Endocytic sites mature by continuous bending and remodeling of the clathrin coat

See allHide authors and affiliations

Science  19 Jun 2015:
Vol. 348, Issue 6241, pp. 1369-1372
DOI: 10.1126/science.aaa9555

Bend me, shape me: Clathrin in action

Endocytic clathrin-coated pits were among the first cellular structures described by electron microscopy over five decades ago. Despite this, the question remains: Does clathrin bind to the membrane as a flat lattice and then bend during coated pit invagination, or does clathrin assemble with a defined curvature as membranes invaginate? Avinoam et al. applied two state-of-the-art imaging approaches to resolve this conflict. They suggest that clathrin assembles into a defined flat lattice early in endocytosis, which predetermines the size of the vesicle. The assembled clathrin coat then rearranges through dynamic exchange of clathrin with the cytosolic pool to wrap around the forming vesicle.

Science, this issue p. 1369


During clathrin-mediated endocytosis (CME), plasma membrane regions are internalized to retrieve extracellular molecules and cell surface components. Whether endocytosis occurs by direct clathrin assembly into curved lattices on the budding vesicle or by initial recruitment to flat membranes and subsequent reshaping has been controversial. To distinguish between these models, we combined fluorescence microscopy and electron tomography to locate endocytic sites and to determine their coat and membrane shapes during invagination. The curvature of the clathrin coat increased, whereas the coated surface area remained nearly constant. Furthermore, clathrin rapidly exchanged at all stages of CME. Thus, coated vesicle budding appears to involve bending of a dynamic preassembled clathrin coat.

Clathrin-mediated endocytosis (CME) is essential for a broad range of cellular processes, including nutrient uptake, signal transduction, synaptic vesicle recycling, and immune responses (1, 2). It initiates with cargo binding by adaptor proteins that recruit clathrin on the plasma membrane, followed by membrane invagination, which leads to dynamin-dependent scission of a coated vesicle (3, 4). Clathrin trimers have a triskelion shape and can assemble into heterogeneous polyhedral coats (5). The relation between coat assembly and membrane invagination is poorly understood. Two alternative models describe the transition from planar membrane to clathrin-coated vesicle (CCV). The first—derived from electron microscopy images showing both relatively flat and invaginated clathrin lattices in cells—suggests that clathrin assembles as a planar lattice that subsequently bends as the membrane invaginates (6) (Fig. 1A). For this to happen, complex rearrangements within the clathrin network must occur during budding. The second model avoids this difficulty by proposing that large, flat clathrin lattices are not precursors of CME and that, at sites of CME, clathrin directly assembles to produce the curved coat as the membrane invaginates (7) (Fig. 1A).

Fig. 1 FM and ET of CCPs.

(A) Schematic representation of the two models for CME. In model 1 (top), clathrin (red) is first recruited to sites of endocytosis and bends during PM (black) invagination. In model 2 (bottom), clathrin assembles during invagination and directly adopts the curvature of the resultant vesicle. θ represents the maximum angle between the PM and the invaginated membrane (purple) (13). (B and C) TIRF microcopy images of representative SK-MEL-2 cells hCLTAEN/hDNM2EN (B) expressing DNM2-GFP and hCLTAEN/TF and (C) 90 s after addition of TF:Alexa488 (green), both expressing CLTA-RFP (movies S1 and S2). Scale bars, 5 μm and, inset, 2 μm. (D) Correlative FM and ET workflow. Multicolor fluorescence image of the prepared EM sample (top). Clathrin (red), dynamin (green), TetraSpec (multispectral) (11, 13). White circles outline the approximate position of fluorescent spots and their corresponding position in a slice through a low-magnification tomogram (middle). Red and green circles (bottom) outline the correlated positions of clathrin and dynamin; radius illustrates correlation accuracy (radius of 92 nm) (13). Two slices through the tomogram are shown. Scale bars, 1 μm and 100 nm. (E) A representative slice through the tomogram showing the polygonal structure of the clathrin coat (fig. S2). Scale bar, 50 nm. (F) Representative tomographic slices through CCPs. Scale bar, 100 nm.

