Imaging effects of hyperosmolality on individual tricellular junctions

A nanoscale electrochemical imaging method was used to reveal heterogeneity present in conductance at epithelial cell junctions under hyperosmotic stress.

To obtain MDCKII cell lines homogeneously overexpressing ILDR1 protein (MDCKII-ILDR1), MDCKII cells were transfected with retrovirus method as previously described. 1 Briefly, the full-length mouse ILDR1 gene (NM_001285788) was cloned into the retroviral vector pQCXIN (Clontech Laboratories, Mountain View, CA). Molecular clones were verified by DNA sequencing. Vesicular stomatitis virus glycoprotein (VSV-G) pseudotyped retroviruses were produced in human embryonic kidney cells 293 (HEK-293) and used to infect MDCKII cells at a titer of 10 6 cfu/mL.
Resulted MDCKII-ILDR1 cells were maintained in the same condition as wild type cells.
Before P-SICM measurement (vide infra), cells in confluence were harvested with trypsin-EDTA (Mediatech, Tewksbury, MA) and seeded with 10 6 cells/cm 2 onto a home-made membrane insert. To prepare the insert, a piece of transparent polyethylene terephthalate (PET) membrane (pore size 0.4 μm, pore density 2 × 10 6 /cm 2 , Corning, NY) was taped to the bottom of a petri dish with a hole in the center, and collagen (Sigma-Aldrich) of 150 μg/cm 2 was then added onto the membrane. Cell monolayers cultured on membrane inserts were maintained at 37 °C in a humidified air-5% CO 2 atmosphere for 4-7 days to achieve confluence that displayed steadystate transepithelial electrical resistance (TEER).

TEER Measurements
To measure the overall structural integrity of tight junctions (TJs) in MDCKII cells, TEER value was measured with an endohm chamber (ENDOHM-24SNAP, World Precision Instrument, Sarasota, FL) connected to an EVOM2 epithelial voltohmmeter (World Precision Instrument). Cells were grown on a porous PET membrane insert (Corning) with 0.4 μm pore size and 4.2 cm 2 exposed culture area (A). The membrane insert with cells was placed in the endohm chamber which set one electrode at each side of the cell monolayer, and the resistance value (R) could be read from voltohmmeter. Background resistance of PET membrane (R 0 ) can be determined by the same fashion with a blank membrane insert. TEER value of the cell monolayer in Ω × cm 2 was calculated as:

Electrochemical Impedance Spectroscopy
Electrochemical impedance spectra of wild type MDCKII or MDCKII-ILDR1 cells grown on membrane insert were measured by a CHI 660C potentiostat (CH Instruments, Austin, TX) in three-electrode configuration. HBSS was filled in both sides of cells. A peak-to-peak 10 mV sinusoidal wave with frequency ranged from 1 Hz to 100 kHz was applied to the cell monolayer, and the current response was recorded at 60 discrete frequencies. From the Nyquist plot obtained (Z Im vs Z Re ), a semicircular response was observed and utilized to calculate the resistance of PET membrane and cell monolayer respectively. To obtain the local conductance map, an FPGA board (Digilent, Inc., Pullman, WA) and its user interface were utilized to control the SICM feedback loop. Briefly, at the beginning and end of each "hop" during the hopping mode of SICM scanning, the feedback control was bypassed and paused for a preset duration during which a 4 V pulse was generated from the FPGA to trigger an external transmembrane potential.

P-SICM Setup and Measurement
The transmembrane potential was applied to WE from an Agilent 33220A function generator (Agilent Technologies, Santa Clara, CA) with respect to ground (RE and CE). Potentiometric signal at UE was recorded with a customized differential amplifier with 100× gain and 20 Hz low-pass filter. A digitizer (Axon Digidata 1440, Molecular Devices, Sunnyvale, CA) was used for monitoring the data from all scanning channels.
The acquired channel data were collected by FPGA and the apparent local conductance (G) of the present probed position can be calculated as: 3 In this equation, E is the electric field between the two probe positions at the beginning and end of each "hop" when V T is applied. E can be determined from dividing the difference of potential deflections at these two probe positions (ΔV close -ΔV far ) by the hopping height (Δz). ρ is the specific resistivity of the bath solution in the top chamber. V e represents the potential range of V T (e.g. V e = 100 mV if a ± 50 mV triangular wave is applied as V T ).
As the nanopipette scan over the sample to acquire topographical information, the local G value can also be simultaneously obtained. Thus, the conductance map can be generated after all pixels of a cell area is scanned.
For studying the effect of hyperosmolality, a P-SICM conductance map was first taken on the cell sample in normal condition (HBSS on both sides here), then the bath solution in the bottom chamber was changed to HBSS with defined concentration of mannitol and a second P-SICM scan was conducted on the same area. The effect can be visualized by comparison between two resulted conductance maps.

