Nanoscale Clustering and Dynamics of Phosphatidylinositol 4,5-Bisphosphate in an Immune Cell Model

Abstract

Mast cells mediate their immuno- and neuro-modulatory effects by releasing granules containing bioactive substances. Phosphatidylinositol 4,5-bisphosphate (PIP₂), enriched at the plasma membrane (PM), is a key signaling lipid involved in numerous physiological functions including the calcium entry needed for antigen stimulated mast cell degranulation. However, functional nanoscale PIP₂ clustering and dynamics have not been previously investigated in immune cells. Using the pleckstrin homology domain from PLCδ (PH) tagged with photoswitchable fluorescent protein Dendra2, clustering was revealed in the mast cell model RBL-2H3, both fixed and live. We also discovered that live RBL-2H3 cells have PH clusters which evolve over timescales of ~100s. Additionally, the distribution of PIP₂, and specifically PIP₂ clusters themselves, are disrupted upon addition of the cationic, lipidic drug cetylpyridinium chloride (CPC). CPC led to smaller, less dense, and more circular clusters. Furthermore, PH molecular mobility increased after the addition of CPC, suggesting interference of this drug with PH binding to PIP₂. In addition to this pharmacological relevance, the physiology of PIP₂ clusters during functional stimulation by antigen was investigated. Antigen stimulation led to increased cluster size, which was counteracted by CPC. In live cells, PH density outside clusters was altered by CPC but not by antigen. CPC increased the proportion of regions of high-density PH compared to all other regions of the PM. While PH diffusion was, interestingly, not affected by antigen, it was increased by CPC, particularly in lower density regions. Under all live cell dynamics observed, PH demonstrated confinement which were consistent with simulated diffusion within potential wells with an elliptical shape. These findings illuminate the nanoscale behavior of PIP₂ in immune cells and the correlation of that behavior with cell function.

1. INTRODUCTION

Despite comprising only ~1% of all plasma membrane (PM) phospholipids (1), phosphatidylinositol 4,5-bisphosphate (PIP₂) regulates a multitude of cellular functions. These functions include cytoskeletal rearrangement (2–5), ion transport (6–8), cellular trafficking (9–11), and cellular signaling (12–15). The synthesis and conversion of PIP₂ is tightly regulated by several lipid kinases and phosphatases as well as enzymes which cleave PIP₂ to generate second messengers (12,16). Spatially and temporally distinct pools of PIP₂ generated by these enzymes appear to regulate the various functions involving PIP₂ (17–20).

It has been proposed that PIP₂ clusters at the nanoscale to regulate function, (21–23) yet a consensus on this proposal has not been reached. Several papers report no evidence of nanoscale PIP₂ clustering (24,25) and others posit sample preparation may artificially induce PIP₂ clustering (26,27). However, PIP₂ appears to cluster in diffraction-limited microscopy (28) and several super resolution studies do provide evidence of PIP₂ nanoscale organization (29–37). Nanoscale PIP₂ clusters (70–90 nm in size) have been observed with electron microscopy in fibroblasts and smooth muscle cells (33). Coclustering of PIP₂ with syntaxin, a membrane protein essential for exocytosis, has been observed in PC-12 cells via confocal microscopy (38) as well as via super-resolution stimulated-emission depletion microscopy and STORM, which revealed their average diameter of ~65 nm (30) and ~73 nm (29), respectively. PIP₂ nanoscale dynamics have also been studied in live HeLa cells (39). Thus, there is evidence that PIP₂ may be nonuniformly distributed in cell membranes, but whether and how clustering occurs in live cells at the nanoscale and how clustering relates to function remain incompletely understood.

PIP₂ clustering is observed in immune cells as well. PIP₂, replenished by Nir3, appears clustered in CD8 T-cells (40) via diffraction-limited microscopy. PIP₂ clusters with B-cell receptors (BCR) during early B-cell activation, after which PIP₂ is depleted within BCR clusters (by PLCγ2) and enriched around the edges of BCR clusters (31). Disrupting this spatiotemporal organization led to the reduced number of and size of BCR clusters (31). These studies were important for investigating the roles of PIP₂ but did not quantify nanoscale PIP₂ clustering in immune cells. Super-resolution quantification of PIP₂ clusters in immune cells has not previously been performed, but evidence suggests that nanoscale dynamics and clustering may be important to immune cell function (41).

Mast cells (MCs) are multi-effector immune and nervous system cells found throughout most tissues of the body (42–44), even the brain (45). Mast cells release granules containing biochemical mediators, such as histamine and serotonin (46), to mediate physiological functions. These include coordination between the innate and adaptive immune system, mood regulation, and parasite and microbial defense (45,47–49). The RBL-2H3 mast cell model is a well characterized model for mast cell function and is functionally homologous to mature human mast cells (50–53), including response to exogenous agents (49,51–56).

Several aspects of MC signaling depend on phosphoinositides. MCs are activated via antigen (Ag) crosslinking of immunoglobulin E (IgE)-bound FcεRI receptors inciting a cascade (57,58), which leads to activation phospholipase Cγ1 (PLCγ1) (59). PLCγ1 then binds PIP₂ with its pleckstrin homology (PH) domains (13): positively-charged Lys/Arginine residues on the PH domain bind to the 4- and 5-monoesters of PIP₂ (60). PLCγ1 then cleaves PIP₂ into the second messengers inositol triphosphate (IP₃) and diacylglycerol (DAG) (13). IP₃ binds its receptors on the endoplasmic reticulum (ER), initiating Ca²⁺ influx (61), subsequent microtubule polymerization and, ultimately, induction of PIP₂ clustering to promote degranulation (62–64). Additionally, PIP₂ forms nanoclusters in the presence of multivalent cations (65). High local concentrations of Ca²⁺ have been shown to influence membrane curvature (62) which promotes degranulation (63). This Ca²⁺ influx is in turn regulated by positively-charged septins and PIP₂ clusters (66).

Due to the number of native positively charged molecules which regulate PIP₂ clustering, functions involving PIP₂ seem to be particularly susceptible to disruption by exogenous cations. For example, PIP₂ binding to partners is disrupted by the antibiotic neomycin (67,68). There are several instances where electrostatic disruption of PIP₂ has consequences for mast cell signaling. It has been demonstrated that sphingosine derivatives D-sphingosine and N,N’-dimethylsphingosine disrupt mast cell degranulation by electrostatically disrupting protein interactions with phosphoinositides (69). Cationic polyamines alter Ca²⁺-triggered vesicle fusion (70). Differentially charged amphipaths modulate mast cell degranulation in opposite ways; for example, the negatively-charged salicylate induces higher degranulation in mast cells while the positively-charged chlorpromazine is a pharmaceutical used to potently decrease degranulation (63). Our lab has found that the cationic antimicrobial cetylpyridinium chloride (CPC) disrupts binding of PIP₂ to certain PIP₂-binding proteins in NIH-3T3 cells and potently inhibits mast cell function (71,72). At the Ag doses and stimulation conditions used in this current study, CPC potently inhibits degranulation (72). This inhibition seems to be electrostatic in nature, as a CPC analog lacking a positively-charged nitrogen did not alter degranulation (72). However, the effects of CPC on the nanoscale properties of PIP₂ in mast cells have not yet been investigated.

CPC is widely used as an antibacterial in over-the-counter personal care (e.g., mouthwashes, lozenges, hair conditioners, deodorants, other pharmaceuticals) (73), food processing (e.g., poultry) (74,75), and cleaning products, often at high (mM) concentrations (74,76). However, we find that CPC inhibits mast cell degranulation at concentrations thousands of times lower (1–10 µM) than that used in commercial products (73–75), and at doses far below where a detergent mechanism of action would occur (77–79). Due to consumer product and food exposures, CPC is likely present in human blood at levels of ~0.3 µM (72). CPC has a lipidic structure with a positively charged headgroup (Figure 1), which may allow it to interact electrostatically with PIP₂ while embedded in the membrane or while membrane adjacent in the cytosol. In this study, we examine an exposure-relevant, non-cytotoxic (including no effects on PM potential or integrity) (71,72,80), low-µM CPC dose to investigate the effects of an exogenous lipidic cation on the nanoscale organization of PIP₂ and propose a mechanism for its inhibition of mast cell function.

