A typical human cell is a crowded soup of thousands of different proteins. One way that the cell organizes this complex mix of contents is by creating separate droplets within the cell, like oil in water. These droplets can form through a process known as liquid-liquid phase separation, or LLPS, where specific proteins gather in high concentrations to carry out their cellular roles.

The critical role of LLPS in cellular organization means that it is widely studied by biologists. To detect LLPS, researchers often subject the cells to treatments designed to hold all the proteins in place, creating a snapshot of their natural state. This process, known as fixing, allows scientists to easily label a protein with a fluorescent tag, take pictures of the cells, and look at whether the protein forms droplets in its natural state. This is often easier to do than imaging cells live, but it relies on LLPS being well-preserved upon fixation.

To test if this is true, Irgen-Gioro, Yoshida et al. looked at protein droplets in live cells, and then fixed the cells to check whether the appearance of the droplets had changed. The images taken showed that fixation could alter the size and number of droplets depending on the protein being studied. To explain why the effects of fixing change depending on the protein, Irgen-Gioro, Yoshida et al. hypothesized that a faster fixation – relative to how quickly proteins can bind and unbind to their droplets – can better preserve the LLPS droplets. They verified their idea using a microscopy technique in which they imaged single molecules, allowing them to see how different fixation speeds relative to protein binding affected the droplets.

The work of Irgen-Gioro, Yoshida et al. identifies an important caveat to using fixation for the study of LLPS in cells. Their findings suggest that researchers should be cautious when interpreting the results of such studies. Given that LLPS in cells is an area of research with a lot of interest, these results could benefit a broad range of biological and medical fields. In the future, Irgen-Gioro, Yoshida et al.’s findings could prompt scientists to develop new fixing methods that better preserve LLPS in cells.

Fixing cells to preserve a snapshot of biomolecular transactions in vivo is a widely used strategy in numerous techniques in biology and medicine. Due to its small size and high reactivity with a wide range of biological entities, paraformaldehyde (PFA) is one of the most commonly used fixatives to create covalent cross-linking between biomolecules, for example, proteins and nucleic acids. PFA nonselectively ‘fixes’ or cross-links molecules in proximity to enable characterization of biomolecular interactions formed in living cells. Examples of popular techniques that use PFA to fix cells include ChIP-sequencing (Robertson et al., 2007; Solomon and Varshavsky, 1985), chromosome conformation capture (3C)-based techniques (Dekker et al., 2002), immunofluorescence (Richter et al., 2018), fluorescence in situ hybridization (FISH) (Moter and Göbel, 2000), cross-linking mass spectrometry (Sutherland et al., 2008), super-resolution expansion microscopy (Chen et al., 2015), and super-resolution localization microscopies such as stochastic optical reconstruction microscopy (STORM) (Rust et al., 2006). Although PFA fixation has been used to faithfully preserve live-cell conditions in many scenarios, a number of studies have uncovered situations in which fixation fails to cross-link DNA–protein interactions formed in living cells. By imaging different transcription factors (TFs) in live and fixed cells, Schmiedeberg et al., 2009 showed that TFs bound to DNA with fast dissociation dynamics (<5 s residence times as determined by fluorescence recovery after photobleaching [FRAP]) are not cross-linked to DNA upon PFA fixation. Using live-cell single-molecule imaging, Teves et al., 2016 showed that TFs stay bound to chromosome during mitosis and fixing cells can artificially deplete transiently bound TFs from mitotic chromosomes. These studies exemplify the fact that fixation, with limited reaction rates, cannot provide an instantaneous snapshot and may miss or obfuscate biomolecular interactions that happen either at or faster than the timescale of fixation. What further complicates the result of cell fixation is that the reactivity and reaction rates of PFA are variable and dependent on its biomolecule substrates (Gavrilov et al., 2015; Shishodia et al., 2018). For example, the efficiency and rates at which PFA reacts with proteins can vary by orders of magnitude (Kamps et al., 2019) and are dependent on their amino acid sequences (Kamps et al., 2019; Metz et al., 2004; Sutherland et al., 2008) and tertiary structures (Hoffman et al., 2015).

Among the numerous biomolecular transactions investigated using fixed-cell imaging is liquid–liquid phase separation (LLPS), a long-observed behavior of polymers in solution (Gibbs, 1879; Graham, 1861; Hyman et al., 2014) that has recently generated much excitement in biological research communities due to its proposed roles in cellular organization and functions (Banani et al., 2017; Boeynaems et al., 2018; Mitrea and Kriwacki, 2016; Shin and Brangwynne, 2017). LLPS is driven by excessive levels of transient, selective, and multivalent protein–protein interactions mediated by intrinsically disordered regions (IDRs) within the proteins of interest (Chong et al., 2018; Kato and McKnight, 2018; Li et al., 2012). Whereas rigorous characterization of LLPS in vivo has been challenging and remains a question under active investigation (McSwiggen et al., 2019b), detection of discrete puncta that have a spherical shape, undergo fusion and fission, and dynamically exchange biomolecules with the surrounding according to FRAP is often considered evidence of putative LLPS in living cells. While such diverse measurements have been widely used for studying proteins under overexpression conditions, far fewer approaches are available to probe LLPS under physiological conditions. Detecting local high-concentration regions or puncta of an endogenously expressed protein using immunofluorescence of fixed cells has been used in many studies as evidence of LLPS (Boija et al., 2018; Guo et al., 2019; Owen et al., 2021; Xie et al., 2022; Yang et al., 2020). Not only is the detection of puncta an inconclusive metric for establishing LLPS, whether a punctate distribution observed in fixed cells actually represents the live-cell scenario remains unclear as fixation has only been assumed, but not directly shown, to faithfully preserve multivalent interactions and LLPS formed in living cells. This knowledge gap motivated us to image cells that overexpress various known IDR-containing proteins before and after fixation to evaluate the ability of PFA fixation to preserve LLPS behaviors. We found that, interestingly, fixation can significantly alter the appearance of droplet-like puncta in cells. Our quantitative image analysis suggests that depending on the LLPS-driving protein, fixing cells can either enhance or diminish the apparent LLPS behaviors in vivo. In certain cases, fixation can even cause droplet-like puncta to artificially appear in cells that have a homogeneous protein distribution and no detectable puncta in the live condition. Conversely, fixation can also cause droplet-like puncta in living cells to completely disappear. Combining experiments that modulate fixation rates, live-cell single-molecule imaging that quantifies protein binding dynamics, and simulations based on a kinetic model, we further demonstrated that protein localization in fixed cells depends on an intricate balance of protein–protein interaction dynamics, the overall rate of fixation, and the difference between protein fixation rates in and out of droplet-like puncta. Our work urges caution in the interpretation of previous claims of in vivo phase separation based solely on immunofluorescence imaging of fixed cells and serves to guide future judicious application of PFA fixation.

Understanding situations in which PFA fixation can properly preserve live-cell conditions is essential in judicious applications of fixation-based biological techniques. Because approaches for rigorous determination of LLPS in vivo remain lacking (McSwiggen et al., 2019b) and detection of local high-concentration regions of an endogenously expressed protein in fixed cells via immunofluorescence has been widely used as evidence for LLPS (Boija et al., 2018; Guo et al., 2019; Owen et al., 2021; Xie et al., 2022; Yang et al., 2020), understanding how well fixation preserves LLPS behaviors is important for justifying the immunofluorescence-based diagnosis method and for studying the functional relevance of LLPS in vivo. In this work, we imaged various LLPS systems in living cells before and after PFA fixation, quantified parameters that describe LLPS appearance in cells, and showed that fixation can either enhance or diminish the apparent LLPS behaviors in vivo. Lowering the PFA concentration and adding GA to PFA did not remove the fixation artifacts. For the first time, our work reveals an important caveat in using fixation-based methods to detect and characterize LLPS in vivo and suggests an advantage of using live-cell imaging to study LLPS systems over fixed-cell experiments. However, not all the proteins we examined have their puncta-forming or apparent LLPS behaviors in cells changed upon fixation. For example, PFA fixation faithfully preserves the appearance of FUS(FL), TAF15(IDR)-FTH1, and EWS::FLI1 puncta in cells (Figure 3). Nevertheless, our work points out a necessity to use live-cell imaging to confirm LLPS behaviors previously characterized with fixed-cell experiments. Live-imaging techniques that allow estimation of protein diffusion coefficients within specific cellular compartments, for example, SPT (Hansen et al., 2018; Heckert et al., 2022) and fluorescence correlation spectroscopy (Lanzanò et al., 2017), can be useful alternative approaches for diagnosing LLPS in vivo without the potential artifact of fixation, as diffusion dynamics are recently shown to be affected by LLPS (Heltberg et al., 2021; McSwiggen et al., 2019a; Miné-Hattab et al., 2021; Chong et al., 2022; Ladouceur et al., 2020).

