Receptor Heterodimerization Modulates Endocytosis through Collaborative and Competitive Mechanisms

Abstract

Recruitment of receptors into clathrin-coated structures is essential to signal transduction and nutrient uptake. Among the many receptors involved in these processes, a significant fraction forms dimers. Dimerization of identical partners has generally been thought to promote receptor recruitment for uptake because of increased affinity of the dimer for the endocytic machinery. But what happens when receptors with substantially different affinities for the endocytic machinery come together to form a heterodimer? Evidence from diverse receptor classes, including G-protein-coupled receptors and receptor tyrosine kinases, suggests that heterodimerization with a strongly recruited receptor can drive significant recruitment of a receptor that lacks direct interactions with the endocytic machinery. However, a systematic biophysical understanding of this effect has yet to be established. Motivated by the potential of such events to influence cell signaling, here, we investigate the impact of receptor heterodimerization on endocytic recruitment using a family of engineered model receptors. As expected, we find that dimerization of a weakly recruited receptor with a strongly recruited receptor promotes incorporation of the weakly recruited receptor to endocytic structures. However, the effectiveness of this collaborative mechanism depends heavily on the relative strengths of endocytic recruitment of the two receptors that make up the dimer. Specifically, as the strength of endocytic recruitment of the weakly recruited receptor approaches that of the strongly recruited receptor, monomers of each receptor compete with heterodimers for space within endocytic structures. In this regime, the presence of the strongly recruited receptor drives a reduction in incorporation of the weakly recruited receptor into clathrin-coated structures. Similarly, as the strength of the dimer bond between the two receptors is progressively weakened, competition begins to dominate over collaboration. Collectively, these results demonstrate that the impact of receptor heterodimerization on endocytic recruitment is controlled by a delicate balance between collaborative and competitive mechanisms.

Significance

Receptor dimers play critical roles in regulating membrane traffic and signal transduction. Homodimers are thought to promote endocytic recruitment. In contrast, much less is known about how a heterodimer modulates endocytic recruitment. In moving toward filling this gap, our work reveals that the formation of heterodimers influences endocytic recruitment through a delicate balance between collaborative and competitive interactions. Specifically, weakly recruited receptors experience increased endocytic recruitment when they bind to a more strongly recruited receptor. However, the two receptor species may also compete for limited space within endocytic carriers, an effect which becomes more significant as the bond between them is weakened. Understanding the influence of heterodimers on endocytic recruitment has the potential to reveal novel mechanisms by which cells process extracellular signals.

Introduction

Uptake of receptors from the plasma membrane provides control over the timing and strength of biochemical signals as well as the cellular response to nutrients (1, 2) and growth factors (3, 4). Multiple endocytic pathways are responsible for receptor uptake, including caveolae-dependent endocytosis, CLIC/GEEC (clathrin-independent carriers/glycosylphosphatidylinositol-anchored proteins enriched early endosomal compartment)-type endocytosis, the IL2Rβ pathway, Arf6-dependent endocytosis, flotillin-dependent endocytosis, phagocytosis, macropinocytosis, and circular dorsal ruffles (5). However, the biochemical signals that drive receptor recruitment into endocytic structures are best understood in the context of clathrin-mediated endocytosis. Despite its involvement in trafficking processes from the endoplasmic reticulum to the Golgi and then to the plasma membrane (6), clathrin plays a major role in the budding of endocytic vesicles from the plasma membrane. In this pathway, also known as receptor-mediated endocytosis, receptor proteins use short cytosolic peptide motifs to bind to endocytic adaptor proteins, which in turn recruit other endocytic coat components including clathrin (7). Some receptors, such as transferrin receptor and low-density lipoprotein receptor, display endocytic recruitment motifs constitutively, whereas others, such as receptor tyrosine kinases (RTKs) and G-protein-coupled receptors (GPCRs), reveal these motifs principally upon activation by ligand binding (8, 9). Additionally, receptors such as transferrin receptor and GABA_A receptor γ2 subunit display a single biochemical motif for internalization, whereas other families of receptors such as the epidermal growth factor receptors (EGFRs) (7, 10) and other RTKs (7) display multiple internalization motifs.

Many receptors function as homodimers, which may be stable and ligand-independent or may form upon ligand binding (11, 12). Formation of a dimer results in the display of multiple endocytic uptake motifs, potentially increasing the strength of binding between receptors and the endocytic machinery. Specifically, Wang et al. (13) have reported that non-ligand-induced dimers of EGFRs are internalized much more strongly than monomers. Additional studies have demonstrated increased endocytic uptake of receptors upon higher-order oligomer formation. For example, Liu et al. (14) showed that clustering of transferrin receptors increases their overall uptake. Similarly, Cureton et al. (15) found that canine parvovirus gains cellular entry by binding to transferrin receptors that are locally concentrated within clathrin-coated endocytic structures. However, receptor heterodimers in which the two partners represent distinct receptor species are also common in some of the most widely studied receptor families, including GPCRs (16) and RTKs (17). Comparatively, little is known about the influence of heterodimers on receptor recruitment into endocytic structures.

Interestingly, heterodimerization events frequently bring together receptor monomers that have very different affinities for the endocytic machinery. For example, two GPCRs, the C5a anaphylatoxin receptor and CC chemokine receptor 5 have been observed to form heterodimers in the presence of C5a, a ligand for C5a anaphylatoxin receptor. Here, ligand binding activates C5a anaphylatoxin receptor, driving it to bind to endocytic structures through interactions with the cytosolic protein, β-arrestin. In contrast, CC chemokine receptor 5 remains inactivated and unable to bind to the endocytic machinery (18). Under these conditions, Hüttenrauch et al. (18) showed that CC chemokine receptor 5, despite its inability to bind directly to endocytic structures, is internalized in complex with C5a anaphylatoxin receptors. Jordan et al. (19) demonstrated a similar effect for another pair of heterodimer-forming GPCRs, opioid δ receptors and β₂-adrenergic receptors. Here, upon activation of opioid δ receptors by etorphine, β₂-adrenergic receptors were internalized, despite a lack of direct interaction with the endocytic machinery (19). Similarly, in the RTK family, EGFRs are strongly internalized by clathrin-mediated endocytosis, whereas ErbB2 is weakly internalized (20). The formation of heterodimers between these receptors is thought to modulate the uptake of both partners. Collectively, these observations suggest that dimerization of a receptor that binds strongly to the endocytic machinery with a receptor that interacts weakly or negligibly can result in substantial endocytic uptake of the weakly interacting receptor. However, no general biophysical understanding of the impact of receptor heterodimerization on endocytic recruitment currently exists.

