Par complex cluster formation mediated by phase separation

Cell polarity underlies many processes in metazoan cells and is characterized by local concentration of specific protein complexes at restricted membrane regions. The Par (partitioning defective) complex—comprising Par3 (Bazooka in Drosophila), Par6, and atypical protein kinase C (aPKC)—is a conserved master regulator of polarity across species and cell types, functioning in asymmetric cell division, epithelial polarity, migration, and neuronal polarization. While biochemical interactions among Par3, Par6, and aPKC have been defined (e.g., Par6–aPKC PB1-mediated heterodimerization, aPKC binding to Par3 CR3, and regulatory phosphorylation events), a key open question has been how Par proteins are recruited and concentrated at highly restricted membrane domains to initiate polarity. In Drosophila neuroblasts and C. elegans zygotes, Par proteins form apical/anterior crescents that consist of numerous micrometer-sized clusters. Par3’s N-terminal domain (NTD) can oligomerize and form filaments in vitro, but how such properties underlie the dynamic, fusing Par clusters observed in vivo and how sharp cortex–cytoplasm concentration gradients are maintained remained unclear. This study investigates whether liquid–liquid phase separation (LLPS) drives local condensation of the Par complex to establish apical-basal polarity.

• Prior work established Par3/Par6/aPKC as core polarity determinants in multiple systems, with defined domain interactions: Par6–aPKC PB1-PB1 heterodimers; aPKC binding Par3 CR3; and regulatory phosphorylation relieving inhibitory interactions to modulate complex assembly and function.
• The Par3–Par6 interaction site has been debated; mammalian Par3 was reported to bind Par6 via PDZ domains, but the precise PDZ site was unclear. Drosophila studies indicated weak binding of Par6 PBM to Baz PDZ1 and PDZ3 (Kd > 50 µM).
• In embryos and epithelia, Par proteins form crescents composed of micrometer-sized Par clusters. Par3 NTD oligomerizes and can form helical filaments in vitro, and NTD-mediated oligomerization is required for clustering and polarity in vivo. However, the mechanism by which such oligomerization yields dynamic, fusing clusters with rapid cortex–cytoplasm exchange remained unresolved.
• Par proteins in cortical clusters are highly dynamic and exchange rapidly with cytoplasmic pools, suggesting a dynamic assembly mechanism compatible with LLPS.
• LLPS has emerged as a general mechanism for forming membrane-less compartments via multivalent interactions, prompting investigation of whether Par complex condensation is LLPS-driven.

• In vivo imaging in Drosophila larval neuroblasts (NBs): Confocal microscopy of endogenous Baz/Par6/aPKC and basal determinants (Mira, Numb) through the cell cycle; analysis of apical/basal crescents and puncta.
• 1,6-hexanediol perturbation: Acute treatment (0–10%) of dissected larval brains to test hydrophobic-interaction–dependent condensate sensitivity and reversibility.
• Mammalian cell assays: Overexpression of tagged mammalian Par3 (full-length, Δ4N12, Par3N), Par6β, and PKCι in COS7 cells; puncta formation, time-lapse imaging of fusion; FRAP to quantify exchange kinetics.
• In vitro LLPS assays: Purified Par3N and Par6β (and mutants) labeled with fluorophores; imaging of droplet formation across concentrations (0.5–25 µM); sedimentation-based partitioning between supernatant and pellet; DIC imaging of droplet fusion; FRAP of droplets.
• Biochemistry and biophysics:
  • GST pull-downs from lysates to map interactions between Par3 fragments (NTD, PDZ1–3, Par3N) and Par6β fragments (PB1, Crib-PDZ, PBM, combinations; full-length).
  • Fluorescence polarization to measure binding affinities between Par3 PDZ domains and PBM peptides from Par6 isoforms; effect of Par3 PDZ3 glycine mutations and 1,6-hexanediol on binding.
  • Size-exclusion chromatography to assess Par6 PB1 oligomerization and salt sensitivity.
  • X-ray crystallography of Par3 PDZ3 bound to Par6β PBM peptide; structure determination and analysis (PDB 6JUE).
• Mutational analyses to test multivalency in LLPS:
  • Par3N deletions (ΔNTD, ΔPDZ1, ΔPDZ3), NTD oligomerization mutants (V13D, D70K).
  • Par6β deletions (ΔPB1, ΔCrib-PDZ, ΔPBM) and PB1 replacement with p62 PB1 (chimera); PBM swaps among Par6 isoforms and chimeras.
  • Replacement of Par3 NTD with FUS low-complexity domains (FUSL 1–214; FUSS 1–141) to modulate LLPS-driving capacity.
• aPKC (PKCι) assays:
  • Recruitment of PKCι (full-length) or PKCι PB1 into Par3N/Par6β condensates in vitro and in cells; effect on LLPS extent (sedimentation; COS7 puncta assay).
  • Testing effects of Par3 CR3 phosphorylation: phospho-mimetic Par3 Δ4N12 S827,829E; co-expression with PKCι WT, kinase-dead K273R, constitutively active A120E; in vitro kinase assays with Phos-tag PAGE to determine phosphorylation in supernatant vs condensate fractions.
• Drosophila genetics and in vivo functional tests:
  • Transgenic expression (UAS/GAL4 or actin Flip-out) of Flag-tagged Baz/Par6 WT and LLPS-deficient mutants (e.g., Baz NTDmu, Baz PDZ3mu, Baz ΔPDZ2; Par6 ΔPB1) in WT NBs.
  • MARCM to analyze baz or par6 mutant NB clones; rescue by expressing Baz/Par6 variants; lineage size quantification.
  • CRISPR/Cas9 GFP knock-in lines expressing endogenous-level GFP-Baz WT, GFP-Baz ΔNTD, or GFP-FUSs-Baz; in vivo FRAP in crescents.
• Quantification:
  • FRAP recovery halftime and fractions; puncta-positive cell counting; sedimentation pellet percentages; calibration-based estimation of intracellular concentrations (COS7: GFP-Par3N; fly NBs: GFP::baz cytoplasm vs crescent).

