Large-scale annotated dataset for cochlear hair cell detection and classification.

Our sense of hearing is mediated by cochlear hair cells, of which there are two types organized in one row of inner hair cells and three rows of outer hair cells. Each cochlea contains 5-15 thousand terminally differentiated hair cells, and their survival is essential for hearing as they do not regenerate after insult. It is often desirable in hearing research to quantify the number of hair cells within cochlear samples, in both pathological conditions, and in response to treatment. Machine learning can be used to automate the quantification process but requires a vast and diverse dataset for effective training. In this study, we present a large collection of annotated cochlear hair-cell datasets, labeled with commonly used hair-cell markers and imaged using various fluorescence microscopy techniques. The collection includes samples from mouse, rat, guinea pig, pig, primate, and human cochlear tissue, from normal conditions and following in-vivo and in-vitro ototoxic drug application. The dataset includes over 107,000 hair cells which have been identified and annotated as either inner or outer hair cells. This dataset is the result of a collaborative effort from multiple laboratories and has been carefully curated to represent a variety of imaging techniques. With suggested usage parameters and a well-described annotation procedure, this collection can facilitate the development of generalizable cochlear hair-cell detection models or serve as a starting point for fine-tuning models for other analysis tasks. By providing this dataset, we aim to give other hearing research groups the opportunity to develop their own tools with which to analyze cochlear imaging data more fully, accurately, and with greater ease.

An optimized approach to study nanoscale sarcomere structure utilizing super-resolution microscopy with nanobodies.

The sarcomere is the fundamental contractile unit in skeletal muscle, and the regularity of its structure is critical for function. Emerging data demonstrates that nanoscale changes to the regularity of sarcomere structure can affect the overall function of the protein dense ~2µm sarcomere. Further, sarcomere structure is implicated in many clinical conditions of muscle weakness. However, our understanding of how sarcomere structure changes in disease, especially at the nanoscale, has been limited in part due to the inability to robustly detect and measure at sub-sarcomere resolution. We optimized several methodological steps and developed a robust pipeline to analyze sarcomere structure using structured illumination super-resolution microscopy in conjunction with commercially-available and fluorescently-conjugated Variable Heavy-Chain only fragment secondary antibodies (nanobodies), and achieved a significant increase in resolution of z-disc width (353nm vs. 62nm) compared to confocal microscopy. The combination of these methods provides a unique approach to probe sarcomere protein localization at the nanoscale and may prove advantageous for analysis of other cellular structures.

Large-scale annotated dataset for cochlear hair cell detection and classification.

Our sense of hearing is mediated by cochlear hair cells, localized within the sensory epithelium called the organ of Corti. There are two types of hair cells in the cochlea, which are organized in one row of inner hair cells and three rows of outer hair cells. Each cochlea contains a few thousands of hair cells, and their survival is essential for our perception of sound because they are terminally differentiated and do not regenerate after insult. It is often desirable in hearing research to quantify the number of hair cells within cochlear samples, in both pathological conditions, and in response to treatment. However, the sheer number of cells along the cochlea makes manual quantification impractical. Machine learning can be used to overcome this challenge by automating the quantification process but requires a vast and diverse dataset for effective training. In this study, we present a large collection of annotated cochlear hair-cell datasets, labeled with commonly used hair-cell markers and imaged using various fluorescence microscopy techniques. The collection includes samples from mouse, human, pig and guinea pig cochlear tissue, from normal conditions and following in-vivo and in-vitro ototoxic drug application. The dataset includes over 90'000 hair cells, all of which have been manually identified and annotated as one of two cell types: inner hair cells and outer hair cells. This dataset is the result of a collaborative effort from multiple laboratories and has been carefully curated to represent a variety of imaging techniques. With suggested usage parameters and a well-described annotation procedure, this collection can facilitate the development of generalizable cochlear hair cell detection models or serve as a starting point for fine-tuning models for other analysis tasks. By providing this dataset, we aim to supply other groups within the hearing research community with the opportunity to develop their own tools with which to analyze cochlear imaging data more fully, accurately, and with greater ease.

The actin cytoskeleton in hair bundle development and hearing loss.

