Cytokine receptor cluster size impacts its endocytosis and signaling

Significance

The IL-2R is a key receptor of the immune system, promoting both immunity and tolerance through activation of signaling cascades on lymphocytes. This receptor induces responses that include T cell expansion, development and suppression of autoreactive lymphocytes, being critical in the prevention of autoimmune diseases. These functions are regulated by the internalization and intracellular trafficking of IL-2R, which undergoes a coat-independent mechanism of endocytosis. Our work shows that clustering of IL-2R at the cell surface is fundamental in the efficient internalization of the receptor and that an optimal cluster size is required for this. Disorganization of IL-2R clustering, through perturbation of the actin cytoskeleton or membrane cholesterol, affects endocytosis and modifies IL-2–induced signaling.

Abstract

The interleukin-2 receptor (IL-2R) is a cytokine receptor essential for immunity that transduces proliferative signals regulated by its uptake and degradation. IL-2R is a well-known marker of clathrin-independent endocytosis (CIE), a process devoid of any coat protein, raising the question of how the CIE vesicle is generated. Here, we investigated the impact of IL-2Rγ clustering in its endocytosis. Combining total internal reflection fluorescence (TIRF) live imaging of a CRISPR-edited T cell line endogenously expressing IL-2Rγ tagged with green fluorescent protein (GFP), with multichannel imaging, single-molecule tracking, and quantitative analysis, we were able to decipher IL-2Rγ stoichiometry at the plasma membrane in real time. We identified three distinct IL-2Rγ cluster populations. IL-2Rγ is secreted to the cell surface as a preassembled small cluster of three molecules maximum, rapidly diffusing at the plasma membrane. A medium-sized cluster composed of four to six molecules is key for IL-2R internalization and is promoted by interleukin 2 (IL-2) binding, while larger clusters (more than six molecules) are static and inefficiently internalized. Moreover, we identified membrane cholesterol and the branched actin cytoskeleton as key regulators of IL-2Rγ clustering and IL-2–induced signaling. Both cholesterol depletion and Arp2/3 inhibition lead to the assembly of large IL-2Rγ clusters, arising from the stochastic interaction of receptor molecules in close correlation with their enhanced lateral diffusion at the membrane, thus resulting in a default in IL-2R endocytosis. Despite similar clustering outcomes, while cholesterol depletion leads to a sustained IL-2–dependent signaling, Arp2/3 inhibition prevents signal initiation. Taken together, our results reveal the importance of cytokine receptor clustering for CIE initiation and signal transduction.

Many proteins are part of multimeric complexes and oligomerize to achieve their function. Clustering of membrane proteins plays critical roles in a variety of cellular events, such as cell homeostasis, signal transduction, protein secretion, or endocytosis. In the latter mechanism, it has been shown that local protein crowding can induce membrane conformational changes, promoting its curvature and hence, participating in the initiation of endocytic pits (1). Several mechanisms of endocytosis exist, including clathrin-mediated endocytosis (CME) and other pathways that are clathrin-independent endocytosis (CIE). In both, various studies have shown the role of the clathrin coat, endocytic adaptors, Bin–Amphiphysin–Rvs (BAR) domain proteins, lipids, and actin on membrane curvature, thus promoting pit formation (2). However, few works investigated the impact of cargo oligomerization on this initial step. It has been previously reported that local clustering of transferrin receptor (TfR) promotes clathrin-coated pit initiation (3), and more recently, in vitro studies further show that cargo crowding can induce membrane bending in CME (1, 4, 5). In the absence of a coat protein, like in most CIE pathways, unraveling how pits are initiated has become a priority but is far from being understood. Interestingly, several CIE cargoes are clustered by extracellular glycolipids or lectins, thereby inducing their invagination and efficient uptake (6, 7). Signaling receptors constitute another family of cargoes for which oligomerization might play a role in their function and trafficking (8). Recently, growing numbers of receptors of this family have been reported to use CIE pathways (9, 10). Cytokine receptors are such examples, being critical in both tolerance and immunity (11).

The interleukin-2 receptor (IL-2R) comprises three chains, IL-2Rα, IL-2Rβ, and IL-2Rγ, which in the presence of the ligand, interleukin-2 (IL-2), heterooligomerize and transduce a signal cascade (Janus kinase [JAK]-signal transducer and activator of transcription [STAT], Mitogen-activated protein kinase [MAPK], Phosphoinositide 3-kinase [PI3K]), leading to T cell proliferation (12). IL-2Rβ and IL-2Rγ are internalized independently of clathrin, either constitutively or ligand induced, and both chains are sorted toward the lysosome for degradation, thereby terminating the IL-2 signal transduction (13). Our group has previously shown that the IL-2R is associated with lipid microdomains rich in cholesterol, termed lipid rafts (14, 15). This CIE mechanism strongly relies on actin, dynamin, and their partners (endophilin, Ras-related C3 botulinum toxin substrate 1 [Rac1], P21 Activated Kinase 1/2 [Pak1/2], cortactin, Neural Wiskott-Aldrich syndrome protein [N-WASP], WAVE-family verprolin-homologous protein [WAVE]) (9, 15–19). Moreover, we have shown that IL-2R is mainly found at the base of membrane protrusions where the endocytic vesicle is initiated (20). This led to our current hypothesis that clustering of IL-2R might induce and/or stabilize the pit that would then mature into a vesicle. In the present study, we used gene editing and live imaging of lymphocytes coupled to a single-molecule analysis to determine the stoichiometry of IL-2R, in its native state, during its endocytosis. Further, we unraveled two key factors, F-actin and cholesterol, that regulate cytokine receptor clustering, endocytosis, and IL-2–mediated signaling.

Results

Generation of Genome-Edited Green Fluorescent Protein (GFP)-IL-2Rγ T Cells.

To study the stoichiometry of IL-2Rγ at the surface of T cells, we generated GFP-IL-2Rγ genome-edited cells using CRISPR-Cas9 technology to edit the genome of Kit225 cells, a human T cell line naturally expressing the three chains of IL-2R. We inserted enhanced green fluorescent protein (eGFP) at the 5′ end of IL2RG, after the signal peptide, thus obtaining an N-terminal fusion protein, GFP-IL-2Rγ (SI Appendix, Fig. S1A). Genome-edited cells provide several advantages by preserving the endogenous expression, localization, and thus, stoichiometry of the protein of interest, which are critical for single-molecule analysis.

Edited single cell–derived clones were acquired by cell sorting, and the first exon of IL2RG, where eGFP was introduced, was analyzed by RT-PCR (SI Appendix, Fig. S1B). Fully edited clones were subsequently analyzed by western blot (SI Appendix, Fig. S1C) and finally sequenced. We selected a clone (clone 3) that was successfully edited at both genomic and protein levels to pursue our investigations. First, we verified that GFP-IL-2Rγ was expressed at the surface of edited cells by performing a surface biotinylation experiment (SI Appendix, Fig. S1D). Then, by flow cytometry, we confirmed that edited cells showed similar expression levels (total and surface) of IL-2Rγ, as well as the two other IL-2R chains (β and α), compared with nonedited Kit225 cells (SI Appendix, Fig. S1E).

