A single-nucleus transcriptomic atlas of human inner ear development

Hearing and balance disorders affect roughly 5 percent of the global population, and for decades the only options were neural prostheses such as cochlear implants or vibrotactile feedback strategies. These technologies improve quality of life but never restore the underlying cellular function. That is now changing, as emerging gene therapies aim to correct the biological defects that cause genetic deafness, with recent multicentre trials for OTOF-related deafness already following patients for up to two and a half years. Yet the success of any gene therapy rests on three pillars of knowledge: knowing which human inner ear cell types express the disease-causing genes, understanding how those cells arise and differentiate during development, and determining whether their regulatory programs can be modulated. On all three fronts, current knowledge has been incomplete.

More than 150 deafness-associated genes have been identified, but their expression across human cell types remains largely descriptive and incompletely mapped. Developmental pathways are well described in animal models, yet species differences may limit their translational relevance to humans. Moreover, no established framework existed for identifying the regulatory programs that control human inner ear cell fate. Single-cell RNA sequencing has transformed our understanding of human developmental biology in organs such as the skin, but in the inner ear field most single-cell studies have focused on the mouse cochlea, leaving human inner ear development, and the distinction between cochlear and vestibular regions, poorly defined. Human pluripotent stem cell-derived inner ear organoids are valuable tools for studying human-specific cell types, but they lack key cell types essential for normal function, particularly the secretory epithelia that maintain the ionic balance of the endolymph.

Now a team led by Wouter H. van der Valk, Shantha Devi Udayappan and Heiko Locher at Leiden University Medical Center, working with colleagues in Belgium and at Boston Children’s Hospital and Harvard Medical School, has addressed these gaps with a resource they call HIEDRA, the Human Inner Ear Development snRNA-seq Atlas. Published in Nature Neuroscience, the atlas is built from single-nucleus RNA sequencing of nine human fetal inner ear samples spanning fetal weeks 7 through 15, a window that captures the critical events of cell type specification, including cochlear elongation into its characteristic spiral shape, the onset of hair cell formation around fetal week 10, and the emergence of structures such as the tectorial membrane, the stria vascularis and Reissner’s membrane. After rigorous quality control, the datasets were integrated into an atlas of 42 annotated clusters containing cochlear and vestibular epithelia, hair cells, mesenchymal populations, neurons, Schwann cells, chondrocytes, melanocytes, endothelial cells, immune cells and cycling cells.

The technical power of HIEDRA lies in its ability to resolve transcriptional identities along known spatial domains. Cochlear epithelial cells segregated along the duct floor into medial, prosensory and lateral domains, destined to form the inner sulcus and interdental cells, the organ of Corti, and the outer sulcus respectively. A similar organization was observed in the cochlear roof, where a medial region gives rise to Reissner’s membrane and a lateral region forms the marginal cells of the stria vascularis. Notably, the gene ANOS1, mutations in which cause hearing loss of unknown mechanism, was highly expressed in this lateral roof region, suggesting that ANOS1 may contribute to pathways involved in stria vascularis development. In the vestibular epithelium, supporting cells separated into ampulla-associated and otolith-associated populations, corresponding to the cristae ampullaris and the utricle and saccule, with differential expression of genes such as NEGR1, KCNIP4 and MECOM reflecting their anatomical origins.

Cochlear and vestibular hair cells, though highly correlated, showed distinct transcriptional signatures. Cochlear hair cells enriched genes including UNC5B and ROBO2, the latter essential for cochlear hair cell innervation, while vestibular hair cells showed enrichment of XIRP2 and GALNTL6, genes also reported in developing mouse cochlear hair cells, suggesting that observed differences may partly reflect spatiotemporal variation in hair cell maturation rather than fully distinct developmental programs. Spiral ganglion-like neurons expressed markers such as PLXDC2, SEMA3A, NTNG1, ZFPM2 and ROR1, whereas vestibular ganglion-like neurons showed enriched expression of PCDH9, TAFA2 and RGS7, consistent with their distinct identities. The mesenchymal compartment revealed otic capsule-like clusters with osteoblast-like gene signatures, otic mesenchyme cell clusters expressing markers such as TBX18, CRYM and the hearing loss-associated genes OTOR, COCH and POU3F4, and shared populations of pericytes, endothelial cells, immune cells, melanocytes and Schwann cells, with melanocytes and endothelial cells increasing over time, consistent with inner ear development.

To validate the atlas, the team generated a separate dataset from a fetal week 12.4 inner ear in which the cochlea and vestibular system were manually separated before nuclei isolation. Jaccard similarity analysis quantified the overlap of gene expression between annotated cell types, and vestibular and cochlear epithelial clusters showed strong concordance. Reference-based mapping of HIEDRA hair cells to the validation dataset confirmed that cochlear hair cells mapped to cochlear hair cells and vestibular hair cells to vestibular hair cells. Cross-species comparisons with publicly available mouse single-cell data from the vestibular crista ampullaris and cochlea reinforced the accuracy of the annotations, and label transfer from a mouse cochlear reference spanning embryonic days 14, 16 and postnatal day 1 increased annotation granularity within the human cochlear epithelium, resolving finer subtypes such as inner and outer pillar cells, Deiters’ cells, inner phalangeal cells, Hensen’s cells and Claudius’ cells.

