bianchi and Gray,2002), and semaphorins (Gu et al.,2003). In addition, morphogens that have recently been noted to participate in axon guidance, including bone morphogenetic proteins (BMPs), sonic hedgehog and wingless/Wnts, may also participate in patterning auditory innervation. Established axon guidance factors such as netrins (Gillespie et al.,2005) and slits and their robo receptors (Webber and Raz,2005) are also expressed in the developing inner ear. This review will provide an overview of the development of topographic innervation in the cochlea and subsequently detail the genes and molecules that define this patterned innervation.
TOPOGRAPHIC INNERVATION OF COCHLEA
Sound perception and frequency discrimination rely on precise topographic projections from the cochlea to the brain. The spiral ganglion neurons are the first “way station” in the processing of auditory information and their peripheral projections to the sensory epithelium are intricately organized. In adults, each type I neuron innervates a single inner hair cell (IHC) via the radial fiber bundle, whereas each type II neuron makes contact with over 30 outer hair cells (OHCs) forming the outer spiral fiber. During development, type I neurons make connections with OHCs as well; these synapses are pruned during the perinatal period (Echteler,1992; Simmons,1994). The neuronal subtypes are differentially distributed across the longitudinal axis of the cochlea with a threefold greater number of type I neurons in the base and a twofold greater number of contacts between OHCs and type II neurons at the apex (Ryugo,1992). The cochlear hair cells are frequency-tuned along the longitudinal axis of the cochlea and the spiral ganglion neurons must preserve this precise peripheral topography in their projections to the cochlear nuclei.
The spiral ganglion neurons are housed in the cochlear modiolus. These neurons are derived from the otic placode, delaminate from the otic vesicle at embryonic day 9 (E9) (Liu et al.,2000; Karis et al.,2001), and undergo terminal mitosis at E11.5 to E15.5 with a peak at E13.5 (Ruben,1967). The spiral ganglion neurons migrate to the ventral otic vesicle and begin to extend dendritic projections to the sensory hair cells in a basal-to-apical gradient, with the earliest fibers arriving at the sensory epithelium at E12.5 (Tello,1931). Generally, axonal projections to the cochlear nucleus emerge shortly after the afferent dendrites.
Various mechanisms have been invoked to account for the topographic innervation of the cochlear sensory epithelium. Carney and Silver (1983) postulated that neuronal precursors leave a trail as they delaminate from the otic vesicle. Their neuronal processes then project back to their area of origin along this path, creating a topographic innervation pattern. Presumably, these tracks would be established by the deposition of extracellular matrix molecules. Several extracellular matrix molecules, including laminin-1, tenascin-C, and fibronectin, have been shown to enhance neurite outgrowth (Hemond and Morest,1991; Woolf et al.,1992; Cosgrove and Rodgers,1997; Legan and Richardson,1997; Whitlon et al.,1999a,1999b). Another theory argues that auditory innervation is initially imprecise and relies on activity-dependent tuning to hone down the afferent connection to their final one-to-one precision. However, other studies have suggested that neural connections between the cochlear sensory epithelium and the brainstem are highly accurate even prior to the onset of hearing (Echteler,1992; Snyder and Leake,1997), suggesting that molecular cues rather than activity regulate the intricate topography of the auditory projections.
AXON GUIDANCE CUES AND GROWTH CONE
Accurate targeting of auditory projections relies on interactions of growth cones on the leading tips of extending neurites with both nondiffusible extracellular matrix molecules and diffusible chemotropic agents expressed along the projected path and at the target epithelium. The end effect of activation of axon guidance cues may vary, alternately providing repulsive or attractive cues to outgrowing axons. In one instance, neurites initially traveling towards a site of slit expression switch gears upon crossing the midline to become repelled by slit (Keleman et al.,2002). This “switch” is mediated by expression of an additional gene. Thus, rather than considering these axon guidance factors as strictly attractive or repulsive, it is more helpful to consider them as providing a combinatorial code that allows for relatively complex patterning of neural circuitry and that will vary depending on the particular signals present in the unique microenvironment.
