A platform to induce and mature biomolecular condensates using chemicals and light

Abstract

Biomolecular condensates are membraneless compartments that impart spatial and temporal organization to cells. Condensates can undergo maturation, transitioning from dynamic liquid-like states into solid-like states associated with neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) and Huntington’s Disease. Despite their important roles, many aspects of condensate biology remain incompletely understood, requiring tools for acutely manipulating condensate-relevant processes within cells. Here we used the BCL6 BTB domain and its ligands BI-3802 and BI-3812 to create a chemical genetic platform, BTBolig, allowing inducible condensate formation and dissolution. We also developed optogenetic and chemical methods for controlled induction of condensate maturation, where we surprisingly observed recruitment of chaperones into the condensate core and formation of dynamic biphasic condensates. Our work provides insights into the interaction of condensates with proteostasis pathways, and introduces a suite of chemical-genetic approaches to probe the role of biomolecular condensates in health and disease.

Introduction

Cell compartmentalization is critical for orchestrating the highly-ordered dynamic biochemical activities important for cellular function. In addition to classical compartments consisting of membrane-surrounded organelles, cells also possess non-membranous compartments called biomolecular condensates. These structures can form when proteins and/or nucleic acids demix and coalesce in solution, often through principles of liquid-liquid phase separation¹. Biomolecular condensates are highly dynamic in space and time, forming and dissolving according to cellular needs, providing at-will compartments for enzymatic reactions, protein sequestration, and other activities.

A key characteristic of proteins promoting biomolecular condensate formation is multivalency, which is achieved through multiple binding domains, oligomerization, or intrinsically disordered regions (IDRs) that promote abundant weak interactions^2–4. The valency and interconnectedness of condensate-forming proteins can also affect the material properties of condensates^5,6, which can range from liquid-like, with high fluidity and exchange of proteins between dilute and concentrated phases, to more gel- or solid-like states, with little exchange between phases. How the material properties of a biomolecular condensate affect a condensate’s function and how the cell interacts with condensates of different material properties are subjects that remain poorly understood.

In healthy cells, biomolecular condensates are involved in stress responses, mRNA processing, DNA nuclear organization, and other processes^7,8. Condensates can mature or age over time, transitioning from dynamic liquid-like states into less dynamic gel- or solid-like states^9,10. Such changes have been linked to disease, including neurodegenerative disease-associated proteinopathies such as ALS and Alzheimer’s disease³. Despite extensive efforts, there remains a lack of understanding of the role of condensates and condensate aging in disease development and interactions with cellular proteostasis pathways, and a lack of therapeutic approaches to ameliorating solid-like condensate states.

While many aspects of condensate biology have been studied in vitro², more recent tools have been developed for dynamic in vivo manipulation of condensates^5,11–17. Methods to form condensates with light include: CRY2olig, a variant of the plant photoreceptor cryptochrome 2 (CRY2) that forms solid-like condensates with light¹¹; ‘optoDroplets,’ consisting of a CRY2 domain fused to the IDR of Fused in Sarcoma (FUS), which forms liquid-like condensates with low intensity light⁵; and ‘Corelets,’ which combine a multimeric domain with FUS IDR in light¹⁵. Several chemical-induced systems have also been developed, including the iPOLYMER system, based on repeats of FRB and FKBP12 that interact upon rapamycin addition¹³, and a metal-ion-induced system that uses a Nck SH3 domain and its binding partner fused to a 6xHis tag¹⁶. These synthetic systems have proven useful but each has specific limitations, such as a requirement for two proteins, poor reversibility, or the ability to form only one type of condensate material state.

To expand the toolset of systems for manipulating condensates, we developed a new platform based on the BTB (Broad-Complex, Tram-track, and Bric a brac) domain from the transcription factor BCL6 (B-cell lymphoma 6). Previously, BCL6 and truncated versions (1–275 aa or 1–250 aa) fused to EGFP were found to oligomerize in the presence of a ligand, BI-3802¹⁸. Oligomerization could be reverted with a second ligand, BI-3812, that competes with BI-3802 for binding but does not induce BCL6 oligomerization¹⁸. A follow-up study showed that the BTB domain alone (1–129 aa) fused to EGFP could form oligomeric foci and act as a reversible oligomerization switch¹⁹. Here, we expand on use of the BCL6 BTB domain to develop BTBolig, a multi-use platform for chemical regulation of biomolecular condensates. We demonstrate use forming and dissolving condensates of different material states: liquid-like, solid-like, or fibril-like, and studying the interaction of different condensate states with cellular proteostasis pathways. Building on this platform, we developed an optogenetic or chemical-genetic method, Condensate State Modulation (CoSMo), allowing controlled transition of condensates from liquid-like to solid-like states. Using CoSMo, we identify and characterize dynamic interactions with chaperone proteins that result in formation of biphasic condensates. Our work develops BTBolig as a versatile platform for exploring biomolecular condensate biology and provides a model system to study condensate maturation and chaperone interactions.

Results

BTB-based platform for manipulating biomolecular condensates

To build a chemical-genetic system for regulation of biomolecular condensates, we explored the BI-3802-dependent oligomerization of BCL6^18,19 (Fig. 1a). We tested four configurations of the BCL6 BTB domain (residues 1–129) fused to a monomeric fluorescent protein, mCherry, for condensate formation with BI-3802 (Fig. 1b), including a single BTB domain (BTB-mCh), as well as a tandem dimer (BTB2x-mCh). In addition, adopting a strategy previously used with CRY2 to generate more liquid-like optoDroplets⁵, we tested a configuration containing the IDR of FUS, BTB-mCherry-FUS_IDR (hereafter, we refer to this mCherry tagged version as ‘BTB_IDR’). To add versatility, we also generated a version, BTB_IDR-FRB, containing a hook protein, FRB, at the C-terminus. FRB undergoes rapamycin-dependent dimerization with FKBP12, providing an inducible means to recruit FKBP-fused cargo into the condensate at specific times. Finally, as studies have indicated that fusion of fluorescent proteins with oligomeric character (such as EGFP) can augment oligomerization phenotypes (in some cases resulting in formation of fibril-like structures^11,20), we also generated a BTB-EGFP fusion. While performing these studies, another group characterized a similar BTB-EGFP construct as a reversible oligomerization switch¹⁹. We include our characterization of BTB-EGFP as we found differences with the prior studies (and extended our studies in different directions), and for direct comparison with the other BTB configurations.

Figure 1. BTBolig platform for inducible condensate manipulation.

a) Schematic of BTBolig system. BI-3802 induces oligomerization resulting in condensate formation. Non-oligomerizing BI-3812 competes with BI-3802 for binding and induces condensate dissolution. b) Constructs tested. ‘BTB’, residues 1–129 of BCL6. c) COS-7 cells expressing BTB-mCh treated with 1 μM BI-3802 do not form condensates. Images representative of 3 independent experiments. Scale bar, 10 μm. d) Imaging and quantification of BTB systems. At left, representative images of the same COS-7 cell expressing each BTB system before drug addition, then sequentially exposed to 1 μM BI-3802 then 10 μM BI-3812. BTB2x-mCh images acquired 8 min post BI-3802 & 16 min post BI-3812; BTB_IDR, 16 min post BI-3802 & 16 min post BI-3812; BTB_IDR-FRB, 4.5 min post BI-3802 & 5 min post BI-3812; BTB-EGFP, 17 min post BI-3802 & 17 min post BI-3812). Scale, 10 μm. Kymographs of the regions in yellow boxes are in center. Graphs at right show the kinetics (t_1/2) of condensate formation and dissolution, mean and s.e.m., n= 20 (2x-mCh), n=15 (BTB_IDR), n=11 (_IDR-FRB), n=12 (EGFP) cells, 3 independent experiments.