We note that fixation-induced changes of LLPS appearance may lead to potential misinterpretation of the functional relevance of LLPS in cellular processes. For example, recent work has uncovered that effective transcriptional activation requires an optimum of TF IDR-IDR interactions within TF hubs formed at target genes and that overly high levels of IDR-IDR interactions pushing the system toward LLPS can repress transcription (Chong et al., 2022; Trojanowski et al., 2022). Future characterization of the functionally optimal interaction level will require quantification of the sizes of hubs or droplet-like puncta while measuring transcription activity. Because a fixation-induced increase or decrease in puncta sizes may lead to inaccurate determination of the functional optimum, scrutiny will be required in choosing between live-cell and fixed-cell imaging methods for quantifying LLPS appearance in these types of studies. Moreover, given that fixation can artificially generate intranuclear puncta of EGFP-EWS(IDR) that is homogenously distributed across the live-cell nucleus (Figure 1A), extra caution is required in interpreting immunofluorescence-detected intracellular puncta of an endogenously expressed protein as LLPS as the same puncta-generating fixation artifact might happen to the protein even when it is not phase separating in living cells. To confirm puncta formation, counterpart live-cell imaging of the endogenous protein will be necessary, which requires engineering the cells, for example, by CRISPR, to fluorescently tag the protein.

To understand the factors that can cause fixation-induced changes of LLPS appearance in the cell, we simulated the changes through kinetic modeling, which reveals that the dynamics of POI binding to and dissociating from puncta, the absolute fixation rate of POI, and different fixation rates of POI in and out of puncta all play a role in inducing the fixation artifacts. Our kinetic model takes previous work studying fixation artifacts in the context of protein–DNA interactions (Poorey et al., 2013; Schmiedeberg et al., 2009; Teves et al., 2016) one step further by considering two fixed states of POI instead of one state, which are fixation in and out of puncta with different rates due to distinct local protein composition and concentrations. We then used live-cell single-molecule imaging experiments to demonstrate that as predicted by our model, a fast overall fixation rate of POI relative to its puncta-binding dynamics can minimize fixation artifacts.

We emphasize that because our four-state model makes no assumptions about any state being phase-separated, the logical implications of our model can extend beyond LLPS to other biomolecular transactions and cellular structures that have been found not well preserved by fixation or immunofluorescence, including localizations of cilia proteins (Hua and Ferland, 2017), clustering of cell membrane receptors (Stanly et al., 2016), splicing speckle formation (Neugebauer and Roth, 1997), and chromatin organization and protein binding (Zarębski et al., 2021; Lorber and Volk, 2022; Lerner et al., 2016; Pallier et al., 2003; Kumar et al., 2008 and Teves et al., 2016). Our model can similarly extend beyond PFA to other fixatives. This is useful because different fixatives have been chosen for studying different types of structures. For example, PFA fixation is often preferable for preserving soluble proteins over dehydration fixatives such as methanol (Stadler et al., 2010 and Schnell et al., 2012), yet methanol fixation can be preferable over PFA for preserving proteins bound to mitotic chromatin (Kumar et al., 2008; Lerner et al., 2016). Generally, our model predicts that fixation artifacts will occur whenever a protein can exist in multiple states that have different rates of fixation, and this artifact is most severe when the fixation is slower than the transition between states. For PFA fixation, because its rate is sensitively dependent on the amino acid sequence of POI, the structure of POI, and POI’s cross-linked partners (Hoffman et al., 2015; Kamps et al., 2019; Metz et al., 2006; Metz et al., 2004), POI in different states likely has different PFA fixation rates regardless of the type of interaction it undergoes.

One distinction between our study and previous studies is that we observe that PFA fixation can enhance apparent protein–protein interactions or LLPS behaviors in the cell, suggesting faster fixation for POI in the bound than dissociated state (${\displaystyle k_3>k_4}$), whereas fixation has only been reported to diminish protein–DNA interactions, suggesting slower fixation for POI in the bound state (${\displaystyle k_3<k_4}$) (Poorey et al., 2013; Schmiedeberg et al., 2009; Teves et al., 2016). We hypothesize that this is because fixing the bound state of an LLPS system (within puncta) is dominated by cross-linking reactions between IDRs enriched in puncta, which have reactive residues better exposed to solvent due to lack of well-defined tertiary structures and thereby likely cross-link faster than structured domains cross-linking to DNA (Hoffman et al., 2015). It will be of future interest to measure fixation rates of different biomolecules including IDRs, structured proteins, and nucleic acids to prove the proposed chemical mechanism underlying fixation artifacts. Since our simulated results highlight the role of absolute fixation rates in the outcome of fixation, another future endeavor will be to design novel fixatives with significantly faster cross-linking rates than biomolecular interactions to eliminate fixation artifacts in the cell.

Figure 1 - Source Data 1, Figure 2 - Source Data 1, Figure 3 - Source Data 1, and Figure 6 - Source Data 1 contain the numerical data used to generate the figures. Custom scripts have been uploaded as source code files.

1. 1. Alberti S
   2. Gladfelter A
   3. Mittag T

   (2019) Considerations and challenges in studying liquid-liquid phase separation and biomolecular condensates

   Cell 176:419–434.

   https://doi.org/10.1016/j.cell.2018.12.035

   • PubMed
   • Google Scholar
2. 1. Altmeyer M
   2. Neelsen KJ
   3. Teloni F
   4. Pozdnyakova I
   5. Pellegrino S
   6. Grøfte M
   7. Rask M-BD
   8. Streicher W
   9. Jungmichel S
   10. Nielsen ML
   11. Lukas J

   (2015) Liquid demixing of intrinsically disordered proteins is seeded by poly (ADP-ribose)

   Nature Communications 6:1–12.

   https://doi.org/10.1038/ncomms9088

   • Google Scholar
3. 1. Banani SF
   2. Lee HO
   3. Hyman AA
   4. Rosen MK

   (2017) Biomolecular condensates: organizers of cellular biochemistry

   Nature Reviews. Molecular Cell Biology 18:285–298.

   https://doi.org/10.1038/nrm.2017.7

   • PubMed
   • Google Scholar
4. 1. Bellapadrona G
   2. Elbaum M

   (2014) Supramolecular protein assemblies in the nucleus of human cells

   Angewandte Chemie 53:1534–1537.

   https://doi.org/10.1002/anie.201309163

   • PubMed
   • Google Scholar
5. 1. Berry J
   2. Weber SC
   3. Vaidya N
   4. Haataja M
   5. Brangwynne CP

   (2015) Rna transcription modulates phase transition-driven nuclear body assembly

   PNAS 112:E5237–E5245.

   https://doi.org/10.1073/pnas.1509317112

   • Google Scholar
6. 1. Boeynaems S
   2. Alberti S
   3. Fawzi NL
   4. Mittag T
   5. Polymenidou M
   6. Rousseau F
   7. Schymkowitz J
   8. Shorter J
   9. Wolozin B
   10. Van Den Bosch L
   11. Tompa P
   12. Fuxreiter M

   (2018) Protein phase separation: a new phase in cell biology

   Trends in Cell Biology 28:420–435.

   https://doi.org/10.1016/j.tcb.2018.02.004

   • PubMed
   • Google Scholar
7. 1. Boija A
   2. Klein IA
   3. Sabari BR
   4. Dall’Agnese A
   5. Coffey EL
   6. Zamudio AV
   7. Li CH
   8. Shrinivas K
   9. Manteiga JC
   10. Hannett NM
   11. Abraham BJ
   12. Afeyan LK
   13. Guo YE
   14. Rimel JK
   15. Fant CB
   16. Schuijers J
   17. Lee TI
   18. Taatjes DJ
   19. Young RA

   (2018) Transcription factors activate genes through the phase-separation capacity of their activation domains

   Cell 175:1842–1855.

   https://doi.org/10.1016/j.cell.2018.10.042

   • PubMed
   • Google Scholar
8. 1. Bracha D
   2. Walls MT
   3. Wei MT
   4. Zhu L
   5. Kurian M
   6. Avalos JL
   7. Toettcher JE
   8. Brangwynne CP

   (2018) Mapping local and global liquid phase behavior in living cells using photo-oligomerizable seeds

   Cell 175:1467–1480.

   https://doi.org/10.1016/j.cell.2018.10.048

   • PubMed
   • Google Scholar
9. 1. Chen F
   2. Tillberg PW
   3. Boyden ES

   (2015) Expansion microscopy

   Science 347:543–548.

   https://doi.org/10.1126/science.1260088

   • Google Scholar
10. 1. Chong S
    2. Dugast-Darzacq C
    3. Liu Z
    4. Dong P
    5. Dailey GM
    6. Cattoglio C
    7. Heckert A
    8. Banala S
    9. Lavis L
    10. Darzacq X
    11. Tjian R

    (2018) Imaging dynamic and selective low-complexity domain interactions that control gene transcription

    Science 361:eaar2555.

    https://doi.org/10.1126/science.aar2555

    • PubMed
    • Google Scholar
11. 1. Chong S
    2. Graham TGW
    3. Dugast-Darzacq C
    4. Dailey GM
    5. Darzacq X
    6. Tjian R

