Biomolecular condensates create phospholipid-enriched microenvironments

Jason G Dumelie; Qiuying Chen; Dawson Miller; Nabeel Attarwala; Steven S Gross; Samie R Jaffrey

doi:10.1038/s41589-023-01474-4

. Author manuscript; available in PMC: 2025 Mar 1.

Published in final edited form as: Nat Chem Biol. 2023 Nov 16;20(3):302–313. doi: 10.1038/s41589-023-01474-4

Biomolecular condensates create phospholipid-enriched microenvironments

Jason G Dumelie ¹, Qiuying Chen ¹, Dawson Miller ^1,², Nabeel Attarwala ^1,², Steven S Gross ¹, Samie R Jaffrey ^1,^*

PMCID: PMC10922641 NIHMSID: NIHMS1947329 PMID: 37973889

SUMMARY

Proteins and RNA are able to phase separate from the aqueous cellular environment to form subcellular compartments called condensates. This process results in a protein-RNA mixture that is chemically different from the surrounding aqueous phase. Here we use mass spectrometry to characterize the metabolomes of condensates. To test this, we prepared mixtures of phase-separated proteins and extracts of cellular metabolites and identified metabolites enriched in the condensate phase. Among the most condensate-enriched metabolites were phospholipids, due primarily to the hydrophobicity of their fatty acyl moieties. We found that phospholipids can alter the number and size of phase-separated condensates, and in some cases alter their morphology. Finally, we found that phospholipids partition into a diverse set of endogenous condensates as well as artificial condensates expressed in cells. Overall, these data show that many condensates are protein-RNA-lipid mixtures with chemical microenvironments that are ideally suited to facilitate phospholipid biology and signaling.

INTRODUCTION

Cells contain various non-membranous organelles that organize proteins into specific intracellular domains, such as nucleoli, nuclear speckles and stress granules. These compartments are biomolecular condensates of proteins, and often RNA, that phase separate from the rest of the cytoplasm^1–8. The condensates assemble through multiple weak interactions between the constituent proteins and between the proteins and RNA^9–11. These interactions promote demixing from the cytoplasm, generating phase-separated liquids, gels or solids.

Many of these condensates contain metabolic enzymes, including some that are not typically thought to have cytosolic substrates. For example, diverse enzymes involved in phospholipid metabolism are found in nucleoli, nuclear speckles and chromatin sites undergoing DNA damage repair^12–16. These enzymes include, among others, phosphatidylinositol phosphate kinases (PIPK) and phosphoinositide-specific phospholipase C (PI-PLC)^12–15. Their activity in nuclear compartments is important for nuclear PI3K/AKT signaling, which regulates cell proliferation and apoptosis¹⁷.

The localization of these enzymes to condensates is unexpected because their phospholipid substrates are located in the plasma membrane. However, some evidence suggests that their substrates might also be present in nuclear condensates. Up to 3% of nuclear phospholipids have been found in non-membranous nuclear compartments¹⁸, which may reflect localization to condensates.

Although condensates are typically thought to comprise proteins and nucleic acids, they might also contain cellular metabolites. Condensates form a separate chemical phase from the cytosol¹⁹. When molecules are present in different phases, they partition between them based on their solubility in each phase²⁰. As a result, protein condensates might have metabolite compositions different from the cytosol.

Indeed, the chemotherapeutic compound cisplatin was recently found to preferentially partition into condensates containing the Mediator protein MED1²¹. This finding suggests that the microenvironments within condensates can have unique chemical properties that promote partitioning of specific small molecules. It remains unknown if cellular metabolites similarly partition into condensates.

Here we test the idea that intracellular metabolites partition into protein condensates resulting in a condensate-specific metabolome. Using untargeted metabolomic analysis, we identify cellular metabolites that are preferentially enriched and de-enriched in condensates prepared in vitro. Most notable among the enriched metabolites were phospholipids, which preferentially partition into condensates formed from different low-complexity domain-containing proteins. Phospholipid partitioning is mediated by their hydrophobic regions and depends on a balance of their hydrophobic and polar properties. We find that phospholipid partitioning occurs in cells based on the presence of phosphoinositides in intracellular condensates. We also find that phosphoinositides enter artificial condensates formed in cells using synthetic proteins, suggesting that phase-separation, rather than specific phospholipid transport pathways, is sufficient to enable phosphoinositide partitioning into condensates. Overall, these data demonstrate a new function for biomolecular condensates in creating unique microenvironments that selectively enrich lipids, thus providing an unanticipated subcellular compartment for lipid signaling and metabolism outside the conventional lipid bilayer.

RESULTS

Metabolomics of phase-separated condensates

Intracellular condensates can be considered a distinct chemical phase within the surrounding cytosol¹⁹. An important chemical phenomenon associated with distinct chemical phases is that small molecules exhibit different levels of partitioning into each phase. This is related to the small molecules’ relative solubility in each phase²⁰. Phase separation creates subcompartments within cells, raising the possibility that metabolites, including phospholipids may differentially partition within these subcompartments. In this way, phase-separated cytosolic condensate could have different metabolomes than the bulk cytosol.

To test this, we examined the partitioning behavior of cellular metabolites between aqueous and condensate phases²². We tested three types of condensates induced in vitro: the full-length SARS CoV-2 nucleocapsid (referred to as “nucleocapsid”), and the low complexity domains of Mediator Complex Subunit 1 (MED1), and heterogeneous nuclear ribonucleoprotein A1 (HNRNPA1) (Fig. 1a). Notably, nucleocapsid forms a cytosolic condensate associated with viral packaging^23–27, while MED1 and HNRNPA1 form nuclear condensates, associated with transcription initiation^1,28 and splicing²⁹, respectively. Nucleocapsid contains an RS (arginine-serine) domain²⁷, MED1 contains a domain enriched in serine and lysine, and HNRNPA1 contains an RGG (arginine-glycine-glycine)-enriched domain³⁰. Each of these chemically distinct low complexity domains form weak, non-specific interactions with RNA^{23–27,31,32}.

Fig. 1 | — a, Proteins used in metabolomics studies. Modular domains (grey) and disordered regions (pink) are displayed over disorder scores (y-axis) for each amino acid (x-axis). Brackets indicate the protein region attached to mCherry. Disorder scores were determined by IUPred3⁶⁷⁠.

b, Condensate metabolomics procedure. Condensate-forming proteins, in the presence of metabolites, were stimulated to form condensates by RNA addition. After a brief incubation, condensate and aqueous phases were separated using centrifugation and analyzed using LC-MS.

c, RNA stimulates nucleocapsid and MED1 phase separation. Nucleocapsid (30 μM, top, red), MED1 (30 μM, middle, red) and HNRNPA1 (30 μM, bottom, red) were incubated (10 min, 25oC) in the absence (left) or presence (right) of RNA (150 nM), then imaged using confocal microscopy. While RNA addition did not impact the number of HNRNPA1 condensates, there were increased numbers of nucleocapsid and MED1 condensates after RNA addition. Scale bar, 5 μm (n=2).

d, Nucleocapsid is enriched in the condensate phase. Coomassie G-250 dye stained gels were used to assess whether centrifugation-separated nucleocapsid concentrates in a distinct phase. Nucleocapsid is almost undetectable in input (left) and aqueous (center) fractions, but is readily detected in the condensate fraction (right). Arrow indicates expected mCherry-nucleocapsid location (n=2).

e, Condensate detection in post-centrifugation fractions. Nucleocapsid (red) in the aqueous (left) and condensate (right) post-centrifuge fractions were imaged by fluorescence microscopy. Fractions were diluted in LC-MS-compatible buffer (1:4) prior to imaging. Condensates are only visible in the condensate fraction. Scale bar, 5 μm (n=2).

To determine if endogenous metabolites preferentially partition into these phase-separated condensates, we employed a mass-spectrometry-based untargeted metabolite profiling approach (Fig. 1b). First, we identified liquid chromatography–mass spectrometry (LC-MS)-compatible buffer conditions that induce phase separation. Crowding agents are commonly used to induce phase separation, but are not compatible with LC-MS³³. Additionally, LC-MS requires volatile buffers, which are typically not used in phase-separation experiments.

To evaluate phase-separation conditions, we used mCherry-tagged recombinant proteins and then stimulated condensate formation using generic phase RNA. We found that ammonium bicarbonate buffer (pH 7.5) containing 50 mM NaCl is compatible with phase separation without the need for added polyethylene glycol, as shown by the appearance of red condensates of mCherry-tagged nucleocapsid, MED1, and HNRNPA1 (Fig. 1c). Due to the lack of crowding agents, phase separation required greater concentrations of protein (30 μM) than typically employed for condensate formation^1,23 and, in the case of MED1 and nucleocapsid, addition of RNA visibly stimulated condensate formation (Fig. 1c).

Notably, the three condensates have distinct biophysical properties. The MED1 condensates appeared spherical and settled on the glass cover (Fig. 1c). These condensates exhibited liquid-like properties with ~50% fluorescence recovery after photobleaching (FRAP) in 25 s (Extended Data Fig. 1a,b). In contrast, HNRNPA1 condensates formed network structures that settled inefficiently on the glass and display ~20% FRAP after 100 s (Fig. 1c; Extended Data Fig. 1a,b) suggesting a more viscous condensate. Finally, nucleocapsid condensates were smaller and did not settle on the glass (Fig. 1c). These condensates moved too quickly to study using FRAP, but rapidly dissolved after the NaCl concentration was increased from 50 mM to 500 mM, consistent with more liquid-like characteristics³⁴ (Supplementary Video 1, Extended Data Fig. 1c). Overall, in addition to having different amino acid compositions, the condensates show different biophysical characteristics which may therefore lead to different condensate metabolomes.

To isolate the phase-separated compartment, centrifugation was used to separate the condensate phase (lower phase) from the surrounding aqueous phase (upper phase) (Supplementary Fig. 1a). As expected, the condensate fraction was enriched in protein and RNA (Fig. 1d, Extended Data Fig. 1d,e, Supplementary Fig. 1b,c) and contained visible condensates based on fluorescence microscopy (Fig. 1e, Supplementary Fig. 1c).

We next asked if specific metabolites partition into condensate phases. Cellular metabolites were extracted from mouse livers (Supplementary Fig. 1d). Soluble metabolites were added to each protein solution, and then RNA was added to stimulate condensate formation. The condensates were separated by centrifugation and metabolites were extracted from equal volumes of the input solution, aqueous fraction, and condensate fraction. The relative concentration of each metabolite was then measured using untargeted metabolite profiling analysis by LC-MS. Notably, the condensate fraction was typically ~2 μl, while the aqueous fraction was ~298 μl. Despite these differences in volume, molecules with equal solubility in both phases would exhibit an identical concentration in each fraction after equal sample volumes are analyzed by LC-MS.

Using this approach, we surveyed >800 metabolites contained in an aqueous normal phase LC-MS in-house database based on mass and retention time. We detected 343 with sufficient ion counts in the input samples to effectively quantify their differential partitioning into condensate phases (Extended Data Fig. 2a, Supplementary Data Set 1). Of these 343 identified metabolites, 247 were designated as high-confidence identifications based on mass accuracy, the lack of confounding structural isomers, and detection in different replicates or chromatographic conditions (see Methods, Supplementary Fig. 2–8). The 96 remaining metabolites were assigned provisional identifications based on lower-confidence mass and retention time matches to an in-house metabolite database (Supplementary Data Set 1). The 247 metabolites with high-confidence identifications included amino acids, nucleotides, lipids and a variety of other polar metabolites (Supplementary Data Set 1). For each protein condensate, we compared the ratio of metabolites in the condensate fraction relative to input samples across three replicates (Supplementary Fig. 9). This metabolite ratio was correlated across replicates (rho 0.47–0.80, Spearman’s rank correlation coefficient). These data support the reproducibility of metabolomic measurements.

We next asked whether any metabolites are enriched in any of the three protein condensates. Initially, we considered a metabolite to be significantly enriched in a fraction if it had a two-fold differential enrichment with an unadjusted p < 0.05. Using this definition, 71 metabolites (21% of those detected) were enriched in the condensate fraction for nucleocapsid, 39 (11%) for MED1 and 68 (20%) for HNRNPA1 (Fig. 2a,b). The reduced number of enriched metabolites in MED1 may reflect higher variation between technical replicates in MED1 experiments (Supplementary Fig. 9b) or differences in protein composition.

The metabolites enriched in the three condensate fractions showed substantial overlap (Fig. 2b,c). Among MED1 condensate-enriched metabolites, 90% were enriched in the nucleocapsid condensate. Among HNRNPA1-enriched metabolites, 68% were enriched in the nucleocapsid condensate. This suggests that despite the different amino acid compositions of their low-complexity domains, all three condensates partition similar metabolites.

Chemical Similarity Enrichment Analysis (ChemRICH)³⁵, identified several metabolite classes enriched in condensates, including phospholipid and phospholipid-like molecules, such as lysophospholipids and unsaturated fatty acids (Fig. 2d). Phospholipids include membrane lipids such as phosphatidylserine, phosphatidylcholine, phosphatidylethanolamine, and phosphatidylinositol. Higher order phosphoinositides (e.g. phosphatidylinositol 4,5-bisphosphate (PIP₂)) were also detected in condensate phase of our metabolomics screen, but could not be detected in input samples, preventing us from performing relative condensate enrichment calculations. The inability to detect these molecules in input samples may be due to their high ionic charge and relatively low abundance³⁶.

We confirmed the enrichment of phospholipids in condensates using reference lipids as mass and retention time standards and isotopically labeled lipids for quantification (see Methods, Supplementary Fig. 5,8 and Supplementary Data Set 1).

We next identified metabolites that were de-enriched from condensates. Measuring de-enrichment is challenging because the condensate phase can be contaminated by small amounts of the aqueous phase. This is because the condensate phase is ~2 μl. This small volume is readily affected by small amounts of residual aqueous phase. For example, a metabolite that is fully depleted from the condensate fraction will only show a two-fold depletion if 2 μl of aqueous phase contaminates the condensate phase.

We therefore relied on a larger number of replicates to assess depletion. In this analysis, we performed six additional metabolomics experiments using the MED1 condensate and four using nucleocapsid and applied multiple-hypothesis adjustments (Extended Data Fig. 2b–c, 3, Supplementary Fig. 10). When all nine MED1 experiments were combined, only ADP-ribose (2.1 fold), nicotinamide adenine dinucleotide (NADH, 2.2 fold) and succinylaminoimidazole carboxamide riboside (2.3 fold) were significantly de-enriched (i.e. >1.95-fold de-enriched, FDR <0.05). In the seven nucleocapsid experiments, fifteen metabolites (4%) were de-enriched from the nucleocapsid condensate fraction, including carnitine (2.1 fold), five relatively polar acylcarnitines (XlogP < 0), S-adenosylmethionine (SAM) (2.0 fold) and histidine (2.2 fold) (Supplementary Table 1). ChemRICH analysis identified various categories of polar molecules among de-enriched metabolites in the two condensates (Supplementary Table 2). Together, these results suggest that subsets of polar metabolites are de-enriched from MED1 and nucleocapsid condensates.

We next asked if RNA affects metabolite partitioning into condensates. This was tested with the nucleocapsid condensate since it phase separates, at least to some extent, both with and without RNA. Similar sets of metabolites partitioned in the no-RNA and RNA-containing condensates (r>0.77, Pearson’s correlation coefficient; Extended Data Fig. 3). This suggests RNA has minimal effects on nucleocapsid condensate metabolite partitioning.

Hydrophobicity and polarity affect phospholipid partitioning

We next wanted to understand which portion of phospholipids and phospholipid-like metabolites contributes to their condensate partitioning.

To determine if hydrophobic regions are needed for partitioning, we compared the enrichment of phospholipids (which contain two fatty acyl moieties), lysophospholipids (which contain one fatty acyl moiety), and glycerophosphoryl species (which contain polar head groups, but no fatty acyl moieties) in the metabolomics data (Fig. 3a, Extended Data Fig. 4a). The glycerophosphoryl species include glycerol with a single phosphate connected to either choline, ethanolamine, inositol, or serine. Glycerophosphoethanolamine was not abundant enough in input samples to be analyzed. None of the other three glycerophosphoryl moieties were significantly enriched in the condensate fraction (Fig. 3a). Thus, glycerophosphoryl moiety portions of phospholipid alone are insufficient to promote condensate enrichment and the hydrophobic regions are required.

Fig. 3 | — a, Phospholipid-like molecules require fatty acyl moieties to partition. The median log₂-fold enrichment in the nucleocapsid (left, purple), MED1 (center, blue) or HNRNPA1 (right, green) condensate fractions is plotted for identified phospholipids (n=51), lysophospholipids (containing one fatty acyl moiety, n=14) and glycerophosphoryl head groups (e.g. glycerophosphocholine, n=3). Each metabolite is represented by a single dot, while the lines within each violin plot represent quartiles. Phospholipids (p=0.004105, nucleocapsid; p=0.02125, MED1; p=0.009171, HNRNPA1) and lysophospholipids (p=0.002941, nucleocapsid; p=0.02125, MED1; p=0.002941, HNRNPA1) partition into condensates, while glycerophosphoryl groups lacking fatty acyl moieties do not. *p<0.05, **p<0.005 two-sided Wilcoxon rank-sum test (n=3 condensate metabolomic experiments).

b, Structure of acylcarnitines, which like lysophospholipids, are amphipathic and contain a fatty acyl moiety.

c, Acylcarnitines with fatty acyl moieties containing sixteen or more carbons partition into condensates. To determine if a minimum fatty acyl moiety length is needed for acylcarnitine partitioning, the size of acylcarnitine fatty acyl moieties (x-axis, carbons) is plotted against their median log₂-fold enrichment in nucleocapsid (purple), MED1 (blue) or HNRNPA1 (green) condensates relative to input sample. Acylcarnitines with fewer than seven carbons are either not enriched in condensates (MED1 or HNRNPA1) or de-enriched (nucleocapsid) in condensates. On the other hand, all acylcarnitines with > 15 carbons are enriched in condensates, with at least 7.5-fold enrichment in all three condensates (n=3).

