Tetraspanin CD82 shapes EGFR signaling outcomes through nanoscale receptor organization

Abstract

Tetraspanins are integral membrane proteins that play a crucial role in organizing and regulating cellular signaling by serving as scaffolds that compartmentalize receptors and other signaling molecules within membrane microdomains. Here, we report how the tetraspanin CD82 modulates the molecular organization and signaling of the Epidermal Growth Factor Receptor (EGFR), a key molecule involved in cellular proliferation, differentiation, and survival. Combining multi-color super-resolution microscopy with advanced image reconstruction and analysis techniques, we demonstrate that CD82 selectively associates with EGFR, promotes receptor oligomerization, and acts as a regulator of ligand-independent receptor phosphorylation in a palmitoylation-dependent manner. Additionally, CD82 promotes a more compact molecular organization of EGFR, which correlates with altered endocytosis and downstream signaling outcomes. These findings underscore the importance of tetraspanins in the spatial and functional regulation of cell surface receptors, with implications for controlling aberrant signaling in disease and positions CD82 as a potential target for developing therapeutic strategies aimed at modulating EGFR signaling by influencing receptor organization.

Introduction

The organization of signaling molecules at the plasma membrane plays a critical role in determining the efficiency, specificity, and amplitude of cellular responses to extracellular stimuli. Understanding the principles governing the molecular organization of the membrane is key to determining how dynamic alterations in membrane organization enable cells to coordinate complex processes. Tetraspanins represent a diverse family of integral membrane scaffold proteins characterized by their unique structure, featuring four transmembrane domains. Expressed across various cell types and tissues, tetraspanins serve as master organizers of membrane microdomains, influencing cellular physiology by forming tetraspanin-enriched microdomains (TEMs). Tetraspanins associate with other tetraspanins and membrane-associated molecules (Charrin et al., 2009; Levy and Shoham, 2005) to regulate a broad range of cellular processes, including cell adhesion, migration, signaling, and 1 vesicular trafficking (Bassani and Cingolani, 2012; Charrin et al., 2009; Termini and Gillette, 2017). This versatility allows tetraspanins to sculpt the molecular landscape of the plasma membrane and orchestrate diverse signaling pathways, including signaling through the ErbB family of receptors.

Signaling through the Epidermal Growth Factor Receptor (EGFR), the prototypical ErbB receptor, is critical to processes ranging from development and wound healing to carcinogenesis and metastasis. EGFR signaling is inherently dependent on its plasma membrane organization, where binding of ligands, such as epidermal growth factor (EGF), to the extracellular domain of EGFR leads to receptor dimerization and autophosphorylation of tyrosine residues within the intracellular domain of EGFR. These sites act as docking sites for adaptor proteins and signaling molecules that elicit diverse downstream cellular responses. When dysregulated, EGFR signaling can be exploited by malignant cells in the development and progression of numerous cancers (Mendelsohn and Baselga, 2003; Scaltriti and Baselga, 2006; Sigismund et al., 2018). As a major player in a range of malignancies, EGFR has been an attractive therapeutic target, with several antibody and small-molecule agents developed over the last two decades (Ciardiello and Tortora, 2008). However, the success of many of these drugs has been limited. Thus, a more detailed understanding of the underlying mechanisms that govern the spatial and temporal regulation of EGFR signaling is needed to better target and disrupt malignant progression.

Among tetraspanins, CD9 and CD82 have been shown to modulate receptor activity through direct interactions, while indirect interactions have been proposed as the basis for modulation of ErbB receptors by tetraspanins CD81 and CD151 (Berditchevski and Odintsova, 2016). Through these interactions, tetraspanins are proposed to regulate the receptor signaling, lateral mobility, and affinity for ligand, thereby enabling cells to precisely regulate downstream processes (Charrin et al., 2009; Hemler, 2003; Hemler, 2005; Levy and Shoham, 2005; Termini and Gillette, 2017). Here, we investigate how tetraspanin scaffolds modulate the membrane organization of EGFR, with a focus on CD82. Prior work has shown that EGFR signaling can be attenuated by tetraspanin CD82 (Li et al., 2013; Odintsova et al., 2000; Wang et al., 2007). Furthermore, studies have shown that CD82 expression can regulate the ubiquitination and degradation of EGFR (Odintsova et al., 2013). More recently, CD82-containing nanodomains were shown to confine unliganded EGFR, impacting receptor mobility and dynamics at the membrane, and modulating ligand affinity (Sugiyama et al., 2023). However, the mechanisms by which the CD82 scaffold functions to modulate the molecular organization of EGFR on the plasma membrane, as well as the impact of this spatial regulation on downstream EGFR signaling outcomes, remains unclear. Coupling super-resolution microscopy with novel image reconstruction techniques and analyses, we find that the CD82 molecular scaffold associates with EGFR and impacts both its oligomeric clustering and molecular packing, leading to distinct signaling outcomes downstream of receptor activation.

Results and discussion

EGFR associates with tetraspanin CD82, which modulates EGFR molecular organization.

Previous studies identified EGFR regulatory interactions with several tetraspanins, including CD9 (Murayama et al., 2008; Tang et al., 2015; Wang and Han, 2015), CD81 (Diao et al., 2012; Meyer et al., 2015), CD82 (Danglot et al., 2010; Odintsova et al., 2000; Odintsova et al., 2013; Odintsova et al., 2003; Wang et al., 2007), and CD151 (Kgk et al., 2020; Wong et al., 2024; Zhu et al., 2021) (reviewed in (Berditchevski and Odintsova, 2016)). Analysis of the tetraspanin surface expression on HeLa cells (Fig. S1 A) identified minimal surface expression of CD9 and CD81 and significant surface expression of CD82 and CD151. Next, we investigated how tetraspanins contribute to the membrane organization of EGFR by quantifying EGFR clustering and tetraspanin CD82 and CD151 association using sequential direct stochastic optical reconstruction microscopy (seq-dSTORM) (Valley et al., 2015b) (Fig. 1 A). Membrane distribution of EGFR was assessed by the Hopkins Statistic (Hopkins and Skellam, 1954), where values near 0.5 indicate a random distribution of molecules and values > 0.5 suggest an increased tendency for clustering. While the high Hopkins Statistic values in Figure 1B likely reflect overcounting from dSTORM localization data, we find EGFR has a higher clustering tendency when cells are treated with EGF, as previously reported (Mudumbi et al., 2024; Nagy et al., 2010; Sako et al., 2000; Valley et al., 2015a). Additionally, we used the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm (Ester et al., 1996) to quantify cluster number, finding that EGF treatment results in fewer overall clusters, as shown previously (Harwardt et al., 2020). These findings are consistent with the merging of individual clusters (Fig. 1 C). To quantify EGFR association with CD82 and CD151, we analyzed the two-color super-resolution images using the bivariate Ripley’s K function, which allows us to detect the association of EGFR with either CD82 or CD151 at a given spatial scale (Wilson et al., 2007; Yang et al., 2007). In Figure 1D, we find that EGFR and CD82 are highly associated under both unliganded and ligand-activated conditions, with an observed decrease in association following EGF treatment that is consistent with previous reports (Odintsova et al., 2000). In contrast, EGFR and CD151 are only associated at distances greater than 50 nm (Fig. 1 E). In fact, at distances less than 50 nm, EGFR and CD151 are randomly associated when unliganded and negatively associated following EGF treatment. Therefore, based on these findings, we focused our studies on tetraspanin CD82 as a potential driver of EGFR organization and downstream signaling.