    (2022) Tuning levels of low-complexity domain interactions to modulate endogenous oncogenic transcription

    Molecular Cell 82:2084–2097.

    https://doi.org/10.1016/j.molcel.2022.04.007

    • PubMed
    • Google Scholar
12. 1. Costantini LM
    2. Fossati M
    3. Francolini M
    4. Snapp EL

    (2012) Assessing the tendency of fluorescent proteins to oligomerize under physiologic conditions

    Traffic 13:643–649.

    https://doi.org/10.1111/j.1600-0854.2012.01336.x

    • PubMed
    • Google Scholar
13. 1. Currie SL
    2. Rosen MK

    (2022) Using quantitative reconstitution to investigate multicomponent condensates

    RNA 28:27–35.

    https://doi.org/10.1261/rna.079008.121

    • PubMed
    • Google Scholar
14. 1. Dekker J
    2. Rippe K
    3. Dekker M
    4. Kleckner N

    (2002) Capturing chromosome conformation

    Science 295:1306–1311.

    https://doi.org/10.1126/science.1067799

    • PubMed
    • Google Scholar
15. 1. Gavrilov A
    2. Razin SV
    3. Cavalli G

    (2015) In vivo formaldehyde cross-linking: it is time for black box analysis

    Briefings in Functional Genomics 14:163–165.

    https://doi.org/10.1093/bfgp/elu037

    • PubMed
    • Google Scholar
16. Book

    1. Gibbons JD
    2. Chakraborti S

    (2014)

    Nonparametric Statistical Inference

    CRC press.

    • Google Scholar
17. Book

    1. Gibbs JW

    (1879)

    On the Equilibrium of Heterogeneous Substances

    heidok.

    • Google Scholar
18. 1. Graham T

    (1861) X. liquid diffusion applied to analysis

    Philosophical Transactions of the Royal Society of London 151:183–224.

    https://doi.org/10.1098/rstl.1861.0011

    • Google Scholar
19. 1. Grimm JB
    2. English BP
    3. Choi H
    4. Muthusamy AK
    5. Mehl BP
    6. Dong P
    7. Brown TA
    8. Lippincott-Schwartz J
    9. Liu Z
    10. Lionnet T
    11. Lavis LD

    (2016) Bright photoactivatable fluorophores for single-molecule imaging

    Nature Methods 13:985–988.

    https://doi.org/10.1038/nmeth.4034

    • PubMed
    • Google Scholar
20. 1. Grimm JB
    2. Xie L
    3. Casler JC
    4. Patel R
    5. Tkachuk AN
    6. Falco N
    7. Choi H
    8. Lippincott-Schwartz J
    9. Brown TA
    10. Glick BS
    11. Liu Z
    12. Lavis LD

    (2021) A general method to improve fluorophores using deuterated auxochromes

    JACS Au 1:690–696.

    https://doi.org/10.1021/jacsau.1c00006

    • PubMed
    • Google Scholar
21. 1. Grünewald TGP
    2. Cidre-Aranaz F
    3. Surdez D
    4. Tomazou EM
    5. de Álava E
    6. Kovar H
    7. Sorensen PH
    8. Delattre O
    9. Dirksen U

    (2018) Ewing sarcoma

    Nature Reviews. Disease Primers 4:5.

    https://doi.org/10.1038/s41572-018-0003-x

    • PubMed
    • Google Scholar
22. 1. Guo YE
    2. Manteiga JC
    3. Henninger JE
    4. Sabari BR
    5. Dall’Agnese A
    6. Hannett NM
    7. Spille J-H
    8. Afeyan LK
    9. Zamudio AV
    10. Shrinivas K
    11. Abraham BJ
    12. Boija A
    13. Decker T-M
    14. Rimel JK
    15. Fant CB
    16. Lee TI
    17. Cisse II
    18. Sharp PA
    19. Taatjes DJ
    20. Young RA

    (2019) Pol II phosphorylation regulates a switch between transcriptional and splicing condensates

    Nature 572:543–548.

    https://doi.org/10.1038/s41586-019-1464-0

    • PubMed
    • Google Scholar
23. 1. Hansen AS
    2. Pustova I
    3. Cattoglio C
    4. Tjian R
    5. Darzacq X

    (2017) Ctcf and cohesin regulate chromatin loop stability with distinct dynamics

    eLife 6:e25776.

    https://doi.org/10.7554/eLife.25776

    • PubMed
    • Google Scholar
24. 1. Hansen AS
    2. Woringer M
    3. Grimm JB
    4. Lavis LD
    5. Tjian R
    6. Darzacq X

    (2018) Robust model-based analysis of single-particle tracking experiments with Spot-On

    eLife 7:e33125.

    https://doi.org/10.7554/eLife.33125

    • PubMed
    • Google Scholar
25. 1. Heckert A
    2. Dahal L
    3. Tijan R
    4. Darzacq X

    (2022) Recovering mixtures of fast-diffusing states from short single-particle trajectories

    eLife 11:e70169.

    https://doi.org/10.7554/eLife.70169

    • PubMed
    • Google Scholar
26. 1. Heltberg ML
    2. Miné-Hattab J
    3. Taddei A
    4. Walczak AM
    5. Mora T

    (2021) Physical observables to determine the nature of membrane-less cellular sub-compartments

    eLife 10:e69181.

    https://doi.org/10.7554/eLife.69181

    • PubMed
    • Google Scholar
27. 1. Hoffman EA
    2. Frey BL
    3. Smith LM
    4. Auble DT

    (2015) Formaldehyde crosslinking: a tool for the study of chromatin complexes

    The Journal of Biological Chemistry 290:26404–26411.

    https://doi.org/10.1074/jbc.R115.651679

    • PubMed
    • Google Scholar
28. 1. Hoops S
    2. Sahle S
    3. Gauges R
    4. Lee C
    5. Pahle J
    6. Simus N
    7. Singhal M
    8. Xu L
    9. Mendes P
    10. Kummer U

    (2006) COPASI -- a complex pathway simulator

    Bioinformatics 22:3067–3074.

    https://doi.org/10.1093/bioinformatics/btl485

    • PubMed
    • Google Scholar
29. 1. Hua K
    2. Ferland RJ

    (2017) Fixation methods can differentially affect ciliary protein immunolabeling

    Cilia 6:5.

    https://doi.org/10.1186/s13630-017-0045-9

    • PubMed
    • Google Scholar
30. 1. Hyman AA
    2. Weber CA
    3. Jülicher F

    (2014) Liquid-Liquid phase separation in biology

    Annual Review of Cell and Developmental Biology 30:39–58.

    https://doi.org/10.1146/annurev-cellbio-100913-013325

    • PubMed
    • Google Scholar
31. Preprint

    1. Jachowicz JW
    2. Strehle M
    3. Banerjee AK
    4. Thai J
    5. Blanco MR
    6. Guttman M

    (2021) Xist Spatially Amplifies SHARP Recruitment to Balance Chromosome-Wide Silencing and Specificity to the X Chromosome

    bioRxiv.

    https://doi.org/10.1101/2021.10.27.466149

    • Google Scholar
32. 1. Kamps JJAG
    2. Hopkinson RJ
    3. Schofield CJ
    4. Claridge TDW

    (2019) How formaldehyde reacts with amino acids

    Communications Chemistry 2:1–14.

    https://doi.org/10.1038/s42004-019-0224-2

    • Google Scholar
33. 1. Kato M
    2. McKnight SL

    (2018) A solid-state conceptualization of information transfer from gene to message to protein

    Annual Review of Biochemistry 87:351–390.

    https://doi.org/10.1146/annurev-biochem-061516-044700

    • PubMed
    • Google Scholar
34. 1. Koga S
    2. Williams DS
    3. Perriman AW
    4. Mann S

    (2011) Peptide-Nucleotide microdroplets as a step towards a membrane-free protocell model

    Nature Chemistry 3:720–724.

    https://doi.org/10.1038/nchem.1110

    • PubMed
    • Google Scholar
35. 1. Kumar S
    2. Chaturvedi NK
    3. Kumar S
    4. Tyagi RK

    (2008) Agonist-Mediated docking of androgen receptor onto the mitotic chromatin platform discriminates intrinsic mode of action of prostate cancer drugs

    Biochimica et Biophysica Acta 1783:59–73.

    https://doi.org/10.1016/j.bbamcr.2007.11.002

    • PubMed
    • Google Scholar
36. 1. Ladouceur AM
    2. Parmar BS
    3. Biedzinski S
    4. Wall J
    5. Tope SG
    6. Cohn D
    7. Kim A
    8. Soubry N
    9. Reyes-Lamothe R
    10. Weber SC

    (2020) Clusters of bacterial RNA polymerase are biomolecular condensates that assemble through liquid-liquid phase separation

    PNAS 117:18540–18549.

