HER2 expression defines unique requirements for flotillin and c-Src in EGFR signaling Free

The epidermal growth factor (EGF) receptor (EGFR) is one of four members of the ErbB family of receptor tyrosine kinases, which also includes HER2 (ERBB2), HER3 (ERBB3) and HER4 (ERBB4) (Lemmon and Schlessinger, 2010). EGFR and other ErbB proteins play important roles in the development and homeostasis of various adult tissues, and ErbB proteins are also central drivers of tumor progression in a wide range of cancers (Sigismund et al., 2018), including breast cancers (Ali and Wendt, 2017; Nakai et al., 2016). Specifically, HER2 drives tumor progression in so-called HER2-positive breast cancers, whereas triple-negative breast cancers lack HER2 expression, instead often relying on EGFR (Cancer Genome Atlas Network, 2012; Hsu and Hung, 2016; O'Reilly et al., 2015; Oh and Bang, 2019). Hence, resolving the mechanisms that regulate EGFR, HER2 and other ErbB proteins is thus important to better understand how these signals can drive tumor progression.

Following binding to growth factor ligands such as EGF, EGFR undergoes asymmetric receptor homodimerization, resulting in activation of its intrinsic kinase domain that in turn triggers trans-autophosphorylation of the C-terminal domain of the receptor (Jura et al., 2009; Zhang et al., 2006). These phosphorylated tyrosine residues contribute to the formation of docking sites for effector proteins such as Grb2 and Shc1, resulting in the activation of intracellular signaling pathways such as the phosphoinositide-3-kinase (PI3K)–Akt, Ras–Erk, signal transducer and activator of transcription (STAT), and phospholipase C γ1 (PLCγ1) pathways (Wee and Wang, 2017).

Initial EGFR signaling occurs upon ligand binding at the plasma membrane. The plasma membrane contains several distinct types of dynamic nanodomains thought to contribute to receptor signaling through scaffolding or spatiotemporal organization of signaling intermediates (Delos Santos et al., 2015; Lu and Fairn, 2018). Upon binding to ligand and concomitant to signal activation, EGFR is recruited to clathrin-coated pits (CCPs), which are 50–200 nm assemblies of clathrin, AP2 and other proteins on the inner leaflet of the plasma membrane, leading to membrane curvature and eventually EGFR internalization into vesicles (McMahon and Boucrot, 2011; Mettlen et al., 2018). Several studies have shown that CCPs not only function in receptor endocytosis, but also have key roles in the regulation of specific aspects of EGFR signaling leading to activation of Akt (Cabral-Dias et al., 2022; Delos Santos et al., 2017; Garay et al., 2015; Pascolutti et al., 2019; Reis et al., 2015; Rosselli-Murai et al., 2018; Sigismund et al., 2008). Akt has three isoforms (Akt1, Akt2 and Akt3); henceforth we refer to Akt activation without distinguishing isoforms.

In addition to CCPs, other structures, such as membrane rafts, may control signaling by receptors including EGFR. Membrane rafts are dynamic membrane nanodomains that are enriched in and highly dependent on cholesterol and glycosphingolipids (Lu and Fairn, 2018; Mukherjee and Maxfield, 2004; Sonnino and Prinetti, 2013). Membrane rafts are a heterogeneous ensemble of nanodomains, at least some of which are driven by specific lipid-binding proteins. Flotillin-1 and -2 are highly similar palmitoylated proteins that oligomerize and bind cholesterol, thus supporting the formation of ∼100 nm protein-driven membrane rafts that scaffold and enrich a number of receptor tyrosine kinases, including EGFR, and select signaling molecules (Kurrle et al., 2012; Meister and Tikkanen, 2014; Staubach and Hanisch, 2011).

There are conflicting results about the requirement for cholesterol and/or flotillin proteins, and thus the involvement of flotillin membrane rafts, in EGFR signaling. EGFR has been observed in membrane fractions that are consistent with membrane rafts, and perturbations of cholesterol alter the membrane mobility of EGFR (Bag et al., 2015; Hiroshima et al., 2021 preprint; Hofman et al., 2008; Irwin et al., 2011; Orr et al., 2005; Pike et al., 2005; Puri et al., 2005; Wang et al., 2009). Some studies have reported that loss of membrane rafts upon cholesterol disruption increases ligand-binding affinity to EGFR and increases EGFR phosphorylation and/or signaling (Chen and Resh, 2002; Hiroshima et al., 2021 preprint; Lambert et al., 2008; Li et al., 2006; Liu et al., 2009; Pike and Casey, 2002; Ringerike et al., 2002; Roepstorff et al., 2002; Turk et al., 2012; Westover et al., 2003; Zhang et al., 2016). In contrast, others have reported that disruption of cholesterol has no effect on EGFR signaling (Sigismund et al., 2008; Ushio-Fukai et al., 2001; Zhang et al., 2016), or instead leads to loss of EGFR signaling (Gibson et al., 2009; Hea et al., 2007; Irwin et al., 2011). Consistent with the latter, disruption of flotillin-1 leads to a loss of EGF-stimulated EGFR clustering and phosphorylation, as well as disruption of downstream signaling (Amaddii et al., 2012).

Considering that perturbations of cholesterol, flotillins and, more generally, membrane rafts have inconsistent effects, as reported in different studies, indicates that some as-yet-unidentified cellular factors or context may define the need for membrane rafts for EGFR signaling. Although EGF specifically binds to EGFR, in the presence of HER2, EGF stimulation strongly favors formation of EGFR–HER2 heterodimers rather than the formation of EGFR homodimers that occurs in the absence of HER2 expression (Li et al., 2012b). Like EGFR homodimerization, EGFR–HER2 heterodimerization also leads to activation of receptor intrinsic kinase activity and receptor phosphorylation; however, EGFR–HER2 heterodimers exhibit differences in the extent of activation of specific signals (Hartman et al., 2012; Wolf-Yadlin et al., 2006) and may also exhibit delayed internalization compared to EGFR homodimers (reviewed by Bertelsen and Stang, 2014). Hence, co-expression of HER2 may underlie unique context-specific properties of EGFR signaling.

We have previously uncovered that HER2 expression renders EGF-stimulated activation of Akt insensitive to clathrin perturbation, whereas clathrin is an essential regulator of EGFR signaling in cells that lack HER2 expression (Garay et al., 2015). This suggests that HER2 expression alters the EGFR requirement for specific clathrin-scaffolded nanodomains; however, it is not known whether HER2 influences EGFR signaling dependence on another subtype of membrane nanodomain. HER2 can interact with flotillin-2 (Asp et al., 2014; Pust et al., 2013). While flotillin-2 is known to regulate HER2 stability at the plasma membrane, flotillin-1 and flotillin-2 have at least partly non-redundant functions (Bitsikas et al., 2014; Kurrle et al., 2012; Otto and Nichols, 2011) despite being largely present in the same plasma membrane structures (Fekri et al., 2019). This suggests that HER2–flotillin interactions may also control EGFR signaling when EGFR is heterodimerized with HER2. Flotillins may also more broadly control membrane lipids and dynamics, but whether flotillins and flotillin membrane rafts contribute to the regulation of EGFR and/or HER2 signaling, and thus underlie context-specific roles of membrane rafts in EGFR signaling, is poorly understood.

