Actin-Regulator Feedback Interactions during Endocytosis

Abstract

Endocytosis mediated by clathrin, a cellular process by which cells internalize membrane receptors and their extracellular ligands, is an important component of cell signaling regulation. Actin polymerization is involved in endocytosis in varying degrees depending on the cellular context. In yeast, clathrin-mediated endocytosis requires a pulse of polymerized actin and its regulators, which recruit and activate the Arp2/3 complex. In this article, we seek to identify the main protein-protein interactions that 1) cause actin and its regulators to appear in pulses, and 2) determine the effects of key mutations and drug treatments on actin and regulator assembly. We perform a joint modeling/experimental study of actin and regulator dynamics during endocytosis in the budding yeast Saccharomyces cerevisiae. We treat both a stochastic model that grows an explicit three-dimensional actin network, and a simpler two-variable Fitzhugh-Nagumo type model. The models include a negative-feedback interaction of F-actin onto the Arp2/3 regulators. Both models explain the pulse time courses and the effects of interventions on actin polymerization: the surprising increase in the peak F-actin count caused by reduced regulator branching activity, the increase in F-actin resulting from slowing of actin disassembly, and the increased Arp2/3 regulator lifetime resulting from latrunculin treatment. In addition, they predict that decreases in the regulator branching activity lead to increases in accumulation of regulators, and we confirmed this prediction with experiments on yeast harboring mutations in the Arp2/3 regulators, using quantitative fluorescence microscopy. Our experimental measurements suggest that the regulators act quasi-independently, in the sense that accumulation of a particular regulator is most strongly affected by mutations of that regulator, as opposed to the others.

Figure 1

Schematic of modeled protein interactions. L is Las17 and F is F-actin. (Blue ovals) Las17. (Gray circles) Arp2/3 complex. (Red circles) Actin monomers. Membrane is green and region where PIP2 is hydrolyzed is dark red. Box A shows the self-recruitment of Las17. Boxes B1–B3 show possible mechanisms for the negative feedback of F-actin on Las17. Box C shows how Las17, Arp2/3 complex, and actin monomers enter the branching mechanism. To see this figure in color, go online.

Figure 2

Schematic of the 3D geometry of the stochastic-growth model. Actin filaments polymerizing from a ring of Las17 push against the membrane, pulling other filaments attached to Sla2 at the membrane back with them. The osmotic pressure is higher in the interior (up arrow), and the force of actin polymerization helps overcome this pressure difference. The turgor pressure is the difference between the interior and exterior osmotic pressures. To see this figure in color, go online.

Figure 3

Oblique snapshot of the stochastic simulation geometry, after 23 s of a wild-type simulation run. Actin filaments are red cylinders, with barbed ends in light green spheres; the membrane is green. The blue disk around the center represents the Las17 region where actin filament branches form. The membrane profile is not explicitly treated by the model but we include an approximation to it to clarify the physical picture. We assumed that the membrane deformation at the center is the average distance from the actin filament pointed ends to the membrane, provided that the number of filaments and the F-actin count exceed the threshold value for force generation (see text). The width of the deformation corresponds roughly to known invagination widths (7). To see this figure in color, go online.

Figure 4

Side-view snapshots of stochastic simulations for different interventions. Color conventions are as in Fig. 3. Each of the rows (a–c) shows the initiation of the simulation, the F-actin peak, disassembly, and near disappearance. Row (d) shows the initial phase and later time points where the F-actin count reaches a steady state. The membrane profile is approximated as in Fig. 3, except that in row (d) we assumed that the actin gel was unable to pull on the membrane. To see this figure in color, go online.

Figure 5

Time courses of L and F from stochastic simulations and experiments (described in detail under Fluorescence Imaging Experiments). (a) Wild-type and las17Δacidic pan1Δacidic mutant, (b) disassembly mutant, sla2Δ mutant, and LatA-treated cells. F-actin count is a measured Abp1 count multiplied by a conversion factor of 8.9 (see text). Model results obtained from 2000 simulation runs, displayed in the same way as the experimental data (see the Supporting Material): plotted points correspond to mode values; error bars are the standard deviation of a distribution of 1000 mode values obtained by bootstrapping. The time courses in this and subsequent figures are aligned with their peaks at time t = 0. Arrows in frame (b) indicate which vertical scale to read. To see this figure in color, go online.

Figure 6

Rate-equation model with branching nucleation. Predicted time courses (solid and dashed lines) of (a) F and (b) L for wild-type (black) and las17Δacidic pan1Δacidic (blue) cells compared with experimental time courses (dots). F is the measured Abp1 count multiplied by a conversion factor of 8.9 (see text). For clarity, the error bars for the experimental data are not indicated here, but they are given in Fig. 5a. To see this figure in color, go online.

Figure 7

(a) Measured time courses of (b) Abp1, (c) Las17, (d) Myo5, (e) Myo3, and (f) Pan1 for wild-type cells and several mutants. ΔA in the legend means Δacidic. Numbers of patches measured (N) in Abp1-GFP are N = 184, 132, 148, 274, 64, and 331, for wild-type, las17ΔApan1ΔA, myo3ΔAmyo5ΔApan1ΔA, las17ΔA, myo3ΔAmyo5ΔA, and pan1ΔA, respectively. Following the same order, in Las17-GFP, N = 197, 202, 181, 151, 88, and 202. In Myo5-GFP, N = 279, 90, 677, 366, 517, and 491. In Myo3-GFP, N = 307, 437, 601, 161, 159, 187, and 627. In Pan1-GFP, N = 206, 263, 183, 347, 151, and 96. In addition, for Myo5-GFP myo5ΔA, N = 1070. Plotted points correspond to mode values; error bars are the standard deviation of a distribution of mode values obtained by bootstrapping (see the Supporting Material). Frame (g) shows representative fluorescence images of GFP-labeled regulators. To see this figure in color, go online.