Early development of the cochlea of the common marmoset, a non-human primate model

Fine-tuned cochlear development is essential for hearing. Owing to the difficulty in using early human fetal samples, most of our knowledge regarding cochlear development has been obtained from rodents. However, several inter-species differences in cochlear development between rodents and humans have been reported. To bridge these differences, we investigated early otic development of a non-human primate model animal, the common marmoset (Callithrix jacchus). We examined 20 genes involved in early cochlear development and described the critical developmental steps for morphogenesis, which have been reported to vary between rodents and marmosets. The results revealed that several critical genes involved in prosensory epithelium specifications showed higher inter-species differences, suggesting that the molecular process for hair cell lineage acquisition in primates differs considerably from that of rodents. We also observed that the tempo of cochlear development was three times slower in the primate than in rodents. Our data provide new insights into early cochlear development in primates and humans and imply that the procedures used for manipulating rodent cochlear sensory cells cannot be directly used for the research of primate cells due to the intrinsic inter-species differences in the cell fate determination program.


Background
Auditory perception is the process by which mechanical sound waves are detected by the inner ear and converted into neuronal electrical impulses perceived by the brain. A coiled organ, the cochlea, plays an essential role in hearing during this sequential process in the inner ear. In the cochlea, hair cells convert mechanosensory sound waves into neural electrical pulses, which are transmitted by the synapses between the hair cells and spiral ganglion neurons and eventually reach the brain's auditory cortex [1,2].
Fine-tuned cochlear development is essential for the acquisition of hearing. This development requires several well-controlled steps, including the formation of coiled structure, specification of sensory epithelium, differentiation of hair cells and spiral ganglion neurons, and their subsequent maturation [3,4]. Most of our knowledge regarding cochlear development has been obtained from rodent models, specifically from mice and rats. However, several inter-species differences in cochlear development between rodents and humans have been reported. For example, the expression pattern of PRPH (peripherin), a neurofilament of the spiral ganglion neurons, differs between primates and rodents during cochlear development [5]. In addition, a human-specific pattern was reported for myelination of spiral ganglion neurons [6]. Despite these inter-species differences in cochlear development between rodents and humans, knowledge regarding cochlear development in humans is scarce because of ethical issues in obtaining human fetal tissues.
To bridge these inter-species differences between rodents and humans, the common marmoset (Callithrix jacchus), a primate model, has been exploited and investigated for cochlear development [7][8][9]. This animal model has also been used in studying deafness-related genes [10][11][12][13][14] and presbycusis [15], as well as in other areas of neuroscience [16]. These previous reports on the primate model have demonstrated the usefulness of the animal model in research on hearing.
In particular, previous reports have shown that cochlear development in this model differs from that of rodent models [7,8]. For example, while Na-K-Cl cotransporter 1 (NKCC1) is well-known as a transporter of the stria vascularis and lateral wall, a distinct transient expression of NKCC1 in the organ of Corti during development has been observed in the common marmoset [7]. Distinct and complex changes in the expression patterns of synaptic vesicle exocytosis-related genes in hair cells during the developmental process of this primate have been reported [8]. However, neither have been reported in rodent models.
Previous embryological studies in this primate have focused only on the fetus after embryonic day 96 (E96), a relatively late stage of cochlear development [7][8][9]. Therefore, early development (before E96) of the cochlea in this animal model has not thoroughly been investigated, although several critical developmental processes are known to occur at this early stage. For example, when cochlear duct coils are formed or how sensory epithelial differentiation occurs in this animal are not well understood.
This study examined the cochlea from E70 to E92 of the common marmoset. We investigated important cochlear developmental steps, including formation of the coiled structure, specifications of the sensory epithelium, differentiation of hair cells, and development of neurons.

