Imaging vesicle formation dynamics supports the flexible model of clathrin-mediated endocytosis

Abstract

Clathrin polymerization and changes in plasma membrane architecture are necessary steps in forming vesicles to internalize cargo during clathrin-mediated endocytosis (CME). Simultaneous analysis of clathrin dynamics and membrane structure is challenging due to the limited axial resolution of fluorescence microscopes and the heterogeneity of CME. This has fueled conflicting models of vesicle assembly and obscured the roles of flat clathrin assemblies. Here, using Simultaneous Two-wavelength Axial Ratiometry (STAR) microscopy, we bridge this critical knowledge gap by quantifying the nanoscale dynamics of clathrin-coat shape change during vesicle assembly. We find that de novo clathrin accumulations generate both flat and curved structures. High-throughput analysis reveals that the initiation of vesicle curvature does not directly correlate with clathrin accumulation. We show clathrin accumulation is preferentially simultaneous with curvature formation at shorter-lived clathrin-coated vesicles (CCVs), but favors a flat-to-curved transition at longer-lived CCVs. The broad spectrum of curvature initiation dynamics revealed by STAR microscopy supports multiple productive mechanisms of vesicle formation and advocates for the flexible model of CME.

Fig. 1. STAR microscopy reveals the dynamics of CCV formation in living cells.

a Principles of STAR microscopy. The evanescent field decay is wavelength dependent, and STAR is based on the changing fluorescence ratio of two spectrally separated fluorophores. STAR imaging of CLCa dual-tagged with iRFP713 and EGFP can be used to study the curvature of clathrin-coated structures. b Current models describing formation of clathrin-coated vesicles (CCVs): i, Constant Curvature Model (CCM); ii, Flat-to-curved transition (FTC); iii, Flat clathrin lattice (FCL) accumulation and dispersion. c 5 μm silica microsphere dual-tagged with Alexa488 and Alexa647, imaged in epifluorescence (E) and TIRF (T), T647/T488 ratio, and theoretical (Theor—blue) and measured (Meas—red) z-position as a function of distance from the center of the bead, mean ± SEM from 31 beads. d Cos-7 cell expressing CLCa-iRFP713-EGFP (STAR) probe imaged simultaneously with 488 and 647 TIRF excitation, arrows indicate example clathrin accumulations, scale bar = 20 μm. Representative of 13 cells and five independent repeats. e Western-blot analysis of STAR probe overexpression in Cos-7 cells. Representative blot from two independent replicates. f Plot of transferrin internalization over time for WT (black) and STAR probe expressing (magenta) Cos-7 cells. Circles and triangles indicate means from three independent replicates at 0, 5, 10, 15, and 20 min for WT Cos-7 (n = 93, 93, 93, 93, 93 cells) and CLCa-STAR OE (n = 89, 91, 93, 86, 92 cells). Black and magenta lines represent means ± SD. There was no main effect of STAR probe expression on TF-568 uptake by two-way ANOVA [Probe, Stimulation]—[F (1, 4) = 0.7909, p = 0.42; F (2.132, 8.526) = 189.5, p < 0.0001]. Šídák’s multiple comparisons test, details in Supplementary Table 1. g Kymograph showing of clathrin accumulation (cyan) and curvature formation (fire), full arrow indicates vesicle formation, dashed arrow indicates flat clathrin lattice assembly. The curvature channel is translated 3 pixels down for better visualization, scale bar = 5 μm. h Quantitative GFP (cyan) and iRFP713 (magenta) intensity and ∆z (orange) traces for representative CCV and FCL, darker shaded band—SD of detected signal. Lighter cyan, magenta, and orange indicate background mean and 2*SD above background, red dashed boxes indicate when GFP signal is significantly higher than threshold. a, b were created with BioRender.com.