We next examined the relationship between alkyl moiety length and condensate enrichment. Phospholipids with shorter fatty acyl moieties (32–35 total carbons in fatty acyl moieties) exhibited median 1.8 to 3.1-fold higher enrichment in the three condensates compared to phospholipids containing longer fatty acyl moieties (>36 total carbons)(Extended Data Fig. 4a,b). These results suggest that while glycerophosphoryl moieties cannot partition in the absence of fatty acyl moieties, but very long fatty acyl moieties reduce partitioning. Thus, a balance of hydrophobicity of the alkyl chains with the polar features of the phospholipid is needed to achieve optimal enrichment in condensates.

In addition to phospholipids, some acylcarnitines were enriched in phase-separated condensates (Fig. 2d, Supplementary Table 1). Acylcarnitines resemble lysophospholipids, but unlike the lysophospholipids in our dataset, they have fatty acyl moieties with < 16 carbons (Fig. 3b). We therefore wondered if minimal fatty acyl moiety lengths could be observed for efficient acylcarnitine partitioning. Acylcarnitines’ fatty acyl moiety lengths were compared with their condensate enrichment (Fig. 3c). No acylcarnitines with < 7 carbons in their fatty acyl moiety were enriched in condensates. On the other hand, all acylcarnitines with > 15 carbons had at least 7.5-fold enrichment in all three condensates. These results suggest that a minimum level of hydrophobicity is required for acylcarnitine partitioning.

To further examine how fatty acyl moiety length impacts lipid condensate enrichment, we used a lipid library containing phospholipids with varying fatty acyl moiety lengths (Extended Data Fig. 5a–c). The library was added to MED1 and partitioning was measured using the same procedure described above. Here, phosphatidylcholine with the shortest fatty acyl moieties, comprising a total of 6 carbons, exhibited negligible enrichment, demonstrating that a minimal level of hydrophobicity is needed for enrichment (Extended Data Fig. 5b). In contrast, lysophosphatidylcholine with 16 carbons and phosphatidylcholine with 18 carbons had intermediate enrichment, with a median of 48 and 24-fold enrichment, respectively. Phospholipids with chain lengths greater than 18 carbons showed maximal enrichment, with a median 96-fold enrichment across all phosphatidylcholines. We also observed that experiments that utilized high concentrations of library (10 μM of each lipid) displayed reduced enrichment of phospholipids with fatty acyl moiety total lengths greater than 18 carbons. This may reflect the ability of lipids at high concentrations to form micelles or vesicles³⁷, which may inhibit condensate partitioning.

We next asked if phospholipid charge, which is determined by the head group, influences partitioning. Phosphatidylserine and phosphatidylinositol carry a net negative charge, while phosphatidylethanolamine, phosphatidylcholine and sphingomyelin carry a net neutral charge. Lipids with any of these head groups partitioned into the three condensates, with some differences in the efficiency of partitioning (Extended Data Fig. 5d). For example, the negatively charged phospholipids had ~2-fold and ~5-fold less enrichment in MED1 and HNRNPA1 condensates, respectively, relative to net0neutral charged phospholipids. This difference in head groups is not observed with the nucleocapsid condensate (Extended Data Fig. 5d).

To further examine the importance of head groups, we used the lipid library described above. Here we again observed relatively similar partitioning of phospholipids with each of the different head groups, including PIP₂ (Extended Data Fig. 5c). Overall, these results suggest that having fatty acyl moieties of some minimal length is more critical for phospholipid condensate enrichment than the type of head group.

Phospholipids partition into diverse condensates

We next wanted to determine if condensate phospholipid partitioning is specific to metabolomic buffer conditions, which are different from the buffers commonly used to induce condensate formation. To address this, we monitored Oregon Green-tagged phosphatidylethanolamine partitioning behavior (Fig. 4a). First, nucleocapsid condensates were prepared using the buffer, protein and RNA concentrations in our metabolomics studies. Consistent with those studies, the phospholipid analogue was ~25-fold enriched in condensates relative to the surrounding aqueous environment (Extended Data Fig. 6a,b). Notably, the Oregon Green dye itself did not enrich in nucleocapsid condensates (Extended Data Fig. 6a,b).

Fig. 4 | — a, Oregon Green phosphatidylethanolamine structure. Oregon Green is a conjugated to 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine. The arrow points to the bond between the Oregon Green dye and the phosphatidylethanolamine.

b, Oregon Green phosphatidylethanolamine is enriched in condensates. To determine if phospholipids are enriched in a wide range of condensates, Oregon Green phosphatidylethanolamine or Oregon Green dye were imaged in six different condensates. Oregon Green phosphatidylethanolamine or dye (2 μM, green) was added to solutions of nucleocapsid, MED1, HNRNPA1, 53BP1, DDX4 or YTHDF2 (3 μM, red) in the presence of generic phage RNA (15 nM) or, for YTHDF2, m6A-containing RNA (300 nM) in buffer containing salts thought to reflect intracellular levels (50 mM Tris pH 7.5, 140 mM KCl, 12 mM NaCl, 0.8 mM MgCl₂, 5% PEG-8000)³⁸⁠. Condensates were imaged after a 10 min incubation. A representative image is displayed for each condition. Oregon Green phosphatidylethanolamine, but not dye, colocalizes with each condensate. Scale bar, 5 μm.

c, Quantification of b. The median ratio of mean fluorescent signal inside condensates to the mean signal outside condensates for Oregon Green dye (orange) or phosphatidylethanolamine (blue) across z-stacks 1–3 μm above the slide surface was plotted for Nucleocapsid (p=0.01216), MED1 (p=2.382e-05), HNRNPA1 (p=0.007237), 53BP1 (p=0.01492), DDX4 (p=0.004712) and YTHDF2 (p=0.03556). Error bars indicate s.e.m. *p<0.05, two-sided Welch’s t-test (n=3 for phospholipid samples, n=2 for dye samples).

Next condensates, were prepared using sodium, potassium and magnesium concentrations that reflect intracellular levels³⁸. These experiments used lower nucleocapsid concentrations nucleocapsid (3 μM) and included the crowding agent polyethylene glycol-8000 (PEG-8000) to induce condensate formation^39,40. Under these conditions, the phospholipid was 36-fold enriched in nucleocapsid condensates (Fig. 4b,c). Again, the Oregon Green dye did not partition into the nucleocapsid condensate in any of these conditions (Fig. 4b,c). Taken together, these results suggest phospholipid partitioning is not specific to the condensate metabolomics’ buffer conditions.

We also wondered if the mCherry tag caused the observed phospholipid partitioning. We generated condensates using nucleocapsid lacking mCherry. These condensates co-localized with Oregon Green-tagged phosphatidylethanolamine, but not with the Oregon Green dye (Extended Data Fig. 6c), suggesting that mCherry does not mediate phospholipid enrichment.

To determine the generality of phospholipid partitioning, we examined HNRNPA1 and MED1 along with condensates formed from (1) full-length YTHDF2 (containing a proline/glutamine/asparagine prion-like low-complexity domain), (2) the low-complexity domains of DDX4 (containing phenylalanine-glycine domains and arginine-glycine domains), and (3) the low-complexity domain of 53BP1 (containing an arginine-glycine-glycine domain). YTHDF2 is associated with processing bodies (P-bodies) and stress granules⁴¹, DDX4 is associated with nuage condensates⁴² and 53BP1 is found in DNA repair compartments⁴³. We purified each of these as an mCherry fusion protein and generated condensates. In each case, Oregon Green phosphatidylethanolamine partitioned into the condensate phase, while Oregon Green dye alone did not (Fig. 4b,c). Enrichment across these condensates varied from 36-fold for nucleocapsid to 85-fold for YTHDF2 condensates. These results indicate that phospholipids preferentially partition into a wide range of protein condensates.

We next wanted to determine if the fatty acyl moieties are necessary for condensate partitioning. We therefore tested if enzymatic cleavage of fatty acyl moieties from Oregon Green-tagged phosphatidylethanolamine prevents condensate partitioning (Extended Data Fig. 7a). To test this possibility, the phospholipid was pretreated with phospholipases that remove either the first or second fatty acyl moieties (phospholipase A1 or A2, respectively) or the head group (phospholipase D). Removing a single fatty acyl moiety fully impaired condensate partitioning (Extended Data Fig. 7b,c). This effect was blocked by EDTA, which inhibits phospholipase A2 or D during the enzymatic treatment step. Overall, these data confirm that phospholipid partitioning into condensates depends on the presence of fatty acyl moieties.

Although our metabolomics data suggested that lysophospholipids partition into condensates, we did not see this effect with the lysophospholipid version of Oregon Green-tagged phosphatidylethanolamine created by treatment with PLA1 and PLA2 (Fig. 3a and Extended Data Fig. 7b,c). The lack of partitioning of the Oregon Green lysophospholipid may, to some extent, reflect its relatively short-chain fatty acyl moieties, each of which contain sixteen carbons. This is at the lower end of the 16 to 22-carbon fatty acyl moieties in the lysophospholipids detected in the metabolomic experiments. Although we found that lysophospholipids with a 16-carbon chain exhibited partitioning into condensates (see Extended Data Fig. 4, 5b), the presence of Oregon Green could make shorter fatty acyl moieties insufficient to drive condensate partitioning. Oregon green is bulky and more polar than naturally occurring head groups, which suggests that both the hydrophobicity and the polarity of the lipid contribute to the overall ability of a phospholipid to partition into condensates..

Phospholipids trigger changes in MED1 phase separation

We next asked if phospholipids affect the process of condensate formation. For these experiments, we decided to use 10 μM of phospholipid. The concentration to use for these experiments is not obvious. Since MED1 condensates are in the nucleus, we estimated the concentration of phospholipids in the nucleus that are not associated with nuclear membranes (see Methods). Briefly, to estimate this concentration we calculated the total nuclear phospholipid concentration based on comparing previously reported dry weight measurements of phospholipids in purified nuclei, with other molecules with known reported nuclear concentrations⁴⁴. We then multiplied this value with the fraction of nuclear phospholipids reported to be outside of the nuclear membrane¹⁸. This analysis indicated that non-membranous nuclear phospholipid levels is ~100 μM. Since individual types of phospholipid will have lower concentrations, and to be more conservative, we studied the effect of adding 10 μM of each phospholipid on condensate formation.

We therefore formed MED1 condensates in the presence of RNA, PEG, endogenous salt concentrations and either Oregon Green-tagged phosphatidylethanolamine or Oregon Green dye (Supplementary Fig. 11). MED1 condensates were then imaged over 60 min (Supplementary Video 2). Throughout this time course, an increased number of MED1 condensates were observed in the sample containing Oregon Green phosphatidylethanolamine. The number and size of MED1 condensates was then quantified after a 15 min incubation. The number of condensates increased ~three-fold with Oregon Green-tagged phosphatidylethanolamine relative to Oregon Green dye (Fig. 5a,b). On the other hand, there was no significant difference in condensate median size (Extended Data Fig. 8a).

Fig. 5 | — a, Oregon Green phosphatidylethanolamine increased the number of MED1 condensates. Condensates were formed with MED1 (3 μM), RNA (15 nM) and either Oregon Green dye or Oregon Green-tagged phosphatidylethanolamine (10 μM). Condensate formation was induced by PEG-8000 (5%) for 15 min (25°C). A representative image is shown, with one condensate expanded (right). Scale bar, 5 μm.

b, Quantification of MED1 condensates that settle onto the glass well bottoms in a. Each image was segmented using ImageJ Robust Automatic Threshold Selection (RAST) as described in Methods. Particles with area > 0.1 μm² were considered condensates. The median condensate number/image across five images/replicate (blue) is plotted for Oregon Green dye or Oregon Green phosphatidylethanolamine (x-axis, p=0.00456). Error bars indicate s.e.m. **p<0.005, two-sided Welch’s t-test (n=6).

c, MED1 condensate formation, as in a, using vehicle (0.5% ethanol) or 10 μM dipalmitoyl versions of phosphatidylethanolamine (DPPE), phosphatidylcholine (DPPC) or phosphatidylserine (DPPS). Scale bar, 5 μm.

d, MED1 condensate formation, as in a, using 10 μM dipalmitoyl versions of phosphoinositides containing zero (PI), two (PIP₂) or three (PIP₃) phosphate groups. Scale bar, 5 μm.

e, MED1 condensate quantification for c and d. The PI sample did not contain ethanol. DPPE (p=0.008272) and DPPS (p=0.03568) increase the number of condensates. Error bars indicate s.e.m. *p<0.05, two-sided Welch’s t-test (n=4 for phosphatidylinositol, n=3 for all other samples).

f, MED1 condensate shape quantification for d. Condensate irregularity was assessed by subtracting one by measured condensate circularity (between 0 and 1). Median condensate irregularity (blue) was plotted for each condition. PIP₂ (p=0.004402) and PIP₃ (p=0.009126) promoted the formation of irregularly shaped condensates, which may reflect smaller tethered condensates. Error bars indicate s.e.m. *p<0.05, **p<0.005, two-sided Welch’s t-test (n=4).

We next asked whether other phospholipids influence the formation of MED1 condensates. Here, we tested the effect of phosphatidylcholine, phosphatidylethanolamine, phosphatidylserine, and phosphatidylinositol, each containing palmitoyl fatty acyl moieties. After fifteen minutes, samples were imaged as described above. Phosphatidylserine addition led to a two-fold increase in the number of MED1 condensates, while phosphatidylethanolamine and phosphatidylcholine addition led to smaller increases (Fig. 5c–e). Phosphatidylinositol addition had a negligible effect on condensate numbers. These results indicate that diverse phospholipids can promote MED1 condensate formation.

We next asked if phosphoinositide phosphorylation impacts condensate formation. Each additional phosphionisitide inositol ring phosphate group adds a negative charge to phosphoinositides, which could markedly alter how they interact with neighboring molecules. For these experiments, we compared MED1 condensate formation after adding phosphatidylinositol (PI), PIP₂, or phosphatidylinositol 3,4,5-triphosphate (PIP₃). Unlike what was observed with other phospholipids, adding either PIP₂ or PIP₃ altered the morphology of MED1 condensates (Fig. 5d). Rather than forming separate distinct spheres, condensates appeared to adhere to each other to form larger branching condensates (Fig. 5d and Supplementary Video 3). While the median size and number of all condensates did not change significantly, more large condensates were observed (> 1 μm) (Extended Data Fig. 8b–d). The shape of each condensate was quantified and a significant reduction in the circularity of MED1 condensates was observed with PIP₂ or PIP₃ addition (Fig. 5f).

To understand the unique effect of these phosphoinositides compared to other phospholipids, we considered studies showing that phosphoinositide-containing membranes can cluster in a divalent ion-dependent manner^45,46. We wondered if a similar phenomenon might be occurring with condensates. Indeed, chelating magnesium with EDTA blocked PIP₂ and PIP₃-induced branching (Extended Data Fig. 8e,f, Supplementary Fig. 12, Supplementary Video 4). This suggests that these branched structures may rely on magnesium to coordinate phosphate species between phosphoinositide molecules.

Phosphoinositides localize to intracellular condensates

Our in vitro experiments suggest that phospholipids partition into a broad range of condensates. However, the conditions of the in vitro phase separation experiments are unlikely to fully recapitulate cellular conditions. We therefore wanted to examine whether phospholipids enter condensates in cells. Given their biological importance, antibodies were previously developed that bind PIP₂ as well as PIP₃. These antibodies have previously been validated^47–49 and further validated by us as described below.

We first asked if PIP₂ and PIP₃ colocalize with nuclear condensates. To address this question, we used immunofluorescence staining with the phosphoinositide-specific antibodies along with antibodies that detect specific nuclear condensates. For these experiments, we used anti-MED1 to detect mediator condensates^1,28 and anti-SON to detect nuclear speckles⁵⁰(Fig. 6a–d). To quantify PIP₂ and PIP₃ enrichment in each condensate, we measured their levels within the condensate relative to regions 1.5–2 μm away from the center of the condensate (Fig. 6b,d). Using this approach, a 3- to 4-fold PIP₂ enrichment was observed in nuclear speckles, while 40–50% enrichment was observed in Mediator condensates. Notably, PIP₂ puncta were also detected outside of these condensate loci, which may reflect its presence in other nuclear condensates (Fig. 6a). In contrast, we did not observe PIP₃ enrichment in either nuclear speckles or mediator condensates (Fig. 6c,d). This may reflect low levels of PIP₃ throughout the nucleus under basal study conditions. Together, these results demonstrate that PIP₂ colocalizes within condensates, with specific nuclear speckle enrichment.

Fig. 6 | — a, PIP₂ (green) was co-immunostained with protein markers (red) for mediator condensates (anti-MED1), nuclear speckles (anti-SON), P-bodies (anti-DCP1) and stress granules (anti-G3BP1). Membrane PIP₂ was removed using 0.2% triton X-100 prior to immunostaining. For stress granule imaging, cells were treated with arsenite (0.5 mM) for 1 hr prior to fixation. Exposure times for PIP₂ imaging were held constant, but are displayed with increased brightness for cytoplasmic granules since there is less PIP₂ in the cytoplasm than the nucleus. Representative images are shown. Scale bar, 5 μm.

b, Background-corrected PIP₂ immunofluorescence signal was measured along a 4 μm line centered in the middle of each condensate. Along that line (x-axis), enrichment at each location relative to the signal 1.5–2 μm from the condensate’s center (y-axis) was plotted for mediator (green), nuclear speckles (brown), P-bodies (red) and stress granules (purple). N=3 biological replicates for mediator condensates (mean n=109 condensates per replicate), nuclear speckles (mean n=518 condensates), and stress granules (mean n=193 condensates), n=2 biological replicates for P-bodies (mean n=86 condensates).

c, Immunofluorescence staining of PIP₃ (green) and protein markers (red) for mediator condensates (anti-MED1), nuclear speckles (anti-SON), P-bodies (anti-DCP1) and stress granules (anti-G3BP1). Membrane PIP₃ was removed using 0.2% triton prior to immunostaining. Scale bar, 5 μm.

d, Quantification of PIP₃ signal in c. N=4 biological replicates for mediator condensates (mean n=83 condensates per biological replicate) and stress granules (mean n=233 condensates), n=3 biological replicates for nuclear speckles (mean n=763 condensates) and P-bodies (mean n=126 condensates).

e, A synthetic protein containing GFP and two copies of the *C. elegans* LAF-1 protein’s RGG domain was over-expressed and PIP₂ or PIP₃ was detected by immunostaining. In a subset of LAF-1 containing cells, condensates co-localized with either phosphoinositide. Scale bar, 5 μm.

f, Median enrichment of anti-PIP₂ (left) or anti-PIP₃ (right) signal in artificial condensates in e is plotted for each cell alongside a violin plot indicating the distribution across all imaged cells. Lines indicate quartiles. For PIP₂, n=17 cells from three experiments and for PIP₃, n=20 cells from three experiments.