Specimens
Cadaverous temporal bone samples of common marmosets at E70 (n = 4), E77 (n = 2), E87 (n = 4), and E92 (n = 3) were used in this study. The animal experiments were approved by the Animal Experiment Committee of Keio University (number: 11006, 08,020) and were performed in accordance with the guidelines of the National Institutes of Health and the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

Tissue preparation
The temporal bone was dissected and fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) overnight immediately after euthanasia. Specimens were embedded in Tissue-Tek O.C.T. compound (Sakura Finetechnical Co., Ltd., Tokyo, Japan) for cross-sectioning. Seven-micrometer sections were used for immunohistochemical analysis.

Immunohistochemistry
After a brief wash with PBS, the sections were heated (80 °C) in 10 µM citrate buffer (pH 6) for 15 min. After another brief wash, the sections were pre-blocked for 1 h at room temperature in 10% normal serum in PBS, incubated overnight with primary antibodies at 4 °C, and then incubated with Alexa Fluor-conjugated secondary antibodies for 60 min at room temperature. Nuclei were counterstained with Hoechst 33,258.

Early morphological development of the common marmoset cochlea
The common marmoset has a gestation period of approximately 143-150 days [17]. Previously, we have reported that cochlear duct formation was complete, with apical, middle, and basal turns, in E96 embryos of the common marmoset. In this study, we first examined morphological changes in the cochlea during early development in the common marmoset using hematoxylin-eosin staining (Fig. 1).
In E70 embryos of the common marmoset, the cochlear duct appeared as a hook-like structure, and the coiled structure was incomplete (Fig. 1A). In E77 embryos, at least a one-turn coiled structure was observed (Fig. 1B). At this stage, a premature bony capsule of the cochlea was also observed. In E87, the cochlear duct was elongated, and a two-turn coiled structure was observed (Fig. 1C). At this stage, the structures of the cochlear bony capsules became more prominent. In E92, an almost entirely elongated coil was formed and a two-and a half-turn coiled structure was observed (Fig. 1D). Our observations indicated that the coiled structure of the common marmoset formed before E70 and its elongation continued until E92.

Development of hair cells
Next, we investigated the initial formation of hair cells by examining the expression of hair cell markers, POU4F3 (POU class 4 homeobox 3) [18] and ATOH1 (atonal bHLH transcription factor 1) [19,20].
In common marmosets, neither POU4F3 nor ATOH1 expression was observed in the E77 cochlear duct, while both of their expression was observed in the vestibular organ ( Fig. 2A). In E87, expression of both POU4F3 and ATOH1 was observed in the SOX2 (SRY-box transcription factor 2) positive-prosensory domain (Fig. 2B-D). In the E87 cochlear duct, one row of inner hair cells was observed between the mid-basal and basal turns. In the basal turn, the development of outer hair cells was observed at this stage. In E92, hair cell development proceeded more apically, and both inner and outer hair cell development was observed in the middle turns ( Fig. 2E-G). In contrast, no expression of hair cell markers was observed in the apical turns at this stage. These observations indicated that hair cell formation in the common marmoset started between E77 and E87 from the basal turns and progressed from the base to the apex.
Ventral expression of SOX2 in the cochlear duct was observed in E70; however, CDKN1B and JAG1 were not observed in these SOX2-positive cells (Fig. 3A). In E77, CDKN1B and SOX2 double-positive cells were observed in the apical turns of the cochlear duct, but JAG1 expression was not observed at this stage ( Fig. 3B and C). In E87, JAG1 expression was observed in the medial-ventral region of the cochlear duct (Fig. 3D). These observations indicated that the prosensory domain specification in this animal started at around E77 at the apical turn and spread to the basal turn. In E92, JAG1 expression was observed in the medial side of the SOX2-positive region in the apical turn, whereas it was broadly distributed in the SOX2-positive region in the basal turn ( Fig. 3D-F). These expression patterns of SOX2 and JAG1 with apical to basal gradients were similar to those previously reported in mice [25][26][27]. In summary, as shown in the schematic diagrams in Fig. 4, CDKN1B expression in the SOX2-positive region preceded JAG1 expression in the common marmoset cochlear sensory epithelium specifications.