To ensure that addition of eGFP did not impact the interaction between IL-2R and its ligand, we first confirmed that incubating GFP-IL-2Rγ edited cells with IL-2 promoted the heterooligomerization of the γ- and β-chains, as seen by coimmunoprecipitation experiments (SI Appendix, Fig. S1F) (21). We then verified that IL-2 induced a faster rate (1.3-fold) of IL-2Rβ and IL-2Rγ endocytosis in both Kit225 nonedited and edited cells, through internalization kinetics experiments measured by flow cytometry (SI Appendix, Fig. S1G) (22). Finally, we confirmed by western blot that IL-2 induced the phosphorylation of tyrosine residues in both nonedited and edited cells, indicating an efficient initiation of the signaling cascade (SI Appendix, Fig. S1H) (12). Altogether, these results show that gene edition of IL-2Rγ did not impair the native function of the receptor and validate the Kit225 GFP-IL-2Rγ cell line as a suitable model to analyze the dynamics and stoichiometry of IL-2Rγ.

GFP-IL-2Rγ Is Organized in Distinct Clusters at the Cell Surface.

To determine the stoichiometry of GFP-IL-2Rγ molecules at endocytic sites in live cells, we used total internal reflection fluorescence (TIRF) microscopy, allowing observation of events occurring at the plasma membrane (PM) with improved axial resolution (Fig. 1 A, i and SI Appendix, Fig. S1A show representative TIRF images of GFP-IL-2Rγ; Movies S1–S15). The dynamics of receptor endocytosis were studied using the Icy platform with the tracking plugin eTrack (19), which relies on the high spatial confinement of putative endocytic sites to resolve their positions over time and accurately obtain the duration and fluorescence intensity of each track (Fig. 1 A, ii) (23). We first calibrated our imaging system for the fluorescence of single eGFP molecules, as previously described (23–25). Briefly, coverslips were coated with purified eGFP molecules at very low density and subsequently imaged at high frequency (20 Hz). Analysis of the intensity profiles of eGFP spots over time showed that the majority of these bleach in an irreversible single-step manner (Fig. 1 A, iii), with the bleaching step corresponding to the intensity of a single eGFP molecule. Kit225 GFP-IL-2Rγ cells were imaged, with or without IL-2, using the same settings, except for acquisition frequency that was set at 1 Hz, and we calculated the number of GFP-IL-2Rγ molecules per track using the preestablished intensity standard of one molecule of eGFP (the complete workflow is represented in Fig. 1A) (a detailed protocol is in ref. 23).

Fig. 1.

GFP-IL-2Rγ is clustered at the cell surface in distinct subpopulations. (A) Schematic representation of the endocytic track detection and the single-molecule calibration technique used to determine the stoichiometry of GFP-IL-2Rγ at the cell surface. (A, i) Live Kit225 edited cells are imaged by TIRF microscopy. (A, ii) GFP-IL-2Rγ putative endocytic tracks are automatically detected, and their fluorescence intensity along time is obtained using the eTrack plugin. (A, iii) A calibration step is performed by imaging purified eGFP at high frequency and determining the fluorescence intensity corresponding to one molecule of eGFP. Finally, the mean of several eGFP bleaching steps is used to calculate the number of GFP-IL-2Rγ molecules per track. (B) Violin plots (Left) and histograms (Right) of the number of GFP-IL-2Rγ molecules per spot at steady state in control and BFA-treated cells, IL-2 starved (−IL-2) or IL-2 stimulated (+IL-2). Histograms represent the distribution of the number of GFP-IL-2Rγ molecules per spot in each condition (mean ± SD; −IL-2, n = 2,805 spots; +IL-2, n = 3,027 spots; BFA −IL-2, n = 3,012 spots; BFA +IL-2, n = 3,596 spots; data obtained from three independent experiments). (C) Distribution of GFP-IL-2Rγ track’s lifetimes fitted with a mixture model comprising one exponential distribution (Exp; red line) and three Gaussians with the following means (± SD): G1 (blue line), 29.1 ± 4.8 s; G2 (green line), 43.9 ± 9.8 s; G3 (yellow line), 149.6 ± 0.6 s. The red dotted line corresponds to the total distribution. (D) Representative images of the modified FRAP experiment (Upper, scale bars: 5 μm) and histograms of the duration of GFP-IL-2Rγ tracks reappearing at the surface of control or BFA-treated cells after photobleaching (Lower). Cells were imaged by TIRF microscopy and photobleached with 100% laser power for 30 s. Following 2 min of recovery, live-imaging movies were acquired normally (1% laser power, 1 frame/s). Data are from two independent experiments, with a total of 535 tracks in control cells and 120 tracks in cells treated with BFA. (E) Distribution of the duration of GFP-IL-2Rγ tracks (noncolocalizing) and GFP-IL-2Rγ/IL-2Rβ colocalizing tracks, represented as a relative frequency of total GFP-IL-2Rγ tracks (mean ± SD; GFP-IL-2Rγ, n = 629 tracks; GFP-IL-2Rγ +IL-2Rβ, n = 229 tracks; data from two independent experiments). (F) Violin plot representing the number of GFP-IL-2Rγ per track, according to their lifetimes. Short-lived tracks have durations of 1 to 25 s (n = 2,419 tracks), long-lived tracks have durations of 30 to 80 s (n = 1,695 tracks), and static tracks have durations of 140 to 150 s (n = 40 tracks).

First, we investigated the number of receptor molecules at steady state by analyzing GFP-IL-2Rγ spots present in the first frame of live-imaging movies. IL-2Rγ was never observed as a monomer, but rather in oligomers containing 2 to 11 molecules (Fig. 1B). The majority of the receptors formed clusters of three to five molecules in the absence of IL-2, yet stimulation with IL-2 shifted the distribution of GFP-IL-2Rγ toward larger clusters with four to six molecules (mean of four molecules −IL-2 and five molecules +IL-2) (Fig. 1B). These results indicate that IL-2R is organized in different cluster populations and suggest that the ligand has an impact on GFP-IL-2Rγ stoichiometry at the PM, perhaps by increasing the proportion of IL-2R ready to be internalized.

At steady state, the total number of cell surface IL-2Rγ results from the balance between receptor biosynthesis (newly secreted population) and endocytosis. To gain a better insight on the endocytic population, we used Brefeldin A (BFA) to inhibit Golgi-mediated protein secretion (26) and thus, exclude GFP-IL-2Rγ being secreted to the PM. Incubation with BFA for 30 min, both in the absence or presence of IL-2, resulted in GFP-IL-2Rγ tracks with a higher number of molecules (Fig. 1B), without significantly altering IL-2R internalization (SI Appendix, Fig. S2B). Importantly, BFA treatment led to a strong reduction (two- to fivefold) of the population of tracks with two to three GFP-IL-2Rγ molecules, as compared with nontreated cells (Fig. 1B). This result indicates that the newly secreted subpopulation of receptors contains up to three molecules and that the endocytic population of GFP-IL-2Rγ corresponds to larger clusters than the secreted ones. This analysis at steady state reveals that IL-2Rγ exists in oligomers at the PM and might be assembled in small clusters during its secretion.

Dynamics of GFP-IL-2Rγ Clusters at the PM.

To investigate a correlation between the oligomerization state of GFP-IL-2Rγ and its endocytosis, we used eTrack to determine the stoichiometry of the receptor (SI Appendix, Fig. S2A) according to its lifetime at the PM (Fig. 1C). We classified the different subpopulations by fitting the distribution of GFP-IL-2Rγ tracks lifetimes with a mixture model defined by the sum of an exponential distribution and three Gaussians (G1, G2, and G3), as previously described (19). First, the exponential distribution represents short-lived tracks (up to 25 s) and accounts for ∼50% of all tracks. These short-lived tracks might represent aberrant detections or newly secreted GFP-IL-2Rγ that is rapidly diffusing at the PM, thus representing nonproductive endocytic tracks (19, 27). Second, the Gaussian distributions found most likely represent endocytic subpopulations with different internalization rates (G1 with 29.1 ± 4.8 s and G2 with 43.9 ± 9.8 s), while G3, lasting the entire length of the movie (149.6 ± 0.6 s), would represent nonterminal static events (Fig. 1C).

