Dcc Mediates Functional Assembly of Peripheral Auditory Circuits

Proper structural organization of spiral ganglion (SG) innervation is crucial for normal hearing function. However, molecular mechanisms underlying the developmental formation of this precise organization remain not well understood. Here, we report in the developing mouse cochlea that deleted in colorectal cancer (Dcc) contributes to the proper organization of spiral ganglion neurons (SGNs) within the Rosenthal’s canal and of SGN projections toward both the peripheral and central auditory targets. In Dcc mutant embryos, mispositioning of SGNs occurred along the peripheral auditory pathway with misrouted afferent fibers and reduced synaptic contacts with hair cells. The central auditory pathway simultaneously exhibited similar defective phenotypes as in the periphery with abnormal exit of SGNs from the Rosenthal’s canal towards central nuclei. Furthermore, the axons of SGNs ascending into the cochlear nucleus had disrupted bifurcation patterns. Thus, Dcc is necessary for establishing the proper spatial organization of SGNs and their fibers in both peripheral and central auditory pathways, through controlling axon targeting and cell migration. Our results suggest that Dcc plays an important role in the developmental formation of peripheral and central auditory circuits, and its mutation may contribute to sensorineural hearing loss.

families of axon guidance molecules in the cochlea thus prompted us to examine the contribution of Dcc in mouse cochlear development.
We first observed two defects in the cochlea of Dcc mutant mice: disrupted SGN assembly in the Rosenthal's canal, and misrouted afferent fibers of SGNs. By tracing the changes developmentally, we found that E16.5 was the earliest time point when the disruption of SG assembly could be observed, excluding any possible origin of SGN delamination defects. Similar disruption of SGN positioning and misrouting of fibers was also observed in the central auditory pathway towards the cochlear nucleus. In addition, the bifurcation pattern of auditory nerve fibers was disrupted in Dcc mutants. Our results revealed a previously unrecognized role of Dcc in regulating the proper organization of not only SGN cell bodies but also their neurites in the developing auditory system. Such disrupted spatial patterning of SGNs and their fibers could be a causal factor that gives rise to hearing impairments seen in clinical cases involving Dcc mutations.

Result
Gene expression analysis of classical axon guidance molecules in mouse cochlea. Previous studies suggest involvements of classical axon guidance molecules in the developing ear 22 . To understand the expression patterns of axon guidance molecules, we have carried out a RNA-sequencing analysis among defined cochlear cell populations at postnatal stages (P3 to P7): hair cells (HCs), supporting cells (SCs) in the greater epithelial ridge (GER), and spiral ganglion neurons (SGNs) in the spiral ganglion (SG) 26 . Our results demonstrated the presence of previously unstudied guidance molecules in the inner ear (e.g. Sema6A-D, Efna5, EphA1) as well as molecules with known patterns of expression in cochlear regions (Table 1). Consistent with our previous in situ hybridization data 23 , Slit2 transcript was highly enriched in the SC population (Table 1). Further, the expression levels of Nrp2 and Efna5 in SGN and HC populations were consistent with their previously reported expression patterns 25,27 .
From the analysis of all families of classical axon guidance molecules (Table 1), we found that six genes were highly specifically expressed in SGNs. These genes included Dcc, Robo4, Slit1, EphA6, EphA8, and Plxnd1. From our previously generated microarray-based gene profile data also at postnatal stages 23 , we found that among these six genes, Dcc exhibited the highest fold difference between the SGN (non-sensory) and HC (sensory) populations (Fig. 1A).
Expression of Dcc in the developing cochlea. In the auditory system, the expression pattern of Dcc has been characterized in the early embryonic and adult vestibulocochlear ganglion and in the ventral cochlear nucleus [28][29][30] . However, its expression during the peripheral auditory circuit development has not been studied. To verify the expression of Dcc in the cochlea within the crucial developmental time window when the peripheral innervations take place, we performed in situ hybridizations in the cochlear whole-mount and cross-sections at the embryonic stage 17.5 (E17.5). Strong Dcc signals were found in the SG, but not in the organ of Corti (oc) or greater epithelial ridge (ger) (Fig. 1B). Cross-sectional views of E17.5 cochlea revealed that Dcc was also expressed in the Reissner's membrane (rm) and spiral ligament (sl) lateral to the stria vascularis (sva) (Fig. 1C). The overall expression pattern of Dcc transcript was consistent with our gene profiling results ( Table 1). The strong expression of Dcc in the SG suggests that it may play a role in regulating the development of SGNs and their fibers.
