It has long been established that dimerization of Interleukin-4 receptor (IL-4R) subunits is a pivotal step for JAK/STAT signal transduction. However, ligand-induced complex formation at the surface of living cells has been challenging to observe. Here we report an experimental assay employing trisNTA dyes for orthogonal, external labeling of eGFP-tagged receptor constructs that allows the quantification of receptor heterodimerization by dual-color fluorescence cross-correlation spectroscopy. Fluorescence cross-correlation spectroscopy analysis at the plasma membrane shows that IL-4R subunit dimerization is indeed a strictly ligand-induced process. Under conditions of saturating cytokine occupancy, we determined intramembrane dissociation constants (Kd,2D) of 180 and 480 receptors per μm2 for the type-2 complexes IL-4:IL-4Rα/IL-13Rα1 and IL-13:IL-13Rα1/IL-4Rα, respectively. For the lower affinity type-1 complex IL-4:IL-4Rα/IL-2Rγ, we estimated a Kd,2D of ∼1000 receptors per μm2. The receptor densities required for effective dimerization thus exceed the typical, average expression levels by several orders of magnitude. In addition, we find that all three receptor subunits accumulate rapidly within a subpopulation of early sorting and recycling endosomes stably anchored just beneath the plasma membrane (cortical endosomes, CEs). The receptors, as well as labeled IL-4 and trisNTA ligands are specifically trafficked into CEs by a constitutive internalization mechanism. This may compensate for the inherent weak affinities that govern ligand-induced receptor dimerization at the plasma membrane. Consistently, activated receptors are also concentrated at the CEs. Our observations thus suggest that receptor trafficking may play an important role for the regulation of IL-4R-mediated JAK/STAT signaling.
The interleukin-4 receptor (IL-4R) subunits are single-pass transmembrane proteins that belong to the important superfamily of hematopoietic cytokine receptors (
). Throughout this family, assembly of ternary ligand-receptor complexes along with conformational changes constitute a prerequisite for proximity-driven cross-activation of receptor-chain-associated, cytoplasmic Janus kinases (JAKs) (
Unlike homotypic receptors such as growth hormone or erythropoietin receptor that form transmembrane domain-mediated dimers, IL-4Rα and IL-2Rγ (common γ-chain, γc) diffuse as monomers in the plasma membrane (
). Thus, at least for the heterotypic receptor types, a diffusion-controlled recruitment of the second receptor chain can be assumed to trigger JAK activation and signal transduction at the plasma membrane. In addition, the IL-2Rγ subunit is shared between different cytokine receptor complexes (
IL-4R is a well-known example of such a heterotypic receptor. IL-4 specifically binds to the IL-4Rα chain, enabling it to recruit either IL-2Rγ to form a type-1 receptor complex or IL-13Rα1 to form a type-2 receptor complex. Alternatively, the type-2 receptor complex can be induced by IL-13 bound to IL-13Rα1 (
An antagonistic IL-4 mutant prevents type I allergy in the mouse: inhibition of the IL-4/IL-13 receptor system completely abrogates humoral immune response to allergen and development of allergic symptoms in vivo.
Using fluorescently labeled IL-4, we previously observed by fluorescence cross-correlation spectroscopy (FCCS) that the IL-13Rα1 chain is recruited into a complex, whereas IL-2Rγ containing type-1 dimers could not be detected in the plasma membrane of HEK293T cells (
). Using synthetic dyes for external receptors labeling in combination with native ligands, we can now detect ligand-induced dimerization for all three IL-4R complexes. Employing a truncated, dominant negative mutant of the IL-4Rα chain even provides sufficient accuracy to quantify for the first time, to our knowledge, two-dimensional dissociation constants in the plasma membrane plane. The affinities are surprisingly weak, pointing toward a cellular concentration mechanism that is necessary to drive dimerization and a stable receptor activation. Based on various additional approaches, we propose that constitutive endocytosis into a novel subpopulation of early endosomes (cortical endosomes, CEs) fulfills this role. These results therefore provide important and unexpected mechanistic insight into IL-4R-mediated JAK/STAT signaling.
Materials and Methods
Molecular biology and cloning
Molecular cloning was performed as described in Worch et al. (
). For the reporter gene assay HEK293T, cells were seeded in a 96-multiwell plate and transfected at 30–40% confluence with 50 ng p3xSTAT6-RE-Luc (STAT6 responsive firefly luciferase), 25 ng pRLTK (constitutive Renilla luciferase), and up to 125 ng of plasmids coding for combinations of IL-4R subunits and STAT6. After 24 h, cells were stimulated with IL-4 or IL-13 (10–20 ng/mL) for 12 h. Before experiment, the cells were washed twice with ice-cold PBS, lysed, and transferred into a white 96-well plate (Corning Inc., Corning, New York). For detection we used luminescent chemicals (Stop&Glo; Promega, Madison, WI) and a Glomax plate reader (Promega). All data points were measured in triplicates.