To demonstrate that the observed short-lived tracks represent the newly secreted population of receptor, we treated cells with BFA and performed a fluorescence recovery after photobleaching (FRAP) experiment by photobleaching the surface receptors and imaging the newly reappearing tracks after 2 min of recovery (Fig. 1D). The majority of new GFP-IL-2Rγ tracks arriving at the cell surface had lifetimes up to 25 s (56% of tracks). In contrast, these short-lived tracks were nearly eliminated in BFA-treated cells (Fig. 1D), suggesting that the short-lived tracks (0 to 25 s) account for the newly secreted population of GFP-IL-2Rγ arriving at the PM. To better define the putative endocytic subpopulation, we analyzed GFP-IL-2Rγ tracks colocalizing with the β-chain of the receptor prelabeled with an antibody recognizing the extracellular part of IL-2Rβ. These colocalized tracks should represent the active endocytic receptor since IL-2 induces the heterooligomerization of the two chains, as well as a more efficient receptor uptake (12). By overlaying the lifetime distributions of GFP-IL-2Rγ tracks, colocalized or not with IL-2Rβ, we observed multiple distribution peaks between 35 and 80 s, mostly present in colocalized tracks (Fig. 1E). Therefore, we presumed these longer-lived tracks (30 to 80 s) to represent the endocytic active GFP-IL-2Rγ population. Altogether, these results allowed us to identify three subpopulations of GFP-IL-2Rγ: a short-lived population, with durations of 0 to 25 s, corresponding to the receptor that is newly secreted and/or diffusing at the cell surface; a long-lived population, with durations of 30 to 80 s, corresponding to the endocytic competent receptor; and finally, a static population, with durations of 140 to 150 s, corresponding to nonproductive endocytic events.

Next, we examined the cluster size of GFP-IL-2Rγ tracks with respect to their duration: short lived, long lived, and static. We observed that the number of molecules per cluster increased with GFP-IL-2Rγ tracks’ lifetimes: a mean of 3 molecules per cluster for short-lived tracks, 4 molecules for long-lived tracks, and for static tracks, we observed a spread distribution with a greater number of molecules (from 4 up to 11 molecules) (Fig. 1F). In agreement with this distribution, fitting the number of GFP-IL-2Rγ molecules per cluster with a mixture model, as previously used for the tracks’ lifetimes, resulted in three Gaussians: G1 with 2.6 ± 0.6 molecules, G2 with 3.9 ± 0.8 molecules, and G3 with 5.5 ± 1.5 molecules (SI Appendix, Fig. S2C). Therefore, we further classified GFP-IL-2Rγ tracks in three main cluster groups: small clusters comprising three or fewer molecules, medium-sized clusters of four to six molecules, and large clusters more than six molecules (Fig. 1F). Representative examples of time-lapse images for each of these clusters, with their fluorescence intensity histogram, are shown in Fig. 2A.

Fig. 2.

Endocytic GFP-IL-2Rγ is composed of medium-sized clusters. (A) Representative images of small, medium, and large GFP-IL-2Rγ clusters along time, together with their respective fluorescence intensity profiles. (B) Proportion of GFP-IL-2Rγ tracks with cluster sizes of three or fewer molecules, four to six molecules, or more than six molecules, represented as a percentage of total GFP-IL-2Rγ tracks (mean ± SEM; −IL-2, n = 3,774 tracks; +IL-2, n = 2,716 tracks; BFA −IL-2, n = 2,689 tracks; BFA +IL-2, n = 4,242 tracks; data from four independent experiments). (C) Duration of GFP-IL-2Rγ tracks from each cluster group, per condition (mean ± SEM in seconds; −IL-2: fewer than or equal to three molecules, n = 2,898 tracks; four to six molecules, n = 841 tracks; more than six molecules, n = 35 tracks; +IL-2: fewer than or equal to three molecules, n = 1,602 tracks; four to six molecules, n = 1,015 tracks; more than six molecules, n = 99 tracks; BFA −IL-2, fewer than or equal to three molecules, n = 1,236 tracks; four to six molecules, n = 1,329 tracks; more than six molecules, n = 124 tracks; BFA +IL-2: fewer than or equal to three molecules, n = 1,375 tracks; four to six molecules, n = 2512 tracks; more than six molecules, n = 355 tracks; data from four independent experiments). ANOVA (Krustal–Wallis) with Dunn’s multiple comparisons. ns, nonsignificant. *P ≤ 0.05; ***P ≤ 0.001. (D) Colocalization of GFP-IL-2Rγ (in green) with EndoA or with Dyn2 (in magenta). Insets show time-lapse frames of interaction and not the full-length GFP-IL-2Rγ tracks. Relative frequency of GFP-IL-2Rγ tracks and distribution of their number of molecules, colocalizing with either EndoA (magenta bars) or Dyn2 (green bars; GFP-IL-2Rγ/EndoA, n = 94 colocalizing tracks; GFP-IL-2Rγ/Dyn2, n = 92 colocalizing tracks). For clarity purposes, images in A and D were processed (background subtraction: 100 pixels; Gaussian blur filter: radius 1). (Scale bars: 5 μm; Insets, 1 μm.)

In the absence of IL-2, the majority of receptors form small clusters (73.7% of the tracks), a lower proportion (25.2%) corresponds to medium-sized clusters, and only 1.1% represents large clusters (Fig. 2B). In the presence of the ligand, we observed a 1.5-fold increase in the percentage of GFP-IL-2Rγ tracks forming clusters of four to six molecules (37.6% of tracks) (Fig. 2B). Since IL-2 acts on the efficiency of IL-2R uptake (1.3-fold increase in receptor endocytosis upon IL-2) (SI Appendix, Fig. S1G), the increase in this cluster subpopulation suggests that the medium-sized clusters are optimal for endocytosis. Coherently, cells treated with BFA (consequently enriching the endocytic population) showed a nearly 1.5-fold increase in the proportion of medium-sized clusters (44.9% −IL-2 and 61.6% +IL-2) in detriment of small clusters (Fig. 2B). In addition, while small clusters have a mean duration of 30 s, independently of IL-2, medium-sized clusters have a shorter mean duration upon IL-2 incubation (from 44 ± 1.2 s −IL-2 to 35 ± 0.7 s +IL-2) (Fig. 2D). This is in agreement with a faster rate of receptor uptake upon ligand binding (SI Appendix, Fig. S1G) and confirms that this cluster subpopulation represents the active endocytic one. Large clusters, with more than six molecules, exhibit long durations (mean of 69 ± 8.6 s −IL-2 and 58 ± 4.4 s +IL-2) (Fig. 2C), suggesting a delayed or stalled internalization. Treatment of cells with BFA led to a decrease in the duration of medium-sized and large clusters, similarly to IL-2 treatment, which can be explained by the exclusion of newly secreted GFP-IL-2Rγ and consequently, an enrichment in the endocytic population (Fig. 2C). To reinforce our conclusion that only larger IL-2R clusters (more than three molecules) are competent for endocytosis, we performed multichannel imaging with two known endocytic factors of CIE, dynamin and endophilin, as good markers of productive CIE pits (9, 15, 19). We transfected Kit225 GFP-IL-2Rγ cells with either Endophilin A (EndoA)-RFP or Dynamin-2 (Dyn2)-mCherry; performed live TIRF acquisitions, calibrating our imaging system for single-particle tracking as previously described; and identified the colocalized tracks using Icy (Fig. 2D and Movies S1–S15). Our data shows that dynamin and endophilin both colocalize with IL-2Rγ clusters comprising more than three molecules, preferentially with medium-sized clusters (four to six molecules, 56 to 73% of colocalized tracks) (Fig. 2D and Movies S1–S15). These data strongly support our conclusion that medium-sized clusters of IL-2Rγ represent the active endocytic population.