The growth cone has emerged as an increasingly independent entity, equipped with its own translational machinery (Steward and Schuman,2001; Piper and Holt,2004). This allows an extending neurite to respond rapidly to environmental stimuli that are present along its trajectory. A large number of gene transcripts have been identified within neuronal processes (Eberwine et al.,2001). For example, transcripts for RhoA, a key regulator of the actin cytoskeleton within the extending neurite, are localized to the growth cone (Wu et al.,2005). Sema-3A, a member of the semaphorin family induces intra-axonal translation of rhoA in extending retinal axons, mediating collapse of the growth cone and axonal repulsion. Netrins and the neurotrophin BDNF also induce intra-axonal mRNA translation (Campbell and Holt,2001). Transcripts for the ephrin receptors have been localized to axonal processes as well (Brittis et al.,2002).
LOCALIZATION OF NEUROTROPIC SIGNALS
Evidence exists for release of tropic factors directly from the sensory hair cells (Bianchi and Cohan,1991,1993; Hemond and Morest,1992). Ectopic hair cells, formed as a result of delivery of the gene Atoh1 (also known as Math1), attract neurite outgrowth toward regions of the cochlear epithelium, which are normally nonsensory. However, there is also evidence that guidance cues are not released exclusively from the hair cell targets. Results from in vitro experiments reveal normal afferent ingrowth toward the sensory epithelium even when hair cells have been eliminated (Sobkowicz,1992). Additionally, mice with deletion of the Atoh1 gene fail to develop differentiated hair cells, yet the spiral ganglion neurites grow toward the sensory epithelium in an orderly pattern with limited mistargeting in the immediate vicinity of the sensory epithelia (Fritzsch et al.,2005).
MORPHOGENS AS AXON GUIDANCE MOLECULES
An area of research that is of current investigational interest is the recycling of morphogens as axon guidance cues (for recent reviews, see Schnorrer and Dickson,2004; Bovolenta,2005; Charron and Tessier-Lavigne,2005). Morphogens are secreted proteins produced by a restricted number of cells in a tissue that, when diffused, bring about specific cellular responses along its concentration gradient (Sanchez-Camacho et al.,2005). Morphogens, specifically BMPs, sonic hedgehog (SHH), and wingless/Wnts, were initially thought of as distinct from more traditional axon guidance factors such as netrins, slits, semaphorins, and ephrins. Morphogens act to determine cell fate by influencing nuclear transcription of target genes. Recent studies demonstrate that Shh, BMP, and Wnt also function as axon guidance cues to control growth cone movement in commissural axons in mouse and chick (Augsburger et al.,1999; Butler and Dodd,2003; Charron et al.,2003; Lyuksyutova et al.,2003; Bourikas et al.,2005) and retinal axons in chick (Trousse et al.,2001; Rodriguez et al.,2005; Schmitt et al.,2005). As Bovolenta (2005) has indicated, the robustness of a molecular mechanism is reflected in its conservation, and the secondary role for morphogens as axon guidance cues is not surprising. Questions remain regarding whether growth cones can respond directly to morphogens (Charron and Tessier-Lavigne,2005), the ways in which an individual growth cone can respond to distinct morphogen concentrations (Bovolenta,2005), and whether morphogens work only in long-range guidance or can also exert influences locally (Schnorrer and Dickson,2004). Regardless, as Shh (Trousse et al.,2001) and Wnt (Rodriguez et al.,2005; Schmitt et al.,2005) have been implicated in retinal axon guidance, it is logical to consider that these morphogens may influence guidance of auditory axons. This is an exciting area for future research, as both Wnt (Dabdoub et al.,2003) and Shh (Riccomagno et al.,2002) influence cell orientation and specification, respectively, in the mammalian cochlea. However, the early role that some of these genes play in inner ear morphogenesis often make it difficult to determine their role in later events such as neurite outgrowth using gene deletion strategies. Creative approaches using partial (Gu et al.,2003) or conditional deletions will allow the elucidation of axonal targeting roles for these genes.
FIBROBLAST GROWTH FACTOR
Growth factors bind to receptors on the cell surface with the primary result of activating cellular proliferation and/or differentiation. Fibroblast growth factors have a prominent role in the development of the skeletal system and nervous system in mammals. Additionally, FGFs are neurotrophic for cells of both the peripheral and the central nervous system (Baird,1994). The ways in which FGFs influence the developing auditory system are only beginning to be explored (Dazert et al.,1998; Hossain and Morest,2000; Mueller et al.,2002; Pirvola et al.,2002; Aletsee et al.,2003).