Figure 1. EGFR selectively associates with tetraspanin CD82, which modulates EGFR molecular organization.

(A) Representative sequential dSTORM image of EGFR (blue), CD82 (red), and CD151 (green) on the surface of a HeLa cell. (B and C) EGFR clustering tendency (B) and number of EGFR clusters per 10³ localizations (C) determined using the Hopkins Statistic and the DBSCAN clustering algorithm, respectively. Data are shown as SuperPlots (Lord et al., 2020) with n_Control =23 and n_+EGF = 20 cells per condition (small symbols) imaged across n_Control = 6 and n_+EGF = 7 independent replicates; means +/− S.D. and Mann-Whitney test results shown are based on averages from independent replicate (large symbols). (D and E) Bivariate Ripley’s K analysis shows likelihood of EGFR association with CD82 (D) or CD151 (E) under basal conditions (solid blue lines) or following 5-minute treatment with 50 nM EGF at 37° C (dashed blue lines); dashed black lines represent bounds of 99% confidence interval around random association. For EGFR-CD82, n_Con = n_+EGF = 6 cells were imaged across 4 independent experiments. For EGFR-CD151, n_Con = 6 and n_+EGF = 4 cells imaged across n_Con = 3 and n_+EGF = 2 independent replicates. (F) dSTORM image of EGFR on the surface of a HEK293 cell overexpressing EGFR and CD82; inset shows representative 2 × 2 μm ROI used for clustering analysis. (G) EGFR clustering tendency quantified using the Hopkins Statistic for control or WT CD82-expressing HEK293 (left) or A431 (right) cells under basal or EGF-activated conditions. (H) Representative diagrams of EGFR clusters identified by DBSCAN in control (top) or CD82-overexpressing (bottom) HEK293 cells under basal conditions (left) or following 5-minute treatment with 50 nM EGF at 37° C (right). (I) Number of clusters per 10³ EGFR localizations identified by DBSCAN analysis of control or CD82-expressing HEK293 cells co-transfected with EGFR.eGFP (left) and control or CD82-expressing A431 cells (right). At least 120 ROIs were analyzed per condition collected from at least 6 cells imaged across at least two independent experiments for each cell line. Exact number of structures, cells, and independent replicates used for analysis are included in the Materials and Methods section. Data in panels (B), (C), (G), and (I) are presented as SuperPlots with larger symbols representing the per cell average of underlying per-ROI metrics depicted by smaller symbols. Error bars represent mean +/− S.D. based on per-cell values; p-values reflect results from generalized linear model analyses, and statistically significant differences are indicated by *p < 0.05, *** p < 0.001, and **** p < 0.0001.

To mechanistically examine how CD82 may be modulating the membrane organization of EGFR, we generated two different cell line models (HEK293 and A431). A431 cells endogenously express abundant EGFR (Merlino et al., 1984) but minimal surface CD82, whereas HEK293 cells express minimal levels of endogenous EGFR and CD82, therefore, these cell models allow us to measure CD82-specific changes without the complication of significant endogenous surface protein (Fig. S1 B). A431 cells were stably generated to overexpress CD82 (Fig. S1 C), whereas HEK293 cells were stably transfected to overexpress both EGFR and CD82 (Fig. S1, D and E). Using these cell models, we quantified EGFR membrane localization data by dSTORM (Fig. 1 F) and assessed EGFR clustering. Similar to HeLa cells, we find high Hopkins Statistic values (~0.9) for both A431 and HEK293 cells, indicating a significant clustering tendency for EGFR (Fig. 1 G). Additionally, we detect a decrease in the overall number of clusters in HEK293 cells following EGF activation using DBSCAN analysis, with CD82-expressing cells having more clusters in both untreated and EGF-activated conditions (Fig. 1, H and I). In contrast, we find that EGF stimulation of A431 cells results in more EGFR clusters, with CD82-expressing cells also trending towards having more EGFR clusters when treated with EGF. These cell line specific shifts in organization likely stem from the significant differences in stoichiometry of EGFR and CD82, with A431 cells having substantially fewer CD82 molecules relative to their exceedingly high expression of EGFR (Fig. S1, F–H). More importantly, it’s critical to note that quantification of protein distributions and cluster analysis at these high molecular densities is particularly susceptible to the overcounting that occurs with dSTORM data; therefore, the application of novel computational techniques was required to more quantitively and accurately resolve receptor organization at the single molecule level.

Bayesian Grouping of Localizations reveals impairments in the molecular organization of palmitoylation-deficient CD82 and its interactions with EGFR.

Using single-molecule localization based super-resolution microscopy to quantify protein cluster sizes of molecules that are densely packed together in close proximity can result in localizations being estimated with relatively poor precision, introducing considerable uncertainty (Endesfelder et al., 2014; Rieger and Stallinga, 2014). In a recent study, we describe the Bayesian Grouping of Localizations (BaGoL) method of inferring the positions of tagged proteins by exploring the possible grouping and combination of localizations from multiple blinking/binding events (Fazel et al., 2022). This advancement in analysis can achieve sub-nanometer precisions under dense labeling conditions. In contrast to the photobleaching-prone fluorophores used in dSTORM, the consistent kinetics of imaging strand binding events used in DNA points accumulation for imaging in nanoscale topography (DNA-PAINT) (Auer et al., 2017; Jungmann et al., 2010) makes this imaging modality particularly well-suited for applying Bayesian inference to group localizations and determine true molecular positions. Therefore, we used DNA-PAINT with BaGoL analysis to quantitatively evaluate the impact of CD82 molecular scaffolding on EGFR membrane organization.