Early patterning of the cochlear duct of the common marmoset
Early patterning of the cochlear duct, such as the organ of Corti, stria vascularis, and Reissner's membrane, is essential for the subsequent development of each compartment of the cochlear duct. This pattern of cochlear development is known to be fine-tuned by several genes [28][29][30]. Next, we evaluated the early patterning of the cochlear duct in this animal model by examining the expression of several markers, including OTX1 (orthodenticle homeobox 1) [31] and OTX2 (orthodenticle homeobox 2) [31] as dorsal markers, PAX2 (paired box 2) [32] as lateral markers, GATA3 (GATA binding protein 3) [33,34] and ISL1 (ISL LIM homeobox 1) [35][36][37] as ventral markers ( Fig. 5 and 6). In E70 embryo of the common marmoset, GATA3 expression was observed in the ventral portion of the cochlear duct, while OTX2 expression was detected in the dorsal portion ( Fig. 5A and D). ISL1 expression was observed in the ventral portion of the cochlear duct ( Fig. 5B) and PAX2 expression was detected in the lateral portion of the cochlear duct (Fig. 5C). OTX1 expression was detected in the dorsal part ( Fig. 5E). At this stage, expression of DLX5 and PAX8 was not detected in the cochlear duct ( Fig. 5F and G).
In E77 cochlea, GATA3 and OTX2 expression was observed in the ventral and dorsal portion of the cochlear duct (Fig. 6A). In the cochlear duct, ISL1 and PAX2 expression was observed in the ventral and lateral portion ( Fig. 6B and C). OTX1 expression diminished and no expression was detected in E77. In in E87 cochlea. GATA3 and OTX2 expression was observed in the ventral and dorsal portion of the cochlear duct (Fig. 6E). ISL1 Fig. 1 Histological analysis of the developing cochlea in the common marmoset using hematoxylin-eosin staining. (A) A cross-sectional view of the common marmoset cochlear duct in an E70 embryo. Cochlear duct formation had already started at E70. However, the coiled structure appeared only as a hook-like structure. (B) A cross-sectional view of the common marmoset cochlea duct in an E77 embryo. At E77, the elongation and coiling of the cochlear duct were more prominent. At this stage, the apical and basal turns can be distinguished. On the modiolar side, spiral ganglion neuron formation was observed at this stage. (C) A cross-sectional view of the common marmoset cochlea duct in an E87 embryo. At E87, the overall coiled structures of the cochlear duct were well developed and approximately two turns of the coiled structure were observed. The scala vestibuli and the scala tympani were not formed. (D) A cross-sectional view of the common marmoset cochlea duct in an E92 embryo. At E92, the coiled structure of the cochlear duct was almost completely formed. At this stage, the immature scala vestibuli and scala tympani were observed in the basal turns   E). In contrast, JAG1 expression was detected more laterally and broadly in the basal turn (Arrow in F and G). The nuclei were counterstained with Hoechst (blue). Scale bar: 50 µm. OC: organ of Corti expression was observed in the ventral portion of the cochlear duct (Fig. 6F). PAX2 expression was detected in the lateral portion of the cochlear duct and hair cells (Fig. 6G). In E92 cochlea, GATA3 and OTX2 expressions were observed in the ventral and dorsal portion of the cochlear duct (Fig. 6H). ISL1 expression was observed in the ventral portion of the cochlear duct (Fig. 6I). PAX2 expression was detected in the lateral portion of the cochlear duct and hair cells (Fig. 6J).
As shown in schematic diagrams, the expression patterns of these well-established conventional markers, which related to the early patterning of the cochlear development in the rodent models, were well conserved in the common marmoset cochlear development (Fig. 7).