Four cell surface receptors (FGFR1–FGFR4) that bind various FGFs have been identified. Each of these receptors has intrinsic tyrosine kinase activity and autophosphorylation of the receptor is the immediate response to FGF binding (McKeehan and Kan,1994). FGFs also bind to cell surface heparan-sulfated proteoglycans with low affinity relative to that of the specific receptors (Ornitz,2000). The role that FGF/proteoglycan binding plays in FGF signaling pathways is not completely understood but may allow the growth factor to remain associated with the extracellular surface of cells.
FGFs have also been shown to enhance cochlear neuron migration and to promote neurite outgrowth (Dazert et al.,1998; Hossain and Morest,2000). Aletsee et al. (2000) demonstrated that exogenous FGF-1 supports rat spiral ganglion neurite extension in vitro. Dazert et al. (1998) showed an addition role for secreted FGF-1 in branching. Transfected cells that secreted FGF1 induced branching and formed bouton-like terminal swellings on the surface of the transfectants. Aletsee et al. (2003) examined the effects of multiple focal sources of FGF-1 on outgrowing spiral ganglion neurites via explants cultured in the presence of FGF-1 covalently coupled to polybead microspheres. Again, the FGF promoted branching in response to the focal sources of FGF-1, but without terminations on the beads. This divergence from results of the Dazert et al. (1998) study was likely due to additional cell surface factors present in the 3T3 cells and not on the microspheres (Aletsee et al.,2003).
The semaphorins are extracellular cues that were first characterized as repulsive cues (initially called collapsin) (Luo et al.,1993). However, more recent studies have revealed that they can also function as chemoattractants (de Castro et al.,1999; Bagnard et al.,2001). Class 3 semaphorins (Sema) have been most extensively characterized in the auditory literature. This class has six members (Committee,1999). Neuropilin-1 (npn1) is a receptor for Sema3A (He and Tessier-Lavigne,1997). While npn1 binds the sema ligands, an intracellular response is not generated unless neuropilin1 forms a receptor complex with its coreceptor, plexin-A1. Neuropilin receptors also bind vascular endothelial growth factors (VEGFs).
Chilton and Guthrie (2003) report expression of multiple npn1 and multiple Sema genes in the chick otic vesicle at stage 19. Expression of npn1 and Sema3A is localized to dorsal and ventrolateral regions of the otic vesicle, which will go on to form the vestibular structures and the cochlea, respectively. In order to segregate effects of NPN-1/VEGF interactions from NPN-1/Sema interactions, Gu et al. (2003) generated Npn1 knockin mice in which NPN1 interactions with the semaphorins was disrupted without affecting its interactions with VEGFs. NPN1Sema− mice have major targeting errors involving the vestibular afferent fibers.
The netrins are secreted proteins that have been implicated in axon guidance. This conserved family is related to laminins and was first identified in vertebrates as providing an attractive cue that guides commissural axons toward the floor plate (Kennedy et al.,1994; Serafini et al.,1994). The netrins are bifunctional, mediating attraction by binding to receptors of the deleted in colorectal cancer (DCC) family (Stein et al.,2001) and repulsion by binding to UNC5 (Leonardo et al.,1997); however, DCC is required for UNC5-mediated repulsion (Leonardo et al.,1997). Netrin1 is expressed in the prenatal (Livesey and Hunt,1997) and postnatal (Gillespie et al.,2005) cochlea. Netrin1-deficient mice have severe morphogenetic abnormalities involving the vestibular system, with reduced anterior semicircular canals and absence of both the posterior and the horizontal semicircular canals (Salminen et al.,2000). The cochlear anatomy was grossly normal; however, the auditory innervation pathways were not delineated in these mice, nor was auditory testing performed. Therefore, an auditory phenotype cannot be fully excluded.
The neurotrophins have been well characterized with regards to their role in neuronal survival. This trophic function can make it somewhat challenging to establish a distinct role for this family in providing tropic cues. Nonetheless, neurotrophin receptor binding does activate modulators of the growth cone's actin cytoskeleton and modulate growth cone trajectories (reviewed in Huber et al.,2003). Studies in dorsal root ganglion (DRG) neurons reveal that neurotrophins modulate the effect of semaphorins on growth cone collapse, presenting yet another modality by which neurotrophins may modulate neurite targeting (Tuttle and O'Leary,1998).