    https://doi.org/10.1073/pnas.2005019117

    • PubMed
    • Google Scholar
37. 1. Lanzanò L
    2. Scipioni L
    3. Di Bona M
    4. Bianchini P
    5. Bizzarri R
    6. Cardarelli F
    7. Diaspro A
    8. Vicidomini G

    (2017) Measurement of nanoscale three-dimensional diffusion in the interior of living cells by STED-FCS

    Nature Communications 8:65.

    https://doi.org/10.1038/s41467-017-00117-2

    • PubMed
    • Google Scholar
38. 1. Lerner J
    2. Bagattin A
    3. Verdeguer F
    4. Makinistoglu MP
    5. Garbay S
    6. Felix T
    7. Heidet L
    8. Pontoglio M

    (2016) Human mutations affect the epigenetic/bookmarking function of HNF1B

    Nucleic Acids Research 44:8097–8111.

    https://doi.org/10.1093/nar/gkw467

    • PubMed
    • Google Scholar
39. 1. Li P
    2. Banjade S
    3. Cheng H-C
    4. Kim S
    5. Chen B
    6. Guo L
    7. Llaguno M
    8. Hollingsworth JV
    9. King DS
    10. Banani SF
    11. Russo PS
    12. Jiang Q-X
    13. Nixon BT
    14. Rosen MK

    (2012) Phase transitions in the assembly of multivalent signalling proteins

    Nature 483:336–340.

    https://doi.org/10.1038/nature10879

    • PubMed
    • Google Scholar
40. 1. Lorber D
    2. Volk T

    (2022) Evaluation of chromatin mesoscale organization

    APL Bioengineering 6:010902.

    https://doi.org/10.1063/5.0069286

    • PubMed
    • Google Scholar
41. 1. McSwiggen DT
    2. Hansen AS
    3. Teves SS
    4. Marie-Nelly H
    5. Hao Y
    6. Heckert AB
    7. Umemoto KK
    8. Dugast-Darzacq C
    9. Tjian R
    10. Darzacq X

    (2019a) Evidence for DNA-mediated nuclear compartmentalization distinct from phase separation

    eLife 8:e47098.

    https://doi.org/10.7554/eLife.47098

    • PubMed
    • Google Scholar
42. 1. McSwiggen DT
    2. Mir M
    3. Darzacq X
    4. Tjian R

    (2019b) Evaluating phase separation in live cells: diagnosis, caveats, and functional consequences

    Genes & Development 33:1619–1634.

    https://doi.org/10.1101/gad.331520.119

    • PubMed
    • Google Scholar
43. 1. Metz B
    2. Kersten GFA
    3. Hoogerhout P
    4. Brugghe HF
    5. Timmermans HAM
    6. de Jong A
    7. Meiring H
    8. ten Hove J
    9. Hennink WE
    10. Crommelin DJA
    11. Jiskoot W

    (2004) Identification of formaldehyde-induced modifications in proteins: reactions with model peptides

    The Journal of Biological Chemistry 279:6235–6243.

    https://doi.org/10.1074/jbc.M310752200

    • PubMed
    • Google Scholar
44. 1. Metz B
    2. Kersten GFA
    3. Baart GJE
    4. de Jong A
    5. Meiring H
    6. ten Hove J
    7. van Steenbergen MJ
    8. Hennink WE
    9. Crommelin DJA
    10. Jiskoot W

    (2006) Identification of formaldehyde-induced modifications in proteins: reactions with insulin

    Bioconjugate Chemistry 17:815–822.

    https://doi.org/10.1021/bc050340f

    • PubMed
    • Google Scholar
45. 1. Miné-Hattab J
    2. Heltberg M
    3. Villemeur M
    4. Guedj C
    5. Mora T
    6. Walczak AM
    7. Dahan M
    8. Taddei A

    (2021) Single molecule microscopy reveals key physical features of repair foci in living cells

    eLife 10:e60577.

    https://doi.org/10.7554/eLife.60577

    • PubMed
    • Google Scholar
46. 1. Mitrea DM
    2. Kriwacki RW

    (2016) Phase separation in biology; functional organization of a higher order

    Cell Communication and Signaling 14:1.

    https://doi.org/10.1186/s12964-015-0125-7

    • PubMed
    • Google Scholar
47. 1. Moter A
    2. Göbel UB

    (2000) Fluorescence in situ hybridization (fish) for direct visualization of microorganisms

    Journal of Microbiological Methods 41:85–112.

    https://doi.org/10.1016/s0167-7012(00)00152-4

    • PubMed
    • Google Scholar
48. 1. Neugebauer KM
    2. Roth MB

    (1997) Distribution of pre-mRNA splicing factors at sites of RNA polymerase II transcription

    Genes & Development 11:1148–1159.

    https://doi.org/10.1101/gad.11.9.1148

    • PubMed
    • Google Scholar
49. 1. Normanno D
    2. Boudarène L
    3. Dugast-Darzacq C
    4. Chen J
    5. Richter C
    6. Proux F
    7. Bénichou O
    8. Voituriez R
    9. Darzacq X
    10. Dahan M

    (2015) Probing the target search of DNA-binding proteins in mammalian cells using TetR as model searcher

    Nature Communications 6:7357.

    https://doi.org/10.1038/ncomms8357

    • PubMed
    • Google Scholar
50. 1. Nott TJ
    2. Petsalaki E
    3. Farber P
    4. Jervis D
    5. Fussner E
    6. Plochowietz A
    7. Craggs TD
    8. Bazett-Jones DP
    9. Pawson T
    10. Forman-Kay JD
    11. Baldwin AJ

    (2015) Phase transition of a disordered nuage protein generates environmentally responsive membraneless organelles

    Molecular Cell 57:936–947.

    https://doi.org/10.1016/j.molcel.2015.01.013

    • PubMed
    • Google Scholar
51. 1. Owen I
    2. Yee D
    3. Wyne H
    4. Perdikari TM
    5. Johnson V
    6. Smyth J
    7. Kortum R
    8. Fawzi NL
    9. Shewmaker F

    (2021) The oncogenic transcription factor FUS-CHOP can undergo nuclear liquid-liquid phase separation

    Journal of Cell Science 134:jcs258578.

    https://doi.org/10.1242/jcs.258578

    • PubMed
    • Google Scholar
52. 1. Pallier C
    2. Scaffidi P
    3. Chopineau-Proust S
    4. Agresti A
    5. Nordmann P
    6. Bianchi ME
    7. Marechal V

    (2003) Association of chromatin proteins high mobility group box (HMGB) 1 and HMGB2 with mitotic chromosomes

    Molecular Biology of the Cell 14:3414–3426.

    https://doi.org/10.1091/mbc.e02-09-0581

    • PubMed
    • Google Scholar
53. 1. Patel A
    2. Malinovska L
    3. Saha S
    4. Wang J
    5. Alberti S
    6. Krishnan Y
    7. Hyman AA

    (2017) Atp as a biological hydrotrope

    Science 356:753–756.

    https://doi.org/10.1126/science.aaf6846

    • PubMed
    • Google Scholar
54. 1. Pollard TD

    (2010) A guide to simple and informative binding assays

    Molecular Biology of the Cell 21:4061–4067.

    https://doi.org/10.1091/mbc.E10-08-0683

    • PubMed
    • Google Scholar
55. 1. Poorey K
    2. Viswanathan R
    3. Carver MN
    4. Karpova TS
    5. Cirimotich SM
    6. McNally JG
    7. Bekiranov S
    8. Auble DT

    (2013) Measuring chromatin interaction dynamics on the second time scale at single-copy genes

    Science 342:369–372.

    https://doi.org/10.1126/science.1242369

    • PubMed
    • Google Scholar
56. 1. Riback JA
    2. Zhu L
    3. Ferrolino MC
    4. Tolbert M
    5. Mitrea DM
    6. Sanders DW
    7. Wei M-T
    8. Kriwacki RW
    9. Brangwynne CP

    (2020) Composition-dependent thermodynamics of intracellular phase separation

    Nature 581:209–214.

    https://doi.org/10.1038/s41586-020-2256-2

    • PubMed
    • Google Scholar
57. 1. Richter KN
    2. Revelo NH
    3. Seitz KJ
    4. Helm MS
    5. Sarkar D
    6. Saleeb RS
    7. D’Este E
    8. Eberle J
    9. Wagner E
    10. Vogl C
    11. Lazaro DF
    12. Richter F
    13. Coy-Vergara J
    14. Coceano G
    15. Boyden ES
    16. Duncan RR
    17. Hell SW
    18. Lauterbach MA
    19. Lehnart SE
    20. Moser T
    21. Outeiro TF
    22. Rehling P
    23. Schwappach B
    24. Testa I
    25. Zapiec B
    26. Rizzoli SO

    (2018) Glyoxal as an alternative fixative to formaldehyde in immunostaining and super-resolution microscopy

    The EMBO Journal 37:139–159.

    https://doi.org/10.15252/embj.201695709

    • PubMed
    • Google Scholar
58. 1. Robertson G
    2. Hirst M
    3. Bainbridge M
    4. Bilenky M
    5. Zhao Y
    6. Zeng T
    7. Euskirchen G
    8. Bernier B
    9. Varhol R
    10. Delaney A
    11. Thiessen N
    12. Griffith OL
    13. He A
    14. Marra M
    15. Snyder M
    16. Jones S

    (2007) Genome-wide profiles of STAT1 DNA association using chromatin immunoprecipitation and massively parallel sequencing

    Nature Methods 4:651–657.

    https://doi.org/10.1038/nmeth1068

    • PubMed
    • Google Scholar
59. 1. Rust MJ
    2. Bates M
    3. Zhuang X

    (2006) Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (storm)

    Nature Methods 3:793–795.

    https://doi.org/10.1038/nmeth929

    • PubMed
    • Google Scholar
60. 1. Sacchetti A
    2. Subramaniam V
    3. Jovin TM
    4. Alberti S

    (2002) Oligomerization of DsRed is required for the generation of a functional red fluorescent chromophore

    FEBS Letters 525:13–19.

    https://doi.org/10.1016/s0014-5793(02)02874-0

    • PubMed
    • Google Scholar
61. Conference

    1. Sage D
    2. Prodanov D
    3. Tinevez JY
    4. Schindelin J

    (2012)

    MIJ: making interoperability between ImageJ and Matlab possible

    ImageJ User & Developer Conference.