Misrouted SGN fibers in the absence of Dcc. Based on the known functional roles of Dcc in the developing nervous system and its highly enriched expression pattern in SGNs, we hypothesized that proper axon guidance may be provided by Dcc signaling during cochlear development. To test this hypothesis, we compared the innervation pattern of SGN fibers in cochlear cross-sections between wild-type and homozygous Dcc mutant (Dcc −/− ) mice. The latter harbors an insertion of a neomycin resistance cassette into exon3 of Dcc gene, which prevents the production of full length proteins 14 . At E14.5, peripheral fibers of SGNs, which would innervate HCs, have not been fully extended out yet in either wild-type or Dcc −/− cochleae (Fig. 1D,G). However, cross-sectional views of Dcc −/− cochleae at E16.5 and E18.5 revealed that ectopic fibers were routed towards the peripheral side of the cochlea, while central auditory nerves did not show any signs of aberrant fiber trajectories (Fig. 1E,F,H,I). High-magnification views of cochlear cross-sections showed exact locations of misrouted SG fibers in ectopic places such as the modiolus (mo) near the Rosenthal's canal (rc) (Fig. 1H,I), spiral ligament (sl), and mesothelial cells of scala vestibule (sv) and scala tympani (st) (Fig. 1J-L), whereas wild-type tissues showed no traces of Tuj1 signals in any of these regions (Fig. 1F).
Misrouted fibers in the peripheral side of the cochlea prompted us to closely examine the pattern of SGN innervation of HCs. At E14.5, when inner radial bundles (i.e. fasciculated afferent/efferent fibers) were not yet visible in a whole-mount preparation, there was no apparent difference in the overall morphology between Dcc −/− and wild-type cochleae ( Fig. 2A,D,G). At E16.5 and E18.5, misrouted fibers originating from the SG region were observed in Dcc −/− but not wild-type cochleae, with some of them apparently overshooting past the sensory epithelium ( Fig. 2B,C,E,F). High-magnification images showed that the misrouted fibers were usually in bundles (Fig. 2H,I). Despite the randomly misrouted fiber bundles, the overall organizational pattern of inner radial bundles was preserved to some degree throughout the whole cochlear length (Fig. 2B,C,E,F).
Next, we examined the SGN innervation pattern in the sensory epithelium. The entire z-stack images of whole-mount cochlea at E18.5 (Fig. 2J,M) were partitioned into two sets, with one consisted of planes from the apex to base of HCs (Fig. 2K,N) and the other from the base of HCs to SGN cell bodies in the Rosenthal's canal (Fig. 2L,O). These images revealed largely normal organization of type I and type II fiber innervation of HCs in the Dcc mutant, as no aberrant fibers were observed around the HC rows (Fig. 2K,N). However, fibers innervating HCs were less densely packed in the Dcc −/− mutant (Fig. 2K,N All the misrouted fibers in the Dcc −/− mutant were observed in the z-stack image representing the space below the HC layers (Fig. 2L,O), indicating a failure of these fibers to enter their target area.
Reduced ribbon synapses due to misrouted afferent fibers in the Dcc −/− mutant. Anatomical studies in the ear have shown that efferent fibers, originating from the superior olivary complex (SOC), descend via the inferior vestibular nerve and eventually travel through the Rosenthal's canal into the organ of Corti 31 . In order to test the possibility of efferent projections contributing to the disorganization of cochlear fibers observed in the Dcc mutant, the pattern of efferent fibers was revealed by depositing a lipophilic dye, DiI, in the path of olivocochlear bundles 32 . No apparent misrouting of efferent fibers was observed in Dcc −/− cochleae ( Fig. 3A-F), indicating that the misrouted fibers revealed by Tuj1 staining (Fig. 3A,D) were all afferent fibers originating from SGN somas. Together, these data suggest that Dcc is important for the proper projection of afferent fibers.