Inner ear hair cells assemble mechanosensitive hair bundles on their apical surface that transduce sounds and accelerations. Each hair bundle is comprised of ∼ 100 individual stereocilia that are arranged into rows of increasing height and width; their specific and precise architecture being necessary for mechanoelectrical transduction (MET). The actin cytoskeleton is fundamental to establishing this architecture, not only by forming the structural scaffold shaping each stereocilium, but also by composing rootlets and the cuticular plate that together provide a stable foundation supporting each stereocilium. In concert with the actin cytoskeleton, a large assortment of actin-binding proteins (ABPs) function to cross-link actin filaments into specific topologies, as well as control actin filament growth, severing, and capping. These processes are individually critical for sensory transduction and are all disrupted in hereditary forms of human hearing loss. In this review, we provide an overview of actin-based structures in the hair bundle and the molecules contributing to their assembly and functional properties. We also highlight recent advances in mechanisms driving stereocilia elongation and how these processes are tuned by MET.

Myosin motors in sensory hair bundle assembly.

Mechanosensory hair bundles are assembled from actin-based stereocilia that project from the apical surface of hair cells in the inner ear. Stereocilia architecture is critical for the transduction of sound and accelerations, and structural defects in these mechano-sensors are a clinical cause of hearing and balance disorders in humans. Unconventional myosin motors are central to the assembly and shaping of stereocilia architecture. A sub-group of myosin motors with MyTH4-FERM domains (MYO7A, MYO15A) are particularly important in these processes, and hypothesized to act as transporters delivering structural and actin-regulatory cargos, in addition to generating force and tension. In this review, we summarize existing evidence for how MYO7A and MYO15A operate and how their dysfunction leads to stereocilia pathology. We further highlight emerging properties of the MyTH4/FERM myosin family and speculate how these new functions might contribute towards the acquisition and maintenance of mechano-sensitivity.

Structural basis for tunable control of actin dynamics by myosin-15 in mechanosensory stereocilia.

The motor protein myosin-15 is necessary for the development and maintenance of mechanosensory stereocilia, and mutations in myosin-15 cause hereditary deafness. In addition to transporting actin regulatory machinery to stereocilia tips, myosin-15 directly nucleates actin filament ("F-actin") assembly, which is disrupted by a progressive hearing loss mutation (p.D1647G, "jordan"). Here, we present cryo-electron microscopy structures of myosin-15 bound to F-actin, providing a framework for interpreting the impacts of deafness mutations on motor activity and actin nucleation. Rigor myosin-15 evokes conformational changes in F-actin yet maintains flexibility in actin's D-loop, which mediates inter-subunit contacts, while the jordan mutant locks the D-loop in a single conformation. Adenosine diphosphate-bound myosin-15 also locks the D-loop, which correspondingly blunts actin-polymerization stimulation. We propose myosin-15 enhances polymerization by bridging actin protomers, regulating nucleation efficiency by modulating actin's structural plasticity in a myosin nucleotide state-dependent manner. This tunable regulation of actin polymerization could be harnessed to precisely control stereocilium height.

DNMT3A Harboring Leukemia-Associated Mutations Directs Sensitivity to DNA Damage at Replication Forks.

PURPOSE: In acute myeloid leukemia (AML), recurrent DNA methyltransferase 3A (DNMT3A) mutations are associated with chemoresistance and poor prognosis, especially in advanced-age patients. Gene-expression studies in DNMT3A-mutated cells identified signatures implicated in deregulated DNA damage response and replication fork integrity, suggesting sensitivity to replication stress. Here, we tested whether pharmacologically induced replication fork stalling, such as with cytarabine, creates a therapeutic vulnerability in cells with DNMT3A(R882) mutations. EXPERIMENTAL DESIGN: Leukemia cell lines, genetic mouse models, and isogenic cells with and without DNMT3A(mut) were used to evaluate sensitivity to nucleoside analogues such as cytarabine in vitro and in vivo, followed by analysis of DNA damage and signaling, replication restart, and cell-cycle progression on treatment and after drug removal. Transcriptome profiling identified pathways deregulated by DNMT3A(mut) expression. RESULTS: We found increased sensitivity to pharmacologically induced replication stress in cells expressing DNMT3A(R882)-mutant, with persistent intra-S-phase checkpoint activation, impaired PARP1 recruitment, and elevated DNA damage, which was incompletely resolved after drug removal and carried through mitosis. Pulse-chase double-labeling experiments with EdU and BrdU after cytarabine washout demonstrated a higher rate of fork collapse in DNMT3A(mut)-expressing cells. RNA-seq studies supported deregulated cell-cycle progression and p53 activation, along with splicing, ribosome biogenesis, and metabolism. CONCLUSIONS: Together, our studies show that DNMT3A mutations underlie a defect in recovery from replication fork arrest with subsequent accumulation of unresolved DNA damage, which may have therapeutic tractability. These results demonstrate that, in addition to its role in epigenetic control, DNMT3A contributes to preserving genome integrity during replication stress. See related commentary by Viny, p. 573.

