Reversible manipulation of liquid-liquid phase separation using a chemical-based chemiDroplet

Liquid-liquid phase separation (LLPS) emerges as a general approach for the generation of biomolecular condensates, which play important roles in various biological processes, including chromatin organization, transcription, translation, protein degradation, and signal transduction. Irregular phase separation is often associated with cancer, neurodegeneration, and other human diseases. Pioneering studies demonstrate that proteins tending to undergo LLPS usually contain tandem structured binding domains or intrinsically disordered regions (IDRs), which establishes the crucial role of multivalent interactions in condensate formation. However, the majority of the studies primarily rely on in vitro reconstitution and cellular fluorescent imaging. Given the highly dynamic nature of protein condensates, precise control of their phase separation behavior is essential to elucidate the function of LLPS in vivo.

Tools allowing the generation and dissolution of condensates at user-defined time points are helpful for researchers to understand their dynamic characteristics and functions. Recently, light-induced tools, including optoDroplet and Corelets, have been developed to trigger droplet assembly in living cells by fusing a light-responsive peptide to various full-length or domains of condensate-forming proteins, including FUS, DDX4, hnRNPA1, and G3BP1{Bracha, 2019 #18}. This method helps comprehend the contribution of multivalent domains and IDRs in LLPS and can be exploited to control protein function, for example, to regulate gene transcription, signal transduction, and stress granule formation. However, the light-induced tool has limitations in prolonged modulation of phase separation to study the cellular function because of the side effects of light stimulation, such as reactive oxygen species (ROS) production and phototoxicity. A chemical-based tool utilizing chemical compounds provides a different approach to tuning condensates for probing the function of phase separation in living cells. However, the identified chemical compounds triggering phase separation suffer from irreversibility and poor mobility. Therefore, it is urgent to develop novel tools to manipulate the intracellular phase separation dynamically.

The authors sought to develop a chemical tool that can be used to manipulate LLPS dynamically and reversibly within the cell. The ideal compound should: 1) efficiently induce protein oligomers; 2) maintain the liquid fluidity of the puncta; 3) be reversed by other compounds to dissolve the condensates; 4) be nontoxic to cells for use in the study of biological functions in cells. Previous studies demonstrated that the compound BI-3802 can bind to the groove structure between two BTB domains of BCL6 dimers, directly in contact with Tyr58 of BTB and induce self-oligomerization of BTB Importantly, the induced oligomers can be competitively reversed by a compound named BI-3812. Inspired by these findings, the authors determined that a BTB-multivalent domain or BTB-IDR fusion protein could confer a chemical dependence on its multivalent interactions, and the compound pairs (BI-3802 and BI-3812) can readily modulate the LLPS of these fused proteins for the study of their functions in vivo.

To this end, the authors first created a fusion construct by adding the BTB-containing domain of BCL6 (BCL6 1-250aa) to the DNA binding domain (DBD, 7-110aa) of IRF3, as DBD is a well-studied multivalent domain required for IRF3 condensate assembly. They expressed this GFP-labelled BCL6 (1-250aa)-DBD fusion protein (known as chemiIRF3_DBD) in HeLa cells and treated the cells with BI-3802. Live cell imaging indicated that BCL6 (1-250aa) alone did not generate puncta within a 10-minute treatment, although it formed clusters upon prolonged treatment. In contrast, chemiIRF3_DBD rapidly formed condensates under BI-3802 treatment. Similar findings were also observed when BCL6 (1-250aa) was fused to the C-terminal (186-320aa) IDR of hnRNPA1 (chemihnRNPA1_C) or C-terminal (266-504aa) IDR of YAP (chemiYAP_C). To verify the condensates are specifically mediated by IDRs and the BCL6 (1-250aa), the authors examined the N-terminal (1-164aa) domain of YAP that does not contain an IDR and found that this fusion protein did not generate puncta, consistent with the previous report that this region of YAP is not required for condensates formation. Additionally, the authors examined chemiIRF3_DBD harboring a BCL6 (1-250aa) R28A mutation that disrupts the oligomerization of the BTB domain and found that it failed to form condensates upon BI-3802 treatment. The authors also found that the formation of puncta of chemiIRF3_DBD occurs in a drug dose- and protein concentration-dependent manner, and the threshold concentration of the protein was estimated to be 1 μmol/L. Collectively, these data suggest that the compound BI-3802 can efficiently trigger the assembly of the BTB domain fusion proteins into condensates.