We next asked whether PIP₂ and PIP₃ colocalize with cytoplasmic condensates. Here, anti-DCP1 antibodies were used to detect P-bodies^51,52 and anti-G3BP1 antibodies to detect arsenite-induced stress granules⁵³(Fig. 6a–d). PIP₂ and PIP₃ enrichment was quantified as described above for nuclear condensates. In the case of these two condensates, a similar 40–50% enrichment was observed for both PIP₂ and PIP₃ in both condensates (Fig. 6a–d). Together, these results suggest that PIP₂ and PIP₃ are enriched in several intracellular condensates, and that the relative phospholipid enrichment varies between physiological condensates.

To test the specificity of antibodies used for intracellular phosphoinositide localization, we used neomycin (2 mg/ml). Neomycin binds to the phosphoinositide phosphate groups, inhibiting the binding of PIP₂ and PIP₃ antibodies^48,54. Neomycin was added during the immunofluorescence blocking and primary antibody steps. In the presence of neomycin, we observed markedly attenuated PIP₂ immunofluorescence staining in endogenous condensates, with at least 95% less background-corrected signal (Extended Data Fig. 9a,b, Supplementary Fig. 13a,b). We also observe a clear intensity reduction for the PIP₃ signal in stress granules, with a less clear difference for the PIP₃ signal intensity in P-bodies. This may reflect the relatively low amount of PIP₃ relative to PIP₂ in cells⁵⁵ or a higher level of background signal from the PIP₃ antibody relative to the PIP₂ antibody. However, despite this background labeling, neomycin selectively reduced the anti-PIP₃ immunofluorescence signal in P-bodies and nuclear speckles (Extended Data Fig. 9c,d, Supplementary Fig. 13c,d). Notably, neomycin did not impact the immunofluorescence staining of antibodies against condensate markers, consistent with its selective binding to phosphoinositides. Overall, these results support the specificity of the observed phosphoinositide enrichment in cellular condensates.

We next wanted to understand if phosphoinositides are directed to condensates via a dedicated phosphoinositide trafficking pathway or through chemical partitioning. To test this, we used artificial condensates, which are unlikely to be targets of a dedicated intracellular PIP₂/PIP₃ trafficking mechanism. We thus expressed GFP that contained two 168-amino acid-long RGG domains from the Caenorhabditis elegans LAF-1 protein. This protein has previously been shown to form condensates in the cytoplasm of U2OS cells⁵⁶. These artificial condensates exhibited enrichment of PIP₂ and PIP₃ (Fig. 6e,f, Extended Data Fig. 10a). Notably, cell-to-cell variation was observed in whether PIP₂ or PIP₃ were enriched in condensates (Fig. 6e,f, Extended Data Fig. 10a,b). About half of cells displayed less than 25% PIP₂ enrichment in condensates and about ~20% had less than 25% PIP₃ enrichment. Thus, PIP₂ and PIP₃ are found in these synthetic condensates, although there is more variability in their enrichment than in endogenous condensates.

DISCUSSION

Here we show that diverse phospholipids can partition into the chemical microenvironment created by phase-separated proteins. We find that phospholipids partition into protein condensates in a manner that depends on an optimal level of hydrophobicity and polarity. We further show that phospholipids can influence the properties of the condensates, altering their size and morphology, raising the possibility that lipid signaling could influence condensate formation and function. We find that phospholipid partitioning into protein condensates likely also occurs in cells and that phospholipids are endogenous components of diverse intracellular condensates, such as nuclear speckles, P-bodies, and stress granules. The phospholipid enrichment in condensates appears to reflect a chemical partitioning process rather than a dedicated lipid trafficking pathway since phosphoinositides can be found in artificially induced cellular condensates created by overexpressing synthetic proteins with low-complexity domains. Overall, these studies reveal that endogenous condensates are lipid-containing structures and that phase separation provides a mechanism for lipid co-localization with their intracellular metabolic enzymes in cells^12–16.

Many intracellular signaling molecules and enzymes are enriched in “microdomains,” which allow signaling to be spatially compartmentalized in cells⁵⁷. For example, cyclic AMP is compartmentalized into specific domains using phosphodiesterases that are positioned to degrade cAMP that diffuses outside those domains⁵⁸. Our data suggest that phase separation provides a distinct mechanism to achieve microdomains in cells. Phase separation creates lipid microdomains, because phospholipids can partition into the chemical environment within these condensates. Additionally, some lipid-modifying enzymes, like the phosphoinositide 3-kinase PIK3C2B⁵⁹, may have evolved to contain low-complexity domains in order to be localized to these microdomains, thus placing enzymes and their substrates in proximity within condensates. Overall, our data reveal that biomolecular condensates are likely to serve as a distinct microdomain that is preferentially suited to lipid biology.

The mechanism for lipid partitioning likely relates to the chemical environment within condensates which appear to be conducive for lipid solvation. Lipids preferentially dissolve in solvents that maximize hydrophobic interactions, while minimizing interactions with ordered water. Condensates contain hydrophobic residues and can have reduced water content, as has been observed with FUS granules which exhibit ~65% water content compared to ~80% water content of the cytosol^60–62. The specific conformation of lipids within the condensates is currently unknown, but they may not be in the form of a lipid bilayer. Instead, the phospholipids may be oriented randomly within the condensate, or alternatively, organized with their fatty acyl moieties facing the inside of the condensate and the polar head groups facing the outside of the condensate like a micelle. Nevertheless, our data suggest that biomolecular condensates should not be considered simply as protein and RNA. They should also be considered as mixtures of protein, RNA, phospholipids, and potentially other metabolites.

Although our data support the idea that lipids can dissolve within condensates, our data does not rule out other mechanisms that may contribute to lipid localization. For example, lipid-binding proteins such as SF-1 may be important for transport of lipids between the plasma membrane and specific intracellular condensates⁶³. Once present in the condensate, the phospholipids may be stable within the condensate microenvironment. The variable presence of lipid-interacting proteins, or variability in protein composition, may help explain why phospholipids, such as PIP₂, are more concentrated in specific condensates, such as nuclear speckles, than other cellular condensates.

The enrichment of phospholipids in condensates may also influence interactions between condensates and phospholipid membranes. Multiple endogenous condensates interact with phospholipid membranes^64–66. It is possible that condensate interactions with membranes facilitate the transfer of phospholipids from membranes into condensates.

METHODS

Protein purification

Sequences encoding full-length SARS-CoV-2 nucleocapsid from (141391, Addgene) and human YTHDF2, as well as the low complexity domains of human MED1 (aa 947–1581), HNRNPA1 (aa 186–320), DDX4 (aa 1–236), and P53BP1 (aa 1208–1977) were amplified via appropriate PCR primers (Supplementary Table 3) and cloned into a modified pET28C vector using golden gate cloning. In this vector, the multiple cloning site was replaced with a sequence encoding a HIS-tag, followed by mCherry, followed by a linker region encoding (GAPGSAGSAAGGSG) previously used in condensate studies²¹, fused to the protein of interest. The nucleocapsid protein lacking an mCherry tag was cloned into an identical vector lacking the mCherry tag. All plasmids were sequenced after cloning to confirm their identity and deposited in Addgene (deposit 82902, plasmids 204402–204408).

Recombinant proteins were expressed in LOBSTR E. coli cells⁶⁸. A fresh bacterial colony grown overnight on an agar plate was used to inoculate 200 ml Luria-Bertani (LB) media containing 50 μg/ml kanamycin. Cells were grown at 37oC to an OD of 0.4–0.6, except MED1 which was grown to 0.8. IPTG (1 mM) was added and bacteria were cultured for an additional 20 hr at 16oC. All following steps were performed at 4oC.

Cells were collected by centrifugation (5000 × g, 10 min) and re-suspended in lysis buffer (50 mM Tris pH 7.5, 500 mM NaCl, 1 × Halt Protease and Phosphotase Inhibitor Cocktail (Thermo Fisher, PI78441), and 1 mM DTT). Lysis buffer for MED1 purification contained 1 M NaCl. Cells were sonicated (15 s on, 45 s off, twelve cycles) and cellular debris was removed by centrifugation (5000 × g, 10 min). The supernatant was added to TALON metal affinity resin (2 ml, TaKaRa, 635652), in a polypropylene column, which had been pre-equilibrated with lysis buffer (10 volumes). The lysate-TALON resin was incubated with rotation (1.5 hr). Unbound protein was allowed to flow through the column by gravity flow and the resin was washed (five times, 10 ml) with lysis buffer containing 20 mM imidazole (Sigma, I2399–100G). The protein was eluted from the resin through incubation with lysis buffer containing 250 mM imidazole (2 ml, 10 min) before being separated from the resin using gravity flow. This incubation and elution step was repeated.

The purified protein was concentrated to a total volume under 1 ml using Amicon Ultra centrifugal filter tubes (Millipore, UFC803024) and dialyzed three times in 250 ml metabolomics buffer (50 mM NH₄HCO₃ pH 7.5, 50 mM NaCl, 1 mM DTT) using passive diffusion through SnakeSkin dialysis tubing (3500 MWCO, Thermo Scientific, 68035). Protein quantity was measured using a Quick Start Bradford assay (Bio-Rad, 500–0205). The final quantity and purity of purified protein was confirmed using protein gels. Briefly, protein (10 μg) was combined with NuPage LDS Sample Buffer (Invitrogen, NP0007) and 5% 2-mercaptoethanol. Samples were incubated (10 min, 70oC) and loaded onto 1.5mm X 10 well NuPage 4–12% Bis-Tris gels (Invitrogen, NP0335BOX) along with Precision Plus Protein All Blue Standards (Bio-Rad, 1610373). Gels were run at 200 V (45 min). Gels were then stained with PageBlue Protein Staining Solution (Thermo Scientific, 24620) following the manufacturer’s instructions and imaged using a ChemiDoc MP Imaging System (Bio-rad) with Image Lab (5.0).

RNA purification

Lambda phage RNA was transcribed using the control plasmid included in the T7-Flash Transcription kit (AmpliScribe, ASF3507) following the manufacturer’s instructions, including DNase I removal of DNA. The RNA was purified using RNA Clean and Concentrator-25 kits (Zymo, R1018).

m6A-containing RNA was transcribed using the protocol described in Ries et al.⁴¹ with a template encoding 10 adenosines: GGACTCGGACTTGGACTCTGGACTTTGGACTTGGACTTGGACTTCGGACTCGGACTTTGGACT. The transcription reaction and RNA purification were conducted as described above, except adenosine was replaced in the transcription reaction with m6A (Selleck, S3190).

RNA quantity was measured using a Nanodrop 2000c (Thermo Scientific). Lambda phage RNA quality was assessed using an 8% agarose gel, prepared using NorthernMax buffer (Invitrogen, AM8671) and 0.5% formaldehyde. Lambda phage RNA was incubated in RNA loading dye (65oC, 5 min, NEB, B0363A) and loaded and run on the gel (1 hr, 110 V). The m6A-containing RNA was incubated in RNA loading dye (5 min, 65oC) and run on a 10% TBE-Urea Gel (1 hr, 150 V, Invitrogen, EC6875BOX). Both gels were washed in TBE and stained with SYBR Gold Nucleic Acid Gel Stain (Invitrogen, S11494) following the manufacturer’s instructions. Gels were imaged on the ChemiDoc MP Imaging System.

Condensate metabolomics

Metabolite extraction from mice livers

Mouse metabolites were collected from livers of eight C57BL/6N female (6–18 months) mice using methanol extraction. The mouse holding room was maintained at 21.5 ± 1°C, relative humidity between 30% and 70%, with a 12:12 hour light:dark cycle. All procedures relating to mouse treatment were approved by the institutional animal care and use committee of Weill Cornell Medicine (Animal protocol number: 0701–575A).

After mouse euthanasia, livers were removed from mice, washed twice in PBS, once in water (both on ice) and then immediately frozen in liquid nitrogen and stored (−80oC). We used 80% methanol/dH₂O to extract metabolites. This method quenches metabolic activity and extracts a broad range of metabolites^69,70. First, 80% methanol (1 mL, −20oC) was added to each liver and incubated (10 min, −20oC). Glass beads were added to livers and then livers were lysed by bead-beating (45 s) using a Tissuelyser cell disrupter (Qiagen). Lysate was incubated (−20oC, 10 min) and centrifuged (16400xg, 5 min) to separate metabolites from macromolecules. Supernatant was collected and 80% methanol (200 μL) was added to the pellet. The incubation, shaking and centrifugation steps were repeated twice. Supernatants were combined and centrifuged (18400xg, 10 min). Supernatants from this final spin were dried using a SpeedVac Concentrator (Savant, SPD131DDA) at 25oC and stored (−80oC). The amount of protein in the pellet was measured using the Quick Start Bradford assay. This protein mass is used as described below to calculate metabolite concentration.

Mouse metabolites were initially re-suspended in condensate buffer (50 mM NH₄HCO₃ pH 7.5, 50 mM NaCl, 1 mM DTT). While the actual metabolite concentration was not measured, the metabolites were resuspended to a protein equivalent concentration of 938 g/L. Metabolites that were not soluble in condensate buffer were removed by centrifugation (2×5 min, 16,000 g each), in which only the supernatant was retained.

Condensate spin-down with metabolites

Purified protein (37.5 μM) was sonicated (10 s) and centrifuged (1 min, 1,000 g) to disrupt any existing condensates and to remove any precipitated proteins. Purified protein (final concentration, 30 μM) was combined with metabolites (final concentration, 150 g/L protein equivalent) and then phage lambda RNA (final concentration, 0.15 μM) in a final volume of 300 μL. Input sample was saved (10 μL) while the sample was incubated (10 min, 25oC). Condensates were separated from the aqueous environment by centrifugation (10 min, 12,500 g, 25oC). The aqueous phase was removed from the condensate phase and then equal volumes (usually ~ 2 μL) of the aqueous fraction, condensate fraction and input sample were processed for metabolomics using identical approaches, as described below. Where shown, representative images of the phases were taken on an iPhone 11. Protein levels in each fraction were evaluated using gels as described above.

Where indicated, RNA was added instead to the nucleocapsid at a concentration of either 0 μM or 0.6 μM. In these experiments, all other conditions, including buffer concentrations, were identical to other condensate metabolomics experiments. Notably, in a different subset of experiments, metabolites were added to MED1 condensates after the 10 min incubation rather than prior to the incubation. Metabolite enrichment in these condensates was highly correlated to the other MED1 condensates (Extended Data Fig. 2c; r=0.92, Pearson’s correlation), suggesting that metabolite addition timing may not be important.

Metabolites were extracted from each fraction and the input for LC-MS as follows. First, samples were diluted in ammonium bicarbonate buffer (50 mM NH₄HCO₃ pH 7.5) and briefly heated (2 min, 65oC) to disrupt condensates before immediately adding 100% methanol (4x volume, −20oC) to precipitate protein and RNA. This heating step does not appear to be necessary for extracting these metabolites and can be excluded (Extended Data Fig. 2b, Supplementary Fig. 10). Protein and RNA were separated from metabolites by vortexing the samples (2 min), followed by incubation (10 min, −20oC) and centrifugation (5 min, 16,400xg). The supernatant was saved and the process was repeated after 80% methanol addition to the pellet (2x, 200 μL). The supernatants were combined and centrifuged (10 min, 18,400xg). The final supernatant was collected and dried using a SpeedVac Concentrator (25oC).

On the day of metabolite analysis, dried-down extracts were reconstituted in 70% acetonitrile (150 μL), at a relative protein concentration of ~2 μg/μL, and the reconstituted extract was injected for LC-MS-based targeted and untargeted metabolite profiling.

LC-MS Reagents

LC-MS grade acetonitrile (ACN), isopropanol (IPA) and methanol (MeOH), 1-butanol (BuOH) were purchased from Fisher Scientific. High purity deionized water (ddH2O) was filtered from Millipore (18 MΩ). OmniTrace glacial acetic acid was obtained from EMD Chemicals. Ammonium acetate, ammonium formate and all other chemicals and standards were obtained from Sigma-Aldrich in the best available grade.

LC-MS metabolomics platform

As a first step to screen the potential enrichment of metabolites into condensates, we used aqueous normal phase chromatography (ANP) which retains polar metabolites and most glycerophospholipids. Sample extracts were analyzed by LC-MS as described previously⁷¹, using a platform comprised of an Agilent Model 1290 Infinity II liquid chromatography system coupled to an Agilent 6550 iFunnel time-of-flight MS analyzer. For all hnRNPA1 and nucleocapsid experiments, as well as MED1 experiments 1–3, ANP chromatography consisted of a Dursan^® coated Diamond Hydride column (Microsolv, 70000–15P-2) using mobile phases of: (A) 50% isopropanol, containing 0.025% acetic acid, and (B) 90% acetonitrile containing 5 mM ammonium acetate. For all other ANP chromatography experiments, we used a Metal-Free Surface, Bio-Inert, Cogent Diamond Hydride column (MicroSolv, 70000–15D-2). To eliminate the interference of metal ions on chromatographic peak integrity and electrospray ionization, EDTA was added to the mobile phase at a final concentration of 5 μM. The following gradient was applied: 0–1.0 min, 99% B; 1.0–15.0 min, to 20% B; 15.0 to 29.0, 0% B; 29.1 to 37min, 99% B.