The possibility of context-specific dependence of EGFR signaling on flotillin membrane rafts also suggests context-specific recruitment to and functional requirement for specific signaling intermediates. Unlike EGFR, HER2 can interact directly with c-Src (also known as SRC) (Muthuswamy and Muller, 1995) via its catalytic kinase domain, leading to c-Src activation and potentiation of signaling elicited by the receptor (Kim et al., 2005; Marcotte et al., 2009). Hence, HER2 co-expression may also define unique requirements for specific signaling intermediates, in addition to defining requirements for specific plasma membrane nanodomains.

Here, we examined how HER2 expression may underlie the context-specific nanoscale organization of EGFR signaling at the plasma membrane. We used MDA-MB-231 triple-negative breast cancer cells and ARPE-19 cells, both of which endogenously express EGFR but exhibit no detectable HER2 (Delos Santos et al., 2017; Garay et al., 2015). We engineered derivatives of these cell lines that express HER2 to allow direct comparison of EGFR signaling in the presence or absence of HER2, likely as a result of EGFR–HER2 heterodimers or EGFR homodimers, respectively, in a common cell background. We find that HER2 expression leads to membrane cholesterol, flotillin-1 and c-Src being functionally required for EGFR signaling and cell proliferation, and establishes the unique recruitment of EGFR and c-Src to plasma membrane flotillin nanodomains. This work resolves important context-specific requirements for flotillin membrane rafts in EGFR signaling.

To directly compare the signaling of EGFR in the presence or absence of HER2 expression, we first generated a derivative cell line of MDA-MB-231 cells stably expressing HER2. Whereas we could not detect HER2 in parental MDA-MB-231 cells, we could readily detect HER2 in MDA-MB-231-HER2 cells by western blotting and immunofluorescence staining (Fig. 1A). The latter revealed expression of HER2 in a majority of MDA-MB-231-HER2 cells.

We next compared the signaling of MDA-MB-231 parental and HER2-expressing cells. Stimulation with EGF triggered a robust increase in the phosphorylation of EGFR (pY1068), Gab1 (pY627) and Akt (pS473) in both cell types (Fig. 1B,C). EGF stimulation of all three signals persisted longer in MDA-MB-231-HER2 cells than in MDA-MB-231 parental cells, although all largely returned to basal levels by 30 min. This suggests that differences in signal activation and regulation occur at the level of EGFR phosphorylation in the context of EGFR homodimers and EGFR–HER2 heterodimers, resulting in changes in the duration of signaling following ligand stimulation.

Prior to stimulation with EGF, MDA-MB-231-HER2 cells had a similar level of EGFR at the cell surface as parental MDA-MB-231 cells (Fig. S1A). In contrast, following 5 min of EGF stimulation, MDA-MB-231-HER2 cells exhibited higher levels of rhodamine–EGF (rho–EGF) associated with the cell surface (as detected by total internal reflection fluorescence microscopy, TIRF-M; Fig. S1B), suggesting that HER2 expression may cause a delay in internalization of EGFR. Consistent with this, we observed that MDA-MB-231-HER2 cells exhibited a delay in arrival of ligand-bound EGFR to early endosomes (Fig. S1C). This in turn suggests that the increased duration of EGFR signaling upon stimulation with EGF in HER2-expressing cells could be related to changes in EGFR membrane traffic, such as retention of EGFR at the cell surface in HER2-expressing cells. We also found that the proliferation and/or survival of MDA-MB-231-HER2 cells was robustly sensitive to the EGFR tyrosine kinase inhibitor gefitinib (Fig. 1E), perhaps more so than for cells lacking HER2 expression (Fig. 1D).

These differences in signaling and/or membrane traffic upon HER2 expression, likely resulting from distinct formation of EGFR homodimers and EGFR–HER2 heterodimers, are consistent with previous reports (Hartman et al., 2012; Wolf-Yadlin et al., 2006) and may reflect changes in the nanoscale organization of receptor signaling at the plasma membrane or other stages of membrane traffic of ligand-bound receptor complexes.

To resolve how differences in EGFR signaling upon co-expression of HER2 may reflect distinct functional requirements for membrane raft nanodomains in EGFR homodimer and EGFR–HER2 heterodimer signaling, we first examined the effect of cholesterol perturbation on EGFR signaling using nystatin treatment. Importantly, nystatin treatment led to a reduction in EGF-stimulated phosphorylation of Akt in MDA-MB-231-HER2 cells (Fig. 2A,C) but was without effect in parental MDA-MB-231 cells (Fig. 2A,B). That the magnitude of phosphorylation of EGFR and Akt after 5 min of EGF stimulation was comparable in cells expressing HER2 or not (Figs 1B, 2A) is consistent with the expression of HER2 altering the type of EGFR dimer formed (homodimers versus heterodimers). Nystatin treatment was without effect on EGF-stimulated EGFR phosphorylation in either cell line (Fig. 2A–C), which suggests that the loss of EGF-stimulated Akt phosphorylation in MDA-MB-231-HER2 cells upon treatment with nystatin is not due to impaired ligand-binding to cell surface EGFR.

Membrane rafts are a heterogeneous collection of dynamic membrane structures that include flotillin- and caveolin-based structures. To determine whether the effect of nystatin treatment in MDA-MB-231-HER2 cells was due to a requirement for flotillin proteins and thus perhaps flotillin nanodomains, we examined the effect of flotillin-1 or flotillin-2 silencing (Fekri et al., 2019). Treatment of MDA-MB-231 cells with flotillin-1 siRNA resulted in silencing of flotillin-1 but not flotillin-2, whereas flotillin-2 siRNA treatment resulted in silencing of both flotillin-1 and flotillin-2 (Fig. S2A). Treatment with flotillin-1 siRNA impaired EGF-stimulated Gab1 and Akt phosphorylation selectively in MDA-MB-231-HER2 cells (Fig. 2D,F) but not in parental MDA-MB-231 cells (Fig. 2D,E). In contrast, treatment with flotillin-2 siRNA was without effect on EGF-stimulated phosphorylation of Gab1 or Akt in either cell type. Since treatment with flotillin-2 siRNA resulted in silencing of both flotillin-1 and flotillin-2, it was initially surprising that flotillin-2 siRNA had a less pronounced effect on EGF-stimulated signaling than flotillin-1 siRNA. However, treatment with flotillin-2 siRNA caused a large increase in EGF-stimulated HER2 phosphorylation (relative to total HER2 levels; Fig. 2D,F), which was not observed following flotillin-1 siRNA treatment. Furthermore, flotillin-1 and -2 siRNA treatments also differed in their effects on total HER2 expression (Fig. S2B). This indicates that while loss of flotillin-1 (upon flotillin-1 siRNA treatment) leads to disruption of EGFR signaling, loss of both flotillin-1 and flotillin-2 together (upon flotillin-2 siRNA treatment) results in differences in HER2 expression and/or amplification of EGF-stimulated signaling at the level of the receptors that may offset some of the defects in signaling observed upon loss of flotillin-1 alone. This is consistent with previous findings of distinct, non-redundant roles of flotillin-1 and -2 (Bitsikas et al., 2014; Kurrle et al., 2012; Otto and Nichols, 2011). This in turn suggests that flotillin-1 is essential for EGF-stimulated activation of signaling, including that leading to activation of the PI3K–Akt pathway, in MDA-MB-231-HER2 cells that express HER2 but not in MDA-MB-231 cells.

