Propagation of trimethylated H3K27 regulated by polycomb EED is required for embryogenesis, hematopoietic maintenance, and tumor suppression

Takeshi Uedaa,1,2, Yuichiro Nakataa, Akiko Nagamachib, Norimasa Yamasakia, Akinori Kanaib, Yasuyuki Seraa, Masato Sasakic, Hirotaka Matsuib, Zen-ichiro Hondad, Hideaki Odae, Linda Wolfff, Toshiya Inabab, and Hiroaki Hondaa,2

aDepartment of Disease Model, Research Institute for Radiation Biology and Medicine, Hiroshima University, Hiroshima 734-8553, Japan; bDepartment of Molecular Oncology, Research Institute for Radiation Biology and Medicine, Hiroshima University, Hiroshima 734-8553, Japan; cDivision of Infection and Host Defense, Faculty of Pharmaceutical Sciences, Tohoku Medical and Pharmaceutical University, Sendai 981-8558, Japan; dHealth Care Center and Graduate School of Humanities and Sciences, Institute of Environmental Science for Human Life, Ochanomizu University, Tokyo 112-8611, Japan; eDepartment of Pathology, Tokyo Women’s Medical University, Tokyo 162-8666, Japan; and fLaboratory of Cellular Oncology, Center for Cancer Research, National Cancer Institute, Bethesda, MD 20892

Edited by Tak W. Mak, The Campbell Family Institute for Breast Cancer Research at Princess Margaret Cancer Centre, University Health Network, Toronto, Canada, and approved July 26, 2016 (received for review January 8, 2016) Polycomb repressive complex 2 (PRC2) catalyzes the monomethylation, disorders, as well as in T-cell leukemia, and mostly predict inactivation dimethylation, and trimethylation of histone H3 Lys27 (H3K27) and acts of PRC2 function (5). We previously identified the EED as a central epigenetic regulator that marks the repressive chromatin mutations impairing PRC2 function in 3.1% of human myeloid domain. Embryonic ectoderm development (EED), an essential compo- disorders (6). Among these mutants, the Ile-to-Met mutation at nent of PRC2, interacts with trimethylated H3K27 (H3K27me3) through aminoacid363(I363M)ofEED,whichislocatedadjacenttothe the aromatic cage structure composed of its three aromatic amino residues constituting the aromatic cage structure, has been shown to acids, Phe97, Trp364, and Tyr365. This interaction allosterically activates have impaired binding ability to an H3K27me3 peptide, where it the histone methyltransferase activity of PRC2 and thereby propagates should interact with EZH2. Overexpression of the I363M-mutated

repressive histone marks. In this study, we report the analysis of protein led to a decrease of global H3K27me3 levels in mouse CELL BIOLOGY knock-in mice harboring the myeloid disorder-associated EED fibroblast cell line NIH 3T3, indicating that this mutant attenuated Ile363Met (I363M) mutation, analogous to the EED aromatic cage the propagation of repressive histone marks through impaired mutants. The I363M homozygotes displayed a remarkable and prefer- integrity of the aromatic cage structure (6). ential reduction of H3K27me3 and died at midgestation. The hetero- Increased susceptibility to hematologic tumors was previously zygotes increased the clonogenic capacity and bone marrow reported with Eed heterozygotes and homozygous hypomorphs repopulating activity of hematopoietic stem/progenitor cells (HSPCs) – Eed and were susceptible to leukemia. Lgals3, a PRC2 target gene encoding (7 9). These mutations blocked the interaction between EED a multifunctional galactose-binding lectin, was derepressed in I363M and EZH2 and/or destabilized the mutated EED (3, 10, heterozygotes, which enhanced the stemness of HSPCs. Thus, our work 11), which are considered to compromise the overall PRC2 complex provides in vivo evidence that the structural integrity of EED to H3K27me3 propagation is critical, especially for embryonic develop- Significance ment and hematopoietic homeostasis, and that its perturbation in- creases the predisposition to hematologic malignancies. Polycomb repressive complex 2 (PRC2) is a central regulator in all forms of histone H3 Lys27 (H3K27) methylation. Interaction of histone methylation | repressive histone marks | hematologic malignancies embryonic ectoderm development (EED) with trimethylated H3K27 (H3K27me3) is required for the allosteric activation of olycomb repressive complex 2 (PRC2) catalyzes the PRC2. We previously identified a myeloid disorder-associated EED Pmonomethylation, dimethylation, and trimethylation of his- Ile363Met (I363M) mutation with impaired binding ability to tone H3 Lys27 (H3K27) and plays a critical role in the epigenetic H3K27me3. By generating I363M knock-in mice, we demonstrated maintenance of repressive chromatin states. Histone-lysine that I363M preferentially dampened the propagation of H3K27me3 N-methyltransferase EZH2 is inactive on its own, even though repressive marks in vivo. The homozygotes caused embryonic the histone-lysine methyltransferase activity of PRC2 resides in lethality, whereas the heterozygotes enhanced hematopoietic the SET domain of EZH2 (1). EZH2 is required to interact with stem/progenitor cell (HSPC) activity, coupled with susceptibility other noncatalytic proteins, including the WD repeat-containing to leukemia. Lgals3, a PRC2 target gene, was derepressed by polycomb protein embryonic ectoderm development (EED) and the I363M, which enhanced the stemness of HSPCs. Our findings zinc finger protein SUZ12, to fully exert histone methyltransferase highlight the significance of the structural integrity of EED in activity (1). In mice, homozygotes for Eed null mutations show cellular homeostasis and tumor suppression. abolished global H3K27 monomethylation, dimethylation, and trimethylation, resulting in lethality by embryonic day (E) 9.5 Author contributions: T.U. and H.H. designed research; T.U., Y.N., A.N., N.Y., A.K., Y.S., Z.-i.H., H.O., and H.H. performed research; L.W. contributed new reagents/analytic tools; T.U., A.K., owing to a defect in primitive streak formation (2, 3). Along with M.S., H.M., T.I., and H.H. analyzed data; and T.U. and H.H. wrote the paper. – its action in PRC2 complex formation, the EED H3K27me3 The authors declare no conflict of interest. interaction allosterically activates the enzymatic activity of PRC2 This article is a PNAS Direct Submission. before propagating H3K27me3-repressive histone marks in a Data deposition: The sequences reported in this paper have been deposited in the DNA positive feedback loop (4). Phe-97, Trp-364, and Tyr-365 in Data Bank of Japan (accession nos. DRX045347–DRX045348 and DRX059643–DRX059650). human EED are required to form the so-called “aromatic cage” 1Present address: Department of Biochemistry, Faculty of Medicine, Kindai University, structures that can recognize H3K27me3 histone marks (4). Osakasayama 589-8511 Japan. Accumulating evidence implicates a genetic loss of PRC2 2To whom correspondence may be addressed. Email: [email protected] or function in patients with hematologic malignancies. Deletions [email protected]. and missense/nonsense mutations of PRC2 components have This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. been shown in myelodysplastic syndrome and myeloproliferative 1073/pnas.1600070113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1600070113 PNAS Early Edition | 1of6 Downloaded by guest on October 1, 2021 formation and the enzymatic activity involving H3K27 monomethylated, AB+/+ KI/+ KI/KI +/+ KI/KI KI/+ +/+ KI/KI KI/+ dimethylated, and trimethylated forms. In this study, to investigate the in vivo effect of the I363M E14.5 E12.5 mutation on disease pathogenesis, we generated and analyzed KI/KI knock-in (KI) mice with the I363M mutant of EED (EED I363M). C D +/+ E +/+ KI/+ KI/KI IP IP We demonstrate that unlike EED deficiency, which abrogates + /KI H3K27me3 (kDa) +/+ KI/ KI H3K27me1, H3K27me2, and H3K27me3, the I363M mutant (kDa) H3K27me2 EZH2 100 IgG IgG Input Input preferentially dampens the propagation of H3K27me3-repressive 100 EZH2 EZH2 57 SUZ12 H3K27me1 EED 70 histone marks. This finding allows us to consider that mice carrying 43 ACTIN 70 H3K4me3 I363M might be an excellent model for analyzing H3K27me3- EED 55 Total H3 preferential roles in vivo. We report the results of our phenotypic, F G H molecular, biochemical, and hematologic analyses of the mutant. Hoechst H3K27me3 1 0.01 H3K27me3 +/+ +/+ 0.8 0.005 Results and Discussion 0.6 0 Forced Expression of EED I363M or Aromatic Cage Mutants Decreases KI/+ 0.4 KI/KI -0.005 H3K27me3 Levels in K562 Cells. To compare the impact of I363M 0.2 +/+ KI/KI on the levels of H3K27me3 with the aromatic residue-mutated 0 -0.01 Average ppm (TSS=0) Average

