Fig. 6: A working hypothesis for PML NB assembly.

a Results of mammalian two-hybrid assay. Relative luciferase activities (RLUs) were used to evaluate the self-interaction of PML and its variant proteins. All data were normalized against the “Vector” group (pACT vector:pBIND-PML interaction = 1), with PML-WT serving as a positive control. All experiments have been conducted in at least three independent replicates. Data are means ± SEM. ***P < 0.001, ****P < 0.0001 (derived from ordinary one-way ANOVA test) were used to show statistical significance. b, c NB formation rescued by ATO. HeLa^Pml−/− cells expressing PML-WT and PML variants were exposed to 2 μM ATO for 1 h before visualization. Scale bar, 5 µm. Statistical analysis of PML NB formation was conducted. Data are means ± SEM. **P < 0.01, ****P < 0.0001 (derived from ordinary one-way ANOVA test). All experiments were performed in independent replicates, with NB counts calculated from ≥ 15 nuclei. All experiments/results displayed in the main figure were conducted with PML isoform IV. d The assembly process of PML NBs was summarized below. Formation of PML monomer: the RING, B1, and B2 domains orderly fold around the PML α3 and distribute on the outer surface of PML molecules. PML dimerization: the resulting PML monomers engage with each other via RB (RING, B1, B2, PML α6)- or CC-interaction interfaces to form a stable dimer. Octopus-like assembly: the CC domain of PML is flexible, and it further recruits and binds PML proteins in an anti-parallel orientation. Of note, although current results favored anti-parallel configuration, parallel CC–CC interaction could not be excluded and needs further characterization in the future. Partner recruitment: SUMO-conjugated PML recruits SIM-containing partners through non-covalent SUMO/SIM interactions. LLPS: by spatial multimerization, millions of PML molecules might polymerize into a higher-order complex, initiating LLPS and ultimately leading to the formation of PML NBs.