In moving toward filling this gap, here, we study a family of model receptors with engineered affinity for clathrin-coated structures (CCSs) and for one another. These model receptors contain intracellular binding motifs for the major clathrin adaptor protein AP2 (21) and extracellular receptor dimerization motifs, both of which have variable affinities. Using this system, we demonstrate that when a receptor with low affinity for CCSs binds a receptor with high affinity for CCSs, localization of the weakly recruited receptor to CCSs increases dramatically. However, CCSs have a limited capacity to accommodate receptors, as illustrated by previous work from the J.S. lab (22) and others (23). This observation suggests that when multiple pools of actively recruited receptors are present at the plasma membrane, they will compete for internalization, as demonstrated by the work of Marks et al. (24). In line with this prediction, we find that competition for internalization limits the ability of receptor dimerization to promote endocytic recruitment. Specifically, competitive interactions begin to dominate over collaborative interactions when both receptors in heterodimers become strongly recruited to CCSs. Similarly, as the heterodimer bond becomes progressively weaker, monomers of the strongly recruited receptor become competitors for the weakly recruited receptor, reducing its incorporation in CCSs. Collectively, these results demonstrate that the impact of receptor heterodimerization on clathrin-mediated endocytosis depends on a delicate balance between the opposing influences of collaboration and competition.

Materials and Methods

Cell culture and transfection

Human retinal pigmented epithelial (RPE) cells (ARPE-19) were purchased from American Type Culture Collection (Manassas, VA) and were cultured in 1:1 F12/Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum (FBS), 20 mM HEPES (Sigma-Aldrich, St. Louis, MO), and 1% penicillin, 1% streptomycin, 1% L-glutamine. These cells were incubated at 37°C with 5% CO₂ and passaged every 48–72 h. For transfection, RPE cells were seeded onto acid-washed coverslips in six well plates (Corning, Corning, NY) at a density of 3 × 10⁴ cells per coverslip for 24 h before being transfected with 1 μg of plasmid DNA using 3 μL FuGENE transfection reagent (Promega, Madison, WI) for single receptor conditions and 1 μg of each plasmid DNA using 6 μL FuGENE transfection reagent for double receptor conditions.

Confocal fluorescence microscopy

A Zeiss Axio Observer Z1 spinning disk confocal microscope (Oberkochen, Germany) with Yokogawa CSU-X1M (Tokyo, Japan) was used to image live cells. Images were collected using a plan-apochromat 100× magnification, 1.4 NA oil immersion objective (Zeiss). Three laser wavelengths of 405, 488, and 561 nm were used for excitation. Three filters centered at 445 nm with a 45 nm width, 525 nm with a 50 nm width, and 629 nm with a 62 nm width were used for emission. A triple-pass dichroic mirror was used to reflect all three lasers (405/488/561 nm). The microscope was equipped with a cooled (−70°C) EMCCD iXon3 897 camera (Andor Technology, Belfast, United Kingdom).

Cells were imaged 16–20 h after transfection by acquiring spinning disk confocal images at the coverslip-proximal plasma membrane surface. Optical properties of the coverslip were matched to those of the immersion oil and chosen appropriately to avoid significant optical interference. To prevent excessive photobleaching, EC-Oxyrase (Oxyrase, Mansfield, Ohio) was added to the phenol-red-free imaging medium with 10% FBS to a final concentration of 0.9 oxyrase units/mL. Data were collected for each condition from at least three wells representing independent transfections. CCSs were detected using particle detection software that has been developed and made publicly available (25). To detect CCSs, a two-dimensional (2D) Gaussian function was fit to the fluorescence intensity profile of each putative punctum in the master channel in which the SD of the Gaussian function (σ) was derived from the point spread function of the microscope. In all experiment conditions, the BFP (clathrin) channel was used as the master channel for puncta detection. Each punctum needed to 1) be a diffraction-limited structure and 2) have an amplitude significantly above the local fluorescence background to be considered as valid. Based on the subpixel location of each punctum in the BFP (clathrin) channel, a 2D Gaussian function was fit to the corresponding punctum in the GFP and RFP channel to identify fluorescence puncta colocalized with each CCS. The centers of these Gaussian functions for colocalized punctum in the GFP and RFP channel were allowed to shift up to 1 SD away from the center of the Gaussian function in the master channel (BFP), equivalent to 146 nm.

Cells were selected based on their expression level of the helper or the competitor receptor such that all cells included in the analysis had a similar mean expression of the helper or the competitor receptor within their respective comparison groups. Further, we ensured that the fluorescence intensity from cells with the lowest receptor expression was significantly above the autofluorescence level. Statistically significant puncta from these selected cells were further filtered to only include CCSs that contained 3–20 pixels per detection. These limits were set to avoid including either nascent vesicles or large clathrin structures.

TIRF microscopy and fluorescence recovery after photobleaching

Total internal reflection fluorescence (TIRF) microscopy was used to image live cells, enabling quantification of plasma membrane intensity over time. An Olympus IX73 microscope body (Tokyo, Japan) was equipped with a Photometrics Evolve Delta EMCCD camera (Tucson, AZ) and Zeiss plan-apochromat 100× magnification, 1.46 NA oil immersion TIRF objective. Three laser wavelengths, 405, 473, and 532 nm, were used for excitation with a 635-nm laser used for autofocus. Photobleaching was performed with a 450-nm laser at ∼70 mW laser power for less than 10 ms. An emission filter centered around 475 nm with a 35 nm width was used to collect fluorescence emission from samples excited by the 405 nm laser. To exclude the signal from the photobleaching laser, an additional long-pass dichroic at 458 nm was used. An emission filter centered around 527 nm with a 70 nm width was used to collect fluorescence emission from samples excited by the 473 nm laser and a dual bandpass filter centered at 583 nm with a 37 nm width and 707 nm with a 51 nm width was used to collect fluorescence emission from samples excited by the 532 nm laser.