Figure 1. Cetylpyridinium chloride (CPC) structure.

Structures of CPC generated using PyMOL (126) with the structure file from PubChem (127). CPC has a lipidic structure with a positive charge from the nitrogen (blue) in the headgroup, which could potentially interact electrostatically with other molecules, and a saturated hydrocarbon chain with potential for hydrophobic or van der Waals interactions with other molecules.

In order to access and quantify nanoscale dynamics and clustering of PIP₂, we utilize Fluorescence Photoactivation Localization Microscopy (FPALM), with achievable resolution on the order of tens of nanometers (81). This allows for nanoscale quantification of clusters smaller than the diffraction limited resolution of a few hundreds of nanometers. While electron microscopy (EM) can achieve even better resolution, EM requires harsh sample processing and is generally incompatible with live cells. FPALM uses localization of individual molecules labeled with fluorescent, photoactivatable probes to build an image capturing the overall cellular distribution (81). This allows cell viability to be maintained and enables live cell imaging (82). Because PIP₂ cluster properties occur at length-scales drastically below what diffraction-limited microscopy can visualize, FPALM is crucial to elucidating the nanoscale organization of PIP₂ in mast cells.

The aim of this study is to quantify the PIP₂ nanoscale distribution in mast cells, test whether PIP₂ clusters, and quantify the nanoscale response of PIP₂ to antigen stimulation of the physiological process of degranulation, and test whether these changes respond to exposure with the positively-charged antibacterial compound CPC. We also explore the impacts of the nanoscale distribution of PIP₂ on mast cell function, using CPC as a tool for probing this functionality.

2. METHODS

2.1. Chemicals and Reagents

Cetylpyridinium chloride (CPC; 99% purity, VWR; CAS no. 123-03-5) was prepared as described in (71) in aqueous solution. Organic solvents with potential for adverse cell effects were avoided in this preparation.

2.2. Cell Culture

RBL-2H3 mast cells were cultured, swapped into phenol red free media for microscopy, and IgE-sensitized (for experiments including Ag-stimulation) as published (72).

2.3. Sample Preparation for Imaging

Cells were transiently transfected with a fluorescently tagged PH-Domain from Phospholipase C delta (Dendra2-PH) via Amaxa electroporation and Cell Line Nucleofector^™ Kit T (Lonza). Cells were plated in 35 mm MatTek dishes (35 mm, coverglass diameter 20 mm) with complete phenol-red free RBL-2H3 media. Following plating, cells were incubated at 37°C/5% CO₂ for approximately 24 hours.

2.3.1. Fixed Samples

For all experiments involving fixed cells, following incubation, MatTek dishes containing cells were removed from the incubator. Samples then underwent two washes of 1 mL of phosphate-buffered saline (PBS) and were fixed at room temperature for 15 minutes in either 1 mL of 4% paraformaldehyde (PFA) (Biotium) in PBS or 3.9% PFA + 0.2% glutaraldehyde (GA) (Electron Microscopy Services) in PBS. Samples then underwent two more washes of 1 mL PBS before a final 1 mL PBS was added to each plate.

2.3.2. CPC Exposure

CPC solution in BSA-Tyrode’s (BT) (71) was freshly made on each experimental day leading up to the end of the cell incubation period. Plates were removed from incubator and media was removed from plates. Subsequently, plates were washed with 1 mL of warm BT. CPC treated plates received 1 mL of 5 µM CPC/BT. Plates were incubated at 37°C/5% CO₂ for 60 minutes and then received additional processing (either Ag-stimulated, fixed, and/or imaged alive). All live-cell plates were imaged on an incubation stage kept at 37°C/5% CO₂. Imaging of live cells lasted no longer than 30 minutes so CPC exposure times for live-cell plates remained between 60–90 minutes.

2.3.3. Antigen Stimulation

RBL-2H3 cells were sensitized with anti-dinitrophenyl (anti-DNP) mouse IgE for 1 hour, as previously described (72), and subsequently stimulated with antigen, DNP-BSA (Life Technologies). To do so, unbound IgE was washed off and cells were stimulated with either 0 ng/mL DNP-BSA/BT (control) or 0.7 ng/mL DNP-BSA/BT for 5 or 10 minutes at either room temperature or 37°C/5% CO₂.

2.4. Illumination Pathway for Cell Selection

Cells were chosen based using the same method as described in (80). The procedure used differs only by use of an LED in place of a mercury lamp.

2.5. FPALM Illumination Pathway

Single-color super-resolution imaging and localization follow previously published methods (32,80,81). Briefly, a 558 nm laser (CrystalLaser LC, CL558–100) and a 405 nm laser (FBB-405-050-FSFS-100, RGBlase LLC) were focused with a convex lens (f = +350 mm, Newport Corporation) into the back port of an inverted microscope (IX71, Olympus) and directed through the back aperture of a 50 x 1.45 NA oil Total Internal Reflection (TIRF) objective lens (Olympus). A power of ~19 mW was used for the 558 nm laser and ~5.5 µW was used for the 405 nm laser with average intensities of ~4.1 kW/cm² and ~0.0019 kW/cm² at the sample, respectively. Fluorescence emission from activated probes was captured through the same objective, passed through a long-pass dichroic (FF562-Di03, Semrock) and a series of notch filters at wavelengths 561 nm (x2) and 405 nm (Stopline, Semrock). Emissions then passed through a 2x telescope (lenses of f = +200mm and f = +400 mm, Newport Corporation) and a bandpass filter (HQ590/75 M, Chroma) resulting in a measured camera pixel size equivalent to 126 nm at the sample. For each acquired ROI, image stacks of 10,000 frames were captured using an EMCCD (iXon + DU897DCS-BV, Andor Scientific) with an EM gain of 200 and frame rate of 31.2 Hz for fixed cells and 100 Hz for live cells. Frame sets were analyzed using standard localization methods written in MATLAB (Mathworks Inc.) including temporal median background subtraction (83), intensity based localization thresholding, 2D Gaussian fitting, tolerancing of poorly fit localizations, and rendering (32,81,84).

2.6. Image Rendering

Images were rendered from single molecule localizations with a brightness scaled inversely with the number of localizations. This was done to account for variation in numbers of localizations from cell to cell, and to account for larger numbers of localizations in live cell acquisitions due to probe replenishment during acquisition. The image in Figure 10B was rendered under different parameters having brightness scaled up by 1.5x as scaling based on localizations overcorrected larger samples with more spread-out localizations. All other representative images were considered similar enough in cell size within experiments that the same rendering parameters were used for all images within a given figure.

Figure 10. Comparison of Dendra2-PH clustering after 5 minutes of antigen stimulation.

(A-B) Representative images of Dendra2-PH in RBL-2H3 cells after 1 hour of IgE sensitization without antigen stimulation (n = 45) and with 5 minutes of 0.7 ng/mL antigen stimulation (n = 47) respectively. Cells were fixed using 4% PFA and imaged using TIRF-FPALM for three biological replicates. Clusters were identified as regions at or above 4x the cell average density containing at least 0.02% of all localizations. (C) Radial distribution functions of stimulated and unstimulated data averaged across cells and clusters. Differences in the distribution are checked for significance using a Komolgorov-Smirnov test represented by ns for no significance. (D-G) Effects of 5 minute Ag stimulation on cluster parameters shown on violin plots. Significance is represented by ns for no significance using Mann-Whitney U-tests.

2.7. Cluster Analysis

Clusters were identified using a grid-based density plot as previously described (32) and as demonstrated in Figure 2. Grids were binned into 20 nm x 20 nm pixels with a width large enough to cover the acquisition ROI. Clusters were identified using a density threshold relative to the cell average density (number of localizations per unit area) of Dendra2-PH (3x and 4x cell average density were used), referred to hereafter as “cell density.” The code used also required clusters to have a minimum of 0.02% of the total number of localizations in the dataset to filter out false clusters due to randomness. Density, area, and perimeter of the clusters were tracked to identify changes across various conditions. A radial distribution function (RDF)(85) which plotted the distribution of PH molecules radially from the center of mass of each cluster was also generated from the average RDFs across clusters and cells in each dataset. When comparing significance between RDFs, normalization is corrected to go to one at long distances (5 µm) as done in (32). This is to avoid false significance readings due to error in normalization by calculated cell averages.