Actin at stereocilia tips is regulated by mechanotransduction and ADF/cofilin.

Stereocilia on auditory sensory cells are actin-based protrusions that mechanotransduce sound into an electrical signal. These stereocilia are arranged into a bundle with three rows of increasing length to form a staircase-like morphology that is required for hearing. Stereocilia in the shorter rows, but not the tallest row, are mechanotransducing because they have force-sensitive channels localized at their tips. The onset of mechanotransduction during mouse postnatal development refines stereocilia length and width. However, it is unclear how actin is differentially regulated between stereocilia in the tallest row of the bundle and the shorter, mechanotransducing rows. Here, we show actin turnover is increased at the tips of mechanotransducing stereocilia during bundle maturation. Correspondingly, from birth to postnatal day 6, these stereocilia had increasing amounts of available actin barbed ends, where monomers can be added or lost readily, as compared with the non-mechanotransducing stereocilia in the tallest row. The increase in available barbed ends depended on both mechanotransduction and MYO15 or EPS8, which are required for the normal specification and elongation of the tallest row of stereocilia. We also found that loss of the F-actin-severing proteins ADF and cofilin-1 decreased barbed end availability at stereocilia tips. These proteins enriched at mechanotransducing stereocilia tips, and their localization was perturbed by the loss of mechanotransduction, MYO15, or EPS8. Finally, stereocilia lengths and widths were dysregulated in Adf and Cfl1 mutants. Together, these data show that actin is remodeled, likely by a severing mechanism, in response to mechanotransduction.

The ATPase mechanism of myosin 15, the molecular motor mutated in DFNB3 human deafness.

Cochlear hair cells each possess an exquisite bundle of actin-based stereocilia that detect sound. Unconventional myosin 15 (MYO15) traffics and delivers critical molecules required for stereocilia development and thus is essential for building the mechanosensory hair bundle. Mutations in the human MYO15A gene interfere with stereocilia trafficking and cause hereditary hearing loss, DFNB3, but the impact of these mutations is not known, as MYO15 itself is poorly characterized. To learn more, we performed a kinetic study of the ATPase motor domain to characterize its mechanochemical cycle. Using the baculovirus-Sf9 system, we purified a recombinant minimal motor domain (S1) by coexpressing the mouse MYO15 ATPase, essential and regulatory light chains that bind its IQ domains, and UNC45 and HSP90A chaperones required for correct folding of the ATPase. MYO15 purified with either UNC45A or UNC45B coexpression had similar ATPase activities (kcat = ∼ 6 s-1 at 20 °C). Using stopped-flow and quenched-flow transient kinetic analyses, we measured the major rate constants describing the ATPase cycle, including ATP, ADP, and actin binding; hydrolysis; and phosphate release. Actin-attached ADP release was the slowest measured transition (∼12 s-1 at 20 °C), although this did not rate-limit the ATPase cycle. The kinetic analysis shows the MYO15 motor domain has a moderate duty ratio (∼0.5) and weak thermodynamic coupling between ADP and actin binding. These findings are consistent with MYO15 being kinetically adapted for processive motility when oligomerized. Our kinetic characterization enables future studies into how deafness-causing mutations affect MYO15 and disrupt stereocilia trafficking necessary for hearing.

Mechanotransduction-Dependent Control of Stereocilia Dimensions and Row Identity in Inner Hair Cells.

