How Science Works | How Cells Form Droplets Without a Surrounding Membrane

A membrane creates a material barrier containing lipids and proteins. A phase boundary arises because the mixture has different preferred compositions on its two sides. These are different physical arrangements, even when both can concentrate particular molecules in a region.

Imagine a transparent diagram with many small protein symbols scattered through a watery background. In one circular region, the same symbols are more numerous, and several RNA strands are also enriched. Draw no double line around the circle. Label the interior “dense phase” and the surroundings “dilute phase”. Add arrows crossing the edge in both directions.

The arrows matter. A compartment can retain its overall composition while individual molecules enter and leave. Persistent organisation does not require permanent residence of every molecule. At equilibrium, exchange continues without a net change in the amount in either phase.

Both phases can contain substantial water. The dense region is not necessarily a dry lump of protein, and the dilute region is not necessarily empty of the molecules found inside. “Dense” and “dilute” describe relative concentrations for specified components.

Oil separating from water is a useful introductory image because it shows a boundary without a manufactured wall. The analogy stops short of the molecular details. Cellular condensates often contain interacting polymers, many components and energy-consuming reactions. Their composition and material behaviour can be much richer than those of a simple kitchen mixture.

Contents · Next chapter

A protein may have several sites that interact with other molecules. This property is called multivalency. If one molecule can make several contacts and its partners can do the same, an extended network becomes possible. Individual contacts can break and reform while the larger assembly persists.

Not every interaction has to be strong. Many individually modest contacts can collectively favour a concentrated state. The outcome depends on their number, arrangement, chemical character and surrounding conditions. Flexible, intrinsically disordered protein regions can participate, but being disordered is neither a universal requirement nor sufficient proof of phase separation.

An original 2012 study of multivalent signalling proteins demonstrated sharp demixing transitions in defined systems containing interacting macromolecules. Changing interaction valency changed the concentrations associated with the transition. In a system involving N-WASP and its biological partners, the transition was associated with a sharp change in activity toward an actin-nucleation factor.

That experiment connects molecular interaction architecture with larger-scale organisation. It does not establish that adding more interaction sites always improves cellular function. More association can also alter mobility, accessibility or the availability of partners elsewhere.

For the broader molecular background, see How Proteins Work. Here, the additional question is collective: under what conditions do many interacting molecules form coexisting compositions rather than simply a succession of small complexes?

Contents · Previous chapter · Next chapter

Consider a deliberately simplified solution with one phase-separating solute and a solvent at fixed temperature. Under specified conditions, there can be a dilute-phase coexistence concentration and a dense-phase coexistence concentration. Between those overall concentrations, the system can accommodate extra solute by changing how much of each phase is present.

The lower coexistence concentration is often discussed as a saturation concentration. In this simple description, increasing the total amount beyond that boundary does not require the surrounding dilute phase to become proportionally more concentrated. Instead, more material can occupy the dense phase.

This is a model with conditions attached. A cell contains many interacting proteins, RNAs, salts and other molecules. Changing one component may change the compositions that coexist. There may also be chemical reactions, transport and continuing energy consumption, so an equilibrium two-component picture cannot be assumed automatically.

A 2020 experimental study of intracellular phase separation showed why this qualification matters. The investigated endogenous condensates did not display one fixed saturation concentration as individual component concentrations changed. Interactions among unlike protein and RNA components contributed strongly to their behaviour.

Thus, a flat dilute-phase concentration can support a suitable simple model, but its absence does not automatically eliminate every multicomponent phase-separation explanation. Researchers must specify the composition changes and the model being tested. A rule extracted from a two-component diagram cannot silently become a definition that every living condensate must obey.

Contents · Previous chapter · Next chapter

Use an original hypothetical system with a dilute concentration of 2 micromolar and a dense concentration of 20 micromolar. Suppose these coexistence concentrations remain fixed while the total concentration is 5 micromolar. Assume equilibrium, a fixed total volume and a negligible contribution from the interface.

Let f be the fraction of total volume occupied by the dense phase. Conservation of solute gives the volume-weighted average: 5 = 20f + 2(1 − f). Expanding gives 5 = 18f + 2, so f = 3/18 = 1/6. About 16.7% of the volume is dense phase.

This calculation is often called a lever-rule calculation. Its meaning is straightforward: the overall concentration must lie between the two phase concentrations, and the volumes determine how strongly each contributes to the average.