We first assessed whether each BTB version could form condensates in COS-7 cells after addition of BI-3802. BTB-mCherry did not form condensates (Fig. 1c), however each of the other systems rapidly formed condensates with 1 μM BI-3802 (Fig. 1d, Supplementary Videos 1–4). BTB2x-mCh and BTB-EGFP condensates formed more slowly (BTB2x-mCh t_1/2, 257 ± 29 s; BTB-EGFP t_1/2, 385 ± 65s), while condensates containing the FUS_IDR formed more quickly (BTB_IDR t_1/2, 128 ± 16 s; BTB_IDR-FRB t_1/2, 132 ± 28 s). The addition of 10 μm BI-3812 to recently- formed condensates resulted in their complete dissolution within seconds to minutes (BTB2x-mCh t_1/2, 581 ± 84 s; BTB_IDR t_1/2, 57 ± 11s; BTB_IDR-FRB t_1/2, 28 ± 2 s; BTB-EGFP t_1/2, 557 ± 95 s) (Fig. 1d). A BTB-EYFP version was also effective in forming condensates, showing similar behavior as BTB-EGFP (Supplementary Fig. 1).

After establishing that each system could form and dissolve condensates, we compared other properties (Fig. 2). Prior to BI-3802 addition, BTB2x-mCh, BTB_IDR, and BTB-EGFP showed diffuse distribution in cells with no condensates present, while approximately half of cells expressing BTB_IDRFRB (those at higher expression levels) showed some preclustering (15.9% of total protein), exclusively in the nucleus (Fig. 2a; Extended Data Fig. 1). With 1 μM BI-3802, 64 ± 11% of BTB2x-mCh and 41 ± 14% of BTB-EGFP protein was recruited into dense puncta within 20 min (Fig. 2a). The fraction of BTB_IDR separating into the dense phase was initially lower (~22% at <20 min after BI-3802 addition), but increased over time (Supplementary Fig. 2a). In comparison, higher levels (57 ± 9%) of BTB_IDR-FRB partitioned into condensates initially, suggesting a lower saturation concentration (Fig. 2a). Using BTB_IDR, we found that BI-3802 could be titrated to tune the extent of condensate formation (Supplementary Fig. 2b). Both BTB_IDR-FRB and BTB-EGFP condensates could be formed and dissolved multiple times without loss of efficacy (Extended Data Fig. 2; Supplementary Video 5), in contrast to prior studies of BTB-EGFP reporting a loss of efficacy with multiple cycles.¹⁹

Figure 2. Comparison of BTBolig system properties.

a) Quantification of % clustered protein before and 15–30 min after 1 μM BI-3802 in COS-7 cells expressing BTB2x-mCh (n=29 cells); BTB_IDR (n=27); BTB-EGFP (n=19); or BTB_IDR-FRB (n=33). Graph shows mean and s.e.m., combined data from 3–5 independent experiments. Pre BI-3802 values for BTB2x-mCh, BTB-EGFP, and BTB_IDR were 0. b) Circularity of cytosolic condensates in COS-7 cells expressing indicated BTB constructs, quantified at 15–30 min post 1 μM BI-3802. Graph shows box plot (defined in Methods), n=109 (BTB_IDR), 246 (_IDR-FRB), 345 (2x-mCh), or 353 (EGFP) condensates from 2 independent experiments. c) Normalized FRAP analysis of BTBolig condensates in COS-7 cells, 5–60 min post addition of BI-3802. Mean and s.e.m., data from 2 independent experiments, BTB2x-mCh, n=6 condensates; BTB_IDR, t_1/2=24.4s, n=11; BTB-EGFP, n=7; BTB_IDR-FRB, t_1/2=42.6s, n=6. Insets show representative regions pre and post bleach. Scale, 2 μm. d) BTB_IDR and BTB_IDR-FRB condensates undergo fusion and coalescence into a sphere, while BTB2x-mCh and BTB-EGFP merge into irregular shapes. Scale bars, 2 μm. Images representative of 5 independent experiments. e) Quantification of merging events. Graph at top shows plot of condensate fusion time vs length of merging condensate diameters. The inverse capillary velocity for BTB_IDR (1.08 s/μm) is calculated from the slope of the plotted line. The data for BTB_IDR-FRB did not fit a line plot. Graph at bottom shows plot of condensate fusion times/diameter vs time after BI-3802 addition. BTB_IDR-FRB condensates showed an increase in this ratio over time. Data represents combined individual fusion events, BTB_IDR, n=24 events from 4 independent experiments; BTB_IDR-FRB, n=17 events from 5 independent experiments.

We next quantified the material states of each condensate system. BTB2x-mCh and BTB-EGFP condensates showed low circularity, poor recovery during fluorescent recovery after photobleaching (FRAP) experiments, and lack of coalescence, indicating a gel-like or solid-like state (Fig. 2b–d). In turn, newly formed (<1 h after BI-3802 addition) BTB_IDR and BTB_IDR-FRB condensates were circular, coalesced and reformed a spherical shape, and showed fast recovery after photobleach (BTB_IDR FRAP t_1/2 = 24.4s; BTB_IDR-FRB FRAP t_1/2 = 42.6s) suggesting a liquid-like state (Fig. 2b–d). To gain a more quantitative understanding of BTB_IDR and BTB_IDR-FRB condensates, we examined condensate merging events. For each event, we plotted the coalescence time versus length of the two merging condensate diameters (Fig. 2e, top). The slope of the fitted line gives a measurement of inverse capillary velocity, corresponding to the ratio of viscosity/surface tension of the merging condensates²¹. The BTB_IDR condensate measurements showed a linear relationship, yielding an inverse capillary velocity of 1.06 s/μm, however BTB_IDR-FRB condensates did not. By plotting the ratio of coalescence time/condensate length we found that the BTB_IDR-FRB condensates were undergoing a time-dependent increase in fusion time, such that equally sized 20–30 min aged condensates were coalescing more slowly than newly-formed condensates (Fig. 2e, bottom). This suggested an increase in viscosity and/or reduced surface tension consistent with condensate maturation. In agreement, 1.5 h and 4 h aged BTB_IDR-FRB condensates showed a reduction in mobility after photobleach compared with newly-formed condensates, while BTB_IDR condensates remained highly mobile at 4 h and even 19 h (Extended Data Fig. 3).

Assessing condensate interactions with cell quality control

We saw an opportunity to use the BTBolig platform to explore how condensates of diverse structural and material properties interact with cell quality control pathways. For example, are more solid-like condensates processed by the cell differently than liquid-like condensates? We formed BTB_IDR, BTB2x-mCh, and BTB-EGFP condensates with BI-3802 and tracked their localization over 45 min, observing large differences in the extent and timing of clearance (Fig. 3a). BTB-EGFP condensates were rapidly cleared, showing mobilization to a perinuclear region over tens of minutes. This phenotype was strikingly different from that observed with either BTB_IDR or BTB2x-mCh condensates, which remained dispersed. At 1.5–3 h post BI-3802, many BTB-EGFP condensates had formed fibril-like structures localizing to a perinuclear region surrounded by vimentin, a protein that forms a cage-like structure around misfolded proteins and serves as an aggresomal marker²² (Fig. 3b). In contrast, BTB2x-mCh condensates remained dispersed at 1.5 h (Fig. 3b). Structural differences between BTB2x-mCh and EGFP-BTB were clear at 3 h, with much of BTB2x-mCh showing compact irregular shapes, while EGFP-BTB assembled into thin fibrils (Fig. 3c), in some cases extending out of the aggresome or localized in the periphery (Supplementary Fig. 3). In contrast, at 1.5–3 h post induction the liquid-like BTB_IDR condensates maintained their rounded shapes, continued to coalesce and grow in size, and showed dispersed localization throughout the cytosol with no vimentin colocalization (Fig 3b,c).