Actin-rich structures, like stereocilia and microvilli, are assembled with precise control of length, diameter, and relative spacing. By quantifying actin-core dimensions of stereocilia from phalloidin-labeled mouse cochleas, we demonstrated that inner hair cell stereocilia developed in specific stages, where a widening phase is sandwiched between two lengthening phases. Moreover, widening of the second-tallest stereocilia rank (row 2) occurred simultaneously with the appearance of mechanotransduction. Correspondingly, Tmc1KO/KO;Tmc2KO/KO or TmieKO/KO hair cells, which lack transduction, have significantly altered stereocilia lengths and diameters, including a narrowed row 2. EPS8 and the short splice isoform of MYO15A, identity markers for mature row 1 (the tallest row), lost their row exclusivity in transduction mutants. GNAI3, another member of the mature row 1 complex, accumulated at mutant row 1 tips at considerably lower levels than in wild-type bundles. Alterations in stereocilia dimensions and in EPS8 distribution seen in transduction mutants were mimicked by block of transduction channels of cochlear explants in culture. In addition, proteins normally concentrated at mature row 2 tips were also distributed differently in transduction mutants; the heterodimeric capping protein subunit CAPZB and its partner TWF2 never concentrated at row 2 tips like they do in wild-type bundles. The altered distribution of marker proteins in transduction mutants was accompanied by increased variability in stereocilia length. Transduction channels thus specify and maintain row identity, control addition of new actin filaments to increase stereocilia diameter, and coordinate stereocilia height within rows.

Author Correction: Defective Gpsm2/Gαi3 signalling disrupts stereocilia development and growth cone actin dynamics in Chudley-McCullough syndrome.

This corrects the article DOI: 10.1038/ncomms14907.

Bipolar filaments of human nonmuscle myosin 2-A and 2-B have distinct motile and mechanical properties.

Nonmusclemyosin 2 (NM-2) powers cell motility and tissue morphogenesis by assembling into bipolar filaments that interact with actin. Although the enzymatic properties of purified NM-2 motor fragments have been determined, the emergent properties of filament ensembles are unknown. Using single myosin filament in vitro motility assays, we report fundamental differences in filaments formed of different NM-2 motors. Filaments consisting of NM2-B moved processively along actin, while under identical conditions, NM2-A filaments did not. By more closely mimicking the physiological milieu, either by increasing solution viscosity or by co-polymerization with NM2-B, NM2-A containing filaments moved processively. Our data demonstrate that both the kinetic and mechanical properties of these two myosins, in addition to the stochiometry of NM-2 subunits, can tune filament mechanical output. We propose altering NM-2 filament composition is a general cellular strategy for tailoring force production of filaments to specific functions, such as maintaining tension or remodeling actin.

Defective Gpsm2/Gαi3 signalling disrupts stereocilia development and growth cone actin dynamics in Chudley-McCullough syndrome.

Mutations in GPSM2 cause Chudley-McCullough syndrome (CMCS), an autosomal recessive neurological disorder characterized by early-onset sensorineural deafness and brain anomalies. Here, we show that mutation of the mouse orthologue of GPSM2 affects actin-rich stereocilia elongation in auditory and vestibular hair cells, causing deafness and balance defects. The G-protein subunit Gαi3, a well-documented partner of Gpsm2, participates in the elongation process, and its absence also causes hearing deficits. We show that Gpsm2 defines an ∼200 nm nanodomain at the tips of stereocilia and this localization requires the presence of Gαi3, myosin 15 and whirlin. Using single-molecule tracking, we report that loss of Gpsm2 leads to decreased outgrowth and a disruption of actin dynamics in neuronal growth cones. Our results elucidate the aetiology of CMCS and highlight a new molecular role for Gpsm2/Gαi3 in the regulation of actin dynamics in epithelial and neuronal tissues.

Harnessing molecular motors for nanoscale pulldown in live cells.

