Polarized localization of phosphatidylserine in the endothelium regulates Kir2.1

Lipid regulation of ion channels is largely explored using in silico modeling with minimal experimentation in intact tissue; thus, the functional consequences of these predicted lipid-channel interactions within native cellular environments remain elusive. The goal of this study is to investigate how lipid regulation of endothelial Kir2.1 — an inwardly rectifying potassium channel that regulates membrane hyperpolarization — contributes to vasodilation in resistance arteries. First, we show that phosphatidylserine (PS) localizes to a specific subpopulation of myoendothelial junctions (MEJs), crucial signaling microdomains that regulate vasodilation in resistance arteries, and in silico data have implied that PS may compete with phosphatidylinositol 4,5-bisphosphate (PIP2) binding on Kir2.1. We found that Kir2.1-MEJs also contained PS, possibly indicating an interaction where PS regulates Kir2.1. Electrophysiology experiments on HEK cells demonstrate that PS blocks PIP2 activation of Kir2.1 and that addition of exogenous PS blocks PIP2-mediated Kir2.1 vasodilation in resistance arteries. Using a mouse model lacking canonical MEJs in resistance arteries (Elnfl/fl/Cdh5-Cre), PS localization in endothelium was disrupted and PIP2 activation of Kir2.1 was significantly increased. Taken together, our data suggest that PS enrichment to MEJs inhibits PIP2-mediated activation of Kir2.1 to tightly regulate changes in arterial diameter, and they demonstrate that the intracellular lipid localization within the endothelium is an important determinant of vascular function.

image of endoplasmic reticulum (ER, yellow) detected via calnexin and IEL (grey), (C) box and whiskers plot of minimum distance of real-world HIEL centers to ER compared to Matlabsimulated HIEL centers. N=1 mouse, n=1 artery, n=2 ROIs, Area= 6.96x10 4 µm 2 , and n=315 HIEL. (D) Representative en face confocal image of interendothelial junctions (green) detected via claudin-5 and IEL (grey), (E) box and whiskers plot of minimum distance of real-world HIEL centers to interendothelial junctions compared to Matlab-simulated HIEL centers. N=6 mice, and n=10 arteries, n=15 ROIs, Area=1.66x10 5 µm 2 , and n=1200 HIEL. Statistical test: Brown-Forsythe and Welch ANOVA. # indicates a p <0.0001 significant difference to real-world HIEL distribution, * indicates a p <0.0001 significant difference to RAND, $ indicates a p <0.0001 significant difference to NC distribution, and & indicates a p <0.0001 significant difference to PC. is the passive diameter. Dotted line indicates the optimal pressure for Eln fl/fl /Cre + arteries. N=3 mice per group, n=4-6 arteries per group. A repeated measures two-way ANOVA was performed, with p<0.0001 significant difference between genotypes for Myogenic Tone and Passive Diameter analyses. A Sidak's multiple comparison post hoc analysis was performed at each pressure where * p<0.050 and *** p<0.001. No significant difference was detected for Active Diameter analysis.     Predictions are then normalized to obtain ratios with respect to IEL length, a metric that can be reproducibly measured in TEM images (Normalized Cases 1-3).