Figure 3. BTB condensates are differentially cleared to the aggresome and can be dissolved at later timepoints by addition of BI-3812.

a) COS-7 cells expressing BTB-EGFP, BTB2x-mCh, or BTB_IDR indicated times before or after 1 μM BI-3802. BTB-EGFP condensates are rapidly cleared to a perinuclear location. Images representative of 3 independent experiments. b) COS-7 cells expressing vimentin-mCh or vimentin-mNeonGreen (mNG) and indicated BTBolig versions, 1.5 h after 1 μM BI-3802. Vimentin forms a cage surrounding the coalesced BTB-EGFP condensates, marking the aggresome. Images representative of 2 independent experiments. c) COS-7 cells expressing BTB-EGFP, BTB2x-mCh, or BTB_IDR 3 h after 1 μM BI-3802, showing shape differences between the three condensate types. Images representative of 3 independent experiments. d) COS-7 cells expressing BTBolig versions and mCherry or mNeonGreen (mNG) labeled vimentin and treated with 1 μM BI-3802 for 24 h, before and after 10 μM BI-3812 resulting in condensate dissolution. Images representative of 2 independent experiments. e) Representative images (left) and quantification (right) of % protein in condensates during condensate dissolution with BI-3812 in experiment shown in (d). The images are a timelapse series of the areas outlined in white in (d). Graph shows mean and s.e.m., n=9 (EGFP), 12 (BTB2x-mCh), or 10 (BTB_IDR) cells from 2 independent experiments. f) BTB-EYFP fiber-like condensates formed after 24h BI-3802 treatment, dissolving after 10 μM BI-3812. Images representative of 2 independent experiments. All scale bars, 10 μm.

At 24 h post BI-3802, the majority of BTB-EGFP and BTB2x-mCh was localized to the aggresome, while BTB_IDR was distributed at sites around the nuclear envelope (Fig. 3d). Addition of BI-3812 dissolved BTB_IDR, BTB-EGFP, and BTB2x-mCh condensates within minutes, even when condensates were localized to the aggresome (Fig. 3d,e). Interestingly, and in contrast to the complete dissolution observed at early timepoints (as in Fig. 1d), at 24 h most condensates contained a small fraction of protein that remained refractory to dissolution (5.3% for BTB_IDR; 10.9% for BTB-EGFP) (Fig. 3e). With BTB-EYFP, even large fiber-like structures that had assembled (some over 2 μm in width) remarkably dissolved with BI-3812 (Fig. 3f; Supplementary Video 6).

Control of protein function using BTB2x-mCh

The rapid induction and reversible nature of the BTBolig platform suggested it could be useful for manipulation of protein function. Light-induced oligomerization tools were previously used to manipulate endocytosis, signaling pathways, and cell death^11,23–28, however chemical-dependent oligomerization has been less commonly used. The rapamycin-dependent iPOLYMER system was used to synthetically generate RNA granules, but requires two scaffold proteins and is not reversible¹³. BTB-EGFP was previously used to activate an EGFR-dependent signaling cascade¹⁹, but as noted above, this protein forms fibril-like structures and is rapidly cleared, factors that could limit many applications.

To test the use of BTB2x-mCh to manipulate function, we examined its ability to induce activity of MLKL (mixed lineage kinase domain-like) protein, involved in the final stages of necroptotic cell death. During necroptosis, MLKL is induced to oligomerize, resulting in plasma membrane (PM) localization and membrane rupture^29–31. Previously, we showed that a FLAG-tagged MLKL (residues 1–140) fused to CRY2olig, which undergoes light-dependent oligomerization, could trigger necroptosis with light³². To test with BTB, we replaced CRY2olig with BTB2x-mCh. In HEK293T cells, untreated FLAG-MLKL(1–140)-BTB2x-mCh showed diffuse cytosolic localization, then rapidly relocalized to the PM upon addition of BI-3802 and induced death (~60% of cells dead after 30–40 min), with no cell death observed in controls lacking BI-3802 (Fig. 4a–c).

Figure 4. Inducible control of oligomerization-dependent activity using BTB2x-mCh.

a) Strategy for BI-3802-induced necroptosis. Oligomerization of FLAG-MLKL(1–140)-BTB2x-mCherry is triggered with BI-3802, leading to PM recruitment and cell lysis. b) HEK293T cells expressing FLAG-MLKL(1–140)-BTB2x-mCh (cyan) and treated with 1 μM BI-3802. FLAG-MLKL is recruited to the PM and results in cell death, visualized with SYTOX Green nuclear stain (purple). Images representative of 3 independent experiments. c) Quantification of cell death (SYTOX Green stain) in HEK293T cells expressing FLAG-MLKL-BTB2x-mCh 30–40 min after addition of 1 μM BI-3802 (n=13 regions) or DMSO control (n=12 regions). Mean and s.e.m., 3 independent experiments. d) Strategy for avidity-induced membrane recruitment using inactive MLKL (SsrA-MLKL(1–140)) and BTB2x-mCh oligomerization. e) SsrA-MLKL(1–140)-BTB2x-mCh expressed in HEK293T cells showing rapid relocalization to the plasma membrane 10 min after addition of BI-3802, which was reversible with 10 μM BI-3812 (imaged 19 min after BI-3812). Images representative of 3 independent experiments. f) Half-time of translocation of SsrA-MLKL(1–140)-BTB2x-mCh to the PM with 0.5 μM BI-3802 addition (n=11 cells), and translocation from the PM to the cytosol with 10 μM BI-3812 (n=9 cells). Mean and s.e.m., combined data from 2 independent experiments. g) Strategy for BI-3802-induced translocation of OCRL to the PM. Drug-induced PM recruitment of the inositol 5-phosphatase domain of OCRL results in depletion of PI(4,5)P2, which can be visualized with a iRFP-PH_PLCδ1 biosensor that translocates from the PM to cytosol. h, i) Representative images (h) and quantification of kinetics (i) of HEK293T cells expressing SsrA-MLKL(1–140)-BTB2x-mCh-OCRL and iRFP-PH_PLCδ1 before treatment, 2 min post 1 μM BI-3802, and 10 min post 10 μM BI-3812. Graphs in (i) show the normalized fluorescent intensity of SsrA-MLKL-BTB2x-mCh-ORCL or iRFP-PH_PLCδ1 at the cytosol as a function of time after addition of each ligand. Data from a single experiment, representative of 2 experimental repeats. All scale bars, 10 μm.

Previously, we used a N-terminal peptide fusion to generate an inactive version of MLKL-CRY2olig (SsrA-MLKL-CRY2olig) that allows oligomerization-dependent recruitment of a cargo to the PM without inducing cell death³². We converted this light-triggered system to a chemical-triggered system, replacing CRY2olig with BTB2x (Fig. 4d). SsrA-MLKL-BTB2x-mCh showed robust recruitment to the PM with BI-3802 (half-time, 101 ± 14 s) that was reversible with 10 μM BI-3812 (half-time, 279 ± 62 s) (Fig. 4e,f). We fused this module to the inositol 5-phosphatase (5-ptase) domain of the phosphatase OCRL (Fig. 4g), which when recruited to the PM results in dephosphorylation of PI(4,5)P2 ³³. Using an iRFP-PH_PLCδ1 sensor that translocates from the PM to the cytosol upon loss of PI(4,5)P2³³, we observed that with 1 μM BI-3802, OCRL was recruited to the PM and iRFP-PH_PLCδ1 translocated from the PM to the cytosol, while 10 μM BI-3812 reversed the process (Fig. 4h,i). Our results demonstrate BTB2x-mCh is effective in inducing cellular responses triggered by oligomerization and can be substituted in place of CRY2olig for applications.

