Plasma Membrane is Compartmentalized by a Self-Similar Cortical Actin Meshwork

A broad range of membrane proteins display anomalous diffusion on the cell surface. Different methods provide evidence for obstructed subdiffusion and diffusion on a fractal space, but the underlying structure inducing anomalous diffusion has never been visualized because of experimental challenges. We addressed this problem by imaging the cortical actin at high resolution while simultaneously tracking individual membrane proteins in live mammalian cells. Our data confirm that actin introduces barriers leading to compartmentalization of the plasma membrane and that membrane proteins are transiently confined within actin fences. Furthermore, superresolution imaging shows that the cortical actin is organized into a self-similar meshwork. These results present a hierarchical nanoscale picture of the plasma membrane.

Popular Summary

Across all biological domains, the cell surface plays several critical roles. In mammalian cells, the surface consists of a lipid membrane that provides a boundary to enclose cell components, controls cellular intake and secretions, and helps transmit environmental signals. For proper physiological function, the proteins within the membrane are segregated and organized into dynamic regions that serve different purposes. At any given moment, however, these proteins move about randomly. Despite such an apparent lack of order, the cell surface organization is tightly regulated, albeit in a fashion that is not well understood. We report a direct visualization of the organization of the cell membrane by the underlying actin cytoskeleton, a complex filament meshwork that spans most of the cell and is one of the main players in membrane organization.

Using superresolution imaging of human embryonic kidney (HEK 293) cells, we simultaneously study the random motion of ion channels on the cell membrane and the dynamic actin meshwork. Actin acts as a transient barrier to protein motion, and it partitions the membrane into compartments of various sizes. Close to the cell membrane, the actin meshwork also forms a statistically self-similar fractal that exhibits similar structural properties at many sizes. This structure enables the compartmentalization of the membrane across multiple length scales.

Direct evidence for the compartmentalization of the cell membrane by an underlying actin meshwork is an important step in mammalian cell biology, and it opens up questions regarding how the cell employs this dynamic organization to respond to outside stimuli and to regulate physiological signals.

Figure 1

Voltage-gated potassium channels Kv1.4 and Kv2.1 undergo subdiffusion in the plasma membrane. (a) Four Kv1.4 representative trajectories obtained by single-particle tracking. (b) Time-averaged MSD (TA-MSD) as a function of lag time $\mathrm{\Delta }$ for 20 individual Kv1.4 trajectories. (c) Ensemble-averaged time-averaged MSD (EA-TA-MSD) over 1312 Kv1.4 trajectories ($n=10$ cells). (d) EA-TA-MSD averaged over 6,385 Kv2.1 trajectories ($n=14$ cells). The dashed lines in panels (c) and (d) are visual guides for linear behavior (free diffusion), i.e., $⟨\overline{{\delta }^2(\mathrm{\Delta })}⟩\sim \mathrm{\Delta }$. Error bars show the standard deviation. (e) Sketch illustrating the construction of turning angles from a particle trajectory. (f,g) Turning angle distributions for Kv1.4 (10 cells, 1312 trajectories) and Kv2.1 (14 cells, 6385 trajectories). Turning angle distributions are constructed for lag times between 20 ms and 1 s. (h) Turning angle distributions for fractional Brownian motion simulations with Hurst exponents 0.3 and 0.4. (i) Turning angle distribution for simulations of obstructed diffusion with obstacle concentrations 33% and 41%, i.e., site percolation. (j) MSD averaged over 3114 $\mathrm{\Delta }\mathrm{C}318$ trajectories ($n=5$ cells). (k) Turning angle distributions for Kv2.1 and $\mathrm{\Delta }\mathrm{C}318$ (5 cells, 3114 trajectories) measured with a lag time of 200 ms.

Figure 2

Cortical actin transiently confines Kv channels. (a) Trajectories of individual Kv2.1 channels (shown in cyan) overlaid on an actin PALM image (shown in red). The scale bar is $2\mu \mathrm{m}$. (b) Enlargements of the areas indicated with yellow arrows in panel (a). The scale bar is 500 nm. The left trajectory shows confinement in a large compartment, the middle one shows hopping between two compartments, and the right one shows confinement in a nanoscale domain. (c)–(e) Mean-square displacements $⟨r^2⟩$ covered by Kv1.4 and Kv2.1 and $\mathrm{\Delta }\mathrm{C}318$ channels in 200 ms as a function of their maximum distance from the nearest actin feature. Error bars indicate standard errors.

Figure 3

Characterizations of actin compartments. (a) Superresolution STORM image of the cortical actin in a HEK cell. The inset shows the conventional TIRF image. The scale bar is $2\mu \mathrm{m}$. (b) Average cross-section profile of 20 filaments aligned by the center of each line. The red line is a Gaussian fit with standard deviation $\sigma =20\mathrm{nm}$. (c) Watershed segmentation (shown in green) of the boxed area overlaid on the STORM image. (d) Compartments determined by watershed are designated with different colors. Scale bars in panels (c) and (d) are $1\mu \mathrm{m}$. (e) Distribution of compartment areas for fixed cells (9 cells, $n=2500$ compartments). Areas are shown in logarithmic scale, and the red line is a log-normal distribution.

Figure 4

Fractality of the cortical actin meshwork. (a) Log-log scatter plot of compartment perimeter vs compartment area. The fitted line corresponds to $L=4.8A^{0.55}$ (Pearson correlation coefficient $\rho =0.98$ in log scales). (b) Representative example of box-counting algorithm in one cell where the exponent yields $d_f=1.75$.