Predicting the incidence of myoendothelial junctions in transmission electron microscopy cross-sections
First, we considered basic descriptive data obtained from en face images (Supplemental Table 2) including the circumference of the artery (C, Supplemental Figure 2A) the average diameter of HIEL (d, Supplemental Figure 2B), and spatial density of HIEL (⍴HIEL, Supplemental Figure 2C), which was calculated using Equation 1.
ÑHIEL and d were quantified via the in-house Matlab program, and C was measured as the width of the artery when prepared en face. For this calculation, we assume the measured circumference of the artery is equivalent to the circumference of the IEL (CIEL).
Next, we considered the transverse view of the artery when prepared for TEM imaging (Supplemental Table 3). The TEM section is a thin cross section sliced from a third order mesenteric artery where the thickness of an individual section is YTEM. Since our quantification of HIEL thus far is from the en face view, we wanted to quantify the IEL en face surface area in 1 TEM section in order to predict the density of HIEL in TEM sections. A single TEM cross-section corresponds to an en face surface area of 20.902µm 2 , which was calculated using Equation 2.
An important note is d is much larger than YTEM, such that the average HIEL will span 30 TEM sections Equation 3.
For conceptual clarity, we will consider 6 TEM sections equally spaced apart along artery length d, which corresponds to an IEL surface area of 627.06µm 2 Equation 4 and an IEL length of 1791.6µm Equation 5.
Using ⍴ HIEL and ATEM, d, we calculated we would detect a total of 9 unique HIEL across a sectioned artery length of d with Equation 6.
The final variable we considered in our prediction is the spatial distribution of holes throughout the IEL. For this, we considered three cases of distribution (Supplemental Table 4): (1) uniform distribution (maximum detection), (2) random distribution, and (3) sparse distribution (minimum detection).
For Case 1, we considered 6 TEM sections over artery length d and assumed each of the 9 HIEL were exactly aligned with the start and end of the sectioned area ( , ) such that the HIEL distribution is uniform, and the number of HIEL detections by the microscope user is 54 (Equation 7).
%&'(,' = 6 × HIEL +',,5' However, due to the nature of TEM, the entire cross-sectional area cannot be imaged due to However, based on the Matlab simulations (Figure 1), we have shown HIEL are randomly distributed. We therefore incorporated a random distribution of HIEL into our prediction. Thus, instead of assuming each HIEL would have 6 detections across each of the 6 sections, we assumed each HIEL had a different number of detections due to a random alignment with the beginning and end of the length sectioned area. The random distribution of detections was considered as follows: 2 HIEL with 6 detections each (12 detections), 1 HIEL with 5 detections each (5 detections), 2 HIEL with 4 detections each (8 detections), 1 HIEL with 3 detections each (3 detections), 1 HIEL with 2 detections each (2 detections), and 2 HIEL with 1 detection each (2 detections), bringing the total to 32 detections (D %&'(,7 , adjusted Equation 7 random distribution). Using the same normalization process as described above, we determined if HIEL are randomly distributed, we expect 17.8 HIEL per 1000 m length of IEL (Equation 9).
Due to the random distribution of HIEL, there are some areas of IEL with relatively few HIEL.
Thus, a third case is considered to reflect a sparse distribution scenario. In this minimum case scenario, each of the 9 HIEL are only aligned with 1 TEM section and thus is only detected once during imaging. For 9 HIEL considered across 6 sections, with each HIEL appearing in only 1 section, there are 9 total detections (D %&'(,8 , adjusted Equation 7 random distribution).
Normalizing these to the IEL within TEM images as for the previous two scenarios, 5 HIEL are expected per 1000 m length of IEL (Equation 10).
Since we have demonstrated the HIEL are randomly distributed, and there are areas sparse in HIEL, the predicted range of HIEL density in the TEM images is 5-17.5 HIEL per 1000 m IEL analyzed.

Vascular cell co-culture
Vascular cell co-culture (VCCC) was created with human coronary artery smooth muscle cells (HCoASMCs) and human umbilical vein endothelial cells (HUVECs), paraffin embedded, and stained as originally described by us. (7, 69) Cell lines were purchased from ATCC and use before 10 passages.

Transfections for immunohistochemistry
HeLa or HEK293T were plated in 6-well dishes and grown until 70-80% confluent. for 30 minutes in a 37°C water bath. The artery was then cannulated on glass canula on a pressure myograph setup. After equilibration to 80mmHg, 10µM TopFluor-PS (Avanti, 810283P) was added to the bath solution and circulated for 30 minutes. The artery was then immediately prepared en face and secured to a microscope slide with a droplet of Prolong Gold with DAPI and a glass coverslip (no fixation step). Arteries were imaged on an Olympus Fluoview 1000 microscope.

VVG stain and quantification
Arteries were isolated from mice and fixed with 4% PFA at 4°C overnight, embedded in 3% agarose/PBS, then sent to the UVA Research Histology Core to be embedded in paraffin wax cut into 5µm cross-sections. The cross-sections were deparaffinized by heating to 60°C for 1 hour in a glass slide holder, followed by dehydration with Histoclear (National Diagnostics, HS-200) and a decreasing ethanol gradient. The Elastic Stain Kit (Verhoeff Van Gieson / EVG Stain) kit was purchased from Abcam (ab150667). Slides were placed into elastic stain solution for 15 minutes and rinsed with cool, running diH2O to remove excess stain. Sections were differentiated by dipping the slides in 2% ferric chloride solution 15 times. The differentiation reaction was stopped by rinsing slides in cool, running diH2O. Excess stain was further removed by dipping the slides in the sodium thiosulfate solution for 1 minute, followed by rinsing in cool, running diH2O. The slides were then placed in Van Gieson solution for 2 minutes for a counterstain. Excess stain was removed by rinsing in two changes of 95% EtOH, followed by a 2-minute incubation in 100% EtOH. The slides were then quickly mounted using VectaMount media and a coverslip. Images were taken using a traditional light microscope and a 20x or 40x objective.
ImageJ was used for image analysis. The stained arterial cross-section image was opened in ImageJ. The colors of the original image were then separated by "H PAS stain" color deconvolution (Image à Color à Color Deconvolution à H PAS). The indigo image was analyzed because it specifically contained the Verhoeff elastin stain. The indigo imaged was converted to a black and white mask, and the threshold of indigo was automatically determined by the ImageJ software (Process à Binary à Convert to Mask). Student's t test.