Interestingly, silencing of flotillin-1 impaired EGF-stimulated HER2 phosphorylation in MDA-MB-231-HER2 cells (Fig. 2D,F). In contrast, flotillin-1 silencing was without effect on EGF-stimulated EGFR phosphorylation in HER2-expressing cells, whereas flotillin-1 silencing led to a significant decrease in EGF-stimulated EGFR phosphorylation in cells lacking HER2 (Fig. 2G). This suggests that similarly to the effects of nystatin, flotillin-1 silencing likely does not impact EGFR cell surface expression levels or ligand binding in HER2-expressing cells, and that the defects in EGFR signaling in HER2-expressing cells observed upon loss of flotillin-1 are likely the consequence of disruption downstream of ligand binding and EGFR phosphorylation. Taken together with the altered EGF-stimulated signaling upon co-expression of HER2, likely reflecting the formation of EGFR–HER2 heterodimers, these experiments suggest that cells co-expressing HER2 require flotillin-1 for multiple facets of signaling, including phosphorylation of HER2 within an EGFR–HER2 heterodimer and/or PI3K–Akt signaling downstream of the receptor.

To exclude the possibility that the effects of HER2 overexpression on EGFR signaling were unique to MDA-MB-231 cells, we examined ARPE-19 cells, which also lack endogenous HER2 expression. We previously engineered a derivative of these cells that stably expresses HER2 (ARPE-19-HER2) (Garay et al., 2015). Consistent with our results in MBA-MB-231 cells, ARPE-19 cells stably expressing HER2 exhibited loss of EGF-stimulated Akt phosphorylation upon flotillin silencing, whereas parental ARPE-19 cells that lack HER2 were unaffected by flotillin silencing (Fig. S3). Furthermore, silencing of flotillin-1 impaired EGF-stimulated Akt phosphorylation in Sk-Br-3 cells, a breast cancer cell line that endogenously expresses EGFR and high levels of HER2 (Fig. 2H). Collectively, these results indicate that cholesterol and flotillins have a selective role in promoting EGFR signaling in cells that express HER2, but not in cells that lack HER2 expression. This in turn suggests that EGFR and/or its signals may exhibit localization and/or scaffolding in flotillin nanodomains upon HER2 co-expression.

To determine how HER2 co-expression may alter the spatial organization of EGFR with respect to flotillin nanodomains and CCPs at the plasma membrane, we used TIRF-M to visualize labeling of the receptor and markers for each plasma membrane nanodomain. Labeling of ligand-bound EGFR was achieved by treatment with 20 ng/ml rho–EGF, while labeling of flotillin and clathrin nanodomains was achieved by immunofluorescence staining using specific antibodies for flotillin-1 (validated previously; Fekri et al., 2019) and the clathrin adaptor protein AP2 (Fig. 3A; Fig. S4A). We also labeled cells with fluorescently conjugated trastuzumab to confirm expression of HER2. MDA-MB-231-HER2 cells exhibited some flotillin-1 structures that were also positive for rho–EGF, whereas these structures were not as readily observed in MDA-MB-231 parental cells that lacked HER2 expression (Fig. 3A).

To quantitatively compare rho–EGF localization in flotillin structures in MDA-MB-231 parental and HER2-expressing cells, we used automated detection and analysis of these diffraction-limited objects using a Gaussian-based modeling approach (Aguet et al., 2013), as we have used previously to detect signaling proteins within various plasma membrane structures (Cabral-Dias et al., 2022; Delos Santos et al., 2017; Lucarelli et al., 2016). This method allows for the detection of diffraction-limited objects in a primary channel, followed by detection of the intensity of fluorescence signal of the primary or other channels by the amplitude of the Gaussian model at the position of each object detected in the primary channel. Measurement of background overlap in images is done by similar analysis of a set of images with randomized positions of the different channels.

We first used this method for detection of flotillin-1 objects. While some level of rho–EGF fluorescence was detected in flotillin-1 objects in MDA-MB-231 cells (Fig. 3B), this was similar to that obtained in randomized image sets (Fig. S4H). In contrast, MDA-MB-231-HER2 cells had significantly more rho–EGF enrichment in flotillin-1 structures than cells lacking HER2 expression, as observed when examining the levels of rho–EGF in flotillin structures in individual cells (Fig. 3B, left) and the mean values from independent experiments (Fig. 3B, middle). Subtraction of the background overlap obtained from randomized images revealed a robust increase in rho–EGF intensity in flotillin-1 structures upon expression of HER2 (Fig. 3B, right). We did not observe any differences in the density (number per area) or intensity of flotillin-1 objects in MDA-MB-231 cells versus MDA-MB-231-HER2 cells (Fig. S4D,E). Since rho–EGF binds directly to EGFR and not to HER2, this indicates that HER2 co-expression alters the distribution of EGFR to cause enrichment of EGFR within flotillin nanodomains, likely as part of an EGFR–HER2 heterodimer.

As HER2-expressing cells had a small but significant increase in cell surface rho–EGF following 5 min of labeling (Fig. S1B) that may be due to delayed EGF-stimulated endocytosis, we next performed similar object-based analysis with rho–EGF as the primary objects. We observed an increase in the detection of rho–EGF objects (Fig. S4B) and in the intensity of rho–EGF in these objects (Fig. S4C) in HER2-expressing cells, likely reflecting a delay in EGFR internalization in HER2-expressing cells. Taken together with the increased detection of rho–EGF within flotillin-1 objects (Fig. 3B), this suggests that although there is increased ligand-bound EGFR associated with flotillin-1 objects in HER2-expressing cells, this may not necessarily represent specific enhanced recruitment of EGFR to flotillin-1 objects in HER2-expressing cells given the increased cell surface abundance of ligand-bound HER2 in this condition.