For immunoblotting, cells were stimulated with IL-4 or IL-13 (Gibco, Grand Island, NY) at 37°C for 30 min, washed with ice-cold PBS (phosphate-buffered saline), and lysed (25 mM Tris pH 7.2, 150 mM NaCl, 5 mM MgCl2, 0.2% NP-40, 1 mM DTT) with 5% Glycerol and the Complete Protease Inhibitor cocktail (Roche). Proteins were detected after blotting with phospho-STAT6 (mouse, 1:2000; Santa Cruz Biotechnology, Dallas, TX) or GAPDH (mouse, 1:4000; Santa Cruz Biotechnology) and visualized with HRP-coupled goat anti-mouse (1:4000; Sigma, St. Louis, MO) using ECL+ (Life Technologies, Norwalk, CT). The STAT6 response kinetics was fitted with a logistic function using the software ORIGIN (OriginLab, Northampton, MA),
with t0 as a typical time constant (half-maximum activation) and A1, A2, and p as parameters.
Synthesis of trisNTA dyes
Synthesis of the trisNTA probe was achieved by coupling of three carboxy functionalized NTA moieties, prepared by alkylation of the amino function of a protected glutamate derivative, to three amino groups of a cyclam scaffold (
). An N-Boc protected 6-amino caproic acid spacer was attached to the fourth amino group. After deprotection, the primary amine of the spacer was conjugated to amino-reactive dyes (QSY-7 (Alexa647, 5-carboxytetramethylrhodamine); Life Technologies). The Ni2+ saturated compounds were purified by HPLC using aqueous 400 mg/L ammonium acetate/acetonitrile gradients and characterized by electrospray ionization-mass spectrometry.
). Primary antibodies: EEA1 (mouse, 1:500; Abcam, Cambridge, UK), phospho-Y497-IL-4Rα (rabbit, 1:1000; Abcam), and Lamp1 (mouse; 1:300; Santa Cruz Biotechnology). Live cell imaging was done at 22°C in air buffer (150 mM NaCl, 20 mM HEPES pH 7.4, 15 mM glucose, 46 mM trehalose, 5.4 mM KCl, 0.85 mM MgSO4, 1.7 mM CaCl2, 0.15 mg/mL bovine serum albumin).
Fluorescence recovery after photobleaching (FRAP)
Experiments were performed at 22°C, on an LSM780 (Zeiss, Jena, Germany) with a 40×, NA 1.2 objective using a multichannel GaAsP detector to collect eGFP emission (490–560 nm). HEK293T cells transiently expressing IL-2Rγ/JAK3-eGFP complexes were analyzed in 8-well chambered Labtek slides (Nalge Nunc, Roskilde, Denmark). The regions of interest (ROIs) were defined perpendicular to the membrane plane (2.5 × 2.5 μm2) (
). JAK3-eGFP was bleached to ∼20% before recovery using the combined excitation of the 405-nm and 488-nm laser lines. We recorded 50–100 frames over a period of 5–10 min. Excitation-induced bleaching in the ROI (area A) during the postbleach period was corrected by a distant, nonbleached control ROI (area B). The intensity curves were corrected for imaging-dependent bleaching, denoting B0 as the prebleach intensity in area B (
Then we normalized the signal to initial prebleach intensity level A0 and C1, the first time-point of the corrected postbleach series
Finally, to derive the time constants t0, D(t) was fitted with a saturating exponential
Stimulated emission depletion (STED) microscopy
Superresolution imaging of JAK3-eYFP positive endosomes was performed on a TCS SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany). eYFP was excited at 514 nm, and excitation depletion was achieved with a 592-nm laser (Leica Microsystems) delivering an ∼350 mW at the object plane. Images were acquired with a 100×, NA = 1.4 oil lens (STED orange, Leica Microsystems). Pixel size for STED imaging was set to 32 nm, resulting in a 2–3-fold oversampling.
Total internal reflection fluorescence (TIRF) microscopy
For time-lapse microscopy, the cells were imaged with a Life Cell AF6000 LX (Leica Microsystems) TIRF microscope equipped with a 37°C incubation chamber. Imaging was done with a HCX PL Apo 100× NA 1.46 oil objective, a QAX filter set for separate excitation and emission of eGFP and A647, and a cooled, high sensitivity electron-multiplying charge-coupled device camera (Andor Technology, Belfast, UK). Time series were taken with an electron-multiplying gain of 300 V and a frame rate of 500 ms.
