Multivalent interactions between molecular components involved in fast endophilin mediated endocytosis drive protein phase separation

Abstract

A specific group of transmembrane receptors, including the β1-adrenergic receptor (β1-AR), is internalized through a non-clathrin pathway known as Fast Endophilin Mediated Endocytosis (FEME). A key question is: how does the endocytic machinery assemble and how is it modulated by activated receptors during FEME. Here we show that endophilin, a major regulator of FEME, undergoes a phase transition into liquid-like condensates, which facilitates the formation of multi-protein assemblies by enabling the phase partitioning of endophilin binding proteins. The phase transition can be triggered by specific multivalent binding partners of endophilin in the FEME pathway such as the third intracellular loop (TIL) of the β1-AR, and the C-terminal domain of lamellipodin (LPD). Other endocytic accessory proteins can either partition into, or target interfacial regions of, these condensate droplets, and LPD also phase separates with the actin polymerase VASP. On the membrane, TIL promotes protein clustering in the presence of endophilin and LPD C-terminal domain. Our results demonstrate how the multivalent interactions between endophilin, LPD, and TIL regulate protein assembly formation on the membrane, providing mechanistic insights into the priming and initiation steps of FEME.

Fig. 1. Endophilin undergoes LLPS in a crowded environment.

a Droplets formed by rat endophilin A1 (25 µM) in the presence of 10% PEG. Left, confocal fluorescence image of LLPS droplets of endophilin doped with Alexa 594-labeled endophilin (1 µM); right, transmission image of the droplets. Scale bar 10 µm. b Phase diagram showing endophilin-PEG LLPS system. The filled circles indicate where liquid-like droplets were observed, whereas the open circles indicate no droplet formation. c Representative confocal images of endophilin droplets before, 0 s after, and 156 s after photobleaching at different PEG concentrations. Unlabeled endophilin concentrations used for droplet formation were 150, 50, 25, and 25 µM for 2.5, 5, 7.5, and 10% PEG, respectively, and Alexa 488-labeled endophilin (1 µM) was used for fluorescence imaging. Scale bar 2 µm. d FRAP time profiles to show the recovery rate at different PEG concentrations. Normalized intensities of the bleached area relative to the unbleached intensities are plotted for each time point. Each data point is an average of three repeats performed on different droplets, error bars indicate standard deviation. The solid lines indicate 2-exponential (2.5%, 5%, and 7.5% PEG) or 1-exponential (10% PEG) fits of the recovery profiles. e Bar plot showing the mobile fractions of protein within the droplets obtained from the exponential fits of the recovery profiles. Error bars indicate standard deviations of three independent FRAP profiles. f Domain structures of full-length endophilin, its ΔH0 mutant, N-BAR only mutant, ΔH0-BAR domain mutant, and ΔN-BAR mutant and their phase behavior in the presence of 10% PEG. The open circles and the filled circles indicate observation of no droplet and droplets in 3 out of 3 trials respectively using different preparations whereas the half-filled circle indicates droplets were observed 2 out of 3 trials. g Fluorescence intensity ratio of Alexa 594-labeled endophilin variants (1 µM) from corresponding protein droplets vs. from the bulk solution. The droplets were formed in the presence of 25 µM of the unlabeled protein-variant and 10% PEG. The bar plots represent mean ± standard error of mean (s.e.m.) from three independent experiments (gray circles) where 10–20 droplets were considered per experiment. All P values (two-tailed) were determined by Student’s t test, N = 3.

Fig. 2. Endophilin binding partners partition into LLPS droplets and exhibit regulatory behavior.