We first used this approach to assess changes in CD82 organization resulting from disruption of the tetraspanin scaffold. Critical to the nanodomain membrane organization of tetraspanin CD82 are the five membrane proximal cysteines that become palmitoylated and promote tetraspanin packing in the membrane (Charrin et al., 2002; Delandre et al., 2009; Termini and Gillette, 2017; Yang et al., 2002; Yang et al., 2004). Previous work from our laboratory using super-resolution microscopy and DBSCAN analysis identified shifts in CD82 nanodomain organization when the five membrane proximal cysteines were mutated (Palm-CD82), supporting the role of palmitoylation in CD82 clustering (Termini et al., 2016). Therefore, we used DNA-PAINT to image WT-CD82 and Palm-CD82 on the surface of stably transfected cells, ensuring similar CD82 expression levels (Fig. S2, A and B). Application of the BaGoL algorithm to DNA-PAINT data showed a marked improvement in resolution, allowing for identification of individual emitters corresponding to single CD82 molecules found in CD82 nanoclusters (Fig. 2, A and B). Analyses of nearest-neighbor distances (NND) within CD82 clusters of pre-BaGoL data (Fig. 2 C, solid lines) show minimal differences between the molecular packing of WT-CD82 and Palm-CD82. Moreover, these differences are only seen at scales below the resolution of DNA-PAINT, likely reflecting a skewing due to overcounted individual molecules. In contrast, analysis of BaGoL-processed data (Fig. 2 C, dashed lines) reveals impaired molecular packing of Palm-CD82 when compared to WT-CD82, finding differences at a range between 10–40 nanometers. Further analysis of BaGoL-DNA-PAINT data indicates that, while WT-CD82 and Palm-CD82 show similar numbers of clusters, Palm-CD82 is clustered to a lesser extent, forming smaller clusters with fewer molecules per cluster (Fig. 2 D). Together, these data demonstrate the utility of BaGoL for analyzing super-resolution imaging data to quantify organizational changes at the single molecule level.

Figure 2. Bayesian Grouping of Localizations reveals impairments in the molecular packing of palmitoylation-deficient CD82 and its interactions with EGFR.

(A and B) Representative DNA-PAINT images of WT CD82 (A) and Palm-CD82 (B) on the surface of transfected A431s. Insets show magnified DNA-PAINT region (top), BaGoL analysis assignment of localizations to predicted true emitter (middle) and resulting BaGoL image (bottom) scale bar − 1 μm. (C) Cumulative Frequency Distribution of CD82 nearest neighbor localization distances (NND) within clusters for (red) WT- and (yellow) Palm-CD82 from DNA-PAINT images before (solid lines) and after (dashed lines) BaGoL analysis. A Kolmogorov-Smirnov was used to confirm significant differences (p < 0.0001) between WT and Palm-CD82. (D) Clustering statistics for WT (red) and Palm- (yellow) CD82 in transfected A431s after BaGoL analysis; data analyzed from same ROIs as (C). Data are presented as SuperPlots with larger symbols representing the per cell average of underlying per-ROI metrics depicted by smaller symbols. Error bars represent mean +/− S.D. based on per-cell values; p-values reflect results from generalized linear model analyses, and statistically significant differences are indicated by *p < 0.05, **p < 0.01. (E) Two-color overlay of BaGoL-analyzed EGFR (magenta) and CD82 (green) on the surface of A431s expressing WT (left) or Palm- (right) CD82 under basal conditions (top) or following 5-minute treatment with 50 nM EGF at 37° C (bottom). White arrows indicate instances in which EGFR is found in close association with CD82, while arrowheads show instances where EGFR is found separate from CD82. (F and G) Bivariate Ripley’s K (F) and pair cross-correlation (G) analyses showing association of EGFR and CD82 under basal (left) and EGF-activated conditions (right) in cells expressing WT- (red) and Palm- (yellow) CD82. Dashed lines represent 99% confidence interval. Data shown were collected from at least 12 ROIs collected from ≥ 6 cells per condition imaged across two to three independent experiments. Exact number of structures, cells, and independent replicates used for analysis are included in the Materials and Methods section.

Next, we used 2-color DNA-PAINT with BaGoL to evaluate the impact of CD82 palmitoylation on CD82-EGFR interactions (Fig. 2 E). Consistent with our seq-dSTORM data (Fig. 1, D and E), we find a strong association between unliganded EGFR and WT-CD82 which is diminished upon EGF treatment, as quantified by bivariate Ripley’s K (Fig. 2 F) and pair cross-correlation analyses (Veatch et al., 2012) (Fig. 2 G). Interestingly, we find that while association with unliganded EGFR is diminished in Palm-CD82 cells relative to WT-CD82, EGF treatment slightly enhances association between Palm-CD82 and EGFR, resulting in nearly identical EGFR-CD82 associations in WT-CD82 and Palm-CD82 cells following receptor activation. Taken together, these single molecule data identify a strong association between unliganded EGFR and CD82 that is dependent upon the palmitoylation state of CD82.

CD82 expression promotes oligomerization and tight molecular packing of EGFR.

We next sought to use the improved resolution and reduced overcounting provided by BaGoL to assess the impact of CD82 scaffolding on the molecular clustering of EGFR (Fig. 3 A). Analysis of EGFR NND revealed that the majority of EGFR molecules are spaced at sub-20 nm distances (Fig. S2 C), suggesting a quantitative estimate of receptor interaction distances that is consistent with the literature (Strauss and Jungmann, 2020). DBSCAN clustering analysis of the data identified an increase in the percentage of EGFR clustered along with a corresponding decrease in the overall number of clusters in cells expressing WT CD82 when compared to both control and Palm-CD82 upon EGF treatment (Fig. 3, B and C). These data suggest that receptor activation results in the formation of larger scale complexes. Accordingly, we find an increase in EGFR cluster sizes in WT CD82 cells upon EGF treatment, as well as larger clusters when compared to control and Palm-CD82 cells (Fig. 3 D). Interestingly, we observe no significant changes in the extent of EGFR clustering or the size of EGFR clusters in control cells following EGF treatment (Fig. 3, B–D). These findings potentially reflect a dynamic re-organization of EGFR into distinct molecular complexes, which could result in seemingly unchanged clustering.

Figure 3. CD82 expression promotes oligomerization and tight molecular packing of EGFR.