EGF binding triggers recruitment of EGFR to CCPs at the plasma membrane, which eventually leads to receptor endocytosis. The increased recruitment of EGFR to flotillin-1 structures in HER2-expressing cells could indicate that in HER2-expressing cells there may be either loss of EGFR recruitment to plasma membrane clathrin structures, reflecting competitive recruitment of a limited pool of EGFR to each nanodomain, or recruitment of EGFR to both types of structure. To resolve this, we used primary detection of clathrin structures via AP2 labeling as we had similarly done for flotillin-1 structures. HER2-expressing cells had elevated levels of rho–EGF within CCPs compared to the MDA-MB-231 parental cells lacking HER2 expression, as observed when examining individual cells within an experiment (Fig. 3C, left) and the mean measurements from independent experiments (Fig. 3C, middle). Subtraction of the background overlap (Fig. S4I) revealed a small but significant increase in rho–EGF intensity in CCP structures upon expression of HER2 (Fig. 3C, right). There was no difference in the number (Fig. S4F) or intensity (Fig. S4G) of AP2 objects (CCPs) between MDA-MB-231-HER2 cells and MDA-MB-231 cells. That there is an increase in ligand-bound EGFR levels within CCPs (after 5 min of EGF stimulation) in HER2-expressing cells suggests that delays in EGFR internalization in these HER2-expressing cells, as previously reported (Bertelsen and Stang, 2014), may not be due to defective recruitment of ligand-bound receptor complexes to CCPs. Instead, this suggests delayed progression of CCPs harboring EGFR to internalized vesicles subsequent to recruitment of the ligand-bound receptor complexes in HER2-expressing cells.

The increased detection of EGFR in flotillin-1 and AP2 (clathrin) objects suggests that HER2-expressing cells could exhibit overlap of flotillin-1 nanodomains and CCPs, with these structures also containing EGFR–HER2 heterodimers. To resolve this, we examined the overlap of flotillin-1 within AP2 objects (CCPs). HER2-expressing cells had elevated levels of flotillin-1 within CCPs compared to cells that lack HER2 expression, as observed when examining individual cells within an experiment (Fig. 3D, left), and the mean flotillin-1 intensity per CCP in independent experiments (Fig. 3D, middle). This higher level of flotillin-1 within CCP structures in HER2-expressing cells was also relative to the background overlap in randomized images (Fig. S4J); subtracting the background overlap obtained from randomized images revealed an increase in flotillin-1 intensity in CCP structures upon expression of HER2 (Fig. 3D, right). This suggests that EGF-stimulated HER2-expressing cells exhibit increased recruitment of flotillin to AP2 (clathrin) structures, which may be related to the increase in ligand-bound EGFR found in clathrin structures in HER2-expressing cells.

To complement the results obtained by using rho–EGF to selectively label ligand-bound EGFR, we also performed similar experiments using an EGFR Fab fragment specific to the ectodomain of the receptor. This strategy allows labeling of cell surface EGFR, whether ligand bound or not (Sugiyama et al., 2022 preprint), and thus allows resolution of how acute EGF stimulation changes EGFR localization within flotillin nanodomains in each cell type. There was no detectable increase of EGFR within flotillin nanodomains upon EGF stimulation in MDA-MB-231 cells lacking HER2 expression (Fig. 3E,F; Fig. S5). In contrast, in MDA-MB-231-HER2 cells, EGF stimulation triggered a significant increase in EGFR enrichment within flotillin-1 objects (Fig. 3E,F; Fig. S5). These results indicate that the recruitment of EGFR to flotillin-1 structures in HER2-expressing cells may not merely be due to prolonged retention of ligand-bound EGFR at the cell surface, but that this may represent selective recruitment of EGFR–HER2 heterodimers to flotillin-1 nanodomains upon EGF stimulation.

Collectively, these experiments indicate that the distinct requirement for flotillin-1 in MDA-MB-231 cells that co-express HER2 correlates with an increased detection of ligand-bound EGFR (and thus likely EGFR–HER2 heterodimers) in flotillin-1 structures compared to that in cells lacking HER2 expression. These results also suggest that a subset of EGFR–HER2 heterodimers are detected within flotillin-1 structures that may also overlap with clathrin structures.

To determine how the alteration in EGFR spatial organization to flotillin nanodomains and the requirement for flotillin-1 in EGFR signaling upon HER2 co-expression broadly impacts cell physiology, we next examined the effect of flotillin-1 silencing on cell proliferation and survival. To do so, we used an automated imaging system and quantified cellular confluence in phase-contrast images over time. MDA-MB-231 (parental) cells exhibited a time-dependent increase in cell confluence, indicating cellular proliferation, which was unaffected by flotillin-1 silencing (Fig. 4A,C). In contrast, the increase in cell confluence reflecting proliferation in MDA-MB-231-HER2 cells was significantly impaired by flotillin-1 silencing (Fig. 4B,D). Collectively, these results indicate that there is a correlation between the selective requirement for flotillin-1 in EGFR signaling to PI3K–Akt, the enhanced localization of EGFR to flotillin-1 structures, and the selective requirement for flotillin-1 for cell proliferation and/or survival in cells co-expressing HER2 compared to cells lacking HER2 expression.

The unique requirement for flotillin-1 for EGFR signaling and enhanced recruitment of the receptor to plasma membrane flotillin structures in HER2-expressing cells suggests that HER2 expression may also induce a unique requirement for specific signaling intermediates that may be scaffolded within flotillin nanodomains. A prime candidate for such a unique signaling intermediate is c-Src, given that this kinase is able to interact directly with HER2 but not with EGFR (Kim et al., 2005; Marcotte et al., 2009). Thus, we next sought to determine whether c-Src may be uniquely required for EGFR signaling upon HER2 co-expression. Silencing of c-Src (Fig. S6A) impaired EGF-stimulated Akt phosphorylation in MDA-MB-231-HER2 cells (Fig. 5B) but was without effect in parental MDA-MB-231 cells lacking HER2 expression (Fig. 5A). Furthermore, and consistent with the effects of flotillin-1 silencing, c-Src silencing led to a reduction of EGF-stimulated HER2 phosphorylation (Fig. 5C), but not of EGF-stimulated EGFR phosphorylation (Fig. 5D). The similarity of the phenotypes observed upon flotillin-1 and c-Src silencing on signaling by EGFR, specific for HER2-expressing cells, suggests that c-Src and flotillins may regulate EGFR signaling by contributing to a common mechanism.

c-Src can interact directly with HER2 via binding to the HER2 kinase domain with a motif surrounding Y877 of HER2 (Kim et al., 2005; Marcotte et al., 2009). Flotillin nanodomains are enriched in gangliosides, and globotriaosylceramide (Gb3) co-immunoprecipitates with c-Src, suggesting formation of a transbilayer-spanning complex (Roy et al., 2020). This may indicate that flotillin nanodomains contribute to the selective engagement of c-Src by EGFR–HER2 heterodimers. To test this possibility, we first examined whether c-Src was recruited to flotillin nanodomains. We transfected cells with eGFP-tagged c-Src (c-Src–eGFP), imaged the resulting cells using TIRF-M and performed automated image analysis that allowed measurement of c-Src fluorescence intensity within flotillin structures (Fig. 6A,B). c-Src fluorescence intensity was increased significantly in flotillin-1 structures by EGF stimulation in MDA-MB-231-HER2 cells (Fig. 6B, right panels), whereas EGF stimulation did not impact c-Src levels in flotillin structures in parental MDA-MB-231 cells lacking HER2 expression (Fig. 6B, left panels). As in earlier figures, in Fig. 6 we present measurements of the levels of c-Src in flotillin objects in individual cells within an experiment (violin plots) as well as the measurements from several independent experiments (bar graphs with individual data points). For these experiments, we selected cells that expressed low but detectable levels of c-Src–eGFP, with similar mean levels of expression among the conditions examined (Fig. S6B). Interestingly, the magnitude of the increase in c-Src–eGFP fluorescence in flotillin-1 structures in HER2-expressing cells is comparable to the EGF-stimulated gain in Fyn fluorescence in CCPs in cells that lack HER2 expression (Cabral-Dias et al., 2022). In cells that lack HER2 expression, this recruitment of Fyn to clathrin structures is required to regulate EGFR signaling, which thus complements the observed EGF-stimulated gain of c-Src enrichment within flotillin-1 structures and the requirement for c-Src upon EGF stimulation in HER2-expressing cells (Figs 5, 6). This indicates that HER2 expression alongside EGF stimulation contributes to the recruitment of c-Src to flotillin structures.