Figure 2. Grid-based cluster identification method.

(A) Representation of localizations (grey points) taken from a raw image. (B) Gridded version of the previous image, separating into equal sized regions to be assessed for densities. (C) Identification of gridded regions which passed (light grey squares) or failed (black squares) a minimum density threshold. (D) Identification of molecules within connected regions which passed a density threshold and subsequent minimum number of molecules threshold (green points).

2.8. Density Heatmaps

Density heatmaps were generated using the same density grids generated during cluster analysis. Density grids are displayed with a colormap that scales by the number of localizations per unit area divided by the cell average number of localizations per unit area. An upper limit of 4x cell average density was implemented on color scaling so that all regions which would pass the initial 4x cell density check for cluster identification could be easily visualized.

2.9. Cluster Property Plots

Cluster data was averaged by cell and error was calculated based on cell-to-cell variability. Violin plots for cluster properties were generated using Graphpad Prism v8.4.3 (Graphpad Software Inc.) and stats tests were performed using a Mann-Whitney test. Radial distribution function (RDF) graphs were plotted using MATLAB (Mathworks Inc.) as an average of the radial distributions across all clusters in all cells for given conditions. Error bars for data points on RDF graphs are standard error of the mean, and were calculated using cell-to-cell variability.

2.10. Density Distributions

Density distributions were generated from the same grids used for cluster analysis. All grids within the cell mask are taken, and the fractional count of the number of bins at each density is recorded, binned relative to cell average density. The log of the fractional count is taken and is plotted for all values greater than the cell density. Density ranges not found in at least 50% of observed cells are excluded. Slopes are compared for significant differences given fit error in Graphpad Prism.

2.11. Trajectory Analysis

Trajectories were calculated using the method of (86) and following similar methods published subsequently (32,84). Identified molecules were tracked frame by frame with a maximum distance of 400 nm between subsequent frames (at a frame rate of 100 Hz) for a localization to be connected to the prior frame as part of a trajectory. To avoid misidentification of trajectories, if other molecules were identified within twice this distance in any pair of subsequent frames, the given trajectory was not progressed. Mean squared displacement of molecules were calculated and used to further calculate the diffusion coefficient of the molecules. Data was averaged per cell. Turn angle histograms for Dendra2-PH trajectories were compared to simulated trajectories under ellipsoidal gradient confinement for 100 simulations per condition using previously published simulation MATLAB scripts (32).

2.12. Diffusion Heatmaps

Diffusion heatmaps were binned into 60 nm x 60 nm bins. For a given trajectory the average position between two successive frames was used as the location for that trajectory step for determining its respective bin. Taking all the trajectories within a bin, the mean trajectory length was determined to create a heatmap of diffusion throughout the cell. The color values of the heatmap were determined corresponding to the trajectory length. An upper limit for color scaling was implemented at 0.25 µm/frame (at 100 Hz) to provide better contrast between high and low diffusion.

2.13. Colocalization Index vs Molecular Step Length

Colocalization index (CI) was calculated using a modified version of code found in (87). The original code compares colocalization between two species; the modified code reassigns the first species as the second and checks for self-colocalization as a local density quantification by molecule. For each molecule with a subsequent trajectory step, the distance to the next step is compared to the CI of the molecule at the original step. CIs are binned and the trajectory lengths corresponding to the CIs found in a bin are averaged. The relationship between the molecular step length and colocalization are plotted on a scatter plot with error bars represented by standard error.

2.14. Statistical Testing

When comparing data sets between two conditions on violin plots, Mann-Whitney U Tests were used. An additional Bonferroni-Holm correction was implemented for any data sets in which multiple subcategories of the overall data were compared to account for the problem of multiple comparisons. When comparing curve fitting parameters between two conditions, an unpaired t-test was used to compare the fit value with standard deviations. When comparing cluster-based radial density distributions, a Komolgorov-Smirnov test was used. When comparing data between three conditions on violin plots, a Kruskal-Wallis test with post-hoc Dunn’s test for multiple comparisons was used. When comparing curve fitting parameters between three conditions, a one-way Anova was used to compare the fit value with standard deviations.

3. RESULTS

To visualize the PIP₂ distribution in our cells, we use Dendra2-PH as a PIP₂ specific binding partner. This probe provides similar distributions of PIP₂ as GloPIP (32), the fluorescent analog of PIP₂ (88). Using Dendra2-PH as a PIP₂ marker, we observe with TIRF-FPALM the presence of high density PIP₂ regions localized within the ventral surface of fixed RBL-2H3 cells (Figure 3) and were present regardless of fixation method (Figure 4). These regions tended to consist of elongated shapes with a high-density core and a decline in density moving outward from this core. These structures were more often seen near the edges of the cell but were still plentiful within more centralized regions of the bottom membrane. This density distribution of Dendra2-PH can be visualized throughout the cell, as well as a zoomed in nanoscale distribution, in Figure 3B. We can visualize the core of clustered regions by the dark red zones and a gradual decrease in density as zones change towards yellow and then further towards blue. Though densities persist above 4x cell average density, a cutoff of four was used in Figure 3B to distinguish regions which would pass the first check in cluster identification. It was found that in RBL-2H3 cells fixed with PFA both with and without GA (n = 40), all had clusters identified under the given parameters. Visual observations of clustering are also reflected within the functional dependence of density on distance from cluster centers, which we plot from r = 0 to r = 500 nm (Figure 3C & 2C). Clusters, first identified as regions of at least 4x the average cell density, persisted at or above this density out to 80 ± 20 nm on average. They also persisted with at least 2x the average cell density out to 220 ± 20 nm on average (Figure 3C). Though there was a difference in distribution going out beyond the size of the clusters (Figure 4C), which resulted in visually sharper images after PFA + GA fixation compared to PFA fixation (Figures 2A–B), quantification of cluster parameters was not significantly different between fixation conditions (Figures 2D–G). Quantifying the position error across frames (Supplemental Figure 1), similar values were found for both PFA fixation (0.052 ± 0.006 µm) and PFA + GA fixation (0.053 ± 0.006 µm), additionally indicating similar levels of fixation.

Figure 3. Quantification of Dendra2-PH clustering within PFA-fixed RBL-2H3 cells.

(A) Representative image of Dendra2-PH in PFA-fixed RBL-2H3 cells with high expression levels (> 100k localizations per sample). Images were acquired using TIRF-FPALM. (B) Density map of Dendra2-PH for sample shown in (A). Values correspond to density relative to average cell density (974 molecules/µm²) with values 4x cell average density and above shown as dark red. Zoomed in regions demonstrating nanoscale clustering are shown for both (A) and (B). Regions which meet the 4x cell average density threshold for cluster identification are indicated by black-magenta arrows; regions which have Dendra2-PH enrichment relative to average but are below 4x cell average density are indicated by purple-white arrows. (C) Radial distribution function (RDF) quantifying average Dendra2-PH distribution around the center of identified Dendra2-PH clusters for (n = 40) high expression cells. Clusters were identified as regions of molecules over 4x cell average density containing at least 0.02% of all localized molecules. An RDF value of one corresponds to a density equal to cell average.

Figure 4. Quantitative comparison of Dendra2-PH clustering between PFA and PFA + GA fixation methods in RBL-2H3 cells.

Representative images of RBL-2H3 cells labeled with Dendra2-PH using TIRF-FPALM after (A) 4% PFA fixation (n = 45) and (B) 3.9% PFA + 0.2% GA fixation (n = 45). Clusters were identified as regions at or above 4x the cell average density containing at least 0.02% of all localizations. (C) Cluster based radial distribution graph of Dendra2-PH using either 4% PFA or 3.9% PFA with 0.2% GA. Values over the range of 0–500 nm are shown. Differences in the distribution are checked for significance using a Komolgorov-Smirnov test represented by *p ≤ 0.05. Regions of significance indicated by gray bars. (D-G) Effects of fixation on cluster parameters between fixation methods are shown on violin plots with (n = 45) cells for each condition across three biological replicates. Significance tests were conducted using Mann-Whitney U-tests with ns representing no significance.