Focus Ion Beam (FIB) for Model Cell Junction
The synthetic model junction for validation of computer vision algorithm (see below) was milled into a 50 nm thick Si 3  All images were converted to TIFF format and arranged using Photoshop (Adobe).

Statistical Analysis
The automated algorithm output G values of each individual cell bodies (CBs), bicellular tight junctions (bTJs) and tricellular tight junctions (tTJs) present in the conductance map. Thus, the increased amount of G value (ΔG) for these features overall after mannitol treatment can be expressed as means ± error of the means.

S2. Validation of the ability of P-SICM to reveal tight junction related properties
To validate that P-SICM can reveal TJ-related functions or properties, P-SICM topography and conductance maps were compared to antibody stained Zonula occludens (ZO)-1 fluorescence images labeled to determine the location of tight junctions in the same area scanned. Topography ( Figure S4a Figure S4c) images were collected. The expanded region shown in Figure S4c indicates the location where P-SICM recordings were made. As shown in Figure S4d, when topographic or conductance images are layered with the fluorescence images, the location of TJs is collocated. From these images we also note that the topographic images are suitable for locating the position of cell body and cell junctions.

S3. Long-duration control experiments as endurance test
To assure the cell viability and stability of TJs under experiments 3.5-4 hours in duration, endurance tests were performed where the MDCKII cells were scanned by P-SICM three times successively (ca. 5 hours imaging (Fig. S5)). The integrity of the MDCKII monolayer before and after this test was evaluated by EIS as shown in Fig.   S6. Both bTJs and tTJs were found to be stable for at least 3 P-SICM scans, which is strongly supports cell viability under conditions employed here (2 P-SICM scans) in examining the effects of hyperosmolality.  Figure S6. Impedance spectra of MDCKII cells before P-SICM scans (0 min, black) and after P-SICM scans (283 min, red).

S4. Automated algorithm for quantification of P-SICM conductance map
The automated algorithm for quantification of P-SICM conductance map was written in Python 3.7 supplemented with image processing module OpenCV (cv2). Fig. S7 shows the overview of extracting junction information. First, the data matrix of P-SICM conductance map is used as the input of the program. The averaged G value of the whole map is calculated and used as the threshold to zero all the pixels below it. The remaining pixels are then scaled to [0,255] which results in a grayscale uint8 image.
To make the junction area more pronounced, before the scaling these remaining pixels can be multiplied by 255. If the pixel is larger than 255, its value is set as 255. The program next removes incoherent non-zero pixels defined as shapes with their number of continuous pixels (i.e. size) too small to be significant. cv2.findContours() function is utilized to find all the shapes present in the uint8 image as well as their number of pixels. Shapes smaller than the threshold size set by the users are eliminated. Median filter is then applied to remove other noises and smoothen the shape of non-zero areas. The image is further filtered by image thresholding which generates a binary image for the convenience of subsequent operations.
cv2.dilate() followed by cv2.erode() are used to eliminate small black pixels in the white objects (also referred to as "hole closing"). Next, skeletonization function 4 is used to extract the skeleton which outlines the medial axes of existing objects. The white pixels of preliminary skeleton image are usually not continuous, and thus the skeleton image is subjected to another set of noise removal functions and pruning.
First, the same function of removing incoherent non-zero pixels is performed and the skeleton images is dilated in advance to avoid excessive loss. "Hole closing" function is again applied to remove small holes in the shapes. Considering that the final skeleton of the junction areas should be a continuous object, a function named "gap closer" here was designed to predict and connect discrete shapes. In

S5. The role of Ca 2+ in hyperosmolar effect
To study the effect of Ca 2+ on tTJ alteration, the sample was bathed in HBSS with ~ 2 μM Ca 2+ and other ion components identical to original one. Mannitol dosages used here were 0 (control), 25 mM, 50 mM and 100 mM.

S6. The role of ILDR1 overexpression in hyperosmolar effect
The barrier properties of transfected ILDR1 cells were evaluated by EIS (Fig. S9). P-SICM measurements were then performed on ILDR1 cells with the same procedure and mannitol dosage as the wild type cells (Fig. S10 and Fig. 8).