Two members of the neurotrophin family, brain-derived neurotrophic factor (BDNF) and neurotrophin 3 (NT-3), are expressed in the inner ear and have been demonstrated to promote neurite outgrowth (Ernfors et al.,1992; Pirvola et al.,1992; Ylikoski et al.,1993; Aletsee et al.,2001). Deletion of BDNF primarily results in loss of vestibular neurons, whereas deletion of NT-3 results in a greater than 80% reduction of spiral ganglion neurons (Ernfors et al.,1995; Bianchi et al.,1996). Loss of trk C, the receptor for NT-3, also results in a reduction of spiral neurons, particularly at the base of the cochlea (Fritzsch et al.,1998). Despite the overall reduction in spiral ganglion neurons, studies in mutant mice have demonstrated that dendritic projections of the surviving spiral ganglion neurons are able to reach their targets in the sensory epithelium (Fritzsch et al.,1997,1999).
Elucidating the role of neurotrophins as tropic factors requires creative techniques for parsing out their effect on growth cone trajectories from the effect on neuronal survival. Recently, Hellard et al. (2004) utilized a strategy for avoiding neuronal loss due to deletion of BDNF by crossing BDNF−/− and bax−/− mice. Mice with deletion of the bax gene do not lose neurons in the absence of BDNF. The double homozygous deletion revealed that while some vestibular fibers still targeted appropriately, others projected outside the cristae. Another useful strategy utilized by Tessarollo et al. (2004) is to express BDNF under the control of the NT-3 (now called Ntf3) promoter. This led to expression of BDNF in the basal turn of the cochlea at an earlier time point than its expression in the vestibular structures. Misexpression of BDNF in the cochlea led to a massive mistargeting of vestibular fibers toward the cochlear epithelium. These experiments reveal a significant role for neurotrophins as tropic as well as trophic factors for the developing vestibulocochlear fibers.
The ephrins are membrane-associated ligands for the Eph family of receptors, the largest family of receptor tyrosine kinases (for an excellent review of eph proteins as regards auditory development, see Cramer,2005). Eph/eph signaling can be bidirectional (directing axonal and dendritic projections) as well as bifunctional (attractive or repulsive). Eph/eph signaling is an excellent candidate for auditory axon guidance given its role in guiding topographic projections in other organs of the nervous system. For example, ephs and their ephrin receptors play a vital role in guiding topographic projections from the retinal ganglion cells to the optic tectum (Feldheim et al.,2000; Hindges et al.,2002).
Expression of the ephs and their receptors in the auditory system has led to speculation on the role of this signaling pathway in patterning auditory neuronal connectivity (Lee et al.,1996; Pickles and van Heumen,1997; Bianchi and Gale,1998; Bianchi and Liu,1999; Cramer et al.,2000). Bianchi and Gray (2002) found that ephrin B1 is expressed on developing auditory nerve fibers and that EphB receptors (particularly EphB2) inhibit neurite outgrowth. This is supported by Brors et al. (2003), who found that EphA4 provides repulsive signals, characterized by turning, stopping, and/or reversal, to developing cochlear ganglion neurites mediated through ephrin-B2 and -B3. These data are made even more significant in light of the recent work of Lee and Warchol (2005) in the chick model. Ephrin A2 was expressed in normal chick acoustic ganglion cells; this expression was lost in a subset of those neurons after treatment with gentamicin. Most interestingly, the spatial and temporal pattern of ephrin A2 loss coincided with the pattern of hair cell loss and regeneration. This finding allows consideration of the possibility that ephrin A2 may be involved in the guidance of ganglion cells to regenerated hair cell targets in a tonotopically appropriate manner. Previous work by Person et al. (2004) in the chick found that EphA4 and ephrin-B2 were expressed in gradients along the tonotopic axis of nucleus laminaris during the period of innervation by nucleus magnocellularis. EphA4 and ephrin-B2 expressions were most robust at the high-frequency end of the nucleus and decreased toward the low-frequency end. In contrast, Siddiqui and Cramer (2005) found the spatial distribution of EphA4 immunolabeling corresponded to the low-frequency regions of the central projections of the auditory nerve, and peripherally within the low-frequency areas of the basilar papilla. EphB2 labeling in this study was complementary to EphA4, with the expression demonstrated in high-frequency areas. These gradients of expression are highly suggestive of a role for EphA4 and EphB2 in patterning tonotopic auditory projections.