    • Google Scholar
62. 1. Schindelin J
    2. Arganda-Carreras I
    3. Frise E
    4. Kaynig V
    5. Longair M
    6. Pietzsch T
    7. Preibisch S
    8. Rueden C
    9. Saalfeld S
    10. Schmid B
    11. Tinevez J-Y
    12. White DJ
    13. Hartenstein V
    14. Eliceiri K
    15. Tomancak P
    16. Cardona A

    (2012) Fiji: an open-source platform for biological-image analysis

    Nature Methods 9:676–682.

    https://doi.org/10.1038/nmeth.2019

    • PubMed
    • Google Scholar
63. 1. Schmiedeberg L
    2. Skene P
    3. Deaton A
    4. Bird A

    (2009) A temporal threshold for formaldehyde crosslinking and fixation

    PLOS ONE 4:e4636.

    https://doi.org/10.1371/journal.pone.0004636

    • PubMed
    • Google Scholar
64. 1. Schnell U
    2. Dijk F
    3. Sjollema KA
    4. Giepmans BNG

    (2012) Immunolabeling artifacts and the need for live-cell imaging

    Nature Methods 9:152–158.

    https://doi.org/10.1038/nmeth.1855

    • PubMed
    • Google Scholar
65. 1. Sergé A
    2. Bertaux N
    3. Rigneault H
    4. Marguet D

    (2008) Dynamic multiple-target tracing to probe spatiotemporal cartography of cell membranes

    Nature Methods 5:687–694.

    https://doi.org/10.1038/nmeth.1233

    • PubMed
    • Google Scholar
66. 1. Shin Y
    2. Brangwynne CP

    (2017) Liquid phase condensation in cell physiology and disease

    Science 357:eaaf4382.

    https://doi.org/10.1126/science.aaf4382

    • PubMed
    • Google Scholar
67. 1. Shishodia S
    2. Zhang D
    3. El-Sagheer AH
    4. Brown T
    5. Claridge TDW
    6. Schofield CJ
    7. Hopkinson RJ

    (2018) Nmr analyses on N-hydroxymethylated nucleobases-implications for formaldehyde toxicity and nucleic acid demethylases

    Organic & Biomolecular Chemistry 16:4021–4032.

    https://doi.org/10.1039/c8ob00734a

    • PubMed
    • Google Scholar
68. 1. Solomon MJ
    2. Varshavsky A

    (1985) Formaldehyde-mediated DNA-protein crosslinking: a probe for in vivo chromatin structures

    PNAS 82:6470–6474.

    https://doi.org/10.1073/pnas.82.19.6470

    • PubMed
    • Google Scholar
69. 1. Stadler C
    2. Skogs M
    3. Brismar H
    4. Uhlén M
    5. Lundberg E

    (2010) A single fixation protocol for proteome-wide immunofluorescence localization studies

    Journal of Proteomics 73:1067–1078.

    https://doi.org/10.1016/j.jprot.2009.10.012

    • PubMed
    • Google Scholar
70. 1. Stanly TA
    2. Fritzsche M
    3. Banerji S
    4. García E
    5. Bernardino de la Serna J
    6. Jackson DG
    7. Eggeling C

    (2016) Critical importance of appropriate fixation conditions for faithful imaging of receptor microclusters

    Biology Open 5:1343–1350.

    https://doi.org/10.1242/bio.019943

    • PubMed
    • Google Scholar
71. 1. Sutherland BW
    2. Toews J
    3. Kast J

    (2008) Utility of formaldehyde cross-linking and mass spectrometry in the study of protein-protein interactions

    Journal of Mass Spectrometry 43:699–715.

    https://doi.org/10.1002/jms.1415

    • PubMed
    • Google Scholar
72. 1. Teves SS
    2. An L
    3. Hansen AS
    4. Xie L
    5. Darzacq X
    6. Tjian R

    (2016) A dynamic mode of mitotic bookmarking by transcription factors

    eLife 5:e22280.

    https://doi.org/10.7554/eLife.22280

    • PubMed
    • Google Scholar
73. 1. Tokunaga M
    2. Imamoto N
    3. Sakata-Sogawa K

    (2008) Highly inclined thin illumination enables clear single-molecule imaging in cells

    Nature Methods 5:159–161.

    https://doi.org/10.1038/nmeth1171

    • PubMed
    • Google Scholar
74. 1. Trojanowski J
    2. Frank L
    3. Rademacher A
    4. Mücke N
    5. Grigaitis P
    6. Rippe K

    (2022) Transcription activation is enhanced by multivalent interactions independent of phase separation

    Molecular Cell 82:1878–1893.

    https://doi.org/10.1016/j.molcel.2022.04.017

    • PubMed
    • Google Scholar
75. 1. Wang Jie
    2. Choi J-M
    3. Holehouse AS
    4. Lee HO
    5. Zhang X
    6. Jahnel M
    7. Maharana S
    8. Lemaitre R
    9. Pozniakovsky A
    10. Drechsel D
    11. Poser I
    12. Pappu RV
    13. Alberti S
    14. Hyman AA

    (2018) A molecular grammar governing the driving forces for phase separation of prion-like RNA binding proteins

    Cell 174:688–699.

    https://doi.org/10.1016/j.cell.2018.06.006

    • PubMed
    • Google Scholar
76. 1. Xie J
    2. He H
    3. Kong W
    4. Li Z
    5. Gao Z
    6. Xie D
    7. Sun L
    8. Fan X
    9. Jiang X
    10. Zheng Q
    11. Li G
    12. Zhu J
    13. Zhu G

    (2022) Targeting androgen receptor phase separation to overcome antiandrogen resistance

    Nature Chemical Biology 18:1341–1350.

    https://doi.org/10.1038/s41589-022-01151-y

    • PubMed
    • Google Scholar
77. 1. Yang P
    2. Mathieu C
    3. Kolaitis RM
    4. Zhang P
    5. Messing J
    6. Yurtsever U
    7. Yang Z
    8. Wu J
    9. Li Y
    10. Pan Q
    11. Yu J
    12. Martin EW
    13. Mittag T
    14. Kim HJ
    15. Taylor JP

    (2020) G3Bp1 is a tunable switch that triggers phase separation to assemble stress granules

    Cell 181:325–345.

    https://doi.org/10.1016/j.cell.2020.03.046

    • PubMed
    • Google Scholar
78. 1. Yewdall NA
    2. André AAM
    3. Lu T
    4. Spruijt E

    (2021) Coacervates as models of membraneless organelles

    Current Opinion in Colloid & Interface Science 52:101416.

    https://doi.org/10.1016/j.cocis.2020.101416

    • Google Scholar
79. 1. Zarębski M
    2. Bosire R
    3. Wesołowska J
    4. Szelest O
    5. Eatmann A
    6. Jasińska-Konior K
    7. Kepp O
    8. Kroemer G
    9. Szabo G
    10. Dobrucki JW

    (2021) Translocation of chromatin proteins to nucleoli-the influence of protein dynamics on post-fixation localization

    Cytometry. Part A 99:1230–1239.

    https://doi.org/10.1002/cyto.a.24464

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Fixation Can Change the Appearance of Phase Separation in Living Cells" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Felix Campelo as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Vivek Malhotra as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Judith Miné-Hattab (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

In the discussion amongst the reviewers, we all agreed that this is a very interesting and important paper. But the paper needs some clarifications and extra work. Specifically, the major points that should be addressed follow.

1) Compare different fixation methods/concentrations, etc. The authors only report on the artifacts caused by PFA at 4% (with 10' incubation time). They should test other concentrations and other fixatives. In particular, work from the Eggeling lab showed a similar fixation artifact on cell surface receptor clustering (https://journals.biologists.com/bio/article/5/9/1343/1227/Critical-importance-of-appropriate-fixation), which was somehow minimized by adding glutaraldehyde to the fixation buffer. The authors should also test this.