(A) Representative BaGoL-analyzed DNA-PAINT images of EGFR on the surface of control (left), WT CD82 (middle), or Palm-CD82 (right) A431 cells under basal conditions (top) or following 5-minute treatment with 50 nM EGF at 37° C (bottom). Inset shows magnified view depicted as a dot plot diagram with overlays indicating clusters containing two points (green) or more than two points (magenta) identified by DBSCAN. (B–D) Percent of EGFR clustered (B), number of clusters per 10³ EGFR localizations (C), and average cluster diameters (D) quantified using DBSCAN analysis of experimental groups represented in (A) and (B). (E) Distribution of EGFR clusters by number of points per cluster; data points represent the fraction of clusters per ROI found to contain two, three, four, or more than four molecules. (F and G) Percentage of total EGFR per ROI found in clusters containing two points (F) or greater than two points (G). (H) Relative frequency distributions of nearest neighbor distances between EGFR molecules within clusters; a Kruskal-Wallis test with Dunn’s multiple comparisons of these data shows statistically significant differences (p < 0.0001) for all relevant pair-wise comparisons. Scatter plots in panels B–G depict data from ≥ 24 ROIs chosen from ≥ 8 cells per condition gathered across at least two independent experiments. Relative frequency distributions depict pooled data for all localizations, with ≥ 68,227 localizations per condition. Exact number of structures, cells, and independent replicates used for analyses are included in the Materials and Methods section. Lines and error bars represent mean +/− S.D. Statistically significance differences are denoted as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 and reflect results from generalized linear model analyses through negative binomial regression.

While canonical models describe EGFR signaling as proceeding primarily through receptor dimerization following ligand binding (Lemmon and Schlessinger, 2010; Low-Nam et al., 2011; Ogiso et al., 2002; Sako et al., 2000; Valley et al., 2015a), a growing body of evidence suggests that pre-formed and inactive dimers, as well as monomers and larger order oligomers, play an important role in regulating receptor activation and signal transduction (Clayton et al., 2007; Gan et al., 2007; Kozer et al., 2013; Kozer and Clayton, 2020; Mudumbi et al., 2024; Nagy et al., 2010; Needham et al., 2016; Yu et al., 2002; Zanetti-Domingues et al., 2018). Consistent with previous reports (Liu et al., 2007; Saffarian et al., 2007; Yang et al., 2009), we find that EGFR primarily forms dimeric clusters (> 50%) across all cell lines, whereas fewer than 20% of clusters consist of more than four molecules (Fig. 3 E). Compared to control and Palm-CD82 cells, dimeric EGFR clusters account for a smaller fraction of EGFR clusters in WT CD82 cells following treatment (~55% compared to ~70%), with a significant increase in the fraction of clusters containing more than four receptors per cluster. Further analysis of EGFR clusters revealed shifts in the percentage of EGFR found in dimeric or oligomeric clusters, depicted in representative ROIs by the green and magenta outlines in Figure 3A, respectively. In contrast to the apparent similarity observed in earlier metrics (Fig. 3, B–D), control cells show an increase in the percentage of EGFR found in dimeric clusters, as well as a concomitant decrease in the percentage of EGFR found in oligomers, upon EGF treatment (Fig. 3, F and G). Expression of CD82, however, results in a distinct shift in clustering in response to EGF, promoting clusters with multiple receptors instead of dimeric clusters, which is consistent with the observed larger cluster sizes in WT CD82 cells (Fig. 3 D). Different models have been proposed for how EGFR dimers can multimerize into oligomers (Huang et al., 2016; Needham et al., 2016). Additionally, scaffold proteins within or adjacent to the membrane have also been postulated to play a role (Mudumbi et al., 2024) and here, we provide evidence for the CD82 scaffold protein as a driver of larger scale EGFR oligomers upon ligand stimulation. The reduced EGFR oligomerization observed with the Palm-CD82 mutant further suggests that CD82-mediated EGFR oligomerization requires the lateral assembly of the CD82 scaffold. Our previous two-color quantum dot tracking indicates that unliganded EGFR dimers are more transient when compared to the longer-lived dimers formed between liganded receptors (Low-Nam et al., 2011; Valley et al., 2015a); together with the present evidence that CD82 enhances the oligomer fraction of EGFR, these findings indicate that future studies should explore how CD82 impacts the lifetime of different EGFR complexes.

Lastly, since close proximity between EGFR monomers is necessary for receptor auto- and cross-phosphorylation downstream of ligand binding, we also investigated how CD82 impacts the EGFR NND within clusters (Fig. 3 F). Here, we find that CD82 expression promotes tighter association of EGFR molecules within a cluster under basal conditions, regardless of palmitoylation status. Intriguingly, while activation with EGF results in tighter receptor packing in both control and WT CD82 cells, Palm-CD82 exhibit a shift away from high density clusters upon EGF treatment. Collectively, these data demonstrate that the CD82 scaffold organizes EGFR by facilitating the tight packing of EGFR molecules within clusters while also promoting larger cluster sizes and greater overall EGFR clustering. Since CD82 enhances the close proximity of EGFR monomers in response to ligand, we next assessed whether this tight packing impacts the activation of the EGFR intracellular kinase domain.

Expression of CD82 results in basally suppressed phosphorylation of EGFR in a palmitoylation-dependent manner.

To evaluate how the CD82 scaffold-mediated changes in EGFR molecular organization impact signal transduction, we carried out biochemical studies evaluating the phosphorylation of EGFR at site-specific tyrosine residues (Fig. 4, A–D and Fig. S3). Under basal conditions, we find that EGFR signaling is attenuated when CD82 is expressed, whereas Palm-CD82 cells demonstrate higher phosphorylation across all tested tyrosine residues in the absence of ligand (Fig. 4, E–H). These data support the notion that the scaffold function of CD82 helps to restrict ligand-independent EGFR phosphorylation and suggest that CD82-associated pre-formed EGFR complexes are predominantly inactive. In contrast, Palm-CD82 expression results in significant activation of EGFR in the absence of ligand, underscoring the importance of the CD82 scaffold function in fine tuning critical cellular responses. Next, we analyzed EGFR activation following treatment with 50 nM EGF, a saturating dose for receptors capable of ligand binding. These data reveal that expression of CD82 results in enhanced ligand-dependent receptor phosphorylation, particularly at Tyr992 and Tyr1068, while Palm-CD82 cells display a global impairment in fold-activation of EGFR following treatment with EGF (Fig. 4, I–L). Collectively, these findings suggest that by dampening basal EGFR signaling, the CD82 scaffold can promote a more robust activation response upon ligand stimulation. A combination of structural studies suggests that EGFR kinase domain interactions with the intracellular membrane surface are autoinhibitory (Endres et al., 2013; Kaplan et al., 2016; Sengupta et al., 2009). As such, EGFR association with CD82 nanodomains may help to promote these juxtamembrane autoinhibitory interactions, which serve to maintain EGFR inactivation even in the presence of dimerization. Moreover, we speculate that when CD82 palmitoylation sites are mutated and CD82 nanodomains are disrupted, the autoinhibitory interaction is released, resulting in enhanced basal signaling. Therefore, we next set out to determine how the altered EGFR signaling observed impacts broader cell physiology, evaluating proliferation under basal conditions in the absence and presence of the EGFR inhibitor, afatinib. In Figure 4M, we find that Palm-CD82 cells are more proliferative when compared to control and CD82-expressing cells, consistent with their enhanced basal EGFR activation at Tyr845. This difference in proliferation is abolished following inhibition of EGFR kinase activity with afatinib, indicating that the enhanced proliferation of Palm-CD82 cells is at least in part driven by their elevated basal EGFR phosphorylation. Thus, when taken together, our data suggest that the scaffolding capacity of CD82 serves to diminish basal signaling from pre-formed EGFR complexes, which in turn enables a more robust activation response in the presence of ligand. Moreover, disruption of the scaffolding capacity of CD82 results in sustained EGFR activation that promotes enhanced cell proliferation independent of ligand.

