LAUSR

In PNAS, Ren et al. (1) report histone demethylase KDM5B (also called JARID1B or PLU1) as a suppressor of acute myeloid leukemia (AML). AML is one of the most frequently diagnosed adult cancers and is the most common type of acute leukemia (2). It is a heterogeneous group of cancers that arise from clonal expansion and differentiation arrest of myeloid progenitors in the bone marrow. Chromosomal translocation is a hallmark event driving AML initiation. Among different AML subtypes, recurring t(5;11)(q35;15.5) that causes fusion of nuclear receptor–binding SET domain protein 1 (NSD1) to nucleoporin-98 (NUP98) occurs in 16% of pediatric AML (3). NUP98-NSD1–positive AML patients exhibit poor prognosis, with 4-y event-free survival of less than 10% (3), which highlights the urgent demand for new treatment strategies for these patients. Ren et al. (1) show that UNC1999, a dual inhibitor of the histone H3K27 methyltransferases enhancer of zeste homolog 1/2 (EZH1/2), significantly inhibited the clonogenic growth and induced the differentiation and apoptosis of murine AML cells driven by NUP98-NSD1. Consistently, UNC1999 treatment prolonged the survival of mice injected with these AML cells. These observations are reminiscent of previous studies that revealed the efficacy of the EZH2 inhibitor on AML driven by MLLAF9, MLL-ENL, MLL-AF10, or MOZ-TIF2 transformation (4, 5). These results support the development of EZH2targeted therapy to treat AML. EZH2 is the catalytic subunit of polycomb repressive complex 2 (PRC2). To search for key downstream regulators that become reactivated and account for the antileukemia effect of EZH2 inhibition, Ren et al. (1) compared overlapping transcriptomic changes of AML cells after treatment with EZH2 inhibitors or down-regulation of embryonic ectoderm development (EED), an essential component of PRC2. Kdm5b was identified as one of the 31 consistently induced genes. Chromatin immunoprecipitation sequencing (ChIP-seq) analysis showed that the transcriptional start site of Kdm5b is marked by EZH2, the H3K27me3 repressive mark catalyzed by EZH2, and the H3K4me3 activating mark (1). These suggest that Kdm5b is a direct target of EZH2 poised for transcriptional activation or repression and could have a tumor-suppressive role in AML. In fact, Ren et al. (1) then demonstrated the tumor suppressor role of Kdm5b in AML through Kdm5b knockdown experiments. Ablation of Kdm5b further enhanced AML clonogenic growth and desensitized AML cells to UNC1999 treatment both in vitro and in vivo. RNA sequencing analyses of AML cells with Kdm5b perturbation consistently showed a negative relationship between KDM5B level and transcriptional activity of genes that promote proliferation and stemness. These findings were consistent with a previous study that indicated the tumor-suppressor role of KDM5B in mixed lineage leukemia (MLL)–rearranged AML (6). Furthermore, Ren et al. (1) showed that low KDM5B expression is significantly correlated with poor survival of human AML patients, reinforcing the notion that KDM5B serves as a tumor suppressor in AML. However, these results contrast with the findings in solid tumors, including melanoma (7, 8) and estrogen receptor–positive breast cancer (9), that indicate the oncogenic and stem-like functions of KDM5B. The mechanisms that dictate the contextdependent roles of KDM5B in cancers remain to be investigated. NUP98-NSD1 fusion is an oncoprotein mainly because it binds to the regulatory elements of the locus harboring the homeobox (HOX) gene cluster. HOX genes encode a highly conserved family of transcription factors that are required to maintain stemness in both hematopoietic stem cell and AML driven by MLL-AF9 and NUP98-NSD1 (10, 11). Ren et al. (1) performed ChIP-seq to show that KDM5B colocalizes with NUP98-NSD1 to the promoterproximal sequence of the Hoxa locus as well as genes that either maintain stemness or drive proliferation, such as Sox4, Ccnd3, and Kras. This implies that KDM5B competes with NUP98-NSD1 for binding to the H3K4me3 transcriptional active mark to repress the effector genes that drive AML. KDM5B is a large multidomain protein with an AT rich interaction domain (ARID) domain for DNA binding, a jumonji C (JmjC) domain that confers the histone demethylase activity to erase H3K4me3, and three plant homeodomain (PHD) fingers. In order to determine the role of different domains in KDM5B to suppress AML, Ren et al. (1) ectopically expressed the wild-type (WT) and a series of KDM5B mutants in the NUP98-NSD1 AML model. Surprisingly, all domains, except the catalytic JmjC domain, abolished the antileukemic activity of KDM5B both in vitro and in vivo. Ren et al. (1) further performed ChIP-seq to show that the WT and catalytic inactive mutant of KDM5B not only substantially overlap globally with each other to promoter-proximal sequences, but also locally to the similar set of stemness and proliferation genes that are also bound by NUP98-NSD1. Such a high similarity of chromatin localization patterns between these two forms of KDM5B