Could the protein aggregates that cause neurodegeneration turn out to be the cellular equivalent of ice cubes? Scientists investigating how proteins naturally transition between gas-like solutions, condensed liquids, solidified gels, and insoluble fibrils are finding increasing evidence that proteins liquefy before aggregating. Paul Taylor and colleagues recently reported that the amyotrophic lateral sclerosis (ALS)-linked protein hnRNPA1 condenses into liquid droplets to promote assembly of stress granules. In another study, Nicolas Fawzi and colleagues detail the structure of the ALS-linked protein FUS. It remains amorphous when it condenses into droplets—i.e., it behaves as a liquid. Simon Alberti and colleagues found that FUS variants associated with disease have a greater penchant for assuming the liquid form and that FUS droplets go on to form membrane-free organelles such as stress granules, which are believed to protect cells under duress. The droplets may have a physiological role as well because they harbor RNA polymerase. In fact, Cliff Brangwynne and colleagues discovered that active biological processes, such as transcription, stabilize liquid-phase bodies, including nucleoli. On the downside, researchers believe FUS might turn toxic if it condenses further into a solid form, while Taylor found that hnRNPA1 formed amyloid-like fibrils when it condensed further. In addition, Peter St. George-Hyslop and colleagues will publish data from four years’ worth of research related to this topic in Neuron on October 29. Do other proteins involved in neurodegeneration undergo similar phase transitions? And do those liquids lead to toxic aggregates?  

On Friday, October 30, at 11 am U.S. Eastern, when Alberti, Brangwynne, Fawzi, St. George-Hyslop, Taylor, and Markus Zweckstetter discussed what liquid phase transitions mean for proteins and for health and disease.

We thank Cell for giving Alzforum readers temporary free access to Paul Taylor's paper.


We thank Neuron for fast-tracking Peter St. George-Hyslop's paper and making it open access.




Many proteins involved in neurodegeneration famously contain disordered domains that don’t seem to settle into a rigid structure. These so-called low-complexity sequence domains contain sticky glutamine and asparagine residues and promote aggregation. A decade ago, scientists discovered that these domains facilitated the formation of stress granules and processing bodies, ephemeral organelles that store mRNAs and related proteins (Gilks et al., 2004; Decker et al., 2007). These organelles assemble when the cell needs them and later disassemble to release mRNAs.

With no membranes to hold their contents, what keeps the granules together? The answer appears to be phase transitions, aka liquid-liquid demixing. According to this concept, when enough individual proteins get together they condense—like morning dew—into specialized liquid droplets (see movie below) within the larger liquid cytosol or nucleoplasm (reviewed in Weber and Brangwynne, 2012).


Droplets of FUS flow, form, and fuse in the lab. [Courtesy of Fawzi Lab/Brown University.]

Cliff Brangwynne of Princeton University, New Jersey, was among the first to describe liquid organelles. He studied P granules in nematodes, which form in the worms’ germ cells to store RNA and RNA-binding proteins. Before a fertilized egg divides for the first time, the P granules cluster at the cell’s posterior end. This happens because granule components continually disperse and recondense, with more condensation happening at the cell’s posterior (Brangwynne et al., 2009). Recently, Brangwynne and colleagues described how the formation of liquid nucleoli and other nuclear droplets in nematode embryos depends on the concentration of their components, cell size, and transcription (Weber et al., 2015; Berry et al., 2015). Remarkably, they find the same mathematics that describes phase transitions of non-biological matter applies to protein phases, as well.

In the September 24 Cell, Paul Taylor of St. Jude Children’s Research Hospital in Memphis, Tennessee, reports similar properties for granules composed of a protein called hnRNPA1. Taylor’s group previously linked mutations in this gene to amyotrophic lateral sclerosis and multisystem proteinopathy (see Mar 2013 news). In the new study, the researchers tried to purify hnRNPA1 and noticed it turned cloudy in suspension, forming droplets that spread and fused on a coverslip just like water on a tabletop. This behavior required the low-complexity part of the protein. Simply increasing the concentration of hnRNPA1 in HeLa cells was enough to seed liquid stress granules. In vitro, disease-linked mutant hnRNPA1 condensed further, leaving solid precipitates on the coverslip. The protein had formed fibrils. Wild-type protein could do that too, but it took longer. Taylor and colleagues speculated that some fibrilization might be beneficial, helping stabilize granules or performing other functions, but excess solidification could turn pathological.

