pubmed.ncbi.nlm.nih.gov

DASC, a sensitive classifier for measuring discrete early stages in clathrin-mediated endocytosis - PubMed

  • ️Wed Jan 01 2020

DASC, a sensitive classifier for measuring discrete early stages in clathrin-mediated endocytosis

Xinxin Wang et al. Elife. 2020.

Abstract

Clathrin-mediated endocytosis (CME) in mammalian cells is driven by resilient machinery that includes >70 endocytic accessory proteins (EAP). Accordingly, perturbation of individual EAPs often results in minor effects on biochemical measurements of CME, thus providing inconclusive/misleading information regarding EAP function. Live-cell imaging can detect earlier roles of EAPs preceding cargo internalization; however, this approach has been limited because unambiguously distinguishing abortive coats (ACs) from bona fide clathrin-coated pits (CCPs) is required but unaccomplished. Here, we develop a thermodynamics-inspired method, "disassembly asymmetry score classification (DASC)", that resolves ACs from CCPs based on single channel fluorescent movies. After extensive verification, we use DASC-resolved ACs and CCPs to quantify CME progression in 11 EAP knockdown conditions. We show that DASC is a sensitive detector of phenotypic variation in CCP dynamics that is uncorrelated to the variation in biochemical measurements of CME. Thus, DASC is an essential tool for uncovering EAP function.

Keywords: cell biology; computational biology; endocytic accessory proteins; endocytosis; human; systems biology; unbiased classification.

© 2020, Wang et al.

PubMed Disclaimer

Conflict of interest statement

XW, ZC, MM, JN, SS, GD No competing interests declared

Figures

Figure 1.
Figure 1.. Conventional threshold-based cut-off vs. DAS derived metrics.

(A) Schematic of abortive coat (AC) and clathrin-coated pit (CCP) evolving from early clathrin nucleation. (B) Lifetime (τ) and intensity maxima (Imax) characteristics of hypothetical ACs and CCPs. ACs are typically assigned by a user-defined lifetime or Imax threshold. (C) Disassembly risk map D(i,t) represented on a gray value scale indicated by the gradient bar. A representative CCP (blue), AC (red) and outlier trace (OT) (pink) are plotted on the D-map. (D) Distribution of N=215,948 counts of d1 values for WT condition in black solid line. AC group near d1<0 as a subpopulation, and CCP group at d1≈0 as another subpopulation. (E) Distribution of N counts of d2 values. Subpopulations of ACs and CCPs present in two modes. (F) Distribution of N counts of d3 values resolves the small subpopulation of OTs. Shaded area in (D–E) as 95% confidence intervals.

Figure 1—figure supplement 1.
Figure 1—figure supplement 1.. Fluctuation and heterogeneity in the intensity and D traces.

(A) Ten randomly selected traces of eGFP-CLCa intensity at CSs in WT cells. (B) Ten randomly selected traces with the same lifetime τ=20 seconds from the same cells as in (A). (C) D values as time series read from the color map corresponding to the three traces in Fig. 1C in the main text. Color scheme: CCP (blue), AC (red) and OT (pink). (D) 20s cohort of CCP, AC and OTs. Same color scheme as in (C). High background (BG) in dashed line observed in OTs.

Figure 2.
Figure 2.. DASC resolves behaviorally distinct ACs and CCPs.