Induced maturation reveals dynamic chaperone interplay

Maturation of biomolecular condensates occurs when a liquid-like condensate becomes more gel- or solid-like, experiencing a loss of free protein exchange between the dense and dilute phases. Condensate maturation has been proposed to lead to reduced enzyme kinetics¹ and can result in cell toxicity or perturbed function^9,34,35. A better understanding of condensate maturation requires versatile tools that allow controlled transition between different material states. To develop such a tool, we built on a prior method, DisCo (Dissociation of Condensates), where we used an inducible protein-protein interaction to recruit a monovalent protein ligand into a condensate, resulting in its dissolution¹⁴. We reasoned that recruitment of a multivalent protein into the BTB_IDR condensate could result in condensate crosslinking and maturation, a method we termed CoSMo, for Condensate State Modulation (Fig. 5a). To first test this approach, we used FRB as a hook (with BTB_IDR-FRB) to recruit a multivalent tandem FKBP upon addition of rapamycin. We coexpressed BTB_IDR-FRB and mCh-FKBP2x in cells, then formed condensates with 1 μM BI-3802. Addition of 333 nM rapamycin to newly-formed condensates (<30 min after BI-3802) induced the recruitment of mCh-FKBP2x, resulting in a loss of condensate circularity and mobility, consistent with a transformation to a more gel-like state (Fig. 5b, Extended Data Fig. 4a–c). The CoSMo approach was effective using different multivalent FKBP-fused ligands, and was tunable with different rapamycin concentrations (Extended Data Fig. 4d–h).

Figure 5. Light-induced CoSMo approach reveals dynamic chaperone-condensate interactions.

a) Schematic of CoSMo approach. b) COS-7 cells expressing BTB_IDR-FRB and mCh-FKBP2x, treated with 1 μM BI-3802 to induce condensates for <30 min, before or after 333 nM rapamycin (rap) (higher magnification at bottom). Images representative of 2 independent experiments. Scale, 10 μm. c) FRAP analysis of CIBN-BTB_IDR-FRB condensates coexpressed with CRY2olig-mEGFP, 30 min post BI-3802 but before blue light. Mean and s.e.m., n=5 condensates from one experiment. d) CIBN-BTB_IDR-FRB condensates in cells coexpressed with CRY2olig-mEGFP and 1 μM BI-3802, before (5 min post BI-3802) and 11.5 min post light (488 nm, 5% laser, 7×50ms pulses, every 30s). Scale bar, 10 μm. Images representative of 5 independent experiments. Image at near right shows mEGFP signal in the yellow boxed region. Far right, timelapse of the region depicted in the white box (CIBN-BTB_IDR-FRB condensate). e) FRAP analysis of monophasic or biphasic CIBN-BTB_IDR-FRB + CRY2olig-miRFP condensates, treated 30 min with 1 μM BI-3802 and light (465 nm, 1.1 mW/cm², 2 s every 2 min). Mean and s.e.m., n=9 condensates from 2 independent experiments (monophasic) or n=4 condensates from one experiment (biphasic). f) COS-7 cells expressing CIBN-BTB_IDR-FRB, CRY2olig-miRFP, and GFP-HSPA1L or mEGFP-HSPA1A treated with 1 μM BI-3802 and blue light. Scale, 10 μm. Yellow box, zoom image of outlined region. Images representative of 5 independent experiments. g) Images (top) and quantification (bottom) showing dynamic changes in BTB/HSP70 condensates. Images were acquired after addition of BI-3802 and light in COS-7 cells expressing EGFP-HSPA1L, CIBN-BTB_IDR-FRB, and CRY2olig-miRFP. Scale, 1 μm. Graphs show fluorescence changes of EGFP-HSPA1L (left) and CIBN-BTBolig-FRB (right) condensates. Data representative of 3 independent experiments. h) BTB/HSP70 condensates in COS-7 cells coexpressing GFP-DNAJB1, mScarlet-HSPA1A, a non-fluorescent (mCherry K70N) CIBN-BTB_IDR-FRB, and CRY2olig-miRFP, treated with BI-3802 and light. Images representative of two independent experiments. Scale, 2 μm.

To test the generality of CoSMo using a different hook and recruited ligand, we turned to the light-induced CRY2-CIBN dimerization system³⁶ and CRY2olig, which with blue light forms rigid condensates and binds CIBN¹¹. We rationalized that CIBN could act as a hook to recruit the oligomeric lit state of CRY2olig into the condensate, inducing condensate maturation with light. We generated mCherry-tagged CIBN-BTB_IDR-FRB (including the FRB domain as this version more easily forms condensates), coexpressed with CRY2olig-mEGFP, and examined condensate formation pre and post BI-3802. Prior to BI-3802 addition, CIBN-BTB_IDR-FRB showed some preclustering, similar to BTB_IDR-FRB (primarily in the nucleus and dependent on expression level) (Supplementary Fig. 4a,b). The coexpressed CRY2olig-mEGFP showed some prerecruitment with the CIBN-BTB_IDR-FRB condensates even without blue light (Supplementary Fig. 4c), likely due to the high protein concentration in the condensate resulting in avidity-mediated CRY2olig/CIBN interaction even in dark. Despite preclustering, condensates prior to light treatment showed fast dynamic exchange (FRAP t_1/2 in dark = 43.3s, 95% recovery after 300s) (Fig. 5c) and circular shapes (Extended Data Fig. 5a), consistent with a liquid-like state.

To initiate CoSMo, we applied blue light to induce CRY2 oligomerization and enhanced CRY2olig/CIBN interaction. Unexpectedly, many of the condensates transformed from monophasic to a biphasic ‘shell’ structure, containing both CIBN-BTB_IDR-FRB (mCherry signal) and CRY2olig-mEGFP (Fig. 5d; Supplementary Video 7). This transition was size-dependent, with smaller condensates (~58%) remaining monophasic (Extended Data Fig. 5b). Both monophasic and biphasic condensates retained their circular shapes after light (Extended Data Fig. 5a), but showed impaired mobility, indicating a shift to a more gel-like state, with biphasic condensates showing more impairment than monophasic (monophasic FRAP t_1/2=43s, 84% recovery after 300s; biphasic FRAP t_1/2=85s, 55% recovery after 400s) (Fig. 5e).

Hollow/biphasic condensates (referred to as anisosomes) were recently reported upon expression of mutant forms of TDP-43, and proposed to be an intermediate step between its liquid-like state and its solid-like state associated with pathologies³⁷. While the TDP-43 anisosomes initially appeared hollow, proteomic studies revealed the presence of chaperones, including the Hsp70s HSPA1L and HSPA1A, enriched in the inner core. Using EGFP-tagged Hsp70 reporters, we found that CIBN-BTB_IDR-FRB condensates were likewise enriched in the core with EGFP-HSPA1L and mEGFP-HSPA1A (Fig. 5f). Hsp70-associated chaperones EGFP-DNAJB1 and EGFP-Hsp90β were also enriched, but not a EGFP control (Extended Data Fig. 5c). EGFP-HSPA1L in the inner phase was dynamic with high mobility (FRAP t_1/2 = 60s; 100% recovery after 400s) (Extended Data Fig. 5d). Upon encountering a second biphasic condensate, both inner and outer layers merged and coalesced, reforming a single biphasic structure (Extended Data Fig. 5e), supporting that the condensates retained some liquid-like properties, despite the increased gelation of the outer layer with light.

Interestingly, in contrast to the TDP-43 anisosomes³⁷, the interaction of BTB/CRY2olig shell and core chaperones was highly dynamic (Fig. 5g; Supplementary Video 8). Over tens of seconds to minutes, chaperones accumulated and grew within the inner core, while the BTB/CRY2olig shell rearranged in response. As there was no change in the total amount of BTB/CRY2 proteins during this time, the outer layer thinned. High-speed imaging (5 frames/s) revealed the appearance of a breach in the thinned outer layer, followed by rapid expulsion of the inner core chaperones and collapse of the BTB/CRY2olig shell into the prior excluded area in the condensate center (Fig. 5h; Supplementary Fig. 5; Supplementary Video 9). The cycle then repeated with new recruitment of chaperone into the core (as in Fig. 5g between 260 and 400s). Biphasic condensates could be sustained for over 2 h with light treatment, and remained dynamic with no aggresomal localization after 24h of light (Extended Data Fig. 6).