Technically, FCCS measurements and data analysis were performed as described in Weidemann et al. (
). However, trisNTA-A647 was used as an external label for the His-tagged IL-4Rα constructs. To minimize signal instabilities, the focal volume was placed in a homogeneous area of the bottom membrane, close to the glass support. Measurements were taken at 22°C with excitation intensities resulting in 2–3 kHz per particle in the A647 channel and 2–4 kHz per particle in the eGFP channel. For Kd determination, see the Appendix. Fitting was performed with the open-source software PyCorrFit (
Receptor trafficking into JAK3-eGFP positive endosomes was measured by the uptake of preloaded external dye trisNTA-A647 in cells expressing His-tagged receptor subunits. Cells were seeded and transfected in 8-well Labtek chambers (Nunc). For IL-2Rγ kinetics, we used pNhis-IL2Rg-N2 (80 ng), pJAK3-eGFP (50 ng), and pIL4Ra-N1 (40 ng). For IL-4Rα kinetics, we used pIL2Rg-N1 (80 ng), pJAK3-eGFP (50 ng), and pNhis-IL4Ra-N1 (40 ng) in the absence or presence of 20 ng/mL IL-4 (Life Technologies). Twenty-four hours posttransfection, cells were washed twice with air buffer and chilled for 20 min on ice. Cells were then stained with 150 nM trisNTA-A647 in air buffer for 20 min on ice, washed once in air buffer, and transferred to 37°C starting endocytosis. At different time-points, cells were fixed in 4% PFA/PEM.
Confocal imaging was performed on a LSM780 (Zeiss) using avalanche photo-diode detectors. The 16-bit images were recorded with a 40×, NA 1.2 water immersion objective, using a 75-μm pinhole at a pixel dwell time of 50 μs using filter settings for eGFP (505–610 nm) and A647 (655 nm, long-pass). Laser powers were adjusted to stay <1 MHz. Each cell was imaged in three different planes at 1.5-μm distance. For image analysis, we manually identified cortical endosomes in the eGFP channel; an automated script then identified the maximum intensity and extracted the total counts of a centered 3×3 pixel bin (MATLAB, Rel. 2010a; The MathWorks, Natick, MA) in the A647 channel. Endosomes showing intensities exceeding the linear range were flagged and excluded. Average loading per cell (binning 50–100 endosomes) was used for statistical analysis. For fitting the average trafficking kinetics, we used a hyperbolic function (ORIGIN, OriginLab, Northampton, MA).
HEK293T cells as a model system for analyzing IL-4R-mediated signaling
We studied IL-4R subunit interactions in HEK293T cells that are amenable to imaging-based biophysical approaches due to their size and adherent growth. HEK293T cells endogenously express type-2 signaling components but not the type-1 components IL-2Rγ and JAK3 or STAT6 (Fig. 1A). Because STAT6 phosphorylation and STAT6-dependent reporter gene assays provide a highly specific readout for IL-4R signaling output (
), we assayed pathway activation in HEK293T cells transiently expressing STAT6. Consistent with the endogenous presence of type-2 IL-4R subunits, STAT6 could be activated by both IL-4 and IL-13 (Fig. 1B, and see Fig. S1 in the Supporting Material).
We tested the activity of the various ectopically expressed receptor constructs by a STAT6-responsive luciferase reporter gene assay (Fig. 1C). HEK293T cells did not show IL-4-dependent STAT6 reporter activation in the absence of STAT6. Coexpression of full-length IL-4Rα chain increased the IL-4 response and the baseline. IL-4Rα cytoplasmic deletion constructs (IL-4Ram266 constructs comprised of the amino acids 1–266) lacking the STAT6 docking site dominantly suppressed the endogenous machinery, indicating trafficking behavior similar to that of the endogenous receptors. Their competitive potential was unaffected by N-terminal His-tags, C-terminal eGFP, or both. In contrast, luciferase expression after overexpression of the second receptor subunits IL-13Rα1 or the pair IL-2Rγ/JAK3 was comparable to stimulation of the endogenous type-2 receptors suggesting that the ligand binding α-chain is limiting. In both cases, the signal was suppressed by truncated tails. Although the overexpression of individual subunits therefore does not negatively interfere with signaling, the deletion constructs outcompete the endogenous type-2 machinery at comparable overexpression levels (Fig. 1C, see also the FCCS quantifications below).
Ligand-induced IL-4R subunit recruitment at the cell surface
To probe surface-expressed receptors by FCCS, we first developed an external labeling procedure. Nonneutralizing external labels offer the possibility to compare cross-correlation amplitudes and hence dimerization levels in the absence and presence of unlabeled, and thus fully active, cytokine ligands. For labeling we synthesized hexahistidine binding dyes, which were introduced recently as a promising tool for receptor research (
). The bifunctional compounds consist of three nitrilotriacetic acid groups attached to a tetravalent, cyclic scaffold (trisNTA). A fourth substitution carries an aliphatic linker with an N-terminal amino group suitable for the conjugation of aminoreactive fluorescent dyes (Fig. 2A). By this modular approach we produced TAMRA, QSY7, and Alexa647 (A647) conjugated versions to validate the hexahistidine binding properties with spectroscopic methods. In agreement with literature values, our probes showed a lower nanomolar affinity toward a recombinant eGFP-H6 in free solution (Fig. 2B and see Fig. S2A) with a fast association and dissociation kinetics in the timescale of seconds (see Fig. S2B).