Density distributions were further quantified in live RBL-2H3 cells compared to fixed (Figure 5). A representative live cell image rendered for two equal time ranges is shown in Figure 5A. We consistently observed changing clusters over the course of imaging throughout the cell. Yellow arrows point to example regions where groups of molecules persist for only the first half of imaging (magenta) or the second half of imaging (green). Cluster formation can be seen by regions which are solely green, and cluster dispersion in regions which are solely magenta. Movement of clusters can be seen in regions with adjacent magenta and green regions and cluster growth/shrinkage is seen in regions of green/magenta which are adjacent to white regions, which represent areas which did maintained high density over time. Figure 5B shows a density map of all molecules of the cell represented in Figure 5A under the same relative density parameters shown in Figure 3B. It can be seen that in the radial distribution functions for live and fixed cell data (Figure 5C), live cell clusters demonstrated a lower radial density surrounding the clusters at short distances compared to fixed cell clusters at the same distances, while more slowly declining to the expected value of one at long distances. As we were only able to identify clusters in ~68% of live cells using a 4x density threshold and ~95% when using a 3x density threshold, clusters in fixed and live cells were quantified by both a 3x and 4x cell average density. Due to the dynamic nature of clusters in live cells, fixed samples will provide a better understanding of cluster parameters. Live samples will be expected to follow trends in cluster behavior, but changes will lose significance as they become averaged over evolving clusters. Trajectory information then becomes more relevant to live data.

Figure 5. Quantification of Dendra2-PH clustering and cluster evolution in live RBL-2H3 cells.

(A) Representative image of Dendra2-PH in live RBL-2H3 cells acquired by TIRF-FPALM. Molecules were divided into the first 50s of imaging (magenta) and the second 50s of imaging (green) to demonstrate the temporal evolution of clusters. Regions of overlap where localizations occurred during both periods of time can be seen in white. (B) Density map of Dendra2-PH for sample shown in (A). Values correspond to density relative to average cell density (2,909 molecules/µm²) with values 4x cell average density and above shown in dark red. (C) Radial distribution function of live and 4%-PFA-fixed cell data. Data is quantified using both a 3x and 4x cell average density threshold, with a minimum of 0.02% localized molecules required for all identified clusters.

To begin looking at the effects of CPC treatments, we will then first focus on clustering in fixed cells. The disruption of Dendra2-PH clustering due to CPC may be seen independent of fixation method (Figures 6 & 7). Clusters being identified in PFA fixed RBL-2H3 cells were also identified in cells exposed to 5 µM CPC for one hour before fixation. Effects on density distribution due to CPC were more difficult to discern within the microscale structures visible by eye (Figures 6A–B & 7A–B). However, quantification of nanoscale structures revealed stark differences between treatment conditions. CPC had a significant effect on the size of identified clusters as defined by their area and perimeter (Figures 6C, 6E, 7C, & 7E), as well as a modest effect on the cluster density relative to cell average (Figures 6D & 7D). Clusters can be seen to be getting smaller, less dense, and less elongated due to CPC exposure (Figures 6C–F & 7C–F). However, we note that cell density remains unchanged across all treatment conditions within given experiments (Supplemental Figure 6) and so clustering differences indicate rearrangement of Dendra2-PH within the cells.

Figure 6. Cationic compound CPC disrupts clustering of Dendra2-PH in PFA-fixed RBL-2H3 cells.

Representative images of Denda2-PH in RBL-2H3 cells using TIRF-FPALM with 1 hour exposure to either (A) 0 µM CPC in BT or (B) 5 µM CPC in BT. Clusters were identified in these cells using a 4x cell average density threshold with at least 0.02% of all localized molecules. (C-F) Cluster area, density, perimeter, and eccentricity for 0 µM CPC and 5 µM CPC data sets with (n = 40) cells for each condition across four biological replicates. Significance is represented by *p ≤ 0.05, ****p ≤ 0.0001 using Mann-Whitney U-tests.

Figure 7. CPC disruption of Dendra2-PH clusters in RBL-2H3 cells persists under PFA + GA fixation.

Representative images of RBL-2H3 cells fixed with 3.9% PFA + 0.2% GA and labeled with Dendra2-PH using TIRF-FPALM after (A) 1 hour exposure to 0 μM CPC in BT and (B) 1 hour exposure to 5 μM CPC in BT. For each condition, (n = 37) cells across three biological replicates were imaged. Clusters were identified as regions at or above 4x the cell average density containing at least 0.02% of all localizations. (C-F) Effects of CPC on cluster parameters between exposure levels are shown on violin plots. Significance is represented by *p ≤ 0.05, ****p ≤ 0.0001 using Mann-Whitney U-tests.

After testing the effects of CPC in fixed cells, the effects of CPC were observed in live RBL-2H3 cells (Figure 8). Cells treated with CPC appeared to more often have areas across the plasma membrane (PM) which were lacking in Dendra2-PH labeling. Clusters were observed using a 3x density threshold as opposed to the 4x density threshold used in fixed samples due to previously demonstrated cluster evolution in live cells (Figure 5A). A vast majority of cells under each condition (97% and 99% for 0 µM and 5 µM CPC treated cells respectively) had clusters identified. In the identified clusters, CPC caused a smaller, but still significant effect on cluster area (Figure 8C) and perimeter (Figure 8E) when compared to fixed samples (Figures 4C, 4E, 5C, 5E). However, the changes in density (Figure 8D) and elongation (Figure 8F) were not significant. The trend agreed with that seen in fixed samples with clusters generally becoming smaller after CPC exposure. To further expand upon changes to the distribution of Dendra2-PH due to CPC, the density distribution of Dendra2-PH inside cells was plotted for areas above cell average density (Figure 8G). These distributions show a more gradual change in densities after a CPC treatment demonstrated by a smaller magnitude of the slope in the linear decay fit (Figure 8H). This suggests CPC influences the organization of PIP₂ within the live cell.

Figure 8. CPC causes disruption of Dendra2-PH distribution and clustering in live RBL-2H3 cells.

Representative images of live RBL-2H3 cells labeled with Dendra2-PH using TIRF-FPALM after (A) 60–90 min exposure to 0 μM CPC in BT (n = 69 of 71 cells) and (B) 60–90 min exposure to 5 μM CPC in BT (n = 69 of 70 cells). Cells were imaged for each condition across five biological replicates; cells were excluded if no clusters were identified. Clusters were identified as regions at or above 3x the cell average density containing at least 0.02% of all localizations. (C-F) Effects of CPC on cluster parameters between exposure levels are shown on violin plots. Significance is represented by *p ≤ 0.05 and ns for no significance using Mann-Whitney U-tests. (G) Density distributions averaged across cells with points plotted for densities greater than or equal to the cell average. Data is binned by increments of 0.5x the cell average density and mean values are plotted. Error bars are plotted as a standard deviation of cell-to-cell variability. Log of the percent of cell area was fitted with a linear decay function. (H) Comparison of the slopes of the linear fits to quantify differences in cell density distributions across treatments. Significance is represented by ****p ≤ 0.0001 using an unpaired t-test.

Beyond clustering properties, mobility of Dendra2-PH in live cells was quantified under the effects of CPC treatments. The diffusion within cells treated with CPC is visualized to be higher throughout the cell compared to control (Figures 7A–B). The quantified diffusion of Dendra2-PH is significantly higher in cells treated with CPC (D = 1.00 ± 0.03 µm²/s) than in cells with no CPC treatment (D = 0.91 ± 0.03 µm²/s) (Figure 9C). Two populations of diffusion can be seen for both conditions (Figure 9C). The relationship between Dendra2-PH diffusion and local density depends on CPC treatment as well (Figure 9D). Low density regions of Dendra2-PH demonstrated a significant difference in diffusion coefficient ±CPC, but significant differences were not observed in higher density regions, represented by higher colocalization index (Figure 9D). In both cases, the highest densities had little or no motion discernable compared to localization uncertainty.

Figure 9. Dendra2-PH diffusion is altered in CPC treated cells.