Figure 4. Palmitoylated-CD82 suppresses basal phosphorylation of EGFR and limits proliferation.

(A–D) Representative 2-color fluorescent western blot of total EGFR (green) and EGFR pY845 (red) (A), EGFR pY992 (green) and GAPDH (red) (B), total EGFR (green) and EGFR pY1045 (red) (C), and total EGFR (green) and EGFR pY1068 (red) (D) in A431 control, WT CD82, or Palm-CD82 cells following treatment with 50 nM EGF for indicated time. (E–H) Relative phosphorylation of EGFR at Y845 (E), Y992 (F), Y1045 (G), and Y1068 (H) normalized to control A431 cells. (I–L) Foldchange phosphorylation of EGFR at Y845 (I), Y992 (J), Y1045 (K), and Y1068 (L) in control, WT CD82, and Palm-CD82 A431 cells following treatment with 50nM EGF for indicated times normalized to replicate-matched untreated values. Representative blots including additional timepoints are included in supplemental Figure S3. Statistics represent one- (E–H) or two-way (I–L) ANOVAs with Tukey’s post-hoc test for multiple comparisons calculated from at least three independent experiments (n ≥ 3). (M) Cell proliferation measured by CyQuant Direct Assay shown for Control, WT CD82, and Palm-CD82 cells grown in complete medium or medium with 500 nM afatinib for indicated times, normalized to signal at Day 0. Data shown represent three independent experiments (n = 3) with statistics for untreated cells and afatinib-treated cells each calculated using two-way ANOVAs with Tukey’s post-hoc test for multiple comparisons. Lines and error bars represent mean +/− S.D. Statistical significance denoted as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

CD82 expression enhances EGFR:Caveolin association and promotes EGF-induced receptor internalization.

Following activation, the dimerization and oligomerization of EGFR can also contribute to signal transduction through receptor internalization and degradation (Burke et al., 2001; Sigismund et al., 2008; Sorkin, 2001). Tetraspanins are well described regulators of the endocytic pathway (Odintsova et al., 2013; Toribio and Yanez-Mo, 2022; Xu et al., 2009), thus we evaluated the impact of CD82 expression on the internalization of EGFR following ligand stimulation. EGFR internalization primarily occurs through clathrin-mediated endocytosis (Rappoport and Simon, 2009; Vieira et al., 1996), however, non-clathrin endocytosis mechanisms including caveolae are also key components of EGFR signal regulation (Renard and Boucrot, 2021; Sigismund et al., 2005). When quantifying EGFR co-localization with clathrin and caveolin under unliganded and EGF-activated conditions (Fig. 5 A), we find increased co-localization of EGFR with clathrin following EGF treatment in control and CD82-expressing cells (Fig. 5 B). With respect to caveolin, increased colocalization between EGFR and caveolin is detected in control cells following EGF treatment (Fig. 5 C). However, more striking is the significant basal association between EGFR and caveolin in the presence of CD82. These data align with the dampened basal EGFR signaling observed in CD82 expressing cells (Fig. 4, E–H) and previous reports suggesting that association with caveolin suppresses EGFR signaling (Lajoie et al., 2007; Liu et al., 2007; Mineo et al., 1999; Park et al., 2000). Furthermore, following treatment with EGF, we find that CD82 expression enhances internalization of EGFR (Fig. 5 D), whereas Palm-CD82 expression attenuates EGFR internalization relative to both CD82-expressing and control cells. Together, these findings suggest that the scaffolding capacity of CD82 can also modulate the shift of EGFR association between caveolin and clathrin, which can impact signaling and internalization.

Figure 5. CD82 expression enhances EGFR:Caveolin association and promotes EGF-induced receptor internalization.

(A) Representative confocal images of caveolin (yellow), clathrin (green), EGFR (magenta), and DAPI (cyan) in Control, WT CD82, and Palm-CD82 A431 cells under basal (left) and EGF-activated (right) conditions; scale bar − 10 μm. Insets show magnified, single-channel and merged images of the ROIs denoted by the dashed boxes, as well as fluorescence intensity profiles for each label along the dashed line; scale bar − 2 μm. (B and C) Quantification of Mander’s M2 coefficient for colocalization of EGFR with clathrin (B) or caveolin (C) in indicated A431 cells under basal conditions or following 5-minute treatment with 50 nM EGF at 37° C. Data shown as a SuperPlot (Lord et al., 2020) of three independent experiments with five ROIs per image selected for analysis from at least eight different fields of view per replicate; statistical analysis performed using mean of each replicate (n = 3). (D) Percent internalization of AlexaFluor647-conjugated EGF in Control, WT CD82, and Palm-CD82 A431 cells determined by flow cytometry after stripping with acetic acid buffer. Data shown represent three independent replicates (n = 3). Lines and error bars represent mean +/− S.D. Statistics calculated using a two-way ANOVA with Tukey’s post-hoc test for multiple comparisons with significance denoted as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Collectively, our biochemical, biophysical, and computational approaches allow us to demonstrate that CD82 modulates EGFR organization in multiple ways and that the specific association between EGFR and CD82 is primarily inhibitory in nature. Increased association of CD82 with EGFR constrains ligand-independent activation of the receptor, dampening basal signaling from preformed EGFR assemblies. Moreover, CD82 promotes EGFR oligomerization and enhances the molecular packing of EGFR, which collectively drives a strong signaling response upon ligand stimulation. Additionally, we find that the palmitoylation-mediated scaffold function of CD82 modulates endocytic trafficking, receptor internalization, and cell proliferation. In the context of EGFR, these findings represent significant steps towards understanding the mechanisms involved in driving EGFR oligomerization and regulating ligand-independent activation of EGFR, which is a proposed mechanism for the development of resistance to EGFR-targeting therapies (Iyer et al., 2024; Shen and Kramer, 2004; Uribe et al., 2021; Valley et al., 2015a). More broadly, our findings provide insights into the mechanisms by which tetraspanins function to segregate membrane receptors to modulate signaling potential in response to stimuli and regulate downstream cellular processes. Importantly, our results suggest that tetraspanins may represent a novel avenue for developing therapeutic modalities that target EGFR molecular organization, which may help to overcome treatment resistance in disease.