In HEK293T, we coexpressed pairs of fluorescently tagged receptor subunits at the cell surface (Fig. 3A) (
). Under certain conditions, ectopic expression of full-length IL-4Rα chains in HEK293T cells can induce an abnormal rounded shape with large amounts of receptors retained in perinuclear membrane compartments, which appeared unsuitable for microscopic FCCS analysis (see Fig. S3). Based on crystal structures of extracellular domains and our own observations (
), we reasoned that the amino acids located C-terminal of the cytoplasmic JAK1 binding site (No. 267-800, mature numbering) might only play a minor role for complex formation. Therefore, we first used a truncated IL-4Rα chain (H6-IL-4Rαm266) that exhibits improved plasma membrane localization and an extracellular accessible hexahistidine stretch (
). The external dye trisNTA-A647 was applied at saturating conditions (10–20 particles in the observation volume as measured by fluorescence correlation spectroscopy corresponding to ∼35–70 nM). Due to the large difference in diffusion time, the unbound dye and the much slower receptor-coupled labels could be clearly discriminated in the correlation functions (see the fast decay in the autocorrelation curve of the red channel, Fig. 3C). The diffusion coefficient of the fast component (Dt = 374 ± 111 μm2/s; n = 220) corresponds well with dye molecules diffusing in bulk solution (see Table S1 in the Supporting Material). The remaining slow fraction (70–80%) was considered for cross-correlation analysis.
The mobility of the eGFP-tagged receptor chains in the plasma membrane showed some variation among the different constructs. (Fig. 3D). Even in the absence of ligand, IL-2Rγ-eGFP/JAK3 complexes exhibit a reduced mobility as compared to IL-13Rα1-eGFP and the truncated construct (positive control, H6-IL-4Rαm266-eGFP), which may reflect cytoskeleton interactions (
). Applying ligand had no effect. In contrast, type-2 dimers showed a consistently reduced mobility upon ligand binding (IL-13Rα1-eGFP). The order of magnitude of this reduction is more than one would expect for a pure dimerization and may include rearrangements of the local lipid environment of the transmembrane receptors (
). Overall, the diffusion coefficients showed significant scatter, due to the various dynamic processes simultaneously observed at the plasma membrane, e.g., endocytic trafficking.
A more robust readout for complex formation is the cross-correlation ratio (CC), which indicates the degree of codiffusion of the labeled H6-IL-4Rαm266 and the eGFP-tagged second receptor unit (Fig. 3E). In FCCS, the actual CC value depends on which color channel was used for normalization. Here we show CCav, the mean of CCG and CCR (see the Appendix), to demonstrate that the results still hold for a potentially biased abundance of receptor subunits in the two color channels. Applying IL-4 increased CCav for IL-2Rγ-eGFP/JAK3 from 9.2 ± 7.0% to 13.6 ± 8.2% (p < 0.05), and for IL-13Rα1-eGFP from 8.3 ± 7.3% to 35.3 ± 7.9% (p < 0.01), whereas IL-13 produced only 17.3 ± 7.0% (p < 0.01) (see Table S2). The CCav values showed a nearly Gaussian distribution across the measured cell populations; their first moment was positioned between zero and the maximum achievable CCav level, as defined by intramolecular cross-correlation (see Fig. S4, A and B). Thus, ligand-induced complex formation was detected at the plasma membrane for all three complexes.
The magnitude of the fluorescence fluctuations in FCCS obeys a reciprocal relationship with the number of particles in the detection area; thus cellular receptor surface densities can be quantified. In our transient expressions we measured cells expressing, on average, several hundred receptors per μm2 (see Table S2). Assuming a steady-state situation in which the law of mass action governs lateral interactions within the membrane, the abundance of each binding partner from autocorrelation and the fraction of double-labeled dimers derived from cross-correlation enabled us to calculate two-dimensional dissociation constants (see Appendix). Because the binding partners are confined within the two-dimensional membrane plane, these Kd values are expressed in dimensions of receptors per area. Because the external cytokine ligand was applied in excess, these values directly reflect the receptor surface density for which recruitment occurs with a probability of 50%.
We obtained Kd,2D = 180 ± 40 receptors/μm2 for the IL-4Rα:IL-4/IL-13Rα1 interaction, 480 ± 100 receptors/μm2 for IL-4Rα/IL-13:IL-13Rα1, and 1000 ± 170 receptors/μm2 for IL-4Rα:IL-4/IL-2Rγ (Fig. 3F and see Table S2). Calibration of the cross-correlation with autocorrelation amplitudes of either the eGFP or the A647 channel produced consistent results (see Fig. S4C). Notably, the potency of the two different ligands IL-4 and IL-13 to induce a type-2 complex reflects their relative signaling strength (see Fig. S1). However, considering physiological surface expression levels of several receptor molecules per μm2 (
), the order of magnitude of these Kd values for all three receptor types is remarkably high, rendering dimerization at the plasma membrane very ineffective and short-lived.