Videos of live cells were taken at the plasma membrane just above the surface of the coverslip 16–20 h after transfection. For fluorescence recovery after photobleaching (FRAP) measurements, images were collected at 3 s intervals for a total of 50 frames, and photobleaching was performed after 5 prebleached frames. To estimate the recovery of model receptors, we first tracked fluorescence intensities in the GFP channel over the entire video both within the FRAP area and on the surrounding plasma membrane. The intensities on the plasma membrane over time were used to correct for general photobleaching caused by the 473 nm imaging laser over time. We fitted a single exponential decay function to the intensities tracked on the plasma membrane and extrapolated the photobleaching-corrected intensities for each frame. Raw intensities within the FRAP area were subsequently normalized against these corrected intensities on the plasma membrane. We then fitted our normalized curves to the following exponential equation, $y\left(t\right)=A\left(1-e^{-\tau t}\right)$ to estimate the recovery half time from the fitted parameter τ with the following equation: $t_{\left(1/2\right)}=\left(ln0.5/-\tau \right)$.

Results

Binding to a strongly recruited receptor promotes endocytic recruitment of a weakly recruited receptor

To study the impact of heterodimerization on the endocytic recruitment of receptors, we constructed a system consisting of two transmembrane proteins, which we will refer to as the model receptor and the helper receptor. Here, the model receptor consisted of the intracellular and transmembrane domains of the transferrin receptor, followed by an extracellular membrane-proximal green fluorescent protein (GFP) domain, Fig. 1, A and B. The helper receptor also consisted of the intracellular and transmembrane domains of the transferrin receptor. However, its extracellular domain consisted of a membrane-proximal monomeric red fluorescent protein (RFP) domain followed by a single-domain antibody, which has nanomolar affinity for GFP (26) (Fig. 1 C). In this way, when the model and helper receptors are expressed simultaneously on the cell surface, their extracellular domains are able to bind to one another, creating a heterodimer.

Figure 1.

Binding to a strongly recruited receptor promotes endocytic recruitment of a weakly recruited receptor. (A–C) Cartoon schematics show the architecture of the YTRF model receptor (A), the CTRD model receptor (B), and the YTRF helper receptor (C). (D–F) Spinning disk confocal images show the plasma membrane of RPE cells stably expressing the clathrin light chain (CLC) tagged with BFP and transiently expressing the YTRF model receptor, M_YTRF (D), the CTRD model receptor, M_CTRD (E), and the YTRF helper receptor, H_YTRF (F). The box in the top image represents the location of the inset and the dashed circles in the insets highlight the diffraction-limited puncta colocalized between receptors and clathrin channels. It is important to note that these structures are much smaller than the circles drawn around them. (G) A cartoon schematic shows that M_CTRD and H_YTRF are incorporated into CCSs as heterodimers, leading to increased endocytic recruitment of M_CTRD. (H) As in (D)–(F), a confocal image shows M_CTRD and H_YTRF coexpressed in an RPE cell. All scale bars in example images, 5 μm; scale bars in insets, 1 μm. To see this figure in color, go online.

Importantly, transferrin receptors undergo ligand-independent constitutive endocytosis in a clathrin-dependent manner. In particular, the YTRF motif located in the cytoplasmic tail of transferrin receptor is recognized by the μ2 subunit of AP2, an essential adaptor protein in the clathrin-mediated internalization pathway (27, 28). Therefore, transferrin receptors containing the wild-type YTRF motif have a high probability of incorporation into CCSs for cellular uptake (2). Previous work has shown that the strength of binding between transferrin receptor and AP2 can be successively reduced through mutations to the internalization sequence, which reduce its hydrophobicity, YTRF > YARI > CTRF > CTRD (29, 30, 31, 32). We adopted this strategy, creating variants of our model and helper receptors, each of which displayed one of these internalization motifs.

After establishing this family of model and helper receptors, we first assessed their extent of colocalization with CCSs. Specifically, we expressed the YTRF model receptor, M_YTRF, in RPE cells. These cells are frequently used in studies of endocytosis because of their large, well-spread lamellipodia, enabling visualization of the plasma membrane (25). Further, our RPE cells stably expressed clathrin light chain (CLC) tagged with blue fluorescent protein (BFP). In these cells, BFP-labeled puncta appear at the plasma membrane surface, indicating the locations of CCSs. After transfection, we collected spinning disk confocal fluorescence images of the plasma membrane of these cells at the coverslip surface. The example image in Figs. 1 D and Document S1. Supporting Materials and Methods, Figs. S1–S11, and Table S1, Document S2. Article plus Supporting Material show that the YTRF model receptor appeared strongly colocalized with clathrin, suggesting that these receptors were efficiently recruited into CCSs. We next examined the endocytic recruitment of the most weakly recruited model receptor, M_CTRD, which displayed a CTRD motif. When the CTRD model receptor was expressed in BFP-CLC RPE cells, recruitment into CCSs appeared to be substantially reduced in comparison to recruitment of the YTRF model receptor, having a diffuse distribution over the plasma membrane surface with little enrichment at the sites of CCSs (Figs. 1 E and S1 B).

In contrast, the YTRF helper receptor, H_YTRF, had a punctate distribution that appeared well colocalized with CCSs (Figs. 1 F and S1 C), similar to the YTRF model receptor (Fig. 1 D). To determine the impact of the YTRF helper receptor on recruitment of the CTRD model receptor into CCSs, we transiently transfected both receptors into BFP-CLC RPE cells (Fig. 1 G). Here, the CTRD model receptor appeared strongly colocalized with the YTRF helper receptor as well as with CCSs, as will be quantitatively evaluated later in this work. The resulting punctate phenotype appeared similar to that of cells expressing the YTRF model receptor, suggesting that incorporation of the CTRD model receptor into CCSs was substantially increased in the presence of the YTRF helper receptor (Figs. 1 H and Document S1. Supporting Materials and Methods, Figs. S1–S11, and Table S1, Document S2. Article plus Supporting Material).