Loss of spatial restriction of SGNs within the Rosenthal's canal in the Dcc −/− mutant. Since
Dcc has been implicated in regulating neuronal migration in various systems such as the spinal cord, forebrain, midbrain, bowel and pancreas 34-37 , we further examined its potential role in regulating SGN migration. In E14.5 Dcc −/− cochlea, SGNs remained within the Rosenthal's canal without any noticeable abnormalities as compared with wild-type cochleae (Fig. 4A,D). However, starting from E16.5 and continuing into E18.5, a number of ectopically located SGNs were distributed throughout the cochlea along both medial-lateral and basal-apical axis, although the majority of SGNs were still correctly located within the Rosenthal's canal (Fig. 4B,C,E,F). We quantified the number of mispositioned SGN cell bodies by staining SGNs with parvalbumin (PV) 23 (Fig. 4G,H). E16.5 and E18.5 mutant cochleae exhibited a comparable number of mispositioned SGNs (Fig. 4I). To determine whether the mispositioned SGNs were possibly dying neurons being separated from the intact SG, we compared the density of apoptotic neurons between wild-type and Dcc −/− cochleae, using anti-Cleaved Caspase3 antibody 38 . There were no enhanced apoptotic signals within or around the SG region ( Fig. 4J-O). However, signs of enhanced apoptosis were visible in the already misplaced SGN cell bodies and fibers in the Dcc −/− mutant (Fig. 4K,N). These data suggest that the partial disassembly of SG in the absence of functional Dcc is not caused by cell death.
The spatial distribution pattern of mis-migrated SGNs appeared to change developmentally. For example, there were significantly fewer mispositioned SGN cell bodies in the lesser epithelial ridge (LER), but more mispositioned SGNs in the greater epithelial ridge (GER), at E18.5 than E16.5 ( Fig. 5A-C). For a systematic analysis of the location of ectopic SGNs, we reconstructed three dimensional (3D) images from confocal z-stacks. The 3D reconstructed images revealed that every single mispositioned SGN cell body ended up in the space dorsal to HC layers regardless of their lateral-medial positions (Fig. 5D-G). The mispositioned SGNs medial to HC rows were located within the spiral lamina canal (slc) where inner radial bundles (irb) passed through (Fig. 5F,G, white and red arrows). The SGN cell bodies in the LER (which is lateral to HC rows) were located on the same dorsal-ventral plane as the inner radial bundles (Fig. 5G, yellow arrow). Therefore, none of the mispositioned SGNs occupied the space within the organ of Corti per se.   Inappropriate SGN exit from the Rosenthal's canal towards the CNS. As a previous report has demonstrated the importance of Dcc in confining spinal interneuron cell bodies and their axons within the central nervous system (CNS) 39 , we hypothesized that the proper organization of the central auditory pathway may also require Dcc. To test this hypothesis, we examined the SG assembly in the Rosenthal's canal in cross-sections of Dcc mutant cochleae. In E18.5 wild-type cochleae, there was no single SGN cell body observed outside the Rosenthal's canal ( Fig. 6A-C,G). However, in the Dcc mutant, some SGN cell bodies drifted away from the Rosenthal's canal along the pathway of the central auditory nerve (can) (Fig. 6D-F,H). The onset of ectopic SGNs in the central auditory nerve pathway (E16.5) was comparable to that in the periphery (Fig. 6I). At E14.5, SGNs were properly confined within the Rosenthal's canal without mispositioned SGNs observed in the central auditory nerve pathway across the cochlea (Fig. 6I).
