Generation of human otic neuronal organoids using pluripotent stem cells

Abstract Otic neurons, also known as spiral ganglion neurons (SGNs) in mammalian cochlea, transmit electrical signals from sensory hair cells to cochlear nuclei of the auditory system. SGNs are sensitive to toxic insults, vulnerable to get irreversible damaged and hardly regenerate after damage, causing persistent sensorineural hearing loss. Yet, to get authentic SGNs for research or therapeutic purpose remains challenging. Here we developed a protocol to generate human otic neuronal organoids (hONOs) from human pluripotent stem cells (hESCs), in which hESCs were step‐wisely induced to SGNs of the corresponding stages according to their developmental trajectory. The hONOs were enriched for SGN‐like cells at early stage, and for both neurons and astrocytes, Schwann cells or supporting cells thereafter. In these hONOs, we also determined the existence of typical Type I and Type II SGNs. Mature hONOs (at differentiation Day 60) formed neural network, featured by giant depolarizing potential (GDP)‐like events and rosette‐organized regions‐elicited calcium traces. Electrophysiological analysis confirmed the existence of glutamate‐responsive neurons in these hONOs. The otic neuronal organoids generated in this study provide an ideal model to study SGNs and related disorders, facilitating therapeutic development for sensorineural hearing loss.

While there are limited replacement or rescue therapies for SGNs dysfunction.
Efforts have been made to maintain and improve long-term survival and function of cultured cells, especially peripheral neurons in vitro, included but not limited to the integration of bio-materials, conductive substrates and induction of neurotrophic factors into the culture systems. 4 with 20 days of neuronal maturation stage. 13 The 3D organisation of differentiated cells has been recognized to better recapitulate in vivo counterparts compared with the 2D manner. 14 For inner ear cells especially hair cells differentiation, Koehler et al. described a 3D chemically defined differentiation procedure, which finally produced well-functioned hair cells from human pluripotent stem cells (hPSCs), and neurons were not further identified. 15 Another differentiation protocol that generate inner ear cells was developed recently with a 2D combined with 3D method, which generate SGN-like cells and hair cell-like cells. 16 SGN-targeted otic organoids have not been developed yet.
The present study generated hPSC-derived otic neuronal organoids (hONOs) harbouring SGN-like cells and glial cells using a de novo 3D protocol without physical transition of formats after the initiation of differentiation (e.g., 2D to 3D). We identified that the hPSC-derived aggregates went through non-neural ectoderm (NNE),

| Immunohistochemistry immunofluorescence
Immunohistochemistry monolayer-cultured cells and 14-μm cyosections were immersed in 4% paraformaldehyde (PFA) for 15 min at room temperature. After three washes with PBS, samples were blocked in 5% bovine serum albumin (BSA; Sigma), 1% donkey serum and 0.1 Triton X-100 (Sigma) in PBS, naming block solution 1, for 1 h at room temperature. Cells were incubated in block solution 1-diluted primary antibodies at 4 C overnight (see Table S1 for details). After three washes of PBS with 0.1% Triton X-100 in PBS at room temperature for 2 h, samples were incubated in block solution 1-diluted secondary antibodies (Invitrogen; 1:1000) and Hoechst 33342 (Sigma; 2 μg/mL) at room temperature for 90 min followed by 1-h wash with PBS at room temperature. Immunostaining images were captured via a confocal microscope (ZEISS LSM 880).

| Morphological subtyping of neurons and glial cells in mature hONOs
To determine the morphology and quantification of neurons and glial cells in well-functioned hONOs, samples were dissociated in Accutase at 37 C for 10 min, centrifuged at 2000 rpm for 5 min, re-suspended and cultured for additional 24 h under 2D condition. Then monolayercultured cells were fixed for the following immunochemistry procedures. Neurons were labelled using TUJ1, glial cells using S100B, and supporting cells using SOX2 primary antibodies. Cell counting was performed using ImageJ software.

| Calcium imaging
To evaluate the calcium activity, whole hONOs at d60 were labelled with 1 μM Fluo-4AM in NIM at 37 C for 1 h according to the recommended instructions. Then calcium fluorescence for each sample was visualized on a ZEISS LSM 880 inverted confocal microscope at room temperature no more than 2 h. Fluo-4AM was excited at 488 nm and time-lapse images were captured every 5-10 s. Glutamatergic neurons were activated by 10 μM glutamic acid (Sigma) for 5 min at room temperature. Calcium responses were quantified by relative fluorescence intensity (ΔF/F) and ΔF/F value ≥0.4 was considered to be calcium-responsive.