Fig. 2. STAR microscopy identifies curved and flat de novo diffraction-limited clathrin accumulations in Cos-7 cells.

a Automated high throughout analysis of live-cell data using CMEanalysis. CLCa-STAR lifetimes in Cos-7 cells separated as positive (left) and negative (right) ∆z, mean ± SEM (EGFP—Cyan, iRFP713—magenta) (CCVs, n = 1225 events; FCLs, n = 328 events; from 11 cells and three independent experiments). b CLCa-STAR intensity (cyan and magenta) and ∆z (fire) for representative CCVs and FCLs across the lifetime cohorts. Scale bar = 1 μm. c Example CCV and FCL signal traces for the events shown in b. CLCa-STAR intensity: EGFP (Cyan), iRFP713 (magenta) ∆z (Orange), darker shaded band represents the SD of detected signal, lighter colors indicate background mean and 2*SD above background. d Histogram of lifetime distribution of Curved CCS and Flat CCS per µm², per minute (mean ± SEM). e Cumulative frequency of Curved CCS and Flat CCS per µm², per minute. Data was not normally distributed as tested by one-sided Shaphiro–Wilk test, black line—median, median for Curved CCS = 0.0065, median for Flat CCS = 0.0015. Medians were significantly different from one another as tested by two-tailed Mann-Whitney test, exact p < 0.0001 (Data in d, e—Curved CCS, n = 1534 events, Flat CCS, n = 414 events, from 13 cells and five independent repeats).

Fig. 3. CCVs start forming through a gradient of clathrin and curvature lifetimes in Cos-7 cells.

a Models of initial curvature generation during CME and representative intensity and ∆z measurements, mean ± SEM (Nucleation—n = 72 events, Constant curvature—n = 46 events, Flat to curved—n = 144 events, from 25 Cos-7 cells and three independent repeats). Red dashed line indicates the time where signal reached over the intensity threshold. b Kymographs for representative events from a (CLCa-STAR, cyan-magenta; ∆z, fire). White dashed line indicates when signal crossed the threshold. Scale bar = 5 μm. c Distribution of beginning of curvature formation in living Cos-7 cells measured as a difference between the time when ∆z reached above threshold and the time when clathrin accumulation reached above threshold (∆z_Beg − CLCa_Beg [s]). d Density scatter plot of beginning of curvature (∆z_Beg − CLCa_Beg [s]) as a function of total CCP lifetime (x axis is logged for better visualization, percentages for quadrants A-I are summarized in f). e, f Summary of distribution of events across three models: Nuc = Nucleation (∆z_Beg − CLCa_Beg < −1s), CCM = Constant Curvature Model (1 s ≤ ∆z_Beg − CLCa_Beg ≤ 4 s), FTC = Flat-to-curved transition (∆z_Beg − CLCa_Beg > 4 s). Data from d was also classified as events between 5 and 20 s as short-lived, 20–50 s as intermediate, and >50 s as longer-lived. Data presents three independent repeats means and SD are reported as a heat map in f. Data was normally distributed (one-sided Shaphiro–Wilk test, p > 0.05 for each lifetime cohort), two-way ANOVA was performed (lifetime cohort—F (2,18) = 93.17, p < 0.0001; PM bending model—F (2,18) = 93.70, p < 0.0001; interaction—F (4,18) = 57.49, p < 0.0001). Tukey’s multiple comparisons test details can be found in Supplementary Table 2, ***p = 0.0005, ****p < 0.0001 (Data in c–f based on n = 1805 events form 25 Cos-7 cells and three independent experiments). a was created with BioRender.com.

Fig. 4. Endogenous CLCa does not explain the variation of curvature formation.