We tested whether light removal would induce dissociation of CRY2olig and reversion of biphasic BTB/CRY2olig condensates to a monophasic state (Extended Data Fig. 7). We treated cells with BI-3802 and blue light to form biphasic condensates enriched with EGFP-HSPA1L, then examined loss of EGFP-HSPA1L or CRY2olig-miRFP during blue light removal. With blue light removal, ~75% of punctate CRY2olig redistributed to the cytosol over 15 min, a time frame matching the CRY2 dark reversion rate³⁶, however BTB condensates remained biphasic and enriched with Hsp70, showing no loss of Hsp70 signal in biphasic structures (Extended Data Fig. 7c–e). These results suggest a difference in the requirements for maintenance of biphasic condensates, compared with the requirements for their formation.

Biphasic condensates generated by BTB_IDR-FRB and rapamycin

Our discovery of biphasic condensates using light-induced CoSMo led us to examine whether newly-formed BTBolig condensates also recruit Hsp70 chaperones. We were unable to observe enrichment or recruitment of HSPA1L during initial formation of BTB_IDR, BTB2x-mCh, BTB-EGFP, or CRY2olig (Extended Data Fig. 8). We also revisited the rapamycin-induced CoSMo platform, rationalizing that biphasic condensates could easily have been missed in initial studies if the condensates were small or rare. We coexpressed an EGFP-HSPA1L reporter with BTB_IDR-FRB and miRFP-FKBP5x, added BI-3802 then rapamycin, then monitored condensates enriched with EGFP-HSPA1L. At <1 h after BI-3802 addition and prior to rapamycin addition, all BTB_IDR-FRB condensates were monophasic (Fig. 6a). Addition of rapamycin at this time resulted in only a very small fraction of condensates (0.6%) that transitioned from monophasic to biphasic, enriched with EGFP-HSPA1L (Supplementary Fig. 6).

Figure 6. Biphasic condensates formed with BTB_IDR-FRB and rapamycin-induced CoSMo.

Panels (a-j), COS-7 cells transfected with BTB_IDR-FRB (mCherry) and miRFP-FKBP5x and in indicated cases, EGFP-HSP1AL, then treated with 1 μM BI-2802 and/or 333 nM rapamycin as specified. a) % spontaneous biphasic BTB_IDR-FRB condensates at indicated times after BI-3802, mean and s.d., n=3 independent experiments. b) Cells also expressing EGFP-HSPA1L, treated 3.5h with BI-3802. Yellow boxed region also pictured in (d). Scale, 10 μm. c) FRAP analysis of monophasic or biphasic BTB_IDR-FRB condensates treated 3–6 h with BI-3802. Mean and s.e.m., n=10 condensates from one experiment (monophasic) or n=12 condensates from 2 independent experiments (biphasic). Inset, representative photobleach images. Scale, 1 μm. d) Condensate from cell in panel(b), pre/post rapamycin. Scale, 1 μm. e) Circularity of biphasic condensates formed as in (b), pre/post rapamycin. Graph shows box plot (defined in Methods), n=33 (pre) or 30 (post) condensates, 2 independent experiments. **, p=.0005, unpaired two-tailed t-test. f) BTB_IDR-FRB condensate size, 3 h after BI-3802, binned according to their states before/after rapamycin (monophasic ‘mono’, biphasic ‘bi’). Graph shows n=68 condensates from 2 independent experiments. ns, not significant, p=.42; ***, p=.000014 and .000026, unpaired two-tailed t-test. g) Cells (also expressing EGFP-HSPA1L) treated with BI-3802 4 h, pre/post rapamycin. Scale, 10 μm. h) FRAP analysis of BTB_IDR-FRB (mCh) or EGFP-HSPA1L condensates in cells incubated 3–6 h with BI-3802, 30 min after rapamycin. Mean and s.e.m., n=12 condensates (BTB_IDR-FRB) or n=10 (HSPA1L) from one experiment. Inset, representative photobleach images. Scale, 1 μm. i) BTB_IDR-FRB condensates, cells treated as in (g). Scale, 2 μm. Graph below quantifies Hsp70 fluorescence. j) BTB_IDR-FRB condensates in cells also expressing EGFP-HSPA1L, treated with BI-3802 for 3.5 h, pre and post rapamycin (added at 25 min). Scale, 2 μm. Graph at right, Hsp70 and BTB_IDR-FRB signal in condensates. k) Condensates from cells expressing CIBN-BTB_IDR-FRB, EGFP-HSPA1L, CRY2olig-miRFP, and miRFP-FKBP5x, treated 3h with BI-3802 and light (465 nm, 2s every 2 min). Rapamycin added at t=0. Scale, 2 μm. l) FRAP analysis of CIBN-BTB_IDR-FRB condensates treated as in (k), before or 30 min after rapamycin. Mean and s.e.m., n=6 condensates (pre rap), n=8 (post rap) from one experiment.

When we repeated these experiments using 3–6 h aged BTB_IDR-FRB condensates, many condensates had spontaneously transitioned (without rapamycin addition) from monophasic to biphasic, enriched with Hsp70 (Fig. 6a,b). At 3.5 h post BI-3802 and pre rapamycin, both monophasic and biphasic condensates showed reduced photobleach recovery, indicating a condensate maturation process had occurred (Fig. 6c). For the condensates that had already transitioned to a biphasic state, addition of rapamycin and recruitment of miRFP-FKBP5x led to further recruitment of Hsp70 and loss of circularity (Fig 6d,e). Rapamycin also caused additional monophasic condensates to transition to a biphasic state, recruiting Hsp70 in the core (Fig. 6f,g; Supplementary Video 10). 30 min after rapamycin, the outer BTB shell was immobile, while Hsp70 in the core exchanged freely (FRAP t_1/2 = 63s; 100% recovery) (Fig. 6h). In contrast to the light-induced CoSMo, the immobile BTB shell did not pop or release the Hsp70, which accumulated to a plateau within ~20 min (Fig. 6i). At 2 h after rapamycin, the biphasic condensates remained localized in the cytosolic periphery despite their rigid shells and loss of shell circularity (Supplementary Fig. 7).

The differences in the dynamic ‘popping and release’ behavior with Hsp70 seen with the light-induced biphasic condensates, but not the rapamycin-induced biphasic condensates, suggested a correlation with their material state. Supporting this idea, we observed some spontaneous biphasic BTB_IDR-FRB condensates that showed dynamic Hsp70 popping and releasing behavior prior to rapamycin addition (akin to what we had observed with the light-induced CoSMo), however after rapamycin the BTB shell became static and the Hsp70 releasing behavior ceased (Fig. 6j). To explore more directly whether a change in material state impacts popping/release behavior, we formed CIBN-BTB_IDR-FRB condensates and sequentially induced recruitment of CRY2olig-miRFP using light, then miRFP-5xFKBP using rapamycin. After CRY2olig recruitment, condensates showed popping/release behavior with Hsp70 and some mobility in FRAP experiments (FRAP t_1/2 = 71s; 65% recovery after 195s), however addition of rapamycin immobilized the BTB shell and ended the popping behavior (Fig. 6k,l). Together, this work supports a model in which a liquid-like or semi-liquid condensate can engage dynamically with chaperones, cycling between monophasic and biphasic states, while rapamycin-induced gelation fixes the shell and blocks monophasic-biphasic cycling (Extended Data Fig. 9).

Discussion

We describe the development of a new platform, BTBolig, that enables chemical-inducible control of condensate formation and dissolution. Advantages include the requirement for only one scaffold protein, the rapid kinetics of condensate formation and dissolution using cell-permeable ligands, and the applicability of the method to multiple rounds of cycling. The system is also modular, allowing generation of condensates with different physical properties using BTB2x-mCh, BTB-EGFP, BTB_IDR, or BTB_IDR-FRB. Each of these systems presents different properties. For example, BTB_IDR condensates remain liquid-like for long durations, while initially liquid-like BTB_IDR-FRB condensates transition over time to gel-like states. Remarkably, we could observe this transition even at early times after condensate formation (as in Figure 2e). We do not know how the FRB domain contributes to condensate gelation—one possibility is that it acts through avidity to recruit endogenous protein ligands. Similar to the process we observe with CoSMo, these ligands may crosslink the scaffold and result in progressive condensate maturation.

