Protein Phase Separation Provides Long-Term Memory of Transient Spatial Stimuli

Abstract

Protein/RNA clusters arise frequently in spatially regulated biological processes, from the asymmetric distribution of P granules and PAR proteins in developing embryos to localized receptor oligomers in migratory cells. This co-occurrence suggests that protein clusters might possess intrinsic properties that make them a useful substrate for spatial regulation. Here, we demonstrate that protein droplets show a robust form of spatial memory, maintaining the spatial pattern of an inhibitor of droplet formation long after it has been removed. Despite this persistence, droplets can be highly dynamic, continuously exchanging monomers with the diffuse phase. We investigate the principles of biophysical spatial memory in three contexts: a computational model of phase separation; a novel optogenetic system where light can drive rapid, localized dissociation of liquid-like protein droplets; and membrane-localized signal transduction from clusters of receptor tyrosine kinases. Our results suggest that the persistent polarization underlying many cellular and developmental processes could arise through a simple biophysical process, without any additional biochemical feedback loops.

Keywords: developmental patterning; optogenetics; protein phase separation; receptor tyrosine kinase signaling.

Figure 1. A mathematical model predicts long-term spatial memory from phase separation

(a) Asymmetric protein clustering occurs as part of polarized intracellular processes. (b) Schematic of simulated experiment where clusters are locally dissolved by a transient stimulus. (c) Still frames from simulation demonstrating the response to the stimulus in b. (d,e) Quantification of the total number of particles (d) and mean cluster size (e) in the stimulated and unstimulated regions during all three stimulation time periods. Mean ± SEM are shown from five independent runs. (f) A simulated photobleaching experiment demonstrates rapid exchange of monomers in and out of clusters. Mean ± SEM for 10 clusters is shown. (g) Modeling suggests that transient, local stimuli can drive persistent asymmetries of dynamic, liquid-like granules. See also Figure S1 and Movies S1–S3.

Figure 2. Developing an optogenetic system for spatial control over liquid droplet disassembly

(a) Constructs used to create the PixELL optogenetic system and (b) schematic of blue light-dissociable intracellular droplets. (c) Representative images of intracellular clusters before and after 450 nm light-induced dissociation. (d) Quantification of photoswitchable clustering during 5 cycles of dissociation and aggregation. Mean ± SEM are shown for 8 representative cells. Images from c are shown as insets to relate intracellular droplet patterns to SNR quantification. (e) Visualization of two PixELL droplet fusion events. (f) Droplet intensity during FRAP experiments indicating photobleaching at t=0 and recovery over 10 min. Mean ± SEM are shown for 5 cells, normalized to initial intensity. See also Figures S2–S3 and Movies S4–S5.

Figure 3. PixELLs exhibit long-term spatial memory of transient stimuli

(a) Schematic and images of spatially-restricted 450 nm light stimulation. Fluorescent images of FUS^N-FusionRed-PixD are shown for cells before, during and after stimulation. (b) Cytoplasmic intensity in regions inside and outside the stimulation mask for 4 cells. Mean ± SEM are shown. (c) Mean cluster size for the cell in a, averaged across 5 clusters inside and outside the stimulation area. (d) Still images showing long-term memory of a nucleus-localized light stimulus. See also Figure S4 and Movie S6.

Figure 4. PixELLs amplify shallow stimulus gradients into all-or-none spatial patterns of droplets

(a) Gradient stimulation of a PixELL-expressing NIH3T3 cell. Fluorescent images of FUS^N-FusionRed-PixD are shown for a representative cell stimulated with a linear gradient of light intensity. (b) Kymograph of maximum FUS^N-FusionRed-PixD fluorescence within each row of the yellow box from a (right), and median blue light intensity measured within the yellow box from a (left). (c) Quantification of the kymograph in b at 35 min, after spatial light pattern is established. A gradual decrease in 450 nm intensity (top panel; blue curve) elicits a sharp, switch-like transition to form bright FUS^N-FR-PixD droplets (bottom panel; red curve). See also Movie S7.

Figure 5. Membrane-localized optoDroplets retain spatial memory of transient stimuli

(a) Schematic of Myr-optoDroplet construct and mode of activation. (b) Still images of Myr-FUS^N-FusionRed-Cry2 for a cell exposed to a transient, local 450 nm stimulus. (c) Quantification of total intensity for membrane regions inside and outside the stimulus mask, respectively. Mean ± SEM are shown for 3 cells. (d) Schematic and still images of Myr-FUS^N-FusionRed-Cry2 localization in the membrane plane for a cell exposed to a local 450 nm stimulus (dashed blue box) followed by global 450 nm illumination. (e) Quantification of total intensity in membrane regions inside and outside the stimulus mask, respectively. Mean ± SEM are shown for 3 cells. See also Movies S8–S9.

Figure 6. Liquid phase separation drives spatial memory in RTK signaling

(a) Schematic showing FGFR1-optoDroplets for inducing RTK clustering and downstream signaling. (b) FGFR1-optoDroplet cells reversibly “cringe” in response to global blue light stimulation. (c) Quantification of change in cell surface area for cell pictured in b. (d) FGFR1-optoDroplet cells retract in response to light, ‘avoiding’ a local light stimulus. (e) FGFR1-optoDroplet cells exhibit persistent local clustering and cytoskeletal contraction even after a switch to global illumination. (f) Quantification of cell surface area within the local-to-global illuminated region (blue box in e) and global-only illuminated region (remainder of cell in e) during local-to-global illumination. See also Figure S5 and Movies S10–S12.