Protein-protein interactions (PPIs) regulate assembly of macromolecular complexes, yet remain challenging to study within the native cytoplasm where they normally exert their biological effect. Here we miniaturize the concept of affinity pulldown, a gold-standard in vitro PPI interrogation technique, to perform nanoscale pulldowns (NanoSPDs) within living cells. NanoSPD hijacks the normal process of intracellular trafficking by myosin motors to forcibly pull fluorescently tagged protein complexes along filopodial actin filaments. Using dual-color total internal reflection fluorescence microscopy, we demonstrate complex formation by showing that bait and prey molecules are simultaneously trafficked and actively concentrated into a nanoscopic volume at the tips of filopodia. The resulting molecular traffic jams at filopodial tips amplify fluorescence intensities and allow PPIs to be interrogated using standard epifluorescence microscopy. A rigorous quantification framework and software tool are provided to statistically evaluate NanoSPD data sets. We demonstrate the capabilities of NanoSPD for a range of nuclear and cytoplasmic PPIs implicated in human deafness, in addition to dissecting these interactions using domain mapping and mutagenesis experiments. The NanoSPD methodology is extensible for use with other fluorescent molecules, in addition to proteins, and the platform can be easily scaled for high-throughput applications.

Mutational Spectrum of MYO15A and the Molecular Mechanisms of DFNB3 Human Deafness.

Deafness in humans is a common neurosensory disorder and is genetically heterogeneous. Across diverse ethnic groups, mutations of MYO15A at the DFNB3 locus appear to be the third or fourth most common cause of autosomal-recessive, nonsyndromic deafness. In 49 of the 67 exons of MYO15A, there are currently 192 recessive mutations identified, including 14 novel mutations reported here. These mutations are distributed uniformly across MYO15A with one enigmatic exception; the alternatively spliced giant exon 2, encoding 1,233 residues, has 17 truncating mutations but no convincing deafness-causing missense mutations. MYO15A encodes three distinct isoform classes, one of which is 395 kDa (3,530 residues), the largest member of the myosin superfamily of molecular motors. Studies of Myo15 mouse models that recapitulate DFNB3 revealed two different pathogenic mechanisms of hearing loss. In the inner ear, myosin 15 is necessary both for the development and the long-term maintenance of stereocilia, mechanosensory sound-transducing organelles that extend from the apical surface of hair cells. The goal of this Mutation Update is to provide a comprehensive review of mutations and functions of MYO15A.

Heritability of non-speech auditory processing skills.

Recent insight into the genetic bases for autism spectrum disorder, dyslexia, stuttering, and language disorders suggest that neurogenetic approaches may also reveal at least one etiology of auditory processing disorder (APD). A person with an APD typically has difficulty understanding speech in background noise despite having normal pure-tone hearing sensitivity. The estimated prevalence of APD may be as high as 10% in the pediatric population, yet the causes are unknown and have not been explored by molecular or genetic approaches. The aim of our study was to determine the heritability of frequency and temporal resolution for auditory signals and speech recognition in noise in 96 identical or fraternal twin pairs, aged 6-11 years. Measures of auditory processing (AP) of non-speech sounds included backward masking (temporal resolution), notched noise masking (spectral resolution), pure-tone frequency discrimination (temporal fine structure sensitivity), and nonsense syllable recognition in noise. We provide evidence of significant heritability, ranging from 0.32 to 0.74, for individual measures of these non-speech-based AP skills that are crucial for understanding spoken language. Identification of specific heritable AP traits such as these serve as a basis to pursue the genetic underpinnings of APD by identifying genetic variants associated with common AP disorders in children and adults.

The 133-kDa N-terminal domain enables myosin 15 to maintain mechanotransducing stereocilia and is essential for hearing.

The precise assembly of inner ear hair cell stereocilia into rows of increasing height is critical for mechanotransduction and the sense of hearing. Yet, how the lengths of actin-based stereocilia are regulated remains poorly understood. Mutations of the molecular motor myosin 15 stunt stereocilia growth and cause deafness. We found that hair cells express two isoforms of myosin 15 that differ by inclusion of an 133-kDa N-terminal domain, and that these isoforms can selectively traffic to different stereocilia rows. Using an isoform-specific knockout mouse, we show that hair cells expressing only the small isoform remarkably develop normal stereocilia bundles. However, a critical subset of stereocilia with active mechanotransducer channels subsequently retracts. The larger isoform with the 133-kDa N-terminal domain traffics to these specialized stereocilia and prevents disassembly of their actin core. Our results show that myosin 15 isoforms can navigate between functionally distinct classes of stereocilia, and are independently required to assemble and then maintain the intricate hair bundle architecture.

