Poster V14 - Sequence divergence across a family of skin disordered proteins enables access to an array of intracellular liquid-liquid phase separation dynamics

Introduction: The recent discovery of intracellular liquid-liquid phase separation (LLPS) dynamics in the skin provided mechanistic insights into the longstanding association between FLG mutations and skin barrier disorders. FLG is an intrinsically-disordered protein (IDP) that undergoes LLPS to form keratohyalin granules (KGs)─liquid-like biomolecular condensates whose assembly and pH-triggered disassembly propels terminal differentiation during skin barrier formation (Fig.1A). Disease-linked FLG truncation potently abrogates intracellular LLPS. Despite therapeutic interest in rescuing KG formation, FLG is a highly repetitive IDP (10-12 repeat domains of 324 residues each) that is incompatible with protein and gene delivery. Beyond skin barrier disorders, heterologous assembly of KG-like condensates may enable the genetically-encoded synthesis of novel cytosolic organelles for cell engineering. RPTN, a FLG paralog, appears to form KG-like condensates that are distinct from those assembled by FLG. This suggests a promising path towards engineering cells with two orthogonal KG-like compartments. Motivated by these opportunities, we are interested in FLG and FLG paralogs that differ at the sequence level but that share related LLPS-promoting sequence features. Our approach exploits their repeat architecture to extract sequence-level information and applies this towards engineering repeat IDPs that mimic FLG and its paralogs, allowing us to dissect their intracellular LLPS behavior in non-skin cells. In our investigation, we have exposed a range of intracellular LLPS dynamics encoded by sequence divergence across this enigmatic family of skin IDPs. These insights and our newly-engineered FLG-like IDPs will stimulate progress towards the engineering of biocompatible intracellular condensates in human cells.

Materials and

Methods: Building on known sequence heuristics for encoding LLPS in IDPs, we developed custom MATLAB scripts to analyze the LLPS-relevant information of FLG and its four paralogs (RPTN, HRNR, FLG2 and TCHH). We calculated moving averages within 10-residue windows for net charge, average Histidine (His) content, and average composite aromatic residue content including Tyrosine (Tyr, Y), Phenylalanine (Phe, F), and Tryptophan (Trp, W). In addition to compositional analyses, multiple sequence alignments informed the selection of one or two minimal repeats for each FLG paralog. These minimal repeats were used to construct genes encoding FLG paralog constructs with a variable number of repeats up to the full-length of the wild-type repeat domain. We constructed sequence-verified repeat genes using recursive directional ligation by plasmid reconstruction and fused the resulting genes with superfolder GFP (sfGFP) for live-cell imaging. To examine wild-type FLG and paralogs, we PCR amplified their largest exons (invariably encoding the entire IDP domain) from human genomic DNA using custom primers for subsequent cloning as fusions to sfGFP. For intracellular live-cell probing, we transfected select repeat genes and a nuclear marker (red-tagged H2B protein) into HeLa cells, assessing standard intracellular LLPS metrics 24 hours after transfection for up to a day of time-lapse imaging.

Results, Conclusions, and Discussions: Exploiting the repetitive architecture of FLG (Fig.1B) and its paralogs (Fig. 2A-B), we extracted LLPS-driving sequence features (charge; His and aromatic content) and selected minimal repeat units for each paralog. HRNR and FLG2 exhibited notable sequence-level differences in the makeup and distribution of their aromatic residues. FLG2 B-type repeats resembled FLG repeats, enriched for His (and low in Tyr, Phe, or Trp). The reverse was seen for HRNR and HRNR-like FLG2 A-type repeats which were abundant in Tyr, Phe, or Trp and low in His (Fig. 2A-B). Informed by these analyses, we selected two separate minimal repeats for FLG2 construction, each representing A-type and B-type repeats. Then, we built repeat IDPs (9 and 14 repeats at ~600 and ~900 residues respectively) that matched the sequence characteristics of the native FLG2 domains (Fig. 2C). A single repeat sufficed to match the sequence features of HRNR and RPTN (not shown). Fusing our FLG2 A- and B-type repeat domains with sfGFP for live-cell imaging, we captured their intracellular behavior in HeLa cells. Imaging revealed a stark contrast in LLPS dynamics between A- and B-type FLG2 domains (Fig. 2D). The His-rich FLG2-(B-type)14 protein remained soluble in the cytosol at all levels, consistent with our published results for truncated FLG variants consisting of 2-4 FLG repeats. Excitingly, the small FLG2-(A-type)9 protein readily formed intracellular droplet-like structures. Time-lapse imaging captured complete fusion events over the course of 4 hours, providing initial evidence of their liquid-like dynamics (Fig. 2E). Like FLG2-(A-type)9 proteins, our HRNR and RPTN mimics were potent drivers of intracellular LLPS (data not shown). Our preliminary data suggest that FLG paralogs exhibit a wide range of intracellular LLPS dynamics. Additionally, these small, engineered proteins inspired by FLG paralogs will serve as outstanding building blocks for the assembly of KG-like condensates. Our ongoing work contrasts the behavior of engineered mimics to full-length FLG and FLG paralogs, which may reveal yet-to-be-appreciated small domains that modulate their intracellular LLPS dynamics. Across this rich library of FLG-like proteins, we are currently probing whether select paralog mimics form condensates that co-exist with KGs, moving towards engineering intracellular multi-condensate systems.

Acknowledgements (Optional):