Quantitative Analysis of Disappearing Clusters by Fluorescence Decay Fitting and Cluster Population Dynamics.

Despite the robustness and accuracy of eTrack for tracking endocytic spots, the plugin cannot determine whether the end of a track corresponds to the effective entry of the molecule cluster into the cell or to the departure of the highly motile cluster or even due to GFP photobleaching. Thus, we analyzed the disappearing tracks with respect to the cluster size by collecting and normalizing to [0,1] (0 = minimum value and 1 = maximum value of each curve) the intensity trace of clusters along their duration (Fig. 3). We fitted all curves with a sigmoid function $f\left(t\right)=\text{high}+\frac{\text{low}-\text{high}}{1+{10}^{\left(\mu -t\right)/\tau }}$, where $\mu$ and $\tau$ are dynamical parameters that correspond, respectively, to the halftime of disappearance and the time constant of disappearance. We obtained a very good fit (Fig. 3A), with the median parameters for the n = 61 fitted curves being $\text{high}=67\%,\hspace{0.17em}\text{low}=16\%,\hspace{0.17em}\mu =51\hspace{0.17em}\text{s},\hspace{0.17em}\text{and}\hspace{0.17em}\tau =11.5\hspace{0.17em}\text{s}.$ The adequate fitting of the disappearance curve with a sigmoid function provides biophysical information on the endocytosis process. First, the high plateau value before the onset of intensity decreasing toward the background value rules out any bleaching or partial disassembly of receptor clusters before their putative entry. Second, the smooth decrease of fluorescence intensity over a few tens of seconds indicates a putative gradual entry of the receptor into the cell and therefore, its progressive disappearance from the TIRF field of view.

Fig. 3.

Quantitative analysis of disappearing clusters by fluorescence decay fitting. (A) Fluorescence intensity of individual GFP-IL-2Rγ clusters (n = 62) is normalized and fitted with a sigmoid function. The parameters $\mu$ and $\tau$ are dynamic parameters of the model that correspond, respectively, to the halftime and the time constant of disappearance (Materials and Methods). The black curve corresponds to the median intensity of normalized traces ($\pm$ SD). The red solid line corresponds to the sigmoid function with median parameters $\text{high}=67\%,\hspace{0.17em}\text{low}=16\%,\hspace{0.17em}\mu =51\hspace{0.17em}\text{s},\hspace{0.17em}\text{and}\hspace{0.17em}\tau =11.5\hspace{0.17em}\text{s}$. For each individual fluorescence trace, the time origin (t = 0) is aligned with the half-disappearance time [i.e., $\text{intensity}\left(t=0\right)=\frac{\text{low}+\text{high}}{2}$]. (B) Disappearance time constant $\tau$ as a function of the cluster size. The red line represents the linear regression $\left(\rho =0.51,\hspace{0.17em}{p}_{\text{value}}=2.2\times \hspace{0.17em}{10}^{-5}\right).$ (C) Halftime of disappearance $\mu$ as a function of the cluster size. The red line represents the linear regression ($\rho =-0.45,\hspace{0.17em}{P}_{\text{value}}=2.2\times {10}^{-4}$).

We further analyzed the correlation between the disappearance time constant $\tau$ and the cluster size (Fig. 3B). We observed a positive correlation between the time required for spot disappearance and the cluster size ($\rho =0.51,\hspace{0.17em}{p}_{\text{value}}=2.2\times \hspace{0.17em}{10}^{-5}$). The slower disappearance of larger clusters could indicate a more productive endocytosis of receptors as observed for CME. We then plotted the halftime of disappearance $\mu$ as a function of the cluster size (Fig. 3C) and observed a negative correlation ($\rho =-0.45,\hspace{0.17em}{p}_{\text{value}}=2.2\times \hspace{0.17em}{10}^{-4}$). Given that the onset of cluster disappearance (and putative entry into the cell) is approximatively given by $\mu -\tau$ (corresponding to an initial 10% decrease of intensity), it means that the onset of larger clusters disappearance happens faster than for small clusters. Once again, this is in line with an early and progressive entry of larger clusters.

In addition, we performed an analysis of the global decay of cluster populations at the cell membrane (SI Appendix, Fig. S3). We measured the total number of clusters per frame with respect to time in groups of cells and fitted the disappearance rate of each cluster population with a monoexponential function in the absence (SI Appendix, Fig. S3A) (n = 6 experiments) or in the presence of IL-2 (SI Appendix, Fig. S3B) (n = 5 experiments). We observed that the disappearance rate significantly decreases with cluster size, from 0.011 ± 0.001 s⁻¹ (SE) for small clusters with three molecules (with or without IL-2) to 0.043 ± 0.012 s⁻¹ (without IL-2) and 0.058 ± 0.012 s⁻¹ (with IL-2) for larger clusters with six molecules. We then compared the computed disappearance rates with the previously measured track durations (Fig. 2C). We observed that the duration of medium-sized cluster tracks (with four to six molecules) was similar to the inverse of the disappearance rate (SI Appendix, Fig. S3 A and B), indicating that the ending of medium clusters corresponds to effective endocytosis. On the other hand, for smaller clusters (three molecules), the inverse of the disappearance rate (∼100 s) was much larger than the measured track duration (∼30 s), indicating that most track terminations should be due to cluster mobility at the cell membrane rather than effective entry into the cell.

Finally, we measured the proportion of cluster sizes with or without IL-2 and found, similarly to the previous analysis of tracks (Fig. 1B), that medium-sized clusters were significantly enriched in the presence of IL-2 (SI Appendix, Fig. S3 D and E). By combining the measured rate of cluster disappearance with their relative proportion in the absence or presence of IL-2, we computed the disappearance rate of IL-2Rγ molecules as the weighted sum of cluster disappearance rates pondered with their relative proportion with or without IL-2 (SI Appendix, Fig. S3F). We found that the significant enrichment of medium-sized clusters with IL-2, together with the increased rate of disappearance for these clusters, led to a significantly higher disappearance rate of IL-2Rγ in presence of the ligand (median = 0.126 vs. 0.08 s⁻¹ without IL-2, P value = 0.017). This faster IL-2Rγ uptake with IL-2 is in line with our previous data (SI Appendix, Fig. S1G).

Altogether, our results point to the existence of different clusters of GFP-IL-2Rγ at the surface of lymphocytes. While the majority of the receptor is present in small clusters of up to three molecules, likely corresponding to secreted and/or diffusive receptor, clusters with four to six molecules are actively internalized, and IL-2 favors their formation. Clusters with more than six molecules show a delayed internalization, perhaps forming large platforms at the PM that are less efficiently endocytosed.