Equilibrium partitioning of receptors between CCSs and the plasma membrane

The data in Fig. 1 suggest that the presence of the helper receptor significantly increases the effective affinity of the model receptor for CCSs. To better understand this phenomenon, we next sought to quantify the effective increase in affinity by applying a thermodynamic model of receptor recruitment into CCSs (22). Briefly, our previous work developed and experimentally validated Eq. 1, which states that the average number of receptors per CCS, <n>, should depend on the maximal number of receptors that the structure can physically accommodate, N_max, the concentration of receptors on the surrounding plasma membrane, C_mem, and the 2D effective dissociation constant for binding between the receptor and the CCS, Kd_eff. Here, small Kd_eff indicates strong binding.

$$\langle n\rangle =\frac{N_{max}{C}_{mem}}{K_{d_{eff}}+C_{mem}}.$$ (1)

The effective dissociation constant, Kd_eff, can be estimated by fitting Eq. 1 to experimental data that measure <n> as a function of C_mem. These data can be extracted from analysis of images similar to those in Fig. 1, as demonstrated later in this work. However, before applying this analysis to our data, we must evaluate its key assumptions. First, Eq. 1 assumes that the partitioning of receptors between CCSs and the surrounding plasma membrane is determined by thermodynamic equilibrium. CCSs are highly dynamic structures, which rapidly initiate, assemble, and depart from the plasma membrane, all within a lifetime of 30–120 s (33). Therefore, receptors can only reach an equilibrium partitioning between CCSs and the surrounding plasma membrane if they diffuse in and out of CCSs on a timescale that is much shorter than the CCS lifetime. We have previously established that model receptors with the strongly recruited YTRF motif fill growing CCSs on the timescale of seconds or less (22). However, it is unclear whether the CTRD model receptor, which is much more weakly recruited, will be incorporated into CCSs on a similar timescale. To address this question, we used TIRF microscopy to track CCSs at physiological temperature, 37°C from initiation to departure in live RPE cells expressing mCherry-CLC and either the YTRF or CTRD model receptor, which are tagged with GFP. CCSs of varying lifetimes were binned into six lifetime cohorts, 12–19 s, 20–39 s, 40–59 s, 60–79 s, 80–99 s, and 100–120 s, as shown in Fig. 2, A and B. For both model receptors, the intensity in the receptor channel (GFP) tracked closely with the intensity in the clathrin channel (mCherry), demonstrating that receptors rapidly populate nascent CCSs as they mature, regardless of the affinity of these receptors for endocytic structures. This observation suggests that the timescale of receptor loading into CCSs is likely limited by the rate of receptor diffusion over the plasma membrane surface rather than by the kinetics of binding to CCSs.

Figure 2.

Equilibrium partitioning of receptors between clathrin-coated structures (CCSs) and the plasma membrane. (A and B) Lifetime cohorts of tracked CCS intensities in receptor (GFP; dashed line) and clathrin light-chain (CLC) (mCherry; solid line) channels are shown. 4500 CCSs from 15 total cells expressing M_YTRF were analyzed in (A), and 7754 CCSs from 16 total cells expressing M_CTRD were analyzed in (B). Notably, the green curves are noisier than the red curves, likely because of the lower signal/noise ratio in this channel. (C) The lifetime distributions of CCSs from cells expressing no model cargo, M_CTRD, or M_YTRF are shown. The average lifetime of all mCherry-CLC positive tracks across all lifetimes was 52 ± 2 s 95% confidence interval (CI) for cells with no cargo expression, 51 ± 1 s 95% CI for cells expressing M_CTRD, and 58 ± 1 s 95% CI for cells expressing M_YTRF (N = 1648, 7754, and 4500 CCSs, respectively). (D and E) Image sequences depict the recovery of the bleached plasma membrane for cells expressing either M_CTRD alone (D) or both M_CTRD and H_YTRF (E). The dashed line shows the edges of the photobleached area. GFP and RFP contrast settings are displayed in arbitrary units directly above each image sequence. Dashed circles in the BFP channel images show the recovered CCSs. All scale bars, 5 μm. (F and G) Normalized recovery curves are shown as mean (circles) ± standard error of the mean (shaded area). The recovery half time (t_1/2), extracted from the exponential fitting (black dashed line), is shown as mean with 95% CI. N = 7 and 9 cells for (F) and (G), respectively. To see this figure in color, go online.

Equation 1 also assumes that the presence of receptors does not alter the key physical properties of CCSs. This assumption was validated in our previous work in which we confirmed that expression of the model receptor does not substantially change the size, number, and dynamics of CCSs (22). This finding is also in agreement with a previous report, which suggested that the assembly of CCSs is largely independent of cargo levels (34). In line with these findings, data in Fig. 2 C demonstrate that expression of either the YTRF or CTRD model receptor shifted the average lifetimes by less than 15% in comparison to cells expressing no cargo. This shift may reflect a decrease in abortive CCSs, as previously reported (33).

Finally, the assembly of receptor heterodimers has the potential to slow down the diffusion of receptor species over the plasma membrane surface. This effect, if severe, could limit the ability of heterodimers to reach an equilibrium partitioning between CCSs and the surrounding plasma membrane. Therefore, we performed experiments to determine the impact of the helper receptor on the diffusion of the model receptor over the plasma membrane surface. To examine diffusion of the model receptor, we used FRAP experiments under TIRF illumination. We performed these experiments in RPE cells expressing BFP-CLC in combination with either 1) the CTRD model receptor (GFP tagged) or 2) the CTRD model receptor and the YTRF helper receptor (RFP tagged) (Fig. 2, D and E). Our data show that in the absence of the YTRF helper receptor, the CTRD model receptor had a recovery halftime of 28 ± 1 s (95% confidence interval (CI)) for an 80 μm² photobleached area (Fig. 2 F). In contrast, in the presence of the YTRF helper receptor, the CTRD model receptor had a 1.6-fold longer recovery halftime of 45 ± 1 s (95% CI) (Fig. 2 G). Assuming that a receptor dimer has approximately twice the membrane footprint of a monomer, this magnitude of increase in the diffusion time is in approximate agreement with Saffman-Delbrück theory, which predicts that the diffusion coefficient scales proportionally to the square root of the receptor size (35). Because recovery time is proportional to the square of diffusion distance, we estimate that the time required for diffusion across an individual CCS, which has an average diameter of 100 nm, would be 10–20 ms for both receptors, 100–1000 times smaller than the lifetime of a CCS. Therefore, the slight increase in diffusion time of the model receptor in the presence of the helper receptor is unlikely to have a strong impact on its ability to reach an equilibrium partitioning between a CCS and the surrounding plasma membrane.