The results reported above suggest that while HER2 may interact with c-Src directly, the interaction of c-Src with HER2 could further require flotillin nanodomains, perhaps as a form of coincidence detection. To test this, we examined the colocalization of GFP-tagged c-Src with ligand-bound EGFR objects (labeled with rho–EGF) in MDA-MB-231-HER2 cells, alongside flotillin-1 silencing, using time-lapse TIRF-M (Fig. 6C). We used automated detection of rho–EGF puncta by Gaussian modeling of the point-spread function of these diffraction-limited objects, followed by quantification of c-Src–eGFP fluorescence within these EGFR objects, similar to the analysis presented in Fig. 3. These experiments revealed that EGF stimulation causes recruitment of c-Src–eGFP to EGFR objects in MDA-MB-231-HER2 cells treated with control (non-targeting) siRNA (Fig. 6D). In contrast, MDA-MB-231-HER2 cells treated with flotillin-1 siRNA exhibited no detectable EGF-stimulated increase in c-Src–eGFP within EGFR objects (Fig. 6D). Quantification of the area under the curve for these time-lapse measurements of c-Src–eGFP recruitment to EGFR puncta showed a significant reduction of this phenomenon in flotillin-1-silenced cells (Fig. 6E). These results suggest that flotillin-1 may serve as a part of a coincidence-detection mechanism alongside HER2 to recruit c-Src to EGFR–HER2 signaling complexes, although this cannot exclude broader changes in c-Src localization upon flotillin-1 silencing that would render c-Src unavailable to engage with EGFR.

Since we observed a unique role for flotillin nanodomains in controlling cell proliferation (Fig. 4), we next sought to determine whether c-Src is also similarly required for control of cell proliferation selectively in HER2-expressing cells. As we had observed in Fig. 4A, MDA-MB-231 (parental) cells exhibited a time-dependent increase in cell confluence, indicating cellular proliferation, which was unaffected by c-Src silencing (Fig. 7A,C). In contrast, the increase in cell confluence reflecting proliferation in MDA-MB-231-HER2 cells was significantly impaired by c-Src silencing (Fig. 7B,D). These results suggest that the unique requirement for c-Src in MDA-MB-231-HER2 cells but not parental MDA-MB-231 cells lacking HER2 for EGFR signaling is mirrored by a unique requirement in HER2-expressing cells for c-Src for cell proliferation.

We found a selective requirement for flotillin-1 and c-Src in EGF-stimulated activation of EGFR signaling leading to Akt phosphorylation in cells expressing HER2 (Fig. S7). Consistent with this, in cells co-expressing HER2, plasma membrane flotillin-1 structures exhibited increased association with EGFR upon EGF stimulation. Cell proliferation and/or survival was selectively dependent on flotillin-1 and c-Src in cells co-expressing HER2. Hence, we propose that HER2 co-expression defines a shift in EGFR signaling to a flotillin membrane raft- and c-Src-dependent mechanism. It is likely that the distinct effects of HER2 expression on EGFR are due to formation of EGFR–HER2 heterodimers, in particular for signaling experiments that were performed following serum starvation and stimulation with acute EGF, which only binds to EGFR.

Many studies have examined whether and how EGFR may be localized to various membrane raft domains. Some studies have relied on isolation of raft-enriched membrane fractions (reviewed by Delos Santos et al., 2015; Lu and Fairn, 2018). Other studies have relied on detection of colocalization of EGFR with membrane raft markers, such as cholera toxin or GM1 ganglioside, using fluorescence (Hofman et al., 2008) or electron (Puri et al., 2005) microscopy methods. The effect of cholesterol perturbation on EGFR mobility can also be measured by single-particle tracking, which allows inference of the localization of EGFR to cholesterol-dependent membrane rafts (Hiroshima et al., 2021 preprint; Orr et al., 2005). Collectively, these studies provide evidence that EGFR resides within dynamic membrane raft domains, at least under some cellular contexts.

In addition to examining membrane raft localization, some previous studies have also investigated the role of membrane rafts in EGFR signaling, largely by cholesterol disruption. These can be broadly classified into three categories based on their findings: (1) that cholesterol disruption increases EGFR ligand binding and/or signaling (Chen and Resh, 2002; Hiroshima et al., 2021 preprint; Lambert et al., 2008; Li et al., 2006; Liu et al., 2009; Pike and Casey, 2002; Ringerike et al., 2002; Roepstorff et al., 2002; Turk et al., 2012; Westover et al., 2003; Zhang et al., 2016), (2) that cholesterol disruption has little or no effect on EGFR signaling (Sigismund et al., 2008; Ushio-Fukai et al., 2001; Zhang et al., 2016) and (3) that cholesterol disruption impairs EGFR signaling (Gibson et al., 2009; Hea et al., 2007; Irwin et al., 2011).

Since these studies suggest that EGFR signaling distinctly requires cholesterol and membrane rafts in some cellular contexts, the expression of other specific proteins may shift EGFR nanoscale organization to membrane rafts, leading to altered requirements for signaling, although the effects of cholesterol disruption may be complex, and additional mechanisms may contribute to cholesterol-dependent signaling regulation. Our results suggest that the availability of HER2 to form heterodimers with EGFR is one of the factors that defines this switch in regulation by cholesterol and/or flotillin-1 membrane rafts. However, few of these previous studies directly examined the relative expression of EGFR and HER2 or examined the effect of alterations in HER2 expression in a common cellular background, as we have done here. Our results show that HER2 co-expression induces cholesterol- and flotillin-1-dependence for EGFR signaling and may regulate EGFR localization to flotillin structures.

Flotillin-1 and -2 each contain an N-terminal prohibitin homology (PHB) domain, also known as an SPFH (stomatin, prohibitin, flotillin, HflK/C) domain. This PHB domain has membrane insertion domains and S-acylation sites, and mediates interaction with cholesterol (Babuke et al., 2009; Kurrle et al., 2012; Li et al., 2012a; Meister and Tikkanen, 2014; Morrow et al., 2002; Neumann-Giesen et al., 2004; Staubach and Hanisch, 2011; Strauss et al., 2010). An additional C-terminal domain mediates oligomerization (Babuke et al., 2009; Otto and Nichols, 2011; Solis et al., 2007). These properties allow flotillins to form membrane nanodomains that are enriched in cholesterol and glycosphingolipids, as well as some signaling proteins.