Diffusion heatmaps of Dendra2-PH in representative live RBL-2H3 cells captured using TIRF-FPALM after 60–90 min exposure to (A) 0 μM CPC in BT (n = 71 cells) and (B) 60–90 min exposure to 5 μM CPC in BT (n = 70 cells). Data was binned into 60 nm x 60 nm regions. (C) Effects of CPC on diffusion coefficient shown on violin plots. Significance is represented by **p ≤ 0.01 using Mann-Whitney U-tests. (D) Plot comparing average trajectory step length to local molecular density for (n = 71) and (n = 70) cells for 0 μM CPC and 5 μM CPC respectively. Mean localization precision representing average estimated error and 2x mean localization precision representing the maximum estimated error are plotted in red. Significance of *p ≤ 0.001 is shown above the curves determined by Mann-Whitney U-tests and corrected by the Bonferroni-Holm multiple comparisons test. Regions of significance indicated by gray bars. (E-F) Turn angle histograms for Dendra2-PH trajectories. Simulated trajectories under ellipsoidal gradient confinement for 100 simulations are overlaid in green.

Histograms of the trajectory turn angles for both untreated and CPC treated samples were also plotted (Figures 7E–F). Brownian motion of the particles would result in a mean turn angle of 90º (89), however, these histograms (Figure 9E–F) had a bias towards higher turn angles, suggesting confinement of the motion. More so, we see in Supplemental Figure 4E, a mean turn angle of ~98º, supporting some sort of confinement. To probe the type of confinement, several types of simulated confinement modeled in (32) were compared against our data. We used the following experimental parameters (diffusion, time step, confinement zone size) to confine the model as much as possible and limit the number of free parameters. The diffusion coefficient was taken to be the mean cell diffusion coefficient from our data (Figure 9C) and a time step equal to our frame rate (100 Hz) was used. The size of the confinement zone was selected to be slightly larger than our identified clusters as a gradient distribution is often seen around clusters. Confinement was simulated to an r₀ of 220 nm and 250 nm for fitting to 0 µM and 5 µM CPC data respectively. An ellipsoidal gradient provided the best fit for the molecular confinement seen in turn angle histograms (green line in Figures 7E–F).

The role of Ag stimulation of mast cells on Dendra2-PH clustering can be seen beginning in Figure 10. These images represent cells stimulated with 0.7 ng/mL Ag for 5 minutes following 1 hour of IgE sensitization, and subsequently fixed. No obvious changes were observed by eye (Figure 10A–B). There was neither a change in the distribution of Dendra2-PH around identified clusters between stimulation conditions (Figure 10C) nor a change in any of the measured cluster parameters (Figure 10D–F) when observing the ventral side of the cell with TIRF-FPALM. With no significant change in cluster properties at the 5-minute time point, a set of live samples were observed over the course of 30 minutes, starting 5 minutes before stimulation, and every subsequent 5 minutes including immediately after stimulation (Supplemental Figure 2). Among images acquired, the 10-minute time point demonstrated the greatest difference in localizations normalized to the initial time point when comparing stimulated to unstimulated samples.

It was seen that upon 10 minutes of Ag stimulation, nanoscale clusters experienced growth, and that this growth was counteracted by CPC exposure when quantified in fixed samples (Figure 11). Additionally, Ag stimulated cells themselves (Figure 11B) spread out relative to unstimulated cells (Figure 11A), and this enlargement was countered by CPC (Figure 11C). Clusters in Ag stimulated cells were found to be larger in both area (Figure 11D) and perimeter (Figure 11F), with no significant change in density (Figure 11E) and elongation (Figure 11G) when compared to unstimulated samples. When exposed to CPC before Ag stimulation, clusters were significantly smaller than those without pre-exposure to CPC in both area (Figure 11D) and perimeter (Figure 11F) and were significantly less elongated (Figure 11G). Pre-exposure to CPC before Ag stimulation also had a significant decrease in cluster area (Figure 11D) and perimeter (Figure 11F) compared to cells which were neither stimulated nor exposed to CPC; however, cluster density (Figure 11E) and elongation (Figure 11G) did not change significantly.

Figure 11. Comparison of Dendra2-PH clustering after 10 minutes of antigen stimulation.

Representative images of RBL-2H3 cells fixed with 3.9% PFA + 0.2% GA and labeled with Dendra2-PH using TIRF-FPALM after (A) 1 hour IgE sensitization replaced by BT after 60 min and 120 min, (B) 1 hour IgE sensitization replaced by BT after 60 min and Ag stimulation after 120 min and C) 1 hour IgE sensitization followed by the treatment by 5 µM CPC CPC after 60 min and Ag stimulation + 5 µM CPC after 120 min. Ag stimulation duration was 10 minutes. Thirty cells across three biological replicates were imaged for each condition. Time points given are with respect to the time at which IgE sensitization began. Clusters were identified as regions at or above 4x the cell average density containing at least 0.02% of all localizations. (D-F) Effects of Ag stimulation CPC pre-exposure on cluster parameters are shown on violin plots. Significance is represented by *p ≤ 0.05, **p ≤ 0.01, ****p ≤ 0.0001 and ns for no significance using a Kruskal-Wallis test followed by a Dunn’s post-hoc test.

Quantifying clusters in live RBL-2H3 cells with CPC pre-exposure and Ag stimulation, we again saw enlargement of Ag stimulated cells (Figure 12B) relative to unstimulated cells (Figure 12A) and an inhibition of this enlargement when pretreating with CPC before Ag stimulation (Figure 12C). The cluster properties showed no significant difference under measured parameters (Supplemental Figure 3) but showed significant changes in density radially outward from clusters (Figure 12D–F). Dendra2-PH in Ag stimulated cells was seen to be distributed similarly to Dendra2-PH in unstimulated cells on short length scales (Figure 12D), whereas there was a significant increase in density compared to both these conditions when cells were pretreated with CPC prior to Ag stimulation (Figure 12E–F). These RDFs demonstrate CPC redistribution of Dendra2-PH in areas where Dendra2-PH is clustered. In addition to distributions around clusters, density distributions plotted for areas above cell average density were seen to vary depending on Ag stimulation and CPC treatment (Figure 12H). Specifically, we see a larger magnitude of slope in Ag stimulated and unstimulated cells when compared to those pretreated with CPC before Ag stimulation (Figure 12I). These data agree with the previously seen effects CPC has on these trends shown in Figure 8G–H and emphasize the impact CPC has on influencing cellular PIP₂ organization.

Figure 12. Comparison of Dendra2-PH distributions in live cells after 10 minutes of antigen stimulation.

Representative images of live RBL-2H3 cells labeled with Dendra2-PH using TIRF-FPALM after 1 hour of IgE sensitization followed by (A) 1 hour in BT, and then 10 minutes with additional BT added (n = 24 cells), (B) 1 hour in BT and then 10 minutes in 0.7 ng/mL Ag (n = 25 cells), and (C) 1 hour in 5 μM CPC and then 10 minutes in 0.7 ng/mL Ag with 5 μM CPC (n = 25 cells). Data was collected from three biological replicates. (D-F) Radial distribution function comparison between each pair of conditions for cells with identified clusters (n = 22/24, 18/25, 20/25 cells with clusters for each of the three respective conditions). Differences in the distribution are checked for significance using a Komolgorov-Smirnov test represented. (G) Significance for each of the RDF plots is represented by *p ≤ 0.05, **p ≤ 0.01, and ns for no significance. (H) Density distributions averaged across cells with points plotted for densities greater than or equal to the cell average. Data is binned by increments of 0.5x the cell average density and mean values are plotted. Error bars are plotted as a standard deviation of cell-to-cell variability. Log of the percent of cell area was fitted with a linear decay function. (I) Comparison of the slopes of the linear fits to quantify differences in cell density distributions across treatments. Significance is represented by **p ≤ 0.01, ****p ≤ 0.0001, and ns for no significance using an unpaired t-test.