(A) k-medoid classification in three-dimensional feature space (d1,d2,d3), where normalized probability densities ρ-(d1,d2), ρ-d2,d3 and ρ-(d1,d3) are shown as three landscape plots. ρ- values are scaled according to gray bar. Examples of CCPs (blue), ACs (red) and OTs (pink) concentrate near the maxima of ρ-. (B) Lifetime distributions of all CCPs (nCCP traces in blue/solid), ACs (nAC traces in red/dashed), OTs (nOT traces in pink/dotted) and all traces (nCCP+nAC+nOT traces in black/solid). Gray region shows lifetime overlap between CCPs and ACs. (C) Imax distributions and overlap. Color scheme same as in (B). Gray region shows Imax overlap between CCPs and ACs. (D) i. DAS plot: ρ-(d1,d2) contour map (values indicated by ‘rainbow’ color bar) with modes for CCPs and ACs indicated by circle and diamond, respectively. Ten representative CCPs and ACs (blue dots) from the lifetime overlap in (B) close to the modes are projected onto d1-d2 coordinate. Traces of the representative CCPs (ii) and ACs (iii) from i. (E) Same as (D) for the representative CCPs and ACs from the Imax overlap in (C). (F) Five repeats of comparing CCP%=nCCP/(nCCP+nAC+nOT)×100% between 12 and another 12 movies of siControl cells imaged on the same day. A total of 24 movies were randomly shuffled to obtained 12-12 pairs. (G) Normalized probability of given p-values using Wilcoxon rank sum test. 1. Blue asters: 1000 repeats of the same shuffle in (F); 2. Red solid line: exact solution to 12-12 random sample test; 3. Light-blue dashed line: simulation of 1000 tests between 12 normally distributed random numbers (NRN) with mean μ=0 and standard deviation σ=1 and another 12 NRNs with μ=0.5 and σ=1. Shaded area as 95% confidence interval obtained from 10 repeated simulations.

Figure 2—figure supplement 1.
Figure 2—figure supplement 1.. ‘Elbow’ method for estimating optimal cluster number.

Total distance from all the traces to their belonging clusters’ centers are plotted for choosing k=1…7 cluster number in k-medoids clustering. The folloing cell types are tested: (i) H1299, (ii) A549, ARPE with (iii) CLCa labeled by mRuby and (iv) α subunit labeled by eGFP, (v) HPV-RPE, and (vi) SKML.

Figure 2—figure supplement 2.
Figure 2—figure supplement 2.. Ambiguity of traces at the hard boundary between AC and CCP populations generated by k-medoids clustering.

(A) All CCPs (blue) and ACs (red) projected in d1-d2 space. Yellow circles represent ‘edge’ traces that fall between the centers of CCP and AC clusters. (B) Lifetime distribution and (C) maximal intensity distribution of AC, CCP and edge traces. (D) 20 second cohorts of (i) CCP and (ii) edge traces. (E) Phenotype of siCALM in CCP% evaluated with (i) and without (ii) edge traces.

Figure 3.
Figure 3.. Validation of DASC.

(A) DAS plots showing ρ-(d1,d2) contour as ‘rainbow’ color map and color bar for αAP2(WT) cells (i) and αAP2(PIP2-) cells (ii). (B) DAS difference plot (difference in d1d2 distribution) of αAP2(PIP2-) minus αAP2(WT) cells as contour in ‘heat’ map. (C) Comparison of DASC-derived metrics for CCP dynamics in αAP2(WT) vs αAP2(PIP2-) cells, (i) CS initiation rate and (ii) CCP% in αAP2(WT) and αAP2(PIP2-) cells. Dots represent jittered raw data from individual movies, box plots show mean as red line and 95% and 1 standard deviation as red and blue blocks, respectively (see Materials and methods). (iii) CCP lifetime distribution of αAP2(WT) vs αAP2(PIP2-) cells. (iv) Imax distribution of ACs in αAP2(WT) vs αAP2(PIP2-) cells. (D) 20 second cohorts from dual channel movies of CLC-mRuby (magenta, solid) and α-AP2-eGFP (green, dashed) for CCPs (i) and ACs (ii). (E) Time derivative of CLC-mRuby (i) and α-AP2-eGFP (ii) intensities for the first 10 seconds in the dual channel cohorts of CCPs and ACs in (D). (F) Time course of invagination depth Δz(t)/h (TIRF characteristic depth h=115nm) for CCPs (blue) and ACs (red) measured by Epi-TIRF. Statistical analysis of the data used the Wilcoxon rank sum test. *** p-value < 0.001, ** p-value < 0.01, * p-value < 0.05, n.s. (non-significant) p-value > 0.05. Shaded area indicates 95% confidence interval for all plots.

