Most plane-polarized tissues are formed by identically oriented cells [1,2]. A notable exception occurs in the vertebrate vestibular system and lateral-line neuromasts, where mechanosensory hair cells orient along a single axis but in opposite directions to generate bipolar epithelia [3][4][5]. In zebrafish neuromasts, pairs of hair cells arise from the division of a non-sensory progenitor [6, 7] and acquire opposing planar polarity via the asymmetric expression of the polarity-determinant transcription factor Emx2 [8][9][10][11]. Here, we reveal the initial symmetry-breaking step by decrypting the developmental trajectory of hair cells using single-cell RNA sequencing (scRNA-seq), diffusion pseudotime analysis, lineage tracing, and mutagenesis. We show that Emx2 is absent in non-sensory epithelial cells, begins expression in hair-cell progenitors, and is downregulated in one of the sibling hair cells via signaling through the Notch1a receptor. Analysis of Emx2-deficient specimens, in which every hair cell adopts an identical direction, indicates that Emx2 asymmetry does not result from auto-regulatory feedback. These data reveal a two-tiered mechanism by which the symmetric monodirectional ground state of the epithelium is inverted by deterministic initiation of Emx2 expression in hair-cell progenitors and a subsequent stochastic repression of Emx2 in one of the sibling hair cells breaks directional symmetry to establish planar bipolarity.

Collective cell rotations are widely used during animal organogenesis. Theoretical and in vitro studies have conceptualized rotating cells as identical rigid-point objects that stochastically break symmetry to move monotonously and perpetually within an inert environment. However, it is unclear if this notion can be extrapolated to a natural context, where rotations are ephemeral and heterogeneous cellular cohorts interact with an active epithelium. In zebrafish neuromasts nascent sibling hair cells invert positions by rotating≤180° around their geometric center after acquiring different identities via Notch1a-mediated asymmetric repression of Emx2. Here we show that this multicellular rotation is a three-phasic movement that progresses via coherent homotypic coupling and heterotypic junction remodeling. We found no correlation between rotations and epithelium-wide cellular flow or anisotropic resistive forces. Moreover, the Notch/Emx2 status of the cell dyad does not determine asymmetric interactions with the surrounding epithelium. Aided by computer modeling, we suggest that initial stochastic inhomogeneities generate a metastable state that poises cells to move, spontaneous intercellular coordination of the resulting instabilities enables persistently directional rotations, whereas Notch1a-determined symmetry breaking buffers rotational noise.

Only a handful of vertebrates are capable of sensing weak electric fields. Two new studies shed light on the development and physiology of electroreceptive organs.