Electron microscopy paired with immunogold labeling is the most precise tool for protein localization. However, these methods are either cumbersome, resulting in small sample numbers and restricted quantification, or limited to identifying protein epitopes external to the membrane. Here, we introduce SUB-immunogold-SEM, a scanning electron microscopy technique that detects intracellular protein epitopes proximal to the membrane. We identified four critical sample preparation factors that contribute to the method’s sensitivity and validated its efficacy through precise localization and high-powered quantification of cytoskeletal and transmembrane proteins. We evaluate the capabilities of SUB-immunogold-SEM on cells with highly differentiated apical surfaces: (i) auditory hair cells, revealing the presence of nanoscale MYO15A-L rings at the tip of stereocilia; and (ii) respiratory multiciliate cells, mapping the distribution of the SARS-CoV-2 receptor ACE2 along the motile cilia. SUB-immunogold-SEM extends the application of SEM-based nanoscale protein localization for the first time to intracellular epitopes on the exposed surfaces of any cell.

This dataset contains all raw SEM micrographs used for the quantification used in the paper by Miller et al.

SUB-immunogold-SEM method

Default permeabilization: After fixation and dissection, the samples were transferred to 2 mL tubes with TBST (150-mM NaCl, 10-mM Tris-HCl, 0.05% Tween 20, pH 7.5). By default, the permeabilization consisted of incubation with 0.05% Triton X-100 in TBST for 20 min at RT (around 25°C) under nutation mixing at 5 rpm (Boekel Scientific, variable speed mini orbitron, #201100), followed by a 5-min TBST wash.

Alternative permeabilization: Dehydration–rehydration: After fixation, the samples were transferred to 2-mL tubes with TBST and placed on ice. The dehydration–rehydration process involved buffer exchange with ice-cold solutions to increase ethanol percentages in Milli-Q water for 5 minutes without mixing. Liquid transfer was performed with a disposable pasteurette pipette, permanently submerging the sample. The buffer sequence was H₂O, 15% ethanol, 30%, 50%, 75%, 95%, 100%, 100%, 95%, 75%, 50%, 30%, 15% ethanol, H₂O, and TBST.

Alternative permeabilization: Saponin: For this experiment, 0.05% Saponin in TBST was used instead of Tween 20 for 20 min at RT under nutation mixing at 5 rpm, followed by a 5-min TBST wash.

Blocking: Samples were blocked in TBST containing 4% BSA for at least 6 hours or overnight at 4 °C.

Immunogold staining: The samples were transferred to 0.3-mL PELCO mini vials (TED PELLA, #21441) sealed with parafilm with primary antibodies in TBST with 1% BSA under nutation mixing at 5 rpm, overnight at 4 °C. The samples were transferred into 2-mL tubes with a micro dissecting spoon (Biomedical Research Instruments, #15-1025), always maintained in liquid, rinsed once, and washed thrice for 15 min with 1% BSA TBST under nutation mixing at 5 rpm. The samples were then transferred to new 0.3-mL PELCO mini vials with 5-nm, 10-nm, or 15-nm gold-conjugated goat anti-rabbit or mouse IgG (BBI: 1:200 in 1% BSA TBST) and incubated under nutation mixing at 5 rpm overnight at 4 °C. After secondary antibody incubation, the samples were rinsed once and washed thrice for 15 min each with 1% BSA TBST in 2-mL tubes under nutation mixing at 5 rpm.

Post-fixation: The samples were then rinsed twice with 0.1-M sodium cacodylate buffer (pH 7.2) and fixed with 10% glutaraldehyde and 4% PFA in 0.1-M sodium cacodylate buffer for at least 24 h at 4 °C without agitation.