Mispositioning of SGNs was also evident at the terminal region of central auditory nerve axons in the cochlear nucleus (Fig. 7A-D). This abnormality was significant at as early as E16.5 (Fig. 7E-G), similar to the onset of developmental defects in the periphery (Fig. 4I). In addition, there were aberrant fiber bundles deviating from the main trajectory of the central axons, whereas the wild-type central axons exhibited a compact trajectory (Fig. 7A,C). In the wild-type cochlea, the central axons bifurcated, with the ascending and descending branches projecting towards the anterior and posterior parts of the ventral cochlear nucleus respectively 40 (Fig. 7B). The bifurcation pattern appeared different in the Dcc mutant, with relatively fewer ascending axons into the anterior ventral cochlear nucleus (Fig. 7B,D,E,F). These results demonstrate that the phenotypic defect in the periphery of Dcc mutant cochlea was also duplicated to some extent in the central auditory pathway, with a similar developmental onset. In addition, Dcc may also play a role in regulating bifurcation of the central axons.

Is Ntn1 the interacting ligand of Dcc in the development of cochlear innervation pattern?
Functional interactions between Ntn1 and Dcc to guide CNS axons have been broadly characterized 41,42 , and neural defects in Dcc and Ntn1 knockouts usually resemble each other 14,43 . We reasoned that if Ntn1 was a ligand for Dcc, it would be expressed by SGNs themselves or in the fiber pathways. By in situ hybridization, we found that Ntn1 was not expressed at appreciable levels in the cochlea at E17.5 (Fig. 8A). The strongest expression of Ntn1 transcript was in fact on the Reissner's membrane (rm) near the SG boundary, part of which remained on the tissue after dissection (Fig. 8A). This expression pattern was consistent with a previous report demonstrating an absence of Ntn1 in the SG, but its presence only in the Reissner's membrane 44 . Further examination of peripheral auditory circuit development in Ntn1 knockout 43 cochleae showed no obvious defects of SG innervation patterns or SG assembly as observed in the Dcc mutant (Fig. 8B,C). These results demonstrate that Ntn1 is not present in a spatial manner consistent with a role in regulating SG development via interacting with Dcc, and that different ligands for Dcc are likely responsible.

Discussion
Our study suggests that Dcc mediates the proper organization of SG assembly and its innervation pattern in the developing auditory circuits. Dcc −/− mice exhibited spatial disorganization of SG cell bodies and fibers in both peripheral and central auditory pathways beginning from E16.5. This distinct developmental time point at which the mispositioning of SGN cell bodies occurs, indicates that the observed phenotypic defects are not due to any abnormality in the initial delamination of neuroblasts from the otocyst when forming the SG aggregate, which occurs at around E10 45 . Although the development of central projections occurs earlier than the peripheral, with the central projections reaching the hindbrain at around E11.5 46 , in the Dcc mutant the central and peripheral auditory projections developed defective phenotypes around the same time. This suggests that Dcc affects a developmental process common for the central and peripheral auditory projections.
Previously we reported mispositioning of SGN cell bodies in Slit2 (and Robo) mutant mice 23 . Although this specific defective phenotype appears similar in Slit2 and Dcc mutants, the developmental contexts that trigger this defect may be different between the two mutants, as the onsets of mis-migration of SGNs are not the same. In the Slit2 mutant, mispositioned SGNs are seen starting from E14.5 when SGNs start to extend their peripheral axons towards the organ of Corti, while in the Dcc mutant mispositioning of SGNs is only observable starting from E16.5. Based on the currently known cellular processes during mouse cochlear development, there exists a possibility that the defect in the Dcc mutant is linked to convergent extension, which also begins at around E16 47 . During the convergent extension process, the cochlear duct extends, resulting in significant changes in cochlear length and width 47,48 . SGNs are inevitably engaged in this process as they elongate and migrate with the growing cochlear duct. Rearrangements of the cytoskeleton and cellular interactions via certain surface molecules are required for SGN elongation and migration 49 . The whole process would most likely bring a physical tension upon the assembly of SGNs. Possibly, in the Dcc mutant, the SGN intercellular interactions become less effective to overcome this tension, resulting in the partial disassembly of SG. If this were to be the case, potential Dcc-interacting ligands, possibly also expressed in SGNs, may increase the adhesion between SGNs. This is reminiscent of a previous report that Dcc can mediate the adhesion of spinal commissural neurons to the substrate via substrate-bound netrin-1 proteins 50 . We thus postulate that Dcc is involved in maintaining strong SGN interactions for a proper confinement of their cell bodies within the Rosenthal's canal.
