Leader cell PLCγ1 activation during keratinocyte collective migration is induced by EGFR localization and clustering

Abstract Re‐epithelialization is a critical step in wound healing and results from the collective migration of keratinocytes. Previous work demonstrated that immobilized, but not soluble, epidermal growth factor (EGF) resulted in leader cell‐specific activation of phospholipase C gamma 1 (PLCγ1) in HaCaT keratinocytes, and that this PLCγ1 activation was necessary to drive persistent cell migration. To determine the mechanism responsible for wound edge‐localized PLCγ1 activation, we examined differences in cell area, cell–cell interactions, and EGF receptor (EGFR) localization between wound edge and bulk cells treated with vehicle, soluble EGF, or immobilized EGF. Our results support a multistep mechanism where EGFR translocation from the lateral membrane to the basolateral/basal membrane allows clustering in response to immobilized EGF. This analysis of factors regulating PLCγ1 activation is a crucial step toward developing therapies or wound dressings capable of modulating this signal and, consequently, cell migration.

Cells at different locations within the collective sheet display different phenotypes during collective migration. 4,5 Once established, leader cells are crucial for coordinated migration and exhibit different signaling patterns, such as elevated Rac1 and PI3K activity. 5 Variations in the microenvironment such as extracellular matrix (ECM) density or the presence of gradients of soluble growth factors can induce or reinforce the behaviors of leader cells. 2,6 However, our work and others have demonstrated that keratinocytes also exhibit increased collective migration and differential signaling in leader cells when treated with a uniform concentration of growth factors in vitro. For example, epidermal growth factor (EGF) and transforming growth factor β1 (TGFβ1) are both found in the wound microenvironment and induce re-epithelialization in vitro. 7 Our findings revealed that EGFinduced wound closure was dependent on the manner of growth factor presentation, with wounds closing significantly faster on EGF covalently immobilized to the culture substrate, a presentation that mimics ECM-entrapped growth factors. 8 The increased closure obtained with immobilized EGF was a result of individual cells at the leading edge having increased migrational persistence and directionality into the wound. 9 We determined that this unique migratory phenotype resulted from the activation of phospholipase C gamma 1 (PLCγ1), which was highly specific to the leader cells and observed only in response to immobilized EGF. 9 Similarly, Chapnick and Liu recently demonstrated that spatial restriction of ERK activation to the leading edge of keratinocytes resulted in highly directional migration reminiscent of TGFβ stimulation. 10 However, while both of these studies highlight that unique intracellular signaling in leader cells has phenotypic consequences for wound healing, neither study determined what conditions or properties enabled this localized signaling.
Here, we examine several hypotheses and identify differences in receptor localization and clustering as responsible for the spatial restriction of pPLCγ1 to the leader cells on immobilized EGF. Elucidation of these mechanisms provides insight into wound biology, as well as important guidance for the development of biomaterials-or tissue engineering-based strategies to improve re-epithelialization.

| Immobilized EGF increases extent of PLCγ1 activation
We have previously demonstrated that the number of pPLCγ1 positive cells is significantly increased for cells on immobilized EGF along the wound edge relative to both cells in the bulk and to cells exposed to control conditions or an equivalent dose of soluble EGF. 9 To further probe this finding, we analyzed the average pPLCγ1 intensity for cells along the wound edge for HaCaTs treated with soluble EGF or immobilized EGF at 4 hr. As expected, the pPLCγ1 intensity was comparable to background levels in control conditions, while pPLCγ1 intensity was significantly elevated by treatment with EGF and was further elevated when cells were exposed to immobilized rather than soluble EGF (Figure 1a, Supporting Information Figure 1). Therefore, in addition to our prior finding that more edge cells demonstrate clear activation of PLCγ1 on immobilized EGF compared to soluble EGF, 9 the extent of activation within an individual cell is also increased with immobilized EGF.