OTOTO post-fixation: In the case of OsO₄ (EMS #19150)/OTOTO (EMS #21900) sequential post-fixation, the samples were first transferred to glass vials using a micro dissecting spoon and washed with 0.1-M sodium cacodylate buffer. The samples were fixed with 1% osmium in 0.1-M sodium cacodylate for 1 h at RT without agitation and protected from light. The samples were then washed four times (5 min each) using 0.1-M sodium cacodylate, then thrice using H₂O. Subsequently, the samples were incubated with 1% thiocarbohydrazide for 20 min at RT without agitation and protected from light. The samples were then washed four times (5 min each) using water, and then thrice using 0.1-M sodium cacodylate. The sequence was repeated making a total of three osmium and two thiocarbohydrazide incubations.

Dehydration: The samples were washed using 0.1-M sodium cacodylate buffer and transferred to a sample holder for a critical drying point in Milli-Q water. The samples were then dehydrated on ice with ice-cold ethanol solution diluted using Milli-Q water (15%, 30%, 50%, 75%, 95%, 100%, and 100%, 5-min incubations), with the sample holder permanently submerged. The sample holder was placed immediately in the chamber and processed for critical drying point (Autosamdri-815A, Tousimis). The cochleae were mounted on studs using silver paint and coated with 2- to 3-nm of palladium (sputter coater EMS150TS, Electron Microscopy Sciences) as described by Grillet¹⁷. The samples were imaged using a 5-kV accelerated voltage and a 100-pA beam current using a concentric BSE detector on an FEI Magellan 400 XHR Field Emission SEM (Stanford Nano Shared Facilities). The microscope is periodically calibrated for measurements using a SIRA-type calibration specimen for ultra-high-resolution modes with a 2% error between 50- and 350-k magnification in our imaging settings. For OTOTO postfixed samples, the beam current was bumped to 200-pA. The gold beads, characterized by their circular shape and defined diameter, were easily identified as strong BSE sources at the cell surface which form a center core surrounded by a lighter halo. In rare cases where the diameter of the core signal appeared larger than the other beads but no lighter signal separated the core signal, it was assigned to a single gold bead. The micrograph contrast was adjusted or pseudo-colored in postproduction using Photoshop (Adobe) to display the gold better when needed. As stage tilting is impossible in backscattered electron imaging mode, hair bundle orientation could not be adjusted for optimal imaging. The distance measurements of the gold to stereocilia tips were performed using ImageJ2 after scale calibration, placing the measuring ends at the center of the gold bead and the pinnacle of the stereocilia tip. These distances were approximations of the absolute distances, as we measured the shortest distance between the gold beads and the stereocilia tip on 2D pictures without considering the stereocilial volume and the perspective distortion of the images. To measure the position of PMCA2a-gold beads along the length of the stereocilia, we used micrographs of OHC and IHC stereocilia taken from a nearly perpendicular perspective. We measured the distance from the base of the stereocilia at the cuticular plate to the gold bead. Additionally, we measured the stereocilia's full height to determine the gold beads' relative position as a percentage of the total stereocilia height. To visualize the distribution of PMCA2a-gold beads, we constructed distribution graphs by segmenting each stereocilium into percentage ranges of its overall height (e.g., 0-10%, 10-20%, and so forth) and quantifying the number of beads within each segment. These data were then graphically represented as mean values with standard deviation (mean ± SD).

Conversion of the SEM-determined width and height dimensions were calculated using the shrinkage factor reported by Miller et al., including the error propagation.

Quantification and statistical analysis

Measurements were taken from distinct samples. Statistical analyses and sample sizes for all the experiments are detailed in the figure legends and Supplementary Table 1. Normality tests determine whether downstream tests should be parametric or nonparametric. Mann–Whitney U tests (Two-tailed) were employed for nonparametric pairwise comparisons of two groups. To compare multiple ages, nonparametric Kruskal–Wallis tests followed by Dunn’s multiple comparison tests were performed. Refer to Supplementary Table 1 for a comprehensive group and statistical description. GraphPad Prism 9.4 for Mac (GraphPad Software, San Diego, CA, USA) was used for the statistical analyses. The micrographs presented are representative examples from two independently replicated experiments.