We also used a lipidomics platform to validate the inferred identifications of hydrophobic metabolites. This lipidomic platform was adapted with minor modifications from a previously reported Agilent application note (A Comprehensive, Curated, High-Throughput Method for the Detailed Analysis of the Plasma Lipidome, https://www.agilent.com/cs/library/applications/an-plasma-lipidomics-6495-lc-ms-ms-5994-3747en-agilent.pdf) and showed similar lipid retention times for fatty acids and other lipids, as reported. This platform was comprised of an Agilent Model 1290 Infinity II liquid chromatography system coupled to an Agilent 6550 iFunnel time-of-flight MS analyzer. An Agilent ZORBAX Eclipse Plus C18, 100 × 2.1 mm, 1.8 μm reversed phase column was used for the separation. Mobile phases consisted of (A) 10 mM ammonium formate with 5 μM Agilent deactivator additive in 5:3:2 water:acetonitrile:2-propanol and (B) 10 mM ammonium formate in 1:9:90 water:acetonitrile:2-propanol. Column temperature was set at 55°C and autosampler temperature was at 20°C. The flow rate was 0.4 mL/min. The following gradient was applied: 0 min, 15% B; 0–2.5 min, to 50% B; 2.5–2.6 min, to 57% B, 2.6–9 min, to 70% B; 9–9.1 min, to 93% B; 9.1–11.1 min, to 96% B; 11.1– 15min, 100% B; 15–20 min, 15% B.

To quantify metabolites concentration in a subset of samples (Supplementary Data Set 1, Tab 13), we applied the same LC separation system coupled to an Agilent 6460A Triple Quadrupole tandem mass spectrometer. Here, the MED1 experiment 4–9 metabolite samples that were prepared as described were used. SPLASH^® LIPIDOMIX^® Mass Spec Standards (Avanti, 330707) were spiked into the samples immediately prior to performing LC-MS/MS. The parameters for the quantification are reported in Supplementary Data Set 1, Tab 12 and representative dynamic multiple reaction monitoring chromatograms are displayed in Supplementary Fig. 8.

Additionally, the same LC separation system coupled to a Bruker Impact II QTOF equipped with a vacuum insulated probe heated electrospray ionization source (VIP-HESI) (Bruker Daltonics, Billerica, USA) was used to identify representative lipid structures using auto-MS/MS with and without scheduled precursor list fragmentation. Fragments were compared with those deposited in LIPID MAPS, HMDB and MassBank^72–74.

Raw data were analyzed using Mass Hunter Qualitative analysis (10.0), MassHunter Profinder 8.0 and MassProfiler Professional (MPP) 15.1 software (Agilent technologies). Metabolite abundance was extracted from the peak area of the ion chromatogram for each compound and identified by comparing the retention time and molecular mass against an in-house Personal Compound Database and Library (PCDL) of ~850 metabolites measured using ANP chromatography^75,76. This PCDL was developed, as recommended by the LC-MS manufacturer, (Agilent) using commercially available Mass Spectrometry Metabolite Library of Standards molecules provided by IROA Technologies, Enzo and Avanti. A mass window of 10 ppm was used for extraction. Where possible, ANP data was reported from the negative mode, since it is typically more sensitive. In some cases, lipid identification was further confirmed using the lipidomics platform described above by comparison with the retention times and masses in a lipidomics PCDL of 665 metabolites developed as recommended by the manufacturer (Agilent).

All of the metabolites were provisionally identified based on a mass and retention time that matched our ANP PCDL. However, to provide additional confidence in the identifications, we ensured that they passed multiple filters prior to providing an identification that was used in downstream analysis. If the provisional identification based on a mass and retention time matching our in-house ANP PCDL met any of the following criteria, they were not assigned high-confidence identifications:

Mean mass accuracy was >10 ppm in both the original ANP data set and in a repeat ANP data set.
Lipids that were not detected in a repeat reversed phase LC-MS analysis.
Lipids with a mean mass accuracy > 10 ppm in the reversed phase LC-MS analysis.
The metabolite could not be resolved from a confounding structural isomer based on ANP retention time.
The metabolite ID assignment is recognized as non-mammalian in origin (these may reflect metabolites from rodent diet).

In cases where metabolites did not pass these filters, we provide the molecular identity as mass@retention time alongside the provisional identification in Supplementary Data Set 1. These provisional identifications were not considered in the analysis performed in Fig. 2, 3, Extended Data Fig. 2–5.

Further validation of metabolite identifications

To further validate the 19 high-confidence, non-lipid metabolite identification assignments that had >3 ppm mass accuracy in all previous experiments, higher-resolution LC-MS and MS/MS fragmentation was performed on mouse liver metabolite extracts. First, ANP LC-MS was perfomed using Metal-Free Surface, Bio-Inert, Cogent Diamond Hydride columns (MicroSolv, 70000–15D-2). QTOF data acquisition parameters and the amount of sample loaded were tested to detect metabolites with <3 ppm accuracy. Using this approach, eight non-lipid metabolite identifications were validated. The retention times in this experiment matched our PCDL and correlated with previous LC-MS experiments (Supplementary Fig. 4a–c, r > 0.95). Additionally, no other peaks were observed that might confound the identifications. The remaining 11 non-lipid metabolites with >3 ppm mass accuracy were mostly highly abundant, low molecular weight metabolites such as lactate, malate and taurine. Auto MS/MS with a preferred ion list and then targeted MS/MS were used to confirm the identity of these 11 metabolites (Supplementary Fig. 6). Noteably, the major fragment ions shown in Supplementary Fig. 6 match those reported in HMDB, Metlin DB and MassBank^72,73,77.

To further validate the five lipids identities, we repeated our reversed-phase LC-MS experiments described above. Here, different QTOF data acquisition parameters were tested and the amount of sample loaded adjusted to detect metabolites with <3 ppm mass accuracy. Using this approach, the identification of all 5 lipid metabolites was validated with <3 ppm mass accuracy. The retention times in this experiment matched our PCDL and correlated with previous LC-MS experiments (Supplementary Fig. 4d, r > 0.95). Additionally, no alternative peaks were observed that might confound identifications.

Condensate metabolomics using a defined library of compounds

To analyze a defined library of lipids at known concentrations, the following molecules were purchased: phosphatidylcholine (3:0/3:0) (Cayman Chemical, 32703), phosphatidylcholine (9:0/9:0) (Cayman Chemical, 10009874), phosphatidylcholine (12:0/12:0) (Echelon Biosciences, L-1112), phosphatidylcholine (16:0/16:0) (Echelon Biosciences, L-1116), phosphatidylcholine (18:0/18:0) (Echelon Biosciences, L-1118), sn-glycero-3-phosphocholine, lysophosphatidylcholine (16:0) (Echelon Biosciences, L-1516), palmitic acid (Sigma Aldrich, P5585), phosphatidylethanolamine (16:0/16:0) (Avanti Polar Lipids, 850705), phosphatidylglycerol (16:0/16:0) (Avanti Polar Lipids, 840455), phosphatidylinositol (16:0/16:0) (Echelon Biosciences, P-0016), PIP₂ (16:0/16:0) (Echelon Biosciences, P-4516), phosphatidylserine (16:0/16:0) (Echelon Biosciences, L-3116).

Each molecule was dissolved in an appropriate organic solvent. Then either 0.33 pmoles, 3.3 pmoles, or 33 pmoles of each molecule was combined in a 1.5 mL Eppendorf tube. The organic solvents were removed using a SpeedVac Concentrator (Savant, SPD131DDA, 25oC) and the libraries were stored (−80oC). Each tube containing a chemical library was used to perform a single condensate metabolomics experiment using the method to perform condensate metabolomics described above.

Measuring condensate fraction RNA and protein levels

Protein and RNA concentrations were measured in input, condensate and aqueous samples. First, equal amounts of each fraction (typically ~2 μL) were diluted in 1 M NaCl, 50 mM NH₄HCO₃ pH 7.5. These diluted samples (40 μL) were loaded into separate wells of a 384-well plate (Mattek Corporation # PBK384G-1.5-C). Simultaneously, a dilution series of protein with known concentration was loaded onto the 384-well plate. Fluorescence (560 nm excitation mission wavelength, 610 nm emission wavelength) was measured on a fluorometer (Molecular Devices, SpectraMax iD3). Using the dilution series to generate a standard curve, the concentration of protein in the input, aqueous and condensate fractions was calculated.

Fractions and input were incubated (10 min, 55oC) with Proteinase K (0.5 mg/mL, Invitrogen, 25530049) to digest protein. RNA in each fraction was purified using an RNA Clean & Concentrator-5 kit (Zymo, R1013) and quantified using Qubit Broad Range assay kit (Invitrogen, Q10211) following the manufacturer’s instructions on a fluorometer (Qubit, Flex Fluorometer) with software version 1.2.0.

Metabolomics fold-change exclusions and analysis

Metabolites that were not reliably quantified in input samples were removed from downstream analysis. This included metabolites with < 1000 median ion counts/sample or > 2.5 standard deviation in log2(ion counts). Metabolites in an individual LC/MS run could not be detected below 675 ion counts. To limit the impact of low ion count measurements on fold enrichment analysis, metabolites with ion counts lower than 675 had their ion count imputed to be 135 (1/5th the threshold for detection).

Metabolites with only provisional identifications were considered “other metabolites” in Fig. 2a,c, Extended Data Fig. 2b,c, 3a and Supplementary Fig. 10a regardless of provisional identification. Due to the lack of high-confidence identification, these metabolites were also excluded from the ChemRICH analysis displayed in Fig. 2d and Supplementary Table 2, as well as Fig. 3, Extended Data Fig. 3b,c, 4, 5d and Supplementary Fig. 10b–f since these analyses depended on correct metabolite identification.

The metabolite “167.0274@4.08, provisional ID=Quinolinic Acid” was removed from Fig. 2a,c. This metabolite had an extreme median log2 de-enrichment of −14.5 from the HNRNPA1 condensate that was not statistically significant (p=0.25).

Metabolomics data analysis and statistics

All graphs were made with ggplot2 (3.3.5)⁷⁸ in R (4.0.4) within RStudio (2021.09.0 Build 351), except for heat maps which used the heatmap.2 algorithm from the gplots library (3.1.3) in R and the Venn diagram which was made by eulerr (7.0.0) in R. All statistical tests were made in R, except tests for individual metabolite enrichment, which were performed using LibreOffice Calc (7.2.7.2), and chemical similarity enrichment analysis. Chemical similarity enrichment analysis, including calculation of false discovery rates, was performed using the web-based ChemRICH platform³⁵ (accessed February 28, 2023) which used unadjusted p-values calculated in LibreOffice.

Except in the case for ChemRICH analysis and for individual metabolites studied in our metabolomics enrichment analysis, p values were not adjusted for multiple hypotheses testing. The use of two-tailed paired t-test in the condensate metabolome data analysis assumed the log-transformed condensate metabolome data followed a normal distribution. This assumption seems reasonable given the similar level of variance in measurements observed at different enrichment levels between samples and experiments (Fig. 2c, Extended Data Fig. 2b,c, Extended Data Fig. 3a, Supplementary Fig. 9). The Welsh’s t-test used in the analysis of fluorescence microscopy data in Fig. 4 and Fig. 5 also assumed the data followed normal distributions. The basis for this assumption was similar, though there was less data to support these assumptions.

Nucleocapsid ChemRICH analysis of de-enriched metabolites, as well as de-enrichment/enrichment analysis of individual metabolites was performed by combining all seven samples containing nucleocapsid, regardless of RNA concentration. The same analysis performed on MED1 also involved using and all nine MED1 samples, regardless of extraction method or the timing of metabolite addition.

Corrections for multiple hypothesis testing on individual metabolites studied in our metabolomics enrichment analysis used the p.adjust() function in R, with method=“BH”.

Intra-nuclear phospholipid concentration calculations

The total phospholipid concentration in rat liver nuclei has been reported as 3.45% by weight, while protein was reported as 74.6%⁴⁴. Assuming the protein concentration in rat liver nuclei is similar to the reported protein concentration in the Xenopus oocyte nucleoplasm of 106 mg/ml⁷⁹, this suggests a rough total nuclear phospholipid concentration of 4.6 mg/ml. Assuming an average molecular weight of 775 Da, the total nuclear phospholipid concentration would be 6 mM. Using approaches to separate the nuclear membrane from the nucleus and by radioiodinating the extracted phospholipids with the lactoperoxidase method, Albi et al. found that 97% of nuclear phospholipids are in the nuclear membrane¹⁸. This suggests that the concentration of non-membrane associated phospholipids in the nucleus is as high as ~180 μM. Since it is difficult to know the exact concentration that would be considered physiologically relevant, we chose to use a more conservative 10 μM for in vitro experiments.

Imaging condensates

Phospholipid reagents

The following were used in imaging experiments: Oregon Green 488 (Life Technologies, D6145), Oregon Green phosphotidylethanolamine (Life Technologies, O12650), Dipalmitoyl phosphatidylinositol (Echelon Bioscience, P-0016), Dipalmitoyl phosphatidylcholine (Echelon Bioscience, L-1116), Dipalmitoyl phosphatidylethanolamine (Avanti Polar Lipids, 850705), Dipalmitoyl phosphatidylserine (Echelon Bioscience, L-3116), PIP₂ (Echelon Bioscience, P-4516) and PIP₃ (Echelon Bioscience, P-3916).

Passivization of glass wells

Imaging was performed using glass plates passivated with PEG-silane⁸⁰. Wells were washed (2x, water) and then incubated (45 min) with 2% Hellmanex III (Millipore Sigma, Z805939). Wells were then washed (3x, water) and incubated with 0.5 M NaOH (30 min). Next, wells were washed (3x, water) and then incubated with 20 g/l PEG-Silane (Laysan Bio, Inc., MPEG-SIL-5000–1g) in 95% ethanol (overnight). Wells were washed in 95% ethanol (3x) and then water (3x), before being incubated with 50 g/l BSA (20 min, VWR, 0332–100G). Finally, wells were washed in water (3x) and the buffer used to form condensates (as noted below, 2x).

Condensate imaging for metabolomics protocol optimization

Condensates were formed in condensate metabolomics buffer (50 mM NH₄HCO₃ pH 7.5, 50 mM NaCl, 1 mM DTT). First, 30 μM of the studied protein were combined with mouse liver extract metabolites (final concentration, 150 g/l protein equivalent) and then phage lambda RNA (final concentration, 0.15 μM) was added (except where noted) to stimulate condensate formation. This sample (10 μl) was loaded onto a 35 mm glass-bottom dish (Mattek, P35G-1.5–14-C) and incubated (10 min, 25°C). Imaging was performed using the Inverted LSM 880 Airyscan NLO laser scanning confocal and multiphoton microscope using AIM application (14.0.20.201) (Zeiss) at the Bio-Imaging Resource Center at Rockefeller University. At least one Z-stacks was imaged near the base of the sample with exposure times that were below pixel saturation. Z-stacks were imaged using a 63X lens with Glycerine Immersion Oil (Zeiss, 462959-9901-000).

To examine the effect of ionic strength on nucleocapsid condensates, imaging was performed as described above. NaCl (1.1 μL, 5M) was added to the sample and the sample was re-imaged after 1 min. Images in an imaging session used an identical exposure time that ensured pixels were not saturated.

Fluorescence recovery after photobleaching

For FRAP analysis, condensates were formed in condensate metabolomics buffer (50 mM NH₄HCO₃ pH 7.5, 50 mM NaCl, 1 mM DTT). First, 30 μM of the studied protein were combined with mouse liver extract metabolites (final concentration, 150 g/l protein equivalent) and then phage lambda RNA (final concentration, 0.15 μM) was added to stimulate condensate formation. The sample (10 μl) was loaded onto 35-mm well dishes. After a brief incubation (10 min, 25oC), 1–3 newly settled condensates were imaged using the Inverted LSM 880 Airyscan NLO laser scanning confocal and multiphoton microscope. Two circular regions within the condensate were selected and one region was bleached while the other was not. Images were taken every four seconds for two minutes, including three images prior to bleaching.

Imaging phospholipids in condensates

Protein tagged with mCherry (3 μM) was combined with Oregon-Green dye or 1,2-dihexadecanoyl-sn-glycero-3-phosphatidylethanolamine (2 μM) and either m6A RNA (300 nM, YTHDF2 experiment) or lambda phage RNA (15 nM, all other experiments) in condensate buffer (50 mM Tris pH 7.5, 140 mM KCl, 12 mM NaCl, 0.8 mM MgCl₂, 5% PEG-8000), except where indicated. Samples (10 μl) were loaded onto 8-well glass chamber slides (Ibidi, 80841) and, after 10 min, imaged.

Imaging related to enrichment coefficient measurements were performed using the Inverted LSM 880 Airyscan NLO laser scanning confocal and multiphoton microscope. Z-stacks were imaged using a 63X lens with Immersion Oil (Zeiss, 444962-0000-000). Two Z-stacks were imaged near the base of the sample with exposure times that were below pixel saturation. Notably, the same concentration of Oregon Green dye had approximately 10-fold more intense fluorescence than Oregon Green phosphatidylethanolamine, so lower exposure times were used when imaging Oregon Green dye. Notably, partition coefficients were calculated based on differences in signal intensity within individual images, not across images. The location of Z stacks was determined exclusively by examining the sample using the mCherry-associated channel, rather than Oregon Green-associated channel.

Nucleocapsid condensates without an mCherry tag

mCherry-tagged protein (3 μM) was combined with either Oregon Green dye or Oregon Green phosphatidylethanolamine (2 μM, unless noted) and lambda phage RNA (15 nM) in condensate buffer (50 mM Tris pH 7.5, 140 mM KCl, 12 mM NaCl, 0.8 mM MgCl₂, 5% PEG-8000) and then loaded (10 μl) onto a 384-well black glass plate (Mattek Corporation # PBK384G-1.5-C). Samples were briefly incubated (10 min, 25oC) and then imaged using a wide-field fluorescent microscope (Eclipse TE2000-E microscope, Nikon) with a 60X lens using Immersion Oil (Nikon, Type F), using NIS-Elements AR 3.22.15 (Nikon) software. Two separate Z-stacks were imaged near the base of the sample for each sample using phase contrast and then fluorescent imaging of the Oregon Green dye. All images used exposure times that ensured pixels were not saturated.

Phospholipase treatment

The phospholipases used during phospholipase treatment experiments included: phospholipase A2 from honey bee venom (Sigma Aldrich, P9279, 250 μg/ml), phospholipase A1 from Aspergillus oryzae (Sigma Aldrich, L3295, 30 mg/ml), phospholipase D from Streptomyces chromofuscus (Sigma Aldrich, P0065, 1250 units/ml).