KI/KI 0 EED proteins, we first established human chronic myeloid leu- 0 500 -500 1000 1500 2000 5000 3750 2500 1250 1250 2500 3750 5000 Average ppm (Peak summit=1) Average -2000 -1500 -1000 kemia K562 cells expressing wild-type (WT) EED [441 amino - - - - acids, NP_003788.2 (human)], I363M, and two aromatic cage I H3K27me3 summit centered (bp) Distance from TSS (bp) 0.01 mutants, Phe97Ala (F97A) and Trp364Ala (W364A) (Fig. 1A). H3K27me1 0.01 H3K27me2 0.01 Input We found that I363M, but not WT, EED led to a reduction in 0.005 0.005 0.005 global H3K27me3 levels, as expected; in addition, forced ex- 0 0 0

pression of the two aromatic cage mutants also displayed more -0.005 +/+ -0.005 +/+ -0.005 +/+ profoundly decreased global H3K27me3 levels (Fig. 1B). KI/KI KI/KI KI/KI -0.01 -0.01 -0.01 Average ppm (TSS=0) Average We examined the mRNA levels of ITLN2, SLC4A1,and 0 0 0 1250 2500 3750 5000 1250 2500 3750 5000 1250 2500 3750 5000 KLHDC8A, which contain gene regions marked by H3K27me3 and -5000 -3750 -2500 -1250 -5000 -3750 -2500 -1250 -5000 -3750 -2500 -1250 EZH2 in K562 cells, according to ENCODE chromatin immuno- Distance from TSS (bp) Distance from TSS (bp) Distance from TSS (bp) precipitation (ChIP)-sequencing datasets (Fig. S1). Among these, the Fig. 2. Embryonic lethality and decreased global H3K27 levels in I363M ITLN2 gene was up-regulated in cells expressing the mutated EED homozygotes. (A) Macroscopic abnormality of EedKI/KI embryos at E14.5. proteins compared with those expressing WT EED, although the (B) Gross examination of whole embryos and placentas at E12.5. No differences in expression of KLHDC8A and SLC4A1 was unaffected (Fig. 1C). The overall appearance were noted among the indicated genotypes. (C) Immuno- ITLN2 gene derepression was apparently correlated with the effect of blot showing comparable protein expression for EED and EZH2 in embry- EED mutants on H3K27me3 levels. Thus, I363M acted as an anti- onic fibroblasts. (D) I363M forms a PRC2 core complex with EZH2 and SUZ12. Cell extracts were immunoprecipitated, followed by immunoblotting with the morphic mutant of EED, analogous to the aromatic cage mutants. indicated antibodies. Three bands of EED correspond to the reported isoforms. (E) Immunoblot showing decreased amounts of H3K27me3 in the mutated Homozygous Mutation of I363M Leads to Embryonic Lethality. To embryonic fibroblasts. (F) Immunofluorescence analyses for H3K27me3 (green) investigate the effect of I363M in vivo, we generated Eed I363M in embryonic fibroblasts. Nuclei were visualized by Hoechst 33342 staining KI mice according to the identical amino acid sequences of (blue). (G) Average profiles of H3K27me3 deposition across a genomic window of ±2 kb (centered to the H3K27me3 peak summit) as shown by ppm deviation from the peak centered value. The average of the peak value was normalized to 1 for comparison. (H) Average binding profiles around the TSS of AB H3K27me3. (I) Average binding profiles around the TSS of H3K27me1 (Left), H3K27me2 (Middle), and Input (Right). The ppm deviation from the TSS (TSS = 0) is shown. The binding profiles were calculated on the Refseq-annotated for H and I.