Another interesting difference observed with the various BTBolig systems is the difference in cell processing and clearance rates. BTB-EGFP is quickly cleared to the aggresome, much faster than BTB2x-mCh despite both forming more gel- or solid-like condensates—a finding that suggests the elongated fibril-like structure may be important for this process. In turn, the liquid-like BTB_IDR condensates are cleared slowly and do not show aggresomal accumulation. Due to its rapid clearance, BTB-EGFP may not be ideal for controlling oligomeric protein activity as previously proposed¹⁹—users may prefer to substitute the BTB2x-mCh module which is equally robust, rapid, and reversible but does not as vigorously engage proteostasis machinery.

A unique feature of the BTBolig system is the ability to dissolve protein from condensates at different times with BI-3812. Remarkably, even compact protein localized to the aggresome or dense fibrils could be rapidly dissolved. BTB-EGFP and BTB2x-mCh will be useful for studying the impact of releasing protein from fibrils or the aggresome at specific times. Surprisingly, while BTB_IDR dissolved rapidly and completely when BI-3812 was added at short times after condensate formation, after extended (24 h) incubations with BI-3802 a small fraction of the condensate remained refractory to dissolution by BI-3812, suggesting the remaining protein may be uniquely compartmentalized or modified.

We report CoSMo as a chemical- or light-dependent approach that allows the controlled transition of condensates from a liquid-like to a solid-like state. We used this tool to observe how changes in condensate material state affect interactions with chaperone proteins, observing recruitment of Hsp70s and other chaperones into the condensate core and formation of biphasic condensates. These studies support a growing understanding of an intricate relationship between condensates and protein chaperones, for example in recent studies implicating a role for Hsp70 chaperones in preventing hardening of TDP-43 anisosomes³⁷, and showing a role for small heat shock proteins in preventing hardening of FUS droplets³⁸. Our studies suggest that biphasic structures form when condensates are undergoing maturation, transitioning between liquid-like and more gel-like states. Supporting this idea, we did not observe biphasic condensates or recruitment of Hsp70s upon initial formation of rigid (BTB-EGFP or BTB2x-mCh) or more liquid-like (BTB_IDR) condensates. Rather, chaperones were recruited and biphasic shell-core structures formed during the increased gelation provided by light-induced CoSMo, or with aged BTB_IDR-FRB condensates transitioning to gel-like states spontaneously or during rap-induced CoSMo. Our study also showed a correlation between large condensate size and biphasic structure, as observed with both light- and rap-induced CoSMo systems (Extended Data Fig. 5b, Fig. 6f). This size dependence may explain a discrepancy: only aged condensates could form biphasic structures with the rapamycin-induced CoSMo approach, but newly-formed condensates could form biphasic structures with light-induced CoSMo. Indeed, light-induced CoSMo led to notably larger condensates at earlier stages after BI-3802 addition (Supplementary Fig. 8), possibly due to the large amount of preclustering in the system. While the precise mechanism of chaperone recruitment and biphasic condensate formation will require further study, our experiments suggest that both condensate size and material state (at an early stage of transitioning from liquid-like to more gel-like) are important factors.

Unique to our study, we characterize dynamic interactions of biphasic BTB condensates with chaperones that is dependent on the rigidity of the condensate outer shell. Both the light- and rapamycin-induced CoSMo approaches trigger recruitment of Hsp70 and formation of biphasic condensates. However, because the light-induced condensate shell is gel-like, it can rearrange in response to the protein accumulating in the core, resulting in outer shell thinning and explosive release of the core components. In contrast, the strong crosslinking induced by rapamycin results in a more solid-like outer shell that becomes rapidly fixed, resulting in sustained Hsp70 recruitment.

In summary, the methodologies described here complement and add to the growing portfolio of synthetic condensate tools. In its present form, one limitation of BTBolig is its lack of orthogonality in mammalian cells. BCL6 has a role in T-cell differentiation, and the BI-3802 ligand can induce BCL6 degradation³⁹. The BCL6 BTB domain interacts with SMRT and N-CoR corepressor complexes⁴⁰, and use of the system in specific cells may affect this pathway. Further work will be needed to develop an orthogonal drug-protein system to avoid off-target effects with the transcriptional factor BCL6. Despite these limitations, the BTBolig platform and CoSMo offer a versatile set of tools to explore condensates. BTBolig can be combined with other IDRs or condensate-forming domains to trigger condensate formation or dissolution under controlled conditions. Implemented in vivo with disease-relevant proteins, the BTBolig system could enable study of how therapeutic disruption at different times could reverse pathological symptoms. As the dense concentration of protein within condensates can alter protein interactions, the BTBolig tools can be easily combined with proteomic methods^41,42 to interrogate interactions or post-translational modifications in dilute and dense states (pre and post condensate formation). Altogether, the combined use of BTBolig and CoSMo will be useful for investigating how changes in condensate material state impact function, exploring the complex relationships between condensates and cellular proteostasis/chaperone pathways, and studying the role of condensate aging and maturation in disease pathologies.

Methods

Plasmids and chemical reagents

Full construct and oligo sequences are provided in Supplementary Table 1. Cloning was carried out using PCR, restriction enzyme digestion, and ligation with T4 DNA Ligase or Quick Ligase (New England Biolabs), or using Gibson assembly. All PCR reactions used Phusion polymerase (New England Biolabs). Gibson assembly reactions were performed by incubating 200 ng of the insert DNA and 200 ng of the linearized vector at 50°C for 30 min in a homemade Master Mix (0.1 M Tris, 10 mM MgCl₂, 0.2 mM dNTPs, 10 mM DTT, 50 mg/ml PEG-8000, 1 mM NAD, 0.8 U T5 exonuclease, 0.5 U Phusion polymerase, 80 U Taq Ligase (MC Labs)). BI-3802 and BI-3812 were generously provided by Boehringer-Ingelheim and used at final concentrations of 1 μM (BI-3802) and 10 μM (BI-3812), unless otherwise indicated. Rapamycin was from Selleckchem; SYTOX Green was from ThermoFisher. mNeonGreen-Vimentin-N-7 (expressing N-terminal vimentin-mNeonGreen) was from Allele Biosciences.

Mammalian Cell culture

Studies used COS-7 or HEK293T cells, cultured in Dulbecco’s modified Eagle medium (DMEM) (Corning) supplemented with 10% fetal bovine serum (FBS) (Sigma) and 1x Penicillin-Streptomycin (Corning) at 37 ° C with 5% CO₂. Cells were transfected using Lipofectamine 2000 (Invitrogen), according to the manufacturer’s recommendations. For biphasic condensate studies using CRY2olig, samples were wrapped in aluminum foil immediately after transfection and kept in dark until the following day. Subsequent sample manipulations were carried out using a red safelight. For light-induced CoSMo experiments, cells were exposed to 488 nm light through the microscope objective (10% laser power, 488 nm, 7–10×100ms pulses, every 1–3 min) during live cell imaging. For extended light treatments as in Extended Data Fig. 6, cells were exposed to 465 nm light through a custom-built LED array (1.1 mW/cm², 2 s pulses delivered every 2–3 min).