Signaling by activated dimers involving EGFR requires formation of an asymmetric kinase dimer, in which one kinase domain serves as the ‘activator’ and the other as the ‘receiver’, such that only the receiver kinase is active (Jura et al., 2009; Zhang et al., 2006). Importantly, in the context of EGFR–HER2 heterodimers, the HER2 kinase preferentially adopts the activator (inactive) position (Macdonald-Obermann et al., 2012; Ward and Leahy, 2015). Thus, phosphorylation of HER2 upon EGF stimulation is expected to result from direct phosphorylation by the EGFR kinase domain or via an indirect mechanism, such as by c-Src or another kinase.

We observed that flotillin-1 and c-Src are required for EGF-stimulated phosphorylation of Akt and of Y1248 on HER2, but not for the phosphorylation of EGFR Y1068. This indicates that perturbations of flotillin-1 or c-Src do not alter EGFR cell surface levels or ligand binding, but instead may alter signaling by EGFR downstream of EGFR–HER2 heterodimerization. One possibility is that cholesterol and/or other lipids in flotillin membrane rafts impact the conformation of the EGFR–HER2 heterodimer, gating phosphorylation of certain receptor residues such as Y1248 on HER2. Interestingly, computational modeling indicates that the HER2 transmembrane domain may bind cholesterol, impacting the preference for the formation of several possible dimer configurations by the HER2 transmembrane domain (Pawar and Sengupta, 2021). This suggests that cholesterol present within flotillin nanodomains can uniquely regulate the conformation of EGFR–HER2 heterodimers to favor EGF-stimulated phosphorylation of Y1248 and other residues essential for signal transduction.

In addition, flotillin nanodomains can also regulate EGFR signaling by enrichment of signaling regulators or intermediates therein. The ganglioside Gb3 is enriched within flotillin nanodomains (Brandel et al., 2021), and despite being on opposite leaflets, Gb3 can co-immunoprecipitate with c-Src (Roy et al., 2020), suggesting formation of transbilayer-spanning complexes. Indeed, various Src-family kinases are associated with flotillins and flotillin nanodomains (Stuermer, 2010). Furthermore, c-Src directly binds to the HER2 kinase domain (Kim et al., 2005; Marcotte et al., 2009). We found selective recruitment of c-Src upon EGF stimulation to plasma membrane flotillin structures, which is consistent with a coincidence-detection mechanism involving flotillin and HER2 for the selective recruitment of c-Src. This may in turn allow c-Src to phosphorylate specific residues on EGFR or HER2, including Y1248 on HER2. However, neither c-Src silencing nor flotillin-1 silencing impaired EGF-stimulated phosphorylation of Y1068 on EGFR, suggesting that flotillin and c-Src are required subsequent to ligand binding and receptor heterodimerization.

Flotillins may also more broadly remodel the plasma membrane lipid composition by altering membrane fluidity, as shown for bacterial flotillins (Zielińska et al., 2020), or regulating availability of lipids such as sphingosine (Riento et al., 2018). These broad effects could account for some of the effects on Akt signaling produced by silencing flotillin-1, as Akt may require membrane raft domains for activation in some contexts (Lasserre et al., 2008). Nonetheless, our results suggest that any such broad changes occurring upon flotillin-1 disruption impact EGFR signaling selectively in the context of HER2 co-expression.

Hence, we propose that the unique requirement for flotillin-1 and c-Src in signaling by EGFR–HER2 heterodimers reflects a distinct requirement for these proteins for phosphorylation of a subset of EGFR or HER2 sites and/or those of receptor-proximal signaling proteins, such as Gab1, that lead to phosphorylation of Akt. This may reflect more direct functions of flotillins as signaling scaffolds for EGFR–HER2 and receptor-proximal signals or could result from broader regulation of membrane properties by flotillins that manifests as EGFR signaling regulation specifically in the context of HER2-expressing cells. In contrast, EGF ligand binding to EGFR may be largely flotillin-1 and c-Src independent.

We found that HER2 expression leads to enhanced detection of ligand-bound EGFR complexes within clathrin structures. Although some studies have reported that flotillins may be present within clathrin endocytic structures (Ge et al., 2008; Ge et al., 2011; Meister and Tikkanen, 2014), others have found them to be largely distinct (Dam et al., 2020; Fekri et al., 2019; Glebov et al., 2006). Our findings suggest that there is some overlap between flotillin-1 and clathrin structures in MDA-MB-231 cells stimulated with EGF, which is enhanced upon expression of HER2. This suggests that EGFR–HER2 heterodimers reside within flotillin nanodomains or CCPs, and that there may be some overlap of these protein structures controlled by expression of specific receptors such as EGFR and/or HER2.

This in turn suggests that recruitment of EGFR–HER2 to clathrin and flotillin structures could be sequential. In such a model, recruitment of EGFR–HER2 heterodimers to flotillin nanodomains may precede the transit of receptor–flotillin complexes to CCPs, which would then lead to internalization. Plasma membrane flotillin structures are short-lived, with ∼60% of structures having lifetimes of less than 15 s (Fekri et al., 2019), which is shorter than the mean lifetime of CCPs (Aguet et al., 2013). It is possible that receptor complexes thus persist in clathrin structures following disassembly or loss of flotillins, or that flotillins internalize alongside EGFR–HER2. Resolving the contribution of flotillins to EGFR and HER2 membrane dynamics may require single-particle tracking of EGFR or other methods to track EGFR dynamics that are beyond the scope of this study. We focused here on how each nanodomain distinctly regulates EGF-stimulated signaling upon expression of HER2.

We and others have reported that plasma membrane clathrin structures are required for regulation of EGF-stimulated Akt phosphorylation (Cabral-Dias et al., 2022; Delos Santos et al., 2017; Garay et al., 2015; Pascolutti et al., 2019; Reis et al., 2015; Rosselli-Murai et al., 2018; Sigismund et al., 2008). Specifically, we have found that in cells expressing EGFR but not HER2, CCPs are enriched in signaling proteins such as TOM1L1, Fyn, and SHIP2 (also known as INPPL1), leading to the selective activation of Akt2 (Cabral-Dias et al., 2022). We have also previously reported that expression of HER2 renders cells insensitive to clathrin perturbation for EGF-stimulated Akt phosphorylation (Garay et al., 2015). This suggests that the signaling requirements of EGFR–HER2 heterodimers that depend on flotillin-1 cannot be supplied by clathrin structures in the absence of flotillin-1. Although a subset of clathrin structures are enriched in Fyn (Cabral-Dias et al., 2022), a Src-family kinase, it is possible that the interaction of c-Src with HER2 is specific to c-Src and not Fyn, making clathrin-localized Fyn unable to compensate for loss of flotillin and/or c-Src for control of EGFR–HER2 signaling.

