Machine-Learning-Based Analysis in Genome-Edited Cells Reveals the Efficiency of Clathrin-Mediated Endocytosis

Abstract

Cells internalize various molecules through clathrin-mediated endocytosis (CME). Previous live-cell imaging studies suggested that CME is inefficient, with about half of the events terminated. These CME efficiency estimates may have been confounded by overexpression of fluorescently tagged proteins and inability to filter out false CME sites. Here, we employed genome editing and machine learning to identify and analyze authentic CME sites. We examined CME dynamics in cells that express fluorescent fusions of two defining CME proteins, AP2 and clathrin. Support vector machine classifiers were built to identify and analyze authentic CME sites. From inception until disappearance, authentic CME sites contain both AP2 and clathrin, have the same degree of limited mobility, continue to accumulate AP2 and clathrin over lifetimes >∼20 s, and almost always form vesicles as assessed by dynamin2 recruitment. Sites that contain only clathrin or AP2 show distinct dynamics, suggesting they are not part of the CME pathway.

Figure 1. Authentic CME sites were identified and characterized by co-labeling of clathrin and AP2

A) Schematics showing TALEN and donor plasmid design for AP2M1 genome editing. B) Epifluorescence and TIRF images of the MDA-MB-231 cell line genome-edited to express CLTA-GFP and AP2M1-RFP. Scale bar = 5µm C) Magnified views of the boxed region in the TIRF images in B). Scale bar = 1µm. D) Histogram of lifetimes for fluorescent spots containing AP2 and clathrin, AP2 alone or clathrin alone, determined from time-lapse TIRF movies with 200ms exposure times, 2s intervals and 4 min duration (19 cells from 3 experiments). E) Average fluorescence intensity profiles for colocalized AP2 and clathrin tracks belonging to different AP2 lifetime cohorts: 20s<t≤40s, 40s<t≤60s and 60s<t≤80s (n = 58, 37 and 16 for each cohort). The tracks are aligned so that the last detected clathrin spot is aligned to time zero. F) A representative montage showing colocalized AP2-RFP and clathrin-GFP. Circular areas of 7-pixel diameter (756nm) are shown for montages.

Figure 2. Support Vector Machine (SVM) classifiers were calculated for identification of authentic CME sites

A-C) Projections of 4-dimensional (4D) features used for SVM classification onto a 2D plane. The line represents a 2D slice of an SVM hyperplane (classifier) at the mean of the other two features orthogonal to the shown plane. Purple squares represent sites with colocalized clathrin and AP2, and the yellow squares are sites with only AP2 or clathrin. A) A projection of features for AP2 sites to a plane showing feature 1 (track lifetime) and 2 (maximum intensity). B-C) A projection of features for clathrin sites to a plane showing two features (1 and 4 in B; 1 and 2 in C). Projections to all possible combinations of two features are shown in Fig. S2.

Figure 3. Authentic CME sites mature over time and recruit dynamin

A) TIRF images of an MDA-MB-231 cell genome-edited to express DNM2-GFP/AP2M1-RFP. Scale bar = 5µm B) Histogram showing track lifetimes for colocalized or non-colocalized AP2 and dynamin2 tracks from time lapse TIRF movies with 200ms exposure times, 2s intervals and 4 min duration (11 cells from 3 experiments). The average lifetimes are shown in Table S2. C). Average profiles for associated AP2-RFP and dynamin2-GFP tracks belonging to different AP2 lifetime cohorts: 20s<t≤40s, 40s<t≤60s and 60s<t≤80s (n = 51, 79 and 26 for each cohort). The tracks are aligned to the time at which the dynamin2 maximum intensity was reached. D) Representative montage for AP2-RFP and dynamin2-GFP recruitment and disappearance. Circular areas of 7-pixel diameter are shown for montages. E) Bar graphs showing the fraction of predicted authentic and false CME sites that recruited dynamin2. F) Average intensity profiles for AP2-RFP in different lifetime cohorts: 0–10s, 10–20s, 20–30s and 30–40s. AP2 tracks followed by dynamin2 recruitment showed steady increases in intensity as a function of lifetime (n=6, 10, 38 and 47 for each cohort). AP2 tracks not followed by dynamin2 recruitment showed similar maximum intensities even if they had different lifetimes (n=71, 22, 9 and 5 for each cohort). The tracks were aligned so that the first time point of spot detection is zero. Intensity profile data for all tracks are displayed in Fig. S4A.

Figure 4. Distinct dynamics of authentic and false CME sites

A) Clathrin and AP2 tracks obtained from 30s-long streaming images of AP2-RFP/clathrin-GFP cells. The area shown is 5.4×5.4µm². B) MSD curves for sites containing both clathrin and AP2, and sites containing only clathrin or AP2. For sites containing both clathrin and AP2, the MSD for newly appearing sites and for pre-existing sites are shown separately. CL stands for clathrin. C) Example of an incipient CME site with colocalized AP2 and clathrin. The site shows very little movement. Ci) Tracking results overlaid with maximum-intensity projections of the TIRF images of a nascent CME site marked by AP2 (left) and clathrin (right). Cii) Montages of AP2 and clathrin for the same CME site. D) TIRF images of an MDA-MB-231 cell genome-edited to express AP2-GFP and AP2-RFP simultaneously. From the left, each panel shows the original RFP TIRF image, the original GFP TIRF image, a false-colored overlay of the two original TIRF images, and a false-colored overlay of the binary image showing the spot detection in each channel. The spots were detected using the ICY spot detection module (Chenouard et al., 2013). Circles in each panel show the position of the RFP or GFP spots that did not colocalize with a spot in the other color. E) Bar graphs showing the fraction of AP2-RFP spots containing AP2-GFP signal as a function of the lifetime. The height encompassing both yellow and green bars shows percentage of the AP2-RFP spots colocalized with AP2-GFP as a function of their lifetimes (18 cells, 4 experiments). The green bars represent predicted CME sites and the yellow bars represent predicted false CME sites.

Figure 5. Models for authentic and false CME sites at the plasma membrane

A) The false CME clathrin population includes single triskelia and cytoplasmically derived CCVs that visit the TIRF illumination field. B) The non-CME AP2 population consists of AP2 molecules that associate with the PM but quickly dissociate because they are not stabilized by other endocytic proteins. C) Authentic CME sites are initiated when clathrin joins AP2 on a PIP2-enriched part of the PM. Authentic CME sites develop as stable loci containing both AP2 and clathrin, and they mature into larger sites over time to produce CCVs (productive events). AP2 molecules are drawn as circles of different size in different panels for clarity.