Cholesterol and the Actin Cytoskeleton Regulate Clustering of GFP-IL-2Rγ.

The PM is partitioned in functional domains with distinct composition, for which membrane cholesterol and the actin cytoskeleton greatly contribute (28, 29). These domains function as barriers to the free diffusion of proteins and lipids, regulating the interactions between membrane-associated components and hence, clustering. In line with this, we decided to study the role of membrane cholesterol and branched actin filaments in GFP-IL-2Rγ clustering.

We used methyl-β-cyclodextrin (MβCD) to selectively extract cholesterol from the PM (30) and disrupt lipid rafts, where IL-2Rα, -β, and -γ chains are commonly found (14, 15, 31). Cells treated with 2 mM MβCD showed a 2.5-fold decrease of PM cholesterol (SI Appendix, Fig. S4E) and an increased number of GFP-IL-2Rγ molecules per cluster compared with nontreated cells (Fig. 4A). This is not due to an increased expression of GFP-IL-2Rγ at the cell surface upon MβCD treatment, as assessed by flow cytometry analysis (SI Appendix, Fig. S4B). Strikingly, in the absence of IL-2, cholesterol depletion leads to a 12-fold increase in the proportion of clusters with seven or more molecules (8.5% in MβCD-treated cells compared with 0.7% in the control), as well as an increase (1.3-fold) in the proportion of medium-sized clusters (Fig. 4B). Treatment with IL-2 and MβCD also leads to a higher number of molecules per cluster, although to a lesser extent than in the absence of IL-2, particularly for large clusters (2.3-fold increase with IL-2) (Fig. 4B). These data indicate that PM cholesterol regulates oligomerization of GFP-IL-2Rγ.

Fig. 4.

Membrane cholesterol and the actin cytoskeleton regulate GFP-IL-2Rγ clustering and diffusion. (A) Distribution of the number of GFP-IL-2Rγ molecules per track in control and MβCD-treated cells, IL-2 starved (−IL-2) or IL-2 stimulated (+IL-2). Histograms represent relative frequency of tracks, and dashed lines connecting the bars show the overall distribution of GFP-IL-2Rγ molecules per track in each condition. (B) Proportion of GFP-IL-2Rγ tracks with cluster sizes of three or fewer, four to six, or more than six molecules, represented as a percentage of total GFP-IL-2Rγ tracks (mean ± SEM). Data in A and B were obtained from three independent experiments (−IL-2, n = 7,035 tracks; +IL-2, n = 6,892 tracks; MβCD −IL-2, n = 3,162 tracks; MβCD +IL-2, n = 2,741 tracks). (C) Duration of GFP-IL-2Rγ tracks from each cluster group per condition (mean ± SEM in seconds; −IL-2: fewer than or equal to three molecules, n = 1,416 tracks; four to six molecules, n = 564 tracks; more than six molecules, n = 12 tracks; +IL-2: fewer than or equal to three molecules, n = 1,489 tracks; four to six molecules, n = 858 tracks; more than six molecules, n = 100 tracks; MβCD −IL-2: fewer than or equal to three molecules, n = 1,305 tracks; four to six molecules, n = 1,490 tracks; more than six molecules, n = 367 tracks; MβCD +IL-2, fewer than or equal to three molecules, n = 1,319 tracks; four to six molecules, n = 1,277 tracks; more than six molecules, n = 145 tracks; data from four independent experiments). (D) Total displacement of GFP-IL-2Rγ tracks (Left) and search radius of that displacement (Right) in control and MβCD-treated cells. Each dot represents one cell, and means (± SEM) are shown in micrometers. Control, n = 31 cells. MβCD, n = 34 cells. Data were obtained from three independent experiments. (E) Distribution of the number of GFP-IL-2Rγ molecules per track in control and CK-666–treated cells, IL-2 starved (−IL-2) or IL-2 stimulated (+IL-2). (F) Proportion of GFP-IL-2Rγ tracks with cluster sizes fewer than or equal to three, four to six, or more than six molecules represented as a percentage of total GFP-IL-2Rγ tracks (mean ± SEM). Data in E and F were obtained from seven independent experiments for control and from three independent experiments for CK-666 (−IL-2: n = 4,969 tracks; +IL-2: n = 4,089 tracks; CK-666 −IL-2: n = 4,126 tracks; CK-666 +IL-2, n = 3,392 tracks). (G) Duration of GFP-IL-2Rγ tracks from each cluster group per condition (mean ± SEM in seconds; −IL-2: fewer than or equal to three molecules, n = 3,986 tracks; four to six molecules, n = 1,864 tracks; more than six molecules, n = 53 tracks; +IL-2, fewer than or equal to three molecules, n = 2,644 tracks; four to six molecules, n = 2,181 tracks; more than six molecules, n = 274 tracks; CK-666 −IL-2: fewer than or equal to three molecules, n = 1,337 tracks; four to six molecules, n = 1,778 tracks; more than six molecules, n = 342 tracks; CK-666 +IL-2: fewer than or equal to three molecules, n = 1,245 tracks; four to six molecules, n = 1,503 tracks; more than six molecules, n = 155 tracks. Data are from three and seven independent experiments for control and CK-666, respectively. (H) Total displacement of GFP-IL-2Rγ tracks (Left) and search radius of that displacement (Right) in control and CK-666–treated cells. Each dot represents one cell, and means (± SEM) are shown in micrometers. Control, n = 32 cells. CK-666, n = 34 cells. Data are obtained from three independent experiments. Statistics in C and G were performed by ANOVA (Krustal–Wallis) with Dunn’s multiple comparisons within each cluster group and in D and H by a two-tailed unpaired t test. ns, nonsignificant. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001.

Analysis of the duration of GFP-IL-2Rγ tracks revealed that cholesterol depletion leads to a decrease in the lifetime of tracks within all cluster groups (Fig. 4C). However, this is not explained by a faster rate of endocytosis since cells treated with MβCD show an impaired internalization of the receptor (SI Appendix, Fig. S4A), as previously suggested (14, 15). We then investigated if the short duration of GFP-IL-2Rγ tracks, observed upon cholesterol depletion, could result from a faster diffusion of the receptor at the PM, which our eTrack method would be unable to follow. Examination of GFP-IL-2Rγ lateral displacement, by high-frequency live acquisitions (33 Hz), shows a significant increase in the total displacement of the receptor (1.2-fold increase), as well as in the search radius of that displacement (1.3-fold increase), in cells treated with MβCD compared with nontreated cells (Fig. 4D). Analysis of the mean-square displacement of the receptor to obtain diffusion coefficients corroborates these results (0.047 μm² s⁻¹ in nontreated cells vs. 0.129 μm² s⁻¹ in MβCD-treated cells) (SI Appendix, Fig. S4H), explaining the reduced lifetime observed for all receptor tracks. Therefore, depletion of cholesterol leads to an increased lateral diffusion of GFP-IL-2R, which might promote the formation of large clusters.