Early development of the spiral ganglion neuron of the common marmoset
Appropriate differentiation of spiral ganglion neurons and their innervation into the cochlear epithelium is essential for proper acquisition of hearing ability. This developmental process of the spiral ganglion neurons is controlled by the fine tuning of expression of transcription factors such as NEUROD1 (neuronal differentiation 1) [38][39][40], POU4F1 (POU class 4 homeobox 1) [41,42], ISL1 [34,36], GATA3 [34,43], and MAFB (MAF bZIP transcription factor B) [44]. Next, we investigated early development of the spiral ganglion neuron of the common marmoset.
In the common marmoset, cochleovestibular ganglion neurons were observed in the E70 cochlea ( Fig. 8A-E). At this stage, the expression of TUBB3 and PRPH can be detected in cochleovestibular ganglion neurons ( Fig. 8 A and D). Both NEUROD1 [38][39][40] and POU4F1 [41,42] expression can be detected in the cochleovestibular ganglion neurons (Fig. 8B). In E70 embryo of the common marmoset, ISL1 expression was observed in the cochleovestibular ganglion neurons (Fig. 8C), whereas no expression of GATA3 and MAFB was detected ( Fig. 8D and E). In the E77 cochlea ( Fig. 8 F-L), spiral ganglion neurons come to be observed apparently and both NEUROD1 and POU4F1 were detected in the spiral ganglion neurons ( Fig. 8G). At this stage, POU4F1 expression being apparent in most of the spiral ganglion neurons; however, NEUROD1 expression was restricted to neurons in the relatively apical turns (Fig. 8G). In contrast, RBFOX3 (RNA binding fox-1 homolog 3) expression was detected at E77 in NEUROD1-negative neurons located in the basal turns (Fig. 8H). After E77, ISL1, GATA3, and MAFB were observed in the spiral ganglion neurons (Fig. 8 J-L). At E87 (Fig. 9), although RBFOX3 expression was broader (Fig. 9B), NEUROD1 expression could not be detected in the spiral ganglion neurons ( Fig. 9E and F). POU4F1 expression was reduced in the basal turns (Fig. 9F). At E92 (Fig. 10), RBFOX3 expression was detected broadly in the spiral ganglion neurons, whereas NEUROD1 and POU4F1 expression was not observed.
As shown in the schematic diagram, changes of the expression patterns of the transcription factors, which are known to be involving neuronal development in mice, were also observed during the development and maturation of spiral ganglion neurons in the development of the common marmoset cochlea (Fig. 11).

Integration of glial cells into the spiral ganglion neuron
Glial cells surrounding the spiral ganglion neurons originate from neural crest cells and migrate to the spiral ganglion [45,46]. After integration of the migrating glial cells into the spiral ganglion, they differentiate to form myelin [47]. Previously, we have observed myelination of the spiral ganglion neurons at E115 [8]; however, migration and integration of glial cells in this primate have not been evaluated. Hence, we investigated the development of glial cells in the spiral ganglion using SOX10 (SRY-box transcription factor 10) and S100B (S100 calcium binding protein B), which have been used for the markers of glial cells derived from migrating neural crest cells in the spiral ganglion neurons [47]. Finally, we investigated the integration of glial cells into the spiral ganglion neurons.
In the common marmoset, SOX10 + /S100B + glial cells [47] were observed at the E70 cochleovestibular ganglion (Fig. 12 A and B). This indicated that the neural crest cells migrated into the ganglion [45,46] before E70 in this animal. At E77 (Fig. 12 C-F), many of the glial cells were located around the spiral ganglion neurons, while several cells were positioned between the neurons. The integration of satellite glial cells was completed at E87 and E92 (Fig. 12 G-L). At E70, SOX2 expression was observed in SOX10-negative cells in the cochleovestibular ganglion; At E77, SOX2 expression in the spiral ganglion was observed only in SOX10 + /S100 + glial cells.