Next, the authors investigated the biophysical properties of chemical-induced condensates. Following fluorescence recovery after photobleaching (FRAP) of chemiIRF3_DBD clusters induced by BI-3802, fluorescence recovery was observed on a rapid time scale. Additionally, time-lapse microscopy demonstrated that the chemiIRF3_DBD condensates readily fused into larger structures over time, suggesting that the condensates are highly active, with rapid diffusion of molecules across condensates and the surrounding contents. These data demonstrate that BI-3802-induced condensates are liquid-like phase-separated structures.

To investigate the reversibility of the BI-3802-induced phase separation, we treated cells expressing chemiIRF3_DBD with BI-3802 for different durations and removed the compound after the condensate’s induction. Fluorescence microscopy demonstrated that chemiIRF3_DBD clusters persisted for 4 hours, suggesting the clusters can be maintained for a long time following the removal of BI-3802 in the cell culture medium. Then, the authors used another compound, BI-3812, which competes for the same binding site on the BTB domain, thus inhibiting oligomerization. After the induction of chemiIRF3_DBD condensates by BI-3802 in HeLa cells, BI-3812 treatment triggered the disassembly of chemiIRF3_DBD droplets within 10 minutes. Similar results were observed using chemiYAP_C. 1,6-Hexanediol (1,6-hex) is a common compound that putatively disrupts weak hydrophobic interactions. To further confirm the role of multivalent interactions for chemiIRF3_DBD phase separation, we treated HeLa cells with 1,6-hex or alongside BI-3812, and found that the addition of BI-3812 and 1,6-hex together but not 1,6-hex alone rapidly dissolved the condensates in 30 seconds. This suggests that BI-3802 induced oligomerization of BCL6 (1-250aa) and, to a lesser extent, the weak hydrophobic interactions are responsible for chemiIRF3_DBD condensate assembly. Importantly, the assembly was repeatable even after sequential activation and inactivation cycles. Collectively, these data suggest that BI-3802-induced phase separation of the fusion proteins is fully reversible by BI-3812.

Phase separation has been shown to be important for transcriptional control. To examine whether chemical-induced LLPS would regulate gene transcription, the authors focused on IRF7 and IRF3, two transcription factors that undergo phase separation and activate gene transcription upon IFN-b signaling upon virus infection. BCL6 (1-250aa) was fused to IRF7 (chemiIRF7), IRF3 (chemiIRF3), and NLS-IRF3 (chemiNLS-IRF3), respectively. The compound BI-3802 induced their phase separation either in the cytosol (chemiIRF3 and chemiIRF7) or in the nucleus (chemiNLS-IRF3 and chemiIRF7) in the absence of virus infection, suggesting LLPS of these fusion proteins can be precisely modulated by the compound. We treated B16F10 cells with BI-3802 for 30 minutes, allowing chemiIRF7 condensate formation, and performed Quantitative RT-PCR (RT-qPCR) to detect the mRNA level of Ifnb after 8 hours. RT-qPCR analysis demonstrated that BI-3802 significantly activated Ifnb expression independent of upstream INF signaling, which was reversed by BI-3812. Similar results were also observed in other fused proteins, including chemiNLS-IRF3 and chemiYAP (BCL6 (1-250aa) fused with YAP). Furthermore, we compared the chemiDroplet and optoDroplet systems for transcriptional activation [6]. We stimulated the cells expressing CRY2-mCherry-IRF7 (optoIRF7) with blue light or treated cells expressing chemiIRF7 with BI-3802 for 30 minutes and examined the mRNA levels of IFNB after 8 hours. The results indicated that chemiIRF7 could efficiently activate IFNB expression, while optoIRF7 did not. This suggests that chemiDroplet has a superiority over optoDroplet in inducing transcriptional activation over a long period. Importantly, fusion with the BTB-containing domain of BCL6 did not impact the intracellular localization of proteins. Both BI-3802 and BI-3812 had no obvious impact on cell viability during the treatment. Together, these data indicate that our chemical-based system can efficiently modulate protein phase separation and gene transcription.

In summary, this study has established a generalized chemical-based system, “chemiDroplet,” by pairing compounds and inducible oligomers to tune phase separation at specific times within live cells. ChemiDroplet maintains liquid mobility of the phase-separated condensates and enables the reversible control of protein LLPS. This system can be readily implemented to examine the contribution of the phase separation domain in LLPS and shed light on the biological functions of LLPS of key proteins, such as IRF7, IRF3, and YAP, in transcriptional regulation. ChemiDroplet will also provide a strategy for comprehending condensate-related physiology and pathology in model organisms. 

---

---