Supervised machine learning for automated classification of human Wharton's Jelly cells and mechanosensory hair cells.

Tissue engineering and gene therapy strategies offer new ways to repair permanent damage to mechanosensory hair cells (MHCs) by differentiating human Wharton's Jelly cells (HWJCs). Conventionally, these strategies require the classification of each cell as differentiated or undifferentiated. Automated classification tools, however, may serve as a novel method to rapidly classify these cells. In this paper, images from previous work, where HWJCs were differentiated into MHC-like cells, were examined. Various cell features were extracted from these images, and those which were pertinent to classification were identified. Different machine learning models were then developed, some using all extracted data and some using only certain features. To evaluate model performance, the area under the curve (AUC) of the receiver operating characteristic curve was primarily used. This paper found that limiting algorithms to certain features consistently improved performance. The top performing model, a voting classifier model consisting of two logistic regressions, a support vector machine, and a random forest classifier, obtained an AUC of 0.9638. Ultimately, this paper illustrates the viability of a novel machine learning pipeline to automate the classification of undifferentiated and differentiated cells. In the future, this research could aid in automated strategies that determine the viability of MHC-like cells after differentiation.

In vivo and in vitro biophysical properties of hair cells from the lateral line and inner ear of developing and adult zebrafish.

Hair cells detect and process sound and movement information, and transmit this with remarkable precision and efficiency to afferent neurons via specialized ribbon synapses. The zebrafish is emerging as a powerful model for genetic analysis of hair cell development and function both in vitro and in vivo. However, the full exploitation of the zebrafish is currently limited by the difficulty in obtaining systematic electrophysiological recordings from hair cells under physiological recording conditions. Thus, the biophysical properties of developing and adult zebrafish hair cells are largely unknown. We investigated potassium and calcium currents, voltage responses and synaptic activity in hair cells from the lateral line and inner ear in vivo and using near-physiological in vitro recordings. We found that the basolateral current profile of hair cells from the lateral line becomes more segregated with age, and that cells positioned in the centre of the neuromast show more mature characteristics and those towards the edge retain a more immature phenotype. The proportion of mature-like hair cells within a given neuromast increased with zebrafish development. Hair cells from the inner ear showed a developmental change in current profile between the juvenile and adult stages. In lateral line hair cells from juvenile zebrafish, exocytosis also became more efficient and required less calcium for vesicle fusion. In hair cells from mature zebrafish, the biophysical characteristics of ion channels and exocytosis resembled those of hair cells from other lower vertebrates and, to some extent, those in the immature mammalian vestibular and auditory systems. We show that although the zebrafish provides a suitable animal model for studies on hair cell physiology, it is advisable to consider that the age at which the majority of hair cells acquire a mature-type configuration is reached only in the juvenile lateral line and in the inner ear from >2 months after hatching.

Electrical tuning and transduction in short hair cells of the chicken auditory papilla.

The avian auditory papilla contains two classes of sensory receptor, tall hair cells (THCs) and short hair cells (SHCs), the latter analogous to mammalian outer hair cells with large efferent but sparse afferent innervation. Little is known about the tuning, transduction, or electrical properties of SHCs. To address this problem, we made patch-clamp recordings from hair cells in an isolated chicken basilar papilla preparation at 33°C. We found that SHCs are electrically tuned by a Ca(2+)-activated K(+) current, their resonant frequency varying along the papilla in tandem with that of the THCs, which also exhibit electrical tuning. The tonotopic map for THCs was similar to maps previously described from auditory nerve fiber measurements. SHCs also possess an A-type K(+) current, but electrical tuning was observed only at resting potentials positive to -45 mV, where the A current is inactivated. We predict that the resting potential in vivo is approximately -40 mV, depolarized by a standing inward current through mechanotransducer (MT) channels having a resting open probability of ∼0.26. The resting open probability stems from a low endolymphatic Ca(2+) concentration (0.24 mM) and a high intracellular mobile Ca(2+) buffer concentration, estimated from perforated-patch recordings as equivalent to 0.5 mM BAPTA. The high buffer concentration was confirmed by quantifying parvalbumin-3 and calbindin D-28K with calibrated postembedding immunogold labeling, demonstrating >1 mM calcium-binding sites. Both proteins displayed an apex-to-base gradient matching that in the MT current amplitude, which increased exponentially along the papilla. Stereociliary bundles also labeled heavily with antibodies against the Ca(2+) pump isoform PMCA2a.