Thus far, the only Eph deletion to be studied in detail with regards to axonal targeting is the EphB receptor mutant (Cowan et al.,2000). These mice had a subtle error in axonal guidance: transient mistargeting of efferent vestibular fibers. However, this initial mistargeting is rectified within 1–2 days and ultimately these projections do reach the appropriate vestibular targets. This resolution of this mistargeting may be due to the presence of redundant Eph/eph signaling.
Slit/robo signaling has been implicated in the regulation of neuronal migration, axon guidance, and dendritic outgrowth (Hu,1999; Wang et al.,1999; Wu et al.,1999; Erskine et al.,2000; Whitford et al.,2002). The slits are large secreted proteins that bind to the transmembrane robo receptors (Brose et al.,1999; Kidd et al.,1999). Three slit homologs and four robo homologs have been described in mammals (Itoh et al.,1998; Kidd et al.,1998; Huminiecki et al.,2002). Slit proteins not only bind to robos but also to other factors involved in neurite guidance such as laminin and netrin (Brose et al.,1999).
The robo gene was first described in Drosophila as a repulsive cue for axons crossing the midline. Mutations in the robo gene result in multiple recrossings of axons that normally remain ipsilateral (Kidd et al.,1998). Brose et al. (1999) demonstrated that slit and robo function in midline guidance in the mammalian visual system as well. Combined deletion of slit1 and slit2 in mice results in a second optic chiasm forming at a more anterior location, with many fibers opting to enter the opposite eye retrogradely rather than crossing at the normal chiasm (Plump et al.,2002). A mutation in the robo3 gene has been recently identified in humans and results in the absence of crossing of the major sensory and motor projections between the brain and spinal cord (Jen et al.,2004). The repulsive effect of slit on axons appears to be mediated through a collapse of growth cones present on the leading tip of extending neurites (Nguyen Ba-Charvet et al.,1999). Subsequent studies have also revealed a role for slit as a chemoattractant (Wang et al.,1999).
Several studies have established expression of slit and robo genes in the inner ears of chick and rodent models (Holmes et al.,1998; Yuan et al.,1999; Holmes and Niswander,2001; Marillat et al.,2002). In situ hybridization studies revealed expression of slit2, slit3, and robo1 in distinct bands along the neural (slits) and abneural (robo1) regions of the sensory epithelium between E13 and P1 (Webber,2005). Slit2 and slit3 are expressed in complementary regions of the inner sulcus during the period of neurite outgrowth (Gao et al.,2006). Slit2 is also expressed by the cells that divide the spiral ganglion from the inner sulcus. Robo1 transcripts are localized to Hensen's cells and the developing spiral ganglion neurons. The expression pattern of the slit and robo genes in the developing cochlea suggests a model in which robo1-expressing neurites are repelled by slit-expressing inner sulcus cells. Functional studies, including cochlear cocultures with slit-transfected cells and analysis of mice with deletions of the slit genes, will help to delineate the functional role of slit/robo signaling in guiding auditory innervation.
Auditory innervation is likely regulated jointly by several gene families that provide long- and short-range cues to neurites extending toward their hair cell targets. It is possible that the morphogens may account for long-range guidance (Schnorrer and Dickson,2004), whereas the ephs/ephrins, slit/robo may provide short-range cues. Molecules that have well-characterized roles in auditory morphogenesis are now emerging as potential axon guidance cues. Many axon guidance cues are bidirectional (i.e., affect axons and dendrites) and bifunctional (i.e., attractive or repulsive), depending on the microenvironment. Some axon guidance molecules play dual roles as morphogens or trophins, calling for innovative strategies to silence these initial roles so that neurotropic effects can be revealed. Various guidance cues interact promiscuously with members of other axon guidance factors. For examples, robo binds the netrin receptor DCC (Yu et al.,2002) or neurotrophins alter the sensitivity of dorsal root ganglion neurons to semaphorins (Tuttle and O'Leary,1998). The complex interplay of these signaling pathways allows for the tightly regulated patterning of tonotopic innervation that is critical for sound perception and frequency resolution. Exploitation of these signaling pathways has the potential to “molecularly enhance” cochlear implant arrays with better frequency resolution. As hair cell regeneration technology progresses, these guidance cues may also prove useful in connecting these new sensory elements with the central auditory system.