In live cells, antigen stimulation had no noticeable effect on Dendra2-PH diffusion, but CPC increased Dendra2-PH diffusion (Figure 13A–C) like what was seen in Figure 9A and Figure 9B. In support of this, we quantify the Dendra2-PH diffusion in Figure 13D to see no significant change upon stimulation, but a significant increase in Dendra2-PH diffusion with CPC pre-treatment prior to antigen stimulation with respect to both stimulated and unstimulated samples. This trend agrees with what was observed in Figure 9C. Furthermore, we quantify trajectory step length with respect to local density, represented by colocalization index (Figure 13E–G), to find that Ag stimulation has no significant effect on diffusion at any length scale (Figure 13E), but that introducing CPC pretreatment before stimulation increases diffusion at low densities relative to both stimulated and unstimulated cases (Figure 13F–G). This agrees with the low-density effects seen in Figure 9D.

Figure 13. Dendra2-PH mobility increases after CPC treatment independent of antigen stimulation.

Diffusion heatmaps of Dendra2-PH in representative live RBL-2H3 cells captured using TIRF-FPALM after 1 hour of IgE sensitization followed by (A) 1 hour in BT, and then 10 minutes with additional BT added (n = 24 cells), (B) 1 hour in BT and then 10 minutes in 0.7 ng/mL Ag (n = 25 cells), and (C) 1 hour in 5 μM CPC and then 10 minutes in 0.7 ng/mL Ag with 5 μM CPC (n = 25 cells). Data was collected over three biological replicates and images were binned into 60 nm x 60 nm regions. (D) Effects of treatments on diffusion shown on violin plots. Significance is represented by ***p ≤ 0.001, ****p ≤ 0.0001 using a Kruskal-Wallis test followed by a Dunn’s post-hoc test. (E-G) Plot comparing average trajectory step length to local molecular density for (n = 24), (n = 25) and (n = 25) cells for IgE, IgE + Ag, and IgE + Ag + CPC respectively. Mean localization precision representing average estimated error and 2x mean localization precision representing the maximum estimated error are plotted in red. Significance of *p ≤ 0.001 is shown above the curves determined by Mann-Whitney U-tests and corrected by the Bonferroni-Holm multiple comparisons test. Regions of significance indicated by gray bars. (H-J) Turn angle histograms for Dendra2-PH trajectories. Simulated trajectories under ellipsoidal gradient confinement and hard-edge ellipse confinement are overlaid in green and magenta respectively, with 100x simulations for each confinement condition. A ratio of each pair of simulations ran was fitted to the live cell distribution and overlaid in cyan.

These datasets, however, were not able to be fitted using a single confinement simulation type (Figure 13H–J) unlike what was seen in Figure 9E–F. Instead, a combination of two confinement types was used to better fit the distribution of the collected data. Both the hard- and soft-edge ellipsoidal confinement types used were parameterized using the experimentally determined diffusion and the frame rate of acquisition. The hard-edge ellipsoidal confinement was parameterized using an r₀ of 200 nm for each of the three conditions, approximating the distance at which the radial density distribution begins to decline more slowly (Figure 12D–F). The soft-edge ellipsoidal gradient is parameterized using an r₀ of 450 nm for each of the three conditions, approximating the distance where the cluster-based distribution has become close to level (Figure 12D–F). Using these conditions, a weighted average of the fits which best matched collected data is shown in cyan (Figure 13H–J). These fits still did not sufficiently coincide with values from collected data within simulated variability but provided a better fit than either of the individual simulations, suggesting that there may be multiple types of confinement which control these molecules, or that there is a confinement which does not fall within the parameters or potential types tested. Nevertheless, the ratios of these confinement types provide information about the differences between each condition. Cells stimulated with Ag without CPC treatment show significantly different ratios of fitted confinement types to cells stimulated with Ag after CPC treatment (Table 2).

Table 2.

Percentage of each simulated confinement was used to provide the best fit of imaging data when using a weighted average of simulations. Error is given by the standard deviation of the fit ratios across 100 simulations.

	Percent Ellipsoidal Gradient	Percent Hard Edge Ellipse
IgE	34 ± 2%	66 ± 2%
IgE + Ag	31 ± 2%	69 ± 2%
IgE + Ag + CPC	38 ± 2%	62 ± 2%

4. DISCUSSION

4.1. Nanoscale Clusters of free PIP₂ are discovered in RBL-2H3 Cells

PIP₂ plays a role in a wide range of cellular processes (90–92). However, the nanoscale distribution of PIP₂ has yet to be completely clarified (12,39,91,93,94), especially in live cells. Therefore, we decided to use super-resolution to observe PIP₂ in a physiologically relevant cell type, the RBL-2H3 mast cell immune cell type.

Whether PIP₂ has a homogeneous distribution is a matter of debate (24,95,96). Using Dendra2-PH, which we have used to study PIP₂ distributions in other cell types previously (32,71,97,98), we set out to examine the density distributions of free PIP₂ which may readily be available for cellular processes such as PLC cleavage for downstream mast cell degranulation (68,99,100). The aim was to study whether there are higher density pools of this lipid which may be used for specific cellular processes such as immune signaling or reactions to exogenous agents. As was seen in Figure 3A–B, there is significant variation in density of the Dendra2-PH probe. This prompted us to use a strict set of cluster identification parameters like those used in (32).

Study of PIP₂ clusters remains of interest for quantifying functionally distinct pools of PIP₂ (17), and PIP₂ clusters have been observed at the nanoscale (29–37) (smooth muscle cells, PC-12, B cells, NIH-3T3, HEK-293, tobacco plant cells, C2C12). However, the clusters have either not been quantitatively characterized or have not been linked to physiological function with the exceptions of PIP₂ clustering’s role in syntaxin-1A sequestration in PC-12 cells (29,30) and quantitative characterizations of PIP₂ in live HeLa cells (37,39). Though cation-promoted PIP₂ clustering affects membrane curvature (62) upon which degranulation depends (63), there is a need to further study PIP₂ clustering at the nanoscale as its functional importance is not well known in immune mast cells (41). We are one of the first few studies of nanoscale PIP₂ clustering relating to function in live cells.

Other researchers observed the distribution of PIP₂ in fixed INS-1 cells was better representative of live cell results when using PFA + GA vs using PFA alone (24). Thus, we carried out experiments using samples with both of these fixation methods. The clusters identified were found not to be significantly affected by the fixation method used (Figure 4D–F); Figure 4C similarly shows density around clusters of 4x cell average persisting for similar distances under both fixation conditions. While the chemical properties of GA may contribute to sharper images (Figure 4B), these 2 fixation methods did not alter the quantitation of our PIP₂ cluster results.

After fixation, we found molecular diffusion as shown by the average movement between frames (Supplemental Figure 1) was not significantly different when comparing fixation methods. This agrees with the results in (101) which demonstrate that for cytoplasmic lipid-anchored proteins, a large majority of the proteins were immobile when fixed using PFA with or without GA. Combining cluster quantification and diffusion results, we believe that the addition of GA provides better conservation of structures during the fixation process even though our identified cluster parameters and the overall level of fixation at the end were not significantly different.

Next, we explored the distribution of PIP₂ in live cells. Previous work examined differences between fixed and live cell PIP₂ distribution in INS-1 cells (24). The authors quantified a homogeneous spatial distribution and fast dynamics when looking at live INS-1 cells. In contrast, we found a clustered distribution (Figures 3, 6, 10) in each of our live cell experiments.

As seen in Figure 5A, we observed a dynamically changing Dendra2-PH cluster distribution in live cells. The diffusion coefficient for PIP₂ falls within a range of 0.1 to 1 µm²/s (12,39,102–105). At this rate of motion, we expect free PIP₂ to have mobility great enough to easily travel beyond the size of our clusters and yet the clusters persist over time. Additionally, many phosphoinositide binding molecules dissociate on the order of milliseconds to seconds (12,103,106,107), allowing exchange of PIP₂ between our probe and the functional proteins to which it may bind. In fact, if there were momentary spontaneous cluster formations, we might expect that clusters may become averaged over with the timescales we imaged on. However, we can visualize in Figure 5A–B that clusters of molecules persist and evolve over time, and that there are high density regions that may persist over the course of the entire 100s acquisition process. Persistent clustering is not specific to cells imaged at RT, as clusters were also observed later in Figures 6 & 9 for cells imaged live at 37°C/5% CO₂. It should be noted, however, that as cluster evolution is observed, cluster density may increase or decrease as the molecules move over a larger area. This is highlighted in Figure 5C, where we identify the number of cells with identified clusters meeting different density conditions, and the distribution around said clusters. To account for the evolution of Dendra2-PH clusters with time, we decreased density thresholds to 3x the cell average when performing cluster analysis. However, our key finding is that PIP₂ forms clusters in live RBL-2H3 cells, which persist on the timescale of minutes.