On the other hand, clathrin becomes dispensable for EGF-stimulated Akt phosphorylation selectively upon expression of HER2 (Garay et al., 2015). This in turn suggests that flotillin-1 and/or flotillin nanodomains may be redundant with the function of clathrin for activation of PI3K–Akt signaling in the context of HER2 expression, perhaps by scaffolding redundant signaling intermediates. While resolving this mechanism is beyond the scope of the current study, we nonetheless establish for the first time that expression of HER2 alters the functional requirements for EGFR signaling and define cholesterol, flotillin-1 and c-Src as EGFR signaling intermediates that are unique to HER2-expressing cells.

Our results suggest that co-expression of HER2 is an important determinant of the need for flotillin-1 and c-Src, and thus perhaps of unique requirements for membrane nanodomains or control of broader membrane parameters such as fluidity, for the resulting control of signal transduction and/or cell survival and proliferation. Consistent with this, flotillin-2 expression is correlated with reduced survival and poor prognosis in HER2-positive breast cancer (Pust et al., 2013) and gastric cancer (Zhu et al., 2013). Furthermore, HER2-positive cancers that acquire resistance to trastuzumab show elevated levels of c-Src activity, which is correlated with poor patient survival (Peiró et al., 2014; Zhang et al., 2011). Although triple-negative breast cancer cells such as the MDA-MB-231 cells that we examine here are defined by a lack of significant expression of HER2, some studies have found that a subset of these tumors may express low but detectable levels of HER2, despite no amplification of ERBB2 (Marra et al., 2020). As such, it may be useful in future studies to expand on this work to examine how even relatively low levels of expression of HER2 can influence EGFR signaling in triple-negative breast cancer and other cancers.

More broadly, EGFR and HER2 are co-expressed and function in many cells and tissues, both in physiological and pathophysiological settings. Our results contribute to the understanding of cellular contexts, in this case HER2 expression, that may specify distinct requirements for membrane rafts for EGFR signaling.

Anti-phospho-EGFR (Y1068; 3777S; 1:1000), anti-phospho-HER2 (Y1248; 2247S; 1:1000), anti-phospho-Gab1 (Y627; 3233S; 1:1000), anti-flotillin-1 (18634S; 1:1000), anti-actin (8456S; 1:1000), anti-c-Src (2123P; 1:1000) and anti-HER2 (2165S; 1:1000) primary antibodies were from Cell Signaling Technology (Danvers, MA, USA); anti-phospho-Akt (S473; 44-621G; 1:1000) primary antibody was from Invitrogen (Waltham, MA, USA); anti-EGFR antibody (CPTC-EGFR-1-c; 1:1000) was from Developmental Studies Hybridoma Bank (Iowa City, IA, USA); anti-Akt (cat. #2920S, 1:1000) and anti-Gab1 (cat. #3232S, 1:1000) primary antibodies were from Cell Signaling Technology (Danvers, MA, USA); anti-flotillin 2 (cat#sc-48398, 1:1000) and anti-vinculin (sc-73614, 1:1000) primary antibodies were from Santa Cruz Biotechnology, Dallas, TX, USA; anti-ErbB2 (CD34) PE-Vio770 used at 1:400 for FACS sorting of HER2-positive stable cells was obtained from Miltenyi Biotec (Germany). Antibodies to detect clathrin (X-22) (Brodsky, 1985) were generated in house. Secondary antibodies were anti-mouse (cat. # 7076S; 1:1000) and anti-rabbit (cat. #7074S; 1:1000) from Cell Signaling Technology (Danvers, MA, USA).

MDA-MB-231, MDA-MB-231-HER2, ARPE-19 and Sk-Br-3 cell lines were obtained and cultured as previously described (Bautista et al., 2018; Bone et al., 2017). Cells were monitored for contamination using DAPI staining at least monthly. The ARPE-19 cells stably expressing HER2 have been described previously (Garay et al., 2015). Cells were maintained using RPMI (Sigma-Aldrich) containing 10% fetal bovine serum (Life Technologies), 100 μg/ml streptomycin and 100 U/ml penicillin (Life Technologies) at 37°C and 5% CO₂. MDA-MB-231-HER2 cells were grown in medium supplemented with 1 ug/ml of puromycin to maintain HER2 expression.

Parental MDA-MB-231 cells were transfected with pBABEpuro-ERBB2 plasmid (see below) using FuGENE HD (Promega) as per the manufacturer’s instructions. Briefly, cells were incubated in Opti-MEM (Life Technologies) containing reagent and plasmid at 37°C for 24 h. HER2-positive cells were selected using medium containing 3 µg/ml puromycin and then maintained in in 1 µg/ml puromycin. To obtain homogenous selection of HER2 stable cells, cell sorting was employed by labeling intact cells with the fluorescently conjugated HER2 antibody ErbB2 PE-Vio770 and then subjecting them to fluorescence-activated cell sorting using a FACS Aria III (BD Biosciences), selecting the cells in the top 10% of HER2 expression to generate MDA-MB-231-HER2 stable cells.

pBABEpuro-ERBB2 was from Addgene (deposited by Matthew Meyerson; Addgene plasmid 40978; RRID:Addgene 40978). mEmerald Hc-SRC-N-18 (endoding c-Src–eGFP) was also from Addgene (deposited by Michael Davidson; Addgene plasmid 54118; RRID:Addgene_54118).

All transient transfections were performed using FuGENE HD as per the manufacturer's instructions. Cells were incubated in OptiMEM with FuGENE HD and DNA mixture at a 3:1 ratio for 4–6 h, after which medium containing the transfection mixture medium was replaced with regular growth medium. Experiments were conducted 16–18 h post transfection.

siRNA gene silencing was performed using RNAiMAX (Thermo Fisher Scientific) and customized targeted siRNA sequences as per the manufacturer's instruction and as previously described (Delos Santos et al., 2017). Cells were seeded on a 6-well plate 24 h before transfection. Briefly, cells were initially washed with 1× PBS before the addition of OptiMEM medium (Thermo Fisher Scientific) for transfection. RNAiMAX and the customized siRNA sequences were added to the cellular medium at a concentration of 220 pmol/l and incubated for 4–6 h before replacing the transfection medium with regular growth medium. Two rounds of transfection were performed, 24 h and 48 h before the experiment was conducted. The siRNA sequences were used as follows (sequence in parentheses indicates duplex overhang): control (sense), 5′-CGUACUGCUUGCGAUACGG(UU)-3′; control (antisense), 5′-CCGUAUCGCAAGCAGUACG(UU)-3′; flotillin-1 (sense), 5′-UGGCCAAGGCACAGAGAGA(UU)-3′; flotillin-1 (antisense), 5′-UCUCUCUGUGCCUUGGCCA(UU)-3′; c-Src (sense): 5′-GCAGAGAACCCGAGAGGGA(UU)-3′; c-Src (antisense), 5′-UCCCUCUCGGGUUCUCUGC(UU)-3′.