Reduced endocytic recruitment of the helper receptor results in reduced recruitment of the weakly recruited model receptor

Having examined its key underlying assumptions, we next used Eq. 1 to evaluate the impact of receptor heterodimerization on endocytic recruitment of model receptors by CCSs. To compare Eq. 1 to our data, we examined CCSs in ∼100 cells for each condition. To estimate the fluorescence intensity of each CCS, we used publicly available software, cmeAnalysis (25). This algorithm identified diffraction-limited puncta in the receptor fluorescence channels (GFP or RFP), which colocalized with diffraction-limited puncta in the CLC (BFP) channel. A 2D Gaussian function was then fit to the fluorescence intensity profile of each punctum in the receptor channels, providing an estimate of its fluorescence emission. To convert these measurements to estimates of the number of individual GFP or RFP-labeled species per CCS, we divided the fluorescence intensity values by the calibrated fluorescence emission of individual GFP or RFP molecules, an approach we have previously validated (22, 36). Importantly, the intensity of GFP fluorescence scales linearly over a large range of fluorophore concentrations (Fig. S2). Thus, for each punctum, we estimated the number of fluorescent protein-tagged receptors that it represented, providing an estimate of <n>. Using the same approach, we quantified the density of receptors on the surrounding plasma membrane, providing an estimate of C_mem. The resulting values of <n> were plotted as a function of C_mem, as shown in Fig. 3 B, for the YTRF helper receptor and the CTRD model receptor, each expressed individually. Notably, the number of receptors per CCS increases approximately linearly with low receptor expression before leveling toward a maximal value representing the physical capacity of a CCS (N_max). Monomeric receptors such as the YTRF helper and the CTRD model receptor also approach saturation at high C_mem. Therefore, the saturation behavior of the curve does not imply formation of higher-order oligomers. However, the existence of such oligomers cannot be ruled out. Importantly, the range of C_mem presented in our data is comparable to the expression level of transferrin receptors in human fibroblast cells, ∼1% of the plasma membrane surface (37). As suggested by the example images in Fig. 1, our analysis showed that the YTRF helper receptor was recruited by CCSs in substantially higher copy number in comparison to the CTRD model receptor when each receptor was expressed individually (Fig. 3 B). N_max was fixed at 150 based on our prior study of a similar model receptor (22). The resulting best fit of Eq. 1 to the data for the YTRF helper receptor occurred for an effective 2D dissociation constant of 1200 ± 200 molecules per μm² (95% CI). In contrast, we found that the best fit value of Kd_eff for the CTRD model receptor expressed alone was substantially larger at ∼4500 ± 200 molecules per μm² (95% CI) as exemplified by the modest colocalization of this receptor with CCSs, as seen in Fig. 1 E.

Figure 3.

Reduced endocytic recruitment of the helper receptor results in reduced recruitment of the weakly recruited model receptor. (A) A cartoon schematic shows the CTRD model receptor bound to the YXXΦ mutants of the helper receptors. (B) <n> versus C_mem plot is shown. The average number of receptors per CCS is shown as a function of the receptor density on the surrounding plasma membrane. Each point is shown as the average of 250 puncta binned by receptor density on membrane surface ± standard error of the mean. Dashed lines indicate model predictions using Eq. 1. This plot shows 61 points representing 15,202 puncta for H_YTRF and 133 points representing 33,290 puncta for M_CTRD. (C) <n> versus C_mem plot in arbitrary units (AU) is shown. The average concentration of the helper receptor was held constant at ∼8000 AU per μm² (Fig. S7). This plot shows 61 points representing 15,202 puncta for H_YTRF, 78 points representing 19,426 puncta for M_CTRD + H_YTRF, 60 points representing 14,933 puncta for M_CTRD + H_YARI, 75 points representing 18,722 puncta for M_CTRD + H_CTRF, and 133 points representing 33,290 puncta for M_CTRD. (D and E) Spinning disk confocal images show the plasma membrane of BFP-CLC RPE cells transiently expressing M_CTRD with either H_YARI (D) or H_CTRF (E). Scale bars in example images, 5 μm; scale bars in insets, 1 μm. (F) Bar plot summarizing the effective dissociation constants is shown as the average, with error bars indicating the 95% CI. To see this figure in color, go online.

We next sought to quantify the impact of heterodimerization between the CTRD model receptor and the YTRF helper receptor on the endocytic recruitment of the CTRD model receptor. This experiment places GFP-tagged receptors in proximity with RFP-tagged receptors, creating the potential for fluorescence resonance energy transfer (FRET), an effect which could decrease the intensity in the GFP channel. Indeed, fluorescence lifetime imaging experiments revealed a decrease in the fluorescence lifetime of the GFP fluorophore of ∼20% when the GFP-tagged model receptor was coexpressed with an RFP-tagged helper receptor (Fig. S3). However, the extent of FRET was not a function of the colocalization with CCSs or the affinity between the model and helper receptors, suggesting that FRET arises from some process that is independent of endocytic recruitment, such as concentration of receptors in small fluid regions of the plasma membrane corralled by actin (38) or weak binding interactions between the fluorophore domains despite the presence of dimer-reducing mutations (see Supporting Materials and Methods). Regardless of its origin, the extent of FRET is comparable for receptors in CCSs and those on the plasma membrane surface. Therefore, to correct for the impact of FRET on our measurements, we divided both <n> and C_mem by 0.8. By correcting <n> and C_mem simultaneously, the slope of the curve changes relatively little in the low receptor concentration regime where the curve is mostly linear. As shown later in this work, the estimates of Kd_eff before and after the FRET correction are similar because Kd_eff values are derived from the linear portion of Eq. 1. Nonetheless, we plot the results of all two-receptor experiments in terms of arbitrary units to reflect the fact that the absolute number of receptors cannot be precisely estimated because of the potential impact of FRET.

Therefore, to estimate Kd_eff of the CTRD model receptor in the presence of the YTRF helper receptor, we examined the relative Kd_eff values derived from the slopes of <n> versus C_mem plots in arbitrary units for 1) the CTRD model receptor expressed alone, 2) the CTRD model receptor coexpressed with the YTRF helper receptor, and 3) the YTRF helper receptor expressed alone (Fig. 3 C). Here, the slope for the CTRD model receptor coexpressed with the YTRF helper receptor is nearly equivalent to that of the helper receptor expressed alone, indicating a substantial increase in the effective affinity of the CTRD model receptor. Multiplying Kd_eff of the CTRD model receptor from Fig. 3 B by the ratio of these Kd_eff values in arbitrary units, we arrive at an estimate of Kd_eff for the CTRD model receptor in the presence of the helper, ∼1400 ± 100 molecules per μm² (95% CI) (Figs. 3 F and S4; Table S1).