Materials and methods

Cell culture

HeLa, A431, and HEK293 cell lines (American Type Culture Collection) were cultured in appropriate medium (DMEM for HeLa and A431 and MEM for HEK293s) supplemented with 10% Fetal Bovine Serum (FBS), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. Cells were incubated at 37° C and 5% CO₂. Cell lines stably expressing mCherry.Control, mCherry.WT-CD82, and mCherry.Palm-CD82 were generated using plasmids previously described (Termini et al., 2014), which were constructed using an mCherry-C1 Vector (Invitrogen). Stable transfection of EGFR.eGFP in HEK293 was carried out using previously generated plasmid (Lidke et al., 2004). Transfections were carried out using the Amaxa Cell Line Nucleofector Kit T for A431 or Kit V for HEK293, according to the manufacturer’s instructions. G418 was used to select for stable mCherry.Control, mCherry.WT-CD82, and mCherry.Palm-CD82 transfectants in both cell lines, while Zeocin was additionally used to select for EGFR.eGFP double transfectants in HEK293. To ensure similar levels of transfected protein within each cell line (e.g., A431 or HEK293) monoclonal colonies were isolated, and fluorescence-activated cell sorting (FACS) was used to ensure similar expression of CD82 between WT-CD82 and Palm-CD82 A431 cells as well as similar expression of EGFR and CD82 between control, WT-CD82, and Palm-CD82 HEK293 cells (apart from minimal CD82 expression in controls) (Figs. S1 and S2). For experiments involving EGFR activation with EGF, cells were treated with 50 nM EGF (Peprotech, #AF-100-15) for 5 minutes at 37° C to ensure a saturating dose for receptors capable of ligand binding.

Super-resolution imaging

seq-dSTORM

Sequential imaging of multiple targets on the same cells was performed as previously described (Valley et al., 2015b). Briefly, cells were plated on eight-well chamber slides (Nunc Lab-Tek II, ThermoFisher #155409) and allowed to adhere overnight. Following treatment with serum-free media or serum-free media with EGF, cells were rinsed twice with phosphate-buffer saline (PBS) and formaldehyde fixed via addition of 4% paraformaldehyde (PFA) + 0.2% glutaraldehyde in PBS for 2 hours at room temperature (“RT”). Cells were then washed twice with PBS and incubated in 0.1% NaBH₄ in PBS for 10 minutes at RT to remove autofluorescence from glutaraldehyde incubation. Cells were again washed twice with PBS prior to blocking with 3% Bovine Serum Albumin (BSA) in PBS for 30 minutes at RT. Immunolabeling with Alexa Fluor 647 anti-human CD82 antibody (BioLegend, ASL-24, 1:125 diluted in 3% BSA/PBS) was performed overnight at 4° C. Cells were washed, post-fixed with 4% PFA, and washed two more times before imaging in a standard dSTORM imaging buffer consisting of 50 mM Tris, pH 8.5 with 10 mM NaCl, 10% glucose (w/v), 168.8 U/mL glucose oxidase (Sigma # G2133), 1404 U/mL catalase (Sigma #C9322), and 20 mM 2-aminoethanethiol (MEA). Labeled cells were imaged using Total Internal Reflection Fluorescence (TIRF) illumination on a custom microscope system with an electron-multiplying charge-coupled device (EMCCD) camera (iXon 897; Andor Technologies), as previously described (Floren et al., 2020). Prior to dSTORM imaging, a brightfield reference image was taken for alignment of subsequent rounds of imaging. 40,000 frames were collected per cell, with an additional brightfield image taken every 2,000 frames for registration and drift correction. Each frame was 256 × 256 pixels with a pixel size of 0.107 μm and an acquisition time of 10 ms per frame. After imaging CD82, cells were photobleached using a 637 nm laser (HL63133DG, Thorlabs) at ~1.7 kW/cm² and 405 nm laser (Crystal laser) at ~0.25 kW/cm² for 5 minutes. Samples were then rinsed three times with PBS and quenched with 0.1% NaBH₄ in PBS for 10 minutes to remove any residual fluorescence. For each subsequent target, the process of labeling, imaging, and photobleaching was repeated, starting with a 1 hour block with 3% BSA/PBS followed by 1 hour incubation with antibodies targeting each target; anti-EGFR-AF647 (SCBT, R-1) and anti-CD151-AF647 (R&D Systems, #FAB1884R) were used at 1:100 and 1:50, respectively, diluted in 3% BSA/PBS.

DNA-PAINT

Multi-color DNA-PAINT imaging of CD82 and EGFR was performed sequentially on a custom-built sequential super-resolution microscope (SeqSRM) as previously described (Schodt et al., 2023a). Cells were plated on piranha-etched 25 mm #1.5 cover glass (Electron Microscopy Sciences #72290-12) and allowed to adhere overnight. Following treatment with serum-free media or serum-free media with EGF, cells were rinsed twice with PBS, fixed in 100% methanol for 20 minutes at −20° C, and washed twice with PBS before blocking for with 3% BSA/PBS for 30 minutes at RT. Cells were labeled with primary antibodies diluted in 3% BSA/PBS + 0.05% Tween20 overnight at 4° C; anti-EGFR (Clone D38B1, CST #4264) was used at 1:400 and anti-CD82 (Clone BL-2, abcam #47153) was used at 1:1,000. The next day, cells were washed 3 × 5 minutes with 3% BSA/PBS + 0.05% Tween. Cells were then incubated for 1 hour at RT with DNA Docking site-conjugated secondary antibodies (anti-Mouse IgG + Docking Site 1, and anti-Rabbit IgG + Docking Site 2) diluted 1:100 in antibody binding buffer (MASSIVE-AB 2-PLEX, Massive Photonics). Cells were then washed 3 × 5 minutes with 1X washing buffer (Massive Photonics) and post-fixed for 20 minutes at RT with 4% PFA, followed by two more washes with 1X washing buffer. Fluorescent excitation was achieved using a 647 nm laser (2RU-VFL-P-500-647-B1R, MPB Communications) at roughly 1 kW/cm², and emission data were collected using an sCMOS camera (C11440-22CU, Hamamatsu). CD82 was imaged first using an ATTO655-conjugated Imager 1 strand diluted to 500 pM in imaging buffer (Massive Photonics). Following completion of CD82 imaging, Imager 1-containing buffer was removed, and cells were washed 3x with washing buffer. EGFR was then imaged on the same cells using an ATTO655-conjugated Imager 2 strand diluted to 100 pM in imaging buffer. For each image, 200 sequences of 1,000 frames with an exposure time of 20 ms were acquired, with alignment/registration taking place between each sequence.