Phase Separation and Fibrilization. Liquid-liquid phase separations (LLPS) of RNA binding proteins (RBPs) may create ideal conditions for fibrilization. Other proteins of low complexity might behave similarly. [Courtesy of Cell, Molliex et al.]

The ALS-linked protein FUS also undergoes phase transitions, as Simon Alberti and colleagues at the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden, reported in August (see Sep 2015 news). This German group saw FUS make little droplets both in vitro and in cell culture. Mutant FUS formed solid, amyloid-like fibrils; so did wild-type, but more slowly. The risk of ending up with fibrils—which these authors believe have no normal function—is the price the cell pays for being able to rapidly morph FUS into liquid when it’s needed in stress granules or other organelles.

What do individual FUS molecules look like in these droplets? Nicholas Fawzi and colleagues at Brown University, in Providence, Rhode Island, offer an answer in the October 15 Molecular Cell. Using NMR spectroscopy, they determined that the low-complexity domain of FUS remains relatively disordered, even in the liquid state, distinguishing them from more static inclusions or hydrogels. Since FUS localizes to sites of transcription in cells, they hypothesized it might interact with RNA polymerase. Sure enough, the carboxyl-terminal portion of RNA polymerase II nucleated FUS condensation and localized to the droplets in vitro.

Further advancing this topic, Peter St. George-Hyslop of the University of Cambridge in the United Kingdom will publish data in Neuron on October 29 on how ALS/FTD mutations in FUS influence phase transitions Moreover, FUS and hnRNPA1 may not be the only proteins changing phase in neurodegeneration. Scientists believe other aggregation-prone proteins may do the same. Indeed, Taylor and colleagues already have hints that the ALS- and frontotemporal dementia-linked protein TDP-43 makes liquid drops. Tau and α-synuclein are famously disordered proteins that would fit the low-complexity criteria for forming liquid phases. A broader phenomenon would raise important questions. What controls these protein phase transitions? Is the process important in ALS and related diseases? And is there ever a good reason for these proteins to turn solid? Markus Zweckstetter and colleagues at the Max Planck Institute for Biophysical Chemistry at the University of Gottingen, have found that liquid phase transition is necessary for amyloid-like assembly of myelin basic protein and for the generation of myelin membranes. Tune in to the October 30 Webinar and learn the latest on this rapidly evolving field.—Amber Dance


Q: Does forming cytoplasmic droplets mean that cytoplasm is not a liquid? Or a different liquid? How aqueous are these droplets?
Fawzi: You can think about it like oil and water—both are liquids, but the oil is not soluble in water and it will collect together. In the droplets we showed, cells have mechanisms for limiting their size (Cliff’s work) and also the droplets in vitro might “age” and actually aggregate into something solid (no longer liquid). Our in vitro droplets are high-concentration, but are at about the estimates of the macromolecule concentration present in the cytoplasm—however, they are packed with the same single protein.

Q: Liquid-liquid phase separation [LLPS] is extremely common for proteins under crystal growing conditions. These conditions are obviously very different from inside the cell, but is there a danger that one can get lots of proteins to phase separate in a non-physiological way in vitro, if one continues to tweak solute concentrations?
Fawzi: These droplets are formed at a very, very low concentration compared to crystal growing conditions—here, proteins at 1 micromolar. Since some of these proteins (e.g., FUS) are present at about this concentration in cells, this phase separation happens at physiological concentrations. However, yes, there is always a danger!