Oregon Green phosphatidylethanolamine (50 μM) was treated with either (1) no enzyme, (2) phospholipase A1, (3) phospholipase A2, or (4) phospholipase D (90 min, 25oC then 20 min, 37oC) in a Tris-based buffer (50 mM NaCl, 50 mM Tris pH 7.5). Control reactions for (3) and (4) were performed under identical conditions, except EDTA (5 mM) was added to the Ca2+-dependent phospholipase A2 or phospholipase D.

Imaging of these treated phospholipids in condensates was repeated as described above, with the following changes: (1) rather than adding Oregon Green dye or Oregon Green phosphatidylethanolamine (2 μM), we added one of the above reactions diluted 25-fold, and (2) rather than add MgCl2, EDTA (5 mM) was added to prevent unwanted enzymatic activity. Finally, we noticed that, similar to Oregon Green dye, Oregon Green phosphatidylethanolamine treated with any of the three phospholipases had approximately 10-fold brighter fluorescence intensity than in the absence of phospholipase activity. As a result, lower exposure times were used for these samples to prevent pixel saturation.

Condensate changes in response to phospholipid addition

Protein tagged with mCherry (3 μM) was combined with indicated phospholipids (10 μM) and lambda phage RNA (15 nM) in condensate buffer, unless indicated in the main text, and then loaded (10 μl) onto a 384-well black glass plate (Mattek Corporation # PBK384G-1.5-C). After 15 min, samples were imaged using the wide-field fluorescent microscope with a 60X lens with immersion oil (Type F). Five separate Z-stacks were imaged near the base of the sample for each sample. All images in a given imaging session used identical exposure times that ensured pixels were not saturated.

Time-lapse movies were made following the same protocol, with the following two exceptions. First, images were taken every 3 min for at least 1 hr after addition of PEG. Second, lamp power was reduced and exposure times increased to reduce bleaching. Where indicated, EDTA (1/10 volume, 5 mM final concentration) was added after allowing PIP₃-MED1-lambda phage RNA condensates to form.

Immunofluorescence

The following primary antibodies were used for immunostaining: rabbit anti-G3BP1 (Life Technologies, 13057–2-AP, 5 μg/ml), rabbit anti-DCP1A (Abcam, ab47811, 10 μg/ml), rabbit anti-MED1 (Abcam, ab64965, 2 μg/ml), rabbit anti-SON (Sigma Aldrich, HPA023535, 1 μg/ml), mouse anti-PIP₂ (Echelon Bioscience, Z-P045, 10 μg/ml), mouse anti-PIP₃ (Echelon Bioscience, Z-P345, 20 μg/ml). The following secondary antibodies were used to detect the primary antibodies: Donkey anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 488 (Life Technologies, A21206, 2 μg/ml), Donkey anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 647 (Life Technologies, A31571, 2 μg/ml).

U2OS (ATCC HTB-96) cells were obtained from ATCC. Cells were maintained in DMEM (11995–065, Thermo Fisher Scientific) with 10% FBS and 100 μg/ml of streptomycin using standard tissue-culture conditions (37oC, 5% CO2). Cellular mycoplasma contamination was routinely tested by Hoechst and DAPI staining. Except where noted, cells were plated to reach 30–50% of confluency the following day on a German Round coverslip, 12 mm (Electron Microscopy Sciences, 7219–12) coated with poly-D-lysine (PG35GC-1.5–14-C). In the case of stress granule imaging, cells were incubated with sodium arsenite (1 hr, 0.5 mM) prior to fixation.

For all endogenous condensates, cells were fixed (4% formaldehyde, 15 min), then washed in PBS (3x). Next, cells were permeabilized and membranes were removed using 0.2% triton in PBS (10 min). Cover slips were washed (3x, 800 μl PBS) and then incubated in blocking buffer (400 μl, 2% FBS in PBS, 45 min). They were then incubated with primary antibodies in blocking buffer for 1 hr to detect PIP₂ (25oC) or overnight to detect PIP₃ (4oC). Cover slips were then washed in PBS (3x), before being incubated with secondary antibodies in blocking buffer (1 hr, 25oC). Although images display phospholipids in green and protein in red for consistency, phospholipids were detected using an antibody labeled with Alexa 647 dye, while proteins were detected with antibody labeled with Alexa 488 dye. After washes in PBS (5x), cells were mounted onto slides using ProLong Diamond Antifade Mountant (ThermoFisher, P36961).

Slides were imaged the following day using the wide-field fluorescent microscope with a 60X lens and immersion oil described above. At least ten separate Z-stacks were imaged for each sample. The decision to image specific cells was done exclusively based on signal from antibodies detecting proteins. All images used identical exposure times for the antibody detecting phospholipids and times were chosen to ensure pixels were not saturated.

In the case of artificial condensates, cells were plated to reach 40–60% of confluency the following day on German Round coverslips as described above. Next, 500 μg of a plasmid encoding two LAF-1 RGG domains surrounding GFP (Addgene, #124934)⁵⁶ was transfected into cells using 1.5 μl of FuGENE HD (Promega, E2311) following the manufacturer’s instructions. Cells were fixed 24–48 hr after transfection and immunofluorescence staining was performed as described above. However, here GFP fluorescence, rather than antibody, was used to detect condensates and the number of images per experiment varied.

Image analysis and processing

Images were processed with the FIJI (2.3.0) distribution of ImageJ2 (1.53q)^81,82. Deconvolution was performed on all images displayed in figures, except Extended Data Fig 1a,c with the ImageJ DeconvolutionLab2 package (2.1.2)⁸³ using ten iterations of its Richardson-Lucy algorithm function⁸⁴. The point spread function was calculated using the PSF Generator (1.1.1.2) based on its Born and Wolf 3D optical model function. The parameters to generate each point spread function matched the experimental condition and accuracy computation was set to “Best.”

Phospholipid or dye enrichment in condensates

To calculate enrichment of dyes or phospholipids in condensates formed in buffer, Robust Automatic Threshold Selection (RAST) was used to segment each image based only on mCherry signal. Next, the average Oregon Green signal was measured either within or outside of mCherry-enriched segments in each image. For each Z-stack, the median ratio of these two signals was calculated for each of the images taken 1 to 3 μm above the glass surface. The enrichment coefficient for each sample replicate was calculated as the median across the two Z-stacks taken for each sample.

FRAP analysis

For FRAP analysis, median fluorescence was monitored for each image in the region of the condensate that was photobleached as well as a nearby sub-region in the same condensate. Fluorescence recovery in a given image was calculated as the percent of median fluorescence from the three time points recovered relative to the fluorescence observed immediately after bleaching. These numbers were normalized relative to the sub-region that was not photobleached.

Changes in condensate after phospholipid addition

To analyze changes in condensates in response to phospholipid addition, the image immediately above the glass was first manually selected from each z-stack. Measurements were only performed on condensates that settled on the well’s glass surface to ensure consistency across samples. The noise in the image’s mCherry signal was calculated as the median standard deviation in signal from four regions lacking visible condensates. Robust Automatic Threshold Selection (RAST) was then performed to segment each image based on mCherry signal with noise set to the image-specific calculated noise, lambda was set to “10” and min was set to “10”. Unless noted, all particles with area > 0.1 μm² were analyzed as condensates. Finally, the size, number and circularity of segmented particles was calculated using the ImageJ2 Analyze Particles function.

Condensate enrichment detected by immunofluorescence

First, an in-focus frame was manually chosen based on the signal from the antibody detecting the condensate marker. A lower signal threshold was then manually selected that would appropriately segment condensates from non-condensates. Based on this threshold, images were segmented and particles fitting the following size and shape criteria were selected as condensates: mediator condensates (0.231 μm2 < area < 1.7325 μm2, circularity > 0.6), nuclear speckles (0.231 μm2 < area < 2.31 μm2, circularity > 0.8), P-bodies (0.231 μm2 < area < 2.31 μm2, circularity > 0.9), stress granules (0.5775 μm2 < area < 4.62 μm2, circularity > 0.8), and LAF-1 RGG synthetic condensates (0.693 μm2 < area < 9.24 μm2, circularity > 0.5).

To measure the background-corrected enrichment of signal near the core of condensates, we first calculated the background signal in the phosphoinositide channel. For the imaging of endogenous proteins, where cells were less confluent, background signal was simply calculated as the median signal in the phosphoinositide channel across all images for a given experiment. For the imaging of synthetic condensates, the background signal was calculated as the median signal per image in the phosphoinositide channel in five manually selected regions outside of cells. Next, the median signal intensity from antibodies recognizing either PIP₂ or PIP₃ in each pixel surrounding the center of every identified condensate was calculated. The background signal from an experiment was subtracted from the signal at each pixel in each image for that experiment. Enrichment at a pixel or region (as indicated) was then calculated for pixels and regions on a horizontal axis intersecting the center of a given condensate. Enrichment was calculated by dividing the intensity at a given pixel or region by the median intensity of pixels located 1.5–2 μm from the condensate’s center.

For LAF-1 artificial condensates, enrichment was measured along a single axis for individual condensates. This enrichment was calculated as the mean anti-phosphoinositide immunostain signal within 0.55 μm of the condensate’s center divided by the mean signal between 0.7 μm and 1.25 μm from the condensate’s center. These distances were chosen by examining the drop-off in the signal from the condensate’s marker protein at different distances from the condensate’s center.

Metaplots of immunofluorescence signal display the median raw signal at locations relative to the condensate centers’ as identified above. Where metaplots are comparisons with a control (e.g. neomycin), the displayed metaplots are from a single experiment performed side-by-side with identical imaging and display settings.

Production of drawings and illustrations.

Illustrations of the condensate metabolomics experiment and in the abstract were generated using BioRender (licenses TM23TI078A, EQ24NBVJZU, respectively). All representations of chemical structures were generated using PubChem. Amino acid disorder scores were calculated using IUPred3⁶⁷.

Reporting Summary.

Further information on research design is available in the Nature Research Reporting Summary linked to this Article.

Extended Data

Extended Data Fig. 3 | — a, To assess whether changes in generic phage RNA concentration alter condensate metabolomes, we compared nucleocapsid condensate metabolomes after varying the added RNA concentration. Median log₂-fold enrichment of phospholipids (blue, n=51), lysophospholipids (green, n=14), fatty acids (black, n=21) and all other metabolites (orange, n=257) in the nucleocapsid condensate fraction relative to input sample with 150 nM RNA (x-axis) was plotted against median log₂-fold enrichment with no RNA (y-axis; left) or 600 nM RNA (y-axis; right). Enrichment with 150 nM RNA correlates with enrichment in no RNA (r=0.78, Pearson’s correlation coefficient) and 600 nM RNA (r=0.85, Pearson’s correlation coefficient) samples. No metabolites were significantly different between 0 and 150 nM or 150 and 600 nM RNA samples (FDR>0.1 for all metabolites, two-sided paired t-test with Benjamini-Hochberg adjustment). N=3 for 150 nM RNA sample, n=2 for other samples.

b, To determine whether RNA concentration alters fatty acyl-dependent phospholipid partitioning, we compared the median log₂-fold condensate enrichment of phospholipids (n=51), lysophospholipids (containing one fatty acyl moiety, n=14) and glycerophosphoryl head groups (e.g. glycerophosphocholine, n=3) in the condensate fraction of nucleocapsid with different RNA concentrations. Phospholipids and lysophospholipids are enriched relative to head groups in condensates formed with 0 nM (green; p=0.004105, phospholipids; p=0.002941, lysophospholipids), 150 nM (purple; p=0.004105, phospholipids; p=0.002941, lysophospholipids, also shown in Figure 3a) or 600 nM (black; p=0.004105, phospholipids; p=0.002941, lysophospholipids) RNA. Individual metabolites are represented by dots, lines in violin plots represent quartiles. *p < 0.05 **p < 0.005, two-sided Wilcoxon rank-sum test (n=3 for 150 nM RNA sample, n=2 for other samples).

c, A heat map was used to further analyze whether RNA concentration impacts phospholipid enrichment in nucleocapsid condensates. Each glycerophosphoryl head group, lysophospholipid and phospholipid’s median log₂-fold enrichment (blue indicates condensate de-enrichment and red indicates enrichment) is plotted for nucleocapsid condensates formed with 0 (left), 150 nM (middle) or 600 nM (right) RNA. Glycerophosphoryl head groups are sorted alphabetically, while lysophospholipids and phospholipids are sorted by the sum of carbons in their fatty acyl moieties after being grouped by head-group as demarked (y-axis).

Extended Data Fig. 4 | — a, Individual phospholipid enrichment in condensates. To examine whether phospholipid head groups and fatty acyl moieties impact enrichment in condensates, we used a heat map. Each glycerophosphoryl head group, lysophospholipid and phospholipid’s median log₂-fold enrichment (blue indicates condensate de-enrichment and red indicates enrichment) is plotted for nucleocapsid (left), MED1 (middle) and HNRNPA1 (right) condensates. Glycerophosphoryl head groups are sorted alphabetically, while lysophospholipids and phospholipids are sorted by the sum of carbons in their fatty acyl moieties after being grouped by head-group as demarked on the y-axis. We observe that longer phosphatidylethanolamines are less enriched in condensates for all three proteins, with similar phenomena observed with other head groups for specific proteins. This suggests that fatty acyl moiety length may reduce enrichment at least with some head groups and in some condensates (n=3 for 150 nM RNA sample, n=2 for other samples).

b, Phospholipid fatty acyl moiety length is inversely correlated with partitioning. To determine if fatty acyl moiety chain length also contributes to partitioning, the total number of carbons in the fatty acyl moieties of lysophospholipids and phospholipids (x-axis, n=65) was plotted against the median log₂-enrichment of that lipid (y-axis) in nucleocapsid (left), MED1 (center) or HNRNPA1 (right) condensates relative to input sample using our metabolomics data. The number of unsaturated bonds is indicated (right) for each lipid with red indicating no unsaturated bonds, blue indicating more than 5 unsaturated bonds, with a gradient representing intermediate numbers of unsaturated bonds. There is no apparent correlation between the level of saturation at a given fatty acyl moiety length and the level of enrichment. On the other hand, there is a noticeable decrease in enrichment of phospholipids with longer fatty acyl moieties (> 36 carbons) for all three proteins (Spearman’s rho < −0.41, p < 0.005, two-sided Spearman’s test). N=3 replicates per protein.

Extended Data Fig. 5 | — a, Chemical library used for studying the effect of length and headgroup on phospholipid enrichment in MED1 condensates.

b, Condensate metabolome experiments were performed using the chemical library from d rather than liver metabolite extract, with each molecule added to a final concentration of 100 nM (blue), 1 μM (green) or 10 μM (orange). The role of phospholipid tail length was assessed by plotting the combined tail length (x-axis) of each phosphatidylcholine against their log₂-fold enrichment in MED1 condensates (y-axis). Choline glycerophosphoryl head group (carbons=0) and the library’s lysophosphatidylcholine (carbons=16) were also included. Results from each replicate are plotted as separate dots and crossbars represents means across replicates. Enrichment is only apparent with fatty acyl moiety chain length >15. Combined fatty acyl moieties lengths of 16 and 18 have reduced enrichment at the lowest concentration, while phospholipids with longer fatty acyl moieties (>23) have higher enrichment at lower concentrations. Error bars indicate s.e.m. (n=3).

c, Condensate enrichment was compared between phospholipids with dipalmitoyl fatty acyl moieties, but different head groups using the library described in a. The log₂-fold enrichment in MED1 condensates (y-axis) is plotted for each phospholipid (x-axis) when metabolites were added 100 nM (blue), 1 μM (green) or 10 μM (orange). Results of individual replicates are plotted as dots (black), while bars indicate the mean. Error bars indicate s.e.m. (n=3).

d, Net neutral charge phospholipids preferentially partition into MED1 and HNRNPA1 condensates. Phospholipids from the metabolomics datasets analyzed in Figure 2 were grouped based on head group (x-axis): sphingomyelins (SM, n=4), phosphatidylcholines (PC, n=4), phosphatidylethanolamines (PE, n=27), phosphatidylinositols (PI, n=5), and phosphatidylserines (PS, n=11). For each group, median log₂-fold enrichment was plotted for nucleocapsid (purple), MED1 (blue; p=0.02637, SM/PS; p=0.02584, PE/PI; p=0.0001876, PE/PS) or HNRNPA1 (green; p=0.01587, SM/PI; p=0.001465, SM/PS, p=0.01587, PC/PI; p=0.001465, PC/PS; p=0.00149, PE/PI; p=6.79e-06, PE/PS) condensates. Violin plots represent the distribution of median enrichment for each head group and lines demarcate quartiles. *p<0.05, **p<0.005, two-sided Wilcoxon rank-sum test (n=3 condensate metabolomic experiments per protein).

Extended Data Fig. 6 | — a, Oregon Green phospholipid is enriched in nucleocapsid condensates under metabolomics buffer conditions. Nucleocapsid (30 μM, red) was combined with 2 μM Oregon Green dye (left, green) or phosphatidylethanolamine (right, green) in LC-MS-compatible buffer (50 mM NH₄HCO₃ pH 7.5, 50 mM NaCl, 1 mM DTT) and then generic phage RNA (150 nM) was added to promote condensate formation. Condensates were imaged after a 10 min incubation. A representative image is displayed for each condition. Oregon Green phosphatidylethanolamine, but not dye, colocalizes with each condensate. Scale bar, 5 μm.

b, Quantification of a. The median ratio of mean fluorescent signal inside condensates to the mean signal outside condensates for Oregon Green dye (Dye, orange) or phosphatidylethanolamine (Dye-phospholipid, blue) in z-stacks 1–3 μm above the slide surface (y-axis) is plotted for each condition (x-axis). Oregon Green phosphatidylethanolamine is enriched relative to dye (p=0.005083). Error bars indicate s.e.m. *p < 0.05, two-sided Welch’s t-test (n=3 imaging experiments).

c, Oregon Green phospholipid is enriched in nucleocapsid condensates in the absence of mCherry. To determine whether mCherry might drive phospholipid partitioning into nucleocapsid condensates, we asked whether Oregon Green phospholipids partition into nucleocapsid condensates that lack mCherry. Oregon Green phosphatidylethanolamine or dye (2 μM, green) was added to solutions of nucleocapsid (3 μM, grey-scale) in the presence of phage RNA (15 nM) in buffer (50 mM Tris pH 7.5, 140 mM KCl, 12 mM NaCl, 0.8 mM MgCl₂, 5% PEG-8000). Nucleocapsid condensates were imaged after 10 min using phase contrast microscopy, while imaging Oregon Green with fluorescence microscopy. Representative images are shown for each condition and two in-frame nucleocapsid condensates are expanded below. Oregon Green phosphatidylethanolamine, but not dye, colocalizes with the nucleocapsid condensate suggesting that the mCherry-tag is not required for phospholipid partitioning. Scale bar, 5 μm (n=3 for phosphatidylethanolamine samples, n=2 for dye samples).