Now increase the overall concentration to 8 micromolar while keeping the model’s two coexistence concentrations unchanged. The equation becomes 8 = 20f + 2(1 − f). Therefore f = 6/18 = 1/3. The dense phase occupies about 33.3% of the volume.

The dense-phase volume doubles, but its concentration remains 20 micromolar by assumption. “More condensed material” therefore need not mean “a more concentrated droplet”. It can mean more total dense-phase volume, distributed among larger droplets, more droplets or both.

The calculation does not predict the number of droplets or how quickly they form. Those questions involve nucleation, growth, coalescence and other dynamics. It also cannot be applied at an overall concentration below 2 micromolar by accepting a negative volume fraction. A negative answer would signal that this two-phase construction is outside its allowed range.

Contents · Previous chapter · Next chapter

In the 5-micromolar example, the dense phase occupies one-sixth of the volume. Does that mean it contains one-sixth of the solute? No: that region has a higher concentration, so it contains more solute per unit volume.

Choose a total volume of one arbitrary volume unit to make the accounting easy. The dense-phase contribution is 20 × 1/6 = 3.333 concentration-volume units. The dilute contribution is 2 × 5/6 = 1.667. Together they give the required total of 5.

The fraction of solute molecules in the dense phase is therefore 3.333/5 = 2/3, or about 66.7%. A region occupying only 16.7% of the volume contains roughly two-thirds of this solute in the hypothetical system.

At an overall concentration of 8 micromolar, the dense phase occupies one-third of the volume. Its contribution is 20/3 = 6.667, so it contains 6.667/8 = 5/6, or about 83.3% of the solute. Again, the molecular fraction is larger than the volume fraction.

Now add a trace client molecule that does not appreciably change the phase behaviour. Suppose its dense-phase concentration is four times its dilute-phase concentration. With dense volume fraction one-sixth, its molecular fraction inside is (4 × 1/6)/(4 × 1/6 + 1 × 5/6) = 4/9, about 44.4%.

A fourfold concentration preference therefore does not place 80% of the client inside. That answer would ignore the unequal phase volumes. Whenever someone reports enrichment, ask whether the quantity is a concentration ratio, a volume fraction or a fraction of all molecules. Those answer different biological questions.

Contents · Previous chapter · Next chapter

If components exchange freely, how can a condensate remain in a particular part of a cell? One possibility is that the local environment changes the conditions favouring assembly. Molecules can leave one region, diffuse and contribute to growth elsewhere. The visible structure can redistribute without travelling as one permanently sealed package.

The original 2009 P-granule study examined RNA- and protein-rich structures in early Caenorhabditis elegans embryos. Researchers observed liquid-like behaviours and linked localisation with regulated dissolution and condensation. This provided a physical explanation for spatial organisation in a particular living system.

A useful diagram has two panels showing the same total set of coloured symbols. In the first, several concentrated regions are distributed across the cell. In the second, some regions have dissolved while others have grown on one side. Draw arrows for molecular redistribution, rather than implying that every original droplet was mechanically carried intact to its final position.

The diagram is a conceptual account, not a claim that transport never contributes to condensate positioning. Real systems may combine changing local interactions, fluid motion, binding sites and active processes. Observing accumulation at one location does not by itself identify which contribution dominates.

This also explains why “self-organisation” should not be read as “unregulated”. A collective physical response can be regulated through the concentrations, interactions and chemical states of its components. Cellular control can act by changing the conditions under which organisation emerges.

Contents · Previous chapter · Next chapter

A microscope records signals, often fluorescence from labelled molecules. A bright region may contain more labelled molecules, but brightness can also depend on labelling, optical geometry and molecular photophysics. Establishing enrichment requires appropriate comparisons before explaining why the enrichment occurs.

Roundness, fusion and fluorescence recovery are useful observations. However, no single feature is a universal certificate of liquid–liquid phase separation. Molecules bound to a cluster of sites can exchange with their surroundings; a fluorescent spot can recover after bleaching without having formed by demixing.

A 2022 study benchmarking live-cell phase-separation measurements explicitly compared phase separation with molecules binding to spatially clustered sites. Classical fluorescence-recovery measurements did not distinguish those alternatives in the tested systems. The researchers used calibrated half-bleaching to examine preferential internal mixing and the role of an interface.