Live-cell imaging of actin dynamics reveals mechanisms of stereocilia length regulation in the inner ear.

The maintenance of sensory hair cell stereocilia is critical for lifelong hearing; however, mechanisms of structural homeostasis remain poorly understood. Conflicting models propose that stereocilia F-actin cores are either continually renewed every 24-48 h via a treadmill or are stable, exceptionally long-lived structures. Here to distinguish between these models, we perform an unbiased survey of stereocilia actin dynamics in more than 500 utricle hair cells. Live-imaging EGFP-ß-actin or dendra2-ß-actin reveal stable F-actin cores with turnover and elongation restricted to stereocilia tips. Fixed-cell microscopy of wild-type and mutant ß-actin demonstrates that incorporation of actin monomers into filaments is required for localization to stereocilia tips. Multi-isotope imaging mass spectrometry and live imaging of single differentiating hair cells capture stereociliogenesis and explain uniform incorporation of (15)N-labelled protein and EGFP-ß-actin into nascent stereocilia. Collectively, our analyses support a model in which stereocilia actin cores are stable structures that incorporate new F-actin only at the distal tips.

Chaperone-enhanced purification of unconventional myosin 15, a molecular motor specialized for stereocilia protein trafficking.

Unconventional myosin 15 is a molecular motor expressed in inner ear hair cells that transports protein cargos within developing mechanosensory stereocilia. Mutations of myosin 15 cause profound hearing loss in humans and mice; however, the properties of this motor and its regulation within the stereocilia organelle are unknown. To address these questions, we expressed a subfragment 1-like (S1) truncation of mouse myosin 15, comprising the predicted motor domain plus three light-chain binding sites. Following unsuccessful attempts to express functional myosin 15-S1 using the Spodoptera frugiperda (Sf9)-baculovirus system, we discovered that coexpression of the muscle-myosin-specific chaperone UNC45B, in addition to the chaperone heat-shock protein 90 (HSP90) significantly increased the yield of functional protein. Surprisingly, myosin 15-S1 did not bind calmodulin with high affinity. Instead, the IQ domains bound essential and regulatory light chains that are normally associated with class II myosins. We show that myosin 15-S1 is a barbed-end-directed motor that moves actin filaments in a gliding assay (∼ 430 nm · s(-1) at 30 °C), using a power stroke of 7.9 nm. The maximum ATPase rate (k(cat) ∼ 6 s(-1)) was similar to the actin-detachment rate (k(det) = 6.2 s(-1)) determined in single molecule optical trapping experiments, indicating that myosin 15-S1 was rate limited by transit through strongly actin-bound states, similar to other processive myosin motors. Our data further indicate that in addition to folding muscle myosin, UNC45B facilitates maturation of an unconventional myosin. We speculate that chaperone coexpression may be a simple method to optimize the purification of other myosin motors from Sf9 insect cells.

Membrane thickness sensitivity of prestin orthologs: the evolution of a piezoelectric protein.

How proteins evolve new functionality is an important question in biology; prestin (SLC26A5) is a case in point. Prestin drives outer hair cell somatic motility and amplifies mechanical vibrations in the mammalian cochlea. The motility of mammalian prestin is analogous to piezoelectricity, in which charge transfer is coupled to changes in membrane area occupied by the protein. Intriguingly, nonmammalian prestin orthologs function as anion exchangers but are apparently nonmotile. We previously found that mammalian prestin is sensitive to membrane thickness, suggesting that prestin's extended conformation has a thinner hydrophobic height in the lipid bilayer. Because prestin-based motility is a mammalian specialization, we initially hypothesized that nonmotile prestin orthologs, while functioning as anion transporters, should be much less sensitive to membrane thickness. We found the exact opposite to be true. Chicken prestin was the most sensitive to thickness changes, displaying the largest shift in voltage dependence. Platypus prestin displayed an intermediate response to membrane thickness and gerbil prestin was the least sensitive. To explain these observations, we present a theory where force production, rather than displacement, was selected for the evolution of prestin as a piezoelectric membrane motor.