These results suggest that when heterodimers form between strongly and weakly recruited receptors, incorporation of the weakly recruited receptor to CCSs increases. To test this idea, we investigated the localization of the weakly recruited CTRD model receptor to CCSs in the presence of helper receptors with YARI and CTRF internalization motifs. These helper receptors are expected to have decreasing affinities for CCSs in comparison to the YTRF helper receptor but to still experience stronger endocytic recruitment than the CTRD model receptor. To verify these assumptions, we began by characterizing the binding affinities between CCSs and each of these helper receptors (Fig. S5, A–D). As expected, these mutant helper receptors decrease monotonically in affinity for CCS with decreasing hydrophobicity of the YXXΦ motif (YTRF > YARI > CTRF > CTRD) (29, 30, 31, 32). We then examined cells expressing both the CTRD model receptor and one of the mutant helper receptors (Fig. 3 A). Confocal images showed that with the presence of a helper receptor of intermediate affinity for CCSs, the YARI helper receptor, recruitment of the CTRD model receptor to CCSs increased substantially in comparison to recruitment in the absence of any helper receptor (Figs. 1 E, 3 D, S1 B, and S6 A). However, when the CTRD model receptor was coexpressed with a helper receptor with even lower affinity for CCSs, the CTRF helper receptor, the CTRD model receptor appeared to be distributed diffusely over the plasma membrane surface rather than being concentrated in CCSs (Figs. 3 E and S6 B). As reflected by these qualitative observations, endocytic recruitment of the CTRD model receptor to CCSs progressively decreased when coexpressed with the YARI and CTRF helper receptors (Figs. 3, C and F and S4; Table S1). Yet even the presence of the CTRF helper receptor slightly increased the recruitment of the CTRD model receptor to CCSs in comparison to its level of recruitment when expressed alone (Figs. 3, C and F and S4; Table S1). However, as the affinity of the model receptor for CCSs increases, increased competition for space within the CCS has the potential to change the relationship between the model and helper receptors. The next section explores this possibility.

Increased endocytic recruitment of the model receptor leads to competition with the helper receptor

Next, we varied the affinity of the model receptor and measured its incorporation into CCSs when coexpressed with a helper receptor of high affinity for CCSs, the YTRF helper receptor (Fig. 4 A). First, we considered the model receptor with a CTRF motif, which had a lower Kd_eff, ∼2500 ± 100 molecules per μm² (95% CI), in comparison to the CTRD model receptor considered in Fig. 3, 4500 ± 200 molecules per μm² (95% CI) (Fig. S5, F–H). When the CTRF model receptor was coexpressed with the YTRF helper receptor, the effective dissociation constant of the CTRF model receptor decreased by 48%, significantly smaller than the 69% decrease experienced by the CTRD model receptor upon coexpression with the same helper receptor at comparable expression level (Figs. 3 F, 4 B, S8; Table S1). This result indicates that increasing the strength of binding between the model receptor and CCSs decreases the impact of the helper receptor on its endocytic recruitment.

Figure 4.

Increased endocytic recruitment of the model receptor leads to competition with the helper receptor. (A) A cartoon schematic shows the YTRF helper receptor bound to the YXXΦ mutants of the model receptors. (B–D) <n> versus C_mem plot in arbitrary units (AU) as in Fig. 3C is shown. The estimated Kd_eff for (B) and (C) are summarized in inset bar plots with 95% CI. For (B) and (C), the average concentration of the helper or competitor receptor was held constant at ∼8000 AU per μm² (Fig. S8). For comparison, the H_YTRF curve is replotted here from Fig. 3C. The plot in (B) shows 61 points representing 15,202 puncta for M_YTRF, 71 points representing 17,713 puncta for M_CTRF + H_YTRF, and 144 points representing 36,099 puncta for M_CTRF. The plot in (C) shows 150 points representing 37,572 puncta for M_YARI, 93 points representing 23,347 puncta for M_YARI + H_YTRF, and 136 point representing 33,982 puncta for M_YARI + C_YTRF. The plot in (D) shows 150 points representing 37,572 puncta for M_YARI, 34 points representing 8566 puncta for M_YARI + H_YTRF – low expression, 31 points representing 7710 puncta for M_YARI + H_YTRF – medium expression, and 28 points representing 7071 puncta for M_YARI + H_YTRF – high expression. (E) Cartoon schematics show relative endocytic recruitment of M_YARI when 1) expressed alone, 2) with H_YTRF, or 3) with C_YTRF. (F) Cumulative <n> versus cumulative C_mem plot in AU is shown. Each point is shown as the average of 250 puncta binned by cumulative receptor density on membrane surface ± standard error of the mean. This plot shows 45 points representing 11,374 puncta for M_YARI + H_YTRF and 65 points representing 16,248 puncta for M_YARI + C_YTRF. The bar graph shows the average cumulative <n> in AU over the entire range of cumulative C_mem on the plot. Error bars represent standard error of the mean. Asterisks represent statistically significant differences in two-tailed t-test with p < 0.05. To see this figure in color, go online.

Next, we considered a model receptor with a YARI motif, which had a Kd_eff of ∼800 ± 50 molecules per μm² (95% CI), which is a considerably stronger affinity in comparison to the CTRD and CTRF model receptors (Fig. S5, E, G, and H). Interestingly, when the YTRF helper was coexpressed with the YARI model receptor, recruitment of the YARI model receptor became weaker, increasing Kd_eff by more than twofold (Fig. 4 C; Table S1). This reduction in endocytic recruitment is likely the result of competition between the model and helper receptors, which reduces the probability that model receptors will be recruited to CCSs. Specifically, because the YARI model receptor is recruited substantially to CCSs in the absence of the helper receptor, there is a greater potential for competition with helper receptor monomers. This interpretation implies that in cells expressing similar levels of the YARI model receptor, those with higher expression of the YTRF helper receptor are more likely to enter the competition regime, resulting in reduced endocytic recruitment of the YARI model receptor. To evaluate this prediction, we divided our data into three groups with high (9:1 on average), medium (4:1 on average), and low (2:1 on average) stoichiometric ratios between the YTRF helper and the YARI model receptors (Fig. 4 D). As expected, the competition experienced by the YARI model receptor is a function of these stoichiometric ratios. Precisely, the YARI model receptor experienced the most competition in cells expressing the highest stochiometric ratio, resulting in a 1.5-fold and twofold reduction in endocytic recruitment of the YARI model receptor in comparison to the group of cells expressing the medium and the low stoichiometric ratios, respectively.