The reader-level lesson is about discrimination, not an invitation to perform an imaging protocol. Ask what a proposed alternative would predict. If both explanations predict recovery of fluorescence, recovery alone cannot decide between them. A useful additional observation must create different expectations for the competing models.

The existing fluorescence correlation spectroscopy guide explains a separate route from intensity fluctuations to model-bounded mobility estimates. Mobility measurements can contribute evidence, but a diffusion coefficient does not by itself name the mechanism that assembled the compartment.

Contents · Previous chapter · Next chapter

Why do some components concentrate together while others remain outside? Multivalency explains the possibility of cooperative assembly, but it does not automatically explain selective membership. Sequence, folded domains, RNA partners and other structural features can influence which contacts are favourable.

A 2026 study using the synthetic micDROP system tested interactions among ten natural intrinsically disordered regions presented on multivalent scaffolds. Many combinations showed broad compatibility, while particular regions displayed stronger self-specific behaviour. The results separate a tendency to condense from the additional requirements for distinct condensate identities.

The boundary around that conclusion is important. A synthetic scaffold changes the context in which a sequence acts. Findings from that system illuminate possible interaction principles; they do not directly catalogue the composition of every native organelle or show that all cellular specificity is encoded in disordered regions alone.

Material behaviour is another question. A condensate may behave as a relatively fluid material on one timescale and resist deformation on another. Networks can reorganise slowly, and assemblies can change with time. Calling something a condensate does not guarantee that it behaves like a simple liquid under every observation.

Nor does concentration guarantee faster chemistry. Enrichment may bring reactants together, but it can also limit access, alter mobility or separate a substrate from a required partner. A proposed functional benefit needs measurements of the relevant activity and controls for changes that occur independently of condensation.

Contents · Previous chapter · Next chapter

Consider three fictional observations. Sample A contains round fluorescent spots and shows fluorescence recovery after bleaching. Sample B has those features plus measurements consistent with coexisting dense and dilute concentrations over a controlled composition range. Sample C changes its spots and its biochemical activity after an interaction-altering perturbation.

For A, write “dynamic enriched structures are consistent with several assembly mechanisms”. The observations are useful, but they leave alternatives such as clustered binding unresolved. More confident language would require additional discriminating evidence.

For B, ask whether total amounts, volumes and phase concentrations satisfy an inventory like the one in Chapters 4 and 5. Also ask which components were changed and whether the assumed coexistence conditions were held fixed. Consistency with mass balance is necessary for the model but does not establish the entire mechanism by itself.

For C, separate two arrows: the perturbation changes assembly, and the perturbation changes activity. Those observations do not yet prove that the assembly change caused the activity change. The same intervention might directly alter a catalytic site, binding interaction or protein abundance.

A stronger investigation looks for ways to distinguish those routes, such as comparisons that preserve relevant biochemical properties while changing collective organisation. The exact control depends on the molecular system. There is no universally clean switch that can be assumed to alter only phase separation.

Finally, record the level of replication. Many droplets from one preparation are observations within that preparation, not automatically many independent biological tests. Count the independent samples relevant to the claim, preserve negative observations and state which alternative explanations remain open. A convincing explanation is built from converging tests rather than a gallery of attractive images.

Contents · Previous chapter · Next chapter

How can a compartment persist while its molecules exchange? Its composition can be maintained by collective interactions and conditions favouring different phases. Individual membership can change even while the overall organisation remains stable.

In the original model, what happens when total concentration rises from 5 to 8 micromolar? Dense-phase volume fraction rises from one-sixth to one-third, with the assumed phase concentrations still 2 and 20 micromolar. The model does not specify droplet number or growth time.

Why can a fourfold-enriched client have fewer than half its molecules inside? The dense phase occupies only one-sixth of the available volume. Concentration preference and available volume jointly determine molecular allocation; the resulting fraction is four-ninths.

Does a fixed saturation concentration define every cellular condensate? No. It is a useful feature of particular simple models, while multicomponent composition changes and cellular activity can complicate that picture. The model must match the system being investigated.

What does fluorescence recovery prove? It supports molecular mobility or exchange under the measurement conditions. It does not independently establish the thermodynamic or molecular mechanism that created the observed structure.

The durable scientific habit is to connect four things: interactions, composition, observable behaviour and discriminating evidence. Membrane-free organisation becomes understandable when those links are explicit. Return to Cell Biology for neighbouring mechanisms, or continue through the How Science Works master guide.

Contents · Previous chapter · Continue to Cell Biology