human and murine EED protein [NP_068676.1 (mouse)] (Fig. + S2). Mice heterozygous for the mutated allele (EedKI/ ) were born C at the expected Mendelian ratio and crossed to generate homo- zygotes (EedKI/KI). Timed pregnancy experiments showed that the EedKI/KI embryos were developmentally arrested with s.c. edema and hemorrhage at around E14.5, and that no EedKI/KI mutants developed afterward (Fig. 2A and Fig. S3A). The live embryos at E12.5–E13.5 were not anemic, and had normal-sized placentas (Fig. 2B). There were no apparent structural defects of the tissues, including the lung, heart, and fetal liver, that could account for the lethality on H&E-stained sections of E13.5 live homozygotes (Fig. S3B). Of note, at E12.5– 13.5, the combined number of EedKI/KI live embryos and absorbed remnants was almost consistent with the Mendelian ratio (Fig. S3A). Fig. 1. Decreased H3K27me3 levels by I363M overexpression. (A) Structural Some of the absorbed embryos proved to be homozygotes on positions of Ile363 and three aromatic cage residues (Phe97, Trp364, and genotyping PCR, suggesting postimplantation embryonic lethal- Tyr365) on EED ( ID code 3IIW) are shown with the side ity of EedKI/KI mutants. chains in red. Histone H3 is shown with carbon atoms in yellow, nitrogen in KI/KI blue, and oxygen in green, as visualized using PyMol. (B) Immunoblot We noted that the size of Eed fetal liver was apparently showing decreased global H3K27me3 levels in K562 transfectants. (C) Rela- smaller, coupled with significantly reduced cell numbers, although tive mRNA levels of the indicated genes in K562 transfectants. surface markers of blood cells among homozygotes and the

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1600070113 Ueda et al. Downloaded by guest on October 1, 2021 corresponding controls were unchanged (Fig. S3 C and D). Thus, AB C we performed reconstitution of peripheral blood cells by fetal liver cell transplantation and found that the EedKI/KI fetal liver cells had a hematopoietic repopulation capacity comparable to + + that of the Eed / controls (Fig. S3E). These results suggest that the reduced number of EedKI/KI fetal liver cells is attributed mostly to a cell-extrinsic mechanism, which is supported by a previous study indicating that homozygous deletion of Eed in D E the hematopoietic organs caused no critical defects in the fetal liver earlier in midgestation (12). Thus, although the cause of embryonic death in EedKI/KI mutants was unclear, our results demonstrate that the aromatic cage integrity of EED is required for proper embryonic development.

KI/KI Marked Reduction of Global H3K27me3 Levels in Eed Cells. Ex- + + pression of EED and EZH2 proteins was detected among Eed / , + FGH EedKI/ , and EedKI/KI at similar levels in mouse embryonic fi- broblasts (MEFs) from embryos of these three genotypes (Fig. 2C). WT EED and I363M equivalently interacted with EZH2 and SUZ12, as demonstrated by immunoprecipitation using the anti-EZH2 antibody (Fig. 2D). Notably, immunoblot analysis of extracted histones demonstrated dramatically reduced levels of + + + H3K27me3 in EedKI/KI MEFs compared with EedKI/ and Eed / MEFs, whereas there were no alterations in global H3K27me1/me2 and H3K4me3 levels among these genotypes (Fig. 2E). Further- I + more, EedKI/ MEFs also showed lower H3K27me3 levels com- +/+

