Poster V7 - Fluorescent protein-dominated shifts in the intracellular liquid-liquid phase separation dynamics of disordered proteins

Introduction: Intrinsically-disordered proteins (IDPs) recently emerged as major drivers of cell physiology. Exploiting their conformational dynamics, IDPs and IDP domains across eukaryotic proteomes fuel the self-assembly of intracellular and extracellular structures. Intracellularly, select IDPs undergo liquid-liquid phase separation (LLPS), reversibly de-mixing into two liquid phases: an IDP-rich biomolecular condensate and a surrounding dilute aqueous phase. Condensate-forming IDPs function as scaffolds that recruit client biomolecules to orchestrate cellular processes. In disease, liquid-like condensates are thought to age through liquid-to-solid transitions, potentially explaining the solid-like IDP aggregates that are pathological hallmarks of human neurodegeneration. The live-cell probing of IDPs and their LLPS dynamics, however, remains challenging. In particular and unlike folded proteins, the conformational dynamics of IDPs are sensitive to molecular-level fusion to protein domains, raising concerns about the use of fluorescent proteins (FPs) in current live-cell imaging approaches. To enable future engineering of IDP-driven condensates and rigorous dissection of endogenous LLPS-driven processes, here we report for the first time a systemic study of how FPs influence the intracellular LLPS of IDPs. By examining a library of engineered IDPs and a neurodegeneration-relevant IDP across bacteria and human cells, we demonstrate that the intracellular LLPS of diverse IDPs is severely disrupted by molecular-level fusions to monomeric FPs. To minimize these FP-dominated shifts, and departing from molecular-level (in cis) tagging, we advance a generalizable client-based approach for in trans delivery of FPs to IDPs and their assemblies. Our findings and tools will accelerate high-fidelity engineering and live-cell probing of intracellular IDPs and biomolecular condensates.

Materials and

Methods: We built a library of genes encoding fusions of three representative FPs to a diverse set of engineered IDP polymers (IDPPs). We focused on FPs that are commonly treated as monomeric, but varying in their dimerization propensity from perfect monomers to weak dimers (wDimer; at high concentration). To enable comparisons across FPs, we built and characterized all possible FP-linker-FP proteins, obtaining calibrated intensity ratios. We chose IDPPs engineered at the repeat level to match the size of human IDPs (Fig. 1A) and to exhibit the two main classes of LLPS: distinguished by whether phase separation occurs above a lower critical solution temperature (LCST) or below an upper critical solution temperature (UCST). By changing repeat number, we modulated the underlying LLPS behavior and varied IDP size relative to FP domains (Fig. 1B). As a neurodegeneration-relevant IDP, we built a gene encoding TDP43-∆NLS1 —with a defective (d) nuclear localization sequence (NLS1). We expressed FP-IDPPs and FP-TDP43 in bacteria (Fig. 1C) and human cells (Fig. 2A), and purified recombinant IDPPs and FP-IDPPs for basic in vitro studies. To enable live-cell intracellular probing of IDP-scaffolds without FP fusions, we genetically coupled the (equimolar) synthesis of IDPPs modified with a 7-residue peptide (ENLYFQS; cTEV) and a fluorescently-tagged client. As clients, we created FP-fusions to a protease-dead variant of the tobacco etch virus protease (dTEVP), which binds with micromolar affinity and high specificity to cTEV peptides (Fig. 2C). Across bacteria and human cells, we assessed standard intracellular LLPS metrics via live-cell imaging using spinning-disk confocal microscopy.

Results, Conclusions, and Discussions: Human IDP regions (IDRs) and phase-separating IDPs predominantly sample sizes that are comparable to FP domains (~300 residues) (Fig. 1A). Our library of LCST- and UCST-exhibiting IDPs models relevant FP-IDP protein architectures (Fig. 1B). In bacteria, fusions of (LCST1)80 (400 residues) to either sfGFP or a fully monomeric variant (msfGFP) led to intracellular condensates, but msfGFP induced a 2-fold decrease in the average dense phase concentration (Fig. 1D-E). Fusions to mRFP1 proved the most disruptive to LCST1 IDPPs (80-160 repeats), suggesting that additional FP surface properties modulate IDP phase behavior. These differences were already apparent qualitatively. LLPS leads to spherical condensates that minimize surface tension. The weakly-dimerizing sfGFP sustained the expected spherical morphology of (LCST1)80 condensates, whereas msfGFP-(LCST1)80 formed small condensates with varied smooth shapes. mRFP1-(LCST1)80 phases polarized to cell poles without clear condensate boundaries (Fig. 1D). We observed similar results for bacterial condensates formed by (UCST1)40 (320 residues) with a 4-fold shift in partition coefficients (Fig. 1F), except that UCST1 IDPPs tolerated fusion to mRFP1 —giving rise to well-defined condensates. For TDP43∆NLS1 (414 residues), we quantified a 2-fold drop in partition coefficient for mRFP1 fusions (Fig. 1G-H). Turning to purified IDPPs, we confirmed that sfGFP-(LCST1)160 matched the LLPS behavior of untagged (LCST1)160, whereas the mRFP1 fusion disrupted it (Fig 1I). Strikingly, in human cells, mRFP1 fusions invariably abolished intracellular LLPS of (UCST1)40 (Fig. 2B) and LCST1 (80-160 repeats; not shown). In contrast, sfGFP or mRFP1 modified with a wDimer peptide (with undetectable dimerization in yeast) led to prominent liquid-like condensates (Fig. 2B). We saw identical observations for LCST1 IDPPs (not shown). Surprisingly, the choice of FP also proved critical for our new trans-tagging approach (Fig. 2C). mRFP1-client masked the intrinsic LLPS behavior of cTEV-(UCST1)40 (Fig. 2D) and cTEV-LCST1 IDPPs (not shown), whereas the sfGFP-client revealed the expected liquid-like condensates. Overall, our data demonstrate that the intracellular LCST and UCST phase behavior of IDPs is highly sensitive to the molecular-level juxtaposition of monomeric FPs. The growing interest in the disordered proteome and the intrinsic vulnerability of intracellular LLPS dynamics will demand continued engineering of advanced for live-cell imaging tools.

Acknowledgements (Optional):