Q: Is a low-complexity domain [LCD] a somewhat conserved sequence, or is it virtually any sequence that appears to have little complexity, despite how conserved or not conserved the sequences are? We’ve seen a few different proteins with low-complexity domains, so how arbitrary is the designation?
Alberti: The definition is quite arbitrary. Usually, it refers to any sequence that has low-sequence complexity. The only attribute that often is conserved is the amino acid composition, although some sequences also contain repeats, which may have functional significance. However, there are different types of low-complexity sequences (LCS), composed of different amino acids. Some LC sequences have a lot of charged residues. These may undergo LLPS, but their propensity to form aggregates may be low. Prion-like LC sequences instead are devoid of charged residues and composed of mostly polar amino acids (proline, serine, glycine, asparagine, glutamine, tyrosine). They seem to undergo LLPS, and at the same time, are also prone to aggregate.

Q: What is the percentage of proline residues in these proteins?
Alberti: The percentage of proline residues is usually quite high, especially in those regions that are intrinsically disordered. Prolines are structure-breakers. Therefore, the addition of prolines may be one of the design features to keep these domains in a disordered state and prevent conversion into a more ordered, amyloid state.

Q: What do you think is the structural feature of the LCD that prevents formation of steric zippers? Do you think there is a specific structural motif that perhaps keeps the steric zipper motif “under control”?
Fawzi: High proline content can help, along with serine and glycine perhaps.

Q: Could a specific interaction between two different proteins facilitate liquid droplet formation? Can cross-LLPS also happen?
Alberti: Work from Rosen and Parker, recently published in Molecular Cell, showed that two proteins could form mixed assemblies (Lin et al., 2015). Stress granules in living cells contain numerous proteins and RNAs, indicating that LLPS can also happen with complex protein mixtures.

Fawzi: In our in vitro work, we find that the C-terminal domain of RNA polymerase II induces FUS low-complexity domain phase separation. See also Mike Rosen’s 2012 Nature paper (Li et al.). 

Q: Are there biological differences in phase transitions mediated by low-complexity sequences (like hnRNPA1) or mediated by multivalency (like [Mike] Rosen’s SH2 domains)?
Alberti: Both types of interactions rely on multivalent, low-affinity interactions. However, interactions mediated by prion-like LC domains may be less specific (e.g., they also bind RNA) than interactions mediated by, for example, SH2 domains. As a consequence, liquid droplets formed by LC domain proteins may be able to accommodate a large number of different proteins/RNAs. This could also translate into biological differences, but this still has to be shown.

Q: Are non-membrane-bound organelles a target for autophagy?
Alberti: Yes. P granules in  C. elegans, for example, are targets for autophagy. The same has been shown for stress granules. Therefore, the formation of membraneless compartments may be a way of disposing of harmful or unwanted collections of macromolecules.

Q: We would be interested to learn which specific RNA species are being added to accelerate phase transitions.
Fawzi: [Thomas] Cech’s group has shown that FUS does not have an apparent specificity for an RNA sequence or structure it can bind, which may explain its broad role in splicing regulation of a large percentage of transcripts (Wang et al., 2015). Therefore, we used a yeast RNA extract.

Q: There is often discussion of LCS being “prion-like.” Does the panel think the disease progression acts in a prionogenic manner with prion-like RNA-binding proteins moving from cell to cell and seeding phase transition in neighboring neurons?
Alberti: This is quite likely. Liquid droplets are metastable and convert with time into more solid structures, such as hydrogels or fibrillar aggregates. Cells normally are able to prevent this from happening. However, with increasing age or because of mutations, these mechanisms weaken. Once highly ordered structures such as fibers have formed, they are difficult to stop and could spread by infecting liquid droplets in the same cells. Once the cell that harbors them dies, they may be released into the extracellular space. The released fibers could then infect neighboring cells.

Q: Prion-like proteins, when expressed in mammalian cells, form clumps that grow, then the bigger aggregates break down and are inherited by the daughter cells where they start another cycle of assembly and seeding, so to speak. At what stage do you think this fragmentation happens, in the liquid transition phase or in when they are in a fully solid phase?
Alberti: Because the liquid phase would be highly unstable once an aggregate has formed, the fragmentation is most likely happening with solid aggregates. However, the fragmentation of fibrils could occur in a liquid phase formed by other proteins. For example, there is evidence that protein quality control compartments such as IPOD and JUNQ properties (for a review see Bagola et al., 2008).