Encouraged by these results, we finally aimed to confirm ligand-induced subunit recruitment for the full-length IL-4Rα chain. Because the type-1 interaction was already weak, we revisited the type-2 configuration. As is the case for the full-length GFP-tagged IL-4Rα constructs themselves, coexpression IL-13Rα1-eGFP with either nontagged or His-tagged IL-4Rα led to the very same rounded shape and massive retention in a perinuclear membrane compartment, presumably the endoplasmic reticulum (see Fig. S3 and Fig. S5A). However, a small fraction of cells showed a stretched appearance with sufficient colocalization at the cell surface for FCCS (see Fig. S5B). A difference in diffusion between the truncated positive control H6-IL-4Rαm266-eGFP (Dt = 0.24 ± 0.1 μm2/s; n = 121 × 15 s) and the full-length H6-IL-4Rα (Dt = 0.2 ± 0.06 μm2/s; n = 67 × 15 s) was small and unchanged when occupied by cytokine ligands. The FCCS measurements suffered from a lower surface density of full-length H6-IL-4Rα for which we hardly reached 100 receptors per μm2 and, thus, stayed systematically at <Kd. Nevertheless, in agreement with previous results, for both ligands we measured a significant, ligand-induced increase in cross-correlation (see Fig. S5C). The magnitude of this shift was comparable with the experiments using the truncated receptors. The intracellular domains thus play, at most, a minor role for ligand-induced IL-4R dimerization.
IL-4R subunits accumulate in early sorting and recycling endosomes stably anchored below the plasma membrane
To avoid confounding effects of endogenous proteins, we first studied the subcellular distribution of ectopically expressed type-1 receptor subunits. Confocal live cell imaging of nonstimulated cells showed that JAK3-eGFP is recruited to the plasma membrane by IL-2Rγ (Fig. 4A, top panels) and is in turn required for efficient IL-2Rγ membrane localization (
In cells where the IL-2Rγ receptors were fully saturated, increasing levels of nonbound JAK-eGFP produced a homogeneous background in the cytoplasm. In contrast, the IL-2Rγ/JAK3 complexes at the plasma membrane were not homogeneously distributed but concentrated into punctate, specklelike structures across a wide range of expression levels (Fig. 4A, right panels). The distribution of IL-2Rγ/JAK3-containing speckles was independent of stimulation and their size and positions were remarkably stable at the timescale of minutes (Fig. 4B, and see Movie S1 and Fig. S6). Because TIRF illumination is confined to ∼100 nm proximal to the glass support, these stable structures must be located in the cell cortex very close to the plasma membrane.
In standard confocal images, the speckles exhibited a pointlike aspect. However, subdiffraction imaging at ∼90-nm resolution using STED microscopy revealed vesicular structures with a diameter of ∼300 nm (Fig. 4C). To characterize the dynamic behavior of IL-2Rγ/JAK3-eGFP complexes within these structures, we used FRAP (Fig. 4D). The observed recovery time constants of 45 s (plain membrane) and 66 s (ROI containing a vesicle) are consistent with lateral, intramembrane diffusion of IL-2Rγ/JAK3-eGFP in the plasma membrane, but not with recovery modes relying on fast dynamic exchange between the receptor-bound and cytoplasmic pools of JAK3, thus resembling the dynamics of JAK1 (
). However, the immobile fraction was increased from 45 to 78% within ROIs containing both membrane and vesicle, reflecting the complete failure of fluorescence recovery in the vesicles during the observation time (Fig. 4D). Thus, IL-2Rγ/JAK3-eGFP complexes accumulate in vesicular compartments anchored just underneath the plasma membrane, which are not in diffusive exchange with the plasma membrane pool.
We next determined the signature of endosomal cofactors (Fig. 5A and see Fig. S7A). To avoid signals from perinuclear compartments associated with trafficking of nascent receptor chains, we used JAK3-eGFP as a bonafide marker to unambiguously identify the subset of IL-2Rγ-containing, plasma-membrane-associated endosomes. Congruent patterns of positive signals were observed with the early endosomal markers mRFP-Rab5 (96 ± 2% of speckles showing costaining) and EEA1 (84 ± 8%) as well as mCherry-Rab11 (91 ± 2%), a marker of the recycling compartment. Based on these fractions, at least 80% of all JAK3 positive endosomes must contain both Rab5 and Rab11. In contrast, colocalization with the late endosomal and lysosomal markers mRFP-Rab7, Lamp1, or Lysotracker was <20% (Fig. 5B and see Table S3).