Live cell imaging experiments

For live cell imaging, cells were plated onto 35-mm glass bottom culture dishes, transfected the following day, then imaged 18–22 h after transfection. Where specified, 1 μM BI-3802, 10 μM BI-3812, and/or 333 nM rapamycin was added directly to the samples during live cell imaging. Imaging was carried out in media or HBSS (Hanks Balanced Salt Solution, 1.26 mM CaCl₂, 0.41 mM MgSO₄, .49 mM MgCl₂, 5.33 mM KCL, 138 mM NaCl, 0.44 mM KH₂PO₄, 0.34 mM Na₂HPO₄, 5.56 mM D-glucose, and 20 mM HEPES), at 33.5 °C. Light-sensitive samples were kept from ambient light using a light-tight stage-top enclosure. Cells were imaged using an Andor Dragonfly 301 spinning disc imaging system, with an Olympus IX73 base and four-line ILE laser merge module and controller. Images were acquired using a 60x UplanSApo 1.35 NA oil objective and collected on a 1024 ×1024 pixel Andor iXon EM-CCD camera. Data was acquired using Fusion 2.3 or iQ3.6 software (Oxford Instruments).

For FRAP experiments, cells were imaged at 33.5 °C with a Olympus IX71 microscope with a spinning disc scan head (Yokogawa). A 60x/NA 1.4 oil immersion objective was used. Laser illumination for FRAP experiments (50% laser power with a 1–2 ms dwell time) was delivered using galvometric laser-scanning mirrors (FRAPPA; Andor Technology). Photobleaching pulses were calibrated so that they bleached no more than 40–85% of the original signal. An AOTF controlled laser launch (Andor Technology/Oxford Instruments) was used to deliver excitation illumination. Images were collected on a 1024 ×1024 pixel iXon EM-CCD camera (Andor Technology/Oxford Instruments).

Image analysis

All images were analyzed using ImageJ or Fiji software. For all experiments, measurements and quantification were from distinct samples (no repeat samples). Initial Z-stacks were compiled into a maximal projection image for analysis. To calculate the amount of protein within condensates in each cell, images were automatically thresholded (Otsu’s method) in ImageJ. We calculated the total fluorescence within the thresholded fraction (area*mean background subtracted fluorescence) divided by the total fluorescence (area*mean background-subtracted fluorescence), for nucleus, cytosol, or entire cell. To calculate the condensate formation and dissolution kinetics, the % of protein within each condensate was plotted in relation to time, then fitted to a sigmoid function using IGOR Pro 6. To calculate condensate circularity and size, each condensate was automatically segmented using ImageJ. Circularity was determined using the equation circularity = 4π(area/perimeter²). To calculate inverse capillary velocity (Fig. 2e), we plotted the time (s) for two condensates to fuse on the y-axis and the length (in μm) of their diameter at the time of initial contact. The inverse capillary velocity is determined from the slope of the fitted line (fit using GraphPad software), and is a function of the ratio of viscosity to surface tension. In the second graph (Fig. 2e) we plotted the ratio of time/length for each observed merging event in relation to time after BI-3802 addition.

For FRAP experiments, each photobleached region in the condensate was selected and the background free integrated density at each time point was measured. To account for global photobleaching effects, the integrated density of the bleached region was normalized to a cytosolic region of a neighboring, non-bleached cell. The normalized values were mapped to an [0,1] interval, setting the zero value as the normalized integrated density at the photobleaching time point and the 1 value as the average of the normalized integrated density before photobleaching. To calculate FRAP t_1/2 rates, we fit the normalized FRAP recovery data to an exponential function y=A(1-e^−(K*x)) and t_1/2 = ln(2)/K, using IGOR Pro or GraphPad software. While half-FRAP can also provide information about biomolecular condensate material state⁴⁸ we did not perform these experiments due to the small size of many condensates and difficulties tracking their orientation in the 3D volume of the cell.

Cell death studies

To quantify cell death in cells expressing FLAG-MLKL(1–140)-BTB2x-mCh, HEK293T cells were incubated with HBSS containing 400 nM SYTOX Green for 20 min in dark, then washed 1x in HBSS, then placed in imaging buffer and treated with 1μM BI-3802. After BI-3802 addition, cells were tracked for 30–40 minutes and the number of cells stained with SYTOX Green normalized to total transfected cells was quantified. The kinetics of the drug-induced recruitment of MLKL to the plasma membrane was quantified by tracking the background-free intensity of mCh at the cytosol. The half-mean time was obtained from fitting a sigmoid function to the kinetics of each cell.

Graphs, statistical tests, and reproducibility

Microsoft Excel, IGOR Pro, and GraphPad Prism 8 software were used to generate graphs for figures. All box plots in figures show all data points, with the median as a black line, box extending between 25^th and 75^th percentiles, and whiskers indicating min and max data points. Statistical tests used an unpaired two-tailed t-test; a p-value of <0.05 was considered significant. All experiments cited as independent experiments in legends were biologically independent experiments. Representative images for experiments were from biologically independent experiments as follows: two independent experiments: 3b, 3d, 3f, 4h, 5b, 5h, 6i, 6j, 6k; three independent experiments: 1c, 1d, 3a, 3c, 4b, 4e, 6g; five independent experiments: 2d, 5d, 5f, 6b, 6d.

Extended Data

Extended Data Figure 1. Preclustering of BTB_IDR-FRB prior to BI-3802 addition.

a) Quantification of % of BTB_IDR-FRB in clusters in COS-7 cells pre and post 1 μm BI-3802 in nucleus and cytosol. Pre-clustering is predominantly nuclear. Mean and s.e.m., n=34 cells from 5 independent experiments.

b) Quantification of data from (a) from cells undergoing nuclear preclustering. A linear relationship between relative expression level (mean intensity/background) and nuclear preclustering (% of protein in cluster) is observed.

Extended Data Figure 2. BTB_IDR-FRB and BTB-EGFP can undergo multiple rounds of condensate induction and reversal without loss of efficacy.

Shown are representative images (top) and quantification of condensate fluorescence (bottom). (a) COS-7 cells expressing BTB_IDR-FRB and EGFP (as a cell-fill) were treated with 1 μM BI-3802 (filled arrowheads) for 10–15 min then 10 μM BI-3812 (open arrowheads) for 7–15 min. Between each cycle, the media was changed (wash). Data is representative of 2 independent experiments. (b) COS-7 cells expressing BTB-EGFP were treated with 1 μM BI-3802 for 10–20 min (filled arrowheads), then incubated with 10 μM BI-3812 for 15–20 min (open arrowheads). Samples were washed in PBS between each treatment cycle. Data is representative of 2 independent experiments. Scale bars, 10 μm.

Extended Data Figure 3. FRAP of BTB_IDR-FRB and BTB_IDR at later timepoints reveals a condensate maturation process for BTB_IDR-FRB.

a) Normalized FRAP analysis of BTB_IDR-FRB condensates (in COS-7 cells incubated with BI-3802 for 1–1.5 h). Graph shows mean and s.e.m., n=26 condensates, from 2 independent experiments. Inset shows representative condensates pre and post photobleach. Scale, 1 μm. b) Normalized FRAP analysis of BTB_IDR-FRB condensates in COS-7 cells incubated with BI-3802 for 3–6 h. Graph shows mean and s.e.m., n=12 condensates from a single experiment. Inset shows representative condensate pre and post photobleach. Scale, 1 μm. c) Normalized FRAP analysis of BTB_IDR condensates 4 h post BI-3802. Graph shows mean and s.e.m., n=11 cells from 2 independent experiments. Inset shows representative condensate pre and post photobleach. Scale, 1 μm. d) BTB_IDR condensates in COS-7 cell treated with BI-3802 for 19 h (added 24 h post transfection). Representative of 3 independent experiments. Scale, 10 μm. e) Normalized FRAP analysis of BTB_IDR condensates 19 h post BI-3802. Graph shows mean and s.e.m., n=10 cells from a single experiment. Inset, representative condensates pre and post photobleach. Scale, 1 μm.