Cell lysis and western blotting was performed as previously described (Bautista et al., 2018). Following treatment and stimulation with EGF as indicated, cells were washed with ice-cold PBS and prepared for lysis by adding 2× Laemmli sample buffer (0.5 M Tris-HCl pH 6.8, 20% glycerol, 10% SDS), 1 mM sodium orthovanadate, 10 mM okadaic acid and 20 mM of protease inhibitor (cat. #PIC001.1, Bioshop Canada, Burlington, ON, Canada), followed by passage of lysates 5× through a 27.5-gauge syringe and heating at 65°C for 15 min. Next, 10% β-mercaptoethanol and Bromophenol Blue (a few drops of a 0.05% stock solution) were added to the lysates prior to use of samples for SDS-PAGE. Molecular weight markers used were Novex Sharp Pre-stained Protein Standard (LC5800) or PageRuler Prestained Protein Ladder (26617) (Thermo Fisher Scientific). Following SDS-PAGE, proteins were transferred to a PVDF membrane and subjected to blocking, then incubated with appropriate antibodies as previously described (Antonescu et al., 2011). Western blot signal was detected using a Bio-Rad ChemiDoc System upon soaking membranes in Luminata Crescendo HRP substrate (Millipore Sigma), ensuring that the signal was not saturated at any pixel. For all western blot images, full images are shown in Fig. S8.

Western blot images were quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA) (Schneider et al., 2012) by signal integration in the area of each lane approximating each band. This measurement for phosphorylated proteins (e.g. phosphorylated Akt, pAkt) was then normalized to the loading control (e.g. actin) signal, and subsequently normalized to the total protein (e.g. Akt) signal, obtained following reblotting. In each experiment, the resulting normalized pAkt:total Akt signal in each condition was expressed as a fraction of the normalized pAkt:total Akt measurement in the control condition stimulated with EGF for 5 min. Statistical analysis was performed by two-way (Fig. 1) or one-way (Figs 2, 5) ANOVA followed by Bonferroni or Tukey post hoc tests, respectively, with P<0.05 used as a threshold for establishing differences between experimental conditions.

Prior to all experiments (except cell proliferation experiments), the MDA-MB-231, MDA-MB-231-HER2, ARPE-19, ARPE19-HER2, or Sk-Br-3 cells used were serum starved for 1 h. For some experiments (Fig. 2A–C), cells were then treated with medium containing 50 µM nystatin (Sigma-Aldrich, St. Louis, MI, USA) or a corresponding volume of dimethyl sulfoxide (vehicle control) for 1 h. Unless otherwise specified, cells were stimulated with 5 ng/ml of EGF (cat. #PHG0311, Thermo Fisher Scientific) for 5 min.

For experiments shown in Fig. 3A–D, Fig. 6C–E, Fig. S1B,C and Fig. S4 involving labeling with rhodamine–EGF (rho–EGF), following serum starvation, cells were stimulated with 20 ng/ml rho–EGF (generated in house; Lucarelli et al., 2017) or (unlabeled) EGF for 5 min prior to fixation/permeabilization with ice-cold methanol at −20°C for 20 min, which has been reported previously to allow detection of flotillin structures (Possidonio et al., 2014). For experiments shown in Fig. 3E,F and Figs S1A, S5 involving labeling with the EGFR Fab, following serum starvation, cells were incubated for 15 min at 37°C with Fab–Cy3B (Fab conjugated to Cy3B; described in Sugiyama et al., 2022 preprint) then stimulated, as indicated, for 5 min with (unlabeled) EGF prior to fixation/permeabilization with ice-cold methanol at −20°C for 20 min. In all labeling experiments, following blocking in 3% BSA solution, cells were incubated with specific primary antibodies at room temperature for 1 h, followed by washing and incubation with appropriate secondary antibodies. Cells were maintained in PBS at 4°C prior to TIRF-M imaging.

For experiments related to Fig. 1A, Fig. 3, Fig. 6A,C and Figs S1B, S4, S5, cell samples were imaged using a Quorum (Guelph, ON, Canada) Diskovery instrument, comprising a Leica DMi8 microscope operating in TIRF mode, equipped with a 63×/1.49 NA TIRF objective with a 1.8× camera relay (total magnification 108×). Images from samples were acquired using a Zyla 4.2-megapixel sCMOS camera using 488 nm, 561 nm or 627 nm laser illumination and 525/50, 620/60 or 700/75 emission filters. Fixed-cell TIRF-M imaging (Fig. 3, Fig. 6A) was done at room temperature with samples mounted in PBS. For live-cell imaging experiments (Fig. 6C), cells were maintained at constant 37°C during imaging, in phenol-free DMEM/F12 media (Gibco) supplemented with 20 mM HEPES and 20 ng/ml EGF, with imaging at rate of 1 frame per 15 s for 5 min. For the experiments shown in Fig. S1A–C, samples were imaged using the same instrument operating in spinning disc confocal model. Each image series was obtained as a z-series with a step size of 800 nm; subsequent analysis was performed on sum projections of these z-series.

Systematic, unbiased detection and analysis of either flotillin or clathrin diffraction-limited structures was performed as previously described (Cabral-Dias et al., 2022; Delos Santos et al., 2017; Lucarelli et al., 2017), using custom software developed in MATLAB (Mathworks Corporation, Natick, MA, USA), as described previously (Aguet et al., 2013). Briefly, diffraction-limited flotillin, clathrin, or rho–EGF structures were detected using a Gaussian-based model method to approximate the point-spread function of flotillin, AP2 or EGF objects (‘primary’ channel). The TIRF-M intensity corresponding to various proteins in a ‘secondary’ (or ‘tertiary’) channel (e.g. rho–EGF, AP2, flotillin or c-Src–eGFP) within each object detected in the ‘primary’ channel was determined by the amplitude of the Gaussian model for the appropriate fluorescence channel for each object position detected in the ‘primary’ channel. The detection of background overlap of secondary or tertiary channels with primary objects was determined by randomizing the position of secondary or tertiary channels relative to the primary channel, accomplished by rotation of the channels 180 degrees relative to one another. Statistical analysis was performed by two-tailed unpaired Student's t-test, with P<0.05 used as a threshold for establishing differences between experimental conditions. Software can be obtained as described previously (Lucarelli et al., 2017).

MDA-MB-231 and MDA-MB-231-HER2 cells were seeded on a 6-well plate and subjected to siRNA transfection as indicated (Figs 4, 7). For these experiments involving transfection, immediately following the second round of transfection, cells were placed in the Incucyte SX5 (Sartorius, Göttingen, Germany) and maintained at 37°C and 5% CO2 for detection and automated analysis of cell growth and proliferation, as we described previously (Lo et al., 2022). For experiments not involving transfection (Fig. 1D,E), 4 h after seeding, cells were incubated in medium containing 10 μM gefitinib (cat. #NU1042-005, Nacalai USS, Inc., San Diego, CA, USA) or vehicle control (DMSO, 0.1% v/v) and then immediately subjected to imaging as described above. The phase-contrast camera was used to take images every 30 min for 48 h. Time-lapse image series were analyzed using the Incucyte Basic Analysis software. To determine cellular confluence, the phase-contrast images were analyzed using the following settings: minimum area filter=25 μm², segmentation adjustment=0.8. Measurements were normalized to initial confluence in phase-contrast images. Statistical analysis was performed by two-way ANOVA followed by Šidák post hoc test, with P<0.05 used as a threshold for establishing differences between experimental conditions.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260133.reviewer-comments.pdf