pared with Eed MEFs, in agreement with a dominant-negative CELL BIOLOGY effect of I363M (Fig. 2E). These results were validated by im- munofluorescent staining (Fig. 2F). Although marked reduction of H3K27me3 levels was confirmed in EedKI/KI MEFs by flow cytometry, the fluorescence levels of H3K27me3 in EedKI/KI cells were higher than those of the fluorescence-conjugated isotype control IgG (Fig. S4A), suggesting that I363M preferentially but Fig. 3. I363M causes an age-dependent expansion of hematopoietic cells with incompletely abrogated the trimethylation form of H3K27. enhanced repopulating activity of hematopoietic stem cells. (A) Immunoblot + ChIP sequencing (ChIP-seq) revealed that the average H3K27me3 showing decreased H3K27me3 in the EedKI/ BM nucleated cells (extracted his- ± tones). (B and C) Increased numbers of total BM nucleated cells (B) and hema- levels across a genomic window of 2kbcenteredtotheH3K27me3 + KI/KI topoietic progenitor cells (C)inagedEedKI/ mice. The total numbers of BM peak summit were profoundly reduced in Eed MEFs compared − − + +/+ nucleated cells and BM progenitor cells (LK; Lin ,Sca-1 ,c-Kit ) from young (age with those in Eed control cells (Fig. 2G). Thus, the integrity of + + + 10–12 wk) and aged (age 1.5 y) EedKI/ and Eed / mice (n = 4 per group) are the aromatic cage structure was shown to be critical for the shown. Mean values are presented; error bars indicate SD. *P < 0.05; **P < 0.005. + + + propagation of H3K27me3-repressive histone marks as proposed (D) Numbers of WBCs, myeloid cells (Mac1 ), B cells (B220 ), and T cells (Thy1.2 ) KI/KI previously (4). In line with this result, Eed cells demonstrated in the peripheral blood from littermates at age 1.5 y (Eed+/+, n = 13; EedKI/+, a more precipitous decrease in H3K27me3 levels at regions es- n = 10). Error bars indicate SD. *P < 0.05; ***P < 0.0005. (E) CFU-GM assay. Here + pecially distal from transcription start site (TSS) ±1kb compared 1 × 104 c-Kit cells were plated initially. Colonies were counted and then replated + + 4 with the Eed / controls, whereas a comparable strong peak en- at 2 × 10 cells every 7 d. Error bars indicate SD. The absolute numbers of col- < < richment with a sharp dip was detected within TSS ±1kb in these onies at the different platings are shown. **P 0.005; ***P 0.0005. (F)Rep- resentative images of colonies from the third plating as shown in E.(G) Cytospins + two genotypes (Fig. 2H). of the EedKI/ colony-forming cells cultured in methylcellulose at the fourth We further investigated the effect of I363M on the distribu- plating. Immature cells were observed. (H) Flow cytometry analysis of colony- tions of H3K27me1 and K27me2 by ChIP-seq. The H3K27me1 forming cells harvested at the first (Left)andthird(Right) platings. Mac1−Gr1−- and K27me2 occupancy around TSS was slightly altered by gated cells were analyzed for expression of CD34 and c-Kit. Compared with the + + + − + I363M, whereas the entire pattern of the cistromes was compa- Eed / controls, more EedKI/ cells showed an immature CD34 c-Kit phenotype. + +/+ KI/KI (I) Increased repopulation ability of EedKI/ LSKs. (Left) Percentages of donor- rable in Eed and Eed MEFs (Fig. 2I). These epigenetic + alterations also might influence the gene expression in part, al- derived (Ly5.2 ) cells in the total WBCs in the peripheral blood. Mean values are presented; error bars indicate the SEM. *P < 0.025. (Right) The absolute number though at this point we are unable to address whether the altered + of donor-derived WBC (Ly5.2 ) at 6 mo after transplantation in the same ex- H3K27me1 and K27me2 arise from a direct effect of the I363M periment. The horizontal line indicates the mean. **P < 0.005. mutation or from a secondary process caused by the decreased amount of H3K27me3. KI/KI The Eed mutants displayed embryonic lethality at a broader cytometry in combination with antibodies against H3K27me3 and spectrum of developmental stages than the Eed null homozygous surface antigen markers. In primary WT mouse BM, we found that embryos (2), possibly highlighting the developmentally preferen- the hematopoietic stem cell (HSC) population, LSK [lineage- tial role of H3K27me3 distinct from H3K27 monomethylation − + + and dimethylation. negative (Lin ), Sca-1 ,c-Kit ] cells, had markedly higher levels of H3K27me3 compared with more differentiated stages (Fig. S4B). KI/+ EedKI/+ Hematopoietic Cells Have a Proliferative Advantage and We detected lower levels of H3K27me3 in Eed LSK cells, as Exhibit Enhanced Repopulating Capacities. Whole bone marrow expected (Fig. S4C). + KI/+ (BM) cells from adult EedKI/ mice displayed less H3K27me3 The total BM cell numbers were normal in young Eed mice + + compared with Eed / controls (Fig. 3A). We analyzed H3K27me3 (6–12 wk), but increased in older mutants after a 1.5-y follow-up levels at the various differentiation stages of BM cells using flow (Fig. 3B). Furthermore, BM hematopoietic progenitors, defined as a

Ueda et al. PNAS Early Edition | 3of6 Downloaded by guest on October 1, 2021 − − + hematopoietic progenitor LK population (Lin ,Sca-1 ,c-Kit )that Bivalent H3K4me3 GO_G protein signaling A coupled to cyclic nucleotide Granulocyte /H3K27me3 genes contains common myeloid progenitors, granulocyte-macrophage second messenger Fingerprint progenitors, and megakaryocyte-erythroid progenitors, underwent a + KI/ NES 1.62 NES 2.09 NES 1.45 significant age-dependent expansion in the Eed BM (Fig. 3C and FDR-q 0.00 FDR-q 0.00 FDR-q 0.02 Fig. S5A), although there were no alterations in the LSK population + Upregulated in KI/+ Upregulated in KI/+ Upregulated in KI/+ between these time points (Fig. S5B). Aged EedKI/ mice (not young) exhibited a significant increase in white blood cell counts, including myeloid cell, B-cell, and T-cell lineages, in the peripheral B C ChIP: H3K27me3 + + Plt8/+ / Suz12 Top 5 common Lgals3 Arf-Ink4a blood compared with the Eed mice (Fig. 3D and Fig. S5C). The KI/+ LSK UP upregulated genes KI/+ Eed * n.s. Eed mice also displayed splenic hypercellularity at a younger age Top 100 3 LSK UP Lgals3 0.5 (Fig. S5D). Although splenomegaly was observed in some older Rftn2 0.4 KI/+ 2 Eed mice at around 1.5 y of age, there was no obvious lineage 19 Bex4 0.3 bias in these enlarged spleens (Fig. S5 E and F). Thus, the I363M Hnf4a 0.2 1

Rnf208 % of Input 0.1 % of Input mutation resulted in leukocytosis and an age-dependent 0.0 0 hypercellularity in the hematopoietic tissues, whereas I363M alone D E +/+ KI/+ +/+ KI/+ was considered insufficient for hematopoietic transformation in mice. * EGFPhigh To evaluate the effect of I363M on the clonogenic potential of 250 EedKI/+ siCtrl EV Lgals3 + n.s. siLgals3 200 9.06 74.9

hematopoietic cells, we subjected c-Kit HSC/progenitor cells from c-Kit

* c-Kit BM to serial hematopoietic granulocyte-macrophage colony- 150 * EGFPlow forming (CFU-GM) assays. On the first and second plating rounds, 100 * there was no difference in the number of colonies between the * +/+ KI/+ KI/+ 50 CD34 Eed and Eed cells; however, Eed cells exhibited contin- Colony number 0 uous growth and generated significantly more concentric colonies 12 3 456G CD34 after the third plating, whereas the colony-replating capacity of Plating round +/+ 7 Eed cells was substantially reduced (Fig. 3 E and F). The replated 6 + F EedKI/ colonies were homogeneous and less differentiated, EV Lgals3 5 EV Lgals3 − − + 4 containing myeloblast-like cells (Fig. 3G)andMac1Gr1 c-Kit 3 − 2 CD34 , defining a more immature phenotype (13, 14) (Fig. 3H) 1 0