Extended Data Fig. 7 | — a, Phospholipase cleavage sites. Arrows indicate location of phospholipase cleavage. R indicates fatty acyl chains.

b, Removing fatty acyl moieties from Oregon Green phosphatidylethanolamine depletes condensate enrichment. Oregon Green phosphatidylethanolamine (green, 2 μM) was pretreated with each of the indicated phospholipases, or vehicle, prior to being combined with nucleocapsid (red, 3 μM) and RNA (15 nM) in buffer lacking divalent cations (50 mM Tris pH 7.5, 140 mM KCl, 12 mM NaCl, 5 mM EDTA, 5% PEG-8000). Samples were imaged by fluorescence microscopy after a 10 min incubation. Pretreatment with any of the three phospholipases leads to reduced Oregon Green signal enrichment in condensates, indicating that both fatty acyl moieties are needed for condensate enrichment. EDTA inhibits the activity of both phospholipase A2 and D. EDTA (5 mM) was added, as indicated, during the phospholipid pretreatment with these phospholipases as controls. EDTA addition during phospholipase treatment with phospholipases A2 or D restored phospholipid condensate enrichment. Representative image are shown for each condition, scale bar, 5 μm.

c, Quantification of b. The median ratio of mean fluorescence signal inside nucleocapsid condensates to the mean signal outside nucleocapsid condensates for Oregon Green phosphatidylethanolamine (blue) across z-stacks 1–3 μm above the slide surface was plotted for each enzymatic treatment. EDTA restores phospholipid enrichment for PLA2 (p=0.002764) and PLD (p=0.04016). Error bars indicate s.e.m. *p<0.05, two-sided Welch’s t-test (n=2).

Extended Data Fig. 8 | — a, Oregon Green phosphatidylethanolamine did not alter MED1 condensate size in Fig. 5a. Microscopy images in the experiment performed in Fig. 5a were segmented using ImageJ RAST (see Methods). Particles > 0.1 μm² were considered condensates. Median area of these condensates across five images/replicate (y-axis, blue) is plotted for Oregon Green dye or Oregon Green phosphatidylethanolamine (x-axis). Error bars indicate s.e.m. NS, not significant, two-sided Welch’s t-test (n=6 imaging experiments).

**b-c**, Quantification of the (b) number and (c) size of MED1 condensates after phosphoinositide addition. The median number of condensates/image (y-axis) and size of condensates (y-axis; p=0.02776, PIP₂) were quantified for each replicate for each type of phosphoinositide imaged in Fig. 5d (x-axis). Error bars indicate s.e.m. NS, not significant *p<0.05, two-sided Welch’s t-test (n=4 imaging experiments).

d, Quantification of the number of large MED1 condensates in Fig. 5d. Particles with area > 0.1 μm² were considered large condensates. The median number of large condensates/image across five images (y-axis) was plotted for each imaged condition (x-axis). More large condensates formed after PIP₂ (p=0.00783) or PIP₃ (p=0.01359) addition than PI addition. Error bars indicate s.e.m. *p<0.05, two-sided Welch’s t-test (n=4 imaging experiments).

e, To determine if interactions between PIP₃ phosphate groups and divalent cations affect MED1 condensate morphology, condensate formation was performed without (top) or with (middle) EDTA (5 mM), or with both EDTA and extra MgCl₂ (bottom, final concentration 5.8 mM) for 15 min (25°C). A representative image is displayed for each condition, with the region around one condensate expanded (right). Scale bar, 5 μm (n=4 imaging experiments).

f, Quantification of MED1 condensate shape changes in e. Irregularity of condensates was quantified by subtracting the measured condensate circularity (which is between 0 and 1) for samples in e from one. Median condensate irregularity across five images/replicate (y-axis, blue) was plotted for each condition (x-axis). MED1 condensates were more circular with EDTA than without EDTA (p=0.005526) or with both EDTA and extra MgCl₂ (p=0.01472). Error bars indicate s.e.m. *p<0.05, two-sided Welch’s t-test (n=4 imaging experiments).

Extended Data Fig. 9 | — a, Neomycin was used to test the specificity of anti-PIP₂ immunofluorescence. Neomycin binds to the head groups of phosphorylated phosphoinositides. Cells were co-immunostained for PIP₂ (green) and SON. Either neomycin (2 mg/ml) or vehicle (PBS) was added to both the pre-antibody incubation with 2% FBS and during the primary antibody incubation. PIP₂ images are displayed with increased brightness for cytoplasmic granules since there is less PIP₂ in the cytoplasm than the nucleus. A representative image is displayed for each immunostain. Scale bar, 5 μm (n=2).

b, Metaplot of PIP₂ signal proximal to condensates imaged in a. PIP₂ signal was measured in the area surrounding each condensate’s center. The median protein (red) and PIP₂ (green) signal intensity across all condensate images is plotted. Each metaplot represents one representative experiment. Both the vehicle and neomycin treatments for each experiment were performed at the same time using the same imaging and display conditions. Neomycin treatment depletes the PIP₂ signal observed in condensates (n=2).

c, To determine if anti-PIP₃ immunofluorescence in condensates is specific to PIP₃, neomycin was used to deplete anti-PIP₃ immunofluorescence. Cells were co-immunostained for PIP₃ (green) and G3BP1 as described in Fig. 6c. In the presence of neomycin, there is still significant amount of background signal across the cell, but the enrichment of PIP₃ in stress granules and P bodies is no longer observed, suggesting that signal is specific. A representative image is displayed for each immunostain. Scale bar, 5 μm (n=2).

d, Metaplot of PIP₃ signal proximal to condensates imaged in c. PIP₃ signal was measured in the area surrounding each condensate’s center. The median protein (red) and PIP₃ (green) signal (intensity) across all condensate images from an experiment is plotted for immunostains with or without neomycin (as indicated), with the distance from the center of each condensate indicated on the x-axis and y-axis. Both the vehicle and neomycin treatments for each experiment were performed at the same time using the same imaging and display conditions. PIP₃ signal in condensates is depleted by neomycin (n=2).

Extended Data Fig. 10 | — a, Metaplot of PIP₂ and PIP₃ signal proximal to condensates imaged in Fig. 6e. The location of condensates was identified in a semi-automated manner. Anti-PIP₂ (above) or anti-PIP₃ (below) immunofluorescence was then measured in the area surrounding each condensate’s center. The median GFP-LAF1 (red, left) and anti-PIP₂ or anti-PIP₃ (green, right) signal intensity across all examined synthetic condensates (from n=3 biological replicates) is displayed with the distance from the center of each condensate indicated on the x-axis and y-axis. Median PIP₂ and PIP₃ immunofluorescence is higher in synthetic condensates formed by LAF-1 RGG domains than in adjacent regions. The number of condensates examined is indicated in each row.

b, PIP₂ and PIP₃ are not enriched in synthetic condensates within select cells. A synthetic, condensate-forming protein containing GFP and two copies of the *C. elegans* LAF-1 protein’s RGG domain was over-expressed from a plasmid. Immunofluorescence was performed as described in Fig. 6a, using antibodies against either PIP₂ (top) or PIP₃ (bottom). Approximately two-third of the cells containing LAF-1 condensates had condensates that did not co-localize with either phosphoinositide and these images are representative of those cells. A region containing condensates is highlighted by a white square and expanded in the adjacent image (right). Scale bar, 5 μm (n=3 experiments).

Supplementary Material

Supplementary Figures and tables

NIHMS1947329-supplement-Supplementary_Figures_and_tables.docx^{(2.1MB, docx)}

Supplementary Video legends

NIHMS1947329-supplement-Supplementary_Video_legends.docx^{(6.8KB, docx)}

Supplementary Video 1

Download video file^{(491.6KB, avi)}

Supplementary Video 2

Download video file^{(626.6KB, avi)}

Supplementary Video 3

Download video file^{(424.4KB, avi)}

Supplementary Video 4

Download video file^{(825.9KB, avi)}

Supplementary Data Set 1

NIHMS1947329-supplement-Supplementary_Data_Set_1.xlsx^{(528.3KB, xlsx)}

ACKNOWLEDGMENTS

We thank all members of the Jaffrey laboratory for comments and suggestions. We thank Dr. Sibylle Mitschka for producing the illustrations in Fig. 1b and in the abstract image. We also thank the Bio-Imaging Resource Center of Rockefeller University for their assistance in performing confocal imaging (RRID:SCR_017791). This work was supported by the National Institutes of Health grants R35NS111631 and R01CA186702 (S.R.J.); R01AR076029, R21ES032347 and R21NS118233 (Q.C.); and NIH P01 HD067244 and support from the Starr Cancer Consortium I13-0037 (S.S.G.).

Footnotes

COMPETING INTERESTS

S.R.J. is scientific advisor to, and owns equity in 858 Therapeutics. The remaining authors declare no competing interests.

DATA AVAILABILITY

Metabolomics data is publicly available at the NIH Common Fund’s National Metabolomics Data Repository (NMDR) website, the Metabolomics Workbench, https://www.metabolomicsworkbench.org where it has been assigned Project ID PR001509. The data can be accessed directly via its Project DOI: 10.21228/M8N71K. Due to the size and lack of available condensate imaging databases, raw imaging data is available upon request to the corresponding author.