a Standard deviation of STAR measurements from Monte Carlo simulations of vesicle assembly with varying percentages of STAR-probes. b Western-blot analysis of CLCa siRNA silencing and STAR probe expression in Cos-7 cells. Representative of two independent replicates. c Cos-7 cells treated with control or CLCa targeting siRNA and expressing CLCa-STAR. Arrows—clathrin accumulations, scale bar = 20 µm, and kymograph showing clathrin accumulation (cyan) and curvature formation (fire), full arrow—vesicle formation, dashed arrow—flat clathrin lattice assembly. The curvature channel is translated 3 pixels down, scale bar = 5 µm. d Histogram of lifetime distribution of curved and flat CCS per µm², per minute (mean ± SEM). e Cumulative frequency of curved CCS and flat CCS per µm², per minute. Black and magenta lines—mean ± SEM, data was normally distributed, except for Ctrl Flat CCSs by one-way Brown-Forsythe ANOVA F = 27.15 (3.000, 64.66), exact p < 0.0001; medians for: Curved CCSs [Ctrl siRNA; CLCa siRNA] = [0.01362, 0.01498], Flat CCS = [0.003975, 0.003975], ****p < 0.0001. P values adjusted for multiple comparison using statistical hypothesis testing. f Distribution of beginning of curvature formation measured as ∆z_Beg − CLCa_Beg [s] g Density scatter plot of ∆z_Beg − CLCa_Beg [s] as a function of total CCP lifetime. h, i Distribution of events across three models and three lifetime cohorts following the classification form Fig. 3, Nuc = Nucleation, CCM = Constant Curvature Model, FTC = Flat-to-curved transition. Data presents three independent repeats means and SD are reported as a heat map in i. Data was normally distributed, two-way ANOVA was performed (lifetime cohort—[F (2, 18) = 54.82, p < 0.0001; F (2, 18) = 13.89, p = 0.0002]; PM bending model—[F (2, 18) = 118.4, p < 0.0001; F (2, 18) = 69.47, p < 0.0001]; interaction—[F (2, 18) = 118.4, p < 0.0001; F (4, 18) = 24.14, p < 0.0001]). Tukey’s multiple comparisons test details in Supplementary Tables 4 and 5 and the comparison in Supplementary Table 6, ns = not significant, **p = 0.0036, ****p < 0.0001. (Data in d–i—Curved CCSs n = [2655, 2506] events, Flat CCSs, n = [849, 1048] events, from refs. ^, cells and three independent repeats).

Fig. 5. Majority of CCVs form simultaneously with clathrin arrival in HUVECs.

a HUVEC expressing CLCa-STAR, arrows = clathrin accumulations, scale bar = 20 μm. Representative of 9 HUVECs and three independent repeats. b Western-blot analysis of STAR probe expression in HUVECs. Representative blot from three independent replicates. c Kymograph showing clathrin accumulation (cyan) and curvature formation (fire), full arrow = vesicle formation, dashed arrow = flat clathrin lattice assembly. Curvature channel is translated 3 pixels to the right, scale bar = 5 μm. d Quantitative intensity and ∆z traces for CCV and FCL, lighter cyan, magenta and orange = background mean and 2*SD above background, darker shaded band represents the SD of detected signal, red dashed boxes = clathrin signal significantly higher than threshold. e CMEanalysis of CLCa-STAR lifetimes separated on whether ∆z was generated (mean ± SEM, EGFP—Cyan, iRFP713—magenta). f Histogram of lifetime distribution of curved and flat CCS per µm², per minute—mean ± SEM. g Cumulative frequency of curved and flat CCS per µm², per minute, mean ± SD, data was normally distributed, two-tailed unpaired t-test, p = 0.0064, *p < 0.01, means: Curved CCS = 0.00104, Flat CCS = 0.00028. (Data in e–g based on CCVs—n = 458 events, FCLs—n = 143 events, from 8 cells and three independent repeats). h Distribution of beginning of curvature formation measured as ∆z_Beg − CLCa_Beg [s]. i Density scatter plot of ∆z_Beg − CLCa_Beg [s] as a function of total CCP lifetime. j, k Distribution across three membrane bending models. Data from i was classified as in Fig. 3. Data represents three independent repeats, means and SD reported as a heat map in k. Data was normally distributed (one-sided Shaphiro-Wilk test, p > 0.05 for each lifetime cohort) and analyzed by two-way ANOVA (lifetime cohort—F (2,18) = 19.55, p < 0.0001; PM bending model—F (2,18) = 15.55, p = 0.0001; interaction—F (4,18) = 6.712, p = 0.0017). Tukey’s multiple comparisons test details in Supplementary Table 7, ns = not significant, * (CCM[5–19,20 s] vs FTC[5–19,20 s])—p = 0.0278, * (Nuc_(20-50s] vs FTC_(20-50s])—p = 0.0423, ***p = 0.0007 (Data in h–k based on n = 283 events form 9 HUVEC cells and three independent repeats).

Fig. 6. CME occurs through flexible model of CCVs formation, whereby the initiation of curvature and clathrin accumulation are not correlated and a single clathrin-centric model of vesicle formation cannot encompass the heterogeneity of CME dynamics.

Moreover, flexible model of endocytosis is consistent between different cell lines and cargos, and the balance of that model could be shifted by features such as PM tension, cargo size, or local concentration of adaptor proteins (Created with BioRender.com).