To study clathrin-coated pit (CCP) maturation, we used well-characterized genome-edited human cell lines expressing fluorescently tagged clathrin light chain A at endogenous concentrations (8, 9) (hCLTAEN) (Fig. 1, B and C). To distinguish late stages of CME, we used cells that coexpressed endogenously tagged dynamin-2 (9) (hCLTAEN/hDNM2EN) (Fig. 1B). To label CCPs that recruit cargo, we incubated cells with fluorescently labeled transferrin (TF), a constitutive cargo of CME (hCLTAEN/TF) (Fig. 1C). CME events imaged by total internal reflection fluorescence (TIRF) microscopy showed a buildup of clathrin fluorescence that reached a plateau before a burst of dynamin preceded the site’s disappearance, as previously described (8, 10) (Fig. 1, B and C, and movies S1 and S2).

To distinguish between the models of coat assembly, we applied a correlative fluorescence microscopy (FM) and electron tomography (ET) method to locate CCPs precisely and to obtain three-dimensional information about their shape (1113) (Fig. 1D and movies S3 to S5). We obtained, in total, 233 tomographic reconstructions of sites where a clathrin signal was detected at the cell periphery (117 from hCLTAEN/hDNM2EN and 116 from hCLTAEN/TF). In all cases, the fluorescent signal correlated with an endocytic site with a clear membrane coat (13). When the coat lay parallel to the imaging plane, the characteristic polygonal architecture of polymerized clathrin was visible (Fig. 1E and fig. S1). We analyzed 199 CCPs and 27 CCVs corresponding to different stages of CME (Fig. 1F) (13).

To characterize changes in membrane shape, we extracted the membrane and coat profiles and analyzed them computationally (12, 13) (Fig. 2). We determined the invagination depth, tip curvature, and coated membrane area. When a constricted neck was detected at the base of the invagination, we measured neck width and height (13) (Fig. 2B and tables S1 and S2). We further determined the maximal angle (θ) between the adjacent plasma membrane (PM) and the invaginating segment of the membrane (Fig. 2B). Independently of invagination size and which model is correct, θ will increase during invagination from 0° to 90° in a U-shaped invagination before tending to 180° at neck constriction and scission (13) (Figs. 1A and 2B and movie S6). θ is therefore a surrogate for CCP stage and was used to order CCPs from early to late.

Fig. 2 Data extraction and analysis.

(A) An oblique tomographic slice containing the long axis of an invagination (left) and the corresponding profile (right) of the outer surface of the coat (red) and the inner leaflet of the PM (black). (B) Schematic of PM (black) and coat (red) profiles illustrating the measured parameters (13). (C) Correlated position of clathrin (red circles) and dynamin (green circles) determined from multiple correlations relative to individual membrane profiles of dynamin-positive CCPs (light gray) and their average membrane and coat profiles (black) (13). Maximum likelihood is in the center of the inner circle. Contour lines are at 10% and 1% of the maximum likelihood, respectively.

We detected no apparent morphological differences between the cargo-labeled CCPs from hCLTAEN/TF and the CCPs observed in hCLTAEN/hDNM2EN cells (fig. S2) and, unless stated otherwise, we pooled the data from these data sets. Dynamin-positive sites consisted exclusively of deep invaginations with a neck and clustered at late stages of CME (~θ > 90°; depth, 68 to 155 nm) (Fig. 3, A and B, and fig. S3). The high precision of correlation allowed us to localize dynamin to the base of the invagination close to the PM (13) (Fig. 2C). Dynamin-associated necks had widths of up to 121 nm, which greatly exceeded the diameter of assembled dynamin rings in vitro (<30 nm) (14); these findings suggested that dynamin recruitment and neck constriction begin before ring formation.

Fig. 3 Changes in membrane shape during invagination.

Analysis of parameters extracted from the data set (Fig. 2B), invagination depth (A), neck width (B), coated surface area (C), and tip curvature (D), each plotted against the angle θ, which represents the growth stage of a CCP (13). CCPs without a neck (black), with a neck (cyan), and associated with dynamin (red). (A) Invagination depth increases during CCP maturation. The neck begins to form when CCPs exceed ~70 nm in depth at ~θ = 90°. Dashed lines: θ = 90° and depth = 70 nm. (B) Neck width decreases during CCP maturation. Dynamin-positive invaginations appear at the onset of neck constriction. NoN, no neck. (C) The coated membrane surface area of CCPs is not significantly different from that of CCVs (CCPs: 28 ± 13 × 103 nm2, CCVs: 25 ± 11 × 103 nm2; P > 0.2, two-tailed t test). (Inset) A sketch of expected results for the first model (red) and the second model (black) (Fig. 1A). Vesicle surface area: range (gray area) and distribution (green scatter plot, mean in red). (D) During maturation, the tip curvature decreases [measured from fitted circles as in (E)], which shows that CCP curvature increases. At late stages, it reaches the curvature range seen in released vesicles. Vesicle radii: range (gray area), distribution (green scatter plot, mean in red), and inset as in (C). (E) Representative profiles of different stages of invagination; coat (red) and PM (black), with circles fitted to determine tip curvature (dashed line).