If competition with the YTRF helper receptor is responsible for reduced recruitment of the YARI model receptor to CCSs, then removal of the nanobody domain from the helper receptor would be expected to further decrease endocytic recruitment of the YARI model receptor. This decrease is expected because it eliminates the ability of the YTRF receptor to act as a helper, such that the YTRF and YARI receptors can only interact competitively. To test this idea, we coexpressed the YARI model receptor with a version of the YTRF helper receptor that lacked the nanobody domain, the YTRF competitor receptor, C_YTRF (Fig. 4 E). Similar to the YTRF helper receptor, the competitor receptor consists of the intracellular and transmembrane domains of the transferrin receptor. However, its extracellular domain only consists of an RFP. As expected, the recruitment of the YARI model receptor into CCSs was significantly inhibited in the presence of the competitor receptor, resulting in a 4.5-fold increase in its effective dissociation constant (Fig. 4 C). Interestingly, the cumulative recruitment to CCSs, the sum of recruitment of both the YARI model receptor and either the YTRF helper or competitor receptor, also slightly decreased when the competitor receptor was present rather than the helper receptor (Fig. 4 F). This result suggests that receptor-receptor binding interactions increase overall receptor recruitment to CCSs, in line with prior findings (14, 15).

Collectively, these results illustrate that the presence of a helper receptor can either increase or decrease endocytic recruitment of the model receptor. In particular, as the affinity of the model receptor for CCSs increases, competition with the helper receptor eventually outweighs the potential of the helper receptor to promote recruitment of the model receptor to CCSs. Specifically, this competition increases with increasing stochiometric ratio of helper to model receptors. Similarly, we would expect that as the binding affinity between the model and helper receptors decreases, competition between the receptors will outweigh their ability to collaborate. This prediction is tested in the next section.

Reduced binding between the model and helper receptors reduces endocytic recruitment of the weakly recruited model receptor

The binding interaction between GFP and GFP nanobody, which drives formation of receptor heterodimers in our model system, is relatively strong, having a dissociation constant of ∼1 nM in solution (26). In contrast, many binding interactions between native membrane proteins are thought to be relatively weak. For example, GPCRs are thought to form weak, transient oligomers at the plasma membrane surface (39). Therefore, to better mimic weak binding interactions between native receptors, we created GFP nanobody mutants with reduced binding affinity for the GFP-tagged model receptor. Specifically, we used the structure of the GFP-GFPnb complex to identify two amino acids on the GFP nanobody, Arg35, and Glu103, which are expected to form essential salt bridges with Glu142 and Arg168 of GFP, respectively (26). To implement this approach, we created mutant versions of the YTRF helper receptor, which contained either a single R35A mutation or both R35A and E103A mutations (Fig. 5 A). These mutant helper receptors are expected to have successively reduced affinity for the model receptor. However, because their YTRF motifs are unmodified, their strong recruitment into CCSs should remain intact.

Figure 5.

Reduced binding between the model and helper receptors reduces endocytic recruitment of the weakly recruited model receptor. (A) A cartoon schematic shows M_CTRD bound to H_YTRF. The zoomed-in image shows the key amino acids in the binding interface between GFP nanobody and GFP. (B) Cartoon schematics show the relative endocytic recruitment of M_CTRD in the presence of H_YTRF with mutations on the GFP nanobody domain, either R35A (top) or R35A and E103A (bottom). (C and D) Spinning disk confocal images show the plasma membrane of BFP-CLC RPE cells transiently expressing M_CTRD with either H_YTRF R35A (C) or H_YTRF R35A E103A (D). Scale bars in example images, 5 μm; scale bars in insets, 1 μm. (E and F) <n> versus C_mem plot in arbitrary units (AU) as in Fig. 3C is shown. The average concentration of the helper or competitor receptor was held constant at ∼5500 AU per μm² (Fig. S11). The best fit Kd_eff values are summarized in (F) with 95% CI. For comparison, curves for M_CTRD and M_CTRD + H_YTRF are replotted here from Fig. 3C. This plot shows 74 points representing 18,579 puncta for M_CTRD + H_YTRF, 114 points presenting 28,495 puncta for M_CTRD + H_YTRF R35A, 143 points representing 35,764 puncta for M_CTRD + H_YTRF R35A E103A, 133 points representing 33,290 puncta for M_CTRD, and 151 points representing 37,679 puncta for M_CTRD + C_YTRF. (G) Cumulative <n> versus cumulative C_mem plot in AU as in Fig. 4F is shown. This plot shows 62 points representing 15,503 puncta for M_CTRD + H_YTRF, 104 points representing 26,116 puncta for M_CTRD + H_YTRF R35A, 108 points representing 26,976 puncta for M_CTRD + H_YTRF R35A E103A, and 90 points representing 22,660 puncta for M_CTRD + C_YTRF. The bar graph shows the average cumulative <n> in AU over the entire range of cumulative C_mem on the plot. Error bars represent standard error of the mean. Asterisks represent statistically significant differences in two-tailed t-test with p < 0.05. To see this figure in color, go online.

Although the three-dimensional affinities between GFP and these mutant versions of the nanobody were not measured, the studies in this section measure the impact of these mutations on the ability of the helper receptor to recruit the CTRD model receptor to CCSs. As detailed below, these studies provide estimates of the impact of the mutations on the 2D binding affinities between GFP and the nanobody. To investigate the impact of these mutations on the recruitment of the model receptor, we cotransfected each of the nanobody mutant YTRF helper receptors with the weakly recruited CTRD model receptor (Fig. 5 B). When cotransfected with the single mutant YTRF helper receptor, the CTRD model receptor was well colocalized with puncta in the clathrin channel, indicating significant recruitment into CCSs (Figs. 5 C and S9A). However, when cotransfected with the double mutant YTRF helper receptor, the CTRD model receptor was distributed much more diffusely over the plasma membrane surface, indicating significantly less recruitment into CCSs (Figs. 5 D and S9 B). Here, the double mutant YTRF helper receptor remained strongly colocalized with clathrin, suggesting that its affinity for CCSs remained high, despite its reduced binding affinity to the model receptor, as expected (Fig. S10).

