Description

Deciphering the transcriptional dynamics of vestibular stereocilia morphogenesis using CellTrails maps

Ellwanger DC¹, Scheibinger M¹, Avenarius MR², Barr-Gillespie PG², and Heller S¹

¹Department of Otolaryngology - HNS, Stanford University School of Medicine, Stanford, CA 94305, USA; ²Oregon Hearing Research Center and Vollum Institute, Oregon Health and Science University, Portland, OR 97239, USA

Key players in the function of the inner ear are cohorts of mechanosensing organelles, called stereociliary bundles, protruding from the apical surface of sensory hair cells. Bundle growth and maturation involves an orchestration of distinct sequential and overlapping cellular processes whose temporal program of gene expression remains to be elucidated.

We profiled 183 selected genes in 1,008 single cells from the developing chicken utricle. Common multivariate data analysis techniques revealed that expression changes during hair bundle assembly are subtle and intricate to extract. We addressed this limitation by developing CellTrails, an algorithm of high sensitivity for the de novo mapping and visualization of branched cellular trajectories. It embeds cells in a lower dimensional manifold using spectral decomposition of a nonlinear cell-cell association index. Applying competitive learning, CellTrails determines transient and quiescent cellular states, enabling the accurate temporal ordering of cells. Expression dynamics are derived by a dynamic time warping strategy accounting for varying sampling probabilities of longitudinal data points.

We identified two sensory cell subtype-associated trajectories towards hair bundle formation. Examination of mature functional cellular features verified the predicted branching point. In situ hybridizations and immunohistochemistry combined with quantitative measurements of bundle lengths validated the identified spatial and temporal expression patterns on RNA and protein level, respectively. Further, CellTrails maps of scRNA-Seq data from murine utricles unveil cross-species similarities. Comparison of CellTrails to recent trajectory reconstruction approaches shows its superiority in cell ordering and branch identification.