and suggesting a greater growth and survival advantage of colonies/ 20000 cells + EV Lgals3 EedKI/ hematopoietic cells. + To test this possibility in vivo, we performed a competitive BM Fig. 4. Gene alterations in EedKI/ hematopoietic stem cells. (A)GSEAfor repopulation assay, in which HSCs (LSK cells) isolated from the expression of genes containing bivalent domain (Left), GO analysis for + + + + donor mice (Eed / or EedKI/ ; Ly5.2 ) were injected into lethally the biological process (Middle), and granulocyte fingerprint (Right) of dif- + ferentially expressed genes. (B) The top five genes commonly up-regulated irradiated recipients (Ly5.1 ) with BM nucleated cells from re- KI/+ Plt8/+ + in Eed and Suz12 LSKs by direct comparison. (C) ChIP-PCR analysis. + + cipient mice (WT; Ly5.1 ). The recipients transplanted with The H3K27me3 level was decreased at the Lgals3 gene locus in EedKI/ c-Kit KI/+ Eed cells exhibited significantly higher chimerism, with in- cells, whereas it was detected comparably at the Arf-Ink4a locus in both + creased donor-derived Ly5.2 cell numbers in the peripheral genotypes. (D) Decreased number of colonies in siLgals3-transfected cells. + + + blood (Fig. 3I). Thus, EedKI/ BM HSCs were shown to have a Here 5 × 103 transduced cultured EedKI/ c-Kit cells were plated in the initial cell-autonomous growth advantage. Our results are similar to plating. Colonies were counted after 5 d of culture and then replated at 4 those of previous studies reporting that the heterozygous null 2 × 10 cells. The average number of triplicated concentric colonies is shown. Error bars indicate SD. *P < 0.05. (E) Lgals3-IRES-EGFP (Lgals3) or empty mutation of Eed confer a growth advantage for BM HSCs (7, 12), + vector (EV)-transduced c-Kit cells were harvested after serial replatings (the − − whereas homozygous deletion of Eed in the hematopoietic cells fifth plating round), after which Mac1 Gr1 -gated cells were analyzed for results in severe HSC exhaustion only in the BM after birth (12). CD34 and c-Kit expression. Compared with the EV-transduced controls, more Therefore, complete loss of function and impaired function of Lgals3-transduced colony-forming cells retained an immature CD34−c-Kit+ EED appear to have different (possibly opposing) effects on phenotype. Most EGFPhigh cells belonged to the CD34−c-Kit+ fraction. BM hematopoiesis. (F) Cytospins of the colony-forming cells in E. Lgals3-transduced cells that retained a more immature morphology are shown. (G) Lgals3-transduced The Eed Mutation Primed Comprehensive Gene Alterations That cells display sustained ability of concentric colony formation. The bar graph × 4 Deregulated Hematopoietic Homeostasis. To determine the mo- shows the average number of colonies from 2 10 cells at the sixth plating lecular basis of the increased hematopoietic activity by the (Left) and representative images of colonies (Right). Error bars indicate SD. I363M mutation, we performed RNA sequencing of LSK cells, followed by Gene Set Enrichment Analysis (GSEA) for genes To focus on the differentially expressed genes associated with involved in hematopoiesis and (GO) biological KI/+ deregulated PRC2 function, we used the reported gene expression process category. Eed LSK cells showed significant enrich- data from mice harboring the Suz12 Plt8 mutation. Mice heterozy- ment of primitive HSC-specific genes harboring bivalent histone Plt8/+ gous for the Suz12 Plt8 allele (Suz12 mice) were reported to modifications, activating H3K4me3, and repressive H3K27me3 show leukocytosis and enhanced HSC repopulating activity, whereas associated with stem cell maintenance (Fig. 4A, Left) (15). Fur- homozygotes showed embryonic lethality, resembling the phenotypes thermore, I363M induced a significant enrichment of genes in- volving cellular signaling and homeostasis, as represented by of Eed I363M mutants (17, 18). By direct comparison of genes up- cyclic nucleotide-mediated signaling, G protein coupled to nu- regulated in LSK cells, we identified 19 up-regulated genes poten- cleotide signaling, and cellular cation homeostasis (Fig. 4A, tially associated with PRC2 deregulation in both mutants (Fig. S6A). + Middle and Table S1). In EedKI/ LSKs, the granulocyte finger- Among these, Lgals3, encoding Galectin-3, was identified as the KI/+ Plt8/+ print (16) was up-regulated and genes involved in immune reg- most up-regulated gene in Eed and Suz12 cells, which is ulation (lymphoid) were down-regulated, which potentially marked by H3K27me3 at TSS ±1kb according to the reported ChIP- contributed to a more statistically significant shift of myeloid seq for LSK (19) (Fig. 4B and Fig. S6A). Galectin-3 is a multifunc- lineage in the peripheral blood and thereby enhanced CFU-GM tional β-galactose–binding lectin reported to be involved in various capacity (Fig. 4A, Right and Table S1). biological processes, including cell growth and differentiation (20).

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1600070113 Ueda et al. Downloaded by guest on October 1, 2021 ChIP experiments followed by PCR (ChIP-PCR), using an A B anti-H3K27me3 antibody, demonstrated significantly lower +/+ (n=15) PB Spleen Gr14 H3K27me3 levels on the promoter region of the Lgals3 gene in 100100 10 3.55 48.9