2) Better relate the experimental observations with the dynamic model. As suggested (see detailed reports below), this could be by measuring diffusion coefficients (e.g. FRAP) and see if that matches with the model's predictions.

3) Improve the description of LLPS in the introduction (see details below).

4) Finally, although we do not require the authors to test how other cellular structures are affected by fixation, the authors should add a short section in the introduction to mention some examples of other kinds of structures that are not well conserved by the fixation. Otherwise, it gives the impression that fixation artifacts exist only for LLPS, unfortunately, the preservation of structures by fixation is not limited to LLPS. They should also discuss the fact that some fixation protocols can destroy some structures while another protocol will preserve them. It is the case for some filament structures such as actin filament which are preserved by methanol fixation but seem altered by PFA fixation. This part would help the reader to understand that the quality of a fixation protocol strongly depends on the type of structure studied.

Reviewer #1 (Recommendations for the authors):

I think this is an important paper that presents an important observation that will be important for the community. My main concern/question is whether the two parts of the paper (experimental observations and computational model) are connected causally or not. I think that the glycine experiments point in that direction, but I fail to see concluding evidence on whether the observed changes in the LLPS structures after fixation are indeed caused by slower/faster fixation rates in/out of the condensates. Along these lines:

– Line 111: "The fact that different phase-separating proteins can have bifurcating behaviors upon fixation is interesting.": I fully agree with this. Have the authors considered coexpressing in the same cells the same protein with two different tags (that behave differently after fixation), such as GFP-TAF15 and dsRed-TAF15; or probably even better, dsRED and Halo TAF15? If the kinetic model represents the experimental situation, wouldn't you expect that fixation leads still to the appearance of small droplets in the dsRED but not in the Halo protein?

– Figure 4: glycine also seems to cause a change in the fraction of droplet protein in live cells (compare left panels in A and B). Could the authors discuss that?

Related to the model:

– Line 154: k2 in the model appears as a volumetric rate (that is, all particles in the droplet have the same escape probability). Would a surface escape rate (only particles at the surface are able to escape) change the results of the model?

– Can the authors provide analytical expressions (I believe that is relatively simple) for the plots in 5C, D as a function of the relative in puncta fix. rate and the relative overall fix. rate?

Reviewer #2 (Recommendations for the authors):

To strengthen the manuscript, the authors should try more protocols of fixation. In the simulation part, they could try to incorporate the diffusion coefficient of the protein of interest and see if it is possible to predict the effect of fixation as a function of the diffusion coefficient.

The manuscript focuses on LLPS but it would be interesting to discuss other artefacts of fixation outside of the LLPS: have they tested the artefacts on structures like filaments, chromatin organization, or other types of structures than LLPS? Or do fixation artefacts only concern LLPS?

Reviewer #3 (Recommendations for the authors):

Proteins that undergo LLPS in living cells show a very dynamic behavior and rapidly move from the biomolecular condensates to the surrounding environment (e.g. cytoplasm or nucleoplasm), as demonstrated using FRAP. This dynamic behavior could explain why when comparing fluorescently tagged IDR proteins in living cells and in fixed cells, one could detect only the "larger" condensates in living cells, while detecting also smaller condensates in the fixed cells. Comparison of the number and size of condensates detected in living cells using conventional confocal microscopy and super-resolution microscopy will help understand whether this is sufficient to increase the number of smaller condensates detected in living cells. If not, this would support the idea proposed by the authors that "when the overall fixation rate is fast compared with the dynamics of targeted interactions, fixation artifacts can be minimized even with unequal fixation rates in and out of puncta."

Not all proteins analyzed showed a different distribution in living versus fixed cells (as shown in Figure 3 for GFP-FUS). The differences in the number and size of condensates observed in living and fixed cells should be correlated with the dynamic of the protein analyzed by FRAP. Are proteins with the highest mobility measured by FRAP corresponding to those that show an increased number/size of puncta upon fixation? Are FUS protein-protein interactions stable and less dynamic compared with the overall fixation rate?

Recommendations for improving the writing and presentation: Defining whether a protein undergoes LLPS is based on different assays in vitro, using recombinant proteins and in cells. The observation that a given protein forms "puncta" inside the cells is generally not accepted as a criterion to establish whether it undergoes LLPS. The measure of the number, size, sphericity, liquid-like dynamic behavior (e.g. by FRAP), and sensitivity to agents such as e.g. hexanediol are all assays required to establish and characterize whether a given protein undergoes LLPS. These aspects should be described in the introduction of the paper. It is a bit simplistic to only focus on the calculation of the number and size of puncta before/after fixation.

Essential revisions:

In the discussion amongst the reviewers, we all agreed that this is a very interesting and important paper. But the paper needs some clarifications and extra work. Specifically, the major points that should be addressed follow.

1) Compare different fixation methods/concentrations, etc. The authors only report on the artifacts caused by PFA at 4% (with 10' incubation time). They should test other concentrations and other fixatives. In particular, work from the Eggeling lab showed a similar fixation artifact on cell surface receptor clustering (https://journals.biologists.com/bio/article/5/9/1343/1227/Critical-importance-of-appropriate-fixation), which was somehow minimized by adding glutaraldehyde to the fixation buffer. The authors should also test this.

We thank the reviewers for this constructive comment and agree that various concentrations of PFA along with PFA supplemented by glutaraldehyde (GA) should be tested for potential fixation artifacts. Following the reviewer’s suggestions, we have performed all the recommended experiments and included new data in the revised manuscript to address this concern.

1) We quantified the artifact resulting from fixation using 0% PFA (PBS buffer only), 1% PFA, 2% PFA, and 8% PFA besides 4% PFA. We showed that all concentrations of PFA other than 0% resulted in statistically significant change in all the parameters that quantify LLPS appearance (“number of puncta”, “surface roughness”, and “punctate percentage”), though a quantitative comparison of parameter changes at different PFA concentrations is difficult due to increased fluorescence quenching effects at higher PFA concentrations. The description of our new data was added to the revised manuscript in Paragraph 3 on Page 3 and in Figure 1 —figure supplement 3.

2) We quantified the artifact resulting from fixation using 4% PFA supplemented with 0.2% GA, the same fixative used in (https://journals.biologists.com/bio/article/5/9/1343/1227/Critical-importance-of-appropriatefixation). We found a statistically significant change in all the LLPS-describing parameters. We also compared the fixation artifact resulting from PFA/GA in combination with the fixation artifact resulting from PFA alone and found no statistically significant differences. The description of our new data was added to the revised manuscript in Paragraph 3 on Page 3 and in Figure 1 —figure supplement 4.

Overall, these new results showed that fixation artifacts for LLPS systems are not unique to fixation with 4% PFA; rather, they can result from a broad range of PFA concentrations and cannot be eliminated by the addition of GA.

2) Better relate the experimental observations with the dynamic model. As suggested (see detailed reports below), this could be by measuring diffusion coefficients (e.g. FRAP) and see if that matches with the model's predictions.

We thank the reviewers for identifying this disconnect and agree that experiments probing the dynamics are needed to bridge the gap between our experimental observations and kinetic model of the fixation effects. Following the reviewers’ suggestions, we have now included a significant amount of new data in the revised manuscript to address this concern.

Our model does not predict the impact of diffusion dynamics on cell fixation, instead, it states that association rate, dissociation rate, in-puncta fixation rate, and out-of-puncta fixation rate together determine the fixation artifact. In particular, our model predicts that a fast overall fixation rate relative to binding dynamics can result in a reduced fixation artifact. In the revised manuscript, we have performed a series of new imaging experiments to prove this important prediction. The two proteins we focused on are Halo-TAF15(IDR) that experiences a significant fixation artifact and TAF15(IDR)-Halo-FTH1 that does not experience any fixation artifact (Figures 2C and 3B in the revised manuscript). Our model predicts that compared with Halo-TAF15(IDR), TAF15(IDR)-Halo-FTH1 has a faster overall fixation rate relative to binding dynamics. Since fixation of both proteins are completed within 1~2 minutes, it is predicted that TAF15(IDR)-Halo-FTH1 binds to its droplet-like puncta less dynamically than Halo-TAF15(IDR) does. We used single-particle tracking (SPT) to measure the residence times of Halo-TAF15(IDR) and TAF15(IDR)-Halo-FTH1 in respective puncta and showed that TAF15(IDR)-Halo-FTH1 binds to its puncta significantly longer than Halo-TAF15(IDR) (Figure 6 and Video 2 in the revised manuscript), suggesting significantly more stable homotypic interactions of the former than the latter and a higher relative overall fixation rate of the former than the latter. In short, our experimental observations are perfectly consistent with the model prediction.