SR image reconstruction and BaGoL implementation

Image reconstruction and analysis of resulting data were performed using custom-written software in MATLAB (MathWorks Inc.) as previously described (Schodt et al., 2023b). Briefly, collected data were analyzed using a 2D localization algorithm based on maximum likelihood estimation (Huang et al., 2011; Smith et al., 2010) which converts pixel values into photo counts. The photon counts are subjected to thresholding and filtering to localize only individual emitter positions across frames, and the accepted emitters are then used to reconstruct the SR image. BaGoL analysis was performed as previously described (Fazel et al., 2022). Localization data from reconstructed DNA-PAINT images were used as inputs for the BaGoL algorithm. Three 5 × 5 μm ROIs were selected from each reconstructed image and processed through the BaGoL algorithm using the following parameters: burn-in chain length = 8000; post-burn-in chain length = 2000; ROI size = 100 nm; ROI overlap = 25 nm; pre-clustering cutoff = 20 nm; and minimum nearest neighbors = 5.

SR image analysis

Spatial analysis of data from one- and two-color experiments was performed using the Single Molecule Imaging Toolbox Extraordinaire (SMITE) software package (Schodt et al., 2023b). One 10 × 10 μm ROI per cell was selected for analysis of EGFR clustering in HeLa cells, while twenty 2 × 2 μm ROIs per image were selected for analysis of HEK293 and A431 data. Three 5 × 5 μm ROIs per image were selected for analysis of BaGoL data, and cells in which any ROI contained more than 15% of localizations with precision greater than 10 nm were systematically excluded from analysis. Clustering tendency was quantified using the Hopkins statistic (Hopkins and Skellam, 1954), which tests for complete spatial randomness of a probe pattern by comparing nearest neighbor distances from random points and randomly chosen probes. The Hopkins statistic was calculated for each ROI using the default parameters in the SMITE analysis suite: Number of test points = 10; Number of tests to average over: 1,000. Cluster characterization for each protein target was assessed by application of the DBSCAN algorithm to the reconstructed SR data using the following parameters: for reconstructed SR data ε = 20 nm and n = 4, and ε = 40 nm and n = 30 were used for EGFR and CD82, respectively; for BaGoL-processed data, ε = 20 nm and n = 2, and ε = 40 nm and n = 3 were used for EGFR and CD82, respectively. To quantify association between proteins imaged in multi-color experiments, pair-wise data sets were analyzed using the Bivariate Ripley’s K and pair-cross correlation algorithms in SMITE.

Flow cytometry

Surface expression measurements

Surface expression of EGFR and CD82 were measured using anti-EGFR-AF647 (SCBT, R-1) and anti-CD82-AF647 (BioLegend, ASL-24). Fluorescence was detected on the Accuri C6 plus flow cytometer (BD Biosciences). These measurements were run in parallel with the Quantum MESF 647 to quantify the number of Molecules of Soluble Fluorochromes (MESF) on the surface of antibody-labeled cells (Bangs Labs, #647A). We used the MESF values along with manufacturer-determined fluorophore-to-protein ratios of the antibodies to determine the number of EGFR and CD82 molecules found on the surface of our cells.

Internalization assay

Internalization of EGFR was measured via flow cytometry using EGF-AF647 (ThermoFisher #E35351) and acetic acid as previously described (Cleyrat et al., 2013). Briefly, cells were trypsinized, washed with ice-cold PBS, and incubated with 10 nM EGF-647 diluted in serum-free DMEM for 1 hour on ice. Cells were then spun down and washed with ice cold PBS to remove unbound EGF. Cells were resuspended with pre-warmed media to allow internalization to proceed. At each timepoint, a portion of cells were moved into ice cold PBS and split into stripped and non-stripped groups. Non-stripped samples were then fixed in 2% PFA for 20 minutes while stripped samples were washed in acid strip buffer (0.5 M NaCl with 0.2 M acetic acid, pH ~ 2.7) for 5 minutes on ice prior to fixation. Cells were then read on Accuri C6 to quantify fluorescence intensity. The fluorescence intensities of the acid-stripped samples and matched non-stripped samples were then used to quantify percent internalized.

Proliferation assay

Changes in cell proliferation were assayed using the CyQUANT Direct Cell Proliferation Assay (ThermoFisher #C35011) per manufacturer’s guidelines. Briefly, 10³ cells per well were plated in 96-well black/clear bottom plates (ThermoFisher #165305) and allowed to adhere overnight in 100 μL complete medium. Per biological replicate, enough wells were seeded to assay each cell line in triplicate for days 0–5. The following day (“Day 0”), media was replaced in all wells, and 100 μL of 2X CyQUANT Direct detection reagent was added to the “Day 0” wells. Cells were then incubated with detection reagent for 1 hour at 37° C, and fluorescence was detected using the BioTek Synergy H1 microplate reader (Agilent) using excitation and emission wavelengths of 480 and 535 nm, respectively. This process was repeated for each subsequent day for at least three biological replicates (n > 3). For afatinib treatment groups, cells were treated with 500 nM afatinib (Selleck Chemicals #S1011) on days 0 and 3.

Western blotting

Western blots were performed as previously described (Termini et al., 2014). The following antibodies were used for western blots at a 1:1,000 dilution: Anti-EGFR (Clone D38B1, CST #4267), Anti-EGFR pY845 (Clone 12A3, SCBT # sc-57542), Anti-EGFR pY992 (ThermoFisher #44-786G), Anti-EGFR pY1045 (Clone 11C2, abcam#24928), and Anti-EGFR pY1068 (Clone 1H12, CST #2236). Anti-GAPDH (ProteinTech #60004-1-IG) and anti-alpha Tubulin (ProteinTech #66031-1-Ig) were used at 1:2,000. Goat anti-Rabbit IgG Alexa Fluor 790 and goat anti-Mouse IgG Alexa Fluor 680 (ThermoFisher #A27041 and A28183) secondary antibodies were used at 1:10,000. Both primary and secondary antibody solutions were diluted in 5% non-fat dry milk in TBS with 0.1% Tween. Fluorescence was detected using the ChemiDoc MP (Bio-Rad), and densitometry was performed using the Image Studio Lite software (LICOR). Quantified intensity measurements of two-color fluorescent blots were used to determine the ratios of phospho-EGFR to total EGFR for all phospho-sites except pY992; as the host species for the pY992 antibody (rabbit) limited two-color detection alongside the total EGFR antibody (also made in rabbit), ratios of pY992 to total EGFR were calculated after first normalizing signal from each antibody to that of GAPDH or alpha-Tubulin from their corresponding membranes. Phospho-EGFR to total EGFR ratios were then used to determine fold change EGFR activation.