Similarly, when coexpressed, IL-13Rα1 and IL-2Rγ were trafficked into the same compartments, suggesting that type-1 and type-2 components follow the same endocytosis pathway (see Fig. S7B). Thus, in HEK293T cells all three IL-4R subunits reside in a subpopulation of regular early sorting and recycling endosomes stably anchored in the cell cortex, which we accordingly termed cortical endosomes (CEs).
IL-4R subunit internalization and JAK/STAT signal transduction
Coexpression studies showed that, under steady-state conditions, IL-4R subunits accumulate in CEs already in the absence of ligand. To test whether the ligand-occupied receptor follows the same endocytosis route, we used our previously described fluorescently labeled IL-4-A647 probe (
). HEK293T cells overexpressing type-1 IL-4R subunits were loaded with ligand at 4°C, washed, and then fluorescence redistribution followed at room temperature. As long as membrane dynamics remained blocked, the distribution of IL-4-A647 bound to full-length IL-4Rα at the cell surface appeared homogenous (Fig. 6A, top row). However, IL-4-A647 accumulated rapidly in the CEs within several minutes of temperature release (Fig. 6A, bottom rows). A similar redistribution kinetics was observed when expressing the truncated receptor IL-4Rαm266 (see Fig. S8A), consistent with observations in BaF3 cells showing that internalization of ligand does not require signaling active receptors (
Using trisNTA-A647 as a ligand independent tracer (Fig. 2) enabled us to study the effect of receptor occupancy on the trafficking kinetics (see Fig. S8, B and C). The N-terminus is located distant from the IL-4/IL-13 binding site; trisNTA labeling and cytokine binding are therefore orthogonal (
). We tracked internalization of receptors preloaded with trisNTA-A647 by photon counting after temperature release (Fig. 6B). The saturation value for dye loaded via H6-IL-4Rα was only 30% of that loaded by H6-IL-2Rγ. A nonspecific background (9.5 ± 4%) was determined by comparing H6-IL-2Rγ with nontagged IL-2Rγ at time-point 30 min. The uptake kinetics could be fitted by saturating exponentials with time constants of 6.6 ± 0.7 min for H6-IL-2Rγ and 8.5 ± 1.3 min for H6-IL-4Rα, which is qualitatively comparable to the internalization of labeled IL-4 ligand by untagged receptor (Fig. 6A). Importantly, H6-IL-4Rα internalization was independent of IL-4 occupancy (Fig. 6C).
To compare trafficking with IL-4R-mediated JAK/STAT pathway activation, we performed a corresponding pulse-chase experiment: HEK293T cells transiently expressing STAT6 were saturated with IL-4 on ice, washed, and lysed after increasing time-points of temperature release (37°C). The lysates were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and probed for STAT6 phosphorylation (Fig. 6D). The response curve followed a sigmoid shape that could be fitted to a logistic function (t0 = 12.6 min). Note that the onset of the response is slightly delayed with respect to receptor trafficking. The response persisted for >30 min without reaching a plateau.
Finally, we assayed for activated JAK/STAT pathway components. Because antibodies for the STAT6 binding site of activated IL-4Rα were not available, we probed for the adjacent phospho-tyrosine Y497, a docking site for the insulin receptor substrate 1/2 (
). IL-4Rα colocalized with JAK3-positive vesicles already in the absence of ligand. However, pY497 staining was IL-4-dependent and preferentially colocalized at the vesicles with both receptor subunits (Fig. 6E). Recalling the concept of signaling endosomes established for several other pathways (
), these observations suggested that CEs might play a similar role for JAK/STAT signal transduction.
Using IL-4R as an example, we report, to our knowledge for the first time, lateral affinities of receptor subunit recruitment for members of the hematopoietic superfamily of cytokine receptors in the plasma membrane. Although ligand-induced dimerization was detected for the full-length IL-4Ra chain, a quantitative Kd determination requires the use of a truncated IL-4Rα construct that shows the same trafficking behavior as the full-length receptor but exhibits improved plasma membrane partitioning. The construct lacks the STAT6 docking sites required for downstream signaling; our reporter gene assay therefore shows that it is fully competitive toward the endogenous type-2 signaling machinery. Because these experiments were performed under excess of stimulating cytokine ligand, the truncated IL-4Rα construct must compete at the level of dimerization and not for capturing ligands from bulk solution. However, the cytoplasmic tail of the truncated IL-4Rα still contains a functional box-1 motif required for JAK1 binding. While JAK3 was supplemented to saturate IL-2Rγ, receptor dimerization was measured in the presence of endogenous levels of JAK1 and Tyk2, for which the exact receptor occupancy (at IL-4Rα and IL13Rα1, respectively) could not be determined. Further studies will be required to address systematically whether JAKs contribute to the stability of the activated IL-4R complex.