We compared incorporation of the CTRD model receptor into CCSs for four cases: 1) CTRD model receptor coexpressed with the unmodified YTRF helper receptor (repeated from Fig. 3 C), 2) CTRD model receptor coexpressed with the single mutant YTRF helper receptor, 3) CTRD model receptor coexpressed with the double mutant YTRF helper receptor, and 4) CTRD model receptor expressed alone (repeated from Fig. 3 C). These data reveal that coexpression with the single mutant helper receptor substantially increased recruitment of the CTRD model receptor to CCSs, decreasing its effective dissociation constant by 64%. The resulting recruitment of the CTRD model receptor into the CCS was only slightly less than it was for coexpression with the unmodified helper receptor, as shown in Fig. 5, E and F and Table S1. However, coexpression with the double mutant helper receptor resulted in a more modest increase in endocytic recruitment of the CTRD model receptor, decreasing its effective dissociation constant by only 18% (Fig. 5, E and F and Table S1).

As the strength of the bond between the helper and model receptors decreases, we would expect the YTRF helper receptor to be recruited into CCSs increasingly as a monomer rather than as a heterodimer, increasing the potential for competition with the model receptor. In the limit of no binding affinity between the two receptors, the CTRD model receptor would face pure competition from the YTRF “helper” receptor. As expected, coexpression of the YTRF competitor with the CTRD model receptor resulted in levels of CTRD model receptor recruitment to CCSs that were even lower than when the CTRD model receptor was expressed alone, driving a 51% increase in its dissociation constant, as seen in Fig. 5, E and F and Table S1.

In Fig. 5 G, we analyzed the cumulative receptor recruitment to CCSs for the same combinations reported in Fig. 5 E. Our results showed that the CTRD model receptor expressed with the unmodified YTRF helper resulted in the highest total number of receptors per CCS, whereas the CTRD model receptor with the YTRF competitor resulted in the lowest. As expected, coexpression of the model receptor with the mutant helpers drove intermediate levels of cumulative recruitment to CCSs. These results further support the idea that assembly of heterodimers enhances cumulative endocytic recruitment of receptors, in agreement with the findings from Fig. 4 F and the literature (14, 15). More broadly, these results illustrate that endocytic recruitment of individual receptor species depends not only upon their affinity for CCSs but also upon the strength of their interactions with other receptors that may also be targets of internalization.

Discussion

Multiple studies have highlighted the significant physiological consequences of heterodimerization between transmembrane receptors, including GPCRs (40, 41) and RTKs (42). Specifically, several studies have reported that receptors that lack a strong biochemical recruitment motif for the endocytic machinery can still be recruited by growing endocytic vesicles when they form a dimer with a second receptor species that is being actively recruited for internalization (18, 19, 20). We observed a similar mechanism within our system in which model receptors that were weakly recruited to CCSs as monomers became efficiently recruited for clathrin-mediated internalization when they formed heterodimers with strongly recruited helper receptors. This collaborative mechanism could provide cells with an indirect route for controlling receptor-mediated signaling events. Specifically, uptake of a strongly internalized receptor may impact signaling events that are associated with receptors that lack direct binding interactions with the endocytic machinery. For example, serotonin receptor 5-HT_1A, a weakly internalized GPCR, forms stable heterodimers with a strongly internalized GPCR, 5-HT₇. Heterodimerization of these receptors in the presence of serotonin drives internalization of 5-HT_1A, resulting in phosphorylation of its downstream signaling partner, Erk. The level of Erk activation by internalized 5-HT_1A was found to be significantly enhanced in the presence of 5-HT₇, demonstrating a critical role for heterodimerization in the cellular response to serotonin (43).

In addition to these collaborative effects, we also found that the presence of a strongly recruited helper receptor can negatively impact incorporation of a weakly recruited model receptor to CCSs, owing to competition between the two receptor species for limited space within endocytic structures (Figs. 4, C and D and 5 E). Competitive interactions between receptors during cellular uptake were first explored by Marks et al. (24), who reported that the internalization of TTMb, a chimeric transmembrane receptor, is substantially reduced by overexpression of receptors that compete for the same interactions with coated vesicle adaptor proteins. More recently, several groups have demonstrated that receptors with bulky extracellular domains can saturate endocytic structures. For example, Mettlen et al. (44) have shown that overexpression of the bulky low-density lipoprotein receptor presents an obstacle to internalization. Similarly, cargo molecules with bulky intrinsically disordered domains are internalized less efficiently in comparison to small cargoes (36). To our knowledge, the current study is the first to characterize the combined effects of collaborative and competitive influences on endocytic recruitment of receptors.

The main result of this study is that the impact of heterodimerization on incorporation of a receptor into CCSs depends on a balance between competitive and collaborative effects. Although our study utilizes model receptors rather than physiological receptors, this basic physical principle may apply to many receptor-receptor interactions in cells. Specifically, when the model receptor has little or no ability to bind to endocytic structures on its own, its endocytic recruitment should increase in the presence of a more strongly recruited helper receptor (Fig. 3 C). However, as the affinity of the weakly recruited model receptor approaches that of the strongly recruited helper receptor, the balance is expected to shift toward competition between them, eventually resulting in decreased incorporation of the model receptor to CCSs in the presence of the helper receptor (Fig. 4 C). Similarly, the competitive effect increases in relative importance with increasing stochiometric ratio of helper to model receptors (Fig. 4 D) and decreasing strength of the heterodimer bond (Fig. 5 E).

Taken together, our results provide a quantitative perspective on the impact of receptor heterodimerization on the localization of receptors to CCSs. By demonstrating that receptors can interact through both collaborative and competitive mechanisms, this work points to an inherent coupling between receptor interactions, membrane traffic, and cell signaling. Moving forward, the findings from this study will be tested using specific families of physiological receptors to better understand routes by which cells receive and process signals that originate at the plasma membrane.

Author Contributions

C.Z. designed experiments, performed experiments, analyzed data, and wrote the manuscript. A.C.M.D., H.A.A., and M.F.L performed experiments and analyzed data. C.C.H. and J.R.H. designed experiments and analyzed data. J.C.S. designed experiments, supervised the research team, and wrote the manuscript. All authors consulted together on the interpretation of the data and helped to refine the manuscript.

Editor: Kalina Hristova.