| Multi-electrode array assay
Single-well plates with 256 TiN electrodes (spacing 200 μm and diameter 30 μm; Multichannel Systems) were used to assess neural network developed in d60 hONOs. After hONOs were mounted on 1% Matrigel-precoated multi-electrode array (MEA) well and cultured for 72 h, MEA recording was performed on a MEA2100 system using the equipped Multi Channel Experimenter software (Multichannel systems). To assess the glutamate reactivity of hONOs, 10 μM glutamic acid (Sigma) was added to NIM and incubated at 37 C for 2 min. Recordings were analysed using Multi Channel Analyser software and electrodes detected at least 5 spikes/min were considered as active.

| Statistical analysis
hONOs were performed at least for three differentiation series derived from H9 and CTS-Q2 ESCs. Data were shown as mean ± SEM. Statistical analysis was performed on Prism 9 GraphPad software using two-tailed unpaired t-test for two groups or one-way ANOVA using Dunnett's multiple comparison for three and more groups, which were indicated in corresponding figure legends. p < 0.05 was assumed to be statistically significant.

| Step-wise induction of NNE-and PPE-like cells
Previous studies illustrated that BMP activation and TGF-β inhibition promised efficient NNE induction of hPSCs in 2D 13 and 3D differentiation patterns. 15 After 3 days of NNE induction, expression of NNE marker genes TFAP2A and DLX3 in aggregates were significantly increased with neglectable expression of pluripotency marker genes, suggesting cells at this stage might be NNE-like cells (Figure 2A

| Identification of ONPs
The developmental stage of SGNs after PPE is ONP. At the end of PPE induction (d8), some SOX1-positive cells emerged and expression of SOX1 significantly increased compared with that on d0, which reminded us to further make sure the identity of this cell population at this stage. Astonishingly, expression of FOXG1 and PAX8 (otic markers), was the highest between d0 to d25 ( Figure 3B). And expression of JAG1, another otic marker, was kept at high level from d3 to d25 ( Figure 3B). These transcriptome results suggested us that ONP stage might start from d8. The positive staining of otic markers (FOXG1, PAX8, GATA3, and JAG1) in d8 hONOs derived from H9 and CTS-Q2 ESCs substantiated the initiation of otic destination from d8 ( Figure 3C-G). On d25, a relatively large amount of ONPs (PAX8 for otic identity, SOX1 and TUJ1 for neuronal progenitor) declared that ONP stage might not terminate at once (Figure 3H,I).

| Neurons in hONOs resemble SGNs cellularly and morphologically
Neuronal progenitors needed to differentiate into mature neurons for physical performances ( Figure 4A). On d30 (5 days after ONP  in d60 hONOs (Figures 4F and S2c). These above results indicated that d60 hONOs maintained mature SGN-like cells. Systematic evaluation of genes about mature neurons, synaptogenesis, glutamatergic neurons, and SGN identity gave out the conclusion that d60-d150 might be the stage when neurons function well ( Figure 4G). Cells dissociated from d90 hONOs were cultured in monolayer pattern to observe the morphology of neurons. Neurons were classified according to polarity and neurite length, where half neurons were bipolar and length of neurites between short and long neurites is significantly different (Figure 4H-N).

| Glia prosperity at the late stage of hONOs generation
Bright field observation of hONOs showed that on d300 and afterwards, hONOs presented many protrusions at the border ( Figures 1A,   5A, and S1a). Considering that mature neurons do not proliferate, we

| Transcriptome identification of pattern profiles during hONO generation
To evaluate the differentiation profiling, we performed bulk RNA-seq of hONOs sampled on different time points, from d0 to d500 ( Figure 6A). PCA and cluster analysis indicated the phase overlap of d60-d90, d3-d25, d300-400 ( Figures 6B and S3). Vast expression of pluripotent markers characterized the stem cell in d0 aggregates. NNE and PPE marker genes were highly expressed from d3 to d25, and ONP marker genes highly expressed until d30. Most neuronal marker genes were enriched on d30-d150 and glia cells on d150-d500 ( Figure 6C). On d60-d150, markers of different kind of neurons (especially glutamatergic neurons) and SGN-specific markers as well as subtyping markers, were enriched expressed ( Figure 6D,E). From d60 and on, markers of glutamatergic receptors as well as neuronal functionrelated genes were enriched identified ( Figure 6F). hONOs at d150-d500 featured as enrichment of genes related with gliogenesis which might partially explain the prosperity of glial cells at this stage ( Figure 6G). From d25 and on, markers of otic mesenchymal cells were enriched and markers of matrix component were enriched from d150 and on ( Figure 6H). To summarize, hONOs are featured as neurogenesis during d60-d150, and as gliogenesis after d150. As the morphology of hONOs changed distinctly, we hypothesized that cellular composition as well as function of hONOs at this time point were completely different from the former time points. We compared the transcriptome profiles of hONOs from d300 with that from d60. Compared with d60, 7449 genes were differently expressed in hONOs of d300 ( Figure S3b