Specializations of intercellular junctions are associated with the presence and absence of hair cell regeneration in ears from six vertebrate classes.

Sensory hair cell losses lead to hearing and balance deficits that are permanent for mammals, but temporary for nonmammals because supporting cells in their ears give rise to replacement hair cells. In mice and humans, vestibular supporting cells grow exceptionally large circumferential F-actin belts and their junctions express E-cadherin in patterns that strongly correlate with postnatal declines in regeneration capacity. In contrast, chicken supporting cells retain thin F-actin belts throughout life and express little E-cadherin. To determine whether the junctions in chicken ears might be representative of other ears that also regenerate hair cells, we investigated inner ears from dogfish sharks, zebrafish, bullfrogs, Xenopus, turtles, and the lizard, Anolis. As in chickens, the supporting cells in adult zebrafish, Xenopus, and turtle ears retained thin circumferential F-actin belts and expressed little E-cadherin. Supporting cells in adult sharks and bullfrogs also retained thin belts, but were not tested for E-cadherin. Supporting cells in adult Anolis exhibited wide, but porous webs of F-actin and strong E-cadherin expression. Anolis supporting cells also showed some cell cycle reentry when cultured. The results reveal that the association between thin F-actin belts and low E-cadherin is shared by supporting cells in anamniotes, turtles, and birds, which all can regenerate hair cells. Divergent junctional specializations in supporting cells appear to have arisen independently in Anolis and mammals. The presence of webs of F-actin at the junctions in Anolis appears compatible with supporting cell proliferation, but the solid reinforcement of the F-actin belts in mammals is associated with its absence.

TRPC3 and TRPC6 are essential for normal mechanotransduction in subsets of sensory neurons and cochlear hair cells.

Transient receptor potential (TRP) channels TRPC3 and TRPC6 are expressed in both sensory neurons and cochlear hair cells. Deletion of TRPC3 or TRPC6 in mice caused no behavioural phenotype, although loss of TRPC3 caused a shift of rapidly adapting (RA) mechanosensitive currents to intermediate-adapting currents in dorsal root ganglion sensory neurons. Deletion of both TRPC3 and TRPC6 caused deficits in light touch and silenced half of small-diameter sensory neurons expressing mechanically activated RA currents. Double TRPC3/TRPC6 knock-out mice also showed hearing impairment, vestibular deficits and defective auditory brain stem responses to high-frequency sounds. Basal, but not apical, cochlear outer hair cells lost more than 75 per cent of their responses to mechanical stimulation. FM1-43-sensitive mechanically gated currents were induced when TRPC3 and TRPC6 were co-expressed in sensory neuron cell lines. TRPC3 and TRPC6 are thus required for the normal function of cells involved in touch and hearing, and are potential components of mechanotransducing complexes.

Oncomodulin identifies different hair cell types in the mammalian inner ear.

The tight regulation of Ca(2+) is essential for inner ear function, and yet the role of Ca(2+) binding proteins (CaBPs) remains elusive. By using immunofluorescence and reverse transcriptase-polymerase chain reaction (RT-PCR), we investigated the expression of oncomodulin (Ocm), a member of the parvalbumin family, relative to other EF-hand CaBPs in cochlear and vestibular organs in the mouse. In the mouse cochlea, Ocm is found only in outer hair cells and is localized preferentially to the basolateral outer hair cell membrane and to the base of the hair bundle. Developmentally, Ocm immunoreactivity begins as early as postnatal day (P) 2 and shows preferential localization to the basolateral membrane and hair bundle after P8. Unlike the cochlea, Ocm expression is substantially reduced in vestibular tissues at older adult ages. In vestibular organs, Ocm is found in type I striolar or central hair cells, and has a more diffuse subcellular localization throughout the hair cell body. Additionally, Ocm immunoreactivity in vestibular hair cells is present as early as E18 and is not obviously affected by mutations that cause a disruption of hair bundle polarity. We also find Ocm expression in striolar hair cells across mammalian species. These data suggest that Ocm may have distinct functional roles in cochlear and vestibular hair cells.