Beyond descriptive classification, the researchers reconstructed developmental trajectories to infer candidate transcriptional drivers of lineage formation. Applying trajectory inference to cochlear sensory differentiation across pseudotime, they identified three transcriptional states: prosensory, marked by ISL1 and JAG1; sensory, marked by GATA3, CDKN1B, SOX2 and LGR5; and hair cell, marked by INSM1, POU3F4, MYO7A, FGF8 and OTOF. Gene set enrichment analysis along the pseudotime trajectory revealed that PI3K-Akt, ERBB, Wnt, AMPK, Hippo and phospholipase D signaling were active across prosensory and sensory states, while TGF-β and Hippo signaling were suppressed in the hair cell state, indicating that their downregulation may be a necessary step for hair cell fate acquisition. Among the genes nominated by gene regulatory network analysis as candidates connected to the stereocilia program, ANKFN1 emerged as the most robust, showing a significant positive association with a stereocilia gene module score in both cochlear and vestibular hair cells, a finding consistent with the known disruption of Ankfn1 homologs in zebrafish causing vestibular defects.

Mapping disease-associated genes onto HIEDRA cell types revealed selective vulnerabilities. Genes linked to syndromic and nonsyndromic hearing loss, Ménière’s disease, presbycusis, vestibular schwannoma, ototoxicity and tinnitus were cross-referenced with cell populations, and the highest gene burden was observed in the cochlear and vestibular sensory units. However, disorder genes were also highly enriched in the secretory units, which include the cochlear lateral roof cells, vestibular dark cells and melanocytes, cell types that maintain endolymph ionic homeostasis, a critical process for normal inner ear function. The team also identified a potential mechanism linking RIPOR2 mutations, which cause adult-onset hearing loss, to otic mesenchyme cell dysfunction, as RIPOR2 was enriched in both vestibular and cochlear otic mesenchyme subtypes in addition to its known role in stereocilia function. These findings highlight that disorders may arise from dysfunction in both sensory and nonsensory lineages, and that so-called deafness genes may also impact vestibular function.

The most striking mechanistic insight came from the analysis of cochlear lateral roof cells and vestibular dark cells, two poorly characterized secretory populations that regulate potassium homeostasis in the endolymph. Trajectory inference identified early and late transcriptional states in both lineages, and pathway enrichment revealed Wnt, TGF-β and Hedgehog signaling among the pathways enriched in the early state that become less engaged in the late state. Hedgehog signaling stood out because of its central role in early inner ear morphogenesis and its limited prior association with secretory cell specification. Critically, both cell types expressed PTCH1, SMO, KIF7, GLI2 and GLI3 but showed little to no GLI1 expression, which implies the absence of ligand-induced activation, suggesting that repression of Hedgehog signaling may be required for secretory epithelial differentiation.

To test this prediction functionally, the team used human inner ear organoids derived from a GMP-compliant induced pluripotent stem cell line. Previous work had shown that these organoids contain vestibular sensory epithelium but that secretory dark cells do not spontaneously develop. Because sensory and secretory cells develop in parallel in the inner ear, the researchers inhibited Hedgehog signaling with cyclopamine from day 15 to day 30 of organoid differentiation, the window in which vestibular hair cells emerge transcriptionally by day 18 and are detectable by immunohistochemistry by day 30. At day 60, control organoids formed SOX2-positive sensory epithelium but lacked DACH1-positive, OTX2-negative dark cells, whereas cyclopamine-treated organoids developed both SOX2-positive sensory epithelium and DACH1-positive, OTX2-negative vestibular dark cell-like epithelium expressing early ion channel markers including SLC26A4, ATP1B2 and BSND. The treated organoids also showed integration of MLANA-positive melanocytes disrupting the LAM-positive basal membrane, all hallmarks of vestibular secretory function. Single-nucleus RNA sequencing of day 90 cyclopamine-treated organoids confirmed the emergence of dark cell-like populations comprising roughly 22 percent of the otic epithelial cells, compared with roughly 43 percent vestibular supporting cells. These results demonstrate that Hedgehog inhibition is critical for dark cell differentiation in human inner ear organoids, validating the developmental insights from HIEDRA and expanding the model’s capacity to capture the major functional cell types of the human inner ear. The atlas is publicly available through the gEAR portal and the UniApp Dataportal, positioning HIEDRA as both a reference atlas and a mechanistic framework for the developing human inner ear, and illustrating how human developmental atlases can directly inform improved experimental models for mechanistic and translational studies.

Subject of Research: Single-nucleus transcriptomic atlas of human inner ear development

Article Title: A single-nucleus transcriptomic atlas of human inner ear development

Article References: A single-nucleus transcriptomic atlas of human inner ear development. (n.d.). https://doi.org/10.1038/s41593-026-02432-8

Image Credits: AI Generated

DOI: 10.1038/s41593-026-02432-8

Keywords: inner ear, single-nucleus RNA sequencing, transcriptomic atlas, Hedgehog signaling, inner ear organoids, hearing loss, vestibular system, gene therapy, cell fate specification, stria vascularis, single-nucleus, transcriptomic