Under this premise, we find that the ranking of Kd,2D values reflects the relative signal potencies of the otherwise identical type-2 receptor complexes induced by IL-4 or IL-13, supporting a direct link between complex stability and signal strength. However, this ranking does not correlate with the affinities of isolated ectodomains in free solution, where the preformed IL-4/IL-4Rα complex recruits IL-2Rγ and IL-13Rα1 ectodomains with a comparably low affinity (0.5–4 μM) (
). It therefore remains unclear how to account for the entropic contributions when translating interactions between soluble forms into the two-dimensional membrane context.
The lateral, intramembrane Kd,2D values measured by FCCS were of the order of several hundred receptors per μm2 and could be only measured in cells expressing correspondingly high surface densities (see Table S2). In our HEK293T model, we verified that IL-2Rγ mRNA is downregulated (Fig. 1A) and that endogenous IL-4Rα protein is below the detection limit (see Fig. S3A). Therefore, a potential background of nontagged receptors, and hence an artificially reduced Kd, could only affect the type-2 configuration for which we nevertheless obtained the highest cross-correlation. Thus, our observation that the affinity of the type-1 complexes is extremely low, is conservative. The fact that the IL-4R type-1 interaction is weak was previously proposed (
). However, with hindsight, these experiments did not distinguish between receptor pools at the cell surface and those within CEs. The absolute Kd values are two orders of magnitude above typical cytokine receptor expression levels (500–5000 molecules per cell with ∼1000 μm2 surface correspond to a density of 0.5–5 receptors per μm2) (
). This discrepancy is striking considering that, even under full ligand occupancy, such affinities would produce only a negligible number of ternary, signaling active IL-4R dimers in the plasma membrane.
The weak affinity for receptor subunit dimerization implies that a transient or locally confined cellular concentration mechanism is required to efficiently drive IL-4R dimerization and establish a robust JAK/STAT response. Here we propose that constitutive internalization might be such a mechanism. Confocal, STED, and TIRF imaging showed that the receptor subunits accumulate in small vesicles stably positioned close to plasma membrane. Marker expression defined these vesicles as a subpopulation of regular early sorting and recycling endosomes that we termed CEs. We then quantified the precise uptake kinetics into CEs for each of the type-1 receptor subunits with rate constants of 6–9 min. As expected from their CE localization in nonstimulated cells, internalization does not require pathway activation because the rate constants of IL-4Rα were unchanged upon ligand binding.
In comparison, the JAK/STAT pathway response is rather slow: the onset of STAT6 phosphorylation is delayed by 1–2 min; the time to reach half-maximum of activation (12.5 min) is of the same order of magnitude as the rate constants for internalization. Even more a surprise, the signal steadily increased far beyond the expected residence time of the complex. IL-4 bound to IL-4Rα has a half-life of ∼5.5 min (
), during the course of the pulse chase experiment (30 min) most of the ligand should be dissociated if the complexes remained at the cell surface. Thus, to maintain such a long-lasting output in response to a transient stimulus, activated receptor complexes must either remain stable or continuously reform within confined subcellular compartments, e.g., in the CEs. Such a scenario is further supported by the preferential localization of activated IL-4Rα chain within CEs.
Recently, it was observed in T-cells that JAK/STAT pathway activation by IL-2Rγ-sharing IL-7R involves partitioning into cholesterol-enriched microdomains, transient cytoskeleton interactions, and the formation of signalosomes in the plasma membrane (
). Although such signalosomes could constitute a potential concentration mechanism for CKR at the plasma membrane, we see no such aggregations in our epithelial HEK293T model. Intriguingly, in the IL-2 context, lipid raft partitioning of IL-2Rγ has also been linked to receptor endocytosis (
). The actin-mediated mode of endocytosis may readily explain the stable localization of CEs within the cell cortex, which may involve a densely packed actin mesh.
Although these cell biological mechanisms have yet to be experimentally verified, our observations raise the possibility that CEs provide a subcellular concentration function for the formation of activated receptor dimers and thus constitute signaling platforms not yet described for this receptor class (
). This mode of activation would intrinsically buffer the JAK/STAT pathway against fluctuating ligand concentrations and lateral heterogeneities associated with a dynamic, microdomain patterning of the plasma membrane.
We thank John J. O’Shea for providing JAK3 constructs, Marino Zerial for providing Rab constructs, Bernard Hoflack for antibodies, and Attila Szanto and Lazlo Nagy for the STAT6-inducible reporter genes. We thank C. Herold for providing a MATLAB script, and K. Crell, S. Herrmann, S. Knappe, S. v. Kannen, R. Perez Palencia, and Bea Scheffer for their committed technical assistance.
R.W. is grateful for receiving a postdoctoral fellowship from the Alexander von Humboldt Foundation (Germany). The use of a Confocor3 was supported by Carl Zeiss AG (Jena, Germany). This work was supported by a Center for Regenerative Therapies Dresden seed grant (to K.K., C.B., P.S., and T.W.) and Deutsche Forschungsgemeinschaft priority program No. TRR67 (to H.G., T.W., and P.S.).