+ 3 KI/ 10

Eed cells compared with controls (Fig. 4C and Fig. S6B), 8080 102 MPD FL4-H (mouse no.8) 101 40.2 7.33 indicating derepression of Lgals3 by the I363M mutation. Tran- 60 100 60 0 1 2 3 4 10 10 10 10 10 KI/+ + FL1-H 104 sient knockdown of Lgals3 in Eed c-Kit cells led to impaired 40 p=0.0003 1.18 16.5 40 3 MOL4070LTR 10

infection MPD 102 colony-formation capability at the successive platings, thereby FL4-H 2020 KI/+ (n=18) (mouse no.12) 101 Tumor-free survival (%) survival Tumor-free 47.1 35.2 100 suppressing the long-term stemness rather than clonogenicity 0 00 10 Mac1 itself (Fig. 4D and Fig. S6C). Furthermore, when retrovirally survival (%) Tumor-free 00100 100200 200300 300 transduced with Lgals3-IRES-EGFP (i.e., internal ribosome entry DaysDays + – C 626.7 D site enhanced green fluorescent protein cassette), the WT c-Kit 31.3 83.6 Evi1 27.7 283.3 EVI1 70.4 BM cells retained a higher population of immature cells with 25 10 20 8 blast-like morphology, in correlation with the expression level of 6 15 10 4 RPKM EGFP after serial replatings (Fig. 4 E and F and Fig. S6D). 2 5 0 0 In line with this observation, concentric colony formation was (Median=0.0188) 25 p=0.0096 sustained for longer periods by Lgals3 overexpression, whereas 20 Runx2 30 EZH2 15 20 10 the numbers of Lgals3-transduced colonies and vector controls 10 5 apparently decreased in a comparable manner (Fig. 4G and 0 0 Relative expression −10(Median=0) Fig. S6E). These results indicate that derepression of Lgals3,a Relative RPKM Spl-1 Spl-2 Spl-3 Spl-5 Spl-6 Spl-7 Spl-8 Spl-9 Spl-4 −20 Spl-10 Spl-11 Spl-12 Spl-13 Thy-14

PRC2 target gene, contributes to the stemness of HSPCs. Human Spl-Ctrl chronic myeloid leukemia cells overexpressing LGALS3 showed Fig. 5. Increased susceptibility to hematologic malignancies by the I363M increased proliferation and cell accumulation in the BM in a xe- – nograft model (21). Clinically, higher expression of LGALS3 is mutation. (A) Survival curves of mice infected with MOL4070LTR. (B) Wright Giemsa staining of peripheral blood (Left) and spleen imprints (Middle), considered a poor prognostic factor for overall survival in pa- along with the flow cytometry analysis (Right), of leukemia that developed + tients with acute myeloid leukemia (22). Thus, deregulated in retrovirus-infected EedKI/ mice diagnosed with a myeloproliferative dis- expression of Lgals3 may contribute to tumor susceptibility, order. Leukemic cells in the spleens of two representative mice were positive malignant transformation, and leukemia development. for Gr1 and/or Mac1 (mice 8 and 12 in Table S2). (C) Relative mRNA levels of Evi1 and Runx2 in the leukemic tissues. Evi1 was up-regulated in four of five CELL BIOLOGY The I363M Mutation Increases Susceptibility to Leukemia and Causes mice with myeloid disease (mice 3, 6, 8, and 12 in Table S2; blue). Spl, spleen; + Myeloid Leukemia with Overexpressed Evi1. EedKI/ mice survived Thy, thymus. (D) RNA-Seq data of 179 human AML cases (TCGA database) +/+ showing a significant inverse correlation between mRNA levels, higher levels as long as Eed controls over 1.5 y. Because the I363M mutation = was originally identified in human myeloid disorders, we investi- of EVI1 (Upper), and lower levels of EZH2 (Lower)incasesofAML.*P 0.0096, two-tailed Fisher’sexacttest. gated whether the I363M mutation influences the susceptibility to hematologic malignancies by retroviral insertional mutagene- KI/+ +/+ sis. To do so, we infected Eed and Eed neonates with a in 179 cases of acute myeloid leukemia (AML) from The replication-competent Moloney murine leukemia retrovirus- Cancer Genome Atlas (TCGA) database, based on data avail- based MOL4070LTR with the ability to induce malignant my- ability (26). We found that EVI1 waselevatedin17casesof eloid disorders and lymphoid leukemias (23). This protocol > KI/+ AML at extremely high levels ( 100-fold of the median reads profoundly accelerated disease onset and mortality in the Eed per kb per million) (Fig. 5D, Upper); however, because all cases neonates (Fig. 5A), and myeloid or T-cell malignancies were of AML had comparable EED and SUZ12 expression levels, we observed, corresponding to the clinical phenotypes of patients could not detect a substantial correlation in gene expression be- with loss-of-function mutations in the PRC2-associated genes (5) KI/+ tween EVI1 and these genes. Nevertheless, the AML cases that (Table S2). More specifically, 5 of 14 virus-infected Eed mice highly expressed EVI1 correlated significantly with lower levels of developed myeloid diseases involving distinct differentiation EZH2 gene expression (Fig. 5D, Lower). These data suggest a – stages, as determined by Wright Giemsa staining (Fig. 5B, synergistic role of EVI1 activation and impaired PRC2 function in Left and Middle) and analysis of surface markers (increased + − + + myeloid diseases. Mac1 Gr1 and Mac1 Gr1 populations) (Fig. 5B, Right and Tumor susceptibility caused by the aromatic cage mutation Table S2). Thus, EED I363M predisposes mice to malig- may be reminiscent of the histone H3 Lys27Met mutations nancy, including myeloid disorders, synergistically with ad- identified in patients with gliomas. These histone mutations halt ditional gene alterations. PRC2 complex formation by increasing the affinity to EZH2, To identify genes cooperating with the I363M mutation in resulting in an arrest of H3K27me3 propagation (27). Thus, mice, we extracted genomic DNA from the leukemic tissue of the virus-infected mice and subjected it to inverse PCR to detect impaired H3K27me3 propagation might be a crucial step in virus integration sites (24). We identified five genes representing promoting solid tumors, not only hematologic neoplasms. + common integration sites in EedKI tumor samples, and then In conclusion, in this study we have highlighted the biological used quantitative PCR to verify that two genes, Evi1 and Runx2, importance of the structural integrity of EED to H3K27me3 were highly expressed in the corresponding tumor tissues (Fig. 5C propagation, especially for embryonic development and he- and Table S2). In fact, Evi1 was strikingly up-regulated in four of matopoietic homeostasis, and demonstrated that this disrup- the five myeloid tumors (Fig. 5C;mice3,6,8,and12inTable S2), tion predisposes mice to hematologic malignancies (Fig. S7). whereas Runx2 was elevated in three of the eight T-cell leukemia Questions remain concerning tissue- and cell type-specificity samples (Fig. 5C;mice1,13,and14inTable S2). This result and context-dependency of EED function, however, because KI/+ supported Eed I363M genetically cooperates with Evi1 and Runx2 despite lower levels of H3K27me3, the Eed mice were able +/+ toward myeloid diseases and T-cell leukemia, respectively. to survive as long as the Eed mice. Interestingly, a recent EVI1, which encodes a zinc finger-containing DNA-binding study reported that H3K27me3 propagation is regulated by protein, has been reported to be highly expressed in myelodysplastic the JARID2 protein, a cofactor of PRC2 (28). Future studies syndromes and myeloid leukemia (25). We accordingly inspected are needed to clarify the perplexing and multifaceted in vivo the gene expression levels of EVI1 and PRC2 core components function of PRC2.

