Control of stereocilia length during development of hair bundles

Fluorescence microscopy <br> Structured illumination (SIM) images were acquired at 26-30°C with a 63X 1.4 NA oil immersion lens on a Zeiss lattice-based Elyra 7 microscope with dual PCO.edge 4.2 sCMOS cameras for detection. Grid selection and z-spacing was guided by the software and kept consistent across images. Post-acquisition processing was performed with software-recommended standard filtering for the 488 channel, without baseline subtraction and with “scale to raw” checked. Contrast was manually adjusted to retain both dim and bright structures due to the high dynamic range of the phalloidin signal. Verification of channel alignment and measurement of the microscope point-spread function was carried out as previously described. <br> Airyscan images were acquired at room temperature using a 63x, 1.4 NA Plan-Apochromat objective on a Zeiss 3-channel LSM980 laser-scanning confocal microscope equipped with an Airyscan.2 detector and run under ZEISS ZEN (v3.1, 64-bit software; Zeiss) acquisition software. Settings for x-y pixel resolution, z-spacing, as well as pinhole diameter and grid selection, were set according to software-suggested settings for optimal Nyquist-based resolution. Raw data processing for Airyscan-acquired images was performed using manufacturer-implemented automated settings. Display adjustments in brightness and contrast and reslices and/or maximum intensity Z-projections were made in Fiji/ImageJ software. <br> For cochlea imaging, for each stain, 2-4 images were acquired from 1-2 cochlea per genotype per age for each experiment, and experiments were conducted three to four times. Ears from control and mutant littermates or from different ages of C57BL/6J mice (of both sexes) were stained and imaged on the same days for each experiment to limit variability. Genotyping was performed either prior to dissection or performed on tails collected during dissection for younger animals ( <br> Image rendering and measurements of stereocilia dimensions <br> Surface rendering was carried out with Imaris software (imaris.oxinst.com; versions 9.7.0, 9.8.0, 9.9.0; Oxford Instruments); the rendered surfaces were stored as sparse voxel octrees, which allowed them to be manipulated and measured using Imaris. A batch processing file was created in Imaris and applied to each dataset. Within each batch, background subtraction was first performed in the image processing tab using a filter width of 1 µm for the phalloidin channel for all images. Volumetric surface areas of the stereocilia in each image were then created from the phalloidin channel using creation parameters guided by the program and edited as needed so that the phalloidin signal for each stereocilium was accurately surrounded by a surface. Due to the high dynamic range of the signal, and the low intensity of the signal in control row 3 or 4 stereocilia, creation parameters were guided to create the most accurate surfaces for row 1 and 2, which often left some low intensity row 3 or 4 stereocilia undetected. Manual cutting of any connected stereocilia surfaces was performed so that each stereocilium was modeled by a unique surface. For any visible stereocilia lacking a surface (most often in rows 3 or 4), a surface was created using the magic wand tool in the Create tab of Imaris. If the intensity of the stereocilium was too similar to the phalloidin intensity of the surrounding cuticular plate, a surface could often not be generated. Thus, row 3 and 4 stereocilia are only sparsely represented in older control bundles. Stereocilia surfaces were labeled as corresponding to each stereocilia row (1, 2, 3, or 4+) and as corresponding to each cell in the image. Measurements of Volume, BoundingBoxOO (object-oriented) Lengths, Number, and Position (x,y) for all labeled surfaces were exported from Imaris to Microsoft Excel. BoundingBoxOO lengths are generated within Imaris by considering the minimal rectangular box which fully encloses each object with no constraints to the orientation; the dimensions of that box are reported as Lengths A, B, and C, with A being the shortest principal axis and C the longest principal axis. To measure length and width of each stereocilium, object-oriented BoundingBoxOO Length C and the average of lengths A and B were used, respectively. For volume, we used the surface volume measurement provided by Imaris, which quantifies the amount of space each object occupies. The number of stereocilia and the sum of stereocilia volume per cell were calculated in Excel from the Imaris output. To measure the center 10 stereocilia, 10 neighboring row 1 and row 2 stereocilia at the center of each bundle were selected within Imaris, and measurements for this subset were exported to Excel. Position (x,y) data could be used to determine neighboring stereocilia and calculate differences in dimensions for adjacent stereocilia within each row or adjacent row 1 and 2 stereocilia (Fig 5). <br> Three-dimensional rendering of stereocilia and EPS8 and GPSM2 tip staining (Fig 11-12) from Airyscan z-stack images was also performed in Imaris using similar methods. Background subtraction was first performed in the image processing tab using a filter width of 1 µm for all channels for all images. Volumetric surface areas of the stereocilia in each image were then created from the phalloidin channel using creation parameters guided by the program and edited as needed so that the phalloidin signal for each stereocilium was accurately surrounded by a surface. Volumetric surface areas of the EPS8 or GPSM2 signals in each image were also created in their respective channels using creation parameters guided by the program and filtered to only include those with at least 0.01 overlap volume ratio with the phalloidin stereocilia surfaces, so that only EPS8/GPSM2 surfaces at stereocilia tips were included. All surfaces were manually edited to split any joining surfaces and create any surfaces that were missed for low intensity phalloidin, EPS8, or GPSM2 signals. Once all surfaces were edited, the two sets of surfaces were merged and the phalloidin and EPS8 or GPSM2 surfaces corresponding to each individual stereocilium were joined in the edit tab, creating a surface that encompassed both the phalloidin signal and tip complex signals for each stereocilium. Each stereocilium surface was then labeled according to its row position and cell in the image. Thus, for each stereocilium we measured both the length (BoundingBoxOO Length C) and the corresponding intensity sum for each channel in the image, with the channel 1 (568 nm) intensity sum measuring EPS8 amount and the channel 2 (488 nm) measuring GPSM2 amount. These measurements were exported to excel, pooled, and then plotted within Prism. <br> Quantitation of proteins at stereocilia tips <br> Airyscan z-stacks were imported into Fiji, which was used for all analysis steps. For analysis of C57BL/6J mice throughout development, CAPZB and EPS8L2 in P21.5 mutants, and rhodamine-actin signal, summed z-projections of Airyscan stacks were made that included row 1 and row 2 tips in the same projection. For samples in which row 1 and 2 were in separate Z planes, separate Z-projections were made for row 1 and row 2 tips. Regions of interest (ROIs) were selected at 10 (or more) row 1, and 10 (or more) row 2 tips per hair bundle. ROIs were circles that encompassed most of each tip. Using Fiji’s Measure function, the area and mean gray value were measured for each ROI; measurements were also made outside the stereocilia and above the epithelium to determine background. The total fluorescence signal in a tip (tip signal) was thus area • mean gray value • minus total signal for the background. Tip signals below the background were assigned the value of 0, which were used in calculating averages. <br> For analysis of mutant mice in Fig 11-12, due to increased length variability, regions of interest (ROIs) were selected at row 1 or row 2 tips within individual x-y slices from the z-stacks. Images were kept as multi-channel stacks and phalloidin was used to guide selection at the tip of each stereocilium. For these images, all row 1 and 2 stereocilia were measured within each cell. Measurements and calculation of tip signal were otherwise carried out as above. Mean intensities were calculated as the average tip intensity for all row 1 and row 2 stereocilia within each cell. Mean intensities were normalized to the average tip intensity for all row 1 and 2 stereocilia from all het cochlea imaged on the same day. Histograms were created in Prism using frequency distributions of the number of values, with a bin width of 0.2 µm. The distributions were fit with a least-squares regression for a sum of two Gaussians. <br>