REFERENCES

1.Sabari BR, Dall’Agnese A, Boija A, Klein IA, Coffey EL, Shrinivas K, Abraham BJ, Hannett NM, Zamudio AV, Manteiga JC, et al. Coactivator condensation at super-enhancers links phase separation and gene control. Science 361, (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Chong S, Dugast-Darzacq C, Liu Z, Dong P, Dailey GM, Cattoglio C, Heckert A, Banala S, Lavis L, Darzacq X, et al. Imaging dynamic and selective low-complexity domain interactions that control gene transcription. Science 361, (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Brangwynne CP, Mitchison TJ & Hyman AA Active liquid-like behavior of nucleoli determines their size and shape in Xenopus laevis oocytes. Proc. Natl. Acad. Sci 108, 4334–4339 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Hnisz D, Shrinivas K, Young RA, Chakraborty AK & Sharp PA A Phase Separation Model for Transcriptional Control. Cell 169, 13–23 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Larson AG, Elnatan D, Keenen MM, Trnka MJ, Johnston JB, Burlingame AL, Agard DA, Redding S & Narlikar GJ Liquid droplet formation by HP1α suggests a role for phase separation in heterochromatin. Nature 547, 236–240 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Strom AR, Emelyanov AV, Mir M, Fyodorov DV, Darzacq X & Karpen GH Phase separation drives heterochromatin domain formation. Nature 547, 241–245 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Altmeyer M, Neelsen KJ, Teloni F, Pozdnyakova I, Pellegrino S, Grøfte M, Rask MBD, Streicher W, Jungmichel S, Nielsen ML, et al. Liquid demixing of intrinsically disordered proteins is seeded by poly(ADP-ribose). Nat. Commun 6, 1–12 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Oshidari R, Huang R, Medghalchi M, Tse EYW, Ashgriz N, Lee HO, Wyatt H & Mekhail K DNA repair by Rad52 liquid droplets. Nat. Commun 11, 1–8 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Guillén-Boixet J, Kopach A, Holehouse AS, Wittmann S, Jahnel M, Schlüßler R, Kim K, Trussina IREA, Wang J, Mateju D, et al. RNA-Induced Conformational Switching and Clustering of G3BP Drive Stress Granule Assembly by Condensation. Cell 181, 346–361.e17 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Sanders DW, Kedersha N, Lee DSW, Strom AR, Drake V, Riback JA, Bracha D, Eeftens JM, Iwanicki A, Wang A, et al. Competing Protein-RNA Interaction Networks Control Multiphase Intracellular Organization. Cell 181, 306–324.e28 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Yang P, Mathieu C, Kolaitis RM, Zhang P, Messing J, Yurtsever U, Yang Z, Wu J, Li Y, Pan Q, et al. G3BP1 Is a Tunable Switch that Triggers Phase Separation to Assemble Stress Granules. Cell 181, 325–345.e28 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Boronenkov IV, Loijens JC, Umeda M & Anderson RA Phosphoinositide Signaling Pathways in Nuclei Are Associated with Nuclear Speckles Containing Pre-mRNA Processing Factors. Mol. Biol. Cell 9, 3547–3560 (1998). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Payrastre B, Nievers M, Boonstra J, Breton M, Verkleij AJ & Van Bergen en Henegouwen PMP A differential location of phosphoinositide kinases, diacylglycerol kinase, and phospholipase C in the nuclear matrix. J. Biol. Chem 267, 5078–5084 (1992). [PubMed] [Google Scholar]
14.Choi BH, Chen Y & Dai W Chromatin PTEN is involved in DNA damage response partly through regulating Rad52 sumoylation. Cell Cycle 12, 3442 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Steinbach N, Hasson D, Mathur D, Stratikopoulos EE, Sachidanandam R, Bernstein E & Parsons RE PTEN interacts with the transcription machinery on chromatin and regulates RNA polymerase II-mediated transcription. Nucleic Acids Res. 47, 5573–5586 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Karlsson T, Altankhuyag A, Dobrovolska O, Turcu DC & Lewis AE A polybasic motif in ErbB3-binding protein 1 (EBP1) has key functions in nucleolar localization and polyphosphoinositide interaction. Biochem. J 473, 2033–2047 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Davis WJ, Lehmann PZ & Li W Nuclear PI3K signaling in cell growth and tumorigenesis. Front. Cell Dev. Biol 3, 24 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Albi E, Mersel M, Leray C, Tomassoni ML & Viola-Magni MP Rat liver chromatin phospholipids. Lipids 29, 715–719 (1994). [DOI] [PubMed] [Google Scholar]
19.Brangwynne CP, Eckmann CR, Courson DS, Rybarska A, Hoege C, Gharakhani J, Jülicher F & Hyman AA Germline P granules are liquid droplets that localize by controlled dissolution/condensation. Science 324, 1729–1732 (2009). [DOI] [PubMed] [Google Scholar]
20.Johansson HO, Karlström G, Tjerneld F & Haynes CA Driving forces for phase separation and partitioning in aqueous two-phase systems. J. Chromatogr. B. Biomed. Sci. App 711, 3–17 (1998). [DOI] [PubMed] [Google Scholar]
21.Klein IA, Boija A, Afeyan LK, Hawken SW, Fan M, Dall’Agnese A, Oksuz O, Henninger JE, Shrinivas K, Sabari BR, et al. Partitioning of cancer therapeutics in nuclear condensates. Science 368, 1386 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Wollny D, Vernot B, Wang J, Hondele M, Safrastyan A, Aron F, Micheel J, He Z, Hyman A, Weis K, et al. Characterization of RNA content in individual phase-separated coacervate microdroplets. Nat. Commun 13, 2626 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Carlson CR, Asfaha JB, Ghent CM, Howard CJ, Hartooni N, Safari M, Frankel AD & Morgan DO Phosphoregulation of Phase Separation by the SARS-CoV-2 N Protein Suggests a Biophysical Basis for its Dual Functions. Mol. Cell 80, 1092–1103.e4 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Perdikari TM, Murthy AC, Ryan VH, Watters S, Naik MT & Fawzi NL SARS-CoV-2 nucleocapsid protein phase-separates with RNA and with human hnRNPs. EMBO J. 39, e106478 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Iserman C, Roden CA, Boerneke MA, Sealfon RSG, McLaughlin GA, Jungreis I, Fritch EJ, Hou YJ, Ekena J, Weidmann CA, et al. Genomic RNA Elements Drive Phase Separation of the SARS-CoV-2 Nucleocapsid. Mol. Cell 80, 1078–1091.e6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Cubuk J, Alston JJ, Incicco JJ, Singh S, Stuchell-Brereton MD, Ward MD, Zimmerman MI, Vithani N, Griffith D, Wagoner JA, et al. The SARS-CoV-2 nucleocapsid protein is dynamic, disordered, and phase separates with RNA. Nat. Commun. 2021 121 12, 1–17 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Lu S, Ye Q, Singh D, Cao Y, Diedrich JK, Yates JR, Villa E, Cleveland DW & Corbett KD The SARS-CoV-2 nucleocapsid phosphoprotein forms mutually exclusive condensates with RNA and the membrane-associated M protein. Nat. Commun 12, 1–15 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Boija A, Klein IA, Sabari BR, Dall’Agnese A, Coffey EL, Zamudio AV, Li CH, Shrinivas K, Manteiga JC, Hannett NM, et al. Transcription Factors Activate Genes through the Phase-Separation Capacity of Their Activation Domains. Cell 175, 1842–1855.e16 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Guo YE, Manteiga JC, Henninger JE, Sabari BR, Dall’Agnese A, Hannett NM, Spille JH, Afeyan LK, Zamudio AV, Shrinivas K, et al. Pol II phosphorylation regulates a switch between transcriptional and splicing condensates. Nature 572, 543 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Chong PA, Vernon RM & Forman-Kay JD RGG/RG Motif Regions in RNA Binding and Phase Separation. J. Mol. Biol 430, 4650–4665 (2018). [DOI] [PubMed] [Google Scholar]
31.Henninger JE, Oksuz O, Shrinivas K, Sagi I, LeRoy G, Zheng MM, Andrews JO, Zamudio AV, Lazaris C, Hannett NM, et al. RNA-Mediated Feedback Control of Transcriptional Condensates. Cell 184, 207–225.e24 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Molliex A, Temirov J, Lee J, Coughlin M, Kanagaraj AP, Kim HJ, Mittag T & Taylor JP Phase Separation by Low Complexity Domains Promotes Stress Granule Assembly and Drives Pathological Fibrillization. Cell 163, 123–133 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Weaver R & Riley RJ Identification and reduction of ion suppression effects on pharmacokinetic parameters by polyethylene glycol 400. Rapid Commun. Mass Spectrom 20, 2559–2564 (2006). [DOI] [PubMed] [Google Scholar]
34.Wang Z, Zhang G & Zhang H Protocol for analyzing protein liquid–liquid phase separation. Biophys. Rep 5, 1–9 (2019). [Google Scholar]
35.Barupal DK & Fiehn O Chemical Similarity Enrichment Analysis (ChemRICH) as alternative to biochemical pathway mapping for metabolomic datasets. Sci. Rep 7, 1–11 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Cheung HYF, Coman C, Westhoff P, Manke M, Sickmann A, Borst O, Gawaz M, Watson SP, Heemskerk JWM & Ahrends R Targeted Phosphoinositides Analysis Using High-Performance Ion Chromatography-Coupled Selected Reaction Monitoring Mass Spectrometry. J. Proteome Res 20, 3114–3123 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Zhang H, Dudley EG & Harte F Critical Synergistic Concentration of Lecithin Phospholipids Improves the Antimicrobial Activity of Eugenol against Escherichia coli. Appl. Environ. Microbiol 83, e01583–17 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Resnick LM, Barbagallo M, Dominguez LJ, Veniero JM, Nicholson JP & Gupta RK Relation of cellular potassium to other mineral ions in hypertension and diabetes. Hypertension 38, 709–712 (2001). [DOI] [PubMed] [Google Scholar]
39.Zamudio AV, Dall’Agnese A, Henninger JE, Manteiga JC, Afeyan LK, Hannett NM, Coffey EL, Li CH, Oksuz O, Sabari BR, et al. Mediator Condensates Localize Signaling Factors to Key Cell Identity Genes. Mol. Cell 76, 753–766.e6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Bratek-Skicki A, Pancsa R, Meszaros B, Van Lindt J & Tompa P A guide to regulation of the formation of biomolecular condensates. FEBS J. 287, 1924–1935 (2020). [DOI] [PubMed] [Google Scholar]
41.Ries RJ, Zaccara S, Klein P, Olarerin-George A, Namkoong S, Pickering BF, Patil DP, Kwak H, Lee JH & Jaffrey SR m6A enhances the phase separation potential of mRNA. Nature 571, 424–428 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Nott TJ, Petsalaki E, Farber P, Jervis D, Fussner E, Plochowietz A, Craggs TD, Bazett-Jones DP, Pawson T, Forman-Kay JD, et al. Phase Transition of a Disordered Nuage Protein Generates Environmentally Responsive Membraneless Organelles. Mol. Cell 57, 936 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Kilic S, Lezaja A, Gatti M, Bianco E, Michelena J, Imhof R & Altmeyer M Phase separation of 53BP1 determines liquid‐like behavior of DNA repair compartments. EMBO J. 38, e101379 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Neitcheva T & Peeva D Phospholipid composition, phospholipase A2 and sphingomyelinase activities in rat liver nuclear membrane and matrix. Int. J. Biochem. Cell Biol 27, 995–1001 (1995). [DOI] [PubMed] [Google Scholar]
45.Bradley RP, Slochower DR, Janmey PA & Radhakrishnan R Divalent cations bind to phosphoinositides to induce ion and isomer specific propensities for nano-cluster initiation in bilayer membranes. R. Soc. Open Sci 7, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Wen Y, Vogt VM & Feigenson GW Multivalent Cation-Bridged PI(4,5)P2 Clusters Form at Very Low Concentrations. Biophys. J 114, 2630–2639 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Thomas CL, Steel J, Prestwich GD & Schiavo G Generation of phosphatidylinositol-specific antibodies and their characterization. Biochem. Soc. Trans 27, 648–652 (1999). [DOI] [PubMed] [Google Scholar]
48.Osborne SL, Thomas CL, Gschmeissner S & Schiavo G Nuclear PtdIns(4,5)P2 assembles in a mitotically regulated particle involved in pre-mRNA splicing. J. Cell Sci 114, 2501–2511 (2001). [DOI] [PubMed] [Google Scholar]
49.Niswender KD, Gallis B, Blevins JE, Corson MA, Schwartz MW & Baskin DG Immunocytochemical detection of phosphatidylinositol 3-kinase activation by insulin and leptin. J. Histochem. Cytochem 51, 275–283 (2003). [DOI] [PubMed] [Google Scholar]
50.Sharma A, Takata H, Shibahara KI, Bubulya A & Bubulya PA Son Is Essential for Nuclear Speckle Organization and Cell Cycle Progression. Mol. Biol. Cell 21, 650 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Cougot N, Babajko S & Séraphin B Cytoplasmic foci are sites of mRNA decay in human cells. J. Cell Biol 165, 31 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Kedersha N, Stoecklin G, Ayodele M, Yacono P, Lykke-Andersen J, Fitzler MJ, Scheuner D, Kaufman RJ, Golan DE & Anderson P Stress granules and processing bodies are dynamically linked sites of mRNP remodeling. J. Cell Biol 169, 871–884 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Tourrière H, Chebli K, Zekri L, Courselaud B, Blanchard JM, Bertrand E & Tazi J The RasGAP-associated endoribonuclease G3BP assembles stress granules. J. Cell Biol 160, 823 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]
54.Schacht J Purification of polyphosphoinositides by chromatography on immobilized neomycin. J. Lipid Res 19, (1978). [PubMed] [Google Scholar]
55.Clark J, Anderson KE, Juvin V, Smith TS, Karpe F, Wakelam MJO, Stephens LR & Hawkins PT Quantification of PtdInsP3 molecular species in cells and tissues by mass spectrometry. Nat. Methods 8, 267–272 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Schuster BS, Reed EH, Parthasarathy R, Jahnke CN, Caldwell RM, Bermudez JG, Ramage H, Good MC & Hammer DA Controllable protein phase separation and modular recruitment to form responsive membraneless organelles. Nat. Commun 9, 1–12 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Zhang JF, Mehta S & Zhang J Signaling Microdomains in the Spotlight: Visualizing Compartmentalized Signaling Using Genetically Encoded Fluorescent Biosensors. Annu Rev Pharmacol Toxicol. 61, 587–608 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Calebiro D & Maiellaro I cAMP signaling microdomains and their observation by optical methods. Front. Cell. Neurosci 8, (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Youn JY, Dyakov BJA, Zhang J, Knight JDR, Vernon RM, Forman-Kay JD & Gingras AC Properties of Stress Granule and P-Body Proteomes. Mol. Cell 76, 286–294 (2019). [DOI] [PubMed] [Google Scholar]
60.Benayad Z, von Bülow S, Stelzl LS & Hummer G Simulation of FUS Protein Condensates with an Adapted Coarse-Grained Model. J. Chem. Theory Comput 17, 525–537 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Murthy AC, Dignon GL, Kan Y, Zerze GH, Parekh SH, Mittal J & Fawzi NL Molecular interactions underlying liquid−liquid phase separation of the FUS low-complexity domain. Nat. Struct. Mol. Biol 26, 637–648 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Century TJ, Fenichel IR & Horowitz SB The concentrations of water, sodium and potassium in the nucleus and cytoplasm of amphibian oocytes. J. Cell Sci 7, 5–13 (1970). [DOI] [PubMed] [Google Scholar]
63.Blind RD, Suzawa M & Ingraham HA Direct modification and activation of a nuclear receptor - PIP2 complex by the inositol lipid kinase IPMK. Sci. Signal 5, (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Lee JE, Cathey PI, Wu H, Parker R & Voeltz GK Endoplasmic reticulum contact sites regulate the dynamics of membraneless organelles. Science 367, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Ma W & Mayr C A Membraneless Organelle Associated with the Endoplasmic Reticulum Enables 3′UTR-Mediated Protein-Protein Interactions. Cell 175, 1492–1506.e19 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Snead WT, Jalihal AP, Gerbich TM, Seim I, Hu Z & Gladfelter AS Membrane surfaces regulate assembly of ribonucleoprotein condensates. Nat. Cell Biol 24, 461–470 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Erdos G, Pajkos M & Dosztányi Z IUPred3: Prediction of protein disorder enhanced with unambiguous experimental annotation and visualization of evolutionary conservation. Nucleic Acids Res. 49, W297–W303 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Andersen KR, Leksa NC & Schwartz TU Optimized E. coli expression strain LOBSTR eliminates common contaminants from His-tag purification. Proteins Struct. Funct. Bioinforma 81, 1857–1861 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Dettmer K et al. Metabolite extraction from adherently growing mammalian cells for metabolomics studies: Optimization of harvesting and extraction protocols. Anal. Bioanal. Chem 399, 1127–1139 (2011). [DOI] [PubMed] [Google Scholar]
70.Ser Z, Liu X, Tang NN & Locasale JW Extraction parameters for metabolomics from cell extracts. Anal. Biochem 475, 22 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Chen Q et al. Rewiring of Glutamine Metabolism Is a Bioenergetic Adaptation of Human Cells with Mitochondrial DNA Mutations. Cell Metab. 27, 1007–1025.e5 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Wishart DS et al. HMDB 5.0: the Human Metabolome Database for 2022. Nucleic Acids Res. 50, D622–D631 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Horai H et al. MassBank: a public repository for sharing mass spectral data for life sciences. J. Mass Spectrom 45, 703–714 (2010). [DOI] [PubMed] [Google Scholar]
74.LIPID MAPS^®. Lipidomics Gateway. https://www.lipidmaps.org/tools/structuredrawing/GP_p.php?headgroup=PS&sn1=18:0&sn2=20:4(5Z,8Z,11Z,14Z).
75.Chen Q et al. Measurement of Melanin Metabolism in Live Cells by [U-13C]-L-Tyrosine Fate Tracing Using Liquid Chromatography-Mass Spectrometry. J. Invest. Dermatol 141, 1810–1818.e6 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Chen Q et al. Accelerated transsulfuration metabolically defines a discrete subclass of amyotrophic lateral sclerosis patients. Neurobiol. Dis 144, 105025 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Smith CA et al. METLIN: A Metabolite Mass Spectral Database. Ther. Drug Monit 27, 747 (2005). [DOI] [PubMed] [Google Scholar]
78.Wickham H ggplot2: Elegant Graphics for Data Analysis. (Springer International Publishing, 2016). doi: 10.1007/978-3-319-24277-4. [DOI] [Google Scholar]
79.Handwerger KE, Cordero JA & Gall JG Cajal bodies, nucleoli, and speckles in the Xenopus oocyte nucleus have a low-density, sponge-like structure. Mol. Biol. Cell 16, 202–211 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
80.Sanulli S & Narlikar GJ Generation and Biochemical Characterization of Phase-Separated Droplets Formed by Nucleic Acid Binding Proteins: Using HP1 as a Model System. Curr. Protoc 1, e109 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
81.Rueden CT et al. ImageJ2: ImageJ for the next generation of scientific image data. BMC Bioinformatics 18, 1–26 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
82.Schindelin J et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 2012 97 9, 676–682 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
83.Sage D et al. DeconvolutionLab2: An open-source software for deconvolution microscopy. Methods 115, 28–41 (2017). [DOI] [PubMed] [Google Scholar]
84.Dey N et al. Richardson–Lucy algorithm with total variation regularization for 3D confocal microscope deconvolution. Microsc. Res. Tech 69, 260–266 (2006). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figures and tables

NIHMS1947329-supplement-Supplementary_Figures_and_tables.docx^{(2.1MB, docx)}

Supplementary Video legends

NIHMS1947329-supplement-Supplementary_Video_legends.docx^{(6.8KB, docx)}

Supplementary Video 1

Download video file^{(491.6KB, avi)}

Supplementary Video 2

Download video file^{(626.6KB, avi)}

Supplementary Video 3

Download video file^{(424.4KB, avi)}

Supplementary Video 4

Download video file^{(825.9KB, avi)}

Supplementary Data Set 1

NIHMS1947329-supplement-Supplementary_Data_Set_1.xlsx^{(528.3KB, xlsx)}

Data Availability Statement

[R1] 1.Sabari BR, Dall’Agnese A, Boija A, Klein IA, Coffey EL, Shrinivas K, Abraham BJ, Hannett NM, Zamudio AV, Manteiga JC, et al. Coactivator condensation at super-enhancers links phase separation and gene control. Science 361, (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R2] 2.Chong S, Dugast-Darzacq C, Liu Z, Dong P, Dailey GM, Cattoglio C, Heckert A, Banala S, Lavis L, Darzacq X, et al. Imaging dynamic and selective low-complexity domain interactions that control gene transcription. Science 361, (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R3] 3.Brangwynne CP, Mitchison TJ & Hyman AA Active liquid-like behavior of nucleoli determines their size and shape in Xenopus laevis oocytes. Proc. Natl. Acad. Sci 108, 4334–4339 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] 4.Hnisz D, Shrinivas K, Young RA, Chakraborty AK & Sharp PA A Phase Separation Model for Transcriptional Control. Cell 169, 13–23 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R5] 5.Larson AG, Elnatan D, Keenen MM, Trnka MJ, Johnston JB, Burlingame AL, Agard DA, Redding S & Narlikar GJ Liquid droplet formation by HP1α suggests a role for phase separation in heterochromatin. Nature 547, 236–240 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R6] 6.Strom AR, Emelyanov AV, Mir M, Fyodorov DV, Darzacq X & Karpen GH Phase separation drives heterochromatin domain formation. Nature 547, 241–245 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R7] 7.Altmeyer M, Neelsen KJ, Teloni F, Pozdnyakova I, Pellegrino S, Grøfte M, Rask MBD, Streicher W, Jungmichel S, Nielsen ML, et al. Liquid demixing of intrinsically disordered proteins is seeded by poly(ADP-ribose). Nat. Commun 6, 1–12 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R8] 8.Oshidari R, Huang R, Medghalchi M, Tse EYW, Ashgriz N, Lee HO, Wyatt H & Mekhail K DNA repair by Rad52 liquid droplets. Nat. Commun 11, 1–8 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R9] 9.Guillén-Boixet J, Kopach A, Holehouse AS, Wittmann S, Jahnel M, Schlüßler R, Kim K, Trussina IREA, Wang J, Mateju D, et al. RNA-Induced Conformational Switching and Clustering of G3BP Drive Stress Granule Assembly by Condensation. Cell 181, 346–361.e17 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R10] 10.Sanders DW, Kedersha N, Lee DSW, Strom AR, Drake V, Riback JA, Bracha D, Eeftens JM, Iwanicki A, Wang A, et al. Competing Protein-RNA Interaction Networks Control Multiphase Intracellular Organization. Cell 181, 306–324.e28 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R11] 11.Yang P, Mathieu C, Kolaitis RM, Zhang P, Messing J, Yurtsever U, Yang Z, Wu J, Li Y, Pan Q, et al. G3BP1 Is a Tunable Switch that Triggers Phase Separation to Assemble Stress Granules. Cell 181, 325–345.e28 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R12] 12.Boronenkov IV, Loijens JC, Umeda M & Anderson RA Phosphoinositide Signaling Pathways in Nuclei Are Associated with Nuclear Speckles Containing Pre-mRNA Processing Factors. Mol. Biol. Cell 9, 3547–3560 (1998). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R13] 13.Payrastre B, Nievers M, Boonstra J, Breton M, Verkleij AJ & Van Bergen en Henegouwen PMP A differential location of phosphoinositide kinases, diacylglycerol kinase, and phospholipase C in the nuclear matrix. J. Biol. Chem 267, 5078–5084 (1992). [PubMed] [Google Scholar]

[R14] 14.Choi BH, Chen Y & Dai W Chromatin PTEN is involved in DNA damage response partly through regulating Rad52 sumoylation. Cell Cycle 12, 3442 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R15] 15.Steinbach N, Hasson D, Mathur D, Stratikopoulos EE, Sachidanandam R, Bernstein E & Parsons RE PTEN interacts with the transcription machinery on chromatin and regulates RNA polymerase II-mediated transcription. Nucleic Acids Res. 47, 5573–5586 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R16] 16.Karlsson T, Altankhuyag A, Dobrovolska O, Turcu DC & Lewis AE A polybasic motif in ErbB3-binding protein 1 (EBP1) has key functions in nucleolar localization and polyphosphoinositide interaction. Biochem. J 473, 2033–2047 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] 17.Davis WJ, Lehmann PZ & Li W Nuclear PI3K signaling in cell growth and tumorigenesis. Front. Cell Dev. Biol 3, 24 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] 18.Albi E, Mersel M, Leray C, Tomassoni ML & Viola-Magni MP Rat liver chromatin phospholipids. Lipids 29, 715–719 (1994). [DOI] [PubMed] [Google Scholar]

[R19] 19.Brangwynne CP, Eckmann CR, Courson DS, Rybarska A, Hoege C, Gharakhani J, Jülicher F & Hyman AA Germline P granules are liquid droplets that localize by controlled dissolution/condensation. Science 324, 1729–1732 (2009). [DOI] [PubMed] [Google Scholar]

[R20] 20.Johansson HO, Karlström G, Tjerneld F & Haynes CA Driving forces for phase separation and partitioning in aqueous two-phase systems. J. Chromatogr. B. Biomed. Sci. App 711, 3–17 (1998). [DOI] [PubMed] [Google Scholar]