| PLCγ1 activation does not result from larger cell areas observed near the wound edge
To understand why the cells on the edge have increased activation of pPLCγ1, we considered some of the prior explanations for wound edge-specific behavior as well as differences between the edge and bulk cells. Edge-specific activation of leader cells has been observed in response to chemotactic gradients 6 ; however, all of the cells in the keratinocyte sheet in our experiments were exposed to a uniform concentration of immobilized EGF. It is widely recognized that keratinocytes at the wound edge undergo hypertrophy, 11,12 and previous reports using uniform stimuli have linked differences in leader cell signaling to variations in cell size 13 or the related property of cell density. 10 To examine the possibility that cell size regulates the edgespecific PLCγ1 phosphorylation found in HaCaTs treated with immobilized EGF (Figure 1b), cell areas were measured based on actin staining. Consistent with prior studies, 11 cells on the wound edge had larger areas (Figure 1c). In addition, cells on immobilized EGF had significantly larger cell area at the edge when compared to all other groups, suggesting a possible link between cell area and PLCγ1 activation.
However, the distribution of cell sizes between all conditions overlapped; in particular, edge cells on immobilized EGF were only slightly larger than edge cells treated with soluble EGF. Therefore, we conducted a detailed analysis of the edge cells on immobilized EGF ( Figure 1d) and determined that there was not a significant difference in cell area between pPLCγ1-positive and pPLCγ1-negative cells ( Figure 1e). This result suggests that increased cell area was not responsible for the increased activation of pPLCγ1 on immobilized EGF.

| PLCγ1 activation requires a decrease in tight junctions with neighboring cells
We next examined whether differences in cell-cell connections could play a role in the observed activation of PLCγ1, as cells at the leading edge need to remodel their tight junctions in order to migrate. 14 Epidermal growth factor receptor (EGFR) activation has been shown to increase tight junction assembly in confluent cells, 15 but did not impact zonula occludens-1 (ZO-1) expression or localization. 16 Alternatively, cytokines that disrupt tight junctions in airway epithelial cells do so through EGFR activation of ERK. 17 However, the role of tight junctions in PLCγ1 activation is not known. Cells were co-stained for ZO-1, one component of tight junctions in keratinocytes, 18 and pPLCγ1 ( Figure 2a). Cells were quantified as pPLCγ1-positive and as ZO-1 positive based on the ratio of membrane: cytoplasmic signal ( Figure 2b). This classification demonstrated that all cells that were pPLCγ1-positive were also ZO-1 negative, and that this was significantly different compared to a random distribution (Figure 2c). To determine if loss of tight junctions was sufficient to induce pPLCγ1 in cells located in the bulk, cells on immobilized EGF were treated with ochratoxin-A, a mycotoxin that has previously been shown to disrupt tight junctions. 19 As expected, treatment with ochratoxin-A resulted in a shift in ZO-1 staining from membrane-localized to diffuse or nearly absent throughout the cell body for bulk cells (Figure 2d). However, there was no increase in pPLCγ1 staining in the bulk despite the loss of tight junctions, suggesting that breakdown of tight junctions alone was not sufficient to induce activation. This may indicate that other cell-cell junctions that are unaffected by ochratoxin are involved in the mechanism. Therefore, we plated cells at a low seeding density and stimulated them with EGF before they had time to reach confluency and form strong cell-cell junctions (Figure 2e). In this setting, pPLCγ1 was again only observed when cells were cultured on immobilized EGF, but not all cells in this condition were pPLCγ1-positive. Therefore, we conclude that the loss of ZO-1 containing tight junctions is necessary, but not sufficient, for PLCγ1 activation.