Q: Can LLPS also explain the spreading of amyloid proteins?
Alberti: Once a protein has converted into an amyloid state, it will convert all the protein in a liquid droplet quite rapidly through conformational conversion.

Q: If wild-type FUS and FUS-G156E are co-incubated or co-expressed, do they form heterotypic droplets, and if so, do these droplets exhibit properties (such as fusion, disassembly, etc.) similar to wild-type protein, mutant protein, or an intermediate phenotype?
Alberti: Indeed, we found that wild-type and G156E do form mixed droplets. The properties of these mixed droplets, at least initially, are indistinguishable from wild-type. We have not done aging experiments, so we do not know what happens with time.

Q: With dark-state exchange saturation transfer nuclear magnetic resonance (DEST NMR)—how can you be sure the protein you’re getting inside the condensed phase is truly condensed phase FUS and not a small fraction of liquid-like FUS which is co-existing with the dense phase?
Fawzi: For these studies, we used direct NMR characterization of a single, very large condensed “droplet” or continuous phase of FUS. The intensity of the NMR signals we see are consistent with the high (7 mM) concentration that we estimate (by dilution spectrophotometry) in the liquid phase separated state. However, we are starting to use DEST to look at the structural details of the conversion from liquid to solid that the other speakers have reported.


Comments on this content

No Available Comments

Make a Comment

To make a comment you must login or register.


News Citations

  1. Disease Mutations Zip Lock Stress Granules in Proteinopathy, ALS
  2. ALS Protein Said to Liquefy, Then Freeze en Route to Disease

Paper Citations

  1. . Stress granule assembly is mediated by prion-like aggregation of TIA-1. Mol Biol Cell. 2004 Dec;15(12):5383-98. Epub 2004 Sep 15 PubMed.
  2. . Edc3p and a glutamine/asparagine-rich domain of Lsm4p function in processing body assembly in Saccharomyces cerevisiae. J Cell Biol. 2007 Nov 5;179(3):437-49. PubMed.
  3. . Getting RNA and protein in phase. Cell. 2012 Jun 8;149(6):1188-91. PubMed.
  4. . Germline P granules are liquid droplets that localize by controlled dissolution/condensation. Science. 2009 Jun 26;324(5935):1729-32. Epub 2009 May 21 PubMed.
  5. . Inverse size scaling of the nucleolus by a concentration-dependent phase transition. Curr Biol. 2015 Mar 2;25(5):641-6. Epub 2015 Feb 19 PubMed.
  6. . RNA transcription modulates phase transition-driven nuclear body assembly. Proc Natl Acad Sci U S A. 2015 Sep 22;112(38):E5237-45. Epub 2015 Sep 8 PubMed.
  7. . Formation and Maturation of Phase-Separated Liquid Droplets by RNA-Binding Proteins. Mol Cell. 2015 Oct 15;60(2):208-19. Epub 2015 Sep 24 PubMed.
  8. . Phase transitions in the assembly of multivalent signalling proteins. Nature. 2012 Mar 7;483(7389):336-40. PubMed.
  9. . Nucleic acid-binding specificity of human FUS protein. Nucleic Acids Res. 2015 Sep 3;43(15):7535-43. Epub 2015 Jul 6 PubMed.

Other Citations

  1. Bagola et al., 2008

External Citations

  1. hnRNPA1
  2. FUS 
  3. paper
  4. paper

Further Reading

No Available Further Reading

Primary Papers

  1. . Phase separation by low complexity domains promotes stress granule assembly and drives pathological fibrillization. Cell. 2015 Sep 24;163(1):123-33. PubMed.
  2. . Residue-by-Residue View of In Vitro FUS Granules that Bind the C-Terminal Domain of RNA Polymerase II. Mol Cell. 2015 Oct 15;60(2):231-41. Epub 2015 Oct 8 PubMed.