Appendix: Kd Determination from FCCS Data
We estimated the two-dimensional dissociation constant for lateral recruitment of receptor subunits under saturating ligand concentrations. Fitting and background correction returns the fluctuation amplitudes GG(0), GR(0), and Gx(0), each of which reflect the inverse average number of observed particles N in the respective detection volume, Veff (
). The two-component fit resolves the fraction f of molecules contributing to fast diffusion in bulk solution (three-dimensional, e.g., unbound ligand trisNTA-A647) and the fraction (1−f) contributing to the slow diffusion within the membrane plane (two-dimensional, e.g., trisNTA-A647 labeled receptor). The average number of receptors diffusing in the membrane plane is
Note that capital letters (G, R) index the color channels (fitting derived parameters) whereas small letters (g, r) index the type of label (true molecular parameters). We treated the eGFP channel accordingly due to a stable fast fraction with a correlation time in the millisecond range, which was absent in the cross-correlation functions, and therefore unrelated to receptor diffusion. We verified with a positive control (PC, trisNTA-A647 labeled NHis-IL4Rm266-eGFP) that this treatment yields the exact same receptor density when measuring the double-labeled receptors simultaneously in both color channels (NRVeff,G/NGVeff,R = 0.99 ± 0.04; mean ± SE; n = 57). The focal volume in the eGFP channel Veff,G was determined with a diffusion standard (AlexaFluor 488; Life Technologies (= A488) in 10 mM Tris, pH 8, D = 435 μm2/s (
The chromatic mismatch between the eGFP and the A647 channel was derived from the ratio of the slow, membrane-related diffusion times measured for the double labeled PC (τdiff,R/τdiff,G = 1.30 ± 0.076; mean ± SE; n = 57). The autocorrelation amplitudes can be calibrated into two-dimensional receptor densities for both color channels:
We defined CC as the ratio of simultaneously recorded cross- and auto-correlation amplitudes (
To derive binding relevant data, CC values must be rescaled to the accessible CC range defined by intramolecular cross-correlation positive control (PC) and a noninteracting negative control (NC) measured under the same instrumental conditions:
Although we show that ligands are unable to recruit a noninteracting NC (Fas receptor chain), the corresponding cellular CCNC values introduced scatter. We therefore checked numerically that the crosstalk-related contribution to the cross-correlation amplitude (5% of the eGFP fluorescence appears in the A647 channel) was negligible by using a recently developed mathematical formalism (
). With CCNC = 0, the measured and normalized CCn values directly represent the fraction ligand bound,
with respect to the total number of particles labeled in the orthogonal color (not used for normalization: G → r and R → g).
The dissociation constant is given by the law of mass action containing the equilibrium (or steady-state) concentrations of free and bound receptors, which can be expressed by particle numbers contained in a uniform observation volume Veff,
The number of complexes can be derived from the normalized cross-correlation ratio CCn. To avoid mixing up differently sized observation volumes, one has to be consistent with the color channel. For example, using CCR,n gives
and rearranged gives
The true particle numbers Nr as well as Ng now both refer to Veff,R and can be directly derived from fluorescence correlation spectroscopy amplitudes. However, due to chromatic mismatch, one must account for the differently sized observation volumes
which is, finally,
Because the problem is symmetrical, the corresponding expression referring to Veff,G can be easily derived by swapping the indices. Here Cg and Cr represent the true calibrated concentrations (receptor densities) as derived from the autocorrelation measurements.
The distinction between Kd,G and Kd,R is of a purely technical nature. In practice, when chromatic mismatch and uneven expression levels do not accidentally compensate each other, CCG,n and CCR,n values from the same binding reaction are expected to be different. Using the channel with larger cross-correlation (CCn,max) is expected to be more reliable in terms of stochastic errors. Nevertheless, both ways of determining the Kd should be consistent. As shown in Fig. S4C, the experimental values for Kd,G and Kd,R show a large degree of correlation (slope ∼1). However, the Kd estimation becomes increasingly uncertain when the measured CC value approaches the minimum (here ∼0%) or maximum (here ∼50%) of the experimentally determined, accessible CC range. The reported final Kd values comprise the arithmetic mean between average Kd,G and Kd,R measured for a population of 20–30 cells.
An antagonistic IL-4 mutant prevents type I allergy in the mouse: inhibition of the IL-4/IL-13 receptor system completely abrogates humoral immune response to allergen and development of allergic symptoms in vivo.
The signaling paradigm for cytokine receptors consists of ligand binding to the extracellular domains of cell-surface-exposed receptors. This first even leads to either dimerization or conformational change of a preformed inactive dimer (1,2), which in turn leads to reciprocal activation of Janus kinase proteins (JAK) that are appended to the cytosolic juxtamembrane regions of receptors. Once JAKs become activated they phosphorylate tyrosines on cytosolic domains of receptors and on JAKs themselves.
If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password