The fixation artifact of TAF15(IDR)-Halo-FTH1 was described in Paragraph 1 on Page 8 and Figures 3B and 3E of the revised manuscript. The description of our SPT results and how they are related to our kinetic model were added to Paragraph 3 on Page 13, Figure 6, and Video 2. We have also updated Figure 5D to show the simulation of cell fixation with diminished rather than enhanced LLPS to be consistent with our observation for Halo-TAF15(IDR). This makes it easier for the readers to compare our experimental observations to our model simulations.

3) Improve the description of LLPS in the introduction (see details below).

We thank the reviewers for this suggestion and have updated the description of LLPS in the introduction in Paragraph 2 on Page 2 as follows:

“Whereas rigorous characterization of LLPS in vivo has been challenging and remains a question under active investigation (McSwiggen et al., 2019b), detection of discrete puncta that have a spherical shape, undergo fusion and fission, and dynamically exchange biomolecules with the surrounding according to FRAP is often considered evidence of putative LLPS in living cells. While such diverse measurements have been widely used for studying proteins under overexpression conditions, far fewer approaches are available to probe LLPS under physiological conditions. Detecting local high-concentration regions or puncta of an endogenously expressed protein using immunofluorescence of fixed cells has been used in many studies as evidence of LLPS ...”

4) Finally, although we do not require the authors to test how other cellular structures are affected by fixation, the authors should add a short section in the introduction to mention some examples of other kinds of structures that are not well conserved by the fixation. Otherwise, it gives the impression that fixation artifacts exist only for LLPS, unfortunately, the preservation of structures by fixation is not limited to LLPS. They should also discuss the fact that some fixation protocols can destroy some structures while another protocol will preserve them. It is the case for some filament structures such as actin filament which are preserved by methanol fixation but seem altered by PFA fixation. This part would help the reader to understand that the quality of a fixation protocol strongly depends on the type of structure studied.

We thank the reviewers for suggesting that we clarify that fixation artifacts are not exclusive to LLPS and that the quality of a fixation protocol strongly depends on the type of structure being studied. We have updated the Discussion section to address this concern in Paragraph 4 on Page 15:

“We emphasize that because our four-state model makes no assumptions about any state being phase-separated, the logical implications of our model can extend beyond LLPS to other biomolecular transactions and cellular structures that have been found not well preserved by fixation or immunofluorescence, including localizations of cilia proteins (Hua and Ferland, 2017), clustering of cell membrane receptors (Stanly et al., 2016), splicing speckle formation (Neugebauer and Roth, 1997), and chromatin organization and protein binding (Zarębski et al., 2021; Lorber and Volk, 2022, Lerner et al., 2016; Pallier et al., 2003; Kumar et al., 2007; and Teves et al., 2016). Our model can similarly extend beyond PFA to other fixatives. This is useful because different fixatives have been chosen for studying different types of structures. For example, PFA fixation is often preferable for preserving soluble proteins over dehydration fixatives such as methanol (Stadler et al., 2010 and Schnell et al., 2012), yet methanol fixation can be preferable over PFA for preserving proteins bound to mitotic chromatin (Kumar et al., 2007 and Lerner et al., 2016)”.

Reviewer #1 (Recommendations for the authors):

I think this is an important paper that presents an important observation that will be important for the community. My main concern/question is whether the two parts of the paper (experimental observations and computational model) are connected causally or not. I think that the glycine experiments point in that direction, but I fail to see concluding evidence on whether the observed changes in the LLPS structures after fixation are indeed caused by slower/faster fixation rates in/out of the condensates.

We thank the reviewer for sharing this concern/question. We’d like to start by explaining the importance of our 4-state kinetic model involving fixed states in and out of puncta. While excellent previous works (Poorey et al., 2013; Schmiedeberg et al., 2009; and Teves et al., 2016) have proposed models that explain their respective observed fixation artifacts, their models are unable to capture how fixation has the capability to both enhance and diminish LLPS behavior that we observed experimentally. The inability of previous models to describe our experimental results motivated us to propose the 4-state kinetic model, which perfectly simulates the bifurcating behaviors. Although there is currently not a method capable of directly measuring fixation rates in and out of puncta in a cell, difference between the two fixation rates is well supported by well-established facts that (1) the dilute and concentrated phases of a LLPS system have different protein composition and concentrations (Currie and Rosen, 2022; Koga et al., 2011; Nott et al., 2015; Yewdall et al., 2021) and (2) fixation rate varies with both protein composition and concentration by orders of magnitude (Hoffman et al., 2015; Kamps et al., 2019; Metz et al., 2006; Metz et al., 2004). We have listed the facts in Paragraph 2 on Page 11 in the revised manuscript. Despite the infeasibility of measuring fixation rates in and out of puncta and showing experimentally the difference causes a fixation artifact, we have performed SPT experiments to measure the binding dynamics of different proteins within respective puncta and showed that a fast overall fixation rate relative to binding dynamics can minimize fixation artifacts, proving a different critical prediction by our model. We believe our new data successfully bridged the previous gap between experiments and the model. Please see Response 2 for the details of our additional experiments to address the concern.

Along these lines:

– Line 111: "The fact that different phase‐separating proteins can have bifurcating behaviors upon fixation is interesting.": I fully agree with this. Have the authors considered coexpressing in the same cells the same protein with two different tags (that behave differently after fixation), such as GFP-TAF15 and dsRed-TAF15; or probably even better, dsRED and Halo TAF15? If the kinetic model represents the experimental situation, wouldn't you expect that fixation leads still to the appearance of small droplets in the dsRED but not in the Halo protein?

We thank the reviewer for this suggestion. Although we agree that this would make a very interesting experiment, this experimental design is unfortunately not feasible in practice. Because TAF15 undergoes strong homotypic interactions (Chong et al., 2018), TAF15 labeled with two different tags, e.g., dsRED and Halo, will colocalize in the same puncta instead of forming separate puncta each containing dsRED-TAF15 or Halo-TAF15 only. Such a homotypic interaction capability is ubiquitous for IDRs with a LLPS potential. Thus, our kinetic model can only account for one protein species in the cell at a time and the model is unsuitable to describe the fixation artifacts for the suggested case.

– Figure 4: glycine also seems to cause a change in the fraction of droplet protein in live cells (compare left panels in A and B). Could the authors discuss that?

We thank the reviewer for this suggestion and have added data and discussion in Paragraph 1 on Page 10 as follows:

“We found that adding 25mM glycine to live U2OS cells that overexpress DsRed2-TAF15(IDR) increases the starting punctate percentage from 18 ± 1.92 to 36 ± 3.82% (quantified from 23 cells), indicating an increase in the degree of LLPS. Although the underlying mechanism of such increase is unclear, we speculate this might be because hydrophobic intermolecular contacts that play an important role in TAF15(IDR) LLPS (Patel et al., 2017) are enhanced by the presence of hydrophobic glycine.”

Related to the model:

– Line 154: k2 in the model appears as a volumetric rate (that is, all particles in the droplet have the same escape probability). Would a surface escape rate (only particles at the surface are able to escape) change the results of the model?

We thank the reviewer for this interesting question. Our model does not explicitly model the spatial distribution of proteins; rather, it only considers the average exchange rates between ${\displaystyle S_1}$ and ${\displaystyle S_2}$ at the population level. This means that rather than separately considering the potentially different dissociation rates of POI molecules on the surface and interior of a punctum, an ensemble dissociate rate of POI molecules from puncta is used here. An in-depth model that distinguishes between surface and interior dissociation rates could certainly provide further insight into the makeup of the ensemble dissociation rate but would ultimately not change the simulation results of our model.

We have added discussion clarifying this point in the revised manuscript in Paragraph 1 on Page 11:

“These are the average exchange rates between ${\displaystyle S_1}$ and ${\displaystyle S_2}$ and do not concern the potential spatial inhomogeneity in the rates at the molecular level. For example, individual POI molecules at the surface and interior of a punctum might dissociate with different rates, but our model does not differentiate these molecules.”

– Can the authors provide analytical expressions (I believe that is relatively simple) for the plots in 5C, D as a function of the relative in puncta fix. rate and the relative overall fix. rate?

We thank the reviewer for this suggestion. We have provided the analytical solutions to the differential equations posed by our model below. Since they turned out to be too cumbersome to provide additional insight to readers, we chose to numerically calculate our results in the manuscript as Figure 5C and 5D.

Here, we present analytical solutions to the following systems of differential equations.