In order to study the role of the actin cytoskeleton, we specifically disrupted actin assembly via the Arp2/3 complex with the inhibitor CK-666 (SI Appendix, Fig. S4F) (32), shown previously to inhibit constitutive IL-2R endocytosis (20). Similarly to cholesterol depletion, we observed an increase in the number of GFP-IL-2Rγ molecules per cluster in cells treated with 100 μM CK-666 (Fig. 4E). Inhibiting Arp2/3 leads to larger GFP-IL-2Rγ clusters (Fig. 4E), without affecting the surface expression of the receptor (SI Appendix, Fig. S4D). Like MβCD, CK-666 treatment mostly induces a higher proportion of clusters with seven or more molecules (9.8-fold increase without IL-2 and 1.6-fold increase with IL-2) (Fig. 4F) and to a lesser extent, medium-sized clusters (1.4-fold increase −IL-2 and 1.3-fold increase +IL-2) (Fig. 4F). IL-2 partly attenuates the effect of CK-666 in the formation of larger clusters (Fig. 4 E and F). Thus, Arp2/3-mediated actin polymerization regulates the cluster size of GFP-IL-2Rγ. Treatment with CK-666 resulted in a decrease in GFP-IL-2Rγ lifetimes within all cluster groups (Fig. 4G), despite the inhibition in receptor endocytosis (SI Appendix, Fig. S4C). Again, this can be explained by an enhanced lateral diffusion of the receptor at the cell surface, as confirmed by the increase in total displacement, search radius, and diffusion coefficient (0.062 μm² s⁻¹ in nontreated cells vs. 0.179 μm² s⁻¹ in CK-666–treated cells) of GFP-IL-2Rγ in cells treated with CK-666 (Fig. 4H and SI Appendix, Fig. S4H).

Interestingly, we performed the same diffusion experiments for TfR, a transmembrane protein excluded from lipid rafts and internalized by a clathrin-dependent route, and observed a different effect of either MβCD or CK-666 treatment than for IL-2R. First, in control conditions, TfR is much less diffusive than IL-2R (by fivefold in the diffusion coefficient, 0.011 μm² s⁻¹) (SI Appendix, Fig. S4H); upon cholesterol depletion, we observed a slight decrease in TfR total displacement, while there was no effect with CK-666 (SI Appendix, Fig. S4 G and H and Movies S1–S15). This result suggests that IL-2R lateral diffusion is different from other nonraft-associated transmembrane proteins.

The cholesterol content of the PM and the actin meshwork underneath it play an important role in the biophysical properties of the lipid bilayer, and their perturbation has been shown to prevent clathrin-coated pit invaginations (33, 34). To determine if cholesterol depletion or disruption of the branched actin cytoskeleton has an impact on the formation of IL-2R–containing pits, we performed a surface immunogold labeling of IL-2Rβ in GFP-IL-2Rγ cells, treated or not with MβCD or CK-666 and incubated with or without IL-2. All conditions were done at 37 °C with a 3-min uptake to slightly allow initiation of endocytic pits, as previously described (20). Cells were fixed, and ultrathin sections (70 nm) were observed by transmission electron microscopy (TEM). We quantified the proportion of IL-2Rβ–labeling beads present in pits or in flat membrane (Fig. 5 and SI Appendix, Fig. S5B). In control conditions, without IL-2, 74% of beads are in flat membrane regions, and only 26% are present in pits. In contrast, IL-2 leads to a 1.4-fold increase in the percentage of IL-2Rβ associated with pits (36% in pits) (Fig. 5 A and B), in agreement with the IL-2–mediated increase on the rate of receptor endocytosis (SI Appendix, Fig. S1G) (13). Cholesterol depletion induces a striking decrease in the percentage of IL-2Rβ present in pits, with 91% of beads associated with flat membrane (Fig. 5 A and B). In addition, MβCD induces an overall flattening of the PM, with cells displaying less evident invaginations, confirming the role of cholesterol on endocytic vesicle formation (SI Appendix, Fig. S5A). Nonetheless, incubation with IL-2 after MβCD treatment led to the partial recovery of IL-2Rβ in endocytic pits (33%) (Fig. 5B). The impact of cholesterol on PM invaginations and on IL-2R–containing pits is consistent with our previous results showing an inhibitory effect of MβCD on IL-2R uptake (SI Appendix, Fig. S4A) and the formation of large GFP-IL-2Rγ clusters, partly counteracted by IL-2 (Fig. 4 A and B). Inhibition of the Arp2/3 complex with CK-666 led to similar results, with 96% of IL-2Rβ found in flat membrane regions and 4% found in endocytic pits in the absence of IL-2. Incubation with the ligand slightly increased the percentage of beads found in endocytic pits of CK-666–treated cells to 10%, although in sharp contrast to 32% in the control (Fig. 5 C and D).

Fig. 5.

Membrane cholesterol depletion and Arp2/3 inhibition preclude IL-2R from endocytic pits. (A) Representative TEM pictures of IL-2Rβ labeled by a specific antibody and protein A coupled to 10-nm gold beads at the surface of control or MβCD-treated Kit225 GFP-IL-2Rγ cells in the absence (−IL-2) or presence (+IL-2) of IL-2. IL-2Rβ can be seen in flat PM regions or in endocytic pits. Magnifications of labeled IL-2Rβ (dashed white squares) are shown below. Scale bars are shown. (B) Quantification of labeled IL-2Rβ at the PM of control or MβCD-treated cells (±IL-2), according to its association with endocytic pits or flat membrane. Results are expressed as a percentage of total beads found at the PM. (C) Representative TEM pictures of IL-2Rβ, labeled as in A, at the surface of control or CK-666–treated Kit225 GFP-IL-2Rγ cells in the absence (−IL-2) or presence (+IL-2) of IL-2. Magnifications of labeled IL-2Rβ (dashed white squares) are shown below. Scale bars are shown. (D) Quantification of labeled IL-2Rβ, according to its PM association, at the PM of control or CK-666–treated cells (±IL-2). Results are expressed as a percentage of total beads found at the PM.

Altogether, these results demonstrate that membrane cholesterol and actin filaments are key regulators of IL-2Rγ clustering and endocytosis. Both cholesterol and branched F-actin control receptor diffusion at the PM, most likely assisting in receptor anchorage and engagement for endocytosis and ultimately preventing the stochastic assembly of larger IL-2Rγ clusters.

Cholesterol and F-Actin Differently Control IL-2R Signaling.

Since cholesterol depletion and F-actin disruption lead to the assembly of very large GFP-IL-2Rγ clusters (more than seven molecules), we then tested if these could have an impact on receptor signaling. In presence of the ligand, the β- and γ-chains of IL-2R oligomerize, inducing activation of Jak1 and Jak3 and the phosphorylation of the receptor chains on several tyrosine residues, followed by initiation of the STAT5, PI3K, and MAPK signaling cascades (12).

We coimmunoprecipitated IL-2Rβ to investigate its interaction with GFP-IL-2Rγ and consequent tyrosine phosphorylation in cells untreated or treated with either MβCD or CK-666. Cholesterol extraction with MβCD led to an increase in GFP-IL-2Rγ coimmunoprecipitated with IL-2Rβ and to a higher phosphotyrosine signal (Fig. 6 A and C). In addition, the corresponding lysates showed an increase in phospho-ERK1/2 and phospho-STAT5 in MβCD-treated cells compared with the control (Fig. 6 A and C). Strikingly, the opposite effect was seen in cells treated with CK-666 (Fig. 6 B and C). We observed a marked decrease in the interaction of GFP-IL-2Rγ and IL-2Rβ and consequently, a reduction in tyrosine phosphorylation. In agreement, the corresponding lysates showed a decrease in STAT5A and ERK1/2 phosphorylation (Fig. 6 B and C). Moreover, when performing IL-2–mediated signaling kinetics, we observed a sustained activation of ERK, with a noticeable phospho-ERK1/2 signal still present at 30 min, in cells treated with MβCD (Fig. 6D). Contrarily, cells treated with CK-666 show an inhibition of signaling, with decreased levels of phospho-STAT5A and phospho-ERK1/2 induced by IL-2 at all time points (Fig. 6E), similarly to our immunoprecipitation results (Fig. 6B) and to previous published data (35).