Ueda et al. PNAS Early Edition | 5of6 Downloaded by guest on October 1, 2021 Materials and Methods Immunofluorescence. For immunostaining, standard protocols were applied, The experiment was approved by the Committee on the Ethics of Animal using trimethyl-histone H3 (Lys27) rabbit monoclonal antibody (9733; Cell Experiments of the Hiroshima University, and all mice were maintained Signaling Technology) and Alexa Fluor 488-conjugated anti-rabbit IgG an- according to the guidelines of the Institute of Laboratory Animal Science, tibody (Invitrogen). The nuclei were counterstained with Hoechst 33342. Hiroshima University. Colony-Forming Assay. Cells were plated in MethoCult GF M3534 (STEMCELL Cell Culture and Transfection. The K562 cells were transfected with episomal Technologies). Colony numbers were counted, and cells were collected, type pEBmulti-Neo-vector (Wako Pure Chemicals) and pEBmulti-Neo-myc-WT pooled, and replated at the indicated cell numbers per 35-mm plate. EED/myc-F97A/myc-I363M/myc-F364A, using Nucleofection (Amaxa), and Competitive BM Reconstitution Assay. Dose of 2 × 103 LSK cells from 10-wk-old then selected by growth in G418 (1 mg/mL) for 7 d. + + + + EedKI/ and Eed / mice (Ly5.2 ) were administered i.v. into 8-wk-old C57BL/ × 5 Nuclear Complex Coimmunoprecipitation. Nuclear protein extraction and 6-Ly5.1 mice irradiated with a dose of 8.5 Gy together with 2 10 BM coimmunoprecipitation assays were performed using the Nuclear Complex competitor cells from C57BL/6-Ly5.1 mice. Co-IP Kit (Active Motif), according to the manufacturer’s protocol. Nuclear × 6 + extract (500 μg) was incubated with either mouse IgG or anti-EZH2 antibody Electroporation of siRNAs into Hematopoietic Cells. Here 1 10 c-Kit BM (AC22; Millipore), and Dynabeads M-280 sheep anti-mouse IgG (11201D; cells were cultured in X-VIVO15 (Lonza) supplemented with cytokines (SCF, Life Technologies). 100 ng/mL; TPO, 100 ng/mL; IL-6, 20 ng/mL) and then nucleofected using the Amaxa Mouse Macrophage Nucleofector Kit (program Y-001) and 100 pmol Acid Extraction of Histones. Cells were lysed in Triton extraction buffer [0.5% mouse Lgals3 pooled siRNA (ON-TARGETplus) or siCONTROL Non-Targeting (vol/vol) of Triton-X100 and protease inhibitor mixture (Roche) in PBS] for Pool (Dharmacon). At 24 h after nucleofection, the cells were plated for the 10 min on ice. The pellet was washed in Triton extraction buffer and then CFU-GM assay. resuspended in 0.2 M HCl for acid extraction overnight at 4 °C. The supernatants were subjected to immunoblotting. MOL4070LTR Retrovirus Injection. Preparation and infection of retrovirus were performed as described by Wolff et al. (23). Newborn pups from the breeding Immunoblotting. The following antibodies were used for immunoblotting: mice were inoculated i.p. with virus solution. Identification of retroviral in- anti-EED (05-1320 or 17–10034), anti-EZH2 (05-1319), and anti–monomethyl- tegration sites was performed as described previously (29). histone H3 (Lys27) (07-448) (Millipore); anti–c-Myc (sc-40) and anti-SUZ12 – (sc-46264) (Santa Cruz Biotechnology); anti–trimethyl-histone H3 (Lys27) Statistics. Mouse survival curves were constructed using Kaplan Meier (9733) (Cell Signaling Technology) and anti-histone H3 (3638; Cell Sig- methodology and compared by the log-rank test using GraphPad Prism naling Technology or 05-499; Millipore); anti–di-methyl-histone H3 (Lys27) software. Unless stated otherwise, other statistical analyses were performed ’ (ab24684) and anti–trimethyl-histone H3 (Lys4) (ab8550) (Abcam); anti- using Student s t test. Lgals3 (12733) (Cell Signaling Technology); and anti–β-actin (Sigma-Aldrich). Standard lysate preparation and electroblotting protocols were followed ACKNOWLEDGMENTS. We thank Yuki Sakai for animal care; Megumi unless stated otherwise. Nakamura, Sawako Ogata, and Rika Tai for mouse genotyping and molecular experiments; and Dr. Tatsuo Miyamoto for technical assistance. We also thank + the RIKEN BioResource Center for providing the B6.Cg-Tg(CAG-Cre)CZ-MO2Osb Retroviral Gene Transduction. c-Kit BM cells were cultured in medium mice (RBRC01828). This work was supported in part by a Grant-in-Aid for containing 100 ng/mL SCF, 100 ng/mL TPO, and 20 ng/mL IL-6. Cells were in- Scientific Research from the Ministry of Education, Science and Culture of fected twice with a retrovirus vector, with a 24-h interval between infections. Japan; the Takeda Science Foundation; the Osaka Cancer Research Founda- + Sorted EGFP cells were used for subsequent experiments. tion; and the Japan Leukemia Research Fund.