Although the neurites extended from the mispositioned SGNs contributed to the overall pattern of disorganized SG fibers in the periphery, a significant portion of defective fibers seemed to originate directly from the SG (Fig. 2H,I). This suggests that there is a direct effect of Dcc signaling on axon guidance. In addition, there could be a secondary effect from SGN mispositioning on fiber misrouting, since the mislocation may result in a failure of proper spatial and temporal exposure to axon guidance cues. Further investigations will be needed to elucidate mechanisms behind Dcc signaling in the organization of SGN fibers.
All the ectopic SGNs in the periphery and CNS were mis-migrated along the auditory nerves or were largely confined to the nerve pathway. The misrouted SG fibers, on the other hand, were less spatially restricted compared with SGN cell bodies, as they were scattered throughout the cochlea. This difference in distribution pattern between mislocated cell bodies and fibers may be due to limited routes by which SGN cell bodies can travel through. Previous studies have revealed that the cochlea is consisted of a web of connective tissue fibers interwoven into a porous meshwork, which is suitable for the exchange of fluids and molecular substances 51,52 . SGN fibers may have been able to pass through these connective tissues, whereas SGN cell bodies may not.
Since Ntn1 expression site is not even close to SGNs and their fibers, it is unlikely that Ntn1 can play a significant role in the development of SGNs and their fibers. Nonetheless, we cannot exclude the possibility that Ntn1 can interact with Dcc in mediating other aspects of cochlear development, e.g. the morphological development of cochlear ducts. Future work is required to identify the molecular partner of Dcc underlying its regulation of SG organization and their fibers. Based on the extremely weak expression of Ntn1 in the cochlea, it is likely that molecules other than Ntn1 serve as Dcc ligands. In fact, other secreted molecules such as Draxin and CBLN4 have recently been reported to bind to Dcc 53,54 . Whether these molecules can serve as Dcc-interacting partners in SG development awaits to be tested in the future.
As previous studies demonstrate that hearing loss can be caused by a reduction in synaptic innervation of hair cells 55,56 , the diminished ribbon synapses observed in the Dcc mutant raises the possibility that Dcc mutations are underlying some cases of sensorineural hearing loss, which is of great interest to be further validated. Because Dcc −/− animals are embryonic lethal, we could not perform hearing tests on these animals. Perhaps Dcc Kanga strain 4 can be utilized to test the direct connection between hearing loss and Dcc, since the homozygous survives into the adulthood and shows similar brain defects as those reported for the targeted allele of Dcc 4 .
Overall, our results suggest that Dcc plays a role in regulating the proper spatial patterning of SGNs during cochlear development and may contribute to sensorineural hearing loss phenotypes. In the future, it will be of particular interest to examine human patients with mirror movements for hearing deficits. Committees at the University of Southern California (USC). The Dcc (Dcc tm1Wbg ) 14 mouse line was maintained on C57BL/6 background. Dcc homozygous mutants were obtained by crossing heterozygous littermates with the plug day designated as E0.5. Dcc +/− and Dcc −/− embryos were genotyped by genomic PCR (gPCR) using wild-type forward (5′ ggccattgaggttccttt3′ ), wild-type reverse (5′ aagacgaccacacgcgag3′ ) and mutant reverse (5′ tcctcgtgctttacggtatc3′ ) primers. Ntn1 mutant 43 embryos were obtained from Dr. Le Ma.