| EGFR localization varies between edge and bulk cells
Tight junctions serve as adhesive contacts among epithelial cells and restrict membrane protein movement between apical and basolateral membranes. 20 Therefore, we next examined whether EGFR localization varied between cells along the leading edge and cells in the bulk. Cells in all three conditions (control, soluble EGF, immobilized EGF) were examined by confocal microscopy for EGFR ( Figure 3a), and z-stack reconstructions analyzed for patterns in EGFR signal per cell. Other reports have suggested that leader cells are associated with differences in the levels of cell surface receptors, which could lead to increased activation in response to a stimulus. 5 Examination of the images suggested that EGFR may be lower at the wound edge; however, due to differences in cell size (i.e., larger area at the edge) it was possible that cells with a weaker intensity/pixel had the same or more EGFR when integrated across the cell volume. In our system, the integrated intensity for EGFR over the entire z-stack of the cell was less than 5% different for edge versus bulk cells (data not shown), suggesting there was not a major difference in expression level. However, when the percentage of the total EGFR signal per cell located in each plane (from the basal to the apical side) was quantified, a dramatic difference in protein localization was observed ( Figure 3b).
For all three treatments, cells in the bulk had evenly dispersed EGFR throughout the height of the cell. In contrast, cells located on the wound edge showed a significantly increased level of EGFR in the planes closest to the basal membrane ( Figure 3c); this pattern did not F I G U R E 1 pPLCγ1 activation is not due to variations in cell area. (a) HaCaTs were seeded as confluent monolayers and treated with vehicle control, soluble EGF, or immobilized EGF for 4 hr after the fence was lifted, and then stained for pPLCγ1 (representative images provided in Supporting Information Figure 1). pPLCγ1 intensity was quantified in cells adjacent to the wound edge. Data presented as individual cells (n = 35-116 cells/condition) with mean ± SD shown as lines. * indicates significantly different relative to control;^indicates significantly different relative to soluble by Tukey-HSD, p < .05. suggesting that EGFR relocation occurs either concurrent with or following the release of cellular junctions. This mechanism of differential receptor localization provides another potential mechanism to regulate leader cell-specific behaviors. Indeed, localization of EGFR to the basal side may play an important role in native wound healing, as the sources of EGF in the wound microenvironment (e.g., macrophages 21 ) are located under the keratinocyte sheet. The basal localization of EGFR may also limit the responsiveness of keratinocytes to topically-applied EGF in wound treatment, potentially explaining the poor efficacy of this approach. 22   Based on the observed differences in EGFR staining using confocal imaging, we next imaged the basal membrane for EGFR using stimulated emission depletion microscopy (STED) in order to gain improved resolution and characterize patterns in cluster size. In the bulk, EGFR was localized primarily to the basolateral junction, and the signal was evenly distributed with few apparent regions of higher concentration indicates significantly different relative to soluble by t test with Welch's correction for unequal variance, p < .05 F I G U R E 5 Methyl-β-cyclodextrin induced EGFR clustering and PLCγ1 activation in response to soluble EGF. (a) Cells were seeded at a low density and treated with vehicle (V) or 50 mM methyl-β-cyclodextrin (MβCD) for 4 hr. Cells were examined by STED for EGFR (red) and pPLCγ1 (green). Scale bar = 5 μm. (b) Quantification of EGFR cluster sizes for each condition. Data presented as individual clusters with the number of clusters noted for each condition and mean ± SD shown as lines. * indicates significantly different relative to vehicle control for that condition, p < .05 by t test with Welch's correction for unequal variance and Bonferroni correction for multiple comparison. (c) Quantification of pPLCγ-1 fluorescent intensity in each condition. Data presented as mean ± SD, n = 9 images/condition. * indicates significantly different relative to vehicle with same EGF treatment;^indicates significantly different relative to control with MβCD; # indicates significantly different relative to control with vehicle; p < .05 by Tukey-HSD (Supporting Information Figure 2). For cells on the wound edge, the distribution of EGFR in control conditions was not restricted to the basolateral membrane, but appeared as scattered EGFR signal without obvious clusters (Figure 4a). Cells treated with soluble EGF showed some areas of potential receptor clusters, while cells on immobilized EGF had large areas of dense EGFR signal (Figure 4a). Previous studies have defined receptor clusters as ranging from 0.07 to 3 μm 2 23,24 ; therefore, we classified any signal with an area less than 0.1 μm 2 as an isolated receptor and quantified the sizes of all clusters (Figure 4b).

