Evaluating phase separation in live cells: diagnosis, caveats, and functional consequences

Abstract

The idea that liquid-liquid phase separation (LLPS) may be a general mechanism by which molecules in the complex cellular milieu may self-organize has generated much excitement and fervor in the cell biology community. While this concept is not new, its rise to preeminence has resulted in renewed interest in the mechanisms that shape and drive diverse cellular self-assembly processes from gene expression to cell division to stress responses. In vitro biochemical data have been instrumental in deriving some of the fundamental principles and molecular grammar by which biological molecules may phase separate, and the molecular basis of these interactions. Definitive evidence is lacking as to whether the same principles apply in the physiological environment inside living cells. In this Perspective, we analyze the evidence supporting phase separation in vivo across multiple cellular processes. We find that the evidence for in vivo LLPS is often phenomenological and inadequate to discriminate between phase separation and other possible mechanisms. Moreover, the causal relationship and functional consequences of LLPS in vivo are even more elusive. We underscore the importance of performing quantitative measurements on proteins in their endogenous state and physiological abundance, as well as make recommendations for experiments that may yield more conclusive results.

Keywords: condensate; fluorescence recovery after photobleaching; liquid–liquid phase separation; phase separation.

Figure 1.

Liquid–liquid phase separation is a function of concentration. (A) A schematic of a phase diagram depicting under what set of environmental conditions (temperature, salt concentration, pH, etc.) the system will remain as a single phase or spontaneously form two phases. An increase in the y-axis would represent any environmental change that would weaken monomer interactions, e.g., increasing temperature. The dashed line depicts how the system responds to increasing protein concentration, further illustrated in B and C. (B) For proteins that can phase separate, at a certain critical concentration (c), droplets form. Past this critical concentration, production of more protein increases droplet size but does not change the concentrations in either phase, until eventually the concentrated phase entirely fills the space whereupon the system returns to the one-phase regime (A). (C) An illustration of the processes depicted in A and B as it occurs in the cell—in this hypothetical example, in the nucleus.

Figure 2.

Evidence for LLPS in cells is largely phenomenological. (A) A bar graph quantifying the use of descriptive or phenomenological criteria in the studies from Table 1, separated into experiments that are performed on the endogenous protein (knock-in, KI) compared with those in overexpression systems (OE). The x-axis is the number of proteins from the 33 studies that were claimed to display that evidence. (B) A simulated example of how diffraction-limited fluorescence imaging can obscure fine features. The top row depicts various simulated structures, and the bottom row is the image acquired by the microscope detector. (C) A bar graph quantifying the use of assays which give direct evidence for LLPS in vivo. “Any direct evidence” is any example which demonstrated at least one of the categories of direct evidence. See Table 1.

Figure 3.

Fluorescence recovery is misleading as an assay for LLPS. (A) A schematic of a Fluorescence Recovery After Photobleaching experiment. Fluorescent molecules in the cell are bleached with a strong laser in one spot and the signal is allowed to recover over time. In simple diffusion, as is expected in a liquid like a phase-separated domain, mixing of bleached and unbleached molecules is only governed by diffusion. In the case where some molecules may bind to an immobile element, diffusing molecules will mix first before the bound molecules can unbind and exchange. (B) Binding and diffusion have different impacts on the rate of recovery and extent of signal recovery. There are many methods to analyze FRAP data, the simplest being measuring the half-life of recovery (t_1/2). If the molecule under study has a high rate of diffusion compared to its binding rate, modulating the size of the bleach spot (dashed circles in A) will not affect the recovery (dashed lines). If diffusion is the limiting factor, as predicted by LLPS, then the size of the bleach spot should affect the t_1/2 of the curve. (C) Reported t_1/2 times from the studies in Table 1. Cases where the same protein or protein domain have been measured more than once are indicated by connected lines. A few such examples have been labeled for reference. Bolded circles represent measurements on endogenous proteins while the other measurements are in overexpression conditions.