[R21] 21.Klein IA, Boija A, Afeyan LK, Hawken SW, Fan M, Dall’Agnese A, Oksuz O, Henninger JE, Shrinivas K, Sabari BR, et al. Partitioning of cancer therapeutics in nuclear condensates. Science 368, 1386 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R22] 22.Wollny D, Vernot B, Wang J, Hondele M, Safrastyan A, Aron F, Micheel J, He Z, Hyman A, Weis K, et al. Characterization of RNA content in individual phase-separated coacervate microdroplets. Nat. Commun 13, 2626 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R23] 23.Carlson CR, Asfaha JB, Ghent CM, Howard CJ, Hartooni N, Safari M, Frankel AD & Morgan DO Phosphoregulation of Phase Separation by the SARS-CoV-2 N Protein Suggests a Biophysical Basis for its Dual Functions. Mol. Cell 80, 1092–1103.e4 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R24] 24.Perdikari TM, Murthy AC, Ryan VH, Watters S, Naik MT & Fawzi NL SARS-CoV-2 nucleocapsid protein phase-separates with RNA and with human hnRNPs. EMBO J. 39, e106478 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] 25.Iserman C, Roden CA, Boerneke MA, Sealfon RSG, McLaughlin GA, Jungreis I, Fritch EJ, Hou YJ, Ekena J, Weidmann CA, et al. Genomic RNA Elements Drive Phase Separation of the SARS-CoV-2 Nucleocapsid. Mol. Cell 80, 1078–1091.e6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R26] 26.Cubuk J, Alston JJ, Incicco JJ, Singh S, Stuchell-Brereton MD, Ward MD, Zimmerman MI, Vithani N, Griffith D, Wagoner JA, et al. The SARS-CoV-2 nucleocapsid protein is dynamic, disordered, and phase separates with RNA. Nat. Commun. 2021 121 12, 1–17 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] 27.Lu S, Ye Q, Singh D, Cao Y, Diedrich JK, Yates JR, Villa E, Cleveland DW & Corbett KD The SARS-CoV-2 nucleocapsid phosphoprotein forms mutually exclusive condensates with RNA and the membrane-associated M protein. Nat. Commun 12, 1–15 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] 28.Boija A, Klein IA, Sabari BR, Dall’Agnese A, Coffey EL, Zamudio AV, Li CH, Shrinivas K, Manteiga JC, Hannett NM, et al. Transcription Factors Activate Genes through the Phase-Separation Capacity of Their Activation Domains. Cell 175, 1842–1855.e16 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Guo YE, Manteiga JC, Henninger JE, Sabari BR, Dall’Agnese A, Hannett NM, Spille JH, Afeyan LK, Zamudio AV, Shrinivas K, et al. Pol II phosphorylation regulates a switch between transcriptional and splicing condensates. Nature 572, 543 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] 30.Chong PA, Vernon RM & Forman-Kay JD RGG/RG Motif Regions in RNA Binding and Phase Separation. J. Mol. Biol 430, 4650–4665 (2018). [DOI] [PubMed] [Google Scholar]

[R31] 31.Henninger JE, Oksuz O, Shrinivas K, Sagi I, LeRoy G, Zheng MM, Andrews JO, Zamudio AV, Lazaris C, Hannett NM, et al. RNA-Mediated Feedback Control of Transcriptional Condensates. Cell 184, 207–225.e24 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R32] 32.Molliex A, Temirov J, Lee J, Coughlin M, Kanagaraj AP, Kim HJ, Mittag T & Taylor JP Phase Separation by Low Complexity Domains Promotes Stress Granule Assembly and Drives Pathological Fibrillization. Cell 163, 123–133 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] 33.Weaver R & Riley RJ Identification and reduction of ion suppression effects on pharmacokinetic parameters by polyethylene glycol 400. Rapid Commun. Mass Spectrom 20, 2559–2564 (2006). [DOI] [PubMed] [Google Scholar]

[R34] 34.Wang Z, Zhang G & Zhang H Protocol for analyzing protein liquid–liquid phase separation. Biophys. Rep 5, 1–9 (2019). [Google Scholar]

[R35] 35.Barupal DK & Fiehn O Chemical Similarity Enrichment Analysis (ChemRICH) as alternative to biochemical pathway mapping for metabolomic datasets. Sci. Rep 7, 1–11 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] 36.Cheung HYF, Coman C, Westhoff P, Manke M, Sickmann A, Borst O, Gawaz M, Watson SP, Heemskerk JWM & Ahrends R Targeted Phosphoinositides Analysis Using High-Performance Ion Chromatography-Coupled Selected Reaction Monitoring Mass Spectrometry. J. Proteome Res 20, 3114–3123 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R37] 37.Zhang H, Dudley EG & Harte F Critical Synergistic Concentration of Lecithin Phospholipids Improves the Antimicrobial Activity of Eugenol against Escherichia coli. Appl. Environ. Microbiol 83, e01583–17 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R38] 38.Resnick LM, Barbagallo M, Dominguez LJ, Veniero JM, Nicholson JP & Gupta RK Relation of cellular potassium to other mineral ions in hypertension and diabetes. Hypertension 38, 709–712 (2001). [DOI] [PubMed] [Google Scholar]

[R39] 39.Zamudio AV, Dall’Agnese A, Henninger JE, Manteiga JC, Afeyan LK, Hannett NM, Coffey EL, Li CH, Oksuz O, Sabari BR, et al. Mediator Condensates Localize Signaling Factors to Key Cell Identity Genes. Mol. Cell 76, 753–766.e6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R40] 40.Bratek-Skicki A, Pancsa R, Meszaros B, Van Lindt J & Tompa P A guide to regulation of the formation of biomolecular condensates. FEBS J. 287, 1924–1935 (2020). [DOI] [PubMed] [Google Scholar]

[R41] 41.Ries RJ, Zaccara S, Klein P, Olarerin-George A, Namkoong S, Pickering BF, Patil DP, Kwak H, Lee JH & Jaffrey SR m6A enhances the phase separation potential of mRNA. Nature 571, 424–428 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R42] 42.Nott TJ, Petsalaki E, Farber P, Jervis D, Fussner E, Plochowietz A, Craggs TD, Bazett-Jones DP, Pawson T, Forman-Kay JD, et al. Phase Transition of a Disordered Nuage Protein Generates Environmentally Responsive Membraneless Organelles. Mol. Cell 57, 936 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R43] 43.Kilic S, Lezaja A, Gatti M, Bianco E, Michelena J, Imhof R & Altmeyer M Phase separation of 53BP1 determines liquid‐like behavior of DNA repair compartments. EMBO J. 38, e101379 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R44] 44.Neitcheva T & Peeva D Phospholipid composition, phospholipase A2 and sphingomyelinase activities in rat liver nuclear membrane and matrix. Int. J. Biochem. Cell Biol 27, 995–1001 (1995). [DOI] [PubMed] [Google Scholar]

[R45] 45.Bradley RP, Slochower DR, Janmey PA & Radhakrishnan R Divalent cations bind to phosphoinositides to induce ion and isomer specific propensities for nano-cluster initiation in bilayer membranes. R. Soc. Open Sci 7, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R46] 46.Wen Y, Vogt VM & Feigenson GW Multivalent Cation-Bridged PI(4,5)P2 Clusters Form at Very Low Concentrations. Biophys. J 114, 2630–2639 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R47] 47.Thomas CL, Steel J, Prestwich GD & Schiavo G Generation of phosphatidylinositol-specific antibodies and their characterization. Biochem. Soc. Trans 27, 648–652 (1999). [DOI] [PubMed] [Google Scholar]

[R48] 48.Osborne SL, Thomas CL, Gschmeissner S & Schiavo G Nuclear PtdIns(4,5)P2 assembles in a mitotically regulated particle involved in pre-mRNA splicing. J. Cell Sci 114, 2501–2511 (2001). [DOI] [PubMed] [Google Scholar]

[R49] 49.Niswender KD, Gallis B, Blevins JE, Corson MA, Schwartz MW & Baskin DG Immunocytochemical detection of phosphatidylinositol 3-kinase activation by insulin and leptin. J. Histochem. Cytochem 51, 275–283 (2003). [DOI] [PubMed] [Google Scholar]

[R50] 50.Sharma A, Takata H, Shibahara KI, Bubulya A & Bubulya PA Son Is Essential for Nuclear Speckle Organization and Cell Cycle Progression. Mol. Biol. Cell 21, 650 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R51] 51.Cougot N, Babajko S & Séraphin B Cytoplasmic foci are sites of mRNA decay in human cells. J. Cell Biol 165, 31 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R52] 52.Kedersha N, Stoecklin G, Ayodele M, Yacono P, Lykke-Andersen J, Fitzler MJ, Scheuner D, Kaufman RJ, Golan DE & Anderson P Stress granules and processing bodies are dynamically linked sites of mRNP remodeling. J. Cell Biol 169, 871–884 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R53] 53.Tourrière H, Chebli K, Zekri L, Courselaud B, Blanchard JM, Bertrand E & Tazi J The RasGAP-associated endoribonuclease G3BP assembles stress granules. J. Cell Biol 160, 823 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]

[R54] 54.Schacht J Purification of polyphosphoinositides by chromatography on immobilized neomycin. J. Lipid Res 19, (1978). [PubMed] [Google Scholar]

[R55] 55.Clark J, Anderson KE, Juvin V, Smith TS, Karpe F, Wakelam MJO, Stephens LR & Hawkins PT Quantification of PtdInsP3 molecular species in cells and tissues by mass spectrometry. Nat. Methods 8, 267–272 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R56] 56.Schuster BS, Reed EH, Parthasarathy R, Jahnke CN, Caldwell RM, Bermudez JG, Ramage H, Good MC & Hammer DA Controllable protein phase separation and modular recruitment to form responsive membraneless organelles. Nat. Commun 9, 1–12 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R57] 57.Zhang JF, Mehta S & Zhang J Signaling Microdomains in the Spotlight: Visualizing Compartmentalized Signaling Using Genetically Encoded Fluorescent Biosensors. Annu Rev Pharmacol Toxicol. 61, 587–608 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R58] 58.Calebiro D & Maiellaro I cAMP signaling microdomains and their observation by optical methods. Front. Cell. Neurosci 8, (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R59] 59.Youn JY, Dyakov BJA, Zhang J, Knight JDR, Vernon RM, Forman-Kay JD & Gingras AC Properties of Stress Granule and P-Body Proteomes. Mol. Cell 76, 286–294 (2019). [DOI] [PubMed] [Google Scholar]

[R60] 60.Benayad Z, von Bülow S, Stelzl LS & Hummer G Simulation of FUS Protein Condensates with an Adapted Coarse-Grained Model. J. Chem. Theory Comput 17, 525–537 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R61] 61.Murthy AC, Dignon GL, Kan Y, Zerze GH, Parekh SH, Mittal J & Fawzi NL Molecular interactions underlying liquid−liquid phase separation of the FUS low-complexity domain. Nat. Struct. Mol. Biol 26, 637–648 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R62] 62.Century TJ, Fenichel IR & Horowitz SB The concentrations of water, sodium and potassium in the nucleus and cytoplasm of amphibian oocytes. J. Cell Sci 7, 5–13 (1970). [DOI] [PubMed] [Google Scholar]

[R63] 63.Blind RD, Suzawa M & Ingraham HA Direct modification and activation of a nuclear receptor - PIP2 complex by the inositol lipid kinase IPMK. Sci. Signal 5, (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R64] 64.Lee JE, Cathey PI, Wu H, Parker R & Voeltz GK Endoplasmic reticulum contact sites regulate the dynamics of membraneless organelles. Science 367, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R65] 65.Ma W & Mayr C A Membraneless Organelle Associated with the Endoplasmic Reticulum Enables 3′UTR-Mediated Protein-Protein Interactions. Cell 175, 1492–1506.e19 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R66] 66.Snead WT, Jalihal AP, Gerbich TM, Seim I, Hu Z & Gladfelter AS Membrane surfaces regulate assembly of ribonucleoprotein condensates. Nat. Cell Biol 24, 461–470 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R67] 67.Erdos G, Pajkos M & Dosztányi Z IUPred3: Prediction of protein disorder enhanced with unambiguous experimental annotation and visualization of evolutionary conservation. Nucleic Acids Res. 49, W297–W303 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R68] 68.Andersen KR, Leksa NC & Schwartz TU Optimized E. coli expression strain LOBSTR eliminates common contaminants from His-tag purification. Proteins Struct. Funct. Bioinforma 81, 1857–1861 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R69] 69.Dettmer K et al. Metabolite extraction from adherently growing mammalian cells for metabolomics studies: Optimization of harvesting and extraction protocols. Anal. Bioanal. Chem 399, 1127–1139 (2011). [DOI] [PubMed] [Google Scholar]

[R70] 70.Ser Z, Liu X, Tang NN & Locasale JW Extraction parameters for metabolomics from cell extracts. Anal. Biochem 475, 22 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R71] 71.Chen Q et al. Rewiring of Glutamine Metabolism Is a Bioenergetic Adaptation of Human Cells with Mitochondrial DNA Mutations. Cell Metab. 27, 1007–1025.e5 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R72] 72.Wishart DS et al. HMDB 5.0: the Human Metabolome Database for 2022. Nucleic Acids Res. 50, D622–D631 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R73] 73.Horai H et al. MassBank: a public repository for sharing mass spectral data for life sciences. J. Mass Spectrom 45, 703–714 (2010). [DOI] [PubMed] [Google Scholar]

[R74] 74.LIPID MAPS^®. Lipidomics Gateway. https://www.lipidmaps.org/tools/structuredrawing/GP_p.php?headgroup=PS&sn1=18:0&sn2=20:4(5Z,8Z,11Z,14Z).

[R75] 75.Chen Q et al. Measurement of Melanin Metabolism in Live Cells by [U-13C]-L-Tyrosine Fate Tracing Using Liquid Chromatography-Mass Spectrometry. J. Invest. Dermatol 141, 1810–1818.e6 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R76] 76.Chen Q et al. Accelerated transsulfuration metabolically defines a discrete subclass of amyotrophic lateral sclerosis patients. Neurobiol. Dis 144, 105025 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R77] 77.Smith CA et al. METLIN: A Metabolite Mass Spectral Database. Ther. Drug Monit 27, 747 (2005). [DOI] [PubMed] [Google Scholar]

[R78] 78.Wickham H ggplot2: Elegant Graphics for Data Analysis. (Springer International Publishing, 2016). doi: 10.1007/978-3-319-24277-4. [DOI] [Google Scholar]

[R79] 79.Handwerger KE, Cordero JA & Gall JG Cajal bodies, nucleoli, and speckles in the Xenopus oocyte nucleus have a low-density, sponge-like structure. Mol. Biol. Cell 16, 202–211 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R80] 80.Sanulli S & Narlikar GJ Generation and Biochemical Characterization of Phase-Separated Droplets Formed by Nucleic Acid Binding Proteins: Using HP1 as a Model System. Curr. Protoc 1, e109 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R81] 81.Rueden CT et al. ImageJ2: ImageJ for the next generation of scientific image data. BMC Bioinformatics 18, 1–26 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R82] 82.Schindelin J et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 2012 97 9, 676–682 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R83] 83.Sage D et al. DeconvolutionLab2: An open-source software for deconvolution microscopy. Methods 115, 28–41 (2017). [DOI] [PubMed] [Google Scholar]

[R84] 84.Dey N et al. Richardson–Lucy algorithm with total variation regularization for 3D confocal microscope deconvolution. Microsc. Res. Tech 69, 260–266 (2006). [DOI] [PubMed] [Google Scholar]

PERMALINK

Biomolecular condensates create phospholipid-enriched microenvironments

Jason G Dumelie

Qiuying Chen

Dawson Miller

Nabeel Attarwala

Steven S Gross

Samie R Jaffrey

SUMMARY

INTRODUCTION

RESULTS

Metabolomics of phase-separated condensates

Fig. 1 |. A method for measuring condensate metabolomes.

Fig. 2 |. Phospholipids are enriched in condensates.

Hydrophobicity and polarity affect phospholipid partitioning

Fig. 3 |. Phospholipid hydrophobicity is important for condensate partitioning.

Phospholipids partition into diverse condensates

Fig. 4 |. Phospholipids partition into diverse types of condensates.

Phospholipids trigger changes in MED1 phase separation

Fig. 5 |. Phospholipids can alter the quantity, size and morphology of MED1 condensates.

Phosphoinositides localize to intracellular condensates

Fig. 6 |. Phospholipids are enriched in multiple condensates in cells.

DISCUSSION

METHODS

Protein purification

RNA purification

Condensate metabolomics

Metabolite extraction from mice livers

Condensate spin-down with metabolites

LC-MS Reagents

LC-MS metabolomics platform

Further validation of metabolite identifications

Condensate metabolomics using a defined library of compounds

Measuring condensate fraction RNA and protein levels

Metabolomics fold-change exclusions and analysis

Metabolomics data analysis and statistics

Intra-nuclear phospholipid concentration calculations

Imaging condensates

Phospholipid reagents

Passivization of glass wells

Condensate imaging for metabolomics protocol optimization

Fluorescence recovery after photobleaching

Imaging phospholipids in condensates

Nucleocapsid condensates without an mCherry tag

Phospholipase treatment

Condensate changes in response to phospholipid addition

Immunofluorescence

Image analysis and processing

Phospholipid or dye enrichment in condensates

FRAP analysis

Changes in condensate after phospholipid addition

Condensate enrichment detected by immunofluorescence

Production of drawings and illustrations.

Reporting Summary.

Extended Data

Extended Data Fig. 1 |. Developing a novel method for measuring condensate metabolomes.

Extended Data Fig. 2 |. Processing and assessing quality of condensate metabolomics data.

Extended Data Fig. 3 |. The role of RNA in metabolite condensate partitioning.

Extended Data Fig. 4 |. The role of fatty acyl moieties in phospholipid partitioning.

Extended Data Fig. 5 |. The role of charge and hydrophobicity in phospholipid partitioning.

Extended Data Fig. 6 |. Phospholipids co-localize with condensates under diverse conditions.

Extended Data Fig. 7 |. Phospholipase treatment inhibits Oregon Green phosphatidylethanolamine condensate partitioning.

Extended Data Fig. 8 |. Phospholipids can alter condensates.

Extended Data Fig. 9 |. Neomycin depletes anti-PIP2 and anti-PIP3 immunostaining.

Extended Data Fig. 10 |. Variable anti-PIP2 and anti-PIP3 antibody signal in synthetic condensates formed by LAF-1 RGG domains.

Supplementary Material

ACKNOWLEDGMENTS

Footnotes

DATA AVAILABILITY

REFERENCES

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases

Extended Data Fig. 9 |. Neomycin depletes anti-PIP₂ and anti-PIP₃ immunostaining.

Extended Data Fig. 10 |. Variable anti-PIP₂ and anti-PIP₃ antibody signal in synthetic condensates formed by LAF-1 RGG domains.