Extended Data Figure 4. Rapamycin-induced CoSMo.

a) Schematic of CoSMo approach with rapamycin (rap). b) COS-7 cells expressing BTB_IDR-FRB and mCh-FKBP2x were treated with 1 μM BI-3802 for <30 min. Condensate circularity was quantified pre or post 333 nM rapamycin. Graph shows box plot (defined in Methods), n=140 (-rap) or 325 (+rap) condensates from one experiment. **, p<.001, unpaired two-tailed t-test. c) FRAP analysis of BTB_IDR-FRB condensates treated as in (b), pre or 10 min post rapamycin. Mean and s.e.m., n=7 (pre) or 10 (post) condensates from 2 independent experiments. d, e) COS-7 cells expressing BTB_IDR-FRB and mEGFP-FKBP2x treated with 1 μM BI-3802 for <30 min, before or 10 min post 333 nM rapamycin. Representative images in (d) (different cells, with higher magnification images at right), representative of 2 independent experiments. Scale, 10 μm. FRAP analysis in (e), mean and s.e.m., n=12 (pre) or n=13 condensates (post) from 2 independent experiments. f) FRAP analysis of condensates in cells expressing BTB_IDR-FRB and mEGFP-FKBP2x, 30–60 min after 1 μM BI-3802 and 15–60 min after adding 10, 60, or 333 nM rapamycin. Mean and s.e.m., n=10 (10 nM), 25 (60 nM), or 13 (333 nM) condensates from 2 independent experiments. g) COS-7 cells expressing BTB_IDR-FRB and miRFP-FKBP5x <30 min after 1 μM BI-3802, before and 10 min after 333 nM rapamycin. Images representative of 3 independent experiments. Scale, 10 μm. Higher magnification region at bottom. h) Condensate circularity with miRFP-FKBP5x recruitment as in (g), pre or 10 min post rapamycin. Data shows box plot (defined in Methods), n=394 (pre) or 359 (post) condensates from 2 experiments. ****, p=6.6×10–13, two-tailed unpaired t-test.

Extended Data Figure 5. Additional quantification of biphasic condensates induced by light using CoSMo approach.

a) Circularity of CIBN-BTB_IDR-FRB condensates coexpressed with CRY2olig-mEGFP, 30 min post BI-3802, before and after light (488 nm, 5% laser power, 7×50ms pulses, every 30s). Post light separated into monophasic or biphasic. Box plot (defined in Methods), n=271 (pre-light), 103 (post, monophasic) or 62 (post, biphasic) condensates from 2 independent experiments. ns, not significant (p=.21 and .12), unpaired two-tailed t-test. b) Quantification of size of CIBN-BTB_IDR-FRB condensates remaining monophasic or transitioning to biphasic, for COS-7 cells also expressing CRY2olig-mEGFP, treated with 1 μM BI-3802 then light. Size quantified before light treatment. Data represents individual points and median, n=59 condensates from 2 independent experiments. **, p=.000004, unpaired two-tailed t-test. c) COS-7 cells expressing CIBN-BTB_IDR-FRB, CRY2olig-miRFP, and GFP-tagged HSP90α, DNAJB1, or EGFP alone, treated with BI-3802 and light to induce biphasic condensates. Yellow boxes denote a higher magnification image of the boxed region. Scale, 10 μm. Images representative of 2 independent experiments. d) Normalized FRAP analysis of EGFP-HSPA1L within the core of biphasic CIBN-BTB_IDR-FRB + CRY2olig-miRFP condensates (treated with BI-3802 and blue light). Mean and s.e.m., n=5 condensates from one experiment. e) Condensates undergoing fusion in cells expressing CIBN-BTB_IDR-FRB (labeled as ‘BTB’), CRY2olig-miRFP, and EGFP-HSPA1L (Hsp70), treated with 1 μm BI-3802 and exposed to blue light. Scale, 2 μm. Images representative of 3 independent experiments.

Extended Data Figure 6. Tracking biphasic condensates generated through light-induced CoSMo with extended light treatment.

a-c) COS-7 cells expressing CIBN-BTB_IDR-FRB, CRY2olig-miRFP, and HSPA1L-EGFP were treated with BI-3802 then blue light (465 nm, 1.1 mW/cm2, 2s pulse every 2 min) for 2 h. Images representative of 3 independent experiments. a) Representative image of cell after 2 h blue light. Condensates remain biphasic and circular and are distributed throughout the cell periphery. Scale, 10 μm. b) Timelapse of 2 h light-treated condensates that continue to merge and coalesce suggestive of a liquid-like state. Scale, 1 μm. c) Timelapse of 2 h light-treated condensates showing popping/release behavior with Hsp70. Scale, 1 μm. d-f) 24 h light treatment. COS-7 cells expressing CIBN-BTB_IDR-FRB (mCh), CRY2oligmiRFP, and HSPA1L-EGFP were treated with 1 μM BI-3802 then blue light for 24 h (465 nm, 1.1 mW/cm2, 2s pulse every 2 min). Images representative of 2 independent experiments. d) Cells show no aggresomal accumulation of CIBN-BTB_IDR-FRB after 24 h light. Scale, 10 μm. e) Timelapse of 24 h light-treated biphasic condensates merging and coalescing. Scale, 1 μm. f) Timelapse of biphasic condensate showing popping and releasing behavior with HSPA1L-EGFP after 24 h of light. Scale, 1 μm.

Extended Data Figure 7. Removal of light after light-induced CoSMo does not result in transition of condensates to monophasic state.

a) COS-7 cells expressing HSPA1L-EGFP, CRY2olig-miRFP, and CIBN-BTB_IDR-FRB (with a K70N non-fluorescent mCherry, to allow quantification of CRY2olig-miRFP signal without overlap with mCherry) were incubated 40 min with 1 μM BI-3802, then 488 nm light (10% laser power, 12×100 ms pulses, every 2 min) for 30 min, followed by 60 min with no 488 nm light to allow dissociation of CRY2olig, then a second 488 nm light treatment. Graphs show quantification of normalized EGFP and CRY2olig signal in clusters during experiment. Data normalized to the amount at the end of the initial light period, set to 1. Graphs show mean and s.e.m, n=5 cells (EGFP) or 4 cells (CRY2olig), of a single experiment, independently performed 2 times with similar results. b) Individual cell data for experiment in (a) showing % of clustered CRY2olig-miRFP at each timepoint. n=4 cells. c) A parallel experiment with identical design as in (a), but using wildtype mCherry to allow visualization of CIBN-BTB_IDR-FRB signal. BTB_IDR-FRB signal in condensates remains constant. Data normalized to the amount at the end of the initial light period, set to 1. Graph shows mean and s.e.m., n=5 cells, of a single experiment, independently performed 2 times with similar results. d,e) Images of BTB_IDR-FRB (mCh) and HSPA1L-EGFP (d) or CRY2olig-miRFP and HSPA1L-EGFP (coexpressed with K70N CIBN-BTB_IDR-FRB) (e) signal in condensates 30 min after initial 488 nm light, or after 60 min without 488 nm light. Condensates do not revert to monophasic. Scale, 2 μm. Representative of 2 independent repeats.

Extended Data Figure 8. HSPA1L does not colocalize with initially formed rigid or liquid-like condensates.

Shown are representative images of COS-7 cells expressing EGFP-HSPA1L (or mScarlet-HSPA1L) and either BTB2x-mCh, CRY2olig-mCh, BTB_IDR, or BTB-EGFP, 5 min to 2 h post inducing condensates (1 μM BI-3802 or light, as appropriate). Scale, 10 μm. Data representative of 2 independent experiments.

Extended Data Figure 9. Model of biphasic condensate behavior.

Light-induced CoSMo results in condensates that retain liquid-like behavior, allowing dynamic interaction with chaperones. Rapamycin-induced CoSMo results in initial recruitment of Hsp70 and rearrangement to a shell/core structure. Subsequent gelation locks the shell structure in place but allows additional Hsp70 recruitment.

Data Availability Statement.

All data analyzed during this study are included in the article and supplementary information. Source data is included with this publication. Imaging data is provided as supplementary videos associated with this publication; original TIFF formats of imaging data (not deposited in source data due to the large file sizes) is available from the corresponding author on request.