• Endogenous Par proteins form discrete apical cortical puncta that condense into crescents at metaphase and disperse after anaphase in Drosophila NBs; puncta are sensitive to 1,6-hexanediol and recover upon washout, consistent with LLPS.
• Par3 open-state fragments (Δ4N12; Par3N comprising NTD and PDZ1–3) form dim puncta in COS7 cells; co-expression with Par6β dramatically increases bright puncta that fuse over time and show rapid FRAP recovery (~75% with t1/2 ~10 s), indicative of liquid-like condensates.
• In vitro, Par3N forms droplets in a concentration-dependent manner; addition of Par6β greatly enhances LLPS, yielding larger and more numerous droplets. LLPS detectable at low concentrations (~0.5 µM by sedimentation), with faster fusion at 25 µM.
• Par3 PDZ3 directly recognizes Par6β PBM with Kd ~1 µM (∼50-fold stronger than Drosophila Baz PDZ3–Par6 PBM). The Par3 PDZ3 G600,602A mutation weakens binding ~20-fold. The crystal structure (PDB 6JUE) shows canonical PDZ–PBM interactions, with Par6β PBM residues -2ITL engaging the PDZ groove.
• Oligomerization increases multivalency and avidity: Par3 NTD oligomerization and Par6β PB1 self-association both enhance Par3–Par6 binding; Par6β PB1 oligomerizes in SEC and is salt-insensitive.
• Multivalency is essential for LLPS:
  • Par3N ΔPDZ3 or ΔNTD markedly impairs LLPS with Par6β, while ΔPDZ1 has minor effect; LLPS is salt-sensitive (consistent with NTD charge interactions).
  • Par6β ΔPBM or ΔPB1 reduces LLPS and its promotion of Par3N LLPS; replacing Par6β PB1 with p62 PB1 restores LLPS; ΔCrib-PDZ mildly reduces LLPS.
  • In cells, corresponding mutations (Par3N NTDmu; Par3N ΔPDZ3; Par6β ΔPB1/ΔPBM) greatly reduce or abolish puncta; Par6γ (PBM conserved) behaves like Par6β; Par6α (divergent PBM) fails unless PBM is swapped to Par6β sequence.
  • Replacing Par3 NTD with LLPS-driving FUS LCDs restores puncta formation proportionally to FUS LLPS strength (FUSL > FUSS), supporting LLPS as the clustering mechanism.
• aPKC (PKCι) is recruited into Par3N/Par6β condensates as a client without altering LLPS extent. Par3 Δ4N12 phospho-mimetic (S827,829E) forms fewer puncta with Par6β, suggesting aPKC phosphorylation antagonizes condensate formation. In vitro, PKCι WT and A120E phosphorylate Par3 in supernatant but not within condensates, indicating suppressed kinase activity inside droplets.
• In vivo condensation and function depend on LLPS:
  • Overexpressed Baz WT, Baz ΔPDZ2, and Par6 WT form apical crescents; Baz NTDmu and Baz PDZ3mu lose apical condensation; Par6 ΔPB1 diffuses cytoplasmically. Baz NTDmu/PDZ3mu disrupt localization of endogenous apical and basal determinants (dominant-negative effects).
  • In baz or par6 mutant NBs, LLPS-competent Baz or Par6 WT rescue apical localization, while LLPS-deficient variants (Baz NTDmu, Baz PDZ3mu; Par6 ΔPB1) fail to localize apically.
  • Lineage outcomes: WT type I NB lineages average ~73 cells; baz mutants ~23 cells. Rescue by Baz WT or Baz ΔPDZ2 largely restores lineage size; Baz NTDmu/PDZ3mu only partial; Par6 does not rescue baz mutants. par6 mutant lineage defects are rescued by Par6 WT but not Par6 ΔPB1.
• Concentration measurements: COS7 average cytoplasmic Par3N (~5 µM) and within puncta (~40 µM); endogenous Baz in NBs averages ~0.14 µM cytoplasm and ~1.32 µM in the apical crescent—consistent with lack of cytoplasmic LLPS but membrane-enriched LLPS at the cortex.
• Endogenous-level knock-ins: GFP-Baz ΔNTD still shows apical condensation but with less concentrated crescents and increased cytoplasmic diffusion vs WT; FUSs-Baz knock-in restores apical localization and normal partner localization; larval brain size reduced in Baz ΔNTD knock-in relative to WT and FUSs-Baz.
• Model: LLPS of Par3/Par6 concentrates the Par complex at the apical cortex in a cell cycle–dependent manner; aPKC is recruited inertly and upon activation phosphorylates Par3 to disperse condensates, coordinating assembly/disassembly during the polarity cycle.