Figure 3—figure supplement 1.
Figure 3—figure supplement 1.. Single frame from movies of αAP2(WT) and αAP2(PIP2-) cells.

Note that CCPs in the αAP2(PIP2-) cells are much dimmer potentially altering the ability to detect valid CS initiation events.

Figure 3—figure supplement 2.
Figure 3—figure supplement 2.. 10 s (i–ii) and 30 s (iii-iv) cohorts of CCPs and ACs obtained from CLC-mRuby and α-AP2-eGFP movies.
Figure 3—figure supplement 3.
Figure 3—figure supplement 3.. DASC combined with EPI-TIRF approach reveals CME invagination kinetics.

(A) Schematic of CCP in EPI and TIRF microscopy at 0, intermediate and maximal invagination depth (∆z). ‘+’ as starting point and ‘*’ as center of mass of CCP during invagination. TIRF characteristic depth h=115nm. (B) i. 10s cohorts of CCPs (blue) and ACs (red) in TIRF channel (solid lines) and EPI (dashed lines). ii. Δz(t)/h time course of CCPs (blue) and ACs (red) derived from EPI-TIRF cohorts in i, Δzmax indicated as dark dot. (C) i-ii Same as (B) for 30s cohorts and Δz(t)/h. (D) i-ii Same as (B) for 40s cohort and Δz(t)/h but without ACs, which are extremely rare in the 40s cohort (Fig. 2B). (E) Maximum intensity Imax distribution of CCPs (blue) and ACs (red). Dashed lines indicate 50<Imax<100. (F) Δz(t)/h time course of the subset of CCPs (blue) and ACs (red) exhibiting 50<Imax<100. (G) Same as (F) but for all CCPs and ACs. Shaded area as 95% confidence interval (Materials and methods).

Figure 4.
Figure 4.. Stage specific phenotypes detected by DASC compared to transferrin uptake measurement.

(A) Schematic of 4 stages of CME: CS initiation, CCP stabilization, CCP maturation and CCV internalization. Stage 1-3 are quantified by CS initiation rate (CS init. in min−1μm−2), CCP% and CCP lifetime distribution. Bulk assays for transferrin receptor uptake (TfReff) measure CCV formation are not stage specific. CCP rate (min−1μm−2) measures the combination of initiation and stabilization. Effects of siRNA knockdown of CALM, epsin1 and Eps15 on (B) CS initiation rate, (C) CCP%, (D) CCP rate, (E) CCP lifetime distribution and (F) TfReff (internalized over surface bound transferrin receptors, error bars as 95% confidence interval and statistical significance explained in Materials and methods). 20 bootstrapped siControl movies from all experimental days as siControl* are compared to the specific siRNA conditions to obtained the DASC determined phenotypes in (B–D). (G) Probability distribution of -log10p of CCP% for 300 bootstraps. The 3 conditions in (B–D) plus siMock (20 more of bootstrapped siControl movies) are shown. Veridical lines indicates location of n.s. (p>0.05) and *** (p<0.001) significance. See details on data pooling in Materials and methods.

Figure 4—figure supplement 1.
Figure 4—figure supplement 1.. Measurements of transferrin receptor uptake and siRNA knockdown efficiency.

(A) Internalized transferrin receptors (TfRint) after 10 min in arbitrary unit (a.u.) after normalization to control. (B) Surface bound transferrin receptors (TfRsuf) (a.u.) after normalization to control. (C) Efficiency of transferrin receptor uptake (TfReff) as ratio of TfRint over TfRsuf after normalization to control. Error bars represent 95% confidence intervals. Statistical significance of TfRint and TfRsuf are obtained using 2-sample t-test. A statistical test for ratios is applied to calculate the significance of TfReff, see Materials and methods. (D) Representative KD efficiency of 11 EAPs shown by western blots.

Figure 5.
Figure 5.. DASC is a sensitive measure of stage-specific defects in CME not detected by bulk measurement of transferrin uptake.