If clathrin directly assembles into the curved coat by addition of triskelia during membrane invagination, then the coated membrane area should grow as sites mature. In contrast, we found that the coated membrane area in CCPs did not appreciably change during budding and did not differ significantly between CCPs and CCVs (Fig. 3C and figs. S1C and S2, C and D). This suggests that enough clathrin is recruited before invagination to coat a complete vesicle. Furthermore, if clathrin directly assembles into the curved coat, then the tip curvature of CCPs would remain constant during budding and would be the same as that in the resultant CCVs. In contrast, our analysis showed that tip curvature continuously increased during invagination (Fig. 3, D and E, and fig. S2E). These results show that the curvature of CCPs increases, whereas the surface area remains largely constant. These properties are only consistent with the model where clathrin is initially deposited on a nearly planar membrane followed by subsequent membrane bending.

In order to bend, flat lattices composed primarily of hexagons must acquire pentagons requiring extensive molecular rearrangements and removal of triskelia (7, 15). It has been reported previously that clathrin exchanges at CCPs (16, 17). However, it is unclear if, and at what stages, this occurs during canonical CME. To test this, we performed fluorescence recovery after photobleaching (FRAP) analysis of individual CCPs (13) while simultaneously imaging clathrin and dynamin (Fig. 4A and movie S7). Clathrin fluorescence recovered rapidly, reaching 60% recovery with a half time of ~2 s, both before and after dynamin recruitment (Fig. 4, B and C, and fig. S4). This indicates that clathrin is exchanging at early and late stages of CME and explains why blocking clathrin turnover arrests CCPs at all stages (18).

Fig. 4 FRAP analysis of individual early and late CCPs.

(A) Images of a representative FRAP experiment (top) and montage of the bleached spot showing the pre-, post-, and recovery after photobleaching of clathrin-RFP in hCLTAEN/hDNM2EN SK-MEL-2 cells. Scale bar, 2 μm. (B and C) Mean fluorescence recovery profiles of clathrin when dynamin was absent [early (B)] or present for at least seven frames [late (C)]; error bars indicate SD. MF, mobile fraction; t1/2, half time; and number of sites (n).

Although fluorescence intensity profiles of individual CCPs display considerable heterogeneity in assembly kinetics and overall persistence at the plasma membrane, recent analysis has shown that clathrin displays an initiation and growth phase, followed by a plateau in which fluorescence intensity is at its maximum (10, 19). Our correlated FM and ET data showed that clathrin recruitment occurs before membrane invagination, which suggests that invagination occurs during the fluorescence intensity plateau (Fig. 3C and fig. S5). In yeast, the coat is also recruited before membrane bending, which suggests that this is a conserved feature (12). Numerous EM studies have documented a population of flat clathrin lattices that could potentially evolve into single vesicles (6, 20). It is likely that these lattices correspond to the early CCPs observed here. Subdomains of the larger clathrin lattices that exist in some cell types may also undergo local rearrangements to produce vesicles (20, 21).

Recruitment of clathrin before membrane bending provides a flat, dynamic array as a platform for cargo recruitment. This implies that the membrane to be internalized and the size of the future vesicle are not determined by clathrin geometry during assembly into a curved cage but rather are selected before invagination during cargo recruitment. Rapid clathrin exchange is consistent with a dynamically unstable lattice—dynamic instability is a common property within networks of low-affinity protein interactions (22). It would allow for stochastic abortion of sites that initiate but fail to cross a growth- or cargo-mediated checkpoint (19, 2325) before investing energy in membrane bending. During invagination, further exchange would allow clathrin reorganization and bending of the lattice into a defined cage that requires active disassembly.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Tables S1 to S2

Movies S1 to S7

References (2631)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank members of the Kaksonen and Briggs laboratories for helpful discussion; B. Podbilewicz, J. Ries, and P. Lénárt for critically reading the manuscript; the EMBL electron and advanced light microscopy facilities; and A. Politi and J. Ellenberg for single-site FRAP software. O.A was supported by the Marie Curie Actions COFUND (European Commission Cofunding of Regional, National, and International Programme) [grant no. 229597; EMBL Interdisciplinary Postdoc (EIPOD)]. Computer codes used are available upon request.
View Abstract

Navigate This Article