Development of hair cells and specification of the prosensory domain in the developing common marmoset cochlea
Development of hair cells and specification of the prosensory domain following the cochlear elongation is the one of the most essential steps in cochlear development. Our observation of the early cochlear development of the common marmoset identified the time points of these critical steps in this model animal. Combining observation in this study with our previous report, [7] hair cells developments of the common marmoset start between E77 and E87 from the basal turns and the progression reach to the apex turn at E101 (Fig. 13A). Preceding to this hair cells development, the elongation of the cochlear duct is observed until E92 (Fig. 13B).
In the human fetus, elongation of the cochlear duct starts at 4 − 5 weeks of gestation [48,49], and the coiled cochlea can be observed at approximately 9 − 10 weeks of gestation [48][49][50]. In humans, specification of the sensory epithelium is observed at 9 weeks of gestation [51], and the differentiation of auditory hair cells begins at 12-14 weeks of gestation [5,52]. Contrastingly, cochlear anlage becomes evident at E10.75 in mice, the first turn of the cochlea appears at E12, and the completely coiled cochlea is observed at E17 [53]. POU4F3 expression is observed at Fig. 9 Spiral ganglion neurons in E87. Expression of PRPH3, RBFOX3, GATA3, ISL1, and MAFB was observed in the E87 spiral ganglion neurons (A-D). POU4F1 expression was observed in all spiral ganglion neuron turns, more in apical turns than in mid-apical turns. POU4F1 expression was observed in parts of the basal turns of spiral ganglion neurons. This indicated that the expression of POU4F1 decreased in basal turns. NEUROD1 expression was not detected in E87 spiral ganglion neurons (E and F). Nuclei were counterstained with Hoechst stain (blue). Scale bar: 50 µm E15.5 [18] and ATOH1 is expressed in a single row of cells from the base to the mid-base of the cochlea at E14.5 [19]. Combining the results of these previous reports with our present observations, we concluded that the E70 cochlea of the common marmoset was equivalent to the human cochlea at approximately 8 weeks of gestation and the E12 mouse cochlea. Similarly, the E87 cochlea of the common marmoset corresponded to the human cochlea at approximately 12 weeks of gestation and the E15 mouse cochlea.
Specification of the prosensory domain, which is essential for the subsequent formation of the organ of Corti, is an important step in cochlear development. Reports show that this specification developed during the elongation of the cochlear duct both in mice (E12.5 − E14.5) [22,25,54] and humans (9 weeks of gestation) [51]. Next, we investigated the timing of the specification of the prosensory domain in this animal using the cochlear prosensory domain markers, SOX2 [21,22], CDKN1B [23], and JAG1 [23][24][25]. This investigation revealed that prosensory domain specification was started at around E77 at the apical turn and spread to the basal turn until E87 in common marmoset (Fig. 13B).
Unlike that observed in mice, CDKN1B expression was followed by JAG1 expression during the specification of the cochlear prosensory domain in the common marmoset ( Fig. 3 and 4). In mice, JAG1 expression is observed as early as in E12.5 and it overlaps with the SOX2 domain [25]. CDKN1b expression is observed at around E13.5-E14.5 [23,25]. This determination of the prosensory domain via JAG1-mediated Notch1 signal activation occurs via lateral induction [26]. It is known that conditional knockout of Jag1 in the cochlea results in downregulation of Sox2 and Cdkn1b; therefore, prior expression of JAG1 is essential for the expression of SOX2 and CDKN1B in mice [25,55]. However, prosensory domain specification occurred in this conditional knockout mouse, indicating the possibility of compensation of JAG1 by other Notch ligands [25].
Our observation that JAG1 expression was preceded by CDKN1B or SOX2 expression suggested that CDKN1B and SOX2 expression was controlled by other factors such as other Notch ligands, and that the role of JAG1 in controlling the expression of CDKN1B or SOX2 may be less dominant in this primate but needs validation. Until now, the expression patterns of JAG1 and CDKN1B in developing human fetuses have not been reported. Thus, we cannot conclude that this inter-species difference between common marmosets and mice will also be observed between humans and mice. However, our observations suggest the possibility of primate-specific control of prosensory domain specification of the cochlea. Nonetheless, further studies are warranted in the future.
Despite the difference in initial expression patterns and the lateral induction process, expression of JAG1, CDKN1B, and SOX2 in the later process of prosensory domain specifications are well conserved between common marmoset and mice [25][26][27]. In both mice at E14.5 and the common marmoset at E87, JAG1 was expressed mainly in the Kölliker's organ. Thus, JAG1 expression showed only a slight overlap with CDKN1B expression, and most of the prosensory domain, as assessed by CDKN1B expression, was JAG1negative. Along with cochlear development, the JAG1-positive region progressed to the lateral side. In the later stages of development, Notch signaling is believed to divert the development of the prosensory domain into hair cells via lateral inhibition [26]. This preservation of expression pattern may suggest that lateral inhibition of cochlear development is well-preserved in mammals.

