The phenylalanine-and-glycine repeats of NUP98 oncofusions form condensates that selectively partition transcriptional coactivators

Recurrent cancer-causing fusions of NUP98 produce higher-order assemblies known as condensates. How NUP98 oncofusion-driven condensates activate oncogenes remains poorly understood. Here, we investigate NUP98-PHF23, a leukemogenic chimera of the disordered phenylalanine-and glycine (FG)-repeats-rich region of NUP98 and the H3K4me3/2-binding PHD finger domain of PHF23. Our integrated analyses using mutagenesis, proteomics, genomics, and condensate reconstitution demonstrate that the PHD domain targets condensates to H3K4me3/2-demarcated developmental genes while FG repeats determine condensate composition and gene activation. FG repeats are necessary to form condensates that partition a specific set of transcriptional regulators, notably the KMT2/MLL H3K4 methyltransferases, histone acetyltransferases and BRD4. FG repeats are sufficient to partition transcriptional regulators and activate a reporter when tethered to a genomic locus. NUP98-PHF23 assembles the chromatin-bound condensates that partition multiple positive regulators, initiating a feed-forward loop of reading-and-writing active histone modifications. This network of interactions enforces an open chromatin landscape at proto-oncogenes, thereby driving cancerous transcriptional programs.反复出现的致癌性NUP98融合基因可形成被称为凝聚体（condensates）的高阶组装体。目前学界对NUP98致癌融合蛋白驱动的凝聚体如何激活致癌基因的机制仍不甚明晰。本研究针对NUP98-PHF23展开探究，该蛋白是由NUP98中富含无序苯丙氨酸-甘氨酸（FG）重复序列的区域，与PHF23中结合H3K4me3/2的PHD锌指结构域所构成的致白血病嵌合蛋白。本研究通过诱变实验、蛋白质组学、基因组学以及凝聚体重构等整合分析手段证实：PHD结构域可将凝聚体靶向至由H3K4me3/2标记的发育基因，而FG重复序列则决定凝聚体的组成与基因激活能力。FG重复序列是形成可分选特定转录调控因子的凝聚体的必要条件，这类凝聚体可特异性招募包括KMT2/MLL H3K4甲基转移酶、组蛋白乙酰转移酶以及BRD4在内的关键调控因子。当FG重复序列被锚定至基因组位点时，其足以完成转录调控因子的分选并激活报告基因。NUP98-PHF23可组装结合染色质的凝聚体，通过分选多种正向调控因子，启动读取并写入活性组蛋白修饰的前馈环路。这一相互作用网络可在原癌基因位点维持开放染色质构象，进而驱动致癌性转录程序的运行。