(A) Δr (percentage difference) and p-values in CCP rate by comparing 11 EAP KD conditions plus siMock (Fig. 4G) to bootstrapped siControl*. Both quantities are obtained 300 times through bootstrapping. Colored boxes with black dotted edges, correspond to the p-values in the vertical axis; boxes outlined by magenta edges, correspond to Δr values in the horizontal axis. Red lines represent means (p- and Δr¯) the black regions represent 95% confidence intervals and 1 standard deviations as colored blocks of the 300 bootstrap results. Legend shows the color for each condition. Significance level is indicated by the dashed horizontal lines. (B) Summary of phenotypes of the 11 conditions evaluated by the Δr¯ in CS initiation rate (CS init.), CCP%, CCP rate, and CCP median lifetime (τCCP), and Δr in transferrin receptor uptake: internalized and efficiency (TfRint and TfReff) relative to control. EAP KD sorted from low Δr¯ CCP rate to high. (C) Principle component analysis (PCA). Projection of 6 variable values from 11 conditions in (B) into principle component space. First and second component (Component 1 and 2) account for 65.90% and 21.16% of total variance, respectively. Projection of original variable axes presented as red vectors with blue arrows.

Figure 5—figure supplement 1.
Figure 5—figure supplement 1.. Shifts in CS initiation rate and CCP% upon EAP knockdowns and correlations between all the variables.

(A) (i) CS initiation rate and (ii) CCP% calculated from 300 bootstraps. Same box plot format as in Figure 5A. (B) Correlation matrix of 6 variables in Figure 5A. Diagonal bar graphs showed histogram of individual variable values. Off-diagonal graphs showed pair-wise Pearson linear correlation coefficient. Implemented in Matlab’s function corrplot.

Author response image 1.
Author response image 1.. Evidence of early appearance of Dyn2 independent on clathrin dynamics during CME.

(A) Short and bright burst of Dyn2 (blue box) in genome edited SKML cells expressing Dyn2-eGFP (adapted from Aguet et al., 2013). (B) Similar burst in H1299 cells (adapted from Srinivasan et al., 2018). DASC determined CCP and AC groups are very similar with regard to the intensity of endogenous Dyn2-eGPF in both (C) SKML and (D) A549 cells.

Similar articles

Cited by

References

    1. Aguet F, Antonescu CN, Mettlen M, Schmid SL, Danuser G. Advances in analysis of low signal-to-noise images link dynamin and AP2 to the functions of an endocytic checkpoint. Developmental Cell. 2013;26:279–291. doi: 10.1016/j.devcel.2013.06.019. - DOI - PMC - PubMed
    1. Avinoam O, Schorb M, Beese CJ, Briggs JAG, Kaksonen M. Endocytic sites mature by continuous bending and remodeling of the clathrin coat. Science. 2015;348:1369–1372. doi: 10.1126/science.aaa9555. - DOI - PubMed
    1. Bassik MC, Kampmann M, Lebbink RJ, Wang S, Hein MY, Poser I, Weibezahn J, Horlbeck MA, Chen S, Mann M, Hyman AA, Leproust EM, McManus MT, Weissman JS. A systematic mammalian genetic interaction map reveals pathways underlying ricin susceptibility. Cell. 2013;152:909–922. doi: 10.1016/j.cell.2013.01.030. - DOI - PMC - PubMed
    1. Beacham GM, Partlow EA, Lange JJ, Hollopeter G. NECAPs are negative regulators of the AP2 clathrin adaptor complex. eLife. 2018;7:e32242. doi: 10.7554/eLife.32242. - DOI - PMC - PubMed
    1. Bendris N, Williams KC, Reis CR, Welf ES, Chen PH, Lemmers B, Hahne M, Leong HS, Schmid SL. SNX9 promotes metastasis by enhancing Cancer cell invasion via differential regulation of RhoGTPases. Molecular Biology of the Cell. 2016:1409–1419. doi: 10.1091/mbc.E16-02-0101. - DOI - PMC - PubMed

Publication types

MeSH terms

Substances