| Neuronal network is well constructed in hONOs
The above results revealed that hONOs (d60-d150) might mainly focused on neuronal function. Next we confirmed the existence of functional neurons and neural network in d60 hONOs via calcium imaging and MEA. There were extensive calcium-responsive regions in spontaneous hONOs, suggesting the presence of functional neurons ( Figure 7A; Video S1). Interestingly, we found many rosette-like calcium-responsive regions under spontaneous conditions ( Figure 7B; Video S2). The time-lapse imaging showed a population of calcium traces (CTRs) that fired in a specific rhythm, that is, the core region fired to peripheral regions at different directions sequentially ( Figure 7B). The CTRs fired in core region (ROI1) were similarly

| DISCUSSION
Until recently, the elucidation of SGNs development and degeneration was hindered due to the lack of human SGNs materials. The present study described a de novo 3D differentiation protocol that differentiates hESCs into otic neuronal organoids majoring in SGN-like cells.
According to the developmental trajectory of SGNs in mammalian inner ear, NNE is the beginning of otic induction. 15 Then NNE develops into PPE and SGN precursors. 17 We sequentially identified NNE, PPE, ONP marker genes expressed in hONOs at corresponding differentiation time points, indicating that the final neurons belong to otic lineages.
Mammalian SGNs in the cochlea contains two basic types, where 95% are Type I SGNs that connect to inner hair cells and 5% are Type II SGNs that connect to outer hair cells. 18,19 Type I SGNs are all bipolar neurons and connect with one inner hair cell. Type I SGNs differ in sound sensitivity and spontaneous electrical activity. [20][21][22][23] Type II SGNs are pseudounipolar neurons and extend long neurites to receive signals from multiple outer hair cells. 24 Previous studies illustrate that Type II SGNs play critic roles in noise-induced sound sensation. 25,26 In this study, half of neurons in mature hONOs were bipolar, proving the SGN identity from the aspect of morphology. In the transcriptome analysis, we found the enriched expression of SGN marker genes as well as markers specific for Type I and Type II SGNs.
Glial cells are shown to be essential to the survival, neural network, and refinement of neurons. 27 Zafeiriou et al. reported the facilitation of glial cells to neuronal maturation in brain organoids. 28 In the mammalian inner ear, there are various types of glial cells, such as Schwann cells, satellite glial cells, supporting cells, astrocytes and so on. 29,30 In previous inner cell differentiation protocols, gliogenesisrelated growth factors were seldom induced in order to produce more hair cells and neurons. 15,16 In our study, to generate sufficient glial cells, we added cytokine CNTF to the medium, which is reported to promote astrogliogenesis and produce astrocytic progenitors from neural stem cells, at the early maturation of neurons. 31 To our expectations, diversified glial lineages, including supporting cells, astrocytes and Schwann cells were enriched in hONOs at late differentiation stage. In the morphologic analysis of glial cells and neurons, some S100B + SOX2 + glial cells extended dendrites along with neurites, suggesting that this type of glial cells might served as Schwann cells, which are related with the myelination of SGNs. 15 On the other hand, the short and expanded dendrites of another kind of S100B + SOX2 + cells did not go with neurites, indicating that they might be supporting cells. The enrichment of glial lineages makes hONOs be a useful source for glia-related studies about inner ear disorders.
Synaptogenesis is acknowledged to compromise the neuron maturation. 32,33 In our study, we determined that synaptogenesis was initiated at d30, and continued to increase until d150, which reminded us that neurons at this stage might form connectivity and exert plasticity. We detected morphological neural network at d60 and calcium activity analysis gave out more solid results about the construction of neural network, including GDP-like events and rosette-organized regions-elicited CTRs. GDP events have been reported to be commonly detected in murine and human brain, marking the development of neural network. 34,35 Interestingly, we found a new pattern of calcium responses in spontaneous condition, that is, the rosetteorganized regions-elicited CTRs. We found the core ROI of rosette manipulated the pattern of CTRs elicited by peripheral ROIs, resembling the signal transduction between neighbour neurons. This phenomenon consistently appeared in all detected hONOs at d60-d90 (n = 4). Further exploration of this phenomenon would be performed in hONOs at all time points. In addition, glutamate responses were educed in mature hONOs evidenced by enhanced frequency and amplitude in consistent with glutamate-triggered performances of murine SGNs explants as previously described. 36 The differentiation protocol was also reproduced using a clinic-grade pluripotent stem cells which added up to its applicability for clinical requirements.
In the present study, we did not characterize SGN subtypes using molecular markers which needed to be completed in the future study. Besides, we only determined the neural function of hONOs at limited time points, and those at the early and late stages are not mentioned.
The further functional analyses including connection with hair cells and cochlear nucleus as well as drug responses needed to be performed.

| CONCLUSIONS
The present study described a de novo 3D differentiation protocol that differentiates hESCs into otic neuronal organoids majoring in