1. Simon JA, Kingston RE (2009) Mechanisms of polycomb gene silencing: Knowns and 18. Lee SC, et al. (2013) Polycomb repressive complex 2 (PRC2) suppresses Eμ-myc lymphoma. unknowns. Nat Rev Mol Cell Biol 10(10):697–708. Blood 122(15):2654–2663. 2. Faust C, Schumacher A, Holdener B, Magnuson T (1995) The eed mutation disrupts 19. Adli M, Zhu J, Bernstein BE (2010) Genome-wide chromatin maps derived from limited anterior mesoderm production in mice. Development 121(2):273–285. numbers of hematopoietic progenitors. Nat Methods 7(8):615–618. 3. Montgomery ND, et al. (2005) The murine polycomb group protein Eed is required for 20. Dumic J, Dabelic S, Flögel M (2006) Galectin-3: An open-ended story. Biochim Biophys global histone H3 lysine-27 methylation. Curr Biol 15(10):942–947. Acta 1760(4):616–635. 4. Margueron R, et al. (2009) Role of the polycomb protein EED in the propagation of 21. Yamamoto-Sugitani M, et al. (2011) Galectin-3 (Gal-3) induced by leukemia micro- repressive histone marks. Nature 461(7265):762–767. environment promotes drug resistance and bone marrow lodgment in chronic mye- 5. Laugesen A, Helin K (2014) Chromatin repressive complexes in stem cells, develop- logenous leukemia. Proc Natl Acad Sci USA 108(42):17468–17473. ment, and cancer. Cell Stem Cell 14(6):735–751. 22. Cheng CL, et al. (2013) Higher bone marrow LGALS3 expression is an independent 6. Ueda T, et al. (2012) EED mutants impair polycomb repressive complex 2 in myelodysplastic unfavorable prognostic factor for overall survival in patients with acute myeloid syndrome and related neoplasms. Leukemia 26(12):2557–2560. leukemia. Blood 121(16):3172–3180. 7. Lessard J, et al. (1999) Functional antagonism of the Polycomb-Group genes eed and 23. Wolff L, Koller R, Hu X, Anver MR (2003) A Moloney murine leukemia virus-based Bmi1 in hemopoietic cell proliferation. Genes Dev 13(20):2691–2703. retrovirus with 4070A long terminal repeat sequences induces a high incidence of 8. Richie ER, et al. (2002) The Polycomb-group gene eed regulates thymocyte differen- myeloid as well as lymphoid neoplasms. J Virol 77(8):4965–4971. tiation and suppresses the development of carcinogen-induced T-cell lymphomas. 24. Jonkers J, Berns A (1996) Retroviral insertional mutagenesis as a strategy to identify Oncogene 21(2):299–306. cancer genes. Biochim Biophys Acta 1287(1):29–57. 9. Sauvageau M, et al. (2008) Quantitative expression profiling guided by common 25. Shih AH, Levine RL (2011) Molecular biology of myelodysplastic syndromes. Semin retroviral insertion sites reveals novel and cell type-specific cancer genes in leukemia. Oncol 38(5):613–620. Blood 111(2):790–799. 26. Cancer Genome Atlas Research Network (2013) Genomic and epigenomic landscapes 10. Denisenko O, Shnyreva M, Suzuki H, Bomsztyk K (1998) Point mutations in the WD40 of adult de novo acute myeloid leukemia. N Engl J Med 368(22):2059–2074. domain of Eed block its interaction with Ezh2. Mol Cell Biol 18(10):5634–5642. 27. Lewis PW, et al. (2013) Inhibition of PRC2 activity by a gain-of-function H3 mutation 11. Sewalt RG, et al. (1998) Characterization of interactions between the mammalian found in pediatric glioblastoma. Science 340(6134):857–861. polycomb-group proteins Enx1/EZH2 and EED suggests the existence of different 28. Sanulli S, et al. (2015) Jarid2 methylation via the PRC2 complex regulates H3K27me3 mammalian polycomb-group protein complexes. Mol Cell Biol 18(6):3586–3595. deposition during cell differentiation. Mol Cell 57(5):769–783. 12. Xie H, et al. (2014) Polycomb repressive complex 2 regulates normal hematopoietic stem 29. Yamasaki N, et al. (2010) Identification of Zfp521/ZNF521 as a cooperative gene for cell function in a developmental-stage-specific manner. Cell Stem Cell 14(1):68–80. E2A-HLF to develop acute B-lineage leukemia. Oncogene 29(13):1963–1975. 13. Matsuoka S, et al. (2001) CD34 expression on long-term repopulating hematopoietic 30. Chow S, et al. (2005) Whole blood fixation and permeabilization protocol with red stem cells changes during developmental stages. Blood 97(2):419–425. blood cell lysis for flow cytometry of intracellular phosphorylated epitopes in leu- 14. Celik H, et al. (2015) Enforced differentiation of Dnmt3a-null bone marrow leads to kocyte subpopulations. Cytometry A 67(1):4–17. failure with c-Kit mutations driving leukemic transformation. Blood 125(4):619–628. 31. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based ap- 15. Sun D, et al. (2014) Epigenomic profiling of young and aged HSCs reveals concerted proach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA changes during aging that reinforce self-renewal. Cell Stem Cell 14(5):673–688. 102(43):15545–15550. 16. Chambers SM, et al. (2007) Hematopoietic fingerprints: An expression database of 32. Li H, Durbin R (2009) Fast and accurate short read alignment with Burrows-Wheeler stem cells and their progeny. Cell Stem Cell 1(5):578–591. transform. Bioinformatics 25(14):1754–1760. 17. Majewski IJ, et al. (2008) Polycomb repressive complex 2 (PRC2) restricts hematopoietic 33. Zhang Y, et al. (2008) Model-based analysis of ChIP-Seq (MACS). Genome Biol stem cell activity. PLoS Biol 6(4):e93. 9(9):R137.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1600070113 Ueda et al. Downloaded by guest on October 1, 2021