Early patterning of the cochlear duct of the common marmoset
For investigating the early patterning of the cochlear duct, we used OTX1 and OTX2 as dorsal markers of the developing cochlear duct, GATA3 and ISL1 as ventral markers, and PAX2 as a lateral marker.
OTX1 and OTX2 encode members of the bicoid subfamily of homeodomain-containing transcription factors, which are required for the normal development of the inner ear [31]. In mouse cochlea, co-expression of OTX1 and OTX2 can be observed in the roof of the cochlear duct until E13, whereas OTX1 expression was not detected in the cochlea at E14 [31]. In contrast, OTX2 expression was observed in the non-sensory area of the roof of the cochlear duct, corresponding to the Reissner's membrane [56]. In the common marmoset, coexpression of OTX1 and OTX2 was detected in the E70 cochlea; however, after E77, only OTX2 expression was observed in the dorsal part of the cochlear duct. Fig. 11 Schema showing the expression patterns of neuronal markers in early neuronal development. The expression patterns of genes involving the early development of spiral ganglion neurons were investigated. In the common marmoset, expressions of NEUROD1 and POU4F1 were observed in E77, and these expressions disappeared as development progressed. In contrast, broader RBFOX3 expression was observed after E87. GATA3 expression and MAFB expression followed ISL expressions in the spiral ganglion neuron of the common marmoset GATA3 encodes a protein that belongs to the GATA family of transcription factors. In the developing cochlea of mouse, GATA3 expression has been reported in the cochlear sensory epithelium and spiral ganglion neurons [33,34]. In the cochlear duct of the E13.5 mouse, GATA3 expression was observed in the prosensory domain and the outer sulcus. ISL1 encodes a member of the LIM/homeodomain family of transcription factors. In the developing cochlea of mouse, ISL1 expression has also been observed in the cochlear sensory epithelium and spiral ganglion neurons [35,36]. Reports show that ISL1 expression is dispensable for the development of the mouse cochlear prosensory region [37]. In the common marmoset, we detected the expression of both GATA3 and ISL1 in the developing cochlear duct. Expression of both was observed in the ventral cochlear sensory epithelium during the periods examined in this study.
PAX2 encodes a paired box protein, which is essential for normal inner ear morphogenesis [32]. In the developing cochlea of mouse, PAX2 expression was reported in the lateral wall of the cochlear duct and presumptive stria vascularis [32]. PAX2 expression in hair cells has also been reported in the developing cochlea of mice [43] and humans [57]. In the common marmoset, we detected the expression of PAX2 in the developing cochlear duct. Relatively broad PAX2 expression was detected in the E70 cochlea; however, PAX2 expression was restricted to the lateral wall of the cochlear duct in E77. In E87 and E97, PAX2 expression was observed on the lateral side of the cochlear epithelium and hair cells.
In addition, we examined the expression patterns of PAX8 (paired box 8) and DLX5 (distal-less homeobox 5). PAX8 encodes a paired box protein and DLX5 encodes a member of the homeobox transcription factor gene family. Both PAX8 and DLX5 are important for the specification of the otocyst and subsequent cochlear development [58,59]. PAX8 expression has been reported in the otic placode, where it cooperates with PAX2 for inner ear morphogenesis and innervation in mice [58]. DLX5 expression has also been reported in developing otic placodes, and it is essential for the morphogenesis of the semicircular canal [59,60]. In the common marmoset, we did not detect the expression of either PAX8 or DLX5 in the developing cochlear duct at E70, whereas they were expressed in some of the vestibular organs ( Fig. 5F and G).
Although the expression patterns of genes investigated in this study were well conserved between the species, inter-species differences were observed in the expression patterns of genes related to prosensory domain specification that proceed simultaneously with this patterning.