4.2. Cetylpyridinium Chloride Disrupts Nanoscale PIP₂ Clusters and Increases Dendra2-PH Diffusion

Cetylpyridinium chloride, a cationic quaternary ammonium compound, has been found to disrupt mast cell degranulation, including under the conditions used in this current study (72), through interference of calcium mobilization in RBL-2H3 and human LAD2 mast cells (108,109) at low micromolar doses. This inhibition may be related to CPC interference with clusters of PIP₂ including interferences with PIP₂-binding partners (71,110), which was observed in NIH-3T3 fibroblasts under conditions similar to those of the current study. As PIP₂ hydrolysis is the precursor to IP₃ signaled Ca²⁺ release (99), we predicted that the calcium disruption in mast cell degranulation may be due to CPC disruption of PIP₂. This hypothesis is further supported by the potential for electrostatic interaction between the negatively charged headgroup of PIP₂ and positively charged CPC.

In order to investigate whether CPC disruption of PIP₂ underlies CPC inhibition of mast cell function, we first aimed to quantify whether CPC influenced Dendra2-PH clustering in RBL-2H3 cells like what was seen for PAmKate-PH in NIH-3T3 cells (71). Exposure to 5 µM CPC (one hour) decreases area, perimeter, density, and elongation of Dendra2-PH clusters in cells, following the CPC treatment, fixed with either PFA alone (Figure 6C–F) or with PFA + GA (Figure 7C–F). These results indicated that CPC does disrupt the ability of PIP₂ to cluster on the nanoscale, regardless of fixation method tested.

In live RBL-2H3 cells imaged at 37ºC and 5% CO₂ transfected with Dendra2-PH, exposure to CPC still decreases cluster area and perimeter (though less significantly), but CPC does not significantly affect cluster density and eccentricity (Figure 8C–F). Importantly, though, the trend of smaller clusters remains. It is here that we emphasize the effects due to cluster evolution. There may be shorter timescale disruptions due to CPC which become averaged over during a live cell acquisition. This becomes especially true when the presence and location of the clusters themselves change over the timescales in which the data is acquired. Additionally, it is somewhat expected that CPC may play a smaller role disrupting unstimulated cells as CPC does not affect spontaneous degranulation in unstimulated mast cells when using doses between 1 pM to 10 µM (72). Thus, CPC may possess multiple molecular targets in the cell, including the disruption of PIP₂ clustering as still seen in Figure 8C & 8E. Independent of the other roles CPC may play, we discover that CPC disrupts clusters of Dendra2-PH.

Independent of the role of PIP₂ clustering, we see significantly different density distributions in CPC treated cells compared to controls (Figure 8G–H). This analysis picks up distributional changes at density levels typically ignored or biased by cluster quantification due to strict density cutoffs. The density distributions follow an exponential decay, which we may observe from noting that the linearly fitted data is a log of the actual values. It is important to note that there is not a significant difference at any specific density, but rather, in the distribution of the densities. Thus, it is more sensitive to smaller changes in the cells’ overall density levels than density thresholded approaches. This methodology provides an additional way to quantify distributional changes of molecules without relying on the need to identify clusters using a strict set of parameters. Moreso, the trend of CPC disruption of Dendra2-PH in RBL-2H3s can be expanded from disruption of clusters to disruption of overall distribution.

Beyond affecting the distribution of Dendra2-PH, CPC also increases the diffusion coefficient of these molecules (Figure 9C). In samples with 0 µM CPC exposure, we see two apparent distributions of diffusion. One is centered around 0.5 µm²/s and the other slightly below 1 µm²/s (Figure 9C), both of which fall within the PIP₂ diffusion of 0.1 to 1 µm²/s measured by other groups (12,39,102–105). We noticed these same two distributions in cells exposed to CPC but with an upward shift in average diffusion. We dissected the effects CPC has on diffusion at different densities in Figure 9D. It is striking that the CPC effects occur most significantly at lower densities (colocalization index); one explanation could be that tighter grouping of Dendra2-PH is preventing CPC from accessing certain areas and preventing disruption. We also note that higher density regions are more sparse, so the error grows. To denote the range at which molecules may be considered stationary, we provide both the mean localization precision and 2x the mean localization precision (plotted over the uncertainty ranges). These values are meant as an upper bound on apparent change in position in a stationary molecule due to localization error. At and below the upper bar (Figure 9D) are values which may be stationary within error, and values at or below the lower bar are considered fully stationary.

Interestingly, it has been shown that positively charged molecules may affect diffusion of negatively charged molecules in membranes (111,112). Similarly, we observed a change in diffusivity of PH upon addition of positively charged CPC (Figures 7C and 11D). PIP₂ has a negatively charged headgroup which controls many cellular functions through interactions with proteins (91,93,94,113,114). Introducing positively charged CPC could lead to screening of the negatively charged PIP₂ headgroups. As PIP₂ has an electrostatic interaction with the positive region of PH (115), CPC may affect PH binding to PIP₂ at the PM. A decrease in PIP₂ association with PH at the PM would be expected to result in increased diffusivity of the PH Domain (116), which we observe (Figures 7C and 11D). In context of our results, CPC sequestration of PIP₂ away from PH would be expected to increase the diffusion of Dendra2-PH. Our increase in Dendra2-PH diffusion therefore suggests an interaction between CPC and PIP₂.

We have previously shown that CPC interferes with MARCKS binding to PIP₂ at the PM (71), where it normally sequesters PIP₂ until needed for cellular functions (117). MARCKS utilizes its polybasic effector domain to sequester multiple PIP₂ molecules at a time, so CPC may disrupt this electrostatic interaction. Electrostatic interactions between PIP₂ might disrupt other proteins which use polybasic domains to function or be recruited to the PM. For example, Cdc42 is a Rho-GTPase who’s polybasic diarginine motif at its C-terminus allows attachment to PIP₂ containing membranes (118) and is necessary for the Ca²⁺ and PIP₂ oscillations produced during mast cell degranulation (119). PIP₂ also co-clusters with the SNARE protein syntaxin-1A via its polybasic domain. These clusters act in concert with synaptotagmin-1 to mediate membrane fusion and exocytosis (120). CPC may electrostatically disrupt interactions between PIP₂ (along with other negatively charged molecules) and the polybasic domains of enzymes involved at different points in mast cell degranulation.

4.3. Antigen Stimulation Induces Nanoscale PIP₂ Cluster Growth and is Countered by Cetylpyridinium Chloride

Perhaps our most pressing question is the functional role of the newly identified PIP₂ clusters. As noted previously, there is evidence for functionally distinct pools of PIP₂ which respond to IgE receptor activation (17). Additionally, IgE receptor FcεRI redistributes after Ag stimulation (121). Furthermore, IgE-FcεRI crosslinking by Ag leads to lipid co-redistribution with clustered receptors (41,122). All of these findings suggest that PIP₂ clustering may play a functional role in mast cell degranulation. Even more so, lipid co-redistribution suggests a potential mechanism for cluster evolution.

We initially investigated the effects due to 5 minutes of 0.7 ng/mL Ag stimulation conducted at room temperature on Dendra2-PH in RBL-2H3 cells (Figure 10). To our surprise, we found no significant difference in any of our cluster parameters (Figure 10D–F) or the radial distribution around these clusters (Figure 10C). To probe when stimulation may have the greatest effect on PIP₂ levels, we tracked cells over the course of 25 minutes of stimulation (Supplemental Figure 2). As PLCγ should deplete free PIP₂ at the PM under antigen stimulation, we expected a greater decrease in localizations relative to the initial imaging of the cell when under the peak effects of the stimulation. We found this biggest decrease to be at 10 minutes (Supplemental Figure 2) and further continued to investigate this question.