| EGFR clustering is induced on immobilized EGF
Few EGFR clusters were observed in cells in the bulk regardless of treatment condition (Supporting Information Figure 2

| Cholesterol depletion promotes EGFR oligomerization and PLCγ1 activation by EGF
As noted above, EGFR clusters were substantially larger in cells treated with immobilized EGF relative to soluble EGF, and, similar to the localization pattern observed for pPLCγ1, this clustering was localized to cells on the edge. Based upon these findings, we hypothesized that induction of EGFR clustering would enable soluble EGF to activate PLCγ1. Methyl-β-cyclodextrin is a cholesterol-depleting agent that disrupts the lipid rafts on the plasma membrane 28 and has previously been shown to induce EGFR clusters in HaCaTs. 29 To maximize the number of cells that were capable of inducing PLCγ1, we exam-  (Figure 5b). We next examined PLCγ1 activation and saw no difference in pPLCγ1 in the control condition, indicating that receptor clustering alone was insufficient to induce EGFR activation of PLCγ1 in the absence of ligand ( Figure 5c).
As predicted, pPLCγ1 was significantly increased in cultures treated with methyl-β-cyclodextrin in combination with soluble EGF. Likewise, methyl-β-cyclodextrin increased receptor oligomerization and pPLCγ1 levels for the immobilized EGF condition. These findings support our hypothesis that induction of EGFR clustering-either by ligand immobilization or through cholesterol depletion-enables EGF to activate PLCγ1 in HaCaTs.

| Conclusions
Combined, our data suggest that differences in EGFR apical/basal localization coincident with the release of cellular junctions are responsible for the edge-specific activation of pPLCγ1, while the ability to form EGFR clusters is responsible for the specificity for immobilized over soluble EGF. While we are unaware of previous work connecting EGFR clusters specifically to PLCγ1 activation, it is known that EGF family ligands have varying affinities for EGFR, with only some ligands having the ability to activate PLCγ1. 30 Ligand affinity has also been shown to impact the stability of receptor dimers, with downstream effects on signal duration and cell fate. 31 These prior findings are consistent with our finding of activation of PLCγ1 in response to EGFR clustering through either immobilized EGF or cholesterol depletion. Our previous work has demonstrated that migration persistence of individual keratinocytes within a collective cell sheet is the strongest predictor of in vitro wound closure 11 and that PLCγ1 activation is responsible for this persistent movement. 9 Thus, PLCγ1 may serve as a target for improving wound re-epithelialization, and understanding the factors that lead to its activation is a crucial step toward developing therapies or wound dressings capable of modulating this signal. Additionally, PLCγ1 is recognized to impact motility in a wide range of cell types, 32-34 therefore, elucidation of the factors that cause edge-specific activation of PLCγ1 may translate to other cell types and tissues.

| MATERIALS AND METHODS
All materials were purchased from Thermo Fisher Scientific (Waltham, MA) unless noted otherwise.

| Experimental conditions
HaCaTs were seeded at a sparse density in 24-well plates prepared as

| Immunofluorescent staining
HaCaT cells were fixed upon fence removal or 4 hr later, following treatment with EGF, ochratoxin A, and/or methyl-β-cyclodextrin.  Supporting Information Figure 2, EGFR clusters were measured by subtracting the fluorescence of nonclustered EGFR from the original image using the confocal image and the "Image Calculator" command, receptor clusters were identified using a (10/255) threshold, and the "Analyze Particles" command was used to measure the cluster sizes.

| Statistical analysis
Data were analyzed with GraphPad Prism 7.0 (La Jolla, CA) and are presented as the mean ± SD, and all experiments were performed at least twice to ensure reproducibility. Statistical significance was determined using ANOVA with post-tests as noted in the figure legends. In all tests, p < .05 was considered statistically significant; calculated p values and adjusted p values for multiple comparisons can be found in Supporting Information Table 1.