Early development of the spiral ganglion neuron of the common marmoset
So far, relatively large inter-species differences between rodents and primates have been reported with respect to the development of spiral ganglion neurons [5,7]. For example, unlike in rodents [5,61], in which PRPH expression is detected in relatively late stages of cochlear development, PRPH was expressed in spiral ganglion neurons prior to hair cell innervation in humans and common marmosets [5,7]. In this study, we investigated the early development of spiral ganglion neurons in the common marmoset.
NEUROD1, a member of the neuroD family of basic helix-loop-helix (bHLH) transcription factors, is also known as one of the proneural genes responsible for the development of neuroectodermal progenitor cells. NeuroD expression has been reported in the developing mouse cochlea [39], and it is essential for neuronal differentiation of spiral ganglion neurons [38,39]. In mice, NeuroD expression was detected in the spiral ganglion in E9.5, and its expression was observed until E13.5 [40]. POU4F1 encodes a member of the POU-IV class of neural transcription factors, POU4F1, also known as brain-specific homeobox/POU domain protein 3A (BRN3A). POU4F1 regulates soma size, target field innervation, and axon pathfinding of spiral ganglion neurons [41]. In mice, POU4F1 expression was observed as early as E9.5, and it was downregulated in the spiral ganglion neurons in E17.5 [42]. RBFOX3 encodes a member of the RNA-binding FOX protein family, which is involved in the regulation of alternative splicing of pre-mRNA. RBFOX3 is also known as the neuronal nuclei (NeuN) antigen, which has been widely used as a marker for post-mitotic neurons [62,63]. RBFOX3 expression has been reported in post-mitotic mature spiral ganglion neurons after E11.5 in mice [42]. The sequential changes in the expression of NEUROD1, POU4F1, and RBFOX3 during cochlear development are well conserved between mice and common marmosets.
Studies have shown that ISL and GATA3 are essential for the proper development of spiral ganglion neurons and sensory epithelium [34]. In the mouse cochlea, Isl1 expression was observed in the inner ear neuronal lineage from embryonic day E8.5 to E9.5, similar to the expression of NeuroD [36]. In contrast, at E9.5, GATA3 expression was detected only in the otic epithelium, and no expression was detected in the neuronal lineage [43]. GATA3 expression was first observed in the selected neuronal lineage at E12.5, which increased all over the spiral ganglion neurons after E13.5 [34]. Previous studies have shown that GATA3 is essential not for differentiation but for the survival of spiral ganglion neurons [34]. MAFB encodes MAFB, a basic leucine zipper (bZIP) transcription factor. MAFB acts downstream of GATA3 to control the differentiation of spiral ganglion neurons and drive the differentiation of the large post-synaptic density characteristic of the ribbon synapse [44]. The sequential expression patterns of ISL1, GATA3, and MAFB were conserved between mice and marmosets.

Integration of glial cells into the spiral ganglion neuron
Finally, we compared the timing of integration of glial cells into the spiral ganglion neuron between the species. In mouse cochlea, SOX10-positive glial cells originating from neural crest cells were observed in the spiral ganglion neurons at E12.75. At this stage, the SOX10-positive glial cells wrapped the spiral ganglion neurons [64]. and also expressed SOX2. As development progressed, the SOX10-expressing cells were integrated throughout the spiral ganglion at E17.5 [64].
The satellite glial cells of the developing cochlea of the human fetus express S100B, which is a member of the S100 family of proteins containing two EF-hand calciumbinding motifs [5]. In humans, both SOX10-and S100Bpositive glial cells can be observed in 9 weeks gestation fetuses [47], which integrate into the spiral ganglion neurons and envelope them between 12 and 14 weeks of gestation in humans [47].
In common marmoset, the integration of satellite glial cells was observed at E87 (approximately 12 weeks of gestation in human and the E15 mouse cochlea). Our observation indicated that glial cell integration of the common marmoset is more similar to human, and this integration was observed before hair cell differentiation in the primate. This timing was relatively earlier than that in mice [64].