Fig. 6.

Cholesterol and actin differently impact IL-2–dependent signal transduction. (A) Lysates of Kit225 GFP-IL-2Rγ control or MβCD-treated cells, stimulated or not with IL-2 for 30 min at 37 °C, were immunoprecipitated (IP) with an antibody against IL-2Rβ (Left). Western blots were probed with antibodies against phosphotyrosine (pTyr; 4G10), IL-2Rγ (AF-284), and IL-2Rβ (C-20). The respective lysates are shown (Right) and were immunoblotted with antibodies against phospho-STAT5A (p-STAT5A), phospho-ERK1/2 (p-ERK1/2), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). (B) Lysates of Kit225 GFP-IL-2Rγ control or CK-666–treated cells, stimulated or not with IL-2 for 30 min at 37 °C, were immunoprecipitated with an antibody against IL-2Rβ (Left). Western blots were probed as in A for both IP and lysates. (C) Quantifications of A and B. IP quantifications (Left) are represented as a percentage of the control, normalized to the amount of IL-2Rβ that was immunoprecipitated. Quantifications of lysates (Right) are shown as a percentage of the control, normalized to the respective levels of GAPDH (mean ± SEM from four independent experiments for MβCD and three independent experiments for CK-666). (D) Lysates from Kit225 GFP-IL-2Rγ cells nontreated (control) or treated with MβCD and stimulated or not with IL-2 for 0, 5, 10, 30, and 60 min at 37 °C. Proteins were detected by western blot using antibodies against p-STAT5A, p-ERK1/2, GAPDH, STAT5A, and ERK2. (E) Lysates from Kit225 GFP-IL-2Rγ control or CK-666–treated cells stimulated or not with IL-2 for 0, 5, 10, 30, and 60 min at 37 °C. Samples were immunoblotted as in D. DMSO, dimethyl sulfoxide.

These results allowed us to conclude that extraction of membrane cholesterol and disruption of F-actin distinctively impact IL-2–mediated signaling, while they lead to seemingly similar cellular effects (IL-2Rγ clustering phenotypes and endocytic default). Cholesterol depletion promotes an association of the β- and γ-chains, consequently leading to sustained STAT5 and ERK signaling. Contrarily, F-actin disruption prevents the IL-2–dependent interaction of β- and γ-chains, thus hindering activation of the signaling cascade. Therefore, whereas depletion of cholesterol likely enhances the formation of large complexes triggering a robust signaling downstream of IL-2Rγ/IL-2Rβ, disruption of the actin cytoskeleton causes an increase in large clusters that are inefficient for IL-2–induced signal transduction.

Discussion

Our results provide further insight into the factors regulating clustering at the PM by revealing the importance of cytokine receptor clustering for CIE and signal transduction. The comprehensive study of the lifetimes of IL-2Rγ endocytic tracks and their single-molecule analysis allowed us to identify three subpopulations of IL-2Rγ at the PM, each with distinctive cluster sizes. Golgi-mediated secretion drives IL-2Rγ to the cell surface, conceivably as preassembled clusters, in agreement with previous studies on the quaternary structure of IL-2R showing that γ_c exists as a stable homotrimer (36) and that this homotypic complex can be seen both at the endoplasmic reticulum and in the Golgi (37). This newly secreted population is represented by small clusters, of up to three molecules, diffusing at the membrane, and that would not be competent for endocytosis, as we almost never observe these small clusters colocalizing with endocytic factors such as dynamin or endophilin (Fig. 2D). Once at the PM, addition of IL-2Rγ molecules to preassembled clusters leads to the formation of medium-sized clusters, with four to six molecules that are endocytic competent (Fig. 7A). IL-2 promotes the formation of medium-sized clusters (Fig. 7B) that are rapidly internalized, as shown both by our endocytic tracking analysis and by the fitting of clusters’ disappearance rate and their proportions. Clustering might arise from the “bridging” of multiple clusters through IL-2 interaction with the different IL-2R chains. Alternatively, it could occur via the reduced diffusion and stabilization of clusters in the presence of the ligand, allowing addition of further receptor molecules. The IL-2–induced reorganization of IL-2Rγ clusters corroborates the more efficient endocytosis of the receptor upon the ligand. The lifetimes of the endocytic IL-2Rγ subpopulation (30 to 80 s) are shorter than previously determined for IL-2Rβ (50 to 120 s) (19). Importantly, the former study analyzed the β-chain of the receptor by overexpressing it on epithelial cells, whereas we report the lifetimes of IL-2Rγ in T cells endogenously expressing the receptor. The physiological context might be critical for an efficient endocytosis and thus, explain the faster receptor uptake in lymphocytes. The lifetimes of medium-sized clusters are partly similar to those of clathrin-coated pit formation (28 to 32 s), with longer-lived clathrin clusters corresponding to larger coated vesicles (24). Nevertheless, a single endocytic pathway can show variable durations of internalization depending on the cargo being uptaked. A third subpopulation, characterized by large clusters with more than six molecules, represents stalled receptor complexes inefficiently endocytosed. In agreement with this, cholesterol or F-actin disruption, two key factors for IL-2R endocytosis, led to a higher proportion of this subpopulation of clusters.

Fig. 7.

Model for IL-2Rγ clustering. (A) At steady state, IL-2Rγ is present at the PM as small clusters (with up to three molecules) representing newly secreted and/or abortive endocytic events, medium clusters (four to six molecules) that are endocytic competent, and large clusters (with more than six molecules) that are stalled. (B) Stimulation with IL-2 promotes clustering of IL-2Rγ, the formation of endocytic competent clusters (medium clusters), and initiation of the signaling cascade by oligomerization of IL-2Rβ and IL-2Rγ. (C) Lipid raft disruption leads to the enhanced lateral diffusion of receptor molecules and their stochastic interaction within the structured compartments defined by the actin cytoskeleton meshwork. This promotes the assembly of large clusters, inhibiting IL-2R endocytosis while inducing robust signaling events arising from the interaction of IL-2Rβ and IL-2Rγ. (D) Disruption of the actin cytoskeleton promotes assembly of large IL-2Rγ clusters due to the increased lateral diffusion of free diffusing and lipid raft–bound receptors, leading to a defect in endocytosis. In this case, the absence of oligomerization between the β- and γ-chains of the receptor prevents the initiation of the IL-2–mediated signaling cascade.