The findings demonstrate that the Par3/Par6 complex undergoes liquid–liquid phase separation to form dynamic cortical condensates that underlie apical crescent assembly in dividing Drosophila neuroblasts. This LLPS mechanism reconciles previous observations of dynamic, fusing Par clusters and rapid exchange with cytoplasmic pools by invoking multivalent interactions: Par3 NTD oligomerization increases valency and scaffolding capacity; Par3 PDZ3 specifically binds Par6 PBM; and Par6 PB1 oligomerization further amplifies avidity. The structural determination of the Par3 PDZ3–Par6β PBM interface and mutational analyses pinpoint the specificity that drives co-condensation. Recruitment of aPKC into Par3/Par6 condensates without enhanced kinase activity provides a means to concentrate aPKC locally while keeping it quiescent; upon activation (e.g., via Cdc42 or cell cycle cues), aPKC phosphorylates Par3 CR3 to disassemble condensates, enabling polarity remodeling across the cell cycle. In vivo genetic and rescue experiments link LLPS competence directly to proper apical localization, asymmetric division, and lineage development, establishing functional relevance. Measurements of endogenous protein concentrations suggest that two-dimensional membrane attachment enriches Par proteins enough to cross the LLPS threshold, explaining why LLPS occurs at the cortex but not in cytoplasm at physiological levels. More broadly, the study supports a general principle whereby polarity complexes assemble into membrane-associated biomolecular condensates through multivalent interactions to achieve localized, dynamic, and regulatable concentration of signaling machinery.

This work establishes liquid–liquid phase separation as the mechanism driving local condensation of the Par3/Par6 complex at the apical cortex during asymmetric division. Key contributions include: identifying cell cycle–dependent Par puncta in NBs; demonstrating Par3 open-state LLPS enhanced by Par6 via specific PDZ3–PBM recognition and oligomerization; structurally resolving the PDZ3–PBM interaction; showing aPKC recruitment as an inactive client and phosphorylation-dependent condensate dispersal; and linking LLPS capacity to proper apical condensation, asymmetric division, and lineage outcomes in vivo. Future directions include elucidating how Par3 autoinhibition is relieved in vivo to permit LLPS, defining upstream signals (e.g., Cdc42, Plk1) that regulate condensate assembly/disassembly spatiotemporally, quantifying endogenous Par6 and aPKC concentrations at the cortex, and testing whether other polarity complexes (e.g., Lgl/Dlg/Scribble; PCP complexes) employ similar LLPS-driven condensation mechanisms.

• Many cellular assays relied on overexpression, which can shift LLPS thresholds and potentially cause non-physiological condensate formation; while endogenous knock-ins were used to address this, overexpression caveats remain for some conclusions.
• Affinity differences between mammalian and Drosophila proteins (e.g., stronger mammalian PDZ3–PBM interaction) may complicate cross-species extrapolation.
• The precise mechanisms relieving Par3 autoinhibition to enable open-state LLPS in vivo are not defined.
• Quantitative endogenous concentrations were measured for Baz; comparable in situ quantification for Par6 and aPKC at the cortex was not reported.
• The study focuses on neuroblasts; generalization to other polarized cell types and tissues, while plausible, requires direct testing.