a Graphical illustration of endophilin droplet formed by LLPS allowing partitioning of endophilin binding proteins as clients into the condensed phase. b Confocal images showing partitioning of fluorescently labeled TIL (Alexa 488) of β1-adrenergic receptor, PRM7 (Alexa 633) of lamellipodin, LPD^850–1250 (Alexa 647), BIN1 isoform 9 (Alexa 488), FBP17 (Alexa 488) and amphiphysin (Amph1) (Alexa 488) into droplets formed by endophilin (25 µM) in the presence of 10% PEG. The corresponding transmission images show the endophilin droplets. Scale bar 10 µm. In the case of Amph1, an enlarged image is shown in the inset (scale bar 2 µm) for the droplet surrounded by dotted lines to illustrate the peripheral distribution of the protein. c Apparent partition coefficients (K_app) for the clients within endophilin droplets as determined from fluorescence intensities inside and outside the droplets. Bar plot represents mean ± s.e.m. from three independent experiments (gray circles) where 10–20 droplets were considered per experiment. d FRAP profiles of the client proteins partitioned within endophilin droplets formed in the presence of 2.5% and 10% PEG. Each data point is an average of three repeats performed on different droplets, error bars indicate standard deviation. The solid lines indicate exponential fits of the recovery profiles. e Effect of client proteins (10 µM) on the endophilin-PEG phase boundary (10% PEG). The open circles indicate no droplets and the filled circles indicate liquid-like droplets were observed 2 out of 2 independent trials whereas the half-filled circles indicate droplets were observed 1 out of 2 trials. f Distribution of Amph1-Alexa 488 (200 nM) and BIN1-Alexa 488 (200 nM) within endophilin droplets (25 µM endophilin, 4% Alexa 594 labeled) formed in the presence of 10% PEG. Scale bar 2 µm. g Fluorescence intensity profile along the dotted white line showing Amph1-Alexa 488 fluorescence intensity is higher along the edges of the droplet whereas BIN1-Alexa 488 intensity is homogeneous within the droplet. h Confocal images of droplets formed in the presence of 25 µM of endophilin, 10% PEG, and various concentrations of Amph1 and BIN1. Scale bars 5 µm. All experiments were performed in 20 mM HEPES buffer, 150 mM NaCl, 1 mM TCEP, 10% (w/v) PEG, pH 7.4, and at room temperature.

Fig. 3. Endophilin undergoes LLPS through multivalent interactions with proline-rich motifs in the absence of PEG.

a Cartoon diagram illustrating that multivalent interaction between SH3 domains of dimeric endophilin and a multiple PRM containing ligand can drive LLPS. b Confocal fluorescence images (top) and transmission images (bottom) of droplets formed by the TIL/endophilin, PRM7/endophilin, and LPD^850–1250/endophilin system. The TIL/endophilin and PRM7/endophilin droplets were formed in the presence of 100 µM of endophilin and 100 µM of either TIL or PRM7 with 1 µM of either TIL-Alexa 488 or PRM7-Alexa 633. The LPD^850–1250/endophilin droplets were formed by mixing 20 µM of LPD^850–1250 and 60 µM of endophilin and contained 1 µM of LPD^850–1250-Alexa 647. Scale bars 20 µm. c Confocal images and intensity profiles from a representative FRAP experiment on a TIL‒endophilin droplet to monitor the mobility of both endophilin (Alexa 594) and TIL (Alexa 488). Recovery time constants (t_1/2) for TIL (green) and endophilin (red) were determined from single-exponential fits (solid lines) of the FRAP data and have been reported as mean±s.d. of three independent FRAP experiments. d, e FRAP studies on PRM7/endophilin and LPD^850–1250/endophilin droplets in the presence of endophilin-Alexa 488, PRM7-Alexa 633, and LPD^850–1250-Alexa 647. Recovery time constants (t_1/2) for PRM7, LPD^850–1250 (cyan), and endophilin (green) have been reported as mean±s.d. of three independent FRAP experiments. Scale bars 2 µm (c–e). f–h Phase diagrams for TIL/endophilin (f), PRM7/endophilin (g), and the LPD^850–1250/endophilin (h) LLPS systems. The red dotted line represents the axis of a 1:1 mixing ratio of both proteins. All experiments were performed in 20 mM HEPES buffer, 150 mM NaCl, 1 mM TCEP, pH 7.4, and at room temperature.

Fig. 4. Endophilin forms clusters on the membrane in the presence of multivalent binding partners.

a Cartoon representation of the interactions between endophilin and proline-rich-motifs of GPCR-TIL and LPD C-terminal domain on cell membrane. b An in vitro model system that has been developed to mimic the endophilin–LPD–GPCR interactions using solid-supported bilayers (SSBs) with conjugated either PRM7-His₆ or LPD^850–1250-His₆ and TIL-His₆ via Ni²⁺-NTA-lipids (right). c Confocal images showing distributions of TIL-Alexa 488 (top), PRM7-Alexa 633 (middle), and LPD^850–1250-Alexa 647 (bottom) on SSBs composed of Ni²⁺-NTA lipid and DOPC (1:99). Images were recorded after incubating the functionalized SSBs with 0, 1, and 2.5 µM endophilin for 30 min. Scale bar 2.5 µm. d Merged images of endophilin-Alexa 594 channel with TIL-Alexa 488 channel (top), PRM7-Alexa 633 channel (middle), and LPD^850–1250-Alexa 647 channel (bottom) in the presence of 2.5 µM endophilin. Scale bar 2.5 µm. e Intensity profiles along the dashed yellow lines shown in c, d showing that clustering of TIL, PRM7, and LPD^850–1250 occurred in the presence of endophilin and endophilin itself colocalized with TIL, PRM7, and LPD^850–1250 in the clusters. f Radially averaged normalized autocorrelation function (G(r)) demonstrating the degrees of clustering in the TIL (left), PRM7 (middle), and LPD^850–1250 (right) channels at 0–2.5 µM of endophilin. The auto-correlation function determines the probability of finding a fluorescent pixel at a given distance r from a center pixel. Solid lines represent the fitting of the auto-correlation plots to a single-exponential function, G(r) = A e^−r/R, to express the extent of clustering in terms of a correlation length (R). All experiments were performed in 20 mM HEPES buffer, 150 mM NaCl, 1 mM TCEP, pH 7.4, and at room temperature.