Altogether, our work indicates that IL-2R cluster size is determinant for its endocytosis, suggesting that clustering might function as a checkpoint mechanism regulating pit invagination, as proposed for CME or for endocytosis of clathrin-independent carriers of GPI-APs (glycosylphosphatidylinositol-anchored proteins) enriched early endosomal compartments (CLIC-GEEC pathway) (3, 6). Regarding IL-2R, we can conceive that both γ- and β-chains have a well-folded extracellular domain and an unstructured intracellular domain. Intrinsically disordered domains have been reported in the cytoplasmic tails of other cytokine receptors and seem to be relevant for intracellular signaling (38, 39). The cytoplasmic tails of β and γ are strongly associated with Jak1 or Jak3 and upon IL-2, with other associated signaling molecules, increasing the size of these membrane nanodomains and the steric pressure between molecules. The asymmetrical organization of the receptor chains might induce a curvature in the lipid bilayer that would be enhanced by protein crowding, as shown in vitro (1), thus promoting membrane curvature toward the cytoplasm to increase membrane surface and reduce steric repulsion. Recently, cholera toxin subunit B has been shown to induce a negative membrane curvature only when bound to three or more of its glycosphingolipid receptors GM1 (40). Larger clusters (more than six molecules) would less efficiently bend the membrane, due to the increased energy cost and steric repulsion restricting interaction with endocytic adaptors. Similar results have been observed for CME, suggesting that increased cargo size can slow down internalization rate and drive the formation of large clathrin-coated pits with low curvature or flat clathrin lattices that are endocytically inactive (27, 41). Another hypothesis is that an endocytic factor, such as dynamin or endophilin, may regulate a cargo loading checkpoint in this CIE pathway, as proposed for TfR (42). Such a mechanism could selectively promote the efficient internalization of clusters comprising four to six cargo molecules while leading to the disassembly and/or diffusion of clusters that do not reach the productive size, as proposed for CME (27).

We identified membrane cholesterol and the branched actin cytoskeleton as key regulators of IL-2Rγ clustering and IL-2–induced signaling. These factors are fundamental in the dynamic maintenance of PM domains, limiting protein diffusion and controlling their partition and function at the cell surface. One can hypothesize that in normal conditions, clustering of IL-2R induces receptor confinement that is necessary for its endocytosis. In the absence of cholesterol or upon actin perturbation, IL-2R is not trapped/confined when clustered, as it has been seen for other raft-associated receptors (e.g., FcεRI) (43). This would take part in the inhibition of endocytosis, and consequently, the receptor would have higher chances of assembling into larger clusters, again delaying uptake (Fig. 7 C and D). The fact that TfR, a nonraft-associated receptor entering by CME, does not have the same diffusion profile as IL-2R reinforces our model in which CIE strongly depends on receptor clustering of raft-associated cargos. Membrane cholesterol and F-actin seem to primarily function to segregate IL-2Rγ molecules, avoiding unwanted receptor interactions and stochastic clustering that could lead to unrestrained signal activation. In our model, concluding on the cross-talk between clustering and signaling is intricate since we observe two different signaling outcomes paired with increased IL-2Rγ clustering. The enhanced association of receptor subunits upon cholesterol depletion favors large clusters, leading to a sustained IL-2–dependent signal (Fig. 7C). In contrast, Arp2/3 inhibition abrogates IL-2–induced signaling (Fig. 7D). A decrease in IL-2–induced STAT5 phosphorylation had already been shown in cells treated with the actin depolymerizing drug cytochalasin D (35). One hypothesis is that IL-2Rγ and IL-2Rβ are initially segregated into distinct lipid rafts, which prevents their interaction in unstimulated conditions. Following IL-2 binding, interaction with the actin meshwork would increase the confinement of these raft domains and promote their coalescence, leading to interaction of the β- and γ- chains and their associated JAKs and inducing the signal cascade. This is in line with works reporting Golgi-derived homoclusters of GPI-APs and the regulation of these nanoclusters by the cortical actin cytoskeleton (44, 45). The fact that homodimers/oligomers of IL-2Rγ have been observed in the Golgi network (37) reinforces the assumption of a distinct microdomain partition of IL-2Rγ and IL-2Rβ. Another possibility is that IL-2Rγ and IL-2Rβ are, at least in part, already associated at the PM, yet their cytoplasmic domains are flexible and kept apart to prevent signal activation, as previously suggested (35). Binding of IL-2 could lead to a conformational change in the preassembled complex, pulling the cytoplasmic tails of IL-2Rγ and IL-2Rβ closer together through local actin polymerization (35) and hence, activating the signaling cascade. Interestingly, a similar actin-mediated anchorage mechanism has been proposed for the IL-7R (46).

Understanding the impact of receptor clustering in signal activation is fundamental, as altered expression or assembly of surface receptors can lead to abnormal signal transduction, being linked with several human diseases such as cancer or immunodeficiences (47, 48). There has been a considerable effort in understanding the role of receptor clustering in cellular trafficking and signaling, underlining the function of organized domains in governing the biophysical and biochemical properties of the PM. Particularly, our work provides insights into the regulation of CIE and cytokine-induced signaling by surface receptor clustering, paving the way for compelling future research.

Materials and Methods

Live-Cell TIRF Imaging.

Live imaging of GFP-IL-2Rγ endocytic tracks was performed as described in ref. 23. Briefly, round 25-mm no. 01 glass coverslips (Fisher Scientific) were precleaned with 70% ethanol followed by acetone and sonicated in an ultrasonic cleaning bath. Coverslips were then coated with Poly-l-lysine (Sigma-Aldrich) to promote cell adherence. Kit225 GFP-IL-2Rγ cells were IL-2 starved 24 h before imaging, and the day of imaging, cells were plated onto precleaned coated coverslips. In experiments with drug treatment, cells were incubated with or without the drug for 30 min at 37 °C, and treatment was maintained throughout image acquisition. Otherwise, in experiments with no drug treatment, cells were immediately imaged in phenol red-free Roswell Park Memorial Institute (RPMI) medium supplemented with 5% fetal bovine serum (FBS). Movies were acquired with a laser-scanning microscope (LSM) 780 Elyra PS.1 TIRF microscope (Zeiss) equipped with an EMCCD Andor Ixon 887 1K camera and using an alpha PIn Apo 100×/1.46 oil objective, a 488-nm (100-mW) HR solid laser line, and a BP 495–575 + LP 750 filter to detect GFP-IL-2Rγ. Image acquisition was done at 1 frame/s during 150 s, with 1% laser power and 150-ms exposure time.

For the modified FRAP analysis, GFP-IL-2Rγ cells, pretreated or not with BFA for 30 min at 37 °C, were first excited with 100% laser power for 30 s to photobleach surface receptors. Following 2 min of recovery, cells were imaged regularly for detection of reappearing GFP-IL-2Rγ. BFA was maintained in the imaging medium throughout acquisition.

For IL-2Rβ/GFP-IL-2Rγ simultaneous two-color live imaging, IL-2Rβ was observed by incubating cells with an anti–IL-2Rβ antibody coupled to Cy3 (561-Cy3) for 2 min at 37 °C in phenol red-free RPMI with 5% FBS. Cells were then carefully washed with the same medium, immediately prior to imaging. Acquisition was done using 488- and 561-nm lasers (both 100 mW) and a dichroic laser-blocking filter (LBF) 488/561, and the parameters of acquisition were as before. For EndoA-RFP or Dnm2-mCherry and GFP-IL-2Rγ simultaneous two-color live imaging in TIRF, we used 488- and 561-nm lasers (both 100 mW) and a dichroic filter LBF 488/561, and the parameters of acquisition were as before.

For displacement analysis, Kit225 GFP-IL-2Rγ cells, pretreated or not with MβCD or CK-666 for 30 min at 37 °C, were imaged with or without incubation of Tf-Cy3 to label TfR using the same TIRF setup as for GFP-IL-2Rγ endocytic track detection, except for acquisition rate that was set to 1 frame/30 ms, for a total of 3 s, to allow xy tracking.

Approximately 8 to 12 cells were acquired per condition per experiment. All live-imaging movies were analyzed using the open-source software Icy (49) (Institut Pasteur).

Additional materials and methods are in SI Appendix, Materials and Methods.

Data Availability

All study data are included in the article and/or supporting information.