Confocal microscopy

Cells were plated on 25 mm #1.5 cover glass (Electron Microscopy Sciences #72290-12) and allowed to adhere overnight. The next day, cells were serum-starved for 1 hour and treated with either serum-free media or serum-free media with 50 nM EGF for 5 minutes at 37° C. Cells were then washed 2x with PBS and formaldehyde fixed with the addition of 4% PFA for 20 minutes at RT. After two PBS washes, cells were blocked/permeabilized in 3% BSA/PBS + 0.2% Tween for 30 minutes and incubated overnight in primary antibodies diluted in 3% BSA/PBS + 0.1% Tween. Primary antibodies used were as follows: Coralite488-conjugated anti-Clathrin light chain (ProteinTech #CL488-66487) at 1:100; AlexaFluor546-conjugated anti-Caveolin (SCBT #sc-53564-AF546) at 1:100; and anti-EGFR (CST #4267) at 1:400. The next day, cells were washed 3 × 5 minutes with 3% BSA/PBS, followed by a 1-hour incubation with Goat anti-Rabbit IgG AlexaFluor 647 (ThermoFisher). Cells were then washed 3 × 5 minutes with 3% BSA/PBS followed by a final PBS wash before being mounted using ProLong Glass mounting medium (ThermoFisher). Medium was allowed to cure for at least 24 hours prior to imaging. Mounted cells were imaged using a Zeiss LSM 800 system (ZEISS) with excitation wavelengths of 405, 488, 561 or 640 nm and a 63X/1.4 numerical aperture oil immersion objective. Co-localization between EGFR and caveolin or clathrin was determined using ImageJ “Fiji” software (Schindelin et al., 2012). Images were pre-processed with Fiji’s built-in background subtraction function, using the sliding paraboloid option with a radius of 0.2 pixels, followed by a Gaussian blur with a sigma of 0.5. Five 10 × 10 μm ROIs were then selected from at least ten fields of view per experiment. Mander’s correlation coefficients were calculated between EGFR and Caveolin or EGFR and Clathrin for each ROI using the BioP JACoP plug-in(Bolte and Cordelieres, 2006).

Statistics

Statistical analysis of nested SR clustering data

Statistical analyses of the dSTORM and DNA-PAINT data in Figures 1–3 were accomplished using generalized linear mixed effects models which estimated and compared differences between treatments/cell lines while accounting for the nested nature of the SR imaging datasets (with clustering metrics potentially being gathered from several ROIs within a cell, and potentially multiple cells being imaged per day) by the inclusion of random effects. These models accommodated for the distributional properties of the various metrics of interest either directly or through transformation. The bulk of these analyses were conducted using a Poisson distribution to model observed counts (e.g. number of clusters) while accounting for the total number of molecules via offset terms. For analysis of count variables displaying overdispersion (e.g. the CD82 clustering data in Figure 2 D), the negative binomial distribution was used. For comparisons of quantitatively scaled metrics, normal distributions were used, potentially after transformation to better accommodate distributional assumptions. For instance, analysis of Hopkins Statistic values was performed after logit transformation. Datasets used in statistical models are available via public repository. Additional details regarding the exact number of structures, cells, and independent experiments used for these analyses are included below.

	Treatment group		ROI Size (in μm)	# ROIs	Total # Cells	# of independent experiments	# of Clusters	# of Localizations
Fig. 1, B & C	Control		10 × 10	1 (per cell)	23	7	12,325	203,287
	+ EGF				20	6	10,862	217,127
Fig. 1, G & I	HEK293 Con	Con	2 × 2	20 (per cell)	11	2	37,088	555,774
		+ EGF			15	2	47,723	789,467
	HEK293 WT CD82	Con			12	2	30,098	432,594
		+ EGF			11	2	34,821	495,728
	A431 Con	Con			6	2	11,697	172,606
		+ EGF			7	2	13,322	179,495
	A431 WT CD82	Con			6	3	12,587	183,043
		+ EGF			11	4	24,142	316,821
Fig. 2, C & D	WT CD82		5 × 5	16	9	4	2,354	25,285
	Palm-CD82			14	6	2	2,266	22,292
Fig. 3	Control	Con	5 × 5	3 (per cell)	8	2	20,185	89,398
		+ EGF			10	2	17,726	72,404
	WT CD82	Con			20	5	67,307	284,767
		+ EGF			12	4	61,637	286,191
	Palm-CD82	Con			27	5	60,930	273,336
		+EGF			13	2	30,231	128,606

General statistics

Analyses for all other figures were performed using Prism 9 or 10 software (GraphPad). All quantified data were collected from at least three independent experiments. Results are shown as mean ± standard deviation. Where indicated, data are plotted as SuperPlots to show the per-replicate averages (larger symbols) used for statistical analyses while also presenting the variability of the underlying data (smaller symbols). For comparisons in Figures 1 B and 1 C, a two-tailed Man-Whitney test was used. The Kolmogorov-Smirnov test was used for comparison of cumulative distributions in Figure 2 C, while a Kruskal-Wallis test with Dunn’s multiple comparisons was used for analysis of Figure 3 H. To test for statistical significance between multiple comparisons in Figures 3 and 5, one- or two-way ANOVAs performed followed by Tukey’s post-hoc test for multiple comparisons. Significant differences are indicated using asterisks as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Supplementary Material

This manuscript contains 3 supplementary figures. Fig. S1 shows flow cytometry data for expression of tetraspanins CD9, CD81, CD82, and CD151 in HeLa cells, as well as expression data for EGFR and CD82 in wild-type and transfected A431 and HEK293s cells used in this study. Fig. S2 includes additional data showing similar expression levels of CD82 and EGFR in A431 cells transfected with mCherry-Palm-CD82, as well as details on nearest neighbor EGFR distances BaGoL DNA-PAINT data. Fig S3 shows Western blots including additional timepoints used in Figure 4.

Data availability statement

The data underlying all figures in this manuscript are available in the article and its online supplemental materials. Additionally, any datasets generated during this study, including the SAS scripts used for statistical analyses of data in Figs. 1–3, are openly available in Dryad at https://doi.org/10.5061/dryad.3n5tb2rz2.