Fig. 5. TIL partitions into pre-existing endophilin–LPD clusters on the membrane and further enhances clustering.

a Confocal images of SSBs with conjugated His₆-tagged TIL (Alexa 488) and PRM7 (Alexa 633) in the presence and absence of endophilin (10% Alexa 594 labeled). Scale bar 2.5 µm. b Fluorescence intensity profiles for the images along the dashed yellow lines shown in a demonstrating the extent of co-localization of TIL and PRM7 within the clusters. c Radially averaged normalized autocorrelation functions and its single exponential fits (solid lines) demonstrating the clustering in the TIL, PRM7, and endophilin channels before (left) and after (right) addition of endophilin. d Protein distribution on the SSBs containing tethered PRM7 and endophilin before and 15 min after addition of TIL (50 nM). e Fluorescence intensity profiles for the images along the dashed yellow lines shown in e showing co-localization of TIL and PRM7 into the clusters. f Extent of clustering in the TIL, PRM7 and endophilin channels before (top) and after (bottom) addition of TIL quantified by radially averaged autocorrelation function and its single-exponential fits. g Distribution of tethered LPD^850–1250 and endophilin on SSBs before (top) and after (bottom) addition of TIL (50 nM). h Fluorescence intensity profiles along the yellow dashed lines shown in g. i Radially averaged auto-correlation functions with fits to show the extent of clustering in endophilin, LPD^850–1250, and TIL channels before (top) and after (bottom) addition of TIL. j Left, cross-correlation functions to compare the extent of co-clustering between pre-existing PRM7 and endophilin on a bilayer before (gray) and after (cyan) addition of TIL. The extent of co-clustering between the added TIL with the pre-existing PRM7 is shown on the right. Solid lines are the guide for the eye. k Cross-correlation analysis for the LPD^850–1250/endophilin on a bilayer before (gray) and after (cyan) addition of TIL. On the right, the extent of co-clustering between the added TIL and the pre-existing LPD^850–1250 is shown. All experiments were performed in 20 mM HEPES buffer, 150 mM NaCl, 1 mM TCEP, pH 7.4, and at room temperature.

Fig. 6. LPD drives membrane adhesion in the presence of endophilin and might stabilize membrane bud formation at FEME sites.

a Time-lapse images of GUVs tubulated in the presence of endophilin before and after the addition of LPD^850–1250 showing recruitment of LPD causes contraction of generated tubules. Top panel, GUVs enriched with endophilin (1 µM, 10% Alexa 488 labeled); middle panel, binding of LPD^850–1250 on the GUVs; bottom panel, endophilin and LPD^850–1250 channels merged. Scale bar 10 µm. b Enlarged lipid channel image from the box enclosed region of a to show long tubules present in the absence of LPD (yellow arrow) and small structures (red arrow) formed after contraction of tubules upon LPD addition. c Quantitative analysis showing changes in tubule length correlates with the extent of LPD recruitment onto GUV membrane. To approximately estimate the tubule length, a radial average of the intensities of the area outside the GUV radius (r > r_GUV) was estimated in the lipid channel. A sum of average intensities along the radius ${\sum }_{r>r_{\mathrm{GUV}}}{Ī}_r$ from a single frame was plotted against time (red circles). To estimate LPD binding over time, the fluorescence intensity of Alexa 647 was estimated from the GUV surface and plotted against time (blue squares). d Representative TEM image showing membrane tubules generate in the presence of endophilin. e Representative images showing that LPD in the presence of endophilin cause membrane adherence. Membrane adherence leads to LUV–LUV adhesion without tubulation (left) and also to tubules adhered along their length (right). All experiments were performed in 20 mM HEPES buffer, 150 mM NaCl, 1 mM TCEP, pH 7.4, and at room temperature. f A proposed model showing how endophilin and LPD multivalent interactions might stabilize negative membrane curvature (dark gray area of the membrane) at the neck region of a membrane bud during FEME.