Tissue dissection, FACS, RNA amplification, and Microarray and RNA sequencing. Preparation of RNA sequencing samples was performed as described previously 26 . P3 to P7 mouse cochleae from Gad1-GFP mice for the purification of SCs in the GER region and PV-Cre;Ai6 mice for the purification of HCs and SGNs were dissected out. SCs are positive for GFP in the Gad1-GFP line, and HCs and SGNs are positive for GFP in PV-Cre;Ai6 mice. The GFP-labeled HC and SGN regions were mechanically separated during the cochlear dissection as there is a clear structural separation between the two regions. The isolated tissues were treated with activated papain (20U/ml; Worthington) for 20 min followed by 2 min crude typsin (0.5 mg/ml, Sigma-Aldrich) to achieve complete dissociation. Then, fluorescence-activated cell sorting (FACS) was performed at the Flow Cytometry Core Facility of USC. Cell suspensions were fed into a BDAriaII sorter and purified using 488 nm laser excitation and a 100-μm cytoNozzle. 2000 cells for each distinct cell population were collected into DMEM plus 10% FBS and pelleted down through centrifuge. RNA was extracted from the collected cells using PicoPure RNA isolation kit (Arcturus). Each RNA sample was then amplified using WT-Ovation Pico amplification kit (Nugen) and sequenced with Illumina HiSeq 2000 (Illumina) at the USC Genomics Center following the manufacturer's instructions. Microarray samples were prepared from P5 to P7 mouse cochleae as described previously 23 . HCs were purified with FASC after staining wildtype cochleae with 5 μM FM1-43 (Invitrogen) or after dissecting tdTomato-labelled HC regions from PV-Cre;Ai14 (with SG trimmed off) mice. There were no differences in fluorescence expression pattern in the cochlea between PV-Cre;Ai14 and PV-Cre;Ai6 animals. For the SGN population, fluorescent SGNs from PV-Cre;Ai14 cochleae, in which the HC region was mechanically trimmed off, were collected. Preparation of RNA samples was similar to that for the RNA-sequencing experiments. Samples were profiled on GeneChip Mouse Genome 430 2.0 Array (Affymetrix) and analyzed within R environment.
Imaging. Confocal z-stack images were obtained using Fluoview1000 (Olympus), projected using National Institutes of Health (NIH) ImageJ and further processed using Inkscape.
Dye labeling. Embryonic mouse head was fixed in 4% PFA for at least overnight. Cochlea was exposed by removing the inner ear bones and a small crystal of lipophilic carbocyanin dye 1,1 ′ -dioctadecyl-3,3,3 ′ ,3′tetramethylindocarbocyanine perchlorate (DiI, Molecular Probes TM ) was placed in the mid turn of the cochlea and incubated at 37 °C in PBS for 4 days. Cochlear nuclei were then sectioned with a sagittal plane cut of the fixed head at 300 μm thickness to include the entire cochlea nucleus in one section. Dissected pieces of cochlear nucleus were then cleared with Clear T57 and imaged with confocal microscopy. For DiI labeling of efferent fibers, small piece of DiI crystal was inserted into the olivocochlear fiber pathway and incubated at 37 °C in PBS for 3 weeks. Cochlea was dissected out and imaged whole-mount with confocal microscopy. Image analysis and statistics. All the confocal image stacks were processed in Fiji (ImageJ) software and 3D reconstructions were performed using Olympus FV10-ASW 3.1 software. The hair cell layers and boundaries of SG were identified based on the immunostaining of Myo6 and Tuj1, respectively, from the image series of the confocal Z-stack. For calculating area occupancy by fibers, whole-mount immunostained z-stacks from the apex to base of HCs were converted to binary images. Pixels of outlined objects in binary images were then counted and subtracted from the pixel number of the total area. Significance was tested using two-tailed Student's t-test.
Mispositioned SG cell bodies were identified and quantified based on SGN cell body morphology by DiI labeling or PV/Tuj-1 immunostaining. For quantification of number of mispositioned SGNs along the central auditory pathway, a circular boundary with 200 μm radius was set, with the center positioned at the most medial edge of the Rosenthal's canal, and any SGNs within this range were counted. For counting ribbon synapses on inner hair cells, whole-mount cochleae stained with anti-CtBP2 were used to acquire confocal z-stacks with 0.3 μm intervals. The total number of ribbon synapses was divided by the total number of IHC nuclei (stained with anti-CtBP2) to obtain the average number of synapses per IHC. Significance was tested using two tailed Student's t-test.