Inter-species difference of cochlear development between rodents and primates
The time course of early cochlear development in the common marmoset is summarized in Fig. 14. The essential developmental steps of early cochlear development include elongation of the cochlear duct, specification of the sensory epithelium, and differentiation of hair cells. This schema indicates the temporal relationship between the crucial steps.
This study confirmed that most of the developmental steps of the common marmoset were conserved in humans or mice. However, several inter-species differences were observed as summarized in the schema (Fig. 15). Two of the most prominent inter-species differences observed in this study were the tempo of the cochlear development, and expression patterns of the genes associated with the sensory epithelium specifications.
So far, although the inter-species differences in the tempos of cochlear development have been underestimated, they may be essential for organogenesis from pluripotent stem cells or organ regeneration [65]. In this study, it was observed that cochlear elongation (from E70 to E92) takes about three weeks, and differentiation of hair cells from basal to apical turn takes about two weeks (from E87 to E101). In contrast, in mice, it has been reported that cochlear elongation took about one week (from E10.75 to E17), [53] and hair cells differentiation needed four days (from E13.5 to E17.5) [19]. This comparison indicates that cochlear development is approximately three times slower in the common marmoset than in mice. Overall, the tempo of cochlear development of the common marmoset was more similar to that of humans than that of mice. In general, the body pattern formation of an embryo, known as segmentation, progresses more slowly in humans than in mice [66,67]. This process is governed by an oscillating biochemical process known as the segmentation clock [68]. Recently, the inter-species difference in the segmentation clock has been investigated. Reports have shown that this inter-species difference in the tempo of the segmentation clock is caused by the difference in the rates of biochemical reactions; the synthesis and degradation of the critical molecules are Fig. 15 Schematic diagrams showing early cochlear development comparing common marmoset, mouse, and human. This study confirmed that most of the early cochlear developmental steps of the common marmoset were conserved in humans or mice. Gene expression patterns in hair cells development and neural development were well-conserved between the species. However, several inter-species differences were also observed. Expression patterns of the genes associated with the sensory epithelium specifications (CDKN1B and JAG1) showed more significant inter-species differences between the common marmoset and mouse. In marmosets, the progression of differentiation of hair cells from basal to apical turn is about three times slower than in mice two or three times slower in humans than in mice [66]. Our observation indicates that a similar phenomenon that explains the inter-species differences in the tempos of the segmentation clock may also underlie cochlear development.
The inter-species differences in the expression patterns of prosensory domain-specific genes are another prominent feature of this study. The molecular mechanism of prosensory domain specification and subsequent hair cell differentiation is of considerable interest in the field of hearing research, as regeneration of hair cells retracing the developmental steps may be utilized for designing therapy for hearing loss [69,70]. Our data imply that the procedures used for manipulating rodent cochlear sensory cells cannot be directly used to research primate cells due to the intrinsic inter-species differences in the cell fate determination program. More detailed studies are needed on the species-specific mechanisms that determine sensory cell fate during cochlear development in primates.

Conclusion
We determined the time course of the essential developmental stages of early cochlear development in the common marmoset, a non-human primate model. In addition, we clarified the similarities and differences in the expression patterns of developmental proteins in the cochlea between previously established rodent models and the primate model. Our data will be helpful for studying early cochlear development in primates and human. SOX10: SRY-box transcription factor 10; S100B: S100 calcium binding protein B.