${\displaystyle \{\begin{array}{rl}\frac{d{S}_1}{dt}& =k_1{S}_2-k_2{S}_1-k_3{S}_1\\ \frac{d{S}_2}{dt}& =-k_1{S}_2+k_2{S}_1-k_4{S}_2\\ \frac{d{S}_3}{dt}& =k_3{S}_1\\ \frac{d{S}_4}{dt}& =k_4{S}_2\end{array}}$

We have

${\displaystyle \{\begin{array}{rl}{S}_1(t)& =-\frac{C_4}{k_2{k}_4}\left(k_1{k}_3+k_2{k}_4+k_3{k}_4+\left(k_2+k_3\right){\xi }_{+}\right){\mathrm{e}}^{t{\xi }_{+}}-\frac{C_3}{k_2{k}_4}\left(k_1{k}_3+k_2{k}_4+k_3{k}_4+\left(k_2+k_3\right){\xi }_{-}\right){\mathrm{e}}^{t{\xi }_{-}}\\ S_2(t)& =\frac{C_4}{k_4}{\xi }_{+}{\mathrm{e}}^{t{\xi }_{+}}+\frac{C_3}{k_4}{\xi }_{-}{\mathrm{e}}^{t{\xi }_{-}}\\ S_3(t)& =C_1+\frac{C_4{k}_3}{k_2{k}_4}\left(k_1+k_4+{\xi }_{+}\right){\mathrm{e}}^{t{\xi }_{+}}+\frac{C_3{k}_3}{k_2{k}_4}\left(k_1+k_4+{\xi }_{-}\right){\mathrm{e}}^{t{\xi }_{-}}\\ S_4(t)& =C_2+C_4{\mathrm{e}}^{t{\xi }_{+}}+C_3{\mathrm{e}}^{t{\xi }_{-}}\end{array}}$

Where C₁, C₂, C₃ and C₄ are constants dependent on the initial conditions and where

${\displaystyle {\xi }_{\pm }=\pm \frac{\sqrt{k_1^2+2k_1{k}_2-2k_1{k}_3+2k_1{k}_4+k_2^2+2k_2{k}_3-2k_2{k}_4+k_3^2-2k_3{k}_4+k_4^2}}{2}-\frac{\left(k_1+k_2+k_3+k_4\right)}{2}}$

Reviewer #2 (Recommendations for the authors):

To strengthen the manuscript, the authors should try more protocols of fixation. In the simulation part, they could try to incorporate the diffusion coefficient of the protein of interest and see if it is possible to predict the effect of fixation as a function of the diffusion coefficient.

The manuscript focuses on LLPS but it would be interesting to discuss other artefacts of fixation outside of the LLPS: have they tested the artefacts on structures like filaments, chromatin organization, or other types of structures than LLPS? Or do fixation artefacts only concern LLPS?

We thank the reviewer for raising this good point. This current work focuses on studying the effects of cell fixation on the appearance of LLPS. Although we do not investigate fixation artifacts of other cellular structures here, we have now discussed previously characterized fixation artifacts on other structures in the revised manuscript, e.g, localizations of cilia proteins (Hua and Ferland, 2017), clustering of cell membrane receptors (Stanly et al., 2016), splicing speckle formation (Neugebauer and Roth=, 1997), and chromatin organization and protein binding (Zarębski et al., 2021; Lorber and Volk, 2022, Lerner et al., 2016; Pallier et al., 2003; Kumar et al., 2007; and Teves et al., 2016). As also mentioned in Response 4, we have discussed this important point in Paragraph 4 on Page 15.

Reviewer #3 (Recommendations for the authors):

Proteins that undergo LLPS in living cells show a very dynamic behavior and rapidly move from the biomolecular condensates to the surrounding environment (e.g. cytoplasm or nucleoplasm), as demonstrated using FRAP. This dynamic behavior could explain why when comparing fluorescently tagged IDR proteins in living cells and in fixed cells, one could detect only the "larger" condensates in living cells, while detecting also smaller condensates in the fixed cells. Comparison of the number and size of condensates detected in living cells using conventional confocal microscopy and super-resolution microscopy will help understand whether this is sufficient to increase the number of smaller condensates detected in living cells. If not, this would support the idea proposed by the authors that "when the overall fixation rate is fast compared with the dynamics of targeted interactions, fixation artifacts can be minimized even with unequal fixation rates in and out of puncta."

We appreciate the reviewer for this interesting idea and agree that it is possible that there are sub-diffraction limit puncta of DsRed2-TAF15(IDR) in cells, which were not detected using conventional confocal microscopy, and these small puncta potentially have more dynamic protein dynamics than the larger puncta and can potentially be affected by fixation more significantly. Although we agree that a comparison of DsRed2-TAF15(IDR) distribution under confocal and super-resolution microscopy would be a valuable set of experiments, performing super-resolution imaging of DsRed2 would require highly specialized equipment (SIM or STED microscope), which we do not have access to. Performing these experiments will require establishment of new external collaborations, which means an unpredictable amount of work and time. Besides, quantifying protein dynamics within sub-diffraction limit puncta would be a very challenging task if feasible at all. Thus, this work is more appropriate for a future study.

Nevertheless, although we are not immediately able to differentiate the fixation artifacts on potential small, confocal-invisible puncta and on large puncta, our confocal microscopy data sufficiently supported our conclusion that the apparent LLPS behavior of DsRed2-TAF15(IDR) is overall enhanced by PFA fixation.

Not all proteins analyzed showed a different distribution in living versus fixed cells (as shown in Figure 3 for GFP-FUS). The differences in the number and size of condensates observed in living and fixed cells should be correlated with the dynamic of the protein analyzed by FRAP. Are proteins with the highest mobility measured by FRAP corresponding to those that show an increased number/size of puncta upon fixation? Are FUS protein-protein interactions stable and less dynamic compared with the overall fixation rate?

We thank the reviewer for these important questions and agree that the severity of the fixation artifact should be correlated with a protein’s dynamics in the cell. Our kinetic model of fixation concerns protein-protein interaction dynamics instead of protein diffusion dynamics. Indeed, the model predicts that proteins with more dynamic interactions have more fixation artifacts, if the proteins of interest have similar absolute fixation rates. Since FRAP dynamics are contributed by both diffusion and interaction dynamics, which are often not separable or quantifiable from FRAP measurements (https://www.cell.com/biophysj/fulltext/S0006-3495(04)74392-1), we instead used SPT to directly measure protein interaction dynamics in the revised manuscript. As described in Response 2 and Response 17, we measured the dissociation rates of two proteins from respective puncta using SPT: Halo-TAF15(IDR), which is poorly preserved by fixation, and TAF15(IDR)-Halo-FTH1, which is well preserved by fixation. We found that the dissociation rate of Halo-TAF15(IDR) is much faster than that of TAF15(IDR)-Halo-FTH1, demonstrating more dynamic interactions of Halo-TAF15(IDR) than TAF15(IDR)-Halo-FTH1. Our observations that Halo-TAF15(IDR) has more dynamic interactions and is more poorly preserved by fixation than TAF15(IDR)-Halo-FTH1 agrees very well with our model’s prediction. Please see Response 2 and Response 17 for more details. Our new data and discussion have been added to the revised manuscript in Paragraph 3 on Page 13 and in Figure 3B, Figure 3E, Figure 6, and Video 2.

Recommendations for improving the writing and presentation: Defining whether a protein undergoes LLPS is based on different assays in vitro, using recombinant proteins and in cells. The observation that a given protein forms "puncta" inside the cells is generally not accepted as a criterion to establish whether it undergoes LLPS. The measure of the number, size, sphericity, liquid-like dynamic behavior (e.g. by FRAP), and sensitivity to agents such as e.g. hexanediol are all assays required to establish and characterize whether a given protein undergoes LLPS. These aspects should be described in the introduction of the paper. It is a bit simplistic to only focus on the calculation of the number and size of puncta before/after fixation.

We thank the reviewer for their suggestions. We agree that although detecting puncta in immunofluorescence images have been the most widely used approach for detecting LLPS for endogenous proteins, many other assays have been used for characterizing LLPS for overexpression systems in cells. We have now included the commonly used LLPS characteristics in cells in the Introduction (Paragraph 2 on Page 2) while describing our motivation of examining the effects of cell fixation on LLPS appearance: “Whereas rigorous characterization of LLPS in vivo has been challenging and remains a question under active investigation (McSwiggen et al., 2019b), detection of discrete puncta that have a spherical shape, undergo fusion and fission, and dynamically exchange biomolecules with the surrounding according to FRAP is often considered evidence of putative LLPS in living cells. While such diverse measurements have been widely used for studying proteins under overexpression conditions, far fewer approaches are available to probe LLPS under physiological conditions. Detecting local high-concentration regions or puncta of an endogenously expressed protein using immunofluorescence of fixed cells has been used in many studies as evidence of LLPS... Not only is the detection of puncta an inconclusive metric for establishing LLPS, whether a punctate distribution observed in fixed cells actually represents the live-cell scenario remains unclear…”

Author details

1. Shawn Irgen-Gioro

   Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, United States

   Contribution

   Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   Contributed equally with

   Shawn Yoshida

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8638-6191

2. Shawn Yoshida

   1. Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, United States
   2. Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, United States

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – review and editing, Writing – original draft

   Contributed equally with

   Shawn Irgen-Gioro

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0866-2741

3. Victoria Walling

   Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, United States

   Contribution

   Investigation

   Competing interests

   No competing interests declared

4. Shasha Chong

   Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, United States

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   schong@caltech.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5372-311X

• 6,217

  Page views

• 856

  Downloads

• 16

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.