To probe both the effects of Ag stimulation and the potential for CPC to disrupt these effects, we imaged a combination of these conditions (Figure 11). We found that 10 minutes of Ag stimulation resulted in an increase in area and perimeter of our nanoclusters with no significant change in density or elongation compared to the unstimulated case (Figure 11D–F). We note, however, that the visually identifiable microdomains of Dendra2-PH appear more continuous compared to the unstimulated case (Figure 11A–B). The redistribution may be a result of PLC cleavage of PIP₂ or a result of recruitment to different sites after antigen stimulation. However, the exact cause is not yet known. Conversely, when we pre-exposed cells to 5 µM CPC for an hour before stimulating with 0.7 ng/mL of Ag for 10 minutes, we found clusters which were not significantly different than those seen in the unstimulated case. This suggests that the CPC is inhibiting the effects Ag stimulation has on these clusters.

As with previous conditions imaged, we expanded into live cell imaging to further probe the effects of ± Ag, ± CPC. No significant changes in cluster properties, in high density regions, were found in live cells due to Ag ± CPC in IgE sensitized cells (Supplemental Figure 5). However, the trend of increasing area/perimeter with Ag stimulation and decreasing area with Ag stimulation following CPC pre-exposure persists. Then, to better quantify the difference in density distributions farther away from cluster centers, we used RDFs (Figure 12D–F). We see that, outside of the clusters, Ag stimulation causes a depletion of Dendra2-PH from unstimulated levels (Figure 12D), which may suggest recruitment towards the nanoscale clusters, as suggested by fixed cell data (Figure 11D & 11F). In Figure 12F, we see a pretreatment with CPC negates and potentially reverses short range changes outside of clusters which would be caused by subsequent Ag stimulation, as seen by the similarity in density compared to unstimulated samples. This treatment also results in a significant increase in density at long distances compared to unstimulated samples. Combining this result with Figure 12E, we observe CPC is causing a general increase in PH density outside clusters, potentially due to depleting PH levels in clusters (Figure 11D & 11F), which results in increased levels of PH in the surrounding area.

In addition to changes in Dendra2-PH clustering, we observed an overall change in Dendra2-PH distribution between Ag stimulated cells with and without CPC treatment (Figure 12H–J). Though the fraction of the cell at a particular density followed an exponential decay relationship with the density relative to cell average (represented by a linear fit of the log of these values), CPC treatment before Ag stimulation slowed the rate of this decay (Figure 12H–I). Looking at individual points, we see this difference is seen primarily at higher densities; regions 1–2x above cell average remained relatively similar (Figure 12H). Compared to the control of IgE sensitization, CPC treatment before Ag stimulation had effects opposite to those caused by Ag stimulation without CPC treatment (Figure 12H). This continues to support the finding that Ag stimulation effects on PIP₂ are inhibited by CPC treatment.

CPC significantly increases the diffusion coefficient of PH in live cells either with or without antigen (Figures 7C & 11D). These results further support the idea of CPC having an electrostatic interaction with PIP₂. As the PH Domain of PLC allows it to anchor to PIP₂-rich membranes for PIP₂ hydrolysis (115,123–125), CPC disruption of PH-PIP₂ interactions is indicated as a prior step to previously observed inhibition of Ca²⁺ release needed for mast cell degranulation (72). We also see that CPC again has significant effects on diffusion only at lower densities (Figure 13F–G) in agreement with what was seen in Figure 9D. Combining this with the CPC disruption of Ag stimulation effects at the lower density regions surrounding clusters, we may confidently conclude CPC is disrupting key pools of PIP₂ relevant to mast cell degranulation.

More research is needed on how this perturbation of PIP₂ cluster distribution affects mast cell function.

4.4. Dendra2-PH Confinement is Well Modeled by Ellipsoidal Potentials

As seen in Figures 7E–F and 11H–J, Dendra2-PH has a turn angle distribution which is mostly level for angles 0–90 degrees (with a slight bump centered around 0 degrees) but slopes upwards for larger turn angles. This is similar to simulated confinement in (32) and agrees with findings in (39) which finds a tendency towards more obtuse angles. The findings of (39) measuring mean turn angles between 97º and 102º overlap with our findings of mean turn angles between 94º to 98º for Dendra2-PH, noting that our results are more similar to their Tubby_c angles than their PH-PLCδ1 values. Similar to what was seen in (39), our results also indicate that there is some degree of confinement of the Dendra2-PH probe used.

To further quantify the confinement observed, we simulated the same confinement types from (32). The two confinement types that appeared to provide the best fits were ellipsoidal gradient and hard-edge ellipse potentials. These potentials are also physically relevant as our identified clusters consistently demonstrated elongation. When comparing 0 µM CPC to 5 µM CPC treatment conditions, we were able to successfully fit the range of turn angles with only an ellipsoidal confinement simulation (Figure 9E–F). We also can model this confinement with a cosine plus gaussian function (Supplemental Figure 4). This suggests that there is a potential acting on these molecules to result in the gradient distribution seen around clusters. We were able to closely fit our turn angles when comparing IgE vs IgE + Ag vs IgE + Ag + CPC conditions, but this required a combination of two potentials (Figure 13H–J). This suggests multiple confinements may be present and acting on Dendra2-PH under these conditions but may also be due to not finding the correct parameters to adequately simulate this confinement. A better fit can be obtained when simulating a cosine plus gaussian function but does not account for types of confinement (Supplemental Figure 5). Even without an ideal fit, this fitting of confinement models allows for comparison between conditions to provide better understanding of changes in confinement between conditions when more than one model is needed to approximate the confinement.

4.5. Dendra2-PH Dynamics Calls for Better Imaging of Bound/Free PIP₂

Many of the molecules we observe are not in clustered regions. This indicates that they undergo free diffusion, similar to the findings in (39). We must consider what this implies for PIP₂. First, the confined subset likely consists of Dendra2-PH following PIP₂ interacting with clusters, as the highest densities were seen to be the only regions with limited diffusion (Figures 7D and 11E–G). Secondly, we observe cluster evolution. The high levels of Brownian diffusion, indicated by a mean turn angle near 90º, suggest that the molecules should end up randomly distributed and non-clustered with time. However, we consistently observe molecules which are clustered, demonstrating that this is not the case. To better understand how these clustered regions are maintained with time, methods for imaging PIP₂ without binding to a probe would be very helpful.

4.6. Summary

Using the probe Dendra2-PH, we drew several conclusions about PIP₂. Free PIP₂ clusters on the nanoscale and these clusters evolve over time. Both clusters, and the overall distribution of PIP₂ get disrupted by CPC. The disruptive effects of CPC work against distributional changes of PH induced by Ag stimulation to the extent of inducing inverse effects. CPC also increases the diffusion of Dendra2-PH, likely by disrupting electrostatic interactions between PIP₂ and other binding partners such as Dendra2-PH. The Dendra2-PH molecule demonstrates some level of confinement, indicated to occur primarily at higher densities (and thus areas of PIP₂ enrichment), and which may be modeled by ellipsoidal potentials. With all these findings specifically quantifying free PIP₂, there is still need for a probe which may image and quantify bound PIP₂. These experiments open up the path to PIP₂ clustering and dynamics quantification in other cell types and in additional cell processes or external factors which may elicit changes to PIP₂ distributions.

Table 1.

Treatments and fixation/live conditions applied to RBL-2H3 cells for the experiment represented by each figure.

	Treatments			Fixation/Live Conditions
Figure	IgE	CPC	Antigen	PFA	PFA + GA	Live Cells (Room Temp)	Live Cells (37ºC/5% CO2)
1				X
2				X	X
3						X
4		X		X
5		X			X
6		X					X
7		X					X
8	X		X	X
9	X	X	X		X
10	X	X	X				X
11	X	X	X				X

Statement of Significance:

Deepening the understanding of lipids in immune cell function is essential in understanding the key processes involved in the responses of the body to external agents and diseases. By furthering our understanding of PIP₂, a key signaling lipid for immune response among other roles, along with the role of PIP₂ in immune cell function, we aim to open doors for new therapies and treatments supporting these responses, and to reduce the use of various chemicals which interfere with these essential functions. By introducing treatments of exogenous agents to our cells, we build an understanding of effects, both good and bad, which exposures at similar levels may have.